Home Search Collections Journals About Contact us My IOPscience

Exponential fading to white of black holes in quantum gravity

This content has been downloaded from IOPscience. Please scroll down to see the full text.

2017 Class. Quantum Grav. 34 105007

(http://iopscience.iop.org/0264-9381/34/10/105007)

View the table of contents for this issue, or go to the journal homepage for more

Download details:

IP Address: 207.162.240.147
This content was downloaded on 07/07/2017 at 18:38

Please note that terms and conditions apply.

You may also be interested in:

The lifetime problem of evaporating black holes: mutiny or resignation

Carlos Barceló, Raúl Carballo-Rubio, Luis J Garay et al.

Thermodynamical aspects of gravity: new insights

T Padmanabhan

The (2 + 1)-dimensional black hole

S Carlip

Coupling a point-like mass to quantum gravity with causal dynamical triangulations

I Khavkine, R Loll and P Reska

Asymptotics with a positive cosmological constant: I. Basic framework

Abhay Ashtekar, Béatrice Bonga and Aruna Kesavan

Quantum gravity: a progress report

S Carlip

Surface tension and negative pressure interior of a non-singular ‘black hole’

Pawel O Mazur and Emil Mottola

Causal structure of analogue spacetimes

Carlos Barceló, Stefano Liberati, Sebastiano Sonego et al.

Toward black hole entropy in shape dynamics

Gabriel Herczeg and Vasudev Shyam
 Classical and Quantum Gravity

Class. Quantum Grav. 34 (2017) 105007 (15pp) https://doi.org/10.1088/1361-6382/aa6962

Exponential fading to white of black holes

in quantum gravity
Carlos Barceló1, Raúl Carballo-Rubio1,2 and Luis J Garay3,4
1
Instituto de Astrofísica de Andalucía (IAA-CSIC), Glorieta de la Astronomía, 18008
Granada, Spain
2
The Cosmology & Gravity Group and the Laboratory for Quantum Gravity &
Strings, Department of Mathematics & Applied Mathematics, University of Cape
Town, Private Bag, Rondebosch 7701, South Africa
3
Departamento de Física Teórica II, Universidad Complutense de Madrid, 28040
Madrid, Spain
4
Instituto de Estructura de la Materia (IEM-CSIC), Serrano 121, 28006 Madrid,
Spain

E-mail: [email protected], [email protected] and [email protected]

Received 30 November 2016, revised 7 March 2017

Accepted for publication 27 March 2017
Published 19 April 2017

Abstract
Quantization of the gravitational field may allow the existence of a decay
channel of black holes into white holes with an explicit time-reversal
symmetry. The definition of a meaningful decay probability for this channel
is studied in spherically symmetric situations. As a first nontrivial calculation,
we present the functional integration over a set of geometries using a single-
variable function to interpolate between black-hole and white-hole geometries
in a bounded region of spacetime. This computation gives a finite result which
depends only on the Schwarzschild mass and a parameter measuring the width
of the interpolating region. The associated probability distribution displays an
exponential decay law on the latter parameter, with a mean lifetime inversely
proportional to the Schwarzschild mass. In physical terms this would imply
that matter collapsing to a black hole from a finite radius bounces back
elastically and instantaneously, with negligible time delay as measured by
external observers. These results invite to reconsider the ultimate nature of
astrophysical black holes, providing a possible mechanism for the formation
of black stars instead of proper general relativistic black holes. The existence
of both this decay channel and black stars can be tested in future observations
of gravitational waves.

Keywords: black holes, white holes, gravitational collapse, Hawking

evaporation, massive stars, quantum gravity, gravitational waves

1361-6382/17/105007+15$33.00 © 2017 IOP Publishing Ltd Printed in the UK 1

Class. Quantum Grav. 34 (2017) 105007 C Barceló et al

1. Introduction

The construction of a well-defined and physically meaningful theory of quantum gravity is

the aim of different research programs [1]. Any attempt to describe the quantum properties of
the gravitational field must inevitably make a series of leaps, by means of the consideration
of additional mathematical structures or self-consistent rules that go further than the classi-
cal description of spacetime geometry embodied in general relativity. The presumed extreme
weakness of observable effects associated with quantization makes this research field highly
speculative from an empirical standpoint. Whatever approach (or just partial construction)
showing hints that quantum gravity effects may be larger, and hence testable in the near future,
would be greatly valuable in that it might help to break this vicious circle.
A well-known consequence of partial (i.e. semiclassical) quantization is the destruction of the
stability of black holes through an extremely faint, but nonzero, evaporation due to the emission
of Hawking radiation [2, 3]. In a series of papers [4–7] the authors have proposed that black
holes might decay instead through a time-symmetric decay channel that outdrives Hawking
evaporation as the dominant channel once non-perturbative quantum gravity effects are taken
into account, leading to white holes as the end product5. In this picture, black holes formed in
the gravitational collapse of massive stars from a finite radius would just represent short-lived
configurations that are followed almost instantaneously by the bounce of the collapsing star back
to its initial radius, at least in an idealized situation with no dissipation. Including dissipation
would lead to dampened oscillations and eventual stabilization of the bouncing structure in a
black ultra-compact horizonless object. This dynamical process points towards the formation of
black stars, supported by quantum vacuum effects [12, 13]. As explained in the conclusions, the
corresponding gravitational wave signature associated with the gravitational collapse of mas-
sive stars would be substantially different from the templates dictated by general relativity. Also
black stars present characteristics that may permit one to distinguish them from black holes [5],
in particular in binary coalescences [14]. The collection of additional gravitational wave data,
following the first events observed [15, 16], will permit one to put to test these features.
Previous work by the authors has focused on the properties of spacetime geometries
describing in effective terms the decay of black holes into white holes in a given interval of
time. While the knowledge of these effective geometries represents a natural starting point
to study the physical implications of this picture, such as the specific form of the associated
gravitational wave signature, it remains to be seen that this decay channel becomes indeed
allowed in quantum gravity. Given the mathematical subtleties involved and the very vague-
ness on the precise meaning of the latter concept, this question is hard to answer in an exhaus-
tive way. The present paper communicates two nontrivial results related to this question: (1)
It is possible to use these effective geometries to obtain a measure of the probability that the
time-symmetric decay of black holes into white holes takes place in an arbitrary time interval,
and (2) The probability distribution on this time interval takes the form of an exponential
decay law, with a mean lifetime of black holes which is in complete accordance with previous
considerations by the authors.
The contents of this paper are up to certain extent inspired by the Euclidean approach to
quantum gravity [17, 18]. This approach aims to construct a consistent path integral through
the consideration of the analytic continuation of Lorentzian metrics to Riemannian metrics,
hence avoiding oscillatory integrals that appear in the Lorentzian case. This motivation leads
to the following formal expression for the amplitude between two configurations h− and h+ of
the spacetime geometry at hypersurfaces Σ− and Σ+, respectively:

