Some Exact Solutions in General Relativity: Petarpa Boonserm

Some Exact Solutions
in
General Relativity
arXiv:gr-qc/0610149 v1 30 Oct 2006
Master’s thesis
by
Petarpa Boonserm
A thesis submitted to the

Victoria University of Wellington
in fulfilment of the requirements
for the degree of
Master of Science in Mathematics
2005
Supervisor: Dr. Matt Visser

Abstract
In this thesis four separate problems in general relativity are considered, divided
into two separate themes: coordinate conditions and perfect fluid spheres. Regard-
ing coordinate conditions we present a pedagogical discussion of how the appropriate
use of coordinate conditions can lead to simplifications in the form of the spacetime
curvature — such tricks are often helpful when seeking specific exact solutions of the
Einstein equations. Regarding perfect fluid spheres we present several methods of
transforming any given perfect fluid sphere into a possibly new perfect fluid sphere.
This is done in three qualitatively distinct manners: The first set of solution generat-
ing theorems apply in Schwarzschild curvature coordinates, and are phrased in terms
of the metric components: they show how to transform one static spherical perfect
fluid spacetime geometry into another. A second set of solution generating theorems
extends these ideas to other coordinate systems (such as isotropic, Gaussian polar,
Buchdahl, Synge, and exponential coordinates), again working directly in terms of the
metric components. Finally, the solution generating theorems are rephrased in terms
of the TOV equation and density and pressure profiles. Most of the relevant calcula-
tions are carried out analytically, though some numerical explorations are also carried
out.
i
ii
Acknowledgements
I would like to thank my supervisor, Dr. Matt Visser for his guidance and support.
I am lucky to have such an active and hard working supervisor. I really appreciate
the effort he made for my thesis, making sure that things got done properly and on
time. I would also like to thank Silke Weinfurtner, who I collaborated with on several
projects, for her help, support, and encouragement.
I am grateful to the School of Mathematics, Statistics, and Computer Science for
providing me with an office and all the facilities, and the Thai Government Scholarship
that provided me with funding.
I also would like to say thanks to my family who were also very supportive and
listened to me.
Finally, I would like to extend my gratitude to Brijesh Dongol for always being
there to listen to me. I am not sure that this thesis would have been finished without
the support he showed.
iii
iv
Preface
This thesis looks at four problems in general relativity:
• Coordinate conditions in general relativity: A method of looking for nice solu-

tions of the Einstein equations.
• Develop several new transformation theorems that map perfect fluid spheres into
perfect fluid spheres.
• Develop a systematic way of classifying the set of all perfect fluid spheres.
• Rephrase all these theorems which apply in Schwarzschild coordinates, directly

in terms of the pressure profile and density profile.
General relativity (GR) or general relativity theory (GRT) is a theory of gravitation

discovered by Albert Einstein. The fundamental physical postulate of GR is that
the presence of matter causes curvature in the spacetime in which it exists. This
curvature is taken to be the gravitational field produced by the matter. Einstein’s
field equation gives the mathematical description of how the matter and curvature are
related. Moreover, once this curvature is given, GR describes how other objects (such
as planets and light beams) move in this gravitational field via the geodesic equation.
In addition, general relativity states that clocks run slower in strong gravitational
fields (or highly accelerated frames), predicting a gravitational redshift. It also pre-
dicts the existence of gravitational lensing, gravitational waves, gravitomagnetism, the
Lense-Thirring effect, and relativistic precession of orbiting bodies.
v
The first problem is looking for nice solutions of the Einstein equations. However,
Before stating Einstein’s equation, we need to briefly describe the concept of special
relativity, otherwise general relativity will be hard to understand.
Structure of the thesis

This thesis has been written with the goal of being accessible to people with a basic
background in general relativity, especially in coordinate conditions in general relativ-
ity and perfect fluid spheres.
This thesis is made up of six chapters and three appendices. Two of appendices are
papers published on work relating to this thesis. One was produced as a collaboration
with Dr. Matt Visser, Silke Weinfurtner, Celine Cattoen and Tristan Faber. The
second was produced in collaboration with Dr. Matt Visser and Silke Weinfurtner. At
the time of writing all papers have been published.
vi
Contents
Acknowledgements iii
Preface v
1 General introduction 1
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Einstein’s equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2.1 General and Special Covariance . . . . . . . . . . . . . . . . . . 2
1.3 Special Relativity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3.1 Minkowski space-time . . . . . . . . . . . . . . . . . . . . . . . 3
1.3.2 The Stress-Energy Tensor . . . . . . . . . . . . . . . . . . . . . 5
1.3.3 Relativistic hydrodynamics . . . . . . . . . . . . . . . . . . . . . 7
1.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.4.1 Postulational formulation of special relativity . . . . . . . . . . 9
1.4.2 The correspondence principle . . . . . . . . . . . . . . . . . . . 10
1.4.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2 Coordinate conditions in general relativity 13

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.2 Independent components: Riemann’s argument . . . . . . . . . . . . . 15
2.2.1 For n = 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.2.2 For n = 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.2.3 For n = 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
vii
2.2.4 For n = 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.3 Lorentzian manifolds . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.4 The block diagonal ansatz . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.4.1 Collected results for the connexion . . . . . . . . . . . . . . . . 25
2.4.2 Back to the Ricci tensor: qualitative results . . . . . . . . . . . 25
2.4.3 Calculating the Ricci tensor - diagonal pieces . . . . . . . . . . . 28
2.4.4 Off-diagonal part of the Ricci tensor . . . . . . . . . . . . . . . 31
2.4.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.5 The ADM (Arnowitt-Deser-Misner) ansatz . . . . . . . . . . . . . . . . 34
2.5.1 ADM Decomposition . . . . . . . . . . . . . . . . . . . . . . . . 34
2.5.2 Reverse-ADM (Kaluza-Klein) Decomposition . . . . . . . . . . . 36
2.6 Static spacetimes: ADM and reverse-ADM (Kaluza-Klein) Decomposition 37
2.7 Zero–shift spacetimes: ADM and reverse-ADM (Kaluza-Klein) Decomposition 38
2.7.1 Systematic derivation . . . . . . . . . . . . . . . . . . . . . . . . 40
2.7.2 Riemann tensor . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
2.8 General stationary spacetime: ADM Decomposition . . . . . . . . . . . 43
2.8.1 Christoffel symbols of first kind . . . . . . . . . . . . . . . . . . 45
2.8.2 Completely general ADM decomposition . . . . . . . . . . . . . 52
2.8.3 The inverse-ADM decomposition . . . . . . . . . . . . . . . . . 52
2.8.4 Christoffel symbols of the 2 metrics −1/N 2 and hij . . . . . . . 55
2.8.5 Collected results for the connection . . . . . . . . . . . . . . . . 56
2.9 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
3 Perfect fluid spheres in general relativity 59

3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
3.2 Six theorems for generating perfect fluid spheres . . . . . . . . . . . . . 60
3.2.1 The first theorem . . . . . . . . . . . . . . . . . . . . . . . . . . 67
3.2.2 The second theorem . . . . . . . . . . . . . . . . . . . . . . . . 71
3.2.3 The third and fourth theorems . . . . . . . . . . . . . . . . . . . 78
3.2.4 Formal properties of the generating theorems . . . . . . . . . . . 81
3.2.5 Seed and non-seed metrics . . . . . . . . . . . . . . . . . . . . . 83
3.2.6 The fifth and sixth theorems . . . . . . . . . . . . . . . . . . . . 86
viii
3.3 Formal properties of the linking theorems . . . . . . . . . . . . . . . . . 88
3.4 Classifying perfect fluid spheres . . . . . . . . . . . . . . . . . . . . . . 93
3.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
4 Solution generating theorems — other coordinates 103

4.1 General diagonal coordinates: . . . . . . . . . . . . . . . . . . . . . . . 103
4.2 Schwarzschild (curvature) coordinates: . . . . . . . . . . . . . . . . . . 104
4.3 Isotropic coordinates: . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
4.3.1 The seventh and eighth theorems . . . . . . . . . . . . . . . . . 106
4.3.2 Two linking generating theorems: . . . . . . . . . . . . . . . . . 110
4.4 Gaussian polar coordinates: . . . . . . . . . . . . . . . . . . . . . . . . 111
4.4.1 The ninth theorem . . . . . . . . . . . . . . . . . . . . . . . . . 112
4.5 Buchdahl coordinates: . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
4.6 Synge isothermal coordinates: . . . . . . . . . . . . . . . . . . . . . . . 117
4.7 Exponential coordinates: . . . . . . . . . . . . . . . . . . . . . . . . . . 119
4.8 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
5 TOV equation - related theorems 123

5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
5.2 TOV equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
5.2.1 Interior equation . . . . . . . . . . . . . . . . . . . . . . . . . . 123
5.3 A perfect fluid sphere in Schwarzschild coordinates . . . . . . . . . . . 126
5.4 Solution generating theorems . . . . . . . . . . . . . . . . . . . . . . . 131
5.4.1 Four theorems . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
5.4.2 Four theorems in terms of pressure and density . . . . . . . . . 132
5.4.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
6 Conclusions 143
6.1 The main analysis: Structure of the thesis . . . . . . . . . . . . . . . . 143
6.2 Further interesting issues . . . . . . . . . . . . . . . . . . . . . . . . . . 144
Bibliography 145
ix
x
List of Figures
1.1 This structure shows the correspondence principle for general relativity. 10
3.1 A Pulsating spherical star cannot emit gravitational waves. . . . . . . . 65

3.2 The solid lines show B(r) for 10 reapplications of Theorem 1 onto the Minkowski metric. The d
3.3 This structure shows that when we apply theorem 1 to Minkowski, we get an Einstein static un
3.4 The solid lines show ζ(r) for 10 reapplications of Theorem 2 onto the Minkowski metric. The d
3.5 This structure shows that when we apply theorem 2 to Minkowski, we get the K-O III geometry
3.6 This structure shows that Theorem 3 and theorem 4 are in general distinct. When we apply the
3.7 This structure shows that when we invoke theorem 5 and apply it to Minkowski space, then ma
3.8 Structure graph for Minkowski space used as seed metric. . . . . . . . . 95
3.9 Numerical implementations of Theorem 3 and Theorem 4 can be used to investigate the behavi
4.1 This structure shows that Theorem 7 and Theorem 8 are in general distinct. When we apply T
4.2 This structure shows that Theorem 9 is idempotent. When we apply Theorem 9 to Minkowski
5.1 The diagram shows that when we write out equation (5.4) in terms of ρ, p, and ub, and projecti
xi
xii
List of Tables
3.1 This table shows several seed and non-seed metrics, which satisfied the definition of seed and no
3.2 Some well-known perfect fluid spheres and their coresponding metrics. Note that we have often
3.3 Some apparently new prefect fluid spheres and their coresponding metrics. Sometimes the relev
3.4 Seed solutions and their generalizations derived via theorems 1–4. The notation “[integral]” den
3.5 Non-seed solutions and their generalizations. . . . . . . . . . . . . . . . 101
5.1 Seed solutions and their generalizations derived via theorems 1B in terms of pressure and densi
xiii
xiv
Chapter 1
General introduction
1.1 Introduction
General relativity (GR) is the theory of space, time and gravitation formulated by
Einstein in 1915. General relativity is a beautiful theory, however it is often regarded
as a very deep and difficult theory. It has been considerably developed since the
late 1950s. Furthermore, in the mid 1960s, the modern theory of gravitational col-
lapse, singularities, and black holes has been developed. In this thesis, we present
the coordinate condition in general relativity, we will discuss the recent advances and
developments in methods of looking for a nice solutions of Einstein equation. In ad-
dition, we also present the perfect fluid spheres in general relativity. Understanding
the concept of perfect fluid spheres are important. They are first approximations to
attempt at building a realistic model for a general relativistic star. We introduce the
“algorithmic” techniques that permit one to generate large classess of perfect fluid
spheres in a purely mechanical way. Furthermore, we will extend these algorithmic
ideas, by proving several solution-generating theorems of varying levels of complexity.
We shall then explore the formal properties of these solution-generating theorems and
then will use these theorems to classify some of the previously known exact solutions.
Indeed, we develop several new transformation theorems that map perfect fluid spheres
into perfect fluid spheres. Additionally, we develop a systematic way of classifying the
set of all perfect fluid spheres. The remaining part of this thesis will be devoted to
1
rephrasing all these theorems, which originally apply in Schwarzschild coordinates,
directly in terms of the pressure profile and density profile.
Now we would like to give a mathematically precise formulation of the ideas in
Einstein’s equation. Before stating Einstein’s equation, we need to describe in brief
the concept of special relativity, otherwise general relativity will be hard to understand.
1.2 Einstein’s equation

To start we give an exposition of elementary topics in the geometry of space, and
the spatial tensorial character of physical laws in prerelativity physics. We formulate
general relativity and provide a motivation for Einstein’s equation, which relates the
geometry of spacetime to the distribution of matter in the universe.
1.2.1 General and Special Covariance

Assume that space has the manifold structure of R3 and the association of points of
space with elements (x1 , x2 , x3 ) of R3 can be accomplished by construction of a rigid
rectilinear grid of metersticks. We call the coordinates of space derived in this manner
as Cartesian coordinates [24]. The distance, S, between two points, x and x̄, defined
in terms of Cartesian coordinates by
S 2 = (x1 − x̄1 )2 + (x2 − x̄2 )2 + (x3 − x̄3 )2 (1.1)
This formula is the distance between two points.

Referred to equation (1.1), the distance between two nearby points is
(δS)2 = (δx1 )2 + (δx2 )2 + (δx3 )2 (1.2)
Therefore, the metric of space is given by
ds2 = (dx1 )2 + (dx2 )2 + (dx3 )2 (1.3)
In the Cartesian coordinate basis, we derive

X
ds2 = hab (dxa )(dxb ) (1.4)
a,b
2
with hab = diag (1, 1, 1). This definition of hab is independent of choice of Cartesian
coordinate system.
When the components of the metric in the Cartesian coordinate basis are constants,
we get
∂a hbc = 0. (1.5)
The space is the manifold R3 which possesses a flat Riemann metric. We are able to
use the geodesics of the flat metric to construct a Cartesian coordinate system. We use
the fact that initially parallel geodesics remain parallel because the cuvature vanishes.
1.3 Special Relativity

1.3.1 Minkowski space-time
We introduce a 4-dimensional continuum called space-time in which an event has
coordinates (t, x, y, z) [19].
Minkowski space-time is defined as a 4-dimensional manifold provided with a flat
metric of signature +2. By the definiton, since the metric is flat, there exists a special
coordinate system covering the whole manifold in which the metric is diagonal, with
diagonal elements equal to ±1. For convenience, we prefer to use the convention that
lower case latin indices run from 0 to 3. The special coordinate system is called a
Minkowski coordinate system and is written
(xa ) = (x0 , x1 , x2 , x3 ) = (c t, x, y, z). (1.6)
In special relativity [24], it is assumed that spacetime has the manifold structure of
R4 .
The spacetime interval, S, between two events x and x̄ defined by
S = −(x0 − x̄0 )2 + (x1 − x̄1 )2 + (x2 − x̄2 )2 + (x3 − x̄3 )2 (1.7)
in units where c = 1.
From the equation (1.7), we are able to defind the metric of spacetime ηab by
3
X
ds2 = ηab dxa dxb (1.8)
a,b=0
3
with ηab = diag (-1, 1, 1, 1), where xa is any global inertial coordinate system.
Therefore, the ordinary derivative operator, ∂a , of the global inertial coordinates
satisfies
∂a ηbc = 0 (1.9)
The curvature of ηab vanishes. In addition, we can parameterize timelike curves by

proper time, τ , defined by
Z
τ = (−ηab T a T b )1/2 dt, (1.10)
where t = arbitrary parameterizaton of the curve, and T a = the tangent to the curve
in this parameterization.
The tangent vector ua to a timelike curve parameterized by τ is defined by the
4-velocity of the curve. The square of any 4-vector is an invariant and so
ua ua = −c2 , (1.11)
where c = 1 so we get,
ua ua = −1. (1.12)
In the absence of external forces, its 4-velocity will satisfy the equation of motion,
ua ∂a ub = 0, (1.13)
where ∂a is the derivative operator associated with ηab . In addition, when forces are
present, the equation (1.13), ua ∂a ub is nonzero. Furthermore, all material particles
have a parameter known as “rest mass”, m, which appears as a parameter in the
equations of motion when forces are present. We can define the energy momentum
4-vector, pa , of a particle of mass m by
pa = m u a . (1.14)
Finally, we can define the energy of particle which is measured by an observer – present
at the site of the particle – as
E = −pa v a , (1.15)
4
where v a is the 4-velocity of the observer.
In special relativity, the energy is the “time component” of the 4-vector, pa . At the rest
frame, a particle with respect to the observer, equation (1.15) reduces to the familiar
formula E = m c2 . When the spacetime metric, ηab is flat, and the parallel transport
is path independent, we are able to define the energy of a particle as measured by an
observer who is not present at the site of the particle and has 4-velocity parallel to
that of the distant observer.
1.3.2 The Stress-Energy Tensor

In special relavity, we define the energy-momentum 4-vector of a particle of mass m
as in equation (1.14).
From [19], the Minkowski line element takes the form
ds2 = −dt2 + dx2 + dy 2 + dz 2 (1.16)
We can write this in tensorial form as
ds2 = ηab dxa dxb , (1.17)
We take ηab to denote the Minkowski metric

 
−1 0 0 0
 0 +1 0 0 
ηab ≡   = diag (−1, +1, +1, +1) (1.18)
 
 0 0 +1 0 
0 0 0 +1
In relativistic units the equation for the proper time satisfies
dτ 2 = −ds2 . (1.19)
Now we present proper time τ relates to coordinate time t for any observer whose
velocity at time t is v, where

dx dy dz
v= , , . (1.20)
dt dt dt
5
So we have,
dτ 2 = −ds2 = −(−c2 dt2 + dx2 + dy 2 + dz 2 )

" 2 2 2 #
dx dy dz
= −dt2 −c2 + + +
dt dt dt
( " 2 2 #)
2
1 dx dy dz
= −c2 dt2 −1 + 2 + +
c dt dt dt
v2

= c2 dt2 1 − 2
c
1/2
v2

dτ = 1− 2 c dt (1.21)
c
The time-component of the energy-momentum vector does represent the energy of the
particle
−1/2
v2

0 E
p = , γ = 1− 2 , and E = mc2 γ. (1.22)
c c
The space-components are the components of the three-dimensional momentum
p = mγv. (1.23)
A perfect fluid is defined to be a continuous distribution of matter with stress

energy tensor Tab of the form
Tab = ρua ub + p(ηab + ua ub ), (1.24)
where ua is the 4-velocity of the fluid, ρ is the mass-energy density in the rest-frame
of fluid, and p is the the pressure in the rest-frame of the fluid.
When there is no external forces, the equation of motion of a perfect fluid is simply
∂ a Tab = 0. (1.25)
6
Consider ∂ a Tab , we can write this in terms of ρ, p, and ua as
∂ a Tab = ∂ a [ρua ub + p(ηab + ua ub )]

= (∂ a ρ)ua ub + (∂ a p)(ηab + ua ub )
+(ρ + p)(∂ a ua )ub + (ρ + p)ua (∂ a ub )
= [(ua ∂a ρ) + (ρ + p)∂ a ua ] ub
+ [(ρ + p)ua ∂a ub + ∂ a p(ηab + ua ub )] (1.26)
For equation (1.26) we can project the resulting equation parallel and perpendicular
to ub , we find:
[(ua ∂a ρ) + (ρ + p)∂ a ua ] = 0, (1.27)
[(ρ + p)ua ∂a ub + ∂ a p(ηab + ua ub )] = 0. (1.28)

dp −
→
In the non-relativistic limit, when p ≪ ρ , uµ = (1, −
→
v ), and v ≪ | ∇p|, equation
dt
(1.27) becomes,
ua ∂a ρ + ρ ∂ a ua = 0
−
→ −
→ →
⇒ ∂t ρ + −
→
v · ∇ ρ + ρ∇ · − v = 0
−
→ −
→
⇒ ∂t ρ + ∇ · (ρ v ) = 0 (1.29)
and equation (1.28) becomes

−
∂→

v −
→ −
→−→ −
→
ρ + ( v · ∇) v = − ∇p (1.30)
∂t
1.3.3 Relativistic hydrodynamics

Ordinary hydrodynamics is dealing with two basic equations [22]:
The equation of continuity
∂t ρ + ∇ · (ρ−
→
v)=0 (1.31)
This relates to the density and velocity of the fluid. It is equivalent to the conservation
of mass.
7
The Euler equation is the fluid mechanics equivalent to Newton’s second law, it
relates the acceleration of a particle following the flow
−
→ dv ∂−
→v −
→→
a = = + (−→
v · ∇)−v (1.32)
dt ∂t
−
→
We can write it in term of force density f and mass density ρ:
−
→
−
→ ∂−
→v −
→ −
→− → f
a = + ( v · ∇) v = (1.33)
∂t ρ
These laws have relativistic generalizations which are:
∇a (ρ v a ) = 0 (1.34)
Aa = V b ∇b V a (1.35)
where Aa is now a 4-vector field of 4-accelerations.
The relativistic continuity equation yields
∂(ρ γ)
= ∇ · (ρ γ −
→
v) (1.36)
∂t
We interpret ρ as proportional to the number density of particles as measured by an
observer moving with the fluid. Indeed, the γ factor is corresponding to the fact that
Lorentz contraction “squashes” in the direction of motion, therefore, as seen by an
observer moving with respect to the fluid the number density of particles is ρ γ.
Now for the 4-acceleration
−
→
Ai = (γ ∂ + γ [−
t
→
v · ∇]) [γ v i ] (1.37)
and
−
→
A0 = (γ ∂t + γ [−
→
v · ∇]) [γ] (1.38)
The standard Newtonian results is reproduced at the low velocity, where γ → 1.
The special theory of relativity only deals with flat spacetime and the motion of
objects is usually treated in terms of Lorentz transformations and translations from
inertial frame to another. The general theory of relativity extends it to deal with
non-inertial frames, and via Einstein’s equations with curved spacetimes as well. We
need to understand the basic concepts of special relativity before turning to gen-
eral relativity—as otherwise the mathematical constructions used in general relativity
would appear rather unmotivated.
8
1.4 Conclusion
1.4.1 Postulational formulation of special relativity
There are two sets of postulates which are useful to generalize the general theory [19].
Postulate I. Space and time are represented by a 4-dimensional manifold provided with a
symmetric affine connection, Γa bc , and a metric tensor, gab , which is satisfied as
follows:
(i) gab is non-singular with signature − + ++;
(ii) ∇c gab = 0;
(iii) Ra bcd = 0.
The Postulate states that Γa bc is the metric connection and that the metric is flat.
Postulate II. There exist privileged classes of curves in the manifold singled out as follows:
(i) ideal clocks travel along timelike curves and measure the parameter τ defined by
dτ 2 = −gab dxa dxb ;
(ii) free particles travel along timelike geodesics;
(iii) light rays travel along null geodesics.
The first part of the second postulate makes physical the distinction between space
and time in the manifold. In Minkowski coordinates, it distinguishes the coordinate x0
from the other three as the “time” coordinate. Furthermore, it states that the proper
time τ which a clock measures is in accordance with the clock hypothesis. The rest of
Postulate II singles out the privileged curves that free particles and light rays travel
along.
9
1.4.2 The correspondence principle
Any new theory is consistent with any acceptable earlier theories within their range of
validity. General relativity must agree on the one hand with special relativity in the
absence of gravitation and on the other hand with Newtonian gravitational theory in
the limit of weak gravitational fields and low velocities by comparing with the speed
of light. This gives rise to a correspondence principle, as in figure (1.1), where arrows
indicate directions of increased specialization.
Figure 1.1: This structure shows the correspondence principle for general relativity.
1.4.3 Discussion
Special relativity is a physical theory published in 1905 by Albert Einstein [2]. The
theory is a special case of Einstein’s theory of relativity where the effects of acceler-
ation and of gravity can be ignored. Special relativity can be cast into an abstract
geometrical form based on the use of a flat 4-manifold called Minkowski space. In-
deed, most of this can easily be handled using high school algebra, however this is
not necessarily the most useful way of doing things — especially when you then want
to generalize to the curved spacetime of the general relativity. Most of the special
10
relativity, especially the foundations, can be adequately treated using matrix algebra
[22].
The theory of special relativity asserts that spacetime is the manifold R4 with a
flat metric of Lorentz signature defined on it. In special relativity, the principle of
general covariance states that the spacetime metric, ηab , is the only quantity relating
to spacetime structure can appear in any physical laws. In addition, the laws of physics
in special relativity satisfy the principle of special covariance with respect to the proper
Poincaré transformation.
In special relativity, energy is described in terms of the “time component” of 4-
vector pa . The continuous matter distributions in special relativity are described by a
symmetric tensor Tab called the stress energy momentum tensor.
In this chapter we introduce the theory of special relativity, which deals with the
motion of objects as seen from inertial frames. General relativity extends this to cover
non-inertial frames as well. As the theory of general relativity relies on the theory of
special relativity, we present the concept of special relativity first, so general relativity
is easier to understand.
11
12
Chapter 2
Coordinate conditions in general

relativity
2.1 Introduction
General relativity is a beautiful physical theory, invented for describing the gravita-
tional field and the equations it obeys. General relativity is phrased in the language
of differential geometry. It is by construction coordinate independent.
General relativity does not have a preferred coordinate system, and the coordinates
must be constructed along with the spacetime. The mathematics and physics cannot
depend on the particular choice of coordinate system you choose to place on the
manifold. However, the choice of the coordinate system can be useful in other ways.
It may simplify the mathematical calculation and the physics interpretation of what
is going on.
For example, if the physics is time independent and spherically symmetric, it is
extremely useful to choose a coordinate system that respects this property.
Reference [19] describes several coordinate conditions that we can consider to solve
the vacuum field equations
Gab = 0 (2.1)
13
for gab . At first sight, the problem seems well posed. Indeed, there are 10 equations for
the 10 unknowns gab . However, the equations are not independent, they are connected
by 4 differential constraints through the contracted Bianchi identities
∇b Gab ≡ 0. (2.2)
We therefore seem to have a problem of under-determinacy, when there are fewer
equations than unknowns. We cannot expect complete determinancy for any set gab ,
since they can be transformed with fourfold freedom by a coordinate transformation:
X a → X ′a = X ′a (x) (a = 0, 1, 2, 3). (2.3)

In fact, we can use this coordinate freedom to impose four conditions on the gab .
These are known as coordinate conditions or gauge conditions. We can give an example
by introducing Gaussian normal coordinates in which
g00 → +1, g0i → 0 (i = 1, 2, 3). (2.4)

The remaining six unknowns gαβ can be determined by six independent equations
in (1).
The assumption of spherical symmetry of a spacetime means that each event be-
longs to a two dimensional submanifold with spherical symmetry. This two dimensional
submanifold can be given the metric
ds2 = R2 (dθ2 + sin2 dφ2 ) (2.5)

where θ and φ are the standard angular coordinates on the sphere. There must exist
(at least locally) two additional coordinates t and r labelling the different spheres.
They can be chosen to be orthogonal to the sphere and also mutually orthogonal,
t to be time-like and r space-like. The metric thus reads
gik = diag −c2 e2φ , e2Λ , R2 , R2 sin2 θ

(2.6)
where Λ = Λ(r, t), φ = φ(r, t) and R = R(r, t). If ∂R 6= 0, it is possible to rescale R

∂r
to be identical with r. However, this special choice of the coordinate system need not
always be useful in particular cases.
14
In general relativity, we express physics in terms of tensors. General relativity
is a geometrical theory. Specifically, we will deal with a Riemannian manifold which
defines a metric. We used the metric to define a distance and length (norm) of vectors.
We have infinitesimal distance or interval which is defined as
(ds)2 = ds2 = gab dxa dxb (2.7)
where, ds in (2.7) is called a “line element”, gab is called a metric form. In addi-
tion, gab is called “covariant” metric and it has an inverse of the form g ab which is a
“contravariant” metric.
2.2 Independent components: Riemann’s argument

The quick way to see how much we can get away with is to simply count the number
of independent components of the metric tensor. In n dimensions, since the metric is a
symmetric T (0,2) tensor, it has n(n+1)/2 independent components. Since there are n
coordinates, we can eliminate up to n of the metric components by suitably choosing
our coordinates. This leaves
n(n+1)/2 - n = n(n-1)/2
true “independent” functions in the metric. This argument is due to Riemann.

