Вы находитесь на странице: 1из 10

NAKED SINGULARITIES

Roger Penrose*
Department of Mathematics
Birkbeck College
London W.C.I., England

I first review what is now the “standard” picture of gravitational collapse to a


black hole-which I shall refer to as the establishment viewpoint. The picture is, in
fact, a very good one, with a remarkable degree of internal consistency. But it is
worth emphasizing that, among other things, this picture does depend on a very
big assumption, that of cosmic censorship.‘ In my own view, this assumption might
well be false in suitable circumstances. I believe that the consequences of possible
violations of cosmic censorship are certainly worth considering seriously, particu-
larly in view of Weber’s observations.
The essentials of the establishment viewpoint are depicted in FIGURE 1. I am
supposing that a collapse has taken place to the extent that one of the criteria for
the applicability of the Hawking-Penrose singularity theorem2 occurs. Let us say
that either there is a trapped surface or, as is the case depicted in FIGURE 1, there is
a future light cone (namely that of P)with the property that sufficient matter crosses
it to cause all the light rays to begin to reconverge somewhere to the future of P.
With the normal assumption that Einstein’s field equations with “reasonable”
sources (i.e., subject to the energy condition) hold and that the solution is ade-
quately general and without closed timelike curves, the existence of some form of
space-time singularity is implied. But the theorem does not say that the singularity
is necessarily hidden inside an absolute event horizon. It is here that we must invoke
the ad hoc assumption of cosmic censorship in order that the establishment picture
be obtained. In effect, this principle states that naked singularities do not develop
out of an initially nonsingular state, but we must be careful in formulating a precise
statement of this, and I shall return to some of the difficulties later. A naked singu-
larity may be said to exist in a space-time M if there is a well-defined external
infinity (that is, 9+, if M is weakly asymptotically simple3) from which timelike
curves may be drawn into the past that terminate on the singularity (FIGURE2). I
shall not make precise, here, the concept of “terminating on a singularity” but refer
instead to Hawking’s concept of asymptotic predi~tabilify.~ This is a precise condi-
tion on a space-time M that may be taken as a statement that M accords with a
form of cosmic censorship. Imposing such a condition, we deduce the existence of an
absolute event horizon E, which encloses but does not intersect the singular region.
This horizon E may be defined as the boundary of the past of the external infinity.
The property of E that its surface area increases with time may now be readily de-
rived. This property has been used (by Floyd and Penrose5 and by Hawking4) to
obtain significant inequalities on the behavior of black holes. Finally, to complete the
establishment picture we invoke the Israel-Carter-Hawking conjecture/theorem,6-8
which asserts that the space-time settles down-in some appropriate but as yet ill-
defined sense-to become that of a Kerr black hole. Unlike cosmic censorship, how-
ever, there is some considerable mathematical support for this particular conjecture.
For the purposes of my discussion, therefore, I shall not call this conjecture into
* Present address: Mathematical Institute, Oxford, England.
125
126 Annals New York t\.caderny of Sciences

- f ;
KERR

C
0
N
._
L
0
L
+
C
:
a
+

FIGURE1. Space-time diagram depicting collapse to a black hole according to the


standard (“establishment”) picture. (In all figures, time proceeds from the bottom of
the page to the top.)
Penrose: Naked Singularities 127

NAKED
P
CLOTHED

FIGURE 2. For asymptotically flat space-times, the difference between naked and
clothed singularities is that in the first case the singularity lies to the past of $+ (future
null infinity) but not in the second case.

question. But one should bear in mind that there is still a possible weakness in the
establishment picture here too.
Let me describe an attempt that I made, about two years ago, to obtain a con-
tradiction with the establishment point of view. I envisage a shell of photons (“null
dust”) that collapses into a region of flat space-time. (The situation is a kind of
generalized time-reverse of Synge’s “instantaneous transformation of a massive
particle into radiation” ”) I assume no symmetry either for the shape of the shell or
for the distribution of energy density on the shell. The shell has merely to be con-
vex-in order that no caustics occur in the remote past on the null hypersurface H ,
representing the history of the shell. This requirement (together with the fact that
the shell is null) ensures that we can choose the space-time inside the shell to be flat
(FIGURE 3).
Consider, now, an expanding null hypersurface N in the flat region, which inter-
sects H in a (spacelike) two-surface S having the topology of a sphere. Extending N
as a null hypersurface through H , we find that the divergence of N jumps down by
an amount proportional to the intensity of energy of the shell at S . This focusing
effect of H on N is smaller the farther out (i.e., further back into the past) that S is
on H . This is because the energy intensity varies as r-2 back along the generators of
H , whereas the divergence of N decreases as r-l. Moving S inwards on H , we expect
to find a location at which the divergence of N is exactly canceled so that the rays of
N that extend into the future from S have zero divergence as they leave S . Let us
suppose that S = So can be so chosen. Then S o will be a marginally trapped surface.
128 Annals New York Academy of Sciences

