PHYSICAL REVIEW D 86, 124048 (2012)

New type of phase transition in gravitational theories

Xián O. Camanho,1 José D. Edelstein,1,2 Gastón Giribet,3 and Andrés Gomberoff4
1
Department of Particle Physics and IGFAE, University of Santiago de Compostela, E-15782 Santiago de Compostela, Spain
2
Centro de Estudios Cientı́ficos CECs, Avenida Arturo Prat 514, Valdivia, Chile
3
University of Buenos Aires FCEN-UBA and IFIBA-CONICET, Ciudad Universitaria, Pabellón I, 1428, Buenos Aires, Argentina
4
Departamento de Ciencias Fı́sicas, Universidad Andres Bello, Avenida República 252, Santiago, Chile
(Received 12 May 2012; published 26 December 2012)
We set forth a new type of phase transition that might take place in gravitational theories whenever
higher-curvature corrections are considered. It can be regarded as a sophisticated version of the Hawking-
Page transition, mediated by the nucleation of a bubble in anti-de Sitter (AdS) space. The bubble hosts a
black hole in its interior, and separates two spacetime regions with different effective cosmological
constants. We compute the free energy of this configuration and compare it with that of thermal AdS. The
result suggests that a phase transition actually occurs above a certain critical temperature, ultimately
changing the value of the cosmological constant. We discuss the consistency of the thermodynamic picture
and its possible relevance in the context of AdS/CFT.

DOI: 10.1103/PhysRevD.86.124048 PACS numbers: 04.50.
h, 04.60.
m

It is a well-known fact that black holes in AdS display

I. INTRODUCTION
Higher-curvature corrections to the Einstein-Hilbert
(EH) action appear in any sensible theory of quantum
gravity as next-to-leading orders in the effective action.
Quadratic terms, for instance, such as the Lanczos-Gauss-
Bonnet (LGB) action [1], appear in bona fide realizations
of string theory [2–5] and M theory [6,7].
Interesting implications of these terms within the con-
text of the AdS/CFT correspondence have been recently
the focus of thorough investigation (see, e.g., Refs. [8–16]).
They allow for the holographic description of a broad
family of quantum field theories (like 4d superconformal
field theories with unequal central charges [17]), as well as
for the study of the fluid/gravity correspondence beyond
the EH gravitational sector.
A remarkable example of higher-curvature gravity is
Lovelock theory [18], the natural extension of general
relativity to higher dimensions. More precisely, it is the
most general theory of gravity yielding second-order field
equations. General relativity and Chern-Simons gravity
[19] are particular cases of Lovelock theories. Although
quantum corrections are not generally of the Lovelock
type, these theories provide a tractable playground that
captures many important features that are rather generic.
Among them, the existence of new branches of black hole
solutions corresponding to vacua with different cosmologi-
cal constants that pop out as soon as higher-curvature terms
are brought into place.
The main motivation of the present work is that of study-
ing transitions between the different vacua of the theory.
This is important to investigate whether a new type of
instability involving nonperturbative solutions occurs in
the theory. Studying such transitions may uncover a mecha-
nism that eventually leads to a built-in mechanism to reduce
the (absolute) value of the effective cosmological constant.
the so-called Hawking-Page (HP) transition [20], i.e., the
semiclassical phase transition between thermal AdS space
at finite temperature and a black hole in AdS with the same
Hawking temperature. This has been further interpreted as
a confinement/deconfinement phase transition in the dual
CFT [21].
A natural question arises regarding the role of the differ-
ent black hole branches and would-be transitions among
them. At first glance it might seem that such phase tran-
sitions are forbidden since distinct branches exhibit a
different asymptotic behavior.
In this article, however, we report on a new type of phase
transition taking place in higher-curvature gravities, which
can be thought of as a sophisticated version of the HP
transition involving different branches. We work out the
simplest example given by LGB gravity—whose ordinary
HP transition was studied in Ref. [22]—since it is enough
to realize that these phase transitions occur in Lovelock
theory as well [23], and possibly in a larger class of higher-
curvature theories.
We consider the theory at finite temperature and observe
that, above a critical temperature, Tc , the higher-curvature
vacuum decays, producing a bubble which hosts a black
hole in its interior. This phase transition is analogous to the
thermalon transition discussed in Ref. [24], where the
materialization of an electrically charged bubble induces
the decay of a de Sitter vacuum into another with a smaller
cosmological constant, but containing a black hole. In the
present case, however, the bubble is not made from matter,
but from the gravitational field itself. It separates two
regions having a different effective cosmological constant
and ultimately changes its value throughout the whole
space, thus changing the asymptotics.
In the canonical ensemble, whether or not the transition
takes place can be decided by evaluating the Euclidean

