Gravitational perturbations of the Schwarzschild spacetime: A practical covariant and

gauge-invariant formalism
Karl Martel and Eric Poisson
Department of Physics, University of Guelph, Guelph, Ontario, Canada N1G 2W1
(Dated: January 23, 2004)
We present a formalism to study the metric perturbations of the Schwarzschild spacetime. The for-
malism is gauge invariant, and it is also covariant under two-dimensional coordinate transformations
that leave the angular coordinates unchanged. The formalism is applied to the typical problem of
calculating the gravitational waves produced by material sources moving in the Schwarzschild space-
time. We examine the radiation escaping to future null infinity as well as the radiation crossing
the event horizon. The waveforms, the energy radiated, and the angular-momentum radiated can
all be expressed in terms of two gauge-invariant scalar functions that satisfy one-dimensional wave
equations. The first is the Zerilli-Moncrief function, which satisfies the Zerilli equation, and which
represents the even-parity sector of the perturbation. The second is the Cunningham-Price-Moncrief
arXiv:gr-qc/0502028v1 8 Feb 2005

function, which satisfies the Regge-Wheeler equation, and which represents the odd-parity sector of
the perturbation. The covariant forms of these wave equations are presented here, complete with
covariant source terms that are derived from the stress-energy tensor of the matter responsible for
the perturbation. Our presentation of the formalism is concluded with a separate examination of
the monopole and dipole components of the metric perturbation.

I. INTRODUCTION this paper we hope to deliver its final chapter. We aim

to present the formalism of metric perturbations of the
Metric perturbations of the Schwarzschild spacetime Schwarzschild spacetime in its most mature and prac-
have been studied for a long time, starting with the pio- tical incarnation yet. Our formalism is covariant and
neering work of Regge and Wheeler [1], Vishveshwara [2], gauge invariant, and it goes beyond the work of Ger-
and Zerilli [3]. The theory was summarized in an influ- lach and Sengupta, and of Sarbach and Tiglio, by the
ential monograph by Chandrasekhar [4], and it has been development of covariant master equations for the even-
applied to many different physical situations (see, for ex- parity and odd-parity sectors of the theory, complete with
ample, Chapter 4 of the book by Frolov and Novikov [5]). covariant source terms that are derived from the stress-
In particular, a useful application has been the compu- energy tensor of the matter which is responsible for the
tation of gravitational waves produced by a point par- perturbation.
ticle moving in the field of a Schwarzschild black hole We have in mind a typical application of the per-
(see Ref. [6], the review article of Ref. [7], and references turbation formalism, the calculation of gravitational
therein). Another has been the simulation of a collision waves produced by material sources moving in the
of two black holes in a “close-limit” approximation (see Schwarzschild spacetime. We are interested in the ra-
the review article of Ref. [8] and references therein). diation that escapes to future null infinity and manifests
Traditionally the perturbation formalism is developed itself as waveforms h+ and h× that are directly observ-
in the standard Schwarzschild coordinates (t, r, θ, φ), and able to gravitational-wave detectors, and we are inter-
in a standard choice of gauge known as the “Regge- ested in the radiation that crosses the black-hole horizon.
Wheeler gauge.” The tradition also makes use of Fourier- This application illustrates well the need for a covariant
transform techniques and presents the perturbation equa- and gauge-invariant formalism, as the two types of radia-
tions in the frequency domain instead of the time do- tion admit different descriptions. The radiation at future
main. Moncrief [9] was the first to present the formalism null infinity is best described by casting the perturba-
in a gauge-invariant package, recognizing the practical tion in an outgoing radiation gauge and expressing it in
advantages that gauge invariance provides: While the a retarded coordinate system (u, r, θ, φ) related to the
Regge-Wheeler gauge is useful for many purposes, it is usual Schwarzschild coordinates by the transformation
not useful for others, and the power to switch from one u = t − r − 2M ln(r/2M − 1). The radiation at the hori-
gauge to another within a gauge-invariant framework is zon, on the other hand, is best described by adopting an
often required. Another refinement of the formalism was incoming radiation gauge and expressing the perturba-
produced by Gerlach and Sengupta [10], who presented tion in an advanced coordinate system (v, r, θ, φ) related
the gauge-invariant perturbation equations in an arbi- to the usual Schwarzschild coordinates by the transfor-
trary coordinate system, thus liberating the formalism mation v = t + r + 2M ln(r/2M − 1). Our formalism
from the usual Schwarzschild coordinates and their poor permits the use of any coordinate system xa = (x0 , x1 )
behavior at the event horizon. that can be obtained from the usual Schwarzschild coor-
This program to translate the traditional perturbation dinates (t, r); we do not, however, consider transforma-
formalism into a covariant, gauge-invariant language was tions of the angular coordinates θA = (θ, φ).
recently revived by Sarbach and Tiglio [11], and with In the formalism developed in this paper, the radiation
 2

at future null infinity and at the horizon are described be clearly delineated. We refer the reader to the litera-
in terms of scalar and gauge-invariant master functions, ture reviewed in this Introduction for additional details;
Ψlm a lm a
even (x ) and Ψodd (x ), which can be computed from another repository of relevant derivations is Martel’s PhD
the metric perturbations. These functions are labeled by dissertation [14]. Throughout this paper we adopt the
spherical-harmonic indices l and m, and by their behavior sign conventions of Misner, Thorne, and Wheeler [15],
under a parity transformation. The function Ψlm a
even (x ) and we set c = G = 1.
is constructed from the even-parity perturbations, and
it is equal to the gauge-invariant function that was first
introduced by Moncrief [9]; it is a close cousin to Zerilli’s II. SCHWARZSCHILD SPACETIME
original master function [3], and it satisfies a covariant
version of Zerilli’s differential equation. The complete The Schwarzschild metric is expressed as
source term for this equation is presented for the first
time in this paper. The function Ψlm a ds2 = gab dxa dxb + r2 ΩAB dθA dθB , (2.1)
odd (x ), on the other
hand, is constructed from the odd-parity perturbations,
in a form that is covariant under two-dimensional co-
and it is equal to the gauge-invariant function that was
ordinate transformations xa → x′a . The coordinates xa
first introduced by Cunningham, Price, and Moncrief [12]
span the submanifold M2 of the Schwarzschild spacetime
and recently revived by Jhingan and Tanaka [13]; it is es-
— the “(t, r) plane” — and lower-case Latin indices a,
sentially the time integral of the original Regge-Wheeler
b, c, etc. run over the values 0 and 1. The coordinates
master function [1], and it satisfies a covariant version of
θA = (θ, φ) span the two-spheres xa = constant, and
the Regge-Wheeler equation. The complete source term
upper-case Latin indices A, B, C, etc. run over the values
for this equation is presented for the first time in this
2 and 3. The full spacetime manifold is M = M2 × S 2 .
paper.
The two-dimensional tensor gab and the scalar r are func-
The paper is organized as follows. In Sec. II we give a tions of the coordinates xa , and ΩAB = diag(1, sin2 θ) is
covariant description of the Schwarzschild spacetime and the metric on the unit two-sphere.
specify our notations and conventions. In Sec. III we in- We shall use three different coordinate systems xa in
troduce the scalar, vector, and tensor spherical harmon- the applications of the perturbation formalism to be pre-
ics that are used in the decomposition of the metric per- sented below. The first is (t, r), the usual Schwarzschild
turbation. In Sec. IV we examine the even-parity sector coordinates. The second is (u, r), where the retarded-
of the perturbation, introduce the Zerilli-Moncrief master time coordinate u is defined by u = t−r −2M ln(r/2M −
function, and derive the one-dimensional wave equation 1). The third is (v, r), where the advanced-time coordi-
that it satisfies. In Sec. V we examine the odd-parity nate v is defined by v = t+r +2M ln(r/2M −1). In these
sector of the perturbation, introduce the Cunningham- coordinates the Schwarzschild metric takes the form
Price-Moncrief master function, and derive the one-
dimensional wave equation that it satisfies. In Sec. VI gab dxa dxb = f dt2 + f −1 dr2 , (2.2)
we describe the behavior of the perturbations near fu- = f du2 − 2 dudr, (2.3)
ture null infinity, construct the radiative part of the per-
= f dv 2 + 2 dvdr, (2.4)
turbation field, extract the waveforms h+ and h× , and
compute the rates at which the radiation carries away en- where f := 1−2M/r and M is the mass of the black hole.
ergy and angular momentum. In Sec. VII we describe the These systems share the property that the scalar r is
behavior of the perturbations near the black hole’s event adopted as one of the coordinates. Our formalism is not,
horizon, and calculate the rates at which they transfer however, limited to these coordinate choices; one retains
energy and angular momentum to the black hole. In the freedom of using any coordinate system whatever, for
Sec. VIII we conclude our presentation and examine the example, harmonic coordinates, isotropic coordinates, or
nonradiating pieces of the metric perturbation, those as- double-null coordinates.
sociated with the low multipoles l = 0 and l = 1. We introduce the dual vector
Various technical details are relegated to the Appen- ∂r
dices. In Appendix A we expand our discussion of vec- ra := , (2.5)
∂xa
torial and tensorial spherical harmonics. In Appendix
B we present the perturbed Ricci tensor for a general which is normal to the surfaces of constant r(xa ); in the
spherically-symmetric background spacetime. In Appen- coordinates of Eqs. (2.2)–(2.4), ra = (0, 1). We use g ab ,
dices C, D, and E we list the perturbation equations the inverse to gab , to raise its index: ra = g ab rb . This
in the commonly used coordinate systems xa = (t, r), allows us to give a covariant definition to the function f
xa = (u, r), and xa = (v, r), respectively. that appears in Eqs. (2.2)–(2.4):
Because the topic of metric perturbations of the 2M
Schwarzschild spacetime is so venerable, we will allow ra ra =: f = 1 − . (2.6)
ourselves in this paper to simply state our results and r
omit most lengthy derivations that lead to those results. We also introduce εab , the (antisymmetric) Levi-Civita
We hope, however, that the path to the results will always tensor on M2 ; in the coordinates of Eqs. (2.2)–(2.4),
 3

