You are on page 1of 6


Scalar, electromagnetic, and Weyl perturbations of BTZ black holes: Quasinormal modes
Vitor Cardoso* and José P. S. Lemos†
CENTRA, Departamento de Fı́sica, Instituto Superior Técnico, Av. Rovisco Pais 1, 1096 Lisboa, Portugal
共Received 13 January 2001; published 22 May 2001兲
We calculate the quasinormal modes and associated frequencies of the Bañados-Teitelboim-Zanelli 共BTZ兲
nonrotating black hole. This black hole lives in 2⫹1 dimensions in an asymptotically anti–de Sitter spacetime.
We obtain exact results for the wave function and quasinormal frequencies of scalar, electromagnetic and Weyl
共neutrino兲 perturbations.

DOI: 10.1103/PhysRevD.63.124015 PACS number共s兲: 04.30.⫺w

I. INTRODUCTION very interesting aspect of these black hole solutions is related

to the AdS-CFT 共conformal field theory兲 conjecture 关9兴. For
When one is describing the evolution of some conserva- instance, due to this AdS-CFT duality, quasi-normal frequen-
tive system, one often considers a small perturbation or a cies in the BTZ black hole spacetime yield a prediction for
small departure from a known solution of the system, and the thermalization time scale in the dual two-dimensional
one generally arrives at a wave equation describing it. For a CFT, which otherwise would be very difficult to compute
system with no explicit time dependence, one finds the nor- directly. If one has, e.g., a 10-dimensional type IIB super-
mal mode solutions of the wave equation, satisfying certain gravity, compactified into a BTZ⫻S 3 ⫻T 4 spacetime, the
boundary conditions, and one can then specify completely scalar field used to perturb the BTZ black hole can be seen as
the perturbation as a linear superposition of these normal a type IIB dilaton which couples to a CFT field operator O.
modes. In this case the operator associated with the pertur- Now, the BTZ in the bulk corresponds to a thermal state in
bation is self-adjoint, the frequencies are real and the modes the boundary CFT, and thus the bulk scalar perturbation cor-
are complete. responds to a thermal perturbation with nonzero 具 O典 in the
However, when one deals with open dissipative systems, CFT.
as is the case in this paper, such an expansion is not possible. There has been some recent work on perturbations of
Instead of normal modes, one considers quasinormal modes Schwarszchild AdS spacetimes: Horowitz and Hubeny 关11兴
共QNM兲 for which the frequencies are no longer pure real, studied the QNM frequencies for scalar perturbations in 4, 5
signaling that the system is losing energy. Although QNMs and 7 dimensions. Wang et al. 关12,13兴 studied scalar pertur-
are in general not complete and therefore insufficient to fully bations and QNMs on a Reissner-Nördstrom geometry, Chan
describe the dynamics 共see 关1,2兴 and references therein兲, they and Mann 关14兴 studied the QNM frequencies for a confor-
nevertheless dominate the signal during the intermediate mally coupled scalar field. For work on BTZ black holes
stages of the perturbation, being therefore extremely impor- such as entropy of scalar fields, see 关15兴 and references
tant. therein.
QNMs of black holes were first numerically computed by In this paper we shall consider the QNMs of the 3D non-
Chandrasekhar and Detweiler 关3兴, and subsequent numerical rotating BTZ black hole 关8兴. The non-rotating BTZ black
simulations 关4–6兴 showed that the amplitude is dominated, at hole metric for a spacetime with negative cosmological con-
intermediate times, by a ringing signal due to the QNMs. stant, ⌳⫽⫺1/l 2 , is given by
Aside from the pure mathematical interest, black hole’s