FIGURE3. An attempt to derive a contradiction with the “establishment” view-


point. A shell of photons collapses into a flat space-time. The assumption that a Kerr
black hole results implies the existence of a certain inequality on the initial geometry.
Penrose: Naked Singularities 129
If moved a little inwards along the generators of H , it becomes a trapped surface
and the Hawking-Penrose theorem applies. Assuming cosmic censorship, we deduce
the existence of an absolute-event horizon E intersecting H in a two-surface S1,the
surface Sr lying farther out along the generators of H than So.Furthermore, the
total surface area A1 of S1 is not less than the area A. of SO.Moreover, the total
surface area of cross-section of E increases from that of SIas we proceed into the
future until, according to the establishment viewpoint, it approaches the final value
A, = g?rm(m + 1 / M 2 - u2)

for the event horizon of the Kerr metric for mass, m, and angular momentum, am.
(Units are chosen so that G = c = 1 . ) This in turn is less than or equal to the
Schwarzschild value
A, = 16nm2
for the same mass. This mass, m, cannot exceed the original total mass, mo,of the
collapsing shell, assuming that no further matter or radiation falls in from infinity.
In general, we would expect the strict inequality m < m,, since outgoing gravita-
tional radiation would carry mass-energy away as the final black hole is formed.
We have, therefore, a string of inequalities
AO < < A, < A, < 167rmo2
implied by assumptions that go to make up the establishment picture of collapse to
a black hole. Furthermore, A. and mo are quantities that can be related to each
other and refer to the geometry of the original setup in flat space. If we could ar-
range such a collapsing shell to have A. > 167rmo2,then we should have a contradic-
tion with the establishment picture. The most reasonable inference about the future
history of the space-time resulting from such a collapse would then be that a naked
singularity develops.
The initial setup is characterized simply by the two-surface Soembedded in the
initial Minkowski space-time. A. is simply the area of SO,while inocan be calculated
from the geometry of the two null hypersurfaces, H and N , in Minkowski space,
which intersect in So.The divergence of N at S measures the intensity of the energy
of the shell at So (since this intensity is exactly sufficient to reduce the divergence to
zero). Scaling this intensity dawn by the inverse square of the appropriate luminosity
parameter on H , we obtain the initial intensity of the shell at infinity. Integrating this
over the sphere at infinity, we obtain rno. For the case when the shell is spherical in
shape (i.e., when H is a light cone) we may obtain the establishment inequality
A0 < 167rmoZin the form of a simple mathematical relation
4*4;rm < (4;r(1 - VIogr)P)2
where r is a smooth positive function defined on the unit sphere (with surface
element P = sin 8 d8 A dp and covariant Laplacian v") the quantity r measuring the
distance of S o from the time axis through the vertex of the cone H . (It is curious that
exactly the same inequality arises for an analogous situation in which incoming
gravitational waves, whose form is that of a curtailed Robinson-Trautman solution,
implode to form a black hole.) We require the condition
v2log r < 1
to ensure positivity of the mass intensity.
Another special case of special interest is given when S lies in a spacelike hyper-
130 Annals New York Academy of Sciences

plane. Gibbons has pointed out that the inequality A . < l6rmO2now becomes
Minkowski’s inequality

