This page intentionally left blank

GRAVITATIONAL COLLAPSE AND SPACETIME

SINGULARITIES

Physical phenomena in astrophysics and cosmology involve gravitational collapse

in a fundamental way. The final fate of a massive star when it collapses under
its own gravity at the end of its life cycle is one of the most important ques-
tions in gravitation theory and relativistic astrophysics, and is the foundation of
blackhole physics.
General relativity predicts that continual gravitational collapse gives rise to
a spacetime singularity, which may be hidden inside an event horizon or visi-
ble to external observers. This book investigates these issues, and shows how
such visible ultra-dense regions arise naturally and generically as an outcome
of dynamical gravitational collapse. Quantum gravity may take over in these
regimes to resolve the classical spacetime singularity. The quantum effects from
a visible extreme gravity region could then propagate to external observers,
providing a useful laboratory for quantum gravity, and implying interesting
consequences for ultra-high energy astrophysical phenomena in the universe.
This volume will be of interest to graduate students and academic researchers
in gravitation physics and fundamental physics, as well as in astrophysics and
cosmology. It includes a review of recent research into gravitational collapse, and
several examples of collapse models are worked out in detail.

Pan k a j S. Jos hi conducts research at the Tata Institute of Fundamental

Research, Mumbai. His research interests include gravitation physics, spacetime
structure and quantum gravity, and cosmology and relativistic astrophysics. He
has published many research papers and books in these areas, and has held vis-
iting faculty positions in several countries, lecturing and doing research on these
topics.
Professor Joshi has an excellent international reputation for his work in the
field of gravitation theory. His extensive analysis of general relativistic gravita-
tional collapse has been widely recognized as providing significant insights into
the final end states of a continual collapse, formation of visible singularities, and
nature of cosmic censorship and blackholes.
 C A M B R I D G E M O N O G R A P H S O N M AT H E M AT I C A L P H Y S I C S
General editors: P. V. Landshoff, D. R. Nelson, S. Weinberg

S. J. Aarseth Gravitational N-Body Simulations

J. Ambjørn, B. Durhuus and T. Jonsson Quantum Geometry: A Statistical Field Theory Approach
A. M. Anile Relativistic Fluids and Magneto-Fluids
J. A. de Azcárrage and J. M. Izquierdo Lie Groups, Lie Algebras, Cohomology and Some Applications in
Physics†
O. Babelon, D. Bernard and M. Talon Introduction to Classical Integrable Systems
F. Bastianelli and P. van Nieuwenhuizen Path Integrals and Anomalies in Curved Space
V. Belinkski and E. Verdaguer Gravitational Solitons
J. Bernstein Kinetic Theory in the Expanding Universe
G. F. Bertsch and R. A. Broglia Oscillations in Finite Quantum Systems
N. D. Birrell and P. C. W. Davies Quantum Fields in Curved Space†
M. Burgess Classical Covariant Fields
S. Carlip Quantum Gravity in 2 + 1 Dimensions
J. C. Collins Renormalization†
M. Creutz Quarks, Gluons and Lattices†
P. D. D’Eath Supersymmetric Quantum Cosmology
F. de Felice and C. J. S. Clarke Relativity on Curved Manifolds†
B. S. DeWitt Supermanifolds, 2nd edition†
P. G. O. Freund Introduction to Supersymmetry†
J. Fuchs Affine Lie Algebras and Quantum Groups†
J. Fuchs and C. Schweigert Symmetries, Lie Algebras and Representations: A Graduate Course for Physicists†
Y. Fujii and K. Maeda The Scalar Tensor Theory of Gravitation
A. S. Galperin, E. A. Ivanov, V. I. Orievetsky and E. S. Sokatchev Harmonic Superspace
R. Gambini and J. Pullin Loops, Knots, Gauge Theories and Quantum Gravity†
M. Göckeler and T. Schücker Differential Geometry. Gauge Theories and Gravity†
C. Gómez, M. Ruiz Altaba and G. Sierra Quantum Groups in Two-dimensional Physics
M. B. Green, J. H. Schwarz and E. Witten Superstring Theory, volume 1: Introduction†
M. B. Green, J. H. Schwarz and E. Witten Superstring Theory, volume 2: Loop Amplitudes, Anomalies and
Phenomenology†
V. N. Gribov The Theory of Complex Angular Momenta
S. W. Hawking and G. F. R. Ellis The Large-Scale Structure of Space-Time†
F. Iachello and A. Arima The Interacting Boson Model
F. Iachello and P. van Isacker The Interacting Boson–Fermion Model
C. Itzykson and J.-M. Drouffe Statistical Field Theory, volume 1: From Brownian Motion to Renormalization
and Lattice Gauge Theory†
C. Itzykson and J.-M. Drouffe Statistical Field Theory, volume 2: Strong Coupling, Monte Carlo Methods,
Conformal Field Theory, and Random Systems†
C. Johnson D-Branes
P. S. Joshi Gravitational Collapse and Spacetime Singularities
J. I. Kapusta Finite-Temperature Field Theory†
V. E. Korepin, A. G. Izergin and N. M. Boguliubov The Quantum Inverse Scattering Method and Correlation
Functions†
M. Le Bellac Thermal Field Theory†
Y. Makeenko Methods of Contemporary Gauge Theory
N. Manton and P. Sutcliffe Topological Solitons
N. H. March Liquid Metals: Concepts and Theory
I. M. Montvay and G. Münster Quantum Fields on a Lattice†
L. O’ Raifeartaigh Group Structure of Gauge Theories†
T. Ortı́n Gravity and Strings
A. Ozorio de Almeida Hamiltonian Systems: Chaos and Quantization†
R. Penrose and W. Rindler Spinors and Space-Time, volume 1: Two-Spinor Calculus and Relativistic Fields†
R. Penrose and W. Rindler Spinors and Space-Time, volume 2: Spinor and Twistor Methods in Space-Time
Geometry†
S. Pokorski Gauge Field Theories, 2nd edition
J. Polchinski String Theory, volume 1: An Introduction to the Bosonic String†
J. Polchinski String Theory, volume 2: Superstring Theory and Beyond†
V. N. Popov Functional Integrals and Collective Excitations†
R. J. Rivers Path Integral Methods in Quantum Field Theory†
R. G. Roberts The Structure of the Proton†
C. Rovelli Quantum Gravity
W. C. Saslaw Gravitational Physics of Stellar and Galactic Systems†
H. Stephani, D. Kramer, M. A. H. MacCallum, C. Hoenselaers and E. Herlt Exact Solutions of Einstein’s Field
Equations, 2nd edition
J. M. Stewart Advanced General Relativity†
A. Vilenkin and E. P. S. Shellard Cosmic Strings and Other Topological Defects†
R. S. Ward and R. O. Wells Jr Twistor Geometry and Field Theories†
J. R. Wilson and G. J. Mathews Relativistic Numerical Hydrodynamics

†
Issued as a paperback
Gravitational Collapse and
Spacetime Singularities
PA N K A J S . J O S H I
Tata Institute of Fundamental Research,
Mumbai, India
CAMBRIDGE UNIVERSITY PRESS
Cambridge, New York, Melbourne, Madrid, Cape Town, Singapore, São Paulo

Cambridge University Press

The Edinburgh Building, Cambridge CB2 8RU, UK
Published in the United States of America by Cambridge University Press, New York
www.cambridge.org
Information on this title: www.cambridge.org/9780521871044

© P. S. Joshi 2007

This publication is in copyright. Subject to statutory exception and to the provision of

relevant collective licensing agreements, no reproduction of any part may take place
without the written permission of Cambridge University Press.
First published in print format 2007

ISBN-13 978-0-511-37712-9 eBook (EBL)

ISBN-13 978-0-521-87104-4 hardback

Cambridge University Press has no responsibility for the persistence or accuracy of urls
for external or third-party internet websites referred to in this publication, and does not
guarantee that any content on such websites is, or will remain, accurate or appropriate.
 To my parents,
Arunadevi Shantilal Joshi
and
Shantilal Ramshankar Joshi
 Contents

Preface page ix

1 Introduction 1

2 The spacetime manifold 10

2.1 The manifold model 10
2.2 The metric tensor 21
2.3 Connection 24
2.4 Non-spacelike geodesics 29
2.5 Spacetime curvature 32
2.6 The Einstein equations 38
2.7 Exact solutions 43

3 Spherical collapse 60
3.1 Basic framework 62
3.2 Regularity conditions 69
3.3 Collapsing matter clouds 71
3.4 Nature of singularities 79
3.5 Exterior geometry 87
3.6 Dust collapse 90
3.7 Equation of state 129

4 Cosmic censorship 135

4.1 Causal structure 136
4.2 Spacetime singularities 149
4.3 Blackholes 161
4.4 Higher spacetime dimensions 169
4.5 Formulating the censorship 175
4.6 Genericity and stability 190

5 Final fate of a massive star 210

5.1 Life cycle of massive stars 213
5.2 Evolution of a physically realistic collapse 215

vii
viii Contents

5.3 Non-spherical models 225

5.4 Blackhole paradoxes 235
5.5 Resolution of a naked singularity 238

References 255
Index 269
 Preface

The physical phenomena in astrophysics and cosmology involve gravitational

collapse in a fundamental way. The final fate of a massive star, when it
collapses under its own gravity at the end of its life cycle, is one of the
most important questions in gravitation theory and relativistic astrophysics
today. The applications and basic theory of blackholes vigorously developed
over the past decades crucially depend on this outcome.
A sufficiently massive star many times the size of the Sun would undergo
a continual gravitational collapse on exhausting its nuclear fuel, without
achieving an equilibrium state such as a neutron star or white dwarf. The
singularity theorems in general relativity then predict that the collapse gives
rise to a spacetime singularity, either hidden within an event horizon of grav-
ity or visible to the external universe. The densities and spacetime curvatures
get arbitrarily high and diverge at these ultra-strong gravity regions. Their
visibility to outside observers is determined by the causal structure within
the dynamically developing collapsing cloud, as governed by the Einstein
field equations. When the internal dynamics of the collapse delays the hori-
zon formation, these become visible, and may communicate physical effects
to the external universe. These issues are investigated here, and the treat-
ment is aimed at showing how such visible ultra-dense regions arise naturally
and generically as the outcome of a dynamical gravitational collapse in Ein-
stein gravity. While it predicts the existence of visible singularities; classical
general relativity may no longer hold in these very late stages of the col-
lapse, and quantum gravity may take over to resolve the classical spacetime
singularity. The quantum effects from a visible, the extreme gravity region
could then propagate to outside observers to provide a useful laboratory for
quantum gravity. Blackholes need not form in such a scenario and there may
be interesting consequences for ultra-high energy astrophysical phenomena
in the universe.
The general theory of relativity, which has strong experimental support, is
used here, and its basics and useful features of spacetimes are reviewed. The
necessary tools are developed as needed, but a prior familiarity with general
relativity would help. It is a pleasure to thank many friends and colleagues

ix
x Preface

for numerous discussions and work as cited, on the themes described here.
Special thanks are due to R. Goswami and I. H. Dwivedi for their ideas and
help and for our studies together. A. Mahajan and S. Khedekar helped with
the manuscript.
 1
Introduction

Gravitation theory and relativistic astrophysics have gone through exten-

sive developments in recent decades, following the discovery of quasars in
the 1960s, and other very high energy phenomena in the universe such as
gamma ray bursts. Compact objects such as neutron stars and pulsars also
display intriguing physical properties, where the effects of strong gravity
fields are seen to play a fundamental role. When the masses and energy
densities involved in the physical phenomena are sufficiently high, as is the
case in the situations above, it has become increasingly clear that the strong
gravitational fields, as governed by the general theory of relativity, play
an important and much more dominant role. This gravitational dynamics
must be taken into account for any meaningful description of these observed
ultra-high energy objects.
A similar situation involving very strong gravitational fields, and which
may be connected to some of the above phenomena, is that of a massive star
undergoing a continual gravitational collapse at the end of its life cycle. This
happens when the star has exhausted its nuclear fuel that provided a balance
against the internal pull of gravity. This phenomenon, dominated essentially
by the force of gravity, is fundamental to basic theory and astrophysical appli-
cations in blackhole physics that have received increasing attention in past
decades, and also in cosmology. In the past two decades, there have been
extensive investigations of gravitational collapse models within the frame-
work of Einstein’s theory of gravity, and these have provided useful insights
into the final fate of a massive star.
This book is about the phenomena of gravitational collapse. Such a col-
lapse of massive matter clouds is at the heart of the physics and astrophysics
of happenings, some of which are mentioned above, where extremely high
mass and energy densities are involved. For example, several models to
explain gamma ray bursts are in terms of a collapsar, where the gravitational
collapse of a single massive star is invoked to understand such a burst of

1
2 Introduction

ultra-high energy. Apart from blackhole physics, gravitational collapse is the

key physical process that is fundamental to the formation of a star itself
from interstellar clouds or nebulae, in the formation of galaxies and clusters
of galaxies, and in structure formation in the universe as a whole. In general,
gravitational collapse of a massive matter cloud would play an important
role in the physical processes and a variety of happenings on a cosmic scale
that involve the force of gravity in an important manner.
A continual gravitational collapse for a massive star would be the situa-
tion when the entire matter cloud collapses and shrinks under the force of its
own gravity. Therefore, gravity overtakes and dominates the other three fun-
damental forces of nature, in particular the weak and strong nuclear forces,
which generically provide the outward pressure in a star to balance it against
the inward pull of gravity of the cloud, in addition to the usual thermal pres-
sures. For massive stars, typically such a collapse takes place when the star
has exhausted its nuclear fuel, and when there is no supporting force left
against the force of its own gravity, which is ever present.
The final outcome of such a collapse depends on the initial mass of the star.
A star with a mass lower than about two to three solar masses will stabilize
as a white dwarf or neutron star after losing some of its original mass. In
these cases, after an initial collapse of the cloud when the star has exhausted
its nuclear fuel, the star again stabilizes at a much smaller radius due to
internal balancing forces provided by either electron or neutron degeneracy
pressures. For heavier stars that are several solar masses, they may again set-
tle to a neutron star final state if the star could throw away the excess mass
in the process of its evolution. However, for more massive stars, none of the
above internal pressures can achieve the required balance, and a continual
gravitational collapse becomes inevitable. The collapse then must proceed
towards creating a spacetime singularity, as predicted by the singularity the-
orems of general relativity theory, which may be hidden within a blackhole
or which may be visible to external observers. A spacetime singularity is a
region where the physical parameters such as mass, energy densities, and
the spacetime curvatures go to their extreme values and blow up, so that the
usual laws of physics break down at such a singularity.
In such extreme regions, however, where the length and time scales are
comparable to the Planck length and time, quantum effects become impor-
tant. These must necessarily be taken into account and combined with the
effects of gravity. At present, we have no mechanism or complete theory to
deal with such quantum effects and the intense force of gravity together in
a unified manner, namely a quantum gravity theory. However, it is widely
believed that a quantum gravity theory, dealing with all forces of nature in
a unified way, would take over from purely classical general relativity when
the collapse reaches extreme matter densities and spacetime curvatures in its
very advanced later stages. In these stages of collapse, it is very likely that
 Introduction 3

when the quantum effects are incorporated together with the gravitational
force, the classical spacetime singularity may be resolved, and may no longer
exist in the full theory.
Gravitational collapse is thus a key phenomena for many astrophysical
processes for stars or other larger systems in the universe. In particular, the
very advanced stages of collapse of a massive star are occurrences in nature
where the effects of both gravity and the quantum would be combined. Even
if the final spacetime singularity, as predicted by classical general relativity,
may be resolved, possibly through quantum gravity effects, such a collapse
will necessarily give rise to spacetime regions of ultra-high mass densities and
curvatures, where the physical effects will be extreme.
The important physical issue would then be whether such extreme gravity
regions formed in the gravitational collapse of a massive star are visible to
external observers in the universe. An affirmative answer here would mean
that the physical phenomena of the gravitational collapse of a massive star
could provide a very good laboratory to study quantum gravity effects in the
cosmos, and this may help towards generating clues for an, as yet, unknown
theory of quantum gravity. A laboratory similar to that provided by the
early universe is then created in the later stages of the continual collapse of a
massive star. An additional feature would be that, whereas the early universe
was a unique event that happened only once, the collapse phenomena would
continue to occur whenever a sufficiently massive star in the universe died
on exhausting its nuclear fuel. If such ultra-strong gravity regions become
visible to external observers in the spacetime, an opportunity to observe the
quantum gravity effects in the universe is provided.
The answer to this is determined by the causal structure of spacetime
in the vicinity of a spacetime singularity. This is actually decided by the
dynamics of the gravitational collapse of the matter cloud, as it evolves from
a regular initial data, defined on an initial surface, from which the collapse
develops. This dynamical evolution is governed by the Einstein equations.
In other words, it is only the study of the collapse dynamics of the matter
clouds that would decide the visibility or otherwise of the ultra-strong gravity
regions. If, as the collapse evolves, the event horizons of gravity develop much
before the spacetime singularity forms, then these extreme gravity regions are
hidden away from the external universe, and a blackhole forms as the collapse
outcome. On the other hand, if such horizons are delayed or fail to develop
during collapse, as governed by the internal dynamics of the collapsing cloud,
then the scenario where the extreme gravity regions are visible to external
observers occurs, and a visible naked singularity forms.
The importance of gravitational collapse processes in relativistic astro-
physics was realized when Datt (1938) and Oppenheimer and Snyder (1939)
used general relativity to study the dynamical collapse of a homogeneous
spherical dust cloud under its own gravity. This model gave rise to the
4 Introduction

concept of a blackhole. The term blackhole itself was popularized only later
in the 1960s. The above work established that, under idealized conditions, a
collapsing cloud of matter with zero pressure will necessarily give rise to a
blackhole. Such a blackhole is a region of spacetime from which no light or
matter can escape away to faraway external observers, and which necessarily
covers the spacetime singularity or the regions of extreme physical condi-
tions from the external universe. Specifically, in order to create a blackhole
as the final state of gravitational collapse of the star, an event horizon must
develop in the spacetime earlier than the time when the final spacetime sin-
gularity forms. Such an event horizon is a one-way membrane such that light
or matter can fall into the region covered by it, but cannot escape away. If
the event horizon developed prior to the formation of the singularity, neither
the singularity nor the collapsing matter that has fallen within it would be
observable to an external observer, and a blackhole is said to have formed
as the final endstate of the collapsing star. All the matter of the star is then
supposed to be crushed into the infinite density singularity at the center of
the blackhole.
How early and when the horizon will actually develop in a realistic col-
lapse is determined by the dynamics of the collapsing matter, the physical
conditions within the star, and the dynamical evolution of the cloud as gov-
erned by the Einstein equations of gravity. Investigations in high energy
astrophysics have already used the concept of a blackhole quite extensively.
However, the actual understanding of the phenomena of gravitational col-
lapse, and the conditions under which it can lead to the blackhole formation,
or otherwise, within the framework of general relativity has progressed only
relatively recently.
Further to the early studies mentioned above, it was generally assumed
that the final endstate of collapse of a massive star will be a blackhole only.
However, several important questions remained unanswered. For example,
what would be the effects of non-zero pressures, which would be certainly
important in the later stages of collapse, towards determining the collapse
endstate, or, how will an inhomogeneous cloud collapse, say with a physically
realistic density profile that is higher at the center and decreases slowly as
one moves away from the center of the star? Early work on gravitational
collapse focused only on simple models with idealized conditions, assuming
a totally homogeneous density within the star, zero pressures, and so on,
which would not be physically realistic. For example, a realistic star must
have non-zero internal pressures, and its density would be typically higher
at the center, as compared with its outer layers.
These physical issues and important questions have been crucial to the
foundations of blackhole physics. But, not much attention could be paid to
them, mainly due to the complexity of the equations of general relativity.
This is because, in general, the Einstein equations are non-linear, second
 Introduction 5

order partial differential equations that are quite difficult to solve. Therefore,
the only model available until the late 1960s for the dynamical gravitational
collapse of a massive matter cloud was that of a homogeneous, pressureless
spherical cloud. In addition, not much attention was paid to these issues
by the general relativists of the 1940s and 50s, who, by and large, did not
consider such ultra-high energy phenomena to be physically realistic or of
much astrophysical significance.
As indicated above, it was only the discovery in the 1960s of very
high energy astrophysical phenomena that generated a keen theoretical inter-
est in the continual gravitational collapse processes. However, mathematical
difficulties and the complexity of gravity theory did not allow much progress.
Then, the cosmic censorship hypothesis was introduced by Penrose (1969),
which conjectured that the outcome of any generic gravitational collapse of
a massive star must lead necessarily only to a blackhole formation as the
collapse final state. This hypothesis thus suggested that the extreme and
ultra-strong gravity regions, or the spacetime singularity, must always nec-
essarily be covered within an event horizon of gravity, and that the external
observers should never be able to see the singularity. This assumption means
that whatever the physical conditions and forces within the massive stars
may be (for example, they may be inhomogeneous in their density distribu-
tion, the pressures may be non-zero, or they may not be totally spherical
and so on), the outcome of their continual collapse must give rise to a black-
hole only. In other words, this amounts to an assumption of the nature of
the allowed dynamical evolutions of the collapsing clouds, namely that the
Einstein equations must permit only those evolutions that create the event
horizon necessarily much prior to the formation of the final singularity or
the ultra-strong gravity regions. Then, the singularity would be necessarily
hidden within the horizon, which is a one-way surface, not allowing it to
be seen by any external observers.
The cosmic censorship conjecture thus implies that no ultra-strong grav-
ity regions forming in continual collapse will be visible to outside observers.
That is, no naked singularity will develop in the collapse, and the event hori-
zon developing in the dynamical collapse will always manage to cover these.
Hence, the outcome of any gravitational collapse is necessarily a blackhole,
and external observers can never see any ultra-strong gravity regions forming
in the collapse, as indicated in Fig. 1.1.
As yet, a specific mathematical formulation for cosmic censorship that has
been properly defined does not exist. Then, a proof of the same would have to
be obtained within the framework of Einstein’s gravity theory. The cosmic
censorship assumption nevertheless provided a major impetus to develop-
ments in blackhole physics, and two parallel streams of developments took
place. On one hand, the theoretical properties of blackholes were devel-
oped extensively, using cosmic censorship as the basic assumption, thus
6 Introduction

Spacetime singularity

Event horizon

Collapsing matter cloud

Initial surface

Fig. 1.1 The final outcome of a generic gravitational collapse must be a

blackhole according to the cosmic censorship conjecture. Then, the eventual
spacetime singularity of the collapse has to be preceded by the event horizon of
gravity.

creating the laws of blackhole thermodynamics and related aspects (see for
example, Hawking and Ellis, 1973). On the other hand, efforts to estab-
lish the censorship hypothesis continued, as it was clear all along that this
assumption was absolutely fundamental to the theory and applications in
blackhole physics, and so it needed a rigorous formulation and proof within
the framework of general relativity. It is widely recognized that a proof
of the censorship conjecture would place blackhole physics and its applica-
tions on a sound footing, whereas its failure would actually throw blackhole
dynamics and related applications into serious doubt. Hence, the validity,
or otherwise, of the cosmic censorship conjecture has remained an issue of
crucial importance for all these years. The efforts to prove it have not suc-
ceeded for the past three decades, and there are even serious difficulties
in formulating any rigorous mathematical version or a statement for this
conjecture.
The theme that the only way out of this impasse is to study rigorously
the dynamical gravitational collapse phenomena within the framework of
Einstein’s theory of gravity is proposed and developed here. This has been
investigated extensively in the last couple of decades, and some of the
issues that have been addressed include: what is the outcome of a continual
gravitational collapse under physically realistic conditions, as governed by
the Einstein equations? Will it be necessarily a blackhole as hypothesized
by the censorship conjecture, or would it give rise to a naked singularity,
where ultra-strong gravity regions forming in collapse are visible to external
observers? In the latter case, would it be possible to observe the quantum
 Introduction 7

gravity effects taking place in these visible ultra-strong gravity regions? Some
of these issues are discussed here.
A detailed study of the collapse phenomena may be the only way towards
any possible physically realistic formulation of censorship, if one exists. Such
a study and investigation of collapse could also lead to novel physical insights
and possibilities emerging out of the intricacies of the gravitational force. It
would appear that beyond the studies so far, mainly of static and stationary
solutions modeling blackholes, investigating dynamical evolutions as permit-
ted by the Einstein equations would offer new insights into the nature of
gravity. This is an arena that has been explored less, and which needs to be
investigated carefully in detail.
To this end, gravitational collapse scenarios with non-zero pressures and
more realistic equations of state for classes of general matter fields are con-
sidered here. A general formalism is developed to treat the spherical collapse
from regular initial data. These considerations also point to why it has not
been possible so far to make any definite progress on the censorship conjec-
ture. It is seen that it is first necessary to acquire a deeper and more extensive
understanding of the dynamical evolutions and gravitational collapse pro-
cesses in general relativity. Recent work on studying and understanding
the final fate of dynamical gravitational collapse in gravitation theory is
discussed. General matter fields are considered so as to include important
physical features in the collapse, such as inhomogeneities in matter distri-
bution, non-vanishing pressures, different forms for the equations of state
of the collapsing matter, and other such aspects. It is seen that in spherical
gravitational collapse, given the matter initial data on an initial surface from
which the collapse develops, there are the rest of the free initial data such
as the velocities of the collapsing shells, and the classes of the dynamical
evolutions as permitted by the Einstein equations, which lead to the final
state that is either a blackhole, or a naked singularity that is a visible ultra-
strong density and curvature region forming in the collapse not covered by
an event horizon. The nature of the outcome depends on the regular initial
data from which the collapse evolves, and the allowed dynamical evolutions
in the spacetime, as permitted by the Einstein equations.
After the basics of the structure and properties of spacetimes and the
essentials of relativity theory are summarized in Chapter 2, the above issues
are discussed in Chapter 3. Collapsing dust clouds, which generalize and
include as a special case the Oppenheimer–Snyder dust collapse models, and
which give an idea of the possible outcomes of gravitational collapse in terms
of a blackhole or a naked singularity, are also discussed in Chapter 3. The
Oppenheimer–Snyder dust collapse scenario is included here as a special case
when the cloud is homogeneous. It is seen, however, that a more realistic den-
sity profile with a density higher at the center and decreasing as one moves
away from center, gives rise to a naked singularity as the collapse endstate
8 Introduction

Spacetime singularity
Event horizon
Outgoing null geodesics

Collapsing matter cloud

Initial surface

Fig. 1.2 If the collapsing cloud is inhomogeneous, with a density higher at the
center, the trapped surface formation and event horizon in the collapse are
delayed to give rise to a naked singularity, where the ultra-strong gravity regions
are visible to outside observers.

(see Fig. 1.2). In general, it is seen that the collapse outcome depends on
the nature of the initial matter profiles and the evolutions allowed by the
Einstein equations. The structure of this spacetime in the homogeneous den-
sity case gives rise to the basic notion and concept of a blackhole. The dust
collapse picture provides a concrete background to the possible final states
of a continual gravitational collapse.
Chapter 4 then studies several useful aspects of spacetime structure,
singularities and collapse, as related to the cosmic censorship hypothesis pos-
sibilities and the structure of naked singularities developing in gravitational
collapse. It is pointed out that while the cosmic censorship does not hold in
general relativity in the obvious sense of ruling out naked singularities from
all physically realistic gravitational collapse models, any definite formulation
of this hypothesis will depend on a detailed analysis of stability and gener-
icity aspects related to collapse scenarios, and the naked singularities and
blackhole phases developing as final outcomes of the gravitational collapse.
Several possibilities towards any plausible formulation are discussed.
In light of the results available so far and the emerging scenario, the key
physical issue is the possible final state of a massive star. The basic problem
to be addressed is: what will the final outcome of the gravitational collapse
of a massive star be when it collapses freely at the end of its life cycle on
exhausting its nuclear fuel under the force of its own gravity? Under realistic
astrophysical conditions, will it turn into a blackhole, or does it terminate as
a naked singularity? Are there any observable consequences in the latter case?
 Introduction 9

These physical questions underlie many considerations here on gravitational

collapse.
While theoretical properties of blackholes have been studied rather exten-
sively, the naked singularity solutions in general relativity, arising out of
dynamical collapse studies are relatively less understood as yet. It is some-
times asked how a naked singularity could arise in the collapse, allowing
the light to escape from the extreme gravity regions even when the gravity
fields are so strong. Some of these aspects are discussed in Chapter 5. Also
explored and pointed out here are the physical features, such as the role of
inhomogeneities and spacetime shear, that lead to a naked singularity rather
than a blackhole as the collapse endstate. The physics that possibly causes
a naked singularity in the collapse, rather than a blackhole, is examined.
While it may be stated that a good understanding now exists on spherical
collapse in general for a generic matter field, non-spherical collapse remains
major uncharted territory. This is also closely related to the stability and
genericity aspects of collapse outcomes, and these issues are discussed here.
The information loss paradox and related issues have highlighted some of
the important problems with the blackhole paradigm, which also include the
existence of an infinite density spacetime singularity at the center of a black-
hole, leading to an instability even at the classical level, and uncertainties
of the correctness, or otherwise, of the cosmic censorship conjecture. Under
such a situation, a possibility worth considering could be the avoidance or
delay of trapped surfaces formation as the star evolves, collapsing under grav-
ity. This is the case when a collapse evolution to a naked singularity takes
place, where the trapped surfaces do not form early enough or are avoided
in the spacetime. In that case, in the late stages of the collapse, the star
could radiate away most of its mass. This could then offer a way out of the
blackhole conundrums, whilst also resolving the singularity problem.
As such, the outcomes of a continual collapse, namely the blackhole and
naked singularity, are very different from each other in nature. The naked
singularity, which is more like an event than an object in many cases, could
have quite different physical properties compared with a blackhole. There-
fore, the implications of the visibility of the ultra-high density and curvature
regions to a faraway observer in the spacetime need to be investigated. Such
a scenario offers an intriguing possibility that the quantum gravity effects
may become observable during the later final stages of the collapse. This is
because the ultra-strong gravity regions where quantum gravity effects take
place are now no longer hidden under the event horizon, but are visible and
can, in principle, communicate with external observers. This may offer inter-
esting connections and pointers towards observational effects of quantum
gravity arising from gravitational collapse. These possibilities are discussed
in Chapter 5, where some implications of loop quantum gravity formalism
from such a perspective are indicated.
 2
The spacetime manifold

Here, the essential fundamentals of general relativity and related mathemat-

ical aspects are described. For further details, see texts such as Weinberg
(1972), Misner, Thorne, and Wheeler (1973), and Wald (1984). Other neces-
sary techniques are developed in later chapters as necessary. While defining
vectors, tensors, and other quantities, we use both a local and a coordinate
free global approach, and indicate how to make a transition from one to the
other representation, which is useful in several situations.
In Section 2.1 the manifold model for spacetime is introduced. Basic defi-
nitions of a differentiable manifold, and various topological and orientability
properties are discussed. The metric tensor and related aspects are con-
sidered in Section 2.2, and the connection on a spacetime is considered in
Section 2.3. Timelike and null geodesics play a basic role in the considerations
here on gravitational collapse. These are a special set of non-spacelike trajec-
tories that represent the motion of freely falling material particles and light
rays, and they clarify many properties of a spacetime. These are discussed
in Section 2.4. The spacetime curvature is considered in Section 2.5, and
the Einstein equations governing the dynamics of matter in the spacetime
are discussed in Section 2.6. Many exact solutions have been found to the
Einstein equations so far; however, the Schwarzschild and Vaidya geometries
are particularly relevant to gravitational collapse scenarios, and Section 2.7
discusses these.

2.1 The manifold model

The universe is modeled as a four-dimensional spacetime M in general rel-
ativity, together with an indefinite Lorentzian metric tensor g , which has
the signature (−, +, +, +). Conditions ensuring physical reasonability to
the spacetime model are generally assumed. These include the space and

10
 2.1 The manifold model 11

time orientability, and necessary topological regularity conditions such as

the Hausdorffness and connectedness. Here, this basic model of the space-
time universe that underlies Einstein’s theory of gravitation is specified. The
manifold model for the universe naturally incorporates the observed continu-
ity of space and time at the classical level, and the basic principle of general
relativity where the locally flat regions combine to produce a globally curved
continuum. This implies that a smooth change of coordinates is possible
when a transition is made from one coordinate patch to another.

2.1.1 Differentiable manifolds

The n-dimensional Euclidian space Rn is a collection of all n-tuples
(x1 , . . . , xn ) such that −∞ < xi < ∞, i = 1, . . . , n, and which has the
natural Euclidian metric. An open ball of radius r around any point x in
Rn is the set of all points y such that | x − y |< r, where the modulus
denotes the positive definite distance as defined by the Euclidian metric on
Rn . The open sets in Rn are sets which can be expressed as a union of such
open balls.
Basically, an n-dimensional differentiable manifold is a set that is locally
similar to an open set of Rn . Therefore, locally Euclidian patches are glued
together smoothly to obtain a space which need not be Euclidian globally.
An n-dimensional, C ∞ , real differentiable manifold is a set M, together
with a collection {uα , φα }, called an atlas for M . Here, the uα values are
subsets of M and the φα values are one–one maps of a given uα onto an open
subset in Rn , which satisfy the following.

(1) The sets uα form a cover for M , that is, any given p in M must be in a
uα for some value of α, and


M= uα . (2.1)
α

(2) Whenever two neighborhoods uα and uβ intersect, that is, uα ∩ uβ = φ,

then the map φα ◦ φ−1β from R to R , which takes points of φβ (uα ∩ uβ )
n n

to points of φα (uα ∩uβ ), is infinitely differentiable in a continuous manner

(a smooth C ∞ -function) as a mapping between two open subsets of Rn
(see Fig. 2.1).

Alternatively, it is possible to consider the map φβ ◦ φ−1

α , and the same
condition again holds. Each uα is called a local coordinate neighborhood
or a chart where p ∈ uα has coordinates of φα ( p) in Rn . The condition
(2) above ensures that whenever an event p ∈ M undergoes a coordinate
12 The spacetime manifold

Rn
f(u)
M
u f f
( p)
p

u
f

q f
(u
)
f
(q)

Fig. 2.1 All events p and q in the manifold have neighborhoods which are
homeomorphic to subsets in Rn . The points p, q ∈ M have coordinates
of φα (p) and φβ (q). Whenever the neighborhoods in M intersect, there should
be a smooth change of coordinates.

change, the change is necessarily smooth. That is, if {xi } and {y i } are
local coordinates of p ∈ M in uα and uβ respectively, then the functions
xi = xi ( y 1 , . . . , y n ) are C ∞ -functions from Rn to Rn . A maximal or complete
atlas is chosen for the spacetime manifold M, that is, if {uα , φα } is an atlas
for M , one selects for M the atlas that consists of all other atlases that are
compatible with {uα , φα }. This implies that their union with {uα , φα } is also
a C ∞ -atlas.
This implies that one has included all possible, mutually compatible coor-
dinate systems for the manifold M. A C r -manifold is defined in a similar way,
where it is required that the transition functions φα ◦ φ−1 β are r-times
continuously differentiable, where a continuous function is denoted by C 0 .
The Euclidian plane R2 , or Euclidian space Rn , is, in itself, a manifold as
it is covered by a single chart Rn , where φ would be the identity map with
the coordinate range −∞ < xi < ∞ for i = 1, . . . , n. Another example of
such a manifold is the two-sphere S 2 defined by

S 2 = {(x1 , x2 , x3 ) ∈ R3 | (x1 )2 + (x2 )2 + (x3 )2 = 1}. (2.2)

The six hemispherical open sets Oi± for i = 1, 2, 3 are given by Oi± =
{(x1 , x2 , x3 ) ∈ S 2 | ±xi > 0}, which cover S 2 . Each Oi± is mapped
onto the open disk {(x, y ) ∈ R2 | x2 + y 2 < 1} by the projection maps such
as f1+ (x1 , x2 , x3 ) = (x2 , x3 ). The overlap functions fi± ◦ ( fj± )−1 are C ∞ -
functions in their domain of definition. Thus, S 2 is a two-dimensional,
C ∞ -manifold that cannot be covered by a single coordinate system. Similarly,
the sphere S n in n-dimensions is also a differentiable manifold.
 2.1 The manifold model 13

2.1.2 Vectors and one-forms

A function f : M → Rn is called differentiable if the map f ◦ φ−1 ∞
α is a C -
n n r
map for all charts φα as a map from R to R ; C -functions can be defined
similarly (Spivak, 1965).
Suppose now M and M  are two differentiable manifolds with φα and ψα
denoting charts of M and M  respectively. A map h : M → M  is called
C r -differentiable if ψα ◦ h ◦ φ−1 r
α is always C -differentiable as a map from
R to R for all α. If the dimension of M is n and that of M  is n with
n n

n > n , then the map h cannot be one–one. However, if h is one–one, onto,

and continuous from M to M  such that h−1 is also a continuous map, then
h is called a homeomorphism. If a homeomorphism and its inverse are both
C r -maps, then it is called a C r -diffeomorphism.
A C k -curve in M is a C k -map from an interval of R into M . A vector
(or a contravariant vector) (∂/∂t)λ(t0 ) , tangent to a C k -curve λ(t) at a point
λ(t0 ), is an operator from the space of all smooth functions on M into R:
   
∂ ∂f f [λ(t + s)] − f [λ(t)]
( f) = = lim , (2.3)
∂t λ(t0 ) ∂t λ(t0 ) s→0 s

where s denotes a small increment of the parameter t. This is d( f ◦ λ)/dt,

which is the derivative of f in the direction of λ(t) with respect to parameter
t. If f = t, where t is the parameter along the curve,
 
∂
(t) = 1. (2.4)
∂t λ

If the xi values are local coordinates in a neighborhood of p = λ(t0 ), then

 
∂f dxi ∂f
= |λ(t0 ) , (2.5)
∂t λ(t0 ) dt ∂xi

where a repeated index means summation over the values 1, . . . , n. (This

summation convention is used throughout.) Therefore, every tangent vector
at p ∈ M is expressed as a linear combination of the coordinate derivatives,
which are (∂/∂x1 )p , . . . , (∂/∂xn )p . Conversely, any linear combination of
these operators that are partial derivatives with respect to coordinates can
be chosen, namely, V i (∂/∂xi )p , with the values of V i being any numbers.
It is then possible to find a curve which admits this linear combination as
a tangent (see for example, Wald, 1984). The vectors (∂/∂xj )p are linearly
independent (if not, then there are numbers V i such that
 
i ∂
V = 0, (2.6)
∂xi p
14 The spacetime manifold

Fig. 2.2 The tangent space Tp at a point p ∈ M , which gives the set of all
directions at that point.

with at least one V i being non-zero, and applying this to the coordinate
functions x1 , . . . , xn gives V i = 0 for all i, a contradiction). Therefore, the
vectors (∂/∂xj ) span the vector space Tp , the space of all tangent vectors at
p (see Fig. 2.2). The vector space structure here is defined by

(αX + βY )f = α(Xf ) + β (Y f ), (2.7)

for α, β ∈ R and X, Y ∈ Tp , where X and Y are vectors at p; Tp is also

called the tangent space at p. The basis {(∂/∂xi )p } is called a coordinate
basis of Tp . A general basis is denoted by {ei }, where i = 1, . . . , n, are
linearly independent vectors. Then, for any vector V ∈ Tp ,

V = V i ei , (2.8)

where the numbers V i are called the components of V with respect to the
basis ei . In a coordinate basis, V i = dxi /dt. Again, {∂/∂xi } forms a basis
of Tp which means the dimension of Tp is n.
For the tangent space Tp at p ∈ M , the vector space of all dual vectors at
p, also called covariant vectors or one-forms at p, can be naturally defined. A
one-form ω at p is a real-valued linear functional on Tp , denoted by ω (X ) ≡
ω, X , and the linearity condition implies

ω, αX + βY = αω, X + βω, Y . (2.9)

Given a tangent space basis {ea }, a unique set of one-forms {ea } is given
by the condition that the one-form eb maps a vector V into V b , that is, the
bth component of V in the basis ea . Therefore,

eb , V = V b , (2.10)

where a, b, . . . and i, j, . . . denote indices for vectors and tensors.

 2.1 The manifold model 15

From the above,

ea , eb = δba , (2.11)

where the right-hand side is the Kronecker delta function. The linear
combinations of one-forms ω and η are defined by

αω + βη, V = αω, V + βη, V , (2.12)

with α, β ∈ R. Then, {ea } is a basis for the space of all one-forms at p

because any one-form ω can be written as ω = ω a ea with

ωa = ω, ea . (2.13)

Therefore, the set of all one-forms at the event p forms a vector space at p,
the dual of Tp , and is denoted by Tp∗ . The basis ea is a dual basis to ea . If
ω ∈ Tp∗ and V ∈ Tp , then

ω, V = ωa ea , V b eb = ωa V b δba = ωa V a . (2.14)

A vector field V on a manifold M is an assignment of a tangent vector Vp

at each p ∈ M . The vector field is said to assign vectors smoothly if, for each
smooth function f on M , the function V ( f ), the directional derivative of f
along the vector Vp , is also smooth on M at each point p. The coordinate
basis vector fields ∂/∂xi are smooth, and so a vector field will be smooth
provided that its coordinate components V i are smooth functions. Given two
vector fields V and W , a new vector field, called their commutator [V, W ],
is defined by
[V, W ]( f ) = V [W ( f )] − W [V ( f )]. (2.15)

The commutator for any two coordinate basis vector fields vanishes. If f
and g are any two smooth functions, it can be seen that [V , W ]( f + g ) =
[V , W ]( f )+[V , W ](g ) and that [V , W ](αf ) = α[V , W ]( f ) for any α ∈ R.
It can be shown that

[V , W ]( f g ) = f [V , W ](g ) + g [V , W ]( f ), (2.16)

which is the product property. By expanding in a coordinate basis it is seen

that the commutator [V , W ] will be a smooth vector field if and only if both
V and W are smooth. Note that [V , V ] = 0 and that [V , W ] = −[W , V ].
Furthermore, the commutator is linear in each of its arguments with respect
to addition, that is

[V 1 + V 2 , W ] = [V 1 , W ] + [V 2 , W ]. (2.17)
16 The spacetime manifold

Any smooth function f on M defines a one-form df , called the differential

of f , by the rule
df, V ≡ V f. (2.18)
Therefore, in a coordinate basis,

∂f
df, V = V a . (2.19)
∂xa
The local coordinate functions (x1 , . . . , xn ) are used to define a set of one-
forms (dx1 , . . . , dxn ), which is a basis dual to the coordinate basis because
 
a ∂ ∂xa
dx , b = = δba . (2.20)
∂x ∂xb

Also,
 
∂ ∂f
df = df, dxa = dxa , (2.21)
∂xa ∂xa

which is the usual definition of the differential df.

If f is a non-constant function, the surfaces f = const. define an (n − 1)-
dimensional submanifold of M . Consider the set of all the vectors V ∈ Tp
such that
df, V = V f = 0, (2.22)
then the vectors V are tangent to curves in the f = const. submanifold,
through p. Therefore, the differential df is normal to the surface f = const.
at p.

2.1.3 Topological structure

A C ∞ -maximal atlas on a spacetime manifold M induces a natural topology
on M, given by the companion Euclidian space by requiring each φα to be
a homeomorphism. Therefore, the open sets in M are pre-images of open
sets in Rn and their unions. Then, the collection {uα } provides a basis for
the spacetime topology, where Rn has its canonical topology, defined by the
metric  1/2
d(x, y ) = (x1 − y1 )2 + · · · + (xn − yn )2 (2.23)
for any x, y ∈ R .
n

Several topological regularity conditions that are assumed for a physically

reasonable spacetime manifold are now given. The spacetime is assumed to be
 q in M , there are disjoint open sets
Hausdorff, that is, given p and q with p =
uα and uβ in M such that p ∈ uα and q ∈ uβ . Physically interesting space-
time examples such as the Schwarzschild geometry and Robertson–Walker
 2.1 The manifold model 17

models are Hausdorff. This is a reasonable requirement on a spacetime which

ensures the uniqueness of limits of convergent sequences, and incorporates
the intuitive notion of distinct spacetime events.
Next, the spacetime M has no boundary. A boundary represents, in a
sense, the ‘edge’ of the universe, which is not detected by any astronomical
observations. Mathematically, it is common to have manifolds without a
boundary. For example, for a two-sphere S 2 in R3 , no point in S 2 is a
boundary point in the induced topology as implied by the natural topology
on R3 , because all neighborhoods of any p ∈ S 2 are contained within S 2 in
this induced topology. Assume M to be connected, that is, M = X1 ∪ X2 ,
with X1 and X2 being two open sets and X1 ∩ X2 = φ is not possible. This is
because disconnected components of the universe cannot interact by means
of any signals, and the observations are confined to the connected component
where the observer is situated. But, M could be either simply connected or
multiply connected. For further discussion on multiply connected spacetimes
and the notion of a wormhole in the Schwarzschild geometry, see Wheeler
(1962, 1964) and Misner, Thorne, and Wheeler (1973). Such wormholes are
like ‘handles’ in the multiply connected topology of space and can connect
widely separated regions in space.
It is known, however, that such wormholes are not stable and collapse
as soon as created, unless the violation of the energy condition in an aver-
aged sense is allowed, thus implying negative energy fields (see for example,
Deutsch and Candelas, 1980; Lee, 1983; Morris, Thorne, and Yurtsever,
1988). Therefore, a wormhole may be stabilized only by shifting the energy
of vacuum to be negative by some quantum processes. The process of topol-
ogy change could also give rise to a multiply connected spacetime. It is
not clear if the topology of space could change while it evolves in time,
and if so, what physical agencies cause it. A topology change can affect
the structure of spacetime severely to cause naked singularities (Joshi and
Saraykar, 1987).
A spacetime is assumed to be non-compact, because compact spacetimes
violate causality and admit closed timelike curves. One could then enter one’s
own past, which is considered to be highly unphysical. Usually, M is also
taken to be paracompact. An atlas {uα , φα } is called locally finite if there is
an open set containing every p ∈ M that intersects only a finite number of
the sets uα . A manifold M is called paracompact if, for every atlas {uα , φα },
there is a locally finite atlas {Oβ , ψβ } with each Oβ contained in some uα .
For a further discussion on these topological concepts, see Simmons (1963)
and Willard (1970).
For a connected, Hausdorff manifold, the paracompactness property is
equivalent to the existence of a countable base for the topology of M . The
existence of a Lorentz metric globally on M implies any Hausdorff manifold
with a C r Lorentz metric tensor must be paracompact (Geroch, 1968b).
18 The spacetime manifold

Let B be the set of all ordered basis {ei } for Tp . If {ei } and {ej
} are in
B , then
ej
= aij
ei . (2.24)

If a denotes the matrix [aij ], then det[a] = 0. An equivalence relation in B

is introduced by the condition that ei ∼ ej
if and only if det[a] > 0. Clearly,
there are exactly two such equivalence classes that are called the orientations
of Tp . By an arbitrary choice, one of these classes is called a positive orien-
tation and the other a negative orientation for Tp . Now, let M be a manifold
and p ∈ M . Suppose for all p there is a neighborhood U of p and n continu-
ous linearly independent vector fields {ξ 1 , . . . , ξ n } such that for any q ∈ U ,
the basis {ξ 1 (q ), . . . , ξ n (q )} belongs to the same equivalence class under a
transformation of coordinates when one chooses another chart V containing
q . Then, under all possible coordinate transformations, the determinant for
the transformation matrix of the basis has the same sign, and M is called an
orientable manifold. This definition could then also be stated as below. An
n-dimensional manifold M is called orientable if M admits an atlas {Ui , φi }
such that, whenever Ui ∩ Uj = ∅, then
 
∂xi
J = det > 0, (2.25)
∂xj

where {xi } and {xj } are local coordinates in Ui and Uj respectively. The
Möbius strip is an example of a non-orientable manifold.

2.1.4 Tensors
Tensors are geometric objects on a spacetime, and are invariant under the
change of coordinates. The stress–energy tensor represents various matter
fields existing on a spacetime, such as the electromagnetic field, dust, and so
on. On the other hand, the global geometry and curvature of the manifold
are described by tensor fields such as the metric tensor and the curvature
tensor. In the general theory of relativity, the form of physical laws remains
unchanged under the general transformation of coordinates (principle of gen-
eral covariance). So, physical fields are represented by various tensor fields on
a spacetime, and the laws governing these are tensor equations, valid under
arbitrary coordinate transformations. In an inertial coordinate frame, these
reduce to special relativity laws.
A tensor T of type (r, s) at p ∈ M is a multilinear real-valued functional
on the Cartesian product, as given by

Tp∗ × · · · × Tp∗ × Tp × · · · × Tp → R, (2.26)

 2.1 The manifold model 19

where there are r-factors of Tp∗ and s-factors of Tp . Therefore, T acts on

one-forms and vectors in general to produce a real number.
If T is a tensor of (r, s) type at p ∈ M , then

T (ω 1 , . . . , ω r , V 1 , . . . , V s ) = T (ωi1 ei1 , . . . , ωir eir , V j1 ej1 , . . . , V js ejs ).

(2.27)
Using multilinearity, the above equation can be written as

T i1 ...ir j1 ...js ωi1 . . . ωir V j1 . . . V js , (2.28)

where
T i1 ...ir j1 ...js ≡ T (ei1 , . . . , eir , ej1 , . . . , ejs ), (2.29)

and {ei } and {ei } are basis vectors at p for Tp and Tp∗ respectively.
The space of all tensors of type (r, s) at p is called the tensor product
Tsr ( p), denoted by,

Tsr ( p) = Tp ⊗ · · · ⊗ Tp ⊗ Tp∗ ⊗ · · · ⊗ Tp∗ , (2.30)

with r-factors of Tp and s-factors of Tp∗ . The dimension of Tsr is nr+s , with n
being the dimension of the manifold. This is a vector space of all (r, s) tensors
over real numbers with the addition of tensors and scalar multiplication
defined in a natural manner. In particular, a vector is a tensor of type (1,0)
and a one-form is a tensor of type (0,1). Using the basis vectors {ei } and
{ei } for the tangent space and cotangent space at p, the set

{ei1 ⊗ · · · ⊗ eir ⊗ e j1 ⊗ · · · ⊗ e js } (2.31)

forms a basis for the tensor product Tsr ( p) with all indices running from 1
to n. Then, any tensor T ∈ Tsr can be expressed as

T = T i1 ...ir j1 ...js ei1 ⊗ · · · ⊗ eir ⊗ e j1 ⊗ · · · ⊗ e js , (2.32)

with the tensor components T i1 ...ir j1 ...js defined as above.

Consider a change of coordinates, giving a change of basis {ei } to {ei }
and {ej } to {ej
}. If a coordinate basis {∂/∂xi } for Tp and a corresponding
basis {dxi } for the cotangent space Tp∗ are chosen, then under a change of


coordinates, the components of T in the new coordinates {xi } are
 
i
1 ...i
r i
1 ∂
i
r ∂
T j1
...js
=T dx , . . . , dx , j
, . . . , j
. (2.33)
∂x 1 ∂x s
20 The spacetime manifold



Since xi are functions of xi , substituting in the above equation for ∂/∂xi


and dxi gives, for the transformed components,



i
1 ...i
r i1 ...ir ∂xi1 ∂xir ∂xj1 ∂xjs
T j1
...js
=T j1 ...js . . .
. . . . (2.34)
∂xi1 ∂xir ∂xj1 ∂xjs

Therefore, the components of a vector V and a one-form ω transform as


i
∂xi i ∂xi
V = V , ω i
= ωi . (2.35)
∂xi ∂xi

For a general set of basis vectors, the transformation of components of a
tensor can be written similarly.
For a tensor of type (r, s), the contraction of T over a contravariant index
and a covariant index is defined to be a tensor C (T ) of type (r − 1, s− 1). For
example, if one contracts over the first contravariant and covariant indices,
this gives

C11 (T ) = T ij...l im...n ej ⊗ · · · ⊗ el ⊗ em ⊗ · · · ⊗ en . (2.36)

Using the equations above for a transformation under the change of basis
vectors, it is seen that the contraction C 1 1 is independent of the basis used
and so invariant under a change of coordinates. Similarly, T can be contracted
over any pair of contravariant and covariant indices.
In the space of all tensors of type (r, s) at p, the addition of two tensors
T and T  , and multiplication by a real number α, are defined as

(T + T  )(ω 1 , . . . , ω r , X 1 , . . . , X s ) = T (ω 1 , . . . , ω r , X 1 , . . . , X s )
+ T  (ω 1 , . . . , ω r , X 1 , . . . , X s ),
(αT )(ω 1 , . . . , ω r , X 1 , . . . , X s ) = αT (ω 1 , . . . , ω r , X 1 , . . . , X s ). (2.37)

The outer product of the two tensors T and S of type (r, s) and (r, s ) can
now be defined in terms of their components to give a new tensor T ⊗ S ,

(T ⊗ S )i1 ...ir+r
j1 ...js+s
= T i1 ...ir j1 ...js S ir+1 ...ir+r
js+1 ...js+s
. (2.38)

This allows new tensors to be constructed out of vectors and dual vectors.
A tensor field of type (r, s) on M is an assignment of a tensor of the same
type at all p in M . It is C k -differentiable if all the components of T have the
same differentiability as functions of the coordinates.
If T is a (0, 2) type tensor, it acts on the pairs of vectors V , W to produce
a real number T (V , W ) = Tij V i W j . Then T is called symmetric if

T (V , W ) = T (W , V ). (2.39)
 2.2 The metric tensor 21

If {ei } is a basis for the tangent space, this implies T (ei , ej ) = T (ej , ei ),
that is
Tij = Tji . (2.40)

Similarly, T is called antisymmetric if

Tij = −Tji . (2.41)

One can formulate this in terms of symmetric and antisymmetric parts of T .

For Tij , its symmetric part is written as

1
T(ij) = (Tij + Tji ), (2.42)
2!

and its antisymmetric part is written as

1
T[ij] = (Tij − Tji ). (2.43)
2!

Then, T is called symmetric if T(ij) = Tij , and antisymmetric if T[ij] = Tij .

In general, for a tensor Ti1 ...ir of type (0, r), T(i1 ...ir ) is defined as the sum
over all permutations of indices i1 , . . . , ir divided by r!. Similarly, T[i1 ...ir ] is
defined as the alternating sum over all permutations of the indices i1 , . . . , ir
divided by r!. Therefore, for example

1 i
T[ijkl] = i
[T + Tklj i
+ Tljk − Tkjl
i
− Tlkj
i
− Tjlk
i
]. (2.44)
3! jkl

In general, a tensor of type (r, s) is called symmetric over a collection of

indices if it equals its symmetric part over these indices, and antisymmetric
tensors are defined in a similar manner.

2.2 The metric tensor

The notion of distance between any two infinitesimally separated points of a
spacetime manifold is defined by the metric tensor. These distances locally
reduce to those given by special relativity, which have a flat metric with an
indefinite signature on the Minkowski spacetime. As special relativity is seen
to be valid by experiments, it must hold when confined to local regions in
the spacetime, corresponding to measurements of space and time intervals
at the laboratory scale. Therefore, the spacetime distances between events
need not be positive definite.
This is carried out by assuming the existence of an indefinite metric tensor
field g defined globally on M as a (0, 2) type, symmetric tensor. Therefore,
22 The spacetime manifold

the metric tensor acts on pairs of vectors to give a number, and is symmetric
in its indices. Choosing a coordinate basis,

g ≡ gij dxi ⊗ dxj , (2.45)

where gij = g (∂/∂xi , ∂/∂xj ). If V and W are any vectors, this gives
g (V , W ) = gij V i W j . This is written conventionally in the form of a distance
between two infinitesimally separated points in the spacetime as

ds2 = gij dxi dxj . (2.46)

For a single vector V , g (V , V ) is the magnitude of V , which is gij V i V j .

The metric tensor is assumed to be non-degenerate, that is, there is no
non-zero vector V = 0 such that g (V , W ) = 0 for all W ∈ Tp . Then, the
matrix [gij ] is non-singular, so there must be an inverse matrix g ij such that

g ij gjk = δki . (2.47)

So the tensors g ij and gij give an isomorphism or a unique correspondence

between the space of covariant and contravariant vectors as

Xi = gij X j , X i = g ij Xj . (2.48)

Similarly, a second rank tensor T can also be written as

T i j = g ik Tkj , T j i = g jk Tki , T ij = g ik g jl Tkl . (2.49)

In particular,
g ik gkm = g i m = δ i m , (2.50)
i i
and the Kronecker delta δm transforms as components of a tensor, so δm and
i
gm are identical tensors.
The tensors T i j , T j i , or T ij are treated as representations of the same
geometric object because these are uniquely associated. Such an isomor-
phism between the covariant and contravariant arguments is equivalent to
the procedure of ‘raising’ and ‘lowering’ of indices as pointed out above. The
multilinear map
g : Tp × Tp → R (2.51)

can also be viewed as a linear correspondence from Tp to Tp∗ in the sense of

the mapping V → g (., V ). The non-degeneracy of the metric implies that
this map is one–one and onto and so g establishes a one–one correspondence
between vectors and dual vectors. The components Vi = gij V j are the one-
form components uniquely associated with the vector components V j .
 2.2 The metric tensor 23

Suppose M is an n-dimensional manifold with g being the metric on it.

Then, at any p ∈ M one can always choose an orthonormal basis {ei } such
that the metric components of the gij values have the diagonal form

gij = diag(+1, . . . , +1, −1, . . . , −1). (2.52)

If the metric has the form gij = (+1, . . . , +1) then it is called positive definite.
Then g (X, X ) = 0 implies X = 0. It is called a Lorentzian metric if the
form is
gij = diag(+1, . . . , +1, −1), (2.53)
with (n − 1) terms being positive. The metric is indefinite in the sense that
the magnitude of a non-zero vector could be positive, negative or zero. The
vector X ∈ Tp is called timelike, null, or spacelike, as defined by

g (X, X ) < 0, g (X, X ) = 0, g (X, X ) > 0, (2.54)

respectively. An indefinite metric divides the vectors in Tp into three disjoint

classes, which are the timelike, null, and spacelike vectors. The null vectors
form a cone in the tangent space Tp that separates the timelike vectors and
the spacelike vectors.
The differentiable manifold of four dimensions, with a globally defined
Lorentzian metric tensor is called a spacetime manifold. The signature of the
metric tensor is defined as the number of its positive eigenvalues minus the
number of negative eigenvalues. Therefore, a spacetime is a four-dimensional
differentiable manifold with a Lorentzian metric globally defined, and which
has the signature +2.
In the special theory of relativity, the spacetime admits a global coordinate
frame covering the entire manifold so that the metric has the form given by
(2.53) globally. The metric coefficients are constants on the manifold, which
is called the Minkowski spacetime. The tangent vector for a particle traveling
with a constant velocity less than that of light through a point p in such a
spacetime is represented by a timelike vector at p. The particle must travel
within the future light cone at p, which satisfies the equation g (X, X ) = 0.
This defines the set of all null vectors at p representing the photon paths.
Now, according to special relativity, no material particles and signals travel
at a velocity more than that of light. Thus, the metric determines the causal
structure of spacetime in the sense that an event p is causally related to
another event q if and only if there is a timelike or null signal between p and
q . All such events lie on or within the double cone at p, defined by the metric
tensor as above.
For the spacetime continuum of general relativity that is non-flat, the
metric coefficients are functions of spacetime coordinates and one has to
solve for the metric as a solution to the Einstein field equations. As for the
24 The spacetime manifold

existence of a Lorentz metric on a spacetime, any C r -paracompact manifold

admits a C r−1 Lorentz metric if and only if it admits a non-vanishing C r−1
line element field (an assignment of a pair of equal and opposite vectors
(V , −V ) globally on M ) at each of its points (see for example, Hawking
and Ellis, 1973). Such a field is always defined for a non-compact manifold
and hence a Lorentz metric always exists.
Let (M, g ) be a spacetime and γ be a C 1 -curve in M . Then γ is a timelike,
null, or spacelike curve if the tangent vector is timelike, null, or spacelike
respectively at all its points. A timelike or null curve is also sometimes called
a non-spacelike curve. The tangent space magnitudes defined by g ,
1/2
X → | g (X, X ) | , (2.55)

are related to the distances on the manifold as below. If X is the tangent

vector along γ such that g (X, X ) has the same sign at all its points, then
the arc length between p = γ (t1 ) and q = γ (t2 ) along the curve is given by
 b
1/2
L(γ ) = s = (| g (X, X ) |) dt. (2.56)
a

The equations (2.56) and (2.55) are equivalent to the expression ds2 =
gij dxi dxj , which represents the infinitesimal arc length along γ .

2.3 Connection
The notion of parallel transport of a given vector X in Euclidian spaces can
be defined by requiring that, in going from a point p to q , both the magnitude
and direction of X must not change. If both these for the tangent vector
remain unchanged along a curve, it is called a straight line, along which
the tangent is parallel transported. In Euclidian space, if a vector is parallel
transported from points p to q along two different curves, the result is the
same, independent of the path taken. However, this is not necessarily the
case for a general affine manifold. For a differentiable manifold, the notion
of the parallel transport of vectors is defined by introducing a connection on
M.
Let X be a vector field on M , the derivative operator ∇X on M then gives
the rate of change of vectors or tensor fields along X at all p ∈ M . If Y is
another vector field at p, then the operator ∇X maps Y into a new vector
field Y → ∇X Y such that the following are satisfied:

(1) ∇X (αY + βZ ) = α∇X Y + β∇X Z for all α, β ∈ R;

(2) ∇f X+gY Z = f ∇X Z + g∇Y Z ;
(3) ∇X ( f Y ) = f ∇X Y + Y X ( f ).
 2.3 Connection 25

A connection ∇ at a point p ∈ M is a rule that assigns, to each vector field

X at p, a differential operator ∇X that maps an arbitrary C r vector field
Y at p into a vector field ∇X Y , such that (1), (2), and (3) are satisfied (for
a further discussion, see Hicks, 1965). The covariant derivative of Y , ∇Y , is
defined as a type (1, 1) tensor field that gives a vector ∇X Y when contracted
with the vector X . In such a case, condition (3) above implies

∇( f Y ) = df ⊗ Y + f ∇Y . (2.57)

A C r connection ∇ on a C k manifold (k ≥ r + 2) is a rule assigning a

connection ∇ to each p ∈ M such that if Y is a C r+1 vector field, then ∇Y
is a C r tensor field of type (1,1),

∇Y = Y i ; j e j ⊗ ei . (2.58)

Here, Y i ; j is often called the covariant derivative of the vector Y i , completely

defined by the n3 connection coefficients Γijk , given by choosing the vector
fields X and Y to be the basis vector fields

∇ej ek ≡ Γijk ei . (2.59)

It is easy to see that this is equivalent to the condition

ei , ∇ej ek = Γi jk . (2.60)

Therefore, in a coordinate basis,

  
∂
dxi , ∇∂/∂x j = Γijk . (2.61)
∂xk

Consider now the vector ∇X Y . Defining

∇∂/∂xi Y ≡ ∇i Y , (2.62)

using the rules defining the connection given above, and the relation

∂ ∂f
X( f ) = Xi i
( f ) = Xi i , (2.63)
∂x ∂x
one obtains   
∂Y k ∂
∇X Y = X i
+ Γk ij Y j . (2.64)
∂xi ∂xk
Comparing this with (2.61)
 
∂
∇X Y = Y k
;i X
i
, (2.65)
∂xk
26 The spacetime manifold

where
∂Y k
Y k ;i ≡ + Γk ij Y j . (2.66)
∂xi
It can be seen that the components of the vector ∇X Y are given as Y k ;i X i ,
and
∂Y i
Y i, j ≡ (2.67)
∂xj
can be defined.


Then, taking the transformation of the coordinates {xi } → {xi } when
the basis vectors transform as ei → ei
, it can be seen that Y i , j does not
transform like the components of a tensor. Similarly, consider the connection
coefficients in the new coordinate system,



Γk i
j
= ek , ∇ei
ej
. (2.68)

Transforming the dashed vectors to the original coordinate system and using
the conditions (2) and (3) above gives, in a coordinate basis,

 

∂xk ∂xi ∂xj k ∂ 2 xk
Γk i
j
=

Γ ij + . (2.69)
∂xk ∂xi ∂xj ∂xi
∂xj

It follows that, because of the presence of the second derivative terms above,
the coefficients Γi jk also do not transform like the components of a tensor.
Consider, however,
   
∂ i
j
∂
∇X Y = (Y i
;jX
j
) = (Y ; j
X ) , (2.70)
∂xi ∂xi

which implies


i ∂xj ∂xi j
j i

Y ; j X = Y ; j
j X . (2.71)
∂x ∂xi

Since the above is true for an arbitrary vector X j , Y i ; j are components of a
tensor.
Further, if Γi jk and Γi¯jk are components of two different connections on M ,
then using the coordinate transformations, it can be seen that the quantities

C i jk = Γi¯jk − Γi jk (2.72)

are components of a tensor.

For a connection ∇ on M , the torsion tensor T is defined by

T (X, Y ) = ∇X Y − ∇Y X − [X, Y ]. (2.73)

 2.3 Connection 27

Writing the components,

T (X, Y ) = (Γi jk − Γi kj )X j Y k ei . (2.74)

This is a type (1, 2) tensor with components

T i jk = Γi jk − Γi kj . (2.75)

A connection is called symmetric when the torsion tensor vanishes,

Γi jk = Γi kj , (2.76)

or [X, Y ] = ∇X Y − ∇Y X . Symmetric connections are always worked with

in this book, and the torsion tensor is assumed to be vanishing.
The notion of connection is generalized to arbitrary tensor fields to obtain
a tensor ∇X T of type (r, s) for any given tensor T of type (r, s) by assuming
first that ∇ is linear and obeys the Leibnitz rule. That is,

∇X (αS + βT ) = α∇X S + β∇X T, α, β ∈ R, (2.77)

and
∇X (S ⊗ T ) = ∇X S ⊗ T + S ⊗ ∇X T (2.78)
for any vector field X and tensor fields S and T . Furthermore, ∇ must agree
with the usual notion of a directional derivative, that is,

∂f
∇X f = df, X = Xf = X i . (2.79)
∂xi
Finally, ∇ must commute with contractions, that is,

(∇a T )i1 ...l...ir j1 ...l...js = ∇a T i1 ...l...ir j1 ...l...js . (2.80)

As shown earlier,

∇X T = T i1 ...ir j1 ...js ;a X a ei1 ⊗ · · · ⊗ eir ⊗ e j1 ⊗ · · · ⊗ e js , (2.81)

with
∇X T i1 ...ir j1 ...js = T i1 ...ir j1 ...js ;a X a . (2.82)
Now, by considering the expansion for ∇i (ej ⊗ ek ) it is seen that

∇a ei = −Γi ac ec , (2.83)

and if ω is a one-form then

∇ej ω = ωk; j ek , (2.84)

28 The spacetime manifold

with
∂ωk
ωk; j ≡
− Γi jk ωi . (2.85)
∂xj
In general, the covariant derivative of a tensor T can be written as

∂T i1 ...ir j1 ...js im
T i1 ...ir j1 ...js ;a = + Γ he T i1 ...e...ir j1 ...js
∂xa m

− e
Γ hjn T i1 ...ir
j1 ...e...js . (2.86)
n

Given a Lorentzian metric on M , the condition ∇X g = 0 defines a unique

torsion-free connection on M . Then,

(∇X g )ij = gij;k X k = 0, (2.87)

which implies that

gij;k = 0. (2.88)
The parallel transport of vectors preserve the scalar product defined by the
metric tensor g , and the connection coefficients Γi jk are determined in terms
of the first derivatives of the metric components. Since all the information on
spacetime structure is supposed to be contained in the ten metric functions
gij , this is to be expected. One way to see this is the following. Using (2.86),
the covariant derivative of the metric can be written as
∂gij
gij;k = − Γm il gmj − Γm jl gmi . (2.89)
∂xk
Now, using the condition gij;k = 0 and defining

gmj Γm il = Γjil , (2.90)

the above equation can be written as

∂gij
gij,k ≡ = Γjil + Γijl . (2.91)
∂xk
Using the above equations and the symmetry property of the connection,

Γjil = 12 (gji,m − gmj,i + gim, j ). (2.92)

This can also be seen by specializing to the frame of free fall. In such a
case, all the connection coefficients vanish and the metric is locally that of
special relativity. Then, gij = ηij and the partial derivatives of gij vanish.
Therefore, from the above equation for gij;k , one obtains gij;k = 0. As this
is a tensor equation, it must hold in all frames in general, and one can again
proceed as earlier.
 2.4 Non-spacelike geodesics 29

2.4 Non-spacelike geodesics

In Euclidian spaces, the line of shortest distance between any two points is
the straight line joining them, along which the tangent does not change in
direction or magnitude. That is, the tangent is parallel transported. Now,
let γ (t) : R → M be a C 1 -curve in M . If T is a C r (r ≥ 0) tensor field on
M , then the covariant derivative of T along γ (t) is defined as

DT
= T i...l k...m;h X h , (2.93)
∂t
where X is the tangent to γ . Then, γ is called a geodesic if its tangent vector
is parallel transported along it. That is, if X is the tangent vector field along
γ , then ∇X X is proportional to X . Therefore, there exists a function f such
that
∇X X = f X. (2.94)
Writing the components, this implies (X i ; j X j )ei = f X i ei always holds,
and so X i ; j X j = f X i along the geodesic curve. But, it is always possible
to reduce f to zero by a suitable choice of the curve parameter t along γ , so
the equation for the geodesic is written as

X i ; j X j = 0, (2.95)

where the X i values are components of the tangent vector to the geodesic.
The parameter t is called an affine parameter along γ , which is an
affinely parametrized geodesic. If {xi } denotes a local coordinate system,
the components X i are written as X i = dxi /dt and the equation for
geodesics is
d2 xi j
i dx dx
k
+ Γ jk = 0. (2.96)
dt2 dt dt
The affine parameter along the geodesic is determined up to an additive and
multiplicative constant. Thus, if t is an affine parameter, then so is t = at + b
and X i ; j X j = 0. Here b = 0 gives a new choice of the initial point γ (0) and
a= 0 implies a renormalization of the vector X .
A geodesic in (M, g ) is called timelike, spacelike, or null if its tangent
vector is timelike, spacelike, or null respectively. Here, the timelike or null
geodesics that represent the paths of particles or photons in the spacetime
are mainly considered. Since the tangent to a geodesic is parallel transported,
a timelike or null geodesic remains the same always and it cannot become
spacelike. In a Riemannian manifold with a positive definite metric, such
geodesics give the curves of shortest distance between its points. However, in
a spacetime with a Lorentzian metric, the non-spacelike geodesics maximize
the distance between the points, as defined by (2.56). If there is a timelike
geodesic between the points p and q , there is no shortest distance geodesic
30 The spacetime manifold

between them because, by introducing null geodesic pieces, one could always
join these points by curves of arbitrary small lengths. On the other hand,
any maximal length curve between p and q must necessarily be a timelike
geodesic.
The geodesic equations above are n equations in n variables xi with i =
1, . . . , n. Thus, the existence theorems for differential equations ensure that,
given xi and dxi /dt, that is, given any initial point p and the value of the
tangent vector X i , a unique geodesic through p with this value of tangent
exists. This can be used to define the exponential map Ep : Tp → M from
the tangent space at p into the spacetime. Under this map, any given tangent
vector X i in Tp is mapped to a point in M , a unit affine parameter distance
away along the unique geodesic determined by p and X i . It is clear that the
exponential map may not be defined on all of Tp because all the geodesics in
M passing through p may not extend to all the values of the affine parameter.
In such a case, M is called geodesically incomplete. On the other hand,
if the exponential map is defined on all of Tp for all points p, then M is
called geodesically complete. Then, all geodesics in M extend for all values
of their affine parameter. Also, the map Ep may not be one–one, because
the geodesics might cross each other. However, it can be shown (Bishop and
Critendon, 1964) that for a sufficiently small neighborhood Np of p there
is a neighborhood of the origin in Tp which is diffeomorphically mapped
onto Np by the exponential map which is one–one and well-defined on this
neighborhood. In such a case, the exponential map can be used to define
the normal coordinates on the neighborhood Np of p. Since Tp is an n-
dimensional vector space equivalent to Rn , the coordinates of any r ∈ Np can
be chosen to be the n-coordinates of the vector X p that is mapped onto it.
This coordinate system has the property that the geodesics are mapped into
straight lines and the connection coefficients vanish at p. Therefore, this
coordinate system turns out to be quite convenient for calculations at the
point p. The neighborhood Np can have a further property that any two
points in it can be joined by a unique geodesic contained totally within Np .
Such a neighborhood of p is called a convex normal neighborhood.
The geodesic equations are derived from the parallel transport required for
the tangent vector. If one requires that the curve must extremize the length,
namely that δl = 0, and works out
 b

dxa dxb
1/2
δl = δ

gab dt dt
dt = 0, (2.97)
a

using the variational methods, it turns out that the resulting equations
are precisely the geodesic equations (2.96). So, the geodesics extremise the
lengths of curves between any two spacetime points. If the events p and q
are timelike related, and if there is a maximum length timelike curve from p
 2.4 Non-spacelike geodesics 31

to q , that curve must be a timelike geodesic. Therefore, in many situations,

the useful way to work out geodesic equations is to choose the Lagrangian
as

L = 12 gab ẋa ẋb , (2.98)

and to write the Lagrange equations, which are the equations of the space-
time geodesics. Then, by comparison with the geodesic equations, it is also
possible to evaluate the quantities Γi jk for the spacetime.
In the Minkowski spacetime, the surface t = 0 is a three-dimensional sur-
face with the time direction always normal to it. Any other surface of
t = const. is also a spacelike surface in the same way. In general, let S
be an (n − 1)-dimensional manifold. If there is a C ∞ -map φ : S → M that
is locally one–one (there is a neighborhood N for every p ∈ S such that φ
restricted to N is one–one) and φ−1 is also C ∞ as defined on φ(N ), then
φ(S ) is called an immersed submanifold of M . If φ is globally one–one, then
φ(S ) is called an embedded submanifold of M . It may also be required that
φ be a homeomorphism with the induced topology on φ(S ) from M . Lower
dimensional embedded submanifolds in M represent well-behaved surfaces
in the spacetime.
A hypersurface S of any n-dimensional manifold M is defined as an (n− 1)-
dimensional embedded submanifold of M , and Vp is denoted by the (n − 1)-
dimensional subspace of Tp of the vectors tangent to S at any p ∈ S . It
follows that a vector n ∈ Tp exists that is unique up to the scale, and that
is orthogonal to all the vectors in Vp . This is called the normal to S at p. If
the magnitude of n is either positive or negative at all points of S without
changing the sign, then n could be normalized so that gab na nb = ±1. If
gab na nb = −1, then the normal vector is timelike everywhere and S is called
a spacelike hypersurface. If the normal is spacelike everywhere with a positive
magnitude, S is a timelike hypersurface, and S is a null hypersurface if the
normal na is null at S .
The timelike geodesics could be used to define the synchronous coordinate
system in the neighborhood of a spacelike hypersurface in the spacetime as
below. Let ∇a be the metric connection that satisfies ∇a gbc = gbc;a = 0. Let
S be a spacelike hypersurface, then, for every p ∈ S , let γ be the unique
timelike geodesic with a tangent na , that is, the congruence of these curves
at points of S is orthogonal to S . Then, in the neighborhood of that portion
of S , the coordinates q → (x1 , . . . , xn−1 , t) are assigned for any point q in the
future of p along γ , where t is the parameter along γ and x1 , . . . , xn−1 are
the spatial coordinates of p. In particular, if the geodesics in the congruence
are parametrized by the proper time t with the magnitude of the tangent
given by −1 along γ , then the spacelike surfaces are given as {t = const.}
surfaces. Note that S can be labeled as the {t = 0} spacelike surface.
32 The spacetime manifold

The synchronous coordinates have the important property that when the
congruence {γ} is orthogonal to S0 , it will also be orthogonal to subsequent
surfaces St given by t = const. Clearly, {γ} is orthogonal to S0 by construc-
tion. To see that this holds for any St for t within the domain of construction,
let X a be any basis vector for the tangent space at a point of St . Then,

nb ∇b (na X a ) = (nb ∇b na )X a + nb na ∇b X a . (2.99)

Since nb is a tangent to the geodesic, the above equals na nb ∇b X a . But, X

and n are coordinate vectors, implying ∇n X = ∇X n, that is,

na ∇ a X b = X a ∇ a nb . (2.100)

This implies

nb ∇b (na X a ) = na X b ∇b na = 12 X b ∇b (na na ) = 0, (2.101)

because na na = −1. Therefore, na X a = 0 in the future of S0 in the domain

of the validity of the synchronous coordinate system.

