Chapter 3

The Kerr solution

As we have seen, the solution of Einsteins equations describing the
exterior of an isolated, spherically symmetric object (the Schwarzschild
solution) is quite simple. Indeed, it has been found in 1916, immediately after the derivation of Einsteins equation. In the case of
a rotating body, instead, the problem (which is very relevant: astrophysical bodies do rotate) is much more difficult: we dont know
any analytic, exact solution describing the exterior of a rotating star
(even if we know approximate solutions).
But we know the exact solution describing a rotating, stationary,
axially symmetric black hole. It is the Kerr solution, derived in
1963 by R. Kerr. We say that this metric describes a black hole,
because it is a solution of Einstein equations in vacuum (T = 0)
and it has a curvature singularity covered by an horizon: like in the
case of Schwarzschild spacetime, everything falling inside the hole
cannot escape.
We stress that while, thanks to Birkhoff theorem, the Schwarzschild
metric for r > 2M describes the exterior of any spherically symmetric isolated object (a star, a planet, a stone, etc.), the Kerr metric
outside the horizon can only describe the exterior of a black hole.1
1 Actually, there is no proof that it cannot exist a stellar model matching with Kerr metric
at the surface of the star, but such a model has never been found, and it is common belief
that it is unlikely to exist.

39

3.1

The Kerr metric in Boyer-Lindquist coordinates

The explicit form of the metric is the following:

"
! 2
dr
2
2
2
ds = dt +
+ d + (r 2 + a2 ) sin2 d2

2Mr
+
(a sin2 d dt)2

(3.1)

where
(r) r 2 2Mr + a2
(r, ) r 2 + a2 cos2 .

(3.2)

The coordinates (t, r, , ) in which it is expressed the Kerr metric

in (3.1) are called Boyer-Lindquist coordinates.
The Kerr metric depends on two parameters, M and a; comparing (3.1) with the far field limit metric of an isolated object (18.3),
we see that M represents the mass of the black hole, and Ma its
angular momentum, as measured from infinity.
Some properties of the Kerr metric can be directly seen by looking
at the line element (3.1):
It is stationary: it does not depend explicitly on time.
It is axisymmetric: it does not depend explicitly on .
It is not static: it is not invariant for time reversal t t.
It is invariant for simultaneous inversion of t and ,
t t
,

(3.3)

as can be expected: the time reversal of a rotating object produces an object which rotates in the opposite direction.
In the limit r , the Kerr metric (3.1) reduces to Minkowski
metric in polar coordinates; then, the Kerr spacetime is asymptotically flat.

40

 In the limit a 0 (with M %= 0), it reduces to the Schwarzschild

metric: r(r 2M), r 2 then
ds2 (1 2M/r)dt2 + (1 2M/r)1dr 2 + r 2 (d2 + sin2 d2 ) .
(3.4)

In the limit M 0 (with a %= 0), it reduces to

r 2 + a2 cos2 2
dr +(r 2 +a2 cos2 )d2 +(r 2 +a2 ) sin2 d2
r 2 + a2
(3.5)
which is the metric of flat space in spheroidal coordinates:
ds2 = dt2 +

ds2 = dt2 + dx2 + dy 2 + dz 2

where

x =
r 2 + a2 sin cos

r 2 + a2 sin sin
y =
z = r cos .

(3.6)

(3.7)

Indeed,

r
sin cos dr + r 2 + a2 cos cos d r 2 + a2 sin sin d
r 2 + a2

r
sin sin dr + r 2 + a2 cos sin d + r 2 + a2 sin cos d
dy =
r 2 + a2
dz = cos dr r sin d
(3.8)

dx =

thus

"
r2
2
2
dx + dy + dz =
sin + cos dr 2
r 2 + a2
#
$
+ (r 2 + a2 ) cos2 + r 2 sin2 d2 + (r 2 + a2 ) sin2 d2
r 2 + a2 cos2 2
=
dr + (r 2 + a2 cos2 )d2 + (r 2 + a2 ) sin2 d2 .
r 2 + a2
(3.9)
2