冉 冊 冉 冊
QNM calculations have been a very active field, and new ⫺1
methods, both numerical and analytical, have been devel- r2 r2
ds ⫽ ⫺M ⫹
dt ⫺ ⫺M ⫹
dr 2 ⫺r 2 d ␾ 2 , 共1兲
oped 共see 关7兴 for a review兲. l2 l2
Up until very recently, all these works dealt with asymp-
totically flat spacetimes. In the past few years there has been
a growing interest in asymptotically AdS 共anti–de Sitter兲 where M is the black hole mass. The horizon radius is given
spacetimes. Indeed, the Bañados-Teitelboim-Zanelli 共BTZ兲 by r ⫹ ⫽M 1/2l. We shall in what follows suppose that the
black hole in 2⫹1 dimensions 关8兴, as well as black holes in scalar, electromagnetic and Weyl 共neutrino兲 fields are a per-
共3⫹1兲-dimensional AdS spacetimes with nontrivial topology turbation, i.e., they propagate in a spacetime with a BTZ
共see, e.g. 关10兴兲, share with asymptotically flat spacetimes the metric. We will find that all these fields obey a wave equa-
common property of both having well defined charges at tion and the associated QNM are exactly soluble yielding
infinity, such as mass, angular momentum, and electromag- certain hypergeometric functions. As for the frequencies one
netic charges, which makes them a good testing ground has exact and explicit results for scalar and electromagnetic
when one wants to go beyond asymptotic flatness. Another perturbations and numerical results for Weyl perturbations.
To our knowledge, this is the first exact solution of QNMs
for a specific model 共see 关1兴兲.
*Email address: In Sec. II we give the wave equation for scalar and elec-

Email address: tromagnetic perturbations, and find the QNMs themselves

0556-2821/2001/63共12兲/124015共6兲/$20.00 63 124015-1 ©2001 The American Physical Society


and their frequencies. In Sec. III we find the wave equation

for Dirac and Weyl 共neutrino兲 perturbations and analyze their
QNMs. 冋 册g m 共 t,r 兲
A ␮ 共 t,r, ␾ 兲 ⫽ h 共 t,r 兲 e im ␾ ,

k m 共 t,r 兲

II. PERTURBING A BLACK HOLE WITH SCALAR where m is again our angular quantum number, and this de-
AND ELECTROMAGNETIC FIELDS composition is similar to the one in eigenfunctions of the
A. The wave equation total angular momentum in flat space 关16兴.
However, going through the same steps one finds that the
In this subsection we shall analyze the scalar and electro- equation for electromagnetic perturbations is the same as the
magnetic perturbations, which as we shall see yield the same one for scalars, Eq. 共5兲. The reason is that in three dimen-
effective potential, and thus the same wave equation. sions the 2-form Maxwell field F⫽F ␮ ␯ dx ␮ ⵩dx ␯ is dual to a
First, for scalar perturbations, we are interested in solu-
1-form d⌽.
tions to the minimally coupled scalar wave equation

⌽ ␮ ; ␮ ⫽0, 共2兲 B. QNMs for scalar and electromagnetic perturbations

Although a precise mathematical definition for a QNM
where a comma stands for ordinary derivative and a semi- can be given, as a pole in the Green’s function 关7兴, we shall
colon stands for covariant derivative. We make the following follow a more phenomenological point of view. A QNM
ansatz for the field ⌽: describes the decay of the field in question. For Eq. 共5兲 it is
defined as a corresponding solution which 共i兲 near the hori-
1 zon is purely ingoing, ⬃e i ␻ r * , corresponding to the exis-
⌽⫽ f 共 r 兲 e ⫺i ␻ t e im ␾ , 共3兲 tence of a black hole, and 共ii兲 near infinity is purely outgoing,
r 1/2
⬃e ⫺i ␻ r * 共no initial incoming wave from infinity is allowed兲.
One can see that the potential V(r) diverges at infinity, so we
where m is the angular quantum number. It is useful to use require that the perturbation vanishes there 共note that r⫽⬁
the tortoise coordinate r defined by the equation dr corresponds to a finite value of r , namely r ⫽0). This
* *
⫽dr/ 关 ⫺M ⫹(r 2 /l 2 ) 兴 , and given implicitly by * *
vanishing of the solution at ⬁ will only be possible for a
discrete set of complex frequencies ␻ called quasinormal
r⫽⫺M 1/2coth共 M 1/2r 兲 , 共4兲 frequencies.
with r ⑀ ]⫺⬁,0] (r ⫽⫺⬁ corresponds to r⫽r ⫹ , and 1. Exact calculation
* *
r ⫽0 corresponds to r⫽⬁).
* Putting l⫽1 and using the coordinate r , the wave equa-
With the ansatz 共3兲 and the tortoise coordinate r , Eq. 共2兲 *
is given by
* tion 共5兲 takes the form