5
See [8–11] for other qualitatively different scenarios that include the formation of white holes.

2
Class. Quantum Grav. 34 (2017) 105007 C Barceló et al

1 g(Σ+) = h+
⟨(1)
h+, Σ+ | h−, Σ−⟩ = ∫
N g(Σ−) = h−
Dg exp(−S EH[g]).

Here S EH[g] is the Einstein–Hilbert action of a Euclidean geometry g satisfying the

boundary conditions specified at Σ− and Σ+, Dg is the measure on the configuration space
of Euclidean geometries, and N is a normalization constant. Dealing with Euclidean geom-
etries helps to improve the convergence properties of the path integral, though equation (1) is
still formal in that it is not well-defined in the absence of further considerations [18].
In this paper we define a suitable version of equation (1) that is valid for a symmetry-
reduced situation. This functional integral will be defined as the sum over a given set of
geometries, invariant under time-reversal and spatial rotations, and interpolating between a
black-hole geometry at Σ− and a white-hole geometry at Σ+. In this integration, we include all
the geometries interpolating between black-hole geometries and white-hole geometries that
use a single-variable interpolating function. The results obtained show that these geometries
provide the dominant contribution in the relevant region of the parameter space, hence proving
that this approximation is self-consistent. On the other hand, the hypersurfaces Σ± are arbi-
trary (but for a set of minimal technical assumptions), which shows explicitly the generality
of the results obtained. These considerations permit going from the ambiguous quantity in
equation (1) to a tractable expression which is evaluated in similar terms to one-dimensional
quantum-mechanical problems, and indeed displays a finite value.
After the evaluation of the relevant functional integrals, the results that follow are analyzed.
The resulting probability distribution exhibits an exponential decay on one of the param­
eters measuring the size of the interpolating region Γ. We show that the mean lifetime in this
exponential-decay law is inversely proportional to the Schwarzschild mass, which in terms
of external observers means that the decay of black hole into white holes is extremely fast
(indeed the fastest possible in practical terms). This result, together with the use of Euclidean
path integrals and the lack of a classical solution joining the black-hole and white-hole geom-
etries at Σ− and Σ+, may suggest the interpretation of this phenomenon as a tunneling effect.
Following this interpretation the effective geometries in the interpolating region Γ, not being
vacuum solutions of the Einstein field equations, would correspond to the classically forbid-
den configurations.

2. Interpolating geometries

Our aim is to evaluate a functional integral over a set of (Euclidean) interpolating geometries
that describe a smooth transition between black-hole and white-hole geometries. The first
step in this program is naturally the definition of the Euclidean geometries to be considered.
Natural units G = ħ = c = 1 are used in the following.
Let us start recalling the form of the time-symmetric Lorentzian geometries interpolating in
a bounded region of spacetime Γ between a black hole with mass M and its time-reversal solu-
tion corresponding to a white hole. The existence of these geometries was discussed in [4, 5],
where it was shown that the interpolating region has to extend further than the Schwarzschild
radius 2M. The fine details concerning the explicit construction of these geometries can be
read in [6]. Here we just mention briefly the main results of these papers that are of direct
relevance for the present discussion.
Before entering in specific details, we want to stress that we are assuming that there is no
microscopic arrow of time in quantum gravity. This motivates studying a time-symmetric
decay channel. This is shared by many approaches to the construction of a theory to quantum
gravity, but for certain notable exceptions [19–22]. Most importantly, the lack of a microscopic
3
Class. Quantum Grav. 34 (2017) 105007 C Barceló et al

arrow of time does not enter in contradiction with the T-violation observed in particle physics,
as explained for instance in [23].
The line element of the geometries to be considered can be written as
[dr − f (u )v(r )dt ]2
gab dx a dx b = −dt 2 +
(2) + r 2 d Ω22 .
1 − 2M /ri
Here we are using coordinates x a = (t , r , θ, ϕ ), d Ω22 is the line element of the 2-sphere with
angular coordinates (θ, ϕ ), r ⩽ ri is the radial coordinate, v(r ) = (2M /ri − 2M /r )1 / 2 and
t
=
u(3) .
∆ (r )
As discussed below, ∆(r ) is a fixed, but unspecified, function that describes the shape of
the spacetime region Γ interpolating between these two classical geometries. For f (u ) = ∓ 1
one would recover two specific patches of the Schwarzschild solution going from r  =  0 to
r  =  ri; these correspond, respectively, to the black-hole and white-hole patches in Painlevé–
Gullstrand coordinates [24]. The function f(u) interpolates between these two limiting values
so that equation (2) represents the geometry outside a stellar structure undergoing gravita-
tional collapse and a subsequent time-symmetric bounce at t  =  0, though we will not need
to consider explicitly the geometry inside the star. Due to this interpolation, the coordinates
being used are not the usual ingoing or outgoing Painlevé–Gullstrand coordinates, but rather
a non-trivial combination of them. The parameter ri  >  2M that marks the extension of the
interpolating region is coincident with the initial radius of the star (this identification between
two in principle different quantities follows from the explicit construction of the interpolating
geometries).
The only restriction on these interpolations is that they are effectively one-dimensional,
which fixes their dependence on the variable u. The single-variable function f(u) interpolates then
between the values corresponding to the black-hole patch in the ingoing Painlevé–Gullstrand
coordinates [f(u)  =  −1] and the white-hole patch in the outgoing Painlevé–Gullstrand coordi-
nates [f(u)  =  1]. In order to make explicit the role played by the function ∆(r ), let us define a
subset Γ of the (t, r) coordinates as
Γ = {(t , r )| t ∈ [−∆(r ), ∆(r )], r ∈ [r0, ri ]}.
(4)
Here r0 ∼ M1 / 3  ri is the radius in which the deviations from the classical collapsing solu-
tion due to quantization are of order unity (equivalently, the radius in which the spacetime
curvature becomes Planckian), leading generally to an effective repulsive potential that modi-
fies the spacetime geometry [25–35]. This radius is around 1013 in natural units for solar-mass
stars, or 10−22 m in SI units. The boundaries of Γ over which boundary conditions on f(u) are
imposed are given by
Σ± = {(t , r ) | t = ±∆(r ), r ∈ [r0, ri ]}.
(5)
The boundary conditions to be satisfied by f(u) take then the simple form
f (u = ±1) = ±1, f ′ (u = ±1) = 0. Both conditions guarantee that the interpolating region
matches smoothly the black-hole and white-hole geometries at Σ− and Σ+, respectively. Note
that time-reversal invariance implies f(−u)  =  −f(u), hence f(0)  =  0. Due to time-reversal
symmetry, each geometry in equation (2) is alternatively specified by a choice of function f(u)
for u ∈ [0, 1], such that | f (u )| ⩽ 1 and satisfying the boundary conditions f(0)  =  0, f(1)  =  1 and
f ′ (1) = 0. These conditions define the functional space F on the interval u ∈ [0, 1].