The metric (3.1) is singular for = 0 and for = 0. By computing the curvature invariants (like for instance R R )
one finds that they are regular at = 0, and singular at = 0.
Thus = 0 is a true, curvature singularity of the manifold,
whereas (as we will show) = 0 is a coordinate singularity.
41

Notice that in the Schwarzschild limit (a = 0), = r 2 = 0 gives

the curvature singularity, while (for r %= 0) = r(r 2M) = 0
gives the coordinate singularity at the horizon.
The metric has the form

with

gtt
0
=
0
gt

0 gt

0 0

0 0
0 0 g
0

!
"
2Mr
gtt = 1

2Mr
a sin2
gt =

+
,
2Mra2
2
2
2
sin sin2 .
g = r + a +

(3.10)

(3.11)

The g component can be rewritten in a different way, which will

be useful later:
g = (r 2 + a2 ) sin2 +

2Mra2 sin4

.
sin2 - 2
(r + a2 cos2 )(r 2 + a2 ) + 2Mra2 sin2

.
sin2 - 2
(r + a2 )2 (r 2 + a2 )a2 sin2 + 2Mra2 sin2
=

.
sin2 - 2
(r + a2 )2 a2 sin2 .
(3.12)
=

Let us compute the inverse metric. To get g , we only have to

invert the t block in (3.10), while the inversion of the r part is
trivial. The metric in the t block is
!
"
gtt gt
(3.13)
gab =
gt g
and its determinant is
2
g = gtt g gt

42

=
=
=
=

!
,
"+
2Mr
2Mra2
4M 2 r 2 a2
2
2
2
sin sin2
1
sin4
r +a +
2

+
,
2
2Mra
2Mr
sin2 sin2 + (r 2 + a2 )
sin2
r 2 + a2 +

.
2Mr 2 - 2 2
sin a sin + r 2 + a2
(r 2 + a2 ) sin2 +

(3.14)
(r 2 + a2 ) sin2 + 2Mr sin2 = sin2

therefore
gab =
and
g
with
g

tt

g t
g

1
sin2

g tt
0
=
0
g t

!
0

0
0

g gt
gt gtt

g t
0

1
0

0 g

"

0
0

,
+
1 2
2Mra2
2
2
sin
=
r +a +

2Mr
a
=

a2 sin2
=
sin2

(3.15)

(3.16)

(3.17)

where we have used the fact that

2Mr
r 2 + a2 cos2 2Mr
a2 sin2
=
=
.
sin2
sin2
sin2

3.2

(3.18)

Symmetries of the metric

Being stationary and axisymmetric, the Kerr metric admits two

Killing vector fields:
k

(3.19)

or equivalently, in coordinates (t, r, , ),

k (1, 0, 0, 0)

m (0, 0, 0, 1) .
43

(3.20)

In the motion of a particle with four-velocity u , there are then two

conserved quantities:
E u k = ut

L = u m = u .

(3.21)

In the case of massive particles, for which the four-momentum is

P = mu , they are the energy at infinity per mass unit and the angular momentum per mass unit, respectively. In the case of massless
particle, we can choose properly the affine parameter (as we will always do in the following) so that the four-momentum coincides with
the four-velocity: P = u ; thus, for massless particles E is the energy at infinity and L the angular momentum.
It can be shown that k , m are the only Killing vector fields of
the Kerr metric; thus, any Killing vector field is a linear combination
of them.

3.3

Frame dragging and ZAMO

Let us consider an observer, with timelike four-velocity u , which

falls into the black hole with zero angular momentum
L = u = 0 .

