Beruflich Dokumente
Kultur Dokumente
discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/307875799
CITATION READS
1 21
3 authors, including:
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Luciano Gabbanelli on 18 May 2017.
Abstract
We analyze in detail a previous proposal by Dvali and Gomez that black holes could
be treated as consisting of a Bose-Einstein condensate of gravitons. In order to do so
we extend the Einstein-Hilbert action with a chemical potential-like term, thus placing
ourselves in a grand-canonical ensemble. The form and characteristics of this chemical
potential-like piece are discussed in some detail. After this, we proceed to expand the
ensuing equations of motion up to second order around the classical Schwarzschild metric
so that some non-linear terms in the metric fluctuation are kept. We argue that the
resulting equations could be interpreted as the Gross-Pitaevskii equation describing a
graviton Bose-Einstein condensate trapped by the black hole gravitational field. Next we
search for solutions and, modulo some very plausible assumptions, we find out that the
condensate vanishes outside the horizon but is non-zero in its interior. Based on hints
from a numerical integration of the equations we formulate an ansatz and eventually find
an exact non-trivial solution for a mean-field wave-function describing the graviton Bose-
Einstein condensate in the black hole interior. Based on this we can rederive some of the
relations involving the number of gravitons N and the black hole characteristics, summa-
rized in its Schwarzschild radius, along the lines suggested by Dvali and Gomez. These
relations are parametrized by a single parameter a dimensionless chemical potential.
August 2016
ICCUB-16-032
1
1 Introduction
In an interesting saga of papers, Dvali, Gomez and others [1,2] have put forward an intriguing
suggestion: black holes (BH) could be understood as Bose-Einstein condensates of gravitons.
If this view is correct, this would reveal a deep quantum nature of such fascinating objects
and could lead to an alternative understanding of some of the most striking features of BH;
for instance, Hawking radiation [4] could be understood as being due to evaporation (leakage
as the authors put it) of the condensate. Besides, this picture brings new ideas about the
Bekenstein entropy [5], the absence of hair [6], as well as the quantum nature of information
storage and the possible information loss in BHs [7].
The main point of these works is that the physics of BH can be understood in this picture
in terms of a single number N , the number of gravitons contained in the Bose-Einstein
condensate (BEC). These condensed gravitons have a wave length N LP , LP being the
Planck length; they have a characteristic interaction strength g 1/N (reminiscent of meson
interactions in large Nc gauge theories) and the leakage leads to a Hawking temperature
of
order TH 1/ N LP , equal to the inverse of . The mass of the BH is M N M P and
its Schwarzschild radius therefore is given by rs NLP , thus agreeing with the Compton
wavelength of the quantum gravitons , in accordance with the uncertainty principle that
dictates rs in the ground state of the quantum system. Therefore, up to various factors of
the Planck mass everything is governed by N , the number of intervening gravitons. Modulo
some assumptions, all these results stem from the basic relation (unless otherwise stated we
work in units where ~ = 1)
N
rg = (1)
M
that relates the number of gravitons N , the mass of the gravitating object M and its gravi-
tational radius rg . For a Schwarzschild BH rg = rs .
We found these results intriguing and we set up to try to understand them in a different
language and, if possible, attempt to be more quantitative. On the process we have separated
slightly from the original line of thought of the authors. The original approach of [1, 2] is
not geometric at all. There is no mention of horizon or metric. Here we will adopt a more
conservative approach. We will assume the pre-existence of a classical gravitational field
created by an unspecified source that generates the Schwarzschild metric. As it is known,
nothing can classically escape from a BH so if we wish to interpret this in potential terms
(which of course is not correct but serves us for the purpose of creating a picture of the
phenomenon) it would correspond to a confining potential. On and above this classical
potential one can envisage a number of quantum fields being trapped. For sure there is a
gravity quantum field present; hence gravitons (longitudinal gravitons that is). Continuing
with our semi-classical analogy, these gravitons have had a long time to thermalize and it is
therefore expected that they can form a Bose-Einstein condensate. Of course, one may argue
at once that gravitons do not have repulsive interactions, at least naively, and that therefore
a BEC is impossible to sustain. To this objection one could reply in a twofold way. First, it
is up to the equations to determine whether such a condensate is possible or not (we will see
below that indeed it is, at least in the case of vanishing angular momentum). Secondly one
might also answer that in theories of emergent gravity, the ultimate nature of gravitons may
be some type of fermionic degrees of freedom (such as e.g. in the model suggested in [3]).
Then, repulsion is assured at some scale.
Therefore we will try to identify a consistent set of equations describing a BEC constructed
2
on top of this classical field1 . We will see that remarkably enough the characteristics of the
BEC are uniquely described in terms of the Schwarzschild radius of the BH and the value
of a dimensionless parameter, interpreted as a chemical potential. A condensate appears not
only to be possible but actually intimately related to the classical field that sustains it that
determines its characteristics. It is therefore tempting to go one step beyond and reverse the
order of the logical implication and eventually attempt to derive the classical field as a sort
of mean field potential a la Hartree-Fock. However we will stop short of doing so here.
In the subsequent we present the results of our investigations.
We will now expand the perturbed Einstein tensor up to second order in h . This will allow
us to retain the leading non-linearities. Four of the components of the Einstein tensor are
non-null but only three are linearly independent; both angular components coincide when
using the form G for the Einstein tensor.
