DAMTP 96{53 PP96{98 gr-qc/9605058

Area spectrum of the Schwarzschild black hole

Department of Physics, University of Maryland, College Park, Maryland 20742{4111, USA Department of Applied Mathematics and Theoretical Physics, University of Cambridge, Silver Street, Cambridge CB3 9EW, United Kingdom (Revised version, September 1996. Published in Phys. Rev. D 54 (1996), pp. 4982{4996.)

Jorma Louko

Jarmo Makelay

Abstract
We consider a Hamiltonian theory of spherically symmetric vacuum Einstein gravity under Kruskal-like boundary conditions in variables associated with the Einstein-Rosen wormhole throat. The con guration variable in the reduced classical theory is the radius of the throat, in a foliation that is frozen at the left hand side in nity but asymptotically Minkowski at the right hand side in nity, and such that the proper time at the throat agrees with the right hand side Minkowski time. The classical Hamiltonian is numerically equal to the Schwarzschild mass. Within a class of Hamiltonian quantizations, we show that the spectrum of the Hamiltonian operator is discrete and bounded below, and can p made positive de nite. The large eigenvalues behave asymptotibe cally as 2k, where k is an integer. The resulting area spectrum agrees with that proposed by Bekenstein and others. Analogous results hold in the presence of a negative cosmological constant and electric charge. The classical input that led to the quantum results is discussed. Pacs: 04.60.Ds, 04.60.Kz, 04.70.Dy, 04.20.Fy

Typeset using REVTEX

On leave of absence from Department of Physics, University of Helsinki. Electronic address: [email protected]
yElectronic address: [email protected]

I. INTRODUCTION
One of the most intriguing problems in black hole thermodynamics is the statistical mechanical interpretation of black hole entropy. One surmises that black hole entropy should re ect an outside observer's ignorance about the quantum mechanical microstates of the hole, but it has proved very di cult to characterize what exactly these quantum mechanical microstates might be. There are hints that the relevant degrees of freedom may live on the horizon of the hole 1{4], in situations where the horizon can be meaningfully de ned. There is also evidence that black hole entropy describes the entanglement between the degrees of freedom in the interior and in the exterior of the hole 5{8]. Recent results from string theory 9] suggest that black hole entropy can be recovered by counting the quantum microstates even in situations where the de nition of these states presupposes no black hole geometry. For reviews, see Refs. 9{11]. Even prior to Hawking's prediction of black hole radiation 12], the anticipated connection between black holes and thermodynamics led Bekenstein 13] to propose that the horizon area of a black hole is quantized in integer multiples of a fundamental scale, presumably of p the order of the square of the Planck length lPlanck = ~Gc?3 :
2 A = k lPlanck ;

(1.1)

where k ranges over the positive integers and is a pure number of order one. This proposal has since been revived on various grounds; see Refs. 14{31], and references therein. Although the horizon of a (classical) black hole is a nonlocal object, its total area is completely determined by the irreducible mass 32], and one can therefore alternatively view the rule (1.1) as a proposal for the spectrum of the quantum irreducible mass operator.1 As the irreducible mass can classically be read o from the asymptotic fallo of the black hole gravitational eld, one expects such an operator to be sensibly de nable even in a quantum theory that only refers to observations made at an asymptotically at in nity. In particular, for a Schwarzschild hole, the irreducible mass coincides with the Schwarzschild mass. The implications of the area spectrum (1.1) for macroscopic physics were recently elaborated on by Bekenstein and Mukhanov 26]. Consider, for concreteness, a Schwarzschild hole. The area is given in terms of the Schwarzschild mass M by

A = 16

where mPlanck = ~cG?1 is the Planck mass. Now, (1.1) and (1.2) imply that M can only take discrete values. When the black hole evaporates, it can thus only make transitions between the mass eigenstates corresponding to these discrete values. As a consequence, the radiation comes out in multiples of a fundamental frequency, which is of the same order as the maximum of Hawking's black-body spectrum, and the corresponding wavelength is of the order of the Schwarzschild radius of the hole. This means that the radiation
1 We

lPlanck mPlanck

M2 ;

(1.2)

thank John Friedman and Pawel Mazur for emphasizing this to us.

will di er from the black-body spectrum in a way that is, in principle, macroscopically observable. For example, if M is of the order of ten solar masses, or 2 1031 kg, then the fundamental frequency is of the order of 0:1 kHz, which is roughly the resolving power of an ordinary portable radio receiver. The discussion can also be generalized to accommodate a nonvanishing angular momentum 13]. Arguments presented in favor of the area spectrum (1.1) include quantizing the angular momentum of a rotating black hole 13], information theoretic considerations 14,18,19,26], string theoretic arguments 15,16], periodicity of Euclidean or Lorentzian time 15,17,27,29{31], a treatment of the event horizon as a membrane with certain quantum mechanical properties 22,24], and a Hamiltonian quantization of a dust collapse 25]. Recently, a membrane model for the horizon 33] recovered an area spectrum that is ner than (1.1), and a calculation within a loop representation of quantum gravity 34] recovered an area spectrum that e ectively reproduces the Planckian spectrum for black hole radiation. The purpose of the present paper is to give a derivation of (e ectively) the spectrum (1.1) from a Hamiltonian quantum theory of spherically symmetric Einstein gravity, with judiciously chosen dynamical degrees of freedom. By Birkho 's theorem 35], the local properties of spherically symmetric vacuum Einstein geometries are completely characterized by a single parameter, the Schwarzschild mass. In a classical Hamiltonian theory of such spacetimes, the true dynamical degrees of freedom are thus expected to contain information only about the Schwarzschild mass and the embedding of the spacelike hypersurfaces in the spacetime. It was demonstrated in Refs. 36{39] that this is indeed the case, under certain types of boundary conditions that specify the (possibly asymptotic) embedding of the ends of the spacelike hypersurfaces in the spacetime. The variables of the reduced theory then consist of a single canonical pair: the coordinate can be taken to be the Schwarzschild mass, and its conjugate momentum carries the information about the evolution of the (asymptotic) ends of the spacelike hypersurfaces in the spacetime. The theory is thus no longer a eld theory, but a theory of nitely many degrees of freedom. This means, in particular, that quantization of the theory can be addressed within ordinary, nite dimensional quantum mechanics. Our Hamiltonian theory of spherically symmetric vacuum spacetimes will be built on two major assumptions. First, we shall adopt for the spacetime foliation the boundary conditions of Ref. 38]. This implies that the classical solutions have a positive value of the Schwarzschild mass, and that the spacelike hypersurfaces extend on the Kruskal manifold from the left hand side spacelike in nity to the right hand side spacelike in nity, crossing the horizons in arbitrary ways. We shall, however, specialize to the case where the evolution of the hypersurfaces at the left hand side in nity is frozen, and the evolution at the right hand side in nity proceeds at unit rate with respect to the right hand side asymptotic Minkowski time. This means that, apart from constraints, the Hamiltonian will consist of a contribution from the right hand side in nity only, and the value of the Hamiltonian is equal to the Schwarzschild mass. The physical reason for this choice is that while our theory will remain that of vacuum spacetimes, we expect these conditions to correspond to physics accessible to an inertial observer at one spacelike in nity, at rest with respect to the hole: the proper time of such an observer is our asymptotic Minkowski time, and the Arnowitt-Deser-Misner (ADM) mass observed is the Schwarzschild mass. Second, we shall adopt as our reduced dynamical variables a canonical pair that is inti3

mately related to the dynamical aspects of the Kruskal manifold. Our con guration variable a can be envisaged as the radius of the Einstein-Rosen wormhole throat 32], in a spacetime foliation in which the proper time at the throat increases at the same rate as the asymptotic Minkowski time at the right hand side in nity. An example of a foliation satisfying these conditions can be constructed by taking the Novikov coordinates 32] and deforming them near the left hand side in nity to conform to our boundary conditions there. The resulting classical theory has two properties whose physical interest should be emphasized. First, every classical solution is bounded, in the sense that the variable a starts from zero, increases to the maximum value 2M , where M is the Schwarzschild mass, and then collapses back to zero. This evolution corresponds to the wormhole throat starting from the white hole singularity, expanding to the bifurcation two-sphere, and then collapsing to the black hole singularity. The spacetime dynamics in these variables is therefore, in a certain sense, con ned to the interior regions of the Kruskal manifold. This property re ects the physics observed by an inertial observer at asymptotic in nity, as such an observer sees her exterior region of the Kruskal manifold as static. Second, as the proper time on the timelike geodesics that form the throat trajectory increases at the same rate as the asymptotic right hand side Minkowski time, one may regard our foliation as a preferred one, by the principle of equivalence, for relating the experiences of an inertial observer at the asymptotic in nity to the experiences of an inertial observer at the throat. Note, however, that as the total proper time from the initial singularity to the nal singularity along the throat trajectory is nite, our choice of the foliation implies that the throat reaches the white and black hole singularities at nite values of the asymptotic right hand side Minkowski time. As the asymptotic right hand side Minkowski time evolves at unit rate with respect to our parameter time, this means that the classical theory is incomplete: the classical solutions cannot be extended to arbitrarily large values of the parameter time, neither to the past nor to the future. We quantize the theory by Hamiltonian methods, treating a as a con guration variable, specifying a class of `reasonable' inner products, and promoting the classical Hamiltonian into a self-adjoint Hamiltonian operator. For certain choices of the inner product the Hamiltonian operator turns out to be essentially self-adjoint, whereas in the remaining cases the class of self-adjoint extensions is parametrized by U (1) and associated with a boundary condition at a = 0. We nd that the spectrum of the Hamiltonian is discrete and bounded below in all the cases. When the Hamiltonian is essentially self-adjoint, the spectrum is strictly positive, and in the remaining cases there always exist self-adjoint extensions for which the spectrum is strictly positive. A WKB estimate for the large eigenvalues of the Hamiltonian yields, via (1.2), the result that the large area eigenvalues are asymptotically given by2
2 A 32 k lPlanck + constant + o(1) ; (1.3) where k is an integer and o(1) denotes a term that vanishes asymptotically at large A. The additive constant depends on the choice of the inner product and, when the Hamiltonian

