Foundations o f Physics Letters, VoL 5, No.
2, 1992
DOES GENERAL RELATIVITY ALLOW AN OBSERVER
TO VIEW AN ETERNITY IN A FINITE TIME?
M a r k L. H o g a r t h
Department of History and Philosophy of Science,
Free School Lane, Cambridge,
CB2 3RH, United Kingdom
Received September 18, 1991; revised November 11, 1991
I investigate whether there are general relativistic spacetimes that al-
low an observer g to collect in a finite time all the data from the
worldline of another observer ~, where the proper length of ~'s world-
line is infinite. The existence of these spacetimes has a bearing on
certain problems in computation theory. A theorem shows that most
standard spacetimes cannot accommodate this scenario. There are
however spacetimes which can: anti-de Sitter spacetime is one exam-
ple.
Key words: general relativity, eternity, anti-de Sitter, Pitowsky.
1. I N T R O D U C T I O N
Any computer primed to perform an infinite number of compu-
tational steps must take an eternity to complete the task, because
completion in a finite time would imply an unbounded signal veloc-
ity - - conflicting with relativity theory. 1 This would seem to suggest
that the full potential of these computers is available only to immor-
tal computer users. But, as Itamar Pitowsky has pointed out (private
communication), there is no reason why the computer user must re-
main beside the computer. If he follows a different worldline his clock
will tick at a rate different to that of the computer's clock, and per-
haps an extreme case could be organized in which the rates are such
173
0894-9875/92/0400-0173506.50/0© 1992 Plenum Publishing Corporation
174 Hogarth
that the finite proper time as measured by the computer user %orre-
sponds" to an infinite proper time as measured by the computer. In
this case, and granting also that the computer can always signal to
the computer user, the computer user will take only a finite time to
view the eternity of the computer's life and with it the results of its
computations.
Are there relativistic spacetimes in which this is allowed to oc-
cur? This is Pitowsky's question, which I will now attempt to answer.
2. T H E S T O R Y O F D A V E , H A L , A N D G O L D B A C H
Let us spell out the details of the problem with the help of a lit-
tle story, in order to darify what is required of those spacetimes that
permit the scenario described above. To begin with, there lives an
immortal computer 2, called HAL, and a mortal computer user, called
Dave. Suppose that Dave is itching to know whether the Goldbach
conjecture is true or false. He turns to HAL for help. HAL is happy
to sacrifice her eternal life to this great question and begins to test
systematically all the even numbers to see if there exists one which is
not the sum of two primes. Meanwhile Dave has done some calcula-
tions of his own and has discovered that there exists in the universe
a special worldline of length a one hour, which contains a point to the
future of HAL's worldline. It so happens that this worldline is close
by, and Dave decides to follow it. "Do keep in touch," he calls to
HAL as he leaves. Throughout his journey Dave receives a stream
of messages from H A L telling him which even numbers have been
tested and whether or not a counter-example to the conjecture has
been found. As the hour draws on, the messages are received at an
ever increasing rate until finally, in the hour's closing moments, the
rate blows up forcing the messages to compact together. A moment
later and the messages cease. Then, suddenly, Dave knows the truth
of the Goldbach conjecture.
Let us call the above scenario with HAL and Dave, (S), and
a spacetime which permits (S) to occur, Pitowsky. Before giving a
precise definition of a Pitowsky spacetilT,e, we need to gain a little
more insight into these spacetimes by attempting to accommodate
(S) in the simplest of relativistic arenas: Minkowski spacetime.
General Relativity and Finite Time 175
Dave ldline
space
Fig. 1. Conformal diagram of Minkowski spacetime.
3. IS M I N K O W S K I SPACETIME PITOWSKY?
Without any loss of generality, let us work in HAL's rest frame.
We need to construct a worldline for Dave such that, after one hour
of his proper time, he is to the future of the whole of HAL's worldline.
However it is immediately clear that no such point exists, except per-
haps the "point" at which HAL ceases to exist, i.e., at x = 0,t = co.
I write "point" because it is added to the usual representation of
Minkowski spacetime and consequently cannot be regarded as a being
a member of it. This point at infinity is labelled q in the conformal
diagram depicted in Fig. 1. If Dave oscillates to-and-fro about the
line x = 0, along a worldline which approaches a null zig-zag line
exponentially quickly, then it is easy to show that his total proper
distance from p(x = 0, t = 0) to q(x = 0, t -- ee) is finite.
But even granting that Dave can reach q in finite time is not
enough to capture the spirit of (S). For supposing Goldbach's conjec-
ture to be true de facto, Dave will never come to know the truth within
Minkowski spacetime (as against the point q). To guard against this
possibility, I will demand of a Pitowsky spacetime that the point q is
not a point at infinity but a normal spacetime point. Imposing this
176 Hogarth
condition means Dave will live through the event q, thus allowing him
to reflect upon all the data he received from HAL prior to then.