4
Class. Quantum Grav. 34 (2017) 105007 C Barceló et al

The particular form of the function ∆(r ) defines the hypersurfaces Σ± in equation (5) or, in
other words, the shape of the interpolating region Γ, and is therefore fixed. The degeneracy in
the choice of ∆(r ) cannot be reduced without further knowledge of quantum gravity. Hence
we consider a general function ∆(r ) from the beginning, and show explicitly that the results
below are robust against changes in this function. The only technical conditions imposed on
∆(r ) is that it is a concave downwards function that decreases monotonically outwards until
reaching a zero value at r  =  ri, namely ∆(ri ) = 0. Under these conditions, the (maximum)
value of this function ∆0 = ∆(r0 ) controls the size of Γ and hence the duration of the trans­
ition between the black-hole and white-hole geometries. It is useful to define a dimensionless
variable x  =  r/ri and a dimensionless function I (x ) = ∆(x )/∆0.
We now define a rule to associate a unique Euclidean metric ḡab to each Lorentzian metric
gab in equation (2), and hence a weight in the Euclidean path integral (1). Given the structure
of the spacetime line element we are considering, it is enough to follow the usual prescription
of analytical continuation to imaginary values of the temporal coordinate t (see, e.g. [36]). Let
us consider a new temporal coordinate T  =  it where i is the imaginary unit; for real t the coor-
dinate T takes purely imaginary values. The standard prescription we follow then is replacing
t by T and performing the analytic continuation of the functions of the time t so that these
functions take real values for T real. Specifically, we define if (−it /∆) = −f¯ (T /∆), where
now f¯ (T /∆) is taken as a real function of the variable u¯ = T /∆(r ). This leads to the following
Euclidean metric ḡab for each Lorentzian metric gab in equation (2):
[dr − f¯ (u¯)v(r )dT ]2
g¯ab dx a dx b = dT 2 +
(6) + r 2 d Ω22 .
1 − 2M /ri
In the following we drop the overline on the metric ḡab, the function f¯ (u¯) and its variable ū,
keeping in mind that we will be working in Euclidean signature with line element (6).

3. Probability amplitude

In this section we define and evaluate the probability amplitude for the transition between
black-hole and white-hole geometries. The geometry in the interpolating region Γ, not being
a vacuum solution of the Einstein field equations, leads to a nonzero value of the Einstein–
Hilbert action S EH[g] = ∫ ε R(g)/16π, where ε is the Riemannian volume form and R(g) the
Γ
Ricci scalar of the Euclidean metric in equation (6). This nonzero value, which can be evalu-
ated directly from the form of the metric in equation (6), permits one to define a nontrivial
weight exp(−S EH[g]) which is understood as the probability measure for a given interpolating
geometry. This leads to the natural definition of the probability amplitude between black-
hole and white-hole geometries at Σ− and Σ+ as the functional integral of this exponential
factor over all the possible interpolating geometries satisfying the boundary conditions. The
Einstein–Hilbert action evaluated on the interpolating geometries can be recast in a remark-
ably simple form after lengthy manipulations (detailed in appendix), leading to the remark-
ably simple equation
1
⟨WH| BH⟩M , ∆0 =
N ′ ∫F Df exp(−S EH[g])
⎛ kI M ∆0 ⎞ ⎡ ⎤
exp⎜− ⎟ Df exp ⎢− du f 2 (u )⎥ .
1 M ∆0 1
=
N ⎝ 1 − 2M /ri ⎠ ∫F ⎣ 1 − 2M /ri ∫0 ⎦ (7)


5
Class. Quantum Grav. 34 (2017) 105007 C Barceló et al

This amplitude is defined in terms of a Gaussian functional integral over single-vari-

able functions f (u ) ∈ F . The numerical constant kI is a function of the area spanned by
I (x ) = ∆(x )/∆0, taking values in the interval kI ∈ [1, 3]. In equation (7) as well as in the fol-
lowing equations, only the leading order in the extremely small parameter r0 /ri is written
explicitly. Subdominant terms on the parameter r0 /ri in the Einstein–Hilbert action have been
explicitly considered and shown to be irrelevant for our conclusions. Also note that, in equa-
tion (7), the pieces of the classical action that do not display dependence on ∆0 have been
absorbed in the definition of the normalization constant.
The evaluation of the Gaussian functional integral in equation (7) can be accomplished
considering a discretization of the interval [0,1] defined in terms of a set of points {ui}iN=+12
and taking the limit N → ∞ after performing the necessary calculations. If we define
fi = f (ui ) ∈ [0, 1], i  =  1,...,N  +  2, the integral inside the exponential in equation (7) is dis-
cretized as
1 1 N +2 2 2 1 N