2 We

shall use the symbol

to denote an asymptotic expansion throughout the paper.

is not essentially self-adjoint, also on the choice of the self-adjoint extension. With two particular choices for the inner product, we can verify (and improve on) the accuracy of this WKB result rigorously; with another two particular choices, we can rigorously verify the accuracy of the leading order term. We can therefore view our theory as producing, from a Hamiltonian quantum theory constructed from rst principles, the area spectrum (1.1) with = 32 . We shall argue that the discreteness of the quantum spectrum is related to the classical incompleteness of the theory. As the variable a classically reaches the singularity at a = 0 within nite parameter time, both in the past and in the future, the classical theory can be thought of as particle motion on the positive half-line in a con ning potential. Whenever the self-adjoint Hamiltonian operator is constructed so that the possible `quantum potential' part does not become signi cant, general theorems guarantee that the spectrum of the Hamiltonian will be discrete 40,41]. In physical terms, wave packets following classical trajectories will be re ected quantum mechanically from the origin, and the quantum dynamics will in this sense have a quasiperiodic character. In contrast, if the spacetime foliation were chosen so that it would take an in nite amount of parameter time for the variable a to reach the singularity, the classical theory could be thought of as particle motion on the full real line in a potential that is con ning on the right but not on the left. In such potentials, the spectrum of a self-adjoint Hamiltonian operator generically has a continuous part, corresponding physically to the fact that wave packets can travel arbitrarily far to the left without being re ected. We shall present a simple example of each of these two types of foliation, hand-picked so that the Hamiltonian operator becomes easily tractable. From the rst example we can reproduce the area spectrum (1.1) with an arbitrary value of the constant ; in the second example, the spectrum of the Hamiltonian operator will be continuous and consist of the full non-negative half-line. In addition to the above results for the vacuum theory, we shall also brie y investigate the inclusion of a xed electric charge and a negative cosmological constant. With the analogous choices for the boundary conditions, the phase space coordinates, and the Hamiltonian quantum theory, we again show that the spectrum of the Hamiltonian operator is discrete and bounded below. The distribution of the large eigenvalues could presumably be analyzed by a suitable generalization of our vacuum techniques; however, we shall not pursue this issue here. The rest of the paper is as follows. In section II we derive the reduced Hamiltonian theory in our phase space variables, starting from Kuchar's reduced phase space variables 38] and performing the appropriate canonical transformation. The theory is quantized in section III, with considerable parts of the technical analysis deferred to the three appendices. In section IV we discuss the inclusion of the electric charge and a negative cosmological constant. Section V contains a brief summary and a discussion. For the remainder of this paper we shall work in natural units, ~ = c = G = 1.

II. CLASSICAL WORMHOLE THROAT THEORY

In this section we present a classical Hamiltonian theory of the Schwarzschild black hole in terms of reduced phase space variables that are associated with a wormhole throat, in a 5

sense to be made more precise below. We rst brie y recall, in subsection II A, Kuchar's Hamiltonian reduction of spherically symmetric vacuum geometries 38]. In subsection II B we derive the throat theory from Kuchar's reduced theory via a suitable canonical transformation.

A. Kuchar reduction
We start from the general spherically symmetric ADM line element

ds2 = ?N 2 dt2 + 2(dr + N r dt)2 + R2 d

(2.1)

where d 2 is the metric on the unit two-sphere, and N , N r , , and R are functions of t and r only. We adopt the fallo conditions of Ref. 38]. These conditions render the spacetime asymptotically at both at r ! 1 and r ! ?1, and they make jrj coincide asymptotically with the spacelike radial proper distance coordinate in Minkowski space. Each classical solution consists of some portion of the Kruskal manifold 32], such that the constant t hypersurfaces extend from the left hand side spacelike in nity to the right hand side spacelike in nity, crossing the horizons in arbitrary ways. In particular, the Schwarzschild mass is positive for every classical solution. The fallo conditions also guarantee that the fourmomentum at the in nities has no spatial component: the black hole is at rest with respect to the left and right asymptotic Minkowski frames. We x the asymptotic values of N at r ! 1 to be prescribed t-dependent quantities, denoted by N (t). The Hamiltonian form of the Einstein action, with appropriate boundary terms, reads then

S=

? dt (N+ M+ + N?M? ) ;

dt

Z 1

?1

_ dr P _ + PR R ? N H ? N r Hr (2.2)

where H and Hr are respectively the super-Hamiltonian constraint and the radial supermomentum constraint. The quantities M (t) are determined by the asymptotic fallo of the con guration variables, and on a classical solution they both are equal to the Schwarzschild mass. We refer to Ref. 38] for the details. Note that when varying the action (2.2), N (t) are considered xed. Through a judiciously chosen canonical transformation, followed by elimination of the constraints by Hamiltonian reduction, Kuchar 38] brings the action (2.2) to the unconstrained Hamiltonian form

S=

_ dt pm ? (N+ + N?) m] ;

(2.3)

where the independent variables m and p are functions of t only. The degrees of freedom constitute thus the single canonical pair (m; p): the theory is no longer that of elds, but that of nitely many degrees of freedom. The variables take the values m > 0 and ?1 < p < 1, and their equations of motion are 6

_ m=0 ; _ p = ?N+ ? N? :

(2.4a) (2.4b)

For our purposes, it will not be necessary to recall the details of the derivation of the action (2.3), but what will be important is the interpretation of the reduced theory in terms of the spacetime geometry. On each classical solution, the time-independent value of m is simply the value of the Schwarzschild mass. By Birkho 's theorem, m thus carries all the information about the local geometry of the classical solutions. The variable p, on the other hand, is equal to the di erence of the asymptotic Killing times between the left and right in nities on a constant t hypersurface, in the convention where the Killing time at the right (left) in nity increases towards the future (past). The two terms in the evolution equation (2.4b) arise respectively from the two in nities. Thus, p contains no information about the local geometry, but instead it carries the information about the anchoring of the spacelike hypersurfaces at the two in nities. Note that p, as the di erence of the asymptotic Killing times, is invariant under the global isometries that correspond to translations in the Killing time.

B. Hamiltonian throat theory

We shall now make two restrictions on the reduced theory (2.3). First, we specialize to N+ = 1 and N? = 0. This means that the parameter time t coincides with the asymptotic right hand side Minkowski time, up to an additive constant, whereas at the left hand side in nity the hypersurfaces remain frozen at the same value of the asymptotic Minkowski time for all t. The action reads then

S=

_ dt (pm ? m) :

(2.5)

The fact that the Hamiltonian now equals simply m reproduces the familiar identi cation of the Schwarzschild mass as the ADM energy, from the viewpoint of asymptotic Minkowski time evolution at one asymptotically at in nity. Our choice as to which of the two in nities has been taken to evolve is of course merely a convention, but the choice of completely freezing the evolution at the other in nity arises from the requirement that our theory describe physics accessible to observers at just one in nity. We shall return to this issue in section V. Second, for reasons that will become transparent below, we con ne the variables by hand to the range jpj < m. This means that in each classical solution, the asymptotic right hand side Minkowski time only takes values within an interval of length 2 m, centered around a value that is diagonally opposite to the non-evolving left end of the hypersurfaces in the Kruskal diagram. In terms of the parameter time t, each classical solution is then de ned only for an interval ? m < t ? t0 < m, where t = t0 is the hypersurface whose two asymptotic ends are diagonally opposite. Consider now the transformation from the pair (m; p) to the new pair (a; pa) de ned by the equations 7

p db 1 ? pa 2m2b ? 1 1 (2.6a) = 2ma ? a + m arcsin (1 ? a=m) + 2 m ; p pa = sgn(p) 2ma ? a2 : (2.6b) The ranges of the variables are a > 0 and ?1 < pa < 1. The transformation is well-de ned, jpj =
one-to-one, and canonical. The new action reads

Z 2m

S=
where the Hamiltonian is given by

dt (pa a ? H ) ; _

(2.7)