The number of free components of the Riemann tensor is strictly limited by all
symmetries when we deal with a [torsion-free] metric connexion.
2.2.1 For n = 1
If n = 1, then this implies there are zero truly independent components in the metric
(see the proof below): In any 1-dimensional Riemann manifold we can always choose
coordinates so that the metric is flat. (A space-time is said to be flat if Ri jkl = 0 at
every point in it.)
15
Proof.
ds2 = f (x) dx2 = gxx (x) dx2 .
Let Z p
y= f (x) dx,
then
dy p
= f (x),
dx
which implies
p
dy = f dx.
So finally
dy 2 = f dx2 = ds2 .
i.e. we always can choose a coordinate y such that ds2 = dy 2 in any 1-dimension
manifold
2.2.2 For n = 2
If n = 2, then this Riemann’s argument implies there is exactly one truly independent
component in the metric: In any 2-dimensional Riemann manifold it is a standard
result that locally we can always choose coordinates to make the metric conformally
flat.
Indeed, when n = 2 the Riemann tensor has only one independent component,
essentially the Ricci Scalar, R. We have (in any coordinate system)
1
Rabcd = R (gac gbd − gad gbc )
2
1
Rab = R gab
2
Theorem: Any 2-dimensional Riemann manifold is locally conformally flat
gij = exp[2φ(x)] δij
16
2.2.3 For n = 3
If n = 3, then Riemann’s argument implies there are exactly three truly independent
components in the metric: In any 3-dimensional Riemann manifold it is a standard
result (typically attributed to Darboux) that locally we can always choose coordinates
to make the metric diagonal:
 
g11 0 0
gij =  0 g22 0 
 
0 0 g33
That is, Riemann 3-manifolds have metric that are always locally diagonalizable. If
n = 3, then the Riemann tensor has only six independent components, essentially Rab .
Indeed
Rabcd = −2{ga[d Rc]b + gb[c Rd]a } − R ga[c gd]b (2.8)
That is
Rabcd = {gac Rbd + gbd Rac − gad Rbc − gbc Rad }

1
− R {gac gbd − gad gbc } (2.9)
2
2.2.4 For n = 4
If n = 4 Riemann’s argument implies there are exactly six truly independent compo-
nents in the metric. This means that for a generic manifold in n = 4 you cannot assert
that  
g11 0 0 0
 0 g 0 0 
22
gij = 
 

 0 0 g33 0 
0 0 0 g44
because this metric has too few free functions (four rather than six). So in general
we cannot in 4 dimensions choose patches to make the metric patchwise diagonal. In
local coordinate patches, there are often non-zero off-diagonal elements. For a truly
17
general metric there must be at least 2 off diagonal elements. For instance, it might
be true that we can always set:
 
g11 g12 0 0
 g12 g22 0 0 
gij = 
 

 0 0 g33 g34 
0 0 g34 g44
This at least has the correct counting properties, and it would be nice to know if this
sort of thing is always possible. Another thing that can always be done locally in 4
dimensions is to set:  
1 0 0 0
 0 g
22 g23 g24 

gij = 


 0 g23 g33 g34 
0 g24 g34 g44
This is called a Gaussian normal coordinate patch.
If n = 4 the Riemann tensor has only twenty independent components. Ten of
them are the Ricci tensor Rab and the other ten are hidden in the “Weyl tensor”.
1
Cabcd = Rabcd + {ga[d Rc]b + gb[c Rd]a } + Rga[c gd]b (2.10)
3
That is
1 1
Rabcd = Cabcd + {gac Rbd + gbd Rac − gad Rbc − gbc Rad } − R{gac gbd − gad gbc } (2.11)
2 6
2.3 Lorentzian manifolds

The Lorentz metric is used to formulate the free-space Maxwell equations in a 4-
dimensional pseudo Riemann space and also used in describing mechanical phenomena
in a mass-free Riemann space. The line element and metric are then assumed to be
ds2 = −(dx0 )2 + (dx1 )2 + (dx2 )2 + (dx3 )2

= −c2 dt2 + (dx2 + dy 2 + dz 2 ) (2.12)
18
This particular metric tensor will be called the Lorentz metric tensor, and the corre-
sponding metric, the Lorentz metric.
 
−1 0 0 0
 0 1 0 0 
gab = 
 

 0 0 1 0 
0 0 0 1
The signature of a metric tensor is the number of positive and negative eigenvalues of
the metric. That is, the corresponding real symmetric matrix is diagonalized, and the
diagonal entries of each sign counted. If the matrix is n × n, the possible number of
positive signs may take any value p from 0 to n. Let q = n − p, then the signature may
be denoted either by a pair of integers such as (p, q), or as the bracketed sum (p + q),
or as a explicit list such as − + ++, or even as the number s = |p − q|. (Unfortunately
all of these different notations are in common use). A Lorentzian metric is one with
signature (p, 1) = (n−1, 1) or sometimes (1, q) = (1, n−1). For a Lorentzian signature:
(0+1) dimension
For a Riemannian geometry in (0+1) dimensions, this is trivial to get Lorentzian
geometry
g11 = − exp[2φ(t)]
(1+1) dimensions
For a Riemannian geometry in (2+0) dimensions we can always locally choose coordi-
nates such that
gij = exp[2φ(t, x)] δij
For Lorentzian geometry in (1+1) dimensions this becomes
gij = exp[2φ(t, x)] ηij
where " #
−1 0
ηij =
0 +1
19
" #
−1 0
gij = exp[2φ(t, x)]
0 +1
so " #
− exp[2φ(t, x)] 0
gij =
0 + exp[2φ(t, x)]
We can find the Ricci scalar n in term of φ, its derivative, η, the Ricci tensor and
Riemann tensor. The Ricci scalar is
n 2 2 o
2 − d 2 φ(t, x) + d 2 φ(t, x)
R=− dx dt (2.13)
exp(2φ(t, x))
And we can calculate the Ricci tensor, its non-zero components are:
d2
2
d
R11 = − φ(t, x) + φ(t, x) (2.14)
dx2 dt2
2 2
d d
R22 = φ(t, x) − φ(t, x) (2.15)
dx2 dt2
Similarly, we can calculate the Riemann tensor, its non-zero components are:
2 2
d d
R1212 = − exp(2φ(t, x)) − φ(t, x) + φ(t, x) (2.16)
dx2 dt2
In Lorentzian geometry, it is possible to make another coordinate choice by setting

" #
−[1 − v 2 ] −v
gij =
−v +1
where v = v(t,x), so " #

−[1 − v(t, x)2 ] −v(t, x)
gij =
−v(t, x) +1
In terms of this coordinate system we have: The Ricci scalar is
2
∂2
2
∂ ∂
R = −2 v(t, x) − 2 v(t, x) − 2v(t, x) v(t, x) (2.17)
∂x ∂t ∂x ∂x2
20
And we can calculate the Ricci tensor, its non-zero components are:
( 2 )
2
2
∂ ∂ ∂
R11 = − (−1 + v(t, x))2 v(t, x) + v(t, x) + v(t, x) v(t, x)
∂x ∂t ∂x ∂x2
(2.18)
( 2 )
∂2
2
∂ ∂
R12 = v(t, x) v(t, x) + v(t, x) + v(t, x) v(t, x) (2.19)
∂x ∂t ∂x ∂x2
2
∂2
2
∂ ∂
R22 =− v(t, x) − v(t, x) − v(t, x) v(t, x) (2.20)
∂x ∂t ∂x ∂x2
Similary, we can calculate the Riemann tensor, its non-zero components are:
2
∂2
2
∂ ∂
R1212 =− v(t, x) − v(t, x) − v(t, x) v(t, x) (2.21)
∂x ∂t ∂x ∂x2
(2+1) dimensions
The simplest special coordinate patch that it is always possible to set up yields
 
−h0 2 0 0
gij =  0 h1 2 0  ,
 
0 0 h2 2
where hi = hi (t, x, y), so

 
−h0 (t, x, y)2 0 0
gij =  0 h1 (t, x, y)2 0 .
 
0 0 h2 (t, x, y)2
Alternatively we can always locally set up patches of the form

 
−[N 2 − v 2 ] v1 v2
gij =  v1 1 0 
 
v2 0 1
21
where vi = vi (t, x, y), N = N(t, x, y), v 2 = v12 + v22 .
So the metric is
 
−[N(t, x, y)2 − (v12 (t, x, y) + v22 (t, x, y))] v1 (t, x, y) v2 (t, x, y)
gij =  v1 (t, x, y) 1 0
 

v2 (t, x, y) 0 1
Coordinate patches of this type are called Painleve–Gullstrand coordinates. Another
alternative, in (2+1) dimensions, is that we can always locally set up patches of the
form  
−[1 − v 2 ] v1 v2
gij =  v1 exp[2φ] 0
 

v2 0 exp[2φ]
where φ = φ(t, x, y), v 2 = exp[−2φ] (v12 + v22 ).
The inverse metric is then:
 
−1 v1 exp[−2φ] v2 exp[−2φ]
g ij =  v1 exp[−2φ] (exp[2φ] − v12 ) exp[−4φ] −v1 v2 exp[−4φ]
 

2
v2 exp[−2φ] −v1 v2 exp[−4φ] (exp[2φ] − v2 ) exp[−4φ]
In (2+1) dimensions we can calculate the Ricci scalar, the Ricci tensor and the Rie-
mann tensor. However, all results are too complex to analyze in detail.
(3+1) dimensions
This is the physically interesting case, and this is the case that will be most difficult
to analyze in detail, and the remainder of the Chapter is devoted to this case.
2.4 The block diagonal ansatz

• It is certainly true that we can always locally set:
 
g11 g12 0 0
 
 g12 g22 0 0 
gµν = 
 0

 0 g33 g34 

0 0 g34 g44
22
• This at least has the correct counting properties.
Now we can take the above as an ansatz and calculate the Ricci tensor, and then the
Einstein tensor, respectively.
Specifically, We write the four coordinates as xµ = (y a , z i ), with
x1 = y 1; x2 = y 2; x3 = z 1 ; x4 = z 2 .
Then  
g11 (y, z) g12 (y, z) 0 0
 
 g12 (y, z) g22 (y, z) 0 0 
gµν (x) = gµν (y, z) =  

 0 0 g33 (y, z) g34 (y, z) 

0 0 g34 (y, z) g44 (y, z)
That is " #
hab (y, z) 0
gµν (y, z) =
0 fij (y, z)
This means the 4 dimensional metric has split up into four 2 × 2 blocks, two of them
trivial (zero). This will make the Ricci tensor relatively easy to calculate.
The inverse tensor is very simple
" #
µν hab (y, z) 0
g =
0 f ij (y, z)
where hab and f ij are two dimensional inverses of the two 2 × 2 matrices hab and fij .
For definiteness, we assume the h subspace has signature (1+1) [one time direc-
tion plus one space direction] while the f subspace has signature (2+0) [two space
directions]. But this subtle point does not really affect the subsequent analysis.
The Christoffel symbols can be computed from

µ µν 1
Γ σρ (y, z) = g gν(σ,ρ) − gσρ,ν (2.22)
2
and now we can spilt them up into little 2 × 2 blocks
 
Γµ 11 (y, z) Γµ 12 (y, z) Γµ 13 (y, z) Γµ 14 (y, z)
 µ
 Γ 12 (y, z) Γµ 22 (y, z) Γµ 23 (y, z) Γµ 24 (y, z)

Γµ σρ (y, z) = 


 Γµ (y, z) Γµ (y, z) Γµ (y, z) Γµ 34 (y, z) 
 13 23 33 
µ µ µ
Γ 14 (y, z) Γ 24 (y, z) Γ 34 (y, z) Γµ 44 (y, z)
23
Some of these blocks should have a simple interpretation in terms of the Christoffel
symbols of the 2 dimensional metrics hab and fij .
For example

a aν 1 ad 1
Γ bc (y, z) = g gν(b,c) − gbc,ν = g gd(b,c) − gbc,d
2 2

1
= had hd(b,c) − hbc,d = [Γ(h)]a bc (y, z). (2.23)
2
That is, this portion of the 4 dimensional Christoffel symbol can be calculated simply
by looking at the 2 dimensional Christoffel symbol corresponding to the two dimen-
sional metric hab . Similarly, we can calculate this:

i iν 1 il 1
Γ jk (y, z) = g gν(j,k) − gjk,ν = g gl(j,k) − gjk,l
2 2

1
= hil hl(j,k) − hjk,l = [Γ(f )]i jk (y, z). (2.24)
2
The complications of course come with the other pieces, such as

i iν 1 ij 1
Γ bc (y, z) = g gν(b,c) − gbc,ν = f gj(b,c) − gbc,j
2 2

1 1
= f ij 0 − hbc,j = − f ij hbc,j . (2.25)
2 2
Similarly, and we can calculate this:

a aν 1 ab 1
Γ jk (y, z) = g gν(j,k) − gjk,ν = h gb(j,k) − gjk,b
2 2

1 1
= hab 0 − fjk,b = − hab fjk,b. (2.26)
2 2
This part of the 4-dimensional Christoffel symbol is rather simple but does not have
a simple interpretation in terms of the 2 dimensional sub-geometries.
There are four other sub-blocks to calculate:

a aν 1 ac 1
Γ bk (y, z) = g gν(b,k) − gbk,ν = g gc(b,k) − gbk,c (2.27)
2 2

1 1 1 1 1
= g ac gcb,k + gck,b − gbk,c = hac hcb,k = hac hcb,k
2 2 2 2 2
24
Whence
1
Γa jc (y, z) = hab hbc,j (2.28)
2
Similarly

i iν 1 ik 1
Γ jc (y, z) = g gν(j,c) − gjc,ν = g gk(j,c) − gjc,k (2.29)
2 2

ik 1 1 1 ik 1 1
= g gkj,c + gkc,j − gjc,k = f fkj,c = f ik fjk,c
2 2 2 2 2
Whence
1
Γi bk (y, z) = f ij fjk,b (2.30)
2
Various symmetries should make all four of these results obvious once any one of
them has been calculated.
2.4.1 Collected results for the connexion

There are a total of 8 sub-blocks to deal with:
Γa bc (y, z) = [Γ(h)]a bc (y, z).

Γi jk (y, z) = [Γ(f )]i jk (y, z).
1
Γi bc (y, z) = − f ij hbc,j .
2
1
Γa jk (y, z) = − hab fjk,b.
2
1
Γa bk (y, z) = hac hcb,k .
2
1 ab
Γa jc (y, z) = h hbc,j .
2
1 ik
Γi jc (y, z) = f fjk,c .
2
1 ij
Γi bk (y, z) = f fjk,b.
2
2.4.2 Back to the Ricci tensor: qualitative results

Once we have all eight sub-blocks, we have the entire 4 dimensional Christoffel symbol.
25
The Ricci tensor of gµν can then be calculated from [17], page 224
Rµν = Γσ µν,σ − Γσ µσ,ν + Γσ ρσ Γρ µν − Γρ σµ Γσ ρν . (2.31)
• It is convenient to first calculate Γσ ρσ
In fact, it is a standard result that

√ 1 g,a
Γσ ρσ = ln

−g ,ρ
= (2.32)
2 g
(see, for instance, equation (8.51a) on page 222 of [17]). Then we can simplify:
1 √ √
−g Γσ µν − Γρ σµ Γσ ρν ,

Rµν = √ ,σ
− ln −g ,µν
(2.33)
−g
(see, for instance, equation (8.51b) on page 222 of [17]). This is the explicit formula
for the Ricci tensor.
To interpret the Ricci tensor, write it in 2 × 2 sub-blocks
 
R11 (y, z) R12 (y, z) R13 (y, z) R14 (y, z)
 
 R12 (y, z) R22 (y, z) R23 (y, z) R24 (y, z) 
Rµν (y, z) = 
 
 R13 (y, z) R23 (y, z) R33 (y, z) R34 (y, z) 

R14 (y, z) R24 (y, z) R34 (y, z) R44 (y, z)
That is " #
Rab (y, z) Rai (y, z)
Rµν (y, z) =
Rjb (y, z) Rij (y, z)
Some of these blocks should have a simple interpretation in terms of the Ricci
tensor of the 2 dimensional metrics hab and fij . For instance we expect
[R(g)]ab = [R(h)]ab +???
[R(g)]ij = [R(f )]ij +???
where the extra terms (the ???) involve y derivatives of f and z derivatives of h. That
is, we expect the full 4 dimensional Ricci tensor of these particular 2 × 2 sub-blocks
26
to be related to the Ricci tensors of the 2 dimensional metrics f and h, but with some
additional terms.
For the off-diagonal block
[R(g)]ia =???
It will probably involve y derivatives of f and z derivatives of h.
Once you have the Ricci tensor, the Ricci scalar is easy
R(g) = g µν Rµν = hab Rab + f ij Rij = R(h)+??? + R(f )+???
That is
R(g) = R(h) + R(f )+???
That is, we expect the full 4 dimensional Ricci scalar to be related to the sum of the
Ricci tensors of the 2 dimensional metrics f and h, but with some additional terms.
Then for the four dimensional Einstein tensor
1
Gµν = Rµν − R gµν (2.34)
2
we can sub-divide the 2 × 2 sub blocks as
1
[G(g)]ab = [G(h)]ab − R(f ) hab +???
2
1
[G(g)]ij = [G(f )]ij − R(g) fij +???
2
But the Einstein tensor of any two dimensional manifold is zero; this is one of the
special features of two dimensions, so
[G(h)]ab = 0 = [G(f )]ij
and we have
1
[G(g)]ab = − R(f ) hab +???
2
1
[G(g)]ij = − R(g) fij +???
2
Finally, for the off-diagonal
[G(g)]aj = [R(g)]aj =???
Of course it is all these ??? pieces that give the important mathematics and physics.
Again, we are not really sure what to expect, they will probably involve y derivatives
of f and z derivatives of h.
27
2.4.3 Calculating the Ricci tensor - diagonal pieces
Start from the general expression for the Ricci tensor of the metric gµν :
1 √ √
−g Γσ µν ,σ − ln −g ,µν − Γρ σµ Γσ ρν .

Rµν = √ (2.35)
−g
Note that, since the metric is block diagonal,
g = det[gµν ] = det[hab ] det[fij ] = h f. (2.36)
Now consider
1 √ √
−g Γσ ab − Γρ σa Γσ ρb

[R(g)]ab = √ ,σ
− ln −g ,ab
(2.37)
−g
which we expand as
1 p
c
1 p
k
p
[R(g)]ab = √ −h f Γ ab + √ −h f Γ ab − ln −h f
−h f ,c −h f ,k ,ab
c d c k j k j d
−Γ da Γ cb − Γ ka Γ cb − Γ ka Γ jb − Γ da Γ jb
Then
1 √ 1 f,c c 1 p
[R(g)]ab = √ −h Γc ab + Γ ab + √ −h f Γk ab
−h ,c 2 f −h f ,k
√ p
− ln −h − ln f
,ab ,ab
c d c k
−Γ da Γ cb −Γ ka Γ cb − Γj ka Γk jb − Γj da Γd jb
which we regroup as
1 √ √
[R(g)]ab = √ −h Γc ab − ln −h − Γc da Γd cb
−h ,c ,ab
1 f,c c 1 p p
+ Γ ab + √ −h f Γk ab − ln f
2 f −h f ,k ,ab
c k j k j d
−Γ ka Γ cb − Γ ka Γ jb − Γ da Γ jb (2.38)
But the first line is just [R(h)]ab , the 2-dimensional Ricci tensor calculated using the
2-metric hab . The other terms can be slightly rearranged as

1 f,c c 1 p
k

[R(g)]ab = [R(h)]ab − (ln f ),ab − Γ ab + √ −h f Γ ab
2 f −h f ,k
c k j k j d
28
Now use colons : to denote 2-dimensional covariant derivatives using the 2-metric h.
Then
1 1 h,k k 1 p k
[R(g)]ab = [R(h)]ab − (ln f ) :ab + Γ ab + √ f Γ ab
2 2 h f ,k
c k j k j d
Now we start to insert the various pieces.

Note that
1 p k 1 1 p kj 1 1 p ij
√ f Γ ab =− √ f f ∂j hab =− √ f f ∂i hab
f ,k 2 f ,k 2 f ,j
We can write this more compactly as

1 p k 1
√ f Γ ab = − ∆f hab (2.41)
f ,k 2
where by ∆f we mean the two-dimensional Laplacian in the f subspace, treating the
h indices as though they were scalars.
Also note that
1
Γj ka Γk jb = f jm fmk,a f kn fnj,b (2.42)
4
which we can re-write as
1 ij
Γj ka Γk jb = f fjk,a f kl fli,b
4
1 −1
tr f f,a f −1 f,b

= (2.43)
4
Next, we start looking at
1
Γc ka Γk cb = − (hcd hda,k ) (f kj hbc,j )
4
1
= − hcd (had,i f ij hbc,j ) (2.44)
4
Similarly,
1
Γj da Γd jb = − hcd (had,i f ij hbc,j ) (2.45)
4
Finally
h,k k 1 h,i ij
Γ ab = − f hab,j (2.46)
h 2 h
29
Now pull all these pieces together. We get
1 1 1
(ln f ):ab − ∆f hab − tr f −1 f,a f −1 f,b

[R(g)]ab = [R(h)]ab −
2 2 4
1 h,i ij 1 cd
− f hab,j + h (had,i f ij hbc,j ) (2.47)
4 h 2
There are still simplifications we could make. For instance

f:a f:ab f:a f:b
(ln f ):ab = = − (2.48)
f :b f f f
Using this we get rid of the logarithm. Then finally

1 f:ab 1 f:a f:b 1 1
− ∆f hab − tr f −1 f,a f −1 f,b

[R(g)]ab = [R(h)]ab − +
2 f 2 f f 2 4
1 h,i ij 1 cd
− f hab,j + h (had,i f ij hbc,j ) (2.49)
4 h 2
For [R(g)]ij we can appeal to symmetry to simply write:
1 h|ij 1 h|i h|j 1 1

− ∆h fij − tr h−1 h,i h−1 h,j

[R(g)]ij = [R(f )]ij − +
2 h 2 h h 2 4
1 f,a ab 1 kl
− h fij,a + f (fik,a hab fjl,b) (2.50)
4 f 2
The vertical bar | now denotes 2-dimensional covariant derivatives in the f direction,
treating the h indices as scalars.
To obtain the Ricci scalar we note
1 ∆h f 1 f:a f :a 1 ab 1
hab [R(g)]ab = hab [R(h)]ab − − h ∆f hab − hab tr f −1 f,a f −1 f,b

+
2 f 2 f f 2 4
|i
1 h|i h 1
+ f ij tr h−1 h,i h−1 h,j

− (2.51)
4 h h 2
with a similar formula holding for hij [R(g)]ij :
1 ∆f h 1 h|i h|i 1 ij 1
f ij [R(g)]ij = f ij [R(f )]ij − − f ∆h fij − f ij tr h−1 h,i h−1 h,j

+
2 h 2 h h 2 4
|a
1 f|a f 1
+ hab tr f −1 f,a f −1 f,b

− (2.52)
4 f f 2
30
Then in the Ricci scalar there are a lot of partial cancellations
R(g) = hab [R(g)]ab + f ij [R(g)]ij
1 ∆h f 1 ∆f h 1 f:a f :a 1 h|i h|i
= R(h) + R(g) − − + +
2 f 2 h 4 f f 4 h h
1 ab 1 ij
− h ∆f hab − f ∆h fij
2 2
1 ab −1 1
+ h tr f f,a f −1 f,b + f ij tr h−1 h,i h−1 h,j

(2.53)
4 4
2.4.4 Off-diagonal part of the Ricci tensor

Now we consider what we can say about the off-diagonal piece [R(g)]ai . Again we will
want to start from the general formula for the Ricci tensor of the metric gµν :
1 √ √
−g Γσ µν ,σ − ln −g ,µν − Γρ σµ Γσ ρν .

Rµν = √
−g
Now consider
1 √ √
−g Γσ ai ,σ − ln −g ,ai − Γρ σa Γσ ρi

[R(g)]ai = √ (2.54)
−g
which we expand as
1 p
c
1 p
k
p
[R(g)]ai = √ −h f Γ ai + √ −h f Γ ab − ln −h f
−h f ,c −h f ,k ,ai
c d c k j k j d
−Γ da Γ ci − Γ ka Γ ci − Γ ka Γ ji − Γ da Γ ji
That is
1 p 1 p
[R(g)]ai = √ −h f Γc ai + √ −h f Γk ab
−h f ,c −h f ,k
1 1
− (ln(−h)),ai − (ln(f )),ai
2 2
−Γ da Γ ci − Γ ka Γk ci − Γj ka Γk ji − Γj da Γd ji
c d c
(2.55)
So that
1 1 1 √
[R(g)]ai = − (ln(f )),ai + (ln(f )),b Γb ai + √ −h Γb ai
2 2 −h ,b
1 1 1 p
− (ln(−h)),ai + (ln(−h)),j Γj ai + √ f Γj ay
2 2 f ,j
c d c k j k j d
−Γ da Γ ci − Γ ka Γ ci − Γ ka Γ ji − Γ da Γ ji (2.56)
31
Now consider √
b 1
−h Γ ai − Γc ab Γb ci
b

Γ ai:b = √ ∂b (2.57)
−h
Here Γb ai is treated as though it were a T11 tensor with respect to the h subspace, the
i index [which lies in the f subspace] is treated as though this quantity were a scalar
with respect to coordinate transformations in the f subspace. This notation, together
with the analogous
1 p
Γj ia|j = √ ∂j f Γj ia − Γkij Γj ka

(2.58)
f
allows us to simplify the off-diagonal Ricci tensor as:
1 1
[R(g)]ai = Γb ai:b + Γj ia|j − (f,a /f ),i − (h,i /h),a
2 2
1 1 1
+ (f,b /f ) Γb ai + (h,j /h) Γj ai − hac,l hcd f lj fji,d (2.59)
2 2 2
To simplify things, we can define two tensors by
1 1
Kabi = − hab,i = − hab|i (2.60)
2 2
1 1
Lija = − fij,a = − hij:a (2.61)
2 2
with the convention that the abc... indices are raised and lowered using h, and the
ijk... indices are raised and lowered using f .
(The reason for this notation is that these quantities are some sort of generalization
of the notion of extrinsic curvature.)
With these conventions we have:
Γa bc (y, z) = [Γ(h)]a bc (y, z).
Γi jk (y, z) = [Γ(f )]i jk (y, z).
Γi bc (y, z) = Kbc i .
Γa jk (y, z) = Ljk a .
Γa bk (y, z) = −K a bk .
Γa jc (y, z) = −K a cj .
Γi jc (y, z) = −Li jc .
Γi bk (y, z) = −Li kb .
32
Which then implies
Rai = −K b ai:b + K b bi:a − Lj ia|j + Lj ja|i

1 f,b b 1 h,j
− K ai − LK j ia − 2Kac j Lji c (2.62)
2 f 2 f
2.4.5 Discussion
This form of the metric is physically interesting because there is a good chance that it
really is general for 4 dimensional spacetime, and general enough to encode the Kerr
metric for a rotating black hole.
For this chapter we simply counted the number of independent components of the
metric tensor. In 1-dimension, there are zero truly independent components in the
metric. Furthermore, in any 1-dimensional Riemann manifold we are able to choose
the coordinates to make the metric flat. For 2-dimensions, Riemann’s argument im-
plies that there is only one truly independent component in the metric. Indeed, any
2-dimensional Riemann manifold has a standard result that locally, we can choose
coordinates to make the metric conformally flat. The Riemann tensor also has one
independent component, essentially the Ricci scalar, R. In addition, for n = 3, then
Riemann’s argument implies that there are exactly three truly independent compo-
nents in the metric. The Riemann 3-manifolds have metrics that are always locally
diagonalizable. For n = 4, the Riemann tensor has only twenty independent compo-
nents. Ten of them are the Ricci tensor Rab , while the other ten are hidden in the
“Weyl tensor”.
We use the Lorentz metric to formulate the free-space Maxwell equations in a
4-dimensional Riemann space, and to describe mechanical phenomena in a vacuum
Riemann space. It is trivial to get a Lorentzian geometry from a Riemannian geometry
in (0+1) dimensions. For a Riemannian geometry in (1+1) dimensions we are able to
find the Ricci scalar, the Ricci tensor and the Riemann tensor. While, for a Riemannian
geometry in (2+1) dimensions it is more complicated to calculate the Ricci scalar, Ricci
tensor and Riemann tensor.
The case for (3+1) dimensions is physically interesting, but it is the most difficult
to analyze in detail. However, we are able to calculate the Ricci tensor and Einstein
33
tensor in the 4 dimensional metric. We split the metric into four 2 × 2 blocks, two of
them trivial (zero), which allows us to calculate the Ricci tensor easily. Furthermore,
the Christoffel symbols can be computed from equation (2.22). For the 4-dimensional
Christoffel symbol, there is a total of 8 sub-blocks to deal with (see section (2.4.1)).
Now we are able to calculate the Ricci tensor of gµν from equation (2.31). We expect
the full 4 dimensional Ricci tensor to be related to the sum of the Ricci tensors of
the 2 dimensional metrics f and h, with some additional terms. In addition, we can
calculate the 4 dimensional Einstein tensor from equation (2.34). This is useful as a
way of looking for nice new solutions of the Einstein equations.
2.5 The ADM (Arnowitt-Deser-Misner) ansatz

Both ADM and reverse-ADM (Kaluza–Klein) decompositions seek to split “spacetime”
into “space” plus “time” in some sort of natural way, and then use this split to describe
spacetime curvature in terms of the curvature of space and some additional structure.
In addition, this decomposes the four-metric of spacetime as a time-varying three-
metric on a three-dimensional domain together with the lapse and shift, which are
scalar-valued and three-vector-valued functions of time and space, respectively.
2.5.1 ADM Decomposition