2.5 Spacetime curvature

The curvature for a spacetime is measured by the non-commutation of the
tangent vectors when these are parallel transported along different curves
to arrive at the same spacetime point. This is measured by the Riemann
curvature tensor, which is defined as a type (1, 3) tensor, R : Tp∗ × Tp × Tp ×
Tp → R. In a coordinate basis, the Riemann tensor can be written as
i
R = Rjkl ei ⊗ ej ⊗ ek ⊗ el . (2.102)

If the vector R(X, Y )Z is defined as

R(X, Y )Z ≡ ∇X (∇Y Z ) − ∇Y (∇X Z ) − ∇[X,Y ] Z, (2.103)

then the components of the Riemann tensor are given by

Ri jkl = ei , R(ek , el )ej . (2.104)

Working out the components gives

∂
R(X, Y )Z = Ri jkl X kY lZ j . (2.105)
∂xi
Now, in order to evaluate (2.104), note that

[∇X (∇Y Z )]i = ∇X (Z i ; j Y j ) = Z i ; jk Y j X k + Z i ; j Y j ;k X k . (2.106)

 2.5 Spacetime curvature 33

Similarly,
[∇Y (∇X Z )]i = Z i ; jk X j Y k + Z i ; j X j ;k Y k . (2.107)
Finally,

−∇[X,Y ] Z = −∇(Y i ; j X j −X i ; j Y j )(∂/∂xi ) Z

= −Z k l Y l ; j X j + Z k l X l ; j Y j . (2.108)

Combining the above equations,

R(X, Y )Z = (Z i ;lk − Z i ;kl )X k Y l . (2.109)

Comparing (2.109) and (2.105),

Z i ;lk − Z i ;kl = Ri jkl Z j , (2.110)

which is the same as

∇k ∇l Z i − ∇l ∇k Z i = Ri jkl Z j . (2.111)

The above equation could also be taken as the defining equation for
the components of the curvature tensor. As shown by the left-hand side
of (2.111), the Riemann curvature tensor provides the measure of non-
commutation of a tangent vector when parallel transported along different
curves to arrive at the same spacetime point.
In place of the vectors X ,Y , and Z the basis vectors ei can now be chosen.
Then,

∇ej ∇ek el = ∇ej (Γa kl ea )

= ej (Γa kl )ea + Γa kl Γh ja eh . (2.112)

Consider the definition of the components of the Riemann tensor as given by

(2.111). In particular, if a coordinate basis is chosen, then [ei , ej ] = 0 and

Ri jkl = ei , ∇ek ∇el ej − ei , ∇el ∇ek ej . (2.113)

Then, using (2.61) and a coordinate basis, the coordinate components of

the Riemann curvature tensor can be given in terms of the coordinate
components of the connection as

∂ Γi lj ∂ Γi kj
Ri jkl = − + Γi ka Γa lj − Γi la Γa kj . (2.114)
∂xk ∂xl
As pointed out earlier, given the metric tensor g on M , there is a unique,
torsion-free connection on M defined by the condition ∇X g = 0, which is
equivalent to the vanishing covariant derivative of the metric tensor, gij;k = 0.
34 The spacetime manifold

Then, parallel transport of vectors preserves the scalar product defined by

g and g (V , V ) = const. along a geodesic γ , where V is the tangent to γ .
Then,

∇X (g (Y , Z )) = Xg (Y , Z )
= ∇X (gij Y i Z j )
= g (∇X Y , Z ) + g (∇X Z, Y ). (2.115)

Evaluating Y (g (Z, X )) and Z (g (X, Y )) and adding the first and subtract-
ing the second from (2.115) gives

g (Z, ∇X Y ) = 12 [Xg (Y , Z ) + Y g (Z, X ) − Zg (X, Y ) + g (Y , [Z, X ])

+ g (Z, [X, Y ]) − g (X, [Y , Z ])]. (2.116)

Choosing the basis vectors ei in place of the vectors X , Y , and Z in (2.116)

gives the connection coefficients in terms of the derivatives of gij and the Lie
derivatives of the basis vectors

g (ei , ∇ej ek ) = gim Γm jk = Γijk . (2.117)

Choosing a coordinate basis with [ei , ej ] = 0 gives the usual Christoffel

symbols
 
1 ∂gij ∂gik ∂gjk
Γijk = + − . (2.118)
2 ∂xk ∂xj ∂xi
It follows then that the Riemann tensor components are expressed in terms
of the metric tensor and its second derivatives when the connection defined
by the metric is used. From now on, ‘the connection’ means this unique
connection as defined by the metric tensor.
The expression, as given by (2.114) and earlier definitions, implies that
the Riemann tensor has the symmetry given by

Ri jkl = −Ri jlk , (2.119)

which is equivalent to Ri j(kl) = 0. Furthermore, the curvature tensor obeys

the cyclic identity Ri [ jkl] = 0, which can be written as

Ri jkl + Ri klj + Ri ljk = 0. (2.120)

The covariant derivatives of the Riemann tensor satisfy the Bianchi identities
given by Ri j[kl;a] = 0, which is the same as

Ri jkl;a + Ri jla;k + Ri jak;l = 0. (2.121)

 2.5 Spacetime curvature 35

A straightforward proof would involve writing down each term above explic-
itly, substituting from (2.114), and then taking a summation. There are
certain additional symmetries that are valid when the connection is the one
induced by the metric. In this case,

Γijk = gil Γl jk , Rijkl = gia Ra jkl , Γl jk = g li Γijk . (2.122)

The Riemann tensor Rijkl defined by the metric has the symmetry

Rijkl = −Rjikl , (2.123)

which means R(ij)kl = 0. Also, in this case the Riemann tensor is symmetric
in the pairs of the first two and last two indices,

Rijkl = Rklij . (2.124)

The spacetime (M, g ) is said to have a flat connection if and only if

Ri jkl = 0, that is, all the components of the Riemann tensor must be vanish-
ing. This is the necessary and sufficient condition for a vector at a point p
to remain unaltered after parallel transport along an arbitrary closed curve
through p. This is subject to the condition that all such curves can be shrunk
to zero, or the spacetime has to be simply connected. In general, the parallel
transport of vectors does not hold in a spacetime manifold in the sense that
given a connection, if a given vector is parallel transported along two differ-
ent spacetime curves to arrive at the same point, the resultant vector will
be different in general. However, when all the components of the Riemann
tensor vanish, it can be shown that whenever a vector is transported from
one point to the other in the spacetime, the result is independent of the
path taken. In such a case, the connection is also said to be integrable and
a necessary and sufficient condition for this to happen is the vanishing of all
the components of the Riemann tensor.
When a symmetric connection is integrable, the manifold is called flat.
Furthermore, in the case of the connection being the metric connection, the
vanishing of all the Riemann tensor components provides a necessary and
sufficient condition for the spacetime metric to be flat, that is, a global
coordinate system in M exists such that the metric reduces to the diagonal
form with values ±1 everywhere.
The Ricci tensor is defined as a type (0, 2) tensor, obtained by contracting
the Riemann tensor,
Rjl = Ri jil . (2.125)
As a consequence of the symmetries discussed above, it follows that the Ricci
tensor is symmetric, and
Ri ikl = 0. (2.126)
36 The spacetime manifold

A further contraction of the Ricci tensor gives the curvature scalar R,

which is defined as
R = g ij Rij . (2.127)
The quantity R has the property that it depends only on the values of gij
and on their derivatives only up to the second order. Furthermore, it is
linear in the second derivatives of the metric components. The total number
of independent scalars that could be constructed from the metric and its
derivatives up to the second order is 14.
As a consequence of the various symmetries described above, the total
number of independent components of Rijkl reduces to 20 when the dimen-
sion of the manifold is four. When the dimension is three, Rijkl has six
independent components essentially given by Rij , and when the dimension
is two, there is only one independent component, which is R.
Another important tensor that can be constructed from Rijkl is the Weyl
tensor, which is also sometimes called the Weyl conformal tensor,
 
Cijkl = Rijkl + gi[l Rk] j + gj[k Rl]i + 13 Rgi[k gl] j . (2.128)

The symmetry properties of the Weyl tensor follow from those of the Riemann
tensor discussed above; it possesses the same symmetries as the Riemann
tensor. Also, it can be verified that the following identically vanishes,

g ik Cijkl = 0. (2.129)

The Weyl tensor is that part of the curvature tensor for which all contractions
vanish for any pair of contracted indices,

C i jil = 0. (2.130)

If the Weyl tensor vanishes throughout the spacetime with Cijkl = 0 at all
points, then it can be shown that the metric gij must be conformally flat.
This means that a conformal function Ω(xi ), 0 < Ω < ∞, exists such that

gij = Ω2 ηij , (2.131)

where ηij is the flat Minkowskian metric. The Weyl tensor is conformally
invariant in the sense that under a conformal transformation gij → g ij =
Ω2 gij ,
i i
Cjkl = Cjkl . (2.132)
It is possible to show that a necessary and sufficient condition for the space-
time metric to be conformally flat is that the Weyl tensor must vanish
everywhere.
 2.5 Spacetime curvature 37

The geodesic deviation equation, which is also called the Jacobi equation,
is now derived. This characterizes the coming together, or moving away,
of spacetime geodesics from each other as a result of the spacetime cur-
vature. Consider a smooth one-parameter family of affinely parametrized
non-spacelike geodesics, characterized by the parameters (t, v ), where t is
the affine parameter along a geodesic and v = const. characterizes different
geodesics in the family with t, v ∈ R. Such non-spacelike geodesics span a
two-dimensional submanifold on which t and v could be chosen as coordi-
nates. The vectors T = ∂/∂t and V = ∂/∂v are then coordinate vectors for
which [T , V ] = 0. Then, since the torsion tensor is vanishing,

∇T V = ∇V T , (2.133)

which implies T i ∇i V = V i ∇i T . Because T is a tangent to the geodesics,

T i ∇i T j = 0. Now, define the operator D by D ≡ T i ∇i . Then,

DV j = V i ∇i T j . (2.134)

Taking another derivative,

D2 V j = DV i ∇i T j + V i D(∇i T j )
= (T k ∇k V i )(∇i T j ) + V i T l ∇l ∇i T j . (2.135)

However, by the definition of the Riemann curvature tensor,

∇l ∇i T j − ∇i ∇l T j = Rj kli T k . (2.136)

Substituting this into (2.135),

D2 V j = (V k ∇k T i )(∇i T j ) + ∇i ∇l T j V i T l + Rj kli T k V i T l
= V k ((∇k T i )(∇i T j ) + (∇k ∇l T j )T l ) + Rj kli T k V i T l
= V k (∇k (T i ∇i T j )) + Rj kli T k V i T l
= Rj kli T k V i T l . (2.137)

The equation
D2 V j = −Rj kil T k V i T l (2.138)

is called the Jacobi equation, or the equation of geodesic deviation. It is

clear from the above that D2 V j = 0 if and only if all the components of
the Riemann tensor vanish. On the other hand, whenever some components
are non-zero, then the neighboring non-spacelike geodesics will necessarily
accelerate towards or away from each other.
38 The spacetime manifold

2.6 The Einstein equations

General relativity is a theory of gravity defined on a spacetime manifold,
where the force of gravitation is described in terms of the curvature of the
spacetime. These curvatures are in turn generated by the matter fields exist-
ing in the spacetime, as governed by the Einstein equations. The Einstein
equations involve second derivatives of the metric tensor. So, it is assumed
that the metric components are at least C 2 -differentiable functions of the
coordinates. All pairs (M  , g  ) that are diffeomorphic to (M, g ) are regarded
as equivalent and (M, g ), which represents this entire equivalence class of
spacetimes with equivalent physical properties, is studied.
The local causality and local conservation of the energy and momentum
are accepted as the basic physical postulates for the spacetimes in general rel-
ativity. The basic mathematical criterion while formulating general relativity
has been that the matter distribution determines the geometry of the space-
time universe in terms of the Riemann curvature tensor. Next, the motion
of any test particles in such a gravitational field is always independent of its
own mass and composition. This is the principle of equivalence, which has
now been verified to a great degree of accuracy to show that any two objects
with different masses and different compositions always arrive at the same
time on the surface of the earth when they have left from the same height. A
logical consequence of this is that any reference frame uniformly accelerated
with respect to an inertial frame of the special relativity is locally identical
to a frame at rest in a gravitational field. Finally, in general relativity, one
postulates the principle of general covariance, namely that all physical laws
are expressed as tensor equations so that they are valid in a general frame
of reference, and are invariant under arbitrary coordinate transformations.
When restricted to the frame of free fall, these must produce the laws of
special relativistic physics.
All matter fields on the spacetime, such as electromagnetic fields, dust,
perfect fluids or scalar fields, are assumed to be represented by a second
rank tensor T ij , called the energy–momentum tensor, in the sense that T ij
vanishes on any open region in the spacetime if and only if all the matter
fields vanished there. Such matter fields then obey tensor equations on the
spacetime and the derivatives involved will only be the covariant derivatives
with respect to the unique connection defined by the metric tensor. This
is because for any other connection defined on M , its difference from the
metric connection, which is a tensor again as shown earlier, could always be
regarded as another physical field on M .
Such a stress–energy tensor T ij then describes the matter fields on the
manifold. For example, for dust, which is the matter distribution composed
of non-interacting material particles, the field is characterized by the proper
density ρ0 of the flow and the four velocities of the particles given by dxi /dτ ,
 2.6 The Einstein equations 39

where τ is the proper time along the timelike trajectory describing the par-
ticle worldline. The simplest second rank tensor constructed from these two
quantities is
T ij = ρ0 ui u j . (2.139)
The component T 00 of this energy–momentum tensor is
dx0 dx0
T 00 = ρ0 . (2.140)
dτ dτ
In a special relativistic frame, this can be interpreted as the relativistic energy
density of matter. It can also be shown that requiring this tensor to have
zero divergence in such a frame gives the conservation of the energy and
momentum.
A perfect fluid is characterized by an additional scalar quantity, the pres-
sure p = p(xi ). In the limit as the pressure vanishes, this must reduce to the
dust form of matter. Furthermore, one also demands the conservation laws
in a special relativistic frame, and that these should reduce to the classical
equations of continuity and the Navier–Stokes equations in the appropriate
limits. Then this energy–momentum tensor is given in a general frame as

T ij = (ρ + p)ui uj + pg ij , (2.141)

which can be taken as the definition of a perfect fluid in general relativity.

In general, the energy–momentum tensors of various fields can be
constructed by using a variational principle where there is a proposed
Lagrangian, and the change in action is considered due to the change in
the metric.
For an arbitrary frame, and for other matter fields such as the electro-
magnetic field, or a charged scalar field, the principle of local conservation
of energy and momentum states that

T ij ; j = 0. (2.142)

The above equation for the stress–energy tensor contains much information
on the matter fields in a spacetime. For example, if the spacetime contains
a Killing vector K i then the above equation could be integrated to give a
conservation law. The conserved vector in such a case is defined as P i =
T ij Kj , and P i ;i = 0 as a consequence of (2.142) and the Killing equation

Ki; j + Kj;i = 0. (2.143)

Then, the integration of P i ;i over a compact region implies that the total
flux over a closed surface of the energy–momentum is zero in the direction
of the Killing vector (Hawking and Ellis, 1973). Even when the spacetime
40 The spacetime manifold

does not admit a Killing vector, given any point p a Riemannian normal
coordinate system at p could be set up so that the metric components have
the Minkowskian values and the connection coefficients Γi jk vanish at p. A
small enough neighborhood of p could then be chosen so that the values of
gij and Γi jk differ by an arbitrarily small amount from the values at p. Using
this fact, it could be shown that isolated test particles should move along
timelike geodesics (Fock, 1939; Dixon, 1970).
Furthermore, all matter fields are assumed to obey the postulate of local
causality, that is, the equations governing the matter fields are such that for
any p ∈ M , there is an open neighborhood U of p in which a signal can be
sent between any two points of U if and only if there is a non-spacelike curve
joining these points. This principle is valid in special relativity and is also
accepted in general relativity.
The above principles effectively imply that it is the spacetime metric, and
the quantities derived from it, that must appear in the equations for phys-
ical quantities, and that these equations must reduce to the flat spacetime
equations when the metric is Minkowskian. This is the basic content of the
general theory of relativity, where the spacetime manifold is now allowed
to have topologies other than R4 , and the metric gij can be non-flat glob-
ally. In general relativity, the matter fields expressed by the stress–energy
tensor are related to the non-flat nature of the spacetime by means of the
Einstein equations, which are the basic equations satisfied by the spacetime
metric. In Einstein’s theory, one does not discuss the physical interaction
of matter fields in a fixed background metric prescribed in advance. Actu-
ally, the gij values are treated as dynamical variables that depend on the
matter content of the spacetime and that are to be solved from the Einstein
equations.
The Newtonian theory gives an important indicator towards obtaining this
relationship between the matter content and the spacetime geometry, where
the gravitational field is described by a potential φ. The tidal acceleration
between nearby particles is given in terms of the separation between them
and the second derivatives of φ. In a curved spacetime manifold, such tidal
accelerations are described by the Jacobi equation (2.138), in terms of the
Riemann curvature tensor. Furthermore, the Poisson equation

∇2 φ = 4πρ (2.144)

must be recovered in the Newtonian limit. Both in special and general theo-
ries of relativity, the matter content is described by the stress–energy tensor
Tij , and the mass–energy density ρ corresponds to the quantity Tij V i V j .
Therefore, each side of Poisson’s equation corresponds to the Riemann ten-
sor as expressed in the Jacobi equation and Tij V i V j respectively. Another
important indicator for the comparison is provided by the Bianchi identities.
 2.6 The Einstein equations 41

Contracting i with k in (2.121) gives

∇a Rjl + ∇i Ri jla − ∇l Rja = 0. (2.145)

Then, raising j and contracting with a,

∇a Ra l + ∇i Ri l − ∇l R = 0. (2.146)

This then gives 2∇i Ri l − ∇l R = 0. Therefore,

∇i Gij = 0, (2.147)

where Gil is the Einstein tensor defined by

Gil ≡ Ril − 12 gil R. (2.148)

From the above considerations, one can make a comparison given by

Rij = 4πTij as field equations for general relativity. This does not, however,
work because the contracted Bianchi identities imply ∇l R = 0, so the trace
T = const. throughout the spacetime. This is an unphysical restriction on
the matter content.
Using the above indicators, Einstein proposed the field equations

Gij = Rij − 12 Rgij = 8πTij . (2.149)

Then, the contracted Bianchi identities actually imply the local conservation
of energy and momentum through the Einstein equations. Taking the trace
of the above,
R = −8πT. (2.150)
Substituting this back in the above gives an alternative form of the Einstein
equations,
Rij = 8π (Tij − 12 T gij ). (2.151)
The definition of the Ricci tensor suggests that the Einstein equations
depend on the derivatives of the metric up to the second order. These equa-
tions are highly non-linear in gij , however, they are linear in the second
derivatives of gij . In fact, the quantities Rij and Rgij are the only second-
rank symmetric tensors that are linear in the second derivatives of the metric,
and involve only up to the second derivatives of gij . Actually, the Einstein
equations are a coupled system of non-linear second order partial differ-
ential equations for gij . This makes the task of solving these extremely
difficult. Several symmetry assumptions on the spacetime generally need
to be imposed in order to work out the metric components as a solution
to the Einstein equations. Some solutions that are useful in the context of
gravitational collapse are discussed later.
42 The spacetime manifold

Given the T ij values, the field equations may be viewed as a set of differ-
ential equations to determine the gravitational potential values, g ij , that fix
the resulting geometry. A particularly important case here is that of vacuum
solutions when T ij = 0. On the other hand, one could arbitrarily specify
the ten metric potentials and then compute the Einstein tensor Gij . Then,
the field equations determine the energy–momentum tensor Tij . However,
in that case, the resulting T ij values turn out to be unphysical most of the
time in that the energy conditions ensuring the positivity of mass–energy
density may be violated. Such a violation of the energy conditions is rejected
on physical grounds as all observed classical fields obey positivity of energy
density, which is closely connected with the physical features of gravitation
theory.
In general, the field equations are ten equations, connecting a total of
twenty quantities, which are the ten components of gij and the other ten
components of Tij . Therefore, the field equations are the conditions placing
constraints on the simultaneous choice of these twenty quantities. If part of
the gravitational potentials and the matter contents are determined from
physical conditions, then such conditions can fully determine the matter and
geometry. In particular, if the vacuum equations are considered,

Gij = Rij − 12 Rgij = 0, (2.152)

then there are ten equations to determine the ten quantities gij . However,
the Bianchi identities
∇j Gij = 0 (2.153)
place four differential constraints on these equations, which are not all
independent. Therefore, there is an indeterminacy in that there are fewer
equations than unknowns to be determined. Furthermore, there is an intrin-
sic gauge freedom available in the general theory of relativity that does not
allow a complete determination of the metric potentials. This is given by the
coordinate freedom that allows a transformation from one set of coordinates


xi to any other set of coordinates xi . However, this coordinate freedom
could be used to impose conditions on the metric components. For example,
choosing the normal coordinates gives g00 = 1 and g0α = 0, α = 1, 2, 3 in this
coordinate system. This leaves six other components to be determined from
the field equations. The issue is closely linked to the Cauchy problem in gen-
eral relativity where the basic problem is that, given initial data on a regular
spacelike hypersurface, one would like to determine its unique evolution in
the future or past.
The Einstein equations can admit a cosmological constant. Note that the
most general second rank tensor which can be constructed out of Gij and gij ,
which is divergence free and involves the derivatives of the metric tensor up
to second order only, is the linear combination Gij + Λgij (Lovelock, 1972),
 2.7 Exact solutions 43

where Λ is a constant. Therefore, addition of such a constant multiple of gij

to the Einstein tensor preserves all the required properties discussed above.
Einstein historically introduced the cosmological term Λ in the equations in
order to generate static cosmological solutions, and wrote the equations as

Rij − 12 Rgij + Λgij = 8πTij . (2.154)

It is seen that for an empty spacetime with Tij = 0, the Einstein equations
then become
Rij = −Λgij . (2.155)

If Λ = 0, then one does not obtain the Newtonian theory in the limit of
slow motions and weak fields. However, if the magnitude of Λ is very small,
then such departures will be quite negligible and approximate agreement
with the Newtonian theory is obtained.

2.7 Exact solutions

Owing to their complexity and the fact that they are a highly non-linear
system of partial differential equations, sufficiently general solutions of the
Einstein equations are difficult to find. Almost all known exact solutions
assume a high degree of symmetry, such as the spherical or axial symmetry,
the existence of various Killing vector fields on the spacetime, or similar other
conditions. These represent special and idealized situations. (For a detailed
discussion on exact solutions with various symmetries see Stephani et al.,
2003).
Such spacetime examples, however, do provide a good idea of what is pos-
sible within the framework of general relativity. They also give an indication
of certain important features that may be present in a situation that is much
more general than the actual solution being studied. An example of this is
the singularity theorems in general relativity. Some exact solutions, such as
the Schwarschild and Friedmann–Robertson–Walker models, exhibited the
occurrence of a spacetime singularity where physical quantities were blow-
ing up. Earlier, it was thought that this was a mathematical artifact due
to the symmetry of these solutions, and singularity would not be present in
sufficiently general models once the symmetries are relaxed. It turned out,
however, that the singularity was a fairly general feature of general relativity
and would occur under rather general physical assumptions, as shown by the
singularity theorems. Some of the solutions also have interesting applications.
For example, the Minkowski spacetime is both the geometry of special relativ-
ity and locally that of any general relativistic model, and the Schwarzschild
spacetime is used to test physical predictions of general relativity.
44 The spacetime manifold

Some of the most interesting and physically relevant solutions to the

Einstein equations are spherically symmetric; these include the Friedman–
Robertson–Walker models in cosmology, and the Schwarzschild solution that
can model an isolated star. In order to understand the possible final fate of
a gravitationally collapsing massive star, we discuss the spherical symmetry
in some detail. Such a symmetry represents a high degree of idealization of
the physical situation, but the advantage is that one can solve it many times
analytically to get exact results. It is also possible that several salient phys-
ical features of the situation are preserved when departures from spherical
symmetry are taken into account.
In fact, the basic motivation for the theory and the idea of a blackhole come
from the case of a spherically symmetric homogeneous dust cloud collapse.
Independently of the interior solution, the metric exterior to such a spherical
body must be the Schwarzschild spacetime, as implied by Birkhoff ’s theorem
(Birkhoff, 1923), which states that the only vacuum, spherically symmetric
gravity field must be static. The spherically symmetric and asymptotically
flat spacetimes such as the Schwarzschild geometry are useful to model the
spacetime outside the sun and stars, and could be used to obtain conclusions
relevant for the experimental verification of general relativity. Such solutions
can also possibly represent the outcome of a complete gravitational collapse
of a massive star. Similarly, the Vaidya geometry can model the field outside
a radiating star. The Schwarzschild and Vaidya geometries have useful appli-
cations, and are particularly relevant to the gravitational collapse scenarios
studied here. Some of the properties of these exact solutions of the Einstein
equations that are referred to in later chapters are reviewed below.

2.7.1 Minkowski spacetime

The Minkowski spacetime mathematically is the manifold M = R4 with the
Lorentzian metric

ds2 = −dt2 + dx2 + dy 2 + dz 2 , (2.156)

with −∞ < t, x, y, z < ∞ giving the range of coordinates. This is a flat

spacetime with all components of the Riemann tensor Ri jkl = 0, and hence
it is the simplest empty spacetime solution to the Einstein equations Gij =
8πTij = 0, which underlies the physics of special relativity. The vector ∂/∂t
provides a time orientation here. If the spherical polar coordinates (t, r, θ, φ)
given by x = r sin θ sin φ, y = r sin θ cos φ, and z = r cos θ are used, the
above metric becomes

ds2 = −dt2 + dr2 + r2 (dθ2 + sin2 θ dφ2 ). (2.157)

 2.7 Exact solutions 45

The range of the coordinates r, θ, φ are 0 < r < ∞, 0 < θ < π , and
0 < φ < 2π . Two such coordinate neighborhoods are needed to cover all
of the Minkowski spacetime.
In coordinates (t, x, y, z ) it is obvious that the geometry is flat, because
all the metric components are constants, so all the connection coefficients
are vanishing. In other coordinate systems, such as spherical coordinates
(t, r, θ, φ), the connection coefficients Γi jk do not all vanish (for example,
Γ1 22 = r); however, all the Riemann curvature tensor components do still
vanish.
The Lorentz transformations on the Minkowski spacetime are defined as
the set of metric preserving isometries that are linear and homogeneous.
Physically, these represent the change of reference frame from one to another
inertial observer, given by the coordinate change


xi → xi = Li j xj . (2.158)

In addition to the metric preserving property, the above implies det

Li j = ±1, so the matrix is non-singular. If det Li j = +1 and L0 0 ≥ 1,
then the Lorentz transformations preserve the orientations in both space
and time. The set of all Lorentz transformations, form a group where the
identity map is δ i j , and the inverse is the inverse matrix. The Lorentz group
is a subgroup of the Poincaré group of transformations, which are general
inhomogeneous mappings that leave the Minkowskian metric invariant. It
consists of a Lorentz transformation together with an arbitrary translation
in space and time. This is a ten-parameter group, consisting of six Lorentz
parameters and four translation parameters, and physically it is a mapping
of one inertial frame into another in a general position in the spacetime.
The geodesics of Minkowski spacetime are the straight lines of the under-
lying Euclidian geometry. Given an event in M , the lines at 45◦ to the time
axis through that event give null geodesics in M . They form the boundary
of the chronological future or past I ± ( p) of an event p, which is the set of all
events that lie in the future (past) of p, and which contain all possible time-
like particle trajectories through p including timelike geodesics. The causal
future J + ( p) is the closure of I + ( p) in Minkowski space, which includes
all the events in M that are either timelike or null related to p by means
of future directed non-spacelike curves from p. The causal past of an event
is dually defined. The spacelike hypersurfaces t = const. in the Minkowski
spacetime give a family of Cauchy surfaces that covers all of M . A Cauchy
surface is a spacelike hypersurface such that all inextendible non-spacelike
curves in M meet this surface once and only once. However, all spacelike
hypersurfaces in M need not be Cauchy surfaces. For example, the family
given by −t2 + x2 + y 2 + z 2 = A = const. with A < 0 are inextendible
spacelike surfaces that are not Cauchy surfaces. All these surfaces are fully
46 The spacetime manifold

contained inside the chronological past or the future of the origin and there
are timelike geodesics outside this past or future cone that do not meet any
of these surfaces.

2.7.2 The ideal points boundary

In general, the spacetime boundary consists of points at infinity and the
spacetime singularities. To understand global properties and the structure
of the infinity of the Minkowski spacetime, the procedure of Penrose (1968)
and the general procedure of Geroch, Kronheimer, and Penrose (1972) can
be used. An arbitrary event p in the Minkowski spacetime is uniquely deter-
mined either by its chronological future I + ( p) or past I − ( p). If a future
directed non-spacelike curve γ has a future end point at p, I − (γ ) = I − ( p).
(By definition, I − (γ ) is the union of all I − (q ) with q being any point on γ .)
On the other hand, if γ is a future inextendible curve without any future end
point, the set I − (γ ) determines a point at infinity of M . (A future or past
inextendible curve, in the context of Minkowski spacetime, is a trajectory
which goes off to infinity in the future or past without stopping anywhere.)
Two such curves γ1 and γ2 determine the same ideal point, or a point at
infinity, if I − (γ1 ) = I − (γ2 ).
Such a procedure defines the future ideal points. Past ideal points are
defined dually using past inextendible non-spacelike curves. In the case of
Minkowski spacetime, there are future directed inextendible curves γ that
are timelike and have the same past, which is the entire spacetime M , that
is, I − (γ ) = M . Hence, all such timelike curves determine a single future
ideal point i+ , called the future timelike infinity. The past timelike infinity
i− is similarly defined. If γ is chosen to be a future endless null geodesic or
a null curve, it is possible to have a situation where I − (γ ) is not the entire
Minkowski spacetime. Certain timelike curves also have this property. For
example, consider the past of the timelike hyperbola

t = sinh λ, x = cosh λ, y, z = 0, −∞ < λ < ∞. (2.159)

Then, I − (γ ) lies completely to the past of the null hypersurface x = t. It

can be shown in general (Geroch, Kronheimer, and Penrose, 1972) that, for
a non-spacelike curve γ , if I − (γ ) = M and if γ is future endless, then a null
hypersurface Sγ exists, the half-space below which coincides with I − (γ ).
If the collection of ideal points can be denoted by I + , then there is a
one–one and onto correspondence between the points of I + and such null
hypersurfaces. Any such null hypersurface is determined by the value of the
time t at which it intersects the time axis, and by the direction of the null
vector at the point of intersection. Since the set of all possible light ray
 2.7 Exact solutions 47

directions at any point is equivalent to the two-sphere S 2 , it follows that I +

is a three-dimensional manifold with topology S 2 × R.
The three-dimensional null hypersurfaces I + and I − are called the future
and past null infinities respectively for the Minkowski spacetime. A general
spacetime would also admit such a boundary construction under certain con-
ditions, such as being asymptotically flat and empty. It can be shown for the
Minkowski spacetime that all complete null hypersurfaces are flat and so are
like the surfaces {x = t}, in which case the topological structure of the null
infinity is clearly I + = S 2 × R. However, the topological structure of null
infinities for a general spacetime need not be the same.

2.7.3 Conformal compactification

A differential structure and a metric on the future null infinity I + can be
introduced. A convenient way to attach the ideal points boundary I + to M
is to use a suitable conformal factor Ω to obtain a transform of the original
spacetime metric ηij ,
gij = Ω2 ηij , Ω > 0. (2.160)

This leaves the causal structure of M invariant, because the null geodesic
paths of the metric ηij and the unphysical metric gij are the same under a
conformal mapping, up to a reparametrization. Therefore, the past of any
non-spacelike curve γ is unchanged and there is a natural correspondence
between ideal points in both the spacetimes. Since light cones are unaltered
by a conformal transformation, the boundary attachment obtained in this
manner is coordinate independent.
In the Minkowski spacetime, the advanced and retarded null coordinates
can be introduced, where

v = t + r, u = t − r, (2.161)

which gives a reference frame, based on null cones, which is most suitable to
analyze the radiation fields.
Under this transformation of coordinates, the metric becomes

ds2 = −du dv + 14 (u − v )2 (dθ2 + sin2 θ dφ2 ), (2.162)

with −∞ < v < ∞ and −∞ < u < ∞. Now, the information at future
null infinity corresponds to taking the limit as v → ∞, which amounts to
moving in the future along u = const. light cones. Similarly, past null infinity
corresponds to u → ∞. This procedure could be made precise in a coordinate
independent way. The Minkowski spacetime M can be compactified by means
48 The spacetime manifold

of a conformal transformation of the metric as given by

Ω2 = (1 + v 2 )−1 (1 + u2 )−1 , (2.163)

and then by adding the closure to add the null infinities. New coordinates p
and q can be introduced by

v = tan p, u = tan q. (2.164)

Then, the corresponding ranges for p and q are

π π π π
− <p< , − <q< , (2.165)
2 2 2 2

and the metric ḡij on the unphysical spacetime M , after the conformal
transformation, is given by

ds̄2 = −dp dq + sin2 ( p − q )(dθ2 + sin2 θ dφ2 ). (2.166)

It is possible to see now that the metric above, with the coordinate ranges as
given above, is a manifold embedded as a part of the Einstein static universe.
To see this, let
T = p + q, R = p − q, (2.167)
and then (2.166) becomes, in (T, R, θ, φ) coordinates,

ds̄2 = −dT 2 + dR2 + sin2 R(dθ2 + sin2 θ dφ2 ), (2.168)

with the coordinate ranges given by

−π < T + R < π, −π < T − R < π. (2.169)

This is precisely the natural Lorentz metric on S 3 × R, which is the Ein-

stein static universe, except that the coordinate ranges are now restricted by
(2.169). In this picture, the future null infinity I + is given by T = π − R
for 0 < R < π , and the past null infinity is given by T = −π + R for
0 < R < π . (For a further discussion see, for example, Wald, 1984.) The
structure of infinity for any spherically symmetric spacetime can be depicted
by a similar Penrose diagram. As mentioned above, I + here is topologically
S 2 × R.

2.7.4 Schwarzschild solution

The Schwarzschild solution gives the geometry exterior to a spherically sym-
metric massive body such as a star and has been used extensively to verify
experimentally the predictions of the general theory of relativity. This is an
 2.7 Exact solutions 49

empty exterior solution where the Ricci tensor vanishes and is matched at the
boundary to an interior solution inside the body. In (t, r, θ, φ) coordinates,
the metric can be given as
   −1
2m 2m
ds = − 1 −
2
dt + 1 −
2
dr2 + r2 dΩ2 , (2.170)
r r

where dΩ2 = dθ2 + sin2 θ dφ2 . The coordinate t is timelike and r, θ, φ are
spacelike coordinates. The radial coordinate r has the property that the
two-sphere given by t = const., r = const. has the two-metric given by

ds2 = r2 (dθ2 + sin2 θ dφ2 ), (2.171)

so the area of any such two-sphere would be 4πr2 .

The coordinate r is restricted to r > 2m because the above metric has
an apparent singularity at r = 2m. The coordinate t has the range −∞ <
t < ∞. The solution is obtained by solving the vacuum Einstein equations
for a spherically symmetric spacetime, where the quantity m is the constant
of integration, with its value determined by using the weak field Newtonian
limit of general relativity. If Φ is the Newtonian gravitational potential, then
in non-relativistic units,

2Φ 2GM
g00  1 + 2
=1− 2 , (2.172)
c c r

where G is the Newtonian constant of gravity, c is the velocity of light, and

M is the point mass at the origin which gives rise to the Newtonian poten-
tial Φ. This determines the constant of integration m in the Schwarzschild
solution as
GM
m= 2 . (2.173)
c
Therefore, this is interpreted as a solution describing the gravitational field
of a point particle with mass m (in relativistic units G = c = 1) situated at
the center.
The Schwarzschild metric is static in the sense that ∂/∂t is a timelike
Killing vector that is a gradient. The metric components gij here are inde-
pendent of time. Also, there are no mixed terms in the metric that involve
both space and time. Therefore, there is no rotation inherent in the space-
time and the metric is stationary. For a detailed discussion on stationary
and static solutions of the field equations, see Stephani et al. (2003). In
the present case, using the Birkhoff theorem, a spherically symmetric vac-
uum solution of Einstein equations must be necessarily static. An important
50 The spacetime manifold

implication of this theorem is that even when a spherically symmetric star

undergoes pulsations or changes in shape, while maintaining the spherical
symmetry, it cannot radiate any disturbances in the exterior, such as grav-
itational waves. Therefore, any spherically symmetric solution of Einstein
equations with Rij = 0 is necessarily the Schwarzschild solution. So, the
Schwarzschild exterior can be used to describe the outside metric for several
situations such as a spherically symmetric star that is either static, or that
undergoes radial pulsations, or a radial spherically symmetric gravitational
collapse.
The spherical symmetry of the Schwarzschild spacetime is seen in that the
metric components g00 and g11 are functions of r alone and not of θ and φ,
as implied by the angular part of the metric. Specifically, the isometry group
of M contains a subgroup which is isomorphic to the group SO(3), and the
orbits of this subgroup are two-dimensional spheres. These isometries are
interpreted as rotations and so the metric remains invariant under rotations
in general for any spherically symmetric spacetime. The parameter m serves
here as the source of the gravitational field, and setting m = 0 gives the flat
Minkowski spacetime. As pointed out above, the comparison with Newto-
nian theory shows that m is to be treated as the gravitational mass of the
body producing the field, as measured from infinity. The spacetime here is
asymptotically flat in the sense that as r tends to infinity, the flat spacetime
metric is recovered, and the gravitational field diminishes to zero.
Generally, the Schwarzschild metric is taken to represent the outside metric
for a star, with coordinate r > r0 , where r0 gives the boundary of the star.
The metric inside r < r0 is a different interior metric determined by the
matter distribution Tij inside the star, and it is matched at the boundary
r = r0 with the Schwarzschild solution. However, in the case of a complete
gravitational collapse, when all the mass collapses at r = 0, it is necessary to
consider the metric above as an empty spacetime solution for all the values
of r. Clearly, the metric has singularities at r = 0 and r = 2m, and hence it
represents only one of the patches 0 < r < 2m or 2m < r < ∞. If one confines
to the manifold given by the latter range of values of r, it is necessary to
determine if the spacetime is extendible, that is, if a bigger spacetime (M  , g  )
exists with M embedded in M  and g = g  on M . That this should be possible
is indicated by the fact that even though the form of the metric is singular at
r = 2m, the curvature scalars are all well-behaved here, and so this could be
merely a singularity due to an inappropriate choice of coordinates. A decision
on whether a given spacetime manifold is maximal or not can be made by
looking at the geodesics. In a maximal manifold, it would be required that
all the geodesics be extended in both the directions to an infinite value of
their affine parameter, or they must terminate at an intrinsic singularity of
the spacetime that is not removable. On the other hand, if this metric is
taken to describe the patch 0 < r < 2m of the spacetime, then it is seen
 2.7 Exact solutions 51

that as r tends to zero, the curvature scalar

m2
Rijkl Rijkl = (2.174)
r6
diverges, so the point r = 0 is a real spacetime singularity. It is not possible
to extend the spacetime across this singularity in a continuous manner. A
maximal extension of the Schwarzschild manifold, which covers both the
patches above, was obtained by Kruskal (1960) and Szekeres (1960).
The vacuum Schwarzschild geometry exterior to the collapsing matter
arises for a dust cloud collapse. In the case of homogeneous collapse, a
Schwarzschild blackhole is necessarily produced in a complete gravitational
collapse, and the event horizon fully covers the resulting spacetime singular-
ity of infinite curvature and density. In other words, as the collapse evolves,
the event horizon develops earlier than the formation of the singularity. This
situation is significant for the cosmic censorship hypothesis and blackhole for-
mation scenarios. The interior metric in this case is exactly that of a closed
Friedmann model in a time reversed sense.
In the extended Schwarzschild manifold, r = 2m is a null hypersurface
and each point on this surface is a two-sphere of area 16πm2 . Note that the
metric component g00 = (1 − 2m/r) is positive for r > 2m; however, g00 < 0
for r < 2m. Therefore, it is no longer possible to use t as a time coordinate in
that range, as the coordinates t and r reverse their roles there, and spacetime
is no longer static. Therefore, the r = 2m surface is also called the static
limit. The vector ∂/∂t with components ξ i = δ i 0 = (1, 0, 0, 0) gives the time
translation, leaving the gij unchanged as it does not involve the time coor-
dinate. Therefore, ξ is a Killing vector that leaves the spacetime geometry
unchanged. One obtains ξ 2 = gij ξ i ξ j = g00 , and for the Schwarzschild metric
ξ 2 vanishes on r = 2m. Hence, at the static limit the timelike Killing vector
becomes null.
The Schwarzschild geometry is an illustration of the basic principle that
Einstein used to formulate gravitation theory, namely that matter tells the
spacetime how to curve. To see this, consider the Schwarzschild solution in a
spacelike surface t = const. and in the equatorial plane θ = π/2. The metric
of this two-dimensional curved surface is given by

dr2
ds2 = + r2 dφ2 . (2.175)
(1 − 2m/r)

The geometry of such a curved surface can be visualized as embedded in the

ordinary Euclidian space. Here, the region 0 < r < rb is to be considered
as filled by the matter that represents the spherical star with a boundary at
r = rb , and the curved surface would then represent the geometry outside
such a star.
52 The spacetime manifold

Consider a static observer along a Killing direction, for whom the four-
velocities are ui = ξ i /|ξ|. Suppose a static source with four-velocity ui1 emits
a photon with four-momentum pi (so pi; j pj = 0 with a suitable parametriza-
tion) and is observed by a static observer with four-velocity ui2 . Now, take
the directional derivative of ξi pi along the geodesic tangent pi ,

(ξi pi ); j p j = ξi; j pi p j + ξi p i ; j p j = 0. (2.176)

The first term vanishes because ξ i is a Killing vector and the second term
vanishes because of the geodesic equation. Therefore, the ratio of energies
measured at these two points by static observers is given by
i  1/2
E1 u pi
= i 1 1/2 . (2.177)
E2 (u pi )2

Using ui = ξ i /|ξ| and ξ i pi = const. along the geodesic,

i  1/2
E1 ξ ξi 1
= 1/2
. (2.178)
E2 (ξ i ξi )2

Since ξ 2 = g00 , this is the gravitational red-shift formula for a static source
and observer in terms of the metric components. It is now seen that if the
observer remains at a finite radius but the source approaches r = 2m, the
red-shift tends to infinity. Therefore, as a particle falls into the blackhole,
approaching r = 2m, the light rays emitted by it are infinitely red-shifted as
observed by a distant static observer in the outside spacetime.

2.7.5 Homogeneous collapse and the blackhole

Consider a spherically symmetric massive star, collapsing gravitationally
when it has exhausted its internal nuclear fuel. In order to investigate the
final state for such a collapse, the dynamical interior for such a star needs to
be considered. As such, there is no unique interior solution available to repre-
sent this situation, which basically depends on the properties of the matter,
the equation of state obeyed, and the physical processes taking place within
the stellar interior. However, assuming the matter to be pressureless and
homogeneous dust allows the problem to be solved analytically, which pro-
vides many important insights (Datt, 1938; Oppenheimer and Snyder, 1939).
In that case, the energy–momentum tensor is given by T ij = ρui u j and the
Einstein equations for the spherically symmetric form of the metric given
above have to be solved. Solving the Einstein equations determines the met-
ric potentials completely and the interior geometry of the star, which is a
 2.7 Exact solutions 53

collapsing dust ball. This is the same line element as that of the closed
homogeneous and isotropic Friedmann models given by
 
dr2
ds = −dt + R (t)
2 2 2 2 2
+ r dΩ , (2.179)
1 − r2

where dΩ2 represents the metric on a two-sphere. The geometry outside

the star is a vacuum and is then the Schwarzschild spacetime as implied
by the Birkhoff theorem. It can be shown that the interior geometry of the
dust cloud is matched at the boundary of the star r = rb with the exterior
Schwarzschild spacetime.
When the collapse is complete, in the case of an empty and asymptoti-
cally flat exterior, the spacetime settles to a vacuum Schwarzschild geometry.
Apart from predicting small observable departures from the Newtonian grav-
ity, the Schwarzschild solution is important for the theory of gravitational
collapse, as sufficiently massive stars unable to support themselves against
the pull of their own gravity at the end of their life cycle could finally settle
to that geometry. The final state of a spherically symmetric homogeneous
dust collapse is necessarily a Schwarzschild configuration that contains a
spacetime singularity hidden within an event horizon. This gives rise to a
blackhole in the spacetime, which is a region from which no causal signals
can reach a faraway observer. This scenario forms the basis of much of the
theory and applications of blackhole physics.
If the star has internal pressures and is radiating, which would be a phys-
ically realistic scenario, then it can be matched to an exterior Vaidya, or a
generalized Vaidya solution, as is discussed later. To understand the vacuum
case, if m is the total mass of the star, then the structure of the resulting
configuration can be discussed by considering the radial null geodesics in this
metric, defined by ds2 = 0 and θ̇ = φ̇ = 0. Taking the positive sign solutions
since the interest here is in the outgoing null geodesics, these are given by

dt r
= . (2.180)
dr r − 2m

Integrating this gives

t = r + 2m ln | r − 2m | + const. (2.181)

In the region r > 2m we have dr/dt > 0 and hence r increases with increasing
t. Thus the above describes the congruence of outgoing radial null geodesics.
The in going null trajectories are given by the negative sign solutions, which
are given by
t = −r − 2m ln | r − 2m | + const. (2.182)
54 The spacetime manifold

In the region below r = 2m, the coordinates r and t change their space-
like and timelike nature and hence the light cones tip over. As a result, no
observer in the region r < 2m can remain at a constant value of r, but they
must move inwards to fall within the intrinsic curvature singularity at r = 0.
Each point in the (t, r) plane represents a two-sphere of area 4πr2 .
This gives an idea of the phenomena happening in the region r < 2m,
namely that any material particle or photon here must fall in the spacetime
singularity, and that it cannot escape to larger values of r to communicate
with external observers in the spacetime. Hence, this is called the black-
hole region in the spacetime. However, the Schwarzschild picture gives the
impression that from the outside r > 2m, no photons or particles could
fall in this blackhole and they will take infinite time to reach the surface
r = 2m. It turns out that this is actually a coordinate defect arising due
to the coordinate singularity at r = 2m, as can be seen by going to the
Eddington–Finkelstein coordinates, where the idea is to choose a new time
coordinate such that the in going null geodesics become straight lines in the
spacetime. It is clear from the Schwarzschild consideration above that the
appropriate change may be given by

t → t + 2m ln(r − 2m), (2.183)

for the r > 2m region. The solution with such a coordinate change is now
regular at r = 2m and the coordinate range is now 0 < r < ∞. This is
called an analytic extension of the Schwarzschild solution. A time-reversed
solution is obtained if a different time coordinate is introduced,

t → t − 2m ln(r − 2m), (2.184)

in which case the outgoing null geodesics are straight lines. A simpler way to
write the metric in the new coordinate system is to introduce the advanced
null coordinate v defined by

v = t + r + 2m ln(r − 2m). (2.185)

The metric then has the form

 
2m
ds = − 1 −
2
dv 2 + 2dv dr + r2 (dθ2 + sin2 θ dφ2 ). (2.186)
r
The radially infalling null geodesics are now given by v = const. It can be
seen that at r = 2m, the radially outgoing photons stay at a constant value
of r, and below this surface they must also fall to the singularity at r = 0.
Further, radially infalling material particles must also fall to the spacetime
singularity within a finite amount of proper time, as measured along their
trajectory.
 2.7 Exact solutions 55

Spacetime singularity
Event horizon
r=0
r = 2m
Apparent horizon

Boundary of the star

Initial surface

Fig. 2.3 Homogeneous dust cloud collapse. The trapped surfaces form when the
star enters a r = 2m radius. The event horizon forms prior to the singularity,
creating a blackhole as the collapse endstate.

The basic features of such a collapsing spherically symmetric homoge-

neous dust cloud configuration are summarized in Fig. 2.3. The gravitational
collapse initiates when the star surface is outside its Schwarzschild radius
r = 2m, and a light ray emitted from the surface of the star can escape to
infinity. However, once the star has collapsed below r = 2m, a blackhole, that
is a region of no escape, develops in the spacetime, which is bounded by the
event horizon at r = 2m. Any point in this empty region below the surface
r = 2m represents a trapped surface (which is a two-dimensional sphere in
spacetime) in that both the outgoing and in going families of null geodesics
emitted from this point converge, and hence no light ray comes out of this
region bounded by r = 2m. Then, the collapse to an infinite density and
curvature singularity at r = 0 becomes inevitable in a finite proper time,
as measured by an observer on the surface of the star. In this case, the
blackhole region in the resulting vacuum Schwarzschild spacetime is given
by 0 < r < 2m and the outer boundary of this region, r = 2m, is called the
event horizon. On the event horizon, only the radial outwards photons stay
where they are, but all the rest of the photons move inwards. No information
from this blackhole region can propagate outside the r = 2m boundary to
any outside observer. In the Schwarzschild geometry, for a source situated
outside r = 2m, part of the photon trajectories emitted with decreasing r
values will go towards the blackhole and fall into the singularity. All the
other null geodesics will escape to infinity and they intersect the future null
infinity. If a source is located below r = 2m, no null geodesic would come out
of the blackhole and they necessarily end up in the singularity in the future.
56 The spacetime manifold

The final state of a complete gravitational collapse, either spherically sym-

metric or otherwise, could possibly be a vacuum spacetime that incorporates
the rotation, and possibly also the electromagnetic fields associated with the
object. It is possible that the charge associated with an astrophysical object
could be quickly neutralized by the surrounding plasma. However, in any
case it will be of interest to obtain all solutions of the Einstein–Maxwell
equations that describe stationary collapsed configurations with charge.

2.7.6 Vaidya metric and the naked singularity

The geometry outside a spherically symmetric star, when the exterior is
taken to be non-empty due to radiation from the star, is given by the Vaidya
metric (Vaidya, 1943, 1951, 1953). Just as astrophysical bodies are found
to be rotating, they radiate energy in the form of electromagnetic radiation.
The Schwarzschild solution does not describe this as it corresponds to an
empty exterior given by Tij = 0. In the case of a normal star, the effect of
radiation on the overall exterior spacetime could be negligible, and effects
such as rotation, magnetic fields, and so on, are considered as small per-
turbations from spherical symmetry. However, the radiation effects would
be important during the later stages of gravitational collapse when the star
could be throwing away considerable mass as radiation, or when abundant
neutrinos are radiated from a collapsing supernova core (see for example,
Kahana, Baron, and Cooperstein, 1984). Such a non-static distribution as
the radiating star would then be surrounded by an ever-expanding zone of
radiation. This radiating system, together with its radiation, could be treated
as forming an isolated object in an otherwise empty, asymptotically flat uni-
verse. Beyond the zone of pure radiation, the spacetime can be described by
the empty Schwarzschild solution.
Here, a spherically symmetric solution to the Einstein equations Gij =
8πTij is sought with a geometrical optics type stress–energy tensor for the
radiation of the form
Tij = σki kj , (2.187)

where ki is a null vector radially directed outwards. The metric is best given
in the null coordinates (u, r, θ, φ) as
 
2m(u)
ds = − 1 −
2
du2 − 2du dr + r2 dΩ2 , (2.188)
r

with m(u) being an arbitrary non-increasing function of the retarded time

u. The above gives the Vaidya metric in the radiation zone, which is to be
matched by the interior metric of the radiating body at the boundary of
 2.7 Exact solutions 57

the star, and is matched by the Schwarzschild metric in the exterior of the
radiation zone.
The form for the energy–momentum tensor σ is defined to be the energy
density of the radiation as measured locally by an observer with a four-
velocity vector v i . Thus, σ is the energy flux as well as the energy density
measured in this frame,
σ ≡ Tij v i v j , (2.189)

with v i vi = −1. Working out the connection coefficients, the Ricci tensor in
null coordinates is given by

2 dm(u) 0 0
Rij = − δ iδ j . (2.190)
r2 du

This implies that the Ricci scalar Ri i = R = 0, and hence the Einstein
equations give
1 dm(u) 0 0
Tij = − δ iδ j , (2.191)
4πr2 du
which is the energy–momentum tensor of a radiating field in the geometric
optics form. From (2.189) and (2.191),

1 dm(u)
σ=− , (2.192)
4πr2 du
which is the expression for the energy density of radiation.
In the case when m(u) = const., the relationship of the null coordinates
used here with the Schwarzschild coordinates (t, r, θ, φ) is not difficult to see.
In such a case, the transformation given by Finkelstein (1958) can be used
to diagonalize (2.188),

u = T − r − 2m log(r − 2m), (2.193)

which gives the Schwarzschild metric in the (T, r, θ, φ) coordinate system.

The energy flux from the star, as seen by an outside observer, was com-
puted by Lindquist, Schwartz, and Misner (1965) by considering radially
moving observers only. As pointed out, σ is the energy flux measured in a
local frame, and if U ≡ v r = dr/dτ for a radially moving observer with
v i vi = −1, v θ = 0 and v φ = 0, then from (2.191) and (2.192), the energy
density q is
1 dm
q=− (γ + U )−2 . (2.194)
4πr2 du
Since q must be positive as it is energy density, it follows from the above
that dm/du ≤ 0. For an observer at rest at infinity, the total luminosity is
58 The spacetime manifold

given by
dm
L∞ (u) = lim 4πr2 q = − , (2.195)
r→∞,U =0 du
that is, it is the negative rate of change of mass of the radiating body.
The surface r = 2m(u) has many interesting properties, as pointed out
by Lindquist, Schwartz, and Misner (1965). Unlike the Schwarzschild case,
where r = 2m is a null hypersurface, for the Vaidya radiating star metric,
this is a spacelike hypersurface. The induced metric on this hypersurface is
given by  
dm
ds |r=2m(u) = −2
2
du2 + r2 dΩ2 . (2.196)
du
This induced metric has the signature (+, +, +) since dm/du < 0. As a
result, the position of light cones on this surface is such that for all timelike
vectors in the forward light cone at all points on this surface, dr/du > 0.
Therefore, no timelike trajectory from the outside region r > 2m(u) can
come and cross this surface to enter inside the r < 2m(u) region. The solution
is of type D in the Petrov classification of spacetimes and possesses a normal
shear-free congruence with a non-zero expansion.
The Vaidya solution has been used extensively in the context of the cosmic
censorship hypothesis, and to study gravitational collapse endstates (see for
example, Papapetrou, 1985; Joshi, 1993, for details). In that case, imploding
radiation shells are considered, rather than the outgoing case considered here.
Then, the function m is taken to be non-decreasing and the advanced null
coordinate t + r is used. For the case of a linear mass function m(u) = λu,
an interesting causality and horizon structure arises, depending on the rate
of collapse ṁ(u), that is, as decided by the magnitude of λ. In this case,
the model is self-similar in that it admits a homothetic Killing vector. When
the collapse is fast enough, the horizon forms well before the singularity to
fully cover it, and a blackhole forms as the collapse endstate. However, for a
slow enough collapse when λ is smaller than a critical value, the horizon and
trapped surface formation is delayed and a naked singularity develops as the
collapse final state (see Fig. 2.4).
The above scenario may be of physical interest if, in the very late stages
of collapse, the star converts itself into some kind of an imploding radiation
ball. On the other hand, it can be argued that real collapsing stars may not
have the form of the energy–momentum tensor of a radiation fluid in the very
late stages of their collapse. In any case, not much is known today on the
actual equation of state for the collapsing star in the very late stages of its
gravitational collapse. The above consideration really shows what is actually
possible within the framework of general relativistic gravitational collapse,
subject to various physical reasonability and regularity conditions, such as
the validity of energy conditions and evolution from regular initial data.
 2.7 Exact solutions 59

Event horizon
Singularity
at
r=0 Outgoing
r = 2m
null geodesic

Incoming matter

Fig. 2.4 Collapse of radiation shells. A thick shell of radiation arrives at

the center in an otherwise empty spacetime. The mass function is m(u) =
λu and for a slow enough collapse, that is, for λ below a critical value,
a naked singularity forms at the center r = 0 from which non-spacelike
curves escape to infinity.

In fact, the homogeneous dust collapse is also a rather special situation, as

discussed in the next chapter. A physically realistic density profile has to be
inhomogeneous, typically higher at the center for an astrophysical object such
as a star, decreasing as one moves away from the center. In such a case, it is
seen that the structure of the trapped surfaces’ geometry changes radically
as soon as inhomogeneities are included into the collapse considerations.
In the case of a homogeneous collapse, the trapped surfaces start forming
early during the evolution of the collapse, and the spacetime singularity
forms later. This results in an event horizon structure that fully covers the
singularity, thus producing a blackhole in the spacetime. However, when the
density is higher at the center, the trapping is delayed and trapped surfaces
form only at the epoch of the onset of the singularity. Such a singularity is
then visible in principle to external observers, and thus a naked singularity
develops as the collapse endstate, rather than a blackhole.
 3
Spherical collapse

A collapsing matter cloud, such as a massive star undergoing a continual

gravitational collapse at the end of its life cycle, is modeled in general rela-
tivity as a dynamical spacetime geometry with a suitable energy–momentum
tensor. The time evolution here is governed by the Einstein equations. The
cloud has a boundary, with its interior collapsing continually as time evolves,
and, at the boundary the interior spacetime is matched to a suitable exte-
rior geometry so as to complete the full model of gravitational collapse. The
physical situation considered here is that of the force of gravity being so over-
whelming that no final, stable configuration, such as a neutron star or white
dwarf, is possible as a collapse endstate, and a continual collapse inevitably
proceeds. If the initial mass of the collapsing star is sufficiently high, then
such a situation is realized.
In such a scenario, the classical theory leads the collapse to the formation
of a spacetime singularity, as predicted by the singularity theorems of general
relativity. The spacetime singularity is a region close to where the densities,
spacetime curvatures, and all other physical quantities grow without bounds.
At the singularity itself these are infinite, and hence, strictly speaking, the
singularity is not part of the spacetime and is regarded as the boundary of
the spacetime manifold. Eventually, as one moves closer to the singularity,
the quantum gravity effects may dominate, which could resolve the classical
singularity. Even if the final singularity is removed due to quantum effects,
the formation of superdense regions of extreme gravity, as signaled by the
occurrence of a spacetime singularity, are really important physically. The
curvatures and physical conditions are extreme in the vicinity of such regions
that come into being as a result of continual gravitational collapse, which
follows from the physical process of the evolution of a massive star at the
end of its life cycle. Both strong gravity, as well as quantum effects, should
come into their own in such regions.

60
 Spherical collapse 61

The singularity theorems predicting the existence and formation of such

ultra-dense regions also allow these to be either visible to external observers,
or otherwise. This aspect is actually determined by the causal structure
of the spacetime in the very late stages of the dynamical collapse, as gov-
erned by the Einstein equations. If the causal structure covers the ultra-dense
regions within the event horizons of gravity, a blackhole develops as the final
state of collapse. On the other hand, when trapped surfaces are delayed in
the continual collapse, a naked singularity forms, which is nothing but a
visible ultra-dense region that may communicate physical effects to outside
observers in the universe. The theoretical and observational properties of
these objects would typically be quite different from each other, and so it
is of much interest to gain an insight into how each of these phases come
about as endstates of a dynamically developing collapse that is governed by
gravity.
The purpose here is to investigate the gravitational collapse scenarios
within the framework of Einstein gravity, and to determine when either a
blackhole or a naked singularity will result in a collapse endstate. For the
sake of focus and clarity, spherical collapse models are considered here in
order to obtain explicit results, and certain more general aspects will be dis-
cussed later. The collapsing matter fields are chosen to be of general type I,
which is a broad class that includes most of the physically reasonable matter
fields. It is shown that, given the initial data for matter (in terms of the ini-
tial density and pressure profiles at an initial surface t = ti from which the
collapse develops), the rest of the initial functions and classes of solutions of
the Einstein equations constructed here exist, such that the spacetime evo-
lution goes to either a blackhole or a naked singularity final state, depending
on the nature of the initial data and the evolutions. The validity of the weak
energy condition and various regularity conditions are preserved. Assump-
tions such as self-similarity of spacetime are not imposed here, in order to
keep the treatment general.
The matter fields treated here are a general and broad class that include
most of the physically reasonable matter such as dust, perfect fluids, massless
scalar fields and such others, and any special restrictions are not imposed on
the form of the matter. In order to investigate the collapse final states, given
the initial data for the matter in terms of the initial density and pressure pro-
files at an initial surface t = ti , classes of solutions to the Einstein equations
are constructed such that the spacetime evolution goes to a final state that
is either a blackhole or a naked singularity, depending on the nature of the
rest of the free initial data functions and allowed evolutions. The method-
ology used here allows the genericity and stability aspects related to the
occurrence of naked singularities in gravitational collapse to be considered
in some detail.
62 Spherical collapse

In Section 3.1, the basic framework for a collapse scenario is described.

Various regularity conditions for gravitational collapse are discussed in
Section 3.2. The dynamical evolution of the collapsing clouds is consid-
ered in Section 3.3. The apparent horizon and structure of a trapped
surface formation is discussed in Section 3.4, which describes the evolv-
ing causal structure of the collapsing cloud that determines the nature of
the singularity in terms of being either visible or covered. The exterior
spacetime and matching conditions are given in Section 3.5. As an exam-
ple to illustrate the formalism here, the gravitational collapse of a dust
cloud with vanishing pressures is discussed in some detail in Section 3.6.
Finally, the issues related to the equation of state and those of the
validity of the energy conditions in the treatment here are considered in
Section 3.7.

3.1 Basic framework

To consider spherical symmetry, let P be any point at a distance r from the
origin O, then the system must be invariant under rotations around O. Such
rotations will generate a two-sphere around O, and the line element on it is
given by
ds2 = r2 (dθ2 + sin2 θ dφ2 ), (3.1)
which is the metric on the two-sphere given by t = const., r = const. in a
general spherically symmetric spacetime. Furthermore, as the metric must
be invariant under the reflections θ → π − θ and φ → −φ, there are no cross
terms in the metric in dθ and dφ. As the line elements must not change with
any change in θ and φ, they have to occur in the metric only in the form of
the two-metric given above. Then, in the (t, r, θ, φ) coordinate system, the
spacetime metric has the form

ds2 = −A dt2 + 2B dt dr + C dr2 + D(dθ2 + sin2 θ dφ2 ). (3.2)

Here, the quantities A, B, C , and D are the functions of t and r that are
to be determined. Introducing a new radial coordinate r = D1/2 and a new
time coordinate t by requiring that dt = F [A dt − B  dr ], where F (t, r ) is
a suitable integrating factor, the line element reduces to

ds2 = −eν dt2 + eψ dr2 + r2 (dθ2 + sin2 θ dφ2 ), (3.3)

where the primes have been dropped. Here, ν = ν (r, t) and ψ = ψ (r, t), and
the quantities eν = 1/F 2 A and eψ = C  − B 2 /A appearing in the metric
are always positive.
In general, the spherical symmetry of a spacetime can be defined using
Killing vectors. There must be three linearly independent spacelike Killing
 3.1 Basic framework 63

vector fields X 1 , X 2 , and X 3 in the spacetime that satisfy the commutator

relations

[X 1 , X 2 ] = X 3 , [X 2 , X 3 ] = X 1 , [X 3 , X 1 ] = X 2 , (3.4)

and their orbits must be closed. Using these properties, the line element
above could be derived rigorously for a spherically symmetric spacetime.
A general formalism for a spherically symmetric gravitational collapse
including pressure, was developed by Misner and Sharp (1964). The
spherically symmetric spacetime is written in comoving coordinates as

ds2 = −e2φ dt2 + eλ dr2 + R2 (t, r) dΩ2 , (3.5)

where dΩ2 = dθ2 + sin2 θ dφ2 is the usual metric on the two-sphere, and φ
and λ are functions of t and r. The stress–energy tensor is that of a perfect
fluid given by
Tij = (ρ + p)ui uj + pgij . (3.6)
The spatial velocities ui are vanishing here and the spatial coordinates of a
given particle remain constant throughout the collapse. A function m(r, t)
is introduced by the definition
 −1
2m
λ
e = 1 + Ṙ −
2
R2 , (3.7)
r

where a dash denotes a derivative with respect to r, and for any function f,
 
˙ −φ ∂f
f =e . (3.8)
∂t

The coordinate t here gives the proper time along the particle world lines.
Integrating the conservation equation T ij ;j = 0 and solving the Einstein
equations, the Misner–Sharp equations for the spherically symmetric collapse
can be written as

ṁ = −4πR2 pṘ, (3.9)

  
1 + Ṙ2 − 2m/r ∂p m + 4πR3 p
R̈ = − , (3.10)
ρ+p ∂R R2
∂m
= 4πR2 ρ. (3.11)
∂R
The above equations, when combined with an equation of state relating ρ
and p, determine the dynamical evolution of the collapse. However, when the
pressure p = 0, the situation is quite complex and numerical computation
64 Spherical collapse

(see for example, May and White, 1966) can be opted for in order to get an
idea of the evolution of the collapsing system.
The general spherically symmetric metric, in a spacetime of dimension N ,
can be written in the form

ds2 = −gab (x0 , x1 ) dxa dxb + R2 (x0 , x1 )dΩ2N −2 , (3.12)

where a and b run from 0 to 1, and

 i−1 
N −2 
2 2
dΩN −2 = sin (θ ) (dθi )2
j
(3.13)
i=1 j=1

is the metric on an (N − 2) sphere with θi being the spherical coordinates.

From this metric, the components of the Einstein tensor can be obtained as
(Rocha and Wang, 2000)

N −2
N
Gab = [gab {2R
R + (N − 3)Q} − 2RR,ab ] , (3.14)
2R 2
1 
N
G22 = − R2  + (N − 3){2R
R + (N − 4)Q} , (3.15)
2
 i−2 

N
Gii (i > 2) = sin2 (θk ) N G22 , (3.16)
k=1

where

Q = 1 + R,a R,a ,
R = g ab R;ab , (3.17)

and
 = g ab ab . (3.18)
Here, ab is the Ricci tensor evaluated from the two-metric gab , and  is the
scalar curvature also evaluated from the two-metric gab .
The spacetime geometry within the spherically symmetric collapsing cloud
is described by comoving coordinates (t, r, θ i ), which are specified below.
The matter field is chosen to be of general type I, which is a broad class
including most of the physically reasonable matter forms, including dust,
perfect fluids, massless scalar fields, and such others. This class of matter
is specified by the requirement that the energy momentum tensor for the
matter admits one timelike and three spacelike eigenvectors (Hawking and
Ellis, 1973; Stephani et al., 2003).
The coordinates (t, r, θ i ) are then chosen to be those along these eigenvec-
tors, which makes the coordinate system comoving, that is, the coordinate
system moves with the matter. The freedom of coordinate transformations
 3.1 Basic framework 65

of the form t = f (t, r) and r = g (t, r) can be used to make the gtr term in
(3.12) and the radial velocity of the matter vanish. In this case, the general
metric in the comoving coordinates (t, r, θ i ) must have three general arbi-
trary functions of t and r, and this can be written in the form (Landau and
Lifshitz, 1975)

ds2 = −e2ν(t, r) dt2 + e2ψ(t, r) dr2 + R2 (t, r)dΩ2N −2 . (3.19)

In this comoving frame, the energy–momentum tensor for any matter field
that is type I is given in a diagonal form
i
T t t = −ρ(t, r), T r r = pr (t, r), T θ θi = pθ (t, r). (3.20)

The quantities ρ, pr , and pθ are the energy density, and radial and tangential
pressures ascribed to the matter field respectively.
The matter cloud can be chosen to have a compact support at an ini-
tial spacelike surface of t = ti , with 0 < r < rb , where rb would denote
the boundary of the cloud. Outside this boundary, the interior solution has
to be matched through suitable junction conditions, with another suitable
spacetime metric to complete the full spacetime geometry.
The matter fields are taken to satisfy the weak energy condition, that is,
the energy density measured by any local timelike observer is non-negative.
This ensures the physical reasonability for the collapsing matter fields consid-
ered here. Another energy condition frequently used is the dominant energy
condition, which demands that, for any timelike observer, the local energy
flow is non-spacelike. These two conditions are frequently regarded as the
main and important energy conditions that are physically reasonable (Hawk-
ing and Ellis, 1973; Wald, 1984). All classical observed matter fields satisfy
these conditions. For the energy conditions to be satisfied, for any timelike
vector V i,
Tik V i V k ≥ 0 (3.21)

and T ik Vk must be non-spacelike. For the energy–momentum tensor (3.20),

these amount to the conditions

ρ ≥ 0, ρ + pr ≥ 0, ρ + pθ ≥ 0, (3.22)
|pr | ≤ ρ, |pθ | ≤ ρ. (3.23)

Now, with the above metric, the following quantities can be evaluated


R,a R,a = Ṙ2 e−2ν − R 2 e−2ψ , (3.24)
R = e−2ν [R̈ + Ṙ(−ν̇ + ψ̇ )] − e−2ψ [R + R (ν  − ψ  )], (3.25)
66 Spherical collapse

and
   
e2ν ν  R e2ψ ψ̇ Ṙ
R,ab = R̈ − ν̇ Ṙ − δa0 δb0 + R − ψ  R − δa1 δb1 + Qab ,
e2ψ e2ν
(3.26)
where   
Qab = Ṙ − ν  Ṙ − ψ̇R δa0 δb1 + δa1 δb0 . (3.27)

Using the above equations, the Einstein equations Gik = Tik now take the
form (in the units 8πG = c = 1)

(N − 2)F  (N − 2)Ḟ
ρ= −2 
, pr = − , (3.28)
2R N R 2RN −2 Ṙ
(N − 2)(pθ − pr ) R pr
ν = − , (3.29)
ρ + pr R ρ + pr
Ġ H
−2Ṙ + R + Ṙ = 0, (3.30)
G H
F
G − H = 1 − N −3 , (3.31)
R
where
G(t, r) = e−2ψ (R )2 , H (t, r) = e−2ν (Ṙ)2 . (3.32)
The arbitrary function F = F (t, r) here has the interpretation of the mass
function for the cloud, which gives the total mass in a shell of comoving radius
r, at an epoch t. The energy condition ρ ≥ 0 implies F ≥ 0 and F  ≥ 0.
Since the area radius vanishes at the center of the cloud, from (3.28), it is
evident that in order to preserve the regularity of density and pressures at
any non-singular epoch t, F (t, 0) = 0, that is the mass function vanishes at
the center of the cloud.
As seen from (3.28), there is a density singularity in the spacetime at R = 0
and at R = 0. However, the latter ones are due to shell-crossings (Yodzis,
Seifert, and Muller zum Hagen, 1973, 1974), which basically indicate the
breakdown of the coordinate system used. These are not generally regarded
as genuine spacetime singularities, because they can be possibly removed
from the spacetime to extend the manifold (Clarke, 1986, 1993). Hence, only
the shell-focusing singularity at R = 0, which is a genuine physical singularity
where all matter shells collapse to a zero physical radius, will be considered
here. This will be discussed in more detail later.
Note that, in general, for a general matter field with non-vanishing pres-
sures as considered here, there are a variety of dynamical time evolutions
possible from the given matter density and pressure profiles as prescribed
on an initial surface (that are called matter initial data here), from which
 3.1 Basic framework 67

the collapse evolves. In particular, even if the cloud commences gravitational

collapse at the initial surface t = ti , there can be classes of solutions of the
Einstein equations where the evolution is such that a bounce is possible at
a later stage for the cloud. Here, mainly the continually collapsing class of
models are considered, because interest is in the physical situation that cor-
responds to the case when the mass of the collapsing cloud, or the star, is so
high that on exhausting its nuclear fuel, the star must undergo a continual
gravitational collapse, completing it in a finite time. Therefore, the contin-
ual collapse condition is included as a part of the framework here. In such
a case, trapped surfaces form as the collapse evolves and a spacetime sin-
gularity necessarily develops as the collapse endstate. The conditions when
the final singularity is necessarily covered within an event horizon of gravity
(as hypothesized by the cosmic censorship) or when it will be naked with
ultra-strong gravity regions being visible to faraway observers then need to
be found. In the case of a bounce or dispersal, no singularity needs to form
in the spacetime, a situation that is not considered here.
The scaling freedom available for the radial coordinate r can be used to
write
R=r (3.33)
at the initial epoch t = ti . It is interesting to note that, if one had wished
to scale the radial coordinate at the initial epoch as R(ti , r) = rβ , with
β being any constant, then the only possible allowed value for β would be
unity. This is because, in the other case, either R would blow up at the
center r = 0 (which is not allowed by regularity conditions, as the Einstein
equations require the metric functions to be at least C 2 ), or R would go to
zero at the center, causing a shell-crossing singularity (which is avoided by
construction) that violates the regularity of the initial data.
A function v (t, r) is now introduced (Joshi and Dwivedi, 1999; Joshi and
Goswami, 2004) as defined by

v (t, r) ≡ R/r. (3.34)

Then, R(t, r) = rv (t, r), and

v (ti , r) = 1, v (ts (r), r) = 0, v̇ < 0. (3.35)

The time t = ts (r), that is v = 0, corresponds to the shell-focusing singularity

at R = 0, which is a genuine spacetime singularity where all the matter shells
collapse to a vanishing physical radius. The condition v̇ < 0 corresponds here
to a continual collapse of the cloud. The description of the shell-focusing
singularity at R = 0 in terms of the function v (t, r) as above has several
advantages. The physical radius goes to zero at the shell-focusing singularity,
but also R = 0 at the regular center of the cloud at r = 0. This is to be
68 Spherical collapse

distinguished from the genuine singularity at the collapse endstate, by the

fact, for example, that the density and other physical quantities including
the curvature scalars all remain finite at the regular center r = 0 of the
cloud, even though R = 0 holds there. This is achieved, as shown below, by
a suitable behavior of the mass function, which should go to a vanishing value
sufficiently fast in the limit of approach to the regular center where (even
though R goes to zero) the density must remain finite. On the other hand,
when the function v (t, r) is used, note that at t = ti , v = 1 on the entire
initial surface, and then as the collapse evolves, the function v continuously
decreases to become zero only at the singularity ts (r). This means that v = 0
uniquely corresponds to the genuine spacetime singularity at R = 0.
From the point of view of dynamic evolution of the initial data prescribed
at the initial epoch t = ti , there are five arbitrary functions of the comoving
shell-radius r, as given by

ν (ti , r) = ν0 (r), ψ (ti , r) = ψ0 (r), R(ti , r) = r,

ρ(ti , r) = ρ0 (r), pr (ti , r) = pr0 (r), pθ (ti , r) = pθ0 . (3.36)

Note that not all the initial data functions above are mutually independent,
as, from (3.29),
  
r
(N − 2)(pθ0 − pr0 ) pr0
ν0 (r) = − dr. (3.37)
0 r(ρ0 + pr0 ) ρ0 + pr0

Therefore, apart from the matter initial data, which describe the initial den-
sity and pressure profiles at the initial epoch t = ti , the rest of the initial
data that are free is ψ0 (r), which essentially describes the velocities of the
collapsing matter shells as discussed later.
Note that in the above, when the pressures are taken to be vanishing, the
scenario reduces to a dust collapse. In this case, the only two free initial
data functions are the initial density and the velocity profile for the cloud.
As discussed later, in this case, for any given initial density profile ρ(r)
for the matter cloud, there are classes of initial velocity profiles ψ (r) that
take the final collapse outcome to either a blackhole, or a naked singularity,
depending on the choice of initial velocities. This is subject to the validity
of the energy conditions and regularity conditions for gravitational collapse.
The Oppenheimer–Snyder homogeneous dust collapse corresponds here to a
special choice of the initial density profile that is a constant function, and
another special choice of the initial velocity function. In fact, if the initial
density profile is chosen to be constant, but the initial velocities are allowed to
take different functional forms, then an initially constant and homogeneous
density can also later inhomogenize, and result in a final evolution that is a
 3.2 Regularity conditions 69

naked singularity. If the pressures are allowed to be non-vanishing, then the

choices in evolution are much wider, as shown below.