d2 f 共r兲
dr 2
⫹ 关 ␻ ⫺V 共 r 兲兴 f 共 r 兲 ⫽0, 共5兲
⳵2 f 共r兲
⳵r 2

⫹ ␻ 2⫺
4 sinh共 M
r 兲

4 cosh共 M 1/2r 兲 2

* 2
where ⫹ f 共 r 兲 ⫽0. 共9兲
cosh共 M 1/2r 兲 2
3r 2 M M2 m2 M m2 Ongoing to a new variable x⫽1/cosh(M 1/2r ) 2 , x ⑀ 关 0,1兴
V共 r 兲⫽ ⫺ ⫺ ⫹ ⫺ , 共6兲 Eq. 共9兲 can also be written as
4l 4 2l 2 4r 2 l2 r2
4x 共 1⫺x 兲 ⳵ 2x f 共 r 兲 ⫹ 共 4⫺6x 兲 ⳵ x ␺ ⫹V̄ 共 x 兲 f 共 r 兲 ⫽0, 共10兲
and it is implicit that r⫽r(r ). The rescaling to the radial
coordinate r̂⫽r/l and to the frequency ␻ ˆ ⫽ ␻ l is equivalent where
to take l⫽1 in Eqs. 共5兲 and 共6兲, i.e., through this rescaling
one measures the frequency and other quantities in terms of
the AdS length scale l. V̄ 共 x 兲 ⫽
4x 共 1⫺x 兲 冋
4 ␻ 2 共 1⫺x 兲
⫺3x⫺x 共 1⫺x 兲

Now, the electromagnetic perturbations are governed by
Maxwell’s equations 4m 2 x 共 1⫺x 兲
⫺ . 共11兲
F ␮ ␯ ; ␯ ⫽0, with F ␮ ␯ ⫽A ␯ , ␮ ⫺A ␮ , ␯ , 共7兲
By changing to a new wave function y 共see 关18兴 for details兲,
where F ␮ ␯ is the Maxwell tensor and A ␮ is the electromag- through
netic potential. As the background is circularly symmetric, it
共 x⫺1 兲 3/4
would be advisable to expand A ␮ in 3-dimensional vector f→ 1/2 y, 共12兲
spherical harmonics 共see 关16兴 and 关17兴兲: x i ␻ /2M


Eq. 共10兲 can be put in the canonical form 关18,19兴 TABLE I. Lowest (n⫽0) QNM frequencies for m⫽0.

x 共 1⫺x 兲 y ⬙ ⫹ 关 c⫺ 共 a⫹b⫹1 兲 x 兴 y ⬘ ⫺aby⫽0, 共13兲 m⫽0

Numerical Exact
with a⫽1⫹im/2M 1/2⫺i ␻ /2M 1/2, b⫽1⫺im/2M 1/2 M 1/2 ␻r ⫺␻i ␻r ⫺␻i
⫺i ␻ /2M , and c⫽1⫺i ␻ /M , which is a standard hyper-
1/2 1/2
geometric equation. The hypergeometric equation has three 2 0.000 1.000 0 1
regular singular points at x⫽0,x⫽1,x⫽⬁, and has two inde- 1 0.000 2.000 0 2
pendent solutions in the neighborhood of each singular point. 5 0.000 10.000 0 10
We are interested in solutions of Eq. 共13兲 in the range 关0,1兴, 10 0.000 20.000 0 20
satisfying the boundary conditions of ingoing waves near x 50 0.000 100.000 0 100
⫽0, and zero at x⫽1. One solution may be taken to be 100 0.000 200.000 0 200
1000 0.000 2000.000 0 2000
y⫽ 共 1⫺x 兲 c⫺a⫺b F 共 c⫺a,c⫺b,c;x 兲 , 共14兲

