a

r
X
i
v
:
g
r
-
q
c
/
0
1
0
5
0
6
3

v
1



1
7

M
a
y

2
0
0
1
1
THE QUANTUM PHYSICS OF BLACK
HOLES: Results from String Theory

Sumit R. Das
Tata Institute of Fundamental Research, Homi Bhabha Road, Mumbai 400005, India; e-mail:
das@theory.tifr.res.in
Samir D. Mathur
Department of Physics, Ohio State University, Columbus, Ohio 43210; e-mail:
mathur@mps.ohio-state.edu
KEYWORDS: black holes, information loss, string theory, D-branes, holography
ABSTRACT: We review recent progress in our understanding of the physics of black holes.
In particular, we discuss the ideas from string theory that explain the entropy of black holes from
a counting of microstates of the hole, and the related derivation of unitary Hawking radiation
from such holes.
CONTENTS
INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
The Entropy Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
Hawking Radiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
The Information Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
Difficulties with Obtaining Unitarity . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
STRING THEORY AND SUPERGRAVITY . . . . . . . . . . . . . . . . . . . . . 9
Kaluza-Klein Mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
11-Dimensional and 10-Dimensional Supergravities . . . . . . . . . . . . . . . . . . . . 11
BRANES IN SUPERGRAVITY AND STRING THEORY . . . . . . . . . . . . . . 13
Branes in Supergravity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
BPS States . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
The Type IIB Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
D-Branes in String Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
Duality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
BLACK HOLE ENTROPYIN STRINGTHEORY: THE FUNDAMENTAL STRING 20
THE FIVE-DIMENSIONAL BLACK HOLE IN TYPE IIB THEORY . . . . . . . 22
The Classical Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
Semiclassical Thermodynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

With permission from the Annual Review of Nuclear and Particle Science. Final version of
this material appears in the Annual Review of Nuclear and Particle Science Vol. 50, published
in December 2000 by Annual Reviews, http://AnnualReviews.org.
Extremal and Near-Extremal Limits . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
Microscopic Model for the Five-Dimensional Black Hole . . . . . . . . . . . . . . . . . 24
The Long String and Near-Extremal Entropy . . . . . . . . . . . . . . . . . . . . . . . 25
A More Rigorous Treatment for Extremal Black Holes . . . . . . . . . . . . . . . . . . 28
The Gauge Theory Picture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
Other Extremal Black Holes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
BLACK HOLE ABSORPTION/DECAY AND D-BRANES . . . . . . . . . . . . . 33
Classical Absorption and grey-body Factors . . . . . . . . . . . . . . . . . . . . . . . . 33
D-brane Decay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
Why Does It Work? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
ABSORPTION BY THREE-BRANES . . . . . . . . . . . . . . . . . . . . . . . . . 39
Classical Solution and Classical Absorption . . . . . . . . . . . . . . . . . . . . . . . . 39
Absorption in the Microscopic Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
Nonextremal Thermodynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
AdS/CFT CORRESPONDENCE AND HOLOGRAPHY . . . . . . . . . . . . . . 41
The Maldacena Conjecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
Calculations Using the Conjecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
Holography and the Bekenstein Entropy Bound . . . . . . . . . . . . . . . . . . . . . . 45
Near-Horizon Limit of 5D Black Hole . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
A Suggestive Model of Holography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
DISCUSSION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
1 INTRODUCTION
Black holes present us with a very deep paradox. The path to resolving this para-
dox may well be the path to a consistent unified theory of matter and quantized
gravity.
In classical gravity, a black hole is a classical solution of the equations of mo-
tion such that there is a region of spacetime that is causally disconnected from
asymptotic infinity (see e.g. Reference [1]). The boundary of such a region is
called the event horizon.
Consider a large collection of low-density matter, in an asymptotically flat
spacetime. For simplicity, we take the starting configuration to be spherically
symmetric and nonrotating (these restrictions do not affect the nature of the
paradox that emerges). This ball of matter will collapse toward smaller radii
under its self-gravitation. At some point, the matter will pass through a critical
radius, the Schwarzschild radius R
s
, after which its further collapse cannot be
halted, whatever the equation of state. The final result, in classical general
relativity, is that the matter ends up in an infinite-density singular point, while
the metric settles down to the Schwarzschild form
ds
2
= −(1 −
2G
N
M
rc
2
)dt
2
+ (1 −
2G
N
M
rc
2
)
−1
dr
2
+ r
2
dΩ
2
. (1)
Here G
N
is Newton’s constant of gravity, and c is the speed of light. The horizon
radius of this hole is
R
s
=
2G
N
M
c
2
→2M, (2)
where the last expression arises after we set G
N
= 1, c = 1. (In what follows, we
adopt these units unless otherwise explicitly indicated; we also set ¯h = 1.)
2
D-Branes and Black Holes 3
Classically, nothing can emerge from inside the horizon to the outside. A test
mass m has effective energy zero if it is placed at the horizon; it has rest energy
mc
2
, but a negative gravitational potential energy exactly balances this positive
contribution. For a rough estimate of the horizon size, we may put this negative
energy to be the Newtonian value −G
N
Mm/r, for which R
s
∼ G
N
M/c
2
.
It may appear from the above that the gravitational fields at the horizon of
a black hole are very large. This is not true. For a neutral black hole of mass
M, the magnitude of the curvature invariants, which are the measure of local
gravitational forces, is given by
[{[ ∼
G
N
M
r
3
. (3)
Thus, at the horizon r = r
H
= 2G
N
M, the curvature scales as 1/M
2
. As a result,
for black holes with masses M G
−1/2
N
, the curvatures are very small and the
spacetime is locally rather close to flat spacetime. In fact, an object falling into a
black hole will not experience any strong force as it crosses the horizon. However,
an asymptotic observer watching this object will see that it takes an infinite time
to reach the horizon. This is because there is an infinite gravitational red-shift
between the horizon and the asymptotic region.
An important point about black hole formation is that one does not need to
crush matter to high densities to form a black hole. In fact, if the hole has mass
M, the order of magnitude of the density required of the matter is
ρ ∼
M
R
3
s

1
M
2
. (4)
Thus, a black hole of the kind believed to exist at the center of our galaxy (10
8
solar masses) could form from a ball with the density of water. In fact, given any
density we choose, we can make a black hole if we take a sufficient total mass
with that density. This fact makes it very hard to imagine a theory in which
black holes do not form at all because of some feature of the interaction between
the matter particles. As a consequence, if black holes lead to a paradox, it is
hard to bypass the paradox by doing away with black holes in the theory.
It is now fairly widely believed that black holes exist in nature. Solar-mass
black holes can be endpoints of stellar evolution, and supermassive black holes
(∼ 10
5
− 10
9
solar masses) probably exist at the centers of galaxies. In some
situations, these holes accrete matter from their surroundings, and the collisions
among these infalling particles create very powerful sources of radiation that are
believed to be the source of the high-energy output of quasars. In this arti-
cle, however, we are not concerned with any of these astrophysical issues. We
concentrate instead on the quantum properties of isolated black holes, with a
view toward understanding the problems that arise as issues of principle when
quantum mechanical ideas are put in the context of black holes.
For example, the Hawking radiation process discussed below is a quantum pro-
cess that is much weaker than the radiation from the infalling matter mentioned
above, and it would be almost impossible to measure even by itself. (The one
possible exception is the Hawking radiation at the last stage of quantum evapora-
tion. This radiation emerges in a sharp burst with a universal profile, and there
are experiments under way to look for such radiation from very small primordial
black holes.)
4 Das & Mathur
1.1 The Entropy Problem
Already, at this stage, one finds what may be called the entropy problem. One
of the most time-honored laws in physics has been the second law of thermody-
namics, which states that the entropy of matter in the Universe cannot decrease.
But with a black hole present in the Universe, one can imagine the following
process. A box containing some gas, which has a certain entropy, is dropped
into a large black hole. The metric of the black hole then soon settles down to
the Schwarzschild form above, though with a larger value for M, the black hole
mass. The entropy of the gas has vanished from view, so that if we only count
the entropy that we can explicitly see, then the second law of thermodynamics
has been violated!
This violation of the second law can be avoided if one associates an entropy to
the black hole itself. Starting with the work of Bekenstein [2], we now know that
if we associate an entropy
S
BH
=
A
H
4G
N
(5)
with the black hole of horizon area A
H
, then in any Gedanken experiment in
which we try to lose entropy down the hole, the increase in the black hole’s
attributed entropy is such that
d
dt
(S
matter
+ S
BH
) ≥ 0
(for an analysis of such Gedanken experiments, see e.g. [3]). Furthermore, an
“area theorem” in general relativity states that in any classical process, the total
area of all black holes cannot decrease. This statement is rather reminiscent of
the statement of the second law of thermodynamics—the entropy of the entire
Universe can never decrease.
Thus the proposal (Equation 5) would appear to be a nice one, but now we
encounter the following problem. We would also like to believe on general grounds
that thermodynamics can be understood in terms of statistical mechanics; in
particular, the entropy S of any system is given by
S = log Ω, (6)
where Ω denotes the number of states of the system for a given value of the
macroscopic parameters. For a black hole of one solar mass, this implies that
there should be 10
10
78
states!
But the metric (Equation 1) of the hole suggests a unique state for the geometry
of the configuration. If one tries to consider small fluctuations around this metric,
or adds in, say, a scalar field in the vicinity of the horizon, then the extra fields
soon flow off to infinity or fall into the hole, and the metric again settles down to
the form of Equation 1.
If the black hole has a unique state, then the entropy should be ln1 = 0, which
is not what we expected from Equation 5. The idea that the black hole configura-
tion is uniquely determined by its mass (and any other conserved charges) arose
from studies of many simple examples of the matter fields. This idea of unique-
ness was encoded in the statement “black holes have no hair.” (This statement
is not strictly true when more general matter fields are considered.) It is a very
interesting and precise requirement on the theory of quantum gravity plus matter
that there be indeed just the number (Equation 5) of microstates corresponding
to a given classical geometry of a black hole.
D-Branes and Black Holes 5
1.2 Hawking Radiation
If black holes have an entropy S
BH
and an energy equal to the mass M, then if
thermodynamics were to be valid, we would expect them to have a temperature
given by
TdS = dE = dM. (7)
For a neutral black hole in four spacetime dimensions, A
H
= 4π(2G
N
M)
2
, which
gives
T = (
dS
dM
)
−1
=
1
8πG
N
M
. (8)
Again assuming thermodynamical behavior, the above statement implies that if
the hole can absorb photons at a given wave number k with absorption cross
section σ(k), then it must also radiate at the same wave number at the rate
Γ(k) =
σ(k)
e
¯ h|k|
kT
− 1
d
d
k
(2π)
d
. (9)
In other words, the emission rate is given by the absorption cross section multi-
plied by a standard thermal factor (this factor would have a plus sign in place of
the minus sign if we were considering fermion emission) and a phase space factor
that counts the number of states in the wave number range

k and

k + d

k. (d
denotes the number of spatial dimensions.)
Classically, nothing can come out of the black hole horizon, so it is tempting to
say that no such radiation is possible. However, in 1974, Hawking [4] found that
if the quantum behavior of matter fields is considered, such radiation is possible.
The vacuum for the matter fields has fluctuations, so that pairs of particles and
antiparticles are produced and annihilated continuously. In normal spacetimes,
the pair annihilates quickly in a time set by the uncertainty principle. However,
in a black hole background, one member of this pair can fall into the hole, where
it has a net negative energy, while the other member of the pair can escape to
infinity as real positive energy radiation [4]. The profile of this radiation is found
to be thermal, with a temperature given by Equation 8.
Although we have so far discussed the simplest black holes, there are black
hole solutions that carry charge and angular momentum. We can also consider
generalizations of general relativity to arbitrary numbers of spacetime dimensions
(as will be required below) and further consider other matter fields in the theory.
It is remarkable that the above discussed thermodynamic properties of black
holes seem to be universal. The leading term in the entropy is in fact given by
Equation 5 for all black holes of all kinds in any number of dimensions. Further-
more, the temperature is given in terms of another geometric quantity called the
surface gravity at the horizon, κ, which is the acceleration felt by a static object at
the horizon as measured from the asymptotic region. The precise relation—also
universal—is
T =
κ

. (10)
1.3 The Information Problem
“Hawking radiation” is produced from the quantum fluctuations of the matter
vacuum, in the presence of the gravitational field of the hole. For black holes of
masses much larger than the scale set by Newton’s constant, the gravitational
6 Das & Mathur
field near the horizon, where the particle pairs are produced in this simple picture,
is given quite accurately by the classical metric of Equation 1. The curvature
invariants at the horizon are all very small compared with the Planck scale, so
quantum gravity seems not to be required. Further, the calculation is insensitive
to the precise details of the matter that went to make up the hole. Thus, if
the hole completely evaporates away, the final radiation state cannot have any
significant information about the initial matter state. This circumstance would
contradict the assumption in usual quantum mechanics that the final state of
any time evolution is related in a one-to-one and onto fashion to the initial state,
through a unitary evolution operator. Worse, the final state is in fact not even
a normal quantum state. The outgoing member of a pair of particles created by
the quantum fluctuation is in a mixed state with the member that falls into the
hole, so that the outgoing radiation is highly “entangled” with whatever is left
behind at the hole. If the hole completely evaporates away, then this final state
is entangled with “nothing,” and we find that the resulting system is described
not by a pure quantum state but by a mixed state.
If the above reasoning and computations are correct, one confronts a set of
alternatives, none of which are very palatable (for a survey see e.g. Reference [7]).
The semiclassical reasoning used in the derivation of Hawking radiation cannot
say whether the hole continues to evaporate after it reaches Planck size, since at
this point quantum gravity would presumably have to be important. The hole
may not completely evaporate away but leave a “remnant” of Planck size. The
radiation sent off to infinity will remain entangled with this remnant. But this
entanglement entropy is somewhat larger [8] than the black hole entropy S
BH
,
which is a very large number (as we have seen above). Thus, the remnant will
have to have a very large number of possible states, and this number will grow
to infinity as the mass of the initial hole is taken to infinity. It is uncomfortable
to have a theory in which a particle of bounded mass (Planck mass) can have an
infinite number of configurations. One might worry that in any quantum process,
one can have loops of this remnant particle, and this contribution will diverge,
since the number of states of the remnant is infinite. But it has been argued
that remnants from holes of increasingly large mass might couple to any given
process with correspondingly smaller strength, and then such a divergence can
be avoided.
Another possibility, advocated most strongly by Hawking, is that the hole does
evaporate away to nothing, and the passage from an intial pure state to a final
mixed state is a natural process in any theory of quantum gravity. In this view,
the natural description of states is in fact in terms of density matrices, and the
pure states of quantum mechanics that we are used to thinking about are only
a special case of this more general kind of state. Some investigations of this
possibility have suggested, however, that giving up the purity of quantum states
causes difficulties with maintaining energy conservation in virtual processes ([9];
for a counterargument, see Reference [10]).
The possibility that would best fit our experience of physics in general would
be that the Hawking radiation does manage to carry out the information of the
collapsing matter [11]. The hole could then completely evaporate away, and yet
the process would be in line with the unitarity of quantum mechanics. The
Hawking radiation from the black hole would not fundamentally differ from the
radiation from a lump of burning coal—the information of the atomic structure
of the coal is contained, though it is difficult to decipher, in the radiation and
D-Branes and Black Holes 7

t ~ M
3

t = 0
A
C
B
r = 0
r ->
Figure 1: Foliation of the black hole spacetime.
other products that emerge when the coal burns away.
1.4 Difficulties with Obtaining Unitarity
Let us review briefly the difficulties with having the radiation carry out the
information.
To study the evolution, we choose a foliation of the spacetime by smooth
spacelike hypersurfaces. This requires that the spatial slices be smooth and that
the embedding of neighboring slices changes in a way that is not too sharp.
As we evolve along this foliation, we see the matter fall in toward the center of
the hole, while we see the radiation collect at spatial infinity. It is important to
realize that the information in the collapsing matter cannot also be copied into
the radiation—in other words, there can be no quantum “Xeroxing.” The reason
is as follows. Suppose the evolution process makes two copies of a state

I
` →[ψ
I
` [ψ
i
`,
where the [ψ
i
` are a set of basis states. Then, as long as the linearity of quantum
mechanics holds, we will find

1
` +[ψ
2
` →[ψ
1
` [ψ
1
` +[ψ
2
` [ψ
2
`
and not

1
` + [ψ
2
` →([ψ
1
` + [ψ
2
`) ([ψ
1
` + [ψ
2
`).
Thus, a general state cannot be “duplicated” by any quantum process.
Figure 1 shows the spacetime in a symbolic way. We use a foliation of spacetime
by the following kind of spacelike hypersurfaces. Away from the black hole, say
for r > 4M, we let the hypersurface be a t = t
0
surface (this description uses the
Schwarzschild coordinates of Equation 1). Inside the black hole, an r =constant
surface is spacelike; let us choose r = M so that this part of the surface is neither
close to the horizon (r = 2M) nor close to the singularity (r = 0). This part of
the hypersurface will extend from some time t = 0 near the formation of the hole
to the value t = t
0
. Finally, we can connect these two parts of the hypersurface
by a smooth interpolating region that is spacelike as well. Each of the spacelike
hypersurfaces shown in
8 Das & Mathur
Figure 1 is assumed to be of this form. The lower one has t = 0, whereas
the upper one corresponds to a time t
0
∼ M
3
, where a mass ∼ M has been
evaporated away as radiation. We assume, however, that at t
0
the black hole is
nowhere near its endpoint of evaporation, either by assuming that a slow dose of
matter was continually fed into the black hole to maintain its size (the simplest
assumption) or by considering a time t
0
where say a quarter of the hole has
evaporated (and modifying the metric to reflect the slow decrease of black hole
mass).
On the lower hypersurface, we have on the left the matter that fell in to make
the hole. There is no radiation yet, so there is nothing else on this hypersurface.
Let us call this matter “A.” On the upper hypersurface, we expect the following
sources of stress energy, in a semiclassical analysis of the Hawking process. On
the left, we will still have the matter that fell in to make the hole, since this part
of the surface is common to both hypersurfaces. On the extreme right, we will
have the Hawking radiation that has emerged in the evaporation process; let us
call this “C.” In the middle are the infalling members of the particle-antiparticle
pairs. These contribute a negative value to the total mass of the system because
of the way the hypersurface is oriented with respect to the coordinate t—this
maintains overall energy conservation in the process. Let us call this part of the
state “B.”
The semiclassical process gives a state for the light matter fields, which is
entangled between components B and C. On the other hand, components A and
B are expected to somehow vanish together (or leave a Planck mass remnant),
since their energies cancel each other. At the end of the process, the radiation C
will have the energy initially present in A. But since C will be entangled with B,
the final state will not be a pure state of radiation.
We can now see explicitly the difficulties with obtaining in any theory a unitary
description of the process of black hole formation and evaporation. In a general
curved spacetime, we should be able to evolve our hypersurfaces by different
amounts at different points—this is the “many-fingered time” evolution of general
relativity extended to include the quantum matter on the spacetime. By using an
appropriate choice of this evolution, we have captured both the infalling matter A
and the outgoing radiation C on the same spacelike hypersurface. If we want the
radiation C to carry the information of the matter A, then we will need “quantum
xeroxing,” which, as mentioned above, cannot happen if we accept the principle
of superposition of quantum mechanics. It would have been very satisfactory if
we just could not draw a smooth hypersurface like the upper one in Figure 1, a
hypersurface that includes both the infalling matter and the outgoing radiation.
For example, we could have hoped that any such surface would need to be non-
spacelike at some point, or that it would need a sharp kink in one or more places.
But it is easy to see from the construction of surfaces described above that all the
hypersurfaces in the evolution are smooth. In fact, the later one is in some sense
just a time translate of the earlier one—the part t = constant in each surface
has the same intrinsic (and extrinsic) geometry for each hypersurface, and the
segment that connects this part to the r = constant part can be taken to be the
same as well. The only difference between the hypersurfaces is that the later one
has a larger r = constant part. One can further check that the infalling matter has
finite energy along each hypersurface and that scalar quantities such as dr/ds are
bounded and smooth along each surface (s is the proper length along the surface).
In the above calculations, spacetime was treated classically, but the conclusions
D-Branes and Black Holes 9
do not change even if we let the evolution of spacelike surfaces be described by the
Wheeler–de Witt equation, which gives a naive quantization of gravity; quantum
fluctuations of the spacetime may appear large in certain coordinates [12], but
such effects cancel out in the computation of Hawking radiation [13].
It thus appears that in order to have unitarity one needs a nonlocal mechanism
(which operates over macroscopic distances ∼ M) that moves the information
from A to C. Even though the spacetime appears to have no regions of Planck-
scale curvature, we must alter our understanding of how information in one set
of low-energy modes (A) moves into another set of low energy modes (C). A key
point appears to be that, in the semiclassical calculation, the radiation C emerges
from modes of the quantum field that in the past had a frequency much higher
than Planck frequency. A naive model of the quantum field would have these
modes at all frequencies, but if the complete theory of matter and gravity has
an inbuilt cutoff at the Planck scale, then the radiation C must have had its
origins somewhere else—possibly in some nonlocal combination of modes with
sub-Planckian energy. If some such possibility is true, we would obtain unitarity,
while also obtaining some nontrivial insight into the high-energy structure of the
quantum vacuum.
Basic to such an approach would be some way of understanding a black hole as
a complicated version of usual matter states, and not as an esoteric new object
that must be added to a theory of “regular” matter. It would still be true that the
final state of a system changes character significantly when its density changes
from that of a star, for instance, to the density at which it collapses to form a
black hole, but the resulting hole should still be described by the same essential
principles of quantum mechanics, density of states, statistical mechanics, etc, as
any other matter system. As we show below, string theory provides not only a
consistent theory of quantized gravity and matter, but also a way of thinking
about black holes as quantum states of the matter variables in the theory.
2 STRING THEORY AND SUPERGRAVITY
In a certain regime of parameters, string theory is best thought of as a theory
of interacting elementary strings (for expositions of superstring theory, see Ref-
erence [14]). The basic scale is set by the string tension T
s
or, equivalently, the
“string length”
l
s
=
1

2πT
s
. (11)
The quantized harmonics of a string represent particles of various masses and
spins, and the masses are typically integral multiples of 1/l
s
. Thus, at energies
much smaller than 1/l
s
, only the lowest harmonics are relevant. The interaction
between strings is controlled by a dimensionless string coupling g
s
, and the above
description of the theory in terms of propagating and interacting strings is a
good description when g
s
< 1. Even in weak coupling perturbation theory,
quantization imposes rather severe restrictions on possible string theories. In
particular, all consistent string theories (a) live in ten spacetime dimensions and
(b) respect supersymmetry. At the perturbative level, there are five such string
theories, although recent developments in nonperturbative string theory show
that these five theories are in fact perturbations around different vacua of a
single theory, whose structure is only incompletely understood at this point (for
10 Das & Mathur
a review of string dualities, see Reference [16]).
Remarkably, in all these theories there is a set of exactly massless modes that
describe the very-low-energy behavior of the theory. String theories have the
potential to provide a unified theory of all interactions and matter. The most
common scenario for this is to choose l
s
to be of the order of the Planck scale,
although there have been recent suggestions that this length scale can be consid-
erably longer without contradicting known experimental facts [15]. The massless
modes then describe the observed low-energy world. Of course, to describe the
real world, most of these modes must acquire a mass, typically much smaller than
1/l
s
.
It turns out that the massless modes of open strings are gauge fields. The
lowest state of an open string carries one quantum of the lowest vibration mode
of the string with a polarization i; this gives the gauge boson A
i
. The effective
low-energy field theory is a supersymmetric Yang-Mills theory. The closed string
can carry traveling waves both clockwise and counterclockwise along its length.
In closed string theories, the state with one quantum of the lowest harmonic
in each direction is a massless spin-2 particle, which is in fact the graviton: If
the transverse directions of the vibrations are i and j, then we get the graviton
h
ij
. The low-energy limits of closed string theories thus contain gravity and are
supersymmetric extensions of general relativity—supergravity. However, unlike
these local theories of gravity, which are not renormalizable, string theory yields
a finite theory of gravity—essentially due to the extended nature of the string.
2.1 Kaluza-Klein Mechanism
How can such theories in ten dimensions describe our 4-dimensional world? The
point is that all of the dimensions need not be infinitely extended—some of them
can be compact. Consider, for example, the simplest situation, in which the 10-
dimensional spacetime is flat and the “internal” 6-dimensional space is a 6-torus
T
6
with (periodic) coordinates y
i
, which we choose to be all of the same period:
0 < y
i
< 2πR. If x
µ
denotes the coordinates of the noncompact 4-dimensional
spacetime, we can write a scalar field φ(x, y) as
φ(x, y) =
¸
n
i
φ
n
i
(x) e
i
n
i
y
i
R
, (12)
where n
i
denotes the six components of integer-valued momenta n along the
internal directions. When, for example, φ(x, y) is a massless field satisfying the
standard Klein-Gordon equation (∇
2
x
+ ∇
2
y
)φ = 0, it is clear from Equation 12
that the field φ
n
i
(x) has (in four dimensions) a mass m
n
given by m
n
= [n[/R.
Thus, a single field in higher dimensions becomes an infinite number of fields in
the noncompact world. For energies much lower than 1/R, only the n = 0 mode
can be excited. For other kinds of internal manifolds, the essential physics is the
same. Now, however, we have more complicated wavefunctions on the internal
space. What is rather nontrivial is that when one applies the same mechanism
to the spacetime metric, the effective lower-dimensional world contains a metric
field as well as vector gauge fields and scalar matter fields.
D-Branes and Black Holes 11
2.2 11-Dimensional and 10-Dimensional Supergravities
Before the advent of strings as a theory of quantum gravity, there was an attempt
to control loop divergences in gravity by making the theory supersymmetric. The
greater the number of supersymmetries, the better was the control of divergences.
But in four dimensions, the maximal number of supersymmetries is eight; more
supersymmetries would force the theory to have fields of spin higher than 2 in the
graviton supermultiplet, which leads to inconsistencies at the level of interactions.
Such D = 4, N = 8 supersymmetric theories appear complicated but can be ob-
tained in a simple way from a D = 11, N = 1 theory or a D = 10, N = 2 theory
via the process of dimensional reduction explained above. The gravity multiplet
in the higher-dimensional theory gives gravity as well as matter fields after di-
mensional reduction to lower dimensions, with specific interactions between all
the fields.
The bosonic part of 11-dimensional supergravity consists of the metric g
MN
and a 3-form gauge field A
MNP
with an action
S
11
=
1
(2π)
8
l
9
p
[

d
11
x

−g[R−
1
48
F
MNPQ
F
MNPQ
] +
1
6

d
11
x A∧F ∧F]], (13)
where R is the Ricci scalar and F
MNPQ
is the field strength of A
MNP
. l
p
denotes
the 11-dimensional Planck length so that the 11-dimensional Newton’s constant
is G
11
= l
9
p
. This theory has no free dimensionless parameter. There is only one
scale, l
p
. Now consider compactifying one of the directions, say x
10
. The line
interval may be written as
ds
2
= e

φ
6
g
µν
dx
µ
dx
ν
+ e

3
(dx
11
− A
µ
dx
µ
)
2
. (14)
In Equation 14, the indices µ, ν run over the values 0 9. The various compo-
nents of the 11-dimensional metric have been written in terms of a 10-dimensional
metric, a field A
µ
and a field φ. Clearly, from the point of view of the 10-
dimensional spacetime, A
µ
∼ G
µ,10
is a vector and φ =
3
4
log G
10,10
is a scalar. In
a similar way, the 3-form gauge field splits into a rank-2 gauge field and a rank-3
gauge field in ten dimensions, A
µν,10
→ B
µν
and A
µνλ
→ C
µνλ
. The field A
µ
behaves as a U(1) gauge field. The bosonic massless fields are thus
1. The metric g
µν
;
2. A real scalar, the dilaton φ;
3. A vector gauge field A
µ
with field strength F
µν
;
4. A rank-2 antisymmetric tensor gauge field B
µν
with field strength F
µνλ
;
5. A rank-3 antisymmetric tensor gauge field C
µνλ
with field strength F
µνλρ
.
At low energies, all the fields in the action (Equation 13) are independent of
x
10
and the 10-dimensional action is
S =
1
(2π)
7
g
2
l
8
s

d
10
x

g(R−
1
2
(∇φ)
2

1
12
e
−φ
F
µνα
F
µνα

1
4
e

2
F
µν
F
µν
+
1
48
e
φ/2
F
µναβ
F
µναβ
) + (15)
12 Das & Mathur
where the ellipsis denotes the terms that come from the dimensional reduction
of the last term in Equation 13. F denotes the field strength of the appropriate
gauge field. The action (Equation 15) is precisely the bosonic part of the action
of Type IIA supergravity in ten dimensions. The scalar field φ is called a dilaton
and plays a special role in this theory. Its expectation value is related to the
string coupling
g
s
= exp ('φ`). (16)
The overall factor in Equation 15 follows from the fact that the 11-dimensional
measure in Equation 13 is related to the 10-dimensional measure by a factor of
the radius of x
10
(which is R), giving
2πR
(2π)
8
l
9
p
=
1
(2π)
7
g
2
l
8
s
, (17)
which defines the string length l
s
. From the 11-dimensional metric, it is clear
that R = g
2/3
l
p
, so that Equation 17 gives l
p
= g
1/3
l
s
.
The 10-dimensional metric g
µν
used in Equations 14 and 15 is called the “Ein-
stein frame” metric because the Einstein-Hilbert term in Equation 15 is canonical.
Other metrics used in string theory, most notably the “string frame” metric, dif-
fer from this by conformal transformations. In this article we always use the
Einstein frame metric.
Although we have given the explicit formulae for dimensional reduction of the
bosonic sector of the theory, the fermionic sector can be treated similarly. There
are two types of gravitinos—fermionic partners of the graviton. One of them has
positive 10-dimensional chirality whereas the other has negative chirality. The
resulting theory is thus nonchiral.
There is another supergravity in ten dimensions, Type IIB supergravity. This
cannot be obtained from D=11 supergravity by dimensional reduction. The
bosonic fields of this theory are
1. The metric g
µν
2. Two real scalars : the dilaton φ and the axion χ
3. Two sets of rank-2 antisymmetric tensor gauge fields: B
µν
and B

µν
with
field strengths H
µνλ
and H

µνλ
4. A rank-4 gauge field D
µνλρ
with a self-dual field strength F
µναβδ
.
Both the gravitinos of this theory have the same chirality.
Because of the self-duality constraint on the 5-form field strength, it is not pos-
sible to write down the action for Type IIB supergravity, although the equations
of motion make perfect sense. If, however, we put the 5-form field strength to
zero, we have a local action given by
S =
1
(2π)
7
g
2
l
8
s

d
10
x[

g(R−
1
2
(∇φ)
2

1
12
e
−φ
H
µνα
H
µνα

1
12
e
φ
(H

µνα
− χH
µνα
)(H
µνα
−χH
µνα
)].
(18)
D-Branes and Black Holes 13
Of course, these supergravities cannot be consistently quantized, since they are
not renormalizable. However, they are the low-energy limits of string theories,
called the Type IIA and Type IIB string.
3 BRANES IN SUPERGRAVITY AND STRING THEORY
Although string theory removes ultraviolet divergences leading to a finite theory
of gravity, such features as the necessity of ten dimensions and the presence of
an infinite tower of modes above the massless graviton made it unpalatable to
many physicists. Furthermore, some find the change from a pointlike particle to
a string somewhat arbitrary—if we accept strings, then why not extended objects
of other dimensionalities?
Over the past few years, as nonperturbative string theory has developed, it has
been realized that the features of string theory are actually very natural and also
perhaps essential to any correct theory of quantum gravity. A crucial ingredient
in this new insight is the fact that higher-dimensional extended objects—branes—
are present in the spectrum of string theory.
3.1 Branes in Supergravity
A closer look at even supergravity theories leads to the observation that the
existence of extended objects is natural within those theories (and in fact turns
out to be essential to completing them to unitary theories at the quantum level).
Consider the case of 11-dimensional supergravity. The supercharge Q
α
is a
spinor, with α = 1 . . . 32. The anticommutator of two supercharge components
should lead to a translation, so we write
¦Q
α
, Q
β
¦ = (Γ
A
C)
αβ
P
A
,
where C is the charge conjugation matrix. Because the anticommutator is sym-
metric in α, β, we find that there are (32 33)/2 = 528 objects on the left-hand
side of this equation, but only 11 objects (the P
A
) on the right-hand side. If
we write down all the possible terms on the right that are allowed by Lorentz
symmetry, then we find [17]
¦Q
α
, Q
β
¦ = (Γ
A
C)
αβ
P
A
+ (Γ
A
Γ
B
C)
αβ
Z
AB
+ (Γ
A
Γ
B
Γ
C
Γ
D
Γ
E
C)
αβ
Z
ABCDE
,
(19)
where the Z are totally antisymmetric. The number of Z
AB
is
11
C
2
= 55, whereas
the number of Z
ABCDE
is
11
C
5
= 478, and now we have a total of 528 objects on
the right, in agreement with the number on the left.
Although, for example, P
1
= 0 implies that the configuration has momentum
in direction X
1
, what is the interpretation of Z
12
= 0? It turns out that this
can be interpreted as the presence of a “sheetlike” charged object stretched along
the directions X
1
, X
2
. It is then logical to postulate that there exists in the
theory a 2-dimensional fundamental object (the 2-brane). Similarly, the charge
Z
ABCDE
corresponds to a 5-brane in the theory. The 2-brane has a 2 + 1 = 3-
dimensional world volume and couples naturally to the 3-form gauge field present
in 11-dimensional supergravity, just as a particle with 1-dimensional world line
couples to a 1-form gauge field as