1550-7998= 2012=86(12)=124048(6) 124048-1 Ó 2012 American Physical Society

CAMANHO et al. PHYSICAL REVIEW D 86, 124048 (2012)
action on two well-defined classical configurations. dr2
According to our free energy computation, the phase tran- ds2 ¼
fðrÞdt2 þ þ r2 d2d
2 ; (3)
fðrÞ
sition actually occurs, the preferable static configuration
above Tc being a black hole surrounded by a bubble. In where d2d
2 is the round metric on a (d
2)-dimensional
Lorentzian signature we can analyze its subsequent dy- sphere. The equations of motion reduce to a single first-
namics. We see that the bubble eventually expands in such order differential equation for f, that can be easily solved
a way that it swallows up the whole spacetime in finite in terms of g ¼ ð1
fÞ=r2 as [26,27]
proper time, thus changing the asymptotics. X
K

½g ¼ ck gk ¼ ; (4)
k¼0 rd
1
A. Higher-curvature corrections
The Lovelock action can be written in terms of the an implicit polynomial solution with up to K branches
vierbein 1-form, ea ¼ ea
dx
, and the curvature 2-form, where  is an integration constant related to the mass of
Rab ¼ d!ab þ !a c ^ !cb , !ab
being the (torsion-free)
the black hole [28],
spin connection,

ðd
2Þ!2 ðd
2Þ!2 d
1 1
d d

M¼  ¼ r  ; (5)
X
K ð2Þd
ð2Þd H
r2H
ck
I¼ I k þ I @; (1)
k¼0 d
2k
rH being the location of the horizon. We focus on the
quadratic theory which is enough to illustrate the
with the bulk contributions, phenomenon reported in this article. The solutions take
the form [25]
Z 0 sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1
Ik ¼ EðdÞ
a1 ...a2k R 1 2 ^ . . . ^ R 2k
1 2k ;
a a a a  
1 @  A
M g ðrÞ ¼
1  1
4 1
d
1 : (6)
2L2 r
where EðdÞa1 ...ak ¼ a1 ...ad e kþ1 ^ . . . ^ e d , a1 ...ad being the
a a

antisymmetric symbol, while K is a positive integer,
K  ½ðd
1Þ=2, and I @ refers to boundary terms to be
discussed below. The coefficients ck are coupling constants
with length dimensions L2ðk
1Þ .
The kth term in the Lagrangian corresponds to the
extension of the Euler characteristic in 2k dimensions.
The zeroth contribution is the cosmological constant term
[we set 2
¼
ðd
1Þðd
2Þ=L2 , i.e., c0 ¼ 1=L2 ], the
first term is the EH action [we normalize the Newton
constant to 16ðd
3Þ!GN ¼ 1, i.e., c1 ¼ 1], and the
second term, quadratic in the Riemann curvature, is the
LGB action (we take c2 ¼ L2 , and call  the LGB
coupling).
Although Lovelock theory yields second-order equa-
tions of motion, they are nonlinear in the curvature. As a
result, the theory admits more than one maximally sym-
metric solution; it has up to K different (A)dS vacua with
effective cosmological constants
i , i ¼ 1; 2; . . . ; K. They
are the solutions of the Kth-order polynomial [25]
X
K Y
K Lovelock theory [29].
½
  ck
k ¼ cK ð