Definition 1. In the case of the ADM decomposition one takes
" #
−[c2 − v 2 ] −vj
gab =
−vi hij
Notation: Latin indices from the beginning of the alphabet (a, b, c, ...) run from 0-3
and refer to space-time; Latin indices from the middle of the alphabet (i, j, k...) run
from 1-3 and refer to space. We use X;a to denote a space-time covariant derivative
and X:i to denote a three-space covariant derivative.
Interpretation: The light cones, defined by ds2 = gab dxa dxb = 0 take the form
hij (dxi − v i dt) (dxj − v j dt) = c2 dt2
34
Here hij is the 3-metric of space. We take
hij = [hij ]−1
to be the inverse 3-metric, and
v 2 = hij vi vj = hij v i v j
Then v i is the 3-velocity of a coordinate system with respect to which c is the “coor-
dinate speed of light”. (Note that in comparison to usual conventions vi is minus the
“shift vector”, while c is equal to the “lapse function”.)
We are able to calculate the inverse metric:
" #
2 j 2
−1/c v /c
g ab =
−v i /c2 hij − v i v j /c2
Note that 3 contravariant vectors that point in the “space” directions are
X a = (0; 1, 0, 0), Y a = (0; 0, 1, 0), and Z a = (0; 0, 0, 1).
So if we want n to be a vector that is “orthogonal” to space, then we need to pick its
covariant components to satisfy
na ∝ (1; 0, 0, 0)
If we want n to be a unit vector that is “orthogonal” to space, then we need to pick
its covariant components to satisfy
na = ± (c; 0, 0, 0)
Its contravariant components are then
1 vi (1; v i )

a
n =∓ ; =∓
c c c
Choosing na to be a f uture directed unit vector orthogonal to the space slices then
fixes
1 vi (1; v i)

a
n = ; =
c c c
and
na = (−c; 0, 0, 0).
35
2.5.2 Reverse-ADM (Kaluza-Klein) Decomposition
Definition 2. The “reverse-ADM decomposition” can be found in Landau-Lifshitz and
consists of setting " #
−c2 −vj
gab =
−vi hij − vi vj /c2
Notation: Again, Latin indices from the begining of the alphabet (such as a, b, c,...)
run from 0-3 and refer to space-time; Latin indices from the middle of the alphabet
(i, j, k, ...) run from 1-3 and refer to space. We use X;a to denote a space-time covariant
derivative and X;i to denote a three-space covariant derivative.
Interpretation: The light cones, defined by ds2 = gab dxa dxb = 0 now take the form
hij dxi dxj = c2 (dt + vi dxi )2
We again say that hij is the 3-metric of space, though it’s a different 3-metric than in
the ADM case. We take
hij = [hij ]−1
to be the inverse 3-metric, and
v 2 = hij vi vj = hij v iv j
Then c is the “two-way coordinate-averaged speed of light”, which is by construction

isotropic. See §84 on pages 251-255 of [14] for details. The physical interpretation of
the 3-velocity v i is not particularly clear, and there seems to be no clean interpretation
of the current formalism in terms of “analogue models” [23]. Nevertheless it is one of
the standard metric decompositions, so let us adopt it and see what happens.
36
2.6 Static spacetimes: ADM and reverse-ADM (Kaluza-
Klein) Decomposition
Our strategy is straightforward:
• Take any static spacetime, and use the natural time coordinate to slice it into
space plus time.
• Use the Gauss-Codazzi and Gauss-Weingarten equations to decompose the
(3+1)-dimensional spacetime curvature tensor in terms of the 3-dimensional spatial
curvature tensor, the extrinsic curvature of the time slice [zero], and the gravitational
potential.
In any static spacetime one can decompose the spacetime metric into block diagonal
form [1, 2, 3]:
ds2 = gµν dxµ dxν

= −c2 dt2 + hij dxi dxj . (2.63)
Here c and hij are assumed to be time-independent, they depend only on position in
space. This is equivalent to taking
" #
−c2 0
gab =
0 hij
Theorem: In any static spacetime coordinates can be found in which the ADM and
reverse-ADM (Kaluza-Klein) decompositions are identical. The proof is by inspection.
Simply set ~v = 0 in the general formalism.
Being static tightly constrains the space-time geometry in terms of the three-geometry
of space on a constant time slice, and the manner in which this three-geometry is
embedded into the spacetime. For example, from [17], page 515, and applying page
518, we have the standard results (which we will re-derive in a slightly more general
context below):
(3+1) (3)
Rijkl = Rijkl . (2.64)
(3+1)
Rtijk = 0. (2.65)
(3+1)
Rtitj = c c:i:j . (2.66)
37
This will generalize when we allow time dependence for c and hij .
Now taking suitable contractions,
(3+1) (3) c:i:j

Rij = Rij − . (2.67)
c
(3+1)
Rti = 0. (2.68)
(3+1)
Rtt = c g ij c:i:j = c (3)
∆c. (2.69)
So that
(3)
(3+1) (3) ∆c
R= R−2
. (2.70)
c
Finally, for the spacetime Einstein tensor (for bits of this see [17], page 552, noting
that in our present case Kij = 0, while in the corresponding formulae of [17] they chose
c = 1)
(3)
(3+1) (3) c:i:j ∆c
Gij = Gij − + gij . (2.71)
c c
(3+1)
Gti = 0. (2.72)
2
(3+1) c (3)
Gtt = + R. (2.73)
2
This decomposition is generic to any static spacetime. (You can check this decom-
position against various standard textbooks to make sure the coefficients are correct.
For instance see [1, 12, 21]).
2.7 Zero–shift spacetimes: ADM and reverse-ADM

(Kaluza-Klein) Decomposition
If the spacetime is not static but the “shift vector” is zero we still have
" #
−c2 0
gab = . (2.74)
0 hij
(Such a decomposition is always possible in a globally hyperbolic spacetime.)

But now we will have to permit time dependence for both c and hij . Note that as
38
long as the shift vector is zero the ADM and reversed-ADM decompositions remain
identical.
In this case it is useful to define a quantity called the extrinsic curvature [17], the
equation (21.67)
1 ∂hij
Kij = − (2.75)
2c ∂t
and then
(3+1) (3)
Rijkl = Rijkl + Kik Kjl − Kil Kjk . (2.76)
(3+1)
Rtijk = c {Kik:j − Kij:k }. (2.77)
(3+1)
Rtitj = c c:ij + “mess”. (2.78)
See [17], equations (21.75) and (21.76). Note (n · n) = −1, and the “n-coordinate” of
equation (21.76) is really “Nt” [that is, “ct”]. The last equation for (3+1) Rtitj is not
given by [17], except implicitly on page 518, and this really is a bit of a mess which
we shall explicitly calculate below.
But some pieces of the Riemann, Ricci, and Einstein can be immediately read off
from [17]. For instance from [17], equation (21.8)
Rti = Gti = N Kij :j − tr(K):i

while from [17], equation (21.77)
1
Gtt = N 2 (3) R + (trK)2 − tr(K 2 )

2
39
2.7.1 Systematic derivation
Christoffel symbols of the first kind:
1 1 1
Γttt = gt(t,t) − gtt,t = gtt,t = − (c2 ),t = −c c,t (2.79)
2 2 2
1 1 1
Γitt = gi(t,t) − gtt,i = − gtt,t = + (c2 ),t = +c c,t (2.80)
2 2 2
1 1 1
Γtti = Γtit = gt(t,i) − gti,t = gtt,i = − (c2 ),i = −c c,i (2.81)
2 2 2
1 1 1
Γtij = gt(i,j) − gij,t = − gij,t = − hij,t = c Kij (2.82)
2 2 2
1 1 1
Γitj = Γijt = gi(t,j) − gtj,i = gij,t = hij,t = −c Kij (2.83)
2 2 2
1 1
Γijk = gi(j,k) − gjk,i = hi(j,k) − hjk,i = (3) Γijk (2.84)
2 2
Raising the first index:
Christoffel symbols of the second kind:
c,t
Γt tt = (2.85)
c
Γi tt = +c hij c,j = +c c,j (2.86)
c,i
Γt ti = Γt it = (2.87)
c
Kij
Γt ij = − (2.88)
c
Γi tj = Γi jr = −c K i j (2.89)
Γi jk = (3) i
Γ jk (2.90)
2.7.2 Riemann tensor

Now insert this into the formula for the Riemann tensor.
a
= −2 Γa b[c,d] − Γa e[c Γe |b|d]

Rbcd (2.91)
where the |b| indicates we do not anti-symmetrize over the b. Lowering the a index
Rabcd = −2 Γab[c,d] + gae,[dΓe |b|c] − Γa e[c Γe |b|d]

(2.92)
40
and rearranging
Rabcd = −2 Γab[c,d] + Γea[c g ef Γ|f b|d]

(2.93)
This is equivalent, for instance, to the unnumbered equation on page 21 of [10], which
can further be simplified to
1
Rabcd = {gad,bc − gbd,ac − gac,bd + gbc,ad } + Γead g ef Γf bc − Γeac g ef Γf bd (2.94)
2
This is equation (11.6) of [10]. So far this is completely general. If gab block diagonal-
izes in the manner gab = −c2 hij then we have
L
1 1
Rabcd = {gad,bc − gbd,ac − gac,bd + gbc,ad } − 2 Γtad Γtbc
2 c
1
+ 2 Γtac Γtbd + hmn Γmad Γnbc − hmn Γmac Γnbd (2.95)
c
Rijkl : If we now look at the space components of the 4-dimensional Riemann tensor
we have
1 1
Rijkl = {gil,jk − gjl,ik − gik,jl + gjk,il} − 2 Γtil Γtjk
2 c
1
+ 2 Γtik Γtjl + hmn Γmil Γnjk − hmn Γmik Γnjl (2.96)
c
But most of this is simply the 3-dimensional Riemann tensor
1 1
Rijkl = (3) Rijkl − 2
Γtil Γtjk + 2 Γtik Γtjl (2.97)
c c
and the rest we recognize in terms of the Kij we defined earlier, so that
Rijkl = (3) Rijkl − Kil Kjk + Kik Kjl (2.98)
which is the explicit way of doing things.

Rtijk : In a similar manner we can evaluate Rtijk as
1 1
Rtijk = {gtk,ij − gik,tj − gtj,ik + gij,tk } − 2 Γttk Γtij
2 c
1 mn mn
+ 2 Γttj Γtik + h Γmtk Γnij − h Γmtj Γnik (2.99)
c
41
Rtijk = (cKik ),j − (cKij ),k − c,j Kik + C,k Kij − cK n k Γnij + c K n j Γnik
so that
Rtijk = cKik,j − cKij,k − c K n k Γnij + c K n j Γnik
that is, in terms of 3-space covariant derivatives
Rtijk = c {Kik:j − Kij:k } (2.100)
This is the second part of what we were trying to prove.

Rtitj : It only remains to calculate Rtitj . (After all Rttij , Rttti , Rtttt and so on are all
zero by symmetries of the Riemann tensor.)
So let us calculate
1 1
Rtitj = {gtj,it − gij,tt − gtt,ij + git,tj } − 2 Γttj Γtit
2 c
1 mn mn
+ 2 Γttt Γtij + h Γmtj Γnit − h Γmtt Γnij
c
So that
1
−hij,tt + (c2 ),ij

Rtitj =
2
1 1
− 2 (−c c,j ) (−c c,i) + 2 (−c c,t ) (c Kij )
c c
+hmn (−cKmj ) (−cKni ) − hmn (c c,m ) Γnij
That is:
Rtitj = (cKij ),t + c c,ij + c,i c,j − c,j c,i − c,t Kij + c2 (Kim hmn Knj ) − (c c,m) Γm ij
So finally
Rtitj = cKij,t + c c:i:j + (K 2 )ij (2.101)
42
Summary: Whenever the metric gab block diagonalizes in the manner gab = −c2 hij ,
L
then even if we allow arbitrary time dependence we have the relativity simple results
(3)
Rijkl = Rijkl − Kil Kjk + Kik Kjl (2.102)
Rtijk = c {Kik:j − Kij:k } (2.103)
Rtitj = cKij,t + c c:i:j + (K 2 )ij (2.104)
where
1 ∂hij
Kij = −
2c ∂t
Of course once we have deduced these results for the Riemann tensor, results for the
Ricci tensor and Einstein tensor can be obtained by simple contraction.
2.8 General stationary spacetime: ADM Decom-

position
For a general stationary spacetime we need to use the full ADM decomposition
" #
−[c2 − v 2 ] −vj
gab =
−vi hij
" #
−1/c2 −v j /c2
g ab =
−v i /c2 hij − v i v j /c2
We do however have the simplification that all quantities are time-independent. (All
time derivatives vanish). Note that the ADM decomposition is in this case quite
distinct from the reverse-ADM decomposition.
We still have, thanks to Dirac
1
Rabcd = {gad,bc − gbd,ac − gac,bd + gbc,ad } + Γead g ef Γf bc − Γeac g ef Γf bd (2.105)
2
43
But now when we insert the inverse metric
1
Rabcd = {gad,bc − gbd,ac − gac,bd + gbc,ad }
2
1 1
− 2 Γtad Γtbc + 2 Γtac Γtbd
c c
−(v /c ){Γtad Γmbc + Γmad Γtbc } + (v m /c2 ){Γmac Γtbd + Γtac Γmbd }
m 2
+(hmn − v m v n /c2 ) Γmad Γnbc − (hmn − v m v n /c2 ) Γmac Γnbd (2.106)
Rijkl : Start by looking at the space components Rijkl . We have
1
Rijkl = {gil,jk − gjl,ik − gik,lj + gjk,il}
2
1 1
− 2 Γtil Γtjk + 2 Γtik Γtjl
c c
−(v m /c2 ){Γtil Γmjk + Γmil Γtjk } + (v m /c2 ){Γmik Γtjl + Γtik Γmjl }
−(v m v n /c2 ) Γmil Γnjk − (v m v n /c2 ) Γmik Γnjl
+hmn Γmil Γnjk − hmn Γmik Γnjl (2.107)
But most of this is simply the 3-dimensional Riemann tensor, and the rest can easily
be regrouped as
(3) 1 1
Rijkl = (Γtil + v m Γmil ) (Γtjk + v n Γnjk ) + 2 (Γtik + v m Γmik ) (Γtjl + v n Γnjl )
Rijkl −
2
c c
(2.108)
Now suppose we define
1
Kij = {Γtij + v m Γmij } (2.109)
c
This is a natural generalization of our previous definition, and would allow us to again
write
(3)
but now for this “improved” definition of Kij . Let’s check what happens. Having de-
layed things as long as possible, we’ll have to start by calculating Christoffel symbols.
44
2.8.1 Christoffel symbols of first kind
(Remember, stationary, no t dependence):
1 1
Γttt = gt(t,t) − gtt,t = gtt,t = 0 (2.111)
2 2
1 1 1
Γitt = gi(t,t) − gtt,t = − gtt,i = + (c2 − v 2 ),i (2.112)
2 2 2
1 1 1
Γtti = Γtit = gt(t,i) − gti,t = gtt,i = − (c2 − v 2 ),i (2.113)
2 2 2
1 1
Γtij = gt(i,j) − gij,t = + (gti,j + gtj,i) (2.114)
2 2
1
= − (vi,j + vj,i ) = −v(i,j)
2
1 1
Γitj = Γijt = gi(t,j) − gtj,i = (git,j − gtj,i)
2 2
1
= − (vi,j − vj,i) = −v[i,j] (2.115)
2
1 1
Γijk = gi(j,k) − gjk,i = hi(j,k) − hjk,i = (3) Γijk (2.116)
2 2
Then we have
1 1 1
{Γtij + v m Γmij } = −v(i,j) + vm Γm ij = − v(i:j)

(2.117)
c c c
so that in this situation
1
Kij = − v(i:j) (2.118)
c
This is compatible with the general definition of extrinstic curvature, see [17] equation
(21.67), note that our v i is minus the shift vector.
Rtijk : In a similar manner we can evaluate Rtijk as

1
Rtijk = {gtk,ij − gik,tj − gtj,ik + gij,tk }
2
1 1
− 2 Γttk Γtij + 2 Γttj Γtik
c c
−(v /c ) {Γttk Γmij + Γmtk Γtij } + (v m /c2 ) {Γmtj Γtik + Γttj Γmik }
m 2
+(hmn − v m v n /c2 ) Γmtk Γnij − (hmn − v m v n /c2 ) Γmtj Γnik (2.119)
45
1
Rtijk = {−vk,ij + vj,ik }
2
1 1
− 2 (Γttk + v m Γmtk ) (Γtij + v m Γmij ) + 2 (Γttj + v m Γmtj ) (Γtik + v m Γmik )
c c
+hmn Γmtk Γnij − hmn Γmtj Γnik (2.120)
Now
1 1 1 1
Γttk + v m Γmtk = − (c2 − v 2 ),k − v m (vm,k − vk,m ) = − (c2 − v 2 ):k − v m (vm:k − vk:m)
2 2 2 2
(2.121)
so that
1 1
Γttk + v m Γmtk = − (c2 ),k + v m (vm:k + vk:m ) = −c {c,k + v m Kmk } (2.122)
2 2
So now
1
Rtijk = {−vk,ij + vj,ik } − (c,k + v m Kmk )Kij + (c,j + v m kmj )Kik − v[m,k] Γm ij + v[m,j] Γm ik
2
(2.123)
This can be rewritten in several ways: If we remain in a coordinate basis
Rtijk = v[j,k],i − v[m,k] Γm ji − v[j,m] Γm ki − (c,k + v m Kmk )Kij + (c,j + v m Kmj )Kik (2.124)
whence
Rtijk = v[j:k]:i − (c,k + v m Kmk )Kij + (c,j + v m Kmj )Kik (2.125)
This is relatively compact, but not so easy to compare to our previous result.
But we can also write
2v[j:k]:i = vj:k:i − vk:j:i = vj:i:k + vm (3) Rm jki − vk:i:j − vm (3) Rm kji

(3)
Rm jki − (3)
Rm kji − (3)
Rm ijk

= vj:i:k + vi:j:k − vk:i:j − vi:k:j + vm
= 2 v(i:j):k − v(i:k):j − vm (3) Rm ijk

(3)
Rm ijk

= 2 [−cKij ]:k − [−cKik ]:j − vm (2.126)
which lets us write
Rtijk = c {[Kik ]:k − [Kij ]:k } − v m

(3)
Rmijk − Kmk Kij + Kmj Kik (2.127)
46
that is, in view of our previous result,
Rtijk = c[Kik:j − Kij:k ] − v m Rmijk (2.128)
This is the tidiest we can make things in a coordinate basis. This is the analogue of
[17], equation (21.76).
For future reference define
Ωij = v[i,j] = v[i:j] (2.129)
and note the identity
Ωjk:i = [cKik ]:j − [cKij ]:k − vm (3) Rm ijk (2.130)
Note that it sometimes easier to work in what is referred to as a “non-coordinate basis”
by defining:
Rnijk ≡ na Raijk (2.131)
in which case
1 1
Rnijk = (1; v m) (Rtijk ; Rmijk ) = [Rtijk + v m Rmijk ] = Kik:j − Kij:k . (2.132)
c c
This means you are not really looking in the “t” direction, but are instead looking in
the “n” direction normal to “space”.
Rtitj : Now for the Rtitj pieces. This is where things might get a little messy. From
Dirac’s general formula, applied to the present situation
1
Rtitj = {gtj,it − gij,tt − gtt,ij + git,tj }
2
1 1
− 2 Γttj Γtit + 2 Γttt Γtij
c c
−(v m /c2 ) {Γttj Γmit + Γmtj Γtit } + (v m /c2 ) {Γmtt Γtij + Γttt Γmij }
+(hmn − v m v n /c2 ) Γmtj Γnit − (hmn − v m v n /c2 ) Γmtt Γnij (2.133)
Because of time independence this greatly simplifies
1 1
Rtitj = {−gtt,ij } − 2 {Γttj Γtit }
2 c
m 2
−(v /c ) {Γttj Γmit + Γmtj Γtit − Γmtt Γtij }
+(hmn − v m v n /c2 )Γmtj Γnit − (hmn − v m v n /c2 )Γmtt Γnij (2.134)
47
so that
1 1
Rtitj = {−gtt,ij } − 2 (Γttj + v m Γmtj ) (Γtit + v m Γmit )
2 c
+(v /c )Γmtt Γtij + hmn Γmtj Γnit − hmn Γmtt Γnij
m 2
(2.135)
Inserting the known values of the Christoffel symbols (calculated above)
1 2 1
Rtitj = (c − v 2 ),ij − (c,j + v m Kmj )(c,i + v m Kmi ) − (v m /c2 ) (c2 − v 2 ),m v(i,j)
2 2
1
+hmn v[m,j] v[n,i] − hmn (c2 − v 2 ),m Γnij (2.136)
2
That is
1 2 1
Rtitj = (c − v 2 ):ij − (c,j + v m Kmj )(c,i + v m Kmi ) + (v m /c) (c2 − v 2 ),m Kij
2 2
+hmn v[m,j] v[n,i] (2.137)
But then
1 2
Rtitj = c c:ij − (v ):ij − v m [c,j Kmi + c,i Kmj ] − v m Kmj v n Kni
2
1
+v m c,m Kij − (v m /c) (v 2 ),m Kij + hmn v[m,j] v[n,i] (2.138)
2
Which still needs to be simplified further.
For instance
1 2 1
(v ):ij = (v k vk ):ij = (v k vk:i):j = (v k :j vk:i ) + (v k vk:i:j ) (2.139)
2 2
But
v:jk vk:i = hkl vk:j vk:i = hkl (−cKkj + v[k:j]) (−cKli + v[l:i] )
= c2 (K 2 )ij − c(KΩ − ΩK)ij + hmn v[m,j] v[n,i] (2.140)
where with obvious notation Ωij = v[i,j] and
(KΩ − ΩK)ij = Kik g kl Ωlj − Ωik g kl Klj (2.141)
48
Aside: Write (K Ω − Ω K)ij as
Kg −1 Ω − Ω g −1 K
then
(Kg −1 Ω − Ω g −1 K)T = ΩT [g −1 ]T K T − K T [g −1 ]T ΩT
= −Ω g −1 K + K g −1 Ω (2.142)
so this object is symmetric, as long as the minus sign is included.

Then
Rtitj = c c:ij − (v k vk:i:j ) − c2 (K 2 )ij + c(KΩ − ΩK)ij

−v m [c,j Kmi + c,iKmj ] − v m Kmj v n Kni
+v m c,m Kij − (v m /c) (v k vk:m ) Kij (2.143)
so that
Rtitj = c c:ij − (v k vk:i:j ) − c2 (K 2 )ij + c(KΩ − ΩK)ij

−v m [c,j Kmi + c,iKmj ] − v m Kmj v n Kni
+v m c,m Kij + (v k v l Kkl ) Kij (2.144)
Now by the Ricci identities the combination v k vk:i:j is symmetric in ij, so that we can
write:
v k vk:i:j = v k vk:(i:j) (2.145)
But now consider

1 1
vk:(i:j) = {vk:i:j + vk:j:i} = {(−cKki + Ωki ):j + (−cKkj + Ωkj ):i }
2 2
1 1
= − {(cKki ):j + (cKkj ):i} + {Ωki:j + Ωkj:i} (2.146)
2 2
Then recall the identity
(3)
Ωjk:i = [cKik ]:j − [cKij ]:k − vm Rm ijk (2.147)
49
to deduce
1 (3)
Rm ijk + (3)
Rm jik

vk:(i:j) = [cKij ]:k + vm (2.148)
2
whence
v k vk:(i:j) = v k [cKij ]:k − v k v l (3)
Rkilj (2.149)
Putting this back into Rtitj we have
Rtitj = c c:ij − v k [cKij ]:k + v k v l (3)

Rkilj − c2 (K 2 )ij + c(KΩ − ΩK)ij
−v m Kmj v n Kni + v m c,m Kij + (v k v l Kkl ) Kij (2.150)
which simplifies to
Rtitj = c c:ij − v k cKij:k − c2 (K 2 )ij + c(KΩ − ΩK)ij

−v m Kmj v n Kni + (v k v l Kkl ) Kij + v k v l (3)
Rkilj (2.151)
We can rewrite the last line in terms of the 4-curvature as
Rtitj = c c:ij − v k cKij:k − c2 (K 2 )ij + c(KΩ − ΩK)ij

+v k v l Rkilj (2.152)
This is the simplest expression we can get in a coordinate basis.
Rninj : Now define in a “non-coordinate basis”.

1
Rninj ≡ na nb Raibj = (1; v i)a (1; v l )b Raibj
c2
1 k k k l

= Rtitj + v R tikj + v R kitj + v v R kilj
c2
1 k k k l

= Rtitj − cv [K ik:j − K ij:k ] − cv [K jk:i − K ij:k ] − v v Rkilj
c2
1 k k l

= Rtitj − cv [K ik:j + K jk:i − 2K ij:k ] − v v R kilj (2.153)
c2
Then
c2 Rninj = c c:ij − v k c Kij:k − c2 (K 2 )ij + c(KΩ − ΩK)ij

−c v k [Kik:j + Kjk:i − 2Kij:k ] (2.154)
50
There are now some simplifications
c2 Rninj = c c:ij + c v k Kij:k − c2 (K 2 )ij + c(KΩ − ΩK)ij
−c v k [Kik:j + Kjk:i] (2.155)
To simplify this we consider the Lie derivative
Lv Kij = v k ∂k Kij + v k ,i Kkj + v k ,j Kik (2.156)
which can also be written in terms of 3-covariant derivatives as
Lv Kij = v k Kij:k + v k :i Kkj + v k ;j Kik (2.157)
so that
Lv Kij = v k Kij:k + hkl [(−cKki + Ωki )Klj + (−cKkj + Ωkj )Kil ] (2.158)
and
Lv Kij = v k Kij:k − 2c (K 2 )ij + (KΩ − ΩK)ij (2.159)
Therefore
c2 Rninj = c c:ij + cLv Kij + c2 (K 2 )ij − c v k [Kik:j + Kjk:i] (2.160)
Summary :
For an arbitrary stationary spacetime. taking
1
Kij = − v(i:j)
c
we have
(3)
Rtijk = c [Kik:j − Kij:k ] − v m Rmijk (2.162)
Rtitj = c c:ij − v k c Kij:k − c2 (K 2 )ij + c (KΩ − ΩK)ij + v k v l Rkilj (2.163)
We can also write this as
Rnijk ≡ na Raijk = Kij:k − Kik:j . (2.164)
and
Rninj ≡ na nb Raibj = c c:ij + cLv Kij + c2 (K 2 )ij − c v k [Kik:j + Kjk:i]. (2.165)
51
2.8.2 Completely general ADM decomposition
For the general ADM case coordinate techniques are too messy to contemplate. The
easy results are on page 514 of [17] and boil down to

1 1
Kij = − ∂t gij + v(i:j) (2.166)
c 2
note that Kij is now a combination of the zero-shift and stationary case.
Then
(3)
and
Rnijk ≡ na Raijk = Kij:k − Kik:j (2.168)
The difficult one is then encoded in page 518, [17] which seems to assert that
Rnink ≡ na nb Raibj = [Ln K]ij + [K 2 ]ij + a(i;j) + ai aj (2.169)
where ai is the space projection of the 4-acceleration a = (n · ∇) n.
2.8.3 The inverse-ADM decomposition

We can decompose the metric as follows
" #
−1/N 2 vi /N 2
gµν = (2.170)
vi /N 2 hij − vi vj /N 2
with
N = N(t, x, y, z), vi = vi (t, x, y, z), and hij = hij (t, x, y, z).
The inverse metric is now

" #
−[N 2 − hij vi vj ] v i
g µν = (2.171)
vi hij
where v i = hij vj and hij = [hij ]−1
52
This is the reverse of ADM formulation
   
g11 g12 g13 g14 −1/N 2 v2 /N 2 v3 /N 2 v4 /N 2
2
h22 − v2 v2 /N 2 h23 − v2 v3 /N 2 h24 − v2 v4 /N 2
   
 g21 g22 g23 g24 
 =  v2 /N
 
gµν =  2

 g
 31 g32 g33   v3 /N
g34   h32 − v3 v2 /N 2 h33 − v3 v3 /N 2 h34 − v3 v4 /N 2 

g41 g42 g43 g44 v4 /N 2 h42 − v4 v2 /N 2 h43 − v4 v3 /N 2 h44 − v4 v4 /N 2
(2.172)
and  P4 
−[N 2 − ij
i,j=1 h vi vj ] v2 v3 v4
v2 h22 h23 h24
 
g µν
 
=  (2.173)

 v3 h32 h33 h34 

v4 h42 h43 h44
For the time being, let vi = 0 then
   
g11 g12 g13 g14 −1/N 2 0 0 0
   
 g21 g22 g23 g24   0 h22 h23 h24 
gµν = 
 g
=  (2.174)
 31 g32 g33 g34  
  0 h32 h33 h34 

g41 g42 g43 g44 0 h42 h43 h44
and
   
g 11 g12
g13
g 14
−N 2
0 0 0
g 21 g 22 g 23 g 24 22 23
h24
   
µν
   0 h h 
g = =  (2.175)

 g 31 g 32 g 33 g 34  
  0 h32 h33 h34 

g 41 g 42 g 43 g 44 0 h42 h43 h44
The 4 dimensional metric can be spilt up into one 3 × 3 block, and two 1 × 3 blocks
with two of them trivial (zero). That is
" #
−1/N 2 0
gµν = (2.176)
0 hij
The inverse tensor is very simple

" #
−N 2 0
g µν = (2.177)
0 hij
53
where −N 2 and hij are two dimensional inverses of the 1 × 1 matrix and 3 × 3 matrices
−1/N 2 and hij , respectively.
The Christoffel symbols can be computed from
1
Γµ σρ = g µν {gν(σ,ρ) − gσρ,ν } (2.178)
2
and you can spilt them up into little sub-blocks
 
Γµ 11 Γµ 12 Γµ 13 Γµ 14
 µ
 Γ 12 Γµ 22 Γµ 23 Γµ 24

Γµ σρ = 

 Γµ µ µ
 (2.179)
 13 Γ 23 Γ 33 Γµ 34 

Γµ 14 Γµ 24 Γµ 34 Γµ 44
54
2.8.4 Christoffel symbols of the 2 metrics −1/N 2 and hij
Some of these blocks should have a simple interpretation in terms of Christoffel sym-
bols of the lower dimensional metrics −1/N 2 and hij .
When t = time and i, j = space, for example
1 1
Γt tt = g tν {gν(t,t) − gtt,ν } = g tt {gt(t,t) − gtt,t }
2 2
1 1 1
= g tt { gtt,t + gtt,t − gtt,t }
2 2 2
1 1 1 Ṅ
= g tt { gtt,t } = − N 2 {(− 2 ),t } = − (2.180)
2 2 N N
1 1
Γt ij = g tν {gν(i,j) − gij,ν } = g tt {gt(i,j) − gij,t }
2 2
tt 1 2 1 1 2
= g {0 − hij,t } = N { hij,t } = N {hij,t} (2.181)
2 2 2
1 1
Γt ti = g tν {gν(t,i) − gti,ν } = g tt {gt(t,i) − gti,t }
2 2
1 1 1
= g tt { gtt,i + gti,t − gti,t }
2 2 2
1 1 1 ∂i N
= g tt { gtt,i } = N 2 {(− 2 ),i} = (2.182)
2 2 N N
1 1
Γi tt = g iν {gν(t,t) − gtt,ν } = g ij {gj(t,t) − gtt,j }
2 2
1 1 1 ∂j N
= g ij {0 − htt,j } = − hij {(− 2 ),j } = −hij (2.183)
2 2 N N
1 1
Γi jk = g iν {gν(j,k) − gjk,ν } = g il {gl(j,k) − gjk,l}
2 2
1
= hil {hl(j,k) − hjk,l } = [Γ(h)]i jk (2.184)
2
1 1
Γi jt = g iν {gν(j,t) − gjt,ν } = g ik {gk(j,t) − gjt,k }
2 2
1 1 1
= g ik { gkj,t + gkt,j − gjt,k }
2 2 2
1 ik 1 ik
= g gkj,t = h hkj,t (2.185)
2 2
55
2.8.5 Collected results for the connection
There are a total of 6 sub-blocks to deal with:
1 1 Ṅ
Γt tt = − N 2 {(− 2 ),t } = − (2.186)
2 N N
1
Γt ij = N 2 { hij,t } (2.187)
2
1 1 ∂i N
Γt ti = N 2 {(− 2 ),i } = (2.188)
2 N N
1 1 ∂i N
Γi tt = − hij {(− 2 ),j } = −hij (2.189)
2 N N
i i
Γ jk = [Γ(h)] jk (2.190)
1 ik
Γi jt = h hkj,t (2.191)
2
2.9 Conclusion
In this section, we analysed a generic spacetime using both ADM and reverse-ADM
(Kaluza-Klein) decompositions which split “spacetime” into “space” plus “time” in a
natural way. We use the split to describe spacetime curvature in terms of the curvature
of space and some additional structure. Furthermore, this decomposition encodes the
four-metric of spacetime in terms of a time-varying three-metric on a three-dimensional
domain together with the lapse and shift.
In addition, we used the Gauss-Codazzi and Gauss-Weingarten equations to decom-
pose the (3+1)-dimensional spacetime curvature tensor in terms of the 3-dimensional
spatial curvature tensor, the extrinsic curvature of the time slice [zero], and the grav-
itational potential. The ADM and reverse-ADM decompositions are identical in any
static spacetime. In addition, we derived the standard results in equation (2.64) to
(2.66) for the spacetime geometry in terms of the three geometry of space on a con-
stant time slice, in which this three geometry is embedded into the spacetime. Finally,
we derived equation (2.71) to (2.73) for the spacetime Einstein tensor.
We consider time dependent but block diagonal metrics in the ADM and reverse-
ADM decompositions. When the “shift vector” is zero, the ADM and reverse-ADM
decompositions are identical. Futhermore, we defined a quantity called the extrinsic
56
curvature in equation (2.75). We are able to calculate the Christoffel symbols of
the first kind and second kind from section (2.7.1). Furthermore, we calculated the
Riemann tensor by using the formula from (2.91) to (2.93). Whenever the metric gab
block diagonalizes in the manner gab = −c2 hij , then even if we allow arbitrary time
L
dependence we have the relatively simple results as equation (2.102) to (2.104).

