AClass of Time-Machine Solutions with a Compact Vacuum Core

Amos Ori
Department of Physics, TechnionIsrael Institute of Technology, Haifa, 32000, Israel
(Received 17 March 2005; published 7 July 2005)
We present a class of curved-spacetime vacuum solutions which develop closed timelike curves at some
particular moment. We then use these vacuum solutions to construct a time-machine model. The causality
violation occurs inside an empty torus, which constitutes the time-machine core. The matter eld
surrounding this empty torus satises the weak, dominant, and strong energy conditions. The model is
regular, asymptotically at, and topologically trivial. Stability remains the main open question.
DOI: 10.1103/PhysRevLett.95.021101 PACS numbers: 04.20.Jb, 04.20.Gz
The problem of time-machine formation is one of the
outstanding open questions in spacetime physics. Time
machines are spacetime congurations including closed
timelike curves (CTCs), allowing physical observers to
return to their own past. In the presence of a time machine,
our usual notion of causality does not hold. The main
question is: Do the laws of nature allow, in principle, the
creation of a time machine from normal initial condi-
tions? (By normal I mean, in particular, initial state with no
CTCs.)
Several types of time-machine models were explored so
far. The early proposals include Godels rotating-dust cos-
mological model [1] and Tiplers rotating-string solution
[2]. More modern proposals include the wormhole model
by Moris, Thorne, and Yurtsever [3], and Gotts solution
[4] of two innitely long cosmic strings. Ori [5] later
presented a time-machine model which is asymptotically
at and topologically trivial. Later Alcubierre introduced
the warp-drive concept [6], which may also lead to CTCs.
The above models, however, all suffer from one or more
severe problems. In some of them [3,6] the weak energy
condition (WEC) [7] is violated, indicating unrealistic
matter-energy content. The WEC states that for any physi-
cal (timelike) observer the energy density is non-negative,
which is the case for all known types of (classical) matter
elds. In other models the CTCs are either preexisting
[1,2,4] and/or come from innity [4] (see [8,9]), and/
or there is a curvature singularity [2].
The only of the above models which does not violate the
WEC, and in which CTCs evolve from normal initial data
(and within a compact region of space) is that of Ref. [5].
But this model, too, is not satisfactory, for the following
reason: The energy-momentum source occupying the time-
machine core, though consistent with the WEC (and also
with the dominant energy condition [10]), does not t any
known type of matter eld. Therefore this conguration
(spacetime plus matter) is not a solution of any prescribed
set of eld equations. Even if we assume there exist
physical matter elds that yield the desired initial congu-
ration, we cannot tell how these elds (and the geometry)
will evolve in time, because we do not have at hand the set
of evolution equations. It therefore leaves open the ques-
tion of whether the system will or will not form CTCs. All
we can say is that the WEC does not preclude the formation
of CTCs.
We would therefore like a good time-machine model to
be made of a well-known matter eld, preferably an ele-
mentary one. Obviously the most elementary eld (in the
present context) is the pure gravitational eld, i.e., a vac-
uum spacetime. It is primarily this issue which we address
in the present paper.
There are several other desired features which we would
like our model to satisfy: We want the spacetime (and
particularly the initial hypersurface) to be asymptotically
at (or, alternatively, asymptotically Friedmann). In addi-
tion, we would like the onset of causality violation to take
place within a nite region of space. That is, we would like
the initial hypersurface to include a compact piece S
0
, such
that the onset of causality violationthe appearance of a
closed causal loopwill be fully dictated by the initial
conditions on S
0
. This requirement makes sense, because
presumably even an advanced civilization will not be able
to control the initial data in the entire space, but only in a
nite region (in the best case). Note that this criterion for
compactness differs from the notion of compact genera-
tion introduced by Hawking [11], as we further discuss
below. We shall refer to the compact region of spacetime
including S
0
and also its future domain of dependence and
its neighborhood, as the core of the time-machine space-
time (in our model this will include the rst causal loop and
the CTCs in its immediate neighborhood).