1 2 (2.8) H = 2 pa + a : a The classical solutions are easily written out in the new variables. The value of H on a classical solution is just m, and by writing the canonical momentum pa in (2.8) in terms of a and a, one recovers the equation of motion for a in the form _ m a2 = 2a ? 1 : _ (2.9) Hence the con guration variable a starts from zero at t = t0 ? m, reaches the maximum value 2m at t = t0 , and collapses back to zero at t = t0 + m. The interest in the variables (a; pa ) is that they have an appealing geometrical interpretation in terms of the dynamics of a wormhole throat in the black hole spacetime. To see this, we recall that the derivation of the reduced action (2.3) from the original geometrodynamical action (2.2) in Ref. 38] relied on the properties of the spacelike hypersurfaces only through their asymptotic behavior, but otherwise left these hypersurfaces completely arbitrary. We can thus exercise this freedom and seek an interpretation of the variables (a; pa) in terms of a suitably chosen foliation. The crucial observation is now that equation (2.9) is identical to the equation of a radial timelike geodesic through the bifurcation two-sphere of a Kruskal manifold of mass m, provided one identi es a as the curvature radius of the two-sphere and the dot as the proper time derivative. This is easily seen for example from the expression of the interior Schwarzschild metric in the Schwarzschild coordinates. With these identi cations, equations (2.6) show that ?p becomes identi ed with the proper time elapsed along this geodesic from the bifurcation two-sphere, with positive (negative) values of ?p yielding the part of the geodesic that is in the black (white) hole interior. Thus, if there exists a foliation consistent with the fallo conditions of Ref. 38], intersecting a timelike geodesic through the bifurcation two-sphere so that ?p agrees with the proper time on the geodesic in this fashion, then the quantity a de ned by (2.6a) is the two-sphere radius along this geodesic. It is easy to see that foliations of this kind do exist. Let us brie y discuss an example that is closely related to the Novikov coordinates 32]. Recall that the geometric idea behind the Novikov coordinates consists of xing a spacelike hypersurface of constant Killing time 8

through the Kruskal manifold, and releasing from this hypersurface a family of freely falling test particles with a vanishing initial three-velocity in the Schwarzschild coordinates. The coordinates ( ; R ) are then de ned so that they follow these test particles: the trajectories are the lines of constant R , and on each trajectory is equal to the proper time. The initial hypersurface is = 0, with R > 0 and R < 0 giving respectively the halves living in the right and left exterior regions. Now, to arrive at a foliation satisfying our requirements, we rst deform the Novikov coordinates near the left hand side in nity to accommodate the condition N? = 0, and we then rede ne R near the in nities in a -independent fashion so as to conform to the radial fallo assumed in Ref. 38] (the right hand side in nity will then have, in the notation of Ref. 38], a fallo with = 1). The distinguished geodesic through the bifurcation two-sphere is given by R = 0, and the coordinate agrees by construction both with the proper time along this geodesic and with the asymptotic Minkowski time at the right hand side in nity. Our interpretation of a gives now a geometrical reason for the restriction jpj < m, which we above introduced by hand. As a radial timelike geodesic from the initial singularity to the nal singularity through the bifurcation two-sphere has the nite total proper time 2 m, foliations satisfying our conditions do not cover all of the spacetime. The foliations only exist for the duration of 2 m in the asymptotic right hand side Minkowski time.3 We summarize. Fix a radial timelike geodesic through the bifurcation two-sphere, and choose any foliation, consistent with our fallo conditions, such that the proper time along the geodesic and the asymptotic right hand side Minkowski time agree on the constant t hypersurfaces. Then, the variable a equals the radius of the two-sphere on the distinguished geodesic. In particular, if the foliation is chosen so that on each constant t hypersurface, the radius of the two-sphere attains its minimum value on the distinguished geodesic, then this geodesic and the ones obtained from it by spherical symmetry form the trajectory of the Einstein-Rosen wormhole throat. This is the case for example in foliations obtained by deforming the Novikov coordinates near the left hand side in nity in the manner discussed above. We shall therefore, with a minor abuse of terminology, refer to a as the radius of the wormhole throat, and to the theory given by (2.7) and (2.8) as the Hamiltonian throat theory. It is important to note that the con guration variable a is bounded on each classical trajectory, reaching the maximum value 2m as the wormhole throat crosses the bifurcation two-sphere. This means, in a certain sense, that the spacetime dynamics in terms of our con guration variable a is con ned `inside' the hole. This is physically appealing from the viewpoint of an observer at in nity: such an observer sees the exterior region of the spacetime as static.
that although the Novikov coordinates ( ; R ) are globally well-de ned on the Kruskal manifold, the hypersurfaces of constant extend from one in nity to the other only for j j < m. For larger values of j j, these hypersurfaces hit a singularity.
3 Note

III. THROAT QUANTIZATION

In this section we shall quantize the Hamiltonian throat theory of section II. We saw above that the classical Hamiltonian is numerically equal to the Schwarzschild mass, and that this Hamiltonian arises as the energy with respect to the Minkowski time evolution at one asymptotically at in nity. The quantum Hamiltonian operator can therefore be viewed as the energy operator with respect to an asymptotic Minkowski frame in which the hole is at rest. In particular, the spectrum of the Hamiltonian operator becomes the ADM mass spectrum of the hole. Our main aim will be a qualitative analysis of this spectrum. We take the states of the quantum theory to be described by functions of the con guration variable a. The Hilbert space is H := L2 (R + ; da), with the inner product ( 1; 2) =
Z 1

da

1 2

(3.1)

^ where (a) is some smooth positive weight function. To obtain the Hamiltonian operator H , we make in the classical Hamiltonian (2.8) the substitution pa ! ?id=(da) and adopt a symmetric ordering with respect to the inner product (3.1). The result is d d ^ 1 (3.2) H = 2 ? 1 da a da + a : For technical reasons, it will be useful to work with an isomorphic theory in which the inner product and the kinetic term of the Hamiltonian take a more conventional form. To 3 achieve this, we write a = x2=3 , = 2 x1=3 2 , and = ?1 . The theory above is then mapped to the theory whose Hilbert space is H0 := L2 (R + ; dx), with the inner product ( 1 ; 2)0 =
Z 1

dx

1 2

(3.3)

The new Hamiltonian operator is 2 2=3 00 ^ 9 d ; (3.4) H0 = 8 ? dx2 + 4x9 + ^ where 0 = d=dx. By construction, H0 is a symmetric operator in H0 . Note that if we had 9 retained dimensions, the `quantum potential' term 8 00 = would be proportional to ~2 . ^ To completely specify the quantum theory, we need to make H0 into a self-adjoint oper9 ator on H0 . The possible ways of doing this depend on the quantum potential term 8 00 = . For concreteness, we shall from now on take = as, where s is a real parameter. The qualitative results would, however, be analogous for any su ciently similar with a power-law asymptotic behavior at a ! 0 and a ! 1. ^ With = as, H0 takes the form 2 2=3 ^ 9 d (3.5) H0 = 8 ? dx2 + 4x9 + r(rx? 1) ; 2 where 10

r = (2s ? 1)=6 for s 2 ; (3.6a) r = (7 ? 2s)=6 for s < 2 : (3.6b) We could have replaced (3.6) by (say) (3.6a) for all s, with (3.5) still holding. However, as ^ (3.5) is invariant under r ! 1 ? r, it will be su cient to analyze H0 for r 1=2; this range for r is recovered through the de nition (3.6). ^ It is easy to see that H0 has self-adjoint extensions for any r. In the terminology of Ref. 41], in nity is a limit point case, whereas zero is a limit point case for r 3=2 and limit circle case otherwise (Ref. 41], Theorems X.8 and X.10, and Problem 7). For ^ r 3=2, H0 is therefore essentially self-adjoint. For 1=2 r < 3=2, on the other hand, ^ the self-adjoint extensions of H0 are characterized by a boundary condition at zero and parametrized by U (1). We now wish to extract qualitative information about the spectrum of the self-adjoint ^ extensions of H0. ^ A rst observation is that the essential spectrum of every self-adjoint extension of H0 is empty (Ref. 40], Theorems XIII.7.4, XIII.7.16, and XIII.7.17). This means that the spectrum is discrete: the spectrum consists of eigenvalues corresponding to genuine, normalizable eigenstates, and the eigenvalues have disjoint neighborhoods. ^ Second, we shall show in appendix A that every self-adjoint extension of H0 is bounded below: the system has a ground state. For r 3=2, the ground state energy is always positive. For 1=2 r < 3=2, the ground state energy depends on the self-adjoint extension, and the situation is more versatile. On the one hand, there is a certain (open) set among the self-adjoint extensions within which the ground state energy is positive. On the other hand, there exist extensions whose ground state energy is arbitrarily negative. Third, we shall show in appendix B that a WKB analysis yields for the squares of the large eigenenergies the asymptotic estimate 2 EWKB 2k + constant + o(1) ; (3.7) where k is an integer and o(1) denotes a term that vanishes asymptotically at large E . The constant depends on r and, for 1=2 < r < 3=2, also on the self-adjoint extension, in a way discussed in appendix B. We shall also be able to rigorously verify the accuracy of this WKB result in the special case r = 5=6, and the accuracy of its leading order term in the special case r = 1. These properties of the spectrum have consequences of direct physical interest. At the low end of the spectrum, the fact that the Hamiltonian is bounded below indicates stability: one cannot extract from the system an in nite amount of energy. At the high end of the spectrum, the asymptotic distribution of the large eigenenergies yields for the black hole area the eigenvalues (1.3): this agrees with Bekenstein's area spectrum (1.1), with = 32 . We shall discuss the physical implications of these results further in section V.

IV. THROAT THEORY WITH CHARGE AND A NEGATIVE COSMOLOGICAL CONSTANT

In this section we shall outline how the throat theory can be generalized to accommodate electric charge and a negative cosmological constant. The classical black hole solutions are 11

in this case given (locally) by the Reissner-Nordstrom-anti-de Sitter metric 35]