where F is a standard hypergeometric function of the second The problem is reduced to that of finding a numerical solu-
kind. Imposing y⫽0 at x⫽1, and recalling that F(a,b,c,1) tion of the polynomial Eq. 共19兲. The numerical roots for ␻ of
⫽⌫(c)⌫(c⫺a⫺b)/⌫(c⫺a)⌫(c⫺b), we get Eq. 共19兲 can be evaluated resorting to numerical computa-
tion. Obviously, one cannot determine the full sum in expres-
a⫽⫺n, or b⫽⫺n, 共15兲 sion 共19兲, so we have to determine a partial sum from 0 to N,
say, and find the roots ␻ of the resulting polynomial expres-
with n⫽0,1,2, . . . , so that the quasinormal frequencies are sion. We then move on to the next term N⫹1 and determine
given by the roots. If the method is reliable, the roots should converge.
We have stopped our search when a 3 decimal digit precision
␻ ⫽⫾m⫺2iM 1/2共 n⫹1 兲 . 共16兲 was achieved. We have computed the lowest frequencies for
some parameters of the angular quantum number m and ho-
The lowest frequencies, namely those with n⫽0 and m⫽0 rizon radius r ⫹ . The frequency is written as ␻ ⫽ ␻ r ⫹i ␻ i ,
had already been obtained by 关20兴 and agree with our results. where ␻ r is the real part of the frequency and ␻ i is its imagi-
nary part.
2. Numerical calculation of the frequencies
In Tables I and II we list the numerical values of the
In order to check our results, we have also computed nu- lowest QNM frequencies, for m⫽0 and m⫽1, respectively,
merically the frequencies. By going to a new variable z and for selected values of the black hole mass.
⫽1/r, h⫽1/r ⫹ one can put the wave equation 共5兲 in the The numerical results agree perfectly with Eq. 共16兲, and
form 共see 关11兴 for further details兲 one sees that the imaginary part of the frequency scales with
the horizon whereas the real part depends only on the angu-
d2 d lar index m.
s共 z 兲 ⌰⫹t 共 z 兲 ⌰⫹u 共 z 兲 ⌰⫽0, 共17兲
dz 2 dz
where ⌰⫽e * a(r), s(z)⫽z ⫺M z , 2
t(z)⫽2M z 4 3

⫺2i ␻ z 2 and u(z)⫽V/ 关 ⫺M ⫹(1/z 2 ) 兴 , with V given by Eq. A. The wave equation
共6兲. Now, z ⑀ 关 0,h 兴 and one sees that in this range, the
We shall develop Dirac’s equation for a massive spinor,
differential equation has only regular singularities at z⫽0
and then specialize to the massless case. The two component
and z⫽h, so it has by Fuchs theorem a polynomial solution.
massive spinor field ⌿, with mass ␮ s obeys the covariant
We can now use Fröbenius method 共see for example 关21兴兲
⬁ Dirac equation
and look for a solution of the form ⌰(z)⫽ 兺 n⫽0 ␪ n( ␻ ) (z

⫺h) (z⫺h) , where ␣ is to be taken from the boundary
TABLE II. Lowest (n⫽0) QNM frequencies for m⫽1.
conditions. Using the boundary condition of only ingoing
waves at the horizon, one sees 关11兴 that ␣ ⫽0. So the final m⫽1
outcome is that ⌰ can be expanded as Numerical Exact
M 1/2 ␻r ⫺␻i ␻r ⫺␻i

⌰共 z 兲⫽ 兺
␪ n( ␻ ) 共 z⫺h 兲 n . 共18兲 1
2 1.000 1.000 1 1
1 1.000 2.000 1 2
5 1.000 10.000 1 10
Imposing now the second boundary condition, ⌰⫽0 at in-
10 1.000 20.000 1 20
finity (z⫽0) one gets
50 1.000 100.000 1 100
⬁ 100 1.000 200.000 1 200