i Þ ¼ 0:
k¼0 i¼1
The theory exhibits degenerate behavior whenever two or
more of these effective cosmological constants coincide.
We are interested in the nondegenerate case.
B. Lovelock black holes
The first ingredient in our discussion is the black hole
solution of the theory. Consider the ansatz
Each of the two branches in (6) is associated with a differ-
ent value of the effective cosmological constant,
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1  1
4

 ¼
: (7)
2L2
Notice that  < 1=4 is needed in order to have real non-
degenerate values.
While g
has a well defined horizon, gþ displays a
naked singularity at the origin provided M Þ 0. This latter
branch is also unstable [25]; the graviton propagator is
proportional to 0 ½
þ  < 0, thus having the wrong sign
with respect to the EH case. In the dual CFT this amounts
to nonunitarity [15]. The stable solution g
also happens to
be the one that is continuously connected to the
Schwarzschild-Tangherlini solution of general relativity
in the  ! 0 limit; namely
 
1 
g

2 1
d
1 þ OðÞ: (8)
L r
These features are also manifest for generic black holes in
(2) The existence of a second vacuum of effective cosmo-
logical constant
þ in the higher-curvature theory was
referred to, in Ref. [25], as the theory having ‘‘its own
cosmological constant problem.’’ Moreover, for small
LGB coupling, the curvature of such vacuum becomes
very large,
þ 
1=ðL2 Þ. One may thereby argue that
considering such a background in the quadratic theory does
not make sense because higher-curvature terms cannot be
neglected. However, far from removing this second vac-
uum, adding higher-order terms further produces a plethora

124048-2
NEW TYPE OF PHASE TRANSITION IN GRAVITATIONAL . . . PHYSICAL REVIEW D 86, 124048 (2012)
of highly curved vacua. With the purpose of understanding We want to explore configurations consisting of a spheri-
the implications of considering one such vacuum in the cal bubble dividing the spacetime in two regions, the outer
theory, we consider spaces that asymptote AdS of typical being taken to asymptote AdS with effective cosmological
radius ð

þ Þ
1=2 . constant
þ (see Fig. 1). Solutions consisting of a spheri-
cally symmetric surface separating regions with different
vacua are known to exist [32,33]. They have been recently
II. GENERALIZED HAWKING-PAGE TRANSITION considered to explore instanton transitions,
þ !

, via
A. Boundary action bubble nucleation [34]. In the present case we are inter-
Let us now discuss the role of the boundary terms, I @ . ested in thermodynamic phase transitions. It is convenient
Their contribution is necessary for the variational principle to break the action in three pieces,
to be well defined. This is analogous to the Gibbons- I ¼ I in þ I  þ I out : (11)
Hawking term in general relativity [30], which in the
first-order formalism can be written as The first term is integrated inside the bubble, while the last
term is integrated outside and includes all the boundary terms
1 Z at infinity necessary to both have a well defined variational
I GH ¼ ab ^ EðdÞ
ab ; (9)
d
2 @M principle and to regularize the infinities. The term in the
middle is integrated on a small region around the bubble. We
where ab is the second fundamental form associated to the consider the limit when its width goes to zero. Proceeding in
extrinsic curvature. Similarly, the boundary term associ- this way, we deal with terms at the boundaries of each region,
ated to the LGB contribution reads [31] which give rise to a finite I  [35,36]. The variation of I 
 
2 Z 2 with respect to the vierbein gives the junction conditions on
IM ¼ ab ^ Rcd
c e ^ ed ^ EðdÞ abcd : the bubble [35,36]. These generalize the Israel conditions of
d
4 @M 3
general relativity.
(10)
In the same way as bulk terms are the dimensional exten- B. The phase transition
sion of the 2k-dimensional Euler characteristic for closed The configuration we will be concerned with is a bubble,
manifolds, the corresponding boundary terms are needed in whose outer region asymptotes AdS with a cosmological
the extension of the Gauss-Bonnet theorem to manifolds constant
þ , while the inner region hosts a black hole
with boundaries. The boundary action is then given by with mass M
, and an effective cosmological constant
I @ ¼ I GH þ L2 I M .