3.2 Regularity conditions

Whilst the basic equations for a general spacetime of dimension N (Goswami
and Joshi, 2006) have been given above, for the rest of this chapter the usual
spacetime dimensions of four (N = 4) shall be used for the sake of clarity and
transparency (Joshi and Dwivedi, 1999; Joshi and Goswami, 2004). Some of
the useful and interesting differences arising when there is a higher spacetime
dimension will be pointed out as necessary.
The regularity of the initial data needs to be ensured in order to make
sure that the gravitational collapse initiates from regular and physically rea-
sonable initial conditions. The initial pressures must thus be taken to have
physically reasonable behavior at the center. Considering that the total force
at the center of the collapsing cloud should be zero, the gradients of initial
pressures vanish at the center. Also, regularity of the initial data requires
that pr0 (0) − pθ0 (0) = 0, that is, at the center the difference between the
radial and tangential pressure vanishes. The metric functions have to be C 2 -
differentiable everywhere as per the requirements of the Einstein equations.
As seen from the equation for ν  above, such a condition is implied by the
requirement that ν  does not blow up at the regular center. This means that
the matter should behave like a perfect fluid at the center of the cloud with
the net force vanishing there.
Note that these regularity conditions do not exclude the collapse models
with either a purely tangential pressure (pr = 0), or purely radial pressure
collapse models (pθ = 0), which will be referred to later in this chapter. This
is because in these cases, the above condition implies that pθ → 0 or pr → 0
respectively, close to the center, where the matter closely approximates dust.
Note that the regularity conditions above give a sufficient criterion for the
regularity of the metric function ν0 (r) at any non-singular initial epoch. It
follows from (3.37) that at the center of the cloud both ν0 and ν0 go to zero.
Hence, ν0 (r) has the form

ν0 (r) = r2 g (r), (3.38)

where g (r) is an arbitrary function which is at least C 1 for r = 0, and

is at least a C 2 -function for r > 0, as the Einstein equations demand the
metric functions to be at least C 2 everywhere. Another regularity condition
frequently used in collapse considerations is that there should be no trapped
surfaces at the initial spacelike surface from which the collapse begins.
Therefore, there are five total field equations with seven unknowns, ρ, pr ,
pθ , ψ , ν , R, and F , which are three matter variables, three metric functions
70 Spherical collapse

and the mass function. This gives the freedom of choice of two free functions.
Selection of these free functions, subject to the weak energy condition and
the given regular initial data for collapse at the initial surface, determines the
matter distribution and metric of the spacetime throughout, and thus leads
to a particular time evolution of the initial matter and velocity distributions
with which the collapse began. As will be shown, it turns out that, given the
matter initial profiles in terms of ρ0 , pr0 , and pθ0 , there are the rest of the
initial data at t = ti , and the classes of solutions to the Einstein equations
that are explicitly constructed here, which give either a blackhole or a naked
singularity as the endstate of collapse. This outcome depends on the nature
of the rest of the initial functions, and the classes of the dynamical evolutions,
as allowed by the Einstein equations.
An important point to be noted here is that, in the description above no
mention has been made so far of the equations of state that the matter must
obey. Typically, these are of the form, pr = pr (ρ) and pθ = pθ (ρ). If these
are specified, then there is no freedom left, and there are seven equations for
seven variables. If this were to be incorporated right away, the only way to
proceed to find the collapse endstate would be to assume a specific equation
of state that the matter must satisfy. Then, one has to examine the collapse
problem accordingly, and examine the nature of the final singularity resulting
from the dynamical evolution, as governed by the Einstein equations. There
have been many collapse studies in the past using this approach, such as for
example, for the dust equation of state for perfect fluids, and others. The
limitation of such an approach, however, is that there is very little existing
knowledge on what a realistic equation of state should be that the matter
has to satisfy at the extreme high densities that a continual collapse realizes
in its advanced stages. For example, even for neutron star densities that are
relatively low compared with those of the later stages of a continual collapse,
there is a great deal of uncertainty on the equation of state for such neutron
matter. As a result, the neutron star mass limits are uncertain. Therefore,
specific or special assumptions used on the equation of state may turn out
to be physically unrealistic or restrictive, and untenable in the final stages
of the collapse. In fact, diametrically opposite views exist on the possible
equation of state in the very late stages of the collapse. For example, while
there are many arguments suggesting that pressures must play an important
role in the later stages of the collapse, the other view is that in such late
stages the matter must necessarily be dust-like (see for example, Hagedorn,
1968; Penrose, 1974a, 1974b).
Under this situation, the approach taken here is that no specific or partic-
ular equation of state is assumed to begin with. All further considerations are
carried out in a general manner, in terms of the allowed initial matter profiles
and the allowed dynamical evolutions of the Einstein equations, towards
determining the blackhole and naked singularity endstates for collapse.
 3.3 Collapsing matter clouds 71

Subsequently, in Section 3.7, the role that the equation of state plays towards
further fine tuning the blackhole and naked singularity outcomes as collapse
endstates is discussed. The advantage of such an approach here is that, first
all the collapse equations are written in general, and only then are different
subcases distinguished, depending on the corresponding equation of state
under consideration. As shall be shown, various important subcases and
equations of state such as dust, perfect fluids, and others are included as
special cases of the treatment given here.

3.3 Collapsing matter clouds

Now the evolution of the matter cloud is considered. The method followed is
outlined below. In the case of a blackhole developing as a collapse endstate,
the spacetime singularity is necessarily hidden behind the event horizon of
gravity. In the case of a naked singularity developing, there are families of
future directed non-spacelike trajectories that terminate in the past at the
singularity, and which can, in principle, communicate information to faraway
observers in the spacetime. The existence of such families confirms the naked
singularity formation, as opposed to a blackhole collapse endstate. The final
singularity produced by the collapsing matter is studied here, and it is shown
that the tangent to the singularity curve at the central singularity at r = 0
is related to the radially outgoing null geodesics from the singularity, if there
are any. By determining the nature of the singularity curve, and its rela-
tion to the initial data and the classes of collapse evolutions, it is possible
to deduce whether the trapped surface formation in the collapse takes place
earlier than the singularity formation epoch, or is delayed. It is this causal
structure of the trapped surface region, and the apparent horizon (that is
the boundary of the trapped region), forming during the collapse that deter-
mines the possible emergence, or otherwise, of non-spacelike curves from the
singularity. This settles the final outcome of the collapse in terms of either
a blackhole or naked singularity formation.
It is necessary to clarify here the meaning of the occurrence of a naked
singularity developing as the end point of the collapse. At times, the non-
existence of trapped surfaces until the formation of the singularity in the
collapse is taken as the signature that the singularity is naked. This has
a reference generally to some slicing of the spacetime, in which the occur-
rence, or otherwise, of the trapped surfaces is examined (see for example,
Shapiro and Teukolsky, 1991, 1992). At times, however, a slicing dependent
definition of naked singularity may not give an accurate picture of what is
happening in the spacetime (Wald and Iyer, 1991). What is meant by the
development of a naked singularity in the collapse is that families of future
directed non-spacelike curves, which in the past terminate at the singularity,
72 Spherical collapse

Exterior
Time
Interior

Boundary of the star

0, pr0, p0
Initial surface

Fig. 3.1 The collapse evolves from regular initial data from an initial space-
like hypersurface. The cloud has a compact support on the spacelike slice and
the interior collapse spacetime is matched to a suitable exterior metric at the
boundary.

exist. No such families exist, originating from the singularity, when the end
product of collapse is a blackhole. This is a definition that is independent of
the coordinates used and the slicing chosen. In the case of a blackhole devel-
oping, the resultant spacetime singularity will be hidden inside an event
horizon of gravity, remaining unseen by external observers. On the other
hand, if the collapse ends in a naked visible singularity, there is a causal
connection between the region of singularity and faraway observers, thus
enabling, in principle, a communication from the super-dense regions close
to the singularity to faraway observers.
Given the matter initial profiles in terms of the initial functions as given
by ρ0 (r), pr0 (r), pθ0 (r) at the initial epoch t = ti from which the collapse
commences, the purpose now is to construct and examine possible evolutions
(classes of solutions to the Einstein equations) of such a matter cloud from
the given initial data, to investigate its final states (see Fig. 3.1).
While constructing the classes of solutions that give the collapse evolutions,
given the matter initial data at t = ti , as much generality as possible is
preserved. The mass function F (t, r) for the collapsing cloud must have the
following general form:
F (t, r) = r3 M(r, v ), (3.39)
where M > 0 is at least a C 1 -function of r for r = 0, and at least a C 2 -
function for r > 0. It is to be noted that F must have this general form,
which follows from the regularity and finiteness of the density profile at the
initial epoch t = ti , and at all other later regular epochs before the cloud
collapses to the final singularity at R = 0. This requires, from the Einstein
equation for energy density, that F must behave as r3 close to the regular
center. Hence, note that since M is a general (at least C 2 ) function, (3.39)
is not really any ansatz or a special choice, but quite a generic class of the
mass profiles for the collapsing cloud, consistent with and allowed by the
regularity conditions. Therefore, no special choice of F is made, but it is
 3.3 Collapsing matter clouds 73

allowed to be a general function as given by the Einstein equation (3.28),

while constructing the classes of solutions that give collapse evolutions from
the given regular initial data.
Then, (3.28) gives

3M + r [M,r + M,v v  ]
ρ(r, v ) = , (3.40)
v 2 (v + rv  )
and
M,v
pr (r, v ) = − . (3.41)
v2
The regular density distribution at the initial epoch is given by

ρ0 (r) = 3M(r, 1) + rM(r, 1),r . (3.42)

It is evident that, in general, as v → 0, ρ → ∞ and pr → ∞. That is, both

the density and radial pressure blow up at the shell-focusing singularity.
It is shown that, given any regular initial density and pressure profiles for
the matter cloud from which the collapse develops, energy profiles or velocity
functions for the collapsing matter shells and classes of dynamical evolutions
as determined by the Einstein equations always exist, so that the collapse
endstate would be either a naked singularity or a blackhole, depending on
the nature of the allowed choices. Therefore, given the matter initial data at
the initial surface t = ti , these evolutions take the collapse to end either as a
blackhole or naked singularity, depending on the choice of the class, subject
to the regularity and energy conditions.
To see this, classes of solutions to the Einstein equations need to be
constructed. A suitably differentiable function A(r, v ) can be defined as

ν  (r, v ) = A(r, v ),v R . (3.43)

That is, A(r, v ),v ≡ ν  /R , and since at t = ti one has R = r, this gives
[A(r, v ),v ]v=1 = ν0 (r). The main interest here is in studying the shell-focusing
singularity at R = 0, which is the physical singularity where all the matter
shells collapse to zero physical radius. Therefore, assume that there are no
shell-crossing singularities in the spacetime where R = 0, and that the func-
tion A(r, v ) is well-defined. From (3.38), the form of ν (t, r) can be generalized
and chosen as the class of functions given by

ν (t, r) = r2 g1 (r, v ), (3.44)

where g1 (r, v ) is a function that is suitably differentiable and g1 (r, 1) = g (r).

It then follows that A(r, v ) has the form

A(r, v ) = rg2 (r, v ). (3.45)

74 Spherical collapse

The class of solutions considered here is with no shell-crossing singular-

ities developing as the collapse evolves, that is R > 0. This is because,
as mentioned above, such singularities are generally weak and can possi-
bly be removed from the spacetime as they are typically gravitationally not
strong, and because spacetime extensions have been constructed in certain
cases (Clarke, 1993). In contrast, in several physically reasonable collapse
models including dust and perfect fluids, the singularity at R = 0 turns out
to be gravitationally strong and is a powerful curvature singularity. Under
this situation, one is interested here only in examining the nature of the
shell-focusing singularities at R = 0, which are genuine curvature singular-
ities arising at the termination of collapse, where the physical radii for all
collapsing shells vanish, and the spacetime necessarily terminates without
extension.
Specifically, R > 0 implies that v + rv  > 0. Since v is necessarily positive
throughout the collapse, it follows that this will be satisfied always whenever
v  is greater than or equal to zero. Even when it is negative, the condition
that the magnitude of rv  should be less than that of v is sufficient to ensure
that there will be no shell-crosses. Later in this section, an expression for
the quantity v , in terms of the initial data and the other free evolutions as
allowed by the Einstein equations is derived. So, it follows that the condition
for avoidance of shell-crossings can be specifically stated in terms of the
behavior of these functions. In particular, it turns out that whenever the
singularity curve ts (r) (which corresponds to R = 0) is increasing at the
center (or when it decreases at a sufficiently slow rate) with a slope greater
than or equal to zero at the origin, the shell-crossing singularities are avoided,
at least in the vicinity of the regular center r = 0. Then, a ball of finite
radius around the regular center that contains no shell-crossings until the
final singularity formation at R = 0 exists.
Coming to the dynamical collapse evolutions, using (3.43) in (3.30), a class
of solutions of the Einstein equations can be given as

G(r, v ) = b(r)e2rA(r,v) . (3.46)

Here, b(r) is another arbitrary function of the shell radius r. By using the
regularity condition on the function v̇ at the center of the cloud,

b(r) = 1 + r2 b0 (r), (3.47)

where b0 (r) is the energy distribution function for the shells. Using (3.43) in
(3.29),
Rpr
pθ = RA,v (ρ + pr ) + 2pr + . (3.48)
R
 3.3 Collapsing matter clouds 75

It can be seen that in general, both the density and radial pressure blow up
at the singularity, so the above equation implies that the tangential pressure
would also typically blow up at the singularity.
Now, using (3.39), (3.43), and (3.46) in (3.31),

R1/2 Ṙ = −eν(r,v) (1 + r2 b0 )Re2r A(r,v) − R + r3 M. (3.49)

The negative sign on the right-hand side of the above equation corresponds
to a collapse scenario where Ṙ < 0. Defining a function h(r, v ) as

e2r A(r,v) − 1
h(r, v ) = = 2A(r, v ) + O(r2 ), (3.50)
r2
(3.31) becomes

v 1/2 v = − e(rA+v) vb0 + e2v (vh + M). (3.51)

Integrating the above equation with respect to v ,

 1
v 1/2 dv
t(v, r) =  . (3.52)
v e(rA+ν) vb0 + e2ν (vh + M)

Note that the variable r is treated as a constant in the above equation. The
above equation gives the time taken for a shell labeled r to reach a particular
epoch v from the initial value v = 1. Expanding the function t(v, r) around
the center of the cloud, provided the functions within the integral above are
sufficiently regular, gives

t(v, r) = t(v, 0) + rX (v ) + O(r2 ), (3.53)

where the function X (v ) is given as


1 1
v 1/2 (vb1 + h1 v + M1 (v ))
X (v ) = − dv , (3.54)
2 v (vb00 + vh0 + M0 (v ))3/2

where

b00 = b0 (0), M0 (v ) = M(0, v ), h0 = h(0, v ),

b1 = b0 (0), M1 (v ) = M,r (0, v ), h1 = h,r (0, v ). (3.55)

Hence, it can be seen that the time taken for a shell labeled r to reach the
spacetime singularity at R = 0 (which is the singularity curve) is given as
 1
v 1/2 dv
ts (r) =  . (3.56)
0 e(rA+ν) vb0 + e2ν (vh + M)
76 Spherical collapse

As it is the continual collapse to be considered here, only those classes of

solutions where ts (r) is finite and sufficiently regular are focused on. This
means that the cloud collapses in a finite amount of time. In the physi-
cal situation of a continual collapse of a massive matter cloud in a finite
amount of time, the function ts (r) has to be finite by definition. As for reg-
ularity, to check the existence conditions for a well-defined, continuous and
C 2 -differentiable singularity curve, a function Q(r, v ) can be defined as

v (N −3)/2
Q(r, v ) =  . (3.57)
e(rA+ν) v (N −3) b0 + e2ν (v (N −3) h + M)

Consider now the following functions,

 1  1
φ1 (r) = Q(r, v ),r dv, φ2 (r) = Q(r, v ),rr dv. (3.58)
0 0

Let A be the rectangular area in the (r, v ) plane defined by the lines

r = 0, r = , v = 0, v = 1. (3.59)

Now, if the following conditions:

(1) Q(r, v ) is a continuous function of r and v in A;

(2) Q(r, v ),r and Q(r, v ),rr are continuous functions of r and v in A;
(3) the integrals φ1 (r) and φ2 (r) converge uniformly in A;

are satisfied, then

d d2
φ1 (r) = [ts (r)], φ2 (r) = [ts (r)], (3.60)
dr dr2
and this implies that the singularity curve ts (r) would be a well-defined
C 2 -function near the center.
Several well-studied collapse models such as dust collapse models and
others, satisfy these or stronger conditions, and so the singularity curve is
well-defined and expandable. Some examples of such singularity curves that
form special cases for the consideration above follow.
In the case of a dust collapse, the expression for the singularity curve is
given by (Goswami and Joshi, 2004a, 2004b)
 1
v 1/2 dv
ts (r)dust =  , (3.61)
0 M(r) + vb0 (r)

where M(r) and b0 (r) are well-defined C 2 -functions of the comoving coor-
dinate r and are well-defined at r = 0. It can then be easily seen that the
 3.3 Collapsing matter clouds 77

singularity curve is differentiable at the center with


dts (r) 1 1 v 1/2 (M1 + vb1 )dv
=− , (3.62)
dr 2 0 (M0 + vb00 )3/2

where

b00 = b0 (0), M0 = M(0),

b1 = b (0), M1 = M (0).

(3.63)

Another such example is that of an Einstein cluster (Ec), which describes a

non-steady spherically symmetrical system of non-colliding particles moving
in such a way that, relative to a suitably moving frame of co-ordinates, their
motion is purely transversal. In this case, the singularity curve is given by
(Mahajan, Goswami, and Joshi, 2005)

 1 v 1/2 v 2 + L(r)
2
dv
r2
ts (r)Ec =     . (3.64)
0 2 2
eν b0 v 3 − r4 v + M(r) v 2 + r2
L(r) L(r)

Here, L(r) is a function of the radial coordinate r only. In the Newtonian

limit, this function corresponds to the angular momentum per unit mass
of the system. Therefore, the function L(r) is called the specific angular
momentum, and has the form

L(r) ≡ r2 l(r). (3.65)

Again, since M(r) and b0 (r) and L(r) are well-defined C 2 -functions of the
coordinate r, it can be seen that the above singularity curve is well-defined
and differentiable at r = 0.
To generalize this, and to give another explicit example, it can be shown
that (Goswami and Joshi, 2006) given any initial data of the form (3.36),
classes of dynamical evolutions that give rise to a well-defined and differ-
entiable singularity curve always exist. By the freedom of choice of free
functions, the evolution functions M(r, v ) and A(r, v ) can be chosen in the
following way,

M(r, v ) = m(r) − pr (r)v 3 , A(r, v ),v = ν0 (r v ),R . (3.66)

The Einstein equations here imply that, for the above class of evolutions,
the radial pressure remains static. However, the tangential pressure blows up
along with the density at the singularity v = 0 and is given by

2pθ (r, v ) = pr + pr (R/R ) + ν0 (rv ),R [ρ(r, v ) + pr ]. (3.67)

78 Spherical collapse

It is now easy to check that the above class of evolutions admits a well-
defined and differentiable singularity curve, as both the functions M(r, v )
and A(r, v ) are well-defined and are C 2 at r = 0 and v = 0.
The point is that, as long as in the construction of classes of the solutions
the functions M and A are taken to be sufficiently regular, then the sin-
gularity curve also becomes regular and expandable, at least up to the first
order. If the singularity curve is not regular even to that extent, it means
that it has no well-defined tangent at the origin, and so the collapse evolu-
tion is not regular enough and may not be of physical interest. In any case,
the purpose here is not to analyze all possible collapse evolutions from a
given initial distribution of matter and velocities. The aim really is to show
that, given any such regular distribution from which the collapse develops,
there exist classes of dynamical evolutions that take the other collapse to
either the naked singularity or the other collapse endstates, depending on
the choice of allowed functions. It is also shown that well-known classes of
collapse models, such as dust, perfect fluids, and others, form special classes
of the treatment given here.
Once a singularity curve that is at least C 2 exists, the function can be
Taylor expanded near the center as

ts (r) = ts0 + rX (0) + O(r2 ), (3.68)

where ts0 is the time at which the central singularity at R = 0, r = 0 develops,

and is given as
 1
v 1/2 dv
ts0 =  . (3.69)
0 vb00 + vh0 + M0 (v )

From the above equation, it is clear that for ts0 to be defined,

vb00 + vh0 + M0 (v ) > 0. (3.70)

In other words, a continual collapse in finite time ensures that the above
condition holds. Also, from (3.51) and (3.53), for small values of r, along
constant v surfaces,

v 1/2 v  = (vb00 + vh0 + M0 (v ))X (v ) + O(r). (3.71)

It is now clear that the value of X (0) depends on the functions b0 , M, and
h, which in turn depend on the initial data at t = ti , the dynamical variable
v , and the evolution function A(r, v ). Therefore, a given set of initial matter
distributions and the dynamical profiles, including the energy distribution of
the shells, completely determine the tangent at the center to the singularity
curve.
 3.4 Nature of singularities 79

3.4 Nature of singularities

The occurrence of spacetime singularities in a general spacetime framework
in collapse and cosmology is discussed in some detail in the next chapter.
Singularities are the boundary points of the spacetime where the normal
differentiability and manifold structures break down. These are the points
where the energy density as given by the Einstein equation (3.28) above, or
the curvature quantities, such as the scalar polynomials constructed out of
the metric tensor and the Riemann tensor, diverge. One example of such a
quantity is the Kretschmann scalar K = Rijkl Rijkl , which is given in the
dust collapse models by

F 2 FF F2
K = 12 − 32 + 48 . (3.72)
R4 R2 R5 R R6

Such singularities are indicated by the existence of incomplete future or past

directed non-spacelike geodesics in the spacetime that terminate at the singu-
larity. Then, it is required that the curvature quantities stated above assume
unboundedly large values in the limit of approach to the singularity along the
non-spacelike geodesics terminating there, in the case of a genuine spacetime
singularity. If such a condition is satisfied, then the singularity should be
considered to be a physically significant curvature singularity.
It is now possible to examine, given the matter initial data at the initial
surface t = ti , how the final fate of collapse is determined in terms of either
a blackhole or a naked singularity. If there are families of future directed
non-spacelike trajectories reaching faraway observers in spacetime, which
terminate in the past at the singularity, then a naked singularity forms as
the collapse final state. In the other case when no such families exist and
the event horizon forms sufficiently earlier than the singularity to cover it, a
blackhole is formed. This is decided by the causal behavior of the trapped
surfaces developing in the spacetime during the collapse evolution and the
apparent horizon, which is the boundary of the trapped surface region in the
spacetime.
In general, the equation of the apparent horizon in a spherically symmetric
spacetime is given as

g ik R ,i R ,k = 0. (3.73)

Therefore, at the boundary of the trapped region the vector R ,i is null.

Substituting (3.19) in (3.73),

R2 e−2ψ − Ṙ2 e−2ν = 0. (3.74)

80 Spherical collapse

Using (3.31), the equation of apparent horizon can be written as

F
=1 (3.75)
RN −3
in a general N -dimensional spacetime, which gives the boundary of the
trapped surface region of the spacetime. For the usual spacetime with N = 4,
this becomes F = R. If the neighborhood of the center gets trapped prior
to the epoch of singularity, then it is covered and a blackhole results, other-
wise it could be naked when non-spacelike future directed trajectories escape
from it.
Therefore, the important point is to determine if there are any future
directed non-spacelike paths emerging from the singularity. To investigate
this, and to examine the nature of the central singularity at R = 0, r = 0,
consider the equation for the outgoing radial null geodesics that is given by

dt
= eψ−ν . (3.76)
dr
It would be desirable to examine if there are any families of future directed
null geodesics emerging from the singularity, thus causing a naked singularity
phase as the collapse endstate. The singularity occurs at v (ts (r), r) = 0,
that is, at R(ts (r), r) = 0. Therefore, if there are any future directed null
geodesics terminating in the past at the singularity, R → 0 as t → ts along
these curves. Now, writing (3.76) in terms of the variables (u = rα , R),
 
dR 1 −(α−1)  Ṙ ψ−ν
= r R 1 + e , (3.77)
du α R

where α is a positive constant to be fixed later. In order to obtain the expres-

sion of the tangent to the null geodesics emerging in the (R, u) plane, a
particular value of α is chosen such that the geodesic equation is expressed
only in terms of known limits. For example, if X (0) = 0, and the functions
M and h are well-defined for 0 ≤ r ≤ rb and 0 ≤ v ≤ 1, α = 5/3 is cho-
sen. Using (3.31) and considering Ṙ < 0, the null geodesic equation can be
obtained in the form
 ⎛ ⎞
dR 3 R vv 1/2 1− F
= + 1/2 ⎝ √ √ R √  ⎠ . (3.78)
du 5 u R
G G+ H
u

If the future directed outgoing null geodesics do terminate at the singular-

ity in the past with a definite tangent, then at the singularity, the tangent
to the geodesics have dR/du > 0 in the (u, R) plane, and they must have a
finite value. In the case of a massive singularity in dimensions greater than
 3.4 Nature of singularities 81

or equal to four, that is, when F (ts (r), r) > 0 for r = 0, all singularities for
r > 0 are necessarily covered since F/R → ∞, and hence dR/du → −∞.
This is when both the pressures pr and pθ are positive with the energy con-
ditions are satisfied. Therefore, in such a case only the central singularity at
R = 0, r = 0 could be naked.
Hence, the central singularity at r = 0, R = 0 needs to be examined
to determine if it is visible or not, and to investigate if any solutions to
the outgoing null geodesics equation exist that terminate in the past at the
singularity, and that go to faraway observers in the future. The conditions
under which this can happen are to be determined. Also, note that since
the singularity curve and the evolution functions are regular, the limit of the
functions H , G, and F/R at r → 0, t → ts0 can be calculated. From (3.46),
as A(r, v ) is a well-defined function, G(ts0 , 0) = 1. Also, from (3.51) at this
point, H ≈ r2 /v . Calculating this limit on the t = ts0 plane from (3.71), at
the point (ts0 , 0), H = 0. Hence, from (3.31), F/R = 0 in this limit.
Let x0 now be the tangent to the outgoing null geodesics in the (R, u)
plane at the central singularity

R dR

x0 = lim lim = . (3.79)

t→ts r→0 u du
t→ts ;r→0

To find out whether the null geodesic equation admits any solution of x0
that is positive and finite at the central singularity, the values of H , G, and
F/R at (ts0 , 0) can be used in (3.78). Also, (3.71) can be used to get the
value of v  v 1/2 on the v = 0 surface at r = 0 (that is, on the point (ts0 , 0)).
Therefore, solving (3.78),

3
M0 (0)X (0),
3/2
x0 = (3.80)
2
and the equation of the radial null geodesic emerging from the singularity is
given by R = x0 u in the (R, u) plane, or in (t, r) coordinates it is given by

t − ts (0) = x0 r5/3 . (3.81)

It now follows that if X (0) > 0, then x0 > 0, and a radially outgoing
null geodesic emerges from the singularity, giving rise to a naked central
singularity. However, if X (0) < 0, then a blackhole solution exists, as there
will be no such trajectories (see Fig. 3.2). If X (0) = 0, then the next higher
order non-zero term in the singularity curve equation will have to be taken
into account, and a similar analysis has to be carried out by choosing a
different value of α.
It can be seen that the above is both a necessary and sufficient condition
for an outgoing radial null geodesic emerging from the singularity to exist.
82 Spherical collapse

Naked singularity Blackhole

E⬘(r) E (r)

Initial
surface
0(r), p0(r), pr0(r)

Fig. 3.2 Given the regular matter initial data, the collapse can evolve either to
a blackhole or a naked singularity final state, depending on the choice of the rest
of the free functions, such as the velocities E(r) of the collapsing shells and the
allowed dynamical evolutions as given by the Einstein equations.

Assume that such a geodesic does exist and that in the (R, u) plane it is
given by the equation R = x0 u, with x0 > 0. Then, at the central singularity
(R = 0, u = 0), the tangent to such a geodesic must be x0 . Also, this tangent
must be the root of the equation
 ⎛ ⎞
 1/2
dR 3 R vv − F/R
⎝ √ 1√
du
−
5 u
+ 1/2 √ ⎠ = 0 (3.82)
(R/u) G G+ H

at the point (R = 0, u = 0). This is possible if and only if

 3/2
3
x0 = M0 (0)X (0) , (3.83)
2

and for the slope to be defined and positive, X (0) ≥ 0.

Next, to see that X (0) > 0 is a sufficient condition for the existence of an
outgoing radial null geodesic emerging from the singularity, consider the case
that the singularity curve has a positive tangent at the central singularity.
 3.4 Nature of singularities 83

Consider then the curve

 3/2
3
t − ts (0) = M0 (0)X (0) r5/3 . (3.84)
2

Along this curve t → ts0 as r → 0. And, as X (0) > 0, this curve is outgoing
in the sense that t increases as r increases along the curve. The quantity
(−gtt dt2 + grr dr2 ) along this curve can be calculated in the vicinity of the
central singularity. Using (3.71), (3.51), and (3.46), for this curve,

− e2ν(ts0 ,0) dt2 + e−2rA(ts0 ,0) R (ts0 , 0) dr2

2

 3
5 3 2/3 
= M0 (0)X (0) r −dr2 + dr2 = 0. (3.85)
3 2

That is, in the vicinity of the central singularity the curve considered above
is null. Therefore, for any given positive value of the tangent to the singu-
larity curve at the central singularity, it is always possible to find a null and
outgoing curve terminating in the past at the central singularity, making the
singularity naked.
Note that basically it is the geometry of the trapped surfaces and the
apparent horizon that decides the visibility, or otherwise, of the spacetime
singularity. Different kinds of collapse evolutions lead to different trapped
surface configurations, thus leading to visibility, or otherwise, of the final
singularity. For example, while in a homogeneous dust collapse, the trapped
surfaces and apparent horizon form early enough to cover the singularity;
when inhomogeneities are included, the trapping is naturally delayed so as
to allow the singularity to be visible.
Hence, some remarks on the nature of the apparent horizon and its relation
to the visibility, or otherwise, of the singularity are made below. To find the
equation of the apparent horizon near the central singularity, let the time
corresponding to a shell labeled by r entering the apparent horizon, in terms
of the variable v , be vah (r). Then, from (3.75), it can easily be seen that
vah (r) is the root of the equation

r2 M(r, v ) − v = 0. (3.86)

Now, using (3.52), the equation for the apparent horizon in the (t, r) plane
can be written as
 vah (r)
v 1/2 dv
tah (r) = ts (r) −  . (3.87)
0 e(rA+ν) vb0 + e2ν (vh + M)
It is obvious that the necessary condition for the existence of a locally naked
singularity is that the apparent horizon curve must be an increasing function
84 Spherical collapse

at the central singularity, in the lowest power of r. If the apparent horizon

curve is decreasing at the singularity, the collapse outcome is necessarily a
blackhole.
In the above, the functions h and M are expanded with respect to r around
r = 0 and the first order terms are considered. At times, however, these are
assumed to be expandable with respect to r2 , and it is argued that such
smooth functions would be physically more relevant. Such an assumption
comes from the analyticity with respect to the local Minkowskian coordi-
nates (see for example, the discussion in Goswami and Joshi, 2004a, 2004b),
and it is really the freedom of definition mathematically. The formalism,
as discussed above, would also work for such smooth functions, which is a
special case of the above discussion.
It can therefore be seen how the initial data, in terms of the free functions
available, determine the blackhole and naked singularity phases as the final
outcome for the collapse. This is because the quantity X (0) is determined by
these initial and dynamical profiles, as given by (3.54). It is clear, therefore,
that given any regular initial density and pressure profiles for the matter
cloud from which the collapse develops, velocity profiles can always be chosen
so that the endstate of the collapse would be either a naked singularity or a
blackhole, and vice versa.
Numerical work on collapse models may provide further insights into
these interesting dynamical phenomena, especially when the collapse is non-
spherical, which remains a major open problem to be considered. Numerical
and some analytical work has been carried out in recent years on spherical
scalar field collapse, and also on some perfect fluid models. While the emer-
gence of the null geodesics from the singularity, thus showing it to be naked
(or otherwise), has been worked out explicitly here in an analytic manner,
the numerical simulations generally discuss the formation, or otherwise, of
the trapped surfaces and apparent horizon. Such considerations may possibly
break down closer to the epoch of the actual singularity formation. In this
case, the actual detection of the blackhole or naked singularity endstates
may not be allowed for, whereas important insights into critical phenom-
ena and dispersal have been gained through numerical methods (Choptuik,
1993). Probably, a detailed numerical investigation of the structure of null
geodesics in collapse models may provide further insights here.
Note that, in the above consideration, the occurrence of a locally naked
singularity only, as opposed to that of a globally naked singularity, has been
deduced. That is, when and in what circumstances the null geodesics escape
from the spacetime singularity going out in the future have been shown, but
the question of when these actually go out of the boundary of the matter
cloud has not been addressed. It is possible, in principle, that the singularity
is only locally visible and trajectories do come out, but then they all fall back
into the singularity again at a later time, without going out of the boundary
 3.4 Nature of singularities 85

Event horizon

Outgoing null
geodesics

Initial surface

Fig. 3.3 Local versus global visibility. If the singularity is only locally visible,
the light rays come out, but they then fall back again at the center without
coming out of the cloud. On the other hand, for a globally visible singularity,
the outgoing rays reach the boundary of the cloud and can reach faraway external
observers.

of the star, thus not allowing the ultra-dense regions to be visible to faraway
external observers (see Fig. 3.3).
This issue has still not been studied for the general class of models such
as the ones considered here that may be of interest. However, for the case
of dust collapse models, this has been studied in some detail (Joshi and
Dwivedi, 1993a), and it is shown that, whenever the singularity is locally
naked, one can always choose the classes of the mass and energy functions
suitably, as one moves away from the center for the larger values of the radial
coordinate r, in such a manner that the singularity becomes globally visible.
The point is, while the local visibility of the central singularity is basically
decided by the conditions near the center, the global visibility depends on the
overall behavior of these functions within the matter cloud, away from the
center. This behavior can still be freely chosen at larger values of r. In other
words, for the dust collapse models, once the singularity is locally visible,
there are always classes of mass and energy functions which can be chosen
in order to make it globally visible. Another important related point is, as
such, there is no scale in the problem, and the size of the collapsing cloud
could be quite large. In such a case, even if the singularity is only locally
visible, it can still be seen for a long enough time by the observers. Therefore,
in principle, a locally naked singularity is also as serious a violation of the
cosmic censorship as a globally visible singularity, and there may not be a
qualitative difference in the two cases in many situations of physical interest
(Penrose, 1979).
86 Spherical collapse

An important and interesting issue frequently mentioned regarding the

occurrence of naked singularities in gravitational collapse is their genericity
and stability. It is argued that if these are not generic or stable, then they
need not be taken seriously. This is a complex issue, because in general rel-
ativity there are no well-defined notions and criteria available for stability
or genericity that can then be applied and tested for a given model. All the
same, a consideration of this issue is indeed important in that, depending
on the collapse situation under discussion, these notions could be formu-
lated in some way or other, to examine if the naked singularities developing
as collapse endstates are ‘generic’ or ‘stable’ in some suitable sense. This
would typically involve taking into account the topology and metric of the
function spaces concerned that define the given collapse scenario. This shall
be discussed in some detail in the next chapter. Presently, however, it may
be noted that, since the spherical collapse has been formulated here in a
general manner for general physically reasonable matter fields, the classes
constructed may provide a good arena to explore and test these important
issues for the cosmic censorship hypothesis.
A related issue is that of non-spherical collapse. It can be asked if the
conclusions available for a spherically symmetric collapse remain the same
and stable under possible non-spherical perturbations. Even though some
non-spherical collapse models have been discussed and investigated, such
as the Szekeres quasi-spherical collapse or some cylindrical collapse models,
these may be regarded as somewhat restrictive in nature. It is not clear as yet
if this issue can be approached in an analytic manner, and detailed numerical
simulations of the collapsing stars could possibly suggest an answer.
Another question that is frequently asked in connection with the occur-
rence of naked singularities as collapse endstates is how to understand this
phenomena physically. A naked singularity signifies the escape of light and
particle trajectories from the ultra-dense spacetime regions. However, gravity
must become very strong in these regions. In such a case, how can anything
escape at all from such a region? Therefore, while a blackhole, which is a
region from which not even light would escape, may appear to be the only
physically reasonable outcome in such situations, the formation of a naked
singularity in the collapse may appear to be counter-intuitive. The point
that comes out from such considerations is that the naked singularities are
more an artifact of general relativity, rather than that of purely Newtonian
physics. Even though the matter density grows higher and higher without
bound and blows up closer to a spacetime singularity, which would denote
the growth of attractive forces of gravity, there are other important purely
general relativistic effects that can delay the formation of trapped surfaces
and apparent horizons, which govern the trapping of light.
An interesting effect that does this is the inhomogeneities and related
spacetime shear. It is intriguing to find that the physical agencies, such
 3.5 Exterior geometry 87

as the spacetime shear and inhomogeneities in matter density distributions

within a dynamically collapsing cloud, could naturally delay the formation
of trapped surfaces during a gravitational collapse. An explicit example of
this will be discussed in Chapter 5. In other words, such physical factors
do naturally give rise to naked singularity phases in a collapse, where the
formation of the apparent horizon and the trapped surfaces is delayed. Even
though the matter densities are arbitrarily large and growing, the shearing
effects could distort the trapped surface geometry in such a manner so as
to avoid the trapping of light, and facilitate the escape of null rays from
such ultra-dense regions. It is thus pointed out how the blackhole and naked
singularity endstates arise naturally in a spherical collapse, as governed by
the geometry of the trapped surfaces.

3.5 Exterior geometry

To complete the full spacetime model, the interior spacetime of the dynamical
collapse needs to be matched to a suitable exterior geometry. As modeling
the collapse of astrophysical objects (such as massive stars) is of interest
here, the collapsing cloud is taken to have a compact support at the ini-
tial surface, with the boundary of the cloud being at some r = rb . If the
pressures at the boundary of the cloud vanish, then it is always possible to
match the interior collapsing spacetime with an empty Schwarzschild exte-
rior. However, in all cases in general, the pressures need not vanish at the
boundary of the cloud. For example, for a dust cloud collapse, the pressure
is zero by assumption, and the collapse can be matched at the boundary to
an exterior Schwarzschild solution. But, for a perfect fluid with an equation
of state p = kρ, the pressure does not have to vanish at the boundary of
the cloud.
Hence in such cases, the general practice is that there would be a boundary
layer, to which the internal geometry is matched, and which could in turn
be matched to an exterior Schwarzschild geometry. This is also appropriate
from an astrophysical perspective, because the exteriors of any realistic stars
would not be really completely empty and vacuum, but would always be
surrounded by a radiation zone, as well as the matter emissions from the
star. Such emissions may be particularly important in gravitational collapse
situations, where the massive star may be emitting matter and radiation
quite significantly as it collapses.
Hence, outlined below is the procedure to match the interior with a general
class of exterior metrics, which is the generalized Vaidya spacetime (Joshi
and Dwivedi, 1999; Wang and Wu, 1999) at the boundary hypersurface Σ
given by r = rb . For the required matching, the Israel–Darmois match-
ing conditions (Israel, 1966a, 1966b) are used, where the first and second
88 Spherical collapse

fundamental forms are matched. That is, the metric coefficients and the
extrinsic curvature are matched at the boundary of the cloud.
Whereas the procedures used below are standard, the particular case
treated here will be described in some detail so as to give the exact pic-
ture of the overall collapse scenario emerging. Note that since the matching
is for the second fundamental form Kij , there is no surface stress energy, or
surface tension at the boundary (see for example, Mazur and Mottola, 2004).
The metric just inside Σ is

ds2− = −e2ν dt2 + e2ψ(t, r) dr2 + R2 (t, r)dΩ2 , (3.88)

which describes the geometry of the collapsing cloud. The metric in the
exterior of Σ is given by
 
2M (rv , V )
ds2+ =− 1− dV 2 − 2dV drv + rv2 dΩ2 , (3.89)
rv

where V is the retarded outgoing null coordinate and rv is the Vaidya radius.
Matching the area radius at the boundary results in

R(rb , t) = rv (V ). (3.90)

Then, on the hypersurface Σ, the interior and exterior metrics are given by

ds2Σ− = −e2ν(t,rb ) dt2 + R2 (t, rb )2 dΩ2 (3.91)

and
 
2M (rv , V ) drv
ds2Σ+ =− 1− +2 dV 2 + rv2 dΩ2 . (3.92)
rv dV

Matching the first fundamental form gives

 
dV eν(t,rb )
=  , (rv )Σ = R(t, rb ). (3.93)
dt Σ 1− 2M (rv ,V )
+ 2 dr v
rv dV

Next, to match the second fundamental forms (extrinsic curvatures) for

the interior and exterior metrics, note that the normal to the hypersurface
Σ, as calculated from the interior metric, is given as
 
ni− = 0, e−ψ(rb ,t) , 0, 0 , (3.94)
 3.5 Exterior geometry 89

and the non-vanishing components of the normal as derived from the

generalized Vaidya spacetime are

1
nV+ = −  , (3.95)
2M (rv , V ) drv
1− +2
rv dV

2M (rv , V ) drv
1− +
rv dV
nr+v =  . (3.96)
2M (rv , V ) drv
1− +2
rv dV

Here, the extrinsic curvature is defined as

1
Kab = Ln gab . (3.97)
2

That is, the second fundamental form is the Lie derivative of the metric
with respect to the normal vector n. The above equation is equivalent to

1 
Kab = gab,c nc + gcb nc,a + gac nc,b . (3.98)
2
 − 
Now, setting Kθθ − Kθθ
+
Σ
= 0 on the hypersurface Σ,

2M (rv , V ) drv
1− +
rv dV
RR e−ψ = rv  . (3.99)
2M (rv , V ) drv
1− +2
rv dV

Simplifying the above equation using (3.93) and the Einstein equations,

F (t, rb ) = 2M (rv , V ). (3.100)

Using the above equation and (3.93),

 
dV eν (R e−ψ + Ṙe−ν )
= . (3.101)
dt Σ 1− F (t, rb )
R(t, rb )

Finally, setting [Kτ−τ − Kτ+τ ]Σ = 0, where τ is the proper time on Σ,

F Re−ν √
M (rv , V ),rv = + √ H ,t + Re2ν ν  e−ψ. (3.102)
2R G
90 Spherical collapse

r=0
Generalized
Vaidya
metric
Outgoing
Interior radiation
geometry

Fig. 3.4 For a general matter field, the interior collapsing metric is matched to
an exterior that is a generalized Vaidya geometry.

Any generalized Vaidya mass function M (v, rv ) that satisfies (3.102) will
then give a unique exterior spacetime with required equations of motion
given by other matching conditions, (3.100), (3.101), and (3.90).
To see that the set of all such functions M (v, rv ) is non-empty, the
examples of a charged Vaidya spacetime M = M (V ) + Q(V )/rv , and the
anisotropic de Sitter spacetime M = M (rv ) are two different solutions of
(3.102) (see for example, Joshi and Dwivedi, 1999; Giambo, 2005). This
gives two unique exterior spacetimes, both of which are subclasses of the
generalized Vaidya metric (see Fig. 3.4).
It is of course also possible to treat this problem from the perspective of
only pure general relativity, and the theoretical cosmic censorship aspect. In
this case, there is no need to cut off the cloud and match it at the boundary.
Suitable fall off conditions can just be imposed for the collapsing matter,
so that faraway the metric becomes Minkowskian at the spatial infinity (see
for example, Choptuik, 1993, for the case of numerical scalar field collapse
models).

3.6 Dust collapse

The dust collapse models have been analyzed extensively towards under-
standing the final fate of gravitational collapse in Einstein theory, and these
can be used to illustrate the ideas above. This class of models have played a
very important and fundamental role in gravitation theory because it is, in
fact, at the very heart of the blackhole paradigm. As pointed out earlier, after
the investigation of Oppenheimer and Snyder (1939), which showed that a
blackhole developed as the final state of a continual gravitational collapse
 3.6 Dust collapse 91

of a homogeneous dust cloud, the cosmic censorship hypothesis was intro-

duced by Penrose in 1969 to suggest that this is the generic outcome for
a collapse in general relativity for any massive star. This conjecture paved
the way for many further developments in blackhole physics. Based on the
assumption of censorship, namely that just as in the case of homogeneous
dust clouds, all stars that are gravitationally collapsing must end up nec-
essarily as blackholes only, much work was carried out on blackhole theory
and its astrophysical applications, while any rigorous formulation and proof
of censorship hypothesis is still awaited.
Though any mathematical formulation or proof of censorship is not avail-
able as yet, further progress on the better understanding of a dust collapse
came only later with work such as the numerical study of Eardley and Smarr
(1979), and the analytic treatment of Christodoulou (1984) and Newman
(1986). A generic analysis of the inhomogeneous dust collapse with general
classes of initial mass and velocity profiles was given by Joshi and Dwivedi
(1993a). Gravitational waves in these models were analyzed by Iguchi, Nakao,
and Harada (1998).
In the past couple of decades, there have been many investigations on
this important class of collapse models, as is discussed below, which has an
interesting mathematical structure as well as a definite physical significance.
The main reason why such a definite progress for these collapse models could
be made was that the full solution of the Einstein equations for the dust
class of spacetimes was already given much earlier by Lemaı̂tre (1933) and
Tolman (1934), and was also discussed by Bondi (1948), which included the
homogeneous dust collapse as a special case. These solutions are known as
the Tolman–Bondi–Lemaı̂tre (TBL) dust collapse models.
As a result of this analysis, it has become clear now that, as soon as one
departs from the assumption of the exact homogeneity of the density distri-
bution in space for the collapsing dust cloud, the final collapse outcome is
no longer necessarily a blackhole. However, the visible ultra-dense regions or
naked singularities could also then develop generically as collapse endstates.
The actual outcome in terms of either a blackhole or naked singularity is
decided by the nature of the regular initial data, in terms of the initial den-
sity and velocity profiles, from which the collapse evolves (see Fig. 3.5, which
shows the marginally bound case).
Here, the occurrence and nature of naked singularities for the inhomoge-
neous gravitational collapse of TBL dust clouds is investigated as a special
case of the discussion so far. It is shown that visible ultra-dense regions form
at the center of the collapsing cloud in a wide class of these models. The
earlier cases considered by Eardley and Smarr (1979), Christodoulou (1984),
and Newman (1986) are included as special cases in the general treatment
given here. This class also contains self-similar, as well as non-self-similar
models.
92 Spherical collapse

 

Space Space
(a) (b)

Fig. 3.5 The nature of the initial density profile affects the final outcome of
collapse. For a constant density (a) a blackhole develops, but for a profile slightly
higher at the center of the cloud (b) the final state is a naked singularity.

It is possible to criticise the dust collapse models in that they incorporate

no pressure, or that they are spherically symmetric. While generalizations are
discussed above and also later, the important point is that this is the model
that has actually inspired the entire blackhole paradigm, as originating from
the Oppenheimer–Snyder work, and hence it deserves a serious and complete
study. As pointed out, it also answers several important questions regarding
the possible endstates of a continual gravitational collapse, and helps to rule
out several possible or plausible formulations of cosmic censorship hypothesis
that have been tried out earlier. Furthermore, there are also arguments to
suggest that dust could possibly be the physically reasonable and realistic
equation of state in the very final stages of gravitational collapse (Hager-
dorn, 1968; Penrose, 1974a), and that at higher and higher densities the
matter may really behave more like the dust equation of state. Again, there
is some case for the argument that, eventually in the final stages of collapse,
the matter distribution should become almost spherically symmetric (see for
example, Nakamura and Sato, 1982). All these factors make a detailed inves-
tigation of dust collapse quite interesting and important. Hence, it is clearly
useful to examine the inhomogeneous dust collapse as modeled by the TBL
spacetimes. Furthermore, a situation analogous to the singularity theorems
might develop here where the conclusions derived under the assumption of
spherical symmetry are preserved, when small perturbations are taken into
account. Thus, spherical symmetry may be a good ansatz to represent certain
important classes of gravitational collapse scenarios.

3.6.1 Tolman–Bondi–Lemaı̂tre (TBL) spacetimes

The TBL metric represents a general solution to the Einstein equations
with a spherically symmetric dust form of matter. In particular, it gives the
geometry for the inhomogeneous dust cloud collapse, and in the comoving
 3.6 Dust collapse 93

coordinates, that is, with ui = δti , the geometry is

R2
ds2 = −dt2 + dr2 + R2 (dθ2 + sin2 θ dφ2 ). (3.103)
1+f

The energy–momentum tensor is that of pressureless dust, as given by

T ij = δti δtj . (3.104)

The Einstein equations then give

F
 = (t, r) = , (3.105)
R2 R
where  is the energy density, R is the area radius of the cloud, which is a
function of both t and r, and

F
Ṙ2 = + f. (3.106)
R
Here, the dot and the prime denote partial derivatives with respect to the
parameters t and r respectively.
As the continual gravitational collapse situation is of concern here, it is
required that Ṙ(t, r) < 0. The quantities F and f are arbitrary functions
of r, as obtained from the integration of the Einstein equations, and, as
seen from the above equations, these have the interpretation of the mass and
velocity functions respectively. The quantity 4πR2 (t, r) gives the proper area
of the mass shells and the area of such a shell at r = const. goes to zero when
R(t, r) = 0. Integration of (3.106) gives

R3/2 G(−f R/F )

t − t0 (r) = − √ , (3.107)
F
where G(y ) is a strictly real positive and bounded function that has the
range 1 ≥ y ≥ −∞, and is given by
 −1 √ √ 
sin y 1−y
G(y ) = − for 1 ≥ y > 0, (3.108)
y 3/2 y
2
G(y ) = for y = 0, (3.109)
3
 √ √ 
− sinh−1 −y 1−y
G(y ) = − for 0 > y ≥ −∞. (3.110)
(−y )3/2 y

The function t0 (r) is a constant of integration. So, there are in all three
arbitrary functions of r, as given by f (r), F (r), and t0 (r). However, the
94 Spherical collapse

remaining coordinate freedom left could be used in the choice of scaling r

in order to reduce the number of such arbitrary functions to two. Therefore,
the area radius R is rescaled using this coordinate freedom, such that

R(0, r) = r. (3.111)

Then, t0 (r) is evaluated by using (3.111) and (3.107) to give

r3/2 G(−f r/F )

t0 (r) = √ . (3.112)
F

The time t = t0 (r) corresponds here to the value R = 0 where the area of
the shell of matter at a constant value of the coordinate r vanishes. It follows
that the singularity curve t = t0 (r) corresponds to the time when the matter
shells meet the real physical singularity, which corresponds to the entire
cloud shrinking to a zero radius. Therefore, the range of the coordinates is
given by
0 ≤ r < ∞, −∞ < t < t0 (r). (3.113)

The case considered by Oppenheimer and Snyder (1939), which discussed the
homogeneous dust cloud collapse, is a special subcase of the above general
dust solution, corresponding to  = (t) only throughout, and with a special
value of the velocity function as given by f (r) = 0. The above comoving
coordinates, tailored to the collapsing matter, which have a somewhat spe-
cial physical significance, have been used here, and these provide a physical
insight into the collapse processes. In the case of a homogeneous density, the
interior of the cloud has a special, more simple form, which is the same as
the time reversed case of the Friedman solution in cosmology.
It is seen, in general, that when the matter density is allowed to be inho-
mogeneous and can change as a function of the coordinate radius r, then,
unlike the collapsing Friedmann case (where the physical singularity R = 0
occurs at a constant epoch of time, say, at t = t0 ), the singular epoch now is
a function of r, as a result of the inhomogeneity in the matter distribution.
The Friedmann case could be recovered from the above equations by setting,
for example, t0 (r) = t0 (r) = 0. This corresponds to the case of a simulta-
neous singularity. It follows that, for an inhomogeneous matter distribution,
different shells of matter arrive at the singularity at different times as given
by ts (r), and a singularity curve rather than a constant singular epoch is
obtained, as was the case for a constant density distribution.
The function f (r) above can be used to classify the spacetime as bound,
marginally bound, or unbound, depending on the range of its values, which are

f (r) < 0, f (r) = 0, f (r) > 0 (3.114)

 3.6 Dust collapse 95

√ F (r) is interpreted as the weighted mass, as

respectively. The function
weighted by the factor 1 + f , within the dust ball B of coordinate radius
r that is conserved in the sense
  r
F (r) 1/2
m(r) = = (1 + f ) (t, r)dv = 4π ρ(r)r2 dr, (3.115)
2 B 0

where (0, r) = ρ(r) is the initial density distribution, from which the grav-
itational collapse of the cloud develops. For physical reasonableness, the
weak energy condition would be assumed throughout the spacetime, that is,
Tij V i V j ≥ 0 for all non-spacelike vectors V i . This implies that the energy
density  is positive everywhere ( ≥ 0), including the region near the center
of the cloud r = 0.
The partial derivatives of R, as given by R and Ṙ, play an important role
in the analysis, and, from the above equations,
     1/2 
3 1
R = rα−1 (η − β )X + Θ − η − β X 3/2 G(−P X ) P+
2 X
≡ rα−1 H (X, r), (3.116)
   1/2   
Λ1/2 1 3
Ṙ = 2
−βX 2
+ P + Θ − η − β X 3/2 G(−P X )
2rX X 2
−N (X, r)
≡ , (3.117)
r
where, using the notation given by Joshi and Dwivedi (1993a),

F f
X = (R/rα ), η = η (r) = r , β = β (r) = r , p = p(r) = rf /F,
F f
(3.118)
F
P = prα−1 , Λ= , (3.119)
rα
and
√
t0 Λ 1+β−η (η − 32 β )G(−p)
Θ ≡ α−1 = + . (3.120)
r (1 + p)1/2 r3(α−1)/2 r3(α−1)/2
The function β (r) is defined to be zero when f is constant and zero. The
factor rα has been introduced here for the sake of convenience in examin-
ing the structure of the naked singularity. The exact value of the positive
constant α ≥ 1 is to be determined and chosen later, and it will depend
on the different models of the spacetime that allow naked singularities. The
functions H (X, r) and N (X, r) are defined by (3.117) and (3.118).
96 Spherical collapse

Using the scaling given by R = r as used above, the energy density  on

the initial hypersurface t = 0 can be written as  = F  /r2 . Since the weak
energy conditions are satisfied and the mass function F is a function of r
only, it follows that F  ≥ 0 throughout the spacetime. The energy density
can be written as
ηΛ
= 2 . (3.121)
R H
Since F  = η Λrα−1 , it follows that H (X, r) ≥ 0 everywhere and η Λ ≥ 0
everywhere as a consequence of the weak energy condition.
In TBL spacetimes singularities occur, as can be seen from the Einstein
equations, at the points where R = 0, which are the shell-focusing singu-
larities, and also at the points where R = 0. At the points where R = 0,
the TBL metric is degenerate and the points R > 0, F  > 0, where R = 0,
are called the shell-crossing singularities, where the nearby shells of matter
cross momentarily (Hellaby and Lake, 1985; Newman, 1986). Such shell-
crossings in TBL spacetimes have been analyzed in detail and their nature
appears to be fairly well understood. Even though the shell-crossing singu-
larities could be locally naked, the important point is that they have been
shown to be gravitationally weak (Newman, 1986). Therefore, it is generally
believed that shell-crossing singularities need not be taken seriously as far
as the cosmic censorship conjecture is concerned. It was pointed out earlier
that the absence of shell-crossings in the spacetime turns out to be related to
the condition that the function t0 (r), giving the proper time for the shells to
fall into the physical singularity, should be a monotonically increasing func-
tion. That is, the shells with increasing r arrive one after the other at the
singularity. The dust density and certain components of the curvature blow
up near such a singularity. However, the causal structure of the spacetime
can be extended through this singularity and the spacetime metric can be
defined also in the neighborhood of such a point in a distributional sense
(Papapetrou and Hamoui, 1967). In the context of such a situation, it is
assumed that there are no shell-crossing singularities in the TBL spacetime,
except probably right at the center of the cloud at r = 0. Instead, the more
serious shell-focusing singularity at R = 0 where the entire cloud collapses
to a zero physical radius is considered, and a physical spacetime singularity
develops.
This does not involve any loss of generality, either for the TBL mod-
els, or the general models discussed earlier, as the basic purpose here is to
examine the formation and local structure of the shell-focusing singularity at
the center of the collapsing cloud. Whereas the existence of shell-crossings
will not affect the qualitative nature of these general conclusions, the above
assumption allows the calculations to be presented in a more transparent
manner. Unlike the shell-crossings, the spacetime metric, however, admits
no extension through a shell-focusing singularity occurring at R = 0, where
 3.6 Dust collapse 97

all the matter shells reduce to a zero size, and therefore this is a genuine
physical singularity. Hence, the occurrence of such shell-focusing singulari-
ties at the center of the collapsing dust cloud are investigated here, and their
nature and structure for the TBL spacetimes are examined. It is clear that
a shell-focusing singularity occurring at r > 0, R = 0 is totally spacelike and
therefore the discussion would be confined to the singularity at r = 0.
The equation for the shell-focusing singularity R(t0 , r0 ) = 0 is as given
earlier, and occurs at r = r0 at the coordinate time t = t0 . The singularity
is called a central singularity if it occurs at r = 0. This central shell-focusing
singularity can be naked, though gravitationally weak (Newman, 1986), for
the class of TBL models considered by Christodoulou (1984), for which the
energy density, which is assumed to be positive everywhere and taken to be
non-zero at r = 0, and the metric functions are even and smooth functions of
t and r. Translated in terms of the parameters defined above, this corresponds
to the case for which η (0) = 3, β (0) = 2, and p(r) is an even smooth function
of r. In terms of the functions F (r) and f (r), this amounts to the conditions

F (r) = r3 F (r), ∞ > F (0) > 0, 0 < p(r) ≤ 1. (3.122)

It was, however, pointed out by Waugh and Lake (1988, 1989) and Ori and
Piran (1987, 1990) that this class of gravitationally weak naked singularities
excludes the self-similar TBL models, where the singularity was shown to
be gravitationally strong along the Cauchy horizon, which is a null geodesic
coming out of the singularity. It was further pointed out that the strong
curvature naked singularity is not necessarily confined to the self-similar
models only, but that there are many classes of TBL models that are non-self-
similar, and the naked singularity occurring there is gravitationally strong
(Dwivedi and Joshi, 1992). In these classes, the collapse terminates in a naked
singularity that is gravitationally strong for a wide range of TBL models that
are non-self-similar in general, and which include all the self-similar models as
a special subclass. In the notation used here, these models are characterized
by the conditions η (0) = 1, with F (r) and f (r) being analytic at r = 0.
Rather general differentiability conditions can be required only on the
functions F (r) and f (r) in that they are assumed to be at least C 1 at the
center r = 0, ∞ > η (0) > 0, and β (0) is finite. In order to ensure the
metric to be C 2 , which is the basic requirement in general relativity, these
differentiability conditions imposed here on the functions f and F are suffi-
cient. Frequently, much stronger conditions are imposed on these functions
and the metric, requiring them to be smooth and analytic. However, these
are considered here to be too strong in general. Anyway, the models with
stronger conditions become special cases of the above. Furthermore, from the
physical reasonableness, one would require F (0) = 0 (otherwise there will
be a massive singularity present already at r = 0), which implies η (0) > 0.
98 Spherical collapse

The condition F being C 1 corresponds to the energy density  not diverging

at the center r = 0 at all times. Note that the function f , and also its first
derivatives (through R ) enter the metric potentials. It might actually be
argued that the above are more general conditions than should actually be
required, as it is customary to assume that the metric is C 2 -differentiable
(which again ensures the above requirements), so that the metric transfor-
mations and other functions connected with the metric are well-defined to
carry out regular physics. Hence, the differentiability requirements here may
be considered physically reasonable and rather general, and which include
practically all the inhomogeneous collapse TBL dust models of interest. In
fact, it could be argued that if the metric is not C 2 -differentiable, but say
only C 1 on the initial surface, it may be considered as being already singular
and not defining a regular initial data on an initial spacelike hypersurface.
In order to represent the gravitational collapse scenario, the energy density
 is taken to have a compact support on an initial spacelike hypersurface. The
TBL spacetime metric above can then be matched at some r = const. = rc
to the exterior Schwarzschild field, and
 
2M drs2
ds = − 1 −
2
dT 2 + + rs2 dΩ2 , (3.123)
rs 1 − 2M/rs

where dΩ2 = dθ2 + sin2 θ dφ2 . The value of the Schwarzschild radial coordi-
nate is rs = R(t, rc ) at the boundary r = rc . Also, m(rc ) = M , where M is the
total Schwarzschild mass enclosed within the dust ball of coordinate radius
r = rc .
The apparent horizon, which indicates the boundary of all trapped surfaces
that define the region that allows no light to escape, forming in the spacetime
during the gravitational collapse in the interior dust ball, lies at R = F (r).
The corresponding time t = tah (r) for the apparent horizon is given by

r3/2 G(−p)
t = tah (r) = √ − F G(−f ). (3.124)
F

The emissions from the shell-focusing singularity R(t0 , r0 ) = 0 for all r0 > 0
would lie in the region above t = tah , that is, t0 > tah for all r0 > 0, with t0
being the time when the singularity at r = r0 occurs. Hence, all radiations
would be future trapped from the shell-focusing singularities at r > 0. At
r = 0 however, t(0) = tah (0) and the singularity could be at least locally
visible. Any light ray terminating at this singularity in the past goes to the
future infinity if it reaches the surface of the cloud r = rc earlier than the
apparent horizon at r = rc . In such a case the singularity would be globally
naked. The nature of this central spacetime singularity has to be examined
in terms of its visibility or otherwise.
 3.6 Dust collapse 99

As seen from the Einstein equations above in the dust case, once the
mass function is specified at the initial epoch, the energy function f (r) fully
specifies the velocity distribution of the in-falling shells. Also, the energy
condition then implies that F  ≥ 0 and the collapsing matter cloud condition
implies that Ṙ < 0. For dust clouds it follows from the equations of motion
that once Ṙ < 0 at the initial epoch from where the collapse commences,
then, at all epochs, the same condition holds, and thus there is a continual
collapse without any reversal until the shell-focusing singularity at R = 0
is reached. In other words, there is no bounce possible in the dust collapse
once the collapse has initiated with Ṙ < 0.
The scaling independence of the comoving coordinate r can again be used
to give

R(t, r) = rv (t, r), (3.125)

where

v (ti , r) = 1, v (ts (r), r) = 0, v̇ < 0. (3.126)

This means the coordinate r has been scaled in such a way that, at the
initial epoch, R = r, and at the singularity, R = 0. Therefore, R = 0 both
at the regular center r = 0 of the cloud, and at the spacetime singularity
where all matter shells collapse to a zero physical radius. The regular center
is then distinguished from the singularity by suitable behavior of the mass
function F (r), so that the density remains finite and regular there at all
times until the singular epoch. The introduction of the parameter v then
allows the spacetime singularity to be distinguished from the regular center,
with v = 1 all through the initial epoch, including the center r = 0, and
which then monotonically decreases with time as the collapse progresses to
the value v = 0 at the singularity R = 0.
From the equations of motion, it is evident that to have a regular solution
over all space at the initial epoch, the two free functions F (r) and f (r) must
have the following forms

F (r) = r3 M(r), f (r) = r2 b(r), (3.127)

where M(r) and b(r) are at least C 1 -functions of r for r = 0, and at least
C 2 -functions for r > 0. This is dictated by the condition that the density
and energy distributions must be regular at the initial epoch and should not
blow up. This is because if the mass function F did not go as a power of at
least r3 closer to the origin, then, as implied from the equations of motion,
the density will be singular at the origin r = 0 as it will diverge there. This
cannot be accepted as regular initial data for the collapse. Similarly, (3.107)
implies that f (r) is determined once the velocity profile is specified, and vice
100 Spherical collapse

versa, for a given initial density distribution. Since the center of the cloud
is taken to be at rest in any spherically symmetric distribution, the leading
term in the energy profile must be at least r or higher. Then again, (3.107)
implies the behavior for f (r) as above.

3.6.2 Collapse endstates

The continual collapse of a dust cloud to a final shell-focusing singularity
at R = 0, where all matter shells collapse to a zero physical radius, is now
considered. In particular, the nature of the central singularity at R = 0, r = 0
is specifically analyzed in detail to determine when it will be covered by
the event horizon, and when it is visible and causally connected to outside
observers. If there are future directed families of non-spacelike curves coming
out from the singularity and reaching faraway observers, then the singularity
will be naked. The absence of such families will give a covered case when the
result is a blackhole.
In the following, the collapse as discussed above is considered and it will
be shown how the initial density and energy distributions, prescribed at the
initial epoch from which the collapse commences, completely determine if
there will be families of non-spacelike trajectories emerging from the singu-
larity. It will be shown that for a generic situation, given an initial density
distribution for the collapse to develop, it is always possible to choose an
energy profile so that the collapse of this density profile ends in a naked
singularity. Alternatively, another class of energy distribution could also be
chosen so that the same density distribution will end up collapsing in order
to create a blackhole. That is, given an initial density profile, the outcome in
terms of either a blackhole or a naked singularity as endstates really depends
on the class of the energy distributions chosen. The converse will also be
seen to hold true, that is, given the initial energy function, classes of density
profiles, subject to the weak energy condition can be chosen so as to give
rise to either the blackhole or naked singularity endstates, depending on the
choice made. These results generalize naturally to any higher-dimensional
dust collapse models in general relativity.
With the regular initial conditions as above, (3.52) can be written as

v 1/2 v̇ = − M(r) + vb(r). (3.128)

Here, the negative sign implies that v̇ < 0, that is, the matter cloud is
collapsing. Integrating the above equation with respect to v ,
 1
v 1/2 dv
t(v, r) =  . (3.129)
v M(r) + vb(r)
 3.6 Dust collapse 101

Note that the coordinate r is to be treated as a constant in the above

equation. Expanding t(v, r) around the center,

X2 (v ) X3 (v )
t(v, r) = t(v, 0) + rX (v ) + r2 + r3 + ···, (3.130)
2 6
where the function X (v ) is given by

1 1
v 1/2 (M1 + vb1 )dv
X (v ) = − , (3.131)
2 v (M0 + vb0 )3/2

where

b0 = b(0), M0 = M(0), b1 = b (0), M1 = M (0). (3.132)

Therefore, the time taken for the central shell to reach the singularity is
given by
 1
v 1/2 dv
ts0 = √ . (3.133)
0 M0 + vb0
From the above equation it is clear that for ts0 to be defined,

M0 + vb0 > 0. (3.134)

In other words, the continual collapse condition implies the positivity of the
above term. Hence, the time taken for other shells to reach the singularity
can be given by the expansion

ts (r) = ts0 + rX (0) + O(r2 ). (3.135)

Also, from the equation for v̇ above and (3.131), for small values of r along
constant v surfaces,

v 1/2 v  = (M0 + vb0 ) (X (v ) + rX2 (v ) + · · · ) . (3.136)

Now, it can easily be seen that the value of X (0) depends on, and is com-
pletely characterized by, the functions M (r) and b(r), which in turn specify
fully the initial mass and energy distributions for the collapsing matter. Spec-
ifying these functions is equivalent to specifying the regular initial data for
the collapse on the initial surface t = 0. In other words, a given set of density
and energy distributions completely determines the slope to the singularity
curve at the origin, which is the central singularity. Also, it is evident that,
given any one of these two profiles, the other one can always be chosen in
such a manner that the quantity X (0) will be either positive or negative.
102 Spherical collapse

In order to determine the visibility, or otherwise, of the central singularity,

the behavior of the non-spacelike curves in the vicinity of the singularity and
the causal structure of the trapped surfaces now need to be analyzed.
The boundary of the trapped surface region of the spacetime is given
by the apparent horizon within the collapsing cloud, which is given by the
equation
F
= 1. (3.137)
R
One now needs to determine when there will be families of non-spacelike
paths coming out of the singularity, reaching faraway observers, and when
there will be none. The visibility, or otherwise, of the singularity is decided
accordingly. Broadly, it can be stated that if the neighborhood of the center
gets trapped earlier than the singularity, then it is covered, otherwise it
is naked with families of non-spacelike future directed trajectories escaping
away from it. By determining the nature of the singularity curve and its
relation to the initial data, it is possible to deduce whether the trapped
surface formation in the collapse takes place before or after the singularity. It
is this causal structure that determines the possible emergence, or otherwise,
of non-spacelike paths from the singularity, and settles the final outcome in
terms of either a blackhole or a naked singularity.
To consider the possibility of the existence of such families, and to examine
the nature of the singularity occurring at R = 0, r = 0 for the scenario under
consideration, consider the outgoing radial null geodesics equation

dt
= eψ. (3.138)
dr

The singularity occurs at a point v (ts (r), r) = 0, which corresponds to

R(ts (r), r) = 0. Therefore, if there are any future directed null geodesics
terminating in the past at the singularity, then R → 0 as t → ts . Now, writ-
ing the null geodesics equation in terms of the variables (u = rα , R) where
α > 1,
 
dR 1 −(α−1)  Ṙ ψ
= r R 1 + e . (3.139)
du α R

Choosing α = 5/3, and using (3.129) together with the collapse condition
Ṙ < 0,
  
dR 3 R v  v 1/2 1 − FR
= + 1/2

 . (3.140)
du 5 u R e−ψ R e−ψ R +
Ṙ

u
 3.6 Dust collapse 103

If the null geodesics terminate at the singularity in the past with a definite
tangent, then at the singularity dR/du > 0 in the (u, R) plane, and this
must have a finite value.
In the case under consideration, all singularities for r > 0 are covered
since F/R → ∞ in the limit of the approach to the singularity, and hence
dR/du → −∞. Therefore only the singularity at the central shell could be
naked.
In order to see the possible emergence of null geodesics from the central
singularity, (3.141) needs to be analyzed. The limits of the concerned func-
tions in (3.141) are calculated at the central singularity. In the TBL case, in
the limit of t → ts , r → 0, e−ψ R → 1. Also, from the equation for v̇ above
and (3.137), in this limit Ṙ → 0. It then follows in general, from the Einstein
equations discussed above, that the term F/R goes to zero in this limit.
It would be interesting to find out when there will be future directed null
geodesics emerging from the central singularity with a well-defined and defi-
nite positive tangent in the (t, r) or (R, u) plane, thus making the singularity
visible. The tangent to the null geodesic at the singularity can be defined as

R dR

x0 = lim lim = . (3.141)

t→ts r→0 u du
t→ts ;r→0

Using (3.141) and (3.137), together with the required limits as above,

3
M0 X (0).
3/2
x0 = (3.142)
2
The necessary and sufficient conditions for a naked singularity to exist, that
is, for the null geodesics with a well-defined tangent to come out from the
central singularity can be deduced. Suppose X (0) > 0, then, from (3.142),
x0 > 0 always; then in the (R, u) plane the equation for the null geodesic
that comes out from the singularity is given by

R = x0 u. (3.143)

In other words, (3.143) is a solution of the null geodesic equation in the limit
of the central singularity. Therefore, given X (0) > 0, a solution of radially
outgoing null geodesics emerging from the singularity can be constructed.
This makes the central singularity visible. In the (t, r) plane, the above null
geodesics near the singularity will be given as

t − ts (0) = x0 r5/3. (3.144)

It follows that X (0) > 0 implies x0 > 0 and radially outgoing null geodesics
emerge from the singularity, giving rise to the central naked singularity.
104 Spherical collapse

On the other hand, if X (0) < 0, then the singularity curve is a decreas-
ing function of r. Hence, the region around the center gets singular before
the central shell, and after that it is no longer in the spacetime. Now, if
there would have been any outgoing null geodesic from the central singular-
ity, it must then go to a singular region or outside the spacetime, which is
impossible. Hence, when X (0) < 0, there is always a blackhole solution.
If X (0) = 0 then the next higher order non-zero term in the singularity
curve equation will have to be taken into account, and a similar analysis by
choosing a different value of α in (3.140) will have to be carried out.
It has thus been shown above that X (0) > 0 is the necessary and sufficient
condition for null geodesics to emerge from the central singularity with a
definite positive tangent. It should be noted, however, that in general the
dependence of R on r along the outgoing null geodesics from the singularity
does not necessarily have to be of a power law form. However, in order
to satisfy the regularity and physical relevance, examining the trajectories
that come out with a regular and well-defined tangent is physically more
appealing, which is the case examined here.
Shown below for completeness is that if null geodesics of any form come
out at all, then those with a definite tangent also must emerge from the
central singularity. Towards this end, consider the equation of the apparent
horizon. From the equations F = R and (3.130), this is given by

X2 (0)
tah (r) = ts0 + rX (0) + r2 + · · · − O(r3 ). (3.145)
2

Since the apparent horizon is a well-behaved surface as one initiates close

to r = 0, for a spherical dust collapse it can be said that the singularity
curve for the collapse and the derivatives around the center are also well-
defined, as the same coefficients are present in both the apparent horizon
equation and (3.131). Also, this shows that whenever X (0) is negative, the
region around the center gets trapped before the central singularity, giving a
sufficient condition for a blackhole to develop. It follows that if null geodesics
are emerging, then at least one of the coefficients X must be non-vanishing
and positive. Then, as already shown, null geodesics with a definite tangent
will emerge from the central singularity.
From the above it follows that in the absence of a null geodesic with a defi-
nite tangent there cannot be any null geodesics emerging from the singularity.
It is also clear now from (3.131) that whether X (0) > 0 or otherwise, it
is fully determined by the regular initial data for the collapse, in terms of
the given initial density and energy distributions for the collapsing shells. It
thus follows that the initial data here completely determines the final fate of
the collapse in terms of the blackhole and naked singularity endstates. This
will be discussed further in the next section.
 3.6 Dust collapse 105

3.6.3 Structure of singularities

The naked singularity forming in the dust collapse has been analyzed exten-
sively, and many of its properties are well-understood. It is known, for
example, that while the models analyzed by Christodoulou (1984) and New-
man (1986) had a naked singularity that was necessarily gravitationally weak,
in general the TBL collapse admits many classes where the naked singular-
ity is a powerfully strong curvature singularity that is not removable. It is
known also that, once the singularity is visible, families of non-spacelike tra-
jectories do come out of it. These are future directed, and in the past they
terminate at the singularity. The global visibility of the singularity has also
been analyzed. It is seen that while the local visibility of the singularity is
decided by the behavior of the density and energy profiles of the cloud close
to the center, the global visibility depends on the nature and behavior of the
mass function F (r) and the velocity function f (r) away from the center of
the cloud (Joshi and Dwivedi, 1993a).
Here, the dust collapse has been treated under fairly general differentiabil-
ity conditions on the mass and velocity functions. Sometimes much stronger
conditions are assumed on these functions, taking them to be necessarily
smooth and analytic. While it may be quite convenient to deal with smooth
and analytic density profiles, especially when it comes to numerical models,
it should not be forgotten that, after all, this is only an extra assumption,
and that the basic equations of general relativity do not demand any such
constraints. Neither is it clear astrophysically that the interiors of the stars
must necessarily have analytic density or energy distributions. In certain
equilibrium cases, the field equations imply that these have to be smooth,
but this need not be true in general, and, the dynamically developing col-
lapse situations especially, could be quite different. The conclusions here also
apply under such stronger conditions.
Consider the scenarios when some of the above assumptions break down.
It is easily seen, using considerations such as those above, that when the
first derivative of density ρ1 is non-vanishing, then the collapse ends in a
naked singularity in all dimensions, including N = 4. Again, the considera-
tions above immediately imply that whenever the spacetime is not marginally
bound, the collapse always results in both the blackhole and naked singu-
larity phases as collapse endstates, depending on the nature of the initial
data, irrespective of whether ρ1 is either zero or non-zero. That is, in a more
generic non-marginally bound case, the condition ρ1 = 0 does not save the
cosmic censorship. However, if, on physical grounds, b1 = ρ1 = 0 are cho-
sen, then from the above discussion it can be seen that X (0) = 0 and cosmic
censorship could be preserved.
Consider the situation when it must be believed somehow that, in the
later stages of the collapse, the form of matter cannot be dust-like, and
106 Spherical collapse

that non-dust forms of matter, and the effects of pressures must be suitably
taken into account. In such a case, as pointed out above, it is seen that even
if only homogeneous initial density profiles (with non-zero initial pressures)
are considered, then the pressure by itself can also cause sufficient distortions
in the formation of the apparent horizon so as to cause a naked singularity
as the endstate of the collapse, rather than a blackhole. However, if the
equation of state is homogeneous, together with the initial data having a
homogeneous density profile, then no naked singularity will appear.
Note that the formalism given here brings out the role of the initial
data in causing the blackhole and naked singularity endstates for the four-
dimensional dust collapse in a clear and transparent manner. To be specific,
it is seen, using (3.131), that given any initial density distribution for the
cloud, a suitable energy profile can be chosen so that the evolution could end
in either a blackhole or a naked singularity, depending on the choice made.
In other words, there is a non-zero measure of energy distributions that will
take the given density profile to a blackhole. The same holds for a naked
singularity to evolve from the same initial density. The converse is also true,
namely that given any initial energy distribution, the density profiles that
give rise to either of these endstates can be chosen.