εtr = εur = εvr = 1. The timelike Killing vector of the Vectorial spherical harmonics come in two types. The
Schwarzschild spacetime is tangent to M2 and is given even-parity harmonics are defined by
by
YAlm := DA Y lm , (3.1)
a ab
t = −ε rb ; (2.7)
while the odd-parity harmonics are
in the coordinates of Eqs. (2.2)–(2.4), ta = (1, 0). We
lm
have ta ta = −f and ta ra = 0, and the vectors ra , ta XA := −εAB DB Y lm . (3.2)
form a basis on M2 . In terms of this basis we have
g ab = f −1 (−ta tb + ra rb ) and εab = −f −1 (ta rb − ra tb ). Their components are listed explicitly in Appendix A.
The covariant derivative operator compatible with gab The vectorial harmonics satisfy the orthogonality rela-
is denoted ∇a ; we thus have ∇a gbc ≡ 0. It is easy to tions
show that for the Schwarzschild solution, Z
A l′ m ′
Ȳlm YA dΩ = l(l + 1) δll′ δmm′ (3.3)
M
∇a ∇b r = gab , (2.8)
r2 and
2 ab
so that r = 2M/r , where  := g ∇a ∇b is the Lapla- Z
′ ′
A lm
cian operator on M2 . We also have ∇a tb = (M/r2 )εab , X̄lm XA dΩ = l(l + 1) δll′ δmm′ , (3.4)
which confirms that ta is a Killing vector. The Riemann
tensor on M2 is Rabcd = (2M/r3 )(gac gbd − gad gbc ). in which an overbar indicates complex conjugation and
We let ΩAB be the inverse to ΩAB , the metric on the dΩ := sin θ dθdφ is an element of solid angle. We also
unit two-sphere. The covariant derivative operator com- have
patible with ΩAB is denoted DA ; we thus have DA ΩBC ≡ Z
0. The Levi-Civita tensor on the unit two-sphere is de- A l′ m ′
Ȳlm XA dΩ = 0, (3.5)
noted εAB , and εθφ = sin θ. The Riemann tensor on the
unit-sphere is RABCD = ΩAC ΩBD − ΩAD ΩBD . which states that the even-parity and odd-parity har-
Covariant differentiation in the Schwarzschild space- monics are always orthogonal. The definitions (3.1) and
time can be defined in terms of covariant differentiation (3.2) for the vectorial spherical harmonics are identical
in the submanifolds M2 and S 2 . If Γabc is the connection to those provided by Regge and Wheeler [1].
associated with ∇a , and if ΓABC is the connection asso- Tensorial spherical harmonics come in the same two
ciated with DA , then it is easy to show that the nonvan- types. The even-parity harmonics are ΩAB Y lm and
ishing components of the spacetime connection are given
by 4 Γabc = Γabc , 4 ΓaBC = −rra ΩBC , 4 ΓABc = r−1 rc δ AB , lm
h 1 i
YAB := DA DB + l(l + 1)ΩAB Y lm , (3.6)
and 4 ΓABC = ΓABC . Using these rules we find that the 2
conservation identities for a stress-energy tensor T αβ in
while the odd-parity harmonics are
the Schwarzschild spacetime take the form
2 lm 1 C
ε DB + εBC DA DC Y lm .


∇b T ab + DB T aB + rb T ab − rra ΩAB T AB = 0 XAB := − (3.7)
r
(2.9) 2 A

and Their components are listed explicitly in Appendix A.

The tensorial harmonics satisfy the orthogonality rela-
4 tions
∇a T aA + DB T AB + ra T aA = 0 (2.10)
r
1
Z
AB l′ m′
when expressed in terms of the submanifold connections. Ȳlm YAB dΩ = (l − 1)l(l + 1)(l + 2) δll′ δmm′ (3.8)
2
and
III. SPHERICAL HARMONICS
1
Z
′ ′
AB l m
X̄lm XAB dΩ = (l − 1)l(l + 1)(l + 2) δll′ δmm′ . (3.9)
In this section we introduce the scalar, vector, and 2
tensor spherical harmonics that are used in the decom- We also have
position of the metric perturbation. The tensorial nature Z
of the spherical harmonics refers to the unit two-sphere, ′
AB l m
Ȳlm
′
XAB dΩ = 0 (3.10)
and in this section we use the metric ΩAB and its inverse
ΩAB to lower and raise all upper-case Latin indices. All
and
tensorial operations (including covariant differentiation)
shall refer to this metric. ΩAB YAB
lm
= 0 = ΩAB XAB
lm
. (3.11)
The scalar harmonics are the usual spherical-harmonic
functions Y lm (θA ). They satisfy the eigenvalue equation The definition (3.6) for the even-parity harmonics does
[ΩAB DA DB + l(l + 1)]Y lm = 0. not agree with that of Regge and Wheeler [1], who work
 4

instead with the set ΩAB Y lm and DA DB Y lm . We find it Under such a transformation the perturbation quantities
lm
more convenient to form the tracefree combinations YAB , change according to (see Appendix B)
which have the property of being (pointwise) orthogonal
to ΩAB Y lm . The definition (3.7) for the odd-parity har- hab → h′ab = hab − ∇a ξb − ∇b ξa , (4.6)
monics also differs from Regge and Wheeler’s, but only 2
by an overall minus sign, which we find convenient to ja → ja′ = ja − ξa − ∇a ξ + ra ξ, (4.7)
r
introduce. l(l + 1) 2
The tensorial harmonics YAB lm
and XAB lm
can be related K → K′ = K + ξ − ra ξa , (4.8)
r2 r
to the spherical-harmonic functions of spin-weight s = 2
±2 [16], and to the pure-spin harmonics used by Thorne G → G′ = G − 2 ξ, (4.9)
r
[17]. These relations are explored in Appendix A.
where we have discarded the spherical-harmonic labels
for brevity (we shall continue with this practice until the
IV. EVEN-PARITY SECTOR end of the section). It is easy to show that the combina-
tions
A. Perturbation fields and gauge transformations
h̃ab := hab − ∇a εb − ∇b εa (4.10)
The even-parity sector refers to those components of
the metric perturbation that can be expanded in terms of and
the even-parity spherical harmonics Y lm , YAlm , ΩAB Y lm ,
lm 1 2
and YAB . Introducing the notation 4 gab = gab + pab , K̃ := K + l(l + 1)G − ra εa (4.11)
4
gaB = paB , and 4 gAB = r2 ΩAB + pAB for the perturbed 2 r
metric, the even-parity sector of the metric perturbation are gauge invariant, where
is
1
εa := ja − r2 ∇a G.
X
pab = hlm
ab Y
lm
, (4.1) (4.12)
2
lm
Equations (4.7) and (4.9) reveal that one can always
X
paB = jalm YBlm , (4.2)
lm choose a gauge in which ja = 0 = G; this is the Regge-
X Wheeler gauge. Equations (4.10)–(4.12) imply that h̃ab =
pAB = r2 K lm ΩAB Y lm + Glm YAB
lm


. (4.3)
hab and K̃ = K in the Regge-Wheeler gauge.
lm

In this section the sums over l are taken to begin at

l = 2; the low multipoles (l = 0 and l = 1) will B. Perturbation equations
be considered separately in Sec. VIII. The fields hlm ab ,
jalm , K lm , and Glm are defined on M2 and they de- The Ricci tensor of the Schwarzschild spacetime van-
pend on the coordinates xa only. They are closely re- ishes, and as a consequence its perturbation is gauge in-
lated to the quantities first introduced by Regge and variant. Its computation can therefore be carried out in
Wheeler [1], who worked exclusively in terms of the usual any convenient gauge, and the Regge-Wheeler gauge is
Schwarzschild coordinates (t, r). In these coordinates clearly convenient. The steps involved are as follows. We
(discarding for brevity the spherical-harmonic labels) we substitute Eqs. (4.1)–(4.3), having set ja = G = 0, into
have htt = f H0 , htr = H1 , hrr = H2 /f , jt = h0 , and the Ricci tensor of Appendix B. We simplify the result
jr = h1 . The function G introduced in Eq. (4.3) is iden- and find that δRab is expanded in terms of Y lm , δRaB
tical to the corresponding Regge-Wheeler quantity, but in terms of YBlm , and δRAB in terms of both ΩAB Y lm
K is different: Khere = KRW − 21 l(l + 1)G; the difference and YABlm
. From the Ricci tensor we compute the Ein-
originates from the fact that Regge and Wheeler work stein tensor and set the result equal to 8πT αβ , which
with ΩAB Y lm and DA DB Y lm instead of ΩAB Y lm and describes the material source for the gravitational per-
lm
YAB . turbations. Each spherical-harmonic component of the
Even-parity gauge transformations are generated by a field equations can then be extracted by involving the
dual vector field Ξα = (Ξa , ΞA ) that is expanded as orthonormality relations (3.3) and (3.8) satisfied by the
X spherical harmonics. At the end of this calculation we
Ξa = ξalm Y lm , (4.4) take advantage of the fact that h̃ab = hab and K̃ = K
lm in the Regge-Wheeler gauge. This allows us to make
the substitutions hab → h̃ab , K → K̃ and therefore to
X
ΞA = ξ lm YAlm ; (4.5)
lm express the gauge-invariant Einstein tensor in terms of
gauge-invariant quantities.
the fields ξalm and ξ lm depend on the coordinates xa only. Our final results are
 5

1 1
1 2
Qab = ∇c ∇(a h̃cb) − gab ∇c ∇d h̃cd −  h̃ab − gab h̃ − ∇a ∇b h̃ + rc ∇(a h̃cb) − gab ∇d h̃cd


2 2 2  r 
1 c
l(l + 1) 1 cd 1 l(l + 1) 2M
− r ∇c h̃ab − gab h̃ + h̃ab − 2 gab rc rd h̃ − + 3 gab h̃
r 2r2 r 2 r2 r
2 3 (l − 1)(l + 2)
− ∇a ∇b K̃ + gab K̃ − r(a ∇b) K̃ + gab rc ∇c K̃ − gab K̃, (4.13)
r r 2r2
1
Qa = ∇c h̃ca − ∇a h̃ + ra h̃ − ∇a K̃, (4.14)
r
2 1 l(l + 1) 2
Q = h̃ − ∇a ∇b h̃ − ra ∇b h̃ab + ra ∇a h̃ −
♭ ab
2
h̃ + K̃ + ra ∇a K̃, (4.15)
r r 2r r
Q♯ = −h̃, (4.16)