This proviso is implicit in the following definition of a Pitowsky
spacetime and, as one might expect, Minkowski spacetime fails to
satisfy its conditions.
4. T H E D E F I N I T I O N OF A PITOWSKY SPACETIME
The discussion above suggests the following preliminary defini-
tion.
(Prelim). Let (-/1/I,g=b) be a spacetime. Then (M,g=b) is
Pitowsky if there exists two future-directed timelike curves
A, # c M which share the same past endpoint, and a point
qC# such that:
(1) = (2)f. dr < oo (a) A c :-(q).
The curves A and # represent, respectively, the worldlines of HAL
and Dave; q represents the point at which Dave has finally gathered
all the data from HAL's worldline. That A and # share the same past
endpoint reflects that HAL and Dave are initially together. Condi-
tions (1) and (2) reflect the longevity of the two observers: HAL lives
forever, while Dave lives for only a finite time. 4 Condition (3) ensures
that HAL can always signal to Dave and, moreover, that all these
signals reach Dave before Dave passes through q.
In fact, the definition of a Pitowsky spacetime has an equivalent
but simpler formulation; call it (Definition).
(Definition). Let (M,g~b) be a spacetime. Then (M,g~b)
is Pitowsky if there exists a future-directed timelike curve
A c M with past endpoint, and a point qCM such that:
(i) J~ dr = ec (ii) A c J-(q).
This follows because all those spacetimes which satisfy the con-
ditions of (Definition) admit a curve with the properties of the curve #
in (Prelim); i.e., a future-directed timelike curve of finite length which
has the same past endpoint, p, say, as A, and which passes through q.
To see this, first notice that since A is timelike (ii) implies that p can
General Relativity and Finite Time 177
\
t
-oo~---- q. ~+oo
pt
Fig. 2. A causally vicious Pitowsky spacetime.
be joined to q by a future-directed timelike curve, u, say. If the length
of u is finite, then ~ is the required curve. If the length is infinite;
then one can construct the required curve by using the well-known
result that any timelike curve with fixed endpoints can be deformed
into a timelike curve of arbitrarily small length (by being deformed
into a "zig-zag-stay-close-to-the-light-cone" curve; cf. Fig. 1).
(Definition) tells us that, roughly speaking, a spacetime is Pitow-
sky if it admits a point to the future of a curve of infinite length.
5. T H R E E E X A M P L E S AND A THEOREM
A simple example of a Pitowsky spacetime is afforded by the
cylindrical spacetime formed by rolling up a temporal segment of two
dimensional Minkowski spacetime. It is easy to see from Fig. 2 that
the conditions of (Definition) are satisfied.
This is perhaps not a very exciting example though, since the
spacetime's closed timelike curves make it rather unphysical. We will
shortly see an example of a causally well-behaved Pitowsky space-
time. However a great many causally well-behaved spacetimes are
not Pitowsky, as shown by the following
Theorem. No globally hyperbolic spacetime is Pitowsky.
(Remark. A spacetime (M, g~b) is strongly causal if about each
point p C M every neighbourhood of p contains a neighbourhood
178 Hogarth
of p which no causal curve intersects more than once (Hawking and
Ellis, p.192). A spacetime (M,g~b) is globally hyperbolic if (M,g~b) is
strongly causal and, for any two points x , y E M, J - ( x ) I-1J+(y) is
compact or empty (ibid., p.206).)
Proof. Let (M, gab) be a globally hyperbolic spacetime. Suppose
there exists a future-directed timelike curve A and a point q such that
A C J-(q). (It will be shown that the length of A must be finite - -
thus preventing (M, g~b) being Pitowsky.) Since strong causality holds
on M , the set J-(q) A J+(p) has an open cover {U~}, where each U~
is a convex normal neighbourhood with compact closure such that A
enters each U~ at most once. From this cover construct another cover
{V~} of J-(q) n J+(p), where each V~ C U~ for some a and such
that, for each Vz with V~ N A ~ 0, the length of A in V~ does not
exceed e. Now, since (M, g~) is globally hyperbolic, J - ( q ) N J+(p) is
compact, which implies that {V~} has a finite subcover. This finite
cover contains A since ), C J-(q) N J+(p). Therefore A can be covered
with a finite number of V~s, each of which contains a segment of A of
finite length. Thus A has finite length and so (M,g~b) is not Pitowsky.
[]
Most of the standard spacetimes - - Minkowski spacetime, Fried-
mann models, Schwarzschild solution - - are globally hyperbolic and
so, by the result above, are non-Pitowsky.