∫
(8)
0
du f 2 (u ) ⟶ ∑ f
N + 2 i=1 i
=
N+2
+ ∑ f 2,
N + 2 i=1 i
where we have used fN + 1 = fN + 2 = 1 as a consequence of the boundary conditions satis-
fied by f(u). The N-dimensional measure for the discretized functional integral is defined as
N
∏i = 1 dfi. The Gaussian functional integral in equation (7) arises taking the N → ∞ limit in
this discretization. This choice of measure is the natural one given the parametrization of
the interpolating geometries. Also this definition is consistent with the potential existence of
diffeomorphisms preserving the structure of the line element (2) and reduce to the identity
transformation at the boundaries Σ±. Under these ‘residual’ diffeomorphisms, the function
f(u) may be non-trivially transformed into another function f ′ (u′) where u′ is defined in the
same way as u. The corresponding path integral measure would be given by ∏iN= 1 df ′i, and the
overall transition amplitude in equation (7) is indeed invariant.
The discretized version of the functional integral in equation (7) is then expressed in terms
of the product of N integrals of the form
⎛ λf2 ⎞ ⎛ λ ⎞
dfi exp⎜⎜− i ⎟ π N+2
erf ⎜ ⎟,
1
∫
(9) ⎟ =
0 ⎝ N + 2⎠ 2 λ ⎝ N+2 ⎠

where erf(x ) is the error function with the usual normalization [37] and λ = M ∆0/ 1 − 2M /ri .
Using the Taylor expansion erf(x ) = 2[x − x 3 /3 + O (x 5)]/ π and the definition of the expo-
nential as the limit exp(x ) = limN → ∞[1 + x /N + O (1/N2 )]N permit showing that
⎡ 1 ⎤
∫ Df exp ⎢−λ du f 2 (u )⎥ = exp(−λ /3).
∫
⎣ ⎦
(10)
F 0

Hence the probability amplitude (7) takes the finite value

⎛ (k + 1/3)M ∆ ⎞
exp⎜− I ⎟.
1 0
⟨WH| BH⟩M , ∆0 =
(11)
N ⎝ 1 − 2M /ri ⎠
It is important to remark that the transition amplitude above has been evaluated exactly.
There is no need to use the well-known saddle-point method or any other approximation
method. Of course, this does not affect the interpretation of this transition amplitude as giving
the probability of the decay of a black hole into a white hole, which stems from its very defini-
tion (and not the specific approach chosen for its evaluation).

6
Class. Quantum Grav. 34 (2017) 105007 C Barceló et al

4. Exponential decay

Equation (11) takes the form that would be expected from a tunneling amplitude, with ∆0
measuring the width of the classically forbidden region. The square of equation (11) gives
then the probability density of tunneling between black-hole and a white-hole geometries for
a specific value of ∆0 ∈ [0, ∞). To obtain the probability that this transition takes place in a
finite interval [0, ∆0] it is then necessary to evaluate the one-dimensional integral
∆0
PBH ▹ WH(M, ∆0) =
(12) ∫0
d∆′0 | ⟨WH| BH⟩M , ∆′0 |2 ,

with the normalization constant N fixed by the normalization condition PBH ▹ WH(M, ∞) = 1.
The evaluation of the integral (12) from equation (11) is straightforward, leading to the fol-
lowing exponential decay law:
⎛ 2(k + 1/3)M ∆ ⎞
BH ▹ WH(M , ∆ 0 ) = 1 − exp⎜− ⎟.
I 0
P
(13)
⎝ 1 − 2M /ri ⎠
The standard normalization of equation (13) ensures, from a mathematical standpoint, that
the transition will always take place if one waits long enough (i.e. infinite time). This does
not imply though that the decay should take place physically: if it happens to have a very long
characteristic time scale, it would be disrupted by other effects; for instance, by the evapora-
tion due to the emission of Hawking radiation (which is not taken into account in the evalu-
ation of the probability amplitude above due to the assumption of time-reversal invariance),
or more drastically by the existence of white hole instabilities as discussed in [7]. Given the
exponential decay form of equation (13), the quantity that determines whether this transition
is physical is the mean lifetime τ defined as usual. Taking into account that kI  +  1/3 is always
positive and of order unity, it will be generally
1 − 2M /ri 1
τ(14)
 ⩽ .
2M 2M

5. Characteristic time scale

The parameter ∆0 is tightly connected to a natural definition of the lifetime of the black hole,
namely the time T ri that the bounce of the star takes as measured by an external observer
situated at the initial radius of the collapsing star r  =  ri [4–6], or equivalently the same time
interval as measured by asymptotic observers at r → ∞, T ∞. The coordinate invariant quanti­ty
T ri is given by twice (due to-time reversal symmetry) the classical collapsing proper time
from ri to r0, T r0i (obtained in a first approximation using the Oppenheimer–Snyder model
[38]), plus an additional term proportional to ∆0. This can be evaluated explicitly from the
line element (2):
0
T ri = T ri + 2∆ 0 .
(15)
The latter term, which could be considered as the quantum-mechanical part, increases line-
arly with the value of ∆0 ∈ [0, ∞), being zero when ∆0 = 0. The value ∆0 = τ given in equa-
tion (14) implies that this contribution is negligible for macroscopic stars, so that the bounce
is an elastic process which does not display an appreciable time delay to be added to T r0i.

7
Class. Quantum Grav. 34 (2017) 105007 C Barceló et al

The same conclusion is valid for asymptotic observers at spatial infinity which measure
the time interval T ∞, straightforwardly obtained from equation (15) using the multiplica-
tive redshift factor characteristic of the Schwarzschild solution 1/ 1 − 2M /ri . Equation (14)
implies then that the contribution to be added to the classical bouncing time scales as 1/2M for
asymptotic observers. In simplified terms, equation (14) should be read as pointing to black
holes as extremely unstable objects. Let us stress that these results cannot be extrapolated to
horizonless objects, due to the use of boundary geometries at Σ∓ that correspond to vacuum
solutions of the Einstein field equations in the entire range r ∈ [r0, ri ]. In astrophysical sce-
narios, ri is given by a O (1) multiple of the Schwarzschild radius 2M, corresponding typically
to the radius of a (proto-)neutron star [39]. This leads to a black hole lifetime T ∞ roughly
proportional to M + O (1/M ), which is to be compared with the extremely larger evaporation
time through Hawking evaporation, proportional to M3.