In this Letter we present a class of vacuum solutions in
which CTCs form at some particular moment. We then use
such a vacuum solution as the core of a new type of time-
machine model which satises all the above requirements.
In this model the evolution starts from a regular initial
spacelike hypersurface (partial Cauchy surface) which
(like spacetime itself ) is asymptotically at, topologically
trivial, and satises the weak, dominant, and strong energy
conditions (the energy conditions). Then CTCs form in the
central region at some particular moment.
PRL 95, 021101 (2005)
P HYS I CAL RE VI E W L E T T E RS
week ending
8 JULY 2005
0031-9007=05=95(2)=021101(4)$23.00 021101-1 2005 The American Physical Society
Our model consists of an empty (i.e., vacuum) torus
which constitutes the time-machine core. This toroidal
vacuum region is immersed in a larger, spherelike, region
of matter satisfying the energy conditions. The matter
region is nite, and is surrounded by an external asymptoti-
cally at vacuum region (specically in our construction it
is the Schwarzschild geometry). The matter in the inter-
mediate range, though satisfying the energy conditions, is
not associated as yet with any specic known matter eld
(though it appears likely that it will be possible to compose
it from, e.g., some combination of neutral and charged
dusts, and electromagnetic elds). Hence we cannot tell
yet what is the matters evolution equation. Nevertheless,
the causality violation occurs in the internal vacuum core,
and it is completely dictated by the initial conditions at the
compact set S
0
(located in the vacuum core too). Here we
shall primarily discuss the vacuum solution at the time-
machine core (which is the main new feature in this
Letter); the structure of the surrounding matter and its
matching to the external asymptotically at universe will
be presented elsewhere [12].
There are two rather general analyses, by Tipler [13] and
by Hawking [11], which put constraints on the creation of a
time machine in a compact region of space without violat-
ing the WEC. At rst sight each of these analyses might
appear to preclude a model like the one presented here. A
closer look, however, reveals that there is no inconsistency.
This was already demonstrated and explained in Ref. [5]
(which, too, presents a time-machine model satisfying the
WEC). In short, Hawkings analysis refers to compactly
generated time machines, which probably is not the case
here (see below). Our model is consistent with Tiplers
analysis because it includes a closed null geodesic (de-
noted N below) which is future incomplete (though no
local irregularity occurs there; See also the discussion in
Ref. [5]).
We turn now to describe the geometry of the time-
machine core. Consider the vacuum solution
ds
2
dx
2
+dy
2
2dzdT +jf(x; y; z) T|dz
2
: (1)
The coordinates (x; y; T) get all real values (though we later
truncate the solution in x and y at the internal boundary of
the matter region), but z is a periodic coordinate, 0 z
L for some L > 0, with z L and z 0 identied. f is
any function (properly periodic in z) satisfying
f
;xx
+f
;yy
0: (2)
[For f not satisfying Eq. (2) the metric (1) represents null
dust with G
zz
given by (1=2)(f
;xx
+f
;yy
) and all other
components vanishing. We shall not discuss this case here.]
[14] It can be shown that this class is locally isometric to
vacuum plane-fronted waves [15]. However, its global
properties are completely different from those of plane-
fronted waves, as we now discuss.
One immediately observes that the metric (1) develops
CTCs at sufciently large T. For each x; y, the metric
function g
zz
f(x; y; z) T is positive (for all z) at small
T, but becomes negative (for all z) at sufciently large T.
Consequently the closed curves of constant x; y; T are
spacelike at small T but become timelikenamely
CTCsat large T. (Below we consider a specic example
in more detail.) Note that the metric is everywhere regular
with det(g) 1, so no local pathology is involved in the
formation of CTCs.