where h̃ := g ab h̃ab and  := g ab ∇a ∇b . The source terms where

are
Z 6M
Λ := (l − 1)(l + 2) + . (4.24)
Q = 8π T ab Ȳ lm dΩ,
ab
(4.17) r
16πr2
Z
Qa = T aB ȲBlm dΩ, (4.18) This is a covariant and gauge-invariant generalization of
l(l + 1) the definition provided by Lousto and Price [18], who
Z
worked with the usual Schwarzschild coordinates and in
Q♭ = 8πr2 T AB ΩAB Ȳ lm dΩ, (4.19)
the Regge-Wheeler gauge. Their even-parity function
32πr4
Z agrees, up to a normalization factor, with the gauge-
Q♯ = T AB ȲAB
lm
dΩ. (4.20) invariant function first introduced by Moncrief [9]. Mon-
(l − 1)l(l + 1)(l + 2)
crief’s function, in turn, is a variant of Zerilli’s original
In Eqs. (4.17)–(4.20) the stress-energy tensor is imag- even-parity function [3] (Zerilli worked in the frequency
ined to be given in fully contravariant form; T ab , T aB , domain instead of the Schwarzschild time domain). Up to
and T AB are then its components in the spacetime co- the same normalization factor the Moncrief and Fourier-
ordinates (xa , θA ). In the event where the stress-energy transformed Zerilli functions are equal up to the pres-
tensor would be given in covariant or mixed form, its ence of source terms; they are equal only where there is
indices would have to be raised with (g ab , r−2 ΩAB ) — no matter. The normalization adopted in Eq. (4.23) will
the inverse Schwarzschild metric — before evaluating the be seen to be convenient when we discuss gravitational
source terms. In Eqs. (4.13)–(4.16) all lower-case Latin radiation at future null infinity (Sec. VI) and at the hori-
indices are lowered and raised with gab and g ab , respec- zon (Sec. VII); our definition of the even-parity master
tively. function is well adapted to the description of radiation
The perturbation equations are not all independent. fields.
By virtue of the Bianchi identities, or the conservation The perturbation equations (4.13)–(4.16) give rise to
equations (2.9) and (2.10), they are related by a wave equation for the function Ψeven (we resume our
practice of omitting the spherical-harmonic labels). The
2 l(l + 1) a 1 a ♭ manipulations are long and tedious, and we shall not
∇b Qab + rb Qab − Q − r Q =0 (4.21)
r 2r2 r present them here (the reader may consult Ref. [14] for
details). We simply state the final result: As a con-
and sequence of the field equations, the even-parity master
2 (l − 1)(l + 2) ♯ function satisfies the Zerilli equation
∇a Qa + ra Qa − Q + Q♭ = 0. (4.22)
r 2r2
( − Veven )Ψeven = Seven , (4.25)
C. Master equation
with potential
The Zerilli-Moncrief function is defined by
36M 2
    
1 2 µ+2 6M 2M
2r

2 a b lm
 Veven = 2 µ + 3 + µ+
Ψlm := K̃ lm
+ r r h̃ − rr a
∇ K̃ lm


, Λ r2 r r4 r
even ab a
l(l + 1) Λ (4.26)
(4.23) and source term
 6

M2
   
4 1 2 24M r M
Seven = ra Qa − Q♯ + −2r2 ra ∇a Q + ra rb Qab + 2rf Q♭ + µ(µ − 2) + 12(µ − 3) + 84 2 Q ,
Λ r (µ + 2)Λ Λ Λ r r
(4.27)

where µ := (l − 1)(l + 2) and Q := gab Qab . The validity where we have discarded the spherical-harmonic labels
of Eqs. (4.25)–(4.27) can be verified by brute-force eval- for brevity (we shall continue with this practice until the
uation of both sides of Eqs. (4.25). The general source end of the section). It is easy to show that the combina-
term for the Zerilli equation has never been presented in tions
the literature; we display it here for the first time, in a 1 1
form covariant under an arbitrary transformation of the h̃a = ha − ∇a h2 + ra h2 (5.7)
coordinates xa used in the two-dimensional submanifold 2 r
M2 of the Schwarzschild spacetime. are gauge invariant. Equation (5.6) reveals that one can
always choose a gauge in which h2 = 0; this is the Regge-
Wheeler gauge. Equation (5.7) implies that h̃a = ha in
V. ODD-PARITY SECTOR the Regge-Wheeler gauge.

A. Perturbation fields and gauge transformations

B. Perturbation equations
The odd-parity sector refers to those components of the
metric perturbation that can be expanded in terms of the The Ricci tensor of the Schwarzschild spacetime van-
lm lm ishes, and as a consequence its perturbation is gauge in-
odd-parity spherical harmonics XA and XAB . Recalling
the notation gab = gab + pab , gaB = paB , and 4 gAB =
4 4 variant. Its computation can therefore be carried out in
r2 ΩAB + pAB for the perturbed metric, the odd-parity any convenient gauge, and as in the preceding section
sector of the metric perturbation is we shall adopt the Regge-Wheeler gauge. We substitute
Eqs. (5.1)–(5.3), having set h2 = 0, into the Ricci tensor
pab = 0, (5.1) of Appendix B. We simplify the result and find that δRab
lm
vanishes, δRaB is expanded in terms of XB , and δRAB
X
paB = hlm lm
a XB , (5.2) lm
lm
in terms of XAB . From the Ricci tensor we compute the
X Einstein tensor and set the result equal to 8πT αβ . Each
pAB = hlm lm
2 XAB . (5.3) spherical-harmonic component of the field equations can
lm then be extracted by involving the orthonormality rela-
In this section the sums over l are taken to begin at l = 2; tions (3.4) and (3.9) satisfied by the odd-parity harmon-
there is no odd-parity perturbation with l = 0, and the ics. At the end of this calculation we take advantage of
case l = 1 will be considered separately in Sec. VIII. the fact that h̃a = ha in the Regge-Wheeler gauge. This
The fields hlm and hlm are defined on M2 and they allows us to make the substitution ha → h̃a and therefore
a 2
depend on the coordinates xa only. They are closely to express the gauge-invariant Einstein tensor in terms of
related to the quantities first introduced by Regge and gauge-invariant quantities.
Wheeler [1], who worked exclusively in terms of the usual Our final results are
Schwarzschild coordinates (t, r). In these coordinates 2
Pa = −h̃a + ∇a ∇b h̃b + rb ∇a h̃b − ra ∇b h̃b


(discarding spherical-harmonic labels) we have ht = h0 r
and hr = h1 . Except for a sign — refer back to the dis- 2 l(l + 1)
b
cussion following Eq. (3.11) — the function h2 is identical − 2 ra rb h̃ + h̃a , (5.8)
r r2
to the corresponding Regge-Wheeler quantity.
Odd-parity gauge transformations are generated by a P = ∇a h̃a , (5.9)
dual vector field Ξα = (0, ΞA ) that is expanded as where  := g ab ∇a ∇b . The source terms are
X
ΞA = ξ lm XAlm
, (5.4) a 16πr2
Z
lm P = T aB X̄B
lm
dΩ, (5.10)
l(l + 1)
in which ξ lm depends on the coordinates xa only. Under 16πr4
Z
P = T AB X̄AB
lm
dΩ. (5.11)
such a transformation the perturbation quantities change (l − 1)l(l + 1)(l + 2)
according to (see Appendix B)
In Eqs. (5.10) and (5.11) the stress-energy tensor is imag-
2 ined to be given in fully contravariant form; T aB and
ha → h′a= ha − ∇a ξ + ra ξ, (5.5)
r T AB are its relevant components in the spacetime coor-
h2 → h′2 = h2 − 2ξ, (5.6) dinates (xa , θA ). In the event where the stress-energy
 7

tensor would be given in covariant or mixed form, its The validity of Eqs. (5.14)–(5.16) can be verified by
indices would have to be raised with (g ab , r−2 ΩAB ) — brute-force evaluation of both sides of Eqs. (5.14). The
the inverse Schwarzschild metric — before evaluating the general source term for the Regge-Wheeler equation has
source terms. In Eqs. (5.8) and (5.9) all lower-case Latin never been presented in covariant form in the literature.
indices are lowered and raised with gab and g ab , respec- We display it here for the first time, and note that in the
tively. usual Schwarzschild coordinates (and in the frequency
The perturbation equations are not all independent. domain), Sodd agrees with the source term presented in
By virtue of the Bianchi identities, or the conservation Eq. (18) of Jhingan and Tanaka [13].
equation (2.10), they are related by
2 (l − 1)(l + 2) D. Regge-Wheeler function
∇a P a + ra P a − P = 0. (5.12)
r r2
The Cunningham-Price-Moncrief function is a close
C. Master equation cousin to the more familiar Regge-Wheeler function [1],
whose covariant and gauge-invariant definition is
The Cunningham-Price-Moncrief function is defined 1 a lm
Ψlm
RW := r h̃a . (5.17)
by r
2r  2  As we shall see in Secs. VI and VII, the Regge-Wheeler
Ψlm
odd := ε ab
∇a h̃ lm
b − ra h̃ lm
b , (5.13) function is not well suited to describe the gravitational
(l − 1)(l + 2) r
radiation field, and in this paper we adopt the function
where εab is the Levi-Civita tensor on the submanifold Ψlm
odd of Eq. (5.13) as the fundamental odd-parity master
M2 . Apart from a different normalization factor, this function. It is straightforward to use the perturbation
is a covariant generalization of the definition provided equations to show that these functions are related by
by Cunningham, Price, and Moncrief [12], who worked 1 a r
with the usual Schwarzschild coordinates. Our definition ΨRW = t ∇a Ψodd + ra P a , (5.18)
(and normalization) agrees with the odd-parity function 2 (l − 1)(l + 2)
considered by Jhingan and Tanaka [13], who also worked where ta = −εab rb is the Killing vector of Eq. (2.7).
with the Schwarzschild coordinates, but in the frequency Outside of sources, and apart from a factor of one-half,
domain. The normalization adopted in Eq. (5.13) will the Regge-Wheeler function is the time derivative of the
be seen to be convenient when we discuss gravitational Cunningham-Price-Moncrief function.
radiation at future null infinity (Sec. VI) and at the hori- The function of Eq. (5.17) also satisfies the Regge-
zon (Sec. VII); our definition of the odd-parity master Wheeler equation,
function is well adapted to the description of radiation
fields. It is noteworthy that it can also be expressed as ( − Vodd )ΨRW = SRW , (5.19)
2r  2  with the potential of Eq. (5.15) and a new source term
Ψlm
odd := εab ∂a hlm
b − ra hb
lm
, given by
(l − 1)(l + 2) r
   
in terms of the original (gauge-dependent) perturbation 1 a
2 3M
SRW = r ∇a P − Pa − 1− P . (5.20)
quantities, and in terms of the partial differentiation op- r r r
erator; the function is nonetheless gauge-invariant and a This covariant form for the source term is also a new
scalar. result.
The perturbation equations (5.8) and (5.9) give rise to
a wave equation for the function Ψodd (we resume our
practice of discarding the spherical-harmonic labels). As VI. RADIATION AT FUTURE NULL INFINITY
in the preceding section we simply state the final result:
As a consequence of the field equations, the odd-parity
To examine the gravitational perturbations near future
master function satisfies the Regge-Wheeler equation
null infinity we adopt the retarded coordinates (u, r, θ, φ)
( − Vodd )Ψodd = Sodd , (5.14) and express the two-dimensional Schwarzschild metric in
the form of Eq. (2.3). In these coordinates, future null
with potential infinity corresponds to taking the limit r → ∞ keeping u
fixed, and our strategy will be to expand the metric per-
l(l + 1) 6M turbations in powers of r−1 . In asymptotically Cartesian
Vodd = − 3 (5.15)
r2 r coordinates the radiative part of the metric would scale
and source term as r−1 ; transforming to spherical coordinates produces
the scalings
2r
Sodd = − εab ∇a Pb . (5.16)
(l − 1)(l + 2) prad
ab = O(r
−1
), prad 0
aB = O(r ), prad
AB = O(r) (6.1)
 8