6. Conclusions

Following these results, the time-symmetric decay into white holes would outdrive Hawking
evaporation, becoming the dominant decay channel for black holes in quantum gravity. Unless
there exists some (symmetry) principle forbidding this decay channel, it will dominate the
physics of black holes. It is interesting to remark that classical and semiclassical arguments
forbid longer time scales, but allow this time-symmetric decay as long as its characteristic
time scale is linear in M [7, 11]. We find the convergence of arguments of diverse nature
concerning the time scale of the time-symmetric decay remarkable, and a strong incentive to
further explore this phenomenon.
The main contribution of this paper is the evaluation of a well-defined probability ampl­
itude for the time-symmetric decay of black holes into white holes, and the subsequent charac-
teristic time scale. Evaluation of transition amplitudes in quantum gravity is notably difficult,
so that we have considered a natural reduction of the number of degrees of freedom involved,
through the use of single-variable interpolating functions in the line element (6). That this
reduction permits carrying out an exact analytical treatment is noteworthy, as also are the
results obtained, namely an exact exponential decay law and a mean lifetime fixed by the
dynamics of general relativity to be given by equation (14). It should be possible to check
whether these features survive in different approaches and quantization schemes [40–43]; let
us stress that time scales linear in M have been indeed shown to arise in the canonical quanti-
zation of null shells [28].
In physical terms, our results point that quantum gravity effects in black holes may be more
dramatic than expected, opening an observational window that is yet to be explored. In the
approximation considered here which neglects dissipative effects (which is indeed a reason-
able approximation due to the short time scale of the process), the collapse of a massive star
from an initial radius ri would be followed by an infinite number of cycles in which the star
bounces back to ri. Undoubtedly, dissipation has to be taken into account in order to elaborate
a satisfactory physical picture, including for instance the very emission of gravitational waves.
Dissipative effects will gradually shrink the initial value of ri down to 2M. If this shrinking
is slow enough (as it should be if dissipative effects represent a small correction), after a
transient composed of a series of dampened oscillations, a stable configuration in the form
of a black ultra-compact horizonless object with radius close to its gravitational radius can
be stabilized due to quantum vacuum effects [12], and a black star is formed [13, 44]. Hence
the decay of black holes into white holes opens the possibility for the formation of black
stars instead of black holes. Both electromagnetic and gravitational wave observations leave

8
Class. Quantum Grav. 34 (2017) 105007 C Barceló et al

wiggle room for black ultra-compact horizonless objects of this kind representing the ultimate
nature of astrophysical black holes, as discussed respectively in [45, 46] and [47] (see also the
general discussion in [48]). Only time (and future observations) will tell whether or not this
possibility stands further theoretical and experimental scrutiny.
In this regard, gravitational wave observations will play an important role. Typical values
of ri makes detection of this decay channel by means of electromagnetic radiation difficult,
due to the interference with many other physical processes. However, macroscopic quantum
effects modifying the local geometry around the Schwarzschild radius would most probably
leave a noticeable imprint (of transient nature) on gravitational radiation. Also the fact that
the interpolating region connects two classical solutions (corresponding to the collapsing and
expanding star, respectively) implies that there should be new emissions of gravitational radia-
tion that are not present in classical general relativity, associated with the expansion of the
(inhomogeneous) star across the white-hole geometry (inhomogeneities in the collapsing star
will be generally amplified in the bounce as shown in cosmological scenarios [49]). These
features, together with the existence of a series of dampened oscillations characterized by the
formation of short-lived black hole horizons, point to a gravitational wave signature showing
periodic features that decay in time, and clearly different to the classical pattern in which only
one cycle (or main signal) would be present. The existence or not of these reverberations in
gravitational wave signals from collapse events provides a crucial way to distinguish between
the formation of black holes or black stars. Regarding the coalescence of already formed
black stars, the existence of a surface outside the would-be horizon may lead (depending on
its reflective properties) to similar reverberant signatures [5, 14].

Acknowledgments

Financial support was provided by the Spanish MINECO through the projects FIS2011-
30145-C03-01, FIS2011-30145-C03-02, FIS2014-54800-C2-1, FIS2014-54800-C2-2 (with
FEDER contribution), and by the Junta de Andalucía through the project FQM219. R.C-
R. acknowledges support, at different stages of the elaboration of this work, from the Math
Institute of the University of Granada (IEMath-GR), the research project MINECO-FEDER
MTM2013-47828-C2-1-P, and the Claude Leon Foundation.

Appendix. Evaluation of the action functional on the interpolating geometries

In this appendix we evaluate the Einstein–Hilbert action on the interpolating geometries,

namely the functional
1
S EH [g] =
(A.1)
16π Γ ∫
ε R(g),

where R(g) is the Ricci scalar of the metric gab with line element
⎡ f 2 (u ) ⎛ 2M 2M ⎞⎤ 2 2f (u )
gab dx a dx b = ⎢1 + ⎜ ⎟⎥ dT −
2M 2M
− − dT dr
⎣ 1 − 2M /ri ⎝ r ri ⎠⎦ 1 − 2M /ri r ri
dr 2
+ + r 2 d Ω22 .
1 − 2M /ri (A.2)

The functional form of R(g) in terms of f(u) is given by

9
Class. Quantum Grav. 34 (2017) 105007 C Barceló et al

4M ⎡ ⎛ ∂f (u ) ⎞2 ∂f (u )
R (g ) = ⎢ + 2
( ) − ( − )⎜ ⎟ − 2(ri − 2r )f (u )
⎝ ∂r ⎠
1 f u r r r
rir 2 ⎣
i
∂r
∂ 2f (u ) (3ri − 4r )∂f (u )/∂T + 2r (ri − r )∂ 2f (u )/∂T ∂r ⎤
− r (ri − r )f (u ) − ⎥.
⎦
(A.3)
∂r 2 2 2M (1/r − 1/ri )

Using the form of the variable u inside the interpolating function, namely u = T /∆(r ),
leads to the relations

∂f (u ) df d ln ∆ ∂ 2f (u ) d2f ⎛ d ln ∆ ⎞2 df ⎡⎛ d ln ∆ ⎞2 d2 ln ∆ ⎤
= −u = u2 2 ⎜ ⎟ +u ⎢⎜ ⎟ − ⎥,
du ⎝ dr ⎠ du ⎣⎝ dr ⎠
,
∂r du dr ∂r 2
dr 2 ⎦
∂f (u ) 1 df (u ) ∂ 2(u ) d ⎛ 1 ⎞ df (u ) d2f (u ) d ⎛ 1 ⎞
= ⎜ ⎟ ⎜ ⎟.
dr ⎝ ∆ ⎠ du du2 dr ⎝ ∆ ⎠
= , +u (A.4)
∂T ∆ du ∂T ∂r