A
µ
dx
µ
. The 5-brane is easily seen to be the
14 Das & Mathur
magnetic dual to the 2-brane, and it couples to the 6-form that is dual to the
3-form gauge field in 11 dimensions.
Thus, it is natural to include some specific extended objects in the quantization
of 11-dimensional supergravity. But how does this relate to string theory, which
lives in ten dimensions? Let us compactify the 11-dimensional spacetime on a
small circle, thus obtaining 10-dimensional noncompact spacetime. Then, if we
let the 2-brane wrap this small circle, we get what looks like a string in ten
dimensions. This is exactly the Type IIA string that had been quantized by the
string theorists! The size of the small compact circle turns out to be the coupling
constant of the string.
We can also choose not to wrap the 2-brane on the small circle, in which case
there should be a two-dimensional extended object in Type IIA string theory.
Such an object is indeed present—it is one of the D-branes shown to exist in
string theory by Polchinski [18]. Similarly, we may wrap the 5-brane on the small
circle, getting a 4-dimensional D-brane in string theory, or leave it unwrapped,
getting a solitonic 5-brane, which is also known to exist in the theory.
Thus, one is forced to a unique set of extended objects in the theory, with
specified interactions between them—in fact, there is no freedom to add or remove
any object, nor to change any couplings.
3.2 BPS States
A very important property of such branes is that when they are in an unexcited
state, they preserve some of the supersymmetries of the system, and are thus
Bogomolny-Prasad-Sommerfield saturated (BPS) states. Let us see in a simple
context what a BPS state is.
Consider first a theory with a single supercharge Q = Q

:
¦Q, Q¦ = 2Q
2
= 2H, (20)
where H is the Hamiltonian. These relations show that the energy of any state
cannot be negative. If
H[ψ` = E[ψ` (21)
then
E = 'ψ[H[ψ` = 'ψ[Q
2
[ψ` = 'Qψ[Qψ` ≥ 0, (22)
where the equality holds in the last step if and only if
Q[ψ` = 0, (23)
that is, the state is supersymmetric. Nonsupersymmetric states occur in a “mul-
tiplet” containing a bosonic state [B` and a fermionic state [F` of the same
energy:
Q[B` ≡ [F`, Q[F` = Q
2
[B >= E[B`. (24)
Supersymmetric states have E = 0 and need not be so paired.
Now suppose there are two such supersymmetries:
Q

1
= Q
1
, Q

2
= Q
2
, Q
2
1
= H, Q
2
2
= H, ¦Q
1
, Q
2
¦ = Z, (25)
where Z is a “charge”; it will take a c number value on the states that we consider
below. In a spirit similar to the calculations above, we can now conclude
0 ≤ 'ψ[(Q
1
± Q
2
)
2
[ψ` = 2E ±2Z. (26)
D-Branes and Black Holes 15
This implies that
E ≥ [Z[, (27)
with equality holding if and only if
(Q
1
−Q
2
)[ψ` = 0, or (Q
1
+Q
2
)[ψ` = 0. (28)
Now we have three kinds of states:
1. States with Q
1
[ψ` = Q
2
[ψ` = 0. These have E = 0 and do not fall into
a multiplet. By Equation 26, they also have Z[ψ` = 0, so they carry no
charge.
2. States not in category 1, but satisfying Equation 28. For concreteness,
take the case (Q
1
− Q
2
)[ψ` = 0. These states fall into a “short multiplet”
described by, say, the basis ¦[ψ`, Q
1
[ψ`¦. Note that
Q
2
[ψ` = Q
1
[ψ`, Q
2
Q
1
[ψ` = Q
2
1
[ψ` = E[ψ`, (29)
so that we have no more linearly independent states in the multiplet. Such
states satisfy
E = [Z[ > 0 (30)
and are called BPS states. Note that by Equation 26, the state with Z`0
satisfies (Q
1
−Q
2
)[ψ` = 0 while the state with Z < 0 satisfies (Q
1
+Q
2
)[ψ` =
0.
3. States that are not annihilated by any linear combination of Q
1
, Q
2
. These
form a “long multiplet” ¦[ψ`, Q
1
[ψ`, Q
2
[ψ`, Q
2
Q
1
[ψ`¦. They must have
E > [Z[ > 0.
In the above discussion, we have regarded the BPS states as states of a quantum
system, but a similar analysis applies to classical solutions. In 10-dimensional su-
pergravity, the branes mentioned above appear as classical solutions of the equa-
tions of motion, typically with sources. They are massive solitonlike objects and
therefore produce gravitational fields. Apart from that, they produce the p-form
gauge fields to which they couple and, in general, a nontrivial dilaton. A brane in
a general configuration would break all the supersymmetries of the theory. How-
ever, for special configurations—corresponding to “unexcited branes”—there are
solutions which retain some of the supersymmetries. These are BPS saturated
solutions, and since they have the maximal charge for a given mass, they are
stable objects. We use such branes below in constructing black holes.
3.3 The Type IIB Theory
Instead of using the 11-dimensional algebra, we could have used the 10-dimensional
algebra and arrived at the same conclusions. In a similar fashion, the existence of
BPS branes in Type IIB supergravity follows from the corresponding algebra. For
each antisymmetric tensor field present in the spectrum, there is a corresponding
BPS brane. Thus we have
1. D(−1)-brane, or D-instantons, carrying charge under the axion field χ;
2. NS1-brane, or elementary string, carrying electric charge under B
µν
;
3. D1-brane, carrying electric charge under B

µν
;
16 Das & Mathur
4. 3-brane, carrying charge under D
µνρλ
;
5. NS5-brane, carrying magnetic charge under B
µν
;
6. D5-brane, carrying magnetic charge under B

µν
;
7. D7-brane, the dual of the D(−1) brane.
We have denoted some of the branes as D-branes. These play a special role in
string theory, as explained below.
3.4 D-Branes in String Theory
Consider the low-energy action of the supergravity theories. We have remarked
above that the equations of motion for the massless fields admit solitonic solu-
tions, where the solitons are not in general pointlike but could be p-dimensional
sheets in space [thus having (p +1)-dimensional world volumes in spacetime]. In
each of these cases, the soliton involves the gravitational field and some other
(p + 1)-form field in the supergravity multiplet so that the final solution carries
a charge under this (p + 1)-form gauge field. In fact, in appropriate units, this
soliton is seen to have a mass equal to its charge and is thus an object satisfying
the BPS bound of supersymmetric theories. This fact implies that the soliton
is a stable construct in the theory. The possible values of p are determined en-
tirely by the properties of fermions in the Type II theory. It turns out that for
Type IIA, p must be even (p = 0, 2, 4, 6), whereas for Type IIB p must be odd
(p = −1, 1, 3, 5, 7). Recalling the massless spectrum of these theories, we find that
for each antisymmetric tensor field there is a brane that couples to it. Because
the brane is not pointlike but is an extended object, we can easily see that there
will be low-energy excitations of this soliton in which its world sheet suffers small
transverse displacements that vary along the brane (in other words, the brane
carries waves corresponding to transverse vibrations). The low-energy action is
thus expected to be the tension of the brane times its area. For a single 1-brane,
for instance, the action for long-wavelength deformations is
S =
T
1
2

d
2
ξ
α

det(g
αβ
), (31)
where ξ
α
, with α = 1, 2, denotes an arbitrary coordinate sytem on the D1-brane
world sheet and g
αβ
denotes the induced metric on the brane,
g
αβ
= ∂
α
X
µ

β
X
ν
η
µν
. (32)
The X
µ
, with µ = 0, , 9, denotes the coordinates of a point on the brane.
This action is invariant under arbitrary transformations of the coordinates ξ
α
on
the brane. To make contact with to the picture discussed above, it is best to
work in a static gauge by choosing ξ
1
= X
0
and ξ
2
= X
9
. The induced metric
(Equation 32) then becomes
g
αβ
= η
αβ
+ ∂
α
φ
i

β
φ
j
δ
ij
, (33)
where φ
i
= X
i
with i = 1, 8, are the remaining fields. The determinant in
Equation 31 may be then expanded in powers of ∂φ
i
. The lowest-order term is
just the free kinetic energy term for the eight scalar fields φ
i
.
It is now straightforward to extend the above discussion for higher-dimensional
branes. Now we can have both transverse and longitudinal oscillations of the
D-Branes and Black Holes 17
brane. For a p-brane we have (9 − p) transverse coordinates, labeled by I =
1, (9 − p), and hence as many scalar fields on the (p + 1)-dimensional brane
world volume, φ
i
. It turns out that the longitudinal waves are carried by a
U(1) gauge field A
α
with the index α = 0, 1, p ranging over the world volume
directions. The generalization of Equation 31 [called the Dirac-Born-Infeld (DBI)
action] is
S =
T
p
2

d
p+1
ξ

det(g
αβ
+F
αβ
), (34)
where F
αβ
is the gauge field strength. Once again, one can choose a static gauge
and relate the fields directly to a string theory description. The low-energy
expansion of the action then leads to electrodynamics in (p +1) dimensions with
some neutral scalars and fermions.
In the above description, we had obtained the branes as classical solutions of the
supergravity fields; this is the analog of describing a point charge by its classical
electromagnetic potential. Such a description should apply to a collection of a
large number of fundamental branes all placed at the same location. But we
would like to obtain also the microscopic quantum physics of a single brane. How
do we see such an object in string theory?
The mass per unit volume of a D-brane in string units is 1/g, where g is the
string coupling. So the brane would not be seen as a perturbative object at weak
coupling. But the excitations of the brane will still be low-energy modes, and
these should be seen in weakly coupled string theory. In fact, when we quantize
a free string we have two choices: to consider open strings or closed strings. If we
have an open string then we need boundary conditions at the ends of the string
that do not allow the energy of vibration of the string to flow off the end. There
are two possibilities: to let the ends move at the speed of light, which corresponds
to Neumann (N) boundary conditions, or to fix the end, which corresponds to
Dirichlet (D) boundary conditions. Of course we can choose different types of
conditions for different directions of motion in spacetime. If the ends of the open
strings are free to move along the directions ξ
α
but are fixed in the other directions
X
i
, then the ends are constrained to lie along a p-dimensional surface that we
may identify with a p-brane, and such open strings describe the excitations of
the p-brane [18]. Because these branes were discovered through their excitations,
which were in turn a consequence of D-type boundary conditions on the open
strings, the branes are themselves called D-branes. (For a review of properties of
D-branes, see Reference [19].)
An interesting effect occurs when two such branes are brought close to each
other. The open strings that begin and end on the first brane will describe
excitations of the first brane, and those that begin and end on the second brane
will describe excitations of the second brane. But, as shown in Figure 2, an open
string can also begin on the first brane and end on the second, or begin on the
second and end on the first, and this gives two additional possibilities for the
excitation of the system. These four possibilities can in fact be encoded into a
22 matrix, with the (ij) element of the matrix given by open strings that begin
on the ith brane and end on the jth brane. This structure immediately extends
to the case where N branes approach each other.
The low-energy limits of open string theories are generally gauge theories. In-
deed the low-energy worldbrane theory of a collection of N parallel Dp-branes
turns out to be described by a non-Abelian U(N) gauge theory in (p + 1) di-
mensions [21]. The N
2
gauge fields are best written as an N N matrix A
α
ab
,
18 Das & Mathur

Figure 2: Open strings ending on D-branes.
where a, b = 1 N. Similarly, there are N
2
scalar fields φ
i
ab
, which transform
according to the adjoint representation of U(N). The coupling constant of the
theory is
g
2
YM
∼ g
s
(l
s
)
p−3
. (35)
Each of these fields has its corresponding fermionic partner. Since BPS states
break half the supersymmetries of the original theory, we have a supersymmetric
Yang-Mills theory with 16 supercharges.
The potential for such a theory turns out to be Tr([φ
i
, φ
j
])
2
, so that in the
ground state one can always choose a gauge so that all the φ
i
s are diagonal. In
fact, the diagonal entries in the matrix φ
i
ab
denote the locations of the branes.
Thus
1
2

i
aa

i
bb
) denotes the center-of-mass transverse coordinate of the pair of
branes labeled by a and b, while d
i
ab
= (φ
i
aa
− φ
i
bb
) denotes the separation along
direction i. A nonzero expectation value for d
i
ab
means that the gauge symmetry
is spontaneously broken.
Generically, the gauge group is broken to [U(1)]
N
—this is the situation when
all the branes are separated from each other, i.e. all the d
i
ab
are nonzero. However,
when some number, say M, of the branes are coincident, the corresponding d
i
ab
are zero for a, b = 1 M, resulting in an enhanced unbroken symmetry, U(M)
[U(1)]
N−M
.
The analog of the DBI action for a collection of N branes is not completely
known at present, although there are some proposals (for discussion see Refer-
ence [22] and references therein). However, the low-energy action is that of a
standard supersymmetric Yang-Mills theory in p + 1 dimensions. In fact, this
is the dimensional reduction of the 10-dimensional supersymmetric Yang-Mills
theory to p + 1 dimensions. The latter has no scalar fields but has 10 com-
ponents of the gauge field, A
µ
, where µ = 0, , 9, each of which is a N N
matrix. Under dimensional reduction, the components parallel to the brane, i.e.
µ = 0, (10−p), , 9, remain gauge fields, whereas the components transverse to
the brane with µ = 1, (9−p) are scalars on the world volume and are renamed
as φ
i
.
The ground states of the above D-branes are BPS states states, which means
that they are stable. Recently, other kinds of non-BPS and unstable D-branes
have been constructed in string theory [20]. In this article, however, we restrict
D-Branes and Black Holes 19
ourselves to BPS branes and their excitations.
Not all the branes that were listed for Type IIA and Type IIB theory are D-
branes. Consider the Type IIA theory. It arises from a dimensional reduction
of 11-dimensional supergravity on a circle. The 11-dimensional theory has 5-
branes and 2-branes, and the 2-brane can end on the 5-brane. If we wrap on
the compact circle both the 5-brane and the 2-brane that ends on it, then in the
Type IIA theory we get a D4-brane and an open string ending on the D4-brane.
But if we do not wrap the 5-brane on the circle, and thus get a 5-brane in the
Type IIA theory, then the open string cannot end on this brane, since there is
no corresponding picture of such an endpoint in 11 dimensions. The physics of
the 5-brane in Type IIA theory is an interesting one, but we will not discuss it
further. Much more is known about the physics of D-branes in Type IIA and IIB
theories, since they can be studied through perturbative open string theory.
Branes of different kinds can also form bound states, and for specific instances
these can be threshold bound states. A useful example is a bound state of Q
1
D1-branes and Q
5
D5-branes. When these branes are not excited, they are in a
threshold bound state. The open strings that describe this system are (a) (1, 1)
strings with both endpoints on any of the D1-branes; (b) (5, 5) strings with both
endpoints on D5-branes; and (c) (1, 5) and (5, 1) strings with one endpoint on
any of the D1-branes and the other endpoint on one of the D5-branes.
3.5 Duality
A remarkable feature of string theory is the large group of symmetries called
dualities (see Reference [16] for review). Consider Type IIA theory compactified
on a circle of radius R. We can take a graviton propagating along this compact
direction; its energy spectrum would have the form [n
p
[/R. But we can also
wind an elementary string on this circle; the spectrum here would be 2πT
s
[n
w
[R,
where n
w
is the winding number of the string and T
s
its tension. We note that
if we replace R →1/2πT
s
R, then the energies of the above two sets of states are
simply interchanged. In fact this map, called T-duality, is an exact symmetry of
string theory; it interchanges winding and momentum modes, and because of an
effect on fermions, it also turns a Type IIA theory into Type IIB and vice versa.
The Type IIB theory also has another symmetry called S-duality, there the
string coupling g
s
goes to 1/g
s
. At the same time, the role of the elementary string
is interchanged with the D1-brane. Such a duality, which relates weak coupling to
strong coupling while interchanging fundamental quanta with solitonic objects, is
a realization of the duality suggested for field theory by Montonen & Olive [23].
The combination of S- and T-dualities generates a larger group, the U-duality
group.
There are other dualities, such as those that relate Type IIA theory to heterotic
string theory, and those that relate these theories to the theory of unoriented
strings. In this article, we do not use the idea of dualities directly, but we note
that any black hole that we construct by using branes is related by duality maps
to a large class of similar holes that have the same physics, so the physics obtained
is much more universal than it may at first appear.
20 Das & Mathur
4 BLACKHOLE ENTROPY INSTRINGTHEORY: THE FUN-
DAMENTAL STRING
String theory is a quantum theory of gravity. Thus, black holes should appear
in this theory as excited quantum states. An idea of Susskind [24] has proved
very useful in the study of black holes. Because the coupling in the theory is
not a constant but a variable field, we can study a state of the theory at weak
coupling, where we can use our knowledge of string theory. Thus we may compute
the “entropy” of the state, which would be the logarithm of the number of states
with the same mass and charges. Now imagine the coupling to be tuned to
strong values. Then the gravitational coupling also increases, and the object must
become a black hole with a large radius. For this black hole we can compute the
Bekenstein entropy from (Equation 5), and ask if the microscopic computation
agrees with the Bekenstein entropy.
For such a calculation to make sense, we must have some assurance that the
density of states would not shift when we change the coupling. This is where
BPS states come in. We have shown above that the masses of BPS saturated
states are indeed determined once we know their charges, which are simply their
winding numbers on cycles of the compact space. Thus, for such states, we may
calculate the degeneracy of states at weak coupling and, since the degeneracy
can be predicted also at strong coupling, compare the result with the Bekenstein-
Hawking entropy of the corresponding black hole. Such states give, at strong
coupling, black holes that are “extremal”—they have the minimal mass for their
charge if we require that the metric does not have a naked singularity.
The extended objects discussed in the previous section have played an impor-
tant role in understanding black holes in string theory.
An example of such an object is a fundamental string in Type IIA or IIB string
theory. Let some of the directions of the 10-dimensional spacetime be compacti-
fied on small circles. Take a fundamental string and wrap it n
1
times around one
of these circles, say along the 9 direction, which has a radius R. This will produce
a rank-2 NS field B
09
with a charge n
1
, which is a “winding charge.” The energy
of the state is E = 2πn
1
T
s
R, which saturates the BPS bound. From the point of
view of the noncompact directions, this looks like a massive point object carrying
electric charge under the gauge field that results from the dimensional reduction
of the rank-2 field. From the microscopic viewpoint, the state of such a string is
unique (it does have a 256-fold degeneracy due to supersymmetry, but we can ig-
nore this—it is not a number that grows with n
1
). Thus, the microscopic entropy
is zero.
If we increase the coupling, we expect a charged black hole. Furthermore, this
is an extremal black hole, since for a given charge n
1
this has the lowest allowed
mass given by the energy given above. However, this black hole turns out to have
a vanishing horizon area. One way to understand this is to note that the tension
of the string “pinches” the circle where the string was wrapped. Thus, entropy is
zero from both the microscopic and the black hole viewpoints, which is consistent
but not really interesting.
To prevent this pinching, we can put some momentum along the string, which
amounts to having traveling waves move along the string. The momentum modes
have an energy that goes as 1/R, so now this circle attains a finite size. If we put
waves that are moving only in one of the directions, we still have a BPS saturated
state, but with a further half of the supersymmetries broken. The total energy of
D-Branes and Black Holes 21
the state is now given by E = 2πn
1
T
s
+n
2
/R, where the second term is now the
contribution from the momentum waves. Because of the winding, the effective
length of the string is L
eff
= 2πn
1
R, so that the momentum may be written as
P = (2πn
1
n
2
)/L
eff
. Thus, for given values of n
1
and n
2
, a large number of states
have the same energy. These correspond to the various ways one can get the
oscillator level n
0
= n
1
n
2
. In addition, it is necessary to consider the fact that
there are eight possible polarizations, and there are fermionic oscillators as well.
The resulting degeneracy of states has been known since the early days of string
theory. For large n
1
and n
2
, the number of states asymptotes to
n(n
1
, n
2
) ∼ exp[2

2

n
1
n
2
]. (36)
From the viewpoint of the noncompact directions, we have an object with two
quantized charges n
1
and n
2
. The charge n
1
is due to the winding of the string,
as mentioned above. The second charge is due to the presence of momentum,
which results in a term in the 10-dimensional metric proportional to dx
9
dx
0
.
Then, by the Kaluza-Klein mechanism explained in the previous section, this
is equivalent to a gauge field A

0
from the viewpoint of the noncompact world.
The corresponding charge is n
2
and is an integer since the momentum in the
compact direction is quantized. At strong coupling, this object is described by
an extremal black hole solution with two charges. The identification of such
fundamental string states with the corresponding classical solution was proposed
by Dabholkar & Harvey [25]. The horizon area is still zero and the curvatures
are large near the horizon.
In an important paper, Sen [27] argued that the semiclassical entropy (for sim-
ilar black holes in heterotic string theory) is given not by the area of the event
horizon but by the area of a “stretched horizon.” This is defined as the location
where the curvature and local temperature reach the string scale. It is indeed
inconsistent to trust a classical gravity solution beyond this surface, since the
curvatures are much larger than the string scale and stringy corrections to su-
pergravity become relevant. There is a great deal of ambiguity in defining the
stretched horizon precisely. However, Sen found that the area of the stretched
horizon in units of the gravitational constant is proportional to

n
1
n
2
, which is
precisely the logarithm of the degeneracy of states given by Equation 36. This
was the first indication that degeneracy of states in string theory may account
for Bekenstein-Hawking entropy. However, in this example, the precise coeffi-
cient cannot be determined, since the definition of the stretched horizon is itself
ambiguous.
It is clear that what we need is a black hole solution that (a) is BPS and (b)
has a nonzero large horizon area. The BPS nature of the solution would ensure
that degeneracies of the corresponding string states are the same at strong and
weak couplings, thus allowing an accurate computation of the density of states.
A large horizon would ensure that the curvatures are weak and we can therefore
trust semiclassical answers. In that situation, a microscopic count of the states
could be compared with the Bekenstein-Hawking entropy in a regime where both
calculations are trustworthy.
It turns out that this requires three kinds of charges for a 5-dimensional black
hole and four kinds of charges for a 4-dimensional black hole.
22 Das & Mathur
5 THE FIVE-DIMENSIONAL BLACK HOLE IN TYPE IIB
THEORY
The simplest black holes of this type are in fact 5-dimensional charged black
holes. Extremal limits of such black holes provided the first example where the
degeneracy of the corresponding BPS states exactly accounted for the Bekenstein-
Hawking entropy [28]. There are several such black holes in Type IIA and IIB
theory, all related to each other by string dualities. We describe below one such
solution in Type IIB supergravity.
5.1 The Classical Solution
We start with 10-dimensional spacetime and compactify on a T
5
along (x
5
, x
6
,
x
9
). The noncompact directions are then (x
0
. x
4
). There is a solution of
Type IIB supergravity that represents (a) D5-branes wrapped around the T
5
,
(b) D1-branes wrapped around x
5
, and (c) some momentum along x
5
. Finally,
we perform a Kaluza-Klein reduction to the five noncompact dimensions. The
resulting metric is
1
ds
2
= −[f(r)]
−2/3

1 −
r
2
0
r
2

dt
2
+ [f(r)]
1/3

dr
2
1 −
r
2
0
r
2
+ r
2
dΩ
2
3
¸
¸
, (37)
where
f(r) ≡

1 +
r
2
0
sin
2
α
1
r
2

1 +
r
2
0
sin
2
α
5
r
2

1 +
r
2
0
sinh
2
σ
r
2

. (38)
Here r is the radial coordinate in the transverse space, r
2
=
¸
4
i=1
(x
i
)
2
, and dΩ
2
3
is the line element on a unit 3-sphere S
3
. The solution represents a black hole
with an outer horizon at r = r
0
and an inner horizon at r = 0. The background
also has nontrivial values of the dilaton φ, and there are three kinds of gauge
fields:
1. A gauge field A
(1)
0
, which comes from the dimensional reduction of the rank-
2 antisymmetric tensor gauge field B

05
in 10 dimensions. This is nonzero,
since we have D1-branes along x
5
.
2. A gauge field A
(2)
0
, which comes from the dimensional reduction of an off-
diagonal component of the 10-dimensional metric g
05
. This is nonzero, since
there is momentum along x
5
.
3. A rank-2 gauge field B

ij
with i and j lying along the 3-sphere S
3
. This is
the dimensional reduction of a corresponding B

ij
in ten dimensions, since
there are 5-branes along (x
5
x
9
).
The presence of these gauge fields follows in a way exactly analogous to the di-
mensional reduction from 11-dimensional supergravity discussed in Equations 13
and 14. The corresponding charges Q
1
, Q
5
, N are given by
Q
1
=
V r
2
0
sinh2α
1
32π
4
gl
6
s
Q
5
=
r
2
0
sinh2α
5
2gl
2
s
N =
V R
2
32π
4
l
8
s
g
2
r
2
0
sinh 2σ, (39)
1
We use the notation of Horowitz et al [26].
D-Branes and Black Holes 23
where V is the volume of the T
4
in the x
6
x
9
directions, R is the radius of the
x
5
circle, and g is the string coupling.
The charge N comes from momentum in the x
5
direction. If we look at the
higher-dimensional metric before dimensional reduction, it is straightforward to
identify the Arnowitt-Deser-Misner (ADM) momentum as
P
ADM
=
N
R
(40)
The ADM mass of the black hole is given by
M =
RV r
2
0
32π
4
g
2
l
8
s
[cosh2α
1
+ cosh 2α
5
+ cosh 2σ]. (41)
5.2 Semiclassical Thermodynamics
The semiclassical thermodynamic properties of the black hole may be easily ob-
tained from the classical solution using Equations 5 and 10 and the relationship
G
5
=

5
g
2
l
8
s
RV
(42)
between the 5-dimensional Newton’s constant G
5
and V, R, g, l
s
.
The expressions are
S
BH
=
A
H
G
5
=
RV r
3
0

3
l
8
s
g
2
coshα
1
coshα
5
coshσ
and (43)
T
H
=
1
2πr
0
coshα
1
coshα
5
cosh σ
. (44)
5.3 Extremal and Near-Extremal Limits
The solution given above represents a general 5-dimensional black hole. The
extremal limit is defined by
r
2
0
→0 α
1
, α
5
, σ →∞
Q
1
, Q
5
, N = fixed. (45)
In this limit, the inner and outer horizons coincide. However, it is clear from the
above expressions that ADM mass and entropy remain finite while the Hawking
temperature vanishes. Of particular interest is the extremal limit of the entropy,
S
extremal
= 2π

Q
1
Q
5
N. (46)
This is a function of the charges alone and is independent of other parameters,
such as the string coupling, volume of the compact directions, and string length.
In the following, we are interested in a special kind of departure from extremal-
ity. This is the regime in which α
1
, α
5
1, but r
0
and σ are finite. In this case,
the total ADM mass may be written in the suggestive form
E =
RQ
1
gl
2
s
+
16π
4
RV Q
5
l
6
s
+E
L
+E
R
. (47)
24 Das & Mathur
Here we have defined
E
L
=
N
R
+
V Rr
2
0
e
−2σ
64π
4
g
2
l
8
s
and E
R
=
V Rr
2
0
e
−2σ
64π
4
g
2
l
8
s
, (48)
and used the approximations
r
2
1
≡ r
2
0
sinh
2
α
1

r
2
0
2
sinh2α
1
=
16π
4
gl
6
s
Q
1
V
,
r
2
5
≡ r
2
0
sinh
2
α
5

r
2
0
2
sinh2α
5
= gl
2
s
Q
5
, (49)
and the expressions for the charges (Equation 39). The meaning of the subscripts
L and R will be clear soon. In a similar fashion, the thermodynamic quantities
may be written as
S = S
L
+S
R
1
T
H
=
1
2

1
T
L
+
1
T
R

, (50)
where
S
L,R
=
RV r
1
r
5
r
0
e
±σ
16π
3
g
2
l
8
s
T
L,R
=
r
0
e
±σ
2πr
1
r
5
. (51)
The above relations are highly suggestive. The contribution to the ADM mass
from momentum along x
5
is E
m
= E
L
+E
R
. In the extremal limit σ →∞, it then
follows from Equations 48 and 40 that E
m
→P
ADM
, which implies that the waves
are moving purely in one direction. This is the origin of the subscripts L and
R—we have denoted the direction of momentum in the extremal limit as left, L.
For finite σ we have both right- and left-moving waves, but the total momentum
is still N/R. The splitting of various quantities into left- and right-moving parts
is typical of waves moving at the speed of light in one space dimension. In fact,
using Equations 50, 51, and 39, we can easily see that the following relations
hold:
T
i
=
1
π