2 2 dR2 M + R2 d 2 ; (4.1a) ds2 = ? 1 ? 2R + Q2 + R2 dT 2 + 2M Q2 R2 R ` 1? R + R 2 + `2 with the electromagnetic potential one-form A = Q dT : (4.1b) R The parameters M and Q are referred to as the mass and the (electric) charge, and the cosmological constant has been written in terms of the positive parameter ` as ?3`?2. For the global structure of the spacetime, see Refs. 42,43]. We shall understand the case of a vanishing cosmological constant as the limit ` ! 1, and in this case the above solution reduces to the Reissner-Nordstrom solution. The Hamiltonian structure of the spherically symmetric Einstein-Maxwell system with a cosmological constant was analyzed by a technique related to Ashtekar's variables in Refs. 44,45]. An analysis via a Kuchar-type canonical transformation and Hamiltonian reduction, both with and without a negative cosmological constant, was given in Ref. 46]. Although the focus of Ref. 46] was on thermodynamically motivated boundary conditions, which con ne the constant t hypersurfaces to one exterior region of the spacetime, the discussion therein generalizes without essential di culty to boundary conditions that allow the constant t hypersurfaces to extend from a left hand side spacelike in nity to the corresponding right hand side spacelike in nity, crossing the event horizons in arbitrary ways but crossing no inner horizons. The new technical issues arise mainly from the fact that with a negative cosmological constant, the left and right in nities are asymptotically anti-de Sitter rather than asymptotically at. The new physical issues arise mainly in the choice of the electromagnetic boundary conditions at the in nities. We shall here concentrate on the theory where the electric charge is xed at the in nities. In the reduced theory, the charge then becomes an entirely nondynamical, external parameter, which we denote by q. On a classical solution, q is equal to Q in (4.1). The action of the reduced theory takes the form (2.3). On the classical solutions, m is equal to the mass parameter M of (4.1a). The canonical conjugate p again equals the di erence in the asymptotic Killing times between the left and right ends of a constant t hypersurface. The range of m is m > mcrit, where the critical value mcrit(q; `) is positive for q 6= 0 and vanishes for q = 0: this restriction arises from the requirement that the classical solutions have a nondegenerate event horizon 42,43,46]. The range of p is ?1 < p < 1. The quantities N determine the evolution of the ends of the hypersurface in the asymptotic Killing time. For ` ! 1, N are simply the asymptotic values of the lapse, whereas for 0 < ` < 1 they are related to the lapse by a factor that diverges at the in nities. Note that in the special case of q = 0 and 0 < ` < 1, we get a theory of vacuum spacetimes with a negative cosmological constant. Mimicking section II, we freeze the evolution of the hypersurfaces at the left hand side in nity by setting N? = 0, and x the evolution at the right hand side in nity to proceed at unit rate with respect to the Killing time by setting N+ = 1. The action is given by (2.5).

12

The value of the Hamiltonian on a classical solution is then equal to the mass parameter. For ` ! 1 this reproduces the identi cation of the mass as the ADM energy from the viewpoint of asymptotic Minkowski time evolution at one in nity, just as in the uncharged case in section II. For 0 < ` < 1, we similarly recover the interpretation of the mass parameter as the ADM-type energy from the viewpoint of asymptotic anti-de Sitter Killing time evolution at one in nity. In analogy with (2.6), we introduce the new variables (a; pa) via the transformation

jpj =

Z a+
a

2mb?1 ? 1 ? q2 b?2 ? b2 `?2 p pa = sgn(p) 2ma ? a2 ? q2 ? a4`?2 ;

db

(4.2a) (4.2b)

where a+ (m; q; `) is the unique positive zero of the right hand side in (4.2b) for q = 0, and the larger of the two positive zeroes for q 6= 0. On a classical solution, a+ is the radius of the event horizon. To make this transformation well-de ned, we again need to restrict by hand the range of p. For q = 0, the upper limit for jpj is obtained from (4.2a) with a = 0. For q 6= 0, the upper limit for jpj is obtained from (4.2a) with a = a? , where a?(m; q; `) is the smaller of the two positive zeroes of the right hand side in (4.2b). On a classical solution, a? is the radius of the inner horizon. With this restriction on the range of p, the transformation (4.2) is well-de ned, one-toone, and canonical. The new action is given by (2.7) with the Hamiltonian 1 p2 + a + q2 + a3 H = 2 aa a `2

(4.3)

and the value of this Hamiltonian on a classical solution is just the mass. As in section II, the theory has an interpretation in terms of a wormhole throat. If there exists a foliation with the appropriate fallo conditions 46], intersecting a timelike geodesic through the event horizon bifurcation two-sphere so that ?p coincides with the proper time from the bifurcation two-sphere along this geodesic, then the quantity a de ned by (4.2a) is the two-sphere radius on this geodesic. If the foliation is chosen suitably symmetric near the speci ed geodesic, we can think of a as the radius of the wormhole throat. We shall now examine the existence of such foliations for the di erent values of the parameters q and `. For q = 0 and 0 < ` < 1, the classical solution is the Schwarzschild-anti-de Sitter hole. The Penrose diagram di ers from that of the Kruskal manifold only in that the asymptotically at in nities are replaced by asymptotically anti-de Sitter in nities, represented by vertical lines 42,43]. Foliations of the desired kind clearly exist: the timelike geodesic starts at the initial singularity with a = 0, reaches the maximum value of a at the bifurcation two-sphere, and ends at the nal singularity with a = 0. The situation is thus qualitatively very similar to that with the Schwarzschild hole. For q 6= 0 and ` ! 1, the classical solution is the Reissner-Nordstrom hole with m > jqj. The Penrose diagram can be found in Refs. 32,47]. Our Kuchar-type Hamiltonian formulation is valid for the part of the spacetime that consists of one pair of spacelikeseparated left and right asymptotically at regions and the connecting region that is bounded in the past and future by the Cauchy horizons. The solutions to the equations of motion 13

obtained from the Hamiltonian (4.3) are periodic oscillations in the interval a? a a+, but our derivation of this Hamiltonian is only valid on each solution between two successive minima of a. Now, it is clear that foliations of the desired kind exist: the timelike geodesic starts with a = a? at the past Cauchy horizon bifurcation two-sphere, reaches a = a+ at the event horizon bifurcation two-sphere, and ends with a = a? at the future Cauchy horizon bifurcation two-sphere. Finally, for q 6= 0 and 0 < ` < 1, the classical solution is the Reissner-Nordstromanti-de Sitter hole, with m so large that a nondegenerate event horizon exists. The Penrose diagram is obtained from that of the Reissner-Nordstrom hole by replacing the asymptotically at in nities by asymptotically anti-de Sitter in nities 42,43]. Our Kuchar-type Hamiltonian formulation is valid for the part of the spacetime that consists of one pair of spacelike-separated left and right asymptotically anti-de Sitter regions and the connecting region that is bounded in the past and future by the inner horizons. The inner horizons are now not Cauchy horizons, as the asymptotically anti-de Sitter in nities render our part of the spacetime not globally hyperbolic. As with the Reissner-Nordstrom hole above, the solutions to the equations of motion obtained from the Hamiltonian (4.3) are periodic oscillations in the interval a? a a+, and our derivation of this Hamiltonian is only valid on each solution between two successive minima of a. Foliations of the desired kind now exist while the timelike geodesic remains su ciently close to the event horizon bifurcation two-sphere. However, it is seen from the Penrose diagram that as the geodesic progresses towards the past and future inner horizon bifurcation two-spheres, there will occur a critical value of the proper time after which the constant t hypersurfaces would necessarily need to become somewhere timelike. Therefore, the throat interpretation can only be maintained in the full domain of validity of the Hamiltonian (4.3) by appealing to a foliation where the constant t hypersurfaces need not be everywhere spacelike. We shall return to this issue in section V. ^ Quantization of the theory proceeds as in section III. The Hamiltonian H0 (3.5) inherits the additional terms q2 + x2 : (4.4) 2x2=3 2`2 The theorems cited in section III show that the additional terms make no di erence for the ^ existence and counting of the self-adjoint extensions of H0, and they also show that the ^ essential spectrum of any self-adjoint extension of H0 is again empty. For q = 0, the proof of the lower bound for the spectrum given in appendix A goes through virtually without change. For q 6= 0, the charge term modi es the small x behavior of the wave functions, and the analysis of the self-adjointness boundary condition is more involved; in particular, there is a qualitative di erence between the cases 7=6 < r < 3=2, r = 7=6, and 1=2 < r < 7=6, arising from whether the next-to-leading term in the counterpart of vE (x) in (A2) dominates the leading order term in the counterpart of uE (x) at small x. However, the modi ed Bessel function asymptotic behavior (A6) and (A9) still holds, as can be shown by applying the series solution method for uE and the `second solution' integral formula for vE (see, for example, Chapter 8 of Ref. 48]). The spectrum of every self-adjoint extension is therefore again bounded below, and certain self-adjoint extensions are strictly positive. One expects that the asymptotic distribution of the large eigenvalues could be investi14

gated via a suitable generalization of the WKB techniques of appendix B. We shall, however, not attempt to carry out such an analysis here.

V. DISCUSSION
In this paper we have considered a Hamiltonian theory of spherically symmetric vacuum Einstein gravity under Kruskal-like boundary conditions. The foliation was chosen such that the evolution of the spacelike hypersurfaces is frozen at the left hand side in nity, but proceeds at unit rate with respect to the asymptotic Minkowski time at the right hand side in nity. The reduced Hamiltonian theory was written in a set of variables associated with the Einstein-Rosen wormhole throat: the con guration variable is the radius of the throat, in a foliation in which the proper time at the throat agrees with the asymptotic right hand side Minkowski time. The classical Hamiltonian is numerically equal to the Schwarzschild mass. We quantized the theory by Hamiltonian methods, taking the wave functions to be functions of the classical con guration variable, and including a general power-law weight factor in the inner product. The classical Hamiltonian was promoted into a self-adjoint Hamiltonian operator. We found that the spectrum of the Hamiltonian operator is discrete and bounded below for all the choices of the weight factor. In the cases where the Hamiltonian operator is essentially self-adjoint, the spectrum is necessarily positive de nite; in the remaining cases, self-adjoint extensions with a positive de nite spectrum always exist. In p all the cases, a WKB estimate gave for the large eigenvalues the asymptotic behavior 2k, where k is an integer, and we were able to rigorously verify the accuracy of this estimate for four particular choices of the weight factor. The resulting spectrum for the area of the black hole agrees with the spectrum (1.1) proposed by Bekenstein and others, with the dimensionless constant taking the value 32 . We also showed that analogous results can be obtained in the presence of a xed electric charge and a negative cosmological constant. It is perhaps worth emphasizing that the basic postulates of our quantum theory consisted of the choice of a Hilbert space and a self-adjoint Hamiltonian operator on it. We did not attempt to de ne more `elementary' operators, such as those for `position' or `momentum', self-adjoint or otherwise, from which the Hamiltonian operator could be constructed. This issue might merit further study within some geometric or algebraic framework of quantization 49{51].4 Even though our theory is that of pure vacuum, our boundary conditions were chosen so as to make the results relevant for physics that is accessible to an inertial observer at a spacelike in nity. Our spacelike hypersurfaces have evolution at only one in nity, and there they evolve at unit rate with respect to the asymptotic Minkowski time. Our classical Hamiltonian is therefore the gravitational Hamiltonian with respect to the proper time of
4 For example, note that the kinetic term of the Hamiltonian operator H (3.2) can be written as ^ 1 ?1=2 p a?1 p ?1=2 , where p : 7! ?i ?1=2 (d=da) 1=2 is a symmetric (but not self-adjoint) ^a ^a ^a 2

momentum operator that can be regarded as conjugate to the position operator a : 7! a . We ^ thank Thomas Strobl for this observation.