␪ n( ␻ ) 共 ⫺h 兲 n ⫽0. 共19兲 1000 1.000 2000.000 1 2000


i ␥ ␮ ⵜ ␮ ⌿⫺ ␮ s ⌿⫽0, 共20兲 R 1 ⫽e i ␯ /2 ⌼ 1 , 共30兲

where ⵜ ␮ is the spinor covariant derivative defined by ⵜ ␮ R 2 ⫽e ⫺i ␯ /2 ⌼ 2 , 共31兲

⫽ ⳵ ␮ ⫹ 41 ␻ ␮ab ␥ [a ␥ b] , and ␻ ␮ab is the spin connection, which
may be given in terms of the tryad e a␮ .
As is well known there are two inequivalent two dimen-
sional irreducible representations of the ␥ matrices in three
r̂ ⫽r ⫹
* *
␮ sr

, 冉 冊 共32兲

spacetime dimensions. The first may be taken to be ␥ 0

⫽i ␴ 2 , ␥ 1 ⫽ ␴ 1 , and ␥ 2 ⫽ ␴ 3 , where the matrices ␴ k are the we get
Pauli matrices. The second representation is given in terms
of the first by a minus sign in front of the Pauli matrices. 共 ⳵ r̂ ⫺i ␻ 兲 ⌼ 2 ⫽W⌼ 1 , 共33兲
From Eq. 共20兲, one sees that a Dirac particle with mass ␮ s in
the first representation is equivalent to a Dirac particle with 共 ⳵ r̂ ⫺i ␻ 兲 ⌼ 2 ⫽W⌼ 2 , 共34兲
mass ⫺ ␮ s in the second representation. To be definitive, we *
will use the first representation, but the results can be inter- where
changed to the second one, by substituting ␮ s →⫺ ␮ s . For
Weyl particles, ␮ s ⫽0, both representations yield the same
results. i⌬ 1/2共 m̂ 2 ⫹ ␮ s2 r 2 兲 3/2
W⫽ . 共35兲
Again, one can separate variables by setting m̂ ␮ s ⌬
r 共 m̂
2 2
⫹ ␮ s2 r 2 兲 ⫹

冋 册
⌿ 1 共 t,r 兲
⌿ 共 t,r, ␾ 兲 ⫽ e ⫺i ␻ t e im ␾ . 共21兲
⌿ 2 共 t,r 兲 Finally, putting Z ⫾ ⫽⌼ 1 ⫾⌼ 2 we have
On substituting this decomposition into Dirac’s equation 共20兲 2
共 ⳵ r̂ ⫹ ␻ 2 兲 Z ⫾ ⫽V ⫾ Z ⫾ , 共36兲
we obtain *

i 共 M ⫺2r 2 兲 r 2␻ with
⫺ r⌿ 2 ⫹i⌬ 1/2⳵ r ⌿ 2 ⫹ ⌿ 2 ⫽ 共 m⫹ ␮ s 兲 ⌿ 1 ,
2⌬ 1/2
⌬ 1/2

共22兲 dW
V ⫾ ⫽W 2 ⫾ . 共37兲
i 共 M ⫺2r 2 兲 r 2␻ *
⫺ r⌿ 1 ⫹i⌬ 1/2⳵ r ⌿ 1 ⫹ ⌿ 1 ⫽ 共 m⫹ ␮ s 兲 ⌿ 2 ,
2⌬ 1/2 ⌬ 1/2 We shall be concerned with massless spinors ( ␮ s ⫽0) for
共23兲 which r̂ ⫽r , and W⫽i⌬ 1/2m̂/r 2 . Thus
* *
where we have put ⌬⫽⫺M r ⫹r /l , we have restored the
冉 冊 冉 冊
2 4 2 1/2
m2 r2 M m r2
AdS length scale l, and in general we follow Chan- V ⫾⫽ ⫺M ⫾ ⫺M . 共38兲
drasekhar’s notation 关22兴. Defining R 1 , R 2 , and m̂ through r2 l 2
r2 l 2

the relations
In the form 共37兲 one immediately recognizes that the two
⌿ 1 ⫽i⌬ ⫺1/4R 1 , 共24兲 potentials V ⫹ and V ⫺ should yield the same spectrum. In
fact they are, in SUSY language, superpartner potentials de-
⌿ 2 ⫽⌬ ⫺1/4R 2 , 共25兲 rived from a superpotential W 共see 关23兴兲. Once again, we can
rescale r and take l⫽1, by measuring everything in terms of
m⫽im̂, 共26兲 l.