E
i
RQ
1
Q
5
=
S
i

2
RQ
1
Q
5
, i = L, R. (52)
Finally, we write the expressions that relate the temperature T and the entropy
S to the excitation energy over extremality ∆E = E
m
− N/R:
∆E =
π
2
2
(RQ
1
Q
5
) T
2
and S = S
extremal
[1 + (
R∆E
2N
)
1
2
]. (53)
5.4 Microscopic Model for the Five-Dimensional Black Hole
This solution and the semiclassical properties described above are valid in regions
where the string frame curvatures are small compared with l
−2
s
. If we require this
to be true for the entire region outside the horizon, we can study the regime of
validity of the classical solution in terms of the various parameters. A short
calculation using Equations 37, 38, and 39 shows that this is given by
(gQ
1
), (gQ
5
), (g
2
N) 1. (54)
D-Branes and Black Holes 25
In this regime, the classical solution is a good description of an intersecting set
of D-branes in string theory.
However, in string theory, the way to obtain a state with large D-brane charge
Q is to have a collection of Q individual D-branes. Thus, the charges Q
1
and
Q
5
are integers, and we have a system of Q
1
D1-branes and Q
5
D5-branes. The
third charge is from a momentum in the x
5
direction equal to N/R. Thus, N is
quantized in the microscopic picture to ensure a single valued wave function.
In the extremal limit (Equation 45), these branes are in a threshold bound
state, i.e. a bound state with zero binding energy. This is readily apparent from
the classical solution. The total energy (Equation 47) is then seen to be the sum
of the masses of Q
1
D1-branes, Q
5
D5-branes and the total momentum equal to
N/R.
The low-energy theory of a collection of Q Dp-branes is a (p + 1)-dimensional
supersymmetric gauge theory with gauge group U(Q). However, now we have
a rather complicated bound state of intersecting branes. The low-energy theory
is still a gauge theory but has additional matter fields (called hypermultiplets).
Instead of starting from the gauge theory itself, we first present a physical picture
of the low-energy excitations.
5.5 The Long String and Near-Extremal Entropy
The low-energy excitations of the system become particularly transparent in the
regime where the radius of the x
5
circle, R, is much larger than the size of
the other four compact directions V
1/4
. Then the effective theory is a (1 + 1)-
dimensional theory living on x
5
. The modes are essentially those of the oscil-
lations of the D1-branes. These, however, have four rather than eight polariza-
tions. This is because it costs energy to pull a D1-brane away from the D5-brane,
whereas the motions of the D1-branes along the four directions parallel to the
D5-branes, but transverse to the D1-branes themselves (i.e. along x
6
x
8
), are
gapless excitations.
Even though a system of static D1- and D5-branes is marginally bound, there
is a nonzero binding energy whenever the D1-branes try to move transverse to
the D5-branes. If we had a single D1-brane and a single D5-brane, the quantized
waves would be massless particles with four flavors. Since we have a supersym-
metric theory, we have four bosons and four fermions. When we have many
D1-branes and D5-branes, it would appear at first sight that there should be
4Q
1
Q
5
such flavors: Each of the D1-branes can oscillate along each of the D5-
branes, and there are four polarizations for each such oscillation. This is indeed
the case if the D1-branes are all separate.
However, there are other possible configurations. These correspond to several
of the D1-branes joining up to form a longer 1-brane, which is now multiply
wound along the compact circle. In fact, if we only have some number n
w
of D1-
branes without anything else wrapping a compact circle, they would also prefer
to join into a long string, which is of length 2πn
w
R. This was discovered in an
analysis of the nonextremal excitations of such D1-branes [29]. S-duality relates
D1-branes to fundamental wrapped strings. If one requires that the nonextremal
excitations of the D1 brane are in one-to-one correspondence with the known
near-extremal excitations of the fundamental string and therefore yield the same
degeneracy of states, one concludes that the energies of the individual quanta of
oscillations of the D1 string must have fractional momenta p
i
= n
i
/(n
w
R), which
26 Das & Mathur
simply means that the effective length of the string is 2πn
w
R.
For the situation we are discussing, i.e. D1-branes bound to D5-branes, the
entropically favorable configuration turns out to be that of a single long string
formed by the D1-branes joining up to wind around the x
5
direction Q
1
Q
5
, i.e. an
effective length of L
eff
= 2πRQ
1
Q
5
[30]. We thus arrive at a gas of four flavors of
bosons and four flavors of fermions living on a circle of size L
eff
. Since the string
is relativistic, these particles are massless. The problem is to count the number of
states with a given energy E and given total momentum P = N/R and see if the
results agree with semiclassical thermodynamics of black holes. This approach to
the derivation of black hole thermodynamics comes from Callan & Maldacena [31]
and Horowitz & Strominger [32], with the important modification of the multiply
wound D-string proposed by Maldacena & Susskind [30]. Our treatment below
follows Reference [33].
5.5.1 Statistical Mechanics in 1 + 1 Dimensions
At weak coupling these bosons and fermions form an ideal gas, which we assume
thermalizes. Consider the general case of such an ideal gas with f flavors of
massless bosons and fermions (each polarization counted once) living on a circle
of size L. For large L, we should be able to treat the system in a canonical
ensemble characterized by an inverse temperature β conjugate to the energy and
a chemical potential α conjugate to the momentum. If n
r
denotes the number of
particles with energy e
r
and momentum p
r
, the partition function Z is
Z = e
h
=
¸
states
exp
¸
−β
¸
r
n
r
e
r
−α
¸
r
n
r
p
r
¸
. (55)
Then α and β are determined by requiring
E = −
∂h
∂β
and P = −
∂h
∂α
. (56)
The average number of particles n
r
in state (e
r
, p
r
) is then given by
ρ(e
r
, p
r
) =
1
e
βer+αpr
±1
, (57)
where as usual the plus sign is for fermions and the minus sign is for bosons.
Finally, the entropy S is given by the standard thermodynamic relation
S = h +αP +βE. (58)
The above quantities may be easily evaluated:
P =
fLπ
8
¸
1
(β + α)
2

1
(β −α)
2

,
E =
fLπ
8
¸
1
(β + α)
2
+
1
(β −α)
2

,
and S =
fLπ
4
[
1
β +α
+
1
β − α
]. (59)
D-Branes and Black Holes 27
Since we have massless particles in one spatial dimension, they can be either
right-moving, with e
r
= p
r
, or left-moving, with e
r
= −p
r
. The corresponding
distribution functions then become
ρ
R
=
1
e
(β+α)er
±1
and ρ
L
=
1
e
(β−α)er
± 1
. (60)
Thus, the combinations T
R
= 1/(β + α) and T
L
= 1/(β − α) act as effective
temperatures for the right- and left-moving modes, respectively, resulting in
1
T
=
1
2

1
T
L
+
1
T
R

. (61)
In fact, all the thermodynamic quantities can be split into left- and a right-moving
pieces: E = E
R
+ E
L
P = P
R
+ P
L
S = S
R
+ S
L
. The various quantities
E
L
, E
R
, P
L
, P
R
, S
L
, S
R
may be read off from Equation 59,
T
R
=

8E
R
Lπf
=
4S
R
πfL
T
L
=

8E
L
Lπf
=
4S
L
πfL
. (62)
5.5.2 Near-Extremal Thermodynamics from the Long String
It is now clear that the statistical mechanics of this free gas in fact reproduces
the thermodynamics of the near-extremal black hole. The relationships of Equa-
tion 62 become the relations Equation 52 if
fL = 8πRQ
1
Q
5
. (63)
Given this agreement, it is obvious that the thermodynamic relations of Equa-
tion 53 are also exactly reproduced. In particular, the physical temperature T
agrees with the Hawking temperature, as is clear from Equations 61 and 50.
Recall that the size of the x
5
circle is 2πR. Thus, one way to satisfy the
relation in Equation 63 is to consider L = 2πR and f = 4Q
1
Q
5
. This would
be the situation if we had Q
1
D1-branes individually wrapping the x
5
direction.
Another way to satsify Equation 63 is to take L = 2πQ
1
Q
5
R and f = 4. This is
the content for the long string, where all the D1-branes join up to form a single
multiply wound string.
It might appear that thermodynamics does not distinguish between these two
configurations. However, the above thermodynamic relations assume that the
system is extensive [30]. One condition for that is that T
L
L, T
R
L 1. It follows
from Equation 62 that this would require
S
L
, S
R
f. (64)
There is an important difference between the two cases. When multiple branes are
each singly wound, the energy of a single particle is n/R for some integer n. Thus,
a near-extremal configuration may be obtained from an extremal configuration by
adding n pairs of left- and right-handed quanta, leading to an excitation energy
∆E = 2n/R. For the long string, similarly, we would have ∆E = 2n/Q
1
Q
5
R,
since the length is now 2πQ
1
Q
5
R. From this it is apparent that for the multiple
singly wound branes,
S
L
= 2π

Q
1
Q
5
(N + n) and S
R
= 2π

Q
1
Q
5
n, (65)
28 Das & Mathur
whereas for the multiply wound long string
S
L
= 2π

Q
1
Q
5
N +n and S
R
= 2π

n. (66)
It is then clear that the extensivity condition (Equation 64) is always satisfied for
the long string, whereas it is violated by the singly wound branes when Q
1
, Q
5
and N are all comparable and large.
The multiply wound string has much lower energy excitations, with a minimum
gap of 1/Q
1
Q
5
R. This is consistent with the conditions for the validity of statis-
tical mechanics for black holes [34]. The point is that for statistical mechanics to
hold, the temperature of a system must be such that the specific heat is larger
than one. It follows from the relations in Equation 53 that this requires (up to
numerical factors of order one)
∆E >
1
RQ
1
Q
5
, (67)
which indicates that the system should have an energy gap 1/Q
1
Q
5
R. This is
exactly what the long string provides.
5.6 A More Rigorous Treatment for Extremal Black Holes
In the above discussion we adopted a simple model of the D1-D5 bound state
and used it to compute the entropy of momentum excitations. We now discuss
more rigorous results on the count of states that can be obtained by using the
fact that we are counting supersymmetric states when we compute the entropy
of an extremal black hole. In fact, such a chain of arguments was used in the
first exact calculation of the microscopic entropy of a black hole by Strominger &
Vafa [28]. We outline the main ideas of such a treatment here. (This subsection
assumes some familiarity with dualities in string theory.)
(A): We count not the actual states of the system but an “index.” The
simplest example of an index is the Witten index of a supersymmetric system
Tr[(−1)
F
e
−βH
] (F is the fermion number and H is the hamiltonian). This index
counts the number of bosonic ground states minus the number of fermionic ground
states. The ground states are maximally supersymmetric states, so if one were
interested in states preserving maximal supersymmetry, then the Witten index
would provide useful information. Note, however, that the index can only give
a lower bound on the number of supersymmetric states, since some states are
counted with a negative sign. As we change the parameters of the theory, states
can appear and disappear in Bose-Fermi pairs, but the index is a robust quantity.
In the black hole context, we are interested not in the ground states of the
system but in states preserving some fraction of the supersymmetry. (In fact,
the states we want carry momentum and thus a nonzero total energy.) We need
a generalization of the Witten index that will carry nontrivial information about
the count of excited states of the system. Of course, we will again obtain a
lower bound to the state count for the given quantum numbers; we then hope
that the bound will approximate the actual state count well enough to permit a
meaningful comparison to the Bekenstein entropy.
(B): Let us consider the black hole discussed by Strominger & Vafa [28]. Take
Type IIB theory but let the complete 10-dimensional spacetime be K3S
1
M
5
,
D-Branes and Black Holes 29
where M
5
are the noncompact directions, K3 is a special kind of compact 4-
dimensional surface, and K3 along with the circle S
1
makes up the five compact
directions. Since one compact direction is S
1
, there is really no difference between
Type IIA and Type IIB theories here; we can go from one to the other by a T-
duality.
What kind of charges can we place on the compact directions to make the
black hole? In analogy to the discussion above of the black hole in T
4
S
1
M
5
,
we can wrap various branes around the cycles of the compact directions, making
sure that they all stretch along the S
1
. Then we can put a momentum charge
along the S
1
direction and obtain a black hole carrying a momentum charge N
p
along with the charges arising from the wrapped branes.
We need some way to describe the charges carried by the branes. Type IIB
theory has 1-, 3-, and 5-dimensional branes. On the other hand, K3 has 0-, 2-,
and 4-dimensional cycles on which something can be wrapped, so we can wrap
all these branes on the K3 S
1
space as desired. As mentioned above, we can
view the theory as Type IIA on K3 S
1
, and then we note that there is a
duality map that relates Type IIA compactified on K3 to heterotic string theory
compactified on T
4
. Type IIA compactified on K3 S
1
maps to heterotic string
theory compactified on T
5
, and here the charges are easy to understand. The
charges that are dual to the wrapped branes in Type IIA now arise in the heterotic
theory from momentum and winding of the elementary heterotic string on the
compact directions of the theory. The possible charge states are characterized
by points on a lattice Γ
21,5
, which means a lattice with 21 positive signature
directions and 5 negative signature directions. The only relevant property of these
charges is that a bilinear invariant, which we may call Q
2
F
, gives the invariant
length squared of a charge vector in this space. This invariant is a generalization
of n
p
n
w
, the product of winding and momentum charges in any one direction;
such a product is invariant under T-duality, since the two kinds of charges get
interchanged.
(C): Since we will be looking for a quantity that will not vary with the con-
tinuous parameters of the system, we can take the size of the S
1
to be much
larger than the size of the K3. Then the physics of the wrapped branes looks like
a (1 + 1)-dimensional sigma model, with the space direction being the S
1
cycle.
It is possible to count the degeneracy of the ground states of the wrapped branes
and infer that the target space of this sigma model must be essentially the sym-
metric product of Q
2
F
/2+1 copies of K3 (Q
2
F
as defined is an even number). Here
the term “essentially” means that this target space may be a deformation of the
stated one, but the topological properties of the space would remain unchanged,
and the space would remain hyper-K¨ ahler, which means that the number of su-
persymmetries would also remain the same. The Q
2
F
/2 + 1 copies of K3 are
analogous to the Q
1
Q
5
strands of the string that we took as the vibrating ele-
ments in the heuristic treatment of Section 5.5 above; each point on each strand
could move on T
4
in that case, and the T
4
is now replaced by K3. The fact that
we take a symmetric product amounts to saying that the strands are identical
and must be so treated in any quantum mechanical count.
We now consider n
p
units of momentum along the S
1
direction of this sigma
model, just as we did in the case of T
4
compactification. If we hold n
p
fixed
and take the size of S
1
to infinity, then we will be looking at modes of very long
wavelength, longer than any length scale that existed in the sigma model. The
30 Das & Mathur
physics of such modes must be a conformal field theory (CFT) in 1+1 dimensions,
and in particular we can separate the excitations into left- and right-moving
modes. If we do not excite the right movers and let the left movers carry the
given momentum, then we will have states with half the maximal supersymmetry:
The right supercharges will annihilate the state and the left ones will not.
If we have such a conformal theory with N = 2 or larger supersymmetry, then
it can be shown that the elliptic genus
Tr[(−1)
J
L
−J
R
e
−βH
y
J
L
]
is also a robust quantity that does not change when the moduli are varied. Here
J
L
and J
R
are integer-valued U(1) charges for the left and right sectors; these are
the U(1) symmetries that rotate the two supercharges of N = 2 supersymmetry
into each other. We want only the count of states when the charges are large,
since only then can we compare the count with the Bekenstein entropy computed
from the classical gravity solution. In a CFT, the degeneracy of states at an
excitation level n above the ground state (in, say, the left sector) is given in
terms of the central charge of the CFT by the formula
d(n, c) ∼ e

nc
6
.
The central charge for a supersymmetric sigma model with a hyper-K¨ ahler space
is easy to calculate. The dimension of the target space must be of the form 4k.
Each bosonic direction contributes 1 to the central charge, and its corresponding
fermion gives an extra 1/2. Thus c = 6k. Putting in the values of k and n = n
p
,
we get
S = lnd(n
p
, Q
F
) ∼ 2π

n
p

Q
2
F
2
+ 1

,
which agrees with the Bekenstein entropy for the classical geometry with the
same charges.
(D): Although the above expression was derived for large n
p
, it is also possible
to compute the elliptic genus for the symmetric product of copies of K3 without
making any approximation. One can relate the elliptic genus when the target
space is a symmetric product of k copies of the space X to the elliptic genus
when there is only one copy of the space [35]. The derivation of this relation
is combinatorial; one symmetrizes over bosonic states and antisymmetrizes over
fermionic states. But the case where the target space is only one copy of X is the
theory of a single string on X, which is a simple CFT to handle, and the elliptic
genus can be explicitly written.
If we apply the above method to the case where the compact space is T
4
instead of K3, we find that the elliptic genus is zero. This happens because there
are fermion zero modes on the torus, which relate a set of bosonic states to an
identical spectrum of fermionic states, and there is a cancellation in the elliptic
genus and we get no useful information about the state count. Maldecena et al
showed [36] that one can use the fact that we have N = 4 supersymmetry in the
CFT to argue that the quantity
Tr[(−1)
J
L
−J
R
J
2
R
e
−βH
y
J
L
]
is a topological invariant in this case. This quantity is nonzero, and it turns out
to give for large charges a state count that reproduces the Bekenstein entropy.
D-Branes and Black Holes 31
5.7 The Gauge Theory Picture
We now show how the above picture of a nonlinear sigma model follows from
gauge theory considerations. As explained above, we are interested in the long-
distance limit of the theory of the D1-D5 system. This system has (1, 1), (5, 5),
and (1, 5) strings and has (4, 4) supersymmetry when viewed as a 1-dimensional
theory living along x
5
. There is a global R symmetry which rotates the various
supercharges: SO(4) = SU(2)
L
SU(2)
R
. In fact, this R symmetry is nothing
but the rotation symmetry in the four transverse directions. The lowest oscillator
modes of the open strings lead to a U(Q
1
)U(Q
5
) gauge theory with the following
supersymmetry multiplets:
1. The states of the (1, 1) and (5, 5) strings along the transverse directions
x
1
x
4
form a “vector multiplet.” There are thus four scalars transforming
as (2, 2) of the R symmetry group and four fermions. The left-handed
fermions transform as (1, 2) of R symmetry, and the right-handed fermions
transform as (2, 1).
2. The states of the (1, 1) and (5, 5) strings with polarizations along the T
4
directions form a “hypermultiplet.” This also has four scalars transforming
trivially under the R-symmetry group and has four fermions. The left-
moving fermions now transform as (2, 1) and the right-moving fermions as
(1, 2).
3. The states of the (1, 5) and (5, 1) strings lead to hypermultiplets whose
transformation properties under R symmetry are the same as those of the
previous hypermultiplets.
We want the branes to be on top of each other. This requires the vector
multiplet scalars to have vanishing expectation values, since these represent the
relative distances between individual branes in the transverse space. On the
other hand, the scalars in the hypermultiplet will have expectation values. This
is known as the Higgs branch of the theory. As a result, the vector multiplets
typically acquire a mass.
2
The hypermultiplet scalars are massless. However,
they are not all independent of each other but are related by a set of conditions.
It is necessary to solve for the independent set of fields and find the low-energy
theory of these fields. Even at the classical level, this is a complicated affair. It
has been pursued elsewhere [37, 38], although a detailed low-energy theory has
not yet been rigorously derived.
Nevertheless, it is possible to obtain an important universal quantity of the
low-energy conformally invariant theory—the central charge. This is because the
central charge is related, by superconformal symmetry, to the anomaly of the R-
symmetry current. The R-symmetry current is anomalous because the left- and
right-moving fermions behave differently under R-symmetry transformations. At
high energies where the theory does not have superconformal invariance but does
have (4, 4) supersymmetry, the anomaly coefficient k may be easily computed
in terms of the number of hypermultiplets and vector multiplets, and it turns
out to be k = Q
1
Q
5
. Because of the well-known ’t Hooft anomaly matching
conditions, this coefficient is the same at low energies, where the theory flows to
2
A U(1) component of the vector multiplet remains massless, representing the overall motion
of the branes in transverse space. However, the U(1) parts of the fields are decoupled from the
SU(Q1) ×SU(Q5) part and can be ignored in this discussion.
32 Das & Mathur
a superconformal theory. Superconformal symmetry then determines the central
charge of the superconformal algebra in terms of the anomaly coefficient c =
6Q
1
Q
5
. The asymptotic density of states in a conformal field theory can be now
determined in terms of the central charge, and for BPS states this leads to the
expression for entropy given above.
There is a slightly different but equivalent approach to the problem. D1-branes
in the D5-branes can be considered as “instanton strings” of the 6-dimensional
U(Q
5
) gauge theory on the D5-brane [39]. Actually these are not really instantons
but rather solitonic objects. Consider configurations of the gauge theory that are
independent of x
5
and time x
0
. Such configurations may be regarded as configura-
tions of a Euclidean 4-dimensional gauge theory living in the x
6
x
9
directions.
Instantons are such configurations, which are self-dual in this 4-dimensional sense.
From the point of view of the 6-dimensional theory, they are thus solitonic strings.
The number of D1-branes is equal to the instanton number Q
1
. These instantons
form a continuous family of solutions with the same energy—the corresponding
parameters form the “instanton moduli space.” The low-energy excitations of
the system are described by the collective excitations or dynamics on this moduli
space. Because of translation invariance of the original configurations along x
5
and x
0
, the collective coordinates are functions of x
5
and x
0
—their dynamics is
thus given by a (1 +1)-dimensional field theory that is essentially a sigma model
with the target space as the moduli space of these instantons. This is the sigma
model discussed in the previous section. In fact, the independent moduli are
obtained by solving a set of equations identical to the equations that determine
the independent set of hypermultiplet scalars.
It turns out that in the orbifold limit, the theory of these instanton strings may
be interpreted in terms of a gas of strings [35, 40] that are wound around the x
5
and move along the T
4
with a total winding number k = Q
1
Q
5
. These strings
can be singly wound or multiply wound. The “long string” used in the previous
sections is nothing but the maximally wound sector. As we have argued, this is
the entropically favored configuration.
In the orbifold limit, however, the supergravity approximation is not valid,
which is why one has to turn on some of the supergravity moduli fields to go to
the correct regime. In fact, by turning on a NS-NS two-form field, one can get rid
of the Coulomb branch altogether, forcing the branes to lie on top of each other,
thus ensuring that we are dealing with the Higgs branch of the theory [42, 41].
The behavior of the system with such moduli turned on has been further analyzed
elsewhere [43].
5.8 Other Extremal Black Holes
There is a wide variety of black hole solutions in string theory (for reviews and
references see e.g. Reference [44]), and numerous studies have related them to
states in string theory. One notable example is a 4-dimensional black hole with
four kinds of charges [45], where once again there is an exact agreement of the
extremal entropy. Near-extremal limits of these black holes have also been stud-
ied; there is, in fact, a natural analog of the long string approximation in which
these black holes are regarded as intersecting branes in 11-dimensional supergrav-
ity [47]. Rotating versions of such 5- and 4-dimensional black holes also provide
exact microscopic derivations of the extremal entropy [48].
D-Branes and Black Holes 33
6 BLACK HOLE ABSORPTION/DECAY AND D-BRANES
In classical gravity, black holes absorb matter via standard wave scattering by an
effective potential arising from the gravitational field. Consider the absorption of
some field. An incident wave is partly scattered back and partly absorbed, and the
absorption cross section can be calculated by solving the relevant wave equation
in the black hole background. Hawking radiation is a quantum process. However,
the radiation rate can be obtained by considering a black hole in equilibrium with
its environment of radiated quanta. The principle of detailed balance then relates
the radiation rate to the absorption cross section by the relation in Equation 9.
We are interested in the decay of a slightly nonextremal black hole by Hawking
radiation. The dominant decay mode is the emission of massless neutral particles.
It is clear that the black hole will evaporate until the excess mass over extremality
has been radiated away, leaving an extremal black hole that is stable. Note
that for such black holes, the specific heat is positive (in contrast to neutral
black holes, which have negative specific heat), so that this evaporation is a well-
defined process. Furthermore, as long as the extremal limit corresponds to a large
black hole, one would expect that at all stages of the decay process, semiclassical
approximations are valid for these large black holes.
In the microscopic models described above, the processes of absorption and
radiation appear to be rather different. Consider a closed string incident on the
intersecting D-brane configurations representing black holes. When the closed
string hits the D-brane, it can either reflect back or split up into open strings
whose ends are attached to the brane. The open strings then do not leave the
brane and hence do not re-emerge in the asymptotic region. Thus, there is a
finite probability that the closed string quantum will be absorbed by the black
hole, which should be calculated using standard rules of quantum mechanics.
Radiation is the reverse of this process: Open strings can join up to form closed
strings, which can then escape from the brane system. At low energies, the
closed string that is propagating in the bulk of spacetime may be replaced by the
quantum of some supergravity field, whereas the open strings may be replaced
by quanta of the low-energy gauge theory living on the brane. We are interested
in this low-energy limit.
6.1 Classical Absorption and grey-body Factors
To compare the absorption or radiation rates of semiclassical black holes and
D-brane configurations, we must compute the classical absorption cross sections
of the relevant fields in the black hole background. We are interested in the
low-energy behavior of these absorption cross sections.
Let φ denote all the supergravity fields collectively. Then the wave equation
satisfied by a test field in the classical background φ
0
can be obtained by substi-
tuting the perturbed field φ = φ
0
+ δφ in the supergravity equations of motion
and retaining the part linear in δφ. The equations of motion would generically
couple the various perturbations, leading to a complicated set of equations. How-
ever, various symmetries of the background classical solution allow us to isolate
special perturbations that satisfy simple decoupled equations.
For the 5-dimensional black hole we are discussing, there are 20 such bosonic
perturbations, called minimal scalars, that satisfy decoupled massless minimally
coupled Klein-Gordon equations. As we shall see, these have the largest absorp-
34 Das & Mathur
tion rate and hence dominate the Hawking radiation. Examples of such fields
are traceless parts of the components of the 10-dimensional metric, which have
polarizations along the T
4
in the (x
6
x
8
) direction.
Scattering of massless particles by 4-dimensional black holes was studied in
the 1970s [49], and the low-energy limit of the s-wave absorption cross section for
massless minimally coupled scalars was found to be exactly equal to the horizon
area for Schwarzschild and charged Reissner-Nordstrom black holes. A similar
calculation for the near-extremal 5-dimensional black hole described above also
showed that the leading term of the absorption cross section is again the horizon
area [50, 33]. It was soon realized, however, that this is in fact a universal result:
For all spherically symmetric black holes in any number of dimensions, the s-
wave cross section of minimally coupled massless scalars is always the area of the
horizon [51].
The actual calculation of the cross section involves solving the wave equation
analytically in various regimes and then matching solutions in the overlapping
regions. As an example, we describe the essential steps that led to the result
obtained in [51].
Consider a metric of the form
ds
2
= −f(r)dt
2
+h(r)

dr
2
+r
2
dΩ
2
p

, (68)
where dΩ
2
p
denotes the line element on a unit p-sphere. Let the horizon be at
r = r
H
so that f(r
H
) = 0. We want to solve the minimally coupled massless
equation ∇
2
φ = 0 for a spherically symmetric scalar field φ(r, t). Using time
translation invariance, we decompose the field into fields of definite frequency,
φ
ω
(r), through φ(r, t) = e
−iωt
φ
ω
(r). The frequency component φ
ω
(r) satisfies
[(r
p
F(r)∂
r
)
2

2
R
2p
(r)]φ
ω
(r) = 0, (69)
where
F(r) = ¦f(r)[h(r)]
p−1
¦
1/2
R(r) = r[h(r)]
1/2
. (70)
Let l denote the largest length scale in the classical solution. We then solve the
equation to lowest order in λ = ωl. Then there are two regions in which the
equation can be easily solved. In the outer region ωr > λ, the equation becomes
that of a scalar field in flat background and may be solved in terms of Bessel
functions. In the near-horizon region ωr < λ, the function R(r) →R
H
= R(r
H
)
may be replaced and the equation may be solved after a simple change of variables.
We must then match the two solutions in the overlapping region.
The physical input is a boundary condition. It is necessary to impose a bound-
ary condition so that there is no outgoing wave at the horizon. In the asymptotic
region, there is both an incoming and an outgoing wave, and matching the so-
lution to the near-horizon solution yields the ratio of the outgoing and incoming
components and hence the absorption probability. This can be converted into an
absorption cross section σ(k) for a plane wave of momentum

k using standard
techniques. The result is, as advertised,
Lim
ω→0
σ(k) =

(p+1)/2
Γ(
p+1
2
)
R
p
H
= A
H
, (71)
where A
H
is the horizon area and the last equality follows from the form of the
metric (Equation 68).
D-Branes and Black Holes 35
For the 5-dimensional black hole, the absorption cross section for such mini-
mally coupled scalars may be calculated in the so-called dilute gas regime defined
by
r
0
, r
N
<r
1
, r
5
r
1
r
5

r
0
r
N
∼ O(1) (72)
when the energy ω is in the regime
ωr
5
<1
ω
T
L
∼ O(1)
ω
T
R
∼ O(1). (73)
The various parameters of the classical solution have been defined in previous
sections. The final result is [52]
σ(ω) = 2π
2
r
2
1
r
2
5
πω
2
(e
ω/T
H
− 1)
(e
ω/2T
L
−1)(e
ω/2T
R
−1)
. (74)
Remarkably, the cross section appears as a combination of thermal factors, even
though it is a result of a solution of the Klein-Gordon equation. When T
L
T
R
and ω < T
L
, one has T
H

1
2
T
R
and the cross section becomes exactly equal
to the area of the horizon. The expression (Equation 74) is called a grey-body
factor because of the nontrivial energy dependence.
Finally, let us briefly discuss the absorption of higher angular momentum modes
and higher spin fields. Writing the relevant wave equations, we find for such
modes an additional “centrifugal” potential near the horizon that suppresses
absorption of these modes by the black hole at low energies [49]. The exceptions
are modes of spin-1/2 fields that are supersymmetric partners of the minimal
scalars. When these modes have their orbital angular momentum, the absorption
cross section is twice the area of the horizon, in agreement with expectations from
supersymmetry.
6.2 D-brane Decay
A first-principles calculation of the decay of a slightly nonextremal 5-dimensional
black hole would involve a strongly coupled gauge theory of the D1-D5 gauge
theory coupled to the bulk supergravity fields. As discussed above, the gauge
theory is difficult to analyze even in weak coupling. However, we have a fairly
good effective theory—the long string model—and weak coupling calculations in
the model give exact strong coupling answers. One might wonder whether the
model may be used to compute decay rates. Once again one may hope that
weak coupling calculations could give reasonable answers for the same reasons
that they succeeded in predicting the right thermodynamics. In this subsection,
we show that this is indeed unambiguously possible for a certain set of modes,
and indeed weak coupling calculations for these modes agree exactly with the
semiclassical results described in the previous subsection.
The theory of the long string is a (1 +1)-dimensional massless supersymmetric
field theory with four flavors of bosons φ
I
, the index I referring to the four
directions of the T
4
along x
6
x
9
and their corresponding fermionic partners.
We want to figure out how these degrees of freedom couple to the supergravity
modes in the bulk. In fact, we are mostly interested in supergravity modes that
are minimal scalars in the dimensionally reduced theory. Out of these 20 scalars,
it is particularly easy to write down the interaction of the long string with the
36 Das & Mathur
transverse traceless components of the 10-dimensional graviton h
IJ
, with indices
I, J lying along the T
4
. At low energy, the relevant action is
T

d
2
ξ∂
α
φ
I

α
φ
J
g
IJ
, (75)
where the ξ
α
denote the coordinates on the long string world sheet and g
IJ
is
the metric on the T
4
. T is a constant. This form of the action follows from the
principle of equivalence. When there is no supergravity background, the action
should be a free action given by Equation 75 with g
IJ
replaced by δ
IJ
. The fields
φ
I
transform as vectors under the local SO(4) transformations of the tangent
space of the T
4
. This requires that at low energies the action in the presence of
a background is given by Equation 75.
If we expand the metric as
g
IJ
= δ
IJ
+


10
h
IJ
(ξ, φ
I
, x
1
x
4
) (76)
we immediately get the coupling of h
IJ
with the long string degrees of freedom.
In Equation 76, we have made explicit the dependence of the supergravity field
on the coordinates on the long string world sheet ξ = (x
0
, x
5
), the T
4
coordinates
φ
I
(which are the fields of the long string theory), and the coordinates transverse
to the black hole (x
1
x
4
). The 10-dimensional gravitational coupling, κ
10
, is
related to the string coupling and string length by
κ
2
10
= 64π
7
g
2
l
8
s
. (77)
The factor


10
in Equation 76 is determined by requiring that the field h
IJ
be
canonically coupled and follows from the bulk supergravity action (Equation 18).
There will be other terms in the action, but these would contain more derivatives
and are suppressed at low energies.
It is clear from Equation 75 that the decay of a nonextremal state into the mode
h
IJ
is dominated by the process of annihilation of a pair of modes of vibration
from the long string, one of which is left-moving and the other right-moving (see
Figure 3). Because we are working in the limit where the radius R of the x
5
circle
is much larger than the size of the T
4
, we can ignore the dependence of h
IJ
on
φ
I
. Furthermore, we are interested in computing s-wave absorption/decay, which
depends on the transverse coordinates only through the radial variable.
The constant T is an effective tension of the long string and depends on the
details of the rather complicated D1-D5 bound state. It appears that the action
we have written has little predictive power unless we figure out these details.
However, it has been recognized [33] that the cross section in question does not
depend on such details. This is because the free kinetic term of the fields φ
I
has to be canonically normalized. Because the interaction is also quadratic in
φ
I
, normalization thus completely removes the constant T from the action. The
amplitudes that result from this action are therefore universal. In the following,
we give the basic steps for calculation of the decay rate when the mode h
IJ
carries no momentum along x
5
and is in an s-wave in the transverse space. This
corresponds to the decay into neutral scalars from the noncompact 5-dimensional
point of view.
Consider the annihilation of two long string modes: (a) one with polarization
I and momenta p = (p
0
, p
5
) [along the (x
0
, x
5
) direction] and (b) one with po-
larization J and momenta q = (q
0
, q
5
) into an h
IJ
bulk mode with momenta
D-Branes and Black Holes 37
D5
D1
Figure 3: Absorption of supergravity mode by D-brane.
k = (k
0
, k
1
, , k
4
, 0) along the (x
0
, , x
5
) directions. Using Equations 75 and
76, we find that the decay rate for this process is
Γ(p, q; k) = (2π)
2
Lδ(p
0
+q
0
−k
0
)δ(p
5
+q
5
−k
5
)

2
(p q)
2
(2p
0
L)(2q
0
L)(2k
0
V LV
4
)
V
4
d
4
k
(2π)
4
.
(78)
The delta functions impose energy conservation and momentum conservation
along the x
5
direction (momentum is not conserved in the other directions be-
cause of Dirichlet conditions). The denominators come from normalization of the
modes: the open string modes are normalized on the circle along x
5
of circum-
ference L = 2πQ
1
Q
5
R whereas the closed string field is normalized in the entire
space with volume V LV
4
, where V
4
is the volume of the noncompact four spatial
dimensions x
1
x
4
.
To obtain the total cross section for production of a scalar, one must now
average over all initial states. Because these states are drawn from a thermal
ensemble, the decay rate is
Γ(k) = (
L

