THE MATHEMATICAL THEORY OF BLACK HOLES

S. Chandrasekhar
The University of Chicago
Chicago, Illinois 60637, USA

In a course of lectures on the 'underlying mathe-

matical structures of classical gravitation theory'
given in 1978, Brandon Carter began with the statement
'If I had been asked five years ago to prepare a course
of lectures on recent developments in classical gravi-
tation theory, I would not have hesitated on the clas-
sical theory of black holes as a central topic of dia-
cussion. However, the most important developments in
gravitational theory during the last three or four
years have not been in the classical domain at all ... '
Carter is undoubtedly right in his assessment that the
mathematical theory of black holes has not been in the
mainstream of research in relativity since 1973. I
therefore find it difficult to understand why the orga-
nizers of this meeting should have chosen precisely .
this topic for the opening talk of this meeting. But I
am grateful to them for their courtesy in assigning to
me this privilege.
The mathematical theory of black holes has held
my attention during the past ten years because of its
classical severity and because also of its relevance
to our understanding of real physical objects in the
universe around us. As I have stated elsewhere, 'the
black holes of nature are the most perfect macroscopic
objects there are in the universe'. They are the most
perfect almost by definition since the only elements
in their construction are our concepts of space and
time. And since the general theory of relativity pro-
vides only a single unique family of solutions for
their description, they are the simplest objects as
5
B. Bertotti et aL (eds.), General Relativity and Gravitation, 5-26.
<iii 1984 by D. Reidel Publishing Company.
•
well.

In my consideration of the mathematical theory of

black holes I shall restrict myself to exhibiting some
r.eatures of the theory which reveal its richness. And
as in a hurried guided tour of ' a museum, I shall draw
your attention only to a very few chosen exhibits: and
I shall not have time to dwell at any length even on
the chosen exhibits .
Let me begin by defining a black hole. For my
purposes, it will suffice to say that a black hole is
a time-independent asyuptotically flat space-time sat=
isfytng Einstein's vacuum field-~quations which parti-
tions the three-dimensional space into two regions: an
external region and an internal region separated by a
smooth convex two-dimensional surface - the event
horizon - 1n such a way that no point in the interior
is communicable to the space outside. By a theorem due
to David Robinson the only solutions of Einstein's
equations, compatible with these requirements, are
those which belong to the Kerr family with its two pa-
rameters, the mass of the black hole and the angular
momentum of the black hole. The static solution with
zero angular momentum is the Schwarz schild solution.
Besides the Schwarzschild and the Kerr solutions, I
shall also consider the Reissner-Nordstr6m solution
which is a solution of the Einstein-Maxwell equations
and represents the charged counte~part of the Schwarz-
schild solution.
May I digress for a moment to point out that it
is now exactly twenty years since Kerr discovered his
solution. In my judgement, it is the most important
discovery in general relativity since its founding;
and I should like to congratulate Kerr on this occasion.
The mathematical properties to which I shall refer
derive from a consideration of the perturbations of
the black holes. It is convenient to consider the per-
turbations as resulting from the incidence of waves of
different sorts such as gravitational waves, electro-
magnetic waves, and massless or massive spin-~ parti-
cles. Stated simply and somewhat narrowly, the goal is
to determine the scattering and the absorption which
the incident waves will experience. The problems of
the incidence of gravitational waves on uncharged
black-holes and of graVitational or electromagnetic
waves on charged black-holes are in some sense the most
relevant since they will result in the perturbation of
the space-time itself and therefore of the metric coef-
ficients as well.
BLACK HOLES

I shall consider first the problem of the

reflexion and transmission of incident gravitational-
-waves by the Schwarzschild black-hole.

The Schwarzschild metric in its standard form (in

the units (c=G=1) is

dS2=~(dt) 2_r2 [(d~) 2+ (de) 2+sin 2e (dIP) 2] , (1)

r

where
(2)

is the 'horizon function'. In considering the perturba-

tionsof this space-time, one must remember that the
metric (1) is written in a coordinate system so adapted
to its spherical symmetry that it takes its simplest
form. The perturbed space-time will not have any spe-
cial symmetry and the coordinate conditions that have
been imposed to achieve the simplicity of the static
metric must be relaxed. However, since we are consider-
ing the perturbations of a spherically symmetric back-
ground, we may without any loss of generality suppose
that while the system is non-stationary it has, never-
theless, symmetry about some chosen axis. And it can
be shown that the metric appropriate to a non-station-
ary axisymmetric space-time can be reduced to the form