. The opposite situation does not possess a smooth
Euclidean section. Across the junction, the vierbein has
(a) β+ to be continuous. Sticking to spherical symmetry, the two
bulk regions are described by a solution of the form (6)
Μ += 0 T+ with Euclidean signature,
dr2
ds2 ¼ f ðrÞdt2 þ þ r2 d2d
2 ; (12)
f ðrÞ
where the  signs denote the outer/inner regions. The
r =0 Thermal anti-de Sitter
notation is consistent with the respective branches. We
r= start by considering arbitrary branch solutions with unde-
8

termined mass parameters on each region. The junction

conditions then fully constrain the allowed possibilities.
(b) β+ The junction is conveniently described by the parametric
equations
β− T+ r ¼ aðÞ; t ¼ T ðÞ; (13)
Μ− Μ+ with an induced metric of the form ds2 ¼ d2 þ
aðÞ2 d2d
2 , which has to be the same from both sides.
r = rH This yields
Black hole inside a bubble
a_ 2
r= a f ðaÞT_ 2 þ ¼ 1; (14)
r= f ðaÞ
8

FIG. 1. Euclidean sections for the two possible states of the where the dot stands for derivatives with respect to the
system: (a) empty thermal AdS, and (b) the bubble hosting a proper time . The function aðÞ appears explicitly in the
black hole in its interior. induced metric, so the radius has to be continuous across

124048-3
CAMANHO et al.
the surface. This condition allows us to write all the
expressions in terms of a and its derivatives and eventually
find a dynamical equation for the bubble itself [33].
We are interested in static configurations which, in view
of (14), translates into
qffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffi
 ¼ f
ðaÞT
¼ fþ ðaÞTþ : (15)
This means that the physical length of the Euclidean time
circle is the same as seen from both sides of the junction.
This matching condition will let us determine the tempera-
ture. Once the periodicity of the inner solution is fixed by
demanding regularity at the black hole horizon, that of the
outer solution gets fully determined,
qffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffi
f
ðaÞ
¼ fþ ðaÞþ ; (16)
such that there is a unique free parameter, the temperature.
While 
is the inverse of the Hawking temperature of the
inner black hole solution, þ is the inverse of the tempera-
ture measured by an observer at infinity.
By resorting to the usual Euclidean time formalism, we
can compute the free energy associated to the bubble
configuration that hosts a black hole, and then compare it
with that of thermal AdS at the same temperature. As we
show in what follows, the computation indicates that the
phase transition occurs above a critical temperature, Tc ðÞ.
The canonical ensemble at temperature 1= is defined
by the path integral over all metrics which asymptote AdS
identified in Euclidean time with period ,
Z ^
Z ¼ Dge
I½g ; (17)
where I^ ¼
iI. The dominant contributions come from
the saddle points. We have then to evaluate the Euclidean
action on a classical solution,
I^ cl ’
logZ ¼ F; (18)
which therefore gives the free energy, F. In the present
case, this basically amounts to computing the difference
between the Euclidean action of the bubble configuration
and that of AdS space identified with the same period in
imaginary time, both represented in Fig. 1.
The Euclidean action is in general divergent due to the
infinite volume of AdS; nevertheless, it can be suitably
regularized by background subtraction, meaning that the
free energy is actually measured with respect to the maxi-
mally symmetric solution. The periodicity at infinity is
fixed by demanding regularity of the black hole solution
at the horizon, r ¼ rH < a, supplemented by the gluing
condition (16), that determines in turn the outer periodicity.
In order to simplify the discussion, we calculate the on-
shell action in terms of two parameters: the position of the
bubble, a, and the temperature, even though they are not
independent from each other.
Unlike the computation of the HP effect in general
relativity, here we have to consider the contribution I^ 
124048-4
PHYSICAL REVIEW D 86, 124048 (2012)
arising on the bubble, when writing the Euclidean action in
the form (11). The boundary term in I^ out regularizes its
divergence by subtracting the background Mþ ¼ 0 with
the same periodicity at infinity. The term I^ in , in turn, is
integrated from the horizon to the location of the bubble.
Finally, I^  is given by
I^  ¼
I^ @
ða; 0 Þ þ I^ @þ ða; 0 Þ; (19)
where the periodicity in Euclidean time is inherited
pffiffiffiffiffiffiffiffiffiffiffiffiffi
from the bulk regions, 0 ¼ f ðaÞ . We can collect
all contributions depending upon the location of the
bubble, I^ bubble , the rest being consequently called
I^ black hole ¼ I^
I^ bubble ¼ 
M