for a closed convex surface in Euclidean three-space, with surface area element o
and extrinsic curvature K . Since this inequality is known to be valid, we see that no
contradiction with the establishment picture is obtained for this situation. Gibbons
has also shown that the general inequality A . Q 16rmo2follows from Minkowski’s
inequality by projection into a Euclidean three-space orthogonal to the time axis.
However, Minkowski’s inequality is not always true for surfaces that are not convex.
In the general situation this projected surface need not be convex. Thus, it is still
just possible that the inequality A . Q l6rmO2might be violated, but it could be vi-
olated only for a rather exotic initial setup.
The implications of all this are unfortunately rather inconclusive. I think it can
be said that some support is offered to the establishment picture by Gibbon’s work,
at least if the collapse is of a kind that does not differ a great deal from that of
spherical symmetry. But it might be argued that an exotic initial setup would in any
case be needed if naked singularities were to be produced. It should be emphasized,
also, that even if a proof of the inequality A . < 16rm0*in all relevant circumstances
were to hand, this would be a very long way off from establishing the principle of
cosmic censorship. It is only that the negation of this inequality would, in effect, es-
tablish the reverse. It might be, for example, that a collapse to a naked singularity
could occur without a trapped surface’s developing at any stage. Finally, it may be
remarked that even if no contradiction with the establishment picture can be ob-
tained using the above argument, the inequalities obtained d o have some physical
implications.If we turn the argument around and assume the establishment viewpoint
to be correct, we obtain inequalities limiting the amount of gravitational radiation
that can be emitted for various configurations of imploding shells of radiation.
One aspect of the naked-singularity problem is the question of the stability of
black holes. One may conceive of a situation in which a “black hole” is present at
first but where the “event horizon” is unstable, the curvature in its neighborhood
increasing to infinity with time. Although much significant work has been done on
this subject e.g. refs. 10-12, one still cannot say with certainty whether the Kerr
metric is stable. One possibility that has been suggested’3 is that for some range of
values k < a / m < 1 , with k some positive number less than one, it might be that
the Kerr solution is unstable, although stable for 0 < a / m < k. If an instability of
this sort occurs, it might not, of course, be the kind that leads to a naked singularity.
It could simply be that any excess angular momentum gets radiated away in gravi-
tational waves.
I think I would be inclined to guess that the Kerr metrics for all 0 < a/ m < 1
are stable against small enough disturbances. But stability against large disturbances
is much more problematical. The question has often been raised as to whether it
might be possible to add high enough angular momentum material to a Kerr black
hole to increase its a / m value to greater than unity. If this could be done, then a
naked singularity would be the result, since for a > m the Kerr metrics represent
naked singularities and not black holes. However, the nearer that a approaches to
m, the “harder” it is for large-angular-momentum prograde material to be swal-
lowed by the black h ~ l e . ’ ~The
J ~ principle of area increase may be invoked5 to
obtain a limit for the ratio of the increment of angular
___ momentum 6(am) to the in-
+
crement of mass 6m in the form 6 (8rm(m l / m z - az) ) > 0. But since the area
principle is based on cosmic censorship, we need to check this formula with an al-
Penrose: Naked Singularities 131

ternative means of derivation. In fact, such a derivation can be obtained, in the case
of test fields, from the fact that the contraction Tabkaof a Killing vector ka with the
test energy-momentum
Tab (with Tab = Tba, VaTab = 0)
gives a quantity subject to a global conservation law. Letting ka be (a/&). and
(d/dcp)” respectively, we obtain an exact conservation of test energy and test angular
momentum about the axis. The area increase inequality now comes down to the fact
that the flux kaTabdSbacross the event horizon, with surface element dSb (taken
future-null) must be positive-an implication independent of any cosmic censorship
assumption. Because of the global conservation laws, this means that the asymptot-
ically measured increment in test energy and test angular momentum must be sub-
ject to the same inequality. Thus, the increment in the total mass and angular
momentum of the black hole should be subject to this inequality-so surface area in-
crease may be reasonably inferred independently of cosmic censorship, with the im-
plication that a gradual “spinning up” of a Kerr solution to values a / m > 1 cannot
be achieved. (It may be remarked, however, that gravitational perturbations could
conceivably present a different situation,’s since no energy-momentum tensor exists
for gravity.)
These considerations have very little to say about large perturbations, however.
We might, for example, envisage two comparable black holes spiraling into one
another. Have we any reason, other than wishful thinking, to believe that a black
hole will be formed, rather than a naked singularity? Very little, I feel; it is really a
completely open question. In fact, one question that should be asked is whether there
is any observational evidence either for or against naked singularities occurring in
nature. There are certainly enough odd things seen in astronomy that one is tempted
to speculate about naked singularities playing a role. But perhaps the question should
be asked the other way: is it possible to explain these observations without invoking
naked singularities? The most glaring place where the answer might appear to be
“no” is in Weber’s results. If we take his conclusions at their face value, then we are
led to ask how it might be possible to beam gravitational radiation very accurately
in the galactic plane. Attempts, such as Misner’s, to explain beaming of this kind in
terms of black holes are certainly worthwhile, but it is very difficult to see how suffi-
cient beaming can be obtained. I feel that it is also worth exploring the possibility
that naked singularities could achieve the required effect.’? Remarkably, the Kerr
metrics with a /m > 1 do exhibit such a beaming. I shall consider a very hypothetical
model that exploits this fact.
Suppose we have a rotating and slowly collapsing body, or collection of bodies,
whose exterior field resembles a Kerr metric with a > m. I am going to suppose
that, for some reason, when general-relativistic effects begin to dominate over all
others, the Kerr configurations become increasingly preferred as approximate ex-
ternal fields. This is at the moment quite unjustified, since no event horizon is present
and there is no other physical reason known for preferring Kerr fields. Nevertheless,
it does not seem to me to be a totally unreasonable assumption in view of other
special mathematical properties possessed by the Kerr field. Then, as the body
shrinks, the external field approaches that of Kerr with a > m ever more closely.
The space-time curvature would become very large at the equatorial region. In the
limit we could envisage the body attaining a disk-like form, the space-time curvature
at its edge becoming virtually infinite-a naked singularity ! Where curvatures are so
large we would expect any disturbance to result in the emission of gravitational
waves. But Carter1*showed that signals (i.e., null geodesics) from the neighborhood
132 Annals New York Academy of Sciences