(3)

where v, ~, ~2' ~3' W, Q2' and Q3 are all functions of

t, r(=x 2), and e(=x 3).
The Schwarzschild line-element, considered as a
special case of the line-element (3) has the metric
coefficients
2V -2~2 ~3
e =e =t:./r2, e =r, e~=rsine, (4)

and
(5 )
8 S. CHANDRASEKHAR

Consequently, a general perturbation of the Schwarz-

schild black-hole will result in w, q2' and q3 becoming
small quantities of the first order of smallness and
the functions v, ~2' ~3' and ~ experiencing small
increments ov, O~2' O~3' ana o~. It is clear that per-
turbations leading to non-vanishing values of w, q2'
and q3 and perturbations leading to increments ov, O~2'
O~3' and o~ are of very different kinds: the former
induce a dragging of the inertial frame and impart a
rotation to the black hole while the latter impart no
such rotation. For this reason we shall call them axial
and polar perturbations, respectively. As one may
expect the two classes of perturbations decouple in
the sense that they can be considered independently of
one another.
In our further considerations it will be assumed
that the various quantities representing the perturba-
tions have a time-dependence given by -.

(0)0); (6)

and that it is with the amplitudes of the perturbations

(with this time-dependent factor) that we shall be con-
cerned. With this assumed time-dependence, we find
that, after the separation of the variables, the various
equations governing the perturbations can be reduced
to a pair of one-dimensional wave-equations of the
forms,

(7)

where the superscripts "+" and "-" distinguish the

polar and the axial perturbations, V(±) represent
short-range potentials (1) (positive everywhere) and
r* is a radial coordinate defined by

(8)

or, explicitly,
r*=r+2Mln(r/2M-1). (9)
The solution to the perturbation problem is thus re-
duced to the elementary one of barrier penetration in
one dimension with which we are familiar in quantum
theory. Precisely, we must seek solutions of the wave
BLACK HOLES 9

equations which satisfy the boundary conditions

Z (±) +e +ior*+R (±) (0) e -iOr*

(10)

And the required reflection and transmission coeffi-

cients for incident gravitational waves are given by

jt' ±) (0) = IR (±) (0) I 2 and tf ±) (0) = IT (±) (0) I 2 • ( 11)

Now the potentials vet) belonging to the axial

and the polar perturbations are given by

v(+)=~ ~4(~2+2)r3+6M~4r2+36M2~2r+72M3
( 12)
r 5 (~2r+6M) 2

and

:5 [(
v (- ) = ~ 2+2) r-6M] , (13 )

where
(14 )

It is, in the first instance, a remarkable (and an

unexpected) fact that in spite of their entirely dis-
similar appearances, the potentials V(+) and v(-) are
very simply related: they are included in the simple
expression,

(15 )

where

and (16 )

The fact that the potentials v(+) and V(-) are related
10 S. CHANDRASEKHAR

in the manner (15) has an important consequence: given

a solution Z(-) belonging to v(-), the expression,

(17 )

provides a solution for Z(+) belonging to V(+). From

the relation (17) it follows that a solution z(+),
derived from a solution Z(-), satisfying the boundary
conditions (10) (appropriate for Z(-») ,satisfies the
boundary conditions (10) (appropriate for Z (+»), 1I1ith
a reflexion and a transmission amplitude that are re-
lated to those for Z(-) by

The reflexion amplitudes thus differ only by a phase

while the transmission amplitudes are the same. We
conclude that the reflexion and the transmission aoef-
fiaients for the axial and the polar perturbations are
the same for all inaident frequenaies.

The principal features of the perturbed Schwarz-

schild space-time we have described are manifested in
the more general context of the perturbed Reissner-
-Nordst~om space-time.
The metric of the Reissner-Nordstrom space-time
(in its standard form) differs from the Schwarzschild
metric (1) only in the definition of the horizon
function, 6: in place of (2), we now have