)
2


−∞
dp
5


−∞
dq
5
ρ(p
0
, p
5
)ρ(q
0
, q
5
)Γ(p, q; k), (79)
where ρ(p
0
, p
5
) denotes the Bose distribution functions discussed in the previous
section. The integral may be evaluated easily and the answer is
Γ(k) = 2π
2
r
2
1
r
2
5
πω
2
1
(e
ω
2T
L
−1)(e
ω
2T
R
− 1)
. (80)
Finally, the decay rate is converted into an absorption cross section by multipli-
cation by an inverse Bose distribution of the supergravity mode:
σ(ω)
D−brane
= 2π
2
r
2
1
r
2
5
πω
2
e
ω
T
−1
(e
ω
T
L
− 1)(e
ω
T
R
− 1)
. (81)
We have expressed the answer in terms of the parameters in the classical solution
by using the length of the effective string,
L = 2πQ
1
Q
5
R =

4
r
2
1
r
2
5
V R
κ
2
. (82)
38 Das & Mathur
The physical temperature T is related to T
L
and T
R
by Equation 61. Thus, when
T
L
T
R
, we have T ∼ 2T
R
. However, we know that the quantities T
L
and T
R
above are in fact exactly the same as the semiclassical quantities and T = T
H
.
For very low energies ω <T
L
, one of the thermal factors above can be simplified.
Using Equations 49 through 51, we see that in this limit
σ(ω) = 2π
2
(r
1
r
5
)
2
T
L
=
(r
1
r
5
)
2
S
L
RQ
1
Q
5
= A
H
. (83)
We have used Equation 52 and the fact that in the regime T
L
T
R
we have S
L
=
S = S
BH
= 2πA
H

2
5
, where κ
2
5
= κ
2
10
/2πRV and κ
2
10
is given by Equation 77.
We have also used Equation 49 to express r
1
, r
5
in terms of g, l
s
, V, Q
1
, Q
5
. This
is exactly the low-energy classical result. The fact that the cross section turns
out to be proportional to the horizon area was already known [31]. The precise
calculation outlined above was performed in Reference [33].
Even before taking this low-energy limit, the entire result (Equation 81) is in
exact agreement with the semiclassical grey-body factor [52].
It is clear from Equation 80 that the thermal factor in the decay rate comes
from the thermal factors of the long string modes. In the strict classical limit,
we can have absorption but no Hawking radiation; this comes about because the
temperature in this limit is zero, so the corresponding thermal factor suppresses
emission completely while absorption is still nonzero [50].
The supergravity mode can also split into a pair of fermions. However, then
the thermal factors that appear in Equation 80 are Fermi-Dirac distributions and
these go to a constant at low energies, rather than diverging as Bose factors do.
As a result, the corresponding cross section is suppressed at low energies. In
a similar way, we may consider emission/absorption of a fermionic supergravity
mode. This would require one left-moving bosonic mode of the long string with
a right-moving fermionic mode. Following the above calculations, it is apparent
that the thermal factor of the left-moving long string mode gives the right powers
of energy to lead to a nonzero low-energy cross section, and the thermal factor for
the right-moving mode provides the thermal factor of the emitted supergravity
fermion—and this is a Fermi-Dirac factor as expected [31].
Classical emission of higher angular momentum modes is suppressed by cen-
trifugal barriers provided by the gravitational field. In the microscopic model,
such emission is again suppressed, but for a different reason. Note that the
bosonic degrees of freedom of the long string are all vectors under the internal
SO(4) of the T
4
but are scalars under the tangent space SO(4) of the transverse
space. Therefore, these can never collide to give rise to modes with nonzero angu-
lar momentum in transverse space. Fermions on the long string, however, carry
transverse tangent space indices (typically as a R-symmetry index) since they are
dimensional reductions of 10-dimensional fermions. Thus, multifermion processes
are responsible for emission of higher angular momentum modes. However, these
can be seen to be necessarily suppressed, since they involve more than two fermion
fields and therefore correspond to higher-dimension operators on the long string
world sheet. Angular-momentum–mode emission is not completely understood in
the long string picture—although the qualitative aspects have been understood
[54].
Similar calculations have been performed for several other situations, e.g. for
absorption/emission of charged scalars from both 5- and 4-dimensional black
holes [53, 52].
D-Branes and Black Holes 39
The low-energy effective action used above follows from a Dirac-Born-Infeld
(DBI) action of the long string in the presence of supergravity backgrounds. One
might wonder whether this DBI action can be used to predict absorption rates
for other supergravity modes. It turns out that this can be done for several other
minimal scalars, namely the dilaton and the components of the NS B field along
the T
4
. There are four other minimal scalars in this background, the couplings of
which cannot be obtained from the DBI action. Perhaps more significantly, there
is a set of other scalar fields that do not obey minimal Klein-Gordon equations—
e.g. the fluctuations of the volume of the T
4
. If we trust the DBI action to obtain
the couplings of these fields, we get answers that do not agree with supergravity
calculations [55]. These cross sections are suppressed at low energies and it is
probably not surprising that the simplest low-energy effective theory does not
work.
6.3 Why Does It Work?
The agreement of the low-energy limit of the absorption cross section—or, equiv-
alently, decay rate—is strong evidence for the contention that Hawking radiation
is an ordinary, unitary quantum decay process. This is, however, surprising—
like the agreement of near-extremal thermodynamics—since our weak coupling
calculation agreed with a strong coupling expectation. The agreement of non-
trivial grey-body factors is even more surprising because now not just a single
number but an entire function of energy is correctly predicted by a weak coupling
calculation.
Weak coupling calculations also gave the correct answer for the extremal en-
tropy, but that case involves BPS states and supersymmetry guarantees the agree-
ment. In nonextremal situations, supersymmetry is broken. However, for small
amounts of nonextremality, we have small excitations over a supersymmetric
background and we may still expect nonrenormalization theorems to ensure that
higher-loop effects vanish at low energies. A complete understanding of this is still
lacking. Explicit one-loop calculations [56] seem to support this scenario. There
is also some evidence for nonrenormalization theorems of the type required [57].
7 ABSORPTION BY THREE-BRANES
The above discussion has addressed extremal solutions with large horizon areas
and their excitations. Extremal solutions with vanishing horizon areas generally
have singular horizons, for which semiclassical results cannot be trusted. There
are, however, several exceptions to this rule. In this section we consider one of
them—the 3-brane in Type IIB supergravity.
7.1 Classical Solution and Classical Absorption
The classical solution for N parallel extremal 3-branes along (x
7
x
9
) is given
by the (Einstein-frame) metric
ds
2
= [f(r)]
−1/2
[−dt
2
+dx
2
] + [f(r)]
1/2
[dr
2
+ r
2
dΩ
2
5
], (84)
where r = (
¸
6
i=1
(x
i
)
2
)
1/2
is the radial coordinate in the transverse space, x =
(x
7
, , x
9
) are the coordinates on the brane, and dΩ
5
is the measure on a unit
40 Das & Mathur
5-sphere. The harmonic function f(r) is given by
f(r) = 1 +

R
r

4
R
4
= 4πg
s
Nl
4
s
. (85)
The dilaton is a constant and the only other nontrivial field is the 4-form gauge
field with a field strength
F
0789r
=
4R
4
g
s
r
5
[f(r)]
−2
. (86)
The horizon is at r = 0 and the horizon area is indeed zero. However, the
spacetime is completely nonsingular. In the near-horizon or near-brane region
r <R, the metric (Equation 86) becomes, in terms of coordinates z = R
2
/r,
ds
2
=

R
z

2
[−dt
2
+ dx
2
+ dz
2
] +R
2
dΩ
2
5
. (87)
This is the metric of the space AdS
5
S
5
, the first factor being composed of
(t, x, z) while the sphere S
5
has a constant radius R. The AdS
5
has a constant
negative curvature 1/R
2
, whereas the sphere S
5
has a constant positive curvature.
Thus, near the horizon the spacetime approaches a “throat geometry,” where the
length of the throat is along the AdS
5
coordinates and the cross sections are
5-spheres of constant radii R.
From Equation 85 it follows that this curvature is much smaller than the string
scale when g
s
N 1. This is thus the regime where semiclassical supergravity is
valid.
It is possible to calculate the classical absorption cross section of various su-
pergravity modes by the extremal 3-brane. For s-wave absorption of minimally
coupled scalars, the leading-order result is [58]
σ =
1
8
π
4
R
8
ω
3
=
1
32π
(κN)
2
ω
3
, (88)
where we have used Equation 85. The cross section of course vanishes at ω = 0,
since the horizon area vanishes. Also note that we are dealing with an extremal
solution that has zero temperature, so we have absorption but no Hawking ra-
diation. It turns out that in this case one can also analytically calculate the
low-energy absorption cross section of such scalars for arbitrary angular momen-
tum [59].
7.2 Absorption in the Microscopic Model
The microscopic theory of N 3-branes is a U(N) gauge theory in 3 + 1 dimen-
sions with 16 supersymmetries, or N = 4 supersymmetry. This is a well-studied
theory and is known to be conformally invariant with zero beta function. It
is the dimensional reduction of N = 1 supersymmetric Yang-Mills in ten di-
mensions. As we will see below, a natural symmetry argument determines the
coupling of some of the supergravity modes to the brane degrees of freedom.
However, if we consider modes like the dilaton and longitudinal traceless parts
of the 10-dimensional graviton, the coupling can be once again read off from the
equivalence principle. For example, the longitudinal graviton has to couple to the
energy-momentum tensor of the Yang-Mills theory. Thus, one can perturbatively
D-Branes and Black Holes 41
calculate the absorption cross section of such modes (which are minimal scalars
from the 7-dimensional point of view) along the lines of the calculation in the
5-dimensional black hole. Again, an exact agreement was found for s-waves [58],
and the agreement was found to extend to all angular momenta [59].
In this case, the reason for the agreement of the weak coupling calculation
with supergravity answers is much better understood. In particular, it is known
that the two-point function of the energy-momentum tensor of N = 4 Yang-Mills
theory is not renormalized, so that the lowest-order result is exact [60]. The
imaginary part of the two-point function in Euclidean space is in turn related
to the absorption cross section, which is thus also protected from higher-order
corrections. There are similar nonrenormalization theorems known for operators
coupling to other modes as well.
7.3 Nonextremal Thermodynamics
The extremal 3-brane of supergravity described above is of course a limit of a
general nonextremal solution with finite horizon area and finite temperature. It
is thus natural to expect that the microscopic description of this is in terms of a
Yang-Mills theory at finite temperature. The thermodynamics of this Yang-Mills
theory was calculated [61] at lowest order in the coupling constant (i.e. a free gas),
and it was found that although the dependence of the thermodynamic quantities
on the energy, volume, and N is the same as in semiclassical thermodynamics of
the corresponding black brane, the precise coefficient differs by a factor of 4/3. We
know of no reason why the thermodynamic quantities are not renormalized as we
go from weak to strong coupling, so the discrepancy is not unexpected. However,
it is significant that the dependences agree—although the dependence on the
energy and the volume is dictated by scaling arguments (this is a conformally
invariant theory), the dependence on N is not. At weak coupling, we know that
there are N
2
degrees of freedom; there is no known argument why this should
continue at strong coupling.
8 AdS/CFT CORRESPONDENCE AND HOLOGRAPHY
We have presented a unitary description of black hole evaporation in terms of the
emission of closed strings from the branes. This description was at weak coupling;
in order to describe absorption by the black hole of quanta with arbitrary energies,
one must imagine that it is continued to strong coupling.
Let us see what form this continuation to higher energies might take. In the
lowest-order calculation, there was only one string interaction, where two open
strings joined up to the emitted closed string. At higher orders in the coupling,
we expect the emitted string to interact several times with the branes before
escaping to infinity as radiation. Let us consider for simplicity the case of the
D3-branes described above. Each such interaction gives one “hole” on the string
world sheet, with the boundary of the hole being constrained to the branes.
From the viewpoint of the dynamics of open strings on the 3-branes, each hole
looks like an open string loop. Thus, as we increase the coupling, we encounter
higher loop processes among the open strings, in the process of emitting a closed
string.
On the other hand, a loop of open strings can be interpreted as a propagation
of a closed string at tree level (see Figure 4). Thus, these open string multiloop
42 Das & Mathur
Figure 4: Open string loop as a closed string exchange. The arrows indicate the
direction of time in each way of looking at the diagram.
processes might be reinterpreted as closed string exchanges between the departing
quantum and the branes. But among such closed string exchange must be the
effect of the gravitational effect of the branes, which, crudely speaking, would
attempt to retard the outgoing quantum, redshifting it to lower energies.
Such a redshift is easily seen if we consider the propagation of the outgoing
quantum in the metric produced by the branes. But now we ask: As we increase
the coupling in the microscopic calculation, should we take into account the
complicated loop processes of the gauge theory (which are significant when the
open string g
o
coupling is not small), as well as the fact that the metric around
the branes will cease to be flat? (This curvature is significant when the closed
string coupling g
c
∼ g
2
o
ceases to be small.) The discussion above indicates that
taking both of these effects into account might amount to overcounting. The
effect of open string loops might well be counted in the propagation of closed
string modes (which include gravitons that can be said to condense to generate
the curvature).
It is hard to answer this question from a direct consideration of loop expansions,
since in the domain of large coupling we are unlikely to have any good expansion
in the number of loops. Thus, the above considerations are somewhat heuristic.
The situation is reminiscent, though, of the appearance of duality in tree-level
string scattering. The t-channel exchange has a sequence of poles, as does the
s-channel scattering, and we confront the issue of whether these are two different
effects to be added together in some way or just dual manifestations of a common
underlying dynamics. In the case of this string scattering, of course, it is now well
understood that the s-channel poles result from an infinite sum over t-channel
poles and are not to be considered as additional contributions to the S-matrix.
8.1 The Maldacena Conjecture
In the much more complicated case of the black hole evaporation process, Mal-
dacena’s bold postulate addressed the relation between microscopic theory of
the branes and the effects of gravity in 10-dimensional spacetime. Maldacena
considered the following limit.
Place together a large number N 1 of D3-branes. Let the string coupling be
weak (g <1), but take the limits of large N and small g in such a way that gN
tends to a finite value. Now consider energies of excitation that are low, so that on
the D3-branes we get open strings but only in their lowest states. The dynamics
of these open strings is, as we saw above, given by a (^ = 4) supersymmetric
Yang-Mills theory. The quantity gN ∼ g
2
YM
N is the ’t Hooft coupling of the
gauge theory. At small ’t Hooft coupling, we can study the large-N gauge theory
D-Branes and Black Holes 43
perturbatively; at large ’t Hooft coupling g
2
YM
N 1 such a perturbative analysis
is not possible.
But for gN 1 we are in a regime where the D3-branes produce an appreciable
gravitational field around themselves, and in fact distort the spacetime into the
form of Equation 84. In particular, as shown above, as we approach the horizon
the geometry becomes the throat geometry AdS
5
S
5
given by Equation 87.
Maldacena postulated that we can consider either (a) the gauge theory on the
branes (which is given by the open string interactions at a perturbative level)
or (b) no branes, but the space AdS
5
S
5
and closed string theory (which has
gravity as a low-energy limit) on this curved manifold. These two theories were
dual descriptions of exactly the same underlying theory.
If the gauge theory of the branes and gravity on the 10-dimensional spacetime
are indeed exactly dual descriptions of the same theory, then we must have a way
of computing the same quantity in two dual ways, and thus to see manifestly the
consequences of this duality. To do this, we need an operational definition of the
dual map, which was provided by Gubser et al [62] and Witten [63]. Let us rotate
the signature of the spacetime to a Euclidean one. In this case, the space spanned
by x, r, which is now Euclidean AdS
5
, has r = 0 as a regular point in the interior
of the space. Thus, there is no singularity that might represent the place where
the D3-branes were placed in a gauge theory description. But this space has a
boundary at r → ∞, which has the topology of S
4
. The conjecture now states
that string theory on this smooth AdS
5
S
5
is dual to ^ = 4 supersymmetric
SU(N) Yang-Mills on the 4-dimensional boundary of the AdS space.
Note that in this formulation of the conjecture, all reference to D3-branes has
disappeared. In fact, a geometry that is AdS
5
S
5
everywhere (and not just at
small r) is an exact solution of string theory without any branes present. We take
the 5-form field strength F present in Type IIB string theory to have a constant
value with integral N on the S
5
; since this field must be self-dual, it also has a
constant value on the AdS
5
. The energy density of this field yields the equal and
opposite curvatures of the S
5
and the AdS
5
spaces; the radius of curvature in
each case is R ∼ (gN)
1/4
l
s
(l
s
is the string length). If we had taken a collection
of D3-branes as above, then the flux of F on the sphere surrounding the branes
would have been N, but only the near-brane geometry would have resembled
AdS
5
S
5
—at large r the space becomes flat. Thus, it seems better to formulate
the duality without reference to any D-branes, although, as we discuss below, this
makes the issues related to black hole information loss somewhat more difficult
to access.
An important fact that led Maldacena to the duality conjecture was the agree-
ment of global symmetries between the gravity theory and the gauge theory.
AdS
5
is a maximally symmetric space with a 15-parameter isometry group. The
gauge theory is in four dimensions, so one may expect a 10-dimensional set of
symmetries, which are the translations and rotations on S
4
. But the ^ = 4
supersymmetric SU(N) Yang-Mills theory is a conformal theory, and the con-
formal group in four dimensions is indeed isomorphic to the 15-parameter group
of isometries of AdS
5
space. Further, the ^ = 4 gauge theory has four super-
charges forming, having an SU(4) symmetry of rotations among themselves. The
gravity theory has an internal space that is S
5
. Thus, when we decompose fields
in harmonics on this space in the process of Kaluza-Klein reduction, the multi-
plets obtained fall into representations of SO(6). But SO(6) ≈ SU(4), so both
the theories have the same R-symmetry group. Further, because of conformal
44 Das & Mathur
invariance, the gauge theory has 16 conformal supercharges in addition to its 16
regular supercharges; this agrees with the 32 supercharges of the gravity theory
on AdS
5
S
5
.
8.2 Calculations Using the Conjecture
Let us return to the issue of how to compare quantities in the two dual theories.
Let Ω be the AdS space, and ∂Ω its boundary. Let us consider for simplicity a
scalar field, the dilaton φ, present in the string theory. If we specify the value
φ
b
of φ on the boundary ∂Ω, then we can solve for φ in the interior of Ω and
compute the action S
SUGRA
of supergravity on this solution. The field φ, on the
other hand, will be dual to an operator in the Yang-Mills theory, and in this case
symmetries fix this operator to be TrF
2
. The duality relation is
e
−S

SUGRA

b
]
=

∂Ω
DA e
−S
YM
+

dξφ
b
(ξ)TrF
2
(ξ)

∂Ω
DA e
−S
YM
. (89)
Here

DA represents the path integral over the gauge theory variables. The
action on the left-hand side is a string action in general, but at the leading order
it can be replaced by the supergravity action.
Using the above relation, it is possible in principle to compute n-point corre-
lation functions from supergravity and compare the results to calculations in the
gauge theory. The gravity theory is simple (perturbative classical supergravity to
lowest order) when N is large (loops of gravity are suppressed by 1/N
2
) and gN
is large (because then the AdS space has a large radius, and stringy corrections,
which depend on l
s
/R, can be ignored to leading order). Thus it is difficult to
compare results on the gravity side with explicit computations in the Yang-Mills
theory, which has a well-known treatment of the large-N limit but no simple way
to handle a regime with g
2
Y M
N ∼ gN 1. However, supersymmetry protects
the values of some quantities from changing when the coupling is varied, and
here agreement is indeed found. Thus, anomalous dimensions of chiral operators
[62, 63] and three-point functions of chiral operators [65] are found to have the
same value when computed from the gauge theory or from the gravity dual using
Equation 89. For quantities that are not protected in this fashion, the gravity
calculation gives a prediction for the strongly coupled gauge theory. Examples of
such predictions are the potential between external quarks placed in the gauge
theory [66], four-point correlation functions, and anomalous dimensions of com-
posite operators (which can be deduced from the four-point functions) [67]. If one
considers the gauge theory at a finite temperature by compactifying the “time”
direction, then one obtains a nonsupersymmetric theory with a discrete spectrum
of glueball states [68]. In this case, the strong coupling calculation has no direct
physical meaning, and only a weak coupling calculation would give the true glue-
ball masses (the coupling flows with scale after supersymmetry is broken; with
supersymmetry intact, it does not, and each value of the coupling gives a dif-
ferent well-defined theory). Nevertheless, there is a surprising agreement of the
qualitative features of the glueball spectrum between the gravity calculation and
lattice simulations [69].
D-Branes and Black Holes 45
8.3 Holography and the Bekenstein Entropy Bound
The AdS/CFT correspondence has been studied extensively over the past two
years [see [70] for review]. This section explains how this correspondence illus-
trates another important property of holography that is in fact its very essence:
the Bekenstein bound [71]. A general argument states that in any theory of grav-
ity, the total entropy S of anything in a large box of volume V and surface area
A is bounded by
S ≤
A
G
, (90)
where G is Newton’s constant. Numerical constants are ignored in this relation.
The form of this bound appears surprising, since in the absence of gravity we know
that entropy is an extensive quantity and should scale as V . However, the result
follows from the existence of black holes. Suppose we start with a state whose
entropy is bigger than this bound, but whose energy is smaller than the mass
of a black hole that fills the entire box. Then we slowly add matter, increasing
the energy. At some stage, gravitational collapse takes over and a black hole is
formed, which then continues to grow until it fills up the box. We know that the
entropy of a black hole is proportional to the area of the horizon measured in units
of Newton’s constant. However, for this black hole, the area of the horizon is the
same as the area of the box A, i.e. the right-hand side of Equation 90. Thus, we
have managed to decrease the entropy in this process and hence have violated the
second law. ’t Hooft [72] and Susskind [73] have proposed a radical interpretation
of the holographic bound (Equation 90), namely, a (d +1)-dimensional theory of
gravity should be equivalent to a d-dimensional theory that lives on the boundary
of space, and this d-dimensional theory must have one degree of freedom per
Planck area. If this is true, the Bekenstein bound naturally follows.
We have seen that supergravity in AdS
5
S
5
is dual to a strongly coupled
large-N Yang-Mills theory in 3 +1 dimensions. Is it then true that the latter has
one degree of freedom per Planck area? This is indeed true [74]. In the AdS
5
S
5
metric (Equation 87) the boundary is at z = 0 and continuum Yang-Mills theory
living on this boundary is dual to supergravity in the whole space. But suppose
we impose an infrared cutoff at z = z
0
. Then it turns out that supergravity in
the region z > z
0
is dual to a Yang-Mills theory with a position space ultraviolet
cutoff δ = z
0
. In fact, renormalization group flows of the gauge theory become
motion in the fifth dimension.
Consider a large patch of this boundary of coordinate size L (in terms of the x
coordinates). It follows from Equation 87 that the physical area of this boundary
is
A ∼
R
3
z
3
0
L
3
. (91)
Using the expression for R in Equation 85 and the fact that the 5-dimensional
Newton’s constant G
5
is given by G
5
∼ G
10
/R
5
∼ (g
2
l
8
s
)/R
5
, we see that
A
G
5
∼ N
2
L
3
z
3
0
. (92)
Now the Yang-Mills theory has N
2
degrees of freedom per unit cell. This is man-
ifest at weak coupling, but the result for the entropy of near-extremal 3-branes
shows that this must also be true in strong coupling. Furthermore, the cutoff of
46 Das & Mathur
this theory is precisely z
0
. Thus, the right-hand side of Equation 92 is indeed the
number of degrees of freedom in this cutoff Yang-Mills theory. This shows that
the gauge theory–supergravity duality provides a concrete demonstration of the
Bekenstein bound.
8.4 Near-Horizon Limit of 5D Black Hole
Another example of the holographic connection is provided by the near-horizon
geometry of the D1-D5 system. This spacetime is AdS
3
S
3
M
4
[75], where M
4
is either T
4
or K
3
on which the five branes are wrapped. The AdS
3
is made up
of time, the transverse radial coordinate r, and x
5
. The boundary theory is once
again a superconformal field theory. However, this is no longer a gauge theory
but rather the low-energy theory of the D1-D5 system discussed in the previous
sections. Once again, the symmetries of the bulk agree with the symmetries of the
conformal field theory. Furthermore, 5D black holes obtained by putting in some
momentum now become the Banados-Teitelboim-Zanelli (BTZ) black holes of 3-
dimensional gravity. In fact, recently the holographic connection has been used
to understand the properties of the D1-D5 system. Details of this development
are supplied elsewhere [70].
8.5 A Suggestive Model of Holography
Unfortunately, despite these agreements between the two dual descriptions, we
do not yet understand the explicit map that relates the gauge theory variables
to the gravity variables. There is, in fact, an earlier example of holography:
(1 + 1)-dimensional string theory. This theory has a nonperturbative definition
in terms of quantum mechanics of N N matrices. One starts with the action
S =

dt[
˙
M
2
− V (M)], noting that when the matrix is large, there is a critical
form for the potential in which successive terms in the loop expansion are com-
parable. The theory can be written in terms of ρ(x, t), the density of eigenvalues
of the matrix [76]. At the critical point, the extra dimension x (which is the
space of eigenvalues) becomes continuous and the collective field ρ(x, t) becomes
a smooth field in 1 + 1 dimensions. Remarkably, in terms of appropriate vari-
ables, these two dimensions (which emerged in totally different ways) exhibit a
local Lorentz invariance [77], though in the presence of a background that breaks
the full symmetry. In fact, at the critical point the loop diagrams of the matrix
model generate discretizations of an oriented world sheet with dynamical world
sheet metric and one scalar field. The usual quantization of noncritical string
theory then converts the scale on the world sheet to an extra dimension in target
space [78], resulting in a (1 + 1)-dimensional field theory. Furthermore, renor-
malization group flow on the world sheet is related to motion in this extra target
space direction [79].
It would be interesting if the gauge theory in 3 + 1 dimensions acquired extra
dimensions in a similar manner to map onto string theory in 10 dimensions. So
far, such an approach has not succeeded, despite several attempts.
9 DISCUSSION
The progress of string theory over the past few years has been truly remarkable.
Consistent quantization of the string forced us to consider a 10-dimensional su-
D-Branes and Black Holes 47
persymmetric theory. Within such a theory, one discovers branes as solitonic
states. Regarding such solitons as elementary quanta turns out to be a duality
transformation, and such duality maps unify all the five perturbative quantiza-
tions of the string, as well as 11-dimensional supergravity. There is no room for
any free parameters in the theory, although many solutions of the field equations
appear equally acceptable at first sight, and we do not yet know which to choose
to describe the low-energy physics of the world that we see around us.
The results on black holes have provided a striking validation of this entire
structure. The entropy of black holes can be deduced from robust thermody-
namic arguments, using just the classical properties of the black hole metric
and Gedanken experiments with low-density quantum matter. A correct quan-
tum theory of gravity should reproduce this entropy by a count of allowed mi-
crostates, and string theory yields the correct microscopic entropy of extremal
and near-extremal holes (the only cases in which we know how to compute with
some confidence). Further, interaction with these microscopic degrees of freedom
yields a unitary description of black hole absorption and Hawking emission for
low-energy quanta from near-extremal holes (again this is the case in which we
may compute with some confidence), thus suggesting that black holes can be
unified into physics without having to modify basic principles of quantum me-
chanics or statistical physics. In no other route that has been attempted for
quantizing gravity has it been possible to arrive at a value for black hole entropy;
the power of supersymmetry and our understanding of the nonperturbative fea-
tures of string theory have allowed us to extract more than just the perturbative
physics of graviton scattering.
What about the information issue in black hole evaporation? The derivation of
unitary low-energy Hawking radiation from microstates certainly gives a strong
indication on the information question, i.e. that information in the matter falling
into the hole is not lost but is recovered in the radiation that emerges. But it
is fair to say that we do not yet have a good understanding of the information
paradox. The paradox presents us with one way of computing Hawking radiation,
via the foliation of spacetime depicted in Figure 1, and this radiation process
appears to be nonunitary. The paradox then challenges us to point out which
of the assumptions that went into the calculations were wrong. It now appears
reasonable that there are nonlocalities in quantum gravity that can extend beyond
Planck scale. Indeed holography, and in particular the Maldacena conjecture,
requires that the degrees of freedom in a volume be encoded by degrees of freedom
at the surface of that volume, so we must give up naive ways of thinking about
locality. But we do not yet understand in detail the mechanism of this nonlocality,
and in particular we lack an explicit description in string theory of the foliation
in Figure 1.
These gaps in our understanding are probably linked to our lack of an adequate
language to understand the variables of string theory in a way that is not based on
perturbation around a given background. In recent years Matrix theory [80] has
attempted such a formulation, but our understanding is still somewhat primitive.
It thus appears that most of the progress in understanding quantum gravity and
the structure of spacetime is yet to come, and string theory should be an exciting
field in the years ahead.
Literature Cited
48 Das & Mathur
1. Misner C, Thorne W, Wheeler J. Gravitation. New York: Freeman (1978); Wald R. General
Relativity. Chicago: Univ. Chicago Press (1984)
2. Bekenstein JD. Phys. Rev. D 7:2333 (1973); Bekenstein JD. Phys. Rev. D 9:3292 (1974)
3. Pelath MA, Wald RM. gr-qc/9901032 (1999)
4. Hawking S. Commun. Math. Phys. 43:199 (1975)
5. Hawking S. Phys. Rev. D 13:191 (1976)
6. Hawking S. Phys. Rev. D 14:2460 (1976)
7. Giddings SB. hep-th/9508151 (1995)
8. Zurek WH. Phys. Rev. Lett. 49:1683 (1982)
9. Banks T, Peskin M, Susskind L. Nucl. Phys. B244:135 (1984)
10. Unruh WG, Wald RM. Phys. Rev. D 52:2176 (1995)
11. ’t Hooft G. Nucl. Phys. B256:727 (1985)
12. Keski-Vakkuri E, Lifschytz G, Mathur SD, Ortiz M. Phys. Rev. D 51:1764 (1995)
13. Keski-Vakkuri E, Mathur SD. Phys. Rev. D 54:7391 (1996)
14. Green M, Schwarz J, Witten E. Superstring Theory, Vols. 1, 2. Cambridge, UK: Cambridge
Univ. Press (1987); Polchinski J. String Theory, Vols. 1, 2. Cambridge, UK: Cambridge
Univ. Press (1998)
15. Arkani-Hamed N, Dimopolous S, Dvali G. Phys. Rev. D 59:086004 (1999); Randall L, Sun-
drum R. Phys. Rev. Lett. 83:3370 (1999)
16. Sen A. hep-th/9802051 (1998)
17. Townsend P. hep-th/9507048 (1995)
18. Polchinski J. Phys. Rev. Lett. 75:4724 (1995)
19. Polchinski J. hep-th/9611050 (1996)
20. Sen A. hep-th/9904027 (1999)
21. Witten E. Nucl. Phys. B460:335 (1996)
22. Tseytlin A. hep-th/9908105 (1999)
23. Montonen C, Olive D. Phys. Lett. B72:117 (1977)
24. Susskind L. hep-th/9309145 (1993)
25. Dabholkar A, Harvey J. Phys. Rev. Lett. 63:478 (1989); Dabholkar A, Gibbons G, Harvey
J, Ruiz F. Nucl. Phys. B340:33 (1990)
26. Horowitz G, Maldacena J, Strominger A. Phys. Lett. B383:151 (1996)
27. Sen A. Nucl. Phys. B440:421 (1995); Sen A. Mod. Phys. Lett. A10:2081 (1995)
28. Strominger A, Vafa C. Phys. Lett. B379:99 (1996)
29. Das SR, Mathur SD. Phys. Lett. B375:103 (1996)
30. Maldacena J, Susskind L. Nucl. Phys. B475:679 (1996)
31. Callan C, Maldacena J. Nucl. Phys. B472:591 (1996)
32. Horowitz G, Strominger A. Phys. Rev. Lett. 77:2368 (1996)
33. Das SR, Mathur SD. Nucl. Phys. B478:561 (1996); Das SR, Mathur SD. Nucl. Phys.
B482:153 (1996)
34. Preskill J et al. Mod. Phys. Lett. A6:2353 (1991)
35. Dijkgraaf R, Moore G, Verlinde E, Verlinde H. Commun. Math. Phys. 185:197 (1997)
36. Maldacena J, Moore G, Strominger A. hep-th/9903063 (1999)
37. Maldacena J. PhD thesis. hep-th/9607235 (1996)
38. Hassan SF, Wadia SR. Phys. Lett. B402:43 (1997); Hassan SF, Wadia SR. Nucl. Phys.
B526:311 (1998)
39. Douglas M. hep-th/9512077 (1995)
40. Dijkgraaf R, Verlinde E, Verlinde H. Nucl. Phys. B486:77 (1997)
41. Dijkgraaf R. Nucl. Phys B543:545 (1999)
42. Seiberg N, Witten E. J. High Energy Phys. 9904:017 (1999)
43. Larsen F, Martinec E. J. High Energy Phys. 9906:019 (1999); Larsen F, Martinec E. J. High
Energy Phys. 9911:002 (1999); Dhar A, Mandal G, Wadia SR, Yogendran K. hep-th/9910194
(1999)
44. Cvetic M. Nucl. Phys. Proc. Suppl. 56B:1 (1997); Youm D. Phys. Rep. 316:1 (1999)
45. Maldacena J, Strominger A. Phys. Rev. Lett. 77:428 (1996); Horowitz G, Lowe D, Maldacena
J. Phys. Rev. Lett. 77:430 (1996)
46. Balasubramanium V, Larsen F. Nucl. Phys. B478:199 (1996)
47. Tseytlin A. Nucl. Phys. B475:149 (1996); Klebanov I, Tseytlin A. Nucl. Phys. B475:179
(1996); Gauntlett J, Kastor D, Traschen J. Nucl. Phys. B478:544 (1996)
48. Brekenridge J, Myers R, Peet A, Vafa C. Phys. Lett. B391:93 (1996); Brekenridge J, et al.
D-Branes and Black Holes 49
Phys. Lett. B381:423 (1996)
49. Page D. Phys. Rev. D 13:198 (1976); Unruh W. Phys. Rev. D 14:3251 (1976)
50. Dhar A, Mandal G, Wadia SR. Phys. Lett. B388:51 (1996)
51. Das SR, Gibbons G, Mathur SD. Phys. Rev. Lett 78:417 (1997)
52. Maldacena J, Strominger A. Phys. Rev. D 55:861 (1997)
53. Gubser S, Klebanov I. Nucl. Phys. B482:173 (1996); Gubser S, Klebanov I. Phys. Rev. Lett.
77:4491 (1996)
54. Mathur SD. Nucl. Phys. B514:204 (1998)
55. Callan CG, Gubser S, Klebanov I, Tseytlin A. Nucl. Phys B489:65 (1997); Klebanov I,
Krasnitz M. Phys. Rev. D 55:3250 (1997); Krasnitz M, Klebanov I. Phys. Rev. D 56:2173
(1997)
56. Das SR. Phys. Rev. D 56:3582 (1997); Das SR. Nucl. Phys. Proc. Suppl. 68:119 (1998)
57. Maldacena J. Phys. Rev. D 55:7645 (1997)
58. Klebanov I. Nucl. Phys. B496:231 (1997); Gubser S, Klebanov I, Tseytlin A. Nucl. Phys.
B499:217 (1997)
59. Klebanov I, Taylor W. Nucl. Phys. B560:207 (1999)
60. Gubser S, Klebanov I. Phys. Lett. B413:41 (1997)
61. Gubser S, Peet A, Klebanov I. Phys. Rev. D 54:3915 (1996)
62. Gubser S, Klebanov I, Polyakov A. Phys. Lett. B428:105 (1998)
63. Witten E. Adv. Theor. Math. Phys. 2:253 (1998)
64. Maldacena J. Adv. Theor. Math. Phys. 2:231 (1998)
65. Freedman D, Mathur SD, Rastelli L, Matusis A. Nucl. Phys. B546:96 (1999); Lee S, Minwalla
S, Rangamani M, Seiberg N. Adv. Theor. Math. Phys. 2:697 (1998)
66. Maldacena J. Phys. Rev. Lett. 80:4859 (1998), S.J. Rey and J. Yee, hep-th/9803001.
67. d’Hoker E, et al. Nucl. Phys. B562:353 (1999); d’Hoker E, et al. hep-th/9911222 (1999)
68. Witten E. Adv. Theor. Math. Phys. 2:505 (1998)
69. Csaki C, Ooguri H, Oz Y, Terning J. JHEP 9901:017 (1999)
70. Aharony O, et al. hep-th/9905111 (1999)
71. Bekenstein J. Phys. Rev. D 27:2262 (1983); Bekenstein J. Phys. Rev. D 49:1912 (1994)
72. ’t Hooft G. hep-th/9310026 (1993); Stephens CR, ’t Hooft G, Whiting B. Class. Quant.
Grav. 11:621 (1994)
73. Susskind L. J. Math. Phys. 36:6377 (1995)
74. Susskind L, Witten E. hep-th/9805114 (1998)
75. Maldacena J, Strominger A. JHEP 9812:005 (1998)
76. Jevicki A, Sakita B. Nucl. Phys. B165:511 (1980)
77. Das SR, Jevicki A. Mod. Phys. Lett. A5:1639 (1990)
78. Das SR, Naik S, Wadia SR. Mod. Phys. Lett. A4:1033 (1989)
79. Das SR, Dhar A, Wadia SR. Mod. Phys. Lett. A5:799 (1990)
80. Banks T, Fischler W, Shenker SH, Susskind S. Phys. Rev. D 55:5112 (1997)