Due to spherical symmetry, the integration of the action functional is reduced to a two-
dimensional integral on the (t, r) variables. It will be useful to perform the integration in terms
of the (u, r) variables instead. The following relation holds for any function h = h(T /∆, r ):
ri ∆(r ) +1 ri
∫r
dr
(A.5)
0
∫−∆(r )
dT h(T /∆, r ) = du ∫−1
dr ∆(r )h(u, r ). ∫r 0

Note also that

1
= det(gab ) =
g(A.6) .
1 − 2M /ri
Let us split the different contributions in order to arrange conveniently the necessary calcul­
ations. Taking into account the 4π multiplicative factor coming from the integration over the
angular variables, one has
7
1
S EH =
(A.7) ∑
4 1 − 2M /ri n = 1
E (n),

where:
ri ∆(r ) 4M
E (1) =
(A.8)
r0
∫ dr r 2 ∫
−∆(r )
dT 2 ,
rir

ri ∆(r ) 4M
E (2) =
(A.9)
r0
∫ dr r 2 ∫
−∆(r )
dT 2 f 2 (u ),
rir

⎛ ∂f (u ) ⎞2
(ri − r )⎜ ⎟,
ri ∆(r ) 4M
E (3) = −
(A.10)
r0
∫
dr r 2
−∆(r )
∫dT
rir ⎝ ∂r ⎠

ri ∆(r ) 8M ∂f (u )
E (4) = −
(A.11)
r0
∫
dr r 2
−∆(r )
∫dT 2 (ri − 2r )f (u )
rir ∂r
,

ri ∆(r ) 4M ∂ 2f ( u )
E (5) = −
(A.12)
r0
∫
dr r 2
−∆(r )
∫dT
rir
(ri − r )f (u )
∂r 2
,

10
Class. Quantum Grav. 34 (2017) 105007 C Barceló et al

ri ∆(r ) 2M (3ri − 4r ) ∂f (u )
E (6) = −
(A.13)
r0
∫ dr r 2
−∆(r )
∫
dT 2
rir 2M (1/r − 1/ri ) ∂T
,

ri ∆( r ) 4M r (ri − r ) ∂ 2f (u )
E (7) = −
(A.14)
r0
∫ dr r 2
−∆(r )
∫
dT 2
rir 2M (1/r − 1/ri ) ∂T ∂r
.

It is now a matter of evaluating these integrals, using equation (A.5). In order to extract the
dependence of these quantities on the dimensionful parameters of the problem, let us define
x  =  r/ri and I (x ) = ∆(x )/∆0.
The first contribution is given by
+1 1 1
(1)
= 4M ∆0
E(A.15) du ∫−1
dx I (x ) = 8M ∆0 ∫r /r
dx I (x ).
0 i
∫r /r
0 i

(1)
This equation shows that E depends only on two dimensionful constants: the Schwarzschild
mass M and ∆0. The dependence is moreover linear in the product M ∆0. The particular value
of the numerical coefficient multiplying this quantity depends on the integrals on the right-
hand side of equation (A.15). Hence changes on the function I(x), or equivalently ∆(r ) will
lead to modifications of dimensionless coefficients, which are of order unity. This structure is
shared by all the terms E (n) of the classical action.
The evaluation of the second contribution is similar to the previous one:
+1 1
(2)
= 4M ∆0
E(A.16) ∫−1
du f 2 (u ) dx I (x ). ∫r /r
0 i

Again, we can see explicitly that changes of ∆(r ) lead to modifications in the dimension-
less numerical factors in the equation above. The third contribution reads
⎛ ∂f (T /∆) ⎞2
r (ri − r )⎜ ⎟
+1 4Mri
E (3) = − ∫−1 0
du ∫r ri
dr ∆(r )
⎝ ∂r ⎠
4M +1 ri ⎛ df ⎞2⎛ d∆ ⎞2
∫ dr ∆(r )r (ri − r )u2⎜ ⎟ ⎜
∫ ⎟
⎝ du ⎠ ⎝ dr ⎠
 =− du
ri −1 r0
+1 ⎛ df ⎞2 1 ⎛ d ln I (x ) ⎞2
= −4M ∆0
−1
∫
du u2⎜ ⎟
⎝ du ⎠ r0 / ri ∫
dx I (x )x (1 − x )⎜
⎝ dx ⎠
⎟ . (A.17)

The fourth contribution is given by

+1 8Mri ∂f (T /∆)
E (4) = − ∫−1 0
du ∫r ri
dr ∆(r )
(ri − 2r )f (T /∆)
∂r
8M +1 df ri d ln ∆
 =
ri −1 ∫
du uf
du r 0 ∫
dr ∆(r )(ri − 2r )
dr
+1 df 1 d ln I (x )
= 8M ∆0
−1
∫
du uf
du r0 / ri ∫
dx I (x )(1 − 2x )
dx
. (A.18)

The fifth contribution is:

11
Class. Quantum Grav. 34 (2017) 105007 C Barceló et al

+1 4M ri ∂ 2f (T /∆)
E (5) = −
−1
∫ r0
du ∫ri
dr ∆(r )
r (ri − r )f (T /∆)
∂r 2
4M +1 d2f ri ⎛ d ln ∆ ⎞2
=− ∫
du u2f 2 dr ∆(r )r (1 − r )⎜
∫ ⎟
ri −1 du r 0 ⎝ dr ⎠
⎡ d2 ln ∆ ⎛ d ln ∆ ⎞2⎤
dr ∆(r )r (1 − r ) ⎢ ⎟ ⎥
4M +1 df ri
+ ∫ ∫ −⎜
⎝ dr ⎠ ⎦
du uf
ri −1 du r 0 ⎣ dr 2
+1 d2f 1 ⎛ d ln I (x ) ⎞2
= −4M ∆0 ∫−1
du u2f
du2
r0 / ri
∫
dx I (x )x (1 − x )⎜
⎝ dx ⎠
⎟