( 1 9)

where Q* denotes the charge of the black hole. But one

important difference between the two space-times is
clear from the outset. The Reissner-Nordstrom space-
-time is a solution of the coupled Einstein-Maxwell
equations and not of Einstein's vacuum-equations. On
this account, one should expect that the incidence of
a pure gravitational-wave on the black hole must result
BLACK HOLES 11

in the excitation of electromagnetic waves and con-

versely.
It is clear that the perturbations will again
fall into two non-combining groups having axial and
polar characteristics. And we find that in spite of
the additional complexities resulting from the require-
ment of treating the coupled system of the Einstein-
-Maxwell equations, the various equations governing
the perturbations can be reduced to two pairs of one-
-dimensional wave-equations with potentials vl±) (i=1,2
and "+" and "-" distinguish the polar and the axial
perturbations) belonging to the same class (15) as in
the context of the Schwarzschild space-time. We find

(i=1 ,2) , (20)

where r* is now defined, in terms ~ given by equation

(19), by the same equation (8); and the potentials,
vi±), occurring in equation (20) are given by

V~±)=±8. ddrfi+8~f~+~2(~2+2)f.,
~ *
~ ~ ~ ~
(21)

where

81=3M-(9M2+4Q;~2)~,
(22)
82=3M+(9M2+4Q;~2)~,
and

(i=1 ,2) • (23)

It follows that the reflexion and the transmission am-

plitudes for the solutions Z{±~ satisfying boundary
conditions analogous to (10), must be related in the
manner (cf. equation (18»
12 s. CHANDRASEKHAR
and T ~ +) (0) =T ~ -) (0) ,
~ ~

(24)
iOi ~Z(~z+2)-2ioei
where e (i=1,2)

However, it should be noted that the reflexion and the

transmission coefficients derived from the amplitudes
Rl±)(0) and Ti±) (0) do not refer to incident gravita-
tional or electromagnetic waves: the amplitudes of
these waves are not given by z~±) and Z!±) but by li-
near combinations,of them. Suppressing the distingui-
shing super-scripts and denoting by H1 and Hz the ampli-
tudes of the prevailing electromagnetic and gravita-
tional waves, we have

and (25)

where
(26)

It is clear that in terms of the relations (25) we can

express the coefficients of reflexion and transmission
of electromagnetic and gravitational waves for any in-
cident arbitrary superposition of such waves. The most
convenient way to express these coefficients is by
means of a scattering matrix, 8 , which relates the
amplitudes of the fluxes (H(~)
1 '
H(l»
z and (H(r)
1 ,
H(r»
z
incident on the potential barriers from the left and
the right, respectively, to the amplitudes of the
fluxes (H~~), H!~» and (H~r), a!r» reflected to the
left and to the right, respectively:
BLACK HOLES 13

+-(r) +-(R.)
HI HI
-+- ( R.) -+-(r)

s· HI
+-(r)
H2
HI
+-(R.)
H2
( 27)

-+-( t) -+-(r)
H2 H2

We find that the scattering matrix, expressed in terms

of the amplitudes Ri(a) and Ti(a) (appropriate for the
solution, Zi) is given by(2)

-
TIl Rll Tn R2l
-
Rll Tll R2l T2l
s= T12 - -
R12 T22 R22
(28)

Rl2 T12 R22 T22

where

(29)

By virtue of these definitions, the scattering matrix

is unitary and time-symmetric. The form of S implies,
consistently with the expectation stated earlier, that
any perturbation of the metric must unavoidably result
in the excitation of both electromagnetic and gravita-
tional waves. Further, from the unitarity and the time
symmetry of the scattering matrix, it follows that the
fraction of the energy, incident as pure gravitational
waves, that is converted into electromagnetic energy
(in the form of reflected electromagnetic waves) must
be the same as the fraction of energy, incident as pure
 S.CHANDRASEKHAR

electromagnetic waves, that is converted into gravita-

tional energy (in the form of reflected gravitational
waves) .

