Black Hole Entropy From Near-Horizon Microstates
Black Hole Entropy From Near-Horizon Microstates
HUTP-97/A106
Andrew Strominger
Abstract
Black holes whose near–horizon geometries are locally, but not necessarily globally, AdS3
(three–dimensional anti-de Sitter space) are considered. Using the fact that quantum
gravity on AdS3 is a conformal field theory, we microscopically compute the black hole
entropy from the asymptotic growth of states. Precise numerical agreement with the
Bekenstein-Hawking area formula for the entropy is found. The result pertains to any
consistent quantum theory of gravity, and does not use string theory or supersymmetry.
1. Introduction
The idea that the black hole entropy should be accounted for by microstates near the
black hole horizon has great appeal and a long history [1-9]. One reason for this is that a
demonstration that the horizon has of order one degree of freedom per Planck area would
provide a statistical explanation of the area formula for the entropy. Such a picture might
also shed light on the information puzzle, in that the information is more safely stored on
the surface than in the causally inaccessible black hole interior.
While much has been learned, attempts at a precise statistical accounting for the
entropy along these lines have been hampered both by the ultraviolet problems in quantum
gravity as well as the infinite number of low-energy modes which are nevertheless high-
frequency because of the large near–horizon redshifts1 . In practice it has not been clear
which modes should or should not be counted, and the results appear cutoff dependent.
Recently a statistical derivation of the black hole entropy has been given for some
cases in string theory [10]. This derivation employed a continuation to a weak-coupling
region where the black hole is treated as a pointlike object and its microstates are counted
by a certain conformal field theory. This construction did not directly address the issue of
whether or not, in regimes for which the black hole is not effectively pointlike and has a
clear horizon, the microstates are in any sense near the horizon.
In this paper we will address this issue in the context of black holes whose near–
horizon geometry is locally AdS3 . This includes many of the string theory examples and
the three–dimensional BTZ black hole [11]. We will microscopically derive the black hole
entropy by counting excitations of AdS3 . Our derivation relies only on general properties
of a diffeomorphism-invariant theory and will not use string theory or supersymmetry.
Our basic result follows quickly from prior results in the literature. Some time ago,
Brown and Henneaux [12] made the remarkable observation that the asymptotic symmetry
group of AdS3 is generated by (two copies of) the Virasoro algebra, and that therefore any
consistent quantum theory of gravity on AdS3 is a conformal field theory 2 . They further
3
computed the value of the central charge as c = √
2G −Λ
, where G is Newton’s constant
and Λ is the cosmological constant. In this paper we simply apply Cardy’s formula [14]
1
With the notable exception of the topological theory of gravity considered in [7] and discussed
below.
2
Maldacena has recently shown, from a totally different perspective, that certain string com-
pactifications to AdS3 are conformal field theories [13].
1
for the asymptotic growth of states for a conformal field theory of central charge c to
microscopically compute the black hole entropy. Precise agreement with the Bekenstein-
Hawking area formula is found.
The conformal field theory that describes the black hole microstates lives on a (1 + 1)–
dimensional cylinder surrounding the black hole. Since all the information is encoded in
this cylinder and remains outside the horizon there is no information loss in this picture.
This paper is organized as follows: In Section 2 the result of Brown and Henneaux is
reviewed. In Section 3 the BTZ black hole is reviewed. In Section 4 other relevant black
holes which approach AdS3 near the horizon are discussed. In Section 5 the entropy is
microscopically computed. In Section 6 we relate the present work to other derivations of
the black hole entropy as well as Maldacena’s recent work on near-horizon dynamics in
string theory [13]. We close with discussion in Section 7.
1 √ 2
Z
S= d3 x −g(R + 2 ) + Sm , (2.1)
16πG ℓ
where Sm is the matter action, the cosmological constant is Λ = − ℓ12 and we have omitted
surface terms. The matter action will not play an important role in the following, but we
include it here to stress the generality of our considerations. In the following we will be
interested in a semiclassical description, which requires that the cosmological constant is
small in Planck units, or equivalently
ℓ ≫ G. (2.2)
where φ has period 2π. AdS3 is the SL(2, R) group manifold and accordingly has an
SL(2, R)L ⊗ SL(2, R)R isometry group. In order to define the quantum theory on AdS3 ,
we must specify boundary conditions at infinity. These should be relaxed enough to allow
2
finite mass excitations and the action of SL(2, R)L ⊗ SL(2, R)R , but tight enough to allow
a well-defined action of the diffeomorphism group. This requires [12]3
−r 2
gtt = + O(1),
ℓ2
gtφ = O(1),
1
gtr = O( 3 ),
r (2.4)
ℓ2 1
grr = 2 + O( 4 ),
r r
1
grφ = O( 3 ),
r
2
gφφ = r + O(1).