1
A classical field is not the same as a quantum condensate, although the latter may trigger the former.
3
Taking into account that Gt t (g ) = 0, we have
r 2 rs2 r 3 2rs r 2 + rs2 r
Gt t g + h = 4
hrr + hrr
r r4 (3)
r 3 3rs2 r + 2rs3 2r 4 6rs r 3 + 6rs2 r 2 2rs3 r
5
hrr 2 hrr hrr = 0
r r5
and analogous equations for the radial and angular components (they are given in Appendix
A).
It is well known that Birkhoffs theorem [8] guarantees that any spherically symmetric
solution of the vacuum field equations must be static and asymptotically flat. So this means
that the solution of the equations G (g + h) = 0 with g being the Schwarzschilds metric,
must necessarily be of the Schwarzschild type again. In conclusion, a perturbation of a
Schwarzschild metric in vacuum must end in another Schwarzschild metric but taking into
account a change in the parameters responsible of inducing the perturbation. Stating this in
other words, a spherically symmetric gravitational field should be produced by some massive
object at the origin characterized by the Schwarzschild radius, hence a perturbation in itself
it can only be produced by an alteration of this radius and therefore a change in its mass.
Indeed, imposing the following ansatz for the components of the perturbation
X B(n) X A(n)
htt = ; hrr = (4)
rn rn
n=1 n=1
and substituting hrr in (3) is possible to obtain the coefficients for each term of the radial
expansion. We obtain
A A2 + 2Ars 2A3 + 3rs A2 + 3rs2 A 5A4 + 8rs A3 + 6rs2 A2 + 4rs3 A
hrr = + + +
r r2 r3 r4
(5)
14A5 + 25rs A4 + 20rs2 A3 + 10rs3 A2 + 5rs4 A
+ + ... ,
r5
where A = A(0) . It is obvious that A = 0 is a solution as implies no perturbation of the
metric at all. But if A 6= 0 the rest of the coefficients of the expansion are fixed once A is
given. The same happens for htt using the radial equation (53) given in Appendix A.
A A3 5A4 + 8rs A3 67A5 + 75rs A4 + 60rs2 A3
htt = +0+ 3 + + + ... . (6)
r r 12r 4 60r 5
Taking into account the accuracy of the expansion in h the components of the metric
become
!
rs A 2GM
gtt + htt 1 + = 1
r r r
! (7)
rs rs2 A A2 + 2Ars 2GM (2GM )2
grr + hrr 1 + + 2 + + =1+ + ,
r r r r2 r r2
M being the new mass associated to the BH and rs = 2GM the perturbed Schwarzschild
radius. Now we will use similar techniques in the quantum case and as we will see the results
are dramatically different. However it is interesting to note that any perturbation h (i.e.
each graviton) has associated a certain amount of energy that is reflected in a change of the
BH mass.
4
2.2 Quantum theory: Einstein equations as Gross-Pitaevskii equations
Perturbing the metric as g = g + h , a completely covariant perturbative expansion for
h is obtained. By construction these equations are non-linear (we keep only the leading
non-linearities, to keep the formalism simple and analytically tractable as we will see later).
We note that the usual Gross-Pitaevskii equation [9] employed to describe Bose-Einstein
condensates is a non-linear Schrodinger equation; i.e. an equation of motion that contains
self-interactions (hence the non-linearity), a confining potential for the atoms or particles
constituting the condensate, and a chemical potential that is conjugate to the number of
particles or atoms contained in the condensate.
Among all these ingredients, the perturbed Einstein equations already contain most of
them. They are already non-linear and while there is no confining potential explicitly in-
cluded (as befits a relativistic theory) they do confine particles, at least classically, because
if the selected background corresponds to a Schwarzschild BH, the strong gravitational field
classically traps particles inside the horizon. However, there is one ingredient still missing,
namely the equivalent of the chemical potential. Therefore we have to extend the formulation
of perturbations around a classical BH solution to the grand-canonical ensemble by adding
to the appropriate action a chemical potential term such as2
1 1
Z Z
4
dV h2 ,
p
Schem.pot. = d x gh h = (8)
2 2
that is conjugate the quantity h2 h h , which should be related to the graviton density
of the condensate inside a differential volume element dV 3 . Taking into account that h is not
directly interacting with the background, but a field that propagates over the background, it
may seem reasonable to take the volume element as the one corresponding to the background.
Still, the new term is completely covariant as can be seen in Appendix B.
Irrespective of this chemical potential term an appropriate action has to be chosen for the
field h . It is only natural to select the Einstein-Hilbert action expanded around the classical
Schwarzschild background g to describe fluctuations of the metric in a reparametrization
invariant way. One can then expand the Einstein-Hilbert Lagrangian MP2 gR(g), with
g = g + h in powers of h up to the desired order (where MP 2 = ~G1 N is the Planck
mass). 4
An oddity of Eq. (8) is that, as defined, has dimensions of [energy]4 . This is because
the field h does not have the canonical dimension; the field that has canonical dimensions
is h MP h . After scaling the two powers of MP , the chemical potential would have
dimensions of [energy]2 . Still, these are not quite the correct dimensions for a chemical
potential. As we shall see below the natural quantity to be identified with a chemical
potential will in fact be the dimensionless combination (r)r 2 that will be conjugate to N ,
the number of gravitons in the BEC condensate with = /Mp2 . Note that we assume that
the chemical potential is r-dependent.