We now turn to the reasons for the equality of

the reflexion and the transmission coefficients asso-
ciated with Z(+) and Z(-) in the context of the Schwarz-
schild black-hole and with Z(+) and Z(-) (i=1,2) in
the context of the Reissner-Nordstrom black-hole. I do
not know if there is any clear physical reason for
these equalities. But the mathematical base for them
can be adequately explained by considering the alter-
native formulation of the perturbation problem in the
framework of the Newman-Penrose formalism. Let me pref-
'ace the explanation by a few remarks concerning the
formalism itself.
In the Newman-Penrose formalism, one sets up at
each point in the space-time a basis for the tangent
space conSisting of two real null-vectors, 1 and n,
and a complex-conjugate pair of null vectors m and m
orthogonal to both & and n. One then projects onto
this basis all the quantities of interest. Thus, by
projecting the Weyl tensor, for example, one replaces
its ten components by five complex scalars ~o' ~1' ~2'
~3' and ~~. When one is dealing with type-D space-times,
as the space-times of black holes are, one chooses for
the real null-vectors, land n, the two shear-free con-
gruences of null geodesics which the space-time allows.
With this choice of basis, the Weyl scalars ~o' ~1'
~3' and ~~ vanish in the stationary background and ~2
is the only non-vanishing scalar. In the perturbed
state, when gravitational waves will be present, ~o,
~1' ~3' and ~~ will cease to vanish. The aspect of the
Newman-Penrose formalism, which makes it so adaptable
to the treatment of the perturbations of type-D space-
-times, is that it provides, directly, equations for
~o and ~~, without any reference to the metric pertur-
bations1 and the radiative aspects of the perturbations
can be ascertained directly by solving for ~o and ~~.
In particular, the reflexion and the transmission coef-
ficients for incident waves can be inferred without
reference to any other features of the perturbation1
and the coefficients one obtains in this manner must
agree with what one derives from other perturbation
schemes. It is this last fact which guarantees the
equality of the reflexion and the transmission coeffi-
cients derived from the separate considerations of
BLACK HOLES 15

axial and polar perturbations. Let me consider how

this arises in the specific mathematical contexts.

In the context of the Schwarz schild space-time

we find that, with the substitution

(30 )

where S+2(8) is a suitably defined angular function,

the radial function Y(r) satisfies the equation

( 31)

where

- d +'
A+--d _1.0, (32)
_ r*

d r8
p=- In (33)
dr* /::,2

Since we should be able to deduce the reflexion and

the transmission coefficients for incident gravitation-
al waves in terms of the solutions of equation (31)
and its complex conjugate, it follows that it must be
possible to transform equation (31), simultaneously,
to either of the two equations governing z(+l and z(-l.
The existence of such dual transformations is guaran-
teed by the coefficient Q in equation (31) satisfying a
certain non-linear ordinary differential equation of
the second order. Thus, by letting (cf. equation (16))

(34)

we can verify that

(35)

By virtue precisely of this equation (and, in particu-

16 s. CHANDRASEKHAR
lar, by the occurrence of S2 on the right-hand side),
we are enabled to make the following dual transforma-
tions,

(36)
and
(37)

where

w(-)=~(r-3M),W(+)=2 ~2r2-3~2Mr-6M2 (38)

r2 r2 (~2r+6f.1)

(39 )

and V(+} and V(-} are given by equations (12) and (13).
The content of equations (36) and (37) is that with
Z(+) and Z(-) defined as solutions of the respective
wave equations, Y given by equation (36) satisfies
equation (31); and conversely with Y defined as a solu-
tion of equation (31), Z(+} and z(-) given by equation
(37) satisfy the respective wave-equations. It is the
existence of these transformations that accounts for
the remarkable relationships that we encountered in
our earlier consideration of the axial and the polar
perturbations of the 5chwarzschild black-hole;
Turning to the Newman-Penrose equations governing
the perturbations of the Reissner-Nordstrom space-time,
we find that their decoupZing depends crucially on two
facts; the spherical symmetry of the background space-
-time which ensures their separability and the special
circumstances that the angular functions, 51 and 52'
belonging to the spin-1 Maxwell-field and spin-2 gravi-
tational-field are related very simply: 51 being a
simple linear combination of 52 and its derivative,
and conversely. (It is by virtue of this last fact
that the common characteristic-value parameter ~ occurs
in the definition of Sl and S2, and of the "mixing"
angle Wgiven by equation (37).) because of these spe-
cial circumstances the equations of the Newman-Penrose
formalism governing the perturbed Reissner-Nordstrom
space-time are reducible to the pair of independent
equations,
BLACK HOLES 17

(i=1,2) (40)

where
d r8
P =-d In -D.' (i,j=1 ,2;i#j), (41)
i r* ~

and
(i=1 ,2;i#j). (42)

And equation (40) allows dual transformations, as in

the context of the Schwarzschild space-time, by virtue
of Qi satisfying a certain nonlinear ordinary differen-
tial equation. Thus, with the definition,

(43 )

Fi satisfies the equation (cf. equation (33))

_1_ (dFi)2_d2Fi +Di F2= Sf +~2(~2+2) (i=1,2), (44)

Fi dr* dr; r8 i Fi

where the Si's are defined in equations (21).