3.6.4 Self-similar collapse

Self-similar models have been discussed and analyzed extensively in general
relativity (see for example, Carr and Coley, 1999, and references therein).
Here, a self-similar example of a spherically symmetric inhomogeneous dust
cloud collapse as described by the TBL spacetimes is considered in some
detail. A class of these models (Joshi and Singh, 1995), to find that the
endstate of the collapse is either a blackhole or a naked singularity, depending
on the parameters of the initial density distribution, which are ρc , the initial
central density of the massive body, and rb , the initial boundary, will be
discussed. The collapse ends in a blackhole if a certain dimensionless quantity
constructed out of this initial data is greater than a critical value, and ends in
a naked singularity if it is less than this number. It is possible to interpret this
result in terms of the strength of the gravitational potential at the starting
epoch of the collapse.
The method to determine either the blackhole or naked singularity final
states in terms of the quantities calculated from the initial data parame-
ters for the density or the velocity profiles of the dust cloud was discussed
above. Essentially, this amounts to examining the behavior of the singular-
ity curve that forms at the termination of the collapse, or equivalently one
has to examine the existence of real positive roots of an equation involving
these initial parameters as shown below. In order to be able to ascertain the
astrophysical implications of such a result, it is necessary to translate the
 3.6 Dust collapse 107

condition for the existence of positive real roots into the actual constraints
on the initial density distribution in the cloud.
This issue is investigated here, and it would be expected that the degree
of inhomogeneity in the matter distribution plays a role in determining the
final fate of the collapse. This is seen by working out the explicit conditions
for collapse to an endstate that is either a blackhole or a naked singularity,
depending on the initial conditions chosen. It turns out that for the class
under consideration here, these outcomes are characterized in terms of the
existence of real positive roots of a quartic equation. This enables the black-
hole or naked singularity configuration as the endstate of the gravitational
collapse to be related in terms of the initial density distribution ρ(r) and
radius rb of the massive body.
As used earlier, the quantity R(t1 , r1 ) denotes the physical radius of a shell
of collapsing matter at a coordinate radius r1 and on the time slice t = t1 .
The quantities F and f are arbitrary functions of r. In further discussion,
the class of solutions is restricted to f (r) ≡ 0, which comprise the marginally
bound TBL models. Similar considerations can be developed for the models
with f > 0 or f < 0. As the collapsing cloud is of interest here, Ṙ(t, r) < 0.
The epoch R = 0 denotes a physical singularity where the spherical shells of
matter collapse to a zero radius, and where the density blows up to infinity.
The time t = t0 (r) corresponds to the value R = 0 where the area of the
shell of matter at a constant value of coordinate r vanishes. This singularity
curve t = t0 (r) corresponds to the time when the matter shells meet the
physical singularity. In the case of a finite cloud of dust, there will be a cut
off at r = rb , where the metric is matched smoothly with a Schwarzschild
exterior, as discussed earlier.
In TBL models, the freedom to relabel the dust shells arbitrarily exists. It
is given by r → g (r) for any shell with r = const. on any t = const. epoch.
Therefore, at any constant time surface, say at t = t0 , R(r, t0 ) can be chosen
to be an arbitrary function of r. This arbitrariness reflects essentially the
freedom in the choice of units. The choice of scaling at t = 0 as given by
R(r, 0) = r is made for the sake of convenience in the calculation. With this
scaling, the Ṙ equation (with f = 0), can be integrated to obtain
3
R3/2 (r, t) = r3/2 − F (r)t, (3.146)
2
and the energy equation is given by

4/3
(r, t) =   , (3.147)
2 G(r) 2 G (r)
t− t−
3 H (r) 3 H  (r)

where G(r) = r3/2 , G (r) = (3/2)r1/2 , and H (r) = F (r).
108 Spherical collapse

Now, write F (r) ≡ rλ(r), and assume λ(0) ≡ λ0 = 0 and is finite, which
is the class of models considered by Dwivedi and Joshi (1992). This means
that, near the origin, F (r) goes as r in the present scaling, and the density
at the center behaves with time as (0, t) = 4/3t2 . This is a general class
of models that includes all self-similar solutions, as well as a wide range of
non-self-similar models, which are found to be quite adequate for the purpose
of the present investigation. The central density becomes singular at t = 0,
and the singularity is interpreted as having arisen from the evolution of the
dust collapse, which had a finite density distribution in the past on an earlier
non-singular initial epoch.
To check if the singularity could be naked, the possibility that future
directed null geodesics would come out of the singularity at t = 0, r = 0
needs to be examined. The equations of the outgoing radial null geodesics in
the spacetime, with k as the affine parameter, can be written as

dK t
+ Ṙ K r K t = 0, (3.148)
dk

dt Kt
= r = R , (3.149)
dr K
where K t = dt/dk and K r = dr/dk are tangents to the outgoing null geo-
desics. The partial derivatives R and Ṙ that occur in the above equations
can be worked out from the solution given above, and these are most suitably
written as
 
 1−η t
R = ηP − √ + η Ṙ, (3.150)
λ r
 
 λ 1−η t
Ṙ = √ +η , (3.151)
2rP 2 λ r

where
rF 
R(r, t) = rP (r, t), η = η (r) = . (3.152)
F
The functions η (r) and P (r) are introduced because they have a well-defined
limit in the approach to the singularity.
If the outgoing null geodesics terminate in the past with a definite tangent
at the singularity (in which case the singularity would be naked), then using
(3.150) and l’Hospital’s rule,

t dt
X0 = lim = lim = lim R , (3.153)
t→0, r→0 r t→0, r→0 dr t=0, r=0
 3.6 Dust collapse 109

where X = t/r is a new variable. The positive function P (r, t) = P (X, r) is

then given by

2 2P 3/2
X− √ =− √ . (3.154)
3 λ 3 λ

Now, Q is defined as Q = Q(X ) = P (X, 0). If the future directed null

geodesics come out of the singularity at t = 0, r = 0, meeting the singularity
in the past with a definite tangent X = X0 , as given above, then it follows
from (3.154) that such a value X0 must satisfy

2
X0 < √ . (3.155)
3 λ0

the definition F = rλ(r), that as r → 0,

Furthermore, it is noted, by using
η → 1. Also, for f = 0, lim Ṙ = − λ0 /Q. Using these results in the expres-
sion for R implies that the condition (3.153) is simplified to the equation

V (X0 ) = 0, (3.156)

where !
λ0
V (X ) ≡ Q + X − X. (3.157)
Q
In order to be the past end point of the outgoing null geodesics, at least
one real positive value of X0 must satisfy (3.156), subject to the constraint√
implied by (3.155), as stated above, and which is given by X0 < 2/3 λ0 .
In general, it was shown by Dwivedi and Joshi (1992) that if the equation
V (X0 ) = 0 has a real positive root, the singularity would be naked. Whenever
this is not realized, the evolution will lead to a blackhole. Such a singularity
could be either locally or globally naked depending on the global features of
the function λ(r).
The sense in which the terms naked singularity and blackhole are used
needs to be clarified. When there are no positive real roots to (3.156), the
central singularity is not naked, because there are no outgoing future directed
null geodesics from the singularity in this case. Furthermore, as discussed
earlier, the shell-focusing singularity R = 0 for r > 0 is always covered.
Hence, in the absence of positive real roots, the collapse will always lead to a
blackhole. On the other hand, if there are positive real roots, it follows that
the singularity is at least locally naked. The global visibility aspect of such
a locally naked singularity will be discussed in the next subsection. Such a
locally naked singularity would also be globally naked when the outgoing
trajectories could reach arbitrarily large values of r, that is, the signals reach
faraway observers. Otherwise, the collapse outcome would still be a blackhole
110 Spherical collapse

when these trajectories that emerge from the singularity fall back to the
singularity again without emerging from the horizon. This is a violation of
weak censorship only. The occurrence of either of these situations will really
depend on the nature of the function λ(r). The conditions under which this
locally naked singularity could be globally naked have been discussed, for
instance, by Joshi and Dwivedi (1993a). The occurrence of positive real roots
implies the violation of strong cosmic censorship, though not necessarily
of weak cosmic censorship. In other words, a blackhole and locally naked
singularity are not mutually exclusive alternatives. It can also be shown that
whenever there is a positive real root as above, a family of outgoing null
geodesics always terminates at the singularity in the past.
The condition for the occurrence of a naked singularity is now examined
in some detail. The condition V (X0 ) = 0 can be written as
 
2
Y3 Y − − α(Y − 2)3 = 0, (3.158)
3
√ 3/2
where Y = λ0 X0 , and α = λ0 /12 have been set. Note that F (r), and
hence λ0 , are always necessarily positive. Using standard results, it can be
shown that this quartic equation has positive real roots if and only if α > α1
or α < α2 , where
26 √
α1 = + 5 3 ≈ 17.3269
3
and
26 √
α2 = − 5 3 ≈ 6.4126 × 10−3 . (3.159)
3
To derive this condition, first note that if it has a real root, it must be posi-
tive, as negative values of Y do not solve this equation. Writing the general
quartic as ax4 + 4bx3 + 6cx2 + 4dx + e = 0 one defines H = ac − b2 , I =
ae − 4bd + 3c2 , J = ace + 2bcd − ad2 − eb2 − c3 , and ∆ = I 3 − 27J 2 . If
∆ < 0, the quartic has two real and two imaginary roots. If ∆ > 0, all roots
are imaginary unless H < 0 and (a2 I − 12H 2 ) < 0, in which case they are
all real. The application of such a procedure to the quartic in (3.158) leads
to the conditions on α given above.
It follows, however,
√ from earlier discussions, that along any such outgoing
null geodesics λ0 X0 = Y < 2/3. Then, (3.155) implies that the larger
range of α for the existence of roots, that is α > 17.3269, is ruled out in
the sense that no outgoing trajectories can meet the singularity with this
larger value of the tangent X0 . This is seen by writing α as a function of
Y , which shows that if α > α1 , then Y > 2. It thus follows that a naked
singularity arises if and only if α < α1 , or equivalently, if and only if λ0 <
0.1809. Whenever the limiting value λ0 does not satisfy this constraint, the
gravitational collapse of the dust cloud must end in a blackhole. The physical
 3.6 Dust collapse 111

interpretation for the quantity λ0 can be obtained from the equation for the
time evolution of the energy density. If the collapse starts at a time −t0 < 0,
and ρc is the initial energy density at the center, then ρc = 4/3t20 . If ρc is
the initial density gradient at the center, then λ0 = 16ρ3c /3ρ2 c . Putting in
the units gives
8πG 16ρ3c
λ0 = . (3.160)
c4 3ρ2
c

Defining β = λ0 /16, it is found that the blackhole arises whenever

0.0113 < β, (3.161)

and the naked singularity results for the values given by β < 0.0113. The
occurrence of one or the other outcome is governed by the conditions of a
combination of the initial central density and the initial density gradient at
the center. The cosmic censorship hypothesis of Penrose could, in the present
context, be translated to the conjecture that values of β smaller than 0.0113
do not occur in a realistic collapse.
Now, the value of the parameters β or λ0 , given the initial density profile
for the collapsing massive star, need to be calculated. Given the initial central
density, the initial density gradient dρ/dr = ρc can be evaluated at the center
as follows. First, note that the expression for dρ/dr can be written, and it is
√ of r → 0, this always goes to a finite quantity that is
seen that in the limit
proportional to 1/ λ0 . Now, given the initial data in the form of the density
distribution ρ(R) for the body at an initial non-singular epoch of time, in
terms of the physical radius R, integration can give the form of the mass
function F (R). Then, the above provides a functional relationship r(R) that
can be inverted, in principle, to express R in terms of r. The mass profile
can then be written explicitly as F (r), and λ0 is evaluated as the limit of
F (r)/r as r → 0.
This model can be generalized in many ways, for instance by considering
the most general class of functions F (r) and f (r). Also, it has been shown
earlier that the pattern of a transition from the blackhole configuration to
the naked singularity configuration persists in models with more general
equations of state. It is desirable to cast results for these models in terms of
constraints on the initial density distribution. For a discussion on non-self-
similar collapse, see Lake (1991).

3.6.5 Strength and global visibility

Here, the aspects related to the structure of the naked singularity, namely
its curvature strength and the visibility for faraway observers, is discussed
in some detail. The question related to when a non-zero measure set of
112 Spherical collapse

non-spacelike trajectories would emerge from the singularity, as opposed to

isolated trajectories emerging, will be discussed in a general manner in the
next chapter.
The curvature strength of the naked singularity is examined, and a wide
class of TBL models that has a strong curvature singularity, an important
indicator for the physical significance is shown. This is carried out in terms
of the strong curvature condition, which ensures that all the volume forms
must be crushed to a zero size in the limit of the approach to the singu-
larity. Also, the divergence of the Kretschmann scalar K = Rijkl Rijkl in
this limit is seen. It is shown that the general class discussed here contains
subclasses of solutions that admit strong curvature naked singularities in
either of the senses stated above. The conditions are also discussed for the
ultra-dense regions to be globally visible. An implication for the fundamental
issue of the final fate of the gravitational collapse is that naked singularities
need not be considered as artifacts of geometric symmetries of spacetime
such as self-similarity, but that they arise in a wide range of gravitational
collapse scenarios once the inhomogeneities in the matter distribution are
taken into account. It follows that these need not be gravitationally weak,
as was widely conjectured towards formulating a censorship statement (see
for example, Joshi, 1993, for more details and references). Discussed fur-
ther in the next chapter will be the stability and genericity aspects for
naked singularities in order to see that these arise from open sets of non-
zero measures of matter and velocity initial data, from which the collapse
develops.
Clearly, any possible rigorous formulation of the cosmic censorship con-
jecture has to be designed so as to avoid the features above. This is why
gravitational collapse studies are crucial for the cosmic censorship hypothesis.
Possibilities in this direction are discussed in the next chapter, while indicat-
ing that the analysis presented here should be useful for any such attempts.
On the other hand, these developments also give important insights into
the formation and structure of naked singularities in continual gravitational
collapse scenarios.
A genuine strong curvature singularity forms as the collapse final state
in the Oppenheimer–Snyder case of a completely homogeneous dust cloud
collapse with zero pressure. The homogeneous case can, however, be viewed
as a subcase of a set of zero measures in the general inhomogeneous class
given by the TBL solutions. It thus becomes imperative to study TBL models
in greater detail to understand the final fate of a collapsing massive body,
when the effects of inhomogeneities are taken into account. Therefore, the
nature and strength of the singularity for a TBL collapse is now examined.
The numerical simulations of Eardley and Smarr (1979) imply that naked
singularities arise in the marginally bound TBL collapse, and subsequently
a class of such models was studied analytically by Christodoulou (1984) to
 3.6 Dust collapse 113

draw similar conclusions. However, these singularities were shown to be grav-

itationally weak, and hence possibly removable, by Newman (1986), who
studied the curvature strengths of these naked singularities and conjectured
that nature avoids strong curvature naked singularities. The occurrence of
naked singularities may not be considered a problem from the point of view
of cosmic censorship if these were always gravitationally weak in a suitable
sense.
The existence of an incomplete non-spacelike geodesic, or an inextendible
non-spacelike curve that has a finite length, as measured by a gener-
alized affine parameter, implies the existence of a spacetime singularity.
The generalized affine length for such a curve is defined as (Hawking and
Ellis, 1973)

  1/2
a
3
i 2
L(λ) = (X ) ds, (3.162)
0 i=0

which is a finite quantity. The X i values are the components of the tan-
gent vector to the curve in a parallel propagated tetrad frame along the
curve. Each such incomplete curve defines a boundary point of the space-
time, which is a singularity. In order to know if this is a genuine physical
singularity, one would typically like to associate it with unboundedly growing
spacetime curvatures. If all curvature components and the scalar polynomi-
als formed out of the metric and the Riemann curvature tensor remain finite
and well-behaved in the limit of approach to the singularity along an incom-
plete non-spacelike curve, it may be possible to remove the singularity by
extending the spacetime when the differentiability requirements are lowered
(Clarke, 1986).
There are several ways to formalize such a requirement. For example, a
parallel propagated curvature singularity is the end point of at least one non-
spacelike curve on which the components of the Riemann curvature tensor are
unbounded in a parallel propagated frame. On the other hand, a scalar poly-
nomial singularity has a scalar polynomial in the metric and Riemann tensor
taking an unbounded large value along at least one non-spacelike curve with
this singular end point. This includes the cases such as the Schwarzschild
singularity where the Kretschmann scalar Rijkl Rijkl blows up in the limit
as r → 0. Will genuine curvature singularities occur in general relativity?
The answer, for the case of parallely propagated curvature singularities, is
provided by a theorem of Clarke (see for example, Clarke, 1993). This theo-
rem shows that for a globally hyperbolic spacetime that is C 0− -inextendible,
when the Riemann tensor is not very specialized in the sense of not being
type D and electrovac at the singular end point, then the singularity must be
a parallely propagated curvature singularity. Similarly, a class of physically
114 Spherical collapse

relevant strong curvature singularities was analyzed by Tipler (1977), Tipler,

Clarke, and Ellis (1980), and Clarke and Królak (1986).
The idea here is to define a physically all-embracing strong curvature sin-
gularity in such a way so that all the objects falling within the singularity
are destroyed and crushed to zero volume by the infinite gravitational tidal
forces. This notion is formulated as below. Let λ(t) be a timelike or null
geodesic that is incomplete at an affine parameter value t = 0. Let K i
denote the tangent vector to λ and µ(t) = Z 1 ∧ Z 2 ∧ Z 3 be a volume form
defined along λ(t) where Z 1 , Z 2 , Z 3 are linearly independent Jacobi vectors
defined along the curve λ orthogonal to K i . (If λ is null then µ(t) is defined
as a two-form.) A real-valued map from the space of all such three-forms can
be defined by (A ∧ B ∧ C ) = det[Ai , B i , C i ]. Denote (µ(t)) by V (t),
which defines a volume element along λ(t) and is independent of the choice
of basis. The singularity at t = 0 is then called a strong curvature singularity
if V (t) = 0 in the limit as t → 0 for all possible µ(t), that is, for all possible
choices of linearly independent Jacobi fields.
This definition effectively captures the notion that all objects falling into
a strong curvature singularity are crushed to zero volume. For a strong cur-
vature singularity, at least one non-spacelike geodesic terminating at the
singularity along which the above curvature condition is satisfied exists. How-
ever, a strong curvature singularity can be defined in a much stronger sense
also (Tipler, 1977), by requiring that the strong curvature condition above
must be satisfied along all non-spacelike geodesics terminating in the sin-
gularity. Necessary and sufficient conditions for the occurrence of strong
curvature singularities are derived by Clarke and Królak (1986), and which
are shown to involve the tetrad components of the Riemann, Ricci and
Weyl tensors, and also the divergence of their integrals along non-spacelike
geodesics running into the singularity. It follows from their analysis that an
incomplete non-spacelike geodesic does not define a strong curvature singu-
larity unless either the Weyl or Ricci tensor components diverge sufficiently
fast along such a trajectory.
A sufficient condition for this to happen is, if t is the affine parameter,

Rij V i V j ≥ A/t2 , (3.163)

for some fixed constant A along the trajectory in the limit of approach to the
singularity as t → 0. This provides a sufficient condition for all the two-forms
µ(k ), defined along the singular null geodesic, to vanish as the singularity
is approached, and implies a very powerful curvature growth establishing
a strong curvature singularity. For timelike geodesics this will imply that
all the volume forms defined by the Jacobi fields along these trajectories
must vanish in the limit of approach to the singularity, or they must vanish
infinitely many times in this limit.
 3.6 Dust collapse 115

To fix these ideas, the case where Rij V i V j = K/t2 is discussed in some

detail. It is convenient to define a length scale y associated with the volume
V (t) by defining y 3 = V . The propagation equation for y (t) along λ(t) is the
Raychaudhuri equation, which is a second order differential equation, given
in this case as
d2 y 1
+ (Rij V i V j + 2σ 2 )y = 0, (3.164)
dt2 3
with σ 2 = σij σ ij being the trace of the shear tensor σij . (For the null case,
a similar equation holds with 1/3 being replaced by 1/2.) Writing F (t) =
(Rij V i V j + 2σ 2 )/3, and ignoring the effects of the shear tensor, which will
enhance the focusing effect considered here, F (t) = A/t2 , where A > 0 is a
fixed constant, is chosen. If a solution of the form y = tα is tried, then the
condition on A is obtained to be A = α − α2 . Since V (t) → 0 in the limit
of approach to the singularity t = 0, α > 0. Again, A > 0 and the above
expression for A implies that 0 < α < 1. Solving for α gives that, in order
for α to be real, A satisfies A ≤ 1/4. The solution for y is then given by
1/2
y = t[1±(1−4A) ]/2
. (3.165)

Therefore, depending on the value of A, it is seen that y ∼ tα with 1/2 ≤

α < 1. The volume V (t) then goes to zero near the singularity at least as
fast as t3/2 .
Basically, the sufficient condition for a singularity to be strong in the
sense of Tipler (1977), is that at least along one null geodesic with the affine
parameter k , with k = 0 at the singularity, in the limit of approach to the
singularity,
lim k 2 Rab K a K b > 0, (3.166)
k→0

where K a and K b are tangents to the null geodesics. This is a sufficient

condition for all two-forms µ(k ) defined along the singular null geodesic to
vanish as the singularity is approached, and implies a very powerful curvature
growth, establishing a strong curvature singularity. For the timelike geodesics
this implies that all the volume forms defined by the Jacobi fields along these
trajectories must vanish in the limit of approach to the singularity, or they
must vanish infinitely many times in this limit.
The important physical consequences of the existence of a singularity are
related to its strength. If the singularity is gravitationally weak, it may
be possible to extend the spacetime. On the other hand, when there is a
strong curvature singularity forming as above, the gravitational tidal forces
associated with it are so strong that any object trying to cross it gets
destroyed. Therefore, as argued by Ori (1992), the extension of spacetime
becomes meaningless for such a strong singularity that destroys to zero size
all the objects terminating at the singularity. From this point of view, the
116 Spherical collapse

strength of the singularity may be considered crucial to the issue of classi-

cally extending the spacetime and thus avoiding the singularity, because for
a strong curvature singularity, no continuous extension of the spacetime may
be possible.
For the general class of TBL models under consideration,

F  (K t )2 F  (K t )2
Ψ ≡ Rab K a K b = = . (3.167)
R2 R R2 R
For radial null geodesics, using l’Hospital’s rule and the TBL equations as
given earlier, and the fact that in the limit of approach to the singularity
r → 0, X → X0 ,

4η0 Λ0
lim k 2 Ψ = √ 2 , (3.168)
k→0 H0 X 02
2 (3α − η0 ) 1 + f0 − N0

where η and Λ are as defined in (3.118) and (3.119). Hence, it is seen from
the definition of Λ in (3.120) that

lim k 2 Ψ = 0 for α < η0 , (3.169)

k→0

lim k 2 Ψ = 0 for α ≥ η0 . (3.170)

k→0

However, from earlier conclusions, a naked singularity occurs only when

α ≤ η0 , therefore the strong curvature condition is satisfied along the singular
geodesics for the classes where α = η0 . As noted earlier, for the special class
considered by Christodoulou (1984) and Newman (1986), α = 7/3 and η = 3,
and hence the naked singularity turns out to be gravitationally weak. On the
other hand, it is clear from the above that, for a wide variety of TBL solutions
satisfying the condition α = η0 , the singularity will be a strong curvature
singularity in the above sense. In general, it is also possible that non-radial
null or timelike curves could terminate at the naked singularity. Then, a
similar calculation along non-spacelike geodesics in general gives

lim k 2 Ψ ∝ rη0 −α r=0 . (3.171)
k→0

So, as discussed above, it can be concluded that the condition for strong
curvature is satisfied along non-spacelike geodesics as well if α = η0 , and if
such families meet the naked singularity in the past.
The Kretschmann scalar Rijkl Rijkl along the geodesics in the TBL
spacetimes goes as
K ∝ r2(η0 −3α). (3.172)
Hence, the singularity is a scalar polynomial singularity as long as α > η0 /3.
 3.6 Dust collapse 117

A self-similar dust collapse example was discussed in an earlier section.

The self-similar TBL models are, in general, defined by the conditions f (r) =
const. and η (r) = 1 = η (0) = α. It follows from the above that the naked
singularity forming in this class will be a strong curvature singularity along
all the families of radial null geodesics. A detailed discussion on the families of
future directed non-spacelike geodesics, terminating at the naked singularity
in the past, is given by Joshi and Dwivedi (1992, 1993b). Various different
families of non-spacelike geodesics do terminate at the naked singularity in
the past in this case, along which the strong curvature condition is also
necessarily satisfied.
As discussed in the previous section, the existence of a real positive root
of the equation V (X ) = 0 shows the singularity to be at least locally naked.
The singularity can be globally visible also, and to examine this, note that
the apparent horizon lies at R(t, r) = F (r) within the collapsing cloud.
Therefore, if a geodesic gets inside the apparent horizon as it is emerging
from the singularity, it becomes ingoing, with R < F along the geodesics
and dR/dr becoming negative. Eventually this trajectory must fall back to
the singularity. Therefore, if a light ray is to reach future infinity in order
for the singularity to be globally naked, it must cross r = rc , which is
the boundary of the dust cloud before the time when the apparent horizon
reaches this boundary as it evolves during collapse. Hence, all escaping non-
spacelike geodesics that reach the boundary r = rc with R(rc ) > F (rc )
would reach the future infinity. Since geodesics emerge from the singularity
with the tangent value X 0 and the apparent horizon has a tangent Λ0 at
the singularity, it follows that X 0 > Λ0 .
As a result, because of the generality of the function F (r), which is the
free initial data subject only to regularity conditions, rc and F (rc ) = 2M ,
with M being the Schwarzschild mass of the cloud, can be chosen suit-
ably such that the geodesics reach the boundary of the cloud r = rc with
R(rc ) > F (rc ), making the singularity globally visible. However, given a
boundary r = rc and F (rc ) = 2M , the singular geodesics that actually
reach future infinity depend on the global properties of the functions F (r)
and f (r).
An explicit class of TBL models where the singularity is shown to be glob-
ally visible is now discussed. Owing to the complexity of the equations, exact
geodesic solutions are few, even for simple forms of the functions F (r) and
f (r), except for the Friedmann models of complete homogeneity. However,
a self-similar marginally bound collapse ( f = 0) can be considered, and its
formalism is illustrated here. This example can be analyzed using a special
null trajectory that is the Cauchy horizon given by X = const., which is the
first null geodesic emerging from the singularity. It is shown that actually
the geodesic equations can be integrated completely for this self-similar case,
in order to obtain radial null families. As pointed out above, the collapse
118 Spherical collapse

will end in a naked singularity when V (X ) = 0 has two real positive and
two complex roots. Let x1 , x2 (x1 > x2 ) be two such positive roots of this
equation. The equation of geodesics, in the form r = r (x), X = x2 is given
by (Joshi and Dwivedi, 1993a)

(x − x2 )n2
r = r(X ) ≡ r(x) = D f1 (x), (3.173)
(x − x1 )n1

where
  
Ax + B
f1 (x) = exp − dx . (3.174)
x2 + D1 x + D2
Here, n1 , n2 , A, B, D1 , and D2 are constants given by
√
Λ0 3 √
x + 4
x − x + Λ0 = (x − x1 )(x − x2 )(x2 + D1 x + D2 ), (3.175)
2
3x3 n1 n2 Ax + B
√ √ = − + 2 , (3.176)
Λ0 3
x + 2 x − x + Λ0
4 x − x1 x − x2 x + D1 x + D2

and D is the constant that labels the different geodesics. The constants n1 ,
and n2 are positive. For the case Λ0 = 7/17,

x1 = 0.658303, x2 = 0.5, n1 = 2.09356, n2 = 1.08511, D1 = 1.36419,

(3.177)
D2 = 1.2509, A = −1.99154, B = −1.26354. (3.178)

It is clear from (3.173) that geodesics reach r = 0 at x = x2 and r = ∞

at x = x1 , making the singularity globally naked. Note that η (r)Λ(r) =
F0 < x2 and therefore all the trajectories that are emitted in the region
x1 > x > x2 reach the future infinity. In fact, x = x1 and x = x2 are also
geodesics that cross the boundary of the cloud and escape to future infinity.
To discuss conditions that ensure the global visibility of the ultra-dense
regions in general, suppose the functions η and β are at least C 0 in the
interval rc ≥ r > 0. Since these involve the first derivatives of f and F as
f  /f and F  /F , this implies that f and F have at least first order continuous
derivatives. The C 2 -differentiability of the metric ensures this. If V (X ) = 0
has only one simple root X = X0 , a family of curves terminates at the
singularity, that is, h0 > 1, with this value of the tangent. Let η (r)Λ(r) <
αX0 for rc ≥ r > 0. Then the singularity would be globally naked. To
see this, consider the equation of geodesics (3.173), where the constant D
labels different geodesics terminating at the singularity, as determined by
the boundary conditions. For a singular geodesic that reaches the boundary
 3.6 Dust collapse 119

of the dust cloud u = uc = rα = rcα with X = (Rc /rcα ) = Xc ,


h0 −1 h0 −1
Xc − X0 = Duc + uc Su(−h0 +1) du, (3.179)
uc

and hence the equation of such a geodesic can be written as

 h0 −1  u
u
X − X0 = (Xc − X0 ) + uh0 −1 Su(−h0 +1) du. (3.180)
uc uc

The event horizon is represented by the geodesic for which Xc = Λ(rc ). Since
it is outgoing, dR/d(rα ) is positive at r = 0 and ejected into the region
R > F where dR/dr is positive. Therefore, all the geodesics that reach
r = rc , where the TBL metric is matched with the Schwarzschild exterior,
with Xc > Λ(rc ) would escape to infinity, while others would become ingoing.
It follows that the geodesics that reach future infinity with their past end
point at the singularity are given by (3.180) with Xc > Λc = Λrc . Hence,
in the case when a family of geodesics terminates at the singularity with a
tangent X = X0 and η (r)Λ(r) < αX0 , for rc ≥ r > 0, the singularity would
be globally visible, as there would always be some geodesics that escape to
infinity.
Consider the case now when the equation V (X ) = 0 has two positive roots
X1 and X2 (X1 > X2 ). In such a case, families of curves would emerge from
the singularity with the tangent either X1 or X2 . Let η (r)Λ(r) < αX2 for
rc ≥ r > 0, then it ensures that some geodesics would cross the boundary
of the cloud with Xc > Λ(rc ) making the singularity globally naked. The
same holds even when more than two positive roots exist. Therefore, if the
family of geodesics do terminate at the singularity with tangent X0 , then
the condition η (r)Λ(r) < αX0 for rc ≥ r > 0 implies the global visibility of
the singularity.

3.6.6 Cosmological constant

For the inhomogeneous dust collapse models, a singularity always develops
necessarily in an initially collapsing configuration, and rebounce or halting
the collapse are not possible. The singularity can be naked, or hidden behind
an event horizon, depending on the nature of the initial density and velocity
profiles for the collapsing cloud.
In the above, the Einstein equations were used with a vanishing cosmo-
logical term. It is possible, however, that this is not the case and there may
be some observational support that the universe is dominated by an energy
component with a negative pressure (Perlmutter et al., 1998, 1999; Zehavi
and Dekel, 1999). A plausible explanation for such a dark energy present
in the universe could be in terms of the presence of a non-zero value of the
120 Spherical collapse

cosmological term Λ. This will give rise to a vacuum energy density, corre-
sponding to a positive sign of Λ. This would represent a spatially uniform
time independent energy density distribution, and its positive value acts as
a globally repulsive force field. Also, the cold dark matter models with a
substantial component supplied by Λ can fit the current observational data
(Ostriker and Steinhardt, 1995).
Therefore, it is useful and interesting to consider the Einstein equations
with a non-zero Λ, to investigate the gravitational collapse, and also the
implications in cosmology. With such a perspective, the structure of a sin-
gularity in the spherically symmetric dust models with a non-vanishing
cosmological constant is discussed here. Dust models with a cosmological
term are discussed in the literature (see for example, Krasinski, 1997). These
models can be matched with the Schwarzschild–de Sitter spacetime at the
boundary of the cloud (Omer, 1965; Lake and Roeder, 1979; Lake, 2000.)
Here, the TBL collapse models are analyzed with a cosmological constant,
and provide the general solution to the Einstein equations with dust as the
source term.
The weak energy condition, Tij V i V j ≥ 0, for all non-spacelike vectors
i
V , is assumed in the spacetime for the matter to be physically reasonable,
especially in the case of the gravitational collapse. The Einstein equations
are analyzed to check whether there are globally regular solutions, as have
been shown to exist in the case of homogeneous density dust solutions with
a positive cosmological constant (Markovic and Shapiro, 2000). The general
solution to the Einstein equations in the case corresponding to the marginally
bound models in the TBL spacetimes, which is when the energy function f =
0, is also derived, and the condition for avoiding shell-crossings is discussed.
The structure of the singularity can be investigated by studying the outgoing
radial null geodesics near it. Then, it can be examined how the given initial
conditions, in the presence of the cosmological term, affect the final state
results in terms of the formation of a blackhole or a naked singularity. It
can also be seen why the naked singularity turns out to be stable to the
introduction of the cosmological term.
Here, the gravitational collapse of a dust cloud in the presence of a cosmo-
logical term is discussed in some detail, and how it may affect the dynamical
evolution of the collapse is examined. In particular, it is found that there
is a non-trivial detailed dynamical structure, especially for the f < 0 case,
and this provides a strong motivation for the present study, namely how the
introduction of a cosmological term non-trivially changes the structure of
the collapse. In most cases, there is no rebounce, and then the nature of the
singularity forming as the endstate of the collapse is studied. At a more tech-
nical level, the change in the dynamical structure of the collapse, when there
is a non-vanishing Λ term present, is reflected in the change in the nature
and structure of the singularity, which is far from obvious. There are some
 3.6 Dust collapse 121

important changes in the collapse dynamics with the introduction of the cos-
mological constant. In particular, the analysis of the roots equation referred
to earlier shows that part of the naked singularity spectrum is covered when
a positive cosmological term is introduced, but it is never removed altogether
irrespective of its value. In this sense, the naked singularity is stable to the
introduction of the cosmological term.
In fact, there has been hope through the last three decades or so that it
may be possible to avoid singularities by the introduction of a cosmological
term with a positive sign (see for example, Hawking and Ellis, 1973, pp. 139,
362), especially in the cosmological case. Hence, it is important to investigate
the collapse dynamics, especially on a large or medium scale in the universe,
in the presence of a cosmological term. In fact, as shown by Deshingkar
et al. (2001), for a certain range of initial data, there is a rebounce possible
due to the presence of the positive cosmological term. Such a possibility will
be especially important in the case of the collapse of either the clusters or
superclusters of galaxies, where the densities may be sufficiently low at the
initial epoch, and such a rebounce can become reality.
Towards the effects of a non-vanishing cosmological term on the final fate
of an inhomogeneous collapsing dust cloud, it is shown that, depending on
the nature of the initial data from which the collapse evolves, and for a
positive value of the cosmological constant, it is possible to have a globally
regular evolution where a bounce develops within the cloud. The initial data
causing such a bounce is characterized in terms of the initial density and
velocity profiles for the collapsing cloud. In the other cases, the result of
the collapse is either the formation of a blackhole, or a naked singularity,
as the endstate of the collapse. It is also found that a positive cosmological
term can cover a part of the singularity spectrum that is visible in the corre-
sponding dust collapse models for the same initial data. The basic set of the
Einstein equations and the regularity conditions for the collapse is discussed,
and then the possibility when there is a non-vanishing cosmological term
present, when an initially collapsing cloud rebounces at a later epoch so that
a singularity does not form, is investigated. Therefore, there is an occurrence
of three phases, the collapse, the reversal, and the subsequent dispersal. A
general solution with a non-zero Λ term for the marginally bound case is
also given.
The model for a self-gravitating, spherically symmetric, inhomogeneous
dust cloud with cosmological constant is given by the metric

R 2
ds2 = −dt2 + dr2 + R2 (dθ2 + sin2 θ dφ2 ), (3.181)
1 + f (r)

where (t, r, θ, φ) is a comoving coordinate system, with notations as earlier.

The equation of state for matter in the interior of the cloud is that of dust,
122 Spherical collapse

with the stress–energy tensor given by Tji = (r, t)δti δjt , with (r, t) being
the energy density of matter. The weak energy condition is assumed for the
matter, which implies that (r, t) ≥ 0. This is equivalent to the strong energy
condition in this case, as the principal pressures are zero.
The Einstein equations in the presence of a cosmological constant can be
written as
F (r) Λ
Ṙ2 = + f (r) + R2 , (3.182)
R 3
F
(t, r) = 2  . (3.183)
R R

The collapse situation with Ṙ < 0 is mainly considered. The two free func-
tions that characterize the dust cloud are those representing the total mass
F (r), and the energy f (r), inside the shell labeled by the comoving coordi-
nate r. The cosmological term Λ can, in principle, be of either sign, however
the recent observations indicated above seem to favor the positive sign. The
above equation for Ṙ can, in principle, be integrated, and after integration, a
constant of integration is obtained. The constant of integration can be fixed
by using the scaling freedom. Here, this is fixed by setting R = r on the
initial hypersurface (t = 0). The two free functions F (r) and f (r) can be
fixed by prescribing the initial density and velocity profiles through (3.183)
and (3.182) respectively. Assume these free functions have the form

F (r) = F0 r3 + Fn r3+n + higher order terms,

(3.184)
f (r) = f0 r2 + fn1 r2+n1 + higher order terms,

where the choice of the first non-vanishing term is made in order to have
regular initial data on the initial surface t = 0 from which the collapse
evolves. From (3.183), it is clear that it is possible to have both shell-crossing
(R (tsc (r), r) = 0), and shell-focusing singularities (R(tsf (r), r) = 0) in these
spacetimes, depending on the dynamics of the shells given by (3.182). Here,
only cases where there are no shell-crossings are considered, as these are gen-
erally not considered to be genuine singularities, and the main interest here
is in studying the nature of the physical singularity corresponding to R = 0
where the matter shells shrink to zero radius. This puts some restrictions on
the initial data.
The metric exterior to the collapsing cloud has the Schwarzschild–de Sitter
form,
ds2 = −gdt2 + g −1 dr2 + r2 (dθ2 + sin2 θ dφ2 ), (3.185)
where g = 1 − (2M/r) − (Λr2 /3). Matching the solutions (3.181) and (3.185)
at the boundary rb of the collapsing dust cloud, 2M = F (rb ) is obtained.
The boundary can be made to bounce from an initially collapsing phase
 3.6 Dust collapse 123

by choosing appropriate initial mass and cosmological terms, as seen below.

This behavior can also be understood by analyzing the curve of the allowed
motion (3.182).
First, the possibility of having globally regular solutions due to the pres-
ence of a cosmological constant is looked for. Equation (3.182), governing
the dynamics of collapsing shells, can be conveniently written as

V (R, r)
Ṙ2 = . (3.186)
3R

Here V (R, r), defined as

V (R, r) = 3F (r) + 3f (r)R + ΛR3 , (3.187)

is an analog of the Newtonian effective potential governing motion of the

shells. The allowed region of motion corresponds to V ≥ 0, as Ṙ2 has to be
greater than zero.
If an initially collapsing state is the starting point, a rebounce will occur if
Ṙ = 0 for a given shell before that shell becomes singular. This can happen
only if the equation
3RṘ2 = V (R, r) = 0 (3.188)

has two real positive roots. In the following, it is convenient to define a

quantity ζ (t, r) = R/r. Equation (3.187) can then be rewritten as

V = 3F + 3f ζr + Λζ 3 r3 , (3.189)

which is a cubic equation with three roots in general. From the theory of
cubic equations, if all the three roots of the equation above are real, then
at least one of them has to be positive and at least one negative. Note that
V (R = 0, r > 0) = 3F > 0. Hence, any regular region between R = 0 and the
first zero of (3.189), that is, ζ1 > 0, always becomes singular during collapse
and so two real positive roots for (3.189) are required for the possibility of
a rebounce. The region between the two real positive roots is a forbidden
region. So, starting on the right side, the collapsing shells bounce back and
then there is continuous expansion. Since one of the real roots has to be
negative and the region between two real positive roots is forbidden, it is not
possible to have oscillating solutions in these spacetimes.
In the case of Λ = 0, it is well-known that a rebounce is not possible, and
the collapse necessarily results in a singularity. The cubic then reduces to a
linear equation and the solution is given as R = Rmax (r) = −F/f . So, only
in the case where f < 0 can Ṙ = 0 for positive R. This corresponds to the
maximum possible physical radius for a given shell, that is, even if an initial
124 Spherical collapse

expansion is the starting point, a given shell will reach the maximum radius
Rmax (r), and then it will recollapse.
In the case when Λ < 0, one, and only one, root is always positive.
The other two roots are negative if 9F 2 < −(4f 3 /Λ), or else they are
complex conjugates. Therefore, any initial configuration becomes singular
in this case. The real positive root in this case gives an upper bound on
the radius R = Rmax (r) of a shell labeled r. This upper bound occurs as
the negative (attractive) contribution from the Λ term keeps on increasing
with an increasing R, while the contribution from gravitational attraction
keeps on decreasing, and so at some point for any value of f , the attraction
due to Λ starts to dominate. Hence, even if there is an initially expanding
configuration, finally there must always be a collapse in this case.
For Λ > 0 and f (r) ≥ 0, it can easily be seen that it is never possible to
have ζ̇ 2 = Ṙ2 /r2 = 0, so a singularity always forms if an initial collapse is
always started from.
For the case when Λ > 0 and f (r) < 0, one root of the cubic above is
always negative. If
4f 3
F2 > − , (3.190)
9Λ
then the other two roots are complex conjugates. So, the singularity always
forms in such a case in an initially collapsing configuration. On the other
hand, if the initial data is such that

4f 3
F2 < − , (3.191)
9Λ

then the other two roots are real and positive. Denote these roots by ζ1
and ζ2 with ζ1 < ζ2 , and the region between these two roots is forbidden.
The entire space of allowed dynamics is given by the two disjoint regions
[0, ζ1 ] and [ζ2 , ∞]. If the initial scale factor ζ0 lies in the first section, i.e. if
ζ0 < ζ1 , then the singularity is always the end point of the collapse. Here, rζ1
represents the upper bound for the physical radius of a shell in this region.
If ζ0 lies in the second section, that is, ζ0 > ζ2 , then there will be a rebounce
from the initial collapsing configuration. After the rebounce, the physical
radius of the shell keeps on increasing forever. There is no upper limit for
the maximum value of ζ in this region and rζ2 gives the lower bound for the
physical radius of a shell, that is, the shell rebounces at R = rζ2 .
From the above discussion, it can be seen that a rebounce is possible only
in the case when Λ > 0 and f < 0, and when the following two conditions
are satisfied:
4f 3
F2 < − (3.192)
9Λ
 3.6 Dust collapse 125

and ζ0 > ζ2 . With the scaling used here, the condition (3.192) can be
written as
   ! 
f 1 9F 2 Λ
1>2 − 2 cos arccos − . (3.193)
r Λ 3 4f 3

The physical quantities such as central density, ρc = F0 /ξ 3 (t), and

curvature scalars,
   
4 2 2 4 2 2 2
ζ̇ + ζ ζ̈ ζ̇ + ζ ζ̇ ζ̈ + ζ ζ̈
Rijkl Rijkl = 12 , Rij Rij = 12 , (3.194)
ζ4 ζ4

and  
ζ̇ 2 + ξ ζ̈
g ij rij = 6 (3.195)
ζ2

stay finite, as ζ (t, r) > 0 for the regular models. For further details see
Deshingkar et al. (2001) and Arun Madhav et al. (2005).
There are several features here that are worth noting, and which have
an interesting physical significance as far as the dynamics of the collapse is
concerned. These also illustrate the effects that a non-vanishing cosmological
term may have towards determining the final fate of a collapsing cloud of
matter. First, with a negative value of the cosmological term, all the solutions
become closed and a singularity always forms in the future even if one starts
with initial expansion. This is to be expected because such a value will
only contribute in a positive manner to the overall gravitational attraction
of matter, and it just acts as a constant positive energy field helping the
collapse. Next, there is a range of initial data where the collapse necessarily
ends in a singularity however large the positive value of the cosmological
term is. This is contrary to the belief sometimes expressed that a positive
cosmological constant can always cause a rebounce, provided it is sufficiently
large. Finally, it is clear from the above that there can be a rebounce only if
the initial density is sufficiently low for a given positive value of Λ. This is so
because the cosmological term becomes dominant with increasing distances,
and gravity dominates at higher densities. Therefore, as the cloud is more
disperse, with a lesser density but a larger size, the effect of the cosmological
term is greater.
It can therefore be seen that while a bounce and a regular solution occur
for a specific range of the initial data, for the majority of the regular initial
data space, the collapse results in a spacetime singularity where the densities
and curvatures blow up. While for Λ = 0 the structure of this singularity
and when it will be naked or covered is known in detail, the above discussion
gives some understanding of the effects of a non-zero Λ towards the structure
of the singularity forming in the collapse. To see this more clearly for a dust
126 Spherical collapse

collapse, the case f (r) = 0 is now analyzed explicitly in some detail. While
f = 0 has been chosen for the simplicity and clarity of the considerations,
similar behavior would also be expected in other cases. Equation (3.182) can
now be written as
  −1/2
F (r) Λ 2
t − tc (r) = ± + R dR, (3.196)
R 3

where tc (r) is an integration function that represents the time at which a

given shell, at a value r, becomes singular, i.e. R(tsf (r), r) = 0. The positive
or negative sign respectively corresponds to the expanding and collapsing
branches of the solution. Here, only the negative sign is considered because
clouds that are collapsing with Ṙ < 0 initially are being considered. The
integral (3.196) can be written as an infinite series in R near the center as
  m 
2 R3/2 ∞
(−1)m (2m − 1)!! ΛR 3
t − tc (r) = −  1+ . (3.197)
3 F (r) m=1
2m (2m + 1)m! 3F (r)

Using the scaling freedom in this solution, R(t = 0, r) = r can be set. This
determines tc (r) as
  m 
2 r3/2 ∞
(−1)m (2m − 1)!! Λr 3
tc (r) = tsf (r) =  1+ .
3 F (r) m=1
2m (2m + 1)m! 3F (r)
(3.198)
From the above expressions,
  1/2  1/2
F rF  3F + ΛR3/2 r
R = R+ 1− . (3.199)
3F 3F 3F + Λr3/2 R

Therefore, the condition R > 0, implying no shell-crossings, can be satisfied

if F  > 0 and 1 − (rF  /3F ) > 0. This means that the total mass inside
a shell and the matter density are increasing and decreasing functions of r
respectively. Incidently, the above expression also gives the conditions for
no shell-cross singularities to occur in the dust collapse when the cosmolog-
ical term also vanishes. It would appear that, in any physically reasonable
collapse picture, these conditions will be satisfied and shell-crossings will
not occur. The weak energy condition, implying positivity of energy density,
guarantees that mass is an increasing function of r. Also, for any realistic
density distributions, it would be physically reasonable that the density is
higher at the center, decreasing away from the center. Therefore, decreasing
density profiles will be worked with here. No shell-crossings thus occur in the
spacetime before the occurrence of the central shell-focusing singularity for
the choice of initial data.
 3.6 Dust collapse 127

As seen below, the behavior of collapsing shells near the center would
depend only on the first non-vanishing derivatives of the density and veloc-
ity profiles. Therefore, the local visibility conditions are unaffected by the
boundary conditions such as the initial choices of the mass function and the
actual value of the Λ term. On the other hand, the global behavior of the
trajectories emerging from the singularity can change due to the addition of
a non-zero cosmological term.
To analyze the nature of the central singularity, the method developed by
Joshi and Dwivedi (1993a), which gives a necessary and sufficient condition
for the local visibility of the singularity is followed. The main idea here is to
see if outgoing radial null geodesics in the spacetime, meeting the singularity
in their past with a well-defined real positive tangent vector in a suitable
plane are possible.
The equation for these null geodesics in the spacetime (3.181), for f (r) = 0
is given by
dt
= R . (3.200)
dr

For convenience, this can be written in the (u, R) plane as

  
dR R F Λ 2
= 1− + R , (3.201)
du αrα−1 R 3

where u = rα , and α ≥ 1 is a constant to be determined later. From (3.201),

it is clear that there are no such outgoing paths existing from the non-central
part of the singularity curve, because the first term under the square root sign
goes to −∞ as t approaches tsf (r). Hence, it is only the central singularity
that can be possibly visible.
Define X = R/u. The purpose here is to check if a well-defined tangent
for (3.201), at r = 0, R = 0 in the limit of t approaching the singular epoch
ts (0) is possible. Using l’Hospital’s rule,

R dR R
X0 = lim = lim = lim
u→0, R→0 u u→0, R→0 du u→0, R→0 αr α−1
  
F Λ
1− + R2 ≡ U (X0 , 0), (3.202)
R 3

where the subscript 0 denotes the value of the quantities at u = 0. The

constant α can be uniquely fixed by demanding that R /rα−1 is non-zero
and finite (Joshi and Dwivedi, 1993a). In this case, R /rα−1 remains finite if
128 Spherical collapse

α = 1 + 2n/3 is chosen, and (3.202) can be written as

   
1 nFn 1 λ0
X− √ 1+ = 0, (3.203)
α 3F0 X X

where λ = F/u and λ0 is the limit as r → 0. If the above has a real positive
root X0 then there will be at least one null geodesic emerging from the
singularity at R = 0, u = 0 with the root X = X0 as a tangent in the (u, R)
plane.
When n < 3, α < 3 and therefore λ0 = 0, and the above equation
reduces to
F
X0 = −
3/2
 n , (3.204)
2F0 1 + ΛF0 /3
which always has a real positive root. Apart from an additional Λ term, this
equation is analogous to the corresponding equation obtained for the TBL
models, and reduces to the same case for Λ = 0. Therefore, when either
the first or the second derivative of density is non-zero, the singularity is
always at least locally visible. The addition of the cosmological term changes
only the value of the tangent (X0 ) to the outgoing radial null geodesics
from the singularity, but not the visibility property itself. Therefore, the
corresponding dust naked singularity spectrum is stable to the addition of a
positive cosmological constant.
In this case, as in the earlier studies on TBL models, the smaller root
will be along the apparent horizon direction and a family of geodesics will
come out along this direction. In this case, set α = 3, and the first term in
(3.201) blows up, and the second term goes to zero such that the product is
F0 . For the value n = 3, which corresponds to the critical case in the TBL
models as discussed
 above, λ0 = F0 is obtained. Introducing X = F0 x2 and
5/2
ξΛ = F3 /F0 1 + ΛF0 /3, the root equation above becomes

2x4 + x3 − ξΛ x + ξΛ = 0. (3.205)

This equation is similar to the corresponding case in the TBL models, with
a modification in the definition of ξ . From the theory of quartic equations,
√
this equation admits a real positive root for ξΛ < ξcrit = −(26 + 15/ 3)/2.
Therefore, for a given central density F0 , and the inhomogeneity parameter
as given by F3 , the naked dust singularity can be partly covered by a positive
Λ, because there is an additional positive term in the denominator in ξΛ . But
it is interesting to note that, however large, a finite Λ term cannot completely
cover the corresponding visible part of dust models.
In a similar manner, it is easy to see that a negative Λ will open up some
covered part in the dust collapse final state spectrum. As such, the discussion
above does not use the positivity of Λ, so the discussion goes through even
 3.7 Equation of state 129

for negative Λ. For all values of n > 3, λ0 = ∞, and there cannot be a real
positive root in these cases. The final singularity is then hidden behind the
event horizon.
It can therefore be seen that the gravitational collapse of a dust cloud
with a non-zero cosmological term can develop into a blackhole, a naked
singularity, and even a globally regular solution, as the final outcome of
a collapse. Each of these outcomes is determined by the choice of initial
parameters, given in terms of the density and velocity profiles of the cloud.
Although for simplicity, this was restricted to the f (r) = 0 case while ana-
lyzing the structure of the singularity, the results can be extended to the
general case. The main aim here has been to examine how the presence of a
non-zero cosmological term affects the dynamical evolution of the collapse.
It has been found that there is a non-trivial detailed dynamical structure,
especially for the f < 0 case, and this provides a strong motivation for
the present study, namely how the introduction of a cosmological term non-
trivially changes the collapse outcomes. In most cases, there is no rebounce,
and then the nature of the singularity forming as the endstate of the col-
lapse is studied. At a more technical level, this is reflected in the change in
the structure of the roots that characterize the outgoing null geodesics, as
discussed above.
In particular, this analysis shows that it is possible to cover a part of the
naked singularity spectrum in the corresponding critical branch of solutions
in the TBL models, when a positive cosmological term is introduced. How-
ever, the naked singularity of dust collapse is never removed, irrespective
of the value of Λ. In this sense, the naked singularity remains stable to the
introduction of a cosmological term.
It can be seen that there are certain important changes in the collapse
dynamics with the introduction of a cosmological term, and the study here
brings this out, allowing the implications of a non-zero Λ towards the final
outcome of gravitational collapse to be understood. These results are of
interest in view of the recent observational claims about a non-vanishing
cosmological constant.

3.7 Equation of state

Here, type I matter fields, which is a rather general form of matter, have
been worked with, and no specific equation of state has been imposed so
far. However, it is important to note that suitable care must be taken in
interpreting these results. Whilst it has been shown that the initial data and
dynamical evolutions chosen do determine the blackhole and naked singu-
larity endstates for collapse, not all these dynamical variables are explicitly
determined by the initial data given at the initial epoch in terms of the matter
130 Spherical collapse

and metric variables. Hence, these functions are fully determined only as a
result of the time development of the system from the initial data, provided
the relation between the density and pressures, that is, a given ‘equation of
state’ is known.
In principle, it is possible to choose these functions freely. For example,
one could specify the matter and velocity profiles at the initial epoch and
also the dynamical evolutions, such as F (v, r) and ν (v, r), only subject to
an energy condition and regularity. Note that v here plays the role of a time
coordinate and this then fully determines the collapse evolution. One can
then calculate the energy density, and the radial and tangential pressures for
the matter. However, in such a case, the resultant ‘equation of state’ could be
quite strange in general. If any equation of state of the form pr = f (ρ) and
pθ = g (ρ) is given, then it is clear from (3.28) and (3.29) that there would be
a constraint on the otherwise arbitrary functions M and A, which specify
the required class, if the solutions of the constraint equations exist. As for the
physics of very high density matter, presently, there is very little idea on what
kind of an equation of state the matter would follow, especially at very high
densities. Closer to the collapse endstates ultra-high energy densities and
pressures are certainly important. Hence, if the possibility that the property
of the matter fields or the equation of state could freely be chosen as above
is allowed for, then the analysis is certainly valid and gives several useful
conclusions on possible collapse endstates. In such a case, it is also possible
that the chosen equation of state will be in general such that the pressures
may explicitly depend not only on the energy density, but also on the time
coordinate.
From such a perspective, it is now shown and pointed out below that the
analysis, as given above, does in fact include several well-known equations
of state and useful classes of collapse models as special cases. It is also seen
that the energy conditions are satisfied throughout the collapse.
Discussed above was the idealized class of dust collapse models where the
pressures were taken to be vanishing. This has been studied extensively and
has yielded many important insights into collapse evolutions. In this special
case, the Einstein equations can be solved completely and some remarks
are made below on the N -dimensional generalization of the usual TBL dust
collapse models. The metric in this case is given as


R2
ds = dt −
2 2
dr2 − R2 (t, r)dΩ2N −2 . (3.206)
1 + r2 b0 (r)

The equations of motion are then written as

(N − 2)F 
=ρ (3.207)
2R(N −2) R
 3.7 Equation of state 131

and
F (r)
Ṙ2 = + f (r). (3.208)
R(N −3)
In the case of dust, the mass function must be F = F (r), and hence the
regularity condition implies that

F (r) = r(N −1) M(r). (3.209)

The energy condition here gives, in the range 0 < r < rb , M(r) ≥ 0 and
3M + rMr ≥ 0. In this case now, the function X (v ) that was discussed
earlier is given as

1 1 v (N −3)/2 (M1 + v (N −3) b1 )dv
X (v ) = − , (3.210)
2 v (M0 + v (N −3) b00 )3/2

and the time taken for the central shell to reach the singularity is
 1
v (N −3)/2 dv
ts0 =  . (3.211)
0 M0 + v (N −3) b00
It is now seen clearly that any given sets of density and velocity profiles at
the initial epoch completely determine the tangent to the singularity curve
at the central singularity. Also, (3.80) becomes

(N −1)/2 N − 1
x0 = M0 X (0). (3.212)
2

It therefore follows that, given any specific density profile of the collapsing
dust cloud, a velocity profile can be chosen so that the endstate of the col-
lapse would be either a naked singularity or a blackhole, depending on the
choice made, such that the energy conditions are also satisfied throughout
the collapse. The converse also holds, namely a given velocity profile can be
chosen for the cloud at the initial epoch, and then there are density pro-
files that will lead the collapse to either the naked singularity or blackhole
final states. As seen, these conclusions hold irrespective of the number of
dimensions of the spacetime. In this sense, this treatment unifies and gen-
eralizes the earlier results of the dust collapse. The interesting point that
comes out here is, given an initial density profile for the collapsing cloud, the
space of velocity profile functions is divided into the regions that lead the
collapse either to a blackhole or a naked singularity evolution, depending on
the choice made, and the converse holds similarly.
While the dust equation of state discussed above is fairly standard and
extensively used, it is widely believed that pressures would play an impor-
tant role in gravitational collapse considerations (Jhingan and Magli, 2000;
132 Spherical collapse

Goncalves and Jhingan, 2001). Discussed below is a class of collapse models

with non-zero pressures, which is, however, idealized in the sense that while
the tangential pressure is allowed to be arbitrary, the radial pressure is taken
to be static. A special case of this situation would be the zero radial pressure
case. These are the purely tangential pressure collapse models that have been
analyzed extensively to investigate the gravitational collapse final states in
terms of blackhole and naked singularity formation.
As pointed out above, if the classes of collapse in which the radial pressure
remains static is considered, the constraint equation for M has the solution

M(r, v ) = m(r) − pr (r)v 3. (3.213)

In addition to this, if there is an equation of state of the form pθ = f (ρ),

then this gives the constraint equation for the function A(r, v ) as

2f (ρ) = pr + pr (R/R ) + A(r, v ),v [ρ(r, v ) + pr ]. (3.214)

For this class of models, the energy conditions are given by

3[m − pr v 3 ] + r[m,r − pr,r v 3 − 3pr v 2 v  ] ≥ 0, (3.215)

3[m − pr v 3 ] + r[m,r − pr,r v 3 − 3pr v 2 v  ] + pr R2 R ≥ 0, (3.216)
 
ρ 1  R
+ (ρ + pr )(A,v + 1) + pr  ≥ 0. (3.217)
2 2 R

As shown in Goswami and Joshi (2002), classes of functions m, pr , and A

exist such that the naked singularity is the endstate for the collapse, and
also the above three energy conditions are satisfied. It follows that, given the
initial matter profiles, classes of collapse evolutions satisfying the energy con-
ditions as seen above exist such that either the blackhole or naked singularity
endstates can result, subject to the above equation of state.
Finally, also discussed here is the perfect fluid with a linear equation of
state. This has been widely used in astrophysical considerations and a linear
equation of state is well-studied. In general, there have been many stud-
ies that discuss the gravitational collapse of fluids (see for example, Cahill
and Taub, 1971; Ori and Piran, 1987, 1990; Foglizzo and Henriksen, 1993;
Harada, 1998; Rocha and Wang, 2000; Harada and Maeda, 2001; Goswami
and Joshi, 2002; Ghosh and Deshkar, 2003; Giambo et al., 2004). Discussed
below is how the formalism outlined above applies to this case in order to find
the blackhole and naked singularity configurations as perfect fluid collapse
endstates.
 3.7 Equation of state 133

For an isentropic perfect fluid, whose pressure is a linear function of the

density only, the equation of state of the collapsing matter is given by

pr (t, r) = pθ (t, r) = kρ(t, r), (3.218)

where k ∈ [−1, 1] is a constant. The case k = 0 gives the dust case, and
k = 1 is the stiff fluid case. Presently only the case of positive pressures is
considered. In this case k > 0, and the energy conditions give

M,v < 0. (3.219)

From the above equation of state and the Einstein equations it can
immediately be seen that the function M is now the solution of the
equation
(N − 1)kM + krM,r + Q(r, v )M,v = 0, (3.220)
where
Q(r, v ) = (k + 1)rv  + v. (3.221)
Now, (3.220) has a general solution of the form

F (X, Y ) = 0, (3.222)

where X (r, v, M) and Y (r, v, M) are the solutions of the system of equations

dM dr dv
− = = . (3.223)
3kM kr Q
Therefore, it can easily be seen that (3.220) admits classes of solutions when
v  > 0. Also, solving the equation for the central shell r = 0, with boundary
conditions ρ → ∞ as v → 0,
m0
M(0, v ) = , (3.224)
v3k
where m0 is a constant. By choosing m0 > 0, the central shell can be made
to satisfy the energy condition ρ(t, 0) > 0 for all epochs. Then, by the
continuity of the density function, it can be said that an -ball exists around
the central shell, for
√ which v  (t, r) > 0 and also ρ(t, r) > 0. But at the
central singularity, vv  ≈ X (0). This implies that classes of solutions exist
that satisfy the energy conditions and also admit a naked singularity as the
collapse endstate. For a further discussion on perfect fluid collapse, and the
details of blackhole and naked singularity formation, see Goswami and Joshi
(2006) and references therein.
It is seen from the above discussion that several well-known classes of
collapse models and equation of states form subcases of the considerations
134 Spherical collapse

given here. Along with these models, the above analysis would work for any
other models with other equations of state, if solutions to the constraint
equations on M and A are permitted. Hence, it follows that the above
provides an interesting framework for the study of dynamical collapse, which
is one of the most important problems in gravitation physics today.
 4
Cosmic censorship

As indicated here, the cosmic censorship hypothesis is fundamental to the

basic theory and applications in blackhole physics. It follows from the pre-
vious considerations that the cosmic censorship, if it does hold, is not valid
in any obvious and plain manner in general relativity, but it has to be care-
fully designed and formulated. This is important for any clear basis and
foundation for blackhole physics. As yet, such a mathematical formulation
is not available, and much work is needed to achieve one. In such a case,
the existence and formation of blackholes as the final state of a gravitational
collapse, whenever a massive star collapses in the universe on exhausting its
nuclear fuel, cannot be taken as automatic.
Basically, cosmic censorship is a statement about the causal structure of
spacetime, as related to the dynamical collapse scenarios that are fundamen-
tal processes in astrophysics and cosmology. Therefore, in this chapter, which
aims to discuss various aspects related to the censorship conjecture and its
possible formulations, are considered in Section 4.1 some basic aspects of
the causal structure of spacetime. The physics related to gravity becomes
much more important in situations such as gravitational collapse and those
of the early universe phases in cosmology. The general theory of relativity
predicts the occurrence of spacetime singularity in such situations, which are
the extreme gravity regions where densities, spacetime curvatures, and other
physical quantities take extreme values. The quantum gravity effects would
become much more prominent in such regions. The occurrence of spacetime
singularities in collapses and cosmology is discussed here in Section 4.2.
The problem of cosmic censorship is then to ensure that the spacetime
singularities developing in a gravitational collapse scenario are necessarily
hidden within the event horizons of gravity, thus ensuring that the final prod-
uct is a blackhole only. Based on this assumption, some aspects of blackhole
physics are discussed in Section 4.3. The validity, or otherwise, of censor-
ship in higher-dimensional spacetimes is discussed in Section 4.4. The major

135
136 Cosmic censorship

difficulties in formulating a censorship hypothesis and any possible proof are

highlighted, and various possible avenues towards developing a mathematical
statement are considered in Section 4.5.
The basic approach here has been that such a formulation, as well as any
possible proof for the censorship, cannot be achieved without an extensive
study and investigation of the gravitational collapse scenarios, such as those
conducted in the previous chapter. Even if naked singularities do form in the
collapse of massive matter clouds in violation of censorship conjecture, their
genericity and stability still remain important issues. There are, however, no
definite criteria in general relativity to test stability and genericity, which
are rather wide spectrum notions in gravitation theory today, and a range of
possibilities are yet to be discussed and explored. These issues are considered
in Section 4.6.
The conclusion that emerges from these considerations on cosmic cen-
sorship here is that one may not hope to formulate this conjecture in any
definitive manner mathematically without an extensive study of gravitational
collapse models in general relativity. On the other hand, a study of the prop-
erties and structure of naked singularities found in the models studied so
far, and especially that of the physical processes in the vicinity of these
extreme gravity regions that are visible, could lead to interesting physical
implications. This latter possibility is discussed in the next chapter.

4.1 Causal structure

The causal relationships between events in a spacetime are determined by
the Lorentzian metric, which decides the past and future light cones at every
event of the manifold. While, locally, the geometry is always that of special
relativity, the global behavior of light cones now depends on the choice of the
metric tensor, which is a function of space and time coordinates. Therefore,
even though locally the speed of light is not to be exceeded, globally the
phenomena, such as the occurrence of closed timelike curves and causality
violations, may be allowed in principle. The possibility of causality viola-
tions really depends on the overall global topology of the spacetime. Various
causality conditions are imposed on the model under consideration in order
to ensure a regular causal structure of the spacetime, as will be specified here.
While the Einstein equations govern the local dynamics of the matter and
spacetime curvature, they say nothing about the global topology or causal
structure of the spacetime universe. These are free to be chosen as per the
appropriate model under consideration and the cosmological observations of
the universe as a whole. As has been realized in the last few decades, these ele-
ments, namely the topology and causal structure, can have significant physi-
cal implications for cosmology and the universe as observed on larger scales.
 4.1 Causal structure 137

The spacetime model worked with must satisfy several physical reasonabil-
ity conditions in order to represent the physical universe as observed. One
of these is that the spacetime should have a regular causal structure. The
physical condition that no material particle signals can travel faster than the
velocity of light fixes the causal structure for the Minkowski spacetime. Also,
in general relativity, which uses the framework of a general spacetime mani-
fold, it was noted that locally the causality relations were the same as in the
Minkowski spacetime, which is the background manifold for special relativ-
ity. However, globally, there could be important differences in the causality
structure and other physical properties, due to a different spacetime topol-
ogy, as compared with that of the Euclidean space R4 of the Minkowski
spacetime (Geroch, 1967).
The strong gravitational fields that arise either in cosmology or gravita-
tional collapse, as signaled by the occurrence of spacetime singularities in
general relativity, also may have significant causality and topology implica-
tions for the structure of the spacetime in the vicinity of these ultra-strong
gravity regions. The causal structure of the spacetime has been studied in
detail, especially from the perspective of the occurrence of spacetime singu-
larities in collapse and cosmology (see for example, Geroch, 1970b; Penrose,
1972; Hawking and Ellis, 1973). Here, in this section, this causal structure
and its relationship with the spacetime topology are discussed. The known
results and definitions required for later chapters are also reviewed and some
results that are of intrinsic interest for the global aspects of spacetimes, and
which emphasize the close interplay between the causal structure and topo-
logy, are given. Basic causal structure ideas and definitions are given, and
topological properties of several spacetime sets and various causality con-
ditions are stated in order to arrive at a unified causality statement for a
reasonable model of spacetime. The results in the present section are valid
for manifolds of arbitrary dimensions greater than or equal to two.
An event p chronologically precedes event q , denoted by p  q , if there
is a smooth, future directed timelike curve from p to q . If such a curve is
non-spacelike, namely timelike or null, p causally precedes q , or p < q . The
chronological future I + ( p) of p is the set of all points q such that p  q . The
chronological past of p is defined similarly. Therefore,

I + ( p) = {q ∈ M | p  q}, (4.1)
I − ( p) = {q ∈ M | q  p}. (4.2)

The causal future ( past) for p is defined similarly:

J + ( p) = {q ∈ M | p < q}, (4.3)

J − ( p) = {q ∈ M | q < p}. (4.4)
138 Cosmic censorship

+
I (S)

p
q

Fig. 4.1 The future of a set S is the union of the futures of all the events in
this set.

The relations  and < are transitive, and for events p, q , and r, p  q and
q < r or p < q and q  r implies p < r (Penrose, 1972). It is seen from this
that
I + ( p) = J + ( p), (4.5)

and also
I˙+ ( p) = J˙+ ( p), (4.6)

where, for a set A, Ā is the closure of A and Ȧ denotes the topological

boundary. The chronological (causal) future of any set S ⊂ M is defined
similarly (see Fig. 4.1),


I + (S ) = I + ( p), (4.7)
p∈S


J + (S ) = J + ( p), (4.8)
p∈S

and the chronological (causal) pasts of the subsets of the spacetimes are
defined similarly. Such dual definitions or results will often be taken as
granted.
Suppose there is a future directed timelike curve from p to q . There is a
local region containing q that the timelike curve must enter in which special
relativity is valid. Therefore, there is a neighborhood N of q such that any
point of N can be reached by a future directed timelike curve from p. It is
thus seen that for any event p ∈ M , the sets I + ( p) and I − ( p) are open in M .
The above also implies that the sets I ± (S ) are open, as they are the union
of open sets in M . However, the sets J ± ( p) are neither open nor closed in
general (see for example, Hawking and Ellis, 1973, or Joshi, 1993). Therefore,
 4.1 Causal structure 139

all the points in the boundary of J + (S ) are not necessarily connected to a

point in S by a null geodesic generator.
In Minkowski spacetime, the set I + ( p) is the set of those points that are
reached by future directed timelike geodesics from p, and the boundary of
this set is generated by null geodesics from p. As seen above, this is not true
for an arbitrary spacetime in general, but locally this property is still valid as
shown by the following result from Hawking and Ellis (1973). If (M, g ) is a
spacetime manifold and N is a convex normal neighborhood of p ∈ M , then,
for any q ∈ N ∩ I + ( p), there is a timelike geodesic from p to q in N , and
the boundary of I + ( p) in N is generated by future directed null geodesics
from p in N .
To derive properties of more general boundaries, a future set can be
defined. A set F of M is called a future set if F = I + (S ) for some sub-
set S of M . An equivalent criterion is I + (F ) ⊂ F . Past sets are defined
similarly. Clearly, future sets are open in M as they are the union of I + ( p)
for all p ∈ S .
To see the properties of future sets, note that if F is a future set, F is the
set of all points x such that I + (x) ⊂ F . To see this, suppose x is such that
I + (x) ⊂ F . Then, a sequence in I + (x) and hence in F can be constructed
with the limit point x. Therefore x ∈ F . Conversely, if x ∈ F , then take
y ∈ I + (x). This gives x ∈ I − (y ), which contains an open neighborhood of
x. Then, this neighborhood contains points of F , that is, y ∈ I + (F ), which
implies I + (x) ⊂ F .
Again, the boundary of a future set F is made of all events x such that
I (x) ⊂ F but x ∈ F . If x ∈ Ḟ , then clearly x ∈ F as F is an open set. But
+

x ∈ F , and as seen above, I + (x) ⊂ F . Conversely, let x ∈ F but I + (x) ⊂ F .