Extremal and Near-Extremal Limits . . . . . . . . . . . Microscopic Model for the Five-Dimensional Black Hole The Long String and Near-Extremal Entropy . . . . . . A More Rigorous Treatment for Extremal Black Holes . The Gauge Theory Picture . . . . . . . . . . . . . . . . . Other Extremal Black Holes . . . . . . . . . . . . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

23 24 25 28 31 32

BLACK HOLE ABSORPTION/DECAY AND D-BRANES . . . . . . . . . . . . .
Classical Absorption and grey-body Factors . . . . . . . . . . . . . . . . . . . . . . . . D-brane Decay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Why Does It Work? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

33
33 35 39

ABSORPTION BY THREE-BRANES . . . . . . . . . . . . . . . . . . . . . . . . .
Classical Solution and Classical Absorption . . . . . . . . . . . . . . . . . . . . . . . . Absorption in the Microscopic Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . Nonextremal Thermodynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

39
39 40 41

AdS/CFT CORRESPONDENCE AND HOLOGRAPHY . . . . . . . . . . . . . .
The Maldacena Conjecture . . . . . . . . . . . . Calculations Using the Conjecture . . . . . . . Holography and the Bekenstein Entropy Bound Near-Horizon Limit of 5D Black Hole . . . . . A Suggestive Model of Holography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

41
42 44 45 46 46

DISCUSSION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

46

1

INTRODUCTION

Black holes present us with a very deep paradox. The path to resolving this paradox may well be the path to a consistent unified theory of matter and quantized gravity. In classical gravity, a black hole is a classical solution of the equations of motion such that there is a region of spacetime that is causally disconnected from asymptotic infinity (see e.g. Reference [1]). The boundary of such a region is called the event horizon. Consider a large collection of low-density matter, in an asymptotically flat spacetime. For simplicity, we take the starting configuration to be spherically symmetric and nonrotating (these restrictions do not affect the nature of the paradox that emerges). This ball of matter will collapse toward smaller radii under its self-gravitation. At some point, the matter will pass through a critical radius, the Schwarzschild radius Rs , after which its further collapse cannot be halted, whatever the equation of state. The final result, in classical general relativity, is that the matter ends up in an infinite-density singular point, while the metric settles down to the Schwarzschild form ds2 = −(1 − 2GN M 2 2GN M −1 2 )dt + (1 − ) dr + r2dΩ2 . 2 rc rc2 (1)

Here GN is Newton’s constant of gravity, and c is the speed of light. The horizon radius of this hole is 2GN M → 2M , (2) Rs = c2 where the last expression arises after we set GN = 1, c = 1. (In what follows, we adopt these units unless otherwise explicitly indicated; we also set ¯ = 1.) h 2

D-Branes and Black Holes

3

Classically, nothing can emerge from inside the horizon to the outside. A test mass m has effective energy zero if it is placed at the horizon; it has rest energy mc2, but a negative gravitational potential energy exactly balances this positive contribution. For a rough estimate of the horizon size, we may put this negative energy to be the Newtonian value −GN M m/r, for which Rs ∼ GN M/c2 . It may appear from the above that the gravitational fields at the horizon of a black hole are very large. This is not true. For a neutral black hole of mass M , the magnitude of the curvature invariants, which are the measure of local gravitational forces, is given by |R| ∼ GN M . r3 (3)

Thus, at the horizon r = rH = 2GN M , the curvature scales as 1/M 2 . As a result, −1/2 for black holes with masses M GN , the curvatures are very small and the spacetime is locally rather close to flat spacetime. In fact, an object falling into a black hole will not experience any strong force as it crosses the horizon. However, an asymptotic observer watching this object will see that it takes an infinite time to reach the horizon. This is because there is an infinite gravitational red-shift between the horizon and the asymptotic region. An important point about black hole formation is that one does not need to crush matter to high densities to form a black hole. In fact, if the hole has mass M , the order of magnitude of the density required of the matter is ρ∼ 1 M ∼ 2. 3 Rs M (4)

Thus, a black hole of the kind believed to exist at the center of our galaxy (108 solar masses) could form from a ball with the density of water. In fact, given any density we choose, we can make a black hole if we take a sufficient total mass with that density. This fact makes it very hard to imagine a theory in which black holes do not form at all because of some feature of the interaction between the matter particles. As a consequence, if black holes lead to a paradox, it is hard to bypass the paradox by doing away with black holes in the theory. It is now fairly widely believed that black holes exist in nature. Solar-mass black holes can be endpoints of stellar evolution, and supermassive black holes (∼ 105 − 109 solar masses) probably exist at the centers of galaxies. In some situations, these holes accrete matter from their surroundings, and the collisions among these infalling particles create very powerful sources of radiation that are believed to be the source of the high-energy output of quasars. In this article, however, we are not concerned with any of these astrophysical issues. We concentrate instead on the quantum properties of isolated black holes, with a view toward understanding the problems that arise as issues of principle when quantum mechanical ideas are put in the context of black holes. For example, the Hawking radiation process discussed below is a quantum process that is much weaker than the radiation from the infalling matter mentioned above, and it would be almost impossible to measure even by itself. (The one possible exception is the Hawking radiation at the last stage of quantum evaporation. This radiation emerges in a sharp burst with a universal profile, and there are experiments under way to look for such radiation from very small primordial black holes.)

4

Das & Mathur

1.1

The Entropy Problem

Already, at this stage, one finds what may be called the entropy problem. One of the most time-honored laws in physics has been the second law of thermodynamics, which states that the entropy of matter in the Universe cannot decrease. But with a black hole present in the Universe, one can imagine the following process. A box containing some gas, which has a certain entropy, is dropped into a large black hole. The metric of the black hole then soon settles down to the Schwarzschild form above, though with a larger value for M , the black hole mass. The entropy of the gas has vanished from view, so that if we only count the entropy that we can explicitly see, then the second law of thermodynamics has been violated! This violation of the second law can be avoided if one associates an entropy to the black hole itself. Starting with the work of Bekenstein [2], we now know that if we associate an entropy AH (5) SBH = 4GN with the black hole of horizon area AH , then in any Gedanken experiment in which we try to lose entropy down the hole, the increase in the black hole’s attributed entropy is such that d (Smatter + SBH ) ≥ 0 dt (for an analysis of such Gedanken experiments, see e.g. [3]). Furthermore, an “area theorem” in general relativity states that in any classical process, the total area of all black holes cannot decrease. This statement is rather reminiscent of the statement of the second law of thermodynamics—the entropy of the entire Universe can never decrease. Thus the proposal (Equation 5) would appear to be a nice one, but now we encounter the following problem. We would also like to believe on general grounds that thermodynamics can be understood in terms of statistical mechanics; in particular, the entropy S of any system is given by S = log Ω, (6) where Ω denotes the number of states of the system for a given value of the macroscopic parameters. For a black hole of one solar mass, this implies that 78 there should be 1010 states! But the metric (Equation 1) of the hole suggests a unique state for the geometry of the configuration. If one tries to consider small fluctuations around this metric, or adds in, say, a scalar field in the vicinity of the horizon, then the extra fields soon flow off to infinity or fall into the hole, and the metric again settles down to the form of Equation 1. If the black hole has a unique state, then the entropy should be ln 1 = 0, which is not what we expected from Equation 5. The idea that the black hole configuration is uniquely determined by its mass (and any other conserved charges) arose from studies of many simple examples of the matter fields. This idea of uniqueness was encoded in the statement “black holes have no hair.” (This statement is not strictly true when more general matter fields are considered.) It is a very interesting and precise requirement on the theory of quantum gravity plus matter that there be indeed just the number (Equation 5) of microstates corresponding to a given classical geometry of a black hole.

the above statement implies that if the hole can absorb photons at a given wave number k with absorption cross section σ(k).3 The Information Problem “Hawking radiation” is produced from the quantum fluctuations of the matter vacuum. in the presence of the gravitational field of the hole. in a black hole background. there are black hole solutions that carry charge and angular momentum.2 Hawking Radiation If black holes have an entropy SBH and an energy equal to the mass M . one member of this pair can fall into the hole.) Classically. in 1974. The vacuum for the matter fields has fluctuations.D-Branes and Black Holes 5 1. However. such radiation is possible. (7) For a neutral black hole in four spacetime dimensions. Although we have so far discussed the simplest black holes. while the other member of the pair can escape to infinity as real positive energy radiation [4]. the emission rate is given by the absorption cross section multiplied by a standard thermal factor (this factor would have a plus sign in place of the minus sign if we were considering fermion emission) and a phase space factor that counts the number of states in the wave number range k and k + dk. which gives 1 dS −1 ) = . Furthermore. then it must also radiate at the same wave number at the rate Γ(k) = e σ(k) h|k| ¯ kT dd k . the pair annihilates quickly in a time set by the uncertainty principle. which is the acceleration felt by a static object at the horizon as measured from the asymptotic region. AH = 4π(2GN M )2 . In normal spacetimes. For black holes of masses much larger than the scale set by Newton’s constant. (8) T =( dM 8πGN M Again assuming thermodynamical behavior. so it is tempting to say that no such radiation is possible. It is remarkable that the above discussed thermodynamic properties of black holes seem to be universal. d − 1 (2π) (9) In other words. the gravitational . The profile of this radiation is found to be thermal. with a temperature given by Equation 8. so that pairs of particles and antiparticles are produced and annihilated continuously. (d denotes the number of spatial dimensions. κ. then if thermodynamics were to be valid. we would expect them to have a temperature given by T dS = dE = dM. where it has a net negative energy. The precise relation—also universal—is κ . Hawking [4] found that if the quantum behavior of matter fields is considered. (10) T= 2π 1. The leading term in the entropy is in fact given by Equation 5 for all black holes of all kinds in any number of dimensions. the temperature is given in terms of another geometric quantity called the surface gravity at the horizon. We can also consider generalizations of general relativity to arbitrary numbers of spacetime dimensions (as will be required below) and further consider other matter fields in the theory. However. nothing can come out of the black hole horizon.

and yet the process would be in line with the unitarity of quantum mechanics. so that the outgoing radiation is highly “entangled” with whatever is left behind at the hole. advocated most strongly by Hawking. is that the hole does evaporate away to nothing. Further. through a unitary evolution operator. the final radiation state cannot have any significant information about the initial matter state. for a counterargument. This circumstance would contradict the assumption in usual quantum mechanics that the final state of any time evolution is related in a one-to-one and onto fashion to the initial state. which is a very large number (as we have seen above). and this contribution will diverge. so quantum gravity seems not to be required. since at this point quantum gravity would presumably have to be important. the remnant will have to have a very large number of possible states. But this entanglement entropy is somewhat larger [8] than the black hole entropy SBH . none of which are very palatable (for a survey see e. One might worry that in any quantum process. Some investigations of this possibility have suggested. where the particle pairs are produced in this simple picture.g. and the passage from an intial pure state to a final mixed state is a natural process in any theory of quantum gravity. and this number will grow to infinity as the mass of the initial hole is taken to infinity. The Hawking radiation from the black hole would not fundamentally differ from the radiation from a lump of burning coal—the information of the atomic structure of the coal is contained. that giving up the purity of quantum states causes difficulties with maintaining energy conservation in virtual processes ([9].6 Das & Mathur field near the horizon. If the above reasoning and computations are correct. is given quite accurately by the classical metric of Equation 1. and then such a divergence can be avoided. Thus. the calculation is insensitive to the precise details of the matter that went to make up the hole. the final state is in fact not even a normal quantum state. Thus. one can have loops of this remnant particle. The hole could then completely evaporate away. The outgoing member of a pair of particles created by the quantum fluctuation is in a mixed state with the member that falls into the hole. The hole may not completely evaporate away but leave a “remnant” of Planck size. though it is difficult to decipher. and the pure states of quantum mechanics that we are used to thinking about are only a special case of this more general kind of state. see Reference [10]). however. since the number of states of the remnant is infinite. if the hole completely evaporates away. The curvature invariants at the horizon are all very small compared with the Planck scale. then this final state is entangled with “nothing. The possibility that would best fit our experience of physics in general would be that the Hawking radiation does manage to carry out the information of the collapsing matter [11]. The semiclassical reasoning used in the derivation of Hawking radiation cannot say whether the hole continues to evaporate after it reaches Planck size. It is uncomfortable to have a theory in which a particle of bounded mass (Planck mass) can have an infinite number of configurations. In this view. in the radiation and . the natural description of states is in fact in terms of density matrices. Another possibility.” and we find that the resulting system is described not by a pure quantum state but by a mixed state. one confronts a set of alternatives. The radiation sent off to infinity will remain entangled with this remnant. Reference [7]). But it has been argued that remnants from holes of increasingly large mass might couple to any given process with correspondingly smaller strength. Worse. If the hole completely evaporates away.

Finally. We use a foliation of spacetime by the following kind of spacelike hypersurfaces. This requires that the spatial slices be smooth and that the embedding of neighboring slices changes in a way that is not too sharp. as long as the linearity of quantum mechanics holds. we choose a foliation of the spacetime by smooth spacelike hypersurfaces. Figure 1 shows the spacetime in a symbolic way. As we evolve along this foliation. we can connect these two parts of the hypersurface by a smooth interpolating region that is spacelike as well.4 Difficulties with Obtaining Unitarity Let us review briefly the difficulties with having the radiation carry out the information. other products that emerge when the coal burns away. we let the hypersurface be a t = t0 surface (this description uses the Schwarzschild coordinates of Equation 1). where the |ψi are a set of basis states.” The reason is as follows. a general state cannot be “duplicated” by any quantum process. let us choose r = M so that this part of the surface is neither close to the horizon (r = 2M ) nor close to the singularity (r = 0). Each of the spacelike hypersurfaces shown in |ψ1 + |ψ2 → (|ψ1 + |ψ2 ) × (|ψ1 + |ψ2 ). say for r > 4M . Then. → |ψ1 × |ψ1 + |ψ2 × |ψ2 . an r =constant surface is spacelike. 1. there can be no quantum “Xeroxing. we will find |ψ1 + |ψ2 and not Thus. This part of the hypersurface will extend from some time t = 0 near the formation of the hole to the value t = t0 . It is important to realize that the information in the collapsing matter cannot also be copied into the radiation—in other words. Inside the black hole. Away from the black hole. To study the evolution. Suppose the evolution process makes two copies of a state |ψI → |ψI × |ψi .D-Branes and Black Holes C B A t~M 3 7 t=0 r=0 r -> Figure 1: Foliation of the black hole spacetime. while we see the radiation collect at spatial infinity. we see the matter fall in toward the center of the hole.

spacetime was treated classically. By using an appropriate choice of this evolution. On the extreme right. which is entangled between components B and C. we have on the left the matter that fell in to make the hole. as mentioned above. But it is easy to see from the construction of surfaces described above that all the hypersurfaces in the evolution are smooth. however. then we will need “quantum xeroxing. In a general curved spacetime. or that it would need a sharp kink in one or more places. Let us call this matter “A. since this part of the surface is common to both hypersurfaces. that at t0 the black hole is nowhere near its endpoint of evaporation. cannot happen if we accept the principle of superposition of quantum mechanics. On the other hand. whereas the upper one corresponds to a time t0 ∼ M 3 . either by assuming that a slow dose of matter was continually fed into the black hole to maintain its size (the simplest assumption) or by considering a time t0 where say a quarter of the hole has evaporated (and modifying the metric to reflect the slow decrease of black hole mass). we will have the Hawking radiation that has emerged in the evaporation process.” The semiclassical process gives a state for the light matter fields. the radiation C will have the energy initially present in A. These contribute a negative value to the total mass of the system because of the way the hypersurface is oriented with respect to the coordinate t—this maintains overall energy conservation in the process.” On the upper hypersurface. let us call this “C.8 Das & Mathur Figure 1 is assumed to be of this form. we should be able to evolve our hypersurfaces by different amounts at different points—this is the “many-fingered time” evolution of general relativity extended to include the quantum matter on the spacetime. we have captured both the infalling matter A and the outgoing radiation C on the same spacelike hypersurface. If we want the radiation C to carry the information of the matter A. so there is nothing else on this hypersurface. in a semiclassical analysis of the Hawking process. The only difference between the hypersurfaces is that the later one has a larger r = constant part. We can now see explicitly the difficulties with obtaining in any theory a unitary description of the process of black hole formation and evaporation.” which. There is no radiation yet. For example. On the lower hypersurface. The lower one has t = 0. One can further check that the infalling matter has finite energy along each hypersurface and that scalar quantities such as dr/ds are bounded and smooth along each surface (s is the proper length along the surface). and the segment that connects this part to the r = constant part can be taken to be the same as well.” In the middle are the infalling members of the particle-antiparticle pairs. components A and B are expected to somehow vanish together (or leave a Planck mass remnant). but the conclusions . At the end of the process. It would have been very satisfactory if we just could not draw a smooth hypersurface like the upper one in Figure 1. the later one is in some sense just a time translate of the earlier one—the part t = constant in each surface has the same intrinsic (and extrinsic) geometry for each hypersurface. But since C will be entangled with B. We assume. since their energies cancel each other. we will still have the matter that fell in to make the hole. In fact. we could have hoped that any such surface would need to be nonspacelike at some point. In the above calculations. where a mass ∼ M has been evaporated away as radiation. the final state will not be a pure state of radiation. a hypersurface that includes both the infalling matter and the outgoing radiation. On the left. Let us call this part of the state “B. we expect the following sources of stress energy.

all consistent string theories (a) live in ten spacetime dimensions and (b) respect supersymmetry. If some such possibility is true. which gives a naive quantization of gravity. The basic scale is set by the string tension Ts or. etc. only the lowest harmonics are relevant. In particular. whose structure is only incompletely understood at this point (for . equivalently. then the radiation C must have had its origins somewhere else—possibly in some nonlocal combination of modes with sub-Planckian energy. The interaction between strings is controlled by a dimensionless string coupling gs . in the semiclassical calculation. we would obtain unitarity. Even in weak coupling perturbation theory. string theory provides not only a consistent theory of quantized gravity and matter. for instance. It thus appears that in order to have unitarity one needs a nonlocal mechanism (which operates over macroscopic distances ∼ M ) that moves the information from A to C. as any other matter system.D-Branes and Black Holes 9 do not change even if we let the evolution of spacelike surfaces be described by the Wheeler–de Witt equation. It would still be true that the final state of a system changes character significantly when its density changes from that of a star. but also a way of thinking about black holes as quantum states of the matter variables in the theory. statistical mechanics. although recent developments in nonperturbative string theory show that these five theories are in fact perturbations around different vacua of a single theory. string theory is best thought of as a theory of interacting elementary strings (for expositions of superstring theory. (11) ls = √ 2πTs The quantized harmonics of a string represent particles of various masses and spins. while also obtaining some nontrivial insight into the high-energy structure of the quantum vacuum. Thus. but if the complete theory of matter and gravity has an inbuilt cutoff at the Planck scale. Even though the spacetime appears to have no regions of Planckscale curvature. but the resulting hole should still be described by the same essential principles of quantum mechanics. to the density at which it collapses to form a black hole. see Reference [14]). but such effects cancel out in the computation of Hawking radiation [13]. A key point appears to be that. and not as an esoteric new object that must be added to a theory of “regular” matter. and the above description of the theory in terms of propagating and interacting strings is a good description when gs 1. the “string length” 1 . the radiation C emerges from modes of the quantum field that in the past had a frequency much higher than Planck frequency. A naive model of the quantum field would have these modes at all frequencies. 2 STRING THEORY AND SUPERGRAVITY In a certain regime of parameters. and the masses are typically integral multiples of 1/ls . quantization imposes rather severe restrictions on possible string theories. As we show below. Basic to such an approach would be some way of understanding a black hole as a complicated version of usual matter states. density of states. quantum fluctuations of the spacetime may appear large in certain coordinates [12]. at energies much smaller than 1/ls . there are five such string theories. we must alter our understanding of how information in one set of low-energy modes (A) moves into another set of low energy modes (C). At the perturbative level.

in which the 10dimensional spacetime is flat and the “internal” 6-dimensional space is a 6-torus T 6 with (periodic) coordinates y i . only the n = 0 mode can be excited. most of these modes must acquire a mass. y) as φ(x.1 Kaluza-Klein Mechanism How can such theories in ten dimensions describe our 4-dimensional world? The point is that all of the dimensions need not be infinitely extended—some of them can be compact.10 Das & Mathur a review of string dualities. the essential physics is the same. unlike these local theories of gravity. . What is rather nontrivial is that when one applies the same mechanism to the spacetime metric. Thus. Now. the simplest situation. in all these theories there is a set of exactly massless modes that describe the very-low-energy behavior of the theory. For other kinds of internal manifolds. which are not renormalizable. which is in fact the graviton: If the transverse directions of the vibrations are i and j. typically much smaller than 1/ls . we have more complicated wavefunctions on the internal space. φ(x. which we choose to be all of the same period: 0 < y i < 2πR. 2. The closed string can carry traveling waves both clockwise and counterclockwise along its length. If xµ denotes the coordinates of the noncompact 4-dimensional spacetime. the effective lower-dimensional world contains a metric field as well as vector gauge fields and scalar matter fields. The massless modes then describe the observed low-energy world. However. String theories have the potential to provide a unified theory of all interactions and matter. we can write a scalar field φ(x. It turns out that the massless modes of open strings are gauge fields. Of course. (12) where ni denotes the six components of integer-valued momenta n along the internal directions. The effective low-energy field theory is a supersymmetric Yang-Mills theory. for example. In closed string theories. although there have been recent suggestions that this length scale can be considerably longer without contradicting known experimental facts [15]. y) = ni φni (x) ei ni y i R . The most common scenario for this is to choose ls to be of the order of the Planck scale. a single field in higher dimensions becomes an infinite number of fields in the noncompact world. Consider. for example. The low-energy limits of closed string theories thus contain gravity and are supersymmetric extensions of general relativity—supergravity. Remarkably. the state with one quantum of the lowest harmonic in each direction is a massless spin-2 particle. When. however. this gives the gauge boson Ai . it is clear from Equation 12 y that the field φni (x) has (in four dimensions) a mass mn given by mn = |n|/R. string theory yields a finite theory of gravity—essentially due to the extended nature of the string. y) is a massless field satisfying the 2 standard Klein-Gordon equation ( x + 2 )φ = 0. to describe the real world. see Reference [16]). For energies much lower than 1/R. then we get the graviton hij . The lowest state of an open string carries one quantum of the lowest vibration mode of the string with a polarization i.

N = 8 supersymmetric theories appear complicated but can be obtained in a simple way from a D = 11. Aµ ∼ Gµ. The bosonic massless fields are thus 1. Such D = 4. (13) where R is the Ricci scalar and FM N P Q is the field strength of AM N P .10 → Bµν and Aµνλ → Cµνλ . 3.2 11-Dimensional and 10-Dimensional Supergravities Before the advent of strings as a theory of quantum gravity. the maximal number of supersymmetries is eight. from the point of view of the 103 dimensional spacetime. which leads to inconsistencies at the level of interactions. Clearly. In a similar way.10 is a vector and φ = 4 log G10. The gravity multiplet in the higher-dimensional theory gives gravity as well as matter fields after dimensional reduction to lower dimensions. The metric gµν . Aµν. there was an attempt to control loop divergences in gravity by making the theory supersymmetric. lp denotes the 11-dimensional Planck length so that the 11-dimensional Newton’s constant 9 is G11 = lp . The line interval may be written as ds2 = e− 6 gµν dxµ dxν + e 3 (dx11 − Aµ dxµ )2. say x10. There is only one scale. 2. N = 2 theory via the process of dimensional reduction explained above. But in four dimensions. A rank-2 antisymmetric tensor gauge field Bµν with field strength Fµνλ . a field Aµ and a field φ. the 3-form gauge field splits into a rank-2 gauge field and a rank-3 gauge field in ten dimensions. 4. At low energies. ν run over the values 0 · · · 9. more supersymmetries would force the theory to have fields of spin higher than 2 in the graviton supermultiplet. The various components of the 11-dimensional metric have been written in terms of a 10-dimensional metric. This theory has no free dimensionless parameter.D-Branes and Black Holes 11 2. A real scalar. The bosonic part of 11-dimensional supergravity consists of the metric gM N and a 3-form gauge field AM N P with an action S11 = 1 [ 9 (2π)8lp √ 1 1 d11x −g[R − FM N P Q F M N P Q] + 48 6 d11x A ∧ F ∧ F ]]. the dilaton φ. lp. the better was the control of divergences. Now consider compactifying one of the directions. A rank-3 antisymmetric tensor gauge field Cµνλ with field strength Fµνλρ . the indices µ. N = 1 theory or a D = 10. 5. The greater the number of supersymmetries. all the fields in the action (Equation 13) are independent of x10 and the 10-dimensional action is 1 1 1 3φ √ S = (2π)1g2 l8 d10x g(R− ( φ)2 − e−φ Fµνα F µνα − e 2 Fµν F µν 7 s 2 12 4 1 φ/2 µναβ + e Fµναβ F )+··· (15) 48 . with specific interactions between all the fields. A vector gauge field Aµ with field strength Fµν . φ 4φ (14) In Equation 14.10 is a scalar. The field Aµ behaves as a U (1) gauge field.

This cannot be obtained from D=11 supergravity by dimensional reduction. giving 1 2πR = . If. The scalar field φ is called a dilaton and plays a special role in this theory.12 Das & Mathur where the ellipsis denotes the terms that come from the dimensional reduction of the last term in Equation 13. it is clear that R = g 2/3lp. the fermionic sector can be treated similarly. One of them has positive 10-dimensional chirality whereas the other has negative chirality. Because of the self-duality constraint on the 5-form field strength. 12 (18) . Two real scalars : the dilaton φ and the axion χ 3. most notably the “string frame” metric. Although we have given the explicit formulae for dimensional reduction of the bosonic sector of the theory. Type IIB supergravity. we put the 5-form field strength to zero. 9 8 (2π)8lp (2π)7g 2ls (17) which defines the string length ls . differ from this by conformal transformations. The action (Equation 15) is precisely the bosonic part of the action of Type IIA supergravity in ten dimensions. There are two types of gravitinos—fermionic partners of the graviton. it is not possible to write down the action for Type IIB supergravity. The metric gµν 2. so that Equation 17 gives lp = g 1/3ls . In this article we always use the Einstein frame metric. There is another supergravity in ten dimensions. however. (16) The overall factor in Equation 15 follows from the fact that the 11-dimensional measure in Equation 13 is related to the 10-dimensional measure by a factor of the radius of x10 (which is R). The 10-dimensional metric gµν used in Equations 14 and 15 is called the “Einstein frame” metric because the Einstein-Hilbert term in Equation 15 is canonical. The bosonic fields of this theory are 1. Its expectation value is related to the string coupling gs = exp ( φ ). Other metrics used in string theory. F denotes the field strength of the appropriate gauge field. although the equations of motion make perfect sense. The resulting theory is thus nonchiral. Both the gravitinos of this theory have the same chirality. A rank-4 gauge field Dµνλρ with a self-dual field strength Fµναβδ . we have a local action given by S= 1 (2π)7 g 2 l8 s √ d10x[ g(R− 1 1 ( φ)2 − e−φ Hµνα H µνα 2 12 1 φ − e (Hµνα − χHµνα )(H µνα − χH µνα )]. From the 11-dimensional metric. Two sets of rank-2 antisymmetric tensor gauge fields: Bµν and Bµν with field strengths Hµνλ and Hµνλ 4.