Thus, while the mathematical structure of the
theory clarifies the various underlying relationships
and in particular the origin of the potential barriers
belonging to the class (15) (which guarantees the equal-
ity of the derived reflexion and the transmission
coefficients), it cannot be said that the deeper (?)
physical significance of the relationships has been
understood.

I now turn to a consideration of the Kerr space-

-time.
The Kerr space-time is stationary and possesses
an axis of symmetry. On these accounts one naturally
expects that the equations of mathematical physics,
written in this geometry, will separate in the vari-
ables t and f. But one does not expect to separate the
variables rand 6; and the fact that they do endows
the Kerr geometry with an aura of the miraculous.
We list below the equations and the authors who
separated them:
18 S. CHANDRASEKHAR

i) The Hamilton-Jacobi equation~ B.Carter (1968)

ii) The sealar wave-equation~ B.Carter (1968)
iii) Maxwell's equations~S.Teukolsky (1971)
iv) Gravitational equations~ S.Teukolsky (1971)
v) Dirac's equation~ S.Chandrasekhar (1976).
The mathematical theory underlying these various equa-
tions is inherently so complex that, within the allot-
ted time, I can hardly give even a glimmering of the
richness of the subject. In my account I shall restrict
myself to the propagation of gravitational waves in
the Kerr space-time~ this is only one aspect of the
larger problem of the metric perturbations.
In Kerr geometry as in Schwarzschild's geometry,
the equations governing the Weyl scalars ~o and ~~
decouple from the rest and can be considered indepen-
dently of the others~ and as Teukolsky showed in 1971,
these equations can be separated. For perturbations
with a t- and a tp -dependence given by

(m=0,±1,±2, ••• ), (45)

one finds that with the substitutions.

1
~ ~ =-_--R_ 2 (r) S_2 (6) , (46)
(p*) ~

where
p=r+iacosS, p*=r-iacosS, (47)

the separated equations are

(48)

(49 )

(50)

(51)

where A is a characteristic-value parameter (determined

by the requirements that S+2 and S_2 are smooth at the
BLACK HOLES 19

t
poles) and !i}n, !i}n, 2'n and 2'nt are differential operators
defined by the equations,

(52)
;t-
~=ae+Q+ncote, 2' =-2' ('IT-e)
n n '
where

(53)

It is clear from the foregoing definitions that ~2R+2

and R_2 satisfy complex-conjugate equations, while
8+ 2 and 8_ 2 satisfy equations, one of which can be
obtained by the other by replacing e by 'IT-e.
The solutions of the pair of equations (48) and
(49) and (50) and (51) satisfy certain remarkable iden-
tities which derive from the following directly veri-
fiable commutation relations:

and

~l2'O ~ ~ (~12'2 +6aocose) = (~12't-6aocOse) ~12'O~2'2. (55)

These commutation relations have far-reaching conse-

quences for the theory. We shall consider some of them.
The commutation relation (54) implies, for example,
that if ~2R+2 is a solution of equation (48),
~2!i}!!i};~!i}!(~2R+2) is a solution of equation (49); and
conversely, if R_2 is a solution of equation (49),
~2~!li6 ~!i}oR_2is a solution of equation (48). In other

words, we may write

(56)

and (57 )
20 S. CHANDRASEKHAR

where ~1 and ~2 are two complex constants. Similarly,

the commutation relation (55) allows us to write

(58 )

and (59)

where 0 1 and 02 are two real constants. Equations (56)-

-(59) are the Teukolsky-Starobinsky identities. At
this point, one may ask why the simple act of taking
the complex conjugate of a solution of equation (48)
(or (49)) is equivalent to applying a fourth-order dif-
ferential operator to it; or, why the simple act of
substituting e by n-e in a solution of equation (50)
(or (51)) is equivalent to applying a fourth-order dif-
ferential operator to it. The answers to these questions
are sheathed in enigmas.
Further consequences of equations (56)-(59) are:

.L\ 2£iJ!£iJ!£iJ;£iJ!L\ 2£iJ/l o ~ ~=~1 ~2

= 1~12 (say) mod L\£iJ~1 £iJo+6icrr-A=0 (60)

and
.l.
=0 2 (say) rnod~~1~+6acrcose+A=0, (61)

where 1~12 and 0 2 are certain 'universal' constants.