(3.22)

This implies that at r , where the metric becomes flat, also

u = 0, and its angular velocity is zero. Such observer is conventionally named ZAMO, which stands for zero angular momentum
observer. The contravariant component of the velocity does not
vanish (except in the limit r ):
u = g t ut %= 0

(3.23)

then the trajectory of the ZAMO has a non-zero angular velocity:

d
=
dt

d
d
dt
d

u
%= 0 .
ut

(3.24)

To compute in terms of the metric (3.1), which is given in covariant form, we use the fact that
u = 0 = g u + gt ut
thus
=

u
gt
=
.
t
u
g
44

(3.25)
(3.26)

We have
gt =

2Mra 2
sin

(3.27)

and, due to (3.12),

g =

.
sin2 - 2
(r + a2 )2 a2 sin2 ,

(3.28)

therefore the angular velocity of a ZAMO is

=

(r 2

2Mar
.
a2 sin2

a2 )2

(3.29)

Notice that
(r 2 + a2 )2 > a2 sin2 (r 2 + a2 2Mr)

(3.30)

thus we always have /(Ma) > 0: the angular velocity has the same
sign as the angular momentum Ma of the black hole, namely, the
motion of the ZAMO is corotating with the black hole.
We can conclude that an observer which approaches a Kerr black
hole with a trajectory which has zero angular velocity at infinity
(and then zero angular momentum) is dragged by the gravitational
field of the black hole, acquiring an angular velocity corotating with
the black hole.

3.4
3.4.1

Horizon structure of the Kerr metric

Removal of the singularity at = 0

To show that = 0 is a coordinate singularity, we make a coordinate transformation that brings the metric into a form which is not
singular at = 0, and then extend the spacetime; such coordinates
are called Kerr coordinates. They are the generalization, to rotating black holes, of the Eddington-Finkelstein coordinates derived in
Schwarzschild spacetime. To begin with, we need to find two families
of null geodesics, one ingoing and one outgoing, and to determine the
corresponding null coordinates (u, v), i.e. the quantities which are
constant in any of these geodesics. In the case of Kerr geometry, the
spacetime cannot be decomposed in a product of two-dimensional
manifolds, thus the study of null geodesics is more complex than in
45

the Schwarzschild case. The Kerr metric admits two special families
of null geodesics, named principal null geodesics, given by
!
" ! 2
"
dx
dt dr d d
r + a2
a

=
, , ,
=
, 1, 0,
, (3.31)
u =
d
d d d d

where the sign plus (minus) corresponds to outgoing (ingoing) geodesics.

In the Schwarzschild limit these are the usual outgoing and ingoing
geodesics u = (1/(1 2M/r), 1, 0, 0), but in the Kerr case they
acquire an angular velocity d/d proportional to a and diverging
when = 0.
We will show explicitly that (3.31) are geodesics later, in the
coordinate frame we are going to define; here we check that they are
null:
g u u = 0 .
(3.32)
We have
dx dx
g
d d

dt
=
d

"2

+(r 2 + a2 ) sin2

"2 ! "2 0
d
dr
+
d
d
"2
!
"2
d
dt
2Mr
2 d

+
.
a sin
d

d d
(3.33)
!

First, we notice that

r 2 + a2 a2 sin2

d
dt
a sin2
=
= .
d
d

(3.34)

Then,
2
(r 2 + a2 )2

2Mr
2
2
2 a
+
(r
+
+
a
)
sin

+
2
2

2
1 (r 2 + a2 )(r 2 + a2 ) + (r 2 + a2 cos2 )(r 2 + a2 2Mr)
=
2
.
(3.35)
+ sin2 a2 (r 2 + a2 ) + (r 2 + a2 cos2 )2Mr = 0

g u u =

and the tangent vector (3.31) is null.

Let us consider the ingoing geodesics, whose tangent vector we
call
"
! 2
a
r + a2

, 1, 0,
;
(3.36)
l =

46

let us parametrize the geodesics in terms of r:

r 2 + a2
d
a
dt
=
= .
(3.37)
dr

dr

We want these geodesics to be coordinate lines of our new system;

thus, one of our coordinates is r, while the others are quantities
which are constant along a geodesic of the family. One of these is
, which is constant along the considered geodesics; the remaining
two coordinates are given by
v t + T (r)
+ (r)

(3.38)

where T (r) and (r) are solutions of2

dT
r 2 + a2
=
dr

a
d
=
dr

so that, along a geodesic of the family,

dv
d
=
0
dr
dr

(3.39)

(3.40)

and the tangent vector of the ingoing principal null geodesics (3.36)
is, in the new coordinates, simply
l = (0, 1, 0, 0) .