for the radiative part of the metric perturbations. Our We turn next to the odd-parity sector of Sec. V. The
goal is to isolate this, and to calculate how much energy gauge conditions imply hu = 0, and the scalings of
and angular momentum is carried away by the radiation. Eq. (6.1) imply that we are looking for the r0 part of
We find it advantageous to work in gauge in which hr and the r1 part of h2 . These are determined by fol-
lowing the same procedure as in the even-parity case, and
ta pab = 0 = ta paB , (6.2) we obtain
where ta is the Killing vector of Eq. (2.7). In spite of (l − 1)(l + 2) u
Z
the fact that ta is not a null vector (except on the even hr = − b(u′ ) du′ + O(r−2 ), (6.9)
2r
horizon, which is well outside our domain of considera-
h2 = b(u)r + O(r0 ), (6.10)
tion), this gauge happens to be a perfectly respectable
“radiation gauge.” where b(u) is not determined by the vacuum field equa-
In this section we take l ≥ 2; as is well known, the tions. We see that the radiative part of the perturbation
low multipoles l = 0 and l = 1 do not contain radiative is contained entirely in the function h2 (u, r). With the
degrees of freedom. We assume that the matter distribu- metric perturbation of Eqs. (6.9), (6.10) we may evaluate
tion responsible for the radiation is confined to a bounded the Cunningham-Price-Moncrief function of Eq. (5.13).
volume, and that our domain of consideration is outside The result is
this volume; we shall therefore be solving the vacuum
field equations. Ψodd = b(u) + O(r−1 ), (6.11)
We begin with the even-parity sector of Sec. IV. The
and we conclude that the radiative part of the odd-parity
gauge conditions imply huu = hur = ju = 0, and the
sector is given by
scalings of Eq. (6.1) imply that we are looking for the r−1
part of hrr , K, and G, as well as the r0 part of jr . These rad, odd
pAB =r
X
Ψlm lm
odd (u, r = ∞)XAB . (6.12)
can be determined by expanding the components of the
lm
metric perturbation in powers of r−1 and substituting
them into the field equations of Eqs. (4.13)–(4.16). The It is obtained by integrating the Regge-Wheeler equation
expansions for hrr , K, and G begin at order r−1 and (5.14) and evaluating the Cunningham-Price-Moncrief
each coefficient is a to-be-determined function of u; the function at r = ∞. Notice that by virtue of Eq. (5.18),
expansion for jr begins instead at order r0 . The field the radiative field could instead be expressed in terms of
equations return Qab , Qa , Q♭ , and Q♯ expanded in powers the u-integral of the Regge-Wheeler function. The need
of r−1 , and setting each coefficient to zero determines the to perform this integration is inconvenient, and it is the
metric perturbation. We obtain simple relationship of Eq. (6.12) that has motivated the
adoption of the Cunningham-Price-Moncrief function as
l(l + 1) u
Z
hrr = − a(u′ ) du′ + O(r−4 ), (6.3) the fundamental odd-parity master function.
r3 The full radiative field is obtained from Eqs. (6.8) and
a(u) (6.12). We have prad rad
ab = 0 = paB and
jr = + O(r−2 ), (6.4)
r X  
prad lm lm lm lm
Z u
l(l + 1) AB = r Ψ Y
even AB + Ψ odd AB ,
X (6.13)
K = a(u′ ) du′ + O(r−4 ), (6.5) lm
2r3
2 ȧ(u) where Ψlm lm lm lm
even ≡ Ψeven (u, r = ∞) and Ψodd ≡ Ψodd (u, r =
G = − + O(r−2 ), (6.6)
(l − 1)(l + 2) r ∞). As expected, the radiative field is transverse, and
tracefree by virtue of Eqs. (3.11). The two fundamen-
where a(u) is a function that is not determined by the tal polarizations of the gravitational wave can be defined
vacuum field equations, and ȧ := da/du. We see that the by h+ := pθθ /r2 and h× := pθφ /(r2 sin θ). Using the
radiative part of the perturbation is contained entirely components of the tensorial spherical harmonics listed in
in the function G(u, r). With the metric perturbation Appendix A, we obtain
of Eqs. (6.3)–(6.6) we may evaluate the Zerilli-Moncrief   2 
function of Eq. (4.23). The result is 1X ∂ 1
h+ = Ψlm
even + l(l + 1) Y lm
2 r ∂θ2 2
lm
Ψeven = − ȧ(u) + O(r−1 ), (6.7)   
(l − 1)(l + 2) im ∂ cos θ lm
− Ψlmodd − Y (6.14)
sin θ ∂θ sin θ
and we conclude that the radiative part of the even-parity
sector is given by and
  
rad, even 1X im ∂ cos θ lm
X
pAB =r Ψlm lm
even (u, r = ∞)YAB . (6.8) h× = lm
Ψeven − Y
lm r sin θ ∂θ sin θ
lm
 2  
It is obtained by integrating the Zerilli equation (4.25) lm ∂ 1 lm
+ Ψodd + l(l + 1) Y . (6.15)
and evaluating the Zerilli-Moncrief function at r = ∞. ∂θ2 2
 9

The energy and angular momentum carried away by We assume that the matter distribution responsible for
the gravitational radiation can be calculated using the the radiation does not come near the event horizon; we
techniques developed by Thorne [17]. We compare our shall therefore be solving the vacuum field equations in
Eq. (6.13) to his Eq. (4.3), taking into account the rela- an empty neighborhood of the event horizon.
tionship between our tensorial harmonics and his pure- We begin with the even-parity sector of Sec. IV. The
spin harmonics (this is spelled out in Appendix A). gauge conditions of Eq. (7.1) imply that hvv = hvr =
Thorne’s mass multipole moments are thus seen to be jv = 0, so that hrr , jr , K, and G are the only non-
proportional to Ψlmeven (u, r = ∞), and his current mo- vanishing components of the metric perturbation. As
ments are proportional to Ψlm odd (u, r = ∞). Substituting stated above, the radiation crossing the event horizon is
these into Thorne’s Eq. (4.16) we obtain described entirely by K and G evaluated at r = 2M . The
  field equations, however, imply that K vanishes on the
dE 1 X horizon. This is verified by expanding hrr , jr , K, and G
= (l − 1)l(l + 1)(l + 2)
du 64π in powers of f = 1 − 2M/r and substituting them into
lm
D
2
E the field equations of Eqs. (4.13)–(4.16). This calcula-
lm 2
× Ψ̇lm even + Ψ̇odd (6.16) tion reveals also that G(v, r = 2M ) = a(v), a function
that is not determined by the vacuum field equations.
for the averaged rate at which the energy escapes to On the other hand, the expansions allow us to evaluate
future null infinity. Substituting instead into Thorne’s the Zerilli-Moncrief function of Eq. (4.23), and the result
Eq. (4.23) returns is Ψeven(r = 2M ) = 2M a(v). We conclude that on the
  horizon, the radiative part of the even-parity sector is
dJ 1 X given by
= (l − 1)l(l + 1)(l + 2)(im)
du 64π
lm
rad, even
X
D E pAB = 2M Ψlm lm
even (v, r = 2M )YAB . (7.2)
× Ψ̄lm Ψ̇ lm
even even + Ψ̄ lm
Ψ̇ lm
odd odd (6.17) lm

for the averaged rate at which the angular momentum It is obtained by integrating the Zerilli equation (4.25)
escapes to infinity. This is the component of the angular- and evaluating the Zerilli-Moncrief function at r = 2M .
momentum vector in the arbitrary z direction which de- We turn next to the odd-parity sector of Sec. V. The
fines the orientation of the angles θ and φ. The overbar gauge conditions imply hu = 0, so that hr and h2 are
indicates complex conjugation, and it is not difficult to the only nonvanishing components of the metric pertur-
show that hdJ/dui is real. The averaging carried out in bation. The radiation crossing the event horizon is de-
Eqs. (6.16) and (6.17) is over a characteristic time scale scribed entirely by h2 evaluated at r = 2M , and the
associated with the gravitational wave. field equations imply that b(v) := h2 (v, r = 2M ) re-
mains as an undetermined function. They also imply
that hr (v, r = 2M ) = c(v), with
VII. RADIATION AT THE EVENT HORIZON
(l − 1)(l + 2) v
Z
c(v) := b(v ′ ) dv ′ .
To examine the gravitational perturbations near 8M 2
the event horizon we adopt the advanced coordinates
(v, r, θ, φ) and express the two-dimensional Schwarzschild These statements are verified by expanding hr and h2
metric in the form of Eq. (2.4). We want to calculate how in powers of f = 1 − 2M/r and substituting them
much energy and angular momentum is transfered to the into Eqs. (5.8) and (5.9). The expansions allow us also
black hole by the perturbation, and we shall do so by to evaluate the Cunningham-Price-Moncrief function of
following the methods devised by Poisson [19]. As in the Eq. (5.13), and the result is
preceding section we impose the gauge conditions
4M dc 1
Ψodd (r = 2M ) = − =− b(v).
ta pab = 0 = ta paB (7.1) (l − 1)(l + 2) dv 2M
on the metric perturbation; ta is still the Killing vector We conclude that on the horizon, the radiative part of
of Eq. (2.7). Recall that this vector is null on the event the odd-parity sector is given by
horizon, and Eq. (7.1), evaluated at r = 2M , is equiva-
lent to Poisson’s Eq. (6.5). Poisson then shows that the rad, odd
X
pAB = −2M Ψlm lm
odd (v, r = 2M )XAB . (7.3)
part of the metric perturbation which is associated with lm
the transport of energy and angular momentum across
the horizon is pAB . [Refer to Poisson’s Eq. (6.10), which It is obtained by integrating the Regge-Wheeler equation
1
establishes the equality between pAB and γAB , the per- (5.14) and evaluating the Cunningham-Price-Moncrief
turbation of the horizon’s intrinsic metric.] function at r = 2M . Notice that by virtue of Eq. (5.18),
In this section we take l ≥ 2, as the low multipoles l = 0 the radiative field could instead be expressed in terms of
and l = 1 do not contain radiative degrees of freedom. the v-integral of the Regge-Wheeler function; this would
 10