The actual equations of motion derived from the previous action may take slightly different
forms depending on the choice of volume elements. For the time being, let us take the simplest
2
Here we choose to formulate the fluctuations of the metric on a fixed background later on we shall revise
this approach.
3
We shall be somewhat more precise about this in Section 5.
4
In order to get Einsteins equations, the usual g measure associated to the full metric has tobe taken
(i.e. dV ) however this is not the only possibility to obtain diff-invariant equations; for instance MP2 gR(g)
also could yield a covariant set of equations.
5
possibility
G (g + h ) = h . (9)
where indices are assumed to be raised with the background metric only. Actually Eq. (9)
can be derived from the action we have chosen only if we neglect terms of the form O(h 2 ).
We will return to this point later.
It is important to keep in mind the following: in the chemical potential part is considered
in the grand canonical ensemble; this implies that this magnitude is an external field and
does not vary in the action; in particular, for these equations of motion, it is independent of
h , so /h = 0. Under these considerations this term resembles the chemical potential
term of the Gross-Pitaevskii equation.
This should be the way of introducing the chemical potential. Otherwise, were not an
external field, it would be necessary to take it into account when performing variations to
derive the equations of motion. An equation of motion would be obtained for and this
equation would nullify automatically the perturbation as well as the chemical potential itself,
i.e. h = 0 and = 0. Therefore, the external field is introduced in the theory as some
kind of Lagrange multiplier.
As Lagrange multipliers have constraint equations, the chemical potential of the theory
may not be arbitrary at all. It must satisfy binding conditions with h . The main difference
is that this constrains are implicit in a GR theory. The restriction of is through the general
covariance conditions for the action. In other words, the diffeomorphism covariance implies
the covariant conservation of the Einstein tensor, and this in itself entails the same for the
chemical potential term,
G = 0 = (h ); = , h + h ; = 0 . (10)
This differential equation of motion for the latter potential is valid up to every order in
perturbation theory.
Hereinafter, we will proceed to find acceptable solutions of these equations and interpret
them. We will separate the possible solutions into two regions: outside and inside the BH
horizon. We certainly expect that the graviton condensate will disappear quickly in the
outside region as r rs and, imposing this as a boundary condition, we will see that in
fact the condensate is identically zero on this side of the horizon. On the contrary, a unique
non-trivial solution will be found in the interior of the BH.
6
limit are, respectively5
1 1 1 2
2
hrr + hrr 2 hrr 2 hrr hrr = htt
r r r r
rs 1 1 rs 1 rs 1 1
3
htt htt 2 hrr + 3 htt 2 htt htt 3 hrr htt + hrr htt + 2 hrr 2 = hrr
r r r r r r r r (11)
2 r
As in the Schwarzschild metric gtt , grr 2 1, the integral should fulfill
Z Z dr
2 2 2
4 dr r htt + hrr < = (Logarithmic divergence) . (14)
0 0 r
For this to happen, if a power law ansatz at infinity is imposed, the exponents must obey
, > 3/2. Then the three equations in (11) are
B B2 AC
(1 ) (1 2 ) =
r +2 r 2+2 r +
A2 B2
rs A B rs rs AB BC
+ +2 +2 + + 2+2
+ ++2
+ 2+2 = +
r r r r r r r r r
(15)
A2 4 B2
rs 2A rs 2B 2rs rs
+ 2 + + + 32
r r +2 r r +2 r r 2+2 r r 2+2
rs AB
(2 + ) + 2 2 + ++2 = 0 .
r r
The requirement of the solution being square-integrable exclude that and could be zero.
This implies automatically the neglection of the terms proportional to the Schwarzschild
5
Here and in successive equations in this perturbative treatment possible terms of the form h2 will be
neglected as we are interested in the simplest approximation to a Gross-Pitaevskii equation. The form of these
terms depends on which convention is chosen for the volume element and to raise/lower indices.
7
radius. Even more, as this numbers and are positive and larger than 3/2, the following
asymptotic identities result
B AC A B BC
(1 ) = + ; = + ; (16)
r +2 r r +2 r +2 r
2 A 2 2B
+2 = 0 . (17)
r +2 r
From the angular equation (17), is mandatory for both terms to contribute at infinity. If this
is not the case, this would nullify A (or B), and then, via both equations in (16), fix B = 0
(or A = 0) and C = 0, i.e h = = 0. The competition of both terms of the angular equation
is possible if and only if = . This condition modifies the angular equation (17) as
2
(A B) =0 = A = B . (18)
r +2
Then from the second equation in (16), the following relationship can be read
1 BC
(A B) =0= . (19)
r +2 r +
This automatically leads to a null chemical potential as the only possible solution for this
region with the proposed ansatz. This is due to the fact that if B = 0, as B is proportional
to A, then A is also zero.
In Appendix C we explore the possibility of exponentially vanishing chemical potential
when r with analogous conclusions. Likewise it can be seen that a much faster Gaussian
decay is also excluded (not reported here).