For generic f the spacetime is curved, with
R
izjz
(1=2)f
;ij
(3)
(and its obvious permutations), and all other components
vanishing, where i; j stand for x and y. The metric becomes
locally at in the degenerate case f 0 [the same holds
for any f f(z)]. This is just the Misner space [16],
generalized to four dimensions in a straightforward man-
ner. The degenerate f 0 spacetime shares some of the
features of the generic-f nondegenerate case, though its
global structure is different and more pathological (see
below).
For concreteness we nowspecialize to a simple example.
We take
f a(x
2
y
2
)=2 (4)
for some constant a >0. This yields an empty curved
spacetime, locally isometric to a linearly polarized plane
wave. We now transform from T to a new time coordinate
t T a(x
2
y
2
)=2 +e
2
for some e > 0, where
2
x
2
+y
2
. This turns out to be
convenient, because the hypersurfaces t const provide a
useful foliation for the analysis below. The line element
becomes
ds
2
dx
2
+dy
2
2dzdt +(e
2
t)dz
2
+2j(2e a)xdx +(2e +a)ydy|dz: (5)
This metric is empty and curved (as before), and again one
nds det(g) 1. Since g
zz
e
2
t, the closed curves
of constant x; y; t are spacelike at t < e
2
and timelike at
t > e
2
(and null at t e
2
). These curves are nongeode-
sic, except for the single closed null geodesic x y t
0, which we denote N.
A hypersurface t const is spacelike when g
tt
< 0 and
timelike whenever g
tt
>0. One nds
g
tt
t +(2e a)
2
x
2
+(2e +a)
2
y
2
e
2
:
Choose now sufciently small positive a and e such that
e >(2e +a)
2
. Then the hypersurfaces t const 0 may
be characterized as follows: (I) For t < 0 they are spacelike
throughout; (II) The hypersurface t 0 is spacelike every-
where, except at the central curve, x y 0 (the geodesic
N), where it is null (g
tt
0).
PRL 95, 021101 (2005)
P HYS I CAL RE VI E W L E T T E RS
week ending
8 JULY 2005
021101-2
(For t > 0 the hypersurfaces t const are mixed: time-
like for sufciently small and spacelike at large .)
In the degenerate case f 0 (the 4d Misner space), the
hypersurface T 0 is null, and all its generators are closed
null geodesics (the curves of constant x; y). This case is
pathological, because the analytic extension beyond T 0
is nonunique [16]. No such pathology occurs in the non-
degenerate case. For a generic f, the closed null geodesics
are isolated [like N in the specic example (4)] and do not
form a hypersurface. Note also that in the degenerate case
f 0 the causal structure is inherently unstable in the
following sense: Adding an arbitrarily small xed number
to g
xz
and/or g
yz
(which vanish otherwise) leaves the
geometry vacuum and locally at, but the hypersurface
of closed null geodesics entirely disappears. [This metric
is isometric to (part of) the Grant spacetime [9], which
includes no closed causal geodesics.] Such a global insta-
bility does not occur in the nondegenerate case, which we
consider throughout this Letter.
The above constructed vacuum solution is now used as
the core of our time-machine spacetime, with x y 0
located at the central circle of the torus. The solution (5) is
truncated at
0
for some
0
> 0, where the matter
region starts.
As was discussed above, we would like the region of
CTCs to evolve, in a deterministic manner, from a compact
spacelike hypersurface S
0
in the vacuum core. This would
be fully accomplished if a region of CTCs were included in
D
+
(S
0
) (D
+
denotes the future domain of dependence [7],
also known as the future Cauchy development). But ob-
viously this can never be the case because by denition D
+
of any spacelike hypersurface cannot include any closed
causal curve. It may be possible, however, that the bound-
ary of D
+
(S
0
) will include a closed causal curve (in fact a
closed null geodesic); And this appears to be the maximum
one can hope for, in terms of the causal relation between S
0
and the region of causality violation. We shall now show
that our model indeed has this desired feature, namely, one
can choose a compact spacelike hypersurface S
0
in the
vacuum core, such that a closed causal curve (the null
geodesic N) appears at the boundary of D
+
(S
0
).