⎡ d2 ln I (x ) ⎛ d ln I (x ) ⎞2⎤
dx I (x )x (1 − x ) ⎢ ⎟ ⎥.
+1 df 1
+ 4M ∆0 ∫−1
du uf
du r0 / ri ∫ ⎣ dx 2
−⎜
⎝ dx ⎠ ⎦ (A.19)

Taking into account the boundary conditions, this can be also written as
+1 ⎛ df ⎞2 1 ⎛ d ln I (x ) ⎞2
E (5) = 4M ∆0
−1
du u2⎜ ⎟∫
⎝ du ⎠ ∫r /r dx I (x )x(1 − x )⎝
0 i
⎜
dx ⎠
⎟

 +1 df 1 d2I (x )
+ 4M ∆0
−1
du uf
du ∫ ∫r /r dx x(1 − x )
0 i dx 2
. (A.20)

Then,
+1 df 1 ⎡ dI (x ) d2I (x ) ⎤
E (3) + E (4) + E (5) = 4M ∆0 ∫−1 du uf
du ∫r /r dx ⎢⎣2(1 − 2x )
0 i dx
+ x (1 − x )
dx 2 ⎦
⎥
+1 df 1 dI (x )
= 4M ∆0 ∫−1
du uf
du ∫r /r dx (1 − 2x )
0 i dx
+ O (r0 /ri ). (A.21)

The sixth contribution is the following:
ri ∆(r ) 2M (3ri − 4r ) ∂f (u )
E (6) = − ∫r 0
dr r 2 ∫−∆(r ) dT rir 2 2M (1/r − 1/ri ) ∂T
2M +1 df ri (3ri − 4r )
=−
ri ∫−1 du
du ∫r 0
dr
2M (1/r − 1/ri )
 2M 1 (3 − 4x ) x
= −2
ri ∫r /r dx
0 i 1−x
2M ⎛ r0 ⎞ r 3
=4 ⎜1 − ⎟ 03 .
ri ⎝ ri ⎠ r i (A.22)

The last contribution to the Einstein–Hilbert action is given by:

ri ∆(r ) 4M r (ri − r ) ∂ 2f (u )
E (7) = − ∫r 0
dr r 2 ∫−∆(r ) dT rir 2 2M (1/r − 1/ri ) ∂T ∂r
 4M +1 ⎛ df d2f ⎞ ri r (ri − r ) d ln ∆
= ∫−1 du⎜ + u 2⎟ ∫r dr
ri ⎝ du du ⎠ 0 2M (1/r − 1/ri ) dr
= 0. (A.23)

12
Class. Quantum Grav. 34 (2017) 105007 C Barceló et al

In order to write down the Einstein–Hilbert action evaluated on the metric gab in equa-
tion (A.2) in a compact form, we will need the relation
1 ⎛ df ⎞ 1 1 1
∫ du uf ⎜ ⎟ = −
(A.24)
0 ⎝ du ⎠ 2 2 0
du f 2 , ∫
obtained integrating by parts:
1 ⎛ df 2 ⎞ 1 1
du u⎜ ⎟ = uf 2 0 −
1
∫
(A.25)
0 ⎝ du ⎠ 0
∫
du f 2 = 1 −
0
du f 2 . ∫
Then,
⎡ 1 1 − 2x dI ⎞⎤
⎛
E (1) + E (2) + E (3) + E (4) + E (5) = 8M ∆0 ⎢ ∫r /r dx⎝I + ⎜ ⎟⎥
⎣ 0 i 2 dx ⎠⎦
⎡ 1 1 ⎛ 1 − 2x dI ⎞⎤
+ 8M ∆0 ⎢ ∫
du f 2 (u ) ∫
dx ⎜I − ⎟⎥
⎣ 0 r0 / ri ⎝ 2 dx ⎠⎦
⎡ 1 ⎛ 1⎞ 1 1 ⎤
= 8M ∆0 ⎢ ∫
dx ⎜2I − ⎟ + du f 2 (u ) + O (r0 /ri )⎥ .
∫
⎣ r0 / ri ⎝ 2⎠ 2 0 ⎦
 (A.26)
1
Let us define the constant kI = 2 ∫ dx (2I − 1/2) which, up to O (r0 /ri ) and taking into
r0 / ri
account the monotonous behavior of the function I(x), can only take the values
I ∈ [1, 3] .
k(A.27)
Putting all together,
M ∆0 ⎡ 1 ⎤
S EH [g] = ⎢⎣kI + ∫0 du f 2 (u ) + O (r0 /ri )⎥
⎦
1 − 2M /ri

r 30(1 − r0 /ri )/r 3i
+ .
2M /ri − 1 (A.28)

This expression is remarkably simple; for instance, it shows no dependence on the deriva-
tive of the function f(u) in the interval u ∈ [0, 1]. The last piece is independent of ∆0 and hence
irrelevant for the evaluation of the probability amplitude on ∆0, as it is absorbed in the defini-
tion of the normalization factor N in equation (7).

References

[1] Carlip S 2001 Quantum gravity: a progress report Rep. Prog. Phys. 64 885–942
[2] Hawking S W 1974 Black hole explosions? Nature 248 30–1
[3] Unruh W G 1976 Notes on black-hole evaporation Phys. Rev. D 14 870–92
[4] Barceló C, Carballo-Rubio R and Garay L J 2014 Mutiny at the white-hole district Int. J. Mod.
Phys. D 23 1442022
[5] Barceló C, Carballo-Rubio R, Garay L J and Jannes G 2015 The lifetime problem of evaporating
black holes: mutiny or resignation Class. Quantum Grav. 32 035012
[6] Barceló C, Carballo-Rubio R and Garay L J 2016 Where does the physics of extreme gravitational
collapse reside? Universe 2 7
[7] Barceló C, Carballo-Rubio R and Garay L J 2016 Black holes turn white fast, otherwise stay black:
no half measures J. High Energy Phys. JHEP1(2016)157