Allowed diffeomorphisms are generated by vector fields ζ a (r, t, φ) which preserves (2.4).
These are of the form
ℓ3 2 + 1
ζ t = ℓ(T + + T − ) + 2
(∂+ T + ∂−2 −
T ) + O( 4 ),
2r r
2
ℓ 1 (2.5)
ζ φ = T + − T − − 2 (∂+ 2 +
T − ∂− 2 −
T ) + O( 4 ),
2r r
1
ζ r = −r(∂+ T + + ∂− T − ) + O( ),
r
∂ ∂
where 2∂± ≡ ℓ ∂t ± ∂φ
and preservation of (2.4) requires that T ± depend on r, φ and t
± t
only as T ± (r, t, φ) = T ( ℓ ± φ) so that ∂± T ∓ = 0.
±
Diffeomorphisms with T = 0 fall off rapidly at infinity and should be considered
“pure gauge transformations”. In the quantum theory the corresponding generators will
annihilate physical states. The diffeomorphisms with nonzero T± modulo the pure gauge
transformations comprise the asymptotic symmetry group. Let us denote the generators
of these diffeomorphisms by
Ln , L̄n −∞<n<∞ ,
+ in( ℓt +φ) − in( ℓt −φ)
where Ln (L̄n ) generates the diffeomorphism with T =e T =e . The
generators obey the algebra
c
[Lm , Ln ] = (m − n)Lm+n + (m3 − m)δm+n,0 ,
12
c
[L̄m , L̄n ] = (m − n)L̄m+n + (m3 − m)δm+n,0 , (2.6)
12
[Lm , L̄n ] = 0,
3
Matter fields, if present, are assumed to fall off rapidly enough so as not to affect the asymp-
totic form of the symmetry generators.
3
with
3ℓ
c= . (2.7)
2G
The explicit computation of the central charge c proceeds [12] by taking the matrix element
of (2.6) in the AdS3 vacuum (2.3). The central charge is then related to an integral of the
vector field parameterizing the diffeomorphism generated by Lm over the vacuum (2.3)
perturbed by L−m . We note that in the semiclassical regime (2.2)
c ≫ 1. (2.8)
(2.6) is of course the Virasoro algebra. Since the physical states of quantum gravity
on AdS3 must form a representation of this algebra, we have the remarkable result [12]
3ℓ
quantum gravity on AdS3 is a conformal field theory with central charge c = 2G . The
conformal field theory lives on the (t, φ) cylinder at spatial infinity.
An important example to which our considerations apply is the BTZ black hole [11,15].
In this section we recall a few of its salient features. Further details can be found in [15].
The metric for a black hole of mass M and angular momentum J is
r2
ds2 = −N 2 dt2 + ρ2 (N φ dt + dφ)2 + dr 2 , (3.1)
N 2 ρ2
with
2 r 2 (r 2 − r+
2
)
N = 2 2
,
ℓ ρ
4GJ
Nφ = − 2 , (3.2)
ρ
ρ2 = r 2 + 4GM ℓ2 − 21 r+
2
,
p
2
r+ = 8Gℓ M 2 ℓ2 − J 2 ,
where φ has period 2π. These metrics obey the boundary conditions (2.4), and the black
holes are therefore in the Hilbert space of the conformal field theory. The Bekenstein-
Hawking black hole entropy is
q
2
Area π 16GM ℓ2 + 2r+
S= = . (3.3)
4G 4G
4
It is convenient to choose the additive constants in L0 and L̄0 so that they vanish for
the M = J = 0 black hole. One then has
1
M= (L0 + L̄0 ), (3.4)
ℓ
r2 2 ℓ2 2
ds2 = − dt + r 2
dφ2
+ dr . (3.6)
ℓ2 r2
1
This is not the same as AdS3 metric (2.3) which has negative mass M = − 8G . Locally they
are equivalent since there is locally only one constant curvature metric in three dimensions.