8
unstable. With the set of solutions for the components of hR obtained for each of the initial
condition imposed, it is possible to compute the quantity dV h2 , defined in Eq. (13), and
the integral extends over the exterior of the BH (i.e. z runs between 0 and 1).
In Figure 1 the behaviour of the integral of h2 outside the BH is presented when the
initial condition for the integration approaches zero at a fixed point near infinity, z 0, i.e.
h (r ) 0. It is observed that this integral is invariably small and actually vanishes
as the initial condition for hrr nullifies. This seems to confirm also numerically that
0.0008
0.0006
g h2
0.0004
dz
0.0002
0.
0.001 0.0008 0.0006 0.0004 0.0002 0.
Decreasing initial condition
Figure 1: Each point represent a magnitude proportional to the integral of h2 outside the event horizon.
The closer the limit to an asymptotically flat space-time is (i.e. decreasing the initial condition near
infinity), the smaller this integral is.
the only possible solution in the outer region is the one with h = 0 and therefore X = 0
throughout the exterior of the BH.
(2z + 1)(z 1)2 rr + z(z 1)3 rr (4z + 1)(z 1)4 rr 2 2z(z 1)5 rr rr = Xtt
2z(z 2)(z 1)rr + z(z 2)(z 1)2 rr z(5z 2)tt 2z 2 (z 1)tt (22)
2z(z 2)(z 1)4 rr rr + z(7z 2)(z 1)2 rr tt z(5z 2)tt tt
2
+z 2 (z 1)3 rr tt z 2 (z 1)tt + 2z 2 (z 1)3 rr tt
2z 2 (z 1)tt tt = 0 .
9
4.1 Behaviour near the event horizon
In order to get a feeling for possible solutions to these equations we consider their linearized
approximation. Only terms linear in tt , rr and their derivatives are kept in the region
z 1. That is
3(z 1)2 rr + (z 1)3 rr = Xtt
The system of equations is algebraic for the variable X (that as we have seen should behave
as a constant as z 1); therefore, it is possible to eliminate X by combining the temporal
and radial equation. In the present ansatz this leads to
B 2 (1 + ) + AB + A2 = 0 . (27)
Together with the angular equation, this determines the solutions up to a constant. There
are two possible solutions for this system
A
i) =0 = 2 A = B = tt = A rr = (28)
(z 1)2
A A
ii) = 1 = 3 A=B = tt = rr = . (29)
(z 1) (z 1)3
In any case, at least one of the perturbations is divergent over the event horizon. Nonetheless,
the first solution appears to be integrable, while this is not the case for the second one. In
case i) X = 1 while if ii) is taken as solution X = 0 (i.e. no chemical potential at all).6
However these solutions do not vanish when z 1 (as one could perhaps have hoped),
thus casting doubts on the relevance of the linearized equations. Let us examine this point
6
Solution ii) represents however a volume-preserving fluctuation at the linear order.
10
and let us only take into consideration the solution i), which is integrable. To make sure that
this solution is little modified for z 1 when non-linearities are switched on, we substitute
the solution back in the non-linear equation and verify that the most singular terms are
still satisfied. Calculations are presented in Appendix D. In addition, this exercise gives one
interesting result: X (1 + A), where A is so far arbitrary, also in sign. Note that at the
linear level we got X = 1 and the fact that the quadratic equation gives an O(A) correction
to this result is consistent as we are implicitly assuming that |h | |g |, i.e. |A| 1.
tt = A w ; rr = B w (32)
while we leave X = X(w) as a free function. This function can be eliminated by combining
Gt t and Gr r , leading to one equation that together with G leads to several constraints
involving the parameters of the ansatz. Of all possibilities, the one that gives the same weight
to all terms in the equations is the one corresponding to = 0, = 2. Then, close to the
origin it would appear as if
tt A ; rr B w2 . (33)
It remains to be seen if this solution matches with the one obtained in the inner vicinity of
the horizon.
tt () = A ; rr () = B 2 ; tt () = 0 . (34)
7
As before linearization is performed only in order to get a feeling of possible solutions; there is no a priori
reason to expect a good description by doing so.
11
We could have equally started the numerical integration routine from a point close to w = 1,
i.e. in the vicinity (the inner vicinity that is) of the horizon. The boundary conditions
according to (28) would then be
(1 )2
tt (1 ) = A ; ;
rr (1 ) = A tt (1 ) = 0 . (35)
2
The respective remaining degrees of freedom are the constants A, A and B , and obviously
they must be chosen in such a way that the solution near w = 1 (z = 1) and the one near
w = 0 (z = ) match.
It turns out that there are unique values of A and A , B that make both solutions match.
Namely, A = A = B . Even more, when the previous boundary conditions are assumed for
the numerical integration, tt as well as the function related to the chemical potential X,
turns out to be constant in the interior of the BH. The constant A itself is arbitrary and is
not determined by the structure of the equations. Changes in A appear as an overall factor
in the solution. For a certain value of A the curves presented in Figure 2 are obtained. The
numerical solution reproduces the general features of the analytical study.