13
Class. Quantum Grav. 34 (2017) 105007 C Barceló et al

[8] Gambini R and Pullin J 2015 Quantum shells in a quantum space-time Class. Quantum Grav.
32 035003
[9] Corichi A and Singh P 2016 Loop quantization of the Schwarzschild interior revisited Class.
Quantum Grav. 33 055006
[10] Bambi C, Malafarina D and Modesto L 2016 Black supernovae and black holes in non-local gravity
J. High Energy Phys. JHEP04(2016)147
[11] De Lorenzo T and Perez A 2016 Improved black hole fireworks: asymmetric black-hole-to-white-
hole tunneling scenario Phys. Rev. D 93 124018
[12] Barcelo C, Liberati S, Sonego S and Visser M 2008 Fate of gravitational collapse in semiclassical
gravity Phys. Rev. D 77 044032
[13] Visser M, Barcelo C, Liberati S and Sonego S 2009 Small, dark, and heavy: But is it a black hole?
(PoSBHGRS,010(2008))
[14] Cardoso V, Hopper S, Macedo C F B, Palenzuela C and Pani P 2016 Gravitational-wave signatures
of exotic compact objects and of quantum corrections at the horizon scale Phys. Rev. D 94 084031
[15] Abbott B P et al 2016 Observation of gravitational waves from a binary black hole merger Phys.
Rev. Lett. 116 061102
[16] Abbott B P et al 2016 GW151226: observation of gravitational waves from a 22-solar-mass binary
black hole coalescence Phys. Rev. Lett. 116 241103
[17] Gibbons G W and Hawking S W 1977 Action integrals and partition functions in quantum gravity
Phys. Rev. D 15 2752–6
[18] Gibbons G W and Hawking S W 1993 Euclidean Quantum Gravity (Singapore: World Scientific)
[19] Wald R M 1980 Quantum gravity and time reversibility Phys. Rev. D 21 2742–55
[20] Penrose R 1980 Singularities and time assymetry General Relativity: an Einstein Centenary Survey
(Cambridge: Cambridge University Press) pp 581–638
[21] Liu C 1993 The arrow of time in quantum gravity Phil. Sci. 60 619–37
[22] Barbour J, Koslowski T and Mercati F 2014 Identification of a gravitational arrow of time
Phys. Rev. Lett. 113 181101
[23] Quinn H 2009 Time reversal violation J. Phys. Conf. Ser. 171 012001
[24] Martel K and Poisson E 2001 Regular coordinate systems for Schwarzschild and other spherical
space-times Am. J. Phys. 69 476–80
[25] Frolov V P and Vilkovisky G A 1979 Quantum gravity removes classical singularities and shortens
the life of black holes The Second Marcel Grossmann Meeting on the Recent Developments of
General Relativity (In Honor of Albert Einstein) (Trieste, Italy, 5–11 July 1979) p 0455
[26] Frolov V P and Vilkovisky G A 1981 Spherically symmetric collapse in quantum gravity
Phys. Lett. B 106 307–13
[27] Hajicek P 2003 Quantum theory of gravitational collapse: (lecture notes on quantum conchology)
Lect. Notes Phys. 631 255–99
[28] Ambrus M and Hajicek P 2005 Quantum superposition principle and gravitational collapse:
scattering times for spherical shells Phys. Rev. D 72 064025
[29] Hayward S A 2006 Formation and evaporation of regular black holes Phys. Rev. Lett. 96 031103
[30] Frolov V P 2014 Information loss problem and a ’black hole‘ model with a closed apparent horizon
J. High Energy Phys. JHEP05(2014)049
[31] Frolov V P 2016 Notes on non-singular models of black holes Phys. Rev. D 94 104056
[32] Kawai H, Matsuo Y and Yokokura Y 2013 A self-consistent model of the black hole evaporation
Int. J. Mod. Phys. A 28 1350050
[33] Bambi C, Malafarina D and Modesto L 2013 Non-singular quantum-inspired gravitational collapse
Phys. Rev. D 88 044009
[34] Torres R and Fayos F 2015 On the quantum corrected gravitational collapse Phys. Lett. B
747 245–50
[35] Arfaei H and Noorikuhani M 2017 Quantum vacuum effects on the final fate of a collapsing ball of
dust J. High Energy Phys. JHEP02(2017)124
[36] Mukhanov V and Winitzki S 2007 Introduction to Quantum Effects in Gravity (Cambridge:
Cambridge University Press)
[37] Abramowitz M and Stegun I A 1964 Handbook of Mathematical Functions: with Formulas,
Graphs, and Mathematical Tables (Applied Mathematics Series) (New York: Dover)
[38] Oppenheimer J R and Snyder H 1939 On continued gravitational contraction Phys. Rev. 56 455–9
[39] Woosley S and Janka T 2005 The physics of core-collapse supernovae Nat. Phys. 1 147

14
Class. Quantum Grav. 34 (2017) 105007 C Barceló et al

[40] Campiglia M, Gambini R, Olmedo J and Pullin J 2016 Quantum self-gravitating collapsing matter
in a quantum geometry Class. Quantum Grav. 33 18LT01
[41] Olmedo J 2016 Brief review on black hole loop quantization Universe 2 12
[42] Achour J B, Brahma S and Marciano A 2016 Spherically symmetric sector of self dual Ashtekar
gravity coupled to matter: anomaly-free algebra of constraints with holonomy corrections
(arXiv: 1608.07314)
[43] Christodoulou M, Rovelli C, Speziale S and Vilensky I 2016 Planck star tunneling time: an
astrophysically relevant observable from background-free quantum gravity Phys. Rev. D
94 084035
[44] Barcelo C, Garay L J and Jannes G 2011 Quantum non-gravity and Stellar collapse Found. Phys.
41 1532–41
[45] Abramowicz M A, Kluzniak W and Lasota J-P 2002 No observational proof of the black hole
event-horizon Astron. Astrophys. 396 L31–4
[46] Abramowicz M A, Bulik T, Ellis G F R, Meissner K A and Wielgus M 2016 The electromagnetic
afterglows of gravitational waves as a test for quantum gravity (arXiv: 1603.07830)
[47] Cardoso V, Franzin E and Pani P 2016 Is the gravitational-wave ringdown a probe of the event
horizon? Phys. Rev. Lett. 116 171101
[48] Visser M 2009 Black holes in general relativity (PoSBHGRS,001(2008))
[49] Brizuela D, Mena Marugan G A D and Pawlowski T 2010 Big bounce and inhomogeneities Class.
Quantum Grav. 27 052001

15