Furthermore, any regular extension of the geometry be-
yond N will include a region of CTCs.
For any
1
> 0 the hypersurfaces t const 0 are all
spacelike at
1
. Also, the hypersurfaces t const <
0 are spacelike even at <
1
. Any such spacelike hyper-
surface may be slightly deformed and still remain space-
like. Consider, in particular, the composed hypersurface
given by t const t
0
< 0 at
1
, by t 0 at

2
, and by some interpolating function t
~
t() in the
range
1

2
, for some parameters 0 <
1
<
2
<

0
. The function
~
t() is chosen to be a monotonic one,
which smoothly bridges between t t
0
at
1
and t
0 at
2
. For sufciently small }t
0
}, the function
~
t()
may be taken to be of sufciently small slope, such that the
hypersurface described by t
~
t() is spacelike in
1

2
. By this we have constructed a composed spacelike
hypersurface in the entire range
0
. We then restrict
this hypersurface to the range
3
for some
2
<
3
<

0
, and denote this compact spacelike hypersurface by S
0
.
Take now any point P at t <0 in a sufciently small
neighborhood of N (located at t 0), and consider
any inextendible past-directed causal curve emanating
from P. (In particular P is located at <
1
and t > t
0
, i.e.,
at the future of S
0
.) Since in the range t
0
t <0 all
hypersurfaces t const are entirely spacelike, one can
easily show that has no other choice but to move towards
smaller t values, until it intersects S
0
. Therefore the point P
belongs to D
+
(S
0
). This means that N resides at the
boundary of D
+
(S
0
).
The compact initial hypersurface S
0
can be extended
through the matter region and the external vacuum region
to spacelike innity, to form a regular, asymptotically at,
partial Cauchy surface, everywhere satisfying the energy
conditions, which we denote . This will be demonstrated
elsewhere [12].
We conclude that in our model the onset of causality
violationthe appearance of the closed null geodesic N
is fully dictated by the initial conditions in the compact
vacuum piece S
0
of the initial hypersurface . In particu-
lar, the details of the initial data at the external vacuum
region, or at the intermediate matter region, cannot inter-
fere with this onset of causality violation. Although the
above construction only demonstrates the inevitable for-
mation of a single closed causal orbit, any smooth (C
1
metric) extension of the geometry at Nto t > 0 will include
a region of CTCs, because g
zz;t
< 0 at N.
Hawking [11] previously introduced another notion of
compactness: A compactly generated time machine is one
in which all null generators of the Cauchy horizon, when
traced back to the past, enter a compact region of space-
time and remain there. He then showed that a compactly
generated time machine (with noncompact initial hyper-
surface) must include a region where the WEC is violated.
Our model evades this WEC violation, probably because it
is not compactly generated. In the intermediate matter
region, the initial data for the metric are constructed to
be regular and smooth (and to satisfy the WEC) [12]. This
holds on and also for some time interval to its future. But
our construction by no means guarantees that the regular
WEC-satisfying metric at the matter region can extend to
all future times. From Hawkings arguments [11] it appears
to follow that the matter geometry, when forced to satisfy
the WEC, must develope some noncompactness (e.g., a
singularity or an internal innity) at later times. Then
these arguments, applied to our construction, may further
suggest that generators of the chronology horizon will
emanate from this noncompact region. But this proposed
scenario must be veried by an explicit study of the struc-
ture of the Cauchy horizon. We emphasize again that
PRL 95, 021101 (2005)
P HYS I CAL RE VI E W L E T T E RS
week ending
8 JULY 2005
021101-3
although probably not compactly generated, our model
does demonstrate the formation of closed causal loops
from the initial data on a compact vacuum region S
0
.
Although it is only a single closed null geodesic which
resides in the closure of D
+
(S
0
), any smooth (C
1
) exten-
sion of the metric to t >0 will include a region of CTCs.