For a general stationary spacetime we need to use the full ADM decomposition.
However, we have the simplification that all quantities are time independent. Indeed,
the case for ADM decomposition is quite distinct from the reverse-ADM decomposi-
tion. The Christoffel symbols first kind are given in section (2.8.1). We derived the
simplest expression for the Riemann tensor in terms of a coordinate basis.
For the general ADM case coordinate techinques are too complicated to contem-
plate (page 514 of [17]). Using the inverse-ADM decomposition we can calculate the
Christoffel symbols from equation (2.178). Furthermore, we can split them up into sub-
blocks in equation (2.179). Some of these blocks can interpreted in terms of Christoffel
symbols of the 2 metrics −1/N 2 and hij see (2.8.4).
57
58
Chapter 3
Perfect fluid spheres in general

relativity
3.1 Introduction
Perfect fluid spheres are interesting because they are first approximations to any at-
tempt at building a realistic model for a general relativistic star [9, 11, 15, 20].
Although they represent a real and important astrophysical situation, explicit and
full general solutions of the perfect fluid constraints in general relativity have only
very recently been developed.
The first static spherically symmetric perfect fluid solution with constant density
was found by Schwarzschild in 1918.
He found two exact solutions to Einstein’s field equation — the “exterior solution”
(relevant outside the star) and the “interior solution” (an approximation to what
goes an inside the star). The so-called Schwarzschild solution is amazingly simple,
especially in view of the field equations which are very complicated. Indeed, the
exterior Schwarzschild solution is not a degenerate case for over-simplified situations
but physically most meaningful. It is this solution by means of which one can explain
most general relativistic effects in the planetary system. The reason is that it describes
the gravitational field outside of a spherically symmetric body like the planets and the
sun.
59
Over the past 90 years a confusing tangle of specific perfect fluid spheres has been
discovered, with most of these examples seemingly independent from each other.
Many “ad hoc” tricks have been explored to solve this differential equation, often
by picking special coordinate systems, or making simple ansatze for one or other of
the metric components. The big change over the last several years has been the
introduction of “algorithmic” techniques that permit one to generate large classess of
perfect fluid spheres in a purely mechanical way. Perfect fluid spheres may be “simple”,
but they still provide a surprisingly rich mathematical and physical structure.
In this chapter we will extend these algorithmic ideas, by proving several solution-
generating theorems of varying levels of complexity. We shall then explore the formal
properties of these solution-generating theorems and then will use these theorems to
classify some of the previously known exact solution, and additionally will generate
several previously unknown perfect fluid solutions.
3.2 Six theorems for generating perfect fluid spheres

Perfect fluid spheres, either Newtonian or relativistic, are the first approximations
in developing realistic stellar models. Because of the importance of these models,
fully general solutions of the perfect fluid constraint in general relativity have been
developed over the last several years.
The central idea is to start solely with spherical symmetry, which implies that in
orthonormal components the stress energy tensor takes the form
 
ρ 0 0 0
 0 p 0 0 
r
Tâb̂ =  (3.1)
 

 0 0 pt 0 
0 0 0 pt
and then use the perfect fluid constraint pr = pt . This is making the radial pressure
equal to the transverse pressure. By using the Einstein equations, plus spherical
symmetry, the equality pr = pt for the pressures becomes the statement
Gθ̂θ̂ = Gr̂r̂ = Gφ̂φ̂ . (3.2)
60
In terms of the metric components, this leads to an ordinary differential equation
[ODE], which then constrains the spacetime geometry, for any perfect fluid sphere.
Over the last 90 years, many “ad hoc” approaches to solving this differential equa-
tion have been explored. We often solve it by picking special coordinate systems, or
making simple ansatze for one or other of the metric components [3, 4, 5, 7, 8, 25]. (For
recent overviews see [9, 11, 20]. For a table of “well-known” perfect fluid solutions see
Table 3.2 on page 99.) The big recent change has been the discovery of “algorithmic”
techniques that permit one to generate large classes of perfect fluid spheres in a purely
mechanical way [13, 16, 18]. In addition, an alternative algorithmic approach is used
to generate any spherically symmetric static solution of Einstein’s equation without
recourse to the details of the material that makes it up.
Solution generating theorems
Algorithmic techniques
We would like to present “Algorithmic” techniques [16] that permit one to generate
large classes of perfect fluid spheres in a purely mechanical way. The condition of
hydrostatic eqilibrium for Newtonian gravity is
dPr m(r)ρ(r)
= −g(r)ρ(r) = −
dr r2
where g(r) is the local acceleration due to gravity, m(r) is the mass contained within
a radius r and ρ(r) is the density.
We can simply calculate the mass within a radius r in terms of the density:
Z r
m(r) = 4π r 2 ρ(r) dr.
0
When we used ρ(r) as a known function, we have to solve a second–order differential

equation:
d2 Pr

2 1 dρ dPr
2
+ − + 4πρ2 = 0
dr r ρ dr dr
61
We can solve for Pr (r) in terms of an integral of the density:
C1 − 4π r2 ρ(r) dr
Z R
Pr (r) = ρ(r) dr + C2 .
r2
We let g(r) be an arbitrary function to solve simple algebra which allows us to find
m(r). The differentiation allows us to find ρ(r) and then Pr (r), respectively:
Z r
2
m(r) = r g(r) = 4π r 2 ρ(r) dr
0
dg
2rg(r) + r 2
⇒ ρ(r) = dr
4πr 2
2
1 g(r) 1
Z
⇒ Pr (r) = − dr − g(r)2 + Pr (0).
2π r 8π
In both the Newtonian and relatistic cases we have to ensure that the density is non-
negative. The pressure is finite and non-negative.
Spherical symmetry
Spherically symmetric means “having the same symmetry as a sphere”. Indeed,
“sphere” means S 2 , not spheres of higher dimension. Rather than the metric on a
differentiable manifold, we are concerned with those metrics that have such symme-
tries. We can characterize symmetries of the metric. By the existence of Killing
vectors, which we can see from the definition below. By demanding spherical sym-
metry we can simplify the equations of motion considerably. It turns out that for
vaccuum spherical symmetry of the metric automatically means the metric is static
and asymptotically flat. This is Birkhoff’s theorem, a proof of which can be found in
[19]. Furthermore, Spherical symmetry can be defined rigorously in terms of Killing
vector fields as follows.
Definition 3. A space-time is said to be spherically symmetric if and only if it admits

three linearly independent spacelike Killing vector fields X a whose orbits are closed
(i.e. topological circles) and which satisfy
[X 1 , X 2 ] = X 3 , [X 2 , X 3 ] = X 1 , [X 3 , X 1 ] = X 2
62
Then there exists a coordinate system in which the Killing vectors take on a stan-
dard form as expressed in the following result.
Proposition 1. In a spherically symmetric space-time, there exists a coordinate sys-

tem (xa ) (called quasi Cartesian) in which the Killing fields X a are of the form
X 0 → 0,
X α → ω α β xβ , ωαβ = −ωβα .
In addition, we know that what the Killing vectors of S 2 are, and that there are
three of them. Therefore, a spherically symmetric manifold is one that has three
Killing vector fields which are the commutator of the Killing vectors which is the same
in either case of those on S 2 .
Schwarzschild solution
The gravitational field of a homogenous spherically symmetric body is derived in
firstly Newton’s and then in Einstein’s gravitational theory. The gravitational field of
a simple Einsteinian model star consists of the exterior and the interior Schwarzschild
solutions. They are joined together at the surface of the star.
The use of arbitrary coordinates is permitted in general relativity. Indeed, the
physical significance of statements about tensor or vector components are other quan-
tities which are not always obvious. However, there are some situations where the
interpretation is almost as straightforward as in special relativity. The center point of
a local inertial coordinate system is the most obvious example. Its coordinate system,
where the principle of equivalence comes from, allows one to treat all local quantities
exactly as in special relativity. Furthermore, Schwarzschild coordinates for a spherical
system turn out to be a second example [17].
Schwarzschild geometry
The geometry of a spherical symmetric vacuum, i.e. vacuum spacetime outside the
spherical black hole is the “Schwarzschild geometry” can be described in terms of the
63
Schwarzschild metric,
dr 2

2 2M
ds = − 1 − dt2 + + r 2 dΩ2 , (3.3)
r 1 − 2M/r
which was derived originally as the external field of a static star. The Schwarzschild
metric seems to have a singularity at the surface where r = 2M due to its coordinates,
in which space and time change their meanings.
Schwarzschild geometry illustrates clearly the highly non-Euclidean character of
spacetime geometry when gravity becomes strong. Furthermore, it illustrates many
of techniques one can use to analyze strong gravitational fields. When appropriately
truncated, it is the spacetime geometry of a black hole and of a collapsing star as well
as of a wormhole [17].
However, that the Schwarzschild exterior solution is the unique spherically sym-
metric vacuum solution, is known as Birkhoff ’s theorem. This is a theorem of general
relativity which states that all spherical gravitaional fields, whether from a star or
from a black hole, are indistinguishable at a large distances. A consequence of this is
that purely radial changes in a spherical star do not affect its external gravitational
field.
That the Schwarzschild geometry is relevant to gravitational collapse follows from
“Birkhoff’s theorem”.
Birkhoff ’s theorem: Let the geometry of a given region of spacetime be spherically
symmetric, and be a solution to the Einstein field equations in vacuum. Then that
geometry is necessarily a piece of the Schwarzschild geometry [17].
In particular, Birkhoff ’s theorem implies that if a spherically symmetric source like
a star changes its size, however does so always remaining spherically symmetric, then
it cannot propagate any disturbances into the surrounding space. From figure 3.1,
shows that a pulsating spherically symmetric star cannot emit gravitational waves. If
a spherically symmetric source is restricted to the region r 6 a for some a, then the so-
lution for r > a must be the Schwarzschild solution, or can be called the Schwarzschild
exterior solution. However, the converse is not true: a source which gives rise to an
exterior Schwarzschild solution is not necessarily spherically symmetric. Some counter-
examples are known. Therefore, in general, a source need not inherit the symmetry of
64
its external field. If we take the limit of equation 3.3 as r → ∞, then the flat space
Minkowski metric of special relativity in spherical polar coordinates is obtained, which
is
ds2 = −dt2 + dr 2 + r 2 (dθ2 + sin2 θ dφ2 ). (3.4)
This shows that a spherically symmetric vacuum solution is necessarily asymptotically
flat [19].
Figure 3.1: A Pulsating spherical star cannot emit gravitational waves.
Firstly, start with some static spherically symmetric geometry in Schwarzschild

(curvature) coordinates
dr 2
ds2 = −ζ(r)2 dt2 + + r 2 dΩ2 (3.5)
B(r)
and then assume it represents a perfect fluid sphere. That is, Gθ̂θ̂ = Gr̂r̂ = Gφ̂φ̂ . While
Gθ̂θ̂ = Gφ̂φ̂ is always fulfilled due to spherical symmetry, we get a nontrivial constraint
from Gθ̂θ̂ = Gr̂r̂ .
We calculate
2Brζ ′ − ζ + ζB
Gr̂r̂ = − (3.6)
r2ζ
65
and
1 B ′ ζ + 2Bζ ′ + 2Brζ ′′ + rζ ′B ′
Gθ̂θ̂ =− (3.7)
2 rζ
We set Gr̂r̂ = Gθ̂θ̂ , which supplies us with an ODE:
[r(rζ)′]B ′ + [2r 2 ζ ′′ − 2(rζ)′]B + 2ζ = 0 . (3.8)
This reduces the freedom to choose the two functions in equation (3.5) to one.
This equation is a first order-linear non-homogeneous equation in B(r). Thus, once
you have chosen a ζ(r) — this equation can always be solved for B(r). Solving for
B(r) in terms of ζ(r) is the basis of [13, 16], (and is the basis for Theorem 1 below).
On the other hand, we can also re-group this same equation as
2r 2 Bζ ′′ + (r 2 B ′ − 2rB)ζ ′ + (rB ′ − 2B + 2)ζ = 0 , (3.9)
which is a linear homogeneous 2nd order ODE for ζ(r), which will become the basis
for Theorem 2 below. Thus, once you have chosen a B(r) — this equation can always
be solved for ζ(r). The question we are going to answer in this section is, how can one
systematically “deform” this geometry while still retaining the perfect fluid property?
That is, suppose we start with the specific geometry defined by
dr 2
ds2 = −ζ0 (r)2 dt2 + + r 2 dΩ2 (3.10)
B0 (r)
and assume it represents a perfect fluid sphere. In the following section, we will show
how to “deform” this solution by applying five different transformation theorems on
{ζ0 , B0 }, such that the outcome still presents a perfect fluid sphere. The outcome
of this process will depend on one or more free parameters, and so automatically
generates an entire family of perfect fluid spheres of which the original starting point
is only one member.
In addition, we analyze what happens if we apply these theorems more than once,
iterating them in various ways. We also try to find the connection between all six
different transformation theorems.
66
3.2.1 The first theorem
The first theorem we present is a variant of a result first explicitly published in [16],
though another variant of this result can also be found in [20]. We first re-phrase the
theorem in [16]. This is slightly different formalism, and demonstrate an independent
way of proving it.
Definition (Idempotence): We call a transformation T idempotent if T2 = T.

Speaking loosely, we will call a theorem idempotent if whenever we apply the trans-
formation more than once, no further solutions will be obtained.
We use our proof to show that the transformation in theorem 1 is (in a certain sense
to be made precise below) “idempotent”. By applying theorem 1 more than once, no
new solutions will be found. Furthermore, the idempotence property of theorem 1 and
theorem 2, which we will soon establish, enables us to divide the class of perfect fluid
spheres into seed metrics and non-seed metrics.
Theorem 1. Suppose {ζ0 (r), B0 (r)} represents a perfect fluid sphere. Define
2 Z ′
ζ0 (r) ζ0 (r) − r ζ0′ (r)

ζ0 (r) 2
∆0 (r) = r exp 2 dr . (3.11)
ζ0 (r) + r ζ0′ (r) ζ0 (r) ζ0 (r) + r ζ0′ (r)
Then for all λ, the geometry defined by holding ζ0 (r) fixed and setting
dr 2
ds2 = −ζ0 (r)2 dt2 + + r 2 dΩ2 (3.12)
B0 (r) + λ ∆0 (r)
is also a perfect fluid sphere. That is, the mapping
T1 (λ) : {ζ0 , B0 } 7→ {ζ0 , B0 + λ∆0 (ζ0 )} (3.13)
takes perfect fluid spheres into perfect fluid spheres. Furthermore a second application
P
of the transformation does not yield new information, T1 = T1 ( λi ) is “idempotent”,
in the sense that
X n Xn o
T1 (λn ) ◦ · · · ◦ T1 (λ2 ) ◦ T1 (λ1 ) = T1 ( λi ) : {ζ0 , B0 } 7→ ζ0 , B0 + λi ∆0 (ζ0 )
i=1
(3.14)
67
We also note that T1 always has an inverse
[T1 (λ)]−1 = T1 (−λ). (3.15)
Proof for Theorem 1. Assume that {ζ0 (r), B0 (r)} is a solution for equation (3.8).
We want to find under what conditions does {ζ0 (r), B1(r)} also satisfy equation
(3.8)? Without loss of generality, we write
B1 (r) = B0 (r) + λ ∆0 (r) . (3.16)
Equation (3.8) can now be used to determine ∆0 (r). Substitute B1 (r) in equation
(3.8)
[r(rζ0)′ ](B0 + λ ∆0 )′ + [2r 2 ζ0′′ − 2(rζ0 )′ ](B0 + λ ∆0 ) + 2ζ0 = 0
we can also re-group this same equation as
[r(rζ0)′ ]B0′ + [2r 2 ζ0′′ − 2(rζ0 )′ ]B0 + 2ζ0 + [r(rζ0)′ ] ∆′0 + 2r 2ζ0′′ − 2(rζ0)′ ∆0 = 0

That ordinary inhomogeneous first-order differential equation in B now simplifies to
[r(rζ0)′ ] ∆′0 + 2r 2 ζ0′′ − 2(rζ0 )′ ∆0 = 0 ,

(3.17)
which is an ordinary homogeneous first-order differential equation in ∆0 .

Now we can calculate ∆0 by following,
[r(rζ0 )′ ] ∆′0 = − 2r 2 ζ0′′ − 2(rζ0 )′ ∆0 .

(3.18)
Rearrange it into
∆′0 − [2r 2 ζ0′′ − 2(rζ0 )′ ]

=
∆0 [r(rζ0 )′ ]
−2rζ0′′ 2
= + (3.19)
(rζ0)′ r
A straightforward calculation, including an integration by parts, leads to
r2 4ζ0′
Z
∆0 (r) = exp dr . (3.20)
[(rζ0)′ ]2 (rζ0)′
68
We can also write this same equation as
r2
Z ′
ζ0 ζ0
∆0 (r) = exp 4 dr
[(rζ0 )′ ]2 ζ0 (ζ0 + rζ0′ )
r2
Z ′ Z ′
ζ0 ζ0 ζ0 ζ0
= exp 2 dr + 2 dr . (3.21)
[(rζ0 )′ ]2 ζ0 (ζ0 + rζ0′ ) ζ0 (ζ0 + rζ0′ )
Adding and subtracting ±2(rζ0′ )/(ζ0 (rζ0 )′ ) to the argument under the integral leads
to
r2
Z ′ Z ′
rζ0′

ζ0 ζ0 ζ0
∆0 (r) = exp 2 dr − 2 dr
[(rζ0 )′ ]2 ζ0 (ζ0 + rζ0′ ) ζ0 (ζ0 + rζ0′ )
Z ′ Z ′
rζ0′

ζ0 ζ0 ζ0
+ exp 2 dr + 2 dr .
ζ0 (ζ0 + rζ0′ ) ζ0 (ζ0 + rζ0′ )
r2
Z ′
ζ0 (ζ0 − rζ0′ )
Z ′
ζ0 (ζ0 + rζ0′ )

= exp 2 dr + 2 dr . (3.22)
[(rζ0 )′ ]2 ζ0 (ζ0 + rζ0′ ) ζ0 (ζ0 + rζ0′ )
We can simplify to,
r2
Z ′
ζ0 (ζ0 − rζ0′ )
Z ′
ζ0
= 2 exp 2 ′
dr + 2 dr .
′
[(rζ0) ] ζ0 (ζ0 + rζ0 ) ζ0
Furthermore,
r2
Z ′
ζ0 (ζ0 − rζ0′ )

= 2 exp 2 ′
dr exp(ln(ζ02 )).
′
[(rζ0 ) ] ζ 0 (ζ 0 + rζ 0 )
Note that, exp(ln(ζ02)) = ζ02 , so we can calculate
2 Z ′
ζ0 (r) ζ0 (r) − r ζ0′ (r)

ζ0 (r) 2
∆0 = r exp 2 dr , (3.23)
ζ0 (r) + r ζ0′ (r) ζ0 (r) ζ0 (r) + r ζ0′ (r)
as previously illustrated.
If we apply this transformation a second time we obtain no additional information.
To see this, consider the sequence
{ζ0 , B0 } 7→ {ζ0 , B1 } 7→ {ζ0 , B2 } . . . (3.24)
69
But at the second step (and all subsequent steps), since ζ0 has not changed, so ∆1 (r) =
∆0 (r). More generally, at all stages of the iteration ∆i (r) = ∆0 (r). We can write this
as !
Yn Xn
T1 (λi ) = T1 λi . (3.25)
i=1 i=1
or in the more suggestive form

n
Y
T1 , T1 (3.26)
i=1
where the symbol , indicates “equality up to relabelling of the parameters”. That

is, transformation T1 is “idempotent” up to relabelling of the parameters (see figure
3.2).
A version of Theorem 1 can also be found in [20]. Specifically, after several manip-
ulations, changes of notation, and a change of coordinate system, the transformation
exhibited in equation (16.11) of [20] can be converted into Theorem 1 above.
Applying theorem 1 to a fixed {ζ0 , B0 } generates a one dimensional space of perfect
fluid spheres, which leads to the corollary below:
Corollary 1. Let {ζ0 , Ba } and {ζ0 , Bb } both represent perfect fluid spheres, then for
all p
{ζ0 , pBa + (1 − p)Bb } (3.27)
is also a perfect fluid sphere, furthermore all perfect fluid spheres for a fixed ζ0 can be
written in this form.
Proof. The result is automatic once you note that for fixed ζ0 the ODE for B is linear,
(though not homogeneous, which is why the two coefficients p and 1 − p are chosen to
add up to 1) .
We defer extensive discussion of the application of this theorem and its corollary
until section 3.4, where we use this and our other generating theorems as a basis
for classifying perfect fluid spheres. At this stage we mention, only as a very simple
example, that T1 applied to Minkowski space results in the geometry of the Einstein
static universe.
70
4
x 10
6
λ1=10
λ2=20
5
λ3=30
B(r) after each application of theorem 1
λ4=40
4 λ5=50
λ6=60
λ7=70
3 λ8=80
λ9=90
λ10=100
2
λonce=550
0
0 1 2 3 4 5 6 7 8 9 10
r
Figure 3.2: The solid lines show B(r) for 10 reapplications of Theorem 1 onto the Minkowski
metric. The dashed line corresponds to a single application with a specific choice for λonce .
It can be seen that 10 applications of Theorem 1 can be expressed by one application.
3.2.2 The second theorem
The second theorem we present is a new transformation theorem. This is a different

formalism, and demonstrates our independent way of proving it. Furthermore, we
would like to show the “idempotence” property of theorem 2 in this section.
Theorem 2. Let {ζ0 , B0 } describe a perfect fluid sphere. Define

Z
r dr
Z0 (r) = σ + ǫ p . (3.28)
ζ0 (r)2 B0 (r)
71
Figure 3.3: This structure shows that when we apply theorem 1 to Minkowski, we get an
Einstein static universe as a new metric.
Then for all σ and ǫ, the geometry defined by holding B0 (r) fixed and setting
dr 2
ds2 = −ζ0 (r)2 Z0 (r)2 dt2 + + r 2 dΩ2 (3.29)
B0 (r)
T2 (σ, ǫ) : {ζ0 , B0 } 7→ {ζ0 Z0 (ζ0 , B0 ), B0 } (3.30)
takes perfect fluid spheres into perfect fluid spheres. Furthermore a second application
of the transformation does not yield new information, T2 is “idempotent” in the sense
that
T2 (σn , ǫn ) ◦ · · · ◦ T2 (σ3 , ǫ3 ) ◦ T2 (σ2 , ǫ2 ) ◦ T2 (σ1 , ǫ1 ) = T2 (σn . . . σ3 σ2 σ1 , ǫn...321 ), (3.31)
where
ǫn...321 = (ǫ1 σ2 σ3 · · · σn ) + (σ1−1 ǫ2 σ3 · · · σn ) + (σ1−1 σ2−1 ǫ3 · · · σn )

+ · · · + (σ1−1 σ2−1 σ3−1 · · · ǫn ). (3.32)
Furthermore, theorem 2 is invertible (as long as σ 6= 0):
[T2 (σ, ǫ)]−1 = T2 (1/σ, −ǫ). (3.33)
72
Reduction of order:
This is a method to find a general solution to a linear differential equation, provided
you already have one particular solution. In this method, we start with an nth –order
linear differential equation. This method is especially useful in solving a second–order
linear differential equation. It reduces the problem to one for solving a first–order
linear differential equation.
Proof for Theorem 2. The proof of theorem 2 is based on the technique of “reduction
in order”. Assuming that {ζ0 (r), B0 (r)} solves equation (3.9), write
ζ1 (r) = ζ0 (r) Z0 (r) . (3.34)
and demand that {ζ1 (r), B0 (r)} also solves equation (3.9). We find
2r 2 B0 (ζ0 Z0 )′′ + (r 2 B0′ − 2rB0 )(ζ0 Z0 )′ + (rB0′ − 2B0 + 2)(ζ0 Z0 ) = 0 (3.35)
we can expand the above equation to
2r 2 B0 (ζ0′′ Z0 + 2ζ0′ Z0′ + ζ0 Z0′′ ) + (r 2 B0′ − 2rB0 )(ζ0′ Z0 + ζ0 Z0′ )

+(rB0′ − 2B0 + 2)(ζ0 Z0 ) = 0 (3.36)
2r 2B0 ζ0′′ + (r 2 B0′ − 2rB0 )ζ0′ + (rB0′ − 2B0 + 2)ζ0 Z0

+(r 2 ζ0 B0′ + 4r 2 ζ0′ B0 − 2rζ0 B0 )Z0′ + (2r 2ζ0 B0 )Z0′′ = 0 (3.37)
This linear homogeneous 2nd order ODE for Z0 now simplifies to
(r 2 ζ0 B0′ + 4r 2 ζ0′ B0 − 2rζ0 B0 )Z0′ + (2r 2 ζ0 B0 )Z0′′ = 0 , (3.38)
which is an ordinary homogeneous second-order differential equation, depending only

on Z0′ and Z0′′ . (So it can be viewed as a first-order homogeneous order differential
equation in Z ′ , which is solvable.) Separating the unknown variable to one side,
Z0′′ 1 B0′ ζ0′ 1

′
= − − 2 + . (3.39)
Z0 2 B0 ζ0 r
73
Re-write Z0′′ /Z0′ = d ln(Z0′ )/dt, and integrate twice over both sides of equation (3.39),
to obtain
r dr
Z
Z0 = σ + ǫ p , (3.40)
ζ0 (r)2 B0 (r)
depending on the old solution {ζ0 (r), B0 (r)}, and two arbitrary integration constants
σ and ǫ.
To see that the transformation T2 defined in Theorem 2 is “idempotent” we first
show
T2 ◦ T2 , T2 , (3.41)
and then iterate. The precise two-step composition rule is

ǫ2
T2 (σ2 , ǫ2 ) ◦ T2 (σ1 , ǫ1 ) = T2 σ2 σ1 , ǫ1 σ2 + . (3.42)
σ1
In this section we consider a two-step composition rule to find idempotence. In
particular, the most outstanding features of these steps are:
• To see “idempotence”, note that for fixed B0 (r) equation (3.9) has a solution
space that is exactly two dimensional.
• Since the first application of T2 takes any specific solution and maps it into the
full two-dimensional solution space, any subsequent application of T2 can do
no more than move one around inside this two dimensional solution space —
physically this corresponds to a relabelling of parameters describing the perfect
fluid metric you are dealing with, not the generation of new solutions.
To be more explicit about this note that at step one

 Z 
 r dr 
ζ0 → ζ1 = ζ0 σ1 + ǫ1 p , (3.43)
 ζ0 (r)2 B0 (r) 
while at the second step

 Z 
 r dr 
ζ1 → ζ2 = ζ1 σ2 + ǫ2 p . (3.44)
 ζ1 (r)2 B0 (r) 
74
That is:
 Z 
 r dr 
ζ2 = ζ0 σ1 + ǫ1 p (3.45)
 ζ0 (r)2 B0 (r) 
 Z 
 r dr 
× σ2 + ǫ2 p R p .
 ζ0 (r)2 B0 (r) [σ1 + ǫ1 r dr/(ζ0 (r)2 B0 (r))]2 
But this can be rewritten as

 Z 
 ǫ2 dZ0 
ζ2 = ζ0 Z0 σ2 +
 ǫ1 Z02 
Z
ǫ2 1
= ζ0 Z0 σ2 − d
ǫ1 Z0

ǫ2 1 1
= ζ0 Z0 σ2 − − . (3.46)
ǫ1 Z0 σ1
Therefore
ǫ2 ǫ2 1
ζ2 = ζ0 − + σ2 + Z0 . (3.47)
ǫ1 ǫ1 σ1
That is

ǫ2 ǫ2 1
Z1 = − + σ2 + Z0 , (3.48)
ǫ1 ǫ1 σ1

Z
ǫ2 r dr
= σ2 σ1 + σ2 ǫ1 + p (3.49)
σ1 2
ζ0 (r) B0 (r)
from which the composition law

ǫ2
T2 (σ2 , ǫ2 ) ◦ T2 (σ1 , ǫ1 ) = T2 σ2 σ1 , ǫ1 σ2 + (3.50)
σ1
follows immediately. (Note that the composition law for T2 is actually a statement
about applying reduction of order to 2nd-order ODEs, it is not specifically a statement
about perfect fluid spheres, though that is how we will use it in the thesis). The
75
general composition law then follows by induction. To find the inverse transformation
we choose σ2 = 1/σ1 and ǫ1 = −ǫ2 , for which
T2 (1/σ1 , −ǫ1 ) ◦ T2 (σ1 , ǫ1 ) = T2 (1, 0) = I. (3.51)
Comment: As other special cases of the composition law we will also mention the
results that !
Y n n
X
T2 (1, ǫi ) = T2 1, ǫi , (3.52)
i=1 i=1
and
T2 (σ, ǫ)n = T2 σ n , ǫ σ n−1 + σ n−3 · · · + σ −(n−3) + σ −(n−1) .