It is then logical to postulate that there exists in the theory a 2-dimensional fundamental object (the 2-brane). we find that there are (32 × 33)/2 = 528 objects on the left-hand side of this equation. Although. The anticommutator of two supercharge components should lead to a translation.1 Branes in Supergravity A closer look at even supergravity theories leads to the observation that the existence of extended objects is natural within those theories (and in fact turns out to be essential to completing them to unitary theories at the quantum level). If we write down all the possible terms on the right that are allowed by Lorentz symmetry. Because the anticommutator is symmetric in α. P1 = 0 implies that the configuration has momentum in direction X 1. Furthermore. the charge ZABCDE corresponds to a 5-brane in the theory. so we write {Qα. The supercharge Qα is a spinor. since they are not renormalizable. called the Type IIA and Type IIB string. A crucial ingredient in this new insight is the fact that higher-dimensional extended objects—branes— are present in the spectrum of string theory. they are the low-energy limits of string theories. . in agreement with the number on the left. these supergravities cannot be consistently quantized. for example. whereas the number of ZABCDE is 11 C5 = 478. 3 BRANES IN SUPERGRAVITY AND STRING THEORY Although string theory removes ultraviolet divergences leading to a finite theory of gravity. with α = 1 . Qβ } = (ΓA C)αβ PA + (ΓA ΓB C)αβ ZAB + (ΓA ΓB ΓC ΓD ΓE C)αβ ZABCDE . some find the change from a pointlike particle to a string somewhat arbitrary—if we accept strings. then we find [17] {Qα. then why not extended objects of other dimensionalities? Over the past few years. β. but only 11 objects (the PA ) on the right-hand side. as nonperturbative string theory has developed. The 5-brane is easily seen to be the . and now we have a total of 528 objects on the right. what is the interpretation of Z12 = 0? It turns out that this can be interpreted as the presence of a “sheetlike” charged object stretched along the directions X 1. Consider the case of 11-dimensional supergravity. However. Qβ } = (ΓA C)αβ PA . The number of ZAB is 11C2 = 55. 3. it has been realized that the features of string theory are actually very natural and also perhaps essential to any correct theory of quantum gravity. such features as the necessity of ten dimensions and the presence of an infinite tower of modes above the massless graviton made it unpalatable to many physicists. just as a particle with 1-dimensional world line couples to a 1-form gauge field as Aµ dxµ . The 2-brane has a 2 + 1 = 3dimensional world volume and couples naturally to the 3-form gauge field present in 11-dimensional supergravity. (19) where the Z are totally antisymmetric. Similarly.D-Branes and Black Holes 13 Of course. . 32. where C is the charge conjugation matrix. X 2.

Then. getting a 4-dimensional D-brane in string theory. Now suppose there are two such supersymmetries: † Q1 = Q1 . there is no freedom to add or remove any object. Nonsupersymmetric states occur in a “multiplet” containing a bosonic state |B and a fermionic state |F of the same energy: Q|B ≡ |F . (22) (23) where the equality holds in the last step if and only if that is. But how does this relate to string theory. we may wrap the 5-brane on the small circle. Q2} = Z. Similarly.14 Das & Mathur magnetic dual to the 2-brane. getting a solitonic 5-brane. (26) . we get what looks like a string in ten dimensions. 2 2 Q1 = H. (20) where H is the Hamiltonian. nor to change any couplings. or leave it unwrapped. These relations show that the energy of any state cannot be negative. Such an object is indeed present—it is one of the D-branes shown to exist in string theory by Polchinski [18]. it is natural to include some specific extended objects in the quantization of 11-dimensional supergravity. if we let the 2-brane wrap this small circle. it will take a c number value on the states that we consider below. which is also known to exist in the theory. in which case there should be a two-dimensional extended object in Type IIA string theory. Thus. we can now conclude 0≤ ψ|(Q1 ± Q2)2 |ψ = 2E ± 2Z. If H|ψ = E|ψ (21) then E = ψ|H|ψ = ψ|Q2|ψ = Qψ|Qψ Q|ψ = 0. Q} = 2Q2 = 2H. We can also choose not to wrap the 2-brane on the small circle. Q † = Q2 . Let us see in a simple context what a BPS state is.2 BPS States A very important property of such branes is that when they are in an unexcited state. In a spirit similar to the calculations above. ≥ 0. Thus. they preserve some of the supersymmetries of the system. thus obtaining 10-dimensional noncompact spacetime. with specified interactions between them—in fact. 2 Q2 = H. (25) where Z is a “charge”. which lives in ten dimensions? Let us compactify the 11-dimensional spacetime on a small circle. 3. the state is supersymmetric. and it couples to the 6-form that is dual to the 3-form gauge field in 11 dimensions. (24) Supersymmetric states have E = 0 and need not be so paired. This is exactly the Type IIA string that had been quantized by the string theorists! The size of the small compact circle turns out to be the coupling constant of the string. one is forced to a unique set of extended objects in the theory. and are thus Bogomolny-Prasad-Sommerfield saturated (BPS) states. Q|F = Q2|B >= E|B . {Q1. Consider first a theory with a single supercharge Q = Q† : {Q.

3 The Type IIB Theory Instead of using the 11-dimensional algebra. for special configurations—corresponding to “unexcited branes”—there are solutions which retain some of the supersymmetries. or D-instantons. there is a corresponding BPS brane. We use such branes below in constructing black holes. D(−1)-brane. or elementary string. D1-brane. take the case (Q1 − Q2)|ψ = 0. States with Q1 |ψ = Q2 |ψ = 0. Q2. These states fall into a “short multiplet” described by. These are BPS saturated solutions. we could have used the 10-dimensional algebra and arrived at the same conclusions. . They are massive solitonlike objects and therefore produce gravitational fields. the existence of BPS branes in Type IIB supergravity follows from the corresponding algebra. By Equation 26. 2. 3. NS1-brane. but a similar analysis applies to classical solutions. a nontrivial dilaton. These form a “long multiplet” {|ψ . Such states satisfy E = |Z| > 0 (30) and are called BPS states. Q1 |ψ . they produce the p-form gauge fields to which they couple and. In 10-dimensional supergravity. Thus we have 1. carrying charge under the axion field χ. Note that Q2 |ψ = Q1|ψ . A brane in a general configuration would break all the supersymmetries of the theory. However. 3. Q2Q1 |ψ = Q2 |ψ = E|ψ . For each antisymmetric tensor field present in the spectrum. Q1|ψ }. they also have Z|ψ = 0. In a similar fashion. 15 (27) (28) 1. they are stable objects. carrying electric charge under Bµν . the branes mentioned above appear as classical solutions of the equations of motion. so they carry no charge. Note that by Equation 26. 1 (29) so that we have no more linearly independent states in the multiplet. the state with Z 0 satisfies (Q1 −Q2 )|ψ = 0 while the state with Z < 0 satisfies (Q1 +Q2 )|ψ = 0. we have regarded the BPS states as states of a quantum system. 2. in general.D-Branes and Black Holes This implies that E ≥ |Z|. typically with sources. but satisfying Equation 28. say. the basis {|ψ . Q2 |ψ . These have E = 0 and do not fall into a multiplet. For concreteness. In the above discussion. Now we have three kinds of states: or (Q1 + Q2 )|ψ = 0. carrying electric charge under Bµν . States not in category 1. They must have E > |Z| > 0. with equality holding if and only if (Q1 − Q2 )|ψ = 0. Q2 Q1 |ψ }. 3. States that are not annihilated by any linear combination of Q1 . and since they have the maximal charge for a given mass. Apart from that.

The possible values of p are determined entirely by the properties of fermions in the Type II theory. Das & Mathur We have denoted some of the branes as D-branes. with µ = 0.16 4. the dual of the D(−1) brane. In fact. 4. (33) where φi = X i with i = 1. 7. Recalling the massless spectrum of these theories. This action is invariant under arbitrary transformations of the coordinates ξ α on the brane. 5. 3. In each of these cases. This fact implies that the soliton is a stable construct in the theory. For a single 1-brane. Now we can have both transverse and longitudinal oscillations of the . The low-energy action is thus expected to be the tension of the brane times its area. 6). We have remarked above that the equations of motion for the massless fields admit solitonic solutions. · · · . the action for long-wavelength deformations is S= T1 2 d2 ξ α det(gαβ ). carrying magnetic charge under Bµν . carrying magnetic charge under Bµν . p must be even (p = 0. gαβ = ∂αX µ ∂β X ν ηµν . 1. · · ·8. 7). we can easily see that there will be low-energy excitations of this soliton in which its world sheet suffers small transverse displacements that vary along the brane (in other words. 2. 6. the soliton involves the gravitational field and some other (p + 1)-form field in the supergravity multiplet so that the final solution carries a charge under this (p + 1)-form gauge field. for instance. are the remaining fields. carrying charge under Dµνρλ . the brane carries waves corresponding to transverse vibrations). NS5-brane. denotes an arbitrary coordinate sytem on the D1-brane world sheet and gαβ denotes the induced metric on the brane. D5-brane. Because the brane is not pointlike but is an extended object. it is best to work in a static gauge by choosing ξ 1 = X 0 and ξ 2 = X 9. this soliton is seen to have a mass equal to its charge and is thus an object satisfying the BPS bound of supersymmetric theories. The determinant in Equation 31 may be then expanded in powers of ∂φi . To make contact with to the picture discussed above. (31) where ξ α. 3. These play a special role in string theory. with α = 1. It turns out that for Type IIA. as explained below. denotes the coordinates of a point on the brane.4 D-Branes in String Theory Consider the low-energy action of the supergravity theories. (32) The X µ . we find that for each antisymmetric tensor field there is a brane that couples to it. where the solitons are not in general pointlike but could be p-dimensional sheets in space [thus having (p + 1)-dimensional world volumes in spacetime]. D7-brane. in appropriate units. 9. whereas for Type IIB p must be odd (p = −1. The lowest-order term is just the free kinetic energy term for the eight scalar fields φi . It is now straightforward to extend the above discussion for higher-dimensional branes. 3-brane. 5. 2. The induced metric (Equation 32) then becomes gαβ = ηαβ + ∂α φi ∂β φj δij .

This structure immediately extends to the case where N branes approach each other. an open string can also begin on the first brane and end on the second. It turns out that the longitudinal waves are carried by a U (1) gauge field Aα with the index α = 0. where g is the string coupling. which corresponds to Neumann (N) boundary conditions. φi . Once again. But the excitations of the brane will still be low-energy modes. Because these branes were discovered through their excitations. For a p-brane we have (9 − p) transverse coordinates. ab . But we would like to obtain also the microscopic quantum physics of a single brane. There are two possibilities: to let the ends move at the speed of light. and this gives two additional possibilities for the excitation of the system. The open strings that begin and end on the first brane will describe excitations of the first brane. the branes are themselves called D-branes. and such open strings describe the excitations of the p-brane [18]. 1.D-Branes and Black Holes 17 brane. How do we see such an object in string theory? The mass per unit volume of a D-brane in string units is 1/g. In the above description. The generalization of Equation 31 [called the Dirac-Born-Infeld (DBI) action] is Tp S= dp+1 ξ det(gαβ + Fαβ ). this is the analog of describing a point charge by its classical electromagnetic potential. In fact. as shown in Figure 2. when we quantize a free string we have two choices: to consider open strings or closed strings. (For a review of properties of D-branes. If the ends of the open strings are free to move along the directions ξ α but are fixed in the other directions X i. Such a description should apply to a collection of a large number of fundamental branes all placed at the same location.) An interesting effect occurs when two such branes are brought close to each other. or to fix the end. with the (ij) element of the matrix given by open strings that begin on the ith brane and end on the jth brane. · · ·p ranging over the world volume directions. we had obtained the branes as classical solutions of the supergravity fields. The low-energy expansion of the action then leads to electrodynamics in (p + 1) dimensions with some neutral scalars and fermions. If we have an open string then we need boundary conditions at the ends of the string that do not allow the energy of vibration of the string to flow off the end. Indeed the low-energy worldbrane theory of a collection of N parallel Dp-branes turns out to be described by a non-Abelian U (N ) gauge theory in (p + 1) dimensions [21]. or begin on the second and end on the first. The low-energy limits of open string theories are generally gauge theories. Of course we can choose different types of conditions for different directions of motion in spacetime. and hence as many scalar fields on the (p + 1)-dimensional brane world volume. (34) 2 where Fαβ is the gauge field strength. labeled by I = 1. So the brane would not be seen as a perturbative object at weak coupling. These four possibilities can in fact be encoded into a 2 × 2 matrix. which corresponds to Dirichlet (D) boundary conditions. then the ends are constrained to lie along a p-dimensional surface that we may identify with a p-brane. and these should be seen in weakly coupled string theory. But. The N 2 gauge fields are best written as an N × N matrix Aα . and those that begin and end on the second brane will describe excitations of the second brane. which were in turn a consequence of D-type boundary conditions on the open strings. one can choose a static gauge and relate the fields directly to a string theory description. see Reference [19]. · · · (9 − p).

The ground states of the above D-branes are BPS states states. U (M ) × [U (1)]N −M . The potential for such a theory turns out to be Tr([φi. i. resulting in an enhanced unbroken symmetry. The latter has no scalar fields but has 10 components of the gauge field. where µ = 0. while di = (φi − φbb ) denotes the separation along aa ab direction i. The coupling constant of the theory is 2 gYM ∼ gs (ls )p−3 . 9. all the di are nonzero. b = 1 · · · N . the gauge group is broken to [U (1)]N —this is the situation when all the branes are separated from each other. the diagonal entries in the matrix φi denote the locations of the branes. Aµ . however. i. this is the dimensional reduction of the 10-dimensional supersymmetric Yang-Mills theory to p + 1 dimensions. φj ])2. Recently. the low-energy action is that of a standard supersymmetric Yang-Mills theory in p + 1 dimensions. other kinds of non-BPS and unstable D-branes have been constructed in string theory [20]. we restrict . the components parallel to the brane. we have a supersymmetric Yang-Mills theory with 16 supercharges.e. · · ·. · · ·(9 − p) are scalars on the world volume and are renamed as φi . say M . The analog of the DBI action for a collection of N branes is not completely known at present. · · · . although there are some proposals (for discussion see Reference [22] and references therein). 9. remain gauge fields.e. the corresponding di ab are zero for a.18 Das & Mathur Figure 2: Open strings ending on D-branes. µ = 0. ab 1 Thus 2 (φi + φi ) denotes the center-of-mass transverse coordinate of the pair of aa bb i branes labeled by a and b. Generically. of the branes are coincident. where a. ab when some number. However. However. Similarly. b = 1 · · · M . Since BPS states break half the supersymmetries of the original theory. In fact. which transform ab according to the adjoint representation of U (N ). Under dimensional reduction. A nonzero expectation value for di means that the gauge symmetry ab is spontaneously broken. there are N 2 scalar fields φi . which means that they are stable. In fact. so that in the ground state one can always choose a gauge so that all the φi s are diagonal. each of which is a N × N matrix. (35) Each of these fields has its corresponding fermionic partner. (10 − p). In this article. whereas the components transverse to the brane with µ = 1.

is an exact symmetry of string theory. It arises from a dimensional reduction of 11-dimensional supergravity on a circle. since there is no corresponding picture of such an endpoint in 11 dimensions. Consider the Type IIA theory. it also turns a Type IIA theory into Type IIB and vice versa. the role of the elementary string is interchanged with the D1-brane. and thus get a 5-brane in the Type IIA theory. A useful example is a bound state of Q1 D1-branes and Q5 D5-branes.5 Duality A remarkable feature of string theory is the large group of symmetries called dualities (see Reference [16] for review). The physics of the 5-brane in Type IIA theory is an interesting one. then in the Type IIA theory we get a D4-brane and an open string ending on the D4-brane. then the open string cannot end on this brane. Much more is known about the physics of D-branes in Type IIA and IIB theories. When these branes are not excited. In this article. We can take a graviton propagating along this compact direction. The 11-dimensional theory has 5branes and 2-branes. we do not use the idea of dualities directly. The combination of S. . If we wrap on the compact circle both the 5-brane and the 2-brane that ends on it. But if we do not wrap the 5-brane on the circle. and those that relate these theories to the theory of unoriented strings. and for specific instances these can be threshold bound states. since they can be studied through perturbative open string theory.D-Branes and Black Holes 19 ourselves to BPS branes and their excitations. We note that if we replace R → 1/2πTsR. the U-duality group. Consider Type IIA theory compactified on a circle of radius R. so the physics obtained is much more universal than it may at first appear. called T-duality. But we can also wind an elementary string on this circle. but we will not discuss it further. There are other dualities. and the 2-brane can end on the 5-brane. which relates weak coupling to strong coupling while interchanging fundamental quanta with solitonic objects.and T-dualities generates a larger group. Branes of different kinds can also form bound states. The open strings that describe this system are (a) (1. then the energies of the above two sets of states are simply interchanged. and (c) (1. its energy spectrum would have the form |np|/R. The Type IIB theory also has another symmetry called S-duality. they are in a threshold bound state. (b) (5. there the string coupling gs goes to 1/gs . 5) and (5. At the same time. but we note that any black hole that we construct by using branes is related by duality maps to a large class of similar holes that have the same physics. In fact this map. 5) strings with both endpoints on D5-branes. 1) strings with one endpoint on any of the D1-branes and the other endpoint on one of the D5-branes. Such a duality. 1) strings with both endpoints on any of the D1-branes. 3. where nw is the winding number of the string and Ts its tension. is a realization of the duality suggested for field theory by Montonen & Olive [23]. Not all the branes that were listed for Type IIA and Type IIB theory are Dbranes. such as those that relate Type IIA theory to heterotic string theory. the spectrum here would be 2πTs |nw |R. and because of an effect on fermions. it interchanges winding and momentum modes.

Thus. which is a “winding charge. Then the gravitational coupling also increases. since the degeneracy can be predicted also at strong coupling. Thus. we may calculate the degeneracy of states at weak coupling and. To prevent this pinching. We have shown above that the masses of BPS saturated states are indeed determined once we know their charges. but with a further half of the supersymmetries broken. The momentum modes have an energy that goes as 1/R. the state of such a string is unique (it does have a 256-fold degeneracy due to supersymmetry. this looks like a massive point object carrying electric charge under the gauge field that results from the dimensional reduction of the rank-2 field. we can put some momentum along the string. black holes that are “extremal”—they have the minimal mass for their charge if we require that the metric does not have a naked singularity. but we can ignore this—it is not a number that grows with n1 ). we expect a charged black hole. we can study a state of the theory at weak coupling. Thus. this black hole turns out to have a vanishing horizon area. entropy is zero from both the microscopic and the black hole viewpoints. For this black hole we can compute the Bekenstein entropy from (Equation 5). This is where BPS states come in. Take a fundamental string and wrap it n1 times around one of these circles. Thus. compare the result with the BekensteinHawking entropy of the corresponding black hole. which amounts to having traveling waves move along the string. for such states. since for a given charge n1 this has the lowest allowed mass given by the energy given above. where we can use our knowledge of string theory. at strong coupling. which is consistent but not really interesting. This will produce a rank-2 NS field B09 with a charge n1 . For such a calculation to make sense. Now imagine the coupling to be tuned to strong values. However. If we increase the coupling. The extended objects discussed in the previous section have played an important role in understanding black holes in string theory. and ask if the microscopic computation agrees with the Bekenstein entropy. which would be the logarithm of the number of states with the same mass and charges. If we put waves that are moving only in one of the directions. Such states give. black holes should appear in this theory as excited quantum states. One way to understand this is to note that the tension of the string “pinches” the circle where the string was wrapped. which are simply their winding numbers on cycles of the compact space. From the point of view of the noncompact directions. and the object must become a black hole with a large radius. which saturates the BPS bound. Thus we may compute the “entropy” of the state. An example of such an object is a fundamental string in Type IIA or IIB string theory. Furthermore. say along the 9 direction. we must have some assurance that the density of states would not shift when we change the coupling. The total energy of . Let some of the directions of the 10-dimensional spacetime be compactified on small circles. which has a radius R. the microscopic entropy is zero. From the microscopic viewpoint. Because the coupling in the theory is not a constant but a variable field.” The energy of the state is E = 2πn1Ts R. this is an extremal black hole. so now this circle attains a finite size. we still have a BPS saturated state.20 Das & Mathur 4 BLACK HOLE ENTROPY IN STRING THEORY: THE FUNDAMENTAL STRING String theory is a quantum theory of gravity. An idea of Susskind [24] has proved very useful in the study of black holes.

This was the first indication that degeneracy of states in string theory may account for Bekenstein-Hawking entropy. since the curvatures are much larger than the string scale and stringy corrections to supergravity become relevant. In an important paper. n2) ∼ exp[2 2 n1 n2 ]. However. At strong coupling. There is a great deal of ambiguity in defining the stretched horizon precisely. the number of states asymptotes to √ √ (36) n(n1 . in this example. In addition. It is clear that what we need is a black hole solution that (a) is BPS and (b) has a nonzero large horizon area. where the second term is now the contribution from the momentum waves. we have an object with two quantized charges n1 and n2 . this object is described by an extremal black hole solution with two charges. The identification of such fundamental string states with the corresponding classical solution was proposed by Dabholkar & Harvey [25]. The BPS nature of the solution would ensure that degeneracies of the corresponding string states are the same at strong and weak couplings. the precise coefficient cannot be determined. A large horizon would ensure that the curvatures are weak and we can therefore trust semiclassical answers. However. Sen [27] argued that the semiclassical entropy (for similar black holes in heterotic string theory) is given not by the area of the event horizon but by the area of a “stretched horizon. Because of the winding. and there are fermionic oscillators as well. the effective length of the string is Leff = 2πn1R. so that the momentum may be written as P = (2πn1n2 )/Leff . Then. These correspond to the various ways one can get the oscillator level n0 = n1 n2 . Thus. as mentioned above. thus allowing an accurate computation of the density of states. The resulting degeneracy of states has been known since the early days of string theory. .D-Branes and Black Holes 21 the state is now given by E = 2πn1Ts + n2 /R. since the definition of the stretched horizon is itself ambiguous. Sen found that the area of the stretched √ horizon in units of the gravitational constant is proportional to n1 n2 . a large number of states have the same energy. which is precisely the logarithm of the degeneracy of states given by Equation 36. The second charge is due to the presence of momentum. In that situation. The charge n1 is due to the winding of the string. The corresponding charge is n2 and is an integer since the momentum in the compact direction is quantized. It is indeed inconsistent to trust a classical gravity solution beyond this surface. The horizon area is still zero and the curvatures are large near the horizon.” This is defined as the location where the curvature and local temperature reach the string scale. it is necessary to consider the fact that there are eight possible polarizations. It turns out that this requires three kinds of charges for a 5-dimensional black hole and four kinds of charges for a 4-dimensional black hole. this is equivalent to a gauge field A0 from the viewpoint of the noncompact world. by the Kaluza-Klein mechanism explained in the previous section. a microscopic count of the states could be compared with the Bekenstein-Hawking entropy in a regime where both calculations are trustworthy. For large n1 and n2 . which results in a term in the 10-dimensional metric proportional to dx9dx0. for given values of n1 and n2 . From the viewpoint of the noncompact directions.

and there are three kinds of gauge fields: 1. Finally. . since there is momentum along x5 . and (c) some momentum along x5. A rank-2 gauge field Bij with i and j lying along the 3-sphere S 3. (38) Here r is the radial coordinate in the transverse space. The background also has nontrivial values of the dilaton φ. · · · x4 ). The solution represents a black hole with an outer horizon at r = r0 and an inner horizon at r = 0. Extremal limits of such black holes provided the first example where the degeneracy of the corresponding BPS states exactly accounted for the BekensteinHawking entropy [28]. since there are 5-branes along (x5 · · · x9 ).22 Das & Mathur 5 THE FIVE-DIMENSIONAL BLACK HOLE IN TYPE IIB THEORY The simplest black holes of this type are in fact 5-dimensional charged black holes. (b) D1-branes wrapped around x5. The noncompact directions are then (x0. 2. we perform a Kaluza-Klein reduction to the five noncompact dimensions. all related to each other by string dualities. r0 1 − r2 1+ 2 r0 sin2 α5 r2   (37) 1+ 2 r0 sinh2 σ r2 . We describe below one such solution in Type IIB supergravity. The corresponding charges Q1 . The resulting metric is1 ds2 = −[f (r)]−2/3 where f (r) ≡ 1+ 2 r0 sin2 α1 r2 r2 1− 0 r2 dr2 2 2 dt2 + [f (r)]1/3  2 + r dΩ3 . which comes from the dimensional reduction of an offdiagonal component of the 10-dimensional metric g05. A gauge field A0 . This is nonzero. 5. 8 32π 4ls g 2 0 (39) We use the notation of Horowitz et al [26]. r2 = 4 (xi )2. x6. This is the dimensional reduction of a corresponding Bij in ten dimensions. which comes from the dimensional reduction of the rank2 antisymmetric tensor gauge field B05 in 10 dimensions. · · · x9 ). 3. There is a solution of Type IIB supergravity that represents (a) D5-branes wrapped around the T 5 .1 The Classical Solution We start with 10-dimensional spacetime and compactify on a T 5 along (x5 . There are several such black holes in Type IIA and IIB theory. The presence of these gauge fields follows in a way exactly analogous to the dimensional reduction from 11-dimensional supergravity discussed in Equations 13 and 14. This is nonzero. Q5. A gauge field A0 . N are given by Q1 = 1 (1) (2) 2 V r0 sinh 2α1 6 32π 4gls Q5 = 2 r0 sinh 2α5 2 2gls N= V R2 2 r sinh 2σ. and dΩ2 i=1 3 is the line element on a unit 3-sphere S 3. since we have D1-branes along x5 .

3 Extremal and Near-Extremal Limits The solution given above represents a general 5-dimensional black hole. Sextremal = 2π Q1Q5 N. (45) In this limit. we are interested in a special kind of departure from extremality. α5 . R is the radius of the x5 circle.D-Branes and Black Holes 23 where V is the volume of the T 4 in the x6 · · · x9 directions. σ → ∞ Q1. If we look at the higher-dimensional metric before dimensional reduction.2 Semiclassical Thermodynamics The semiclassical thermodynamic properties of the black hole may be easily obtained from the classical solution using Equations 5 and 10 and the relationship G5 = 8 4π 5g 2ls RV (42) between the 5-dimensional Newton’s constant G5 and V. 2πr0 cosh α1 cosh α5 cosh σ 3 RV r0 AH = cosh α1 cosh α5 cosh σ 8 G5 8π 3ls g 2 (43) (44) 5. 8 32π 4g 2ls (41) 5. The extremal limit is defined by 2 r0 → 0 α 1 . N = fixed. and g is the string coupling. The charge N comes from momentum in the x5 direction. In the following. This is the regime in which α1 . ls. it is clear from the above expressions that ADM mass and entropy remain finite while the Hawking temperature vanishes. Of particular interest is the extremal limit of the entropy. volume of the compact directions. However. Q5. but r0 and σ are finite. In this case. and string length. g. 2 6 gls ls (47) . α5 1. (46) This is a function of the charges alone and is independent of other parameters. The expressions are SBH = and TH = 1 . it is straightforward to identify the Arnowitt-Deser-Misner (ADM) momentum as PADM = N R (40) The ADM mass of the black hole is given by M= 2 RV r0 [cosh 2α1 + cosh 2α5 + cosh 2σ]. the inner and outer horizons coincide. R. such as the string coupling. the total ADM mass may be written in the suggestive form E= RQ1 16π 4RV Q5 + + EL + ER .

The meaning of the subscripts L and R will be clear soon. In a similar fashion. and 39. R. TL TR r0e±σ .4 Microscopic Model for the Five-Dimensional Black Hole This solution and the semiclassical properties described above are valid in regions −2 where the string frame curvatures are small compared with ls . A short calculation using Equations 37. (g 2N ) 1. we can study the regime of validity of the classical solution in terms of the various parameters. 2πr1r5 (50) TL. If we require this to be true for the entire region outside the horizon. 38. (52) Ti = π RQ1 Q5 2π 2RQ1Q5 Finally.R = RV r1r5 r0e±σ 8 16π 3g 2ls 1 1 = TH 2 1 1 + . This is the origin of the subscripts L and R—we have denoted the direction of momentum in the extremal limit as left. we can easily see that the following relations hold: Si Ei 1 = . The contribution to the ADM mass from momentum along x5 is Em = EL +ER.24 Here we have defined EL = 2 N V Rr0 e−2σ + 4g 2l8 R 64π s Das & Mathur and ER = 2 V Rr0 e−2σ . but the total momentum is still N/R. and 39 shows that this is given by (gQ1).and right-moving parts is typical of waves moving at the speed of light in one space dimension.R = (51) The above relations are highly suggestive. (gQ5). In fact. For finite σ we have both right. the thermodynamic quantities may be written as S = S L + SR where SL. 2N ∆E = (53) 5. 2 V r2 2 2 ≡ r0 sinh2 α5 ∼ 0 sinh 2α5 = gls Q5 . In the extremal limit σ → ∞.and left-moving waves. which implies that the waves are moving purely in one direction. 4g 2l8 64π s (48) and used the approximations 2 2 r1 ≡ r0 sinh2 α1 ∼ 2 r5 6 2 16π 4gls Q1 r0 sinh 2α1 = . 2 (49) and the expressions for the charges (Equation 39). it then follows from Equations 48 and 40 that Em → PADM . using Equations 50. 51. we write the expressions that relate the temperature T and the entropy S to the excitation energy over extremality ∆E = Em − N/R: π2 (RQ1 Q5) T 2 2 R∆E 1 and S = Sextremal[1 + ( ) 2 ]. (54) . L. i = L. The splitting of various quantities into left.

The total energy (Equation 47) is then seen to be the sum of the masses of Q1 D1-branes. Thus. If we had a single D1-brane and a single D5-brane.e. the charges Q1 and Q5 are integers. i. The modes are essentially those of the oscillations of the D1-branes. whereas the motions of the D1-branes along the four directions parallel to the D5-branes. The low-energy theory of a collection of Q Dp-branes is a (p + 1)-dimensional supersymmetric gauge theory with gauge group U (Q). the quantized waves would be massless particles with four flavors. along x6 · · · x8 ).D-Branes and Black Holes 25 In this regime. if we only have some number nw of D1branes without anything else wrapping a compact circle.5 The Long String and Near-Extremal Entropy The low-energy excitations of the system become particularly transparent in the regime where the radius of the x5 circle. are gapless excitations. and there are four polarizations for each such oscillation. Since we have a supersymmetric theory. we have four bosons and four fermions. we first present a physical picture of the low-energy excitations.and D5-branes is marginally bound. is much larger than the size of the other four compact directions V 1/4. However. however. However. but transverse to the D1-branes themselves (i. The third charge is from a momentum in the x5 direction equal to N/R. in string theory. When we have many D1-branes and D5-branes. This is readily apparent from the classical solution. These.e. R. the classical solution is a good description of an intersecting set of D-branes in string theory. there is a nonzero binding energy whenever the D1-branes try to move transverse to the D5-branes. If one requires that the nonextremal excitations of the D1 brane are in one-to-one correspondence with the known near-extremal excitations of the fundamental string and therefore yield the same degeneracy of states. The low-energy theory is still a gauge theory but has additional matter fields (called hypermultiplets). However. the way to obtain a state with large D-brane charge Q is to have a collection of Q individual D-branes. 5. which is now multiply wound along the compact circle. there are other possible configurations. these branes are in a threshold bound state. it would appear at first sight that there should be 4Q1Q5 such flavors: Each of the D1-branes can oscillate along each of the D5branes. This is because it costs energy to pull a D1-brane away from the D5-brane. Then the effective theory is a (1 + 1)dimensional theory living on x5 . S-duality relates D1-branes to fundamental wrapped strings. In the extremal limit (Equation 45). which is of length 2πnw R. one concludes that the energies of the individual quanta of oscillations of the D1 string must have fractional momenta pi = ni /(nw R). a bound state with zero binding energy. Instead of starting from the gauge theory itself. These correspond to several of the D1-branes joining up to form a longer 1-brane. Thus. This was discovered in an analysis of the nonextremal excitations of such D1-branes [29]. In fact. and we have a system of Q1 D1-branes and Q5 D5-branes. which . now we have a rather complicated bound state of intersecting branes. they would also prefer to join into a long string. N is quantized in the microscopic picture to ensure a single valued wave function. Even though a system of static D1. Q5 D5-branes and the total momentum equal to N/R. have four rather than eight polarizations. This is indeed the case if the D1-branes are all separate.

which we assume thermalizes. For the situation we are discussing.e. Since the string is relativistic.e. an effective length of Leff = 2πRQ1Q5 [30]. Our treatment below follows Reference [33]. 2 8 (β + α) (β − α)2 1 1 f Lπ + . ∂α (56) The average number of particles nr in state (er . the entropically favorable configuration turns out to be that of a single long string formed by the D1-branes joining up to wind around the x5 direction Q1Q5 . with the important modification of the multiply wound D-string proposed by Maldacena & Susskind [30].1 Statistical Mechanics in 1 + 1 Dimensions At weak coupling these bosons and fermions form an ideal gas. eβer +αpr ± 1 (57) where as usual the plus sign is for fermions and the minus sign is for bosons. these particles are massless. For large L. If nr denotes the number of particles with energy er and momentum pr . Finally. r (55) Then α and β are determined by requiring E=− ∂h ∂β and P =− ∂h . the entropy S is given by the standard thermodynamic relation S = h + αP + βE. i. the partition function Z is Z = eh = exp states −β r nr er − α nr pr . The above quantities may be easily evaluated: 1 1 f Lπ − . This approach to the derivation of black hole thermodynamics comes from Callan & Maldacena [31] and Horowitz & Strominger [32]. D1-branes bound to D5-branes. pr ) = 1 . 4 β+α β−α (58) P = E = and S = (59) .26 Das & Mathur simply means that the effective length of the string is 2πnw R.5. We thus arrive at a gas of four flavors of bosons and four flavors of fermions living on a circle of size Leff . i. pr ) is then given by ρ(er . 2 8 (β + α) (β − α)2 1 1 f Lπ [ + ]. Consider the general case of such an ideal gas with f flavors of massless bosons and fermions (each polarization counted once) living on a circle of size L. The problem is to count the number of states with a given energy E and given total momentum P = N/R and see if the results agree with semiclassical thermodynamics of black holes. we should be able to treat the system in a canonical ensemble characterized by an inverse temperature β conjugate to the energy and a chemical potential α conjugate to the momentum. 5.