S.
Remarkably enough, a neat result comes out after a quite
lengthy calculation—that nicely carries on to the generic
Lovelock theory [23]—once the junction conditions are
imposed,
I^ bubble ¼ þ Mþ

M
; (20)
which is the exact value needed to correct the on-shell bulk
action such that the thermodynamic interpretation is safely
preserved. In fact, due to this contribution, the total action
takes the form
I^ ¼ þ Mþ
S: (21)
That is, the bubble contributes as mass—carrying the mass
difference between the two solutions—but does not con-
tribute to the entropy. From the Hamiltonian point of view
this is naturally understood as follows. The canonical
action vanishes in this case; the only possible contributions
come from boundary terms both at infinity and at the
horizon, yielding respectively þ Mþ and the entropy,
which are nothing but the total charges of the solution.
The junction conditions simply imply the continuity of
canonical momenta [37].
Equation (21) shows that the junction conditions are
important to guarantee the consistency of the thermody-
namic picture. They also imply
þ dMþ ¼ 
dM
¼ dS; (22)
so that the first law of thermodynamics holds both for the
whole configuration (þ and Mþ ) as well as for the black
hole (
and M
).
C. Bubble nucleation
Having proven the consistency of the thermodynamic
picture in the case of the bubble configuration by deriving
(21) and (22), we are ready to address the question of
global thermodynamic stability. This amounts to analyzing
the free energy associated to the bubble configuration.
The free energy, F, as a function of the temperature
1=þ displays a critical temperature above which it
becomes negative and, thus, the phase transition occurs
NEW TYPE OF PHASE TRANSITION IN GRAVITATIONAL . . . PHYSICAL REVIEW D 86, 124048 (2012)
F
0.4 of a potential barrier. It will therefore either collapse or
Tc eventually expand in such a way that it engulfs the whole
14
spacetime in finite proper time, thus changing its asymp-
12
0.3
totic behavior [23].
10

6
4
0.2 2 We presented a novel mechanism for phase transitions
0
0.00 0.05 0.10 0.15 0.20
0.1
0.0
2 4 6 8 10
FIG. 2 (color online). Free energy versus temperature in 5d
for  ¼ 0:04,0.06, 0.09 (positivity bound), 0.219 [maximal
FðT ¼ 0Þ], and  ! 1=4 (from right to left). The  dependence
of the critical temperature is displayed in a separate box.
(see Fig. 2). If the free energy is positive, however, the
system is metastable. It decays by nucleating bubbles with
a probability given, in the semiclassical approximation,
by e
þ F . Therefore, after enough time, the system will
always end up in the stable, black hole solution. This is
reminiscent of the HP transition, except for the fact that,
here, the thermal AdS vacuum decays into a black hole
belonging to a different branch. Tc ðÞ is monotonically
decreasing, the phase transition becoming increasingly
unlikely the closer we come to the EH—classical—limit.
In this sense, it is a quantum mechanical phenomenon.
These high temperature configurations, or thermalons,
correspond to a black hole and a bubble in equilibrium,
connecting inner and outer solutions of different branches
g
and gþ . Such solutions exist just for positive values of
. The infinite temperature limit corresponds to planar
black holes, in which the junction conditions lead to a
remarkably simple relation that is valid also for general
Lovelock gravity [23], þ Mþ ¼ 
M
. The correspond-
ing free energy is always negative,
ðd
3Þ!2 rd
1
d