We shall call them the Starobinsky constants.
From the manner in which equations (60) and (61)
have been derived, it would appear that there is no al-
ternative to the evaluation of the Starobinsky constants
except by a direct (but enormously tedious) reduction
of the left-hand side of these equations and the suc-
cessive application of the modulo conditions. By such
reductions, one finds that

( 62)

and
BLACK HOLES 21

0 2=12(1+2)2-80 2 1 b2(51+6)-12a~

+1 44 a 4 a 4 = I~ I 2'_1 440 2M2 , (63 )

where it may be recalled- that
(64)
At this point one may ask: why does the evaluation of
1~12 require so much detailed calculation; and what,
indeed, is the origin of the simple relation (63)
between the Starobinsky constants 0 2 and 1~12? Let me
digress briefly to answer, in part, these last ques-
tions.

An evaluation of 1~12, 'without tears', emerges from

a consideration of the so-called 'algebraically spe-
cial' perturbations. These are perturbations in which
only one of the scalars, ~o and ~4' is non-vanishing,
while the other vanishes. From equations (46), (56)
and (57), it follows that the requirement, ~oiO
(Le., R+dO), and ~4::::0 (Le., R_ 2::::O), for example, can
be met if and only if
~l =0, (65)

(66)
and ~2R+2 satisfies equation (48).
The general solution of equation (66) can be rea-
readily written by making use of the fact that

~-;. ili 2 ( ---

= -- d -io ) , (67)
o ~ dr*

where (68)
and r* is a new variable derived in accordance with
the equation (3) ,

dr ili
--*-
2
dr = --f'..
(69 )

It follows from equation (67) that

~r ( n +ior*) =nr n-l e +iOr* .
ore (70)
22 S. CHANDRASEKHAR

Accordingly, the general solution of equation (66) is

given by

(71)

where A, B, C, and D are constants of integration. The

requirement now that ~2R+2 satisfies equation (48)
leads to the conditions

AA+21aB=O,
6(M-iaa 2)A+(A+2)B+4iaC=O,
(72)
6a 2A+4iaa 2B-(A+2)C-6iaD=O,
2a 2B+.2(M+iaa 2)C-AD=O. (72)

The determinant of this system of equations must there-

fore vanish~ and a simple evaluation of the determi-
nant leads to the condition,
(73)

We have thus an alternative derivation (without tears!)

of the Starobinsky constant.

Returning to equations (56)and (57), we observe

that in view of ~ 2R+_2 and R-2 satisfying complex-con-
jugate equations, we may so normalize them, relative-
ly, tnat tney satisfy the complex-conjugate equations,

and (74 )

where we~ave written P+2 and P- 2 in place of ~2R+2

and R_2 to signify the chosen normalization of these
functions. It is important to emphasize that even
after this normalization, there is an ambiguity in the
solutions for P+2 and P- 2 which the physical problem
demands. For, it is clear on general grounds that the
solution of the entire perturbation-problem must be
unique apart from a single real constant of proportion-
ality. Therefore, there can be no ambiguity in P+2 ,
BLACK HOLES 23

once P-2 has been specified in accordance with some

selected boundary condition. In the context of equa-
tions (74), with the chosen relative normalization of
P+ 2 and P- 2 , the ambiguity resides in the relative
phases of P+ 2 and P- 2 • The resolution of this ambiguity
will consist in determining the real and the imaginary
parts of ~ (consistent with 1<G12 given by equation (62));
but it will still leave open the question of the numer-
ical scale-factor by which P- 2 must be multiplied to
obtain the proper solution of the physical problem.
The solution of these two problems - that of determin-
ing the real and the imaginary parts of ~ and that of
ascertaining the numerical scale-factor - are necessary
prerequisites for determining the reflexion and trans-
mission coefficients, for incident gravitational-
-waves, in terms of the solutions of equations (48) and
(49); and they require the consideration of the entire-
ty of the equations provided by the Newman-Penrose
formalism.
The Newman-Penrose formalism provides a formidable
array of equations: a set of 18 complex equations ex-
pressing the components of the Riemann tensor (derived
from the Ricci identity)in terms of the directional
derivatives of the spin coefficients (the Ricci rota-
tion-coefficients in the context) and the Weyi and
the Ricci scalars; a set of 10 complex equations (or 8
equations for a vacuum) expressing the Bianchi identi-
ties in terms of the directional derivatives of the
Weyl and the Ricci scalars and the spin coefficients;
and a set of 12 complex equations expressing the Lie
brackets of the basis vectors with the structure con-
stants written out in terms of the spin coefficient;
altogether an array of 40 (or 38, for the vacuum)
coupled complex-equations. The consideration of this
array of equations, for a complete treatment of the
gravitational perturbation of the Kerr space-time, is
a task of prolixious complexity. I shall not attempt
to describe how this task has been accomplished; but I
will state how the solution to the two problems I have
enunciated earlier emerges from an integrability con-
dition. The mere statement of the integrability condi-
tion and of the identities which follow from it will
give some indication of the nature of the realm of
mathematical analys.is that one has yet to explore.
First, I must introduce a 'bracket notation' in
which the equations of the theory are most compactly
written. We define various 'brackets' by
24 S. CHANDRASEKHAR