The fact that the weak, dominant, and strong energy
conditions are satised suggests that the initial conditions
required for our model are physically acceptable. This does
not mean that we shall be able to practically initiate such
initial conditions in the foreseeable future. However, per-
haps an advanced civilization will be able to do this (and
perhaps even natural processes, involving large gravitating
masses in rapid motion, may lead to such conditions).
There still remains, however, the issue of stability.
Several analyses [11,17,18] indicated possible instabilities
of various time-machine solutions to classical perturba-
tions and/or quantum-mechanical uctuations (see how-
ever [19]). Whereas these analyses mostly referred to
compactly generated models, some of the arguments for
quantum instabilities apply to noncompactly generated
models as well (see, in particular, the discussion in [18]).
These instabilities may raise doubts on whether a model
like the one presented here can be implemented in reality.
Yet, it appears that so far these indications for instability do
not rule out the possibility of actual time-machine con-
struction. The strength of the quantum instability is not
clear yet, and, more importantly, it is not known yet what
will be the outcome of this instability (namely, what will be
the spacetime conguration that eventually forms).
Perhaps we shall have to await the formulation of the full
theory of quantum gravity before we know whether quan-
tum instabilities provide chronology protection or not (see
discussion in [20]). The situation with regards to classical
instabilities is somewhat different: Their occurrence still
needs be established, especially in the noncompactly gen-
erated case and beyond the context of geometrical optics. If
classical instabilities are found to be inevitable, their
strength and their outcome can be explored by evolving
the classical Einstein equations for slightly perturbed ini-
tial conditions.
I thank Serguei Krasnikov for interesting and useful
discussions.
[1] K. Godel, Rev. Mod. Phys. 21, 447 (1949).
[2] F. J. Tipler, Phys. Rev. D 9, 2203 (1974).
[3] M. S. Morris, K. S. Thorne, and U. Yurtsever, Phys. Rev.
Lett. 61, 1446 (1988).
[4] J. R. Gott, Phys. Rev. Lett. 66, 1126 (1991).
[5] A. Ori, Phys. Rev. Lett. 71, 2517 (1993).
[6] M. Alcubierre, Classical Quantum Gravity 11, L73 (1994).
[7] S. W. Hawking and G. F. R. Ellis, Large Scale Structure of
Space-Time (Cambridge University Press, Cambridge,
England, 1975).
[8] A. Ori, Phys. Rev. D 44, R2214 (1991).
[9] J. D. E. Grant, Phys. Rev. D 47, 2388 (1993).
[10] Y. Soen and A. Ori, Phys. Rev. D 54, 4858 (1996).
[11] S. W. Hawking, Phys. Rev. D 46, 603 (1992).
[12] A. Ori (unpublished).
[13] F. J. Tipler, Phys. Rev. Lett. 37, 879 (1976); Ann. Phys.
(N.Y.) 108, 1 (1977).
[14] The above vacuum solution can be extended as follows: To
the metric (1) add g
xz
wy, g
yz
wx, and also add
w
2
(x
2
+y
2
) to g
zz
, for any w 0.
[15] H. Stephani, D. Kramer, M. MacCallum, C. Hoenselaers,
and E. Herlt, Exact Solutions of Einsteins Field Equations
(Cambridge University Press, Cambridge, 2003), 2nd ed.
[16] C. W. Misner, in Relativity Theory and Astrophysics I:
Relativity and Cosmology, edited by J. Ehlers (American
Mathematical Society, Providence, 1967).
[17] S. W. Kim and K. S. Thorne, Phys. Rev. D 43, 3929
(1991).
[18] B. S. Kay, M. J. Radzikowski, and R. M. Wald, Commun.
Math. Phys. 183, 533 (1997).
[19] S. V. Krasnikov, Phys. Rev. D 54, 7322 (1996).
[20] M. Visser, gr-qc/0204022.
PRL 95, 021101 (2005)
P HYS I CAL RE VI E W L E T T E RS
week ending
8 JULY 2005
021101-4