4
X , tt , rr
It is most interesting to plot the results of the numerical integration for ht t and hr r (i.e.
considering h = (z 1) and then raising one index, taking into account that in the
current approach h = h g ). In Figure 3, a plot of the numerical solution reveals that
to a very good approximation throughout the interior of the BH horizon
ht t = hr r
= constant . (36)
12
1.0
0.8
0.6
ht t , hrr
0.4
0.2
0.0
0.2 0.4 0.6 0.8
r
w=
rs
Figure 3: The graph shows that ht t (represented by a gray solid line) and hr r (by a black dashed
line), as obtained by numerical integration of the O(h 2 ) equations, are constant and equal.
ht t = a ; hr r = b ; h = h = 0 . (37)
if indices are lowered and raised with the full metric itself. Otherwise, if the background
metric is used to raise and lower indices, one obtains
g = diag (1 + a)gtt , (1 + b)grr , g , g . (39)
And the corresponding volume elements are r 2 / (1 a)(1 b) or r 2 / (1 + a)(1 + b), re-
p p
spectively. If one assumes constant values for a and b, as the solution in the previous subsec-
tion suggests, the corresponding O(h 2 ) equations simplify enormously and reduce to the
three following ones
and
a = b ; X = 1 + . (41)
Therefore we are retrieving the solution found in the previous subsection: plays the role of
the integration constant A.
Now let us see that the problem actually admits an exact solution if we assume that the
solution is of the form ht t = hr r = with being a constant. In fact, there are several
possible solutions depending on how one chooses to treat the separation between background
and fluctuation metrics.
13
If one chooses to consistently raise and lower indices with the background metric only,
one has to plug Eq. (39) into the Einstein tensor and gets the following closed algebraic
equation:
1
= X = 1 + + . . . (42)
1 + r2
that agrees with the O(h 2 ) perturbative solution found above. Note that X = 0 implies
= 0 but not the other way round.
However, in deriving the equations of motion we have actually neglected a possible con-
tribution of the corresponding volume measures, assuming that Eq. (9) holds exactly. If we
decide to raise and lower indices with the full metric, it is natural to keep the volume element
as the one given by g, g being the full metric. The LHS of the equation of motion now
reads
Gt t = Gr r = 2 (43)
r
(this is obtained after substituting Eq. (38) into the Einstein tensor). The RHS of the
equation also gets modified. On the one hand g is replaced by g, as mentioned, that
depends on h . On the other hand, because the indices of h are raised with the full
metric (that depends on h itself), deriving h2 w.r.t. h is not totally immediate. All
things considered, the resulting equation of motion for a constant is now
3
2
= 2 (44)
r 2
that can be interpreted as a mean field-like Gross-Pitaevskii equation for the condensate wave
function . Using the fact that X r 2 , we (re)obtain that X is a constant and
1 3
X= 3 1 + . . . . (45)
1 2 2
The previous results are valid to all orders in the expansion. Note that h2 = ht t2 +hr r2 = 22 .
Before moving on, it is mandatory to make a comment on the covariant conservation of our
equations. At the end of Section 2.2 we have pointed out that the diffeomorphism invariance
entails a differential equation for the chemical potential, namely (10). In Appendix E a short
discussion of this equation is presented.
We see that the actual relation between and X depends from the precise way the
fluctuations around the background metric are assumed to interact with the metric itself. In
any case the qualitative result that there is a normalizable solution that can be interpreted as
the collective wave function of a graviton condensate remains. A unique relation is obtained
between this (constant) wave function and a dimensionless chemical potential.
14
5 2.0
X , htt , hrr
1.5
0
h2
1.0
-5
0.5
-10
0.0
0.2 0.4 0.6 0.8 0.2 0.4 0.6 0.8
r r
w= w=
rs rs
Figure 4: Curves for htt (black solid), hrr (dashed) line and the dimensionful chemical potential
(dotted line). On the right plot, h2 is shown. As already emphasized, a whole family of solutions are
obtained, parametrized by a real arbitrary constant. One can simply trade this constant for the value
of .
perturbation h is null from the events horizon onwards, the endpoint on the integration
limit can be chosen as the Schwarzschild radius. This way, the integral is
Z rs h i
dr r 2 (ht t )2 + (hr r )2 h2 rs3 . (46)
0
1 1
h h h , (47)
2
with being the graviton wave length, has the right ingredients to play the role of probability
density in the present context. Note that in the BEC = rs . If we assume this, then
3
r
8 2 2 2 1
N= MP rs h rs = N LP (48)
3 |h| 8
that agrees nicely with [2] under the maximum packaging condition = rs . Recall that
h = is a constant that is entirely determined by the value of the dimensionless chemical
potential X. The rest of relations can be basically derived from this.
15
Indeed within the present philosophy we assign to each graviton in the BEC an energy
= 1/ = 1/rs . While there is no formally conserved current, the total energy of the system
will be given, by analogy with the case of a massless neutral scalar field,
1
Z
E= dV 2 (49)
2
by
1
Z Z
2 2
E= dV h = dV h . (50)
2
If the energy is a constant and given by 1/rs , then Eq. (13) has to be interpreted as the
number of gravitons in the BEC, i.e. the integral of the graviton density in the interior of
the BH.