we obtain B. QNMs for Weyl perturbations

i⌬ 1/2 Similarly, the wave equation 共36兲 for Weyl 共until recently
共 ⳵ r ⫺i ␻ 兲 R 2 ⫽ 共 m̂⫺i ␮ s r 兲 R 1 , 共27兲 also called neutrino兲 perturbations may be put in the form

冋 冉
i⌬ 1/2 ⳵ r2 Z ⫾ ⫹ ␻ 2 ⫺m
共 ⳵ r ⫹i ␻ 兲 R 1 ⫽ 共 m̂⫹i ␮ s r 兲 R 2 . 共28兲 * cosh共 M 1/2r 兲 2

* 2
sinh共 M 1/2
r 兲 2
⫾M 1/2 * Z ⫾ ⫽0. 共39兲
Defining now ␯ , ⌼ 1 ⌼ 2 , and r̂ through the relations cosh共 M 1/2
r 兲2

冉 冊
␮ sr Going to a new independent variable, x⫽⫺sinh(M 1/2r ), x
␯ ⫽arctan , 共29兲 *
m̂ ⑀ 关 ⬁,0兴 , we can write


冋 册
TABLE III. Lowest QNM frequencies for m⫽1.
␻ 2 共 1⫹x 2 兲 m 2 mx
⫺ ⫾
M M M 1/2 m⫽1
共 1⫹x 2 兲 Z ⬙ ⫹xZ ⬘ ⫹ 2
Z⫽0. Numerical
共40兲 M 1/2 ␻r ⫺␻i

2 0.378 2.174
By changing the wave function Z to ␹ 5 0.316 5.027
10 0.224 10.006
1/2x⫺2m)/2M 1/2⫺x/2] 50 0.099 50.001
␹ ⫽e [(M arctan共 x 兲 , 共41兲
100 0.071 100.000
500 0.0316 500.000
we have

冉 冊 冉 冊
2m ␻2
共 1⫹x 兲 ␹ ⬙ ⫹
⫹x ␹ ⬘ ⫹ ␹ ⫽0. 共42兲 We have computed the scalar, electromagnetic and neu-
M 1/2 M trino QNM of BTZ black holes. These modes dictate the late
time behavior of the fields. In all cases, these modes scale
On putting s⫽(1⫹iz)/2, s ⑀ 关 21 ,i⬁ 兴 , we have again the with the horizon radius, at least for large black holes and,
hypergeometric equation 共13兲, with a⫽i ␻ /M 1/2, b⫽ since the decay of the perturbation has a time scale ␶
⫺i ␻ /M 1/2, and c⫽ 21 ⫾im/M 1/2, so that the solution to the ⫽1/␻ i , this means that the greater the mass, the less time it
wave equation is again specified around each singular point, takes to approach equilibrium. We have also found that for
and is given by the analytic continuation of the standard large black holes, the QNM frequencies are proportional to
hypergeometric function to the complex plane 关18,19兴. the black hole radius. Since the temperature of a BTZ black
Since infinity is located at s⫽ 21 , there is no easy way to is proportional to the black hole radius, the QNM frequen-
determine the QNM frequencies, so we have to resort to cies scale with the temperature, as a simple argument indi-
numerical calculations. If we put Eq. 共35兲 in the form 共17兲 cates 关11兴. For the study of QNM of 共3⫹1兲-dimensional
one again sees that it has no essential singularities, so the spherical, as well as the toroidal black holes found by Lemos
numerical method just outlined in the previous section may 关24兴 共see 关25兴兲.
be applied. Moreover, since V ⫹ and V ⫺ have the same spec- Is there, for small black holes, any relation with these
trum 关23兴 and the same QNM frequencies 关22兴 we need only quasinormal modes and critical phenomena as speculated by
to work out the frequencies for one of them. In Table III we Horowitz and Hubeny 关11兴? Though that would be an ex-
present the numerical results for the QNM frequencies for tremely interesting relation, we still are not able to answer
neutrino perturbations and for selected values of the black that.
hole mass.
For large black holes one can see that the imaginary part
of the frequencies scale with the horizon (r ⫹ ⫽M 1/2), just as This work was partially funded by FCT through project
in the scalar and electromagnetic case. We have also com- Sapiens. One of us 共V.C.兲 acknowledges financial support
puted some higher modes, and the real part of the frequency from the Portuguese FCT through PRAXIS XXI program.
␻ r , does not seem to depend on which mode we are dealing J.P.S.L. thanks Observatório Nacional do Rio de Janeiro for
with, just as in the scalar and electromagnetic case. hospitality.