D-Branes and Black Holes 27 Since we have massless particles in one spatial dimension. TRL 1.2 Near-Extremal Thermodynamics from the Long String It is now clear that the statistical mechanics of this free gas in fact reproduces the thermodynamics of the near-extremal black hole. This would be the situation if we had Q1 D1-branes individually wrapping the x5 direction. When multiple branes are each singly wound. similarly. all the thermodynamic quantities can be split into left. with er = −pr . PL .and right-handed quanta. The relationships of Equation 62 become the relations Equation 52 if f L = 8πRQ1Q5. respectively. PR. since the length is now 2πQ1Q5R. they can be either right-moving. SR may be read off from Equation 59. SL = 2π Q1Q5 (N + n) and SR = 2π Q1Q5 n.and left-moving modes. the physical temperature T agrees with the Hawking temperature. It might appear that thermodynamics does not distinguish between these two configurations. This is the content for the long string. The various quantities EL . The corresponding distribution functions then become ρR = 1 e(β+α)er ±1 and ρL = 1 e(β−α)er ±1 . we would have ∆E = 2n/Q1 Q5 R. a near-extremal configuration may be obtained from an extremal configuration by adding n pairs of left. Lπf πf L (62) 5. Another way to satsify Equation 63 is to take L = 2πQ1Q5 R and f = 4. TL TR (61) In fact. (60) Thus. Thus. However. Recall that the size of the x5 circle is 2πR. it is obvious that the thermodynamic relations of Equation 53 are also exactly reproduced.5. SL. the combinations TR = 1/(β + α) and TL = 1/(β − α) act as effective temperatures for the right. TR = 8ER 4SR = Lπf πf L TL = 8EL 4SL = . It follows from Equation 62 that this would require SL . SR f. Thus. ER. One condition for that is that TL L. as is clear from Equations 61 and 50. one way to satisfy the relation in Equation 63 is to consider L = 2πR and f = 4Q1Q5 . (64) There is an important difference between the two cases. with er = pr . (63) Given this agreement. resulting in 1 1 = T 2 1 1 + . In particular.and a right-moving pieces: E = ER + EL P = PR + PL S = SR + SL . (65) . From this it is apparent that for the multiple singly wound branes. the above thermodynamic relations assume that the system is extensive [30]. leading to an excitation energy ∆E = 2n/R. the energy of a single particle is n/R for some integer n. or left-moving. For the long string. where all the D1-branes join up to form a single multiply wound string.

then the Witten index would provide useful information.28 whereas for the multiply wound long string √ SL = 2π Q1 Q5N + n and SR = 2π n. We now discuss more rigorous results on the count of states that can be obtained by using the fact that we are counting supersymmetric states when we compute the entropy of an extremal black hole.6 A More Rigorous Treatment for Extremal Black Holes In the above discussion we adopted a simple model of the D1-D5 bound state and used it to compute the entropy of momentum excitations. In fact. such a chain of arguments was used in the first exact calculation of the microscopic entropy of a black hole by Strominger & Vafa [28]. It follows from the relations in Equation 53 that this requires (up to numerical factors of order one) ∆E > 1 . that the index can only give a lower bound on the number of supersymmetric states. with a minimum gap of 1/Q1Q5 R. whereas it is violated by the singly wound branes when Q1. the states we want carry momentum and thus a nonzero total energy. The ground states are maximally supersymmetric states. .) We need a generalization of the Witten index that will carry nontrivial information about the count of excited states of the system. we will again obtain a lower bound to the state count for the given quantum numbers. (In fact. so if one were interested in states preserving maximal supersymmetry. In the black hole context. however. the temperature of a system must be such that the specific heat is larger than one. states can appear and disappear in Bose-Fermi pairs. we are interested not in the ground states of the system but in states preserving some fraction of the supersymmetry. Note.) (A): We count not the actual states of the system but an “index. This is exactly what the long string provides. (B): Let us consider the black hole discussed by Strominger & Vafa [28]. 5. This index counts the number of bosonic ground states minus the number of fermionic ground states. Take Type IIB theory but let the complete 10-dimensional spacetime be K3×S 1 ×M 5 . RQ1 Q5 (67) which indicates that the system should have an energy gap 1/Q1Q5 R. As we change the parameters of the theory. The multiply wound string has much lower energy excitations.” The simplest example of an index is the Witten index of a supersymmetric system Tr[(−1)F e−βH ] (F is the fermion number and H is the hamiltonian). Das & Mathur (66) It is then clear that the extensivity condition (Equation 64) is always satisfied for the long string. Q5 and N are all comparable and large. We outline the main ideas of such a treatment here. This is consistent with the conditions for the validity of statistical mechanics for black holes [34]. The point is that for statistical mechanics to hold. we then hope that the bound will approximate the actual state count well enough to permit a meaningful comparison to the Bekenstein entropy. (This subsection assumes some familiarity with dualities in string theory. since some states are counted with a negative sign. Of course. but the index is a robust quantity.

and then we note that there is a duality map that relates Type IIA compactified on K3 to heterotic string theory compactified on T 4. The possible charge states are characterized by points on a lattice Γ21. and the T 4 is now replaced by K3. and 4-dimensional cycles on which something can be wrapped. and K3 along with the circle S 1 makes up the five compact directions. just as we did in the case of T 4 compactification. the product of winding and momentum charges in any one direction. Type IIA compactified on K3 × S 1 maps to heterotic string theory compactified on T 5 . (C): Since we will be looking for a quantity that will not vary with the continuous parameters of the system. This invariant is a generalization of np nw . What kind of charges can we place on the compact directions to make the black hole? In analogy to the discussion above of the black hole in T 4 × S 1 × M 5 . The only relevant property of these charges is that a bilinear invariant. We need some way to describe the charges carried by the branes. It is possible to count the degeneracy of the ground states of the wrapped branes and infer that the target space of this sigma model must be essentially the symmetric product of Q2 /2+1 copies of K3 (Q2 as defined is an even number). The charges that are dual to the wrapped branes in Type IIA now arise in the heterotic theory from momentum and winding of the elementary heterotic string on the compact directions of the theory. 3-. Here F F the term “essentially” means that this target space may be a deformation of the stated one. longer than any length scale that existed in the sigma model. and the space would remain hyper-K¨hler. which we may call Q2 .D-Branes and Black Holes 29 where M 5 are the noncompact directions. each point on each strand could move on T 4 in that case. Type IIB theory has 1-. but the topological properties of the space would remain unchanged. If we hold np fixed and take the size of S 1 to infinity.5 above. since the two kinds of charges get interchanged. The fact that we take a symmetric product amounts to saying that the strands are identical and must be so treated in any quantum mechanical count. We now consider np units of momentum along the S 1 direction of this sigma model. The . 2-. The Q2 /2 + 1 copies of K3 are F analogous to the Q1Q5 strands of the string that we took as the vibrating elements in the heuristic treatment of Section 5. so we can wrap all these branes on the K3 × S 1 space as desired. K3 is a special kind of compact 4dimensional surface. we can take the size of the S 1 to be much larger than the size of the K3. there is really no difference between Type IIA and Type IIB theories here. we can wrap various branes around the cycles of the compact directions. K3 has 0-. such a product is invariant under T-duality. As mentioned above. with the space direction being the S 1 cycle. and 5-dimensional branes.5. gives the invariant F length squared of a charge vector in this space. we can view the theory as Type IIA on K3 × S 1. which means a lattice with 21 positive signature directions and 5 negative signature directions. Since one compact direction is S 1. Then we can put a momentum charge along the S 1 direction and obtain a black hole carrying a momentum charge Np along with the charges arising from the wrapped branes. On the other hand. then we will be looking at modes of very long wavelength. which means that the number of sua persymmetries would also remain the same. making sure that they all stretch along the S 1. and here the charges are easy to understand. Then the physics of the wrapped branes looks like a (1 + 1)-dimensional sigma model. we can go from one to the other by a Tduality.

the degeneracy of states at an excitation level n above the ground state (in. If we do not excite the right movers and let the left movers carry the given momentum.30 Das & Mathur physics of such modes must be a conformal field theory (CFT) in 1+1 dimensions. If we have such a conformal theory with N = 2 or larger supersymmetry. and there is a cancellation in the elliptic genus and we get no useful information about the state count. This quantity is nonzero. we find that the elliptic genus is zero. The dimension of the target space must be of the form 4k. and its corresponding fermion gives an extra 1/2. and in particular we can separate the excitations into left. and the elliptic genus can be explicitly written. Each bosonic direction contributes 1 to the central charge. We want only the count of states when the charges are large. This happens because there are fermion zero modes on the torus. which relate a set of bosonic states to an identical spectrum of fermionic states. .and right-moving modes. d(n. Thus c = 6k. In a CFT. The derivation of this relation is combinatorial. One can relate the elliptic genus when the target space is a symmetric product of k copies of the space X to the elliptic genus when there is only one copy of the space [35]. one symmetrizes over bosonic states and antisymmetrizes over fermionic states. which is a simple CFT to handle. Here JL and JR are integer-valued U (1) charges for the left and right sectors. we get S = ln d(np . these are the U (1) symmetries that rotate the two supercharges of N = 2 supersymmetry into each other. (D): Although the above expression was derived for large np . If we apply the above method to the case where the compact space is T 4 instead of K3. Maldecena et al showed [36] that one can use the fact that we have N = 4 supersymmetry in the CFT to argue that the quantity 2 Tr[(−1)JL −JR JR e−βH y JL ] is a topological invariant in this case. since only then can we compare the count with the Bekenstein entropy computed from the classical gravity solution. the left sector) is given in terms of the central charge of the CFT by the formula nc 2π 6 . then we will have states with half the maximal supersymmetry: The right supercharges will annihilate the state and the left ones will not. 2 which agrees with the Bekenstein entropy for the classical geometry with the same charges. Putting in the values of k and n = np . then it can be shown that the elliptic genus Tr[(−1)JL −JR e−βH y JL ] is also a robust quantity that does not change when the moduli are varied. c) ∼ e The central charge for a supersymmetric sigma model with a hyper-K¨hler space a is easy to calculate. QF ) ∼ 2π np Q2 F +1 . and it turns out to give for large charges a state count that reproduces the Bekenstein entropy. it is also possible to compute the elliptic genus for the symmetric product of copies of K3 without making any approximation. say. But the case where the target space is only one copy of X is the theory of a single string on X.

4) supersymmetry when viewed as a 1-dimensional theory living along x5 . 5) strings along the transverse directions x1 · · · x4 form a “vector multiplet. 3. Because of the well-known ’t Hooft anomaly matching conditions. this coefficient is the same at low energies. 5) strings and has (4. 5). 1). This is known as the Higgs branch of the theory. the anomaly coefficient k may be easily computed in terms of the number of hypermultiplets and vector multiplets. the vector multiplets typically acquire a mass. (5. It has been pursued elsewhere [37. we are interested in the longdistance limit of the theory of the D1-D5 system. The leftmoving fermions now transform as (2. and (1. The states of the (1.” There are thus four scalars transforming as (2. 1) strings lead to hypermultiplets whose transformation properties under R symmetry are the same as those of the previous hypermultiplets. by superconformal symmetry. they are not all independent of each other but are related by a set of conditions.” This also has four scalars transforming trivially under the R-symmetry group and has four fermions. 2) of R symmetry. the U (1) parts of the fields are decoupled from the SU (Q1 ) × SU (Q5 ) part and can be ignored in this discussion. Nevertheless. 5) strings with polarizations along the T 4 directions form a “hypermultiplet. although a detailed low-energy theory has not yet been rigorously derived. We want the branes to be on top of each other. where the theory flows to 2 A U (1) component of the vector multiplet remains massless. 2) of the R symmetry group and four fermions. The left-handed fermions transform as (1. 38]. As a result. There is a global R symmetry which rotates the various supercharges: SO(4) = SU (2)L × SU (2)R.D-Branes and Black Holes 31 5. The R-symmetry current is anomalous because the left. As explained above. However. this is a complicated affair. and it turns out to be k = Q1Q5 . it is possible to obtain an important universal quantity of the low-energy conformally invariant theory—the central charge. 2. However. The states of the (1. The lowest oscillator modes of the open strings lead to a U (Q1 )×U (Q5) gauge theory with the following supersymmetry multiplets: 1. Even at the classical level.2 The hypermultiplet scalars are massless.7 The Gauge Theory Picture We now show how the above picture of a nonlinear sigma model follows from gauge theory considerations. This requires the vector multiplet scalars to have vanishing expectation values. to the anomaly of the Rsymmetry current. 4) supersymmetry. 2). representing the overall motion of the branes in transverse space. This system has (1. 1). In fact. The states of the (1. 1) and (5. At high energies where the theory does not have superconformal invariance but does have (4. 5) and (5. This is because the central charge is related. . 1) and (5. since these represent the relative distances between individual branes in the transverse space. 1) and the right-moving fermions as (1. It is necessary to solve for the independent set of fields and find the low-energy theory of these fields. the scalars in the hypermultiplet will have expectation values. this R symmetry is nothing but the rotation symmetry in the four transverse directions. and the right-handed fermions transform as (2. On the other hand.and right-moving fermions behave differently under R-symmetry transformations.

The asymptotic density of states in a conformal field theory can be now determined in terms of the central charge. Superconformal symmetry then determines the central charge of the superconformal algebra in terms of the anomaly coefficient c = 6Q1 Q5. In the orbifold limit.g. where once again there is an exact agreement of the extremal entropy. This is the sigma model discussed in the previous section. the supergravity approximation is not valid. one can get rid of the Coulomb branch altogether. D1-branes in the D5-branes can be considered as “instanton strings” of the 6-dimensional U (Q5 ) gauge theory on the D5-brane [39]. the independent moduli are obtained by solving a set of equations identical to the equations that determine the independent set of hypermultiplet scalars.and 4-dimensional black holes also provide exact microscopic derivations of the extremal entropy [48]. The number of D1-branes is equal to the instanton number Q1 . In fact. and for BPS states this leads to the expression for entropy given above. These strings can be singly wound or multiply wound.32 Das & Mathur a superconformal theory.8 Other Extremal Black Holes There is a wide variety of black hole solutions in string theory (for reviews and references see e. . this is the entropically favored configuration. Near-extremal limits of these black holes have also been studied. From the point of view of the 6-dimensional theory. the collective coordinates are functions of x5 and x0 —their dynamics is thus given by a (1 + 1)-dimensional field theory that is essentially a sigma model with the target space as the moduli space of these instantons. and numerous studies have related them to states in string theory. Instantons are such configurations. which are self-dual in this 4-dimensional sense. Rotating versions of such 5. there is. Reference [44]).” The low-energy excitations of the system are described by the collective excitations or dynamics on this moduli space. 41]. forcing the branes to lie on top of each other. The “long string” used in the previous sections is nothing but the maximally wound sector. which is why one has to turn on some of the supergravity moduli fields to go to the correct regime. a natural analog of the long string approximation in which these black holes are regarded as intersecting branes in 11-dimensional supergravity [47]. One notable example is a 4-dimensional black hole with four kinds of charges [45]. There is a slightly different but equivalent approach to the problem. Actually these are not really instantons but rather solitonic objects. Such configurations may be regarded as configurations of a Euclidean 4-dimensional gauge theory living in the x6 · · · x9 directions. they are thus solitonic strings. In fact. Because of translation invariance of the original configurations along x5 and x0. These instantons form a continuous family of solutions with the same energy—the corresponding parameters form the “instanton moduli space. in fact. Consider configurations of the gauge theory that are independent of x5 and time x0 . thus ensuring that we are dealing with the Higgs branch of the theory [42. As we have argued. however. by turning on a NS-NS two-form field. 40] that are wound around the x5 and move along the T 4 with a total winding number k = Q1 Q5 . The behavior of the system with such moduli turned on has been further analyzed elsewhere [43]. the theory of these instanton strings may be interpreted in terms of a gas of strings [35. It turns out that in the orbifold limit. 5.

the processes of absorption and radiation appear to be rather different. as long as the extremal limit corresponds to a large black hole. leading to a complicated set of equations. It is clear that the black hole will evaporate until the excess mass over extremality has been radiated away. Note that for such black holes. We are interested in this low-energy limit. However. Then the wave equation satisfied by a test field in the classical background φ0 can be obtained by substituting the perturbed field φ = φ0 + δφ in the supergravity equations of motion and retaining the part linear in δφ. When the closed string hits the D-brane. whereas the open strings may be replaced by quanta of the low-energy gauge theory living on the brane. For the 5-dimensional black hole we are discussing. semiclassical approximations are valid for these large black holes. which have negative specific heat). so that this evaporation is a welldefined process. However. black holes absorb matter via standard wave scattering by an effective potential arising from the gravitational field. the radiation rate can be obtained by considering a black hole in equilibrium with its environment of radiated quanta. As we shall see. Consider a closed string incident on the intersecting D-brane configurations representing black holes. leaving an extremal black hole that is stable. the specific heat is positive (in contrast to neutral black holes. which can then escape from the brane system. there is a finite probability that the closed string quantum will be absorbed by the black hole. and the absorption cross section can be calculated by solving the relevant wave equation in the black hole background. 6. The principle of detailed balance then relates the radiation rate to the absorption cross section by the relation in Equation 9. Consider the absorption of some field. these have the largest absorp- . The dominant decay mode is the emission of massless neutral particles. that satisfy decoupled massless minimally coupled Klein-Gordon equations. In the microscopic models described above. Hawking radiation is a quantum process. called minimal scalars. the closed string that is propagating in the bulk of spacetime may be replaced by the quantum of some supergravity field.1 Classical Absorption and grey-body Factors To compare the absorption or radiation rates of semiclassical black holes and D-brane configurations. The equations of motion would generically couple the various perturbations. one would expect that at all stages of the decay process. which should be calculated using standard rules of quantum mechanics. various symmetries of the background classical solution allow us to isolate special perturbations that satisfy simple decoupled equations. We are interested in the low-energy behavior of these absorption cross sections. At low energies. Let φ denote all the supergravity fields collectively. An incident wave is partly scattered back and partly absorbed. it can either reflect back or split up into open strings whose ends are attached to the brane. we must compute the classical absorption cross sections of the relevant fields in the black hole background. there are 20 such bosonic perturbations.D-Branes and Black Holes 33 6 BLACK HOLE ABSORPTION/DECAY AND D-BRANES In classical gravity. Thus. The open strings then do not leave the brane and hence do not re-emerge in the asymptotic region. Radiation is the reverse of this process: Open strings can join up to form closed strings. We are interested in the decay of a slightly nonextremal black hole by Hawking radiation. Furthermore.

Γ( p+1 ) H 2 (71) where AH is the horizon area and the last equality follows from the form of the metric (Equation 68). Limω→0 σ(k) = 2π (p+1)/2 p R = AH . the swave cross section of minimally coupled massless scalars is always the area of the horizon [51]. the equation becomes that of a scalar field in flat background and may be solved in terms of Bessel functions. In the outer region ωr > λ. (69) (70) Let l denote the largest length scale in the classical solution. In the asymptotic region.34 Das & Mathur tion rate and hence dominate the Hawking radiation. The actual calculation of the cross section involves solving the wave equation analytically in various regimes and then matching solutions in the overlapping regions. We must then match the two solutions in the overlapping region. We then solve the equation to lowest order in λ = ωl. As an example. Consider a metric of the form ds2 = −f (r)dt2 + h(r) dr2 + r2 dΩ2 . Examples of such fields are traceless parts of the components of the 10-dimensional metric. we describe the essential steps that led to the result obtained in [51]. In the near-horizon region ωr < λ. 33]. which have polarizations along the T 4 in the (x6 · · · x8) direction. It was soon realized. The frequency component φω (r) satisfies [(rpF (r)∂r )2 + ω 2 R2p(r)]φω (r) = 0. through φ(r. The physical input is a boundary condition. there is both an incoming and an outgoing wave. . This can be converted into an absorption cross section σ(k) for a plane wave of momentum k using standard techniques. that this is in fact a universal result: For all spherically symmetric black holes in any number of dimensions. p (68) where dΩ2 denotes the line element on a unit p-sphere. A similar calculation for the near-extremal 5-dimensional black hole described above also showed that the leading term of the absorption cross section is again the horizon area [50. and matching the solution to the near-horizon solution yields the ratio of the outgoing and incoming components and hence the absorption probability. φω (r). Scattering of massless particles by 4-dimensional black holes was studied in the 1970s [49]. however. The result is. as advertised. where F (r) = {f (r)[h(r)]p−1}1/2 R(r) = r[h(r)]1/2. we decompose the field into fields of definite frequency. t). t) = e−iωt φω (r). Using time translation invariance. and the low-energy limit of the s-wave absorption cross section for massless minimally coupled scalars was found to be exactly equal to the horizon area for Schwarzschild and charged Reissner-Nordstrom black holes. the function R(r) → RH = R(rH ) may be replaced and the equation may be solved after a simple change of variables. It is necessary to impose a boundary condition so that there is no outgoing wave at the horizon. Let the horizon be at p r = rH so that f (rH ) = 0. We want to solve the minimally coupled massless equation 2φ = 0 for a spherically symmetric scalar field φ(r. Then there are two regions in which the equation can be easily solved.

Writing the relevant wave equations. the absorption cross section for such minimally coupled scalars may be calculated in the so-called dilute gas regime defined by r0 r1 ∼ ∼ O(1) (72) r0. one has TH ∼ 2 TR and the cross section becomes exactly equal to the area of the horizon. Out of these 20 scalars. Finally. When these modes have their orbital angular momentum. As discussed above. TR (73) The various parameters of the classical solution have been defined in previous sections. let us briefly discuss the absorption of higher angular momentum modes and higher spin fields. However. even though it is a result of a solution of the Klein-Gordon equation. rN r1. In this subsection. The exceptions are modes of spin-1/2 fields that are supersymmetric partners of the minimal scalars. We want to figure out how these degrees of freedom couple to the supergravity modes in the bulk. we are mostly interested in supergravity modes that are minimal scalars in the dimensionally reduced theory. the index I referring to the four directions of the T 4 along x6 · · · x9 and their corresponding fermionic partners. The expression (Equation 74) is called a grey-body factor because of the nontrivial energy dependence. The final result is [52] 2 2 σ(ω) = 2π 2r1 r5 πω (eω/TH − 1) . the cross section appears as a combination of thermal factors. Once again one may hope that weak coupling calculations could give reasonable answers for the same reasons that they succeeded in predicting the right thermodynamics. One might wonder whether the model may be used to compute decay rates.D-Branes and Black Holes 35 For the 5-dimensional black hole. the absorption cross section is twice the area of the horizon. In fact. we find for such modes an additional “centrifugal” potential near the horizon that suppresses absorption of these modes by the black hole at low energies [49]. The theory of the long string is a (1 + 1)-dimensional massless supersymmetric field theory with four flavors of bosons φI . r5 r5 rN when the energy ω is in the regime ωr5 1 ω ∼ O(1) TL ω ∼ O(1). 6. and indeed weak coupling calculations for these modes agree exactly with the semiclassical results described in the previous subsection. we show that this is indeed unambiguously possible for a certain set of modes. we have a fairly good effective theory—the long string model—and weak coupling calculations in the model give exact strong coupling answers. 2 (eω/2TL − 1)(eω/2TR − 1) (74) Remarkably. the gauge theory is difficult to analyze even in weak coupling. When TL TR 1 and ω TL. it is particularly easy to write down the interaction of the long string with the . in agreement with expectations from supersymmetry.2 D-brane Decay A first-principles calculation of the decay of a slightly nonextremal 5-dimensional black hole would involve a strongly coupled gauge theory of the D1-D5 gauge theory coupled to the bulk supergravity fields.

In Equation 76. This corresponds to the decay into neutral scalars from the noncompact 5-dimensional point of view. p5) [along the (x0 . The 10-dimensional gravitational coupling. one of which is left-moving and the other right-moving (see Figure 3). is related to the string coupling and string length by (77) √ The factor 2κ10 in Equation 76 is determined by requiring that the field hIJ be canonically coupled and follows from the bulk supergravity action (Equation 18). When there is no supergravity background. This form of the action follows from the principle of equivalence. If we expand the metric as √ (76) gIJ = δIJ + 2κ10hIJ (ξ. This is because the free kinetic term of the fields φI has to be canonically normalized. φI .36 Das & Mathur transverse traceless components of the 10-dimensional graviton hIJ . the action should be a free action given by Equation 75 with gIJ replaced by δIJ . The amplitudes that result from this action are therefore universal. This requires that at low energies the action in the presence of a background is given by Equation 75. which depends on the transverse coordinates only through the radial variable. Because we are working in the limit where the radius R of the x5 circle is much larger than the size of the T 4 . J lying along the T 4. There will be other terms in the action. we give the basic steps for calculation of the decay rate when the mode hIJ carries no momentum along x5 and is in an s-wave in the transverse space. the T 4 coordinates φI (which are the fields of the long string theory). normalization thus completely removes the constant T from the action. x1 · · · x4 ) we immediately get the coupling of hIJ with the long string degrees of freedom. In the following. 10 . However. It appears that the action we have written has little predictive power unless we figure out these details. Consider the annihilation of two long string modes: (a) one with polarization I and momenta p = (p0. T is a constant. Furthermore. Because the interaction is also quadratic in φI . it has been recognized [33] that the cross section in question does not depend on such details. x5). The constant T is an effective tension of the long string and depends on the details of the rather complicated D1-D5 bound state. (75) where the ξ α denote the coordinates on the long string world sheet and gIJ is the metric on the T 4 . but these would contain more derivatives and are suppressed at low energies. κ10. x5) direction] and (b) one with polarization J and momenta q = (q 0. the relevant action is T d2ξ∂α φI ∂ αφJ gIJ . we are interested in computing s-wave absorption/decay. It is clear from Equation 75 that the decay of a nonextremal state into the mode hIJ is dominated by the process of annihilation of a pair of modes of vibration from the long string. q 5) into an hIJ bulk mode with momenta 8 κ2 = 64π 7g 2ls . with indices I. we have made explicit the dependence of the supergravity field on the coordinates on the long string world sheet ξ = (x0 . we can ignore the dependence of hIJ on φI . At low energy. The fields φI transform as vectors under the local SO(4) transformations of the tangent space of the T 4 . and the coordinates transverse to the black hole (x1 · · · x4).

(81) We have expressed the answer in terms of the parameters in the classical solution by using the length of the effective string. (79) where ρ(p0. k1. The integral may be evaluated easily and the answer is 2 2 Γ(k) = 2π 2r1 r5 πω 1 . p5)ρ(q0. we find that the decay rate for this process is Γ(p. the decay rate is converted into an absorption cross section by multiplication by an inverse Bose distribution of the supergravity mode: σ(ω) D−brane = 2 2 πω 2π 2r1 r5 2 (e TL − 1)(e TR − 1) ω eT − 1 ω ω . · · · . q. Because these states are drawn from a thermal ensemble. L = 2πQ1Q5R = 2 2 8π 4r1 r5 V R . where V4 is the volume of the noncompact four spatial dimensions x1 · · · x4. the decay rate is Γ(k) = ( L 2 ) 2π ∞ −∞ dp5 ∞ −∞ dq5 ρ(p0. ω ω 2 (e 2TL − 1)(e 2TR − 1) (80) Finally. q. k = (k0. p5) denotes the Bose distribution functions discussed in the previous section. k4. κ2 (82) . To obtain the total cross section for production of a scalar. x5) directions. Using Equations 75 and 76. 0) along the (x0. q5)Γ(p. k).D-Branes and Black Holes D5 37 D1 Figure 3: Absorption of supergravity mode by D-brane. one must now average over all initial states. (2p0L)(2q0L)(2k0V LV4) (2π)4 (78) The delta functions impose energy conservation and momentum conservation along the x5 direction (momentum is not conserved in the other directions because of Dirichlet conditions). The denominators come from normalization of the modes: the open string modes are normalized on the circle along x5 of circumference L = 2πQ1Q5R whereas the closed string field is normalized in the entire space with volume V LV4. · · · . k) = (2π)2Lδ(p0 + q0 − k0)δ(p5 + q5 − k5) V4d4k 2κ2 (p · q)2 .

we can have absorption but no Hawking radiation. r5 in terms of g. However. and the thermal factor for the right-moving mode provides the thermal factor of the emitted supergravity fermion—and this is a Fermi-Dirac factor as expected [31]. these can never collide to give rise to modes with nonzero angular momentum in transverse space. where κ2 = κ2 /2πRV and κ2 is given by Equation 77. The fact that the cross section turns out to be proportional to the horizon area was already known [31]. 5 5 10 10 We have also used Equation 49 to express r1. . e. However. but for a different reason. Even before taking this low-energy limit. Therefore. RQ1Q5 (83) We have used Equation 52 and the fact that in the regime TL TR we have SL = S = SBH = 2πAH /κ2 . In the strict classical limit. Thus. when TL TR. Q5. Classical emission of higher angular momentum modes is suppressed by centrifugal barriers provided by the gravitational field. we have T ∼ 2TR. multifermion processes are responsible for emission of higher angular momentum modes. Fermions on the long string. In a similar way. Angular-momentum–mode emission is not completely understood in the long string picture—although the qualitative aspects have been understood [54]. carry transverse tangent space indices (typically as a R-symmetry index) since they are dimensional reductions of 10-dimensional fermions. Similar calculations have been performed for several other situations.and 4-dimensional black holes [53. As a result. ls. In the microscopic model. Note that the bosonic degrees of freedom of the long string are all vectors under the internal SO(4) of the T 4 but are scalars under the tangent space SO(4) of the transverse space. so the corresponding thermal factor suppresses emission completely while absorption is still nonzero [50]. Q1. For very low energies ω TL . 52]. Thus. since they involve more than two fermion fields and therefore correspond to higher-dimension operators on the long string world sheet. rather than diverging as Bose factors do. The supergravity mode can also split into a pair of fermions. then the thermal factors that appear in Equation 80 are Fermi-Dirac distributions and these go to a constant at low energies. This is exactly the low-energy classical result. Following the above calculations. the corresponding cross section is suppressed at low energies. V. The precise calculation outlined above was performed in Reference [33]. it is apparent that the thermal factor of the left-moving long string mode gives the right powers of energy to lead to a nonzero low-energy cross section. however. the entire result (Equation 81) is in exact agreement with the semiclassical grey-body factor [52]. we see that in this limit σ(ω) = 2π 2(r1r5 )2TL = (r1r5)2 SL = AH .g. we know that the quantities TL and TR above are in fact exactly the same as the semiclassical quantities and T = TH . these can be seen to be necessarily suppressed. However. this comes about because the temperature in this limit is zero. It is clear from Equation 80 that the thermal factor in the decay rate comes from the thermal factors of the long string modes. This would require one left-moving bosonic mode of the long string with a right-moving fermionic mode. Using Equations 49 through 51. we may consider emission/absorption of a fermionic supergravity mode. such emission is again suppressed. for absorption/emission of charged scalars from both 5. one of the thermal factors above can be simplified.38 Das & Mathur The physical temperature T is related to TL and TR by Equation 61.

however. several exceptions to this rule.g. equivalently. · · · . unitary quantum decay process. however. and dΩ5 is the measure on a unit . In this section we consider one of them—the 3-brane in Type IIB supergravity. x9) are the coordinates on the brane. However. 7. for which semiclassical results cannot be trusted. 6. but that case involves BPS states and supersymmetry guarantees the agreement. One might wonder whether this DBI action can be used to predict absorption rates for other supergravity modes. Perhaps more significantly. The agreement of nontrivial grey-body factors is even more surprising because now not just a single number but an entire function of energy is correctly predicted by a weak coupling calculation. Weak coupling calculations also gave the correct answer for the extremal entropy. surprising— like the agreement of near-extremal thermodynamics—since our weak coupling calculation agreed with a strong coupling expectation. x = i=1 (x7. Extremal solutions with vanishing horizon areas generally have singular horizons. If we trust the DBI action to obtain the couplings of these fields. This is. for small amounts of nonextremality. decay rate—is strong evidence for the contention that Hawking radiation is an ordinary. there is a set of other scalar fields that do not obey minimal Klein-Gordon equations— e. It turns out that this can be done for several other minimal scalars. In nonextremal situations. the couplings of which cannot be obtained from the DBI action.D-Branes and Black Holes 39 The low-energy effective action used above follows from a Dirac-Born-Infeld (DBI) action of the long string in the presence of supergravity backgrounds. namely the dilaton and the components of the NS B field along the T 4 . There is also some evidence for nonrenormalization theorems of the type required [57]. There are. Explicit one-loop calculations [56] seem to support this scenario. we have small excitations over a supersymmetric background and we may still expect nonrenormalization theorems to ensure that higher-loop effects vanish at low energies. 5 (84) where r = ( 6 (xi )2)1/2 is the radial coordinate in the transverse space. we get answers that do not agree with supergravity calculations [55]. A complete understanding of this is still lacking. There are four other minimal scalars in this background. These cross sections are suppressed at low energies and it is probably not surprising that the simplest low-energy effective theory does not work. supersymmetry is broken. the fluctuations of the volume of the T 4 .3 Why Does It Work? The agreement of the low-energy limit of the absorption cross section—or.1 Classical Solution and Classical Absorption The classical solution for N parallel extremal 3-branes along (x7 · · · x9) is given by the (Einstein-frame) metric ds2 = [f (r)]−1/2[−dt2 + dx2 ] + [f (r)]1/2[dr2 + r2dΩ2]. 7 ABSORPTION BY THREE-BRANES The above discussion has addressed extremal solutions with large horizon areas and their excitations.