give rise to Poisson’s Eq. (7.3), a less convenient expres- the fields ha and G are not defined. The freedom to
sion. perform a gauge transformation is contained in Ξa =
The full radiative field is obtained by adding Eqs. (7.2) ξa Y 00 , ΞA = 0, and the perturbations transform as
and (7.3). We have
X  hab → h′ab = hab − ∇a ξb − ∇b ξa , (8.2)
prad
AB = 2M Ψlm lm lm lm
even YAB − Ψodd XAB , (7.4) 2
K → K ′ = K − ra ξa . (8.3)
lm r
where Ψlm lm lm lm
even ≡ Ψeven (v, r = 2M ) and Ψodd ≡ Ψodd (v, r = There is no analogue here of the gauge-invariant quanti-
2M ). This result should be compared with Eq. (6.13). ties h̃ab and K̃ that were introduced in the general case.
The rates at which the gravitational perturbation trans- The relevant field equations for hab and K are the Qab
fers energy and angular momentum to the black hole can and Q♭ equations of Eqs. (4.13) and (4.15), respectively,
now be calculated using the method described in Sec. VII in which we set l = 0, h̃ab = hab , and K̃ = K. The Qa
of Poisson [19]. Our Eq. (7.4) replaces his Eq. (7.5), and and Q♯ equations of Eqs. (4.14) and (4.16), respectively,
the rest of the calculation is identical. The final results are not defined.
are We now solve the perturbation equations in the ab-

dE

1 X sence of material sources; in this case Qab = 0 = Q♭ .
= (l − 1)l(l + 1)(l + 2) We adopt the advanced coordinates (v, r, θ, φ) and the
dv 64π
lm metric of Eq. (2.4). For convenience we rescale the value
of Y 00 and set it equal to unity. We choose ξa so as to
D E
2 lm 2
× Ψ̇lm even + Ψ̇ odd (7.5)
enforce the gauge conditions hrr = K = 0. This does not
fully determine the gauge, because these conditions are
and
preserved under an additional transformation generated
by ξv = −f α(v) and ξr = α(v), where f = 1 − 2M/r
 
dJ 1 X
= (l − 1)l(l + 1)(l + 2)(im) and α(v) is an arbitrary function. This transformation
dv 64π
lm corresponds to a redefinition of the advanced-time coor-
D E
× Ψ̄lm lm lm lm
even Ψ̇even + Ψ̄odd Ψ̇odd . (7.6) dinate v, of the form v → v ′ = v + ξ v = v + α(v), and
it changes the nonvanishing perturbations according to
These equations replace Poisson’s Eqs. (7.8) and (7.9). hvv → h′vv = hvv + 2f α̇(v) and hvr → h′vr = hvr − α̇(v),
Notice the similarity between Eqs. (7.5) and (6.16), and where an overdot indicates differentiation with respect to
between Eqs. (7.6) and (6.17). In Eq. (7.6), J represents v.
the component of the hole’s angular-momentum vector We now involve the field equations. We notice first
in the arbitrary z direction which defines the orientation that the Qrr = 0 equation implies that hvr is in fact in-
of the angles θ and φ. The overbar indicates complex dependent of r; since this function of v can be altered
conjugation. The averaging carried out in Eqs. (7.5) and at will by a gauge transformation, we may set hvr = 0.
(7.6) is over a characteristic time scale associated with The gauge is now fully determined, and the sole remain-
the gravitational perturbation. ing perturbation field is hvv . By virtue of the Qvr = 0
equation we have that rhvv must be a function of v only,
but the Qvv = 0 equation constrains this function to be a
VIII. LOW MULTIPOLES constant. We therefore have hvv = (constant)/r, and this
completes the integration of the perturbation equations.
To conclude our presentation of the perturbation for- The vacuum, monopole perturbation of a
malism we now handle the special cases l = 0 and l = 1. Schwarzschild black hole is given by
Our discussion here will essentially reproduce Zerilli’s
2δM
Appendix G [3], but we will frame it in the language pvv = , (8.4)
developed in this paper. For concreteness, and for sim- r
plicity, we will focus mostly on the task of integrating the where δM is a constant; all other components vanish.
vacuum field equations for these low multipoles. This addition to gvv simply shifts the Schwarzschild mass
parameter from M to M + δM . The perturbation must
respect Birkhoff’s theorem, and the perturbed metric is
A. Monopole perturbation
indeed another Schwarzschild solution.
When l = 0 the only relevant spherical harmonic is
Y 00 , which is a constant. It follows that YA = XA =
B. Odd-parity dipole perturbation
YAB = XAB = 0, and the perturbation is of even parity.
The only nonvanishing metric perturbations are
The only surviving odd-parity spherical harmonics for
pab = hab Y 00 , pAB = r2 KΩAB Y 00 ; (8.1) l = 1 are XA 1m
, which are obtained from Y 1m using
 11

1m
Eq. (3.2). The tensorial harmonics XAB vanish, and the the unit radial vector Ωi as in Eq. (8.7), and writing the
only surviving components of the metric perturbation are summation over m as an implicit summation over i, we
X obtain our final expression,
paB = h1m 1m
a XB . (8.5)
m 2
pvA = − εAB (δJ)i ΩiB . (8.8)
r
The perturbations h1m
acan be alteredP
by a gauge trans-
formation generated by Ξa = 0, ΞA = m ξ 1m XA1m
; they To give an interpretation to this result we perform a
change according to transformation from the spherical coordinates (r, θA ) to
the Cartesian coordinates xi = rΩi . The metric per-
2 turbation transforms as pvi = r−1 pvA ΩA A
ha → h′a = ha − ∇a ξ + ra ξ. (8.6) i , where Ωi :=
r j
ΩAB δij ΩB . This gives pvi = −2r−2 εAB ΩA B j
i Ωj (δJ) . We
A B
There is no analogue here of the gauge-invariant fields h̃a now involve the easily-established identity εAB Ωi Ωj =
introduced in Sec. V for the general case. The relevant εijk Ωk , where εijk is the permutation symbol (a fully an-
field equation for ha is the Pa equation of Eq. (5.8), in tisymmetric tensor with entries −1, 0, or 1). This gives
which we set l = 1 and h̃a = ha . The P equation of
Eq. (5.9) is not defined. 2 2
pvi = − 2
εijk (δJ)j Ωk = − 3 (δJ × r)i . (8.9)
We now solve the perturbation equations in the ab- r r
sence of material sources; in this case Pa = 0. We once
This form for the metric tensor allows us to identify
more adopt the advanced coordinates (v, r, θ, φ) and the
δJ with the angular-momentum vector of the spacetime.
metric of Eq. (2.4). For convenience we replace the set
The perturbed metric therefore describes a slowly rotat-
of complex harmonics Y 1m with the set Y m given by
ing black hole; it is the Kerr metric linearized with re-
(Y 0 , Y s , Y c ) = (cos θ, sin θ sin φ, sin θ cos φ) spect to its angular-momentum parameter.
=: (Ω3 , Ω2 , Ω1 ). (8.7)

This set is real, but not normalized; summation over m C. Even-parity dipole perturbation
will be replaced by a summation over the abstract index
m = (0, s, c). Notice that Ωi represents the Cartesian The even-parity spherical harmonics for l = 1 are Y 1m ,
components of a unit radial vector; in ordinary vectorial YA1m ,and ΩAB Y 1m ; YAB
1m
vanishes. The metric pertur-
notation, Ω = r/r, where r = (x, y, z) and r = |r|. We bation is then
shall also need ΩiA := ∂Ωi /∂θA , the derivatives of Ωi with X
respect to the angles θA . pab = h1m
ab Y
1m
, (8.10)
We choose ξ so as to enforce the gauge condition m
hr = 0. This does not fully determine the gauge, because
X
paB = ja1m YB1m , (8.11)
this condition is preserved under an additional transfor- m
mation generated by ξ = α(v)r2 , where α(v) is an arbi- X
trary function. This transformation changes hv accord- pAB = r2 ΩAB K 1m Y 1m , (8.12)
m
ing to hv → h′v = hv − α̇(v)r2 , where an overdot indicates
differentiation with respect to v.
and the fields G1m are not defined. Gauge transforma-
We now involve the field equations. We note first that P 1m 1m
tions are generated by Ξa = m ξa Y and ΞA =
the Pr = 0 equation implies that hv must have the form P 1m 1m
ξ Y . The perturbations change according to
of hv = k1 (v)r2 + k2 (v)/r, where k1 and k2 are arbitrary m A

functions. The first term can be removed by a gauge hab → h′ab = hab − ∇a ξb − ∇b ξa , (8.13)
transformation, and we are left with hv = k2 (v)/r and a
2
fully determined gauge. The Pv = 0 equation then im- ja → ja′ = ja − ξa − ∇a ξ + ra ξ, (8.14)
plies that k2 must in fact be a constant. We therefore r
have hv = (constant)/r, and this completes the integra- 2 2
K → K ′ = K + 2 ξ − ra ξa . (8.15)
tion of the perturbation equations. r r
The vacuum, odd-parity, dipole perturbation of a
Schwarzschild black hole is given by There is no analogue here of the gauge-invariant fields h̃ab
and K̃ introduced in Sec. IV for the general case. The
2X relevant field equations for hab , ja , and K are the Qab ,
pvA = (δJ)m XA
m
, Qa , Q♭ equations of Eqs. (4.13)–(4.15), in which we set
r m
l = 1, h̃ab = hab − ∇a jb − ∇b ja , and K̃ = K − 2r−1 ra ja .
where (δJ)m are the constants identified in the preced- The Q♯ equation of Eq. (4.16) is not defined for l = 1.
ing paragraph, and XA m
= −εAB DB Y m are the real odd- We now solve the vacuum field equations, Qab = Qa =
parity harmonics; all other components of the metric per- Q♭ = 0. We adopt the advanced coordinates (v, r, θ, φ),
turbation vanish. Relating the spherical harmonics to the real spherical harmonics of Eq. (8.7), and a gauge in
 12