(3.41)

We can now compute the metric tensor in the coordinate system

We recall that, in Boyer-Lindquist coordinates,
(v, r, , ).
! 2
"
dr
2Mr
2
2
2
+ d +(r 2 +a2 ) sin2 d2 +
(a sin2 ddt)2 .
ds = dt +

(3.42)
We have
r 2 + a2
r 2 + a2
dr ; dt = dv
dr

a
a
d = d + dr
; d = d dr ,

dv = dt +

(3.43)

2 Notice that (3.39) have an unique solution, with the only arbitrariness of the choice of
because the right-hand sides of (3.39) depend on r only.
the origins of v and ,

47

then
(r 2 + a2 )2 2
r 2 + a2
dvdr
dr
+
2
2

a2
= (r 2 + a2 ) sin2 d2 + (r 2 + a2 ) 2 sin2 dr 2

2
2 a
2
2(r + a ) sin drd ,
(3.44)

dt2 = dv 2
(r 2 + a2 ) sin2 d2

dr 2

+ d2 do not change (r, are also coordinates in the new

frame), the parenthesis in the last term of (3.42) reduces to
r 2 + a2 a2 sin2
dt a sin2 d = dv a sin2 d
dr

(3.45)
= dv a sin2 d dr ,

thus
2Mr
2Mr 2 2Mr 2 4 2
(dt a sin2 d)2 =
dv +
a sin d

4Mr
4Mr
+ 2Mr dr 2 4Mr a sin2 dvd

dvdr +
a sin2 ddr

(3.46)
and, putting all together, we have that the coefficient of dvdr is
2

r 2 + a2 4Mr

= 2,

(3.47)

is
the coefficient of ddr
2(r 2 + a2 )

a
4Mr
sin2 +
a sin2 = 2a sin2 ,

(3.48)

and the coefficient of dr 2 is

2Mr 2
(r 2 + a2 )2 r 2 + a2 2 2

+
a sin +
(r + a2 cos2 )
2
2
2

(r 2 + a2 )(r 2 + a2 2Mr) r 2 + a2 2Mr 2 2

+
a sin
=

2
2
r 2 + a2 a2 sin2

= 0.
(3.49)
=

48

Therefore,
!

"
2Mr
ds = 1
dv 2 + 2dvdr + d2

(r 2 + a2 )2 a2 sin2
+
sin2 d2

4Mra 2
2a sin2 drd
(3.50)
sin dvd .

are called Kerr coordinates. They reThe coordinates (v, r, , )

duce to the Eddington-Finkelstein coordinates for a = 0. In this
frame, it is easy to show that l are tangent vector to geodesics;
indeed, since l = (0, 1, 0, 0),
2

l l; = l l = rr = 0

rr = 0

gr,r = grr, = 0
(3.51)
= 0 and gvr , gr
do not

and this is the case, because in (3.50) grr

depend on r.
In the Kerr coordinates, differently from the the Boyer-Lindquist
coordinates, the metric is not singular at = 0. Thus, after changing coordinates to the Kerr frame, we can extend the manifold, to
include also the submanifold = 0, and removing the corresponding coordinate singularity.
We note, for later use, that being
gvr = 1

grr = gr = 0

2
gr
= a sin ,

(3.52)

we have
l = (1, 0, 0, a sin2 ) .

(3.53)

Notice also that, as we have shown above,

(r 2 + a2 )2 a2 sin2
sin2

2Mr 2 4
a sin
= (r 2 + a2 ) sin2 +

g =

(3.54)

and
.
2Mr
2 = 2Mr dv 2 + a2 sin4 d2 2a sin2 dvd
(dv a sin2 d)

(3.55)
49

therefore the metric in Kerr coordinates can also be written in the

simpler form
ds2 = dv 2 + 2dvdr + d2 + (r 2 + a2 ) sin2 d2 2a sin2 drd
2Mr
2.
+
(3.56)
(dv a sin2 d)