Mþ
F¼
H
2
¼
; (23)
ð2Þ
d L þ ðd
2Þ
thus implying that the preferable classical solution is
always the thermalon and the transition always occurs for
high enough temperature.
The junction conditions considered above determine not
only the equilibrium configuration but also, in Lorentzian
signature, the effective potential felt by the bubble and,
consequently, its subsequent dynamics. The scalar field
aðÞ specifying the location of the bubble sits at the top
III. DISCUSSION
0.25 that is a distinctive feature of higher-curvature theories
of gravity. These theories have several branches of asymp-
totically (A)dS solutions that might admit an interpretation
as different phases of the dual field theory. Phase transi-
tions among these are driven by the mechanism described
in the present article. Mimicking the thermalon configura-
T
12 tion [24], a bubble separating two regions of different
cosmological constants pops out, generically hosting a
black hole.
This configuration is thermodynamically preferred
above some critical temperature. The corresponding phase
transition can be interpreted as a generalized HP transition
for the high-curvature branches, driving the system
towards the EH branch. From the holographic point of
view this looks like a confinement-deconfinement phase
transition in a dual CFT, involving an effective change in
the ’t Hooft coupling, both phases being strongly coupled.
Whether a phenomenon like this takes place in a 4d CFT,
particularly within the framework of the fluid/gravity
correspondence—where both phases might be character-
ized by different transport coefficients—or whether it is
overtaken by higher-curvature corrections is an open ques-
tion at this point.
The bubble configuration, being unstable, dynamically
changes the asymptotic cosmological constant, transition-
ing towards the stable horizonful branch of solutions, the
only one usually considered as relevant. This is then a
natural mechanism for the system to select the general
relativistic vacuum among all possible ones. We are aware
of the fact that the vacuum
þ in the LGB theory exhibits
ghosts. The phenomenon presented in this article, however,
takes place in the Lovelock theory as well, where there are
further healthy vacua than the one connected to the EH
action [15].
We think that this mechanism is quite general and
deserves further investigation.
ACKNOWLEDGMENTS
We wish to thank Stanley Deser, Juan Maldacena,
Carlos Núñez and Josep M. Pons for their interesting
comments. G. G. thanks Ceci Garraffo, Elias Gravanis,
and Steve Willison for previous collaborations in the
subject. The work of X. O. C. and J. D. E. was supported
in part by MICINN and FEDER (Grant No. FPA2011-
22594), by Xunta de Galicia (Consellerı́a de Educación
and Grant No. PGIDIT10PXIB206075PR), and by the
Spanish Consolider-Ingenio 2010 Programme CPAN