which hrr = hr = K = 0. This gauge is preserved under The vacuum, even-parity, dipole perturbation of a
a transformation generated by Schwarzschild black hole can be presented in a gauge in
which the nonvanishing components are
2M
ξv = α1 (v) + α2 (v)r,
r 6M 2M
pvv = −2rα̈i Ωi + α̇i Ωi − 2 αi Ωi (8.16)
ξr = α1 (v), r r
ξ = α1 (v)r + α2 (v)r2 ,
and
where α1 (v) and α2 (v) are arbitrary functions. This has 2M
the following effect on the remaining components of the pvA = −r2 α̈i ΩiA + αi ΩiA , (8.17)
r
metric perturbation:
where the vector α(v) is an arbitrary function of the
2M 2M
hvv → h′vv = hvv − 2α̇2 r − (2α̇1 − α2 ) + 2 α1 , advanced-time coordinate v. This perturbation, however,
r r can be removed by a gauge transformation generated by
hvr → h′vr = hvr − (α̇1 + α2 ),
2M 2M
jv → jv′ = jv − α̇2 r2 − (α̇1 + α2 )r − α1 . Ξv = −rα̇i Ωi + αi Ωi , (8.18)
r r
Ξr = αi Ωi , (8.19)
The Qrr = 0 equation implies that hvr is a function of v
only, and this component can be set equal to zero by a ΞA = −r2 α̇i ΩiA + rαi ΩiA . (8.20)
gauge transformation. The remaining gauge freedom is
now restricted by α2 = −α̇1 . The Qr = 0 equation im- The perturbed metric is therefore a Schwarzschild solu-
plies that jv must be of the form jv = k1 (v)r2 + k2 (v)/r, tion expressed in a coordinate system which differs from
where k1 and k2 are arbitrary functions of v. With this the original system (v, r, θA ). These coordinates are not
information and the Qv = 0 equation we find that hvv inertial. In the limit M → 0 the perturbed metric is
must of the form hvv = −k2 /r2 +3k̇2 /r+k3 +2k1 r, where recognized as the metric of flat spacetime expressed in
k3 is an additional arbitrary function. The Qvr = 0 a light-cone coordinate system centered on an acceler-
equation now reveals that k3 = 0, the Qvv = 0 equa- ated world line (see Ref. [20]); the acceleration vector
tion yields 2M k1 = −k̈2 , and the task of integrat- is α̈i (v). For M 6= 0 the perturbed metric describes a
ing the perturbation equations is completed. Defining Schwarzschild black hole moving on “the same” acceler-
k(v) := k2 (v)/(2M ), we have obtained ated world line.

6M 2M
hvv = −2k̈(v)r + k̇(v) − 2 k(v)
r r Acknowledgments
and
This work was supported by the National Science and
2 2M Engineering Research Council of Canada.
jv = −k̈(v)r + k(v)
r
for the nonvanishing perturbation fields. The remaining
gauge freedom is contained in APPENDIX A: COMPONENTS OF THE
VECTOR AND TENSOR SPHERICAL
2M HARMONICS, AND RELATIONSHIP WITH
ξv = −α̇(v)r + α(v), OTHER SPHERICAL HARMONICS
r
ξr = α(v),
We first list the components of the tensorial harmonics
ξ = −α̇(v)r2 + α(v)r,
introduced in Sec. III. According to Eq. (3.1) we have
where we have set α := α1 and α2 = −α̇. Under this
∂ lm
transformation the perturbations change according to Yθlm = Y ,
∂θ
6M 2M ∂ lm
hvv → h′vv = hvv + 2α̈(v)r − α̇(v) + 2 α(v) Yφlm = Y .
r r ∂φ
and According to Eq. (3.2) we have
2M
jv → jv′ = jv + α̈(v)r2 − α(v). 1 ∂ lm
r Xθlm = − Y ,
sin θ ∂φ
Setting α(v) = k(v) produces h′vv = jv′ = 0, and we ∂
conclude that the perturbation is pure gauge. Xφlm = sin θ Y lm .
∂θ
 13

4
According to Eq. (3.6) we have gaB = paB ,
4
 2
∂ 1
 gAB = r2 ΩAB + pAB ,
lm
Yθθ = + l(l + 1) Y lm ,
∂θ2 2
 2  where (gab , r2 ΩAB ) are the components of the back-
lm ∂ cos θ ∂ ground metric and (pab , paB , pAB ) are the components
Yθφ = − Y lm ,
∂θ∂φ sin θ ∂φ of the perturbation. In this Appendix we allow the back-
ground metric to be completely general, so long as it is
 2 
lm ∂ ∂ 1 2
Yφφ = + sin θ cos θ + l(l + 1) sin θ Y lm . spherically symmetric; we do not restrict it to be the
∂φ2 ∂θ 2
Schwarzschild metric. We raise lower-case Latin indices
And according to Eq. (3.7) we have with g ab , the inverse to gab , and we raise upper-case Latin
 2  indices with ΩAB , the inverse to ΩAB . The inverse per-
lm 1 ∂ cos θ ∂
Xθθ = − − Y lm , turbed metric is thus
sin θ ∂θ∂φ sin θ ∂φ
∂2 1 ∂2 4 ab
= g ab − pab ,
 
lm 1 ∂ g
Xθφ = sin θ 2 − − cos θ Y lm ,
2 ∂θ sin θ ∂φ2 ∂θ 4 aB 1
g = − 2 paB ,
∂2 r
 
lm ∂
Xφφ = sin θ − cos θ Y lm . 1 AB 1
∂θ∂φ ∂φ 4 AB
g = 2Ω − 4 pAB ,
r r
lm lm
The tensorial harmonics YAB
and XAB
can be related
to the spherical-harmonic functions of spin-weight s = up to terms quadratic in the perturbations.
±2 [16]. Let mA and m̄A be a complex orthonormal Covariant differentiation with respect to the coordi-
basis on the unit two-sphere, with mA = 2−1/2 (1, i sin θ). nates xa on the submanifold M2 is indicated with ∇a :
The relationship is then ∇a gbc = 0. Covariant differentiation with respect to the
1p coordinates θA on the unit two-sphere is indicated with
lm
YAB = (l − 1)l(l + 1)(l + 2) DA : DA ΩBC = 0. Quantities which depend only on xa
2
are covariantly constant relative to the connection ΓABC ;
× −2 Y lm mA mB + 2 Y lm m̄A m̄B


for example DA r ≡ 0. Quantities which depend only
and on θA are covariantly constant relative to the connection
ip Γabc ; for example ∇a ΩAB ≡ 0.
lm
XAB = − (l − 1)l(l + 1)(l + 2) A straightforward calculation returns the components
2
of the perturbed connection, which we denote 4 Γαβγ +
× −2 Y lm mA mB − 2 Y lm m̄A m̄B ,


δΓαβγ . We obtain
where s Y lm are the spin-weighted spherical harmonics.
These equations can be compared with Eqs. (2.38e) and δΓabc = C abc ,
(2.38f) of Ref. [17]. This reveals that our tensorial har- 1
1
δΓabC = DC pab + ∇b paC − ∇a pbC − rb paC ,
monics are intimately related to the “pure-spin” harmon- 2 r
ics used by Thorne. The relationship is a 1
DB p C + DC p B − ∇ pBC + rrm ΩBC pam ,
a a a


δΓ BC =
lm
q
E2,lm
2
YAB = 12 (l − 1)l(l + 1)(l + 2) TAB 1
δΓAbc = ∇b pcA + ∇c pbA − DA pbc ,


2r 2
and
A 1 1
DC pbA − DA pbC + ∇b pAC − 3 rb pAC ,


δΓ bC =
q
lm 1 B2,lm 2
XAB = 2 (l − 1)l(l + 1)(l + 2) TAB . 2r r
1 1
E2,lm B2,lm δΓABC = 2 C ABC + rm ΩBC pmA ,
Notice that the pure-spin harmonics TAB and TAB r r
are normalized on the unit two-sphere. Our convention
here differs from Thorne’s, who inserts a factor of r−1 in where ra := ∇a r,
mA and m̄A in order to normalize them on a two-sphere
of radius r. 1
C abc := ∇c pab + ∇b pac − ∇a pbc ,


2

APPENDIX B: PERTURBATION OF THE RICCI and

TENSOR FOR A GENERAL
SPHERICALLY-SYMMETRIC BACKGROUND 1
C ABC := DC pAB + DB pAC − DA pBC .


SPACETIME 2

The Ricci tensor for the perturbed spacetime is equal

The perturbed spacetime metric is written as
to the background Ricci tensor Rαβ plus its perturbation
4
gab = gab + pab , δRαβ . We obtain
 14

2 1 1 1
δRab = ∇m C mab + rm C mab − ∇a ∇b pmm − 2 DM DM pab + 2 DM ∇a pbM + ∇b paM


r 2 2r 2r
1 1 1
− 2 ∇a ∇b p M + 3 ra ∇b p M + rb ∇a p M − 4 (ra rb − r∇a ∇b r)pMM ,
M M M


2r 2r r
1  1  1
1
m m
δRaB = DB ∇m p a − ∇a p m + ra p m − paB − ∇m ∇a pmB − ra ∇m pmB − rm ∇a pmB
m


2 r 2 r
1 1 M 1
− 2 ra rm + r∇a ∇m r p B + 2 D DB paM − DM paB + 2 ∇a DM pMB − DB pMM

m


r 2r 2r
1 M M


− 3 ra DM p B − DB p M ,
r 
 1  1 1
δRAB = ΩAB rrm ∇n pmn − g mn pkk + (rm rn + r∇m ∇n r)pmn − DA DB pmm + ∇m DA pmB + DB pmA


2 2 2
1 1 1 1 1  1 
+ rm ΩAB DM pmM − pAB + 2 DM C MAB − 2 DA DB pMM + rm ∇m pAB − ΩAB pMM
r 2 r 2r r 2
2 m  1 M

− 2 r rm pAB − ΩAB p M ,
r 2
where  := g ab ∇a ∇b . These expressions can be simplified by involving Eqs. (2.6) and (2.8) when the background
spacetime is the Schwarzschild spacetime.
Under a gauge transformation generated by the dual vector field Ξα = (Ξa , ΞA ), the components of the metric
perturbation change according to

pab → p′ab := pab − ∇a Ξb − ∇b Ξa ,

2
paB → p′aB := paB − ∇a ΞB − DB Ξa + ra ΞB ,
r
pAB → p′AB := pAB − DA ΞB − DB ΞA − 2rrm Ξm ΩAB .

It can be shown that when the background Ricci tensor vanishes, δRab , δRaB , and δRAB are all invariant under this
transformation. We use this property in Secs. IV and V.

APPENDIX C: PERTURBATION EQUATIONS IN (t, r) COORDINATES

In this Appendix the two-dimensional Schwarzschild metric is written in its standard form

ds2 = −f dt2 + f −1 dr2 ,

with f = 1 − 2M/r. In these coordinates the wave operator is given by

1 ∂2
 
∂ ∂
Ψ = − + f Ψ.
f ∂t2 ∂r ∂r

We set λ := l(l + 1) = µ + 2 and µ := (l − 1)(l + 2) = λ − 2.