If we want an explicit time coordinate, we can define

t v r

(3.57)

so that the metric (3.56) becomes

ds2 = dt2 + dr 2 + d2 + (r 2 + a2 ) sin2 d2 2a sin2 drd
2Mr
2.
+
(dt + dr a sin2 d)
(3.58)

3.4.2

The horizon

Here we study the submanifold

= r 2 + a2 2Mr = 0 ,

(3.59)

which is a coordinate singularity in Boyer-Lindquist coordinates

! 2
"
dr
2Mr
2
2
2
+ d +(r 2 +a2 ) sin2 d2 +
(a sin2 ddt)2 .
ds = dt +

(3.60)
When a2 > M 2 , the equation = 0 has no real solution. In this case
there is no horizon, and the Kerr solution does not describe a black
hole. In this situation, the singularity = 0 is not covered by any
horizon (naked singularity), and this would bring to paradoxes
in our universe. For this reason, and for the reason that known
astrophysical processes are believed to give rise to black holes with
|a| < M, this situation is generally considered unphysical. Here and
in the following, then, we will restrict our analysis to the case
a2 M 2
(the limiting case a2 = M 2 is called extremal black hole).
We have
(r) = (r r+ )(r r )
50

(3.61)

(3.62)

with

M 2 a2

M M 2 a2

r+ M +

(3.63)

solutions of Eq. (3.60). The surfaces of the coordinate singularity

= 0 are then r = r+ and r = r .
Let us consider now the surfaces r constant = 0, whose
normal is
n = , = (0, 1, 0, 0) .
(3.64)
From (3.64), (3.16)
n n g = g rr =

(3.65)

Thus, on the surfaces r = r+ and r = r , where = 0, n n = 0,

and these surfaces are null hypersurfaces, i.e. horizons. Being r+ >
r , we can say that r = r+ is the outer horizon, and r = r is the
inner horizon. Actually, we should have used the Kerr coordinates
to make this computation, but the result would be the same; indeed,
the surfaces r = const. are the same in the two frames, and the sign
of n n is the same as well. It can be easily checked that the result
it the same by computing n n in Kerr coordinates.
The two horizons separate the spacetime in three regions:
I. The region with r > r+ . Here the r = const. hypersurfaces are
timelike. The r limit, where the metric becomes flat, is
in this region; so we can consider this region the exterior of the
black hole.
II. The region with r < r < r+ . Here the r = const. hypersurfaces are spacelike. An object which falls inside the outer
horizon, can only continue falling to decreasing values of r, until
it reaches the inner horizon and pass to region III.
III. The region with r < r . Here the r = const. hypersurfaces are
timelike. This region contains the singularity, which we will
study in section 3.6.
In the case of extremal black holes, when a2 = M 2 , the two horizons
coincide, and region II disappears.

51

If we consider the outer horizon r+ as a sort of surface of the

black hole, then we could conventionally consider the angular velocity at r = r+ of an observer which falls radially from infinity i.e., an observer with zero angular momentum, or ZAMO - as a sort
of angular velocity of the black hole. The angular velocity of a
ZAMO is given by (3.29):
=

d
2Mar
.
= 2
2
dt
(r + a )2 a2 sin2

(3.66)

At r = r+ , = 0 thus
=

2Mar+
H
2
(r+
+ a2 )2

(3.67)

which is a constant. In this sense, we can say that a black hole

rotates rigidly.
The quantity H = (r+ ) can be expressed in a simpler way. We
have
(3.68)
(r+ M)2 = M 2 a2
therefore

2
+ a2 = 2Mr+
r+

and
H =

3.5

a
a
= 2
.
2Mr+
r+ + a2

(3.69)
(3.70)

The infinite redshift surface and the ergosphere

While in Schwarzschild spacetime the horizon is also the surface

where gtt changes sign, in Kerr spacetime these surfaces do not coincide. We have that
$
1# 2
2Mr
gtt = 1 +
=
r 2Mr + a2 cos2

1
(3.71)
= (r rS+ )(r rS ) = 0

when r = rS+ and when r = rS , where

rS M M 2 a2 cos2 .
52

(3.72)

These surfaces are called infinite redshift surfaces, because if a source

located on a point Pem near the black hole emits a light signal with
frequency em , it will be observed at infinity with frequency
1
gtt (Pem )
obs =
em
(3.73)
gtt (Pobs )
thus if at Pem gtt = 0, obs = 0.
The coefficient of r 2 in (3.71) is negative, so gtt < 0 outside
0 inside that interval. On the other hand, being
[rS , rS+ ], and gtt >

M 2 a2 cos2 M 2 a2 , the horizons, located at

(3.74)
r = M M 2 a2 ,
fall inside the interval [rS , rS+ ]:
rS r < r+ rS+ .