It is possible to calculate the classical absorption cross section of various supergravity modes by the extremal 3-brane. x. in terms of coordinates z = R2/r. gs r5 (86) The horizon is at r = 0 and the horizon area is indeed zero. ds2 = R z 2 2 [−dt2 + dx2 + dz 2] + R2 dΩ5. the spacetime is completely nonsingular. the first factor being composed of (t. The harmonic function f (r) is given by f (r) = 1 + R r 4 4 R4 = 4πgs N ls . It turns out that in this case one can also analytically calculate the low-energy absorption cross section of such scalars for arbitrary angular momentum [59]. since the horizon area vanishes. Das & Mathur (85) The dilaton is a constant and the only other nontrivial field is the 4-form gauge field with a field strength F0789r = 4R4 [f (r)]−2. 7. 8 32π (88) where we have used Equation 85. (87) This is the metric of the space AdS5 × S 5 . It is the dimensional reduction of N = 1 supersymmetric Yang-Mills in ten dimensions. For example. Thus.” where the length of the throat is along the AdS5 coordinates and the cross sections are 5-spheres of constant radii R. The cross section of course vanishes at ω = 0. As we will see below. if we consider modes like the dilaton and longitudinal traceless parts of the 10-dimensional graviton. However. near the horizon the spacetime approaches a “throat geometry. so we have absorption but no Hawking radiation. z) while the sphere S 5 has a constant radius R. the longitudinal graviton has to couple to the energy-momentum tensor of the Yang-Mills theory. This is thus the regime where semiclassical supergravity is valid.2 Absorption in the Microscopic Model The microscopic theory of N 3-branes is a U (N ) gauge theory in 3 + 1 dimensions with 16 supersymmetries. the metric (Equation 86) becomes. Also note that we are dealing with an extremal solution that has zero temperature. For s-wave absorption of minimally coupled scalars. However. or N = 4 supersymmetry. whereas the sphere S 5 has a constant positive curvature.40 5-sphere. Thus. From Equation 85 it follows that this curvature is much smaller than the string scale when gs N 1. This is a well-studied theory and is known to be conformally invariant with zero beta function. In the near-horizon or near-brane region r R. a natural symmetry argument determines the coupling of some of the supergravity modes to the brane degrees of freedom. one can perturbatively . the coupling can be once again read off from the equivalence principle. the leading-order result is [58] σ= 1 1 4 8 3 π R ω = (κN )2ω 3. The AdS5 has a constant negative curvature 1/R2.

The imaginary part of the two-point function in Euclidean space is in turn related to the absorption cross section. and N is the same as in semiclassical thermodynamics of the corresponding black brane. Thus. It is thus natural to expect that the microscopic description of this is in terms of a Yang-Mills theory at finite temperature. At weak coupling. This description was at weak coupling. we expect the emitted string to interact several times with the branes before escaping to infinity as radiation. in the process of emitting a closed string. From the viewpoint of the dynamics of open strings on the 3-branes. Again. a free gas). we know that there are N 2 degrees of freedom. a loop of open strings can be interpreted as a propagation of a closed string at tree level (see Figure 4). the precise coefficient differs by a factor of 4/3. as we increase the coupling. the reason for the agreement of the weak coupling calculation with supergravity answers is much better understood. However. volume. there was only one string interaction. Let us consider for simplicity the case of the D3-branes described above. where two open strings joined up to the emitted closed string.D-Branes and Black Holes 41 calculate the absorption cross section of such modes (which are minimal scalars from the 7-dimensional point of view) along the lines of the calculation in the 5-dimensional black hole. we encounter higher loop processes among the open strings. There are similar nonrenormalization theorems known for operators coupling to other modes as well.3 Nonextremal Thermodynamics The extremal 3-brane of supergravity described above is of course a limit of a general nonextremal solution with finite horizon area and finite temperature. an exact agreement was found for s-waves [58]. In the lowest-order calculation. it is significant that the dependences agree—although the dependence on the energy and the volume is dictated by scaling arguments (this is a conformally invariant theory). each hole looks like an open string loop. The thermodynamics of this Yang-Mills theory was calculated [61] at lowest order in the coupling constant (i. so the discrepancy is not unexpected. in order to describe absorption by the black hole of quanta with arbitrary energies. On the other hand. the dependence on N is not. so that the lowest-order result is exact [60].e. Each such interaction gives one “hole” on the string world sheet. with the boundary of the hole being constrained to the branes. 8 AdS/CFT CORRESPONDENCE AND HOLOGRAPHY We have presented a unitary description of black hole evaporation in terms of the emission of closed strings from the branes. At higher orders in the coupling. it is known that the two-point function of the energy-momentum tensor of N = 4 Yang-Mills theory is not renormalized. We know of no reason why the thermodynamic quantities are not renormalized as we go from weak to strong coupling. and the agreement was found to extend to all angular momenta [59]. Let us see what form this continuation to higher energies might take. and it was found that although the dependence of the thermodynamic quantities on the energy. In this case. 7. In particular. these open string multiloop . Thus. there is no known argument why this should continue at strong coupling. which is thus also protected from higher-order corrections. one must imagine that it is continued to strong coupling.

but take the limits of large N and small g in such a way that gN tends to a finite value. The effect of open string loops might well be counted in the propagation of closed string modes (which include gravitons that can be said to condense to generate the curvature). The quantity gN ∼ gYM N is the ’t Hooft coupling of the gauge theory. The t-channel exchange has a sequence of poles. the above considerations are somewhat heuristic. of the appearance of duality in tree-level string scattering. as well as the fact that the metric around the branes will cease to be flat? (This curvature is significant when the closed 2 string coupling gc ∼ go ceases to be small. Now consider energies of excitation that are low. as we saw above. should we take into account the complicated loop processes of the gauge theory (which are significant when the open string go coupling is not small). The dynamics of these open strings is.) The discussion above indicates that taking both of these effects into account might amount to overcounting. Maldacena considered the following limit. crudely speaking. and we confront the issue of whether these are two different effects to be added together in some way or just dual manifestations of a common underlying dynamics. 8.42 Das & Mathur Figure 4: Open string loop as a closed string exchange. we can study the large-N gauge theory . given by a (N = 4) supersymmetric 2 Yang-Mills theory. In the case of this string scattering. since in the domain of large coupling we are unlikely to have any good expansion in the number of loops. Let the string coupling be weak (g 1). so that on the D3-branes we get open strings but only in their lowest states. as does the s-channel scattering. of course. The arrows indicate the direction of time in each way of looking at the diagram. which. At small ’t Hooft coupling.1 The Maldacena Conjecture In the much more complicated case of the black hole evaporation process. Place together a large number N 1 of D3-branes. would attempt to retard the outgoing quantum. redshifting it to lower energies. But now we ask: As we increase the coupling in the microscopic calculation. Such a redshift is easily seen if we consider the propagation of the outgoing quantum in the metric produced by the branes. though. It is hard to answer this question from a direct consideration of loop expansions. Maldacena’s bold postulate addressed the relation between microscopic theory of the branes and the effects of gravity in 10-dimensional spacetime. The situation is reminiscent. it is now well understood that the s-channel poles result from an infinite sum over t-channel poles and are not to be considered as additional contributions to the S-matrix. Thus. But among such closed string exchange must be the effect of the gravitational effect of the branes. processes might be reinterpreted as closed string exchanges between the departing quantum and the branes.

These two theories were dual descriptions of exactly the same underlying theory. r. If the gauge theory of the branes and gravity on the 10-dimensional spacetime are indeed exactly dual descriptions of the same theory. The gauge theory is in four dimensions. all reference to D3-branes has disappeared. To do this. although. having an SU (4) symmetry of rotations among themselves. and the conformal group in four dimensions is indeed isomorphic to the 15-parameter group of isometries of AdS5 space. An important fact that led Maldacena to the duality conjecture was the agreement of global symmetries between the gravity theory and the gauge theory. there is no singularity that might represent the place where the D3-branes were placed in a gauge theory description. The energy density of this field yields the equal and opposite curvatures of the S 5 and the AdS5 spaces. In fact. In particular. But for gN 1 we are in a regime where the D3-branes produce an appreciable gravitational field around themselves. Thus. Maldacena postulated that we can consider either (a) the gauge theory on the branes (which is given by the open string interactions at a perturbative level) or (b) no branes. which are the translations and rotations on S 4 . it also has a constant value on the AdS5 . because of conformal . the space spanned by x. In this case. the multiplets obtained fall into representations of SO(6). If we had taken a collection of D3-branes as above. we need an operational definition of the dual map. But this space has a boundary at r → ∞. the radius of curvature in each case is R ∼ (gN )1/4ls (ls is the string length). it seems better to formulate the duality without reference to any D-branes. the N = 4 gauge theory has four supercharges forming. but only the near-brane geometry would have resembled AdS5 × S 5—at large r the space becomes flat. We take the 5-form field strength F present in Type IIB string theory to have a constant value with integral N on the S 5 . at large ’t Hooft coupling gYM N 1 such a perturbative analysis is not possible. Thus. Let us rotate the signature of the spacetime to a Euclidean one. as we approach the horizon the geometry becomes the throat geometry AdS5 × S 5 given by Equation 87. then we must have a way of computing the same quantity in two dual ways. then the flux of F on the sphere surrounding the branes would have been N . so both the theories have the same R-symmetry group. since this field must be self-dual. Further. But the N = 4 supersymmetric SU (N ) Yang-Mills theory is a conformal theory. Note that in this formulation of the conjecture. but the space AdS5 × S 5 and closed string theory (which has gravity as a low-energy limit) on this curved manifold. as we discuss below. and in fact distort the spacetime into the form of Equation 84. The gravity theory has an internal space that is S 5. when we decompose fields in harmonics on this space in the process of Kaluza-Klein reduction. a geometry that is AdS5 × S 5 everywhere (and not just at small r) is an exact solution of string theory without any branes present. The conjecture now states that string theory on this smooth AdS5 × S 5 is dual to N = 4 supersymmetric SU (N ) Yang-Mills on the 4-dimensional boundary of the AdS space.D-Branes and Black Holes 43 2 perturbatively. as shown above. AdS5 is a maximally symmetric space with a 15-parameter isometry group. But SO(6) ≈ SU (4). Thus. which is now Euclidean AdS5 . this makes the issues related to black hole information loss somewhat more difficult to access. which was provided by Gubser et al [62] and Witten [63]. which has the topology of S 4. has r = 0 as a regular point in the interior of the space. and thus to see manifestly the consequences of this duality. so one may expect a 10-dimensional set of symmetries. Further.

and in this case symmetries fix this operator to be T rF 2 . and here agreement is indeed found. this agrees with the 32 supercharges of the gravity theory on AdS5 × S 5 . The action on the left-hand side is a string action in general. the gravity calculation gives a prediction for the strongly coupled gauge theory. Thus. 63] and three-point functions of chiral operators [65] are found to have the same value when computed from the gauge theory or from the gravity dual using Equation 89. with supersymmetry intact. Let Ω be the AdS space. it is possible in principle to compute n-point correlation functions from supergravity and compare the results to calculations in the gauge theory. Nevertheless. and anomalous dimensions of composite operators (which can be deduced from the four-point functions) [67]. 8. supersymmetry protects the values of some quantities from changing when the coupling is varied.44 Das & Mathur invariance. anomalous dimensions of chiral operators [62. Using the above relation. The gravity theory is simple (perturbative classical supergravity to lowest order) when N is large (loops of gravity are suppressed by 1/N 2) and gN is large (because then the AdS space has a large radius. can be ignored to leading order). The field φ. the dilaton φ. In this case. and stringy corrections. and each value of the coupling gives a different well-defined theory). which depend on ls/R. If one considers the gauge theory at a finite temperature by compactifying the “time” direction. If we specify the value φb of φ on the boundary ∂Ω. The duality relation is e Ω −SSUGRA [φb ] = ∂Ω DA e−SYM + dξφb(ξ)T rF −SYM ∂Ω DA e 2 (ξ) . it does not. will be dual to an operator in the Yang-Mills theory. Examples of such predictions are the potential between external quarks placed in the gauge theory [66]. and only a weak coupling calculation would give the true glueball masses (the coupling flows with scale after supersymmetry is broken. then we can solve for φ in the interior of Ω and compute the action SSUGRA of supergravity on this solution. and ∂Ω its boundary. but at the leading order it can be replaced by the supergravity action. the gauge theory has 16 conformal supercharges in addition to its 16 regular supercharges. For quantities that are not protected in this fashion. which has a well-known treatment of the large-N limit but no simple way 2 to handle a regime with gY M N ∼ gN 1. then one obtains a nonsupersymmetric theory with a discrete spectrum of glueball states [68]. present in the string theory. on the other hand.2 Calculations Using the Conjecture Let us return to the issue of how to compare quantities in the two dual theories. four-point correlation functions. Let us consider for simplicity a scalar field. . (89) Here DA represents the path integral over the gauge theory variables. Thus it is difficult to compare results on the gravity side with explicit computations in the Yang-Mills theory. there is a surprising agreement of the qualitative features of the glueball spectrum between the gravity calculation and lattice simulations [69]. However. the strong coupling calculation has no direct physical meaning.

z0 Using the expression for R in Equation 85 and the fact that the 5-dimensional 8 Newton’s constant G5 is given by G5 ∼ G10/R5 ∼ (g 2ls )/R5. However. However. we have managed to decrease the entropy in this process and hence have violated the second law. We have seen that supergravity in AdS5 × S 5 is dual to a strongly coupled large-N Yang-Mills theory in 3 + 1 dimensions. Then we slowly add matter. Furthermore. The form of this bound appears surprising. Numerical constants are ignored in this relation. If this is true. We know that the entropy of a black hole is proportional to the area of the horizon measured in units of Newton’s constant. which then continues to grow until it fills up the box. Suppose we start with a state whose entropy is bigger than this bound.e. In fact. the area of the horizon is the same as the area of the box A. Thus. but whose energy is smaller than the mass of a black hole that fills the entire box. the total entropy S of anything in a large box of volume V and surface area A is bounded by A (90) S≤ . gravitational collapse takes over and a black hole is formed. This section explains how this correspondence illustrates another important property of holography that is in fact its very essence: the Bekenstein bound [71]. and this d-dimensional theory must have one degree of freedom per Planck area.3 Holography and the Bekenstein Entropy Bound The AdS/CFT correspondence has been studied extensively over the past two years [see [70] for review]. for this black hole. Then it turns out that supergravity in the region z > z0 is dual to a Yang-Mills theory with a position space ultraviolet cutoff δ = z0 . a (d + 1)-dimensional theory of gravity should be equivalent to a d-dimensional theory that lives on the boundary of space. renormalization group flows of the gauge theory become motion in the fifth dimension. i. ’t Hooft [72] and Susskind [73] have proposed a radical interpretation of the holographic bound (Equation 90). At some stage. the result follows from the existence of black holes.D-Branes and Black Holes 45 8. G5 z0 (92) Now the Yang-Mills theory has N 2 degrees of freedom per unit cell. It follows from Equation 87 that the physical area of this boundary is R3 (91) A ∼ 3 L3 . namely. the Bekenstein bound naturally follows. A general argument states that in any theory of gravity. we see that L3 A ∼ N2 3 . But suppose we impose an infrared cutoff at z = z0 . since in the absence of gravity we know that entropy is an extensive quantity and should scale as V . In the AdS5 ×S 5 metric (Equation 87) the boundary is at z = 0 and continuum Yang-Mills theory living on this boundary is dual to supergravity in the whole space. the right-hand side of Equation 90. the cutoff of . but the result for the entropy of near-extremal 3-branes shows that this must also be true in strong coupling. increasing the energy. G where G is Newton’s constant. Consider a large patch of this boundary of coordinate size L (in terms of the x coordinates). Is it then true that the latter has one degree of freedom per Planck area? This is indeed true [74]. This is manifest at weak coupling.

At the critical point. 8. where M 4 is either T 4 or K 3 on which the five branes are wrapped. This theory has a nonperturbative definition in terms of quantum mechanics of N × N matrices.5 A Suggestive Model of Holography Unfortunately. 8. Details of this development are supplied elsewhere [70]. The boundary theory is once again a superconformal field theory. such an approach has not succeeded. t) becomes a smooth field in 1 + 1 dimensions. Furthermore. at the critical point the loop diagrams of the matrix model generate discretizations of an oriented world sheet with dynamical world sheet metric and one scalar field. This shows that the gauge theory–supergravity duality provides a concrete demonstration of the Bekenstein bound. The usual quantization of noncritical string theory then converts the scale on the world sheet to an extra dimension in target space [78]. The AdS3 is made up of time. There is. Once again. in fact. resulting in a (1 + 1)-dimensional field theory.46 Das & Mathur this theory is precisely z0 . recently the holographic connection has been used to understand the properties of the D1-D5 system. Consistent quantization of the string forced us to consider a 10-dimensional su- . This spacetime is AdS3 ×S 3 ×M 4 [75]. the right-hand side of Equation 92 is indeed the number of degrees of freedom in this cutoff Yang-Mills theory. noting that when the matrix is large. One starts with the action ˙ S = dt[M 2 − V (M )]. t). Remarkably. The theory can be written in terms of ρ(x. 5D black holes obtained by putting in some momentum now become the Banados-Teitelboim-Zanelli (BTZ) black holes of 3dimensional gravity. 9 DISCUSSION The progress of string theory over the past few years has been truly remarkable. despite several attempts. though in the presence of a background that breaks the full symmetry. there is a critical form for the potential in which successive terms in the loop expansion are comparable. Furthermore. these two dimensions (which emerged in totally different ways) exhibit a local Lorentz invariance [77]. and x5 . In fact. It would be interesting if the gauge theory in 3 + 1 dimensions acquired extra dimensions in a similar manner to map onto string theory in 10 dimensions. the extra dimension x (which is the space of eigenvalues) becomes continuous and the collective field ρ(x. we do not yet understand the explicit map that relates the gauge theory variables to the gravity variables. in terms of appropriate variables. Thus. So far. the symmetries of the bulk agree with the symmetries of the conformal field theory. the transverse radial coordinate r. this is no longer a gauge theory but rather the low-energy theory of the D1-D5 system discussed in the previous sections.4 Near-Horizon Limit of 5D Black Hole Another example of the holographic connection is provided by the near-horizon geometry of the D1-D5 system. despite these agreements between the two dual descriptions. In fact. the density of eigenvalues of the matrix [76]. However. renormalization group flow on the world sheet is related to motion in this extra target space direction [79]. an earlier example of holography: (1 + 1)-dimensional string theory.

i. But it is fair to say that we do not yet have a good understanding of the information paradox. The paradox presents us with one way of computing Hawking radiation. The paradox then challenges us to point out which of the assumptions that went into the calculations were wrong. interaction with these microscopic degrees of freedom yields a unitary description of black hole absorption and Hawking emission for low-energy quanta from near-extremal holes (again this is the case in which we may compute with some confidence). and string theory yields the correct microscopic entropy of extremal and near-extremal holes (the only cases in which we know how to compute with some confidence). Further. In recent years Matrix theory [80] has attempted such a formulation. and this radiation process appears to be nonunitary. Literature Cited . although many solutions of the field equations appear equally acceptable at first sight. Regarding such solitons as elementary quanta turns out to be a duality transformation. Indeed holography. But we do not yet understand in detail the mechanism of this nonlocality. It now appears reasonable that there are nonlocalities in quantum gravity that can extend beyond Planck scale. In no other route that has been attempted for quantizing gravity has it been possible to arrive at a value for black hole entropy. using just the classical properties of the black hole metric and Gedanken experiments with low-density quantum matter. and we do not yet know which to choose to describe the low-energy physics of the world that we see around us. What about the information issue in black hole evaporation? The derivation of unitary low-energy Hawking radiation from microstates certainly gives a strong indication on the information question.e.D-Branes and Black Holes 47 persymmetric theory. and in particular the Maldacena conjecture. A correct quantum theory of gravity should reproduce this entropy by a count of allowed microstates. and string theory should be an exciting field in the years ahead. The entropy of black holes can be deduced from robust thermodynamic arguments. requires that the degrees of freedom in a volume be encoded by degrees of freedom at the surface of that volume. so we must give up naive ways of thinking about locality. via the foliation of spacetime depicted in Figure 1. There is no room for any free parameters in the theory. These gaps in our understanding are probably linked to our lack of an adequate language to understand the variables of string theory in a way that is not based on perturbation around a given background. as well as 11-dimensional supergravity. the power of supersymmetry and our understanding of the nonperturbative features of string theory have allowed us to extract more than just the perturbative physics of graviton scattering. and in particular we lack an explicit description in string theory of the foliation in Figure 1. thus suggesting that black holes can be unified into physics without having to modify basic principles of quantum mechanics or statistical physics. It thus appears that most of the progress in understanding quantum gravity and the structure of spacetime is yet to come. Within such a theory. one discovers branes as solitonic states. The results on black holes have provided a striking validation of this entire structure. and such duality maps unify all the five perturbative quantizations of the string. that information in the matter falling into the hole is not lost but is recovered in the radiation that emerges. but our understanding is still somewhat primitive.

Phys. Das SR. hep-th/9908105 (1999) 23. Mathur SD. Rev. B475:149 (1996). D 51:1764 (1995) 13. hep-th/9507048 (1995) 18. ’t Hooft G. Nucl. et al. UK: Cambridge Univ. High Energy Phys. Dijkgraaf R. B256:727 (1985) 12. Douglas M. Gauntlett J. Strominger A. 9911:002 (1999). Seiberg N. A10:2081 (1995) 28. 1. Tseytlin A. Phys. Wald RM. Commun. Horowitz G. Phys. B526:311 (1998) 39. B375:103 (1996) 30. UK: Cambridge Univ. Math. Olive D. Unruh WG. Rev. Phys. Dijkgraaf R. Lett. Rev. Lett. Nucl. Rev. Press (1998) 15. Ortiz M. Vols. New York: Freeman (1978). Lett. 43:199 (1975) 5. B383:151 (1996) 27. D 52:2176 (1995) 11. Moore G. Phys. Chicago Press (1984) 2. Dijkgraaf R. Maldacena J. Tseytlin A. Wald R. Lett. B478:561 (1996). A6:2353 (1991) 35. String Theory. Dhar A. hep-th/9508151 (1995) 8. Thorne W. Hassan SF. Nucl. Phys. Dvali G. Maldacena J. Hawking S. Keski-Vakkuri E. Lett. Phys. Brekenridge J. Mathur SD. Das SR. B482:153 (1996) 34. Lett. Rep. Maldacena J. Schwarz J. Phys. Vafa C. Larsen F. B478:199 (1996) 47. Nucl. 83:3370 (1999) 16. Nucl. Nucl. Peskin M. J. Nucl. B486:77 (1997) 41. Hassan SF. Susskind L. Strominger A. Das SR. Rev. Phys. Phys. Sundrum R. Sen A. Wheeler J. Mandal G. Rev. Cambridge. Rev. 77:428 (1996). Moore G. hep-th/9802051 (1998) 17. Phys. D 9:3292 (1974) 3. D 14:2460 (1976) 7. Verlinde E. Phys. Susskind L. Gravitation. Superstring Theory. B475:179 (1996). Phys. Dabholkar A. Lett. D 54:7391 (1996) 14. hep-th/9903063 (1999) 37. Larsen F. Phys. B475:679 (1996) 31. Polchinski J. Gibbons G. 77:2368 (1996) 33. Misner C. hep-th/9512077 (1995) 40. Maldacena J. Ruiz F. Witten E. Phys. Rev. Nucl. Lett. Lifschytz G. Phys. Hawking S. Strominger A. hep-th/9611050 (1996) 20. Phys. Susskind L. Nucl. Phys B543:545 (1999) 42. Strominger A. Bekenstein JD. Lowe D. Phys. Maldacena J. Mathur SD. Rev. Mathur SD. Nucl. hep-th/9607235 (1996) 38. Witten E. Peet A. J. D 59:086004 (1999). Harvey J. General Relativity. B391:93 (1996). B440:421 (1995). Maldacena J. Press (1987). Tseytlin A. Sen A. Nucl. Nucl. Rev. Wadia SR. Lett. Suppl. Youm D. . Harvey J. 49:1683 (1982) 9. Strominger A. Phys. Phys. 9906:019 (1999). Phys. Phys. Maldacena J. Rev. 316:1 (1999) 45. Phys. Sen A. Nucl. Phys. 63:478 (1989). hep-th/9309145 (1993) 25. D 13:191 (1976) 6. Phys. Phys. Yogendran K. 56B:1 (1997). High Energy Phys. Sen A. Lett. gr-qc/9901032 (1999) 4. Keski-Vakkuri E. Wald RM. B402:43 (1997). hep-th/9904027 (1999) 21. Phys. B472:591 (1996) 32. B340:33 (1990) 26. Balasubramanium V. B244:135 (1984) 10. Verlinde H. 75:4724 (1995) 19. Nucl. B379:99 (1996) 29. 2. Zurek WH. Pelath MA. Brekenridge J. Cambridge. Phys. Arkani-Hamed N. Proc. Horowitz G. Nucl. B72:117 (1977) 24. Martinec E. Cvetic M. B460:335 (1996) 22. Dimopolous S. Polchinski J. Vols. Townsend P. B478:544 (1996) 48. Mod. Phys. Phys. 185:197 (1997) 36. Randall L. Nucl. Lett. D 7:2333 (1973). Lett. Mod. Phys. Nucl. Verlinde E. Phys. J.48 Das & Mathur 1. Rev. Traschen J. Larsen F. Kastor D. Montonen C. Phys. Phys. Banks T. Rev. Math. Myers R. Preskill J et al. Chicago: Univ. Phys. Rev. Phys. Callan C. Witten E. Lett. Commun. Dabholkar A. Phys. Mathur SD. Klebanov I. Phys. Phys. Martinec E. 9904:017 (1999) 43. Polchinski J. Hawking S. Phys. Bekenstein JD. Wadia SR. Lett. Lett. Vafa C. Phys. Green M. PhD thesis. Horowitz G. High Energy Phys. Wadia SR. Rev. Giddings SB. 1. Phys. 2. 77:430 (1996) 46. Verlinde H. hep-th/9910194 (1999) 44.

Strominger A. Tseytlin A. Rev. 52. Rey and J. Wadia SR. Klebanov I. Mandal G. 62. Das SR. Krasnitz M. Klebanov I. Phys. Rev. 73. Theor. et al. D 55:3250 (1997). 68. D 56:2173 (1997) Das SR. 77:4491 (1996) Mathur SD. hep-th/9803001. Mathur SD. 69. 54. 2:253 (1998) Maldacena J. Sakita B. Quant. D 14:3251 (1976) Dhar A. 63. Rev. JHEP 9901:017 (1999) Aharony O. Phys. B165:511 (1980) Das SR. Rev. 79. D 55:861 (1997) Gubser S. Phys. 67. Fischler W. Theor. Gibbons G. Math. B560:207 (1999) Gubser S. Phys. Nucl. B381:423 (1996) Page D. Naik S. Adv.D-Branes and Black Holes 49. Nucl. Proc. Klebanov I. B562:353 (1999). 70. D 55:7645 (1997) Klebanov I. A5:1639 (1990) Das SR. Klebanov I. Nucl. Wadia SR. Rev. Nucl. Phys. Stephens CR. Lett. Klebanov I. 2:505 (1998) Csaki C. Rev. A5:799 (1990) Banks T. 80. Lett. 50. Phys B489:65 (1997). Strominger A. Rev. Mathur SD. Rev. Phys. 51. B413:41 (1997) Gubser S. Phys. Taylor W. Dhar A. Rev. Lett. 68:119 (1998) Maldacena J. Yee. B428:105 (1998) Witten E. Phys. Phys. 77. Math. Phys. Rev. D 13:198 (1976). Suppl. Witten E. JHEP 9812:005 (1998) Jevicki A. Krasnitz M. Math. Phys. hep-th/9911222 (1999) Witten E. Phys. Rastelli L. Rev. Mod. Phys. Phys. Matusis A. Wadia SR. Nucl. Nucl. 61. Phys. Nucl. 66. Adv. Phys. Adv. Phys. Phys. 80:4859 (1998). Susskind S. Theor. Grav. Peet A. Math. Phys. Phys. Nucl. D 27:2262 (1983). Phys. Mod. 65. Phys. Klebanov I. Phys. 57. B388:51 (1996) Das SR. Phys. Rev. Phys. d’Hoker E. J. Phys. Tseytlin A. Klebanov I. Nucl. 74. Terning J. Gubser S. Phys. Adv. ’t Hooft G. Seiberg N. Unruh W. B496:231 (1997). Klebanov I. Phys.J. Phys. B546:96 (1999). Class. et al. Lett. Phys. Bekenstein J. 71. hep-th/9310026 (1993). A4:1033 (1989) Das SR. hep-th/9905111 (1999) Bekenstein J. Ooguri H. 53. Oz Y. Polyakov A. 49 56. Mod. Lett. hep-th/9805114 (1998) Maldacena J. d’Hoker E. Rangamani M. D 56:3582 (1997). 2:231 (1998) Freedman D. Lee S. Lett. Rev. 58. Klebanov I. B499:217 (1997) Klebanov I. Gubser S. D 49:1912 (1994) ’t Hooft G. 72. Lett. Phys. Jevicki A. 64. 2:697 (1998) Maldacena J. et al. 78. D 55:5112 (1997) . 59. 11:621 (1994) Susskind L. Whiting B. D 54:3915 (1996) Gubser S. Lett. Gubser S. 55. Theor. Lett 78:417 (1997) Maldacena J. B482:173 (1996). Math. 76. B514:204 (1998) Callan CG. 60. Phys. Lett. Phys. Phys. 75. Rev. 36:6377 (1995) Susskind L. Minwalla S. Shenker SH. S. Phys. Nucl.

Sign up to vote on this title
UsefulNot useful

Master Your Semester with Scribd & The New York Times

Special offer: Get 4 months of Scribd and The New York Times for just $1.87 per week!

Master Your Semester with a Special Offer from Scribd & The New York Times