Then a sequence in I + (x) converging to x can clearly be constructed, which
implies x ∈ F . Therefore, x must be in the boundary of F .
A set S is called achronal if no two points of S are timelike related, that
is, I + (S ) ∩ S = ∅. If F is a future set, the following result from Hawking
and Ellis (1973) shows that the boundary of F is a well-behaved achronal
manifold. Let F be a future set, then the boundary of F is a closed, achronal
C 0 -manifold that is a three-dimensional embedded hypersurface.
The following result from Penrose (1972) then shows that the achronal
boundary of a future set F is always generated by null geodesics that are
either past endless or always have a past end point on F . Let S ⊂ M and
p ∈ I˙+ (S ) − S . Then, a null geodesic contained in the boundary of I + (S )
exists with a future end point p, and which is either past endless or has a
past end-point on S .
From the above, it follows now that if q ∈ J + ( p) − I + ( p), then any
non-spacelike curve joining p and q must be a null geodesic, and that the
boundary of I + ( p) or J + ( p) is generated by null geodesics that have either
a past end point at p or are past endless.
140 Cosmic censorship

Causal relations in a spacetime are defined by the existence of smooth

non-spacelike curves between pairs of events. It is, however, useful to extend
this to define causality by means of continuous curves. This is carried out
by requiring that pairs of points on a curve are locally joined by a smooth
timelike or causal curve. To be precise, a continuous curve λ is called a
future directed timelike (or non-spacelike) curve if each x ∈ λ is contained in
a convex normal neighborhood N , such that if λ(t1 ), λ(t2 ) ∈ N with t1 < t2 ,
then there is a smooth future directed timelike (non-spacelike) curve in N
from λ(t1 ) to λ(t2 ). Such curves are regarded as equivalent under a one–one
continuous reparametrization.
It is useful to introduce the notion of future and past inextendible non-
spacelike curves, which are effectively the trajectories that have no future
or past end points. Let λ be a non-spacelike curve. Then p ∈ M is called a
future end point of λ if, for every neighborhood N of p, a value of the curve
parameter t exists such that for all t > t , λ(t) ∈ N . The past end point
is defined similarly. It is clear that if λ has an end point, it must be unique
because M is Hausdorff. The curve λ is called future or past inextendible if
it has no future or past end point respectively in M .
An inextendible causal curve might be running off to infinity, or it might
end up in a spacetime singularity, or it could enter a compact set in which
it could be trapped to go round and round for ever.
Note that the study of casual relations in a spacetime (M, g ) is equivalent
to that of the conformal geometry of M . Let (M, g ) be the physical space-
time, and consider the set of all conformal metrics ḡ , where ḡ = Ω2 g , with
Ω being a non-zero C r -function. Then if p  q or p < q in (M, g ), the same
relation is preserved in (M, ḡ ) for any conformal metric ḡ . Therefore, casual
relationships are invariant under a conformal transformation of the metric.
However, non-spacelike geodesics in (M, g ) will no longer be geodesics in
(M, ḡ ) unless they are null. The null geodesics are conformally invariant up
to a reparametrization of the affine parameter along the curve. Therefore,
specifying causal relations in M fixes the spacetime metric up to a conformal
factor. Let p be an event in M , and N be a convex normal neighborhood of
p. Minkowskian coordinates {xi } can be introduced in N , in which case it
follows from the above that the set of events in N that are causally connected
to p satisfy
−(x0 )2 + (x1 )2 + (x2 )2 + (x3 )2 ≤ 0. (4.9)

The boundary of these points define the null cone in the tangent space
Tp . Suppose now that W and Z are any two non-null vectors in Tp ,
then,

g (W , Z ) = 1
2
[ g (W + Z, W + Z ) − g (W , W ) − g (Z, Z )] . (4.10)
 4.1 Causal structure 141

Now, if X, Y ∈ Tp are a timelike and a spacelike vector respectively, then

the equation

g (X + λY , X + λY ) = g (X, X ) + 2λg (X, Y ) + λ2 g (Y , Y ) = 0 (4.11)

has two distinct roots λ1 and λ2 , as g (X, X ) < 0 and g (Y , Y ) > 0. The
knowledge of the null cone then implies that λ1 and λ2 can be determined
in principle. But λ1 λ2 = g (X, X )/g (Y , Y ). Hence, the null cone gives the
ratio of the magnitudes of a timelike and a spacelike vector. Therefore, each
term in the equation above is determined up to a factor, and so g (W , Z ) is
determined up to a factor.
The local causality principle implies that, over small regions of space and
time, the causal structure is the same as in the special theory of relativity.
However, on a larger scale global pathological features, such as the violation
of time orientation, possible non-Hausdorff nature or non-paracompactness,
disconnected components of spacetime, and such others, may show up.
Such pathologies are to be ruled out by means of reasonable topological
assumptions, and one would like to ensure that the spacetime is causally
well-behaved. This is carried out by means of introducing various causal-
ity conditions such as the non-occurrence of closed timelike or non-spacelike
curves (causality), and the stability of this condition under small perturba-
tions in the metric (stable causality). In fact, Carter (1971) pointed out that
there is an infinite hierarchy of such causality conditions for a spacetime.
It would appear reasonable to demand that physically realistic spacetimes
do not allow either closed timelike or closed non-spacelike curves, as this
would give rise to the phenomenon of entering one’s own past. However,
general relativity and Einstein’s equations as such do not rule out such a
possibility on their own. For example, the Gödel universe (Gödel, 1949) has
closed timelike curves through each point of the spacetime. Again, the global
topology of M can cause closed timelike curves. For example, for the cylinder
M = S 1 × R, obtained from the Minkowski spacetime by identifying t = 0
and t = 1 hypersurfaces with the metric given by ds2 = −dt2 + dx2 , the
circles x = const., are closed timelike curves. In fact, for all p ∈ M here,
I + ( p) = I − ( p) = M . Such examples could be discarded as mathematical
pathologies in the spacetime topology, and the Gödel universe may be termed
unrealistic because it is a rotating model that does not correspond to the
observed universe.
More difficult to rule out are the Kerr solutions of a spinning gravitational
source (Kerr, 1963), which contain closed timelike curves if the rotation is
sufficiently fast in comparison with the value of the mass parameter. These
could possibly represent the final fate of a massive collapsing star that is
rotating. If a star failed to get rid of enough spin during the process of col-
lapse, it would give rise to a time machine in the spacetime. Wormholes in
142 Cosmic censorship

a spacetime representing the multiply connected nature of the topology of

space could also give rise to closed timelike curves, as indicated earlier. The
physical significance and acceptability of such causality violations have been
examined by Morris, Thorne, and Yurtsever (1988), and by Friedman et al.
(1990). It follows from these considerations that such wormhole spacetimes
do admit unique solutions to at least simple field equations, such as a single
non-interacting scalar field. This shows that even though the causality vio-
lation is perceived as contradictory to the predictability requirements in the
spacetime, in general this need not be so, as there are wormhole spacetimes
where this is not the case.
Spacetimes with closed non-spacelike curves are avoided by requiring M
to satisfy the causality condition; that is, (M, g ) does not admit any closed
timelike or null curves. A spacetime M is said to be chronological when it
admits no closed timelike curves, that is, p ∈ I + ( p) for all p ∈ M . When M
admits no closed non-spacelike curves, it is said to be causal. If M = S 1 × R
and the metric is chosen to be ds2 = dt dx, then M is chronological but not
causal. The circles x = const. are null geodesics here.
The causally well-behaved nature of M turns out to be closely related to
the topological structure of M in that if M is chronological, M cannot be
compact. To see this, suppose M is compact. The sets {I + ( p) | p ∈ M }
cover M . Now, compactness implies that a finite set of points p1 , . . . , pn
exists, such that the set I + ( p1 ) ∪ · · · ∪ I + ( pn ) covers M . Now, p1 , . . . , pn
must be in the cover implies that p1  pi1 for some i1 in 1, . . . , n. Therefore,
p1  pi1  . . .  pin−1 , where all n-points have been exhausted. Hence,
pik  pik for some ik that violates chronology.
Even though a spacetime may be causal, it could be on the verge of vio-
lating causality. Consider (M, g ) given by ds2 = dt dx + t2 dx2 . To see the
behavior of light rays, consider the null geodesics of this spacetime that are
given by dt/dx = −t2 . At t = 0, dt/dx = 0, and so one arm of the light cone
will lie along the x-axis. In this situation, there is a spacetime that is causal,
but a non-spacelike curve from p can enter arbitrary small neighborhoods of
p. For distinct events p and q along the x-axis, I + ( p) = I + (q ). To avoid such
a causal pathology, the distinguishing condition can be imposed, namely that
for all p and q in M , I + ( p) = I + (q ) implies p = q and I − ( p) = I − (q ) implies
p = q . A similar but stronger condition is strong causality (Penrose, 1972),
which states that for all events p ∈ M , every neighborhood of p contains
a neighborhood of p that no non-spacelike curve in M intersects more than
once. If strong causality is violated at p, then there are non-spacelike curves
from neighborhoods of p that come arbitrary close to intersecting themselves.
The relevance of strong causality is that it determines the topology of
spacetime. First, note that the Alexandrov topology or the interval topology
on M is defined by taking the collection {I + ( p) ∩ I − (q ) | p, q ∈ M } as the
basis for the topology on M . The following from Penrose (1972) can then
 4.1 Causal structure 143

be shown. Let (M, g ) be a strongly causal spacetime. Then, the manifold

topology on M is the same as the Alexandrov topology.
Therefore, for a strongly causal spacetime, the causal relations determine
the topological structure of M . It can also be shown that the Alexandrov
topology will be Hausdorff for a strongly causal spacetime. Next, strong
causality imposes a regularity on M in the sense that if γ is any future
inextendible non-spacelike curve, then γ cannot be either totally or partially
future imprisoned in any compact set S in the sense that γ will enter and
remain within S , or it will continually re-enter S . Therefore, for a strongly
causal M , the future endless curve γ must go either to infinity or terminate in
a spacetime singularity, that is, it must go to the boundary of the spacetime
in either case.
Even though (M, g ) is strongly causal, it is possible to create examples
where a small perturbation in the metric tensor components gij will give rise
to closed timelike curves. Therefore, a strongly causal spacetime could still
be on the verge of causality violation. Now, general relativity is supposed
to be a classical approximation of some, as yet unknown, quantum theory
of gravity in which a precise measurement of the metric components at a
single event would not be possible. Therefore, one would like the causality
of spacetime to be preserved under small perturbations in the metric. This
is achieved by requiring the spacetime to be stably causal. Let (M, g ) be a
spacetime, then another metric g  > g exists, such that there is no closed
non-spacelike curve in g  if there was none in g . Here, g  > g means that
every non-spacelike vector in g is always timelike in g  . It is of course clear,
by definition, that if g  < g then g  will have no closed non-spacelike curves
if g had none.
Of the infinite hierarchy of causality requirements available for a spacetime,
stable causality may be considered to be the most relevant physically and
the one that provides a unified criterion for basic causal regularity. Such a
spacetime is causal and its stability is ensured in a suitable manner. A stably
causal M is strongly causal and hence causality determines the topology of
the spacetime. Finally, a stably causal spacetime admits the existence of a
global time function on M , that is, a smooth scalar field f , for which the
gradient is always timelike (Hawking and Ellis, 1973). Such a function assigns
a value of ‘time’ to each point in M so that this time is strictly increasing
along all timelike curves. This is precisely the property that is desired for a
global assignment of a time coordinate in M . Of course, such a time function
is not unique.
A spacetime (M, g ) is called globally hyperbolic if the sets J + (x) ∩ J − (y )
are compact for all x and y in M , and M is causal. For a globally hyperbolic
spacetime, the light cones are necessarily closed. To see this, suppose J + (x)
is not closed, and let y ∈ J + (x), but y ∈ J + (x). Let z ∈ I + (y ). Then,
y ∈ J + (x) ∩ J − (z ), but y ∈ J + (x) ∩ J − (z ). This is a contradiction as
144 Cosmic censorship

q
+
t=1
I ( p)
∩
–
I (q)

p t=0

Σ t = –1

Fig. 4.2 A globally hyperbolic spacetime has a fixed topology M = Σ×R, and is
foliated entirely by spacelike surfaces Σ. For any events p and q, the intersection
I + (p) ∩ I − (q) is compact. Such a spacetime obeys strong cosmic censorship and
allows no naked singularities.

J + (x) ∩ J − (z ) is closed as it is compact. It can also be shown that if S is

a compact set, J + (S ) will be closed in M , and if S1 and S2 are any two
compact subsets, J + (S1 ) ∩ J − (S2 ) must be compact.
From earlier discussions, it can be shown that if M is globally hyperbolic,
then it is strongly and stably causal. In fact, global hyperbolicity is a rather
strong condition on M that uniquely fixes the overall topology of the space-
time (see Fig. 4.2). The spacetime then has a very regular global behavior,
and all the various pathological features stated in the above are removed.
Physically interesting spacetimes such as the Schwarzschild solution, the
Friedmann–Robertson–Walker cosmological solutions, and the steady state
models are all globally hyperbolic.
Another strong motivation for global hyperbolicity is the cosmic censorship
hypothesis, which rules out timelike or naked singularities. It turns out that
if even the locally naked singularities are ruled out, the resulting spacetime
structure must be globally hyperbolic. In the absence of cosmic censorship,
the deterministic structure or global hyperbolicity breaks down as a Cauchy
horizon develops.
The global hyperbolicity of M is closely related to the future or past devel-
opment of initial data from a given spacelike hypersurface in the spacetime.
Let S be a closed achronal set. The edge of a set S , E (S ), is defined as a set
of points x ∈ S such that every neighborhood of x contains y ∈ I + (x) and
z ∈ I − (x) with a timelike curve from z to y that does not meet S . When
S is a closed achronal set without an edge, the result is that S is a three-
dimensional, embedded, C 0 -submanifold of M . A partial Cauchy surface S is
defined as an acausal set without an edge. Therefore, no non-spacelike curve
intersects S more than once and S is a spacelike hypersurface. The future
 4.1 Causal structure 145

domain of dependence of S , denoted by D+ (S ), is defined as the set of all

points x ∈ M such that every past inextendible non-spacelike curve from
x intersects S . It is clear that S ⊂ D+ (S ) ⊂ J + (S ) and as S is achronal,
D+ (S ) ∩ I − (S ) = ∅. The past domain of dependence D− (S ) for a partial
Cauchy surface S is defined similarly. The full domain of dependence for S
is defined as
D(S ) = D+ (S ) ∪ D− (S ). (4.12)
A partial Cauchy surface is called a Cauchy surface or a global Cauchy sur-
face if D(S ) = M . Clearly, for a Cauchy surface S , edge(S ) = ∅. Every
non-spacelike curve in M must meet S once (and exactly once) if S is a
Cauchy surface. The relationship between the global hyperbolicity of M and
the notion of a Cauchy surface is that M is globally hyperbolic if and only
if it admits a spacelike hypersurface S that is a Cauchy surface for M .
Note that even when M is globally hyperbolic, all spacelike surfaces in M
need not be Cauchy surfaces. For example, for the Minkowski spacetime, the
spacelike surfaces t = const. are global Cauchy surfaces, but the hyperboloids

t2 − x2 − y 2 − z 2 = const. (4.13)
are not, as the past or future null cones of the origin are boundaries of the
domain of dependence for these spacelike surfaces. Again, if a point from the
Minkowski spacetime is deleted, the resulting M admits no Cauchy surface
and M is not globally hyperbolic.
It was shown by Geroch (1970a) that a globally hyperbolic M has a unique
topological structure, and it admits no topology change in that M is then
homeomorphic to R × S , where S is a three-dimensional submanifold and
for each t ∈ R, {t} × S is a Cauchy surface for M .
The basic idea of the proof for the above involves introducing a finite
measure µ on M so that µ(M ) = 1. Then, a function h : M → R is
introduced by
µ(J + ( p))
h+ ( p) = . (4.14)
µ(J − ( p))
The function h− is defined similarly. The sets h± = const. are seen to
be Cauchy surfaces for M . Such functions are called causal functions on
a spacetime.
It can be asked if the converse is true in some sense, that is, whether the
direct product spacetimes (M, g ) with M = S × T , where each S × {t} is
spacelike and each {x} × T is timelike, are always globally hyperbolic. The
answer is no; refer to Clarke and Joshi (1988), who discuss a spacetime that
is a direct product as above, but not globally hyperbolic.
Next, let S be a partial Cauchy surface. Then N = D+ (S ) ∪ D− (S ) = M
and N must be a proper subset of M . The boundary of N in M can be
146 Cosmic censorship

divided into two regions H + (S ) and H − (S ) that are respectively called the
future and past Cauchy horizons of S ,

H + (S ) = {x | x ∈ D+ (S ), I + (x) ∩ D+ (S ) = ∅}. (4.15)

The past horizon H − (S ) is defined in a similar manner. Even though M

may not be globally hyperbolic and S is not a Cauchy surface, the region
Int(D+ (S )) or Int(D− (S )) is globally hyperbolic in its own right and the
surface S serves as a Cauchy surface for the manifold Int(N ). Therefore,
H + (S ) or H − (S ) represent the failure of S to be a global Cauchy surface
for M .
The set H + (S ) is achronal and closed. To see this, suppose x, y ∈ H + (S )
with x  y . Then, x ∈ I − (y ) and a neighborhood Nx ⊂ I − (y ) exists. Let
p ∈ Nx ∩ I + (x). Then, p  y and every past directed timelike curve from
p can be extended to be a past directed curve from y , which must meet
S . Therefore, p ∈ D+ (S ), which is a contradiction to D+ (S ) ∩ I + (x) = ∅.
Hence, no two points of H + (S ) are timelike related. Next, let x be a limit
point of a sequence of points {xn } in H + (S ). Suppose γ is a past directed
timelike curve not meeting S , with a future end point at x. Then, for any
y ∈ γ , I + (y ) contains all points {xn } for some n ≥ k . Then, coming from
xn to y and following γ , gives a non-spacelike curve from xn not meeting S ,
which is contradictory to xn ∈ D+ (S ).
As in the case of the boundary of a future set, H + (S ) is generated by
null geodesics that are either past inextendible in H + (S ) or have a past
end point on the edge of S (Geroch, 1970a). The basic physical relevance
of global hyperbolicity on M is that it implies a deterministic structure for
the spacetime in the sense characterized by the above results, given initial
data defined on a Cauchy surface S . The notion of global hyperbolicity was
introduced by Leray (1952), and it is seen from considerations on the Cauchy
problem in general relativity (see for example, Wald, 1984, for a review), that
if N is a globally hyperbolic subset of M , the wave equation for a δ -function
source at any p ∈ N has a unique solution that vanishes outside N − J + ( p).
The definition of global hyperbolicity given by Leray (1952) involves the
space C ( p, q ), which is the set of all C 0 non-spacelike curves from p to q
that are the same up to a reparametrization. Defining a C 0 -topology on this
space of curves, by stating that the two curves are nearby if their points
in M are close enough, allows the global hyperbolicity to be characterized
in term of C ( p, q ) as seen by Seifert (1967). Let M be a strongly causal
spacetime. Then, M is globally hyperbolic if and only if C ( p, q ) is compact
for all p, q ∈ M .
An important property of globally hyperbolic spacetimes, or any globally
hyperbolic subset in M , which is relevant for the singularity theorems is
the existence of maximum length non-spacelike geodesics between pairs of
 4.1 Causal structure 147

causally related events. In a complete Riemannian manifold with a positive

definite metric, any two points can be joined by a geodesic of minimum
length, and, in fact, such a geodesic need not be unique. The analog of this
result for Lorentzian metrics was given by Avez (1963) and Seifert (1967). If
(M, g ) is globally hyperbolic and p, q ∈ M such that p < q , then there is a
non-spacelike geodesic from p to q whose length is greater than or equal to
that of any other future directed non-spacelike curve from p to q .
Therefore, in globally hyperbolic spacetimes, there is a finite upper bound
on the proper time lengths of non-spacelike curves between two chronologi-
cally related events. It is clear that there is no lower limit of lengths for such
curves except zero, because the chronologically related events can always be
joined using broken null curves that could give an arbitrary small length curve
between them. Similarly, if S is a Cauchy surface in a globally hyperbolic
spacetime, then for any point p in the future of S , there is a past directed
timelike geodesic from p orthogonal to S that maximizes the lengths of all
non-spacelike curves from p to S .
The ideal points boundary of the spacetime (that is, the points at infinity
and singularities) are now discussed. Although the points at infinity and sin-
gularities in a spacetime are not regular points for M , they can be attached
to M as an additional boundary. Such a boundary construction including
both the points at infinity and singularities was provided by Geroch, Kron-
heimer, and Penrose (1972), based only on the causal structure of spacetime.
Here, this attached causal boundary or ideal points will be classified using
the properties of causal functions.
A non-empty subset P of M is said to be a past set if there is some
A ⊂ M such that I − (A) = P . If P cannot be expressed as the union of
two proper past subsets, then it is called an indecomposable past set (IP).
If P is an IP and if there is some x ∈ M with I − (x) = P , then P is
known as a proper IP, or PIP. If an IP set is not a PIP, then it is termed
a terminal IP, or TIP. The definitions of future sets, indecomposable future
sets (IFs), PIFs, and TIFs are similar. For many of the statements and
propositions the dual, or similar definitions, are taken for granted. Geroch,
Kronheimer, and Penrose (1972) proved that a set P ⊂ M is an IP if and
only if there is a future directed timelike curve γ such that I − (γ ) = P . Now,
let M + be the union of M̂ and M̌ , which are unions of all IPs and IFs in
M respectively. Then, avoiding duplication in M + , M ∗ can be defined as
the quotient space M + /Rh , where Rh is the intersection of all equivalence
relations R ⊂ M + × M + for which M + /R is Hausdorff. In this case, M ∗
can be viewed as a spacetime with boundary M ⊂ M ∗ , and the topology
of M can be looked upon as the induced topology of M ∗ . Throughout the
discussion here, M has been assumed to be distinguishing.
A point x ∈ M is said to be a regular point if it is represented by a PIP or
PIF. All other points in M ∗ are represented by TIPs or TIFs and are called
148 Cosmic censorship

the ideal or boundary points of M . A curve γ in M is taken as a continuous

map of a general interval into M . The result of Geroch, Kronheimer, and
Penrose (1972) is useful here. Let P = I − (γ ) be an IP where γ is a future
directed timelike curve. Then P is a PIP if and only if γ has a future end
point, and P is a TIP if and only if γ is future inextendible without a future
end point.
Now, the PIPs (and PIFs) can be characterized using the causal functions
introduced earlier. Let γ be a future directed timelike curve. Then, I − (γ ) is
a PIP if and only if h− attains its maximum value along γ . To see this, if
I − (γ ) is a PIP, then γ has a future end point p as seen above. It follows that
I − (γ ) = I − ( p). Now, as M is distinguishing, h− is strictly increasing along
γ , and hence has a maximum value at the future end point p. Conversely,
let h− attain its maximum h− ( p) for some p ∈ γ . Now, if p is not the future
end point of γ , then some q  p exists with q ∈ γ , and h− (q ) > h− ( p) by
the strict increasing nature of h− , which contradicts the hypothesis.
The above characterization can be stated in the following manner also. For
a future directed timelike curve γ , I − (γ ) is a PIP if and only if h+ attains
its minimum value along γ . It now follows that for such a curve, I − (γ ) is a
TIP if and only if either h+ → 0 along γ , or h+ → k with 0 . . . k . . . 1 along
γ , with h+ ( p) = k for any p ∈ γ . Along future endless curves in globally
hyperbolic spacetimes, h+ → 0, whereas the other situation will be realized
when points have been amputated from the spacetime, and the timelike curve
converges to such a cut without having a future end point. Therefore, the
causal boundary points in M ∗ are defined by precisely those timelike curves
along which h± do not realize their extremum values.
Note that the ideal points in M ∗ − M include both the points at infinity
and singularities characterized by the TIPs and TIFs of the spacetime. A TIP
is non-singular if a timelike curve generating it that has an infinite length
exists. All other ideal points are singularities of the spacetime. Now, let x
be a future (past) ideal point in M ∗ defined by a TIP, say I − (γ ) (for a TIF,
say I + (λ)). Then x will be called a future (past) 0-ideal point if h+ (or h− )
converges to zero along the curve, and x will be called a future (past) k-ideal
point if h+ (or h− ) converges to k with 0 . . . k . . . 1 along γ (or λ).
Therefore, causal functions classify all the ideal points for M into the
following eight categories:
(a) future (past) 0-ideal points that are singularities;
(b) future (past) k -ideal points that are singularities;
(c) future (past) 0-ideal points at infinity;
(d) future (past) k -ideal points at infinity.
As mentioned earlier, endless timelike curves in a globally hyperbolic
spacetime define ideal points of type (c). Timelike curves falling into a
Schwarzschild singularity provide an example for ideal points of type (a).
 4.2 Spacetime singularities 149

Finally, it is shown here that the naked singularities as defined by Penrose

(1974a) are all the boundary points of type (b). It needs to be decided
whether the singularities are always hidden behind an event horizon. The
question is whether naked or timelike singularities, which are defined as
follows, could also occur. Let x ∈ M ∗ be a TIP given by I − (γ ). Then, x is
called a future naked singularity if the future endless timelike curve γ has a
future singular end point and a point p ∈ M exists such that I − (γ ) ⊂ I − ( p).
Now let x be a future singular point defined by a future endless timelike
curve γ . As shown earlier, as γ is a TIP, either h+ → 0 along γ or h+ → k
with 0 . . . k . . . 1. Now, consider any sequence {pn } on γ with pn  pn+1
for all n. Then h+ ( pn ) is strictly decreasing along this trajectory. Since
I − (γ ) ⊂ I − ( p), it follows that pn  p for all n. Therefore, h+ ( pn ) is
bounded below by a non-zero fixed number h+ ( p), implying that h+ → 0.
This shows that x is an ideal point of type (b).

4.2 Spacetime singularities

Now the nature and existence of spacetime singularities in a general space-
time are considered. After Einstein proposed the general theory describing
the gravitational force in terms of spacetime curvature, and proposed the
field equations relating the geometry and matter content of the space-
time manifold, the earliest solutions found for the field equations were the
Schwarzschild metric representing the gravitational field around an isolated
body such as a spherically symmetric star, and the Friedmann cosmologi-
cal models. Each of these solutions contained a spacetime singularity where
the curvatures and densities were infinite and the physical description would
break down. In the Schwarzschild solution such a singularity was present at
r = 0, whereas in the Friedmann models it was found at the epoch t = 0,
which is the beginning of the universe and the origin of time where the scale
factor S (t) also vanishes and all objects are crushed to zero volume due to
infinite gravitational tidal forces.
Even though the physical problem posed by the existence of such a strong
curvature singularity in these solutions was realized, initially this phenomena
was not taken seriously. It was generally thought that the existence of such
a singularity must be a consequence of the very high degree of symmetry
imposed on the spacetime while deriving these solutions, which led to many
interesting physical applications. Subsequently, the distinction between a
genuine singularity and a mere coordinate singularity became clear, and it
was realized that the singularity at r = 2m in the Schwarzschild spacetime
was a coordinate singularity that could be removed by a suitable coordinate
transformation. It was clear, however, that the genuine curvature singularity
at r = 0 could not be removed by any such coordinate transformation.
150 Cosmic censorship

The hope was then that, when more general solutions were considered
with a lesser degree of symmetry requirements, such singularities would be
avoided.
This issue was sorted out when a detailed study of the structure of a
general spacetime and the associated problem of a spacetime singularity was
taken up by Hawking, Penrose, and Geroch (see for example, Penrose, 1968;
Geroch, 1971; Hawking and Ellis, 1973). It was shown by this work that a
spacetime will admit singularities within a very general framework, provided
that it satisfies certain reasonable assumptions, such as the positivity of
energy, a suitable causality assumption, and a condition such as the existence
of trapped surfaces. It thus follows that the spacetime singularities form a
general feature of the relativity theory. In fact, these considerations also
ensure the existence of singularities in other theories of gravity that are
based on a spacetime manifold framework and satisfy the general conditions
stated above.

4.2.1 The definition of a singularity

First, the meaning of a singular spacetime is discussed in some detail, and
the notion of a singularity is specified. It turns out that it is the notion of
geodesic incompleteness that characterizes the notion of a singularity in an
effective manner for a spacetime, and enables its existence to be proved by
means of general theorems. A variety of ways in which a spacetime exhibits
singular behavior, and the related notions of singular TIPs and TIFs are
then discussed.
The gravitational focusing caused by spacetime curvature in congruences
of timelike and null geodesics turns out to be the main cause of the exis-
tence of a singularity in the form of non-spacelike incomplete geodesics in
a spacetime. There are many theorems that establish the existence of the
non-spacelike geodesic incompleteness for the spacetime, either in the past
or future. However, these provide no information on the nature of these
singularities or their properties. In particular, these singularities could be
covered inside an event horizon of gravity, or they could be visible to exter-
nal observers if the trapped surfaces are delayed while the singularity forms
during dynamical processes in the spacetime.
The issue of the physical nature of a singularity is thus of much interest.
There are many types of singular behaviors possible for a spacetime, and
some of these could be regarded only as mathematical pathologies in the
spacetime, rather than having any physical significance. This will be espe-
cially so if the spacetime curvatures and other similar physical quantities
remained finite along an incomplete non-spacelike geodesic in the limit of
approach to the singularity. The criterion of Tipler, Clarke, and Ellis (1980)
 4.2 Spacetime singularities 151

as to when a singularity should be considered to be physically important in

terms of the curvature growth along singular geodesics will also be specified
here.
When should it be said that a spacetime manifold (M, g ) is singular, or
that it contains a spacetime singularity? As pointed out, several examples
of singular behavior in the spacetime models of general relativity are known.
Important exact solutions of the Einstein equations, such as the Friedmann–
Robertson–Walker cosmological models and the Schwarzschild spacetime,
contain a spacetime singularity where the energy density or the space-
time curvatures diverge strongly, and the usual description of the spacetime
breaks down.
In the Schwarzschild spacetime, there is an essential curvature singularity
at r = 0 in the sense that, along any non-spacelike trajectory falling into
the singularity, as r → 0, the Kretschmann scalar α = Rijkl Rijkl → ∞.
Also, all future directed non-spacelike geodesics that enter the horizon at
r = 2m must fall into this curvature singularity within a finite value of the
proper time (finite value of the affine parameter in the case of null geodesics).
Therefore, all such curves are future geodesically incomplete.
In the Friedmann–Robertson–Walker models, the Einstein equations imply
that if ρ + 3p > 0 at all times, where ρ is the total energy density and p is
the pressure, there is a singularity at t = 0 that could be identified as the
origin of the universe. If ρ + p > 0 at all times, then it is seen that along all
the past directed trajectories meeting this singularity ρ → ∞, and also that
the curvature scalar R = Rij Rij → ∞. Again, all the past directed non-
spacelike geodesics are incomplete in the above sense. Therefore, there is an
essential curvature singularity at t = 0 that cannot be transformed away by
any coordinate transformation. In fact, similar behavior has been generalized
to the class of spatially homogeneous cosmological models as shown by Ellis
and King (1974), which satisfy the positivity of energy conditions ρ ≥ 0, ρ ≥
3p ≥ 0 and 1 ≥ 3dp/dρ ≥ 0.
The existence of such singularities, where the curvature scalars and densi-
ties diverge, implies a genuine spacetime pathology where the usual laws of
physics must break down. The existence of the geodesic incompleteness in
these cases implies that, for example, a timelike observer suddenly disappears
from the spacetime after a finite amount of proper time.
Of course, singular behavior can also occur without bad behavior of the
curvatures. A simple example is the Minkowski spacetime with a point
deleted. With such a hole in the spacetime, there will be, for example, time-
like geodesics running into the hole, and hence they will be future incomplete.
This is clearly an artificial situation which one would like to rule out in gen-
eral by requiring that the spacetime is inextendible, that is, it cannot be
isometrically embedded into another larger spacetime manifold as a proper
subset.
152 Cosmic censorship

It is, however, possible to give a non-trivial example of the singular behav-

ior of the above type, where a conical singularity exists in the spacetime
as shown by Ellis and Schmidt (1977). Here, the spacetime is inextendible
and the curvature components do not diverge in the limit of approach to the
singularity. This is a behavior similar to that occurring in a Weyl type of
solution. The metric is given by

ds2 = −dt2 + dr2 + r2 (dθ2 + sin2 θ dφ2 ), (4.16)

with the range of coordinates given by −∞ < t < ∞, 0 < r < ∞, 0 < θ < π ,
but with 0 < φ < a, with φ = 0 and φ = a identified and a =  2π . There is a
conical singularity at r = 0 through which the spacetime cannot be extended
and the singular boundary is related to the timelike two-plane r = 0 of the
Minkowski spacetime.
The important question that arises is whether such singularities develop
even when a spacetime of generality is considered and if so, under what con-
ditions. In order to consider this question, it is first necessary to characterize
more precisely what is meant by a spacetime singularity.
While trying to characterize a spacetime singularity, the first point to note
is that by very definition, the metric tensor has to be well-defined at all the
regular points of the spacetime. Since this is no longer true at a spacetime
singularity such as those discussed above, a singularity cannot be regarded
as a regular point of the spacetime, but must be treated as a boundary
point attached to it. This situation causes difficulty when one attempts to
characterize a singularity by the criterion that the curvatures must blow up
near the singularity. The trouble is, since the singularity is not a part of the
spacetime, it is not possible to define its neighborhood in the usual sense in
order to discuss the behavior of curvature quantities in that region.
Characterizing the singularity could be attempted in terms of the diver-
gence of the components of the Riemann curvature tensor along non-spacelike
trajectories of the spacetime. The trouble with this is that the behavior of
such components will, in general, change with the change of frames used, and
this approach is not really of much help. The curvature scalars or the scalar
polynomials in the metric and the Riemann tensor could be used, and they
could be required to achieve unboundedly large values. This is the behavior
encountered in the Schwarzschild and the Friedmann models. However, it is
possible that such a divergence of curvature scalars occurs only at infinity for
a given non-spacelike curve. In general, it looks reasonable to demand that
some sort of curvature divergence must take place along the non-spacelike
curves that encounter a spacetime singularity. However, a general charac-
terization of the singularity in terms of the curvature divergence runs into
various difficulties. For example, for the plane wave vacuum solutions, the
polynomials in the curvature scalars vanish, but the curvature tensor is still
 4.2 Spacetime singularities 153

allowed to be singular (Penrose, 1965). Another example is the Taub–NUT

type of solutions given by Misner (1963, 1967). Here, the spacetime curva-
tures are bounded and the manifold is inextendible, but it is both null and
timelike geodesically incomplete.
Considering these as well as similar situations, the occurrence of non-
spacelike geodesic incompleteness has been generally agreed upon as the
criterion for the existence of a singularity for a spacetime. This criterion does
not cover all possible types of singular behavior. For example, Geroch (1968a)
has given an example of a spacetime that is geodesically complete, but which
contains a future inextendible timelike curve with a bounded acceleration
and with a finite proper length. This could correspond to a rocket ship with
enough fuel to disappear suddenly from the universe after a finite proper
time. Also, one would not like to term all the geodesically incomplete models
as containing a physically genuine singularity, especially if the curvatures are
finite everywhere throughout the spacetime. This includes the Taub–NUT
case mentioned above. In order to call a singularity physically genuine, one
would like to demand some sort of curvature divergence along the incomplete
non-spacelike geodesic. On the other hand, if there is a powerful curvature
divergence along an incomplete non-spacelike geodesic, one would certainly
like to call such a singularity physically significant.
It is clear, however, that if a spacetime manifold contains incomplete
non-spacelike geodesics, there is a definite singular behavior present in the
spacetime. In such a case, a timelike observer or a photon suddenly disap-
pears from the spacetime after a finite amount of proper time, or after a finite
value of the affine parameter. The singularity theorems that result from an
analysis of gravitational focusing and global properties of a spacetime prove
this incompleteness property for a wide class of spacetimes under a set of
rather general conditions.

4.2.2 Gravitational focusing

Gravitational focusing in a spacetime plays an important and key role in
the formation of trapped surfaces as the gravitational collapse develops. It
is discussed how the matter fields with positive energy density affect the
causality relations in a spacetime and how they cause focusing in the families
of timelike and null trajectories. The essential phenomenon that occurs here
is that matter focuses the non-spacelike geodesics of the spacetime into pairs
of focal points or the conjugate points. The basic property of conjugate points
is that if p < q are two conjugate points along a non-spacelike geodesic,
then p  q . Now, there are null hypersurfaces, such as the boundary of
the future I + ( p) for a point p, such that no two points of the hypersurface
could be joined by a timelike curve. Therefore, the null geodesic generators
154 Cosmic censorship

of such surfaces cannot contain any conjugate points, and these must leave
the hypersurface before encountering a conjugate point. This puts strong
constraints on the nature of such surfaces, and the singularity theorems result
from an analysis of these limits.
Consider a congruence of timelike geodesics in the spacetime. This is a
family of curves such that precisely one timelike geodesic trajectory passes
through each point p. Choosing the curves to be smooth, this defines a
smooth timelike vector field on the spacetime. On the other hand, a given
smooth vector field on the spacetime specifies a congruence of curves in the
manifold.
Let V i denote the timelike tangent vector to the congruence. Choosing the
parameter to be the proper time along such timelike trajectories, this can be
normalized to be a unit tangent vector,

V i Vi = −1. (4.17)

The spatial part, hij , of the metric tensor can be defined as

hij = gij + Vi Vj . (4.18)

Then, hi j = δ i j + V i Vj = g ik hkj , and

hij V i = hij V j = hi j Vi = hi j V j = 0. (4.19)

Therefore, hi j can be called the projection operator onto the subspace of Tp ,

orthogonal to the vector V i . The indices of h are now raised and lowered,
just as in the case of the metric tensor,

hij hj k = (gij + Vi Vj )(g j k + V j Vk ) = gik + Vi Vk = hik , (4.20)

and also
hij hij = hi i = δ i i + V i Vi = 3. (4.21)

For the given congruence of timelike geodesics, the expansion, shear, and
rotation tensors are respectively defined as

θij = V(k;l) hki hl j , (4.22)

σij = θij − 13 hij θ, (4.23)
ωij = hki hl j V[k;l] . (4.24)

Here, the volume expansion θ is defined as

θ = θij hij = V(k;l) hlk = ∇k V k = V k

;k . (4.25)
 4.2 Spacetime singularities 155

Furthermore, note that σij and ωij are purely spatial quantities in the sense
that
σij V i = ωij V i = 0. (4.26)
Also, note that
σ i i = hij σij = θ − 13 hij hij = 0. (4.27)
The covariant derivative of V is then expressed as

∇j Vi = Vi;j = 13 θhij + σij + ωij . (4.28)

This is verified by direct substitution from (4.22), (4.23), and (4.24).

Now, the geodesic equations imply that

V k ∇k ∇j Vi = V k ∇j ∇k Vi + Rilkj V l V k . (4.29)

Using the fact that V k is a tangent to the geodesics, that is, ∇j (V k ∇k Vi ) =

0, the above equation can be written as

V k ∇k ∇j Vi = −(∇j V k )(∇k Vi ) + Rilkj V l V k . (4.30)

Taking trace in the above,

dθ
= V k ∇k V i ;i = −(V k ;i V i ;k ) − Rlk V l V k , (4.31)
dτ
where τ is the affine parameter along the geodesic.
Using (4.28) in the above and the anti-symmetry properties of the tensor
ωij , after some simplification,

dθ
= −Rlk V l V k − 13 θ2 − σij σ ij + ωij ω ij (4.32)
dτ
can be obtained, which can be written as

dθ
= −Rlk V l V k − 13 θ2 − 2σ 2 + 2ω 2 . (4.33)
dτ
Equation (4.33) above is called the Raychaudhuri equation (Raychaudhuri,
1955), which describes the rate of change of the volume expansion as the
timelike geodesic curves in the congruence are moved along.
The second and third terms on the right-hand side involving θ and σ are
always positive. Consider now the term Rij V i V j , which by the Einstein
equations can be written as

Rij V i V j = 8π [Tij V i V j + 12 T ]. (4.34)

156 Cosmic censorship

The term Tij V i V j above represents the energy density measured by a time-
like observer with the unit tangent V i , which is the four-velocity of the
observer. For all reasonable classical physical fields this energy density is
generally taken as non-negative, and it is assumed that for all timelike
vectors V i ,
Tij V i V j ≥ 0, (4.35)
is satisfied. Such an assumption is called the weak energy condition. On the
other hand, it is also considered reasonable to believe that the matter stresses
will not be so large as to make the right-hand side of (4.34) negative. This
will be satisfied when
Tij V i V j ≥ − 12 T (4.36)
is satisfied. Such an assumption is called the strong energy condition, and it
implies that for all timelike vectors V i ,

Rij V i V j ≥ 0. (4.37)

By continuity, it can be argued that the same will then also hold for all null
vectors.
Both the strong and weak energy conditions will be valid for well-known
forms of matter, such as the perfect fluid, provided that the energy density ρ
is non-negative and that there are no large negative pressures that are bigger
or comparable to ρ, when converted into physical units.
An additional energy condition often required by the singularity theo-
rems is the dominant energy condition, which states that in addition to the
weak energy condition, the pressure of the medium must not exceed the
energy density. This can be equivalently stated as, for all timelike vectors
V i , Tij V i V j ≥ 0 and the vector T ij Vi is a non-spacelike vector. Such a
condition would be satisfied provided that the local speed of sound does not
exceed the local speed of light.
With the strong energy condition being satisfied, the Raychaudhuri equa-
tion implies that the effect of matter on the spacetime curvature causes a
focusing effect in the congruence of the timelike geodesics due to gravita-
tional attraction. This, in general, causes the neighboring geodesics in the
congruence to cross each other, which gives rise to caustics or conjugate
points. This separation between nearby timelike geodesics is governed by the
geodesic deviation equation,

D2 Z j = −Rkil
j
V k Z iV l, (4.38)

where Z i is the separation vector between nearby geodesics of the congruence.

Solutions of the above equation are called the Jacobi fields along a given
timelike geodesic.
 4.2 Spacetime singularities 157

Suppose now γ is a timelike geodesic. Then, two points p and q along γ are
called conjugate points if a Jacobi field along γ exists that is not identically
zero, but vanishes at p and q . From the derivation of the Raychaudhuri
equation given above, it is clear that the occurrence of conjugate points
along a timelike geodesic is closely related to the behavior of the expansion
parameter θ of the congruence. In fact, it can be shown that the necessary
and sufficient condition for a point q to be conjugate to p is that for the
congruence of timelike geodesics emerging from p, θ → −∞ at q (see for
example, Hawking and Ellis, 1973). The conjugate points along the null
geodesics are also similarly defined. Consider, for example, a congruence
of null geodesics emanating from a point p. If infinitesimally nearby null
geodesics of the congruence meet again at some other point q in the future,
then p and q are said to be conjugate to each other.
Similarly, let S be a smooth spacelike hypersurface, that is, it is an
embedded three-dimensional submanifold. Consider a congruence of time-
like geodesics orthogonal to S . Then, a point p along a timelike geodesic γ
of the congruence is said to be conjugate to S along γ if a Jacobi vector field
along γ exists that is non-zero at S , but vanishes at p. This means that there
are two infinitesimally nearby geodesics orthogonal to S that intersect at p.
Again, the equivalent condition for this to happen, in terms of the parameter
θ, is that the expansion θ for the congruence orthogonal to S tends to −∞
at p. If V i denotes the unit timelike tangent vector field of the congruence
of timelike geodesics, where V i denotes the normal to S , then the extrinsic
curvature χij of S is defined as

χij = ∇i Vj , (4.39)

which is evaluated at S . Clearly, χij V i = χij V j = 0. Also, the hypersurface

orthogonality of the congruence implies that ωij = 0. As a result, χij = χji ,
that is, this is a symmetric tensor. The trace of the extrinsic curvature,
denoted by χ is given by

χ = χi i = hij χij = θ. (4.40)

Therefore, χ = θ at S , where θ is the expansion of the congruence orthogonal

to S .
The behavior of the expansion parameter θ is governed by the Raychaud-
huri equation as pointed out above. For example, consider the situation
when the spacetime satisfies the strong energy condition and the congruence
of timelike geodesics is hypersurface orthogonal. In such a case, ωij = 0
and the corresponding term, ω 2 , vanishes in (4.33). Then, the expression for
the covariant derivative of ωij implies that it must also vanish for all future
158 Cosmic censorship

times. It follows from the above discussion that

dθ θ2
≤− , (4.41)
dτ 3

which means that the volume expansion parameter must necessarily be

decreasing along the timelike geodesics. If θ0 denotes the initial value of
the expansion, then the above can be integrated as θ−1 ≥ θ0−1 + τ /3. It
is clear from this that, if the congruence is initially converging and θ0 is
negative, then θ → −∞ within a proper time distance τ ≤ 3/ | θ0 |.
The following can then be seen from the above discussion. Let M be a
spacetime satisfying the strong energy condition, and let S be a spacelike
hypersurface with θ < 0 at p ∈ S . If γ is the timelike geodesic of the
congruence orthogonal to S passing through p, then a point q conjugate to
S along γ exists within a proper time distance τ ≤ 3/ | θ |, provided γ can
be extended to that value of the proper time.
Suppose now that the trace of the extrinsic curvature χij (which is also
sometimes called the second fundamental form of the surface S ) is negative
everywhere on S , that is, θ = χ < 0 on S , and it is bounded above by a
negative value θmax . In this case, it is clear from the above that all the timelike
geodesics of the congruence orthogonal to S will contain a point conjugate
to S within a proper time distance τ ≤ 3/ | θmax |. Therefore, let M be a
spacetime satisfying the above conditions, and let S be a spacelike surface in
M . Let the trace of the extrinsic curvature χ = θ < 0 on S and be bounded
above by a negative value θmax . Then, all the timelike geodesics orthogonal to
S have a point p conjugate to S within a proper time distance τ ≤ 3/ | θmax |,
provided the geodesics can be extended to that value of the proper time.
Consider now the congruence of timelike geodesics passing through a point
p. As shown by Lemma 4.5.2 of Hawking and Ellis (1973), for any convex
normal neighborhood of p, the trajectories of this congruence are orthogonal
to the spacelike surfaces of the proper time τ = const. along the geodesics.
Therefore, the congruence is hypersurface orthogonal, ωij = 0, and it will also
be zero for all future times. Then, the above discussion again implies that if
the strong energy condition holds for all timelike vectors V i , and if p ∈ γ with
θ = θ0 < 0 at some point q in the future of p along this timelike geodesic, then
γ contains a point r conjugate to p within a proper time distance τ ≤ 3/ | θ0 |
from q , provided it can be extended to that value of the proper time.
The basic implication of the above results is that, once a convergence
occurs in a congruence of the timelike geodesics, the conjugate points or the
caustics must develop in the spacetime. These can be interpreted as the singu-
larities of the congruence. Such singularities could occur even in Minkowski
spacetimes and similar other perfectly regular spacetimes. However, when
combined with certain causal structure properties of the spacetime, this
 4.2 Spacetime singularities 159

implies the existence of spacetime singularities in the form of geodesic

incompleteness. The gravitational focusing for congruences of null geodesics,
or for null geodesics orthogonal to a spacelike two-surface, can be discussed
similarly.
In general, note that even if Rij V i V j = 0 throughout the spacetime, if
σ > 0 then a net focusing effect again results. This will be so if Rijkl V j V l =
2

0 at least at one point in the spacetime. It is then possible from Hawking

and Ellis (1973) to show that if λ(t) is a non-spacelike geodesic, complete
both in the future and the past with a range of the affine parameter t over
(−∞, +∞), with Rij V i V j + 2σ 2 being continuous and non-negative, and
if the latter quantity is positive for at least one value of t, then λ(t) must
contain a pair of conjugate points in the interval (−∞, +∞).
The condition required above amounts to a statement that the non-
spacelike trajectory must pass through some matter or radiation at least
once throughout its history. This will happen if Rij V i V j = 0 at some point
on the trajectory, or the Weyl tensor is non-zero at this point in a suit-
able manner. A precise condition to ensure this is that every non-spacelike
geodesic in M must contain a point at which K[i Rj]el[m Kn] K e K l = 0, where
K is the tangent to the non-spacelike geodesic. This is called the generic
condition. Therefore, every timelike and null geodesic that is both future
and past complete must contain a pair of conjugate points if the spacetime
satisfies the generic condition.
Globally hyperbolic spacetimes have been discussed, and an important
property of these is that if N is a globally hyperbolic subset, and if p  q
for p, q ∈ N , then a timelike geodesic from p to q exists that maximizes the
lengths of all the non-spacelike curves from p to q . Such a maximal curve is
related to the existence of conjugate points in that, if N is a globally hyper-
bolic subset of and p, q ∈ N , with γ (t) being a timelike geodesic maximizing
the lengths of all non-spacelike curves from p to q , then γ (t) contains no
points conjugate to p between p and q . The point is that, if there were a con-
jugate point r to p between p and q , then it can be shown using variational
arguments, that a longer non-spacelike curve from p to q could be obtained
by ‘rounding off the corner’ at r, generated due to the conjugate point r.
This is contradictory to the maximality of the curve γ (t).

4.2.3 Existence of singularities

There are several singularity theorems available that establish the non-
spacelike geodesic incompleteness for a spacetime under different sets of
conditions and that are applicable to different physical situations. However,
the most general of these is the Hawking–Penrose theorem (Hawking and
Penrose, 1970), which is applicable in both the collapse situation and the
cosmological scenario. The main idea of the proof of such a theorem is now
160 Cosmic censorship

discussed. Using the causal structure analysis, it is shown that there must be
maximal length timelike curves between certain pairs of events in the space-
time. As pointed out above, a causal geodesic that is both future and past
complete must contain pairs of conjugate points if the spacetime satisfies
the generic condition and an energy condition. This is then used to draw
the necessary contradiction in order to show that M must be non-spacelike
geodesically incomplete. This result can be stated as in the following.
A spacetime (M, g ) cannot be timelike and null geodesically complete if
the following are satisfied:
(1) Rij K i K j ≥ 0 for all non-spacelike vectors K i ;
(2) the generic condition is satisfied, that is, every non-spacelike geodesic
contains a point at which K[i Rj]el[m Kn] K e K l = 0, where K is the
tangent to the non-spacelike geodesic;
(3) the chronology condition holds;
(4) in M , either a compact achronal set without edge or a closed trapped
surface exists, or a point p exists such that for all past directed null
geodesics from p, θ must eventually be negative.
The main idea of the proof is that it can be shown that the following three
cannot hold simultaneously: (a) every inextendible non-spacelike geodesic
contains pairs of conjugate points, (b) the chronology condition holds, (c) an
achronal set S exists in the spacetime such that E + (S ) or E − (S ) is compact,
where E + (S ) and E − (S ) denote the future and past of the edge of set S .
If this is shown, then the theorem is proved, as (3) is the same as (b),
(4) implies (c), and (1) and (2) imply (a). First, note that (a) and (b)
imply the strong causality of the spacetime (see Prop. 6.4.6 of Hawking and
Ellis, 1973). Next, it can be shown that if S is a future trapped set and if
strong causality holds on I + (S ) then a future endless trip γ exists such that
γ ⊂ Int D+ (E + (γ )). Now, T = J − (γ ) ∩ E + (S ) is defined, T turns out to be
past trapped, and hence λ, a past endless causal geodesic in Int(D− (E − (T )),
exists. Then, a sequence {ai } receding into the past on λ, and a sequence
{ci } on γ to the future is chosen. The sets J − (ci ) ∩ J + (ai ) are compact and
globally hyperbolic, so a maximal geodesic µi from ai to ci exists for each i.
The intersections of µi with the compact set T have a limit point p and a
limiting causal direction. The causal geodesic µ with this direction at p must
have a pair of conjugate points. This is then shown to be contradictory to
the maximality property of the geodesics stated above.
There is a more general way in which the singular points in a spacetime
can be defined using the terminal indecomposable pasts (TIPs) and terminal
indecomposable futures (TIFs), as discussed earlier (Penrose, 1974a, 1979).
The spacetime is assumed to be strongly causal. Here, a curve means a map
γ from an interval [0, a) of the real line into M , where a could possibly be
infinity. Therefore, the curve starts at an initial point γ (0) with a definite
 4.3 Blackholes 161

p q

I –(p) I –(g)
I –(q)

Fig. 4.3 The event p is a regular point in M and the set I − (p) defines a PIP. For
a non-spacelike curve γ going to infinity, I − (γ) gives an ∞-TIP. If λ is a finite
length curve going into a spacetime singularity, then I − (λ) is a singular TIP.

tangent but has no end point, as the interval is open at a. Such a curve will
be called extendible if it is possible to extend the map γ to an end point γ (a)
in M , and otherwise it is called inextendible. Of particular interest here are
the inextendible non-spacelike curves. The TIPs and TIFs are generated by
future directed and past directed timelike curves respectively, and they give
all the boundary points of spacetime that include both the singularities and
points at infinity. Such a boundary point is called ∞-TIP, which is a point
at infinity if it is generated by some timelike curve of infinite proper time
length in the future. A singular TIP is one that is not generated by any such
timelike curve of infinite length (see Fig. 4.3). Similarly, ∞-TIFs and singular
TIFs can be defined. The existence of a singular TIP defines a singularity of
spacetime, giving a class of future directed inextendible timelike curves that
have a finite proper time length, but no future end point.
As pointed out by Clarke (1986), the basic requirement for the ideal end
point of a timelike curve to be called a singularity, rather then a regular
boundary point, is that there should be no extension of the spacetime possible
in which the curve in question could be continued. If such an extension
existed, then the singularity would be similar in some sense to the coordinate
singularity in the Schwarzschild geometry at r = 2m. Therefore, the question
of singularity depends on what type of extension is allowed for the spacetime.
Therefore, a boundary point is called a C k -singularity of the spacetime if
there is no C k -extension of M that removes it. Clarke then defines the index
k as a measure of the strength of the singularity in the sense that the smaller
the k , the stronger the singularity.

4.3 Blackholes
Assuming cosmic censorship in the form of an asymptotic predictability in
the spacetime, much of the theoretical basic blackhole physics is carried out
in an asymptotically flat and predictable model.
162 Cosmic censorship

The notion of asymptotic flatness has played an important role in gravi-

tation theory. Consider a spherically symmetric star with a vacuum outside,
where the metric satisfies the empty space Einstein equations Rij = 0, given
by the Schwarzschild solution. Then, far away from the star, as r → ∞, the
components gij tend to the Minkowskian values. Such isolated systems and
the behavior of gravitational field and metric components far away at infinity
are of much interest in general relativity. Even though the actual universe
is not asymptotically flat as there would be matter present at all distances,
such an approximation is useful to model the geometry of an individual star
and to study its gravitational field.
It was pointed out by the studies of Bondi, van der Burg, and Metzner
(1962), and Sachs (1962) that the characteristic or null surfaces play an
important role in understanding the asymptotic properties of gravitational
fields for such isolated systems. They used characteristic surfaces to study the
metric components and curvature tensor properties in the asymptotic limit.
A coordinate free construction of null infinity and the notion of asymptotic
flatness for a general spacetime were introduced by Penrose (1965, 1968)
by means of conformal compactification of the spacetime. Here, the main
idea is to attach a boundary to the spacetime in such a manner that its
properties coincide with the geometric properties of the boundary I + or
I − for the Minkowski spacetime discussed earlier. A general spacetime is
then called asymptotically flat if it admits such a boundary attachment.
Just as in the Minkowski spacetime, a conformal transformation Ω on the
original spacetime M can be introduced so that Ω → 0 near infinity, and
the new unphysical spacetime (M , Ω2 gij ) is compactified. In (M , Ω2 gij ), the
boundary surface of M corresponds to the infinity of the spacetime M .
Specifically, a spacetime (M, gij ) is called asymptotically flat if a new,
unphysical spacetime (M , ḡ ) with a boundary I exists such that M − I
is diffeomorphic to M with Ω > 0 and ḡij = Ω2 gij , with the following
conditions:

(1) the new unphysical manifold M is smooth everywhere including the

boundary;
(2) the conformal factor Ω is smooth everywhere and Ω = 0 on I ;
(3) all the maximally extended null geodesics in M have a future and a past
end point on I ;
(4) there is a neighborhood of I in M where gij satisfies the vacuum Einstein
equations Rij = 0.

Such a construction of conformal compactification for a spacetime turns

out to be particularly useful in order to study isolated sources in otherwise
empty spacetimes. The surface I can be thought of as infinity in the sense
that the affine parameter along every null geodesic in M grows unboundedly
 4.3 Blackholes 163

large near I . The null geodesics of these two conformally related spacetimes
are completely identical as point sets, as mentioned earlier. However, the
affine parameters along the geodesics in M and M are related as dv̄ =
Ω2 dv . Therefore, the affine parameter along the null geodesics in M must
blow up near I + in the future, and similar behavior holds near the past
infinity.
The definition above is quite stringent in that it assumes that every null
geodesic has two end points: in the future and in the past at I . Although this
is satisfied in the Minkowski spacetime, this does not hold in spacetimes such
as the Schwarzschild and Reissner–Nordström cases, which contain event
horizons and a blackhole region. Here, the future directed null geodesics
that enter the blackhole must end in the singularity at r = 0 and cannot
have an end point at I + . One would like to include these spacetimes in
the general asymptotically flat class, and so a spacetime M is defined to
be weakly asymptotically simple and empty (or a WASE spacetime) if an
asymptotically flat spacetime M in the above sense exists such that there
is a neighborhood of I in M that is isometric to an open set in M . This
definition covers the Schwarzschild and the Reissner–Nordström cases and
also the Kerr solutions.
Some important general properties of blackholes are now discussed. For
a detailed treatment on blackhole physics and results such as the blackhole
uniqueness theorems and details, see Hawking and Ellis (1973) and Wald
(1984). As stated earlier, the fundamental motivation for the concept of
a blackhole comes from a spherically symmetric homogeneous dust collapse
that has two important features. First, for a star undergoing a complete grav-
itational collapse, a region of trapped surfaces forms below r = 2m, from
which no light rays escape to an observer at infinity. Therefore, a blackhole
forms in the spacetime. Second, the ultimate fate of the star undergoing
the collapse is an infinite curvature singularity at r = 0, which is com-
pletely hidden within the trapped surface region and the blackhole. Hence,
no emissions or light rays from the singularity could go out to any observer
at infinity, and the singularity is causally disconnected from the outside
spacetime.
The question now is whether these conclusions can be generalized for a non-
spherically symmetric collapse, and whether they are valid at least for small
perturbations from exact spherical symmetry. It is known from Hawking
and Ellis (1973), using the stability of a Cauchy development in general
relativity, that the formation of trapped surfaces, and hence of a blackhole,
is a stable property when departures from spherical symmetry are taken into
account. Considering a spherically symmetric collapse evolution from given
initial data on a partial Cauchy surface S , the formation of trapped surfaces
T in the form of all the spheres with r < 2m in the exterior Schwarzschild
geometry is found. The stability of the Cauchy development then implies
164 Cosmic censorship

that, for all initial data sufficiently near to the original data in the compact
region J + (S ) ∩ J − (T ), the trapped surfaces must still occur. Then, the
curvature singularity of a spherical collapse also turns out to be a stable
feature as implied by the singularity theorems, which show that the closed
trapped surfaces always imply the existence of a spacetime singularity under
reasonable general conditions.
There is no proof available, however, that the singularity will continue
to be hidden within the blackhole and remain causally disconnected from
outside observers, even when the collapse departs from the homogeneous
dust cases, or is not exactly spherical. If the singularity became visible to
external observers, the predictability in the spacetime will be undermined
because new information could come from the singularity where the densities
and curvatures could be arbitrarily large.
Hence, in order to generalize the notion of blackholes to gravitational
collapse situations other than exactly spherically symmetric homogeneous
dust cases, it becomes necessary to rule out such naked singularities by means
of an explicit cosmic censorship assumption. This could be stated as follows:
if S is a partial Cauchy surface, then there are no naked singularities to the
future of S that can be seen from the future null infinity I + . This is true for
the spherical homogeneous dust collapse, where the breakdown of physical
theory at the spacetime singularity does not disturb the prediction in the
future for the outside asymptotically flat region.
This assumption is made precise by considering the spacetimes (M, g )
that admit a weakly asymptotically simple and empty conformal completion
(M , ḡ ). Then, (M, g ) is said to be future asymptotically predictable from a
partial Cauchy surface S if

I + ⊂ D̄+ (S, M ), (4.42)

that is, the future null infinity I + is contained in the closure of D+ (S ) in

the conformal manifold.
Future asymptotic predictability ensures the cosmic censorship condition
in the form that there are no singularities in the future of S that are ‘naked’,
that is, visible from the future null infinity I + . In the spherical homogeneous
dust collapse, the resulting spacetime is future asymptotically predictable
and the censorship holds. Whether this is respected in any other general
situations is not known, either as a proof for the future asymptotic pre-
dictability for general spacetimes, or of any other suitable version of the
cosmic censorship hypothesis.
A blackhole region in the spacetime, which can be denoted by B , for a
future asymptotically predictable spacetime M is defined as

B = M − J − (I + ). (4.43)
 4.3 Blackholes 165

I+ I+

p
r
q

Fig. 4.4 No non-spacelike curves from the blackhole region B can reach the
future infinity and the events faraway in the spacetime, so it is cut off from the
outside universe. The events p and q are in a blackhole, whereas the event r that
is outside the blackhole can send signals to infinity.

Therefore, this is a spacetime region from which no null or timelike curves

can reach an observer at infinity (see Fig. 4.4). The boundary of B in M ,
given by
H = J˙− (I + ) ∩ M, (4.44)

is called the event horizon. As pointed out earlier, this horizon must be an
achronal surface generated by null geodesics that could have past end points
in M , but that have no future end points. For the Minkowski spacetime,
J − (I + ) = M and there is no blackhole. However, for the Schwarzschild case,
J − (I + ) is the region for the spacetime exterior to r = 2m, and the event
horizon is given by the null hypersurface r = 2m, which is the boundary of
the blackhole region 0 < r < 2m.
By definition, no spacetime singularities are visible at null infinity in an
asymptotically predictable spacetime. In fact, this is true also for trapped
surfaces in this case, provided either the weak or strong energy condition
is satisfied, which implies that all trapped surfaces must be fully contained
within the blackhole region and not visible from I + (Wald, 1984). Specifi-
cally, let (M, gij ) be future asymptotically predictable from a partial Cauchy
surface S , and Rij K i K j ≥ 0 for all null vectors K i . Then, if T is a closed
trapped surface in D+ (S ), T ∩ J − (I + ) = ∅.
Let (M, gij ) be an asymptotically flat spacetime with the associated
unphysical conformal spacetime (M , ḡij ). Suppose p ∈ I + and q ∈ M ∩
J − ( p). Let γ be the future directed null geodesics generator of I + through
p, and let r ∈ γ be any point. Then, q ∈ I − (r) in M . Therefore,
J − (I + ) = I − (I + ) in M , and hence J − (I + ) is open in M . Therefore, the
blackhole region B = M − J − (I + ) is closed in M . This implies that the
event horizon is contained in B . As such, the blackhole region B need not
166 Cosmic censorship

be connected in M , and while considering isolated blackholes forming

out of a gravitational collapse in M , a connected component of B is
worked with.
Any event p on the event horizon H lies on the boundary of the blackhole
region, and so any small perturbation could make p enter J − (I + ), causally
connecting to infinity. Then, the spacetime is no longer asymptotically pre-
dictable. This situation is avoided by further demanding that, for the partial
Cauchy surface S ,
J + (S ) ∩ J − (I + ) ⊂ D+ (S ). (4.45)
This effectively means that a neighborhood of the event horizon could also
be predicted from S , and is equivalent to the condition that the spacetime
exterior to the blackhole region is globally hyperbolic.
In the case of collapsing dust, the event horizon is a null hypersurface
generated by those null geodesics that just reach the surface of the star
when it crosses the radius r = 2m, with m being the Schwarzschild mass of
the star. The area of the horizon increases monotonically until the horizon
reaches the surface of the star. Outside, this area is a constant given by
A = 16πm2 . For the Kerr spacetime, the horizon is defined by r = r+ , the
area being obtained by setting t = const., r = r+ , which gives the metric on
the surface. Then,

√  
A= g dθ dφ = 8πm m + (m2 − a2 )1/2 , (4.46)

where m and a denote the mass and angular momentum parameters

respectively. Therefore, the area of the horizon is a non-decreasing
function.
In fact, for strongly asymptotically predictable spacetimes in general, the
area of a blackhole horizon must either remain constant or must increase,
provided Rij K i K j ≥ 0 for all null vectors K i (Hawking, 1971). The basic
argument leading to the proof of this result describes the evolution of the
event horizon. First, note that the horizon H is generated by future inex-
tendible null geodesics generators, as H is the boundary of the past of I + ,
and so in order to have a future endpoint, a generator must intersect I + ,
which is not possible. Next, for all the null geodesic generators of H, the
expansion θ must be non-negative everywhere, θ ≥ 0. This is because, if at
some p ∈ H, if θ < 0, H can be deformed in the neighborhood of p so that
θ < 0 still in this neighborhood, and it enters the past of J − (I + ). Now,
choose a spacelike two-surface T in J˙− (I + ) in this neighborhood, then T
intersects J − (I + ). Then, generators of J˙+ (T ), visible from I + have past
endpoints at T and are orthogonal to T . However, θ < 0 implies that they
must have a point conjugate to T within a finite affine distance and so cannot
remain in the null boundary all the way to infinity, which is a contradiction.
 4.3 Blackholes 167

Therefore, the area of a two-dimensional cross section of generators cannot

decrease as H evolves to the future. From this, it is then possible to deduce
the result that the area of the event horizon must be non-decreasing in the
future.
In particular, if two or more blackholes merge to form a single blackhole,
the area of its boundary must be greater than or equal to the sum of the
original blackhole areas. An interesting implication of this result was pointed
out by Hawking (1971), to get an upper limit on the energy that can be
emitted in gravitational radiation when two blackholes coalesce, at most half
the initial energy could be released in blackhole collisions.
According to the area theorem above, for all physically allowed processes,
the total area of the blackholes cannot decrease, that is, δA ≥ 0. This is very
similar to the second law of thermodynamics, which says that for all physical
processes, the total entropy of all the matter in the universe is non-decreasing,
that is, δS ≥ 0. Consider, for example, the Schwarzschild blackhole. The area
for its event horizon is given by A = 16πm2 . The only way to reduce this
area is to extract mass from the blackhole, which is impossible because no
particles or photons can cross the event horizon to come out. On the other
hand, the area could be increased by throwing particles in, which would
increase the mass (see for example, Bekenstein, 1973, for details on the
relationship between entropy and area of the event horizon).
Similar analogies have been developed between other laws of thermody-
namics and the laws of blackhole physics. In fact, it has been shown by
Bardeen, Carter, and Hawking (1973) that for any stationary axisymmet-
ric blackhole in an asymptotically flat spacetime, it is possible to define
a quantity called surface gravity that is constant on the horizon. There-
fore, just as the temperature is constant throughout a body in thermal
equilibrium (zeroth law of thermodynamics), so is the surface gravity con-
stant over the horizon of a stationary blackhole. Again, for a Schwarzschild
blackhole, the surface gravity turns out to be 1/4m, and it is impossible to
reduce it to zero by any physical process, just as it is impossible to achieve
the zero temperature for a system by any physical process (third law of
thermodynamics).
As such, the thermodynamic temperature of a blackhole in classical rela-
tivity will be absolute zero, since it is a perfect absorber that does not emit at
all. So, it would appear that the surface gravity may not represent physical
temperature. However, it was shown by Hawking (1975) that when quantum
particle creation effects are taken into account, a blackhole actually radiates
with a black body spectrum at a temperature proportional to the surface
gravity. In this sense, the surface gravity represents the thermodynamic
temperature of a blackhole.
These results on blackholes assume the spacetime to be asymptotically flat.
In reality, however, the universe is not so, in view of the observed distribution
168 Cosmic censorship

of matter at the largest possible scales. Hence, it would be desirable to

have the laws of blackholes given in a more general spacetime framework.
Blackholes in a general globally hyperbolic spacetime were defined by Tipler,
Clarke, and Ellis (1980), and Joshi and Narlikar (1982), who examined the
laws of blackhole physics in such a framework. It is seen that for a closed
globally hyperbolic spacetime with a compact Cauchy surface and a strong
curvature crushing singularity in the future, the sum of the areas of the
blackholes must decrease in the future. It is thus seen that in the situation of
the spacetime not being asymptotically flat, the behavior of blackholes and
the laws governing them may change.
Apart from general relativity, there is a motivation to consider blackholes
in Newtonian theory also, in the sense that the gravitational field of a star
could be so strong that it might stop even light (considered as particles) from
escaping. For example, consider a spherically symmetric mass distribution
of uniform density with radius R and total mass M . For a particle with a
velocity v away from the center and at a distance r, its total energy, which is
the sum of its kinetic and potential energies, is conserved. Suppose now the
velocity v0 of the particle is such that it is able to escape to infinity where it
has a vanishing velocity. Since the total energy will be zero at r = ∞ with
v = 0, which is conserved,

2GM
v02 = . (4.47)
R

When the radial velocity of the particle is less than v0 , it must fall back to
the body, otherwise it escapes to infinity. Therefore, if the mass distribution
and radius of the body were such that c2 = 2GM/R, where c is the velocity
of light, then for any larger mass or smaller radius of the body, even light
will not escape. This was realized by Laplace in 1798, who pointed out that,
for a star with a density the same as the Sun but radius 250 times larger, no
light could escape from its surface.
It is clear that a blackhole could not be observed directly, but gravitational
effects exhibited by such an object must be looked for. Although there is no
conclusive evidence available for the existence of blackholes at the moment,
presently the best candidates seem to be the binary stars in which one of the
partners is visible and the other is supposed to be a blackhole. Such a black-
hole would suck matter from its visible component, in the process forming
an accretion disk around the blackhole. Before the in-falling matter spirals
down the blackhole, the inner, hot regions are believed to produce intense
bursts of X-rays formed by synchrotron radiation. Therefore, the discovery of
the X-ray source Cygnus XI in 1971, which shows rapid variations, indicates
possible evidence for blackholes. Further to this, several other X-ray binaries
have been proposed as possible candidates for blackholes.
 4.4 Higher spacetime dimensions 169

4.4 Higher spacetime dimensions

Although there is no satisfactory proof or mathematical formulation of cen-
sorship available despite many efforts, as the discussion above shows, there
are many classes of dynamical collapse models investigated so far that lead to
a blackhole or a naked singularity as the collapse endstate (see for example,
Clarke and Joshi, 1988; Newman and Joshi, 1988; Lake, 1992; Rendall, 1992;
Szekeres and Iyer, 1993; Rein, Rendall, and Schaeffer, 1995; Wald, 1997; Pen-
rose, 1998; Gundlach, 1999; Królak, 1999; Jhingan and Magli, 2000; Joshi,
2000; Celerier and Szekeres, 2002; Harada, Iguchi, and Nakao, 2002; Giambo
et al., 2003; Debnath, Chakraborty, and Barrow, 2004; Maeda, 2006).
One of the possibilities to recover cosmic censorship could be to con-
sider the possibility that one may actually be living in a higher-dimensional
universe. The recent developments in string theory and other field theo-
ries indicate that gravity is possibly a higher-dimensional interaction, which
reduces to the general relativistic description at lower energies. Hence, it
may be that censorship is restored in higher spacetime dimensions due to
extra physical effects arising from the transition itself to a higher-dimensional
spacetime continuum. The recent revival of interest in such a possibility is
motivated partly by the Randall–Sundrum brane-world models (Randall and
Sundrum, 1999).
It is shown below that certain naked singularities arising in a dust collapse
from smooth initial data, which include those obtained by Eardley and Smarr
(1979), Christodoulou (1984), and Newman (1986), are removed when the
transition to higher-dimensional spacetimes is made. The cosmic censorship
is then restored for dust collapses of these classes, which will always produce
a blackhole as the collapse endstate for dimensions N ≥ 6, under various con-
ditions to be motivated physically, such as the smoothness of the initial data
from which the collapse develops. This is of interest because the gravitational
collapse of a dust cloud has served, over the last several decades, as a basic
and fundamental paradigm in blackhole physics. It is seen explicitly that the
above mentioned naked singularities of the dust collapse, for both marginally
bound and non-marginally bound dust, are removed on going to higher (≥ 6)
dimensions, if only smooth and analytic initial profiles are allowed for.
It is pointed out here that the increase in dimensions deforms the apparent
horizon, which is the outer boundary of the trapped surfaces, in such a
manner that the formation of the trapped surfaces is advanced. Hence, a
critical dimension exists, beyond which the neighborhood of the center gets
trapped before the central singularity and the final outcome of collapse is
necessarily a blackhole (Goswami and Joshi, 2004b).
Consider a general spherically symmetric comoving metric in N ≥ 4
dimensions that has the form

ds2 = −eν(t,r) dt2 + e2ψ(t,r) dr2 + R2 (t, r)dΩ2N −2 , (4.48)

170 Cosmic censorship

where  i−1 

N −2 
2
dΩ2N −2 = sin (θ ) (dθi )2
j
(4.49)
i=1 j=1

is the metric on the (N − 2) sphere. The energy–momentum tensor of the

dust has the form
i
T t t = ρ(t, r), T r r = 0, Tθ θi = 0. (4.50)

Take the matter field to satisfy the weak energy condition, that is, the
energy density measured by any local observer is non-negative, and so for
any timelike vector V i ,

Tik V i V k ≥ 0, i.e. ρ ≥ 0. (4.51)

In the case of a finite collapsing cloud, there is a finite boundary 0 < r < rb ,
outside which the cloud is matched to a Schwarzschild exterior. The range of
the coordinates for the metric is then 0 < r < rb , and −∞ < t < ts (r), where
ts (r) corresponds to the singular epoch R = 0. Solving the N -dimensional
Einstein equations, the generalized Tolman–Bondi–Lemaitre (TBL) metric
can be obtained as


R2
ds = −dt +
2 2
dr2 + R2 (t, r)dΩ2N −2 , (4.52)
1 + f (r)

where f (r) is an arbitrary function of the comoving radius r, and f (r) > −1.
The equations of motion are then given by

(N − 2)F  F (r)
= ρ, Ṙ2 = + f (r). (4.53)
2R(N −2) R R(N −3)
Here, F (r) is an arbitrary function of the comoving coordinate r and has
the interpretation of mass function of the dust cloud, as discussed earlier,
and f (r) is the energy function, which specifies the velocity profiles for the
collapsing shells for any given mass function. The energy condition then
implies F  ≥ 0. Consider the shell-focusing naked singularity at R = 0. The
continual collapse condition is given as Ṙ < 0.
Using the scaling independence of the comoving coordinate r,

R(t, r) = rv (t, r), (4.54)

where
v (ti , r) = 1, v (ts (r), r) = 0, v̇ < 0. (4.55)
This means the coordinate r has been scaled in such a way that at the initial
epoch, R = r, and at the singularity R = 0. As noted previously, R = 0 both
 4.4 Higher spacetime dimensions 171

at the regular center r = 0 of the cloud, and at the spacetime singularity,

where all matter shells collapse to a zero physical radius. The regular center
is then distinguished from the singularity by a suitable behavior of the mass
function F (r) so that the density remains finite and regular there at all times
until the singular epoch. The introduction of the parameter v as above then
allows the spacetime singularity to be distinguished from the regular center,
with v = 1 at the initial epoch, including the center r = 0, and which then
decreases monotonically with time as the collapse progresses to the value
v = 0 at the singularity R = 0.
In order to ensure the regularity of the initial data, it is seen from the
equations of motion that at the initial epoch the two free functions F (r) and
f (r) must have the forms

F (r) = r(N −1) M(r), f (r) = r2 b(r), (4.56)

where F (r) and f (r) are the mass function and the energy function
respectively.
Following Goswami and Joshi (2006), now assume that the initial density
and energy functions ρ(r) and f (r) are smooth and even, ensuring their
analytic nature. Note that the Einstein equations do not impose any such
restriction, which has to be physically motivated, and this implies a certain
mathematical simplicity in the arguments to deal with a dynamical collapse
situation. It follows that both M(r) and b(r) are now smooth C ∞ -functions,
which means the Taylor expansions of these functions around the center must
be of the form

M(r) = M0 + r2 M2 + r4 M4 + · · · , (4.57)

and
b(r) = b0 + r2 b2 + r4 b4 + · · · , (4.58)

that is, all odd terms in r vanish in these expansions, and the presence of
even terms only would ensure smoothness.
To predict the final state of the collapse for a given initial mass and velocity
distribution, the singularity curve resulting from the collapse of successive
matter shells, and the apparent horizon developing in the spacetime are
studied below (see also Goswami and Joshi, 2004a). A decreasing apparent
horizon in the (t, r) plane then is a sufficient condition for a blackhole, as it
shows the entrapment of the neighborhood of the center before the central
singularity.
Consider the continual collapse of the dust cloud to a final shell-focusing
singularity at R = 0, where all matter shells collapse to a zero physical
radius. With the regular initial conditions as above, the Einstein equation
172 Cosmic censorship

for Ṙ can be written as


v (N −3)/2 v̇ = − M(r) + v (N −3) b(r). (4.59)

Here, the negative sign implies that v̇ < 0, that is, the matter cloud is
collapsing. Integrating the above equation with respect to v gives
 1
v (N −3)/2 dv
t(v, r) =  . (4.60)
v M(r) + v (N −3) b(r)

Note that the coordinate r is to be treated as a constant in the above

equation. Expanding t(v, r) around the center,

X2 (v ) X3 (v )
t(v, r) = t(v, 0) + rX (v ) + r2 + r3 + ··· . (4.61)
2 6
Now, from the equation for the form of the mass function above and (4.58),
X (v ) = 0. The function X2 (v ) is then given as
 1
v (N −3)/2 (M2 + v (N −3) b2 )dv
X2 (v ) = − 3/2
, (4.62)
v (M0 + v (N −3) b0 )

where

b0 = b(0), M0 = M(0), b2 = b (0), M2 = M (0). (4.63)

Note that the value of X2 is determined fully by the initial values of the mass
function F (r) and the energy function b(r). Therefore, the time taken for
the central shell to reach the singularity is given by
 1
v (N −3)/2 dv
ts0 =  . (4.64)
0 M0 + v (N −3) b0
From the above equation it is clear that for ts0 to be defined,

M0 + v (N −3) b0 > 0, (4.65)

that is, the continual collapse condition implies the positivity of the above
term. Hence, the time taken for other shells close to the center to reach the
singularity can now be given by

X2 (0)
ts (r) = ts0 + r2 + ··· . (4.66)
2
In order to determine the visibility, or otherwise, of the singularity at R =
0, the causal structure of the trapped surfaces, and the nature and behavior
 4.4 Higher spacetime dimensions 173

of the null geodesics in the vicinity need to be analyzed. If future directed

null geodesics with a past end point at the singularity that go out to faraway
observers in spacetime exist, then the singularity is naked. Otherwise, a
blackhole results as the endstate of a continual collapse. The boundary of
the trapped region of the spacetime is given by the apparent horizon within
the collapsing cloud, which is given by the equation

F
= 1. (4.67)
RN −3
Broadly, it can be stated that if the neighborhood of the center gets
trapped earlier than the singularity, then it is covered, otherwise it is naked
with families of non-spacelike future directed trajectories escaping away from
it. For example, it follows from the above equation that along the singularity
curve t = ts (r) (which corresponds to R = 0), for any r > 0, F (r) goes
to a constant positive value, whereas R → 0. Hence, it follows that trap-
ping already occurs before the singularity develops at any r > 0 along the
singularity curve ts (r).
Now, there is a need to determine when there will be families of non-
spacelike paths coming out of the central singularity at r = 0, t = ts (0),
reaching outside observers, and when there will be none. The visibility,
or otherwise, of the singularity is decided accordingly. By determining the
nature of the singularity curve and its relation to the initial data, it is pos-
sible to deduce whether the trapped surface formation in the collapse takes
place before or after the central singularity. It is this causal structure that
determines the possible emergence, or otherwise, of non-spacelike paths from
the singularity, and settles the final outcome in terms of either a blackhole
or a naked singularity.
From (4.67),
vah (r) = [r2 M(r)]1/(N −3). (4.68)
Using the above equation in (4.60),
 vah (r)
v (N −3)/2 dv
tah (r) = ts (r) −  . (4.69)
0 M(r) + v (N −3) b(r)

As the behavior of the apparent horizon is under consideration close to the

central singularity at r = 0, R = 0 (all other points r > 0 on the singularity
curve are already necessarily covered), the upper limit of integration in the
above equation is small, and hence it is possible to expand the integrand in
a power series in v , and keep only the leading order term, which amounts to

X2 (0) 2 1/(N −3)

tah (r) = ts0 + r2 + · · · − r(N −1)/(N −3) M0 . (4.70)
2 N −1
174 Cosmic censorship

Since the apparent horizon is a well-behaved surface for a spherical dust

collapse, it can be said that the singularity curve for the collapse and its
derivatives around the center are also well-defined, as the same coefficients
are present in both (4.70) and (4.61). It is now possible to analyze the effect
of the number of dimensions on the nature and shape of the apparent hori-
zon. As the number of dimensions is increased, and reaches dimensions higher
than five, the negative term in (4.70) starts to dominate, thus advancing the
trapped surface formation in time. This equation shows the behavior of the
apparent horizon curve for the same initial data in different dimensions. As
seen here, in the usual four-dimensional spacetime, the initial profile ensures
an increasing apparent horizon and it can be explicitly shown that the end-
state is a naked singularity. But, in six and more spacetime dimensions, the
negative term starts to dominate, purely as a result of the increase in dimen-
sions. This causes the trapped surfaces to form sufficiently early to cover the
singularity. It follows that the number of spacetime dimensions causes an
intriguing effect on the causal structure and nature of the apparent horizon,
as can be clearly seen. Therefore, for smooth initial data and for dimen-
sions higher than five, the apparent horizon becomes a decreasing function
of r near the center. This implies that the neighborhood of the center gets
trapped before the central singularity and the central singularity is then
always covered, as opposed to the four-dimensional models of Eardley and
Smarr (1979), Christodoulou (1984) and Newman (1986), in which case a
positive X2 (0) makes the apparent horizon curve increase and hence ensures
a naked singularity (see Fig. 4.5).