As seen above the dimensionless chemical potential X has a rather peculiar behaviour. It
has a constant value inside the BH and it appears to be exactly zero outside. This behaviour is
summarized in Figure 5. Let us now for a moment forget about the geometrical interpretation
of BH physics and let us treat the problem as a collective many body phenomenon. From
this figure it is clear why gravitons are trapped behind the horizon: the jump of the chemical
potential at r = rs would prevent the particles inside to reaching infinity. However this is
not completely true as the picture itself suggests that one of the modes can scape at a time
without paying any energy penalty if the maximum packaging condition is verified. Let us
do a semiclassical calculation inspired by this picture; using M MP N:
dM 1 dN 1 dN
MP = . (51)
dt 2 N dt 2rs dt
To estimate dN/dt (which is negative) we can use geometrical arguments to determine the
flux. If we assume that for a given value of rs only one mode can get out (as hypothesized
above) and that propagation takes place at the speed of light, elementary considerations8
lead to
dM 3 1
2. (52)
dt 2 rs
This agrees with the results of [2] for instance Eq. (35) and yields T 1/rs . The
approximations made in the discussion could only modify the coefficient by a numerical
factor of O(1).
Within this picture, several questions concerning the long-standing issue of loss of infor-
mation may arise as the outcoming state looks thermal [11] but apparently is not; or at least
not totally so. However we shall refrain of dwelling on this any further at this point.
The main result from the previous rather detailed analysis is that the BH is able to sustain
a graviton BEC (and surely similar BECs made of other quanta). But is that condensate
really present? Our results do not answer that question, but if we reflect on the case where
the limit N 0 is taken, without disturbing the BH geometry (i.e. keeping rs constant)
this requires taking 0 in a way that the ratio N /|| is fixed. Then one gets X = 1.
This value appears to be universal and independent of any hypothesis. The metric is 100%
Schwarzschild everywhere. We conclude that a BH produces necessarily a (trapping) non-
zero chemical potential when the physical system is expressed in terms of the grand-canonical
ensemble. Outside the BH, X = 0.
8
To determine the rate of variation of N we have to multiply the surface (4rs2 ) times the flux; i.e. the
density of the mode times the velocity, assumed to be c = 1 in our units. Since the density of the mode is
constant in the interior, it is just 3/4rs3 .
16
0
-1
, X
-2
-3
-4
0.0 0.5 1.0 1.5 2.0
r
w=
rs
Figure 5: The behaviour of both chemical potentials, in solid line and the dimensionless X in
dashed line, across the horizon of the BH (dotted line). In this case the jump at the horizon amounts
to X = 0.2.
Another way of reaching this conclusion is by taking a closer look to our exact equations
in the previous section. If one sets X = 0 then necessarily = 0 and one gets the classical
Schwarzschild solution everywhere, but the reverse is not true. One can have = 0 but
this does not imply X = 0. Let us emphasize that these results go beyond the second order
perturbative expansion used in parts of this article.
As gravitons cross the horizon and are trapped by the BH classical gravitational field they
eventually thermalize and form a BEC. The eventual energy surplus generated in this process
is used to increase the mass and therefore
the Schwarzschild radius of the BH. is now non-
zero; it is directly proportional to N , the number of gravitons, and the chemical potential
departs from the value X = 1, presumably increasing it. As soon as 6= 0 the metric inside
the BH is not anymore Schwarzschild (but continues to be Schwarzschild outside). Note that
the metric is destabilized and becomes singular for = 1, so surely this is an upper limit.
Yet another possible interpretation, that we disfavour, could be the following. Each BH
has associated a given constant value of X, hence of . Then Eqs. (48) would imply that
after the emission of each graviton, the value of rs is readjusted. The problem with this
interpretation is that it would require a new dimensionless magnitude (X) to characterize a
Schwarzschild BH; something that most BH practitioners would probably find hard to accept.
In either case the metric in the BH interior is different from the Schwarzschild one.
17
Schwarzschild metric at quadratic order (i.e. including the leading non-linearities) we found
that the chemical potential necessarily vanishes in the exterior of the BH. On the contrary, in
the interior we have found two set of solutions, one of them has to be discarded as producing a
non-normalizable result. The other one leads to a non-zero chemical potential in the interior
of the BH that behaves as 1/r 2 . There is a finite jump of the chemical potential at the BH
horizon.
Surprisingly or maybe not so this solution modifies the coefficient of the tt and rr
terms in the Schwarzschild solution, but not its functional form. Of course if there is no
chemical potential at all, the modification vanishes, in accordance with well known theorems.
However, if the former is non zero, the modification is also necessarily non-zero. From the
existence and knowledge of this solution, most relations obtained in [1, 2] can be rederived.
We have seen that there is some ambiguity in the way the effect of the background metric
is taken into account and probably more work is needed to produce a unique fully convincing
answer. These ambiguities get reflected in the relation between the number of gravitons and
the Schwarzschild radius, where the value of the (constant) condensate enters.
The relation between the number of gravitons and the geometric properties of the BH
involve an a priori independent and tuneable parameter, the dimensionless chemical potential
X (related to the mean-field wave function of the condensate ). We find this somewhat
strange as this would be a new black hole parameter. Therefore we favour the universal value
X = 1 as discussed in the text.
As should be obvious to the reader who has followed our discussion, our approach is some-
what different from the one developed in the initial papers by Dvali, Gomez and coworkers.