15

an inertial observer at the in nity, at rest with respect to the hole. It is thus reasonable to think of the eigenvalues of the Hamiltonian operator as the possible outcomes that an asymptotic observer would in principle obtain when measuring the ADM mass of the hole. In a given (pure) quantum state, the probability for obtaining a given eigenvalue is determined by the component of the state in the respective eigenspace in the standard way. Although we are here for concreteness using language adapted to a Copenhagen-type interpretation, a translation into interpretations of the many-worlds type could easily be made. One can also make a case that our throat variable a depicts in a particularly natural way the dynamical aspects of the Kruskal manifold. Classically, the wormhole throat begins life at the white hole singularity, expands to maximum radius at the bifurcation two-sphere, and collapses to the black hole singularity. The dynamics of a is therefore, in a certain sense, con ned to the interior regions of the Kruskal manifold, and these are precisely the regions that do not admit a timelike Killing vector. From a physical viewpoint, using a variable with this property is motivated by the fact that an inertial observer at a spacelike in nity sees her exterior region of the Kruskal manifold as static. Further, our foliation made the proper time at the throat increase at the same rate as the asymptotic right hand side Minkowski time: by the principle of equivalence, one may see this as the preferred condition for relating the experiences of an inertial observer at the asymptotic in nity to the experiences of an inertial observer at the throat. We recall, in contrast, that the reduced phase space variables of Refs. 36{38] re ect more closely the static aspects of the Kruskal manifold. Yet another set of variables has been discussed in Refs. 52,53]. With this physical picture, the properties we obtained for the spectrum of the Hamiltonian operator acquire consequences of direct physical interest. At the low end of the spectrum, the fact that the Hamiltonian is bounded below indicates stability: one cannot extract from the system an in nite amount of energy. At the high end of the spectrum, in the semiclassical regime of the theory, the discreteness of the spectrum in accordance with the area quantization rule (1.1) yields the macroscopically observable consequences discussed by Bekenstein and Mukhanov 26]. It should be emphasized, however, that these arguments operate at a somewhat formal level, as our theory does not describe how the quantum black hole would interact with other degrees of freedom, such as departures from spherical symmetry or matter elds. On the grounds of classical positive energy theorems, one may feel inclined to exclude by at quantum theories in which the ground state energy of an isolated self-gravitating system is negative. Among our theories, this would amount to a restriction on the self-adjoint extension in the cases where the Hamiltonian operator is not essentially self-adjoint. However, given the freedom that we have already allowed in the choice of the the inner product, it would be a relatively minor further generalization to add to our Hamiltonian operator the identity operator with some real coe cient, and to take the coe cient as a new parameter in the quantum theory.5 The classical limit of the quantum theory would still be correct, provided the new parameter is understood to be proportional to Planck's constant. When the Hamiltonian operator is not essentially self-adjoint, any given self-adjoint extension can
5 We

thank John Friedman for this observation.

16

then be made positive de nite by choosing the new parameter su ciently large. Note, however, that with any xed value of the new parameter, there still exist self-adjoint extensions whose ground state energy is arbitrarily negative. In section II, we obtained the classical Hamiltonian throat theory by rst going from the geometrodynamical Hamiltonian variables to Kuchar's reduced Hamiltonian theory, and then performing a suitable canonical transformation. The interpretation of our variable a as the radius of the wormhole throat was only introduced after the fact, by appealing to a particular choice of the spacetime foliation. We took the same route in the presence of charge and a negative cosmological constant in section IV. We have not discussed here how to derive a throat theory directly from the unreduced geometrodynamical Hamiltonian theory by introducing a gauge and performing the Hamiltonian reduction, but with the appropriate gauge choice, the resulting theory should by construction be at least locally identical to ours. The only case where one anticipates a di erence in the global properties is in the presence of both a nonvanishing charge and a negative cosmological constant. In this case, we saw in section IV that while our throat theory is valid on each classical solution between two successive minima of a, the wormhole throat interpretation could be maintained for all of this interval only by appealing to a foliation that is not everywhere spacelike. A direct Hamiltonian reduction of the geometrodynamical theory in the corresponding wormhole-type gauge would thus only yield our theory in a more limited domain, valid on each classical solution in a certain interval around a maximum of a. Our choice of freezing the evolution of the spacelike hypersurfaces at the left hand side in nity was motivated by the desire to have a theory that would describe physics accessible to observers at just one in nity. For a vanishing charge and cosmological constant, this motivation can be implemented at the very beginning by setting up the Hamiltonian theory not on the Kruskal manifold, but instead on the R P 3 geon 54]. To see this, recall 54] that the R P 3 geon is the quotient of the Kruskal manifold under a freely acting involutive isometry: this isometry consists of a re ection of the Kruskal diagram about the vertical timelike line through the bifurcation point, followed by the antipodal map on the twosphere. The R P 3 geon has thus only one exterior region, identical to one of the Kruskal exterior regions. Further, the R P 3 geon possesses a distinguished R P 2 of timelike geodesics through the image of the Kruskal bifurcation two-sphere: in the Penrose diagram 54], these geodesics go straight up along the `boundary' of the diagram. The existence of the distinguished geodesics re ects the fact that translations in the Killing time on the Kruskal manifold do not descend into globally de ned isometries of the R P 3 geon. Now, Kuchar's canonical transformation and Hamiltonian reduction generalize readily to the R P 3 geon 39]. The reduced action is obtained from (2.3) by setting N? = 0, and the momentum p is now equal to the di erence of the Killing times between the distinguished timelike geodesics and the single spacelike in nity. Setting N+ = 1, we are led to the action (2.5). The variable a de ned by (2.6) is now equal to the curvature radius of the distinguished R P 2 of timelike geodesics. The above interpretation of our theory in terms of the R P 3 geon generalizes immediately to accommodate a negative cosmological constant. For a nonvanishing charge, on the other hand, there exists again an analogous involutive isometry that can be used to quotient the manifold, but the electric eld is invariant under this isometry only up to its sign. This re ects the fact that Gauss's theorem prohibits a regular spacelike hypersurface with 17

just one asymptotic in nity from carrying a nonzero charge. The R P 3 geon interpretation does therefore not extend to the charged case with a conventional implementation of the electromagnetic eld. As we have seen, the central input in our classical theory was to parametrize the geometry in terms of the radius of the wormhole throat in a judiciously chosen foliation: our variable a is the two-sphere radius on a radial geodesic through the event horizon bifurcation twosphere, in a foliation such that the proper time along the distinguished geodesic agrees with the asymptotic Killing time at the right hand side in nity. One possible generalization would be to relax the requirement that the variable `live' in the interior regions of the manifold, and use instead timelike geodesics that do not pass through the bifurcation two-sphere. To examine this, let us for concreteness set the charge and the cosmological constant to zero, and let us generalize the canonical transformation (2.6) to

jpj =

2m 1+(R0 )2 ] Z
a

q q

db 2mb?1 ? 1 + (R0 )2

?1

(5.1a)

(5.1b) pa = sgn(p) 2ma ? a2 1 + (R0 )2 ?1 ; where R0 is a real-valued parameter. The transformation (2.6) is recovered as the special case R0 = 0. The ranges of the new variables are again a > 0 and ?1 < pa < 1, but the restriction for p is now jpj < 1 + (R0 )2 3=2 m. The action is given by (2.7) with the Hamiltonian " # 2 1 (5.2) H = 2 pa + a 2 : a 1 + (R0 ) On a classical solution, the variable a is now equal to the two-sphere radius on a radial timelike geodesic whose trajectory is given by R = R0 , where R is the Novikov space coordinate in the notation of Ref. 32]. Foliations that make this interpretation possible clearly exist. Simple examples are obtained by deforming the Novikov foliation near the left hand side in nity as in section II, to accommodate the boundary condition N? = 0, and (for R0 6= 0) also near the throat, to prohibit the constant t hypersurfaces from reaching the singularity before the geodesic at R = R0 . Upon quantization along the lines of section II, we nd that the spectrum depends on the parameter R0 only through an overall factor: if ( the eigenvalues are denoted by EkR0 ) , where k ranges over the nonnegative integers, we have
( (0) (5.3) EkR0 ) = 1 + (R0 )2 ?3=4Ek : A much wider generalization would be to relax the requirement, which we above motivated by the equivalence principle, that the proper time along the throat trajectory agree with the asymptotic Killing time. If one allows this freedom, it is not di cult to come up with examples of foliations in which the Hamiltonian takes a mathematically simple form. As an illustration, let us exhibit two examples in the case of vanishing charge and cosmological constant.