(3.53)
Now as long as σ > 1 then for sufficiently large n we see
T2 (σ, ǫ)n ≈ T2 σ n , σ n−1 ǫ = σ n−1 T2 (σ, ǫ) , T2 (σ, ǫ),

(3.54)
where at the last step we have used the fact that the overall multiplicative factor σ n−1
can simply be reabsorbed into a redefinition of the time coordinate. Because of this
result, we see that for fixed σ > 1 (and fixed but arbitrary ǫ) repeated numerical
applications of T2 (σ, ǫ) will have a well-defined limit. In figure 3.4 we have tested the
composition law numerically.
A strictly limited version of theorem 2, with little comment on its importance, can
be found in [20]. Specifically, after several manipulations, changes of notation, and a
change of coordinate system, the transformation exhibited in equation (16.12) of [20]
can be seen to be equivalent to the sub-case σ = 0, ǫ = 1 of theorem 2 above.
For some purposes it is more useful to rephrase theorem 2 as below:
Corollary 2. Let {ζa , B0 } and {ζb , B0 } both represent perfect fluid spheres, then for
all p and q
{p ζa + q ζb , B0 } (3.55)
is also a perfect fluid sphere. Furthermore, for fixed B0 all perfect fluid spheres can be
written in this form.
76
8
σ1=0.25, ε1=1
σ2=0.50, ε2=2
7
σ3=0.75, ε3=3
σ4=1.00, ε4=4
6
σ5=1.25, ε5=5
ζ(r) after each application of theorem 1
σ6=1.50, ε6=6
5 σ7=1.75, ε7=7
σ8=2.00, ε8=8
4 σ9=2.25, ε9=9
σ10=2.50, ε10=10
3 σonce= 3.46,
εonce=591.73
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
r
Figure 3.4: The solid lines show ζ(r) for 10 reapplications of Theorem 2 onto the Minkowski
metric. The dashed line corresponds to a single application with a specific choice for σonce and
ǫonce . It can be seen that 10 applications of Theorem 2 can be expressed by one application.
Proof. The result is automatic once you note that for fixed B0 the 2nd order ODE for
ζ is linear and homogeneous.
We defer extensive discussion about the application of these theorems and corollary
until section 3.4, which cover the following:
• A simple example involving T2 applied to either the Einstein static universe or

the anti-de Sitter universe results in Schwarzschild’s stellar solution (position-
independent density).
• Similarly, corollary 2 applied to a combination of the Einstein static universe and
77
anti-de Sitter space is another way of obtaining Schwarzschild’s stellar solution.
Note that when comparing the results of our computations with the extant literature
we shall adopt the naming conventions of the review article by Delgaty and Lake [9].
For example, some of the less standard names are:
• K-O III: The third spacetime considered by Korkina and Orlyanskii.
• M-W III: The third spacetime considered by Matese and Whitman.
• B-L: The spacetime considered by Buchdahl and Land.
• Kuch: One of several spacetimes considered by Kuchowicz.
• Heint: One of several spacetime considered by Heintmann.
References to the primary literature can be found in Delgaty and Lake [9]. We also
add several new abbreviations:
• Martin: One of several spacetimes considered by Martin and Visser.
• S1: A special name for a particularly simple spacetime.
• P1-P8: Special names for several apparently new perfect fluid spacetimes.
See in particular tables (3.2) and (3.3) for the spacetime metrics corresponding to
these geometries.
3.2.3 The third and fourth theorems

Having now found the first and second generating theorems, is now useful to define
two new theorems by composing them. What we are going to show in the following
section is how to deform the geometry while still retaining the perfect fluid property.
To further illustrate the formalism, we will show how to classify perfect fluid spheres
by using two new generating theorems. We present theorem 3 by taking a perfect fluid
sphere solution {ζ0 , B0 } and applying theorem 1 onto it which gives us a new perfect
fluid sphere {ζ0 , B1 }. The new B1 is given in equation (3.11). We now continue with
78
Figure 3.5: This structure shows that when we apply theorem 2 to Minkowski, we get the
K-O III geometry.
applying theorem 2, again we get a new solution {ζ̃, B1 }, where ζ̃ now depends on
the new B1 . All together we can consider this as a single process by introducing the
following theorem:
Theorem 3. If {ζ0 , B0 } denotes a perfect fluid sphere, then for all σ, ǫ, and λ, the
three-parameters geometry defined by
 Z 2
 r dr  dr 2
ds2 = −ζ0 (r)2 σ + ǫ p dt2 + + r 2 dΩ2
 ζ0 (r) 2 B0 (r) + λ ∆0 (r)  B0 (r) + λ∆ 0 (r)
(3.56)
is also a perfect fluid sphere, where ∆0 is
2 Z ′
ζ0 (r) ζ0 (r) − r ζ0′ (r)

ζ0 (r) 2
∆0 (r) = r exp 2 dr . (3.57)
ζ0 (r) + r ζ0′ (r) ζ0 (r) ζ0 (r) + r ζ0′ (r)
That is
T3 = T2 ◦ T1 : {ζ0 , B0 } 7→ {ζ0 , B0 + λ ∆0 (ζ0 )}

7→ {ζ0 Z0 (ζ0 , B0 + λ∆0 (ζ0 )), B0 + λ∆0 (ζ0 )}. (3.58)
Now we would like to present another new theorem. Instead of starting with the-
orem 1 we could first apply theorem 2 on {ζ0 , B0 }. This gives us a new perfect fluid
79
sphere {ζ1 , B0 }, where ζ1 = ζ0 Z0 (ζ0 , B0 ) is given by equation (3.40). We now continue
with theorem 1 which leads to {ζ1 , B̃} where B̃ depends on the new ζ1 . Again, we can
consider this as a single process by introducing the following theorem:
Theorem 4. If {ζ0, B0 } denotes a perfect fluid sphere, then for all σ, ǫ, and λ, the
three-parameters geometry are defined by
 Z 2
2 2
 r dr 
2 dr 2
ds = −ζ0 (r) σ+ǫ p dt + + r 2 dΩ2 (3.59)
 ζ0 (r) 2 B0 (r)  B0 (r) + λ∆0 (ζ1, r)
is also a perfect fluid sphere, where ∆0 (ζ1 , r) is defined as

2 Z ′
ζ1 (r) ζ1 (r) − r ζ1′ (r)

ζ1 (r) 2
∆0 (ζ1 , r) = r exp 2 dr , (3.60)
ζ1 (r) + r ζ1′ (r) ζ1 (r) ζ1 (r) + r ζ1′ (r)
depending on ζ1 = ζ0 Z0 , where as before

Z
r dr
Z0 (r) = σ + ǫ p . (3.61)
ζ0 (r)2 B0 (r)
That is
T4 = T1 ◦ T2 : {ζ0 , B0 } 7→ {ζ0 Z0 (ζ0, B0 ), B0 }

7→ {ζ0 Z0 (ζ0 , B0 ), B0 + λ∆0 (ζ0 Z0 (ζ0 , B0 ))}. (3.62)
In addition, we analyze what happens, if we apply theorems 3 or 4 more than once.

Some general comments about both new transformation theorems:
• Theorem 3 and theorem 4 are in general distinct, which can be traced back to
the fact that theorem 1 and theorem 2 do not in general commute.
• Theorem 3 and theorem 4 are in general not idempotent. That means when
we apply theorem 3 more than once further distinct solutions will be obtained.
Similarly for theorem 4.
80
In addition, these comments can be described in terms of the equations below.
T3 6, T4 ; T3 ◦ T3 6, T3 ; T4 ◦ T4 6, T4 . (3.63)
The best way to verify this is to try a few specific examples. There may be some
specific and isolated special metrics for which theorem 3 and theorem 4 happen to be
degenerate, or idempotent, and finding such metrics is important for our classification
programme.
Figure 3.6: This structure shows that Theorem 3 and theorem 4 are in general distinct.
When we apply theorem 3 to Minkowski, we get the Schwarzschild Interior geometry, while
applying theorem 4 to the Minkowski, we get Martin 3 .
3.2.4 Formal properties of the generating theorems

The solution generating theorems we have developed interact in a number of interest-
ing ways and exhibit numerous formal properties that will be useful when classifying
generic perfect fluid spheres. To start with, theorem 3 (T3 ) and theorem 4 (T4 ) can
be expressed in terms of theorem 1 (T1 ) and theorem 2 (T2 ):
T3 (g) := (T2 ◦ T1 ) g = T2 (T1 (g)) ;

(3.64)
T4 (g) := (T1 ◦ T2 ) g = T1 (T2 (g)) .
81
where g is a metric representing a perfect fluid sphere. Having this in mind, some of
the new solutions generated by starting from some specific solution can be identified.
For example:
T4 ◦ T1 ≡ T1 ◦ T2 ◦ T1 ≡ T1 ◦ T3 , (3.65)
or
T3 ◦ T3 ≡ T2 ◦ T1 ◦ T2 ◦ T1 ≡ T2 ◦ T4 ◦ T1 . (3.66)
The idempotence of T1 and T2 in this formalism is:
(T1 ◦ T1 ) g =T1 (T1 (g)) , T1 (g) ;

(3.67)
(T2 ◦ T2 ) g =T2 (T2 (g)) , T2 (g) .
Taken together, it is possible to simplify all formulae wherever T1 and T2 appear next
to each other more than once. The following examples should demonstrate how this
works:
(T2 ◦ T3 ) g ≡ (T2 ◦ T2 ◦ T1 ) g , (T2 ◦ T1 ) g ≡ T3 (g) ;

(3.68)
(T1 ◦ T4 ) g ≡ (T1 ◦ T1 ◦ T2 ) g , (T1 ◦ T2 ) g ≡ T4 (g) ,
and in the same way
T4 ◦ T3 ≡ T1 ◦ T2 ◦ T2 ◦ T1 , T1 ◦ T2 ◦ T1 ≡ T1 ◦ T3 ≡ T4 ◦ T1 ;
(3.69)
T3 ◦ T4 ≡ T2 ◦ T1 ◦ T1 ◦ T2 , T2 ◦ T1 ◦ T2 ≡ T2 ◦ T4 ≡ T3 ◦ T2 .
These relationships can be used to structure the solution set generated starting from
any particular perfect fluid sphere, and moreover to classify which metrics can be
produced by our theorems, and which ones cannot. For example, the idempotence
property of theorem 1 and theorem 2 enables us to divide the class of perfect fluid
spheres into seed metrics and non-seed metrics. Seed metrics can never be generated
by using one of the two theorems T1 or T2 , while non-seed metrics are connected to
other simpler metrics via one of these theorems. We formalize this in the following
subsection.
82
3.2.5 Seed and non-seed metrics
Definition (Seed metric): Take a metric g (or a parameterized class of metrics)
and apply theorem 1 or theorem 2 on it. Three different cases are possible:
• Each of the applications supplies us with a new solution. [T1 (g) 6, g 6, T2 (g).]
We define a metric with this behaviour as a seed metric. (We shall soon see
several examples of this behaviour.) For example, figure 2.5 show that Minkowski
is a seed metric. When we apply theorem 1 to it, we get Einstein static. While
applying theorem 2 to Minkowski, we get K-O III instead.
• Only one of the applications supplies us with a new solution, while the other one
gives us the same metric we started with. [T1 (g) , g or T2 (g) , g.] These met-
rics are non-seed metrics. (We shall soon see several examples of this behaviour.)
For example, figure 2.5 show that Einstein static and K-O III are both non-seed
metric. The reason is only one of the applications gives us a new solution, while
the other one still give us the same metric which are Einstein static and K-O
III, respectively.
• Both applications give us the same metric we started with. [T1 (g) , g , T2 (g).]
Metrics of this type are fixed points of the transformation process and we then
also have T3 (g) , g , T4 (g). While we have encountered numerical examples
that seem to exhibit this behaviour, we have no analytic proof for the existence
of non-obvious fixed-point metrics. There is one obvious but not particularly
useful example of a fixed point class of metrics. If we take the ODE in equation
(3.8), and write down its most general solution as a function of the arbitrary
parameters ζ(r), then any of our solution generating theorems applied to this
most general solutions will at most move us around in the parameter space
characterizing the most general solution — the most general solution of equation
(3.8), or equivalently equation (3.9), is thus an infinite-parameter fixed point of
the generating theorems. But apart from this obvious example, it is unclear
whether other fixed point classes of metric exist.
Classifying seed and non-seed metrics are very important. In particular, the most
outstanding features of this algorithm are:
83
• We developed a tool to generate new solutions for a perfect fluid sphere, which
does not require to solve the Einstein equations.
• We also established the relationships among the generating theorems. Before all
metrics seem to have nothing more in common than presenting a perfect fluid
sphere.
• We invented the concept of a seed metric, which cannot be generated by our the-
orems. Starting with such a metric, a class of non-seed metric can be produced.
• This can be used to structure the solutions generated by a perfect fluid sphere
and classify which metrics can and cannot be produced by our theorems.
Examples of some seed and non-seed metrics
Seed metric Non-Seed metric

Minkowski B-L
Exterior Schwarzschild Interior Schwarzschild
de Sitter Einstein static
Tolman IV (A = 0) Tolman IV , Tolman V (n = +1) , Tolman VI
M-W III K-O III
Heint IIa (C=0) Heint IIa
S1 Wyman IIa, Wyman III
Table 3.1: This table shows several seed and non-seed metrics, which satisfied the definition
of seed and non-seed metrics (see Table 3.2 for further details).
We can nevertheless develop several formal lemmata regarding fixed-point metrics.

For instance
Lemma 1. Suppose we have a metric such that ∀ σ, ǫ, λ
T3 (σ, ǫ, λ) g , g, (3.70)
84
where we recall that , denotes equality up to redefinition of parameters. Then in
particular
T1 (λ) g , g , T2 (σ, ǫ) g, (3.71)
and so
T4 (σ, ǫ, λ) g , g. (3.72)
Conversely, suppose we have a metric such that
T4 (σ, ǫ, λ) g , g, (3.73)
then
T1 (λ) g , g , T2 (σ, ǫ) g, (3.74)
and so
T3 (σ, ǫ, λ) g , g. (3.75)
Proof. Trivial, note that T3 (σ, ǫ, λ = 0) = T2 (σ, ǫ) and T3 (σ = 0, ǫ = 0, λ) = T1 (λ).

Then recall T4 = T1 ◦ T2 . Similarly for the converse.
Lemma 2. Suppose we have a metric g such that
T4 g , T3 g (3.76)
and then define g ′ by

T4 g , g ′ , T3 g. (3.77)
Then we have
T4 g ′ , g ′ , T3 g ′ (3.78)
so that g ′ is a “fixed point” of both T3 and T4 .
Proof. Note
T4 g ′ , T4 ◦ T3 g , T1 ◦ T2 ◦ T1 g , T1 ◦ T3 g , T1 ◦ T4 g , T4 g = g ′ (3.79)
and similarly for T3 :
T3 g ′ , T3 ◦ T4 g , T2 ◦ T1 ◦ T2 g , T2 ◦ T4 g , T2 ◦ T3 g , T3 g = g ′ (3.80)
85
Several other formal theorems along these lines can be constructed, but these seem
the most important results.
Finally, among the formal properties enjoyed by the generating theorems, we men-
tion the fact that theorems 3 and 4 are “conjugate” to each other in the following
sense
T4 ≡ T1 ◦ T2 = T1 ◦ T2 ◦ T1 ◦ [T1 ]−1 = T1 ◦ T3 ◦ [T1 ]−1 , (3.81)
and similarly (when the appropriate inverse [T2 ]−1 exists)
T3 ≡ T2 ◦ T1 = T2 ◦ T1 ◦ T2 ◦ [T2 ]−1 = T2 ◦ T4 ◦ [T2 ]−1 . (3.82)
We can write this as

T3 (σ, ǫ, λ) ∼ T4 (σ, ǫ, λ), (3.83)
where ∼ denotes the concept of “similarity” under conjugation by invertible generating

theorems.
In these properties we would like to ensure that
• We wish to emphasise that similarity ∼ is a statement that holds for particular

and fixed values of the parameters (σ, ǫ, λ), as opposed to , which denotes
equivalence under redefinition of parameters.
• If one is working numerically, it is much easier to ask questions involving sim-

ilarity ∼. For analytic work, it is typically easier to ask questions involving
equivalence ,.
3.2.6 The fifth and sixth theorems
Two linking theorems

The last two solution generating theorems we shall present are slightly different from
those developed so far: Consider a perfect fluid sphere solution {ζ0 , B0 } and extend
it to a new perfect fluid sphere {ζ0 Z0 , B0 + ∆}. Previously, we had either set Z0 = 1
and obtained theorem 1, or we had set ∆ = 0 and obtained theorem 2. In other words,
we only changed one metric component at a time. (From this point of view theorem 3
86
and theorem 4 are, strictly speaking, not new theorems, in that they are replaceable
by iterations of theorem 1 and theorem 2 and vice versa.) We now investigate what
happens if we place no a priori restrictions on Z and ∆, and allow both metric com-
ponents to vary simultaneously. The differential equation (3.8) for this problem now
becomes
[r(rζ0Z0 )′ ] ∆′ + 2r 2 (ζ0 Z0 )′′ − 2(rζ0Z0 )′ ∆

+ r 2 ζ0 B0′ + 4r 2 ζ0′ B0 − 2rζ0B0 Z0′ + 2r 2 ζ0 B0 Z0′′ = 0.

(3.84)
Note that if ∆ = 0 this becomes equation (3.17), while if Z0 = 1 this becomes equation
(3.38). In general, this ODE of first-order in ∆, and — as long Z is not a constant —
inhomogeneous. In terms of ∆ this ODE can be solved explicitly and the result stated
as a new theorem:
Theorem 5. Suppose {ζ0 , B0 } describes a perfect fluid sphere, and let Z0 (r) be arbi-
trary. Define
(
[(4r 2 ζ0′ B0 + r 2 ζ0 B0′ − 2rζ0 B0 ) Z0′ + 2r 2 ζ0 B0 Z0′′ ] {ζ0 + rζ0′ }2
Z
∆(λ, r) = ∆0 (r) λ −
r 3 (rζ0Z0 )′ ζ02 Z02
)
(ζ0 Z0 )′ ζ0 Z0 − r (ζ0 Z0 )′
Z
exp −2 dr dr , (3.85)
ζ0 Z0 ζ0 Z0 + r (ζ0 Z0 )′
where
2
(ζ0 Z0 )′ ζ0 Z0 − r (ζ0 Z0 )′
Z
ζ 0 Z0 2
∆0 (r) = r exp 2 dr . (3.86)
ζ0 Z0 + r (ζ0 Z0 )′ ζ0 Z0 ζ0 Z0 + r (ζ0 Z0 )′
Then for all λ, the geometry defined by an arbitrary chosen Z0 (r) and setting
dr 2
ds2 = −ζ0 (r)2 Z0 (r)2 dt2 + + r 2 dΩ2 (3.87)
B0 (r) + ∆(λ, r)
corresponds to a perfect fluid sphere. That is, the mapping
T5 (λ) : {ζ0 , B0 } 7→ {ζ0 Z0 , B0 + ∆(λ, ζ0)} (3.88)
takes perfect fluid spheres into perfect fluid spheres.
87
Note that if Z0 (r) = 1 this simply reduces to theorem 1. Re-arranging equation
(3.84) in terms of Z0 leads to a second-order inhomogenous differential equation, which
cannot in general be solved for a prescribed ∆, unless one imposes further constraints.
So further exploration in that direction is moot. There is however a related theorem
(which may be easier to understand) in terms of parametric derivatives:
Theorem 6. Let {ζ(µ), B(µ)} denote a one-parameter class of perfect fluid spheres,
so that the differential equation (3.8) is satisfied for all µ. Then
′ ′′
′ dB 2 ′′ ′ dB dζ 2
[r(rζ) ] + [2r ζ − 2(rζ) ] + 2r B
dµ dµ dµ
′
dζ dζ
+(r 2 B ′ − 2rB) + (rB ′ − 2B + 2) = 0. (3.89)
dµ dµ
In particular if dζ/dµ = 0 this reduces to the ODE (3.17), while if dB/dµ = 0 this
reduces to the ODE (3.9).
This is simply an alternative viewpoint on the previous theorem, emphasising the

differential equation to be solved.
We again defer extensive discussion to the next section, but that this stage point
out that if we invoke theorem 5 and apply it to Minkowski space, then making the
choice Z0 = 1 + r 2 /a2 leads to the general Tolman IV metric — that is:
T5 (Minkowski; Z0 = 1 + r 2 /a2 ) = (Tolman IV). (3.90)
Even before we systematically start our classification efforts, it is clear that the solution
generating theorems we have established will inter-relate many of the standard perfect
fluid spheres.
3.3 Formal properties of the linking theorems

Before turning to issues of systematic classification of perfect fluid metrics, we wish
to establish a few formal properties of the linking theorems. To simplify the notation,
88
Figure 3.7: This structure shows that when we invoke theorem 5 and apply it to Minkowski
space, then making the choice Z0 = 1 + r 2 /a2 leads to the general Tolman IV metric.
let us define the differential expression
D(ζ, B) ≡ [r(rζ)′]B ′ + [2r 2 ζ ′′ − 2(rζ)′]B + 2ζ

= 2r 2 Bζ ′′ + (r 2 B ′ − 2rB)ζ ′ + (rB ′ − 2B + 2)ζ. (3.91)
Then the condition for a perfect fluid sphere is simply
D(ζ, B) = 0 (3.92)
Now define
D(ζ0 , B0 + ∆0 ) = D(ζ0 , B0 ) + D1 (ζ0 ; ∆0 ) (3.93)
Then it is easy to check that
D1 (ζ0 ; ∆0 ) = [r(rζ0 )′ ] ∆′0 + 2r 2 ζ0′′ − 2(rζ0)′ ∆0

(3.94)
The ODE for theorem 1, where we assume {ζ0 , B0 } is a perfect fluid sphere, is then
D1 (ζ0 ; ∆0 ) = 0. (3.95)
Now define
D(ζ0 Z0 , B0 ) = Z0 D(ζ0, B0 ) + D2 (ζ0 , B0 ; Z0 ), (3.96)
then it is easy to check that
D2 (ζ0 , B0 ; Z0 ) = (r 2 ζ0 B0′ + 4r 2 ζ0′ B0 − 2rζ0 B0 )Z0′ + (2r 2 ζ0 B0 )Z0′′ . (3.97)
89
The ODE for theorem 2, where we assume {ζ0 , B0 } is a perfect fluid sphere, is then
D2 (ζ0 , B0 ; Z0 ) = 0. (3.98)
Now let us consider any simultaneous shift in ζ and B, as considered in theorem 5.