We assume from the start the existence of a classical geometry background that acts as
confining potential for the condensate. The fact that the functional form of the metric per-
turbation induced by the condensate is exactly the same as the original background, of course
gives a lot of credence to the possibility of deriving the latter from the former in a sort of
self-consistent derivation. We have not explored this possibility in detail yet.
It is quite plausible that one could entertain the presence of condensates of other quantum
fields inside the BH horizon (why only gravitons?). While we do not expect much of a
conceptual difference, it would be very interesting to see the similarities and differences with
the case of quantum gravitons.
Acknowledgments
We acknowledge the financial support of projects FPA2013-46570-C2-1-P (MINECO) and
2014SGR104 (Generalitat de Catalunya). This research was supported in part though the
E.U. EPLANET exchange program FP7-PEOPLE-2009-IRSES, project number 246806. The
work of J.A. is partially supported by grants Fondecyt 1150390 and CONICYT-PIA-ACT14177
(Government of Chile).
References
[1] G. Dvali and C. Gomez, Black Holes 1/N Hair. Phys. Lett. 719, 419 (2013);
G. Dvali and C. Gomez, Landau-Ginzburg Limit of Black Holes Quantum Portrait: Self
Similarity and Critical Exponent. Phys. Lett. B 716, 240 (2012);
G. Dvali, D. Flassig, C. Gomez, A. Pritzel, N. Wintergerst, Scrambling in the Black Hole
Portrait. Phys. Rev. D 88, 124041 (2013);
18
G. Dvali and C. Gomez, Black Holes as Critical Point of Quantum Phase Transition.
Eur. Phys. J. C 74, 2752 (2014).
[2] G. Dvali and C. Gomez, Black Holes Quantum N-Portrait. Fortsch. Phys. 61, 742 (2013).
[4] S.W. Hawking, Particle Creation by Black Holes. Commun. Math. Phys. 43, 199 (1975).
[5] J.D. Bekenstein, Black Holes and Entropy. Phys. Rev. D 7, 2333 (1973);
J.D. Bekenstein, Generalized second law of thermodynamics in black-hole physics. Phys.
Rev. D 9, 3292 (1974);
S.W. Hawking, Black holes and thermodynamics. Phys. Rev. D 13, 191 (1976).
[7] S.W. Hawking, The Unpredictability of Quantum Gravity. Commun. Math. Phys. 87,
395 (1982).
For an excellent review on information loss paradox, see, J. Preskill, Do Black Holes De-
stroy Information? International Symposium on Black Holes, Membranes, Wormholes,
and Superstrings, Houston Advanced Research Center (1992). arXiv:hep-th/9209058.
[8] G.D. Birkhoff and R. Langer, Relativity and Modern Physics (Harvard Univ. Press,
1923).
[9] F. Dalfovo, S. Giorgini, L.P. Pitaevskii and S. Stringari, Theory of Bose-Einstein con-
densation in trapped gases. Rev. Mod. Phys. 71, 463 (1999).
19
Appendix A
Radial and angular components of the Einstein tensor
The two remaining components of the Einstein tensor are
r 1 3 2 2
5 4 2 3
Gr (g + h ) = 6 r s r r s r htt r 2rs r + rs r htt
r 2rs r 5 + rs 2 r 4
r 4 3rs r 3 + 3rs 2 r 2 rs 3 r hrr + rs r 3 htt 2 r 5 rs r 4 htt htt
(53)
rs r 3 2rs 2 r 2 + rs 3 r hrr htt + r 5 3rs r 4 + 3rs 2 r 3 rs 3 r 2 hrr htt
+ r 4 4rs r 3 + 6rs 2 r 2 4rs 3 r + rs 4 hrr 2 = 0
and
1 5 2 4 3 3
G (g + h ) : 8 2r s r 3r s r + rs r htt
4r 12rs r 7 + 12rs 2 r 6 4rs 3 r 5
+ 2r 7 7rs r 6 + 8rs 2 r 5 3rs 3 r 4 htt + 2r 8 6rs r 7 + 6rs 2 r 6 2rs 3 r 5 htt
+ 2rs r 5 7rs 2 r 4 + 9rs 3 r 3 5rs 4 r 2 + rs 5 r hrr
+ 2r 7 9rs r 6 + 16rs 2 r 5 14rs 3 r 4 + 6rs 4 r 3 rs 5 r 2 hrr
+ 2rs r 5 htt 2 + 2r 7 7rs r 6 + 5rs 2 r 5 htt htt + r 8 2rs r 7 + rs 2 r 6 htt 2
+ 2r 8 4rs r 7 + 2rs 2 r 6 htt htt 2rs r 5 6rs 2 r 4 + 6rs 3 r 3 2rs 4 r 2 hrr htt
(54)
2r 7 8rs r 6 + 12rs 2 r 5 8rs 3 r 4 + 2rs 4 r 3 hrr htt
2r 8 8rs r 7 + 12rs 2 r 6 8rs 3 r 5 + 2rs 4 r 4 hrr htt
+ rs r 6 3rs 2 r 5 + 3rs 3 r 4 rs 4 r 3 hrr htt
r 8 4rs r 7 + 6rs 2 r 6 4rs 3 r 5 + rs 4 r 4 hrr htt
4rs r 5 18rs 2 r 4 + 32rs 3 r 3 28rs 4 r 2 + 12rs 5 r 2rs 6 hrr 2
7 6 2 5 3 4 4 3 5 2 6
4r 22rs r + 50rs r 60rs r + 40rs r 14rs r + 2rs r hrr hrr = 0 .