关1兴 H. R. Beyer, Commun. Math. Phys. 204, 397 共1999兲. 关9兴 J. Maldacena, Adv. Theor. Math. Phys. 2, 253 共1998兲.
关2兴 E. S. C. Ching, P. T. Leung, W. M. Suen, S. S. Tong, and K. 关10兴 J. P. S. Lemos, in ‘‘Proceedings of the Xth Portuguese As-
Young, Rev. Mod. Phys. 70, 1545 共1998兲. tronomy and Astrophysics’’ Meeting 共World Scientific, Sin-
关3兴 S. Chandrasekhar and S. Detweiler, Proc. R. Soc. London gapore, in press兲, gr-qc/0011092.
A344, 441 共1975兲. 关11兴 G. T. Horowitz and V. Hubeny, Phys. Rev. D 62, 024027
关4兴 S.L. Detweiler and E. Szedenits, Astrophys. J. 231, 211
关12兴 B. Wang, C. Y. Lin, and E. Abdalla, Phys. Lett. B 481, 79
关5兴 Sources of Gravitational Radiation, edited by S. L. Smarr
共Cambridge University Press, Cambridge, England, 1979兲. 共2000兲.
关6兴 P. Anninos, D. Hobill, E. Seidel, L. Smarr, and W. M. Suen, 关13兴 B. Wang, C. M. Mendes, and E. Abdalla, Phys. Rev. D 63,
Phys. Rev. Lett. 71, 2851 共1993兲. 084001 共2001兲.
关7兴 K. D. Kokkotas and B. G. Schmidt, Living Rev. Relativ. 2, 2 关14兴 S. F. J. Chan and R. B. Mann, Phys. Rev. D 55, 7546 共1997兲.
共1999兲. 关15兴 I. Ichinose and K. Satoh, Nucl. Phys. B447, 340 共1995兲.
关8兴 M. Bañados, C. Teitelboim, and J. Zanelli, Phys. Rev. Lett. 69, 关16兴 A. R. Edmonds, Angular Momentum in Quantum Mechanics
1849 共1992兲. 共Princeton University Press, Princeton, NJ, 1957兲.


关17兴 R. Ruffini, Black Holes: les Astres Occlus 共Gordon and Physicists 共Academic, New York, 1995兲.
Breach, New York, 1973兲. 关22兴 S. Chandrasekhar, The Mathematical Theory of Black Holes
关18兴 A. F. Nikiforov and V. B. Uvarov, Special Functions of Math- 共Oxford University Press, New York, 1983兲.
ematical Physics 共Birkhäuser, Boston, 1988兲. 关23兴 F. Cooper, A. Khare, and U. Sukhatme, Phys. Rep. 251, 267
关19兴 M. Abramowitz and I. A. Stegun, Handbook of Mathematical 共1995兲.
Functions 共Dover, New York, 1970兲. 关24兴 J. P. S. Lemos and V. T. Zanchin, Phys. Rev. D 54, 3840
关20兴 T. R. Govindarajan and V. Suneeta, Class. Quantum Grav. 18, 共1996兲.
265 共2001兲. 关25兴 V. Cardoso and J. P. S. Lemos 共in preparation兲.
关21兴 G. B. Arfken and H. J. Weber, Mathematical Methods for