124048-5
CAMANHO et al.
(CSD2007-00042). The work of G. G. was supported by
NSF-CONICET, PIP, and PICT grants from CONICET
and ANPCyT. The work of A. G. was partially supported
by Fondecyt (Chile) Grant No. 1090753. X. O. C. and
[1] C. Lanczos, Ann. Math. 39, 842 (1938).
[2] B. Zwiebach, Phys. Lett. 156B, 315 (1985).
[3] E. S. Fradkin and A. A. Tseytlin, Nucl. Phys. B261, 1
(1985).
[4] D. J. Gross and E. Witten, Nucl. Phys. B277, 1 (1986).
[5] D. J. Gross and J. H. Sloan, Nucl. Phys. B291, 41 (1987).
[6] S. Ferrara, R. R. Khuri, and R. Minasian, Phys. Lett. B
375, 81 (1996).
[7] I. Antoniadis, S. Ferrara, R. Minasian, and K. Narain,
Nucl. Phys. B507, 571 (1997).
[8] M. Brigante, H. Liu, R. C. Myers, S. Shenker, and S.
Yaida, Phys. Rev. D 77, 126006 (2008).
[9] M. Brigante, H. Liu, R. C. Myers, S. Shenker, and S.
Yaida, Phys. Rev. Lett. 100, 191601 (2008).
[10] A. Buchel and R. C. Myers, J. High Energy Phys. 08
(2009) 016.
[11] D. M. Hofman, Nucl. Phys. B823, 174 (2009).
[12] J. de Boer, M. Kulaxizi, and A. Parnachev, J. High Energy
Phys. 06 (2010) 008.
[13] X. O. Camanho and J. D. Edelstein, J. High Energy Phys.
06 (2010) 099.
[14] M. F. Paulos, J. High Energy Phys. 05 (2011) 043.
[15] X. O. Camanho, J. D. Edelstein, and M. F. Paulos, J. High
Energy Phys. 05 (2011) 127.
[16] J. de Boer, M. Kulaxizi, and A. Parnachev, J. High Energy
Phys. 07 (2011) 109.
[17] A. Buchel, R. C. Myers, and A. Sinha, J. High Energy
Phys. 03 (2009) 084.
[18] D. Lovelock, J. Math. Phys. (N.Y.) 12, 498 (1971).
124048-6
PHYSICAL REVIEW D 86, 124048 (2012)
J. D. E. would like to thank the FCEN-UBA and UNAB
for hospitality during part of this project. X. O. C. is
thankful to the Front of Galician-speaking Scientists for
encouragement.
[19] J. Zanelli, Classical Quantum Gravity 29, 133001 (2012).
[20] S. W. Hawking and D. N. Page, Commun. Math. Phys. 87,
577 (1983).
[21] E. Witten, Adv. Theor. Math. Phys. 2, 505 (1998).
[22] R. G. Cai, Phys. Rev. D 65, 084014 (2002).
[23] X. O. Camanho et al. (to be published).
[24] A. Gomberoff, M. Henneaux, C. Teitelboim, and F.
Wilczek, Phys. Rev. D 69, 083520 (2004).
[25] D. Boulware and S. Deser, Phys. Rev. Lett. 55, 2656
(1985).
[26] J. T. Wheeler, Nucl. Phys. B268, 737 (1986).
[27] J. T. Wheeler, Nucl. Phys. B273, 732 (1986).
[28] D. Kastor, S. Ray, and J. Traschen, Classical Quantum
Gravity 27, 235014 (2010).
[29] X. O. Camanho and J. D. Edelstein, arXiv:1103.3669
[Classical Quantum Gravity (to be published)].
[30] G. W. Gibbons and S. W. Hawking, Phys. Rev. D 15, 2752
(1977).
[31] R. C. Myers, Phys. Rev. D 36, 392 (1987).
[32] E. Gravanis and S. Willison, Phys. Rev. D 75, 084025
(2007).
[33] C. Garraffo, G. Giribet, E. Gravanis, and S. Willison, J.
Math. Phys. (N.Y.) 49, 042502 (2008).
[34] C. Charmousis and A. Padilla, J. High Energy Phys. 12
(2008) 038.
[35] S. C. Davis, Phys. Rev. D 67, 024030 (2003).
[36] E. Gravanis and S. Willison, Phys. Lett. B 562, 118 (2003).
[37] M. Bañados, C. Teitelboim, and J. Zanelli, Phys. Rev. Lett.
72, 957 (1994).