Event horizons

Singularities

Apparent horizons

(a) (b)

Fig. 4.5 Typical apparent horizon behavior in a blackhole (a) and a naked sin-
gularity (b) collapse. Whereas in the blackhole case the apparent horizon is
decreasing away from the center, in the naked singularity situation it typically
increases in time away from the center.
 4.5 Formulating the censorship 175

Specifically, suppose there is a future directed outgoing null geodesic energ-

ing from the central singularity at R = 0, r = 0. If (t1 , r1 ) is an event along
the null geodesic, then t1 > ts0 and r1 > 0. But for any such r1 , the trapped
region already starts before t = ts0 , hence the event (t1 , r1 ) is already in
the trapped region and the geodesic cannot be outgoing. Therefore, there
are no outgoing paths from the central singularity, making it covered. It
now follows in general, that is for both marginally bound and non-marginal
cases, that for a dust collapse with smooth initial profiles, the final outcome
is always a blackhole for any spacetime dimensions N ≥ 6, and all the naked
singularities occurring in lower dimensions are removed.
In five dimensions, there is an interesting scenario arising from Banerjee,
Debnath, and Chakraborty (2003). As can be seen from (4.70), there is
a critical value of X2 (0), below which the apparent horizon is decreasing,
resulting in a blackhole endstate. Otherwise, a naked singularity can result.
It is interesting to note also that the above results hold if the initial profiles,
instead of being absolutely smooth C ∞ -functions, are taken to be sufficiently
smooth, namely at least C 2 -functions.
Note that it may still be possible to have both blackhole and naked singu-
larity outcomes as the collapse endstates, when the initial data and the metric
are not assumed necessarily to be C ∞ , or at least C 2 , analytic functions.
Hence, such smoothness conditions need to be investigated and carefully
probed further to see if they could be strongly motivated from a physical
perspective.

4.5 Formulating the censorship

It is clear from the discussions so far that the assumption of cosmic censorship
is crucial and necessary to basic results in blackhole physics. In fact, when the
gravitational collapse is considered in a generic situation, the very existence
of blackholes requires this hypothesis (Penrose, 1969).
If the cosmic censorship is to be established by means of a rigorous proof,
this of course requires a much more precise formulation of the hypothesis. The
statement that the result of a complete gravitational collapse must always
be a blackhole and not a naked singularity, or that all singularities of the
collapse must be hidden in blackholes, causally disconnected from observers
at infinity, is not rigorous enough. This is because, under completely general
circumstances, the censorship or asymptotic predictability is false as it is
possible to choose a spacetime with a naked singularity that would be a
solution to Einstein’s equations if

1
Tij ≡ Gij . (4.71)
8π
176 Cosmic censorship

Therefore, at the minimum, certain conditions on the stress–energy tensor

are required, for example, an energy condition. However, it turns out that to
obtain an exact characterization of the restrictions that should be required
on matter fields in order to prove a suitable version of the cosmic censorship
hypothesis is an extremely difficult task and no such specific conditions are
available presently. In other words, no plausible restrictions have been able
to ensure any provable version of the censorship conjecture.
The requirements in blackhole physics and general predictability argu-
ments have led to several different formulations of the cosmic censorship
hypothesis. The version known as the weak cosmic censorship refers to the
asymptotically flat spacetimes, the ones that are Minkowskian faraway from
the source, and has reference to the null infinity, which is reached by null
geodesics an infinite affine distance away. Weak censorship, or asymptotic
predictability, effectively postulates that the singularities of gravitational
collapse cannot influence events near the future null infinity I + . If S is the
partial Cauchy surface on which the regular initial data for the collapse is
defined, this is the requirement that I + is contained in the closure of D+ (S ).
Therefore, the data on S predict the entire future for faraway observers.
The other version, called the strong cosmic censorship, is a general
predictability requirement on any spacetime, stating that all physically rea-
sonable spacetimes must be globally hyperbolic (see for example, Geroch and
Horowitz, 1979; Hawking and Israel, 1979a, 1979b; Penrose, 1979).
The weak cosmic censorship or the strong asymptotic predictability
requirement means that the region of spacetime outside a blackhole must
be globally hyperbolic (Tipler, Clarke, and Ellis, 1980). A precise formula-
tion of this version of censorship will consist of specifying exact conditions
under which the spacetime would be strongly asymptotically predictable. In
its weak form, the censorship conjecture does not allow causal influences
from the singularity to asymptotic regions in the spacetime to an observer at
infinity, that is, the singularity cannot be globally naked. However, it could
be locally naked in that an observer within the event horizon and in the
interior of the blackhole could possibly receive particles or photons from the
singularity.
Clearly, such a requirement must be formulated more precisely. For exam-
ple, the metric on space should approach that of the Euclidean three-space at
infinity, and matter fields should satisfy suitable fall off conditions at spatial
infinity. Also, the null generators of I + may need to be complete, and the
exact meaning of ‘physically reasonable’ matter fields has to be specified. In
fact, as far as the cosmic censorship hypothesis is concerned, it is a major
problem in itself to find a satisfactory and mathematically rigorous formula-
tion of what is physically desired to be achieved (see for example, Penrose,
1998). Developing a suitable formulation would probably be a major advance
towards the solution of the main problem.
 4.5 Formulating the censorship 177

Since the interest here is mainly in the gravitational collapse scenario, it is

required that the spacetime contains a regular initial spacelike hypersurface
on which the matter fields, as represented by the stress–energy tensor Tij ,
have a compact support and all physical quantities are well-behaved on this
surface. Also, it is generally required that the matter satisfies a suitable
energy condition, and that the Einstein equations are satisfied. Then, it can
be said that the spacetime contains a naked singularity if there is a future
directed non-spacelike curve which reaches a faraway observer or infinity in
the future, and in the past it terminates at the singularity.
The main difficulty in proving the weak censorship appears to be that the
event horizon is a feature depending on the whole future behavior of the solu-
tion over an infinite time period, whereas the present theory of quasi-linear
hyperbolic equations guarantee the existence and regularity of the solutions
over a finite time internal only (Israel, 1984). In this connection, the results of
Christodoulou (1986) on generic spherically symmetric collapses of massless
scalar fields are relevant, where it is shown using global existence theorems
on partial differential equations that global singularity free solutions exist for
weak enough initial data. In any case, even if it is true, the proof for a suit-
able version of the weak censorship conjecture would seem to require much
more knowledge of the general global properties of Einstein’s equations and
solutions than is known presently. As such, various attempts to formulate
and prove the weak censorship have not succeeded so far.
It would appear that sufficient data are not yet available on the various
possibilities present for gravitationally collapsing configurations that would
enable us to decide one way or the other on the issue of the censorship
hypothesis. In this situation, a detailed investigation of the collapse scenar-
ios and models within general relativity really appears necessary, and the
possibilities arising in order to have insights into the issue of the final fate of
the gravitational collapse need to be examined.
For example, shell-crossing naked singularities have been shown to occur
in the spherical collapse of perfect fluids (Yodzis, Seifert, and Muler zum
Hagen, 1973, 1974), where shells of matter implode in such a way that fast
moving outer shells overtake the inner shells, producing a globally naked
singularity outside the horizon. These are the singularities where shells of
matter pile up to give two-dimensional caustics and the density and some
curvature components blow up. The general point of view, however, is that
such singularities need not be treated as serious counter-examples to the
censorship hypothesis, as these are merely consequent to the intersection of
matter flow lines. This gives a distributional singularity that is gravitationally
weak in the sense that the curvatures and tidal forces remain finite near it
(Tipler, Clarke, and Ellis, 1980).
On the other hand, there are shell-focusing naked singularities, as dis-
cussed earlier in detail, occurring at the center of the spherically symmetric
178 Cosmic censorship

collapsing configurations of dust, perfect fluids or radiation shells. These have

to be taken more seriously, and they can be ruled out only by saying that
the dust or perfect fluids are not really ‘fundamental’ forms of matter. How-
ever, if the cosmic censorship is to be established as a rigorous theorem, this
objection has to be made much more precise in terms of a clear restriction on
the stress–energy tensor, because these are forms of matter that otherwise
satisfy reasonability conditions. These include the dominant energy condi-
tion (provided there are no large negative pressures) or a well-posed initial
value formulation for the coupled Einstein matter field equations. Also, these
forms of matter are widely used in discussing various astrophysical processes.
For a review of various related developments and a detailed discussion, see
Joshi (1993) and Krasinski (1997).
Discussed briefly now is the hypothesis of strong cosmic censorship (Pen-
rose, 1979). Unlike the weak conjecture, the strong version demands that the
singularities should not be visible, even to the observers within the black-
hole. That is, they cannot be even locally naked, but are always spacelike and
the spacetime must be globally hyperbolic. Therefore, unless they actually
encounter them, the observer never sees the singular regions. The argument
given in favor of such a strong principle is that if cosmic censorship is really
a basic principle of nature, there should not be any special role given to the
observer at infinity because physical laws operate at a local level. Again, this
principle is to be carefully formulated because, by suitable cuts and identi-
fications in the Minkowski spacetime, it is easy to generate an inextendible
non-globally hyperbolic spacetime. In other words, any non-trivial topology
for the spacetime would give rise to violations of global hyperbolicity, and
therefore of the strong cosmic censorship. For any certain proposed formu-
lations of strong censorship and difficulties encountered, see Penrose (1979).
Again, no general proof is available for strong censorship, but it is sometimes
argued that the Cauchy horizons, forming as a result of non-global hyperbol-
icity of the spacetime, turn out to be unstable in certain cases. For example,
in the Reissner–Nordström case, the Cauchy horizon exhibits a so-called ‘infi-
nite blue-shift’ instability (Chandrasekhar and Hartle, 1983). However, again
the evidence is not uniform here. For example, Morris, Thorne, and Yurtsever
(1988) have shown that for the wormhole spacetime that they constructed,
the Cauchy horizon is immune to the Taub–NUT type of instability, and they
conjecture that it is fully stable, thus providing a counter-example to strong
censorship. An additional condition that is often required in both the weak
and strong formulations of the censorship is that stable spacetimes do not
admit naked singularities. This requires a suitable criteria for the stability of
the spacetimes, which is again a major difficult problem in general relativity.
Various different formulations of the censorship principle have been tried
out, based on different motivations. Therefore, a class of censorship conjec-
tures suggested all naked singularities must be in some sense gravitationally
 4.5 Formulating the censorship 179

weak (Tipler, Clarke, and Ellis, 1980; Israel, 1986a, 1986b; Newman, 1986;
Newman and Joshi, 1988; Nolan, 1999). Some of these suggestions are dis-
cussed here. In order to avoid the difficulties associated with the question as
to which forms of matter and equations of state should be considered rea-
sonable, it is also suggested that one examines first a purely vacuum version
of the censorship. That will show whether pure gravity allows naked singu-
larities. In fact, Geroch and Horowitz (1979) have detailed several possible
approaches to the censorship formulations, and pointed out difficulties in
each case.
To summarize the situation, while the cosmic censorship hypothesis is a
crucial assumption underlying all blackhole physics, gravitational collapse
theory, and many important related areas in gravity physics, no proof and
formulation is available today. The first major task here is actually to formu-
late rigorously a satisfactory version of the hypothesis. The proof of cosmic
censorship would confirm the already widely accepted and applied theory
of blackholes, while its overturn would throw the blackhole dynamics into
serious doubt. Therefore, cosmic censorship turns out to be one of the most
important issues for general relativity and gravitation theory today. Even if
true, a proof for this conjecture does not seem possible unless some major
theoretical advances of the mathematical techniques and the understanding
of the global structure of the Einstein equations are made.
This situation leads to the conclusion that the first and foremost task here
is to carry out a detailed and careful examination of gravitational collapse
scenarios that possibly give rise to a naked singularity formation. Until this is
done properly, trying out different formulations for censorship may not help,
because without really knowing what is involved in the collapse processes,
one may be in for a complete surprise as far as the final fate of a gravitational
collapse is concerned. It is only such investigations of general collapse situ-
ations that could indicate which theoretical advances to expect for a proof,
and which features to avoid while formulating the cosmic censorship.
Under this situation, while no mathematical formulation or proof of cen-
sorship is currently available, another alternative is to ask if it would be
possible to impose suitable physical constraints on the gravitational collapse
so as to ensure the validity of the censorship. In other words, physically real-
istic collapses should satisfy the censorship. Then again, the precise physical
conditions under which the censorship is supposed to be holding are to be
specified. The advantage then is that even if a certain set of physical condi-
tions did not work towards proving the censorship, there is still the option
of trying out another set of physical constraints to continue further efforts.
Eventually, this may lead to an appropriate mathematical formulation of
the censorship conjecture to be established. Many natural looking physical
conditions can be proposed and tried out, with these being indicated as a
remedy to rule out naked singularities.
180 Cosmic censorship

Discussed below are several such physical constraints on a realistic gravita-

tional collapse scenario, and the implications they have towards determining
the final fate of the collapse. In particular, the motivation is to rule out a
naked singularity as the final state by imposing such conditions. It turns out
that a naked singularity cannot be ruled out with the help of such condi-
tions considered so far. But the advantage of such an analysis is, first, that
it clarifies the situation as to what the conditions can possibly achieve. Sec-
ond, it serves as a pointer to something deeper that should be looked for if a
censorship is to be established. This also could imply that, in fact, naked sin-
gularities do develop in wide classes of gravitational collapse scenarios under
realistic physical conditions. An attempt is also made here to get an insight
into why many of these conditions have not worked, or are unlikely to work
in establishing a censorship, and why further more subtle alternatives must
be explored. Finally, it is indicated that hope appears to lie in a detailed
genericity and stability analysis only.

1. A suitable energy condition must be obeyed.

This is one of the basic conditions assumed in the classical gravity descrip-
tion, and it should be satisfied by the matter fields constituting the star, at
least until the collapse has proceeded to such an advanced stage so as to enter
a phase governed by quantum gravity. This is the stage when the classical
description starts breaking down in one way or another.
As noted earlier, if completely arbitrary matter fields are allowed for, it
is quite easy to produce naked singularities. For example, one possibility is
to start with a geometry allowing families of future directed non-spacelike
geodesics, which are future endless, but which terminate in the past at the
singularity. Then, define the matter fields to be given by Tij ≡ (1/8π )Gij .
It is thus obvious that in gravitational collapse consideration must be given
to scenarios where matter fields do satisfy reasonable physical conditions. It
would be hoped that a suitable energy condition would be one of these, as all
observed classical fields do obey such a condition. A further motivation would
be the energy conditions that have been used extensively in the singularity
theorems in general relativity, and which predict the existence of singularities
in gravitational collapse and cosmology.
It would be nice to see if the censorship is obeyed once the matter fields
have been assumed to satisfy suitable energy conditions. It turns out, how-
ever, that there are several classes of collapse models where collapsing matter
does satisfy a proper energy condition, but the collapse leads to an endstate
that is a naked singularity.
Actually, there are classes of collapse models where satisfying the energy
condition appears to be aiding the naked singularity formation as the final
state of collapse, in turn making the naked singularity physically more
 4.5 Formulating the censorship 181

interesting and serious. An example of this is the spherically symmetric

self-similar collapse of a perfect fluid. The general form of the metric is

ds2 = −e2ν(r,t) + e2ψ(r,t) + r2 S 2 (r, t)(dθ2 + sin2 θ dφ2 ), (4.72)

where the metric functions are taken to depend on X = t/r due to self-
similarity. The outgoing null geodesics (Joshi and Dwivedi, 1992, 1993a)
can be worked out from the naked singularity, which turn out to be related
to the density and pressure distributions in the spacetime via the Einstein
equations. These are then given by

r = D(X − X0 )2/(H0 −2). (4.73)

Here, H0 is the limiting value of the quantity H = (η + p)e2ψ , with η

and p corresponding to density and pressure, and D > 0 is the constant of
integration. The weak energy condition is then equivalent to the statement
that H0 > 0, which in turn ensures, from the above geodesics equation, that
families of null geodesics, as opposed to single isolated curves, emerge from
the naked singularity at t = 0, r = 0, which is a node in the (t, r) plane.
It should be noted that the Einstein equations do not require or impose an
energy condition on matter distributions. It is a criterion motivated purely
on physical grounds. This then suggests another possibility, namely that if
somehow in the later stages of a collapse the energy conditions are violated
through whatever agency, then there may be a hope to preserve the cosmic
censorship. In the equation for null geodesics, if the energy conditions are
violated, this would correspond to a negative value of H0 , then there are no
outgoing null geodesics families from the singularity, and the censorship is
essentially preserved.

2. The collapse must develop from regular initial data.

This is one of the most important physical constraints necessary for any
possible version of a cosmic censorship statement. In general, the regularity
conditions on the initial data for the collapse can come in many forms. If
realistic collapse scenarios of matter clouds such as gravitationally collapsing
massive stars are to be modeled, then the densities, pressures, and other
physical quantities must be finite and regular at the initial spacelike surface
from which the collapse develops. That is, the initial surface should not admit
any density or curvature singularities in the initial data so as to represent a
collapse from regular matter distribution.
Generally, this is ensured by imposing the usual differentiability conditions
on the functions involved, together with requirements of finiteness and regu-
larity. It is known from the gravitational collapse analysis so far, that regular
182 Cosmic censorship

distributions of initial densities and pressures (for example, they should be

finite and suitably differentiable on the initial surface) do give rise to both
naked singularities and blackholes, depending on the nature of the regular
initial data from which the collapse evolves, as discussed in Chapter 3. It
turns out that given such matter initial data, there are still sufficient num-
bers of free functions available to choose in the Einstein equations, subject
to the weak energy condition and suitable matching to the exterior of the
collapsing cloud, so that the evolution can end in either the blackhole or
naked singularity outcomes as desired.
At times, more stringent requirements are imposed on the initial data,
for example, a complete smoothness of densities, pressures, and the metric
functions could be asked for. Usually, there are two motivations for this.
One could be the requirements while calculating the analytic or numerical
evolutions where smoothness (which is the same as demanding the analyticity
of these functions) simplifies the analysis considerably. At other times, it is
argued that astrophysically reasonable initial data must be analytic. In the
case of the collapse of a dust cloud, this amounts to demanding analyticity
of the density function. The initial density ρ(r) then must contain no odd
powers in r, and
ρ(r) = ρ0 + ρ2 r2 + ρ4 r4 + · · · (4.74)

at the initial surface t = ti , which gives an analytic density profile. It is

known, however, that even in the case of smooth density profiles with only
the even terms being non-vanishing, the marginally bound dust evolution can
end in a naked singularity (for example, when ρ2 = 0) that is gravitationally
strong. That is, sufficiently fast divergence of curvatures does take place in
the limit of approach to the singularity.

3. Singularities from realistic collapses must be gravitationally strong.

This has been one of the most useful physical requirements, which was
explored rather thoroughly in order to develop a formulation for the cos-
mic censorship conjecture. The idea has been that any singularity that will
develop from a realistic collapse has got to be physically serious in various
aspects, including powerful divergences in all important physical quantities
such as densities, pressures, curvatures and others, at least at the classical
level. A typical condition for the singularity to be gravitationally strong is,
in addition to the divergences such as above, the gravitational tidal forces
must diverge and all physical volumes are crushed to zero size in the limit of
approach to the naked singularity. A sufficient condition for this to happen is

1
Rij V i V j ∝ , (4.75)
k2
 4.5 Formulating the censorship 183

where k is the affine parameter along the non-spacelike geodesics energing

from the singularity, with k = 0 at the singularity, and V i is the tangent
vector to these curves emanating from the naked singularity. The above
curvature condition corresponds to that developed through the analysis of
Clarke, Królak, and Tipler (see Joshi, 1993, and Nolan, 1999, for a further
discussion).
The singularity developing within the blackhole formed out of the stan-
dard dust cloud collapse as investigated by Oppenheimer and Snyder (1939)
is gravitationally strong in the above sense. Now, if it could be established
that whenever naked singularities form in the gravitational collapse, they
are always gravitationally weak, in the sense of important divergences such
as these above not being present in the limit of approach to the singular-
ity, then such singularities could be removable from the spacetime, and it
may be possible to extend the spacetime. Such removable naked singularities
should no longer be regarded as physically genuine, and progress has then
been made towards the cosmic censorship in some form, such as that the
naked singularity could develop in the gravitational collapse; however, they
would be gravitationally weak and removable always.
This possibility has been investigated thoroughly, and it is known now that
gravitationally powerfully strong naked singularities actually do result from
the collapse from the regular initial data (including smooth analytic density
profiles), for several reasonable forms of matter such as dust, perfect fluids,
Vaidya radiation collapses, and several other forms of matter that satisfy
suitable energy conditions. In the next chapter, one such example, in the
case of the Szekeres quasi-spherical collapse models will be discussed. At such
naked singularities, the densities, curvature scalars such as the Kretschmann
scalar, and gravitational tidal forces diverge most powerfully as characterized
above. This is as powerfully strong as the divergences observed at physical
singularities such as the big-bang in cosmology.

4. The matter fields must be sufficiently general.

If a naked singularity formed in the collapse for certain special forms of

matter, such as dust or collapsing radiations only, that would not be of
much interest. For example, the role of pressure cannot be underestimated
in a realistic collapse and so it would be nice to know if matter with pressure
would necessarily give rise to a blackhole only on undergoing gravitational
collapse. If this were the case, matter fields giving rise to a naked singularity
could then be ruled out as special or unphysical, in formulating the censor-
ship, even if they satisfied an energy condition or the collapse developed from
regular initial data.
It is now known, however, as discussed in Chapter 3, that naked singu-
larities are not special to any particular form of matter field, and several
184 Cosmic censorship

physically important equations of state are also included. The collapse can
be studied for a general form of matter, the so-called type I matter fields (all
the known physical forms of matter, such as dust, perfect fluids, and massless
scalar fields are included in this class), subject to the weak energy condition.
The result is, given an arbitrary but regular distribution of matter on the ini-
tial surface, that there are always evolutions available from this initial data
that would either result in a blackhole or naked singularity, depending on the
allowed choice of free functions available from the Einstein equations. More
specifically, in a spherically symmetric collapse with a type I general matter
field, given the distribution of density and the radial and tangential pressure
profiles on the initial surface from which the collapse develops, it is then
possible to choose the free function describing the velocities of the in-falling
shells in such a manner so as to have a blackhole or a naked singularity as
the final end product, depending on this choice.

5. The collapsing cloud must obey a realistic equation of state.

It is conjectured at times that even though naked singularities may develop

for general matter fields, they must go away once a physically reasonable and
realistic equation of state is chosen for the collapsing cloud.
This is a very difficult argument to formulate. First, naked singularities do
form in the collapse of several well-known equations of state, such as dust,
perfect fluids, or in-flowing radiation shells. Second, it is quite difficult to
make any guesses as to the state of the matter, or the realistic equation of
state within a collapsing body such as a massive star that is in its advanced
stages of collapse. Third, the collapsing cloud may not have a single equation
of state, which might actually be changing as the collapse evolves. There have
been suggestions that strange quark matter may be a good approximation
to the collapsing star in its final stages, and the collapse was then examined
in a Vaidya geometry, which again results in blackhole or naked singularity
phases as usual. Therefore, such a choice of equation of state does not remove
naked singularities. At the other extreme, as mentioned earlier, there are also
arguments such as those given by Hagerdorn and Penrose, that the equation
of state, in the very final stages of the collapse, closely approximates that
of dust. In other words, at higher and higher densities, matter may behave
more and more like dust. The point is, if pressures are not negative, then
they may also contribute positively to the collapse just to add to the dust
effect, and may not alter qualitatively the conclusions arrived at in the dust
case. In such a case, the dust collapse situation, which has been investigated
rather thoroughly, would imply that both blackholes and naked singularity
phases would develop in the gravitational collapse, depending on the initial
density and velocity distributions.
 4.5 Formulating the censorship 185

An interesting feature is that, while there are several widely used and
familiar equations of state available that result in the formation of naked
singularities as the final fate of the collapse, there is still not a single equation
of state available so far that ensures necessarily that the end product will be
a blackhole only. Under this situation, it is quite possible that the physics
that causes collapse endstates may not be directly related to the equation of
state or the form of matter collapsing.
A more promising alternative could be to work with a rather different
representation of the matter, such as the one given by the Einstein–Vlasov
statistical description (see for example, Rendall, 2005).

6. All radiations from a naked singularity must be infinitely red-shifted.

In certain subcases of dust collapses resulting in a naked singularity, it

is seen that the red-shift along the null geodesics emerging from the naked
singularity diverges in the limit of approach to the naked singularity. This
has given rise to the possibility that, even if a naked singularity forms in the
collapse, no energy could escape. In this sense, the naked singularity may
be invisible to any external observers for all practical purposes. Of course,
it has to be noted that even if true generally, this does not help the cosmic
censorship hypothesis in the actual sense as, basically, cosmic censorship
is about the question of principle in gravitation theory, namely whether
singularities forming in the gravitational collapse are causally connected to
an external observer, or not, via non-spacelike trajectories.
In any case, it is useful to explore such a possibility, because it may give
some information on the structure of the naked singularity, at least in cer-
tain special models, and if true generally, then it will provide some kind of a
physical formulation for the cosmic censorship. It is, however, possible that
establishing in general that no energy can come out from a naked singularity
can turn out to be very difficult. There could be several reasons for this.
First, it is very tricky to apply the conventional definition of red-shift, which
corresponds to a regular source and observer, to emissions from a naked sin-
gularity. A singularity, by definition, is not a point of the spacetime. Second,
even if there was no escape of energy along the null geodesics, the possibility
of mass emission via timelike or non-spacelike non-geodesic families of paths
emerging from the naked singularity remains open in the very late stages of
the collapse in a burst-like fashion. In the case of such a violent event being
visible, particles escaping with ultra-relativistic velocities cannot be ruled
out from the neighborhood of the naked singularity.
It may also be noted that the classical possibilities regarding the prob-
able light or particle emission, or otherwise, from a naked singularity may
not possibly offer a serious physical alternative, one way or the other. The
reason is that the classical general relativity may break down eventually in
186 Cosmic censorship

the very late stages of the collapse, once the densities and curvatures are
sufficiently high so that quantum gravity effects become important in the
process of a continual collapse. Such quantum effects would come into play
much before the actual formation of the classical naked singularity, which
itself may possibly be smeared out by the quantum effects.
The key question then is that of the possible visibility, or otherwise, of
these extreme strong gravity regions, which do develop in the vicinity of
the classical naked singularity. It is then the causal structure, that is, the
communicability, or otherwise, of these extreme strong gravity regions where
quantum gravity should prevail, which would make the essential difference as
far as the physical consequences of a naked singularity are concerned, rather
than various purely classical aspects such as red-shift.
As seen above, the physical conditions are not able effectively to rule out
naked singularities, which in turn may lead to some possible formulation of
the cosmic censorship conjecture, either a physical or a mathematical one.
With each of the above conditions, there are counter-examples that obey
such a physical constraint, but which produce a naked singularity as the
endstate of a dynamical collapse.
There are three further possibilities that are under active investigation
today towards a possible formulation of the censorship conjecture, and which
may offer a better hope for a cosmic censorship hypothesis. These are now
briefly discussed below.

7. Will quantum gravity remove naked singularities?

It is sometimes argued that after all the occurrence of singularities is a

classical phenomena, and that whether they are naked or covered should not
be relevant – quantum gravity, a final theory in which there should be no
singularities will remove them all anyway. But this is missing the real issue,
it would appear. It is possible that in a suitable quantum gravity theory,
the singularities will be smeared out (although this has not been realized so
far), and there are indications that, in quantum gravity also, the singularities
may not fully go away. However, it appears that the real issue is whether the
extreme strong gravity regions formed due to gravitational collapse are vis-
ible to faraway observers or not. The collapse certainly proceeds classically,
until quantum gravity starts governing the situation at scales of the order
of the Planck length, that is, until the extreme gravity configurations have
developed due to collapse. It is the visibility, or otherwise, of such regions
that is under discussion.
The point is, classical gravity implies necessarily the existence of strong
gravity regions, where both classical and quantum gravity come into their
own. In fact, if naked singularities do develop in the gravitational collapse,
then, in a literal sense, one comes face-to-face with the laws of quantum
 4.5 Formulating the censorship 187

Outgoing
light rays

Quantum gravity
fuzzy region
Collapsing Collapsing
star star

Fig. 4.6 Quantum effects may resolve the spacetime singularity, either covered
or naked. If the collapse outcome is a naked singularity, there may be the oppor-
tunity to observe the quantum gravity effects taking place in the ultra-strong
gravity regions in the universe.

gravity whenever such an event occurs in the universe (see for example, Vaz
and Witten, 1994, 1995, and Wald, 1997). Then, the gravitational collapse
phenomena could provide a possibility of actually testing the laws of quantum
gravity, and every time a massive star collapses in the universe, there is
potentially a laboratory to test the laws of quantum gravity (see Fig. 4.6).
For an earlier discussion on quantum effects near singularities, see Kodama
(1979) and Hiscock, Williams, and Eardley (1982).
In the case of a blackhole developing in the collapse of a finite sized object
such as a massive star, such strong gravity regions have got to be necessarily
hidden behind an event horizon of gravity, which would be well before the
physical conditions became extreme. Then the quantum effects, even if they
caused qualitative changes closer to the singularity, will be of no physical
consequence. This is because no causal communications are then allowed
from within such horizons. On the other hand, if the causal structure were
that of a naked singularity, then communications from such a quantum grav-
ity dominated extreme curvature ball would be visible in principle, either
directly or via secondary effects such as shocks produced in the surrounding
medium. Some of these issues will be discussed further in the next chapter.

8. Should all naked singularities produced by matter fields be considered to be

unphysical?

There has been a suggestion that all naked singularities, whenever they are
produced by matter fields such as dust, perfect fluids, and such others, should
188 Cosmic censorship

be rejected as being only ‘matter singularities,’ which should have nothing to

do with pure gravity. From such a perspective, the naked singularities caused
by massless scalar fields will be of some concern, which is included in the type
I matter fields discussed above. While realistic stars are not made up of mat-
ter fields such as a massless scalar field, in the very final stages of the collapse
such matter forms may have an important role to play in some manner.
It is known, however, that matter forms, such as those above with various
equations of state, have been used extensively in astrophysical studies and it
would be difficult to reject these and their logical consequences outright in
collapse studies. After all, the classic gravitational collapse scenario, which
is really at the foundation of blackhole physics and its chief motivator, is the
homogeneous dust collapse model, as studied by Oppenheimer and Snyder
(1939). Now, in the same models, when a density gradient is put in at the
center, which is a physically more realistic situation, then a naked singu-
larity rather than a blackhole results. The structure of the event and the
apparent horizons then change drastically so as to expose the singularity to
an external observer. All realistic stars will typically have a higher density
at the center, falling off at some rate moving away from the center. In this
sense, one may want to regard the naked singularity developing due to this
density gradient, to be at least as physical as the blackhole. In general rel-
ativity, there have been very many studies with dust, perfect fluids, and
other forms of matter for several decades, and one might want to accept
the logical outcomes available within those collapse scenarios for further
studies.
Again, if arguments, such as those given earlier, are considered in favor of
the equations of state such as dust in the final phases of the collapse, the
outcomes of such a collapse could be taken physically more seriously.

9. Are naked singularities stable and generic?

It would appear that this is the key issue on which any possible future
formulation and proof of the censorship has to crucially depend. Even if
naked singularities do develop in collapse models, if they were not generic
and stable in some suitably well-defined sense, that would make a good case
for censorship. For example, most of the current classes of naked singu-
larities discussed here are within the framework of a spherically symmetric
collapse. While there are some indications that naked singularities do develop
in non-spherical collapses as well, as discussed in the next chapter, such a
non-spherical collapse remains largely uncharted territory, and it would be
essential to examine it rather thoroughly.
The key question to be resolved here is, whilst it is known that physically
reasonable initial data do give rise to naked singularities, will the initial data
subspace, which gives rise to the naked singularity as the collapse endstate,
 4.5 Formulating the censorship 189

have a non-zero measure in a suitably defined sense? As is well-known, how-

ever, the formulation of the stability concept in general relativity is a rather
complicated issue, as there are no well-defined formulations or criteria to test
stability. Before the censorship can be tested, a satisfactory formulation for
the stability criterion has to be arrived at within the framework of general
relativity. Also, the issue of what is a suitable measure in the initial data
space can be a complicated one. Only after making some reasonable progress
here could testing these questions on the naked singularity formation start.
While discussing stability and genericity, great care has to be taken on the
criterion used to test them, as sometimes a criterion that also makes black-
holes unstable, while trying to show the instability of naked singularities can
be used.
In the absence of such well-defined criteria against which to test the avail-
able naked singularity models, various attempts have been made to examine if
they would be stable to some kinds of perturbations. These attempts include
perturbing the density profiles to include pressures, trying to see how the
density gradients at various levels affect the global versus the local visibility
of the naked singularity, imposing symmetry conditions such as self-similarity
and then seeing how the conclusions change on relaxing the self-similarity
condition, studying how certain perturbations grow in the limit of approach
to the Cauchy horizon, which is the first ray coming out of the naked singu-
larity, and such others. While these attempts do not provide any definitive
conclusions regarding the stability, or otherwise, of the naked singularity,
they certainly provide a good insight into the phenomena of the blackhole and
naked singularity phases to show what is possible in gravitational collapse.
Given the complexity of the Einstein field equations, if a phenomenon
occurs so widely in spherical symmetry, it is not unlikely at all that the same
would be repeated in more general situations as well. In fact, before the
advent of well-known singularity theorems in general relativity, it was widely
believed that the singularities found in more symmetric situations such as
the Schwarzschild or Friedmann–Robertson–Walker cosmological models will
go away once the spacetimes are general enough. As is well-known, the
singularity theorems then established that spacetime singularities occur in
rather general spacetime settings without symmetry assumptions, and under
a broad set of physical conditions. Therefore, the singularities that mani-
fested earlier in symmetric situations were indicative of a deeper phenomena.
Such a possibility cannot again be ruled out in the case of the occurrence of
naked singularities.
Clarified above were the basic philosophy and motivation for cosmic cen-
sorship and the crucial role they play in blackhole physics. Some of the
approaches that have been tried out so far to formulate or prove the censor-
ship were then outlined. It turns out that none of the physical constraints or
natural looking physical conditions devised so far are really able to ensure the
190 Cosmic censorship

validity of the cosmic censorship hypothesis. In fact, one tends to conclude

that naked singularities can actually develop in physically realistic gravita-
tional collapse situations. It then follows that more radical options, some of
which were discussed above, must be formulated and tried out if the cen-
sorship conjecture is to be preserved. In fact, it would appear that only one
of these, namely the one involving the stability and genericity of naked sin-
gularities, can be a potentially promising alternative as far as any possible
proof of the censorship hypothesis is concerned.
As it is attempted to work towards censorship along one of these, or other,
paths, it would in fact be important and quite interesting to really under-
stand why naked singularities do actually develop in a gravitational collapse.
As pointed out above, several important physical constraints on collapsing
clouds do not appear to work towards helping the censorship hypothesis.
It then becomes an intriguing question as to what is the physical agency
that seems to be causing a naked singularity in the gravitational collapse in
a rather natural manner within the framework of general relativity. Some
work has been carried out in that direction, and it turns out that while
the gravitational collapse proceeds, the shearing effects and inhomogeneity
within the cloud could play a basic role in delaying the formation of trapped
surfaces and the apparent horizon in a natural manner. This in turn exposes
the singularity to outside observers, depending on the rate of growth of the
shear in the limit of approach to the center. While shear can be one physi-
cal agency to delay trapping, there can be a naked singularity in shear-free
collapse as well. When looked at from such a perspective, it may be thought
that both blackholes and naked singularities are rather natural consequences
of a gravitational collapse in classical general relativity.
It is then possible that the cosmic censorship conjecture does not hold
classically, but may hold quantum mechanically in some sense that is yet
to be figured out. It may be possible then that, for a star going into the
final state of a naked singularity configuration, the quantum gravity induced
particle creation may take over to create a thunderbolt-like burst of energy,
thus clearing up the naked singularity. Such a scenario would be of physical
interest because a naked singularity may have theoretical and observational
properties quite different from a blackhole endstate, and it might make the
communications from extreme strong gravity regions dominated by quantum
gravity possible.

4.6 Genericity and stability

Are naked singularities developing in a gravitational collapse stable and
generic? This is one of the most important questions as far as any possi-
ble proof of the cosmic censorship is concerned. If naked singularities formed
in the collapse are not stable or generic in a suitable sense, they may not be
 4.6 Genericity and stability 191

taken seriously. The key difficulty to addressing this question is that there
are no well-defined criteria or formalisms available in the general theory of
relativity to test the stability and genericity. As opposed to Newtonian the-
ory, there is no well-defined notion for stability available in Einstein’s theory.
Under this situation, there are a variety of ways in which the above question
can be asked and treated, and there is no unique answer available in a general
manner. Many times, it is the physics of the situation that guides the path
adopted, as pointed out above. Nevertheless, this is an important and basic
issue to try to answer.
The singularity theorems establish the existence of spacetime singularities
in the form of incomplete non-spacelike geodesics, both for gravitational
collapse and cosmology. These theorems, however, give no information on the
nature of the singularities, such as whether they occur in the past or future,
the possible growth of curvature in the limit of approach to the singularity,
and whether they will be covered by an event horizon hidden from all outside
observers, or whether they may be causally connected to external observers
in the universe. Hence, while dealing with gravitational collapse scenarios it
becomes inevitable to assume the cosmic censorship in the form of a future
asymptotic predictability of the spacetime, in order to make any progress in
the theory and applications of blackholes. It has to be ensured, through the
censorship conjecture, that the singularities of the collapse are necessarily
hidden inside an event horizon. This assumption essentially relates to the
causal structure in the later stages of a gravitational collapse.
Here, the occurrence of naked singularities arising in the gravitational
collapse of massive matter clouds such as stars have been discussed. The
examination of such dynamical collapse scenarios, concerned with the evo-
lution of regular initial data from a well-behaved initial value surface, imply
the following basic conclusions. First, such a naked singularity forms in the
dynamical evolution of several forms of matter, such as the collapse of in-
flowing radiation, dust, or perfect fluids. Second, a non-zero measure set
of non-spacelike trajectories, in the form of families of non-spacelike curves,
is emitted from the naked singularity, as opposed to a single null geodesic
escaping, which corresponds to a single wave front emerging. Finally, such a
singularity is physically significant in the sense that it is a powerfully strong
curvature singularity as the curvatures diverge rapidly along all the trajec-
tories meeting the naked singularity in the past. It can also be noted that
in dust collapse (see Fig. 4.7), and in similar models, these are seen to arise
from a non-zero measure set of initial data sets (Saraykar and Gate, 1999;
Mena, Takavol, and Joshi, 2000).
How seriously such a naked singularity is to be taken, is it generic and
stable, and what are the implications towards the formulation and proof of
the cosmic censorship hypothesis? It may be noted that the gravitational
collapse situations investigated so far have been spherically symmetric. Is it
192 Cosmic censorship

0.96

0.94

M3

0.92

0.9
33 34 35 36 37
E5 Naked singularity
Blackhole

Fig. 4.7 The blackhole and naked singularity phases in a dust collapse, in terms
of the initial density and velocity profiles (from Mena, Tavakol, and Joshi, 2000).

possible that the naked singularities occurring are artifacts of this assumed
symmetry? As discussed later, there are results that show that naked sin-
gularities occur in some non-spherical collapse models as well, and so their
existence need not be due to the assumed spherical symmetry only. Further-
more, as was shown by the singularity theorems, the singularities developing
in spherical situations still persist, even when small perturbations are taken
into account. It is possible that a similar situation may arise here also, and
so the detailed investigation of the spherically symmetric scenario becomes
quite important.
In any case, if the cosmic censorship is generically correct, it has to hold in
spherical symmetry as well. Hence the results here on the spherical collapse
show that only a substantially fine-tuned version of the censorship holds if
one exists.

4.6.1 Topology on the space of metrics

The stability or genericity of a property in spacetimes can be characterized
in the following way. If a spacetime (M, g ) possesses a certain property,
 4.6 Genericity and stability 193

then for this property to be stable, it would be expected that all the nearby
spacetimes would have the same property. For example, one says that the
spacetime is causal, and not admitting any closed non-spacelike curves is
a stable property, if for all nearby metrics ḡ on the same manifold, the
spacetime (M, ḡ ) also remains causal. Similarly, for a spacetime having naked
singularities, one would like to check if these naked singularities are stable. If
(M, g ) has naked singularities, but all nearby (M, ḡ ) have none, then these
are not stable.
The issue then is to make this notion of ‘nearby’ metrics more precise, and
for that one has to have a topology on the space of all metrics ḡ for a given
spacetime M . One way to do this is to define a C 0 -topology, by defining the
nearby metrics to be those that are nearby in their values at the spacetime
coordinates, and the open balls can be defined by the requirement that for
a given  > 0, the ball consists of all metrics ḡ , such that |gij − ḡij | < .
The issue then would be, should it be required only for the values to be
nearby, or must the derivatives also be nearby? In general relativity, the
metric tensor is required to be at least C 2 . If the first derivatives are also
required to be close, then a C 1 -topology should be used, and if the second
derivatives are required to be close, then a C 2 -topology has to be used. A
concerned property, such as the spacetime having, or not having, a naked
singularity should then be examined in such a context with the appropriate
topology defining the nearby metrics. There is no uniqueness in defining such
a topology on the space of all metrics, and a C ∞ -topology could be used on
the space of all metrics. The main difficulty in a stability and genericity
analysis in general relativity is then that of identifying a suitable physically
relevant topology for the space of all metrics, and then it has to be examined
if there are open sets with non-zero measures with a given property.
As there is no uniqueness in the above procedures, the typical approach
has been to examine perturbations of a given situation, as directed by the
physical requirements of the model. For example, in a gravitational collapse,
for the homogeneous dust collapse, one may try and see if the final state of
the collapse remains unchanged if perturbations that are inhomogeneities in
a given homogeneous profile are introduced, or one may want to perturb the
original solution to a new solution that allows for small non-zero pressures
for a given initial data from which the collapse evolves. Very many such
analyzes have been carried out for gravitational collapses, as indicated in the
previous chapter and in the discussion here.

4.6.2 Censorship and genericity issues

The cosmic censorship hypothesis, as suggested by Penrose (1969, 1979),
emphasizes that the criteria of stability and genericity must be satisfied in
194 Cosmic censorship

some well-defined sense. For example, it could be required that spacetimes

admit no naked singularities with respect to the changes in the initial data,
or the equation of state. Therefore, the strong censorship hypothesis can
be stated as saying that stable spacetimes must be globally hyperbolic, or
that they do not admit locally naked singularities. A similar statement for
the weak censorship would be that stable spacetimes do not admit globally
naked singularity. Hence, one would like to know if the gravitational col-
lapse scenarios discussed in the previous chapter, and the naked singularities
forming are stable in a suitable sense, say under small departures from the
spherical symmetry or changes in the equation of state.
As pointed out above, a general analysis on the question of stability is,
however, a rather complicated issue because the stability theory in general
relativity is a largely uncharted domain on which little is known. For exam-
ple, an implication of the strong censorship principle, as stated above, would
be that singularities that are spacelike in nature must remain spacelike after
a small perturbation in the spacetime. However, some care has to be taken in
formulating such a statement. The reason is that the Schwarzschild singular-
ity could be thought of as being unstable in this sense, because the addition of
even a small amount of charge or angular momentum changes the character
of the singularity and the nature of the solution in a basic way. In this case,
the solution changes to the Reissner–Nordström or Kerr spacetime where
the singularity is not necessarily spacelike. The addition of a small charge
changes the model so that the singularity is locally naked. It is only when a
generic type of perturbation is introduced in the spacetime that the spacelike
nature of the singularity may be retained. It is not clear, however, what such
a generic perturbation would mean in general, and basically a well-defined
stability theory in the framework of general relativity is needed here.
A possible approach to this, as indicated earlier, is to examine the stability
of the Cauchy horizons that must form whenever the strong cosmic censorship
is violated, that is, whenever global hyperbolicity of the spacetime breaks
down. The Reissner–Nordström case provides an important clue here. In this
case, the future Cauchy horizon extends all the way to spatial infinity, and
one way of specifying this could be the following. Given any partial Cauchy
surface S , the Cauchy horizon H + (S ) associated with it has the property
that, given any point p ∈ H + (S ), the set I − ( p) ∩ S is non-compact. In such
a case, even a small perturbation in the initial data on the partial Cauchy
surface will grow and diverge at H + (S ), causing a blue-shift instability. The
reason for this is that the signals from faraway regions on S are infinitely
blue-shifted at H + (S ), and in fact, weak field perturbations would diverge
along H + (S ) (Simpson and Penrose, 1973; McNamara, 1978; Chandrasekhar
and Hartle, 1983). In such a situation, a curvature singularity, rather than a
Cauchy horizon may really occur. This would imply that H + (S ) is not stable
against sufficiently small perturbations in the initial data. This analysis of
 4.6 Genericity and stability 195

the Reissner–Nordström case provides an indicator that the Cauchy horizons,

when they form, could be unstable. In fact, such an instability is seen to be
occurring for the wider class of spacetimes of the Kerr–Newman family, of
which the Reissner–Nordström situation discussed above is a special case.
Care has to be taken in formulating this, because this is a different type of
strong censorship violation compared with that occurring in the dynamically
developing collapse geometries such as the Vaidya model or the Tolman–
Bondi–Lemaı̂tre case, where the set discussed above is always compact, for
any p ∈ H + (S ). It could then be concluded that the best hope for the
censorship lies in analyzing the genericity and stability properties of the
currently known classes of collapse models that lead to the formation of
naked singularities, rather than blackholes as the final state of collapse.
The point of view that emerges then is, as far as the occurrence is con-
cerned, that both blackholes and naked singularities appear to be basic
properties of the gravitational collapse, as a consequence of the dynamics
of the Einstein equations, and emerge in a natural manner as a logical conse-
quence of the general theory of relativity. However, the crucial issue is that of
the genericity and stability of the naked singularities. It would appear that
the real hope for the censorship lies in investigating in detail the stability
properties of the collapse models that develop into a naked singularity.
Discussed below in some detail are some of the issues, such as those
mentioned above, that are related to the genericity, with the help of a higher-
dimensional model. This is related to the question of the validity of the cosmic
censorship in a higher-dimensional collapse scenario. As it is possible to work
out explicitly when the families of non-spacelike geodesics can come out from
the singularity, or otherwise, such genericity issues can now be discussed in
some detail.
The purpose has been to examine if it is possible to recover the cosmic
censorship when a transition is made to a higher-dimensional spacetime,
by studying the spherically symmetric dust collapse in an arbitrary higher
spacetime dimension. Earlier, it was shown how certain classes of naked sin-
gularities are removed once the transition to a higher-dimensional spacetime
is made. It would be of interest to see to what extent such a result is gen-
eralized in the dust collapse towards the naked singularity removal, while
commenting on certain genericity aspects.
If only blackholes are to result as the endstate of a continual gravitational
collapse, several conditions must be imposed on the collapsing configuration,
some of which may appear to be restrictive. This needs to be studied carefully
if these can be suitably motivated physically in a realistic collapse scenario.
The approach developed here generalizes and unifies the earlier available
results on a higher-dimensional dust collapse. Furthermore, the dependence
of a blackhole or a naked singularity as collapse outcomes on the nature
of the initial data from which the collapse develops, comes out explicitly.
196 Cosmic censorship

The method used allows the genericity and stability aspects related to the
occurrence of naked singularities in a gravitational collapse to be considered.
The main motivation for studying a higher-dimensional collapse is that, while
the censorship may fail in the four-dimensional manifold of general relativity,
it can possibly be restored due to the extra physical effects arising from the
transition itself to a higher-dimensional spacetime continuum.
There have been several investigations in recent years on the spherically
symmetric collapse of dust in higher-dimensions. The recent revival of inter-
est in this problem is motivated, to an extent, by various higher-dimensional
theories, including the string paradigm to unify the forces of nature, and the
brane-world scenarios. Different specialized subcases of the general problem
of dust collapse in higher spacetime dimensions have been considered. For
example, the marginally bound case in a general spacetime dimension was
studied by Ghosh and Beesham (2001). The same case was also studied with
an added and physically motivated assumption on initial density profiles,
that the first derivatives of the initial density distribution for the collapsing
cloud must be vanishing (Sil and Chaterjee, 1994; Patil, Ghate, and Saraykar,
2001; Banerjee, Debnath, and Chakraborty, 2003). Also, the non-marginally
bound case with the geometric assumption that spacetime is self-similar, was
examined by Ghosh and Banerjee (2003) for a five-dimensional model.
These studies do provide an idea of what is possible in a gravitational
collapse as far as its endstate spectrum is concerned. It is obvious from the
discussion so far that any possible proof of the censorship must be inspired by
additional physical inputs into the current framework of thinking, with one
of these being a possible transition to higher spacetime dimensions. Any such
alternatives would be worth exploring due to the fundamental significance
of the censorship in blackhole physics. If naked singularities did develop in
realistic gravitational collapses of massive objects, they may have properties
that would be rather different from those of blackholes, both theoretically as
well as observationally, and a comparison of these two cases may prove quite
interesting.
The effect of dimensions on the final fate of the evolution of the matter
cloud that collapses from a given regular initial data is now examined. A
spherically symmetric dust collapse is considered with N ≥ 4 dimensions.
There have been suggestions that if the in-falling velocity of the matter shells
is so high that the effects of pressures are negligible, then dust may be a
good approximation in the final stages of a collapse. Dust collapse is worth
investigating in any case, as it has continued to serve as a basic paradigm in
blackhole physics.
To focus the discussion, consider a model initial density profile given by

ρ2 ρ3
ρ(ti , r) = ρ0 + rρ1 + r2 + r3 + · · · , (4.76)
2! 3!
 4.6 Genericity and stability 197

and the function M(r) is written as

∞
2ρn
M(r) = Mn rn , Mn = , (4.77)
n=0
(N − 2)(N + n − 1)n!

along with an energy profile as specified by

b2
b(r) = b0 + rb1 + r2 + ··· . (4.78)
2!

First, consider the marginally bound class of collapse models for a trans-
parent understanding of the problem. This is the case when the energy
function b(r) above vanishes identically for the collapsing shells. In this case,
the first non-vanishing coefficient Xn (0), where n > 0, could be worked out
as discussed in Chapter 3. This is given by
 
n! Mn
Xn (0) = − . (4.79)
N −1 M0
3/2

Now, it is evident that whenever ρ1 < 0, there will be a naked singularity

in all dimensions, whereas ρ1 > 0 always results in a blackhole. The case
ρ1 < 0 corresponds to the physical situation when the density decreases
with increasing comoving radius r. This is physically realistic as the density
would typically be expected to be highest at the center and then gradually
decrease outwards in any realistic configuration such as a massive star. Fur-
thermore, note that the above conclusion is not dependent on the magnitude
of ρ1 , but only on its sign, that is, the density should decrease away from
the center with the density gradient being non-zero. Therefore, it becomes
clear that it is the density inhomogeneity that delays the formation of the
trapped surfaces, thus causing a naked singularity. This is closely connected
to the non-vanishing spacetime shear, and in the next chapter it will be dis-
cussed how the inhomogeneities and related shear distort the geometry of
the trapped surfaces.
Now assume that the initial density distribution has all odd terms in r
vanishing, that is, it admits no ‘cusps’ at the center and that it is either
sufficiently differentiable, or is a smooth and analytic function of r. In this
case, ρ1 = 0. Then, from (3.71) in the neighborhood of the singularity, the
behavior of v is given by

 2/(N −1)
N − 1
lim lim v = M0 X2 (0) r4/(N −1) . (4.80)
t→ts r→0 4
198 Cosmic censorship

Also, in the same limit, the function F/RN −3 has the form

F r2 M0
lim lim = . (4.81)
t→ts r→0 RN −3 v (N −3)
Therefore, it is clear from (4.80) and (4.81) that if N > 5, then for the
limt→ts , limr→0 , F/R → ∞ and thus the endstate of collapse will always
be a blackhole, as discussed previously. It follows that for a marginally
bound dust collapse, with ρ1 = 0, that is, when the initial density pro-
file is sufficiently differentiable and smooth, the cosmic censorship holds in a
higher-dimensional spacetime with N = 6, or higher.
In the above, the spacetime dimension was taken to be six or higher. Now
consider the case when the spacetime dimension is five, but still with an
analytic initial density profile. In the case of a five-dimensional marginally
bound collapse with ρ1 = 0, the tangent to the outgoing radial null geodesic
at the singularity in the (R, u) plane can be written as
  
 1 − F/R2
x20 = M0 X2 (0)   . (4.82)
1 + F/R2

The sufficient condition for the existence of an outgoing null geodesic from
the singularity is that x0 > 0, which in the above case amounts to

M2
ξ≡ < −2. (4.83)
M20
But again, the outgoing null geodesic should be within the spacetime, that
is, the slope of the geodesic must be less than that of the singularity curve,
   
dt dt
lim lim ≤ . (4.84)
t→ts r→0 dr null dr sing

From the above equation,

M2
ξ= ≤ −8. (4.85)
M20

Therefore, from (4.83) and (4.85), it can be seen that, for an outgoing null
geodesic from the singularity to exist, ξ ≤ ξc = −8, in which case the result
is a naked singularity, otherwise a blackhole results as the collapse endstate.
Note that this situation has an interesting parallel to the four-dimensional
collapse scenario, where a similar critical value exists. However, it is for
the coefficient ρ3 , when both ρ1 and ρ2 are vanishing. Therefore, with the
increase of the spacetime dimension by one, the criticality separating the
 4.6 Genericity and stability 199

blackhole and naked singularity phases shifts at the level of the second density
derivative from the earlier third density derivative.
An interesting observation that could be made here is that for ξ < −2
there is an increasing apparent horizon at the singularity. The apparent
horizon is given by R = F , so it initiates at the central singularity r =
R = 0 and, in the above case, it is increasing in time (as opposed to the
Oppenheimer–Snyder case of a homogeneous dust collapse). Therefore, for
the range −2 > ξ > −8, no trapped surface is formed before the singularity
epoch, but there is still a blackhole as the collapse endstate. This confirms
that the absence of a trapped surface until before the singularity is necessary,
but not a sufficient condition for the formation of a naked singularity. This
is relevant, especially for numerical collapse simulations, where the criterion
for a naked singularity formation is often taken to be just the absence of
trapped surfaces on an evolving sequence of spacelike surfaces, in a particular
slicing of the spacetime. The above example shows that the mere absence of
trapped surfaces cannot be taken as proof that the collapse terminates in a
naked singularity.
It is useful to note here that the well-known Oppenheimer–Snyder class of
collapse solutions is a special case of a marginally bound dust collapse in four
dimensions, in which case the initial density profile is homogeneous, that is,
Mn (n > 0) = 0 for all n. The point is, if the initial density is homogeneous,
but if the collapse is not marginally bound, then the non-zero energy function
f could inhomogenize the collapse at later epochs. In the present case, as
f = 0, at all later epochs the density also remains a function of time only,
that is, it is homogeneous at all later times as well, and it can clearly be
seen that the final outcome of this class of collapse is always a blackhole.
Furthermore, from (3.68) and (3.53) it can be seen that

ts (r) = ts0 , v (t, r) = v (t). (4.86)

As the scale function v is independent of r, all the shells collapse simultane-

ously to the singularity. The time taken to reach the singularity is given as

2
ts0 = . (4.87)
(N − 1)M0
1/2

Therefore, in going to higher-dimensions, for a given density, the time taken

to reach the singularity will reduce.
It is, however, interesting to note that even if an initially homogeneous
density profile is started with, but if non-zero initial radial and tangential
pressures of the form

pr (t, r) = 1, pθ0 (r) = 1 + pθ2 r2 + pθ3 r3 + · · · , (4.88)

200 Cosmic censorship

are allowed for, then,

 1
1 v (N +5)/2 ( pθ3 )
X (0) = −  . (4.89)
3 0 v (N −1) pθ2 − 13 + 32

Therefore, it is seen that a negative pθ3 coefficient does lead to a naked

singularity.
This is similar to the case of a collapse that is not marginally bound, where
an initially homogeneous density profile can turn inhomogeneous at later
epochs due to the non-vanishing shell velocities (Joshi and Dwivedi, 1993a).
In the same way, in the case above, the non-vanishing pressures could also
inhomogenize the initially homogeneous density distribution at later epochs
to cause eventually a naked singularity as the collapse final state. It has to
be noted, all the same, that the equation of state in situations such as above
could be considered to be somewhat peculiar, although the matter is fully
normal, which satisfies the positivity of the energy condition, and the regu-
larity conditions for the collapse are fully satisfied. To state this differently,
it can be argued that the models where only purely tangential pressures are
taken to be non-vanishing may not be considered to be physically realistic.
If the equation of state is chosen to be, say p = kρ, k > 0, or any homoge-
neous equation of state, then, when the initial density profile is taken to be
homogeneous, then so will the initial pressures, and then the collapse will
end up in a blackhole only, and no naked singularity will arise.
Basically, the purpose of the above discussion was to point out how both
blackholes and naked singularity final states arise rather naturally in a grav-
itational collapse. While one or the other of these can be created or avoided
by means of one or the other conditions on the collapse, these conditions are
typically somewhat restrictive, and generically both these outcomes seem to
arise rather naturally as the collapse endstates. Discussed here were various
special subcases of a higher-dimensional collapse scenario that result either
in a blackhole or a naked singularity, depending on the values and behavior
of the parameters involved.
It is necessary, however, to look at the situation in a collective manner if
any insight on the genericity and stability aspects connected to the naked
singularities forming in the gravitational collapse are to be gained. There
may be different kinds of stabilities involved. For example, it can be asked
here if the conclusions will be stable to non-spherical perturbations, or when
will forms of matter more general than dust be considered, and so on. Such
issues are worth a detailed investigation, and will be crucial in the important
problem of collapse endstates.
At a somewhat different, but still quite interesting level, the stability of
these endstates with respect to the perturbations in the initial data space
that determines the final outcome of collapse can be investigated. As pointed
 4.6 Genericity and stability 201

out, this is a function space consisting of all possible mass functions F and
energy functions f . It is worth knowing how, for example, a naked singularity
endstate would be affected when one moves from a given density and energy
profile, which gave rise to this state, to a nearby density or energy profile in
this space of all initial data. The issue of how given density and energy distri-
butions determine the final collapse state has been discussed quite extensively
in the usual four-dimensional dust collapse models, although a somewhat
different methodology was used. These results were completed to give a full
and general treatment of the four-dimensional case, and the typical result
is that given any density profile, one could choose the energy profile (and
vice versa), so that the collapse endstate would be either a blackhole or a
naked singularity, depending on this choice.
As can be seen from the considerations here, these results are generalized to
the case of a higher-dimensional collapse situation, and the method allows a
more definite statement on the genericity of naked singularity formation to be
made. As seen from the discussion above and in Chapter 3, the quantity X (0)
is fully determined from the initial data functions and their first derivatives.
Once it is positive, the collapse ends in a naked singularity and a negative
value gives the blackhole final state. It follows by continuity that, given a
density profile, if the energy profile chosen is such that the collapse ends in a
naked singularity, that is if X (0) > 0, then there is a whole family of nearby
velocities such that this will continue to be the case, and then the naked
singularity forms an open subspace in the initial data space. The same of
course holds for blackhole formations, and both these are neatly separated
open regions in the initial data space. But if, on physical grounds, it is taken
that both ρ1 and b1 must vanish, then the dust treatment gives X (0) = 0,
and the cosmic censor may be restored.
It may be argued that if all the assumptions such as those discussed above
can be suitably motivated physically, then it may be possible to restore
cosmic censorship in a higher-dimensional spacetime for the gravitational
collapse of dust. These conditions will now be discussed in some detail. That
the equation of state must be dust-like in the final phases of the collapse is
a strong assumption, but it is not a possibility that can quite be ruled out,
as discussed earlier. After all, very little is known on the equations of state,
especially what it would be like in the advanced stages of the collapse. Also,
it is not ruled out that, in the late stages of the collapse, the configuration
is like a marginally bound one, especially in the vicinity of the singularity.
The introduction of pressures may, or may not, change such a scenario.
In this case then, it may be possible to recover the censorship if one moves
to a higher-dimensional spacetime arena. This is subject to the validity of
several extra physical inputs, as described above. On the other hand, once
more general situations of either a non-marginally bound case, or with a
more general form of matter, or without restrictive extra assumptions on
202 Cosmic censorship

the nature of the initial density profiles are moved to, then generically both
the blackhole and naked singularity phases could result as endstates of the
collapse in a higher-dimensional spacetime scenario as well. In this way, a
dynamical collapse in general relativity offers a rich spectrum of possibilities
to investigate.

4.6.3 Scalar field collapses

If it is accepted that naked singularities do occur for a wide range of collapse
models, the cosmic censorship requirement could then be interpreted as a
question, namely whether the forms of matter, such as the dust, perfect
fluid, or in-flowing radiation and others, must break down and cease to be
good approximations in the very late stages of the collapse. In fact, these may
not be regarded as fundamental forms of matter even at the classical level,
and are only approximations to the more basic entities such as a massless or
massive scalar field (in the eikonal approximation). Therefore, the question
whether naked singularities occur for a scalar field coupled to gravity, or for
similar matter fields other than dust, perfect fluids, or collapsing radiation
could be asked.
Much attention has been given in past years to analyzing the collapse of a
scalar field both analytically (see Christodoulou, 1986, 1994, 1999; Roberts,
1989; Traschen, 1994; Brady, 1995a, 1995b, and references therein), and
also numerically (Abraham and Evans, 1993; Choptuik, 1993; Evans and
Coleman, 1994; Gundlach, 1995, 1999). This is a model problem of a sin-
gle massless scalar field that is minimally coupled to a gravitational field,
and it provides possibly one of the simplest scenarios for investigating the
nonlinearity effects of general relativity. On the analytic side, the results of
Christodoulou show that when the scalar field is sufficiently weak, a regu-
lar solution, or global evolution for an arbitrary long time of the coupled
Einstein and scalar field equations exists. During the collapse, there is a
convergence towards the origin, and after a bounce the field disperses to
infinity. For strong enough fields, the collapse is expected to result in a black-
hole. For self-similar collapse, the results show that the collapse will result
in a naked singularity. However, the initial conditions that led to the for-
mation of a naked singularity are a set of measure zero here, and hence
the naked singularity formation may be a non-generic phenomenon in these
models.
Such an approach helps one to study the cosmic censorship problem as the
evolution problem in the sense of examining the global Cauchy development
of a self-gravitating system outside an event horizon. A dynamical version
of the cosmic censorship can be suggested that, given reasonable initial data
that is asymptotically flat, and assuming some suitable and reasonable energy
conditions, a global Cauchy evolution of the system outside the event horizon
 4.6 Genericity and stability 203

exists in the sense that the solution exists for arbitrary large times for an
asymptotic observer. For a discussion of such an approach in the context
of self-gravitating scalar fields, see Malec (1995). The problem of the global
existence of solutions is discussed by Malec, and an explicit example of an
initial configuration that results in a naked singularity is found at the center
of symmetry.
Scalar field collapse has also been numerically studied, as mentioned above.
A family of scalar field solutions was considered where a parameter p char-
acterized the strength of the field. The numerical calculations showed that,
for blackhole formation, there is a critical limit p → p∗ and the mass of the
resulting blackholes satisfy a power law Mbh ∝ ( p − p∗ )γ , where the critical
exponent γ has a value of about 0.37. It was then conjectured that such
a critical behavior may be a general property of a gravitational collapse,
because similar behavior was found in some other cases, including imploding
axisymmetric gravitational waves. Also, the case of the collapse of radiation
with an equation of state p = ρ/3 was considered, assuming self-similarity
for the solutions. It is still not clear if the critical parameter γ will have the
same value for all forms of matter chosen, and further investigations may be
required to determine this issue. As the parameter p moves from the weak
to the strong range, very small mass blackholes can form. This has relevance
to the censorship because, in such a case, one can probe and receive mes-
sages from arbitrarily near to the singularity, and this is naked singularity
like behavior. Attempts have also been made to construct models analyti-
cally that may reproduce such a critical behavior assuming self-similarity,
and solutions were constructed that have dispersal, together with solutions
with blackholes or naked singularities.