We have
D(ζ0 Z0 , B0 + ∆0 ) = D(ζ0 Z0 , B0 ) + D1 (ζ0 Z0 ; ∆0 ) (3.99)

= Z0 D(ζ0 , B0 ) + D2 (ζ0 , B0 ; Z0 ) + D1 (ζ0 Z0 ; ∆0 ).
But now let us write
D1 (ζ0 Z0 ; ∆0 ) = Z0 D1 (ζ0 ; ∆0 ) + D12 (ζ0 ; Z0 , ∆0 ), (3.100)
where a brief computation yields
D12 (ζ0 ; Z0 , ∆0 ) = r 2 ζ0 Z0′ ∆′0 + [2r 2 ζ0 Z0′′ + 4r 2 ζ0′ Z0′ − 2rζ0 Z0′ ]∆0 . (3.101)
Then all in all
D(ζ0 Z0 , B0 + ∆0 ) = Z0 D(ζ0 , B0 ) + Z0 D1 (ζ0 ; ∆0 )

+D12 (ζ0 ; Z0 , ∆0 ) + D2 (ζ0 , B0 ; Z0 ). (3.102)
So if {ζ0 , B0 } and {ζ0Z0 , B0 + ∆0 } are both perfect fluid spheres we must have
Z0 D1 (ζ0 ; ∆0 ) + D12 (ζ0 ; Z0 , ∆0 ) + D2 (ζ0 , B0 ; Z0 ) = 0. (3.103)
This is
Z0 × (ODE for theorem1) + (cross term) + (ODE for theorem2) = 0. (3.104)
The cross term vanishes if either Z0 = constant or ∆0 = 0 in which case we recover

the usual theorem 1 and theorem 2. If we do things in the opposite order
D(ζ0 Z0 , B0 + ∆0 ) = Z0 D(ζ0 , B0 + ∆0 ) + D2 (ζ0 , B0 + ∆0 ; Z0 ); (3.105)
= Z0 D(ζ0, B0 ) + Z0 D1 (ζ0 , ∆0 ) + D2 (ζ0 , B0 + ∆0 ; Z0 ). (3.106)
90
We now have to compute
D2 (ζ0 , B0 + ∆0 ; Z0 ) = D2 (ζ0 , B0 ; Z0 ) + D21 (ζ0 , B0 ; Z0 , ∆0 ), (3.107)
and after a brief computation
D21 (ζ0 , B0 ; Z0 , ∆0 ) = 2r 2 ζ0 ∆0 Z0′′ + [r 2 ζ0 ∆′0 + 4r 2ζ0′ ∆0 − 2rζ0∆0 ]Z0′

= D12 (ζ0 , B0 ; Z0 , ∆0 ). (3.108)
Thus the cross term is the same, no matter how you calculate it, and we still have the
identity
D(ζ0 Z0 , B0 + ∆0 ) = Z0 D(ζ0 , B0 ) + Z0 D1 (ζ0 ; ∆0 )

+D12 (ζ0 ; Z0, ∆0 ) + D2 (ζ0 , B0 ; Z0 ). (3.109)
Provided ζ0 , B0 and {ζ0 Z0 , B0 + ∆0 } are both perfect fluid spheres we again deduce
Z0 D1 (ζ0 ; ∆0 ) + D12 (ζ0 ; Z0 , ∆0 ) + D2 (ζ0 , B0 ; Z0 ) = 0. (3.110)
Now this gives us another way of looking at theorem 3 and theorem 4. For theorem 3
we first apply theorem 1 so we have the two equations
D1 (ζ0 ; ∆0 ) = 0; (3.111)
and
D2 (ζ0 , B0 + ∆0 ; Z0 ) ≡ D12 (ζ0 ; Z0 , ∆0 ) + D2 (ζ0 , B0 ; Z0 ) = 0. (3.112)
Conversely, for theorem 4 where we first apply theorem 2 we see that we need to solve
D2 (ζ0 , B0 ; Z0 ) = 0; (3.113)
and
D1 (ζ0 Z0 ; ∆0 ) ≡ Z0 D1 (ζ0 ; ∆0 ) + D12 (ζ0 ; Z0 , ∆0 ) = 0. (3.114)
For theorem 5 we pick Z0 arbitrarily, and solve the single ODE
D1 (ζ0 Z0 ; ∆0 ) + D2 (ζ0 , B0 ; Z0 ) = 0. (3.115)
91
This is a single first-order linear inhomogeneous ODE for ∆0 , and hence solvable. (In
particular this makes it clear that theorem 5 is an inhomogeneous version of theorem
1 with a carefully arranged “source term” D2 (ζ0 , B0 ; Z0 ). While theorem 5 is not
“idempotent” it does satisfy the important formal property that:
Lemma 3.
T1 ◦ T5 , T5 (3.116)
which in particular tells us that the output from theorem 5 is never a seed metric.
Proof. Applying theorem 5 we need to solve
D1 (ζ0 Z0 ; ∆0 ) + D2 (ζ0 , B0 ; Z0 ) = 0 (3.117)
in order to map
{ζ0, B0 } → {ζ0 Z0 , B0 + ∆0 } (3.118)
Now apply theorem 1 to {ζ0Z0 , B0 +∆0 }, this means we have to solve the homogeneous
ODE
D1 (ζ0 Z0 ; ∆1 ) = 0 (3.119)
But then, using properties of first-order ODEs
{D1 (ζ0 Z0 ; ∆0 ) + D2 (ζ0, B0 ; Z0 ) = 0} ⊕ {D1 (ζ0 Z0 ; ∆1 ) = 0} (3.120)
⇒ {D1 (ζ0 Z0 ; ∆0 + ∆1 ) + D2 (ζ0 , B0 ; Z0 ) = 0} (3.121)
which is the ODE from theorem 5 back again.
(The net result of this observation, as we shall see in the next section, is that
theorem 5 can be used to connect one seed metric with the “descendants” of another
seed metric.)
92
3.4 Classifying perfect fluid spheres
We will now see the power of these transformation theorems (solution generating
theorems) by using them in a number of different ways:
• To generate several new perfect fluid spheres
• To relate various perfect previously known fluid spheres to each other.
• And also to classify the geometries we encounter.
First some minor warnings: Despite comments made in [9], Kuch 2 I ≡ Tolman V;
and Kuchb I b is a perfect fluid for general values of its parameters. Furthermore
RR–I ≡ Einstein static; RR–V = Tolman V (n = −5/4). If we had not noted these
degeneracies then our tables below would have been more extensive, but would have
conveyed no extra information.
Starting with the metric for any known perfect fluid sphere and successively ap-
plying theorem 1 and theorem 2 numerous times will supply us with endless “new”
perfect fluid sphere solutions. There are so many types of “new” solutions such as:
• Some of these “new” solutions might already be findable in the literature, some
of them might be truly novel.
• Some of these solutions can be written down in a fully explicit form.
• Some solutions are explicit but not elementary, in the sense that while the metric
components can be exhibited as specific and explicit integrals, these integrals
cannot be done in elementary form.
• Some solutions are so complex that present day symbolic manipulation programs
quickly bog down.
(For specific symbolic computations we have used a vanilla installation of Maple. See
appendix C for some sample worksheets).
We present several tables and diagrams to summarize the situation. Two tables
are used to provide the names and explicit metrics for the perfect fluid spheres we
93
consider. Two other tables are used to describe the inter-relationships of these perfect
fluid spheres under T1 , T2 , T3 , and T4 . Recall that a seed metric is one for which
theorem 1 and theorem 2 both yield metrics distinct from the seed: T1 (g) 6, g 6, T2 (g).
In contrast for non-seed metrics one or the other of these theorems is trivial, either
T1 (g) , g or T2 (g) , g.
In these tables, we can see the notation “[integral]” means that it is definitely a
novel perfect fluid solution. However, the metric components involve an explicit inte-
gral that does not appear to be do-able by elementary methods, and is so complicated
that it does not seem worthwhile to even write it down.
By considering that theorem 1 and theorem 2 have “idempotent” properties , and
the fact that theorem 3 and theorem 4 can be expressed in terms of the first two
theorems, it is possible to structure, and therefore to graphically visualize the rela-
tionship between all metrics generated from a given seed metric. We demonstrate
this behaviour starting with Minkowski spacetime as seed metric where we are start-
ing with this trivial seed to check if it is possible to create endless new perfect fluid
spheres. The first few steps can be carried out explicitly, and show that Minkowski
space generates several well known interesting perfect fluid models.
In figure 3.8 all the ideas from the previous sections are used. Each box represents a
specific metric (perfect fluid sphere) while the arrows correspond to the application of
the different theorems. The horizontal arrows correspond to an application of theorem
1 and the vertical arrows to an application of theorem 2. In addition, the vectors
pointing along the diagonals can either indicate an application of theorem 3 or theorem
4. A dotted arrow corresponds to the application of theorem 3 while the dashed arrow
represents an application of theorem 4.
Figure 3.2 shows that after applying theorem 1 to the Minkowski seed metric, we
get the Einstein static universe. By the idempotence of theorem 1 n applications of
T1 to the seed metric still results in the Einstein static. Similary, any number of
applications of theorem 2 after the first (which leads to the K–O III solution) does
not give us any further new solutions (see figure 3.4) so we can see form this figure
that theorem 2 is “idempotent”. In addition, we can also see that the first and second
generating theorems are not commutative. Application of theorem 1 and theorem 2 in
94
Figure 3.8: Structure graph for Minkowski space used as seed metric.
that order to the Minkowski seed metric results in the Schwarzschild Interior geometry,
whereas application of theorem 1 after theorem 2 gives us the Martin 3 solution, which
is quite distinct from the three-parameter Schwarzschild interior solution. Indeed, it
seems as if it is possible to create endless new solutions for a perfect fluid sphere out
of the Minkowski metric (or any other of the seed metrics). After several iterations
the calculations become more complex, and can no longer be carried out analytically.
We then resort to numerical computation to find out whether theorem 3 and theorem
4 have some sort of numerical limit, a numerical fixed point, or not. Depending on
the choice for (λ, σ, ǫ) both theorems converge very quickly. Specifically, we used
95
Matlab to numerically analyze the evolution of ζ(r) and B(r) after applying theorem 3
several times to the Minkowski metric. Numerical results are summarized in figure 3.9.
This figure indicates that theorem 3 and theorem 4 both appear to have well defined
numerical limits, though we have no direct analytic solution for the fixed point metric.
Figure 3.9: Numerical implementations of Theorem 3 and Theorem 4 can be used to

investigate the behavior for n applications, if the number of applications goes to infinity.
For specific choices of parameters λi , σi , and ǫi , multiple reapplications of Theorem 3 and
Theorem 4 approach a limit, in the sense that both metric components are converging towards
there fixed-points.
96
3.5 Discussion
We have developed several transformation theorems that map perfect fluid spheres to
perfect fluid spheres using Schwarzschild coordinates. We used these transformations
as a basis for classifying different types of perfect fluid sphere solutions. Applying
these theorems on a known perfect fluid sphere, different solutions are often obtained.
Some of the solutions are already known in the literature but most of them are novel.
Moreover, we can classify which metrics can be produced by our theorems, and which
ones cannot. The first theorem we presented is in a slightly different formalism, and
demonstrates an independent way of proving it. Indeed, the second theorem we pre-
sented is a new transformation theorem. The idempotence property of Theorem 1 and
Theorem 2 can divide the class of perfect fluid spheres into seed and non-seed met-
rics. Classifying seed and non-seed metrics are important because we have developed
a tool to generate new solutions for a perfect fluid sphere, which does not require us
to directly solve the Einstein equations. Indeed, the whole procedure was set up in
such a way that we are implicitly and indirectly solving the Einstein equations, but
the utility of the transformation theorems is that one does not have to go back to first
principles for each new calculation.
We established two new theorems by composing first and second generating theo-
rems. They are in general distinct, which can be traced back to the fact that Theorem
1 and Theorem 2 do not, in general, commute. Indeed, Theorem 3 and Theorem 4
are, in general, not idempotent. That means when we apply Theorem 3 or Theorem
4 more than once, further solutions will in general be obtained. In addition, we can
see the power of these transformation theorems as we can generate several new perfect
fluid spheres and also can classify the geometries we encountered.
Furthermore, we have also established several relationships among the generating
theorems. Previously, all metrics seemed to have nothing more in common than pre-
senting a perfect fluid sphere. We have developed the concept of a seed metric, which
is one that cannot be generated by our theorems. Starting with a seed metric and
applying our theorems, it is possible to structure, therefore to visualize, the relation-
ship between all metrics generated from a given seed metric in a graph. Based on this
example it is possible to create endless new solutions out of the Minkowski metric. We
97
also used a numeric program to investigate whether Theorem 3 and Theorem 4 have
fixed point limits. Both seem to converge very quickly. The situation is summarized
graphically in figures (3.1) to (3.9), and in tabular form in tables (3.1) to (3.5).
98
Some selected perfect fluid solutions
Name Metric
Minkowski −dt2 + dr2 + r2 dΩ2
Einstein static −dt2 + (1 − r2 /R2 )−1 dr2 + r2 dΩ2
de Sitter −(1 − r2 /R2 ) dt2 + (1 − r2 /R2 )−1 dr2 + r2 dΩ2
p 2
Schwarzschild Interior − A − B 1 − r2 /R2 dt2 + (1 − r2 /R2 )−1 dr2 + r2 dΩ2
Schwarzschild Exterior −(1 − 2m/r) dt2 + (1 − 2m/r)−1 dr2 + r2 dΩ2
S1 −r4 dt2 + dr2 + r2 dΩ2
K-O III −(A + Br2 )2 dt2 + dr2 + r2 dΩ2
Kuch1 Ib −(Ar + Br ln r)2 dt2 + 2dr2 + r2 dΩ2
B–L −A(r2 /a2 )dt2 + 2(1 + r2 /a2 )−1 dr2 + r2 dΩ2
2
/a2
−B 2 1 + r2 /a2 dt2 + (1−r21+2r 2 2 2

Tolman IV /b2 )(1+r 2 /a2 ) dr + r dΩ
2 2
−B 2 1 + r2 /a2 dt2 + 1+2r /a 2 2 2

Tolman IV (b → ∞) 1+r 2 /a2 dr + r dΩ
2
Tolman V −B 2 r2(1+n) dt2 + (2 − n2 ) [1 − Ar2(2−n )/(2+n) ]−1 dr2 + r2 dΩ2
Tolman V (A → 0) −B 2 r2(1+n) dt2 + (2 − n2 )dr2 + r2 dΩ2
Tolman VI −(Ar1−n + Br1+n )2 dt2 + (2 − n2 )dr2 + r2 dΩ2
h i1/2 2
p
Tolman VII −B 2 cos ln 1 − 2r2 /a2 + r4 /b4 + r2 /a2 − b2 /a2 +θ
+(1 − 2r2 /a2 + 4 4 −1 2 2
r /b ) dr + r dΩ
2

−n2 −(n2 −8n+8)/n 2
Tolman VIII −A2 r2(n−1)(n−4)/n (n2 −4n+2)(n 2 −8n+8) + Br + Cr−2(n −4n+2)/n
−1
−n2 −(n2 −8n+8)/n 2
+ (n2 −4n+2)(n 2 −8n+8) + Br + Cr−2(n −4n+2)/n dr2 + r2 dΩ2
− (1 − 2m/r) dt2 + (1 − 2m/r) 1 + C(2r − 2m)2 dr2 + r2 dΩ2
−1
Kuch 68 II
p h p p i2
Kuch 68 I − A 1 − 2m/r + B r2 /m2 + 5r/m − 30 + 15 1 − 2m/r ln{1 − r/m − r(r − 2m)/m} dt2
+(1 − 2m/r)−1 dr2 + r2 dΩ2
M–W III −Ar(r − a) + 1−r7/4 2
2 /a2 dr + r dΩ
2 2
√
Kuch I b −r2 [A + B arctanh(a/ a2 + r2 )]2 dt2 + 2(1 + r2 /a2 )−1 dr2 + r2 dΩ2
Heint IIa (C=0) −(1 + ar2 )3 dt2 + [(1 + ar2 )/(1 − ar2 /2)]dr2 + r2 dΩ2
h i−1
3ar 2 Cr 2
Heint IIa −(1 + ar2 )3 dt2 + 1 − 2(1+ar 2) + √
(1+ar 2 ) 1+4ar 2
dr2 + r2 dΩ2
Table 3.2: Some well-known perfect fluid spheres and their coresponding metrics. Note
that we have often reparameterized these metrics to make them easier to deal with, and so
their appearance (but not the substance) may differ from other sources [9, 11, 20]. Metric S1
is a special case of K–O III, Tolman V, and Tolman VI, notable for its extreme simplicity.
99
Some apparently new perfect fluid solutions
Name Metric
2A+2B ln r+B
Martin 1 −(Ar + Br ln r)2 dt2 + 2 2A+2B ln r+B−Cr 2 dr
2
+ r2 dΩ2
2 7/3
−1
(r−a)r
Martin 2 −Ar(r − a)dt2 + 47 1 − ar 2 − B (4r−3a) 4/3 dr2 + r2 dΩ2
Martin 3 −(1 + ar2 )2 dt2 + [1 − br2 /(1 + 3ar2 )2/3 ]−1 dr2 + r2 dΩ2
2
2 −2
−(1 + ar2 )2 A + B √ (1+ar )
dt2 + [1 − br2 /(1 + 3ar2 )2/3 ]−1 dr2 + r2 dΩ2
R
P1 rdr
1−br 2 /(1+3ar 2 )2/3
−1/2 2
2 2A+2B ln r+B
rdr dt2
R
P2 −(Ar + Br ln r) σ + ǫ (Ar + Br ln r) −2
2 2A+2B ln r+B−Cr2
2A+2B ln r+B 2 2 2
+2 2A+2B 2 dr + r dΩ
ln r+B−Cr √ 2
2 3 (5+2ar 2 ) 1−ar 2 /2
P3 −(1 + ar ) A + B (1+ar 2 )3/2
dt2 + [(1 + ar2 )/(1 − ar2 /2)]dr2 + r2 dΩ2
2
−r4 A + B √ dr dt2 + (1 + λr2/3 )−1 dr2 + r2 dΩ2
R
P4
r 3 1+λr 2/3
h √ √ i2
P4 −r4 A + B 15 16 λ3
tanh −1
(1/ 1 + λr 2/3 ) − 1
16 1 + λr 2/3 (8r−2 − 10λr−4/3 + 15λ2 r−2/3 ) dt2
+(1 + λr2/3 )−1 dr2 + r2 dΩ2
√ √ 2
P5 −r A r − a + B r + a + 1−r7/4 2 2
2 /a2 dr + r dΩ
2
 2
7/3
−1
dr 7 r2 (r−a)r
dr2 + r2 dΩ2
R
P6 −r(r − a) A + B r
2 7/3
 +
4 1− a2 − B (4r−3a) 4/3
r (r−a)r
(r−a) 1− a 2 −B (4r−3a)4/3
h √ i2
a2 −2r 2 1+2r 2 /a2
−B 2 1 + r2 /a2 dt2 + 2
+ r2 dΩ2
R
P7 A+B √
b2 −r 2 (a2 −r 2 )3/2
rdr (1−r 2 /b2 )(1+r 2 /a2 ) dr
 2
h i−1
r dr 3ar 2 Cr 2
−(1 + ar2 )3 A + B  dt2 + 1 − dr2 + r2 dΩ2
R
P8 r
2(1+ar 2 ) + √
(1+ar 2 ) 1+4ar 2
3ar2 Cr2
(1+ar 2 )2 1− 2(1+ar2 ) + √
(1+ar2 ) 1+4ar2
Table 3.3: Some apparently new prefect fluid spheres and their coresponding metrics.
Sometimes the relevant integrals cannot be done in elementary form. When they can be
done they are explicitly shown.
100
Some seed geometries and their descendants
Seed Theorem1 Theorem2 Theorem 3 Theorem 4

Minkowski Einstein static K-O III interior Schwarzschild Martin 3
exterior Schwarzschild Kuch68 II Kuch 86 I [integral] [integral]
de Sitter Tolman IV interior Schwarzschild P7 [integral]
Tolman V (A = 0) Tolman V Tolman VI Wyman III Wyman IIa
S1 Tolman V (n = +1) K–O III P4 Martin 3
M–W III Martin 2 P5 P6 [integral]
Heint IIa (C=0) Heint IIa P3 P8 [integral]
Table 3.4: Seed solutions and their generalizations derived via theorems 1–4. The notation
“[integral]” denotes a metric so complicated that explicitly writing out the relevant integral
is so tedious that it does not seem worthwhile.
Some non-seed perfect fluid geometries and their descendants
Base Theorem1 Theorem2 Theorem 3 Theorem 4

Tolman IV Tolman IV P7 P7 [integral]
B-L B-L Kuchb I b Kuchb I b [integral]
Heint IIa Heint IIa P8 P8 [integral]
Tolman VI Wyman IIa Tolman VI [integral] Wyman IIa
Kuch1 Ib Martin 1 Kuch1 Ib P2 Martin 1
K–O III Martin 3 K–O III P1 Martin 3
Table 3.5: Non-seed solutions and their generalizations.
101
102
Chapter 4
Solution generating theorems —

other coordinates
Consider other coordinate systems: This chapter generalizes the theorems which we
derived for Schwarzschild coordinates to a number of other coordinate systems. This
may allow us to develop extra relations between the known solutions.
In all cases we give the metric and the pressure isotropy condition Gr̂r̂ = Gθ̂θ̂ .
Finch and Skea [11] estimate that about 55% of all work on fluid spheres is carried out
in Schwarzschild coordinates, that isotropic coordinates account for 35% of research,
and that the remaining 10% is spread over multiple special cases (see below).
4.1 General diagonal coordinates:

Consider the metric:
dr 2
ds2 = −ζ(r)2 dt2 + + R(r)2 dΩ2 (4.1)
B(r)
and then assume it represents a perfect fluid sphere. That is, Gθ̂θ̂ = Gr̂r̂ = Gφ̂φ̂ . The
equation Gθ̂θ̂ = Gφ̂φ̂ is always fulfilled due to spherical symmetry.
Calculate
ζB(R′ )2 + 2BRζ ′ R′ − ζ
Gr̂r̂ = (4.2)
R2 ζ
103
and
1 2Bζ ′R′ + 2ζR′′B + ζB ′ R′ + 2Rζ ′′ B + Rζ ′ B ′
Gθ̂θ̂ = − (4.3)
2 Rζ
We set Gr̂r̂ = Gθ̂θ̂ , which supplies us with an ODE. This ODE reduces the freedom to
choose the three functions in equation (4.1) to two:
[R(Rζ)′ ]B ′ + [2RR′′ ζ + 2R2 ζ ′′ − 2RR′ ζ ′ − 2(R′ )2 ζ]B + 2ζ = 0. (4.4)
This is first order-linear non-homogeneous in B(r), and second-order linear homoge-

neous in ζ(r). Note in particular it is linear in both ζ(r) and B(r) and the comment
below equation (16.10) in “exact solutions” [20] (while true) completely misses the
mark. There is nothing particularly nice about the R(r) dependence.
4.2 Schwarzschild (curvature) coordinates:

This is the basis of the theorems which were previously derived above.
Consider the metric
2 2 dr 2
2
ds = −ζ(r) dt + + r 2 dΩ2 (4.5)
B(r)
and the ODE
[r(rζ)′]B ′ + [2r 2 ζ ′′ − 2(rζ)′]B + 2ζ = 0. (4.6)
This is first order-linear non-homogeneous in B(r), and second-order linear homoge-
neous in ζ(r). Solving for B(r) in terms of ζ(r) is the basis of the [16], (and is the
basis for Theorem 2 in chapter 3).
Sometimes people prefer to write B(r) in terms of m(r)
dr 2
ds2 = −ζ(r)2 dt2 + + r 2 dΩ2 (4.7)
2m(r)
1− r
or even µ(r) = m/r 3 , but there is no qualitative gain in doing so.

If we re-group things in terms of ζ we find
2r 2 ζ ′′ + (r 2 B ′ − 2rB)ζ ′ + (rB ′ − 2B + 2)ζ = 0, (4.8)
104
which is a linear homogeneous 2nd order ODE. This is the basis of Theorems 1 and 3
in the above section.
4.3 Isotropic coordinates:

Consider the metric
1
ds2 = −ζ(r)2 dt2 + {dr 2 + r 2 dΩ2 } (4.9)
ζ(r)2 B(r)2
Calculate
Gr̂r̂ = (ζ ′ )2 B 2 − (B ′ )2 ζ 2 + 2B ′ Bζ 2 /r (4.10)
and
Gθ̂θ̂ = −(B ′ )2 ζ 2 + ζ 2BB ′′ − (ζ ′)2 B 2 + B ′ Bζ 2/r (4.11)
We set Gr̂r̂ = Gθ̂θ̂ , which supplies us with an ODE. This ODE reduces the freedom to
choose the two functions in equation (4.9) to one:
′ 2
ζ B ′′ − B ′ /r
= . (4.12)
ζ 2B
R
There are several ways of improving this. If we write ζ(r) = exp( g(r)dr) then we
have an algebraic equation for g(r):
r
B ′′ − B ′ /r
g(r) = ± . (4.13)
2B
Conversely, the isotropy condition can be written in terms of B(r) as:
B ′′ − B ′ /r − 2g 2 B = 0. (4.14)
This is about as simple as you can make things.

R
There is an improvement in writing B(r) = exp(2 h(r)dr) so that
g(r)2 = 2h(r)2 + h′ (r) − h(r)/r (4.15)
which is the basis of [18].
105
4.3.1 The seventh and eighth theorems
Theorem 7. In Isotropic coordinates, if {ζ(r), B(r)} describe a perfect fluid then so
does {ζ(r)−1, B(r)}. This is the Buchdahl transformation in disguise.
That is, the geometry defined by holding B0 (r) fixed and setting
2 1 2 ζ0 (r)2
ds = − 2
dt + 2
{dr 2 + r 2 dΩ2 } (4.16)
ζ0 (r) B0 (r)
is also a perfect fluid sphere. Alternatively, the mapping
T7 : {ζ0 , B0 } 7→ ζ0−1, B0

(4.17)

Proof for Theorem 7: Assuming that {ζ0 (r)−1 , B0 (r)} solves equation (4.12).
We consider L.H.S. of the equation (4.12),
′ 2 ′ !2 !2 2 ′ 2
ζ0 ζ0−1 −ζ0′ ζ0−2 −ζ0′ ζ0
ζ0 → ζ1 : → −1 = −1 = = (4.18)
ζ0 ζ0 ζ0 ζ0 ζ0
This shows that {ζ(r), B(r)} describe a perfect fluid then so does {ζ(r)−1, B(r)}.
The transformation T7 defined in Theorem 7 is a “square root of unity” in the
sense that:
T7 ◦ T7 = I, (4.19)
To see this, consider the sequence
{ζ0, B0 } 7→ {ζ1 , B0 } 7→ {ζ2 , B0 } . . . (4.20)
In this section we consider a two-step composition to check that T7 is a square root of

unity. In particular, to be more explicit about this note that at step one
ζ1 = ζ0−1 (4.21)

ζ2 = ζ1−1 = (ζ0−1)−1 = ζ0 (4.22)
For reasons of convenience and without loss of generality, we can write for ζ(r), after
applying thorem 1 n-times, as : ζ(r) = ζ(r)−1, for n = odd number and ζ(r) = ζ(r),
for n = even number.
106
Theorem 8. Let {ζ0 (r), B0 (r)} describe a perfect fluid sphere. Define

rdr
Z
Z0 = σ + ǫ . (4.23)
B0 (r)2
Then for all σ and ǫ, the geometry defined by holding ζ0 (r) fixed and setting
1
ds2 = −ζ0 (r)2 dt2 + {dr 2 + r 2 dΩ2 } (4.24)
ζ0 (r)2 B0 (r)2 Z0 (r)2
T8 (σ, ǫ) : {ζ0 , B0 } 7→ {ζ0 , B0 Z0 (B0 )} (4.25)
in order”. Assuming that {ζ0 (r), B0 (r)} solves equation (4.14), write
B1 (r) = B0 (r) Z0 (r) . (4.26)
and demand that {ζ0 (r), B1 (r)} also solves equation (4.14). We find
(B0 Z0 )′′ − (B0 Z0 )′ /r − 2g 2(B0 Z0 ) = 0. (4.27)
(B0′′ Z0 + 2B0′ Z0′ + B0 Z0′′ ) − (B0′ Z0 )/r − (B0 Z0′ )/r − 2g 2 (B0 Z0 ) = 0. (4.28)

′′
B0 − B0′ /r − 2g 2B0 Z0 + 2B0′ Z0′ + B0 Z0′′ − (B0 Z0′ )/r = 0.

(4.29)
A linear homogeneous 2nd order ODE for B now simplifies to
(2B0′ − B0 /r)Z0′ + B0 Z0′′ = 0 , (4.30)

on Z0′ and Z0′′ . (So it can be viewed as a first-order homogeneous order differential
equation in Z ′ , which is solvable.) Separating the unknown variable to one side,
Z0′′ B0′ 1
′
= −2 + . (4.31)
Z0 B0 r
107
Re-write Z0′′ /Z0′ = d ln(Z0′ )/dr, and integrate twice over both sides of equation (4.31),
to obtain
r dr
Z
Z0 (r) = σ + ǫ . (4.32)
B0 (r)2
depending on the old solution {ζ0 (r), B0 (r)}, and two arbitrary integration constants
σ and ǫ.
“Idempotence” Proof for Theorem 8. To see that the transformation T8 defined in

Theorem 8 is “idempotent” we first show
T8 ◦ T8 , T8 , (4.33)

ǫ2
T8 (σ2 , ǫ2 ) ◦ T8 (σ1 , ǫ1 ) = T8 σ2 σ1 , ǫ1 σ2 + . (4.34)
σ1
In this section we consider a two-step composition rule to find idempotence. In par-
ticular, the most outstanding features of these steps are:
• To see “idempotence”, note that for fixed ζ0 (r) equation (4.14) has a solution

 Z 
 r dr 
B0 → B1 = B0 σ1 + ǫ1 , (4.35)
 B0 (r)2 

 Z 
 r dr 
B1 → B2 = B1 σ2 + ǫ2 . (4.36)
 B1 (r)2 
108
That is:
 Z 
 r dr 
B2 = B0 σ + ǫ1 (4.37)
 1 B0 (r)2 
 

 


 Z 

 
 r dr 
× σ2 + ǫ2 Z .