18

As the rst example, suppose that p is restricted by hand to have the range jpj < ?1 m, where is a positive constant. We perform the canonical transformation

where the ranges of the new canonical variables are Hamiltonian takes the form

p ; 2m p p p = 2= m sin 2m ;
= 2= m cos
p

(5.4a) (5.4b)

> 0 and ?1 < p < 1. The

(5.5)

We can identify 2 as the two-sphere radius on a radial geodesic through the bifurcation two-sphere in a foliation where the proper time along this geodesic is = ?sgn(p)
2 Zm

H=

1 2

(p2 + 2) :

p2

p db 1 ; 2mb? ? 1

(5.6)

with given by (5.4a). R To quantize this theory, we adopt the inner product ( 1 ; 2 ) = 01 d 1 2 . We de ne ^ the Hamiltonian operator H by spectral analysis as the positive square root of some positive ^ ^ de nite self-adjoint extension of HSHO, where HSHO := 1 ?(d=d )2 + 2 is the simple 2 harmonic oscillator Hamiltonian operator on the positive half-line. The following statements ^ about HSHO can now be veri ed: (i) the self-adjoint extensions are speci ed by the boundary condition cos( ) ? sin( )d =d = 0 at the origin, with the parameter satisfying 0 < ; (ii) the spectrum of each self-adjoint extension is purely discrete; (iii) the eigenfunctions are parabolic cylinder functions 55], and for = 0 and = =2 they reduce respectively to the odd and even ordinary harmonic oscillator wave functions; (iv) if k denotes the eigenvalues, with k ranging over the nonnegative integers, we have for = 0 and = =2 the respective 1 exact results k = 2k + 3 and k = 2k + 2 , and for other values of the asymptotic large k 2 ? 1 expansion k 2k + 2 + ?1 cot( )k?1=2 + o k?1=2 ; (v) the absence of negative eigenvalues is equivalent to the condition that not lie in the interval ?2?3=2 ?1 ?(1=4)]2 < tan( ) < 0. ^ The resulting spectrum for H therefore agrees asymptotically with the area quantization ^ rule (1.1) for any that makes HSHO positive de nite. The numerical constant takes the value 32 . As the second example, suppose that p retains the full range ?1 < p < 1, and perform the canonical transformation

where the ranges of the new canonical variables are Hamiltonian takes the form 19

m2 ; cosh p 2m p = sinh 2p ; m
=

(5.7a) (5.7b)

> 0 and ?1 < p < 1. The

We can identify 2p as the two-sphere radius on a radial geodesic through the bifurcation two-sphere in a foliation where the proper time along this geodesic is = ?sgn(p)
2 Zm 2p

H = ( 2p2 + 2)1=4 :

(5.8)

p db 1 ; 2mb? ? 1

(5.9)

with given by (5.7a). R In the quantum theory we now adopt the inner product ( 1 ; 2 ) = 01 ?1d 1 2 . The ^ operator Hexp := ? (d=d )]2 + 2 is essentially self-adjoint, its discrete spectrum is empty, and its essential spectrum consists of the non-negative half-line.6 We can therefore de ne ^ ^ 1=4 the Hamiltonian operator H by spectral analysis as (Hexp) . It follows that the spectrum ^ of H is now continuous and consists of the non-negative half-line. These examples suggest that the continuity versus discreteness of the spectrum is related to the question of whether the wormhole throat reaches the initial and nal singularities within nite parameter time. There are general grounds to expect this to be the case. A classical theory in which the two-sphere radius reaches zero within nite parameter time is singular, in the sense that the classical solutions cannot be continued arbitrarily far into the past and future. If one quantizes such a theory so that the classical Hamiltonian is promoted into a time-independent, self-adjoint Hamiltonian operator, then the unitary evolution generated by the Hamiltonian operator remains well-de ned for arbitrarily large times. If one starts with an initial wave function that is a wave packet following some classical trajectory, the quantum time evolution will force the wave packet to be re ected from the classical singularity. The re ection is an entirely quantum mechanical phenomenon, and the quantum dynamics acquires in this sense a quasiperiodic character. On the other hand, if the classical solutions require an in nite amount of time to reach the singularity, one generically expects 41] that in the quantum theory a wave packet initially following a classical trajectory will just keep following this trajectory, with some spreading, for arbitrarily large times. It is clear that these arguments apply without change also in the presence of a negative cosmological constant. In the presence of a nonvanishing charge, an analogous discussion applies with the singularity replaced by the inner horizon. One may hold mixed feelings about a wormhole throat quantum theory that introduces a quantum mechanical bounce at a classical singularity or at an inner horizon. On the one hand, singularities and inner horizons are places where the classical theory behaves poorly, and one anticipates quantum e ects to be important. On the other hand, an outright bounce may appear an uncomfortably orderly quantum prediction, given that (semi)classical intuition associates singularities and inner horizons with collapses and instabilities. Related
40] after bringing and the inner product to a standard form by the substitution = e . Note that if we had retained dimensions, this substitution would need to include a dimensionful constant. ^ Hexp
6 These statements follow in a straightforward manner from Chapter VIII of Ref.

20

discussion, in this and related contexts, can be found in Refs. 56{62]. While we view the model of the present paper as a useful arena where these issues can be addressed in relatively explicit terms, the model is undoubtedly dynamically too poor to support con dent conclusions about the physical reasonableness of a discrete versus continuous black hole spectrum. It would be substantially more interesting if our techniques could be generalized to models containing degrees of freedom that carry Hawking radiation.

ACKNOWLEDGMENTS
We are indebted to John Friedman, Stephen Winters-Hilt, and Ian Redmount for making us aware of their previous, related work on throat variables for the Schwarzschild geometry 63,64], and to John Friedman and Stephen Winters-Hilt for discussions and collaboration in an early stage of this work. We would also like to thank Andrei Barvinsky, David Brown, Valeri Frolov, Gary Horowitz, Ted Jacobson, Ian Kogan, Karel Kuchar, Gabor Kunstatter, Pawel Mazur, Diego Mazzitelli, Yoav Peleg, Jonathan Simon, Thomas Strobl, and especially Don Marolf, for helpful discussions and comments. J. M. is grateful to the DAMTP relativity group for hospitality during his visit. J. L. was supported in part by the NSF grant PHY9119726. J. M. was supported by the Finnish Cultural Foundation.
^ APPENDIX A: SEMI-BOUNDEDNESS OF H0

^ In this appendix we shall show that every self-adjoint extension of the Hamiltonian H0 (3.5) is bounded below, and that certain self-adjoint extensions are strictly positive. We shall discuss separately the cases r 3=2, 1=2 < r < 3=2, and r = 1=2.

1.

3=2

^ For r 3=2, H0 is essentially self-adjoint. Let be an eigenfunction with energy E . With a suitable choice of the overall numerical factor, is real-valued and has the small x expansion (x) = xr 1 + O(x2) :
?

(A1)

Both and 0 are therefore positive for su ciently small x. If now E 0, the eigenvalue equation shows then that is increasing for all x > 0. This implies that cannot be normalizable, which contradicts the assumption that is an eigenfunction. Hence the spectrum is strictly positive.

2. 1=2 < r < 3=2

^ For 1=2 < r < 3=2, the self-adjoint extensions of H0 form a family characterized by a boundary condition at x = 0, and parametrized by U (1). To nd these extensions, we note 21

^ that for any E , the di erential equation H0 = E has two linearly independent solutions, denoted by uE (x) and vE (x), with the asymptotic small x behavior

uE (x) = xr 1?+ O(x2) ; vE (x) = x1?r 1 + O(x2) :

(A2a) (A2b)

For real E , both u and v are real-valued. Using the techniques of Ref. 41], it is easily ^ shown that the eigenfunctions of a given self-adjoint extension of H0 take, up to overall normalization, the form
E

= cos( )uE + sin( )vE ;

(A3)

where 2 0; ) is the parameter specifying the self-adjoint extension. The condition (A3) can be written without explicit reference to the solutions (A2) as
r?1 0 = x!0 (2r ? 1) cos( )xr?1 ? sin( )x2(1?r) d(x ) lim dx

(A4)

We now proceed to obtain a lower bound for the eigenenergies. Consider rst an extension in the range 0 =2. Comparing the eigenvalue di eren1 ^ tial equation to the corresponding equation with the term 2 x2=3 omitted from H0 (3.5) and the energy set to zero, one sees that the prospective eigenfunctions with E 0 are bounded below by the function cos( )xr + sin( )x1?r , which does not vanish exponentially at large x. However, as the potential increases without bound as x goes to in nity, every eigenfunction must vanish exponentially at large x. Hence the spectrum is strictly positive. Consider then an extension in the remaining range =2 < < . Let be an eigenfunction with energy E < 0. Writing y = (?8E=9)1=2 x, the eigenfunction equation reads

d2 r(r ? 1) + 1 + 3y 0= ? 2+ dy y2 8E 2

"

2=3

(A5)

The last term in (A5) is asymptotically small at large negative E , uniformly in the interval y 2 (0; M ], where M is an arbitrary positive constant. Omitting this last term gives an equation whose linearly independent solutions are y1=2Ir?(1=2) (y) and y1=2I(1=2)?r (y), where I is a modi ed Bessel function 65]. Therefore has at large negative E the asymptotic behavior
1 (y=2)1=2 cos( )?(r + 2 )(?2E=9)?r=2Ir?(1=2) (y)
h

+ sin(

3 )?( 2

? r)(?2E=9)(r?1)=2 I

(1=2)?r (y )

(A6)

uniformly for y 2 (0; M ]. The coe cients of the two Bessel functions in (A6) have been xed by comparing the small y expansions of (A3) and (A6) 65]. By the asymptotic behavior of the Bessel functions at large argument 65], we can now choose M so that y1=2I(1=2)?r (y) is positive and increasing for y M=2. For future use, we make this choice so that M > 1. Then, the second term in (A6) dominates the rst term 22

~ at large negative E , uniformly for M=2 y M . Therefore there exists a constant E < 0, ~ dependent on r and , such that and d =dy are positive at y = M whenever E < E . As jr(r ? 1)j < 1, equation (A5) then shows, by virtue of the choice M > 1, that diverges ~ at large y whenever E < E . As is by assumption normalizable, we thus see that the ~ eigenenergies are bounded below by E . Note that the lower bound for the eigenenergies is not uniform in . For xed r and ^ any given E , there exists a unique self-adjoint extension of H0 such that E is in the spec^ 0 = E has for any E a normalizable trum. This is because the di erential equation H solution that is unique up to a multiplicative constant, and matching the small x behavior of this solution to (A3) uniquely speci es the value of . One can thus nd extensions with arbitrarily negative ground state energy.