4.6.4 Families of non-spacelike curves from a singularity

The purpose here has been to discuss the genericity and stability properties
related to cosmic censorship and naked singularities. From such a perspec-
tive, it is relevant to ask about the nature of trajectories coming out from
the singularity when it is visible. As such, in Chapter 3, a wide variety of
situations in which a naked singularity forms in a gravitational collapse was
discussed. However, the existence of only a radial null geodesic coming out
from the singularity towards showing its visibility was demonstrated. If only
a single null geodesic emerged from the singularity, it may not be called
generic enough, as in the case of only a single photon escaping, it will be
non-visible to an external observer for all practical purposes. On the other
hand, the existence of families of future directed non-spacelike paths could
make the singularity visible to outside observers. Also, any material particles
would escape from the vicinity of the singularity only if there are timelike
curves escaping from these ultra-dense regions.
204 Cosmic censorship

The causal structure of spacetime near a singularity and the nature of tra-
jectories emerging from it are analyzed below in some detail. If null and other
timelike paths also emerge from the singularity, then in principle, particle
and energy emission from such ultra-dense regions is allowed. Such emissions
are basically governed by the nature of non-spacelike paths near the singu-
larity. These trajectories are examined and it is shown that if a null geodesic
emerges, then families of future directed non-spacelike curves that also neces-
sarily escape from the naked singularity exist. The existence of such families
is crucial to the physical visibility of the ultra-dense regions.
Here, no underlying symmetries are assumed for the spacetime, and some
earlier considerations on the nature of causal trajectories emerging from a
naked singularity are generalized and clarified. Singularities are the regions
where the physical conditions such as densities and curvatures are at their
extreme. While the big-bang singularity of cosmology is visible in principle,
and gave rise to the universe as a whole, it cannot actually be seen. On
the other hand, when a massive star dies and collapses continually under
gravity, the eventual spacetime singularity can be either hidden within an
event horizon of a blackhole, or it could be visible to outside observers,
depending on how the collapse of the cloud evolves. While a naked singularity
forming in the collapse could provide an opportunity for the physical effects
taking place in these extreme regions to be observable to outside observers
in the universe, the actual visibility of such extreme gravity regions will
depend on the nature and structure of non-spacelike paths emerging from
the singularity.
If a continual collapse leads to a naked singularity formation then, even if
quantum gravity resolves it eventually, the point is that the causal structure
of spacetime in the vicinity of the ultra-dense regions allows them to be seen
by an external observer. Therefore, the quantum effects taking place in the
regions with arbitrarily high matter densities and curvatures can be seen by
the external observers. Any physical effects emerging will be again governed
by the existence of families of non-spacelike paths from the vicinity of the
singularity. It is therefore important to understand the structure of such
families within a gravitational collapse framework.
The important physical issue then is whether such a naked singularity
forming in the gravitational collapse could radiate away energy and par-
ticles. This depends crucially on the existence and structure of families of
non-spacelike trajectories emerging from its vicinity. Also, the actual physical
appearance and size of the singularity will be determined by the non-radial
null trajectories, and the energy emission, if any, will be governed by the
timelike curves and other non-spacelike trajectories escaping from the sin-
gularity. For this reason, several authors have considered the possibility of
non-radial null geodesics emerging from a naked singularity in the context
of spherically symmetric dust collapse models (Deshingkar and Joshi, 2001;
 4.6 Genericity and stability 205

Mena and Nolan, 2001, 2002; Deshingkar, Joshi, and Dwivedi, 2002). Also,
families of non-spacelike and timelike geodesics have been worked out in
self-similar perfect fluid collapses (Joshi and Dwivedi, 1992, 1993b), and for
Vaidya radiation collapse models (Dwivedi and Joshi, 1989, 1991). Most of
these considerations have been in the framework of spherically symmetric
spacetimes, at times together with other symmetry conditions, such as self-
similarity of the models imposed, and within the framework of a specific
matter model.
A general consideration of the nature of non-spacelike trajectories near a
naked singularity will be of much interest from such a perspective. Here, the
non-spacelike trajectories from a naked singularity in general are examined,
and it is shown that if a radial null geodesic emerges, then large families of
non-spacelike curves also necessarily emerge from the singularity. It is thus
seen that the existence of a radial null geodesic is sufficient to ensure the
existence of families of timelike and non-spacelike trajectories escaping, and
in this sense a single photon escaping in a radial direction from the singu-
larity is never an isolated phenomenon. This generalizes and clarifies earlier
considerations in this direction, without assuming any symmetry conditions
on the underlying spacetime or assuming a specific matter model for the
collapse.
When the collapse ends in a naked singularity, the causal structure near the
singularity is such that a null geodesic trajectory γ emerges from it, as shown
earlier. Specifically, γ is future directed, which in the past terminates at the
singularity, and is therefore a past incomplete null geodesic. To examine in
general the possible existence and nature of non-spacelike curves emerging
from this naked singularity, the causal boundary construction developed by
Geroch, Kronheimer, and Penrose (1972), where the spacetime M is taken
to satisfy a suitable causality condition such as strong causality, which rules
out the existence of closed timelike curves, is used here. In this procedure,
a boundary is attached to the regular spacetime manifold, which includes
spacetime singularities as well as the points at infinity.
Note that a boundary attachment to the spacetime manifold is essential to
treat the regular spacetime events, together with its singularities and points
at infinity in a unified manner. There are different ways to attach a boundary
to the spacetime, and they do not necessarily all give the same result. Here
the rather basic approach, given by Geroch, Kronheimer, and Penrose (1972),
is used as it depends essentially only on the causal structure of the spacetime,
which is much more fundamental than, for example, the differential structure
of the spacetime manifold. Also, from a physical point of view, each ideal
point here is directly associated with the region of spacetime that it can
influence, or that it would be influenced by.
An open set W in the spacetime is called a future set if it contains its
own future, that is, I + (W ) ⊂ W . Furthermore, a future set W is called an
206 Cosmic censorship

indecomposable future set (IF) if it cannot be expressed as the union of two

proper subsets that are themselves future sets. Indecomposable past sets
(IPs) are similarly defined. The idea of the causal boundary construction
is to divide the collection of IFs and IPs into two classes, namely the one
representing regular points of the spacetime, and the other class giving all
its boundary points or the ideal points, which include spacetime singularities
as well as points at infinity. The collection of IFs (or IPs) can be divided
into the two parts as follows. For a set W that is an IF, if an event in the
spacetime p ∈ M exists such that W = I + ( p), then W is called a proper
IF or a PIF. All other IFs are called terminal IFs or TIFs, and represent
spacetime singularities and the points at infinity.
Consider now the set I + (γ ) where γ is any null geodesic curve emerging
from the singularity. It is then a future set, because for any p ∈ I + (γ ),
I + ( p) ⊂ I + (γ ). It can be seen that I + (γ ) is an indecomposable future
set, or an IF. To show this, a somewhat modified version of the proof of
Theorem 2.1 of Geroch, Kronheimer, and Penrose (1972) is used. Suppose
I + (γ ) = A ∪ B with A and B both being future sets. If neither A is fully
contained in B nor vice versa, then two events x and y can be found such
that x ∈ A − B and y ∈ B − A. As x, y ∈ I + (γ ), there are points x , y  ∈ γ
such that x ∈ I + (x ) and y ∈ I + (y  ). But, x and y  are causally related, so
suppose now that x is in the past of y  on γ . Then there is a null geodesic
from x to y  , and there is a timelike curve from y  to y , as above. This
implies that there must be a timelike curve from x to y (see for example,
Hawking and Ellis, 1973, p. 183). It follows that y ∈ I + (x ). As x ∈ I + (x ),
this implies that x, y ∈ I + (x ). Therefore, x lies in the intersection of the
sets I − (x) and I − (y ), which is an open set and so contains a neighborhood
N of x . Let z be an event in I + (x ) ∩ N , then z ∈ I + (γ ) and so has to
be in one of the future sets A or B . Suppose it is in A, then since there
are future directed timelike curves from z to both x and y , it follows that
both x, y ∈ A, which is a contradiction. Hence, it follows that I + (γ ) has
to be an IF. Since γ is a past incomplete null geodesic, there is no regular
point p ∈ M such that I + ( p) = I + (γ ), and hence it follows that I + (γ ) is
necessarily a TIF.
The TIF set I + (γ ) here represents a boundary point of the spacetime
that is the naked singularity. While the naked singularity formation as the
endstate of a continual gravitational collapse has been investigated exten-
sively in the last decade or so (especially within the framework of spherically
symmetric collapses and for certain non-spherical examples), the main tech-
nique there has been to show that a radial null geodesic coming out in the
future and terminating in the past at the spacetime singularity exists (see
for example, Joshi and Dwivedi, 1993a; Joshi and Goswami, 2004). These
results basically show that the gravitational collapse from regular initial mat-
ter profiles could result in either the blackhole or naked singularity endstates,
 4.6 Genericity and stability 207

depending on the nature of the initial data from which it evolves and the
dynamical evolutions of the collapsing cloud, as allowed by the Einstein
equations.
However, as remarked above, if the visibility and other related physical
characteristics of a naked singularity that formed in the gravitational collapse
are to be explored, then it is important to examine and understand the
structure of the families of non-spacelike curves from the singularity. Again,
if non-spacelike curves emerge from the singularity, but do not go out of the
boundary of the collapsing cloud, then the singularity will be only locally
visible, and outside observers would not be able to see it. It is necessary
therefore to understand the structure of non-spacelike curves from a naked
singularity in general.
It is now possible to do this. Since the set S = I + (γ ) is a TIF, it fol-
lows from Geroch, Kronheimer, and Penrose (1972) that in this case, a past
inextendible timelike curve λ must exist, such that S = I + (λ). In the case
of the collapse ending in a naked singularity and a radial null geodesic γ
escaping from it, it can thus be seen that the set I + (γ ) is a TIF, and so, by
the above result, there is a timelike curve λ generating this TIF, in the sense
that S = I + (λ). Since both the non-spacelike trajectories γ and λ represent
the same ideal or boundary point of the spacetime that is the naked singular-
ity, and since I + (γ ) = I + (λ) by definition, it follows that the future directed
timelike curve λ must terminate in the past at the naked singularity. In other
words, it has been shown that a timelike curve λ, which escapes away to the
future, and which terminates in the past at the naked singularity exists.
It follows that if p ∈ λ and q is any other event such that q ∈ I + ( p),
then there are timelike curves from the naked singularity to q . This proves
the existence of families of infinitely many future directed non-spacelike tra-
jectories escaping away from the naked singularity. In general, if λ is any
other future directed non-spacelike curve such that I + (λ ) = I + (λ), then it
follows that they all represent the same TIF, which is the naked singularity,
and that λ terminates in the past at this singularity.
Therefore, it can be seen that there is an infinity of future going non-
spacelike curves that emerge from the singularity, if a single null geodesic has
emerged. These include timelike curves as well as non-radial non-spacelike
geodesics. It is seen that the usual method employed to show the existence
of a naked singularity in the collapse, which establishes the existence of a
radial null geodesic escaping away, is sufficient to lead to the existence of
infinite families of future going non-spacelike curves from the naked singu-
larity, as shown here. In the present consideration, it is no longer required to
have any special symmetry assumptions on the spacetime, such as spherical
symmetry, self-similarity, or others, or any specific form of matter model
such as the dust equation of state, which are usually assumed in such
discussions.
208 Cosmic censorship

In particular, this also clarifies and generalizes the earlier results on dust
collapses and other models mentioned above, which have focused on non-
spacelike null geodesics. The null geodesics of the spacetime have, of course,
a special role to play as far as the visibility of the singularity is concerned.
From such a perspective, the existence of radial versus the non-radial families
of null geodesics from a naked singularity will be briefly discussed. Suppose
a radial null geodesic emerges from the naked singularity S developing in a
continual collapse. In this case, as seen above, a timelike curve λ generating
the TIF set I + (λ) that represents the boundary point S exists. All other
future directed non-spacelike curves γ that satisfy I + (λ) = I + (γ ) generate
the same TIF representing the boundary point S , and they give the families
of particle or photon trajectories escaping away from the naked singularity.
The boundary of this future set, which is a TIF, is a three-dimensional null
hypersurface that is ruled by the radial, as well as non-radial, null geodesics
generators γ , which are all incomplete when extended in the past, and which
all have the property that I + (λ) = I + (γ ). This shows that the existence
of a radial null geodesic is sufficient also to give families of non-radial null
geodesics emerging from the singularity. This generalizes the earlier results
on the existence of non-radial null geodesics from the singularity for the
spherically symmetric dust collapse, when a radial null geodesic emerges
from the naked singularity.
It is seen that once a singularity is naked, it gives rise to infinitely many null
as well as timelike curves to escape away from. In this sense, the emission of
paths representing particle or photon trajectories from the naked singularity
is a generic phenomena. This is essential and is a necessary condition for
the naked singularity to give rise to any physical effects that may possibly
be observed by external observers. In the present consideration, the global
visibility of the singularity, that is, the situation in which once the families
of non-spacelike curves emerge from the naked singularity when they will
actually cross the boundary of the cloud to escape to an outside observer,
has not been discussed. It is known, however, in several cases including
spherical dust collapses, that whenever a singularity is locally naked, then
the rest of the free functions in the model can be chosen so as to make it
globally visible.
It is known, for example, in the case of dust collapses, that once the
singularity is locally naked, the choice of a suitable behavior of the mass
function (which is a free function, subject only to some physical conditions
such as an energy condition and regularity of the initial data) away from the
center, allows the null rays to emerge from the boundary of the cloud. It
may also be noted that in various classes of self-similar collapses, once the
singularity is locally naked it becomes necessarily globally visible (Joshi and
Dwivedi, 1992, 1993b). In any case, as there is no scale in the problem, once
the singularity is locally visible, an observer within a large enough blackhole
 4.6 Genericity and stability 209

will still be able to see it for a long enough time. In such a scenario, the
escape of rays outside the boundary of the cloud would not be crucial. A
discussion on global visibility in a more general context of perfect fluids has
recently been given by Giambo (2006).
 5
Final fate of a massive star

The considerations on gravitational collapse so far have been with the moti-
vation to address the physical questions such as the role of collapse in
astrophysics and cosmology. Many of the cosmic processes, such as the
birth of stars, the formation of galaxies, and others, are not well under-
stood today, but it is clear that gravitational collapse will play a major role
there. Hence, understanding the dynamics of the collapse is important, as
has been attempted here in various cases.
The important question of the final fate of massive stars at the end of
their life cycle, when they have used all their nuclear fuel, and when grav-
ity becomes the sole and key governing force, has drawn much attention
for many decades. The importance of this issue was highlighted by Chan-
drasekhar (1934), who pointed out that the life history of a star of small
mass must be essentially different from that of a star of large mass, and that
while a small mass star can pass into a white dwarf stage, a star of large mass
cannot go to this state, and one is left speculating on other possibilities. The
question as to what happens when a massive star, heavier than a few solar
masses, collapses under its own gravity has been a fundamental key problem
in astronomy and astrophysics. If the star is sufficiently massive, beyond
the white dwarf or neutron star mass limits, then a continued gravitational
collapse must ensue without achieving any equilibrium state, when the star
has exhausted its nuclear fuel. To understand the possible endstates of such
a continual gravitational collapse, the dynamical collapse scenarios must be
studied within the framework of a gravitation theory such as Einstein’s the-
ory. The theory of singularities discussed earlier then implies that, under
rather general physical conditions, a spacetime singularity must develop.
Considerations here show that, according to Einstein’s theory of gravity,
such a star in a continual collapse can end up either in a blackhole final state,
from which no communications in the form of light or particles would come
out, or it can go to a naked singularity, where the collapse proceeds but the

210
 Final fate of a massive star 211

trapped surfaces do not form early enough to cover the singularity. In such
a case, the extreme gravity regions can communicate and send out physical
effects to the external universe, and it is also suggested that a huge amount of
energy could possibly be released, in principle, during the final stages of the
collapse from the regions close to the classical singularity (Joshi, Dadhich,
and Maartens, 2000). According to these suggestions, an enormous amount of
energy may probably be generated, either by some kind of a quantum gravity
mechanism, or by means of an astrophysical process where the region could
simply turn into a fireball-like situation, creating shocks into the surround-
ing medium. One way to study the structure of these extreme regions is to
examine the complete spectrum of non-spacelike geodesics through which
this energy could escape. Even if a fraction of the energy so generated is able
to escape to a distant observer, it cannot be ruled out that an observational
signature may be generated. It therefore becomes important to look into
these possibilities in some detail, and to consider likely observable signatures
of each of the blackhole and naked singularity final states of the collapse,
from the perspective of a faraway observer.
The theoretical properties and possible observational signatures of a
blackhole and naked singularity would be significantly different from each
other, and this could be of potential interest from the perspective of astro-
physical observations. An immediate distinction is, for example, in the case
where the collapse ends in a blackhole, an event horizon develops well before
the occurrence of the singularity, and thus the regions of extreme physical
conditions are always hidden from the outside world. But if the collapse devel-
oped into a globally naked singularity, then the energy of the region neigh-
bouring the singularity can escape via the available non-spacelike geodesics
paths or via other non-geodetic, non-spacelike trajectories to a distant
observer.
In this chapter, various aspects on gravitational collapse that have emerged
from the analysis so far are considered. An important issue is the final fate
of a non-spherical collapse. In earlier treatment, it was shown that strong
curvature naked singularities arise in a variety of situations involving dust,
perfect fluids, and other forms of matter. There are many interesting ques-
tions that are under active investigation at the moment. For example, could
naked singularities generate bursts of gravity waves? What kind of quantum
effects will take place near a naked singularity? What will be the generic
outcome for the case of a non-spherical collapse? Many of these issues could
have interesting physical implications. The possibility that the ultra-high
energy astrophysical phenomena, such as the gamma ray bursts, may have a
strong connection to the physics and dynamics of the gravitational collapse
of massive stars cannot be ruled out. In fact, many of the current gamma
ray burst models involve a collapsar, emphasizing the role of the massive
star collapse. Another intriguing possibility is that a naked singularity may
212 Final fate of a massive star

possibly provide some kind of observational signatures for the quantum grav-
ity effects taking place in ultra-strong gravity regions. This would then be
an exciting prospect in view of the current lack of knowledge on quantum
gravity.
In Section 5.1, the life cycle of massive stars is discussed. The large mass
stars, several times heavier than the Sun, follow a characteristically different
life cycle from stars of about one solar mass. Such massive stars typically
live a much shorter life than small mass stars, and undergo a catastrophic
gravitational collapse at the end of their life cycle. In Section 5.2, how a
physically realistic collapse must evolve and the different dynamical forces
at work to govern the final fate of such a collapse are discussed. To give an
explicit example, it is shown how the spacetime shearing forces, caused by the
inhomogeneities in the matter distribution of the star, delay the formation of
trapped surfaces during the evolution of the collapse. This allows the ultra-
strong regions of gravity and the spacetime singularity to become visible
to faraway observers, and it is seen how a naked singularity forms rather
naturally as a collapse endstate.
For any realistic consideration of gravitational collapse, departures from
spherical symmetry have to be taken into account. While this problem is
much more complex and not much work is still carried out on this, the Szek-
eres collapse models are discussed in Section 5.3. These are not spherically
symmetric, and have no Killing vectors. It is shown that these again admit
both blackhole and naked singularity final states. While much analytic and
numerical work remains to be carried out on non-spherical collapse, this
model explicitly illustrates and shows that the naked singularity final state
for the collapse is not necessarily limited to spherical symmetry, and that
these can result in a non-spherical collapse as well.
If both blackhole and naked singularity final states do occur in physi-
cally realistic gravitational collapse scenarios, the main issue would then be
the outcome that nature may prefer for the final state of a collapsing star.
While, mathematically, many evolutions may be possible, in physically real-
istic situations only certain evolutions may be actually realized. From such a
perspective, in Section 5.4 the paradoxes that are associated with the black-
hole formation as the collapse final state, when a massive star undergoes
gravitational collapse at the end of its life cycle, are discussed. Finally, in
Section 5.5 it is discussed how the quantum effects, which will be prominent
in the late stages of the collapse, can play a crucial role in creating a massive
emission of mass and energy in the very late stages of the collapse. It could
then be asked if a ‘quantum star’ comes into being in the very final stages of
the collapse, which allows for a powerful emission of mass and energy. Such
a burst-like emission can have observational features on the one hand and,
on the other, it would help resolve the naked singularity formation, thus
allowing the classical singularity to be removed.
 5.1 Life cycle of massive stars 213

5.1 Life cycle of massive stars

Here, the life of a star that comes into existence by gravitational contraction
and collapse within a dense interstellar gas cloud is outlined. This helps in
understanding the final fate of a massive star that has exhausted its nuclear
fuel that provided the internal pressure against the inwards pull of gravity.
As the dense interstellar gas compresses, possibly due to a variety of rea-
sons, in the process the central temperatures of the material rises to ignite
a nuclear fuel burning cycle. The major bulk of the cloud is made up of
hydrogen, and this now starts burning into helium. This produces energy
and internal pressure to counter the internal gravity pull, and the gravita-
tional contraction is halted. The star enters a quasi-static period when it
supports itself against gravity by means of thermal and radiation pressures.
Such a phase may continue for billions of years, depending on the original
mass of the star. If M < M (where M denotes the mass of the Sun,
∼ 2 × 1030 kg), this period would typically be longer than 1010 years. How-
ever, such a lifetime decreases and becomes much shorter if M > 10M , and
it has to be less than 2 × 107 years. That is, much more massive stars burn
out their nuclear fuel much faster. (For a review, see for example, Blandford
and Thorne, 1979; Blandford, 1987, and references therein).
The star spends this larger portion of its life on the main sequence. This
evolution of the star is actually a balance between the nuclear burning in its
interior and gravitational collapse, and, once much of the hydrogen in the
core is exhausted providing no further pressure against the inwards pull of
gravity, then the collapse must continue further if the star is still sufficiently
massive. In such a process, the core temperatures rise again to initiate ther-
monuclear reactions converting, in the second phase, the helium that formed
earlier into carbon, and the core stabilizes again. For a heavy enough star,
this process repeats itself and finally a large core of stable nuclei, such as
iron and nickel, is built up.
The outcome is that the final state for such an evolution is either an
equilibrium star, or a state of continual gravitational collapse if there are no
internal forces available that are strong enough to build up a pressure within
to resist the pull of gravity of the massive star. The key factor towards
deciding the possibility of such a stability is the equation of state for the
cool matter of the star in its ground state, that is when all possible nuclear
reactions have taken place within the star and no further energy can be
derived from such a burning. The support against the pull of gravity in such
a case must then come either from electron degeneracy pressure, or from
neutron degeneracy pressure. If the electron degeneracy pressure can balance
the gravity, a white dwarf comes into being. If somewhat higher forces are
required, these can be still provided by the neutron degeneracy pressure, and
a neutron star forms.
214 Final fate of a massive star

The equation of state for an ideal electron Fermi gas was approximated
by Chandrasekhar (1931, 1934), who showed that there is a maximum mass
limit for the mass of a spherical, non-rotating star to achieve a white dwarf
stable state. This is given by
 2
2
Mc ∼ 1.4 M , (5.1)
µe

where µe is the constant mean molecular weight per electron. Subsequently,

considerable work has been carried out on equations of state for matter
at nuclear densities (see for example, Arnett and Bowers, 1977) and it is
seen that the maximum mass for non-rotating white dwarfs lies in the range
1.0M –1.5M , depending on the composition of matter. Similar consid-
erations can be made for neutron stars, and this gives the mass range as
1.3M –2.7M .
The radius of a typical white dwarf is ∼104 km and it has a central den-
sity of ∼103 kg cm−3 . For a neutron star, these numbers are ∼10 km and
∼1012 kg cm−3 respectively. The maximum mass upper limits stated above
are raised somewhat when either the rotation or differential rotation of the
star is taken into account. However, under very general circumstances, a
firm upper limit of about 5M is obtained, beyond which the degeneracy
pressures cannot support the star in any case (Harrison et al., 1965; Hartle,
1978). By now very many examples of white dwarfs are known to exist in the
universe. The discovery of pulsars provided strong support for the existence
of neutron stars that must be rotating with periods of fractions of a second
in order to produce the observed pulsar signals.
It follows that if a star has a mass higher than, or about, 5M , it must
then enter a state of perpetual gravitational collapse and contraction once
it has exhausted all its nuclear fuel, and no equilibrium configurations such
as those given above are possible. Of course, there is a possible escape from
the continual collapse if the star manages to throw away most of its mass by
some process during this evolution, and settles again below the neutron star
or the white dwarf limits. In fact, mass ejection is observed in a supernova
explosion for the star. During the process of the collapse, the iron core of the
star would convert to the formation of a neutron core and a neutron star is
born, at least momentarily. Then, the neutron degeneracy pressure develops,
which provides a balancing force against further collapse. Then, the core
collapse is halted or at least slowed down at nuclear densities, and a shock
wave is generated, which propagates outwards in the envelope of the star. In
this case, while the inner core remains a neutron star, the outer parts are
then driven away by the shock, thus releasing enormous mass and energy,
which is believed to be a supernova explosion. However, the theory for such
ejection of matter is not still well-understood.
 5.2 Evolution of a physically realistic collapse 215

In any case, it does not seem likely that all such massive stars would be
able to throw away almost all, or a very major part, of their mass in such
a process. That is because, for stars having tens of solar masses, this would
amount to throwing away almost ninety percent of the mass of the star. In
a supernova, typically only the outer layers of the star are blown off, and
no suitable mechanism that could achieve such a high degree of efficiency
for the mass ejection from the star is envisaged as yet. In the case of a
massive star, in the course of its normal life of nuclear burning, not a very
large portion of its mass can be removed through radiation. Then, once the
gravitational collapse initiates, which is a catastrophic process, a star that
has lived millions of years collapses gravitationally in a matter of seconds.
Now, during such a catastrophic collapse, if the shock that is produced could
not blow off almost all the outer layers, these layers would fall on the newborn
neutron star that has momentarily formed at the core in the process of the
collapse, and the further collapse continues again as the pull of gravity would
exceed the balancing neutron degeneracy pressures.
It is thus seen that the evolution of a massive star, when it exhausts its
nuclear fuel, causes an inevitable continual collapse at the end point of a
stellar evolution. The general relativity theory then implies that a spacetime
singularity must form of necessity in such a scenario. The basic ingredients
of such spacetime singularities were discussed earlier. The cosmic censorship
conjecture then asserts that any such singularities forming in gravitational
collapse must be covered necessarily within an event horizon of gravity, invis-
ible to any external observer, and that in general the final state of such an
evolution must then be a blackhole. This requires an appropriate formulation
of the concepts of a blackhole as well as the cosmic censorship, within the
framework of an appropriate spacetime geometry. To decide on this issue,
the dynamical gravitational collapse has to be analyzed to see how the struc-
ture of the horizons evolve. It is within such a perspective that the analysis
here on gravitational collapse has taken place, and it can be seen that the
dynamical gravitational collapse processes are entirely fundamental in this
way to the basics and applications of blackhole physics.

5.2 Evolution of a physically realistic collapse

It was shown earlier that a continued collapse of a massive matter cloud
could terminate either in a blackhole or a naked singularity, depending on
the matter initial data and the allowed evolutions as decided by the Einstein
equations. These results were obtained under physical conditions such as
the validity of the energy condition, regularity of the initial data, and such
others. However, the actual outcome of a physically realistic collapse still
remains to be investigated. While, mathematically, many models may be
216 Final fate of a massive star

permitted by general relativity, nature might select only a few of these for
the actual evolution of a massive star. Therefore, further examination of the
physical forces operating within a star as the collapse proceeds is required.
This issue is discussed here in the context of a spherical collapse by means
of an explicit example. This also throws some light on why naked singularities
form at all in a gravitational collapse. The key physical features that possibly
cause the development of a naked singularity, rather than a blackhole, as the
endstate of a gravitational collapse are investigated, and it is seen that suffi-
ciently strong shearing effects near the singularity can delay the formation of
the apparent horizon. This exposes the singularity to an external observer,
in contrast to a blackhole, where it is hidden behind an event horizon due to
the early formation of the trapped surfaces. The final outcome of a gravita-
tional collapse in general relativity is an issue of much importance from the
perspective of blackhole physics and its astrophysical implications, and one
needs to understand the key physical characteristics and dynamical features
in the collapse that give rise to a naked singularity, rather than a blackhole.
Here, the treatment of Joshi, Dadhich, and Maartens (2002) is followed.
It is seen that it is the inhomogeneity and related shearing effects within
the cloud that, if sufficiently strong near the central worldline of the col-
lapsing cloud, would delay the formation of the apparent horizon so that
the singularity becomes visible, and communication from the extreme strong
gravity regions to outside observers becomes possible. When the inhomo-
geneity and related shear forces are weak, or for the extreme case of no shear
in a fully homogeneous collapse, the collapse necessarily ends in a blackhole,
because an early formation of the apparent horizon leads to the singularity
being hidden inside an event horizon.
For the spherical gravitational collapse of a massive matter cloud, the
interior metric in comoving coordinates is

ds2 = −e2ν(t,r) dt2 + e2ψ(t, r) dr2 + R2 (t, r)dΩ2 . (5.2)

The matter shear is

 
Ṙ 
σab = e−ν − ψ̇ 13 hab − na nb , (5.3)
R

where hab = gab + ua ub is the induced metric on the three-surfaces orthogonal

to the fluid four-velocities ua , and na is a unit radial vector. The initial data
for collapse are the values on t = ti of the three metric functions, the density,
the pressures, and the mass function that arises from integrating the Einstein
equations, as discussed earlier,

F (ti , r) = ρ(ti , r)r2 dr, (5.4)
 5.2 Evolution of a physically realistic collapse 217

where 4πF (ti , rb ) = M , which is the total mass of the collapsing cloud, and
where r > rb is a Schwarzschild spacetime. As earlier, the rescaling freedom
in r is used to set
R(ti , r) = r , (5.5)
so the physical area radius R increases monotonically in r, and with Ri = 1,
there are no shell-crossings on the initial surface from which the collapse
develops. Of interest here is the central shell-focusing singularity at R =
0, r = 0, which is a gravitationally strong singularity, as opposed to the
shell-crossing singularities that are weak, and through which the spacetime
may sometimes be extended.
The evolution of the density and radial pressure are given by

F Ḟ
ρ= , pr = . (5.6)
R2 R R2 Ṙ
The central singularity at r = 0, where density and curvature are infinite,
is naked if there are outgoing non-spacelike geodesics that reach outside
observers in the future and terminate at the singularity in the past. Outgoing
radial null geodesics of the metric (5.2) are given by

dt
= eψ−ν . (5.7)
dr
First consider the case of a homogeneous density collapse, ρ = ρ(t). Writ-
ing f = e−2ψ R2 − 1, the Einstein equations give f − e−2ν Ṙ2 = −F/R.
Then, as discussed in Chapter 3, the geodesic equations can be written as
 ! 
dR f + F/R R
= 1− , (5.8)
du 1+f αrα−1

where u = rα (α > 1). If there are outgoing radial null geodesics terminating
in the past at the singularity with a definite tangent, then at the singularity,
dR/du > 0. Now in the case of homogeneous density, the entire mass of the
cloud collapses to the singularity simultaneously at the event (t = ts , r = 0),
so that F/R → ∞. Then in that case, from (5.8), dR/du → −∞, so that
no radial null geodesics can emerge from the central singularity. It can be
shown that all the later epochs t > ts are similarly covered.
It has thus been shown that for a spherical gravitational collapse with
homogeneous density, the final outcome is necessarily a blackhole. The
pressures, however, can be arbitrary on which no conditions are imposed.
This conclusion does not require homogeneity of the pressures pr and
pθ , and is independent of their behavior. This generalizes the well-known
Oppenheimer–Snyder–Datt result of the special case of dust, where the
218 Final fate of a massive star

homogeneous cloud always collapses to form a blackhole. An immediate con-

sequence is that if the final outcome of a spherical gravitational collapse is
not a blackhole, then the density must be inhomogeneous necessarily. In any
physically realistic scenario, the density will typically be higher at the center,
so that generically the collapse is inhomogeneous. An inhomogeneous density
profile is thus a necessary condition for a naked singularity to develop as the
collapse endstate.
To understand more clearly the role that the inhomogeneities play in
delaying the trapped surfaces during the collapse, consider now a collapsing
inhomogeneous dust cloud (p = 0), with density that is higher at the center.
The metric is the TBL geometry, given by (5.2) with ν = 0, e2ψ = R2 /(1+f ),
and
F (r)
Ṙ2 = f (r) + . (5.9)
R
These models are fully characterized by the initial data, specified on an
initial surface t = ti from which the collapse develops, which consist of two
free functions that are the initial density ρi (r) = ρ(ti , r) (or equivalently,
the mass function F (r)), and f (r), which describes the initial velocities of
the collapsing matter shells. At the onset of the collapse, the spacetime is
singularity-free, so that from (5.6),

F (r) = r3 F̄ (r) , 0 < F̄ (0) < ∞. (5.10)

The initial density ρi (r) is

ρi (r) = r−2 F  (r). (5.11)

The shell-focusing singularity appears along the curve t = ts (r), and is

defined by
R(ts (r), r) = 0. (5.12)

As the density grows without bound, trapped surfaces develop within the col-
lapsing cloud. These can be traced explicitly via the outgoing null geodesics,
and the equation of the apparent horizon, t = tah (r), which marks the bound-
ary of the trapped region and is given by (see also the discussion on apparent
horizons in Section 3.6)

R(tah (r), r) = F (r). (5.13)

If the apparent horizon starts developing earlier than the epoch of the sin-
gularity formation, then the event horizon can fully cover the strong gravity
regions including the final singularity, which will thus be hidden within a
 5.2 Evolution of a physically realistic collapse 219

blackhole. On the other hand, if trapped surfaces form sufficiently later dur-
ing the evolution of the collapse, then it is possible for the singularity to
communicate with outside observers.
For the sake of clarity, consider a marginally bound collapse, f = 0,
although the conclusions can be generalized to hold for the general case.
Then, (5.9) can be integrated to give

3
R3/2 (t, r) = r3/2 − (t − ti )F 1/2 (r), (5.14)
2
and (5.12) and (5.13) lead to
 3 1/2
2 r
ts (r) = ti + , (5.15)
3 F (r)
2
tah (r) = ts (r) − F (r). (5.16)
3
The central singularity at r = 0 appears at the time

2
t0 = ts (0) = ti + √ , (5.17)
3ρc

where ρc = ρi (0). Unlike the homogeneous dust case of Oppenheimer and

Snyder, the collapse is not simultaneous in comoving coordinates, and the
singularity is described by a curve, the first point being (t = t0 , r = 0).
For inhomogeneous dust, (5.3) and (5.14) give

1 r
(3F − rF  ) .
2
σ 2 ≡ σab σ ab =  2 (5.18)
2 4
6R R F
Here, a generic inhomogeneous mass profile can be chosen to have the form

F (r) = F0 r3 + F1 r4 + F2 r5 + · · · (5.19)

near r = 0, where F0 = ρc /3. The homogeneous Oppenheimer–Snyder dust

collapse has Fn = 0 for all n > 0, and (5.18) then implies σ = 0. The converse
is also true in this case, namely, if a vanishing shear σ = 0 is imposed, then
Fn = 0 necessarily. Whenever there is a negative density gradient, that is,
when there is a higher density at the center, then Fn = 0 for some n > 0, and
it follows from (5.18) that the shear is then necessarily non-zero. Here, note
that if the density profile is required to be analytic, all the odd terms F2n−1
in the mass function can be set to be zero. This, however, is not required by
the analysis here, which is independent of any assumptions on Fn .
It is therefore important to evaluate the effect of such a shear on the
development and time evolution of the trapped surfaces. In other words,
220 Final fate of a massive star

there is a need to determine the behavior of the apparent horizon in the

vicinity of the central singularity at R = 0, r = 0. For this purpose, let the
first non-vanishing derivative of the density at r = 0 be the nth one (n > 0),
that is,
F (r) = F0 r3 + Fn rn+3 + · · · , Fn < 0 , (5.20)
near the center. By (5.18) and (5.16),
 
n 2 Fn 2 9 
1 − 3F0 (t − ti ) + F0 (t − ti ) r2n + O r2n+1 ,
2 1/2 2
σ (t, r) =
6F0 4
(5.21)
2 Fn n 
tah (r) = t0 − F0 r3 − 3/2
r + O rn+1 . (5.22)
3 3F 0

The time dependent factor in square brackets on the right of (5.21) decreases
monotonically from 1 at t = ti to 0 at t = t0 . Therefore, the qualitative role
of the shear in singularity formation can be seen by looking at the initial
shear. The initial shear σi = σ (ti , r) on the surface t = ti grows as rn ,
n ≥ 1, near r = 0. A dimensionless and covariant measure of the shear is
the relative shear, |σ/Θ|, where

Ṙ Ṙ
Θ=2 + (5.23)
R R
is the volume expansion. It follows that

σ
−nFn n

= √ r [1 + O(r)]. (5.24)
Θ i 3 6F 0
It can now be seen how such an initial shear distribution determines the
growth and evolution of the trapped surfaces, as prescribed by the apparent
horizon curve tah (r), given by (5.22). If the initial density profile is assumed
to be smooth at the center, then ρi (r) = ρc + ρ2 r2 + · · · , with ρ2 ≤ 0,
which corresponds to F (r) = F0 r3 + F3 r5 + · · · , with F2 ≤ 0. Now suppose
that ρ2 (and hence F2 ) is non-zero. Then, (5.22) implies that the apparent
horizon curve initiates at r = 0 at the epoch t0 , and increases near r = 0
with increasing r, moving to the future (see Fig. 5.1). Note that as soon
as F2 is non-zero, even with a very small magnitude, the behavior of the
apparent horizon changes qualitatively. Rather than going back into the past
from the center, as would happen in the homogeneous collapse case with
F2 = 0, it is now future pointed. This leads to a locally naked singularity as
the collapse endstate. The singularity may be globally naked and visible to
faraway observers, depending on the nature of the density function at large
values of r.
 5.2 Evolution of a physically realistic collapse 221

1.2e-06

1e-06

8e-07
F3 = –30
6e-07
tah
4e-07
F3 = –15
2e-07
F3 = –2
0
F3 = –0.1
–2e-07
0 0.001 0.002 0.003 0.004 0.005
r

Fig. 5.1 The apparent horizon behavior.

A naked singularity occurs when a comoving observer at fixed r does not

encounter any trapped surfaces until the time of the singularity formation,
whereas, for a blackhole, the trapped surfaces form before the singularity.
Therefore, for a blackhole to form,

tah (r) ≤ t0 for r > 0, near r = 0. (5.25)

In the general case, where there is not a necessarily smooth initial density,
this condition is violated for n = 1, 2, as follows from (5.22). The apparent
horizon curve initiates at the singularity r = 0 at the epoch t0 , and increases
with increasing r, moving to the future, i.e. tah > t0 for r > 0 near the cen-
ter. The behavior of the outgoing families of null geodesics has been analyzed
in detail in these cases, and it is known that the geodesics terminate at the
singularity in the past, which results in a naked singularity, as discussed in
Section 3.6. In such cases, the extreme strong gravity regions can commu-
nicate with outside observers. For the case where n = 3, (5.25) shows that
it is possible to have a blackhole if F3 ≥ −2F0 , or a naked singularity if
5/2

F3 < −2F0 . For n ≥ 4, (5.25) is always satisfied, and a blackhole forms.

5/2

When the dust density is homogeneous, the apparent horizon starts devel-
oping earlier than the epoch of the singularity formation, and the singularity
then is fully hidden within a blackhole. There is no density gradient at the
center, and no shear present within the cloud. On the other hand, if a den-
sity gradient is present at the center, then the trapped surface development
is delayed due to the presence of the shear, as seen above, and, depending
on the ‘strength’ of the density gradient and shear at the center, this may
allow the singularity to be visible. It is the rate of decrease of the shear, as
222 Final fate of a massive star

the center r = 0 on the initial surface t = ti is approached, given by (5.24),

which determines the endstate of the collapse.
The basic point is that when the shear falls rapidly enough to zero at
the center, the result is necessarily a blackhole. If the shear falls more slowly,
there is a naked singularity. It is thus seen that naked singularities are caused
by sufficiently strong shearing forces near the singularity, as generated by the
inhomogeneities in the density distribution of the collapsing configuration.
When the shear decays rapidly near the singularity, the situation is effectively
like the shear-free and homogeneous density case, with a blackhole endstate.
It provides a useful insight to see that, when a blackhole forms, the appar-
ent horizon typically comes into being as a finite-sized surface, at a finite r,
then moving to the center r = 0. This is what happens, for example, in the
Oppenheimer–Snyder blackhole formation in a homogeneous dust collapse.
In such cases, the event horizon, which does typically start at a point, could
have formed earlier than the apparent horizon. On the other hand, in the
case of a naked singularity, it follows from (5.16) and (5.22), that the appar-
ent horizon starts at r = 0, and then is future directed in time, that is, tah
grows with increasing coordinate radius r along the apparent horizon curve
R = F . These two behaviors of the apparent horizon curve are very differ-
ent and are governed by the shearing effects. A comoving observer will not
encounter any trapped surfaces until the time of the singularity formation
in the naked singularity case, whereas, in the blackhole case, the apparent
horizon typically develops before the epoch of the singularity formation. This
is what is meant by the delayed formation of the apparent horizon, caused
by the shearing effects or inhomogeneity present within the cloud.
The relation between density gradients and shear may be understood via
the non-local or free gravitational field. Density gradients act as a source for
the electric Weyl tensor as given by Maartens and Bassett (1998)

Db Eab = 13 Da ρ , (5.26)

where Da is the covariant spatial derivative. Note that the magnetic Weyl
tensor vanishes for spherical symmetry. In turn, the gravito-electric field is
a source for shear, or equivalently, the shear is a gravito-electric potential,

2 2
uc ∇c σab + Θσab + σac σ c b − σ 2 hab = −Eab . (5.27)
3 3

Therefore, density gradients may be directly related to shear as

Da ρ = −4σDa σ − 2ΘDb σab − 3Db (uc ∇c σab ) − 3σa b Dc σbc − 3Db (σac σ c b ),
(5.28)
 5.2 Evolution of a physically realistic collapse 223

where the shear constraint Db σab = 2/3Da Θ has been used. Equation (5.28)
explicitly links the behavior of the density gradients and the shear near the
center, which was discussed above. The free gravitational field that mediates
this link can also provide a covariant characterization of the singularity for-
mation. By (5.24) and (5.27), the relative gravito-electric field E/Θ2 (where
E 2 = 1/2E ab Eab ) near r = 0 is given at t = ti by
 
E −7nFn n
= √ r [1 + O(r)] . (5.29)
Θ2 i 18 6F0

Therefore, naked singularities in a spherical dust collapse are signaled by

a less rapid fall-off of the relative gravito-electric field as the singularity
is approached. Equations (5.24) and (5.29) provide two equivalent ways of
expressing this result. This specifies how much shear is sufficient to create
a locally naked singularity. It is sometimes asked, in the case of a continual
collapse when the densities and curvatures grow without bound, and when
the gravitational fields grow to extreme, how the trapping of light can be
delayed and how such extreme gravity regions can be visible. The above
consideration provides an insight into this, namely that the shearing effects
due to inhomogeneity, which are purely general relativistic effects, cause
the delay of trapped surfaces by distorting the geometry of the trapped
surfaces and the apparent horizon, thus allowing the light to escape, even
from extreme gravity regions.
For the case of a dust collapse, the role of inhomogeneity or shear in
deciding the endstate of the collapse is fairly transparent. To understand
how shear affects the formation of the apparent horizon for general mat-
ter fields with pressures included is much more complicated, in particular
because F = F (t, r) in a general case, whereas Ḟ = 0 for dust. In fact, even
in certain general classes of non-dust models with non-zero pressures, it is
possible to characterize the collapse covariantly. In the above, it was shown
that homogeneous density implies a blackhole endstate. The next logical step
would be to consider models for which the initial density is homogeneous.
For example, if the mass function is

F (t, r) = f (r) − R3 (t, r), f (r) = 2r3 , (5.30)

then (5.6) shows that ρi and (pr )i are constants. The density and pressure
may however develop inhomogeneities as the collapse proceeds, depending
on the choice of the remaining functions, including in particular the initial
velocities of the collapsing shells. The collapse may then end up in either a
blackhole or a naked singularity (for a discussion on this for the case of a
dust collapse, see Joshi and Dwivedi, 1993a). In fact, it can be shown that
zero shear implies a blackhole for these models. By (5.3), (5.6), and (5.30),
224 Final fate of a massive star

the shear-free condition leads to R /R = 1/r, and (5.6) then shows that
ρ = ρ(t), i.e. the density evolution is necessarily homogeneous. As shown
above, the collapse thus necessarily ends in a blackhole. For the class of
models given by (5.30), whenever the collapse ends in a naked singularity,
the shear must necessarily be non-vanishing. Although this class of models is
somewhat special, the result indicates that the behavior of the shear remains
a crucial factor even when pressures are non-vanishing.
It would appear that the only way a singularity becomes visible is when
suitable modifications occur in the geometry of trapped surfaces and the
apparent horizon, and a delay in the trapped surface formation is necessary.
As has been shown here, the shear provides a rather natural explanation for
the occurrence of locally naked singularities in certain collapse models. Suffi-
ciently strong shearing forces in spherical collapsing dust do affect the delay
in the formation of the apparent horizon, thereby exposing the strong gravity
regions to the outside world, and leading to a naked singularity formation.
When shear decays rapidly near the singularity, the situation is effectively
like the shear-free case, with a blackhole endstate. The important point is
that naked singularities can develop in quite a natural manner, very much
within the standard framework of general relativity, as governed by shearing
effects.
In the case of a spherical dust collapse, shear and density inhomogeneity
are equivalent, that is, one implies the other. Although shear contributes
positively to the focusing effect via the Raychaudhuri equation

1 1
Θ̇ + Θ2 = − ρ − 2σ 2 , (5.31)
3 2

its dynamical action can make the collapse incoherent and dispersive.
Depending on the rate of fall-off of shear near the singularity, its disper-
sive effect can play the critical role of delaying the formation of the apparent
horizon, without directly hampering the process of collapse. The dispersive
effect of shear always tends to delay the formation of the apparent horizon,
but it is able to expose the singularity only when the shear is strong enough
near the singularity.
Here, the effects of inhomogeneities and shear are analyzed for certain
collapse models. The implication is certainly not that it is only the shear
that can cause the changes in the trapped surfaces’ geometry. The important
point to emphasize is that modifications in the trapped surfaces’ geometry
can occur rather naturally in the process of a dynamical collapse, due to
physical agencies such as inhomogeneities and spacetime shear. In general,
the phenomena such as trapped surfaces formation and apparent horizons
are independent of any spacetime symmetries, and it would seem clear that
a naked singularity will not develop in a more general gravitational collapse
 5.3 Non-spherical models 225

as well, unless there is a suitable delay of the apparent horizon. This suggests
that the forces such as shear will continue to be pivotal in determining the
final fate of physically realistic gravitational collapse scenarios, independent
of any spacetime symmetries. In any case, the main purpose here has been to
try to understand and find the physical mechanism that leads the collapse to
the development of a naked singularity rather than a blackhole, in some of the
well-known classes exhibiting such behavior. It is found that the shear and
inhomogeneity provide a covariant dynamical explanation of the phenomenon
of naked singularity formation in a spherical gravitational collapse.
The basic question is: what governs the geometry of the trapped surfaces,
or the formation, or otherwise, of the naked singularities in gravitational
collapse? In other words, what is it that causes the naked singularity rather
than a blackhole to develop as the final end product of the collapse? It turns
out from the above that physical agencies such as inhomogeneities in matter
profiles, as well as spacetime shear, play a key role in distorting the trapped
surface geometry, and could delay the trapped surface formation during the
collapse, thus giving rise to a naked singularity, rather than a blackhole, as
the collapse final state. This, in a way, provides the physical understanding of
the phenomena of blackhole and naked singularity final states in the collapse.
It is seen that in a spherical collapse, these phases are generic, and are seen
to be determined by the nature of the initial data from which the collapse
develops, and in terms of the allowed dynamical evolutions. Physical agencies
such as inhomogeneities and shear may cause them, and given the initial
data, there are non-zero measure classes of evolutions that evolve into either
of these outcomes.

5.3 Non-spherical models

In terms of analytic calculations, not much is known about the non-spherical
gravitational collapse. There are indications, however, in view of the proper-
ties of the Kerr geometry, which is a solution to the Einstein equations for a
rotating particle, that to avoid naked singularity formation the object should
not be rotating very fast. In Kerr geometry, the metric is characterized in
terms of the mass of the particle and its angular momentum. If the angular
momentum is larger than the mass, then a naked singularity naturally forms
in the spacetime. It should be noted, however, that there is no interior met-
ric yet known for a Kerr exterior solution. The ideal situation would be the
existence of a non-spherical collapsing cloud, the exterior of which is given
by the Kerr solution, and then to investigate the final state of the collapse.
A spherically symmetric homogeneous dust cloud collapse goes through
the phases of implosion and the subsequent formation of a horizon and a
spacetime singularity completely hidden within the horizon and a blackhole.
226 Final fate of a massive star

The consequences when the collapse is no longer homogeneous, and of differ-

ent equations of state for the matter, are investigated within the spherically
symmetric framework in the previous treatment.
Small perturbations over the spherically symmetric situation were taken
into account in the work of Doroshkevich, Zel’dovich, and Novikov (1966),
de la Cruz, Chase, and Israel (1970), and in the perturbation calculations
of Price (1972). The main outcome of these works is that the basic result of
the spherically symmetric collapse situation remains unchanged, at least in
the sense that an event horizon will continue to form in the advanced stages
of the collapse. It was indicated by the work of Doroshkevich, Zel’dovich,
and Novikov (1966) that, as the collapse progresses and the star reaches the
horizon, the perturbations remain small and there are no forces arising to
destroy the horizon. This analysis contained the basic idea of the no-hair
theorems for blackholes that were developed later, as it is pointed out that,
from the point of view of an outside observer, non-spherical perturbations in
the geometry and the electromagnetic field die down as the star approaches
the horizon. These findings were supported by the numerical calculations
of de la Cruz, Chase, and Israel (1970) and Price (1972), where linearly
perturbed collapse models were integrated to show that the perturbations
died out with time as far as they could be followed. It may be noted, however,
as pointed out by Israel (1986a), that a result showing that a small change
in the initial data perturbs the solution only slightly would be desired. The
above works lead to such a result for bounded time intervals only, whereas the
existence of a horizon depends on the entire future behavior of the solution.
Detailed numerical models for stellar collapse may help reach the desired
solution.
One would then like to inquire further, namely, do horizons still form
when the fluctuations from the spherical symmetry are large, and when the
collapse is highly non-spherical? This issue is again related basically to the
nature and evolution of the trapped surface geometry discussed earlier, as
the collapse develops, but now in the case of a non-spherical geometry. It is
known, for example, that when there is no spherical symmetry, the collapse of
infinite cylinders do give rise to naked singularities in general relativity, which
are not covered by horizons (Thorne, 1972; Misner, Thorne, and Wheeler,
1973; see also Herrera and Santos, 2005; Nakao et al., 2007). However, the
situation of greater physical interest would be that of finite systems, possibly
in an asymptotically flat spacetime. Not much is known about this, except
the hoop conjecture of Thorne (1972), which characterizes the final fate of
a non-spherical collapse in the following statement. The horizons of gravity
form when and only when a mass M gets compacted in a region whose
circumference in every direction obeys

C ≤ 2π (2GM/c2 ). (5.32)
 5.3 Non-spherical models 227

Therefore, unlike the cosmic censorship conjecture, the hoop conjecture now
no longer rules out all the naked singularities, but only makes a definite
assertion on the occurrence of the event horizons in the gravitational collapse.
It is thus seen that the hoop conjecture allows for the occurrence of naked
singularities in general relativity, at least when the collapse is sufficiently
aspherical, and especially when one or two dimensions are sufficiently larger
than the others. A known example of this is given by Lin, Mestel, and Shu
(1965), which discusses gravitational collapse for uniform spheroidal objects
from the perspective of instability in Newtonian gravity (see also, Thorne,
1972; Shapiro and Teukolsky, 1991, 1992, for a discussion). Here, a non-
rotating homogeneous spheroid collapses, maintaining its homogeneity and
spheroidicity, but its deformations grow as the collapse progresses. If the
initial condition is that of a slightly oblate spheroid, the collapse results in a
pancake singularity through which the evolution could proceed and continue.
However, for a slightly prolate spheroidal configuration, the matter collapses
to a thin thread that ultimately results in a spindle singularity. This is more
serious in nature in that the gravitational potential and force and the tidal
forces blow up, as opposed to only the densities blowing up, in a mere shell-
crossing singularity. Even in the case of an oblate collapse, the passing of
matter through the pancake causes prolateness, and subsequently a spindle
singularity again results without the formation of any horizon.
It was indicated by the numerical calculations of Shapiro and Teukolsky
(1991) that a similar situation is maintained in general relativity, also in con-
formity with the hoop conjecture. They evolved collissionless gas spheroids
in full general relativity that collapse in all cases to singularities. When the
spheroid is sufficiently compact, a blackhole that contains the singularity
forms, but when the semi-major axis of the spheroid is sufficiently large, a
spindle singularity results without an apparent horizon forming. These have
to be treated as numerical results, as opposed to a full analytic treatment,
which need not be in contradiction to a suitably formulated version of the
cosmic censorship. The definition of naked singularity here was basically in
terms of the non-occurrence of trapped surfaces in a certain family of non-
spacelike surfaces. Sometimes this may, or may not, indicate a genuine naked
singularity, and a blackhole may still form. However, this gives rise to the
possibility of the occurrence of naked singularities in the collapse of finite
systems in asymptotically flat spacetimes, which could be in violation of the
weak cosmic censorship, but possibly in conformity with the hoop conjecture.
A somewhat broader statement in a similar spirit to the hoop conjecture
is the event horizon conjecture of Israel (1984, 1986a, 1986b). A general
statement for this conjecture is given by the requirement that an event hori-
zon must form whenever a matter distribution (satisfying appropriate energy
conditions) has passed a certain critical point in its gravitational collapse,
namely the formation of a closed trapped surface. A strong motivation for
228 Final fate of a massive star

believing in this conjecture is that, unlike the cosmic censorship hypothesis,

no counter-examples have been found so far. For example, the inequality
A < 4π (2M )2 must hold if the event horizon conjecture is true, for the area
A of a closed trapped surface that forms at time t = 0 (say) in the gravita-
tional collapse of a mass M . The inequality depends only on the initial data
on the spacelike surface t = 0 and hence there is no need to trace future
evolution of the system in order to verify it. Clearly, further investigations
in a non-spherical collapse are needed to probe such a conjecture.
In the case of the exact spherical symmetry holding for the spacetime, it
is known (see for example, Leibovitz and Israel, 1970) that an event horizon
must form when a star collapses to a sufficiently small radius, and when the
positivity of energy is satisfied. The only alternative to this is that the star
must radiate away all of its mass in the process of the collapse. As discussed in
Section 5.1 above, this appears very difficult, at least using purely classical
or astrophysical processes, for the stars having tens of solar masses. Even
for a non-spherical collapse, which involves only small perturbations from
the spherical symmetry, the numerical calculations mentioned above seem to
indicate that a non-singular event horizon must develop during the collapse.
As noted by Jang and Wald (1977), without an event horizon to stop the
flow of outgoing radiation, the gravitational mass of the star must become
negative.
In fact, for all the classes of spacetimes discussed in Chapter 3, an event
horizon always forms in the process of the collapse, even though it fails
to cover the singularity completely. This results in the formation of naked
singularities in the spacetime. Note that merely the formation of an event
horizon is not a sufficient condition for the avoidance of the naked singularity.
If the trapped surfaces and event horizon form late enough as the collapse
develops, a strong curvature singularity can still be visible, as discussed here.
Not only should the event horizon develop, but it must develop early enough,
as is the case for the example in the Oppenheimer–Snyder collapse, in order
to create a blackhole as the gravitational collapse final state.
Of course, in such a case, when the event horizons form accompanied by
the naked singularities, the usual interpretation of blackhole physics becomes
unclear due to the back reaction on the metric as a result of the possible
emission from the naked singularities. For example, in the detailed analysis
for the case of radiation collapse using the Vaidya metric, the horizon could
be fully covered by the emissions from the strong curvature naked singularity,
some of which may fall back inwards, thus affecting the area and structure
of the horizons.
As pointed out by Israel (1984), there are two main aspects of the for-
mulation of such an event horizon conjecture. First, one has to characterize
the formation of a trapped surface in terms of the initial data on a spacelike
surface in the form of a statement, such as a closed two-surface S will be
 5.3 Non-spherical models 229

trapped if the gravitational mass interior to S exceeds a certain critical value

(defined suitably in terms of the geometry of S ). The works such as those of
Schoen and Yau (1983) and Ludvigsen and Vickers (1983) would be relevant
here. Whereas this could be carried out for the spherical symmetric case,
the situation could be fairly complicated in general, because one would need
to characterize the concept of ‘mass interior to S ’ much more precisely. The
other aspect is that of time evolution of such a trapped surface, which should
extend for a finite distance in the future to generate a spacelike three-cylinder,
the sections of which are trapped surfaces with bounded dimensions. Assum-
ing energy conditions, the regularity of the energy-momentum tensor would
be desired on this three-cylinder. If these bounds and extension depend only
on the initial geometry and the mass within the trapped surface, then, as
pointed out by Israel (1986a, 1986b), the three-cylinder could be extended
infinitely into the future, provided it encounters no singularities in its future
development.
Now, the occurrence and nature of naked singularities forming in a class of
non-spherical models known as the Szekeres spacetimes are examined. These
represent irrotational dust collapses, they have no Killing vectors, and are
generalizations of the Tolman–Bondi–Lemaitre spacetimes. It is seen that in
these spacetimes, naked singularities exist that satisfy the limiting focusing
condition and the strong limiting focusing condition. This generally ensures
that various classes of naked singularities forming in the Szekeres models and
also in the TBL collapse spacetimes, as discussed earlier, are strong curvature
naked singularities.
A future incomplete (or past incomplete) causal geodesic terminates in
a strong curvature singularity in the future (or past) if, for every point
q ∈ λ, the expansion θ of the congruence of the future directed (or past
directed) causal geodesics originating from q and infinitesimally neighboring
λ diverges. If the strong curvature condition on a spacetime is defined as hold-
ing if all future and past incomplete null geodesics generating an achronal set
terminate in a strong curvature singularity, then Królak (1983) tried to prove
that, under a strong curvature condition on the spacetime, the cosmic cen-
sorship should hold. Such attempts, however, did not succeed and it turned
out that such a conjecture does not hold, as shown by further work. Also,
the theorems proved by Królak (1983, 1986) and later by Królak (1999)
required an extra restrictive assumption on the causal structure of space-
times. Furthermore, explicit examples of naked strong curvature singularities
were found in Królak’s sense (Eardley and Smarr, 1979; Christodoulou, 1984;
Newman, 1986), and also in Tipler’s sense (Tipler, Clarke, and Ellis, 1980),
and by Dwivedi and Joshi (1989, 1991) and Joshi and Dwivedi (1993a).
These demonstrated the occurrence of strong curvature naked singularities
in the TBL spacetimes representing a spherically symmetric inhomogeneous
collapse of dust, and also in the Vaidya radiation collapse models.
230 Final fate of a massive star

The TBL dust collapse spacetimes, even though they generalize the homo-
geneity assumption, are special in some ways. They are spherically symmetric
and they have matter in the form of irrotational pressureless dust. It is
interesting to know whether naked strong curvature singularities occur in
more general situations than this. Here, it is discussed how naked strong
curvature singularities occur in the Szekeres spacetimes that do not have
any Killing vectors (Joshi and Królak, 1986). This result shows that naked
strong curvature singularities do not arise necessarily as a result of the spher-
ical symmetry. Nevertheless, the Szekeres spacetimes have the same special
form of matter as the TBL spacetimes, which is again irrotational pressure-
less dust. Moreover, Szekeres spacetimes are also special because they can
be matched to the Schwarzschild spacetime and they cannot contain any
gravitational radiation (Bonnor, 1976).
The Szekeres spacetime (Szekeres, 1975) is a solution of the Einstein
equations representing irrotational dust, where

Gab = Tab = ρua ub , ua ua = 1, (5.33)

and the units have been chosen so that c = 8πG = 1. The metric has the
diagonal form given by

ds2 = dt2 − S 2 dr2 − Y 2 (dx2 + dy 2 ), (5.34)

where (r, x, y ) are comoving spatial coordinates. The solution is given below
for the case Y  = ∂Y /∂r = 0,

R(t, r) P (r, x, y )Y  (t, r)

Y = , S=  , (5.35)
P (r, x, y ) 1 + f (r)

where f (r) > −1 and

P = a(r)(x2 + y 2 ) + 2b1 (r)x + 2b2 (r)y + c(r), (5.36)

1
ac − b21 − b22 = , (5.37)
4
F (r)
Ṙ2 = f + . (5.38)
R
Here F (r) is again an arbitrary function of r as in the TBL collapse case,
R denotes the physical radius for the cloud, and the dot denotes the partial
derivative with respect to the time coordinate t.
As earlier, to ensure physical reasonability of the collapse models, some
regularity conditions are imposed, and a class of Szekeres collapse spacetimes
is dealt with. First, the metric is taken to be C 1 everywhere in the spacetime.
Then, the function P must be non-zero everywhere, and its derivative with
 5.3 Non-spherical models 231

respect to r must be continuous and vanishing at r = 0. Also, take the metric

to be locally Euclidean at r = 0. Then, it is necessary to set

f (0) = 0. (5.39)

The function R0 (r) = R(r, 0), which again indicates the physical radius of
the shells, is a monotonically increasing function of r. The freedom in the
choice of the radial coordinate r can be used to do the scaling as

R0 (r) = r. (5.40)

The dust density ρ is given by

P F  − 3F P 
ρ= . (5.41)
P 2 R2 Y 
Although for P > 0, the surfaces r = const., t = const. are spheres, the
solution is not spherically symmetric here because the spheres are no longer
concentric in this case, and their centers are given by (−a−1 b1 , −a−1 b2 ).
Szekeres has also analyzed the singularities and their causal structure in these
spacetimes. When R = 0, the singularity is of the first kind, and when Y  = 0
the singularity is of the second kind. The singularities of the second kind are
familiar shell-crossing singularities that also occur in TBL spacetimes. As in
the TBL spacetimes, the shell-crossing singularities in Szekeres spaces can
also be both locally and globally naked (Szekeres, 1975). However, they are
generally believed to be mild and they will not be considered here. These
singularies can be eliminated by imposing a regularity condition,

Y  > 0. (5.42)

This is no loss of generality, as the main interest here is in examining the

properties of the shell-focusing singularities in terms of visibility, or other-
wise. Szekeres (1975) has also shown that whenever r > 0, the shell of the
dust always crosses the apparent horizon before collapsing to a singularity,
and therefore for r > 0, the singularity cannot be naked, and all these points
hide behind the event horizon. Therefore, the singularity of the first kind
can be naked only when r = 0, which is called the central singularity. This
situation is again analogous to the TBL case. It will be shown that, as in
TBL spacetimes, naked strong curvature singularities do occur in Szekeres
spacetimes. Consider the case of a gravitational collapse, that is, Ṙ < 0. For
simplicity, only the class of marginally bound collapse models is considered,
that is,
f (r) = 0, (5.43)
232 Final fate of a massive star

as mentioned above. Then, the function R(r, t) is given by

  2/3
3 F
R=r 1− t . (5.44)
2 r3

The analysis here can again be similar to that in the TBL collapse models
case, as given by Joshi and Dwivedi (1993a). A set of new functions can be
introduced,

R F F
X= α
, η = r , Λ = α,
r F r
1 − 3η
1
P
Θ= , L=r , (5.45)
r3(α − 1)/2 P

where α ≥ 1, and√the unique value of the constant α is determined by the

condition that Θ/ X does not vanish or does not go to infinity identically as
r → 0 in the limit of approach to the central singularity along any X = const.
direction. It will be assumed that the above functions are at least C 2 . Partial
derivatives R and Ṙ that are useful in the analysis of the singularity are
then given by

Λ N
Ṙ = − , R = rα−1 H, Ṙ = − , (5.46)
X r

where
1 Θ
H = ηX + √ , (5.47)
3 X
√  
Λ 2
N = − 2 Θ − ηX 3/2
. (5.48)
2X 3

The tangents K a = dxa /dk for the outgoing radial (x = const., y = const.)
null geodesics can be written as

dt P
Kt = = , (5.49)
dk Y
dr P
Kr = = , (5.50)
dk PY Y 
dx
Kx = = 0, (5.51)
dk
dy
Ky = = 0, (5.52)
dk
 5.3 Non-spherical models 233

where P satisfies the differential equation

 
dP Ẏ  Ẏ 1
+P 2
− − = 0. (5.53)
dk Y Y  Y 2 PY 2

The parameter k is an affine parameter along the null geodesics.

If the future directed outgoing null geodesics are to terminate in the past
at the central singularity at r = 0, which occurs at some time t = t0 at which
R(t0 , 0) = 0, then along such geodesics we must have R → 0 as r → 0. The
following equation is satisfied along the null geodesics
 
dR 1 dt 
= Ṙ + R (5.54)
du αrα−1 dr
   √
Λ H (X, u) XΛ
= 1− + L ≡ U (X, u), (5.55)
X α α

where u = rα . Note that when one writes the quantity U , one is essentially
working out the necessary condition for the null geodesics to emerge from
the singularity at R = 0, r = 0, with a well-defined tangent given by dR/dt.
Therefore, if the trajectories are emerging with a well-defined tangent from
the singularity, then the quantity U (x, u) is well-defined by definition. For
more details, see, for example, Joshi and Dwivedi (1993a). Consider the limit
X0 of the function X along the null geodesic terminating at the singularity
at R = 0, u = 0. Using the l’Hospital rule,

R dR
X0 = lim = lim = lim U (X, u) = U (X0 , 0).(5.56)
R→0, u→0 u R→0, u→0 du R→0, u→0

The necessary condition for the existence of the null geodesic outgoing from
the central singularity is the existence of the positive real root X0 of the
equation
V (X ) ≡ U (X, 0) − X = 0. (5.57)

By the regularity conditions, limr→0 L = 0. Consequently, the necessary

condition for the existence of the naked singularity in the marginally bound
case of the Szekeres spacetime is the existence of the positive real root of the
equation
  
Λ0 H (X, 0)
1+ − X = 0, (5.58)
X α

where
η0 = η (0), Λ0 = Λ(0), Θ0 = Θ(0). (5.59)
234 Final fate of a massive star

This is exactly the same equation as in the marginally bound TBL case.
Consequently, the same analysis as by the TBL case given earlier by Joshi
and Dwivedi (1993a) applies here. Here, only a few key results will be sum-
marized. To show that the singularity is naked, there is a need to prove that
a solution of the geodesic equation exists such that the tangent X0 is real-
ized at the singularity. It can be proved that there is always at least a single
null geodesic outgoing from the central singularity. Therefore, the existence
of the real and positive root of the (5.58) is both a necessary and sufficient
condition for the existence of a naked singularity.
Some further remarks will now be made on the strength of the singularity,
which will be investigated in some detail. Let M be the spacetime manifold,
and let J (k ) be the quantity along a null geodesic in the spacetime as given
below, with λ : (k0 , 0] → M , being parametrized by an affine parameter k ,
 0
J (k ) = Rab K a K b dk  . (5.60)
k0

Say that the limiting focusing condition holds if J (k ) is unbounded in the

interval (k0 , 0], and say that the strong limiting focusing condition holds if
J (k ) is non-integrable on an interval (k0 , 0]. It is proved in Clarke and Królak
(1986), that the limiting focusing condition implies that λ terminates in a
strong curvature singularity in the sense of Królak, in the future, whereas
the strong limiting focusing condition implies that λ terminates in Tipler’s
strong curvature singularity in the future.
To find out whether the naked singularity satisfies the strong limiting
focusing condition, the limit limk→0 k 2 Rab K a K b along the future directed
null geodesics emerging from the singularity will be investigated. Using the
l’Hospital rule, regularity conditions, and (5.53),



k 2 F  − 3F PP (K t )2 η0 Λ0 H0
lim k 2 Rab K a K b = lim

 = 2 ,(5.61)
k→0 k→0 R R −RP
2  P
X0 (α − N0 )2

where N0 = N (X0 , 0). Therefore, if Λ0 = 0 the naked singularity satisfies

the strong limiting focusing condition.
Suppose then that Λ0 = 0. Equation (5.53) can be written in the form
  
d(ln K t ) Λ 1 dr
= N− L . (5.62)
dk X r dk

Since as k → 0, N → 0 because Λ0 = 0 and L → 0, by the regularity

conditions, the right-hand side of the above equation is integrable on the
 5.4 Blackhole paradoxes 235

interval (k0 , 0] with respect to k . Therefore the limit limk→0 K t exists, and

Kt
Kr = . (5.63)
rα−1 (H − XL)
Therefore if α = 1 the limit limk→0 K r also exists. Suppose that α = 1, and
consider the limit limk→0 kRab K a K b . Applying the l’Hospital rule twice and
using (5.53),

η0 Λ0 H0
lim kRab K a K b = lim K r . (5.64)
k→0 X02 (α − N0 )2 k→0
Therefore when α = 1 the above limit is finite and the naked singularity just
satisfies the limiting focusing condition (Rab K a K b diverges logarithmically).
If α > 1 the above limit diverges and the naked singularity also satisfies the
limiting focusing condition (but not the strong focusing condition, unless
Λ0 = 0).
The above results show that under the regularity conditions, and for a
marginally bound collapse, any central naked singularity in the Szekeres
spacetime is always a strong curvature singularity in Królak’s sense. Fur-
thermore, as the considerations above reduce to the TBL case under various
appropriate limits, the above also implies that many classes of naked sin-
gularities occurring in the TBL dust collapse models are strong curvature
naked singularities necessarily. The significance of this is that these are
physically serious curvature singularities, which are not removable from the
spacetime. As discussed earlier in Chapter 4, all volume forms along non-
spacelike geodesics falling into such a singularity go to a vanishing value. In
other words, all physical objects are crushed to a zero volume as they fall
into a strong curvature singularity.

5.4 Blackhole paradoxes

While the Einstein gravity has been a highly successful theory of gravitation,
it generically admits the occurrence of spacetime singularities under fairly
general physical conditions. These are extreme gravity regions in the space-
time, where the matter densities and spacetime curvatures, as well as various
physical quantities, typically blow up and the theory must break down. Such
singularities may develop in cosmology, indicating a beginning for the uni-
verse, or in a gravitational collapse that ensues whenever a massive star
exhausts its nuclear fuel and undergoes the process of a continual collapse.
It is expected that possibly a future theory of quantum gravity may resolve
these singularities, where all known laws of physics break down. There have
been very many attempts over past decades in this direction, by trying to
construct a quantum theory of gravity that includes the string theory and
236 Final fate of a massive star

the loop quantum gravity formalisms. All the same, it would appear that the
goal of achieving a fully consistent and complete quantum theory of gravity
is yet to be reached. The possibility of singularity resolution through quan-
tum gravity is clearly linked to such efforts, and much work may be needed
before it can be fully realized.
The singularity theorems predicting the occurrence of the spacetime singu-
larities, however, contain three main assumptions under which the existence
of a singularity is predicted in the form of geodesic incompleteness in the
spacetime. These are in the form of a typical causality condition that ensures
a suitable and physically reasonable global structure of the spacetime, an
energy condition that requires the positivity of energy density at the classi-
cal level as seen by a local observer, and finally a condition demanding that
trapped surfaces must exist in the dynamical evolution of the universe, or
in the later stages of a continual gravitational collapse. A trapped region in
the spacetime consists of trapped surfaces, which are two-surfaces, such that
both in going as well as outgoing wavefronts normal to it must converge.
Such trapped surfaces then necessarily give rise to a spacetime singularity
either in a gravitational collapse or in cosmology.
For the same reason, the process of trapped surface formation in a gravita-
tional collapse is also central to blackhole physics. The role of such a trapping
of light and matter within the framework of Einstein’s theory of gravitation
was highlighted by Datt (1938) and Oppenheimer and Snyder (1939), within
the context of the continual collapse of a massive matter cloud. They studied
the collapse of pressureless dust clouds using general relativity, and showed
that it leads to the formation of an event horizon, and a blackhole as the
collapse endstate, assuming that the spatial density distribution within the
star was strictly homogeneous, that is, ρ = ρ(t) only. In the later stages
of the collapse, an apparent horizon develops that is the boundary of the
trapped region, thus giving rise to an event horizon. Once the collapsing star
has entered the event horizon, the causal structure of the spacetime there
would then imply that no non-spacelike curves from that region would escape
away, and the star is cut off from the faraway observers, thus giving rise to
a blackhole in the spacetime.
While blackhole physics has led to several interesting theoretical, as well
as observational, developments and discussions, it is necessary, however, to
study more realistic models of the gravitational collapse in order to put
blackhole physics on a sound footing. This is because the Oppenheimer–
Snyder–Datt scenario is rather idealized and pressures would play an
important role in the dynamics of any realistic collapsing star. Also, density
distributions within the star could not be completely homogeneous in any
physically realistic model, but would be higher at the center of the cloud,
with a typical negative gradient as one moved away from the center. The
 5.4 Blackhole paradoxes 237

study of a dynamical collapse within Einstein’s gravity is, however, a diffi-

cult subject because of the non-linearity of the Einstein equations. It was
hence proposed in 1969 by Penrose that any physically realistic continual
gravitational collapse must necessarily end in a blackhole final state only.
This is the cosmic censorship conjecture as discussed earlier. Although cos-
mic censorship has played a rather crucial role as a basic assumption in all
the physics and astrophysical applications of blackholes so far, no proof or
any mathematically rigorous formulation for it is available as yet, despite
many attempts. Hence, this has been widely recognized as the single most
important problem in the theory of blackholes, and the gravitation physics
of today. In order to make any progress on the censorship hypothesis, or
to understand the final endstate of a massive collapsing star it is necessary
to study realistic gravitational collapse scenarios within the framework of
Einstein’s gravity theory (Joshi, 1993).
Such a study of gravitational collapse is also warranted because of several
deep paradoxes that are associated with the blackholes, which have been
widely discussed. First, all the matter entering a blackhole, must, of necessity,
collapse into a spacetime singularity of infinite density and curvatures at the
center of the cloud, where all known laws of physics must break down. It
is not clear how such a model can be stable at the classical level. Second,
there is a need to formulate mathematically and prove the cosmic censor
hypothesis, that a generic gravitational collapse gives rise to a blackhole
only. As is clear from the discussions so far, this is turning into a formidably
difficult problem, and gravitational collapse studies so far have pointed to
several detailed collapse models that show that the final fate of a collapsing
star could be either a blackhole or a naked singularity, depending on the
nature of the initial data from which the collapse evolves, and the possible
evolutions allowed by the Einstein equations, as discussed here. As opposed
to a blackhole final state, the naked singularity is a scenario where ultra-
strong density and curvature regions of spacetime, forming as result of the
collapse, would be visible to faraway observers in the universe, in violation
of the cosmic censorship. Finally, it is well-known that a blackhole would
create information loss, violating the unitarity principle, thus creating a
basic contradiction with fundamental principles of quantum theory.
Within the context of such a scenario, it may be worth investigating
the possibility that the singularity problem in general relativity, as well as
the blackhole paradoxes, can be resolved, by possibly avoiding the trapped
surface formation in the spacetime during the process of a dynamical grav-
itational collapse. Note that within a cosmological context, singularity free
solutions have been discussed for example, (see for example, Senovilla, 2006,
and references therein). It would be possible to construct classes of collaps-
ing solutions for the Einstein equations, where the trapped surface formation
238 Final fate of a massive star

could be delayed or avoided during the collapse. This could then offer a some-
what natural resolution to the issues such as those above. The main issue
would then be that of finding a mechanism, either at a classical or quantum
level, so that much of the matter of the collapsing star can escape away and
would be thrown out during the final stages of the gravitational collapse,
thus resolving the problems such as those above, and also that of the infinite
density spacetime singularity. Towards realizing such a scenario, the usual
physical reasonability conditions could be imposed, and while regularity of
the initial data could be required, as well as the weak energy condition, the
pressure could be allowed to be negative.
Such a possibility at a quantum level is discussed in some detail in the
next section, where the role of quantum effects towards generating an out-
wards flow of matter and radiation from the collapsing cloud in the very
late stages of the collapse is explored. In any case, the occurrence of a
spacetime singularity indicates the breakdown of the Einstein gravitation
theory in these extreme regimes. The singularity problem and various black-
hole paradoxes at classical and quantum levels were mentioned above. Note
that quantum corrections could generate a strong negative pressure in the
interior of the cloud in the very late stages of the collapse, where the
classical theory should break down. The collapsing star could then radi-
ate away most of its matter as the process of the gravitational collapse
evolves, so as to avoid the formation of trapped surfaces and the spacetime
singularity.

5.5 Resolution of a naked singularity

While a naked singularity may form as the endstate of a continual gravita-
tional collapse within the framework of classical general relativity, the very
final stages of the collapse should be dominated by the quantum effects.
Therefore, these must be taken into account when determining the final fate
of the collapse in any physically realistic scenario. Here, some possibilities in
this direction, especially from the perspective of whether the singularity can
be resolved once the quantum gravity effects are taken into consideration,
are explored.
A class of collapsing scalar field models with a non-zero potential, where
the weak energy condition is satisfied by the collapsing configuration are
constructed and studied. It is seen that the endstate of the collapse at the
classical level can be either a blackhole or a naked singularity, and that phys-
ically it is the rate of collapse that governs these outcomes of the dynamical
evolution. This feature is similar to the Vaidya radiation collapse models (see
Joshi, 1993, for details and references), where again it is the rate of collapse
that determines either the blackhole or naked singularity outcomes.
 5.5 Resolution of a naked singularity 239

There have been considerations of models of four-dimensional gravity cou-

pled to a scalar field with potential V (φ), and modifications and numerical
studies of such a scenario have been discussed in some detail (Alcubierre
et al., 2004; Dafermos, 2004; Garfinkle, 2004; Gutperle and Krauss, 2004;
Hertog, Horowitz, and Maeda, 2004; Hubeny et al., 2004). Some of these
models satisfy the positive energy theorem, but would violate the energy
conditions within an asymptotically anti-de Sitter framework. While the
cosmic censorship violation is not clear, these investigations provide infor-
mation on some aspects of self-interacting scalar fields. Though the basic
question of cosmic censorship in scalar field collapse models remains open
to further analysis, the case of a massless scalar field has been analyzed to
see that a naked singularity develops, but not generically (Christodoulou,
1994, 1999).
Considered below is the scalar field collapse with the potential to examine
in some detail the final state for such a collapse. Some implications of the
loop quantum gravity formalism, when applied to and considered within such
a collapsing cloud in the later stages of the collapse, especially to examine
whether these quantum effects help resolve the classical naked singularity
will be discussed. It is possible that the occurrence of a naked singularity as
the outcome of the collapse may offer the possibility of observing quantum
gravity effects taking place in such visible ultra-strong gravity regions. No
such possibilities exist if the collapse necessarily always ends in a blackhole.
Whereas the quantum effects would be certainly important in the vicinities
of the ultra-strong gravity regions that a spacetime singularity creates, these
will be of no observational consequence if the singularity is always covered
within a blackhole, where all physical effects are hidden within the event
horizon and cannot be seen by any external observer. On the other hand,
if the collapse terminates in a naked singularity, the ultra-strong gravity
regions have a causal connection to the outside universe, and there is a
possibility of observing the quantum gravity effects taking place in these
extreme gravity situations, which are created by the very late stages of the
collapse.
As discussed earlier, the generic conclusion in the studies on the gravita-
tional collapse of various matter fields such as dust, perfect fluids, radiation
collapse and other forms of matter is, depending on the nature of the regular
initial data in terms of the initial distributions of density and pressures of
the matter and other collapse parameters from which the collapse develops,
that the final outcome could be either a blackhole or a naked singularity.
Special importance is, however, sometimes attached to the investigation of
the collapse of a scalar field, because one would like to know if the cosmic
censorship is preserved in the collapse for fundamental matter fields that are
derived from a suitable Lagrangian. Also, in the later stages of the collapse
it is not certain what the equation of state for the matter would be, and it
240 Final fate of a massive star

is possible that the matter at these epochs could be of a fundamental form,

such as a Klein–Gordon or Maxwell field.
From such a perspective, a class of continual collapse models of a scalar
field with potential are constructed and studied here (see also Goswami and
Joshi, 2004c). It is required that the weak energy condition is preserved
throughout the collapse, although the pressures are allowed to be negative
closer to the singularity. The interior collapsing sphere is matched, using the
Israel–Darmois conditions, with a generalized Vaidya exterior to complete
the model. It is found that there are classes of collapse models included here
for which no trapped surfaces form in the spacetime as the collapse evolves,
and the singularity that develops as the collapse endstate is visible. In these
classes, it can be seen that naked singularities are created through a scalar
field collapse with potential from generic initial conditions that violate the
cosmic censorship. It is seen that physically it is the rate of collapse that
governs the formation of either a blackhole or a naked singularity as the
collapse endstate.
A spherically symmetric homogeneous scalar field, Φ = Φ(t), with a poten-
tial V (Φ) is discussed. This ensures that the interior spacetime must
have a Friedmann–Robertson–Walker (FRW) metric. Furthermore, the
marginally bound (k = 0) case is chosen. Then, the interior metric is of
the form
 
ds2 = −dt2 + a2 (t) dr2 + r2 dΩ2 , (5.65)

where dΩ2 is the line element on a two-sphere. In this comoving frame, the
energy-momentum tensor of the scalar field is given as

 
1 2
T t
t = −ρ(t) = − Φ̇ + V (Φ) , (5.66)
2
 
1 2
T rr θ φ
= T θ = T φ = p(t) = Φ̇ − V (Φ) , (5.67)
2

with all other off-diagonal terms being zero.