of the singular ring can escape to infinity only in directions very close to the equa-
torial plane-the singularity itself being “visible” only from the equatorial plane.
Thus, we have the potentiality for a very strong beaming effect for the gravitational
waves-a beaming that could, in principle, be sufficient t o remove the mass-loss
conflict in Weber’s observations.
Of course this is a highly speculative picture. However, I d o not feel that it can
be rejected out of hand. The existence of closed timelike curves in the Kerr solutions
could have been a worrying feature, but the presence of the disk of rotating matter
removes these. The main queries are whether an a > m collapse would take place
at all, and whether Kerr metrics are likely to arise in such situations if it does. It is,
of course, possible that the beaming effect occurs with other naked singularities so
that the restriction to Kerr might not be necessary. Unfortunately, very little is
known about this. It is therefore worthwhile to point out some of the very remarka-
ble properties that the Kerr metric actually does possess independently of the a:m
ratio, since they may possibly indicate a preference that nature could have for these
metrics in highly general-relativistic situations.
In the first place the metrics are asymptotically flat, with two commuting Killing
vectors (stationary and axi-symmetric) . The Weyl tensor is type { 22 1, which implies
the existence of a symmetric “Killing spinor,” K A B , satisfying

this means that the complex quantity T A T B K A B is conserved along null geodesics,
?y* being a parallelly propagated spinor pointing along the null geodesic. The con-

servation of arg(TATBKAB)implies that the propagation of the polarization plane


along the null geodesic can be calculated directly without i n t e g r a t i ~ nThe
. ~ ~ integra-
tion of the geodesic equations (null or non-null) can reduced be to quadratures (as
Carter first showed1*).This fact can be derived from the existence of a symmetric
Killing tensor, Kab

in addition to the two Killing v e c t ~ r sR. ~ ~ ~ ~has shown that K,a can be ob-
. ~Floydz1
tained from a different type of “Killing tensor,” a skew tensor Hab satisfying
VaHbc = VLaHbcl
where
Kab = HacHb”.
The existence of such Hat,, together with the asymptotic flatness and vacuum
conditions, is actually sufficient to characterize the Kerr family of space-time metrics
uniquely. We have, in fact,
Ha& == iKABtA‘B‘ - i€At(KA‘B’
and
KAB = $-1130(A1g)