[p]±=P+ 2±P- 2 [S]±=S+2±S_2'

[~p] ±=~·P+2±~OP-2 [..<l'S] ±=~S+2~~S_2' (75)
[~pl ±=~flp ±~ ~p " [ZZS] ±=ff1 ff2 S 2~~.1:·S
J 0 0 +2 0 0 -& + 1 2 -2'

In many ways the central problem in the solution

of the Newman-Penrose equations is the integrability
condition of the equations,

t, 3/2 'Oar -!F=- [{ Q[ZZS] - -2ao [Zs] - coss}r [p] +

-ia{ Q[ZZ~ +cosS+2(QsinS-aocos2S)[~S]+
-6ao [S} +sinScosS J [p] -] , (76)
and
~~ =i [{ r~~Pr -2 (r 2o+K) [!iIp] - +6ro [p] -} [S] +

-ia{ K[~£i)p]--2ro[~pJ +} [S]-COSs] , (77)

where ~ is some function defined in the theory. ~

shall leave it to your imagination as to how one is to
confront the problem presented by these equations. But
the remarkable fact is that the integrability condition
of equations (76) and (77) determines precisely the
real and the imaginary parts of~ and the scale factor.
In fact, we find

where (78)
and

t,2~O=P+2S+2 and ~4 1 P 2S 2' (79)

4(p*)4 - -

where S+2 and S_2 are the normalized proper solutions

of equations (50) and (51) (in which case D1=D2=D in
equations (58) and (59) with D given by equation (63)).
The solution of the integrability condition of
equations (75) and (76) provides at the same time an
explicit evaluation of the following two indefinite
integrals:
BLACK HOLES 25

al =I::.!J(-i [p1- +~ r [pJ +) dr , (80)

~l + r 1 I::. 3/2

and
(81)

where

We find, in fact, two alternative expressions foral

and <~ given by

t (~l +r 1 )al=K[!iI!iIP) +-2ro[!ilP ] + ,

o (Aa,2-Ga 2 )al=i{ rK[!iI!ilP] --2 (r 2o+K) [!ilp) - +Gro [p] -} (83)

and

-t l~1+rl)Y'=Q~S]+cose+2(Qsine-aocos2e)[.!l'S]+
-Gao [s] +sinecose. (84)

I may also note that other equations of the Newman-Pen

rose formalism lead to the following differential equ~
tions for al and Y' :

=2 ( [p] ++i 4AO (85)

~l +r 1
and

~ [d~ (Q dde t) - (4a,~~2 _Q2)Y']

=-2 /[sfcos e + 40(Aa,2-Ga 2 ) [S]-I. (8G)

a (Cl1 +r 1)
26 S. CHANDRASEKHAR

One does not need to be told that the identities ex-

pressed by equations (83) and (84) are astonishing;they
provide a measure of the' mathematical richness of the
theory.
As I said at the outset, it has not been my in-
tention in this lecture to give any coherent account
of what one may properly consider as the content of
the mathematical theory of black holes. I have tried,
rather, to suggest that there are aspects of the theo-
ry which require a deeper understanding (4).

NOTES
(1) Short-range potentials are potentials whose inte-
grals over the range of r* are finite.
(2) The quantities distinguished by a tilde refer to
the time-reversed situation. Thus R(o)=R(-o).
(3) The introduction of this variable r* is crucial to
much of the subsequent developments.
(4) For additional details on the subject matter dealt
with as well as for references to the literature
see the author's recently published book on 'The
Mathematical Theory of Black Holes' (Oxford, Cla-
rendon Press, 1983).