The even-parity perturbations are htt , htr , hrr , jt , jr , K, and G. Under a gauge transformation generated by ξt ,
ξr , and ξ they change according to htt → h′tt = htt + ∆htt , etc., with

∂ 2M f
∆htt = −2 ξt + 2 ξr ,
∂t r
∂ ∂ 2M
∆htr = − ξt − ξr + 2 ξt ,
∂r ∂t r f
∂ 2M
∆hrr = −2 ξr − 2 ξr ,
∂r r f
∂
∆jt = − ξ − ξt ,
∂t
∂ 2
∆jr = − ξ − ξr + ξ,
∂r r
 15

2f λ
∆K = − ξr + 2 ξ,
r r
2
∆G = − 2 ξ.
r
The gauge-invariant fields are

∂ 2M f ∂2 ∂
h̃tt = htt − 2 jt + 2 jr + r2 2 G − M f G,
∂t r ∂t ∂r
∂ ∂ 2M ∂2 r − 3M ∂
h̃tr = htr − jt − jr + 2 jt + r2 G+ G,
∂r ∂t r f ∂t∂r f ∂t
∂ 2M ∂2 2r − 3M ∂
h̃rr = hrr − 2 jr − 2 jr + r2 2 G + G,
∂r r f ∂r f ∂r
2f ∂ λ
K̃ = K − jr + rf G + G.
r ∂r 2
The field equations are

∂2 3r − 5M ∂ f ∂ (λ + 2)r + 4M µ
Qtt = − K̃ − K̃ + h̃rr + h̃rr + 2 K̃,
∂r2 r2 f ∂r r ∂r 2r3 2r f
∂2 r − 3M ∂ f ∂ λ
Qtr = K̃ + 2
K̃ − h̃rr − 2 h̃tr ,
∂t∂r r f ∂t r ∂t 2r
∂2 (r − M )f ∂ 2f ∂ f ∂ λr + 4M f2 µf
Qrr = − 2 K̃ + 2
K̃ + h̃ tr − h̃ tt + 3
h̃ tt − 2
h̃rr − 2 K̃,
∂t r ∂r r ∂t r ∂r 2r r 2r
∂ ∂ 1 ∂ 2M
Qt = h̃rr − h̃tr + K̃ − 2 h̃tr ,
∂t ∂r f ∂t r f
∂ ∂ ∂ r−M (r − M )f
Qr = − h̃tr + h̃tt − f K̃ − 2 h̃tt + h̃rr ,
∂t ∂r ∂r r f r2
∂2 ∂2 ∂2 1 ∂2 ∂2 2(r − M ) ∂ r − 3M ∂
Q♭ = − 2 h̃rr + 2 h̃tr − 2 h̃tt − 2
K̃ + f 2
K̃ + 2
h̃tr − h̃tt
∂t ∂t∂r ∂r f ∂t ∂r r f ∂t r2 f ∂r
(r − M )f ∂ 2(r − M ) ∂ λr2 − 2(2 + λ)M r + 4M 2 λr2 − 2µM r − 4M 2
− h̃ rr + K̃ + h̃ tt − h̃rr ,
r2 ∂r r2 ∂r 2r4 f 2 2r4
1
Q♯ = h̃tt − f h̃rr .
f

The conservation (Bianchi) identities read

∂ tt ∂ tr 2(r − M ) tr λ
0 = Q + Q + Q − 2 Qt ,
∂t ∂r r2 f 2r
∂ tr ∂ rr M f tt 2r − 5M rr λ f
0 = Q + Q + 2 Q + 2
Q − 2 Qr − Q♭ ,
∂t ∂r r r f 2r r
∂ t ∂ r 2 r µ
0 = Q + Q + Q + Q♭ − 2 Q♯ .
∂t ∂r r 2r
The Zerilli-Moncrief function is
  
2r 2f ∂
Ψeven = K̃ + f h̃rr − r K̃ ,
µ+2 Λ ∂r

where Λ = µ + 6M/r, and its source term is

  
4 r 1 ♯ 2 ∂ 1 24M rr
Seven = Q − Q + 2r2 f f Qtt − Qrr + Q + 2rf Q♭
Λ r (µ + 2)Λ ∂r f Λ
M2
  
r M tt 1 rr
− µ(µ − 2) + 12(µ − 3) + 84 2 fQ − Q .
Λ r r f
 16

The odd-parity perturbations are ht , hr , and h2 . Under a gauge transformation generated by ξ they change
according to ht → h′t = ht + ∆ht , etc., with

∂
∆ht = − ξ,
∂t
∂ 2
∆hr = − ξ + ξ,
∂r r
∆h2 = −2ξ.

The gauge-invariant fields are

1 ∂
h̃t = ht − h2 ,
2 ∂t
1 ∂ 1
h̃r = hr − h2 + h2 .
2 ∂r r
The field equations are

∂2 ∂2 2 ∂ λr − 4M
Pt = − h̃r + 2 h̃t − h̃r − h̃t ,
∂t∂r ∂r r ∂t r3 f
∂2 ∂2 2 ∂ µf
Pr = 2
h̃r − h̃t + h̃t + 2 h̃r ,
∂t ∂t∂r r ∂t r
1 ∂ ∂ 2M
P = − h̃t + f h̃r + 2 h̃r .
f ∂t ∂r r

The conservation (Bianchi) identity reads

∂ t ∂ r 2 r µ
0= P + P + P − 2 P.
∂t ∂r r r
The Cunningham-Price-Moncrief function is
 
2r ∂ ∂ 2
Ψodd = h̃t − h̃r − h̃t ,
µ ∂r ∂t r

and its source term is

 
2r 1 ∂ r ∂ 2M
Sodd = P + f Pt + 2 Pt .
µ f ∂t ∂r r

APPENDIX D: PERTURBATION EQUATIONS IN (u, r) COORDINATES

In this Appendix the two-dimensional Schwarzschild metric is written in the form

ds2 = −f du2 − 2 dudr,

with f = 1 − 2M/r. In these coordinates the wave operator is given by

∂2
 
∂ ∂
Ψ = −2 + f Ψ.
∂u∂r ∂r ∂r

We set λ := l(l + 1) = µ + 2 and µ := (l − 1)(l + 2) = λ − 2.

The even-parity perturbations are huu , hur , hrr , ju , jr , K, and G. Under a gauge transformation generated by ξu ,
ξr , and ξ they change according to huu → h′uu = huu + ∆huu , etc., with

∂ 2M 2M f
∆huu = −2 ξu − 2 ξu + 2 ξr ,
∂u r r
∂ ∂ 2M
∆hur = − ξu − ξr + 2 ξr ,
∂r ∂u r
 17

∂
∆hrr = −2 ξr ,
∂r
∂
∆ju = − ξ − ξu ,
∂u
∂ 2
∆jr = − ξ − ξr + ξ,
∂r r
2f 2 λ
∆K = − ξr + ξu + 2 ξ,
r r r
2
∆G = − 2 ξ.
r
The gauge-invariant fields are
∂ 2M 2M f ∂2 ∂ ∂
h̃uu = huu − 2 ju − 2 ju + 2 jr + r2 2 G + M G − M f G,
∂u r r ∂u ∂u ∂r
2
∂ ∂ 2M 2 ∂ ∂ ∂
h̃ur = hur − ju − jr + 2 jr + r G + r G − M G,
∂r ∂u r ∂u∂r ∂u ∂r
2
∂ ∂ ∂
h̃rr = hrr − 2 jr + r2 2 G + 2r G,
∂r ∂r ∂r
2 2f ∂ ∂ λ
K̃ = K + ju − jr + rf G − r G + G.
r r ∂r ∂u 2
The field equations are
∂2 2 ∂ 1 ∂ f ∂ 2 ∂ λr + 4M
Quu = − K̃ − K̃ + h̃rr + h̃rr − h̃ur + h̃rr ,
∂r2 r ∂r r ∂u r ∂r r ∂r 2r3
2
∂ 2 ∂ r−M ∂ f ∂ 1 ∂ 1 λ+4 f µ
Qur = K̃ + K̃ − 2
K̃ − h̃rr + h̃uu + 2 h̃uu − 2
h̃ur + 2 h̃rr + 2 K̃,
∂u∂r r ∂u r ∂r r ∂u r ∂r r 2r r 2r
∂2 r−M ∂ (r − M )f ∂ 1 ∂ 2f ∂ f ∂
Qrr = − 2 K̃ − K̃ + K̃ − h̃uu + h̃ur − h̃uu
∂u r2 ∂u r2 ∂r r ∂u r ∂u r ∂r
µr + 4M 2f f2 µf
+ 3
h̃ uu + 2
h̃ ur − 2
h̃rr − 2 K̃,
2r r r 2r
∂ ∂ ∂ 2 r−M
Qu = h̃rr − h̃ur + K̃ + h̃ur − h̃rr ,
∂u ∂r ∂r r r2
∂ ∂ ∂ ∂ 2(r − M ) (r − M )f
Qr = − h̃ur + h̃uu + K̃ − f K̃ − h̃ur + h̃rr ,
∂u ∂r ∂u ∂r r2 r2
∂2 ∂2 ∂2 ∂2 ∂2 r−M ∂ 2 ∂ 2 ∂
Q♭ = − 2 h̃rr + 2 h̃ur − 2 h̃uu − 2 K̃ + f 2 K̃ + h̃rr − K̃ − h̃uu
∂u ∂u∂r ∂r ∂u∂r ∂r r2 ∂u r ∂u r ∂r
2 2
2(r − M ) ∂ (r − M )f ∂ 2(r − M ) ∂ λ λr − 2µM r − 4M
+ 2
h̃ur − 2
h̃rr + 2
K̃ + 2 h̃ur − h̃rr ,
r ∂r r ∂r r ∂r r 2r4
Q♯ = 2h̃ur − f h̃rr .
The conservation (Bianchi) identities read
∂ uu ∂ ur M uu 2 ur λ 1
0 = Q + Q − 2 Q + Q − 2 Qu + Q♭ ,
∂u ∂r r r 2r r
∂ ur ∂ rr M f uu 2M ur 2 rr λ f
0 = Q + Q + 2 Q + 2 Q + Q − 2 Qr − Q♭ ,
∂u ∂r r r r 2r r
∂ u ∂ r 2 r ♭ µ ♯
0 = Q + Q + Q + Q − 2Q .
∂u ∂r r 2r
The Zerilli-Moncrief function is
  
2r 2 ∂ ∂
Ψeven = K̃ + h̃uu − 2f h̃ur + f 2 h̃rr + r K̃ − rf K̃ ,
µ+2 Λ ∂u ∂r
where Λ = µ + 6M/r, and its source term is
   
4 r 1 ♯ 2 2 ∂ ∂ uu ur 24M rr
Seven = Q − Q + −2r −f f Q + 2Q + Q + 2rf Q♭
Λ r (µ + 2)Λ ∂u ∂r Λ
 18

M2
  
r M
− µ(µ − 2) + 12(µ − 3) + 84 2 f Quu + 2Qur .
Λ r r

The odd-parity perturbations are hu , hr , and h2 . Under a gauge transformation generated by ξ they change
according to hu → h′u = hu + ∆hu , etc., with

∂
∆hu = − ξ,
∂u
∂ 2
∆hr = − ξ + ξ,
∂r r
∆h2 = −2ξ.