3. r = 1=2
^ For r = 1=2, the self-adjoint extensions of H0 form again a family parametrized by U (1). The boundary condition characterizing the extensions takes the form (A3), where 2 0; ), but now with ? uE (x) = x1=2 1 + O(x2) ; (A7a) vE (x) = uE (x) ln x + O(x5=2 ) : (A7b) Condition (A4) is replaced by d(x?1=2 ) ? sin( )x?1=2 0 = x!0 cos( ) + sin( ) ln x] x lim : (A8) dx For the extension with = 0, one sees as above that the spectrum is strictly positive. Consider then an extension in the range 0 < < . Let be an eigenfunction with energy E < 0, and proceed as above. Equation (A6) is replaced by 9 where K is the second modi ed Bessel function 65] and is Euler's constant. At large negative E , the term proportional to y1=2I0(y) dominates, and one can argue as above. Hence the spectrum is bounded below. As in the case 1=2 < r < 3=2, for any given energy E there exists a self-adjoint extension such that E is in the spectrum. One can thus again nd extensions with arbitrarily negative ground state energy.
^ APPENDIX B: LARGE EIGENVALUES OF H0

?8E

?1=4

1 y1=2 cos( ) ? sin( ) + 2 ln (?2E=9) I0(y) ? sin( )K0(y) ; (A9)

In this appendix we analyze the asymptotic distribution of the large eigenvalues of the ^ self-adjoint extensions of H0 (3.5). The idea is to match a Bessel function approximation at small argument to a WKB approximation in the region of rapid oscillations. We shall again discuss separately the cases r 3=2, 1=2 < r < 3=2, and r = 1=2. 23

1.

3=2

^ We begin with the case r 3=2, where H0 is essentially self-adjoint. We shall throughout denote by an eigenfunction with energy E > 0. Consider rst at small argument. Setting z = (8E=9)1=2 x, the eigenfunction equation reads
2 0 = ? d 2 + r(r ? 1) ? 1 + 3z2 dz z2 8E
"

2=3

(B1)

The last term in (B1) is asymptotically small at large E , uniformly in the interval z 2 (0; ME 1=2 ], where M is an arbitrary positive constant. Omitting this last term gives an equation whose linearly independent solutions are z1=2 Jr?(1=2) (z) and z1=2 Nr?(1=2) (z), where J and N are the Bessel functions of the rst and second kind, and only the former solution is normalizable at small x 65]. The asymptotic large E behavior of is therefore

/ x1=2 Jr?(1=2)

"

(8E )1=2 x 3

(B2)

valid uniformly in any bounded region in x. Here, and from now on, we use the symbol / to denote the asymptotic form at large E , up to a possibly E -dependent coe cient. Introducing two constants 1 and 2 that satisfy 0 < 1 < 2, and using the asymptotic large argument behavior of J 65], we can rewrite (B2) as (8E )1=2 x r ?2 / cos 3
" #

(B3)

valid uniformly for 1 x 2 . Consider then the region of rapid oscillations. We take E so large that the eigenvalue equation has two turning points, and we denote the larger turning point by x0 . The region of rapid oscillations at large E is su ciently far left of x0 but su ciently far right of the origin. As decays exponentially right of x0 , the WKB approximation to the wave function in the region of rapid oscillations is (see, for example, Ref. 66])
WKB

= p2

(x)]?1=2 cos

Z x0
x

dx0 p2 (x0) ? 4

(B4)

where 2 2E ? x2=3 ? 9r(r ? 1) 1=2 : p2 = 3 4x2 The evaluation of the integral in (B4) is discussed in appendix C. We nd (8E )1=2 x ? E 2 + + O ?E ?1=2 WKB / cos 3 2 4 24
" #

(B5)

(B6)

valid uniformly for the WKB estimate

2.

Comparing (B3) and (B6) yields for the large eigenenergies

2 (B7) EWKB 2k + r + 1 + o(1) ; 2 where k is a large integer, and o(1) indicates a term that goes to zero at large E . Note that we have not attempted to control how far the integer k is from the number of the eigenvalue that equation (B7) is meant to approximate. To obtain a formula that gives an approximation to the kth eigenenergy in the limit of large k, one may need to add to the right hand side of (B7) some even integer. As the potential term in the Hamiltonian is smooth without oscillations or step-like behavior, and as jp02p?2 j vanishes for large E uniformly in 1 x 2 , one expects the 2 WKB estimate to the large eigenenergies to be an accurate one. We shall not attempt to investigate the accuracy rigorously for general r, but we shall see below that the accuracy can be veri ed by independent means in the special case r = 5=6, and, to leading order, also in the special case r = 1.

2. 1=2 < r < 3=2

^ For 1=2 < r < 3=2, we recall from appendix A that the self-adjoint extensions of H0 are speci ed by the boundary condition (A3) at small x. Fixing and proceeding as above, we see from the small argument behavior of the Bessel functions 65] that equation (B2) is replaced by
1 (z=2)1=2 cos( )?(r + 2 )(2E=9)?r=2 Jr?(1=2) (z)
h

3 + sin( )?( 2 ? r)(2E=9)(r?1)=2 J(1=2)?r (z) :

(B8)

When = 0, the second term in (B8) vanishes, and we can proceed as above, with x0 now denoting the larger one of the two turning points for r > 1 and the unique turning point for r 1. The WKB estimate for the large eigenenergies is again given by (B7). When 6= 0, on the other hand, the second term in (B8) dominates the rst term at large E , and r in (B3) is replaced by 1 ? r. The WKB estimate (B7) is therefore replaced by
2 1 (B9) EWKB 2k ? r ? 2 + o(1) : In the special case r = 5=6, we can verify the p accuracy of these WKB results rigorously. 1=6 U ?? 1 E 2 ; 2(x2=3 ? E ) , where U is the parabolic The eigenfunctions are now = x 2 cylinder function that vanishes at large values of its second argument 55]. The boundary condition (A4) reads 1 1 0 = cos( )U ? 2 E 2; ? 2E ? 2 sin( )U 0 ? 2 E 2; ? 2E

(B10)

where the prime denotes the derivative of U with respect to its second argument. Using Olver's asymptotic expansions of parabolic cylinder functions (Ref. 67], formula (9.7), and the discussion of the derivative on p. 155), we nd 25

E 2 2k ? 2=3 + O(E ?8=3) E 2 2k + 2=3 + O(E ?4=3) E 2 2k + 2=3 + O(E ?1=3)

for = 0 ; for = =2 ; for 0 6= 6= =2 :

(B11a) (B11b) (B11c)

This corroborates the WKB results (B7) and (B9) for r = 5=6, and gives an improved bound for the error term. In the special case r = 1, the theorem in x7 of Ref. 68] yields the rigorous asymptotic estimate

E 2 2k + O(ln E ) :

(B12)

(The leading order term of (B12) also follows from Ref. 40], p. 1614.) This corroborates the leading order term in our WKB results (B7) and (B9) for r = 1. Finally, we note in passing that for r = 5=6, the boundary condition (B10) and Olver's expansions 67] yield the asymptotic relation 31=6 ?(1=3)]2 + O (?E )?5=3 tan( ) ? 4=3 (B13) 0 2 (?E0 )1=3 for the parameter and the ground state energy E0 , valid in the limit of large negative E0.

3. r = 1=2
For r = 1=2, equation (B8) is replaced by
? 8E ?1=4 1=2 1 z cos( ) ? sin( ) + 2 ln (2E=9) J0(z) + ( =2) sin( )N0 (z) 9

;
(B14)

where is Euler's constant as before. For any , the term proportional to z 1=2 J0 (z) dominates at large energies, and equation (B3) holds with r = 1=2. The WKB estimate for the eigenenergies is thus given by (B7) with r = 1=2, for any .

APPENDIX C: EVALUATION OF THE WKB INTEGRAL

In this appendix we outline the evaluation of the integral in (B4) at large E . We shall throughout assume that E is so large that a classically allowed domain exists (which is a restriction only for r > 1), that x is in the classically allowed domain, and that x > 1, where the constant 1 was introduced in appendix B. Returning to the variable a = x2=3 of the main text, and writing b = (x0 )2=3 , the integral in the exponent in (B4) takes the form

S :=

Z a0
a

9 (r ? 1) db 2Eb ? b2 1 ? 4b3r(2E ? b) ; 26

(C1)

where a0 = x2=3 is the (larger) turning point. At large E , a0 2E + O(E ?3). 0 We x a constant 3 > 0, take E so large that ja0 ? 2E j < 3 =E , and restrict a to 2 be in the interval 1=3 a 2E ? 3=E . We can then replace the upper limit of the integral in (C1) by 2E ? 3=E , with the error in S being of order O(E ?1). The second term under the second square root in (C1) is now uniformly of order O(E ?1). As the rst square root is of order O(E ), we can expand the second square root in the Taylor series and truncate the series after the third term, with the error in S being of order O(E ?1). In the truncated integrand, the third term is proportional to b?11=2 (2E ? b)?3=2 , which is uniformly of order O(E ?3=2), and this term can thus be omitted with the consequence of making an error of order O(E ?1=2) in S . We therefore have 9p(r ? 1) r + O(E ?1=2) : (C2) 8b2 2Eb ? b2 a The integral in (C2) is elementary. The contribution from the second term turns out to be of order O(E ?1=2), and in the rst term the upper limit can be replaced by 2E with the consequence of making an error of order O(E ?1). We thus obtain

Z 2E ? 3 =E

db

2Eb ? b2 ?

p 1 1 2 1 2 2 (E ? a) 2Ea ? a + 2 E arcsin (1 ? a=E ) + 4 p

E 2 + O(E ?1=2) :
2

(C3) was (C4)

2 2 Finally, we restrict a to be in the interval 1=3 a 2=3 , where the constant introduced in appendix B. A large E expansion of (C3) then gives 2 Ea3 S ? 83 + E + O(E ?1=2) ; 2

which leads to (B6).