It may be noted that the comoving coordinate system chosen has a par-
ticular physical significance as compared with an arbitrary system, and the
quantities ρ and p are interpreted as the density and pressure of the scalar
field respectively. It is then easily seen that the scalar field behaves like a
perfect fluid, as the radial and tangential pressures are equal. Take the scalar
field to satisfy the weak energy condition, that is, the energy density as mea-
sured by any local observer is non-negative, and for any timelike vector V i ,
Tik V i V k ≥ 0. This amounts to ρ ≥ 0, ρ + p ≥ 0.
 5.5 Resolution of a naked singularity 241

The dynamical evolution of the system is now determined by the Einstein

equations, which for the metric (5.65) become (in the units 8πG = c = 1),

F −Ḟ
ρ= p= , (5.68)
R2 R R2 Ṙ
F
Ṙ2 = . (5.69)
R

For general spherically symmetric spacetimes, F (t, r) has the interpretation

of the mass function for the cloud, with F ≥ 0. The quantity R(t, r) = ra(t)
is the area radius for a shell labeled by the comoving coordinate r. In order
to preserve the regularity of the initial data, F (ti , 0) = 0, that is, the mass
function should vanish at the center of the cloud. From (3.28) it is evident
that on any regular epoch F ≈ r3 near the center, in order for the initial
data to be singularity free.
The Klein–Gordon equation for the scalar field is given by

d  3 
a Φ = −a3 V (Φ),Φ . (5.70)
dt

Since the aim here is to construct a continual collapse model, the class with
ȧ < 0, which is the collapse condition implying that the area radius of a
shell at a constant value of comoving radius r decreases monotonically, is
considered. In general, there may be classes of solutions where a scalar field
may disperse also (see for example, Choptuik, 1993). The objective, however,
is to examine whether the singularities forming in the scalar field collapse
could be naked, or would be necessarily covered within a blackhole, and if so
under what conditions. The singularity resulting from a continual collapse is
given by a = 0, that is, when the scale factor vanishes and the area radius
for all the collapsing shells becomes zero. At the singularity ρ → ∞.
The key factor that decides the visibility, or otherwise, of the singularity is
the geometry of the trapped surfaces that may form as the collapse evolves,
that is the two-surfaces in the spacetime from which both outgoing and in
going wavefronts necessarily converge. The boundary of the trapped region
in a spherically symmetric spacetime is given by the equation F = R, which
describes the apparent horizon for the spacetime. The spacetime region where
the mass function F satisfies F < R is not trapped, while F > R describes
a trapped region.
In terms of the scale variable a, the mass function can be written as
 " #
3 1 3 1 2 2
F =r a Φ(a) ,a ȧ + V (Φ(a)) . (5.71)
3 2
242 Final fate of a massive star

This is because from the above Einstein equation for density, the mass
function can be solved as
1
F = ρ(t)R3 . (5.72)
3
From (5.72), it can be seen that

F 1
= ρ(t)r2 a2 . (5.73)
R 3
The above relation decides the trapping, or otherwise, in the spacetime as
the collapse develops. The classes of collapse solutions for a scalar field with
potential, and the trapping, or otherwise, as the collapse develops will be
constructed and investigated. This is relevant from the perspective of the
cosmic censorship conjecture in order to understand the development of a
blackhole or a naked singularity as the collapse outcomes.
Towards such a purpose, consider the class of models where, near the
singularity, the divergence of the density is given by ρ(t) ≈ 1/a(t). Then,
using (5.71), the above condition implies, near the singularity,

1 1
Φ(a)2,a ȧ2 + V (Φ(a)) = . (5.74)
2 a
Now, solving the equation of motion (5.69),
√
a
ȧ = − √ . (5.75)
3

The negative sign implies a collapse scenario where ȧ < 0. Using (5.75) and
(5.74) in the latter part of (5.68), Φ can be solved as

Φ(a) = − ln a. (5.76)

Note that as the singularity is approached, a → 0 implies Φ → ∞, that is,

the scalar field blows up at the singularity. Finally, using (5.75) and (5.76)
in (5.70), the potential V can be solved as

5
V (Φ) = eΦ . (5.77)
6
Therefore near the singularity,
1 2
ρ(t) ≈ , p(t) ≈ − . (5.78)
a(t) 3a(t)

It is seen that in the limit of approach to the singular epoch (t = ts ) F/R = 0

for all shells and there is no trapped surface developing in the spacetime.
 5.5 Resolution of a naked singularity 243

In the model above the weak energy condition is satisfied as ρ > 0 and
ρ + p > 0, although the pressure would be negative. Also, it should be noted
that the pressure does not have to be negative from the initial epoch, because
the specific behavior of ρ in (5.78) has been required only near the singularity.
It is possible to choose a V (Φ) such that at the initial epoch 1/2Φ̇2 > V (Φ),
and then the pressure would be positive. But, near the singularity, V (Φ)
would behave according to (5.78), and hence the pressure should decrease
monotonically from the initial epoch and tend to −∞ at the singularity.
If, from an epoch t = t∗ (or equivalently for some a = a∗ ) the density
starts growing as a−1 , then integrating (5.75), the singular epoch can be
obtained as √
ts = t∗ + 2 3a∗ . (5.79)
Therefore, the collapse reaches the singularity in a finite comoving time,
where the matter energy density as well as the Kretschman scalar κ =
Rijkl Rijkl diverge. Note that from the equation of motion (3.31) it follows
that the metric function a is given by
 2
√ 1 ∗
a(t) = a∗ − √ (t − t ) . (5.80)
2 3
This completes the interior solution within the collapsing cloud, thus giving
the required construction.
It can be seen from the above considerations that the absence, or otherwise,
of the trapped surfaces and the behavior of the pressure crucially depend on
the rate of divergence of the density ρ near the singularity. To examine this
more carefully, near the singularity set ρ = a−n with n > 0, as it is known
that ρ(t) must diverge as a(t) goes to zero in the limit of approach to the
singularity. In this case, solving the Einstein equations gives

(n − 3) −n
p= a . (5.81)
3
The corresponding values of Φ and V (Φ) are
√  n  √nΦ
Φ = − n ln(a), V (Φ) = 1 − e . (5.82)
6
Again, calculating F/R in this general case,

F 1
= a2−n . (5.83)
R 3
Therefore, it can be seen that for low enough divergences (0 < n < 2) no
trapped surfaces form and there are negative pressures near the singularity.
For 2 ≤ n < 3, trapped surfaces do form, but the pressure still remains
244 Final fate of a massive star

negative at the singularity. For n ≥ 3, p ≥ 0 and there are trapped surfaces

forming in the spacetime as the collapse advances in time. Conversely, a non-
negative pressure always ensures trapped surfaces in a homogeneous scalar
field collapse. Therefore, the role of the chosen potential V (Φ) is to control
the divergence of the density near the singularity, which in turn governs the
development, or otherwise, of trapped surfaces. Equation (5.82) shows the
behavior of the functions V (Φ) with respect to Φ, for different values of n.
It is seen that the naked singularity arises from a non-zero measure open set
of initial conditions (n < 2), whereas the rest of the initial data set produces
a blackhole as the final endstate of the collapse.
In the above class of solutions, the behavior is described mainly near the
singularity. However, note that an exact solution is given, and the analysis
holds even if right from initial epoch the density behavior is as above. This
gives a global exact solution where the weak energy condition is satisfied.
As such, the treatment holds for any general class of solutions where near
singularity behavior is as described.
To complete the model, this interior spacetime needs to be matched to an
exterior spacetime. For the required matching, the Israel–Darmois conditions
are used, where the first and second fundamental forms (the metric coeffi-
cients and the extrinsic curvature respectively) are matched at the boundary
of the cloud. Whereas the procedures used below are standard, the particular
case treated here is described in some detail in order to give the picture of
the emerging overall collapse scenario. A useful fact is that since the second
fundamental form Kij is matched, there is no surface stress–energy or surface
tension at the boundary (Israel, 1966a, 1966b; Mazur and Mottola, 2004).
The spherical ball of a collapsing scalar field is matched to a generalized
Vaidya exterior geometry (Wang and Wu, 1999; Joshi and Dwivedi, 1999) at
the boundary hypersurface Σ given by r = rb . Then the metric just inside
Σ is
 
ds2− = −dt2 + a2 (t) dr2 + rb2 dΩ2 , (5.84)
while the metric in the exterior of Σ is
 
2M (rv , v )
ds+ = − 1 −
2
dv 2 − 2 dv drv + rv2 dΩ2 , (5.85)
rv

where v is the retarded null coordinate and rv is the Vaidya radius. Matching
the area radius at the boundary,

rb a(t) = rv (v ). (5.86)

Then on the hypersurface Σ, the interior and exterior metrics are given by

ds2Σ− = −dt2 + a2 (t)rb2 dΩ2 (5.87)

 5.5 Resolution of a naked singularity 245

and  
2M (rv , v ) drv
ds2Σ+ =− 1− +2 dv 2 + rv2 dΩ2 . (5.88)
rv dv
Matching the first fundamental form on this hypersurface,
 
dv 1
=  , (rv )Σ = rb a(t). (5.89)
dt Σ 1− 2M (rv ,v)
+ 2 dr v
rv dv

To match the second fundamental form (extrinsic curvature) for interior and
exterior metrics, note that the normal to the hypersurface Σ, as calculated
from the interior metric, is given by ni− = [0, a(t)−1 , 0, 0], and the non-
vanishing components of the normal derived from the generalized Vaidya
metric are
 −1/2
2M (rv , v ) drv
n+ = − 1 −
v
+2 , (5.90)
rv dv

2M (rv , v ) drv
1− +
rv dv
nr+v =  . (5.91)
2M (rv , v ) drv
1− +2
rv dv
Here the extrinsic curvature is defined as
1
Kab = Ln gab , (5.92)
2

that is, the second fundamental form is the Lie derivative of the metric with
respect to the normal vector n. The above expression is equivalent to

1
Kab = [gab,c nc + gcb nc ,a + gac nc ,b ]. (5.93)
2
 − 
Setting Kθθ − Kθθ
+
Σ
= 0 on the hypersurface Σ,

2M (rv , v ) drv
1− +
rv dv
rb a(t) = rv  . (5.94)
2M (rv , v ) drv
1− +2
rv dv

Simplifying the above using (5.89) and (5.69) on the boundary, one gets

F (t, rb ) = 2M (rv , v ). (5.95)

246 Final fate of a massive star

Using the above equation and (5.89),

 
dv 1 + rb ȧ
= (5.96)
dt Σ F (t, rb )
1−
rb a(t)

Finally, setting [Kτ−τ − Kτ+τ ]Σ = 0, where τ is the proper time on Σ,

M (rv , v ),rv = F/2rb a + rb2 aä. (5.97)

The above equations, together with (5.94), completely specify the matching
at the boundary of the collapsing scalar field.
It is known (see for example, Wang and Wu, 1999), that a generalized
Vaidya spacetime describes the matter that is a combination of matter fields
of Type I and Type II. Specifically, the energy–momentum tensor Tik for
(n)
the matter can be written as a linear superposition of two tensors T ik and
(m)
T ik given by
(n)
T ik = µli lk , (5.98)
(m)
T ik = (ρ + P )(li nk + ni lk ) + P gik . (5.99)

Here, lk and nk are null vectors defined by

 
1 2M
0
lk = δ k , nk = 1− δ0k + δ1k . (5.100)
2 rv
(n)
The tensor T ik represents the matter field that moves along the null hyper-
surface v = const. The other tensor describes the matter moving out along
timelike trajectories. The matching conditions uniquely√describe
√ the√quanti-
ties µ and ρ on the matching surface, given by µ = (2/9 a)( 3+ rb a) and
ρ = (2/3a). The pressure component is determined by the specific choice of
mass function satisfying the matching equations. It can be seen that these
quantities are well-defined everywhere, except at the singularity, hence it can
be deduced that the exterior spacetime has no source or singularities.
The above provides the full spacetime. However, one could also consider
the fields that asymptotically fall off to vanishing values, rather than match-
ing the collapsing cloud to an exterior metric such as above. In this case,
suitable fall off conditions which are not discussed in details here, need to
be prescribed. Although the matching used above is in a somewhat general
setting, particular examples which form subcases have been discussed ear-
lier. These serve to show that the set of all functions M (v, rv ) that satisfies
(3.102), is non-empty. Such examples include the charged Vaidya space-
time M = M (v ) + Q(v )/rv and also the anisotropic de-Sitter spacetime
 5.5 Resolution of a naked singularity 247

M = M (rv ), which are two different solutions of (3.102) (see for exam-
ple, Joshi and Dwivedi, 1999; Wang and Wu, 1999; Giambo, 2005). These
give two unique exterior spacetimes, both of which are subclasses of the
generalized Vaidya geometry described above.
Now it can be seen that at the singular epoch t = ts , 2M (rv , v )/rv →
0. Therefore the exterior metric around the singularity smoothly trans-
forms to

ds2 = −dv 2 − 2 dv drv + rv2 dΩ2 . (5.101)

This describes a Minkowski spacetime in retarded null coordinates. Hence, it

can be seen that the exterior generalized Vaidya metric, together with the sin-
gular point at (ts , 0), can be smoothly extended to the Minkowski spacetime
as the collapse completes. From (3.101), it can be seen that the trajectory
that emerges from the singularity (before it evaporates into free space) in the
generalized Vaidya geometry is a null geodesic. It follows that non-spacelike
trajectories can emerge from the singularity that develops as the collapse end
point. Hence, a naked singularity is produced in the collapse of the scalar
field with potential for a non-zero measure set of initial conditions, and the
occurrence of trapped surfaces in the spacetime is avoided.
An interesting observation to be made is that the null geodesic from the
singularity lies completely in the spacetime that is exterior to the collapsing
cloud. This is different, for example, from the globally naked singularities
developing as collapse endstates in the TBL dust collapse models, where the
null geodesic from the singularity starts within the interior spacetime and
then crosses the boundary of the collapsing cloud to emerge into the exterior.
In the present case, the complete interior spacetime collapses simultaneously
at the epoch t = ts and no future directed non-spacelike geodesic from the
singularity can lie in the interior of the cloud. Also, it should be pointed out
that the proof that all simultaneous singularities in non-extendible space-
times must be covered, does not hold in this case, as the spacetime here
(together with the singularity) is extendible to a Minkowski spacetime when
the collapse completes.
Quantum effects in the gravitational collapse of the scalar field model
considered above, which classically leads to a naked singularity, are now
discussed. It is seen that non-perturbative semi-classical modifications near
the singularity, based on loop quantum gravity, give rise to a strong out-
ward flux of energy. This leads to the dissolution of the collapsing cloud
before the singularity can form. Quantum gravitational effects thus resolve
the naked singularity by avoiding their formation. Furthermore, quantum
gravity induced mass flux has a distinct feature that may lead to a novel
observable signature in astrophysical bursts. The visible singularities, pre-
dicted by classical general relativity as gravitational collapse endstates, can,
248 Final fate of a massive star

in principle, be directly observed by an external observer, unlike their black-

hole siblings. Hence, it is of interest to examine these from such a perspective
as there are many classes where, given the initial density and pressure profiles
for a matter cloud, the collapse evolution leads to naked singularity forma-
tion, subject to an energy condition and astrophysically reasonable equations
of state.
Since singularities originate in the regime where the classical general rel-
ativity is expected to be replaced by quantum gravity, whether a quantum
theory of gravity could resolve their formation has remained an outstanding
problem. Also, with the lack of observable signatures from the Planck regime,
naked singularities could, in fact, be a boon for a quantum theory of gravity.
Because, the singularity is visible, any quantum gravitational signature orig-
inating in the ultra-high curvature regime near a classical singularity can, in
principle, be observed, thus providing a rare test for quantum gravity.
One of the non-perturbative quantizations of gravity is given by the loop
quantum gravity formalism (see for example, Ashtekar, 1986, 1991; Ashtekar
and Lewandowski, 2004). Its key predictions include the Bekenstein–Hawking
entropy formula (Ashtekar et al., 1998), and its application to symmetry
reduced mini-superspace quantization of homogeneous spacetimes is called
loop quantum cosmology (Bojowald, 2001, 2002, 2005), which has applica-
tions towards the resolution of the big-bang singularity, initial conditions for
inflation (Tsujikawa, Singh, and Maartens, 2004), and possible observable
signatures in cosmic microwave background radiation.
As the dynamics of a generic collapse would be complex, it is useful to work
here with a collapse scenario such as that of a scalar field. It serves as a good
toy model to gain insights into the role of quantum gravity effects closer to
a naked singularity, at the late stages of a gravitational collapse. One of the
simplest settings is to consider an initial configuration of a homogeneous and
isotropic scalar field Φ = Φ(t) with a potential V (Φ), as discussed above,
and the canonical momentum PΦ . In this case, as seen above, the fate of
the singularity being naked or covered depends on the rate of the gravita-
tional collapse. For an appropriately chosen potential, formation of trapped
surfaces can be avoided even as the collapse progresses, resulting in a naked
singularity with an outward energy flux, which would be, in principle, observ-
able. Since the interior of the homogeneous scalar field collapse is described
by a Friedmann–Robertson–Walker (FRW) metric, techniques of loop quan-
tum cosmology can be used to investigate the way in which quantum gravity
modifies the collapse.
Consider the classical collapse of a homogeneous scalar field Φ(t) with
potential V (Φ) and the canonical momentum for the marginally bound (k =
0) case. The interior metric is given by
 
ds2 = −dt2 + a2 (t) dr2 + r2 dΩ2 , (5.102)
 5.5 Resolution of a naked singularity 249

with the classical energy density and pressure of the scalar field of

ρ(t) = Φ̇2 /2 + V (Φ), p(t) = Φ̇2 /2 − V (Φ) . (5.103)

The dynamical evolution of the system is obtained from the Einstein

equations that yield

Ṙ2 R = F (t, r), ρ = F,r /κaR2 , p = −Ḟ /κR2 Ṙ. (5.104)

Here, κ = 8πG, and F (t, r) = (κ/3)ρ(t)r3 a3 has the interpretation of the

mass function of the collapsing cloud, with F ≥ 0, and R(t, r) = ra(t) is the
area radius of a shell labeled by the comoving coordinate r. In a continual
collapse, the area radius of a shell at a constant value of the comoving radius
r decreases monotonically. The spacetime region is trapped, or otherwise,
depending on the value of the mass function. If F is greater (or less) than R,
the region is trapped (or untrapped). The boundary of the trapped region
is given by F = R. The collapsing interior can be matched at some suitable
boundary r = rb to a generalized Vaidya exterior geometry, as discussed
earlier.
The form of the potential that leads to a naked singularity is determined
by writing the energy density of the scalar field in a generic form as ρ =
ln−4 a−n , where n > 0 and l is a proportionality constant. Using the energy
conservation equation, this leads to the pressure p = [(n − 3)/3] ln−4 a−n .
On substituting (5.103) in these (Goswami and Joshi, 2004),
 √
Φ = − n/κ ln a, V (Φ) = (1 − n/6)ln−4 e κn Φ . (5.105)

Then it is easily seen that F/R = (κ/3)ln−4 a2−n r2 . Therefore, in the collaps-
ing phase as a → 0, whether or not the trapped surfaces form is determined
by the value of n. As discussed above, for 0 < n < 2, if no trapped surfaces
exist initially then no trapped surfaces would form until the epoch a(t) = 0
√ 2/n
with a(t) = 1 − n t/2 3 .
The absence of trapped surfaces is accompanied by a negative pressure,
implying that for a constant value of the comoving coordinate r, Ḟ is nega-
tive and so the mass contained in the cloud of that radius keeps decreasing.
This leads to a classical outward energy flux. As the collapse proceeds, the
scale factor vanishes in finite time and the physical densities blow up, leading
to a naked singularity. Since no trapped surfaces form during the collapse,
the outward energy flux will, in principle, be observable. However, near the
singularity, when the energy density is close to Planckian values, this clas-
sical picture has to be modified and the scenario incorporating quantum
gravity modifications into classical dynamics needs to be investigated. A
non-perturbative semi-classical modification, based on loop quantum grav-
ity for the interior, was considered by Goswami, Joshi, and Singh (2006),
250 Final fate of a massive star

Singularity

Quantum gravity epoch

t = a*

Classical epoch
t =0

Fig. 5.2 Classical and quantum epochs in gravitational collapse.

which is discussed below in some detail. The underlying geometry for the
FRW spacetime in loop quantum cosmology is discrete and both the scale
factor and the inverse scale factor
 operators have discrete eigenvalues. In
particular, a critical scale a∗ = jγ/3P exists below which the eigenval-
ues of the inverse scale factor become proportional to the positive powers
of the scale factor (see Fig. 5.2). Here γ ≈ 0.2375 is the Barbero–Immirzi
parameter, P is the Planck length, and j is a half-integer free parameter
that arises because the inverse scale factor operator is computed by tracing
over SU(2) holonomies in an irreducible spin j representation. The value
of this parameter is arbitrary and is constrained only by phenomenological
considerations.
The change in the behavior of the classical geometrical density (1/a3 ) for
scales a ∼ a∗ can be approximated by

dj (a) = D(q ) a−3 , q := a2 /a2∗ , a∗ := jγ/3 P , (5.106)

with
$  
6
D(q ) = (8/77) q 3/2 7 (q + 1)11/4 − |q − 1|11/4
 %6
− 11q (q + 1)7/4 − sgn (q − 1)|q − 1|7/4 . (5.107)

For a  a∗ , dj ∝ (a/a∗ )15 a−3 , and for a  a∗ it behaves classically with

dj ≈ a−3 . The scale at which transition to the behavior of the geometrical
density takes place is determined by the parameter j .
At the fundamental level, the dynamics in the loop quantum regime is
discrete, however, recent investigations pertaining to the evolutionof coher-
√
ent states showed that for scales a0 = γP ∼ a ∼ a∗ = jγ/3P ,
dynamics can be described by modifications to the Friedmann dynamics on
a continuous spacetime with the modified matter Hamiltonian,

HΦ = dj (a) PΦ2 /2 + a3 V (Φ), (5.108)

 5.5 Resolution of a naked singularity 251

and the modified Friedmann equation,

ȧ2 /a2 = (κ/3)(Φ̇2 /2D + V (Φ)), (5.109)

which is obtained by the vanishing of the total Hamiltonian constraint and

the Hamilton equations, which are Φ̇ = dj (a)PΦ , ṖΦ = −a3 V,Φ (Φ). These
also lead to the modified Klein–Gordon equation,

Φ̈ + 3ȧ/a − Ḋ(q )/D(q ) Φ̇ + D(q ) V,Φ (Φ) = 0 . (5.110)

Since, at classical scales (a  a∗ ) D ≈ 1, the modified dynamical equations

reduce to the standard Friedmann dynamical equations. For scales a ∼ a∗ ,
the Φ̇ term acts like a frictional term for a collapsing phase.
Note that since semi-classical modifications for the inhomogeneous case
are still not known, a complete quantum analysis of the interior and exterior
cannot be carried out, and the exterior is assumed to remain classical. In
any case, the quantum effects are supposed to be important mainly inside
the cloud, which is collapsing and facing a singular fate. Also, as a contin-
uous spacetime can be approximated to a scale factor a0 , the matching of
the interior and exterior spacetimes remains valid during the semi-classical
evolution.
The modified energy density and pressure of the scalar field in the semi-
classical regime can be similarly obtained from the eigenvalues of the density
operator, and by using the stress–energy conservation equation

ρeff = dj (a) HΦ = Φ̇2 /2 + D(q ) V (Φ), (5.111)

and
 
2 1 Ḋ(q ) Φ̇2 Ḋ(q )
peff = 1− − D(q ) V (Φ) − V (Φ) . (5.112)
3 (ȧ/a) D(q ) 2 3(ȧ/a)

It is then straightforward to check that peff is generically negative for a ∼ a∗ ,

and for a  a∗ it becomes very strong. For example, at a ∼ a0 , peff ≈ −9ρeff .
This is much stronger than its classical counterpart p = [(n − 3)/3] ρ with
0 < n < 2. Therefore, a strong burst of outward energy flux can be expected
in the semi-classical regime. Furthermore, for a  a∗ , D(q )  1 and the
Klein–Gordon equation yields Φ̇ ∝ a12 . Hence, from (5.111) it can be easily
seen that the effective density, instead of blowing up, becomes extremely
small and remains finite.
The modified mass function of the collapsing cloud can be evaluated
using (3.31) and (5.109),

F = (κ/3)(d−1 2 3 3
j Φ̇ /2 + a V (Φ)) r . (5.113)
252 Final fate of a massive star

In the regime a ∼ a0 , d−1 2 12

j Φ̇ becomes proportional to a , the potential term
becomes negligible, and thus the mass function becomes vanishingly small
at small scale factors.
The picture emerging from the loop quantum modifications to the collapse
is then as follows. First, before the area radius of the collapsing shell reaches
R∗ = ra∗ at t = t∗ , the collapse proceeds as per classical dynamics, and
as smaller scale factors are approached Φ̇ and the energy density ρ ∝ a−n
increase. The mass function is proportional to an−3 and (as 0 < n < 2) it
decreases with decreasing scale factor, so there is a mass loss to the exterior,
which is also explained by the existence of a negative classical pressure. Next,
as the collapsing cloud reaches R∗ , the geometric density classically given by
a−3 , modifies to dj and the dynamics is governed by the modified Friedmann
and Klein–Gordon equations. The scalar field that experienced anti-friction
in the classical regime, now experiences friction leading to a decrease in Φ̇.
Finally, the slowing down of Φ decreases the rate of the collapse, and the
formation of the singularity is delayed. Eventually, when the scale factor
becomes less than a0 , this leads to a breakdown of the continuum space-
time approximation and semi-classical dynamics. Discrete quantum geometry
emerges at this scale and the dynamics can only be described by the quan-
tum difference equation. The naked singularity is thus avoided until the scale
factor at which a continuous spacetime exists.
If the evolution of the area radius in time as the collapse proceeds is con-
sidered, the semi-classical evolution closely follows the classical trajectory
until the time t∗ . Within a finite time after t∗ , the classical collapse leads to
a vanishing R and a naked singularity. However, the area radius never van-
ishes in the loop modified semi-classical dynamics, and the naked singularity
does not form as long as the continuum spacetime approximation holds. The
evolution of the energy density in Planck units can be examined, and while
the energy density would blow up classically, it remains finite and in fact
decreases in the semi-classical regime.
The phenomena of delay and avoidance of the naked singularity in contin-
uous spacetime is accompanied by a burst of matter to the exterior. If the
mass function at scales a  a∗ is Fi and its difference with a mass of the
cloud for a < a∗ is ∆F = Fi − F , then the mass loss can be computed as
 
∆F ρeff d−1
j
= 1 − n−4 3−n . (5.114)
F (ai ) l ai

For a < a∗ , as the scale factor decreases, the energy density and mass in the
interior decrease and the negative pressure increases strongly. This leads to a
strong burst of matter. The absence of trapped surfaces enables the quantum
gravity induced burst to propagate via the generalized Vaidya exterior to an
observer at infinity.
 5.5 Resolution of a naked singularity 253

As for the evolution of the mass function in the semi-classical regime,

∆F /Fi approaches unity very rapidly. This feature is independent of the
choice of parameter j . The choice of potential causes mass loss to the exterior
in the classical collapse also, but it is much smaller and, in any case, the
classical description cannot be trusted at an energy density greater than
the Planck value, when the quantum effects as described above must be
considered. Basically, it can be seen that the loop quantum evolution leads
to the dissolution of all the mass of the collapsing shell.
Interestingly, for a given collapsing configuration, the scale at which the
strong outward flux initiates depends on the loop parameter j that controls
a∗ . If j is large, then the burst occurs at an earlier area radius and vice versa.
For all such choices, ∆F /Fi → 1, but the outgoing flux profiles change. The
loop quantum burst has a distinct signature in that at a ∼ a∗ the flux
decreases for a short period and then rapidly increases. Since the causal
structure of classical spacetime is such that trapped surface formation is
avoided, this quantum gravitational signature can be, in principle, observed
by an external observer as a slight dimming and subsequent brightening of
the collapsing star. This peculiar phenomena is directly related to the peak in
the function dj (a), and depends solely on the value of the parameter j . If this
is compared with the other phenomenological applications mentioned above,
the effect could not be masked by the role of other loop quantum parameters
in a more general setting. This phenomena is possibly a direct probe to
measure j , and an observer can estimate this loop quantum parameter by
observing the flux profile of the burst, based on this mechanism, and by
measuring the variation in luminosity of the collapsing cloud.
During such a burst, most of the mass of the cloud is ejected and this
may dissolve the singularity. Therefore, non-perturbative semi-classical mod-
ifications may not allow the formation of the naked singularity as the
collapsing cloud evaporates away due to super-negative pressures in the
late regime, as generated by the quantum effects. Such super-negative
pressures would also exist for arbitrary matter configurations, which indi-
cates that the results such as above could hold in a more general setting
as well.
In this sense, loop quantum effects then imply a quantum gravitational
cosmic censorship, alleviating the naked singularity problem. Note that the
semi-classical effects do not show that the singularity is absent, it is only
avoided to scale factor a0 , below which the semi-classical dynamics and
matching may break down. If, for a given choice of initial data, semi-classical
dynamics is unable to dissolve the singularity completely, the final fate of the
naked singularity must be decided by using full quantum evolution. Even
in such cases, there are valuable insights from semi-classical loop quan-
tum effects, with the possibility of phenomenologically constraining the j
parameter.
254 Final fate of a massive star

In the toy model considered above, it was shown that the classical outcome
and evolution of the collapse is radically altered by the non-perturbative
modifications to the dynamics, as induced by the quantum effects. These
considerations are, of course, within the mini-superspace setting, and the
general case of inhomogeneities and anisotropies remains open. However, the
possibility of such observable signatures in astrophysical bursts originating
from the quantum gravity regime near the spacetime singularity is indeed
intriguing, indicating that gravitational collapse scenarios can be used as
probes to test quantum gravity models.
 References

Abraham, A. and Evans, C. R. (1993). ‘Critical behavior and scaling in vacuum

axisymmetric gravitational collapse’. Phys. Rev. Lett., 70, 2980.
Alcubierre, M., Gonzalez, J., Salgado, M. and Sudarsky, D. (2004). ‘The cosmic
censor conjecture: is it generically violated?’, gr-qc/0402045.
Arnett, W. D. and Bowers, R. L. (1977). ‘A microscopic interpretation of neutron
star structure’. Astrophys. J. Suppl., 33, 415.
Ashtekar, A. (1986). ‘New variables for classical and quantum gravity’. Phys.
Rev. Lett., 57, 2244.
Ashtekar, A. (1991). ‘Old problems in the light of new variables’. In Concep-
tual Problems of Quantum Gravity, ed. A. Ashtekar and J. Stachel. Boston:
Birkhauser.
Ashtekar, A. and Lewandowski, J. (2004). ‘Background independent quantum
gravity: a status report’. Class. Quant. Grav., 21, R53.
Ashtekar, A., Pawlowski, T. and Singh, P. (2006). ‘Quantum nature of the big
bang: an analytical and numerical investigation’, gr-qc/0604013.
Ashtekas, A., Baez, J., Corichi, A. and Krasanou, K. (1998). ‘Quantum geometry
and black hole entropy’. Phys. Rev. Lett., 80, 904.
Avez, A. (1963). ‘Essais de geometrie Riemannienne hyperbolique globale.
Applications à la relativité générale’. Ann. Inst. Fourier (Grenoble), 132, 105.
Banerjee, A., Debnath, U. and Chakraborty, S. (2003). ‘Naked singularities in
higher dimensional gravitational collapse’. Int. J. Mod. Phys. D., 12, 1255.
Bardeen, J. M., Carter, B. and Hawking, S. W. (1973). ‘The four laws of black
hole mechanics’. Commun. Math. Phys., 31, 161.
Bekenstein, J. D. (1973). ‘Black holes and entropy’. Phys. Rev. D., 7, 2333.
Birkhoff, G. D. (1923). Relativity and Modern Physics. Cambridge, MA: Harvard
University Press.
Bishop, R. L. and Critendon, R. J. (1964). Geometry of Manifolds. New York:
Academic Press.
Blandford, R. D. (1987). ‘Astrophysical black holes’. In Three Hundred Years
of Gravitation, eds S. W. Hawking and W. Israel. Cambridge: Cambridge
University Press.

255
256 References

Blandford, R. D. and Thorne, K. (1979). ‘Black hole astrophysics’. In General

Relativity: An Einstein Centenary Survey, eds S. W. Hawking and W. Israel.
Cambridge: Cambridge University Press.
Bojowald, M. (2001). ‘Absence of a singularity in loop quantum cosmology’.
Phys. Rev. Lett., 86, 5227.
Bojowald, M. (2002). ‘Inflation from quantum geometry’. Phys. Rev. Lett., 89,
261301.
Bojowald, M. (2005). ‘Loop quantum cosmology’. Living Rev. Rel., 8, 11.
Bondi, H. (1948). ‘Spherically symmetric models in general relativity’. Mon.
Not. R. Astron. Soc., 107, 400.
Bondi, H., van der Burgh, M. G. J. and Metzner, R. (1962). ‘Gravitational
waves in general relativity VII: waves from axi-symmetric isolated systems’.
Proc. Roy. Soc. Lond. A., 269, 21.
Bonnor, W. B. (1976). ‘Non-radiative solutions of Einstein’s equations for dust’.
Commun. Math. Phys., 51, 191.
Brady, P. R. (1995a). ‘Analytic example of critical behaviour in scalar field
collapse’. Class. Quant. Grav., 11, 1255.
Brady, P. R. (1995b). ‘Self-similar scalar field collapse: naked singularities and
critical behavior’. Phys. Rev. D., 51, 4168.
Cahill, M. E. and Taub, A. H. (1971). ‘Spherically symmetric similarity solutions
of the Einstein field equations for a perfect fluid’. Commun. Math. Phys.,
21, 1.
Carr, B. J. and Coley, A. (1999). ‘Self-similarity in general relativity’. Class.
Quant. Grav., 16, R31.
Carter, B. (1971). ‘Causal structure in space-time’. Gen. Relat. Grav., 1, 349.
Celerier, M. and Szekeres, P. (2002). ‘Timelike and null focusing singularities
in spherical symmetry: a solution to the cosmological horizon problem and a
challenge to the cosmic censorship hypothesis’. Phys. Rev. D., 65, 123516.
Chandrasekhar, S. (1931). ‘The maximum mass of ideal white dwarfs’. Astrophys.
J., 74, 81.
Chandrasekhar, S. (1934). ‘Stellar configurations with degenerate cores’. Obser-
vatory, 57, 373.
Chandrasekhar, S. and Hartle, J. (1983). ‘On crossing the Cauchy horizon of a
Reissner: Nordström black hole’. Proc. Roy. Soc. Lond. A., 384, 301.
Choptuik, M. W. (1993). ‘Universality and scaling in gravitational collapse of a
massless scalar field’. Phys. Rev. Lett., 70, 9.
Christodoulou, D. (1984). ‘Violation of cosmic censorship in the gravitational
collapse of a dust cloud’. Commun. Math. Phys., 93, 171.
 References 257

Christodoulou, D. (1986). ‘Global existence of generalized solutions of the spheri-

cally symmetric Einstein scalar equations in the large’. Commun. Math. Phys.,
106, 587.
Christodoulou, D. (1994). ‘Examples of naked singularity formation in the
gravitational collapse of a scalar field’. Ann. Maths., 140, 607.
Christodoulou, D. (1999). ‘The instability of naked singularities in the gravita-
tional collapse of a scalar field’. Ann. Maths., 149, 183.
Clarke, C. J. S. (1986). ‘Singularities: global and local aspects’. In Topological
Properties and Global Structure of Space-time, eds P. G. Bergmann and V. de
Sabbata. New York: Plenum Press.
Clarke, C. J. S. (1993). The Analysis of Spacetime Singularities. Cambridge:
Cambridge University Press.
Clarke, C. J. S. and Joshi, P. S. (1988). ‘On reflecting space-times’. Class. Quant.
Grav., 5, 19.
Clarke, C. J. S. and Królak, A. (1986). ‘Conditions for the occurrence of strong
curvature singularities’. J. Geo. Phys., 2, 127.
Dafermos, M. (2004). ‘On naked singularities and the collapse of self-gravitating
Higgs fields’, gr-qc/0403033.
Datt, B. (1938). ‘Über eineklasse von lösungen der gravitationsgleichungen der
relativität’. Z. Physik., 108, 314.
Debnath, U., Chakraborty, S. and Barrow, J. D. (2004). ‘Quasispherical
gravitational collapse in any dimension’. Gen. Relat. Grav., 36, 231.
de la Cruz, V., Chase, J. E. and Israel, W. (1970). ‘Gravitational collapse with
symmetries’. Phys. Rev. Lett., 24, 423.
Deshingkar, S. and Joshi, P. S. (2001). ‘Structure of nonspacelike geodesics in
dust collapse’. Phys. Rev. D., 63, 024007.
Deshingkar, S. S. , Joshi, P. S. and Dwivedi, I. H. (1999). ‘Physical nature of the
central singularity in spherical collapse’. Phys. Rev. D., 59, 044018.
Deshingkar, S., Joshi, P. S. and Dwivedi, I. H. (2002). ‘Appearance of the central
singularity in spherical collapse’. Phys. Rev. D., 65, 084009.
Deshingkar, S. S., Chamorro, A., Jhingan, S. and Joshi, P. S. (2001).
‘Gravitational collapse and cosmological constant’. Phys. Rev. D., 63, 124005.
Deutsch, D. and Candelas, P. (1980). ‘Boundary effects in quantum field theory’.
Phys. Rev. D., 20, 3063.
Dixon, W. G. (1970). ‘Dynamics of extended bodies in general relativity:
momentum and angular momentum’. Proc. Roy. Soc. Lond. A., 314, 499.
Doroshkevich, A. G., Zel’dovich, Ya. B. and Novikov, I. (1966). ‘Gravitational
collapse of non-symmetric non-rotating masses’. Sov. Phys. JETP, 22, 122.
258 References

Dwivedi, I. H. and Joshi, P. S. (1989). ‘On the nature of naked singularities in

Vaidya space-times’. Class. Quant. Grav., 6, 1599.
Dwivedi, I. H. and Joshi, P. S. (1991). ‘On the nature of naked singularities in
Vaidya space-times II’. Class. Quant. Grav., 8, 1339.
Dwivedi, I. H. and Joshi, P. S. (1992). ‘Cosmic censorship violation in non-self-
similar Tolman–Bondi models’. Class. Quant. Grav., 9, L69.
Eardley, D. M. and Smarr, L. (1979). ‘Gravitational collapse of dust spheres:
time coordinate conditions’. Phys. Rev. D., 19, 2239.
Ellis, G. F. R. and King, A. R. (1974). ‘Was the big bang a whimper?’. Commun.
Math. Phys., 38, 119.
Ellis, G. F. R. and Schmidt, B. (1977). ‘Singular space-times’. Gen. Relat. Grav.,
8, 915.
Evans, C. R. and Coleman, J. S. (1994). ‘Critical phenomena and self-similarity
in the gravitational collapse of radiation fluid’. Phys. Rev. Lett., 72, 1782.
Finkelstein, D. (1958). ‘Past-future asymmetry of the gravitational field of a
point particle’. Phys. Rev., 110, 965.
Fock, V. (1939). ‘Sur le mouvement des masses finies d’aprés la théorie
gravitation einsteinienne’. J. Phys. USSR., 1, 81.
Foglizzo, T. and Henriksen, R. (1993). ‘General relativistic collapse of homothetic
ideal gas spheres and planes’. Phys. Rev. D., 48, 4645.
Friedman, J., Morris, M. S., Novikov, I. D., Echeverria, F., Klinkhammer, G.
and Thorne, K. S. (1990). ‘Cauchy problem in space-times with closed timelike
curves’. Phys. Rev. D., 42, 1915.
Garfinkle, D. (2004). ‘Numerical simulation of a possible counterexample to
cosmic censorship’, gr-qc/0403078.
Geroch, R. (1967). ‘Topology in general relativity’. J. Math. Phys., 8, 782.
Geroch, R. (1968a). ‘What is a singularity in general relativity?’. Ann. Phys.,
48, 526.
Geroch, R. (1968b). ‘Spinor structure of space-times in general relativity I’. J.
Math. Phys., 9, 1739.
Geroch, R. (1970a). ‘Domain of dependence’. J. Math. Phys., 11, 437.
Geroch, R. (1970b). ‘Singularities’. In Relativity, eds S. Fickler, M. Carmeli and
L. Witten. New York: Plenum Press.
Geroch, R. (1971). ‘Space-time structure from a global view point’. In General
Relativity and Cosmology, Proceedings of the International School in Physics
‘Enrico Fermi’, ed. R. K. Sachs. New York: Academic Press.
Geroch, R. and Horowitz, G. (1979). ‘Global structure of space-times’. In General
Relativity: An Einstein Centenary Survey, eds S. W. Hawking and W. Israel.
Cambridge: Cambridge University Press.
 References 259

Geroch, R., Kronheimer, E. and Penrose, R. (1972). ‘Ideal points in space-time’.

Proc. Roy. Soc. Lond. A., 327, 545.
Ghosh, S. G. and Banerjee, A. (2003). ‘Non-marginally bound inhomogeneous
dust collapse in higher dimensional spacetime’. Int. J. Mod. Phys. D., 12, 639.
Ghosh, S. G. and Beesham, A. (2001). ‘Higher dimensional inhomogeneous dust
collapse and cosmic censorship’. Phys. Rev. D., 64, 124005.
Ghosh, S. G. and Deshkar, D. W. (2003). ‘Gravitational collapse of perfect fluid
in self-similar higher dimensional space-times’. Int. J. Mod. Phys. D., 12, 913.
Giambo, R. (2005). ‘Gravitational collapse of homogeneous scalar fields’. Class.
Quant. Grav., 22, 2295.
Giambo, R. (2006). ‘Global visibility of naked singularities’, gr-qc/0603120.
Giambo, R., Giannoni, F., Magli, G. and Piccione, P. (2003). ‘New solutions
of Einstein equations in spherical symmetry: the cosmic censor to the court’.
Commun. Math. Phys., 235, 545.
Giambo, R., Giannoni, F., Magli, G. and Piccione, P. (2004). ‘Naked singularities
formation in the gravitational collapse of barotropic spherical fluids’. Gen. Rel.
Grav., 36, 1279.
Gödel, K. (1949). ‘An example of a new type of cosmological solution of Einstein’s
field equations of gravitation’. Rev. Mod. Phys., 21, 447.
Goncalves, S. M. C. V. and Jhingan, S. (2001). ‘Singularities in gravitational
collapse with radial pressure’. Gen. Rel. Grav., 33, 2125.
Goswami, R. and Joshi, P. S. (2002). ‘What role do pressures play in determining
the final end state of gravitational collapse?’. Class. Quant. Grav., 19, 5229.
Goswami, R. and Joshi, P. S. (2004a). ‘Spherical dust collapse in higher
dimensions’. Phys. Rev. D., 69, 044002.
Goswami, R. and Joshi, P. S. (2004b). ‘Cosmic censorship in higher dimensions’.
Phys. Rev. D., 69, 104002.
Goswami, R. and Joshi, P. S. (2004c). ‘Naked singularity formation in scalar
field collapse’, gr-qc/0410144.
Goswami, R. and Joshi, P. S. (2006). ‘Spherical gravitational collapse in N-
dimensions’, gr-qc/0608136.
Goswami, R., Joshi, P. S. and Singh, P. (2006). ‘Quantum evaporation of a naked
singularity’. Phys. Rev. Lett., 96, 031302.
Gundlach, C. (1995). ‘Choptuik spacetime as an eigenvalue problem’. Phys. Rev.
Lett., 75, 3214.
Gundlach, C. (1999). ‘Critical phenomena in gravitational collapse’. Living Rev.
Rel., 2, 4.
Gutperle, M. and Krauss, P. (2004). ‘Numerical study of cosmic censorship in
string theory’. hep-th/0402109.
260 References

Hagerdorn, R. (1968). ‘Hadronic matter near the boiling point’. Nuovo Cimento
A., 56, 1027.
Harada, T. (1998). ‘Final fate of the spherically symmetric collapse of a perfect
fluid’. Phys. Rev. D., 58, 104015.
Harada, T. and Maeda, H. (2001). ‘Convergence to a self-similar solution in
general relativistic gravitational collapse’. Phys. Rev. D., 63, 084022.
Harada, T., Iguchi, H. and Nakao, K. I. (2000). ‘Naked singularity explosion’.
Phys. Rev. D., 61, 101502.
Harada, T., Iguchi, H. and Nakao, K. (2002). ‘Physical processes in naked
singularity formation’. Prog. Theor. Phys., 107, 449.
Harrison, B. K., Thorne, K. S., Wakano, M. and Wheeler, J. (1965). Gravitation
Theory and Gravitational Collapse. Chicago: University of Chicago Press.
Hartle, J. (1978). ‘Relativistic stars, gravitational collapse, and black holes’. In
Relativity, Astrophysics and Cosmology, ed. W. Israel. Dordrecht: Reidel.
Hawking, S. W. (1971). ‘Gravitational radiation from colliding black holes’. Phys.
Rev. Lett., 26, 1344.
Hawking, S. W. (1975). ‘Particle creation by black holes’. Commun. Math. Phys.,
43, 199.
Hawking, S. W. and Ellis, G. F. R. (1973). The Large Scale Structure of Space-
time. Cambridge: Cambridge University Press.
Hawking, S. W. and Israel, W. (1979a). General Relativity: An Einstein
Centenary Survey. Cambridge: Cambridge University Press.
Hawking, S. W. and Israel, W. (1979b). ‘An introductory survey’. In General
Relativity: An Einstein Centenary Survey, eds S. W. Hawking and W. Israel.
Cambridge: Cambridge University Press.
Hawking, S. W. and Penrose, R. (1970). ‘The singularities of gravitational
collapse and cosmology’. Proc. Roy. Soc. Lond. A., 314, 529.
Hellaby, J. and Lake, K. (1985). ‘Shell crossings and the Tolman model’.
Astrophys. J., 290, 381.
Herrera, L. and Santos, N. O. (2005). ‘Cylindrical collapse and gravitational
waves’. Class. Quant. Grav., 22, 2407.
Hertog, T., Horowitz, G. and Maeda, K. (2004). ‘Generic cosmic-censorship
violation in anti-de Sitter space’. Phys. Rev. Lett., 92, 131101.
Hicks, N. J. (1965). Notes on Differential Geometry. Princeton, NJ: Van
Nostrand.
Hiscock, W. A., Williams, L. G. and Eardley, D. M. (1982). ‘Creation of particles
by shell-focusing singularities’. Phys. Rev. D., 26, 751.
Hubeny, V., Lio, X., Rangamani, M. and Shenker, S. (2004). ‘Comments on
cosmic censorship in AdS/CFT’, hep-th/0403198.
 References 261

Iguchi, H., Nakao, K. I. and Harada, T. (1998). ‘Gravitational waves around a

naked singularity: odd-parity perturbation of Lemaitre–Tolman–Bondi space-
time’. Phys. Rev. D., 57, 7262.
Israel, W. (1966a). ‘Singular hypersurfaces and thin shells in general relativity’.
Nuovo Cimento B., 44, 1.
Israel, W. (1966b). ‘Singular hypersurfaces and thin shells in general relativity’.
Nuovo Cimento B., 48, 463.
Israel, W. (1984). ‘Does a cosmic censor exist?’. Found. Phys., 14, 1049.
Israel, W. (1986a). ‘Must non-spherical collapse produce black holes? A
gravitational confinement theorem’. Phys. Rev. Lett., 56, 789.
Israel, W. (1986b). ‘The formation of black holes in non-spherical collapse and
cosmic censorship’. Can. J. Phys., 64, 120.
Jang, P. S. and Wald, R. (1977). ‘The positive energy conjecture and the cosmic
censor hypothesis’. J. Math. Phys., 18, 41.
Jhingan, S. and Magli, G. (2000). ‘Black holes versus naked singularities
formation in collapsing Einstein clusters’. Phys. Rev. D., 61, 124006.
Joshi, P. S. (1981). ‘On higher order causality violations’. Phys. Lett. A., 85,
319.
Joshi, P. S. (1993). Global Aspects in Gravitation and Cosmology. Oxford:
Clarendon Press, Oxford University Press.
Joshi, P. S. (2000). ‘Gravitational collapse: the story so far’. Pramana,
55, 529.
Joshi, P. S. and Dwivedi, I. H. (1992). ‘The structure of naked singularity in
self-similar gravitational collapse’. Commun. Math. Phys., 146, 333.
Joshi, P. S. and Dwivedi, I. H. (1993a). ‘Naked singularities in spherically sym-
metric inhomogeneous Tolman–Bondi dust cloud collapse’. Phys. Rev. D., 47,
5357.
Joshi, P. S. and Dwivedi, I. H. (1993b). ‘The structure of naked singularity in
self-similar gravitational collapse II’. Lett. Math. Phys., 27, 235.
Joshi, P. S. and Dwivedi, I. H. (1999). ‘Initial data and the end state of
spherically symmetric gravitational collapse’. Class. Quant. Grav., 16, 41.
Joshi, P. S. and Goswami, R. (2004). ‘Role of initial data in spherical collapse’.
Phys. Rev. D., 69, 064027.
Joshi, P. S. and Królak, A. (1996). ‘Naked strong curvature singularities in
Szekeres spacetimes’. Class. Quant. Grav., 13, 3069.
Joshi, P. S. and Narlikar, J. V. (1982). ‘Black hole physics in globally hyperbolic
spacetimes’. Pramana, 18, 385.
Joshi, P. S. and Saraykar, R. V. (1987). ‘Cosmic censorship and topology change
in general relativity’. Phys. Lett. A., 120, 111.
262 References

Joshi, P. S. and Singh, T. P. (1995). ‘Phase transition in gravitational collapse

of inhomogeneous dust’. Phys. Rev. D., 51, 6778.
Joshi, P. S., Dadhich, N. and Maartens, R. (2000). ‘Gamma-ray bursts as the
birth-cries of black holes’. Mod. Phys. Lett. A., 15, 991.
Joshi, P. S., Dadhich, N. and Maartens, R. (2002). ‘Why do naked
singularities form in gravitational collapse?’. Phys. Rev. D., 65,
101501.
Kahana, S., Baron, E. and Cooperstein, J. (1984). ‘Successful supernovae, the
anatomy of shocks: neutrino emission and the adiabatic index’. In Problem
Collapse and Numerical Relativity, eds D. Bancel and M. Signore. Dordrecht:
Reidel.
Kerr, R. P. (1963), ‘Gravitational field of a spinning particle as an example of
algebrically special metrics’. Phys. Rev. Lett., 11, 237.
Kodama, H. (1979). ‘Inevitability of a naked singularity associated with the
blackhole evaporation’. Prog. Theor. Phys., 62, 1434.
Krasinski, A. (1997). Inhomogeneous Cosmological Models. Cambridge: Cam-
bridge University Press.
Królak, A. (1983). ‘A proof of the cosmic censorship hypothesis’. Gen. Relat.
Grav., 15, 99.
Królak, A. (1986). ‘Towards the proof of the cosmic censorship hypothesis’.
Class. Quant. Grav., 3, 267.
Królak, A. (1999). ‘Nature of singularities in gravitational collapse’. Prog. Theor.
Phys. Suppl., 136, 45.
Kruskal, M. D. (1960). ‘Maximal extension of Schwarzschild metric’. Phys. Rev.,
119, 1743.
Lake, K. (1991). ‘Naked singularities in gravitational collapse which is not self-
similar’. Phys. Rev. D., 43, 1416.
Lake, K. (1992). ‘Precursory singularities in spherical gravitational collapse’.
Phys. Rev. Lett., 68, 3129.
Lake, K. (2000). ‘Gravitational collapse of dust with a cosmological constant’.
Phys. Rev. D., 62, 027301.
Lake, K. and Roeder, R. C. (1979). ‘Some remarks on surfaces of discontinuity
in general relativity’. Phys. Rev. D., 17, 1935.
Landau, L. and Lifshitz, E. (1975). The Classical Theory of Fields. New York:
Pergamon Press.
Lee, C. W. (1983). ‘The topology of geodesically complete space-times’. Gen.
Relat. Grav., 15, 21.
Leibovitz, C. and Israel, W. (1970). ‘Maximum efficiency of energy release in
spherical collapse’. Phys. Rev. D., 1, 3226.
 References 263

Lemaı̂tre, G. (1933). ‘L’univers en expansion’. Ann. Soc. Sci. Bruxelles I A.,

53, 51.
Leray, J. (1952). Hyperbolic Differential Equations. Princeton: Institute for
Advanced Study (Preprint).
Lin, C. C., Mestel, L. and Shu, F. H. (1965). ‘The gravitational collapse of a
uniform spheroid’. Astrophys. J., 142, 1431.
Lindquist, R. W., Schwartz, R. A. and Misner, C. W. (1965). ‘Vaidya’s radiating
Schwarzschild metric’. Phys. Rev. B., 137, 1364.
Lovelock, D. (1972). ‘The four-dimensionality of space and the Einstein tensor’.
J. Math. Phys., 13, 874.
Ludvigsen, M. and Vickers, J. A. G. (1983). ‘A simple proof of the positivity of
Bondi mass’. J. Phys. A., 15, L67.
Maartens, R. and Bassett, B. (1998). ‘Gravito-electromagnetism’. Class. Quant.
Grav., 15, 705.
Madhav Arun, T., Goswami, R. and Joshi, P. S. (2005). ‘Gravitational collapse
in asymptotically anti-de Sitter/de Sitter backgrounds’. Phys. Rev. D., 72,
084029.
Maeda, H. (2006). ‘Final fate of spherically symmetric gravitational collapse of
a dust cloud in Einstein–Gauss–Bonnet gravity’. Phys. Rev., D73, 104004.
Mahajan, A., Goswami, R. and Joshi, P. S. (2005). ‘Cosmic censorship in higher
dimensions. II.’. Phys. Rev. D., 72, 024006.
Malec, E. (1995). ‘Global solutions of a free boundary problem for selfgravitating
scalar fields’, gr-qc/9506005.
Markovic, D. and Shapiro, S. L. (2000). ‘Gravitational collapse with a cosmo-
logical constant’. Phys. Rev. D., 61, 084029.
May, M. M. and White, R. H. (1966). ‘Hydrodynamical calculations of general
relativistic collapse’. Phys. Rev., 141, 1232.
Mazur, P. O. and Mottola, E. (2004). ‘Gravitational vacuum condensate stars’.
Proc. Nat. Acad. Sci., 111, 9545.
McNamara, J. M. (1978). ‘Instability of black hole inner horizons’. Proc. Roy.
Soc. Lond. A., 358, 499.
Mena, F. C., Tavakol, R. and Joshi, P. S. (2000). ‘Initial data and spherical dust
collapse’. Phys. Rev. D., 62, 044001.
Mena, F. C. and Nolan, B. (2001). ‘Non-radial null geodesics in spherical dust
collapse’. Class. Quant. Grav., 18, 4531.
Mena, F. C. and Nolan, B. (2002). ‘Geometry and topology of singularities in
spherical dust collapse’. Class. Quant. Grav., 19, 2587.
Misner, C. (1963). ‘The flatter regions of Newman, Unti and Tamburino’s
generalized Schwarzschild space’. J. Math. Phys., 4, 924.
264 References

Misner, C. (1967). ‘Taub-NUT space as a counterexample to almost anything’. In

Relativity Theory and Astrophysics I: Relativity and Cosmology, ed. J. Ehlers.
Lectures in Applied Mathematics, Vol. 8. Providence: American Mathematical
Society.
Misner, C. and Sharp, D. H. (1964). ‘Relativistic equations for adiabatic,
spherically symmetric gravitational collapse’. Phys. Rev. B., 136, 571.
Misner, C., Thorne, K. and Wheeler, J. (1973). Gravitation. San Francisco:
Freeman.
Morris, M. S., Thorne, K. S. and Yurtsever, U. (1988). ‘Wormholes, time
machines, and the weak energy condition’. Phys. Rev. Lett., 61, 1446.
Nakamura, T. and Sato, H. (1982). ‘General relativistic collapse of non-rotating,
axisymmetric stars’. Prog. Theor. Phys., 67, 1396.
Nakao, K., Kurita, Y., Morisawa, Y. and Harada, T. (2007). ‘Relativistic
gravitational collapse of cylindrical dust’. Prog. Theor. Phys., 117, 75.
Newman, R. P. A. C. (1986). ‘Strengths of naked singularities in Tolman-Bondi
space-times’. Class. Quant. Grav., 3, 527.
Newman, R. P. A. C. and Joshi, P. S. (1988). ‘Constraints on the structure of
naked singularities in classical general relativity’. Ann. Phys., 182, 112.
Nolan, B. C. (1999). ‘Strengths of singularities in spherical symmetry’. Phys.
Rev. D., 60, 024014.
Omer, G. C. (1965). ‘Spherically symmetric distributions of matter without
pressure’. Proc. Nat. Acad. Sci., 53, 1.
Oppenheimer, J. R. and Snyder, H. (1939). ‘On continued gravitational
contraction’. Phys. Rev., 56, 455.
Ori, A. (1992). ‘Inner structure of a charged black hole: an exact mass-inflation
solution’. Phys. Rev. Lett., 67, 789.
Ori, A. and Piran, T. (1987). ‘Naked singularity in self-similar spherical
gravitational collapse’. Phys. Rev. Lett., 59, 2137.
Ori, A. and Piran, T. (1990). ‘Naked singularities and other features of self-
similar general relativistic gravitational collapse’. Phys. Rev. D., 42, 1068.
Ostriker, J. P. and Steinhardt, P. J. (1995). ‘The observational case for
a low density universe with a nonzero cosmological constant’. Nature,
377, 600.
Papapetrou, A. (1985). ‘Formation of a singularity and causality’. In A Random
Walk in Relativity and Cosmology, eds N. Dadhich, J. Krishna Rao, J. V.
Narlikar, C. V. Vishveshwara. New Delhi: Wiley Eastern.
Papapetrou, A. and Hamoui, A. (1967). ‘Surfaces caustiques dégénérées dans la
solution de Tolman. La singularité physique en relativité générale’. In Ann.
Inst. H Poincare AVI, p. 343.
 References 265

Patil, K. D., Ghate, S. H. and Saraykar, R. V. (2001). ‘Spherically symmet-

ric inhomogeneous dust collapse in higher dimensional spacetime and cosmic
censorship hypothesis’. Pramana, 56, 502.
Penrose, R. (1965). ‘Zero rest mass fields including gravitation: asymptotic
behaviour’. Proc. Roy. Soc. Lond. A., 284, 159.
Penrose, R. (1968). ‘Structure of space-time’. In Battelle Rencontres, eds
C. Dewitt and J. A. Wheeler. New York: Benjamin.
Penrose, R. (1969). ‘Gravitational Collapse: the role of general relativity’. Riv.
Nuovo Cimento Soc. Ital. Fis., 1, 252.
Penrose, R. (1972). Techniques of Differential Topology in Relativity. Philadel-
phia: AMS Colloquium Publications, SIAM.
Penrose, R. (1974a). ‘Gravitational collapse’. In Gravitational Radiation and
Gravitational Collapse (IAU Symposium No. 64), ed. C. DeWitt-Morettee.
Dordrecht: Reidel.
Penrose, R. (1974b). ‘Singularities in cosmology’. In Confrontation of Cos-
mological Theories with Observational Data, ed. M. S. Longair. Dordrecht:
Reidel.
Penrose, R. (1979). ‘Singularities and time asymmetry’. In General Relativity:
An Einstein Centenary Survey, eds S. W. Hawking and W. Israel. Cambridge:
Cambridge University Press.
Penrose, R. (1998). ‘The question of cosmic censorship’. In Black Holes and
Relativistic Stars, ed. R. M. Wald. Chicago: University of Chicago Press, p.
103.
Perlmutter, S., Aldering, G., Della Valle, M., et al. (1998). ‘Discovery of a super-
nova explosion at half the age of the Universe and its cosmological implications’.
Nature, 391, 51.
Perlmutter, S., Aldering, G., Goldhaber, G., et al. (1999). ‘Measurements of
omega and lambda from 42 high redshift supernovae’. Astrophys. J., 517, 565.
Price, R. H. (1972). ‘Non-spherical perturbations of relativistic gravitational
collapse. I. Scalar and gravitational perturbations’. Phys. Rev. D., 5, 2419.
Randall, L. and Sundrum, R. (1999). ‘An alternative to compactification’. Phys.
Rev. Lett., 83, 4690.
Raychaudhuri, A. K. (1955). ‘Relativistic cosmology’. Phys. Rev., 98, 1123.
Rein, G., Rendall, A. and Schaeffer, J. (1995). ‘A regularity theorem for solutions
of the spherically symmetric Vlasov–Einstein system’. Commun. Math. Phys.,
168, 467.
Rendall, A. (1992). ‘Cosmic censorship and the Vlasov equation’. Class. Quant.
Grav., 9, L99.
Rendall, A. (2005). ‘The nature of spacetime singularities’, gr-qc/0503112.
266 References

Roberts, M. D. (1989). ‘Scalar field counterexamples to the cosmic censorship

hypothesis’. Gen. Relat. Grav., 21, 907.
Rocha, J. F. V. and Wang, A. (2000). ‘Collapsing perfect fluid in higher-
dimensional spherical spacetimes’. Class. Quant. Grav., 17, 2589.
Sachs, R. (1962). ‘Gravitational waves in general relativity, VIII. Waves in
asymptotically flat space-time’. Proc. Roy. Soc. Lond. A., 270, 103.
Saraykar, R. V. and Ghate, S. H. (1999). ‘C1-stability of naked singularities
arising in an inhomogeneous dust collapse’. Class. Quant. Grav., 16, 281.
Schoen, R. and Yau, S. -T. (1983). ‘The existence of a black hole due to
condensation of matter’. Commun. Math. Phys., 90, 575.
Seifert, H. J. (1967). ‘Global connectivity by timelike geodesics’. Zs. F. Naturfor.,
22a, 1356.
Senovilla, J. M. M. (2006). ‘A singularity theorem based on spatial averages’,
gr-qc/0610127.
Shapiro, S. L. and Teukolsky, S. A. (1991). ‘Formation of naked singularities:
the violation of cosmic censorship’. Phys. Rev. Lett., 66, 994.
Shapiro, S. L. and Teukolsky, S. A. (1992). ‘Gravitational collapse of rotating
spheroids and the formation of naked singularities’. Phys. Rev. D., 45, 2006.
Sil, A. and Chaterjee, S. (1994). ‘Singularity structure of a self-similar Tolman
type model in a higher dimensional spacetime’. Gen. Relat. Grav., 26, 999.
Simmons, G. F. (1963). Introduction to Topology and Modern Analysis. Tokyo:
McGraw-Hill Kogakusha Ltd.
Simpson, M. and Penrose, R. (1973). ‘Internal instability in a Reissner–
Nordstrom black hole’. Int. J. Theor. Phys., 7, 183.
Spivak, M. (1965). Calculus on Manifolds. New York: W. H. Freeman.
Stephani, H., Kramer, D., MacCallum, M., Hoenselaers, C. and Herlt, E.
(2003). Exact Solutions of Einstein’s Field Equations. Cambridge: Cambridge
University Press.
Szekeres, P. (1960). ‘On the singularities of a Riemannian manifold’. Publ. Math.
Debrecen, 7, 285.
Szekeres, P. (1975). ‘Quasispherical gravitational collapse’. Phys. Rev. D., 12,
2941.
Szekeres, P. and Iyer, V. (1993). ‘Spherically symmetric singularities and strong
cosmic censorship’. Phys. Rev. D., 47, 4362.
Thorne, K. S. (1972). ‘Non-spherical gravitational collapse: a short review’. In
Magic without Magic: John Archibald Wheeler, ed. J. Clauder. New York: W.
H. Freeman.
Tipler, F. (1977). ‘Singularities in conformally flat spacetimes’. Phys. Lett. A.,
64, 8.
 References 267

Tipler, F., Clarke, C. J. S. and Ellis, G. F. R. (1980). ‘Singularities and hori-

zons’. In General Relativity and Gravitation Vol 2, ed. A. Held. New York:
Plenum.
Tolman, R. C. (1934). ‘Effect of inhomogeneity on cosmological models’. Proc.
Natl. Acad. Sci. USA, 20, 169.
Traschen, J. (1994). ‘Discrete self-similarity and critical point behavior in
fluctuations about extremal black holes’. Phys. Rev. D., 50, 7144.
Tsujikawa, S., Singh, P. and Maartens, R. (2004). ‘Loop quantum gravity effects
on inflation and the CMB’. Class. Quant. Grav., 21, 5767.
Vaidya, P. C. (1943). ‘The external field of a radiating star in general relativity’.
Curr. Sci., 12, 183.
Vaidya, P. C. (1951). ‘The gravitational field of a radiating star’. Proc. Indian
Acad. Sci. A., 33, 264.
Vaidya, P. C. (1953). ‘Newtonian time in general relativity’. Nature, 171, 260.
Vaz, C. and Witten, L. (1994). ‘Evaporation of a naked singularity in 2D gravity’.
Phys. Lett. B., 325, 27.
Vaz, C. and Witten, L. (1995). ‘Soliton-induced singularities in 2D gravity and
their evaporation’. Class. Quant. Grav., 12, 2607.
Wald, R. M. (1984). General Relativity. Chicago: University of Chicago Press.
Wald, R. M. (1997). ‘Gravitational collapse and cosmic censorship’, gr-
qc/9710068.
Wald, R. M. and Iyer, V. (1991). ‘Trapped surfaces in the Schwarzschild
geometry and cosmic censorship’. Phys. Rev. D., 44, R3719.
Wang, A. and Wu, Y. (1999). ‘Generalized Vaidya solutions’. Gen. Relat. Grav.,
31, 1, 107.
Waugh, B. and Lake, K. (1988). ‘Strengths of shell-focusing singularities in
marginally bound collapsing self-similar Tolman space-times’. Phys. Rev. D.,
38, 1315.
Waugh, B. and Lake, K. (1989). ‘Shell-focusing singularities in spherically
symmetric self-similar space-times’. Phys. Rev. D., 40, 2137.
Weinberg, S. (1972). Gravitation and Cosmology. New York: John Wiley.
Wheeler, J. A. (1962). ‘Geometrodynamics’. New York: Academic Press.
Wheeler, J. A. (1964). ‘Geometrodynamics and the issue of final state’. In Rela-
tivity, Groups and Topology, eds C. Dewitt and B. Dewitt. New York: Gordon
and Breach.
Willard, S. (1970). General Topology. Reading, Mass.: Addison-Wesley.
Yodzis, P., Seifert, H. -J. and Muller zum Hagen, H. (1973). ‘On the occurrence
of naked singularities in general relativity’. Commun. Math. Phys., 34, 135.
268 References

Yodzis, P., Seifert, H. -J. and Muller zum Hagen, H. (1974). ‘On the occur-
rence of naked singularities in general relativity II’. Commun. Math. Phys.,
37, 29.
Zehavi, I. and Dekel, A. (1999). ‘Constraints on the cosmological constant from
flows and supernovae’. Nature, 401, 252.
 Index

achronal boundary, 139 near a singularity, 204

achronal set, 139 causality condition, 142
advanced null coordinates, 47 caustics, 156
affine parameter, 29
chart, 12
Alexandrov topology, 142
Christoffel symbols, 34
anisotropic de Sitter metric, 90
apparent horizon, 71, 79, 241 chronological future (past), 45, 137
in dust collapses, 98 chronology condition, 142
in higher-dimensions, 169, 173 commutator, 15
arc length, 24 comoving coordinates, 64
asymptotically flat spacetime, 162 conformal compactification, 47,
atlas, 12 163
conformal flatness, 36
Bianchi identities, 35 conformal transformation, 140
Birkhoff’s theorem, 44
congruence
blackhole, 3, 4, 54
blackhole region, 164 hypersurface orthogonal, 32
blue-shift instability, 178, 194 of non-spacelike curves, 154
bouncing models, 67 of null geodesics, 157
brane-world models, 169 of timelike geodesics, 154
conical singularity, 152
C k -curve, 13 conjugate points, 157
C k -singularity, 161 connection, 25
C r -function, 13
flat, 35
C r -manifold, 12
Cauchy horizon, 146 integrable, 35
instability, 178, 195 torsion-free, 28, 34
Reissner−Nordström case, 195 contravariant vector, 13
in TBL metrics, 195 convex normal neighborhood, 30
Cauchy surface, 45, 145 coordinate singularity, 149
causal boundary, 147, 205 cosmic censorship, 164
causal functions, 145
in blackhole physics, 179
causal future (past), 45, 138
and causal structure, 135
causal relations, 23
and spacetime topology, 142 and cosmological term, 129
causal structure and global hyperbolicity, 144
and naked singularity, 205 in higher dimensions, 169, 195

269
270 Index

cosmic censorship (Cont.) extendible curves, 161

from quantum gravity, 253 extrinsic curvature, 89, 157
in quantum gravity, 190
vacuum version of, 179 first fundamental form, 88
cosmic censorship hypothesis, 5 free gravitational field, 222
cosmological constant, 43, 120 Friedmann models, 53
and bouncing models, 121 future asymptotic predictability, 164
and singularity avoidance, 121 future end point, 140
covariant derivative, 25 future imprisoned curves, 143
curvature scalar, 36 future set, 139
future (past) timelike infinity, 46
dark energy, 119
generic condition, 159
diffeomorphism, 13
geodesic deviation equation, 37, 156
differentiable manifold, 11
geodesic equations, 30
dispersal, 67
geodesic incompleteness, 150
distinguishing condition, 142
geodesics, 29
domain of dependence, 145
affinely parametrized, 29
dominant energy condition, 65, 156
complete, 30
dual basis, 15
incomplete, 30
dust collapse, 7
maximal, 146
bound, unbound, and marginally
timelike, null, spacelike, 29
bound, 94
global hyperbolicity, 144
and cosmic censorship, 91
global time function, 143
with cosmological constant, 121 Gödel universe, 141
in five dimensions, 198 gravitational collapse, 1
generic analysis of, 91 and energy conditions, 180
in higher dimensions, 131, 197 and equation of state, 184
self-similar, 106 of infinite cylinders, 226
initial conditions, 68
Eddington–Finkelstein coordinates, 54 Misner–Sharp equations, 63
edge of a set, 144 and negative pressures, 238
Einstein cluster, 77 physical constraints, 179
Einstein equations, 41 with purely tangential pressure, 69,
Einstein static universe, 48 132
Einstein tensor, 41 and quantum effects, 3
Einstein–Vlasov description, 185 role of inhomogeneities, 225
embedded submanifold, 31 self-similar, 58
energy–momentum tensor, 38 and spacetime singularity, 60
dust, 39 gravitational focusing, 150, 154
perfect fluid, 39 gravitational red-shift, 52
equation of state, 70
dust, 92 homeomorphism, 13
event horizon, 4, 55, 165 homogeneous collapse, 52
event horizon conjecture, 227 hoop conjecture, 226
expansion, 154 hypersurface, 31
exponential map, 30 timelike, null, spacelike, 31
 Index 271

ideal points, 46, 147 global visibility, 112

immersed submanifold, 31 gravitationally strong, 97
indecomposable past set, 147 gravitationally weak, 97
inextendible causal curves, 46 and inhomogeneities, 216
inextendible curve, 140, 161 non-spacelike curves from, 208
initial data for collapse, 182 physical causes for, 225
Israel–Darmois conditions, 88 and quantum gravity, 186
and spacetime shear, 216
Jacobi equation, 37 in the Vaidya metric, 58
Jacobi fields, 156 negative pressure, 119
and absence of trapped surfaces,
Killing equation, 40 249
Killing vector, 40 and loop quantum gravity, 249
Klein–Gordon equation, 241 negative pressure in collapse, 244
Kretschmann scalar, 79 neutron star, 213
non-spacelike curve, 24
limiting focusing condition, 234 non-spherical collapse, 225
line element field, 24 normal coordinates, 30
local causality, 40 null curve, 24
local causality principle, 141 null infinity, 47
local conservation of energy and null vectors, 23
momentum, 39
local coordinate neighborhood, 11 one-forms, 14
loop quantum gravity, 236, 248 Oppenheimer–Snyder–Datt collapse
discrete regime, 250 special case of TBL metric, 199
Hamiltonian, 251 Oppenheimer–Snyder–Datt model,
Lorentz group, 45 236
Lorentz transformations, 45
Lorentzian metric, 10 parallel transport, 24
partial Cauchy surface, 144
mass function, 66 past set, 147
massive stars Penrose diagram, 48
continual collapse of, 215 perfect fluid collapse, 133
final fate of, 210 PIP and PIF, 148
life cycle of, 213 Planck length, 2
and quantum gravity, 212 Poincaré group, 45
matching conditions, 244 points at infinity, 46
metric tensor, 21 principle of equivalence, 38
non-degenerate, 22 principle of general covariance, 38
positive definite, 23 projection operator, 154
signature of, 23
Minkowski spacetime, 23, 45 quantum effects
semi-classical modifications, 251
naked singularity, 3, 5, 177 quantum effects in collapse, 248
central, 81 quantum gravity, 2, 236
dissolution of, 212 observational effects, 239
272 Index

Raychaudhuri equation, 155, 224 regular points, 147

regularity conditions, 69 weakly asymptotically simple and
retarded null coordinates, 47 empty, 163
Ricci tensor, 35 spacetime manifold, 10, 23
Riemann curvature tensor, 32 flat, 35
rotation of a congruence, 154 orientable, 18
spacetime metric
scalar field collapse, 202 spatial part of, 154
with a potential, 239 spacetime singularity, 2, 149
quantum effects in, 248 definition of, 150
Schwarzschild spacetime, 44, 49 essential, 151
analytic extension of, 54 removable, 50, 183
second fundamental form, 89, 158 spacetime topology, 16, 137
self-similar collapse, 108, 181 connected, 17
shear, 154 and global hyperbolicity, 145
and delay of trapped surfaces, 216 Hausdorff, 17
and distortion of apparent horizon, non-compact, 17
220 paracompact, 17
shell-crossing spherical symmetry, 63
avoidance of, 126 stable causality, 143
extension of, 96 static limit, 51
shell-crossing singularity, 66 static metric, 49
shell-focusing singularity, 66 string theory, 236
singularity strong causality, 142
curvature strength of, 112 strong cosmic censorship, 176, 178
globally visible, 84 violations of, 178
locally visible, 84 strong curvature singularity, 229
removable, 150 condition for, 183
scalar polynomial, 113 in the sense of Królak and Clarke,
stability and genericity, 191 114
singularity curve, 171 in the sense of Tipler, 114
singularity theorems, 2, 160 strong energy condition, 156
smooth density profile, 182 strong limiting focusing condition,
space of causal curves, 146 234
spacelike curve, 24 supernova explosion, 214
spacetime surface gravity, 167
asymptotically predictable, 164 synchronous coordinates, 31
boundary, 17, 46 Szekeres collapse models, 230
causal structure of, 136
compact, 142 tangent space, 14
deterministic, 146 Taub–NUT solutions, 153
extendible, 50 TBL collapse models, 93
finite measure on, 145 TBL metric, 93
geodesically incomplete, 30 tensor, 18
inextendible, 151 addition, 20
and maximal length property, 147 anti-symmetric, 21
 Index 273

contraction, 20 vacuum energy density, 120

outer product, 20 Vaidya metric, 44, 56
symmetric, 21 charged, 90
symmetric part of, 21 generalized, 88, 244, 246
tensor fields, 20 generalized mass function, 90
tensor product, 19 vector field, 15
timelike and spacelike vectors, 23 volume expansion, 154
timelike curve, 24
TIP and TIF, 160
topology weak cosmic censorship, 176
on the space of causal curves, 146 weak energy condition, 65, 156
on the space of metrics, 193 Weyl tensor, 36
torsion, 26 electric, 222
trapped surfaces, 55, 236, 241 magnetic, 222
absence of, 243 white dwarf, 213
type I matter fields, 64 wormhole, 17