However they are inequivalent globally. The family of metrics (3.1) can be obtained from
(2.3) by a discrete identification followed by a coordinate transformation. Hence the black
holes are topologically inequivalent to AdS3 . If one attempts to deform the black hole
solution to AdS3 by varying M , naked singularities are encountered between M = 0 and
1
M = − 8G . We will view the nonzero mass black holes as excitations of the vacuum
described by the M = 0 black hole.
This raises the question of the role played by the AdS3 vacuum. A beautiful answer to
this question was given in [16] which however requires supersymmetry. In a supersymmetric
conformal field theory the Ramond ground state, which has periodic boundary conditions
and is annihilated by the supercharge, has M = 0. We identify this with the zero-mass
black hole (3.6). The Neveu-Schwarz ground state has antiperiodic boundary conditions
and is not supersymmetric. It has a mass shift
c
L0 = L̄0 = − . (3.7)
24
Using the formula (2.7) for c and (3.4) for M we find the energy of the Neveu-Schwarz
ground state is
1
M =− , (3.8)
8G
which leads us to identify it with AdS3 . Further evidence for this identification comes from
the fact that the covariantly constant AdS3 spinors are antiperiodic under φ → φ + 2π
(because it is a 2π rotation), while the covariantly constant spinors in the extremal ℓM = J
black hole geometries are periodic [16]. Additionally it can be seen in euclidean space [17]
5
that the coordinate transformation relating (2.3) to (3.6) is exponential in t and φ, just
like the map from the plane to the cylinder.
In a theory with local dynamics, a non-extremal BTZ black hole in empty space
will Hawking radiate. However unlike the higher dimensional examples, there is (with
appropriate boundary conditions at infinity) a stable, non-extremal endpoint corresponding
to a black hole in thermal equilibrium with a radiation bath [18]. This is possible because an
infinite radiation bath in AdS3 has finite energy due to an infinite temperature redshift at
infinity. The generic hamiltonian eigenstate of the AdS3 conformal field theory presumably
corresponds to such an equilibrium state.
4. Other Examples
The Bekenstein-Hawking entropy of a black hole depends only on the area of its
horizon. In order to understand it only the near-horizon geometry of the black hole is
relevant. Hence the considerations of this paper apply to any black hole whose near
horizon geometry is AdS3 up to global identifications. There are many examples of this.
One example, considered in the string theory context in [10], is black strings in six
dimension with charges Q1 , Q5 and longitudinal momentum n. The near horizon geometry
of this black string is locally AdS3 × S 3 × M 4 with M 4 either K 3 or T 4 [19]. This may
be regarded as quantum gravity on AdS3 with an infinite tower of matter fields from both
massive string states and Kaluza-Klein modes of the S 3 × M 4 compactification. Therefore
the states are a representation of the Virasoro algebra (2.6).
The longitudinal direction along this black string lies within the AdS3 . The black
string can be periodically identified to give a black hole in five dimensions. The near-
horizon geometry is then a BTZ black hole. For n = 0 one obtains the M = J = 0 black
hole, while the generic black string yields the generic BTZ black hole.
An example involving four-dimensional extremal charged black holes is considered in
[20], where the U (1) arises from an internal circle. The near horizon geometry involves a
U (1) bundle over AdS2 . The geometry of the bundle is a quotient of AdS3 . The discrete
identification group lies entirely within one of the two SL(2, R)s. Such geometries are
considered in [21].
6
5. Microscopic Derivation of the Black Hole Entropy
We wish to count the number of excitations of the AdS3 vacuum with mass M and
angular momentum J in the semiclassical regime of large M . According to (3.4) and (2.7)
large M implies
nR + nL ≫ c, (5.1)
where nR (nL ) is the eigenvalue of L0 (L̄0 ). The asymptotic growth of the number states
of a conformal field theory with central charge c is then given by [14]
r r
cnR cnL
S = 2π + 2π . (5.2)
6 6
in exact agreement with the Bekenstein-Hawking result (3.3) for the BTZ black hole.
The microscopic derivation of the previous section rests on a key assumption: the
required 2 + 1 quantum theory of gravity must exist. The details of the theory are not
important, except in that the states must as discussed behave properly under diffeomor-
phisms. These are several constructions of such theories in which microscopic derivations
of the entropy have previously been given. It is instructive to compare these with the
present derivation.
The first is Carlip’s derivation [7] of the entropy in pure 2 + 1 gravity with no matter.