Here stands for any of the angular coordinates: or . It is worth to note that if a
rescaling on the Schwarzschild radius rs = 2GM is done, fixing it to zero, then we end up
with Minkowski space-time in the three components of G: (3), (53) and (54).
Appendix B
General covariance of the action
In the following part, the invariance under the gauge group of diffeomorphisms of the chemical
potential term in the action (8) will be shown. Lets start by pointing out that both metrics,
g and g , as well as the perturbation, transform covariantly. In fact, the full metric g =
20
g + h , under an infinitesimal displacement D in the coordinates of the form D []x = ,
changes as
g = ; + ; = g, + , g + , g = L g . (55)
The same gauge transformation rules the background metric under the general coordinate
transformation mentioned above. This is,
g = L g . (56)
As the perturbation is defined by h = g g , under the same perturbation of the
coordinate system, from the subtraction of (55) and (56), the following equality is obtained
h = (g g ) = h, + , h + , h = L h .
The latter identity implies h has also a Lie derivative transformation. The fact that the
perturbation transforms in the same way than the metric tensor, and the chemical potential
behaves as a scalar under a general coordinate transformation, ensure automatically the
general covariance of the theory.
Appendix C
Exponentially decaying ansatz
Lets change the ansatz and impose an exponential decreasing solution for the perturbations
at infinity:
r rs 1
Gr (r ) = A e r B e r = C e r B e r (59)
r3 r r2
2 rs 2 2 r 2 rs 2
G (r ) = + + 2 Ae + + B e r = 0 . (60)
r3 r r3 r
As decreasing solutions are expected, the parameters , and must be not null. This
allows one to neglect all the terms besides the principal ones inside each parenthesis. In the
angular equation there is no way for both terms to compete between each other
e r
G
: 2 A 2 e r 2 B = 0. (61)
r
In conclusion, depending on the fact whether if is bigger or not than , A = 0 or B = 0.
For the first case, if A = 0, the temporal equation
e r
Gt t : B = A C e(+) r (62)
r
21
fixes B = 0. Being both, A and B null, the chemical potential disappears form the theory
in this region. For the second case, if B = 0, the same equation Gt t nullifies A (as no null
solutions for C are expected). However, no perturbation makes the chemical potential sense-
less. To sum up, this ansatz implies that the only solution at infinity is a null perturbation,
h = 0, and the disappearance of the chemical potential from the theory.
Appendix D
O( 2) solution
The solution of the linearized differential equations leads to the expressions tt = A ; rr =
A
(z1) 2
2 . Let us now plug this solution in the non-linear O( ) system and see if the solution
survives or how could it get modified. Near z 1 the full temporal, radial and angular
equations are
3(z 1)2 rr + (z 1)3 rr 5(z 1)4 rr 2 2(z 1)5 rr rr = Xtt (63)
2(z 1)rr + (z 1)2 rr + 3tt + 2(z 1)tt 4(z 1)3 rr 2 2 (z 1)4 rr rr + 3tt tt
2
5(z 1)2 rr tt 2(z 1)3 rr tt (z 1)3 rr tt + 2(z 1)tt tt + (z 1)tt = 0,
(65)
respectively. Replacing the possible solution, tt = A this eliminates any derivative of tt
A
and rr = (z1)2 into these equations, we get in turn
(z 1)(A) (z 1)3 A2 A
2
4
= X (z 1)
(z 1) (z 1) (z 1)2
(67)
A + A2 (1 + A)A XA
= = =
(z 1) (z 1) (z 1)
22
Appendix E
General covariance condition
In this appendix, a derivation of the differential equation for the chemical potential is pre-
sented. We are going to derive this equation for a theory where only the background metric is
considered for the integration measure and to raise and lower indices. Covariant conservation
of the Einstein tensor implies automatically the same for the RHS of our equations
G = 0 = h = h + m hm + m h m
(70)
In the last equality, the covariant derivative is expressed in terms of the Christoffel symbols
of the theory. Performing the sum in m = t, r because m appears always in the perturbation
and only exists in the temporal and radial components, and doing the same procedure with
= t, r, , the following relation is obtained
h
0 =r h r + t t ht t + t tt h t + t tr h r + r rt h t r r hr r + r rr h r
i (71)
+ t h t + r h r + t h t + r h r
where we have taken into account that there is only radial derivatives. The latter equation is
a set of 4 equations, = t, r, , . The only non trivial equation is the radial one, i.e. = r.
Dropping all the null Christoffel symbols and the off-diagonal elements of the h matrix,
makes the equation reads as
h i
0 =r (hr r ) + t tr ht t hr r + r hr r + r hr r (72)
If the perturbations are equal and constant, ht t = a = hr r = b = , the first term inside the
brackets is null. The Christoffel symbols, r = 12 g g,r = 1r and r = 21 g g,r = 1r
makes our expression become
2
(r ) + = 0. (73)
r
The integration is direct and the only freedom is a boundary condition for the differential
equation,
= 0 r 2 = X = 0 . (74)
This value should coincide with the value of our constant dimensionless chemical potential
X = 0 = 1 + .
23