The gauge-invariant fields are

1 ∂
h̃u = hu − h2 ,
2 ∂u
1 ∂ 1
h̃r = hr − h2 + h2 .
2 ∂r r
The field equations are

∂2 ∂2 2 ∂ 2 µ
Pu = − h̃r + 2 h̃u − h̃r − 2 h̃u − 2 h̃r ,
∂u∂r ∂r r ∂u r r
∂2 ∂2 2 ∂ µ µf
Pr = h̃ r − h̃u + h̃u − 2 h̃u + 2 h̃r ,
∂u2 ∂u∂r r ∂u r r
∂ ∂ ∂ 2M
P = − h̃r − h̃u + f h̃r + 2 h̃r .
∂u ∂r ∂r r
The conservation (Bianchi) identity reads

∂ u ∂ r 2 r µ
0= P + P + P − 2 P.
∂u ∂r r r
The Cunningham-Price-Moncrief function is
 
2r ∂ ∂ 2
Ψodd = h̃u − h̃r − h̃u ,
µ ∂r ∂u r

and its source term is

 
2r ∂ u ∂ u ∂ r 2M u
Sodd = − P +f P + P + 2 P .
µ ∂u ∂r ∂r r

APPENDIX E: PERTURBATION EQUATIONS IN (v, r) COORDINATES

In this Appendix the two-dimensional Schwarzschild metric is written in the form

ds2 = −f dv 2 + 2 dvdr,

with f = 1 − 2M/r. In these coordinates the wave operator is given by

∂2
 
∂ ∂
Ψ = 2 + f Ψ.
∂v∂r ∂r ∂r

We set λ := l(l + 1) = µ + 2 and µ := (l − 1)(l + 2) = λ − 2.

The even-parity perturbations are hvv , hvr , hrr , jv , jr , K, and G. Under a gauge transformation generated by ξv ,
ξr , and ξ they change according to hvv → h′vv = hvv + ∆hvv , etc., with

∂ 2M 2M f
∆hvv = −2 ξv + 2 ξv + 2 ξr ,
∂v r r
 19

∂ ∂ 2M
∆hvr = − ξv − ξr − 2 ξr ,
∂r ∂v r
∂
∆hrr = −2 ξr ,
∂r
∂
∆jv = − ξ − ξv ,
∂v
∂ 2
∆jr = − ξ − ξr + ξ,
∂r r
2f 2 λ
∆K = − ξr − ξv + 2 ξ,
r r r
2
∆G = − 2 ξ.
r
The gauge-invariant fields are

∂ 2M 2M f ∂2 ∂ ∂
h̃vv = hvv − 2 jv + 2 jv + 2 jr + r2 2 G − M G − M f G,
∂v r r ∂v ∂v ∂r
∂ ∂ 2M ∂2 ∂ ∂
h̃vr = hvr − jv − jr − 2 jr + r2 G + r G + M G,
∂r ∂v r ∂v∂r ∂v ∂r
2
∂ 2 ∂ ∂
h̃rr = hrr − 2 jr + r G + 2r G,
∂r ∂r2 ∂r
2 2f ∂ ∂ λ
K̃ = K − jv − jr + rf G + r G + G.
r r ∂r ∂v 2
The field equations are

∂2 2 ∂ 1 ∂ f ∂ 2 ∂ λr + 4M
Qvv = − 2
K̃ − K̃ − h̃rr + h̃rr + h̃vr + h̃rr ,
∂r r ∂r r ∂v r ∂r r ∂r 2r3
∂2 2 ∂ r−M ∂ f ∂ 1 ∂ 1 λ+4 f µ
Qvr = K̃ + K̃ + K̃ − h̃rr − h̃vv − 2 h̃vv − h̃vr − 2 h̃rr − 2 K̃,
∂v∂r r ∂v r2 ∂r r ∂v r ∂r r 2r2 r 2r
∂2 r−M ∂ (r − M )f ∂ 1 ∂ 2f ∂ f ∂
Qrr = − 2 K̃ + K̃ + K̃ + h̃vv + h̃vr − h̃vv
∂v r2 ∂v r2 ∂r r ∂v r ∂v r ∂r
µr + 4M 2f f2 µf
+ 3
h̃vv − 2 h̃vr − 2 h̃rr − 2 K̃,
2r r r 2r
∂ ∂ ∂ 2 r−M
Qv = h̃rr − h̃vr − K̃ + h̃vr + h̃rr ,
∂v ∂r ∂r r r2
∂ ∂ ∂ ∂ 2(r − M ) (r − M )f
Qr = − h̃vr + h̃vv − K̃ − f K̃ + h̃vr + h̃rr ,
∂v ∂r ∂v ∂r r2 r2
∂2 ∂2 ∂2 ∂2 ∂2 r−M ∂ 2 ∂ 2 ∂
Q♭ = − 2 h̃rr + 2 h̃vr − 2 h̃vv + 2 K̃ + f 2 K̃ − 2
h̃rr + K̃ − h̃vv
∂v ∂v∂r ∂r ∂v∂r ∂r r ∂v r ∂v r ∂r
2(r − M ) ∂ (r − M )f ∂ 2(r − M ) ∂ λ λr2 − 2µM r − 4M 2
− 2
h̃vr − 2
h̃rr + 2
K̃ − 2 h̃vr − h̃rr ,
r ∂r r ∂r r ∂r r 2r4
Q♯ = −2h̃vr − f h̃rr .

The conservation (Bianchi) identities read

∂ vv ∂ vr M vv 2 vr λ 1
0 = Q + Q + 2 Q + Q − 2 Qv − Q♭ ,
∂v ∂r r r 2r r
∂ vr ∂ rr M f vv 2M vr 2 rr λ f
0 = Q + Q + 2 Q − 2 Q + Q − 2 Qr − Q♭ ,
∂v ∂r r r r 2r r
∂ v ∂ r 2 r µ
0 = Q + Q + Q + Q♭ − 2 Q♯ .
∂v ∂r r 2r
The Zerilli-Moncrief function is
  
2r 2 2 ∂ ∂
Ψeven = K̃ + h̃vv + 2f h̃vr + f h̃rr − r K̃ − rf K̃ ,
µ+2 Λ ∂v ∂r
 20

where Λ = µ + 6M/r, and its source term is

   
4 r 1 ♯ 2 2 ∂ ∂ vv vr 24M rr
Seven = Q − Q + 2r +f f Q − 2Q + Q + 2rf Q♭
Λ r (µ + 2)Λ ∂v ∂r Λ
M2
  
r M
− µ(µ − 2) + 12(µ − 3) + 84 2 f Qvv − 2Qvr .
Λ r r

The odd-parity perturbations are hv , hr , and h2 . Under a gauge transformation generated by ξ they change
according to hv → h′v = hv + ∆hv , etc., with

∂
∆hv = − ξ,
∂v
∂ 2
∆hr = − ξ + ξ,
∂r r
∆h2 = −2ξ.

The gauge-invariant fields are

1 ∂
h̃v = hv − h2 ,
2 ∂v
1 ∂ 1
h̃r = hr − h2 + h2 .
2 ∂r r
The field equations are

∂2 ∂2 2 ∂ 2 µ
Pv = − h̃r + 2 h̃v − h̃r − 2 h̃v + 2 h̃r ,
∂v∂r ∂r r ∂v r r
∂2 ∂2 2 ∂ µ µf
Pr = h̃ r − h̃v + h̃v + 2 h̃v + 2 h̃r ,
∂v 2 ∂v∂r r ∂v r r
∂ ∂ ∂ 2M
P = h̃r + h̃v + f h̃r + 2 h̃r .
∂v ∂r ∂r r
The conservation (Bianchi) identity reads

∂ v ∂ r 2 r µ
0= P + P + P − 2 P.
∂v ∂r r r
The Cunningham-Price-Moncrief function is
 
2r ∂ ∂ 2
Ψodd = h̃v − h̃r − h̃v ,
µ ∂r ∂v r

and its source term is

 
2r ∂ v ∂ ∂ r 2M v
Sodd = P + f Pv − P + 2 P .
µ ∂v ∂r ∂r r

[1] T. Regge and J. A. Wheeler, Phys. Rev. 108, 1063 Netherlands, 1998).
(1957). [6] K. Martel, Phys. Rev. D 69, 044025 (2004).
[2] C. V. Vishveshwara, Phys. Rev. D 1, 2870 (1970). [7] M. Sasaki and H. Tagoshi, Living Rev. Relativity 6
[3] F. J. Zerilli, Phys. Rev. D 2, 2141 (1970). (2004), 6. [Online article]: cited on February 7, 2008,
[4] S. Chandrasekhar, The mathematical theory of black http://www.livingreviews.org/lrr-2003-6.
holes (Clarendon Press, Oxford, England, 1983). [8] R. Gleiser, C. Nicasio, R. Price, and J. Pullin, Phys.
[5] V. P. Frolov and I. D. Novikov, Black hole physics: Basic Rept. 325, 41 (2000).
concepts and new developments (Kluwer, Dordrecht, The [9] V. Moncrief, Ann. Phys. (N.Y.) 88, 323 (1974).
 21

[10] U. H. Gerlach and U. K. Sengupta, Phys. Rev. D 22, tation (Freeman, San Francisco, 1973).
1300 (1980). [16] J. Goldberg, A. MacFarlane, E. Newman, F. Rohrlich,
[11] O. Sarbach and M. Tiglio, Phys. Rev. D 64, 084016 and E. Sudarshan, J. Math. Phys. 8, 2155 (1967).
(2001). [17] K. S. Thorne, Rev. Mod. Phys. 52, 299 (1980).
[12] C. T. Cunningham, R. H. Price, and V. Moncrief, Astro- [18] C. O. Lousto and R. H. Price, Phys. Rev. D 55, 2124
phys. J. 224, 643 (1978). (1997).
[13] S. Jhingan and T. Tanaka, Phys. Rev. D 67, 104018 [19] E. Poisson, Phys. Rev. D 70, 084044 (2004).
(2003). [20] E. Poisson, Phys. Rev. D 69, 084007 (2004).
[14] K. Martel, Ph.d. thesis, University of Guelph (2003).
[15] C. W. Misner, K. S. Thorne, and J. A. Wheeler, Gravi-