27

REFERENCES
1] S. Carlip and C. Teitelboim, Class. Quantum Grav. 12, 1699 (1995). (gr-qc/9312002) 2] S. Carlip and C. Teitelboim, Phys. Rev. D 51, 622 (1995). (gr-qc/9405070) 3] S. Carlip, Phys. Rev. D 51, 632 (1995). (gr-qc/9409052) 4] A. P. Balachandran, L. Chandar, and A. Momen, Nucl. Phys. B461, 581 (1996). (grqc/9412019) 5] L. Bombelli, R. Koul, J. Lee, and R. D. Sorkin, Phys. Rev. D 34, 373 (1986). 6] M. Srednicki, Phys. Rev. Lett. 71, 666 (1993). 7] V. Frolov and I. Novikov, Phys. Rev. D 48, 4545 (1993). (gr-qc/9309001) 8] V. P. Frolov, Phys. Rev. Lett. 74, 3319 (1995). (gr-qc/9406037) 9] G. T. Horowitz, \The origin of black hole entropy in string theory," Report UCSBTH-9607, gr-qc/9604051, to appear in the proceedings of the Paci c Conference on Gravitation and Cosmology, Seoul, Korea, February 1-6, 1996. 10] D. N. Page, in Proceedings of the 5th Canadian Conference on General Relativity and Relativistic Astrophysics, University of Waterloo, 13{15 May, 1993, edited by R. B. Mann and R. G. McLenaghan (World Scienti c, Singapore, 1994). (hep-th/9305040) 11] J. D. Bekenstein, in General Relativity , Proceedings of the Seventh Marcel Grossman Meeting, Stanford, California, 1994, edited by R. Ru ni and M. Keiser (Word Scienti c, Singapore, 1995). (gr-qc/9409015) 12] S. W. Hawking, Commun. Math. Phys. 43, 199 (1975). 13] J. D. Bekenstein, Lett. Nuovo Cimento 11, 467 (1974). 14] V. F. Mukhanov, Pis'ma Zh. Eksp. Teor. Fiz. 44, 50 (1986) JETP Lett. 44, 63 (1986)]. 15] I. Kogan, Pis'ma Zh. Eksp. Teor. Fiz. 44, 209 (1986) JETP Lett. 44, 267 (1986)]. 16] P. O. Mazur, Gen. Relativ. Gravit. 19, 1173 (1987). 17] P. O. Mazur, Phys. Rev. Lett. 57, 929 (1986); 59, 2380 (1987). 18] V. F. Mukhanov, in Complexity, Entropy, and the Physics of Information , SFI Studies in the Sciences of Complexity, Vol. III, edited by W. H. Zurek (Addison-Wesley, New York, 1990). 19] U. H. Danielsson and M. Schi er, Phys. Rev. D 48, 4779 (1993). 20] J. Garc a{Bellido, \Quantum Black Holes," Report SU-ITP-93/4, hep-th/9302127. 21] Y. Peleg, \Quantum dust black holes," Report BRX-TH-350, hep-th/9307057. 22] M. Maggiore, Nucl. Phys. B429, 205 (1994). (gr-qc/9401027) 23] I. Kogan, \Black Hole Spectrum, Horison Quantization and All That," Report OUTP{ 94{39P, hep-th/9412232. 24] C. O. Lousto, Phys. Rev. D 51, 1733 (1995). (gr-qc/9405048) 25] Y. Peleg, Phys. Lett. B 356, 462 (1995). 26] J. D. Bekenstein and V. F. Mukhanov, Phys. Lett. B 360, 7 (1995). (gr-qc/9505012) 27] P. O. Mazur, Acta Phys. Pol. B 26, 1685 (1995). (hep-th/9602044) 28] P. O. Mazur, \Gravitation, the Quantum, and Cosmological Constant," e-print hepth/9603014. 29] H. A. Kastrup, \On the quantum levels of isolated spherically symmetric gravitational systems," Report PITHA 96/16, gr-qc/9605038. 30] A. Barvinsky and G. Kunstatter, \Exact Physical Black Hole States in Generic 2-D Dilaton Gravity," e-print hep-th/9606134. 28

31] A. Barvinsky and G. Kunstatter, \Mass Spectrum for Black Holes in Generic 2-D Dilaton Gravity," e-print gr-qc/9607030. 32] C. W. Misner, K. S. Thorne, and J. A. Wheeler, Gravitation (Freeman, San Francisco, 1973). 33] C. O. Lousto, Phys. Lett. B 352, 228 (1995). (gr-qc/9411037) 34] M. Barreira, M. Carfora, and C. Rovelli, Gen. Relativ. Gravit., to be published. (grqc/9603064). 35] D. Kramer, H. Stephani, E. Herlt, and M. MacCallum, Exact Solutions of Einstein's Field Equations , edited by E. Schmutzer (Cambridge University Press, Cambridge, England, 1980), Sec. 13.4. 36] T. Thiemann and H. A. Kastrup, Nucl. Phys. B399, 211 (1993). (gr-qc/9310012) 37] H. A. Kastrup and T. Thiemann, Nucl. Phys. B425, 665 (1994). (gr-qc/9401032) 38] K. V. Kuchar, Phys. Rev. D 50, 3961 (1994). (gr-qc/9403003) 39] J. Louko and B. F. Whiting, Phys. Rev. D 51, 5583 (1995). (gr-qc/9411017) 40] N. Dunford and J. S. Schwartz, Linear Operators (Interscience, New York, 1963), Vol. II. 41] M. Reed and B. Simon, Methods of Modern Mathematical Physics (Academic, New York, 1975), Vol. II. 42] K. Lake, Phys. Rev. D 19, 421 (1979). 43] M. Ba~ados, C. Teitelboim, and J. Zanelli, Phys. Rev. D 49, 975 (1994). (gr-qc/9307033) n 44] T. Thiemann, Int. J. Mod. Phys. D 3, 293 (1994). 45] T. Thiemann, Nucl. Phys. B436, 681 (1995). 46] J. Louko and S. N. Winters-Hilt, Phys. Rev. D 54, 2647 (1996). (gr-qc/9602003) 47] S. W. Hawking and G. F. R. Ellis, The Large Scale Structure of Space-time (Cambridge University Press, Cambridge, England, 1973). 48] G. Arfken, Mathematical Methods for Physicists , 2nd edition (Academic, New York, 1970). 49] N. M. J. Woodhouse, Geometric Quantization (Clarendon Press, Oxford, 1980). 50] C. J. Isham, in Relativity, Groups and Topology II: Les Houches 1983, edited by B. S. DeWitt and R. Stora (North-Holland, Amsterdam, 1984). 51] A. Ashtekar, Lectures on Non-Perturbative Canonical Gravity (World Scienti c, Singapore, 1991). 52] M. Cavaglia, V. de Alfaro, and A. T. Filippov, Int. J. Mod. Phys. D 4, 661 (1995). (gr-qc/9411070) 53] M. Cavaglia, V. de Alfaro, and A. T. Filippov, \Quantization of the Schwarzschild Black Hole," Report DFTT 50/95, gr-qc/9508062. 54] J. L. Friedman, K. Schleich, and D. M. Witt, Phys. Rev. Lett. 71, 1486 (1993). 55] Handbook of Mathematical Functions, edited by M. Abramowitz and I. A. Stegun (Dover, New York, 1965). 56] M. J. Gotay and J. Demaret, Phys. Rev. D 28, 2402 (1983). 57] E. Poisson and W. Israel, Phys. Rev. D 41, 1796 (1990). 58] A. Bonanno, S. Droz, W. Israel, and S. M. Morsink, Phys. Rev. D 50, 7372 (1994). (gr-qc/9403019) 59] D. Marolf, Class. Quantum Grav. 12, 1199 (1995). (gr-qc/9404053) 60] G. T. Horowitz and D. Marolf, Phys. Rev. D 52, 5670 (1995). (gr-qc/9504028) 61] G. T. Horowitz and R. Myers, Gen. Relativ. Gravit. 27, 915 (1995). (gr-qc/9503062) 29

62] S. Droz, W. Israel, and S. M. Morsink, Phys. World 1996(1), 34 (1996). 63] J. L. Friedman, I. H. Redmount, and S. N. Winters-Hilt (unpublished). 64] I. H. Redmount, talk given at the 3rd Midwest Relativity Meeting, Oakland, Michigan, 1993. 65] I. S. Gradshteyn and I. M. Ryzhik, Table of Integrals, Series and Products, 4th edition (Academic Press, New York, 1980). 66] A. Galindo and P. Pascual, Quantum Mechanics (Springer, Berlin, 1991), Vol. II. 67] F. W. J. Olver, J. Res. Nat. Bur. Stand. 63B, 131 (1959). 68] F. V. Atkinson, Univ. Nac. Tucuman. Revista A 8, 71 (1951).

30