Looking at solutions farther afield, however, I claim: anti-de
Sitter spacetime is Pitowsky. ~ This spacetime - - which is, inciden-
tally, stably causal and ipso facto causally well-behaved - - can be
covered (ibid., p.131) by a single "spherical polar" coordinate system
(t, r, 0, ¢), in which case the line element assumes the form
ds 2 = cosh2r dt 2 _ dr 2 _ sinh2r(d92 + sin29 d¢2).
As will now be shown, we may restrict attention to the plane
0 -- ¢ = 0. In this case the line element simplifies to
d s 2 = c o s h 2 r d t ~ - d r 2.
In keeping with the notation of (Definition), let p and q be the
points (0, 0), (0, 3~r), respectively, and let A be the future-inextendible
timelike curve with past endpoint p whose tangent vector satisfies
dt 2½
dr cosh r"
General Relativity and Finite Time 179
q(0, ~ )
At ~ null geodesic
t=~
Jl t=~---2
At 2
1 ~(0,0) r
>
Fig. 3. Anti-de Sitter spacetime.
Claim: p, q, A satisfy the conditions of (Definition). To satisfy
(i) we must show that the length, L(A), of A is infinite. We have
~dr= // (cosh2r( dt~)- 1)½dr = # d~ = oo.
Next, to satisfy (ii), we need to show that A C J-(q). Setting
ds = 0, we find that the light cone at q is given by dr~dr = + l / c o s h r
(see Fig. 3). The total decrease in t along q's past light cone is
At1 = fo~°dt
~r dr = fo °° cosh
1 r dr:[2arctan(e~)lr=~r2"
Thus every null geodesic with future endpoint q is contained in
the set {(t,r)br < t _< 73 r , 0 < r < oo}. For any point p = p(t, r) with
t < ~-, the timelike curve through p given by r = constant intersects
the past light cone of q. Hence p E J-(q). But this applies to every
point p E A since the t coordinate along A never exceeds
At2 = fo°° dt dr = fo~ 2½ 2½
dr cosh-----r - 2 ~r (< ~').
(ii) is therefore also satisfied and so anti-de Sitter spacetime is
Pitowsky.
The maximally extended Reissner-Nordstrhm solution is also
Pitowsky. This can be seen by examining Fig. 4, which shows part of
180 Hogarth
1| ~~ Singularity
!,q ~i~----at r=0
Cauchy horizon-~,
'~'
Asymptotically
flat region
Fig. 4. Part of the conformal diagram of the maximally extended
Reissner-NordstrSm solution.
the spacetime's conformal diagram. The future-inextendib!e timelike
curve t is contained in an asymptotically flat region and is of infinite
length. The spacetime is Pitowsky because A lies to the past of q, a
point on the Cauchy horizon. 6
6. T W O C O N C L U D I N G REMARKS
It is curious to note that although HAL solves the immortal
question, she never actually possesses the answer; that is left for Dave,
the mortal bystander. Of course this is because HAL can only ever
see part of her life (her past), while Dave can see the whole of it.
It would be interesting to add to our list of three Pitowsky space-
times some spacetimes with distinctly physical features. However, I
take it as agreed that the enactment of (S) would be hopelessly imprac-
tical in our universe even if the spacetime structure were sympathetic
to it. It is therefore difficult to see how extending this list would add
much, if anything at all, to the discussion. No matter: This nega-
tive point is eclipsed by the positive demonstration that the curious
events of (S) can find a home in a causally well-behaved general rela-
tivistic model. This kind of demonstration - - like GSdel's refutation
of the necessity of absolute time - - serves to show us that time has
possibilities far beyond those imagined by common sense.
General Relativity and Finite Time 181
ACKNOWLEDGEMENTS
This paper has profited greatly from David B. Malament's inves-
tigations of Pitowsky spacetimes and I wish to thank him for sharing
his ideas with me. We discovered independently the (Definition) and
the Theorem, but the anti-de Sitter spacetime example was due solely
to Malament. I am grateful to Itamar Pitowsky who brought the
problem to my attention, and to Jeremy Butterfield and John Norton
for helpful comments on previous drafts of the paper.
REFERENCE
S.W. Hawking and G.F.R. Ellis, The Large Scale Structure of Space-
Time (Cambridge University Press, Cambridge, 1973).
NOTES
1. It is conceivable that a computer whose dimensions shrunk suffi-
ciently rapidly to zero could perform an infinite number of compu-
tations in a finite time - - although presumably the ensuing singu-
larity would prevent any data being transmitted beyond the event
horizon. I will ignore this possibility.
2. If the concept of "an immortal computer" is thought too fantas-
tic, replace it with "an infinite number of mortal computers lying
temporally end-to-end."
3. Here and hereafter "length" means "proper length."
4. The exact length here is unimportant to the subsequent investiga-
tion; all that matters is that it is finite.
5. I am treating anti-de Sitter spacetime in its universal covering form.
6. This observation is actually nothing new. See ibid., p.161.