This theory has no local degrees of freedom, and can be recast as a topological Chern-
Simons theory [22,23]. There are nevertheless boundary degrees of freedom at the black
hole horizon which are described by a 1 + 1 conformal field theory. Enumeration of the
boundary states yields the Bekenstein-Hawking entropy. This boundary conformal field
theory has a different central charge (c ∼ 6 for large black holes) and so is not the same
as the conformal field theory discussed here. Nevertheless it seems likely there is some
connection between the approaches, which needs to be better understood. Relevant work
in this direction appeared in [24] where it was shown that, for a boundary at infinity rather
7
than the horizon, the Chern-Simons theory could be recast as a Liouville conformal field
theory living on the boundary. An adaptation to string theory was discussed in [25].
The second example is the five dimensional string theory black hole with charges
Q1 , Q5 and n. It was indeed found in [10] that the black hole states are described by
a c = 6Q1 Q5 conformal field theory. This agrees with (2.7) since in the effective three-
√ √
dimensional theory ℓ = 2πα′ g(Q1 Q5 )1/4 /V 1/4 and G−1 = 2V 1/4 (Q1 Q5 )3/4 /πα′ g.4
What is the relation between these two derivations of the black hole conformal field
theory? The string theory black hole has two descriptions: one as a (string–corrected) su-
pergravity solution, and the other as a bound state of Q1 D-onebranes and Q5 D-fivebranes
with momentum n. For large gQ, but small string coupling g, string perturbation theory
about the supergravity solution generally provides a good description. For small gQ, D-
brane perturbation theory is generally good. It might appear there is no overlap in the
regions of validity of the two descriptions. However, D-brane and supergravity descriptions
have an overlapping region of validity for large gQ in the near–horizon small r region [26].
The supergravity picture is good because the horizon is a smooth place even though r is
small. The D-brane picture is also useful because small r means low energies and higher
dimension corrections to the D-brane gauge theory are suppressed. However, since gQ is
large, it is a large-N gauge theory in this region.
This observation was made more precise in [13] (along with interesting generalizations
to AdSn )with the introduction of a certain scaling limit (see also [19,27,25]). In this limit,
on the supergravity side, only the near–horizon theory of strings on AdS3 remains, while
on the D-brane side one has only the conformal field theory limit of the D-brane gauge
theory. It was further noted [13] that the equivalence of these two theories was consistent
with the global SL(2, R)L ⊗ SL(2, R)R symmetry of the ground state. With the AdS3
boundary conditions defined in [12], one can go a step further and relate the action of the
full local conformal group on both sides.
Of course, string theory enables one to go well beyond the considerations of this paper.
For example one can not only determine the central charge of the theory (which is all that
is needed for the entropy), but the exact conformal theory and degeneracies at every mass
level.
4
V is the four-volume associated with the compactification to six dimensions and we use units
in which the ten-dimensional Newton’s constant is given by G = g 2 8π 6 (α′ )4 .
8
7. Discussion
Where exactly are the states accounting for the black hole entropy? The states of
the AdS3 conformal field theory are associated to the (t, φ) cylinder, and not with any
particular value of the radius r. The dynamics of the theory can be described in terms of
evolution on this cylinder without introducing fields which depend on r. In this description
nothing crosses the horizon, and there is no information loss.
The disappearance of r in the description of the quantum theory is not surprising in
the context of pure 2 + 1 gravity, since that is a topological theory with only boundary
degrees of freedom. However our discussion also applies to string theory in which one
might expect new degrees of freedom at every value of r. This is a concrete example of
the holographic principle advocated in [28,29].
It would certainly be of interest to understand in greater detail the nature of the
quantum states and dynamics in explicit examples. Consider the string theory black hole
made from the compactification of Q5 NS fivebranes and Q1 fundamental strings. In this
case the near-horizon string theory can be represented by a semi-conventional worldsheet
conformal field theory: a level Q5 SL(2, R) WZW model, with discrete identifications from
compactification [30,20,8]. The spectrum of this theory is not well-understood because it
is noncompact. However perhaps it can be better organized utilizing the action of the full
conformal group on the SL(2, R)=AdS3 target space. The Hilbert space for pure 2 + 1
gravity also needs to be better understood, perhaps using the representation as a Liouville
theory [24].
Acknowledgements
I would like to thank J. Maldacena for stimulating conversations. This work was
supported in part by DOE grant DE-FG02-96ER40559.
9
References
10