 

B0 (r)2 [σ1 + ǫ1 r dr/B0 (r)2 ]2 

 

 


  Z
ǫ2
 dZ0 
B2 = B0 Z0 σ2 +
 ǫ1 Z02 
Z
ǫ2 1
= B0 Z0 σ2 − d
ǫ1 Z0

ǫ2 1 1
= B0 Z0 σ2 − − . (4.38)
ǫ1 Z0 σ1
Therefore
ǫ2 ǫ2 1
B2 = B0 − + σ2 + Z0 . (4.39)
ǫ1 ǫ1 σ1
That is
ǫ2 ǫ2 1
B1 = − + σ2 + Z0 , (4.40)
ǫ1 ǫ1 σ1
from which the composition law follows:

ǫ2
T8 (σ2 , ǫ2 ) ◦ T8 (σ1 , ǫ1 ) = T8 σ2 σ1 , ǫ1 σ2 + (4.41)
σ1
Note that this almost identical to the computation performed in the Schwarzschild
coordinates because idempotence of the theorem is a result of “reduction of order” not
something special to general relativity.
109
4.3.2 Two linking generating theorems:
Having now found the first and second generating theorems, is now useful to find the
connection between them. We start by taking a perfect fluid sphere solution {ζ0 , B0 }
and applying theorem 7 onto it which gives us a new perfect fluid sphere {ζ1 , B0 }. The
new ζ1 = ζ0−1. We now continue with applying theorem 8, again we get a new solution
{ζ0−1, B1 }, where B1 now depends on the B0 . All together we can consider this as a
single process by introducing the following corollary:
Corollary 3. If {ζ0, B0 } denotes a perfect fluid sphere, then for all σ and ǫ, the
two-parameters geometry defined by
2 1 2 ζ0 (r)2
ds = − 2
dt + 2 2
{dr 2 + r 2 dΩ2 } (4.42)
ζ0 (r) B0 (r) Z0 (r)
is also a perfect fluid sphere, where Z0 is

rdr
Z
Z0 (r) = σ + ǫ . (4.43)
B0 (r)2
That is
T8 ◦ T7 : {ζ0 , B0 } 7→ {ζ0−1, B0 } 7→ {ζ0−1 , B0 Z0 (B0 )} (4.44)
Instead of starting with theorem 7 we could first apply theorem 8 on {ζ0 , B0 }. This
gives us a new perfect fluid sphere {ζ0 , B1 }, where B1 = B0 Z0 (B0 ) is given by equa-
tion (4.32). We now continue with theorem 7 which leads to {ζ1 , B1 }, where ζ1 = ζ −1.
Again, we can consider this as a single process by introducing the following corollary:
Corollary 4. If {ζ0, B0 } denotes a perfect fluid sphere, then for all σ and ǫ, the
two-parameters geometry defined by
1 ζ0 (r)2
ds2 = − dt2
+ {dr 2 + r 2 dΩ2 } (4.45)
ζ0 (r)2 B0 (r)2 Z0 (r)2
is also a perfect fluid sphere, where Z0 is

rdr
Z
Z0 (r) = σ + ǫ . (4.46)
B0 (r)2
110
That is
T7 ◦ T8 : {ζ0 , B0 } 7→ {ζ0, B0 Z0 (B0 )} 7→ {ζ0−1, B0 Z0 (B0 )} (4.47)
There are general comments about the relationship between both transformation
theorems.
• These two transformation operations are independent of each other.
• The transformation T7 defined in Theorem 7 is a “square root of unity”.
• Theorem 8 is in general idempotent. That means when we apply theorem 8 more

than once no further solution will be obtained.
• Theorem 7 and theorem 8 are in general commute.
In addition, these comments can be described in terms of the equations below.
T7 6, T8 ; T7 ◦ T7 , I; T8 ◦ T8 , T8 ; T7 ◦ T8 , T8 ◦ T7 . (4.48)
4.4 Gaussian polar coordinates:

Consider the metric
ds2 = −ζ(r)2 dt2 + dr 2 + R(r)2 dΩ2 (4.49)
and then assume it represents a perfect fluid sphere. That is, Gθ̂θ̂ = Gr̂r̂ = Gφ̂φ̂ . Now
Gθ̂θ̂ = Gφ̂φ̂ is always fulfilled due to spherical symmetry.
We calculate
2ζ ′RR′ − ζ + ζ(R′)2
Gr̂r̂ = − (4.50)
R2 ζ
and
ζ ′ R′ + R′′ ζ + ζ ′′ R
Gθ̂θ̂ = − (4.51)
Rζ
111
Figure 4.1: This structure shows that Theorem 7 and Theorem 8 are in general distinct.
When we apply Theorem 7 to seed metric, we get a new solution, while applying Theorem 8
to seed metric, we get the other new solution. Furthermore, the diagram shows that Theorem
7 and Theorem 8 in general commute.
We set Gr̂r̂ = Gθ̂θ̂ , which supplies us with an 2nd order ODE for ζ(r), which reduces
the freedom to choose the two functions to one:
ζ ′′R2 − ζ ′RR′ + ζ[1 − (R′ )2 + R′′ R] = 0 (4.52)
We can rewrite this equation as

′
1 − (R′ )2 + R′′ R

′′ ′R
ζ −ζ +ζ =0 (4.53)
R R2
4.4.1 The ninth theorem

Theorem 9. Suppose {ζ0 (r), R0 (r)} represents a perfect fluid sphere. Define

R0 (r)dr
Z
Λ0 (r) = σ + ǫ . (4.54)
ζ0 (r)2
112
Then for all σ and ǫ, the geometry defined by holding R0 (r) fixed and setting
ds2 = −ζ0 (r)2 Λ0 (r)2 dt2 + dr 2 + R0 (r)2 dΩ2 (4.55)
T9 (σ, ǫ) : {ζ0 , R0 } 7→ {ζ0 Λ0 (ζ0 , R0 ), R0 } (4.56)
Proof for Theorem 9. The proof of theorem 9 is again based on the technique of “re-
duction in order”. Assuming that {ζ0 (r), R0 (r)} solves again equation (4.52), write
ζ1 (r) = ζ0 (r) Λ0 (r) . (4.57)

and demand that {ζ1 (r), R0 (r)} also solves equation (4.52). We find
′
1 − (R′ )2 + R′′ R

′′ ′R
(ζ0 Λ0 ) − (ζ0 Λ0 ) + (ζ0 Λ0 ) =0 (4.58)
R R2
′
1 − (R′ )2 + R′′ R

R
(ζ0′′ Λ0 + 2ζ0′ Λ′0 + ζ0 Λ′′0 ) − (ζ0′ Λ0 + ζ0 Λ′0 ) + (ζ0 Λ0 ) = 0 (4.59)
R R2
′
1 − (R′ )2 + R′′ R ′

′′ ′R ′ ′ ′′ ′ R
ζ0 − ζ0 + ζ0 Λ0 + 2ζ0 Λ0 + ζ0 Λ0 − ζ0 Λ0 = 0. (4.60)
R R2 R
A linear homogeneous 2nd order ODE for ζ now simplifies to
R′ ′
(2ζ0′ − ζ0 )Λ + ζ0 Λ′′0 = 0 , (4.61)
R 0
on Λ′0 and Λ′′0 . (So it can be viewed as a first-order homogeneous order differential
equation in Λ′0 , which is solvable.) Separating the unknown variable to one side,
Λ′′0 ζ0′ R0′
= −2 + . (4.62)
Λ′0 ζ0 R0
113
Re-write Λ′′0 /Λ′0 = d ln(Λ′0 )/dr, and integrate twice over both sides of equation (4.62),
to obtain
R0 (r) dr
Z
Λ0 (r) = σ + ǫ . (4.63)
ζ0 (r)2
depending on the old solution {ζ0 (r), R0 (r)}, and two arbitrary integration constants
σ and ǫ.
“Idempotence” Proof for Theorem 9. To see that the transformation T9 defined in

Theorem 3 is “idempotent” we first show
T9 ◦ T9 , T9 , (4.64)

ǫ2
T9 (σ2 , ǫ2 ) ◦ T9 (σ1 , ǫ1 ) = T9 σ2 σ1 , ǫ1 σ2 + . (4.65)
σ1
In this section we consider a two-step composition rule to find idempotence. In par-
ticular, the most outstanding features of these steps are:
• To see “idempotence”, note that for fixed R0 (r) equation (4.53) has a solution

 Z 
 R0 (r) dr 
ζ0 → ζ1 = ζ0 σ1 + ǫ1 , (4.66)
 ζ0 (r)2 

 Z 
 R0 dr 
ζ1 → ζ2 = ζ1 σ2 + ǫ2 . (4.67)
 ζ1 (r)2 
114
That is:
 Z 
 R0 (r) dr 
ζ2 = ζ0 σ1 + ǫ1 (4.68)
 ζ0 (r)2 
 

 


 Z 

 
 R0 (r) dr 
× σ2 + ǫ2 Z .

 

ζ0 (r)2 [σ1 + ǫ1 R0 (r) dr/ζ0 (r)2 ]2 

 

 


 Z 
 ǫ2 dZ0 
ζ2 = ζ0 Z0 σ2 +
 ǫ1 Z02 
Z
ǫ2 1
= ζ0 Z0 σ2 − d
ǫ1 Z0

ǫ2 1 1
= ζ0 Z0 σ2 − − . (4.69)
ǫ1 Z0 σ1
Therefore
ǫ2 ǫ2 1
ζ2 = ζ0 − + σ2 + Z0 . (4.70)
ǫ1 ǫ1 σ1
That is
ǫ2 ǫ2 1
ζ1 = − + σ2 + Z0 , (4.71)
ǫ1 ǫ1 σ1
from which the composition law follows:

ǫ2
T9 (σ2 , ǫ2 ) ◦ T9 (σ1 , ǫ1 ) = T9 σ2 σ1 , ǫ1 σ2 + (4.72)
σ1
Again, note simularities to what happens in Schwarzschild coordinates.
115
Figure 4.2: This structure shows that Theorem 9 is idempotent. When we apply Theorem
9 to Minkowski as a seed metric, we get the K-O III as a new solution.
4.5 Buchdahl coordinates:

Consider the metric
ds2 = −ζ(r)−1 dt2 + ζ(r)dr 2 + ζ(r)R(r)2dΩ2 (4.73)
and then assume it represents a perfect fluid sphere. That is, Gθ̂θ̂ = Gr̂r̂ = Gφ̂φ̂ . Note
Gθ̂θ̂ = Gφ̂φ̂ is always fulfilled due to spherical symmetry. We calculate
1 −4 ζ 2 (R′ )2 + 4 ζ 2 + (ζ ′)2 R2

Gr̂r̂ = (4.74)
4 ζ 3 R2
and
1 4 ζ 2 R′′ + (ζ ′)2 R

Gθ̂θ̂ = − (4.75)
4 ζ3 R
the freedom to choose the two functions to one:
′ 2
ζ 2[1 − (R′ )2 + RR′′ ]
=− . (4.76)
ζ R2
This coordinate system is a sort of cross between isothermal and proper radius.
Theorem 10. If {ζ(r), R(r)} describes a perfect fluid then so does {ζ(r)−1, R(r)}.
This is the Buchdahl transformation in yet another disguise.
The geometry defined by holding R0 (r) fixed and setting
1 R0 (r)2 2
ds2 = −ζ0 (r) dt2 + dr 2 + dΩ (4.77)
ζ0 (r) ζ0 (r)
116
T10 : {ζ0 , R0 } 7→ ζ0−1 , R0

(4.78)
4.6 Synge isothermal coordinates:

Consider
ds2 = −ζ(r)−2 {dt2 − dr 2 } + {ζ(r)−2 R(r)2 dΩ2 } (4.79)
and then assume it represents a perfect fluid sphere. That is, Gθ̂θ̂ = Gr̂r̂ = Gφ̂φ̂ . While
Gθ̂θ̂ = Gφ̂φ̂ is always fulfilled due to spherical symmetry. We calculate
3R2 (ζ ′ )2 − 4RζR′ ζ ′ + (R′ )2 ζ 2 − ζ 2

Gr̂r̂ = − (4.80)
R2
and
3R(ζ ′)2 − 2ζR′ ζ ′ − 2R ζζ ′′ + ζ 2 R′′
Gθ̂θ̂ = − (4.81)
R
the freedom to choose the two functions in equation
′
1 − (R′ )2 + RR′′

′′ ′R
ζ −ζ −ζ = 0. (4.82)
R 2R2
This is very similar equation to Gaussian polar.
Theorem 11. Suppose {ζ0 (r), R0 (r)} represents a perfect fluid sphere. Define

R0 (r)dr
Z
A0 (r) = σ + ǫ . (4.83)
ζ0 (r)2
Then for all σ and ǫ, the geometry defined by holding R0 (r) fixed and setting
1 R0 (r)2
ds2 = −
2 2
dΩ2

2 2
dt − dr + 2 2
(4.84)
ζ0 (r) A0 (r) ζ0 (r) A0 (r)
117
T11 (σ, ǫ) : {ζ0 , R0 } 7→ {ζ0 A0 (ζ0 , R0 ), R0 } (4.85)
in order”. Assuming that {ζ0 (r), R0 (r)} solves equation (4.82), write
ζ1 (r) = ζ0 (r) A0 (r) . (4.86)
and demand that {ζ1 (r), R0 (r)} also solves equation (4.82). We find
′
1 − (R′ )2 + RR′′

′′ ′R
(ζ0 A0 ) − (ζ0 A0 ) + (ζ0 A0 ) =0 (4.87)
R 2R2
′
1 − (R′ )2 + RR′′

R
(ζ0′′ A0 + 2ζ0′ A′0 + ζ0 A′′0 ) − (ζ0′ A0 + ζ0 A′0 ) + (ζ0 A0 ) = 0 (4.88)
R 2R2
′
1 − (R′ )2 + RR′′ ′

′′ ′R ′ ′ ′′ ′ R
ζ0 − ζ0 + ζ0 A0 + 2 ζ0 A0 + ζ0 A0 − ζ0 A0 = 0. (4.89)
R 2R2 R
A linear homogeneous 2nd order ODE for ζ now simplifies to
R′ ′
)A + ζ0 A′′0 = 0 ,
(2ζ0′ − ζ0 (4.90)
R 0
on A′0 and A′′0 . (So it can be viewed as a first-order homogeneous order differential
equation in A′0 , which is solvable.) Separating the unknown variable to one side,
A′′0 ζ0′ R0′
= −2 + . (4.91)
A′0 ζ0 R0
Re-write A′′0 /A′0 = d ln(A′0 )/dr, and integrate twice over both sides of equation (4.91),
to obtain
R0 (r) dr
Z
A0 (r) = σ + ǫ . (4.92)
ζ0 (r)2
depending on the old solution {ζ0 (r), R0 (r)}, and two arbitrary integration constants
σ and ǫ.
118
4.7 Exponential coordinates:
You can always choose to effectively use gtt itself as one of the coordinates, by picking
the functional form of gtt to be any arbitrarily specified function f (z). For instance,
choose
gtt = − exp(−2z) (4.93)
then a useful choice is
dz 2

2 2
ds = − exp(−2z) dt + exp(+2z) + R(z)2 dΩ2 (4.94)
B(z)
and then assume it represents a perfect fluid sphere. That is, Gθ̂θ̂ = Gẑ ẑ = Gφ̂φ̂ . Note
Gθ̂θ̂ = Gφ̂φ̂ is always fulfilled due to spherical symmetry. Calculate
(BR2 − (R′ )2 B + 1) exp(−2z)

Gẑ ẑ = (4.95)
R2
and
1 (B ′ R′ + 2R′′ B + 2BR) exp(−2z)
Gθ̂θ̂ = − (4.96)
2 R
We set Gẑ ẑ = Gθ̂θ̂ , which supplies us with an 1st order ODE for B, which reduces
the freedom to choose the two functions in the metric to one for which the isotropy
constraint becomes
B ′ [RR′ ] + B[4R2 − 2(R′ )2 + 2RR′′ ] + 2 = 0 (4.97)
Theorem 12. Suppose {B0 (z), R0 (z)} represents a perfect fluid sphere. Define
R0 (z)2
Z
R0 (z)
Υ0 (z) = exp −4 dz . (4.98)
(R0′ (z))2 R0′ (z)
Then for all λ, the geometry defined by holding R0 (z) fixed and setting
dz 2

2 2 2 2
ds = − exp(−2z)dt + exp(+2z) + R0 (z) dΩ (4.99)
B0 (z) + λΥ0 (z)
T12 (σ, ǫ) : {B0 , R0 } 7→ {B0 Υ0 (R0 ), R0 } (4.100)
119
Proof for Theorem 12. Assume that {B0 (z), R0 (z)} is a solution for equation (4.97).
We want to find under what conditions does {B1 (z), R0 (z)} also satisfy equation
(4.97)? Without loss of generality, we write
B1 (z) = B0 (z) + λ Υ0 (z) . (4.101)
Equation (4.97) can now be used to determine Υ0 (z). Substitute B1 (z) in Equation
(4.97)
(B0 + λ Υ0 )′ [R0 R0′ ] + (B0 + λ Υ0 )[4R02 − 2(R0′ )2 + 2R0 R0′′ ] + 2 = 0 (4.102)
[R0 R0′ ]B0′ + [4R02 − 2(R0′ )2 + 2R0 R0′′ ]B0 + 2

+ [R0 R0′ ] Υ′0 + 4R02 − 2(R0′ )2 + 2R0 R0′′ Υ0 = 0

(4.103)
That ordinary inhomogeneous first-order differential equation in B now simplifies to
[R0 R0′ ] Υ′0 + 4R02 − 2(R0′ )2 + 2R0 R0′′ Υ0 = 0 ,

(4.104)
which is an ordinary homogeneous first-order differential equation in Υ0 .

Now we can calculate Υ0 by following,
[R0 R0′ ] Υ′0 = − 4R02 − 2(R0′ )2 + 2R0 R0′′ Υ0 .

(4.105)
Rearrage it into
Υ′0 − [4R02 − 2(R0′ )2 + 2R0 R0′′ ]

=
Υ0 [R0 R0′ ]
R0 R′ R′′
= −4 ′ + 2 0 − 2 0′ (4.106)
R0 R0 R0
A straightforward calculation, including an integration by parts, leads to
R02
Z
R0
Υ0 (z) = exp −4 dz . (4.107)
(R0′ )2 R0′
120
4.8 Discussion
We have developed several transformation theorems that map perfect fluid spheres
to perfect fluid sphere using other coordinates. We derived the seventh and eighth
theorems using Isotropic coordinates. The transformation in Theorem 7 is a “square
root of unity”. Indeed, Theorem 8 is alomost identical to the calculation performed
in the Schwarzschild coordinates. In addition, the idempotence of this theorem is a
result of “reduction of order”, which is not special to general relativity. We are able to
classify the connection between both transformation theorems. The two transforma-
tion operations are independent of each other. Furthermore, both theorems in general
commute. Figure 4.1 shows that after applying Theorem 7 to the seed metric, we get a
new solution while applying Theorem 8 to seed metric, we get the other new solution.
By the idempotence of Theorem 8 n applications of T8 to the seed metric still results
in the same non-seed metric. Furthermore, the diagram shows that Theorem 7 and
Theorem 8 in general commute.
In Gaussian polar coordinates, we established another transformation theorem,
which has many similarities to the results in Schwarzschild coordinates. Figure 4.2
shows that Theorem 9 is idempotent. When we apply Theorem 9 to Minkowski as a
seed metric, we get the K-O III as a new solution.
Furthermore, we have also established the transformation theorem using Buchdahl
coordinates. Theorem 10 is the Buchdahl transformation in yet another disguise.
In Synge isothermal coordinates, we presented Theorem 11 is a transformation the-
orem. However, this transformation is almost identical to the computation performed
in the Schwarzschild coordinates.
The last solution generating theorem we presented using exponential coordinates.
Indeed, this Theorem 12 also map perfect fluid spheres to perfect fluid spheres.
121
122
Chapter 5
TOV equation - related theorems
5.1 Introduction
The Tolman-Oppenheimer-Volkoff (TOV) equations describe the hydrostatic equilib-
rium of perfect fluid stars without rotation in general relativity. We can solve the
TOV equations by constructing the equation of state for the entire region of nonzero
density starting from the higher density at center to the surface density. In this chap-
ter, we can rephrase Theorems 1, 2, 3 and 4 directly in terms of the TOV equation.
In addition, we also can rephrase all these theorems, which apply in Schwarzschild
coordinates directly in terms of the pressure profile and density profile.
5.2 TOV equation

5.2.1 Interior equation
In general relativity [24], continuous matter distributions and fields again are described
by a stress-energy tensor Tab . The stress tensor of a perfect fluid is given by
T ab = ρ ua ub + p (g ab + ua ub ) (5.1)
where ρ(r) is density of mass-energy distribution in rest-frame of fluid, u is 4-velocity

of fluid, and p is isotropic pressure in rest-frame of fluids.
123
We know that
Gab = 8πGN Tab (5.2)
We set GN = 1 for convenience, and note that:
∇a Gab = 0 (5.3)
We know that as contracted Bianchi identity.
The equation of motion for the matter is
∇b T ab = 0 (5.4)
We define
dxa
ua = (5.5)
dτ
This is the 4-velocity of a fluid element. We also define
d2 xa dua
Aa = = (5.6)
dτ 2 dτ
This is the 4-accerelation.
−
→
In the rest frame of the fluid element ua = (c, 0 ) and, Aa = (0, → −
a ).
a
It follows that ua A = 0.
It is an invariant quantity. In other words, the 4-acceleration of a particle is always
orthogonal to its 4-velocity. Writing
d
= (u ∇) (5.7)
ds
We see
d d d
0= (−1) = (gab ua ub ) = 2 ua ua = 2 ua (u∇)ua = 2 uaAa = 0 (5.8)
ds ds ds
Now consider ∇b T ab ; we can write this as
∇b T ab = ∇b ρua ub + p(g ab + ua ub )

= (∇b ρ)ua ub + (∇b p)(g ab + ua ub )

+(ρ + p)(∇b ua )ub + (ρ + p)ua (∇b ub )
= (∇b ρ)ub + (ρ + p)(∇ · u) ua

+ (ρ + p)Aa + ∇b p(g ab + ua ub )

= 0 (5.9)
124
Writing out equation (5.4) in terms of ρ, p, and ub , and projecting the resulting
equation parallel and perpendicular to ua , we can spilt up the equation into two
equations:
(∇b ρ)ub + (ρ + p)(∇ · u) ua = 0,

(5.10)
(ρ + p)Aa + ∇b p(g ab + ua ub ) = 0.

(5.11)
with ua ua = −1, and Aa ua = 0, because they are orthogonal to each other.
Figure 5.1: The diagram shows that when we write out equation (5.4) in terms of ρ, p, and
ub , and projecting the resulting equation parallel and perpendicular to ua , we can spilt up
the equation in to two equations.
Now use (∇b p) g ab ua = (∇b p)ub , and (∇b p) ua ub ua = −(∇b p)ub . This means
(∇b p) g abua and (∇b p) ua ubua have the same magnitude but opposite direction. We
conclude
∇ · (ρu) + p ∇ · u = 0 (5.12)
and
(g ab + ua ub ) ∇b p
Aa = − . (5.13)
ρ+p
125
5.3 A perfect fluid sphere in Schwarzschild coordi-
nates
In the Schwarzschild cooridinates [24], the metric of an arbitrary static, spherically
symmetric spacetime can be put into the simple form
ds2 = −f (r) dt2 + h(r) dr 2 + r 2 (dθ2 + sin2 θ dφ2 ) (5.14)
when GN = 1, then Gab = 8πTab .

The Einstein equations yield,
1
8πρ = G00 = R00 + (R0 0 + R1 1 + R2 2 + R3 3 )
2
= (rh2 )−1 h′ + r −2 (1 − h−1 ), (5.15)
1
8πp = G11 = R11 − (R0 0 + R1 1 + R2 2 + R3 3 )
2
= (rf h)−1 f ′ − r −2 (1 − h−1 ), (5.16)
1 d
8πp = G22 = (f h)−1/2 [(f h)−1/2 f ′ ]
2 dr
1 1
+ (rf h)−1 f ′ − (rh2 )−1 h′ . (5.17)
2 2
We assume h is −1
2m(r)
h(r) = 1 − , (5.18)
r
and if we write
f = exp(2φ), (5.19)
the equation (5.16) becomes
dφ [m(r) + 4 πp(r) r 3 ]
= 2 . (5.20)
dr r [1 − 2m(r)/r]
126
Proof for hydrostatic equilibrium . We substitute equations (5.18) and (5.19) into equa-
tion (5.16) we get,
8πp = (rf h)−1 − r −2 (1 − h−1 )

f′ 1 (1 − h−1 )
= −
f rh r2
2m(r) 2m(r)
dφ 1 − r
= 2 − r2
dr r r
dφ r − 2m(r) 2m(r)
= 2 −
dr r2 r3
dφ r − 2m(r) 2m(r)
2 2
= 8πp +
dr r r3
m(r)
dφ 4πp + 3
= r
dr r − 2m(r)
r2
[m(r) + 4 πp r 3 ]
= 2 (5.21)
r [1 − 2m(r)/r]
as previously illustrated.
To see what happens to the pressure and density it will be useful to consider the
TOV equation.
dp −[ρ(r) + p(r)] [m(r) + 4πp(r) r 3 ]
= (5.22)
dr r 2 [1 − 2m(r)/r]
Proof. Consider the stress energy of a perfect fluid is in the equation (5.1) and it
satisfies the equation (5.4) which yield,
(p + ρ)ua ∇a ub + (gab + ua ub )∇a p = 0 (5.23)
Now consider,
(e1 )b ∇a T ab = 0, (5.24)
where, (e1 )b = h1/2 (dr)b = h−1/2 (∂/∂r)b

We can reduce the equation (5.24) into the equation (5.25) because (e1 )b is pointing
127
in the r-direction, which is perpendicular to ua .
(e1 )b [(p + ρ) ua ∇a ub + (δab + ua ub) ∇a p] = 0 (5.25)

h−1/2 (0, 1, 0, 0)b[(p + ρ) ua ∇a ub + (δab + ua ub) ∇a p] = 0
dp
h−1/2 (p + ρ) (ua ∇a ub )b + h−1/2 = 0
dr
Now we consider,
(ua ∇a ub ) = ua ∂a ub + Γbca uc = ua Γbca uc

1 b 1 1 bc
= Γtt = g [−gtt,e + gte,t + get,t ]
f f 2
1 be 1 1 1 b
= g (f ),e = ∇f
f 2 2 f
dφ
= ∇b φ = (5.26)
dr
Now we substitute equation (5.20) into equation (5.25) we get,
dφ dp
h−1/2 (p + ρ) + h−1/2 = 0
dr dr
dp dφ
= −(p + ρ)
dr dr
dp [m(r) + 4πp(r) r 3 ]
= −(p + ρ) (5.27)
dr r 2 [1 − 2m(r)/r]
Equation (5.27) is known as the Tolman-Oppenheimer-Volkoff equation of hydrostatic

equilibrium.
This equation applies only for a perfect fluid (pr = pt ) in Schwarzschild coor-
dinates. There are generalizations appropriate to other coordinate systems. In the
non-relativistic Newtonian approximation this simplifies to (p ≪ ρ, 2m(r)/r ≪ 1):
dp [ρ(r) m(r)]
=− (5.28)
dr r2
which is something that is very easy to derive in Newton’s theory of gravity.
If we assume ρ(r) and hence m(r) are given, then the TOV is a nonlinear first-order
128
ODE for p(r) called a Riccatti equation. Given p(0) a unique solution p(r) necessarily
exists, and can be used to reconstruct the function ζ(r) via:
[m(r) + 4 πp(r) r 3 ]
Z
ζ(r) = exp dr (5.29)
r 2 [1 − 2m(r)/r]
Proof. Firstly, start with some static spherically symmetric geometry in Schwarzschild
dr 2
ds2 = −ζ(r)2 dt2 + + r 2 dΩ2 (5.30)
B(r)
and then assume it represents a perfect fluid sphere.
When
2Bζ ′r − ζ + Bζ
8πp = Grr =
r2ζ
2B ζ ′ 1 B
8πp = − 2+ 2
r ζ r r
′
2B ζ 1 B
= 8πp + 2 − 2
r ζ r r
′
ζ 4πpr (1 − B)
= + (5.31)
ζ B 2Br
when B = 1 − 2m(r)/r, then we get
ζ′ 4πpr 1 − (1 − 2m(r)/r)
= +
ζ (1 − 2m(r)/r) 2 (1 − 2m(r)/r)r
′ 3
ζ m(r) + 4πp r
= 2
ζ r (1 − 2m(r)/r)
d ln(ζ) m(r) + 4πp r 3
= 2
dr r (1 − 2m(r)/r)
m(r) + 4πp r 3
Z
ln(ζ) = dr
r 2 (1 − 2m(r)/r)
[m(r) + 4πp r 3 ]
Z
exp(ln(ζ)) = exp dr
r 2 [1 − 2m(r)/r]
[m(r) + 4πp r 3 ]
Z
ζ = exp dr (5.32)
r 2 [1 − 2m(r)/r]
as required.
129
In this way, starting from ρ(r) you can construct first the function m(r) = 4π ρ(r)r 2 dr,
R
then (via the TOV) the pressure profile p(r), and finally via the above the gravitational
redshift ζ(r).
Suppose in contrast that the pressure profile p(r) is given, then the TOV equation
can be rearranged into a nonlinear first order PDE for m(r):
1 dm(r) r 2 [1 − 2m(r)/r] dp
= −p(r) − (5.33)
4πr 2 dr [m(r) + 4π p(r) r 3 ] dr
Proof. We substitute equations (5.18) and (5.19) into equation (5.15) to get,
dm(r)
= 4πρ(r)r 2 , (5.34)
dr
and rearrange in term of ρ(r) so,
1 dm(r)
ρ(r) = (5.35)
4πr 2 dr
rearrange the equation (5.22) into,
r 2 [1 − 2m(r)/r] dp
ρ(r) = −p(r) − (5.36)
[m(r) + 4π p(r) r 3 ] dr
now substitute equation (5.35) into equation (5.36), we get
1 dm(r) r 2 [1 − 2m(r)/r] dp
= −p(r) − (5.37)
4πr 2 dr [m(r) + 4π p(r) r 3 ] dr
This is an Abel equation (2nd type, class A). There is no simple general solution.
However given p(r) and the initial data m(0) = 0 this can in principle be solved to
determine m(r). Once this is done ζ(r) is again reconstructed via
[m(r) + 4πp(r) r 3 ]
Z
ζ(r) = exp dr (5.38)
r 2 [1 − 2m(r)/r]
130
5.4 Solution generating theorems
5.4.1 Four theorems
The first four theorems we presented in chapter 3 were first explicitly published in
[6]. We now rephrase the theorems in slightly different formalism, and demonstrate
an independent way of proving them. Indeed, we can re-write these theorems directly
in terms of pressure and density profiles p(r) and ρ(r).
Firstly, start with some static spherically symmetric geometry in Schwarzschild
dr 2
ds2 = −ζ0 (r)2 dt2 + + r 2 dΩ2 (5.39)
2m0 (r)
1− r
and then assume it represents a perfect fluid sphere. Our original four theorems can
be written in the form
Theorem 1. Suppose {ζ0 (r), m0 (r)} represents a perfect fluid sphere. Define
2 Z ′
ζ0 (r) ζ0 (r) − r ζ0′ (r)

ζ0 (r) 3
m1 (r) = r exp 2 dr . (5.40)
ζ0 (r) + r ζ0′ (r) ζ0 (r) ζ0 (r) + r ζ0′ (r)
Then for all λ, the geometry defined by holding ζ0 (r) fixed and setting
dr 2
ds2 = −ζ0 (r)2 dt2 + + r 2 dΩ2 (5.41)
2[m0 (r) + λ m1 (r)]
1− r
is also a perfect fluid sphere.
Theorem 2. Let {ζ0 (r), m0 (r)} describe a perfect fluid sphere. For all σ and ǫ, the
geometry defined by holding m0 (r) fixed and setting
 Z 2
 r dr  dr 2
ds2 = −ζ0 (r)2 σ + ǫ dt2 + + r 2 dΩ2 (5.42)
2m (r)
q
2m 0 (r) 1− 0
ζ0 (r)2 1 −
 
r r
is also a perfect fluid sphere.
131
Theorem 3. Let {ζ0 (r), m0 (r)} describe a perfect fluid sphere. For for all σ, ǫ, and
λ, the three-parameters geometry defined by
 Z 2
 r dr 
ds2 = −ζ0 (r)2 σ+ǫ q dt2
2[m0 (r) + λ m1 (r)] 
ζ0 (r)2