where the Weyl conformal spinor is


*AWCD = $o(AoWlCLD),

with o A i A = 1 . Floyd points out that the vector


Fa = HabPb
Penrose: Naked Singularities 133

is parallelly propagated along geodesics (tangent vector P)and conserved on col-


lisions. The squared length FaFQis “Carter’s fourth integral,” given by &Papb,
and is constant along geodesics. In some sense Fa seems to be giving a sort of meas-
ure of angular momentum that retains a measure of conservation about any axis, not
just the rotation axis. The existence of such a quantity (together with asymptotic
flatness and vacuum) occurs only for the Kerr metrics. But the physical implications
of this fact are not yet clear.
To return to the question of cosmic censorship in a more general context, it
must be said that still practically nothing is known about this. It is not even clear
how one could formulate the principle in plausible but precise terms. Many solutions
of Einstein’s equations are known that do possess naked singularities (cf. the recent
solution of Tomimatsu and Sato,22for example, or the CurzonZ3metric). It might
even be possible to produce exact solutions representing a collapse of some special
matter distribution to such a singularity. But this would in itself tell us rather little.
We would have to know whether such behavior was “stable” in the sense of being
qualitatively unchanged when the initial conditions are perturbed in some small but
finite way.
As a postscript to this paper, I should also mention that some interesting solu-
tions representing collapsing dust and possessing naked singularities have been
given by Yodzis and These seem perhaps to be stable under perturba-
tions of the initial conditions, but we need also to ask whether they are, in some
appropriate sense, also stable under perturbations of the equations of state, e.g. what
happens if pressure is introduced? These authors point out that Hagerdorn has
suggested that there should be a maximum possible pressure to density ratio, so
that when the density dominates this value, the pressures should become insignifi-
cant. Further work along such lines would seem to be of great relevance to the naked-
singularity problem.
It is sometimes said that if naked singularities do occur, then this would be
disastrous for physics. I do not share this view. We already have the example of the
big-bang singularity in the remote past, which seems not to be avoidable. The “dis-
aster” to physics occurred right at the beginning. Surely the presence of naked sin-
gularities arising occasionally in collapse under much more ‘(controlled” circum-
stances would be the very reverse of a disaster. The effects of such singular occur-
rences could then be accessible to observation now. Theories of singularities would be
open to observational test. The initial mystery of creation, therefore, would no
longer be able to hide in the obscurity afforded by its supposed uniqueness.
REFERENCES
1. PENROSE,
R. 1969. Rivista del Nuovo Cimento. Serie I, 1, Numero Speciale:
252.
2. HAWKING, S. W. & R. PENROSE. 1970. Proc. Roy. SOC.(Lond.) A 314: 529.
3. PENROSE, R. 1968. In Battelle Rencontres C. M. DeWitt & J. A. Wheeler, Eds.
W. A. Benjamin, Inc. New York, N.Y.
4. HAWKING. S. W. 1972. Commun. Math. Phvs. 25: 152.
5. PENROSE, R. & R. M. FLOYD. 1971. NaturePhys. Sci. 229: 177.
6. ISRAEL, W. 1967. Phys. Rev, 164: 1776.
7. CARTER, B. 1971. Phys. Rev. Lett.78: 331.
8. HAWKING, S. W. & G . F. R. ELLIS. 1973. The Large-scale Structure of Space-
Time (Cambridge Univ. Press.) Cambridge, England.
9. SYNGE,J. L. 1957. Proc. Roy. Irish Acad. A 59: 1.
10. DOROSHKEVITCH, A. G., Y A . B. ZEL’DOVICH & I. D. NOVIKOV. 1965. Zh
Eksperim. i Teor. Fiz. 49: 170; 1965. Soviet Phys. JET P 22.
1 1 . VISHVESHWARA, C. V. 1970. Phys. Rev. D 1: 2870.
12. PRICE,R. H. 1972. Phys. Rev. D . 5: 2439.
134 Annals New York Academy of Sciences
13. SMARR, L. 1972. Surface geometry of charged rotating black holes. Preprint.
14. BARDEEN, J . M. 1969. Nature 224: 1263.
15. WALD,R. 1972. (Preprint).
16. PRESS,W. H. & S. TEUKOLSKI.1972. Preprint.
17. PENROSE.
- -. .. _ R. ~ 1972. Nature 236: 377.
18. CARTER, B-. 1968; Phys. Rev. 174: 1559.
19. WALKER,M. & R. PENROSE. 1970. Commun. Math. Phys. 18: 265.
20. HUGHSTON. L. P.. R. PENROSE.P. SOMMERS & M. WALKER. 1972. Commun.
~ Math. Phys. 27i303.
21. FLOYD,R. 1973. The Dynamics of Kerr Fields. (Ph.D. Thesis, London Univ.
London, England.
22. TOMIMATSU. A & H. SATO. 1972. Phvs. Rev. Lett. 29: 1344.
23. CURZON,H.’ 1924. Proc. Lond. Math. SOC.23: 1344.
24. YODZIS,P., H.-J. SEIFERT& H. MULLERZUM HAGEN. 1973. On the occurrence
of naked singularities of general relativity. Preprint.

Вам также может понравиться