1− r
dr 2
+ + r 2 dΩ2 (5.43)
2[m0 (r) + λ m1 (r)]
1− r
is also a perfect fluid sphere, where m1 (r) is defined as above.
Theorem 4. Let {ζ0 (r), m0 (r)} describe a perfect fluid sphere. For all σ, ǫ, and λ,
the three-parameters geometry defined by
 Z 2
2 2
 r dr  dr 2
ds = −ζ0 (r) σ + ǫ dt2 + +r 2 dΩ2
2[m (r) + λ m̃ (r)]
q
2m0 (r) 1− 0 1
ζ0 (r)2 1 −
 
r r
(5.44)
is also a perfect fluid sphere, where m̃1 (r) is defined as
2 Z ′
ζ (r) ζ(r) − r ζ ′ (r)

ζ(r) 3
m̃1 (r) = r exp 2 dr , (5.45)
ζ(r) + r ζ ′(r) ζ(r) ζ(r) + r ζ ′(r)
depending on ζ = ζ0 ζ1 , where
 Z 
 r dr 
ζ1 (r) = σ + ǫ q . (5.46)
 2 2m0 (r) 
ζ0 (r) 1− r
5.4.2 Four theorems in terms of pressure and density

We can re-write these theorems in terms of pressure and density profiles p(r) and ρ(r).
Consider Theorem 1. By definition ζ is unaffected under the transformation. But
we have a formula for ζ(r) in terms of p(r) and m(r). Let m2 (r) = m0 (r) + λm1 (r)
and p2 (r) be the mass profile and pressure profile after applying theorem 1. Then
[m0 (r) + 4πp0 (r) r 3 ] [m2 (r) + 4πp2 (r) r 3 ]
= (5.47)
r 2 [1 − 2m0 (r)/r] r 2 [1 − 2m2 (r)/r]
132
which we can rearrange to yield
[1 − 2m2 (r)/r]
m2 (r) + 4πp2 (r) r 3 = [m0 (r) + 4πp0 (r) r 3 ] (5.48)
[1 − 2m0 (r)/r]
Furthermore, we can rearrange it into
[1 − 2m2 (r)/r]
4πp2 (r) r 3 = −m2 (r) + [m0 (r) + 4πp0 (r) r 3 ] (5.49)
[1 − 2m0 (r)/r]
Now we substitute m2 (r) = m0 (r) + λm1 (r) into equation (5.49),
[1 − 2(m0 (r) + λm1 (r))/r]
4πp2 (r) r 3 = −m0 (r) − λm1 (r) + [m0 (r) + 4πp0 (r) r 3 ]
[1 − 2m0 (r)/r]
(5.50)
After a bit of algebra:
−λm1 (r) + 4πp0 (r) − 8πp0 (r) r 2 m0 (r) − 8λπp0 (r) r 2 m1 (r)
4πp2 (r) r 3 =
[1 − 2m0 (r)/r]
−λm1 (r)(1 − 8πp0 (r) r 2 ) 4πp0 (r) r 3 [1 − 2m0 (r)/r]
= +
[1 − 2m0 (r)/r] [1 − 2m0 (r)/r]
λm1 (r)(1 + 8πp0 (r) r 2 )
= 4 πp0(r) r 3 − (5.51)
[1 − 2m0 (r)/r]
rearrange it in terms of p2 (r),
1 + 8πp0 (r) r 2

1
p2 (r) = p0 (r) − λm1 (r) (5.52)
4πr 3 1 − 2m0 (r)/r
This allows us to rewrite Theorem 1 as follows:
Theorem 13 (Theorem 1B:). Suppose we are given a pair of functions p0 (r) and
ρ0 (r) that satisfy the TOV.
We use these functions to construct m0 (r), ζ0 (r), and m1 (r) as follows:
Z r
m0 (r) = 4π ρ0 (r̄) r̄ 2 dr̄ (5.53)
0
[m0 (r) + 4πp0 (r) r 3 ]

Z
ζ0 (r) = exp dr (5.54)
r 2 [1 − 2m0 (r)/r]
133
2 Z ′
ζ0 (r) ζ0 (r) − r ζ0′ (r)

ζ0 (r) 3
m1 (r) = r exp 2 dr (5.55)
ζ0 (r) + r ζ0′ (r) ζ0 (r) ζ0 (r) + r ζ0′ (r)
Then for all λ, the pair of functions
1 + 8πp0 (r)r 2

1
p2 (r) = p0 (r) − λm1 (r) (5.56)
4πr 3 1 − 2m0 (r)/r
λ dm1 (r)
ρ2 = ρ0 (r) + (5.57)
4πr 2 dr
also satisfies the TOV and so defines a perfect fluid sphere.
Similar logic can be applied to theorem 2. Note that in that case you are considering
the transformation
m0 (r) → m2 (r) = m0 (r) (5.58)
 Z 
 r dr 
ζ0 (r) → ζ2 (r) = ζ0 (r) σ + ǫ q (5.59)
2m0 (r) 
ζ0 (r)2 1 −

r
So that m0 (r) remains fixed. Now from the Gr̂r̂ component of the Einstein tensor:
consider the equation below
ζ′ m(r) + 4πp(r) r 3
= 2 (5.60)
ζ r (1 − 2m(r)/r)
and rearrange this equation in term of p(r)
ζ′ 2
r (1 − 2m(r)/r) = m(r) + 4πp(r) r 3
ζ
3 2 ζ′
4πp(r) r = −m(r) + r (1 − 2m(r)/r)
ζ
m(r) 1 ζ′
p(r) = − + (1 − 2m(r)/r) (5.61)
4πr 3 4πr ζ
when
m(r) m(r)/3 1
3
= 3
= ρ̄(r) (5.62)
4πr (4π/3) r 3
134
Now we have p(r) equal to
1 1 ζ ′ (r)
p(r) = − ρ̄(r) + (1 − 2m(r)/r) (5.63)
3 4πr ζ(r)
Then in particular
1 1 ζ ′ (r)
p2 (r) = − ρ̄0 (r) + (1 − 2m0 (r)/r) 2 (5.64)
3 4πr ζ2 (r)
so that
  Z ′
 
ζ0 (r) σ + ǫ qr dr 
2 2m0 (r) 
ζ0 (r) 1 −

1 1 2m0 (r) r
p2 (r) = − ρ̄0 (r) + 1− r  Z 
3 4πr  
ζ0 (r) σ + ǫ qr dr
2m0 (r) 
ζ0 (r)2 1 −

r
(5.65)
That is
2m0 (r) ζ0′ (r)

1 1
p2 (r) = − ρ̄0 (r) + 1−
3 4πr r ζ0 (r)
 
 
ǫ q r
2m0 (r) 
ζ0 (r)2 1 −

1 2m0 (r)  r
+ 1− r Z  (5.66)
4πr  
σ+ǫ qr dr
2m0 (r) 
ζ0 (r)2 1 −

r
′
1 (1 − 2m(r)/r) ζ0 (r) , so that
we know p(r) = − 31 ρ̄(r) + 4πr
ζ0 (r)
r  Z −1
ǫ 2m0 (r)
 r dr 
p2 (r) = p0 (r) + 1− r ζ0 (r)−2 σ+ǫ q (5.67)
4π 2m0 (r) 
ζ0 (r)2 1 −

r
This leads us to:
135
Use these functions to construct m0 (r) and ζ0 (r) as follows:
Z r
m0 (r) = 4π ρ0 (r̄) r̄ 2 dr̄ (5.68)
0
[m0 (r) + 4πp0 (r) r 3 ]

Z
ζ0 (r) = exp dr (5.69)
r 2 [1 − 2m0 (r)/r]
Then for all σ and ǫ, the pair of functions
ρ2 (r) = ρ0 (r) (5.70)
r  Z −1
ǫ 2m0 (r)  r dr 
p2 (r) = p0 (r) + 1− ζ0 (r)−2 σ+ǫ q (5.71)
4π r 2m0 (r) 
ζ0 (r)2 1 −

r
also satisfies the TOV and so defines a perfect fluid sphere.
Having now found these theorems 1B and 2B, is now useful to define a new theorem
by composing them. All together we can consider this as a single process by introducing
the following theorem:
Use these functions to construct m0 (r), ζ0 (r), and m1 (r) as follows:
Z r
m0 (r) = 4π ρ0 (r̄) r̄ 2 dr̄ (5.72)
0
[m0 (r) + 4πp0 (r) r 3 ]

Z
ζ0 (r) = exp dr (5.73)
r 2 [1 − 2m0 (r)/r]
2 Z ′
ζ0 (r) ζ0 (r) − r ζ0′ (r)

ζ0 (r) 3
m1 (r) = r exp 2 dr (5.74)
ζ0 (r) + r ζ0′ (r) ζ0 (r) ζ0 (r) + r ζ0′ (r)
136
Then for all σ, ǫ, and λ, the pair of functions
1 + 8πp0 (r) r 2

1
p2 (r) = p0 (r) − λm1 (r)
4πr 3 1 − 2m0 (r)/r
r  Z −1
ǫ 2m2 (r)  r dr 
+ 1− ζ0 (r)−2 σ + ǫ q (5.75)
4π r  2 2m2 (r) 
ζ0 (r) 1 − r
λ dm1 (r)
ρ2 = ρ0 (r) + (5.76)
4πr 2 dr
where m2 (r) = m0 (r) + λm1 (r), also satisfy the TOV and so define a perfect fluid
sphere.
Use these functions to construct m0 (r), ζ0 (r), and m1 (r) as follows:
Z r
m0 (r) = 4π ρ0 (r̄) r̄ 2 dr̄ (5.77)
0
[m0 (r) + 4πp0 (r) r 3 ]

Z
ζ0 (r) = exp dr (5.78)
r 2 [1 − 2m0 (r)/r]
2 Z ′
ζ (r) ζ(r) − r ζ ′ (r)

ζ(r) 3
m̃1 (r) = r exp 2 dr , (5.79)
ζ(r) + r ζ ′(r) ζ(r) ζ(r) + r ζ ′ (r)
depending on ζ = ζ0 ζ1 , where
 Z 
 r dr 
ζ1 (r) = σ + ǫ q . (5.80)
 2 2m0 (r) 
ζ0 (r) 1− r
Then the pair of functions
1 + 8πp0 (r)r 2

1
p2 (r) = p0 (r) − λm1 (r) (5.81)
4πr 3 1 − 2m0 (r)/r
λ dm̃1 (r)
ρ2 = ρ0 (r) + (5.82)
4πr 2 dr
also satisfy the TOV and so define a perfect fluid sphere.
137
Some seed geometries and their descendants in term of pressure and density via Theorem 1B
Seed p0 (r) pTh1 (r) ρ0 (r) ρTh1 (r)

λ 3λ
Minkowski 0 − 4π 0 4π
3500 λ (r−a)2
exterior Schwarzschild 0 π r2 0 − 3500 λ (r−a)(3r−5a)
πr 2
λ(R2 −3r 2 ) 2 2
r −3R4 −6r 4 )
de Sitter 3
− 8πR 2 p0 (r) − 4π(−R 2 +2r 2 )
3
8πR2 ρ0 (r) + 4λπ (7R(−R 2 +2r 2 )2
! !
(2n+n2 ) (2n+n2 )
− −
2 (n+2) (n+2)
−1+n2
Tolman V (A = 0) − 8r2(n+1)
π(n2 −2) p0 (r) − λr
4 π(n+2)2
(2n+3)
8r 2 π(−2+n2 ) ρ0 (r) + λr (3+2n)(2−n)
4 π(n+2)3
1 5λ 5λ
S1 2πr 2 p0 (r) − 36 πr (4/3)
0 108 πr (4/3)
( 1 ) (r−a)(2a+3r) 3 (a2 +4 r 2 ) 1 2
r ( 3 ) (12r −21ar+10a )
2
M–W III − (a−6r) (a+2r)
56 π r 2 a2 p0 (r) + λ 5r 3 4 56 πr 2 a2 ρ0 (r) + 9λπ 5 7
9 (a+r)(−4r+3a)( 3 ) (−4r+3a)( 3 )
2
9a(−1+ar ) λ(7ar 2 +1) 3a (3+ar 2 ) λ(1+3ar 2 )
Heint IIa (C = 0) − 16π(1+ar 2 )2 p0 (r) − √ √
12 3 π(1+ar 2 )2 −4ar 2 −1 16π(1+ar 2 )2 ρ0 (r) − √ 3
4 3π(1+ar 2 )2 (−4ar 2 −1)( 2 )
Table 5.1: Seed solutions and their generalizations derived via theorems 1B in terms of
pressure and density profile.
Seed pTh2 (r) ρTh2 (r)

Minkowski ǫ 2 0
4 π σ+ ǫr2
1
exterior Schwarzschild 2σ
√
r(r−2a)
√ √ 0
2 π(r 2 +5ra−30 a2 + ǫ
+15 r(r−2a) a2 ln(r−a+ r(r−2a)))
de Sitter p0 (r) + ǫ ρ0 (r)
√

σ 2
4π R2 −r 2 R
+ √ ǫ R2 2
−R +r
ǫ
2−n2
Tolman V (A = 0) p0 (r) + q
1
ρ0 (r)
2 πr 2 2B 2 nr (1+n) −ǫ
2−n2
ǫ
S1 p0 (r) + 4 πr 4 (σ−
0
ǫ
2r 2
)
ǫ
M–W III p0 (r) + √
2 7 πAraσ(r−a)
ρ0 (r)
√ +7 πra ǫ
a2 −r 2
ǫ
Heint IIa (C=0) p0 (r) + (1+ar 2 )(7/2)

2ǫ (5+2ar 2 )
ρ0 (r)
8π √ σ+ √
4−2ar 2 27a (1+ar 2 )(3/2) 4−2ar 2
pressure and density profile.
138

ǫ
Minkowski p0 (r) + √ 4 πσ
− 2 λπǫ
ρTh1 (r)
1−2λr2
exterior Schwarzschild [integral] ρTh1 (r)

λ(R2 −3 r 2 )
de Sitter p0 (r) − 4 π(−R2 +2 rr 2) ρTh1 (r)
r2 2 (−R2 +r2 )
ǫ 1− R 2 + 2λr
(−R2 +2r2 )
+  Z 
 
rR2
 σ 
4 π(R2 −r 2 )2 
 R2 +ǫ s dr 
2 2λr2 (−R2 +r2 )

(R2 −r2 ) 1− r 2 +
(−R2 +2r2 )
 
R
(1+n)(−2n)
λr (2+n) (3+2n)
Tolman V (A = 0) p0 (r) − 4 π(2+n) 2 ρTh1 (r)
s
2(2−2n2 )
1 2λr (2+n)
ǫ (2−n2 )
+ (2+n)2
−  Z 
 
r(−1−2n)
4 πB 2 r (1+n)2 
 
 σ+ǫ v
u
u 2(2−2n2 )
dr 

 
2λr (2+n)
u
B 2t 1 +
(2−n2 ) (2+n)2
λ
S1 p0 (r) + 12 r (4/3) π
+ ρTh1 (r)
ǫ
6σ
2πr 4 √ +ǫ − r12 + 5λ
3 − 1
+λ
9 − 1
− √ λ
arctanh √ 3
9−2 λr(2/3) 6r(4/3) 2r(2/3) 3 9−2 λr (2/3) 9−2 λr(2/3)
λ r (1/3) (2a−3r)

M–W III p0 (r) − 9 5 (4/3) ρTh1 (r)
r (−4r+3a)
(a2 −r2 ) 2 λr(7/3) (r−a)5(2/3)
ǫ 7a2
+ (4/3)
45 (−4r+3a)
+  Z 
 
1
 
 σ+ǫ
2πAr(r−a) s dr 
(a2 −r2 ) 2 λr(7/3) (r−a)5(2/3)

A(r−a) +
7a2
 
45 (−4r+3a)(4/3)
λ(7ar 2 +1)
Heint IIa (C=0) p0 (r) − √
−4ar 2 −1
12 π s
ρTh1 (r)

1 4 λr2
ǫ (1+ar 2) (2−ar 2 )− √
3 −12ar2 −3
−  Z 
 
r
 
4 π(1+ar 2 )3 
 σ+ǫ v
u ! dr 

(1+ar2 )t (2−ar2 )− 4 λr2
 u 
√
3 −12ar2 −3
pressure and density profile. The notation “[integral]” denotes a metric so complicated that
explicitly writing out the relevant integral is so tedious that it does not seem worthwhile.
139

λ(5 ǫ r 2 +6 σ)
Minkowski pTh1 (r) 4 π(2 σ+3 ǫ r 2 )(5/3)
exterior Schwarzschild pTh1 (r) [integral]
de Sitter pTh1 (r) [integral]
Tolman V (A = 0) pTh1 (r) [integral]
λ(10 σr 2 −3 ǫ)
S1 pTh1 (r) 4 π(6 σr 2 −ǫ)(5/3)
M–W III pTh1 (r) [integral]
Heint IIa (C=0) pTh1 (r) [integral]
pressure and density profile. The notation “[integral]” denotes a metric so complicated that
explicitly writing out the relevant integral is so tedious that it does not seem worthwhile.
5.4.3 Discussion
In this chapter we have rephrased the theorems 1, 2, 3, and 4 of chapter 3 directly
in terms of the TOV equation, and consequently in terms of the pressure profile and
density profile. We have done this by rephrasing theorem 1 in a slightly different
formalism, and developing an independent proof that is easy to rewrite in terms of
the pressure and density profiles — ρ(r) and p(r) — resulting in the theorem we call
theorem 1B. Similar logic, involving the transformation T2 , can be applied to theorem
2 — now leading to theorem 2B. With these new theorems 1B and 2B in hand, we
can compose them in various ways to obtain theorems 3B and 4B.
We present several tables that exhibit seed solutions and the geometries generated
from them by combining all the various theorems. Table (5.1) shows some seed ge-
ometries and their descendants as obtained via theorem 1B, and similarly table (5.2)
140
shows some seed geometries and their descendants obtained via theorem 2B. Finally
table (5.3) and (5.4) deal with theorems 3B and 4B. We emphasise that theorem 2B
does not affect the density profile, and only the pressure profile changes.
141
142
Chapter 6
Conclusions
As the range of problems tackled is so diverse, it is hard to think of a concise way of

summarising all of them. Indeed, to a large extent much of the work has already has
summarised them at the end of each section. Instead of explaining the problems again
and how far through completion they are, I would like to leave with list of things that
I believe are useful to understand the overall concept of the thesis.
This thesis has been written with the goal of making it accessible to people with
a basic background in general relativity, especially in coordinate conditions, perfect
fluid spheres in general relativity.
6.1 The main analysis: Structure of the thesis

• The thesis summarized the basic concept of special relativity to make general
relativity easier to understand. General relativity is the relativistic theory of
gravitation. It is required from any new physical theory to satisfy the so called
principle of correspondence.
• Coordinate conditions are important for the long time scale simulations of rel-
ativistic systems. We demonstrated the importance of coordinate conditions in
general relativity and proposed a method of looking for nice solutions of the
Einstein equations. In addition, a number of examples are given to illustrate the
143
method.
• We introduced “algorithmic” techniques that permit one to generate large classess

of perfect fluid spheres in a purely mechanical way. Furthermore, we extended
these algorithmic ideas, by proving several solution-generating theorems of vary-
ing levels of complexity. We have developed several transformation theorems
that map perfect fluid spheres to perfect fluid spheres using Schwarzschild co-
ordinates. Furthermore, we used these transformations as a basis for classifying
different types of perfect fluid sphere solutions.
• In chapter 4, we have developed several transformation theorems that map per-

fect fluid spheres to perfect fluid sphere using other coordinates.
• We rephased Theorems 1, 2, 3 and 4 in chapter 3 directly in terms of the TOV

equation.
6.2 Further interesting issues

I would like to leave with the list of things that I believe are interesting to continue to
analyze. In addition, all the lists below are further issues that might be interesting to
looking at.
• For a Riemannian geometry in (2+1) dimensions, we can calculate the Ricci
Scalar, the Ricci tensor and the Riemann tensor. However, all results are too
difficult to analyze in detail. This is one of the physically interesting cases, and
this is a case that I would like to examine in more detail.
• Similar for a Riemannian geometry in (3+1) dimensions. This is also one of

the physical interesting case, and this is the case that will be most difficult to
analyze in detail. This is a challenging case that I would like to work through in
more detail.
• We analysed a generic spacetime using both ADM and reverse-ADM (Kaluza–

Klein) decompositions. However, we found that for the general ADM case co-
ordinate techniques are too complicated to contemplate (page 514 of [17]). The
144
inverse ADM decomposition can calculated the Christoffel symbols from equa-
tion (2.178). This is so interesting to find the techniques for contemplation.
• The other things that would be interesting is consider other coordinate systems:
Generalize the theorems which we derived for Schwarzschild coordinates or co-
ordinates in Chapter 4 to a number of other coordinate systems. This may allow
us to develop extra relations between the known solutions.
• For Theorem 8 in Isotropic coordinate, it would be interesting to see what would

happen when we feed this theorem a Minkowski space seed. In addition, this is
interesting to find simple ways of generating interesting solutions.
• For Theorem 9 in Gaussian polar coordinate, Theorem 10 in Buchdahl coordi-

nates, Theorem 11 in Synge isothermal coordinates, and Theorem 12 in Expo-
nential coordinates, it would be interesting to see what would happen when we
feed this theorem a Minkowski space seed. In addition, this is intersting to find
the way that we can do to R(r) to make its ODE look nicer.
All the above suggestions would be interesting and feasible, although some of them
would be more tedious to extend to standard results in general relativity. However,
this thesis would provide a platform for better understanding coordinate conditions
and the perfect fluid sphere in general relativity. This thesis developed a way of looking
for nice new solutions of the Einstein equations. Furthermore, the goal of this thesis
was to classify the different types of perfect fluid sphere solutions. This thesis has
hopefully served as a brief introduction the “Algorithms” ideas by showing how to
prove several solution-generating theorems of varying levels of complexity. Finally, we
illustrated the formalism to develop extra relations between the known solutions for
Schwarzschild coordinates and other coordinates. The generating theorems in perfect
fluid spheres lead to additional useful “exact solutions”.
145
146
Bibliography
[1] R. Adler, M. Bazin, and M. Schiffer. Introduction to general relativity. McGraw-

Hill Book company, 1965.
[2] Various authors. Special relativity. 2005. [http://en.wikipedia.org/wiki/Special

relativity].
[3] S. Berger, R. Hojman, and J. Santamarina. General exact solutions of Ein-

stein equations for static perfect fluids with spherical symmetry. J. Math. Phys.,
28:2649, 1987.
[4] H. Bondi. Spherically symmetrical models in general relativity. Mon. Not. Roy.
Astron. Soc., 107:410, 1947.
[5] H. Bondi. Massive spheres in general relativity. Mon. Not. Roy. Astron. Soc.,
282:303–317, 1964.
[6] P. Boonserm, M. Visser, and S. Weinfurtner. Generating perfect fluid spheres in

general relativity. Phys. Rev., D71(124037), 2005. [arXiv:gr-qc/0503007].
[7] H. A. Buchdahl. General relativistic fluid spheres. Phys. Rev., 116:1027–1034,

1959.
[8] H. A. Buchdahl. General relativistic fluid spheres II: General inequalities for
regular spheres. Ap. J., 146:275–281, 1966.
147
[9] M. S. R. Delgaty and K. Lake. Physical acceptability of isolated, static, spher-
ically symmetric, perfect fluid solutions of einstein’s equations. Comput. Phys.
Commun., 115:395, 1998. [arXiv:gr-qc/9809013].
[10] P. A. M. Dirac. General theory of relativity. Princeton University Press, 2 edition,

1975.
[11] M. R. Finch and J. E. F. Skea. A review of the relativistic static fluid sphere. 1998,
unpublished. [http://www.dft.if.uerj.br/usuarios/JimSkea/papers/pfrev.ps].
[12] V. Fock. The Theory of Space, Time, and Gravitation. Pergamon, New York,
1964.
[13] K. Lake. All static spherically symmetric perfect fluid solutions of Einstein’s
equations. Phys. Rev., D67:104015, 2003. [arXiv:gr-qc/0209104].
[14] L. D. Landau and E. M. Lifshitz. The classical theory of fields, volume 2.

Butterworth-Heinemann, Oxford, 4 edition, 1975.
[15] D. Martin and M. Visser. Bounds on the interior geometry and pressure profile of
static fluid spheres. Class. Quant. Grav., 20:3699, 2003. [arXiv:gr-qc/0306038].
[16] D. Martin and M. Visser. Algorithmic construction of static perfect fluid spheres.
Phys. Rev., D69:104028, 2004. [arXiv:gr-qc/0306109].
[17] C. W. Misner, K. S. Thorne, and J. A. Wheeler. Gravitation. (W. H. Freeman

and Company, San Francisco), 1973.
[18] S. Rahman and M. Visser. Spacetime geometry of static fluid spheres. Class.
Quant. Grav., 19:935, 2002. [arXiv:gr-qc/0103065].
[19] R.D’Inverno. Introducing Einstein’s Relativity. (Clarendon Press, Oxford), 2002.
[20] H. Stephani, D. Kramer, M. MacCallum, C. Hoenselaers, and E. Herlt. Exact

solutions of Einstein’s field equations. (Cambridge University Press), 2003.
[21] J. L. Synge. Relativity: the General Theory. North-Holland, Amsterdam, 1964.
148
[22] M. Visser. Math 465: Notes on general relativity. Victoria University, 2005.
[http://www.mcs.vuw.ac.nz/courses/MATH465/2005T2/Lecture-Notes/].
[23] M. Visser, C. Barcelo, and S. Liberati. Analogue models of and for gravity. Gen.
Rel. Grav., 34:20, 2002. [arXiv:gr-qc/0111111].
[24] R. M. Wald. General Relativity. The University of Chicago Press, 1984.
[25] M. Wyman. Radially symmetric distributions of matter. Phys. Rev., D75:1930,

1949.
149

Some Exact Solutions in General Relativity: Petarpa Boonserm

Uploaded by

Copyright:

Available Formats

Some Exact Solutions in General Relativity: Petarpa Boonserm

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Some Exact Solutions in General Relativity: Petarpa Boonserm

Uploaded by

Copyright:

Available Formats

Some Exact Solutions

A thesis submitted to the

Supervisor: Dr. Matt Visser

This thesis looks at four problems in general relativity:

• Coordinate conditions in general relativity: A method of looking for nice solu-

• Rephrase all these theorems which apply in Schwarzschild coordinates, directly

General relativity (GR) or general relativity theory (GRT) is a theory of gravitation

Structure of the thesis

2 Coordinate conditions in general relativity 13

3 Perfect fluid spheres in general relativity 59

4 Solution generating theorems — other coordinates 103

5 TOV equation - related theorems 123

3.1 A Pulsating spherical star cannot emit gravitational waves. . . . . . . . 65

1.2 Einstein’s equation

1.2.1 General and Special Covariance

S 2 = (x1 − x̄1 )2 + (x2 − x̄2 )2 + (x3 − x̄3 )2 (1.1)

This formula is the distance between two points.

(δS)2 = (δx1 )2 + (δx2 )2 + (δx3 )2 (1.2)

Therefore, the metric of space is given by

ds2 = (dx1 )2 + (dx2 )2 + (dx3 )2 (1.3)

In the Cartesian coordinate basis, we derive

1.3 Special Relativity

The curvature of ηab vanishes. In addition, we can parameterize timelike curves by

1.3.2 The Stress-Energy Tensor

ds2 = −dt2 + dx2 + dy 2 + dz 2 (1.16)

We can write this in tensorial form as

ds2 = ηab dxa dxb , (1.17)

We take ηab to denote the Minkowski metric

In relativistic units the equation for the proper time satisfies

dτ 2 = −ds2 = −(−c2 dt2 + dx2 + dy 2 + dz 2 )

The space-components are the components of the three-dimensional momentum

A perfect fluid is defined to be a continuous distribution of matter with stress

Tab = ρua ub + p(ηab + ua ub ), (1.24)

∂ a Tab = ∂ a [ρua ub + p(ηab + ua ub )]

[(ρ + p)ua ∂a ub + ∂ a p(ηab + ua ub )] = 0. (1.28)

and equation (1.28) becomes

1.3.3 Relativistic hydrodynamics

(i) gab is non-singular with signature − + ++;

(ii) free particles travel along timelike geodesics;

(iii) light rays travel along null geodesics.

Coordinate conditions in general

X a → X ′a = X ′a (x) (a = 0, 1, 2, 3). (2.3)

g00 → +1, g0i → 0 (i = 1, 2, 3). (2.4)

ds2 = R2 (dθ2 + sin2 dφ2 ) (2.5)

gik = diag −c2 e2φ , e2Λ , R2 , R2 sin2 θ

where Λ = Λ(r, t), φ = φ(r, t) and R = R(r, t). If ∂R 6= 0, it is possible to rescale R

(ds)2 = ds2 = gab dxa dxb (2.7)

2.2 Independent components: Riemann’s argument

true “independent” functions in the metric. This argument is due to Riemann.

gij = exp[2φ(x)] δij

Rabcd = {gac Rbd + gbd Rac − gad Rbc − gbc Rad }

2.3 Lorentzian manifolds

ds2 = −(dx0 )2 + (dx1 )2 + (dx2 )2 + (dx3 )2

gij = exp[2φ(t, x)] ηij

In Lorentzian geometry, it is possible to make another coordinate choice by setting

where v = v(t,x), so " #

where hi = hi (t, x, y), so

Alternatively we can always locally set up patches of the form

2.4 The block diagonal ansatz