(3.75)

They coincide at = 0, , i.e. on the symmetry axis, while at the

ergosphere
horizon
r+
rS+

Figure 3.1: The ergosphere and the outer horizon

equatorial plan rS+ = 2M and rS = 0.

Therefore, there is a region outside the outer horizon where gtt >
0 3 . This region, i.e.
(3.76)
r+ < r < rS+
is called ergoregion, and its outer boundary r = rS+ is called ergosphere. Notice that, being the ergosphere outside the outer horizon,
3 This

does not happen in Schwarzschild spacetime, where gtt > 0 only inside the horizon

53

an observer from infinity can go inside the ergoregion and come back
to infinity.
In the ergoregion the killing vector k = (1, 0, 0, 0) becomes
spacelike:
(3.77)
k k g = gtt > 0 .
We define a static observer an observer (i.e. a timelike curve) with
tangent vector proportional to k . The coordinates r, , are constant along its worldline, therefore this observer is still in the BoyerLindquist coordinate system (3.60). Such an observer cannot exist
inside the ergosphere, because k is spacelike there; in other words,
an observer inside the ergosphere cannot stay still, but is forced to
move.
A stationary observer is an observer which does not see the metric
change in its motion. Then, its tangent vector must be a killing
vector, i.e. it must be a combination of the two killing vectors of
the Kerr metric, k = /t and m = /:
u =

k + m
= (ut , 0, 0, u ) = ut (1, 0, 0, )
|k + m|

(3.78)

where we have defined the angular velocity of the observer

u
d
= t.
(3.79)
dt
u
In other words, the worldline has constant r and . The observer
can only move along a circle, with angular velocity . Indeed, in
such orbits it does not see the metric change, being the spacetime
axially symmetric.
A stationary observer can exist provided
.
(3.80)
u u g = (ut )2 gtt + 2gt + 2 g = 1

i.e.

2 g + 2gt + gtt < 0 .

(3.81)

To solve (3.81), let us consider the equation

2 g + 2gt + gtt = 0
whose solutions are
=

gt

2
gt
gtt g

54

(3.82)

(3.83)

The discriminant is (using eq.(3.14))

2
gtt g = sin2 .
gt

(3.84)

Thus, a stationary observer cannot exist when < 0, i.e. inside the
horizon r < r < r+ .
Being (see (3.30))
g =

sin2 2
[(r + a2 )2 a2 sin2 ] > 0 ,

(3.85)

the coefficient of 2 in (3.81) is positive, and the inequality (3.81)

is satisfied, outside the outer horizon (where r > r+ , so > 0 and
then < + ), for
< < + .
(3.86)
On the outer horizon r = r+ , = 0 and = + , so (3.81) has
no solution, whereas equation (3.82), corresponding to a stationary
null worldline (for instance, a photon), has one solution only; the
only possible stationary null worldline on the horizon has
=

gt
= H
g

(3.87)

i.e. it has the ZAMO angular velocity. This is another reason why
the angular velocity of the ZAMO at the horizon is considered as the
black hole angular velocity: it is the only possible angular velocity
of a stationary particle on the outer horizon.
On the infinite redshift surface, gtt = 0 so (being gt < 0)
2
2
gt gt
= 0.
(3.88)
=
g
As expected, for r rS+ 0, and = 0 belongs to the interval
(3.86), thus the static observer (which has = 0) is allowed, while
for r < rS+ > 0 and the static observer is not allowed.

3.6

The singularity of the Kerr metric

Let us consider the curvature singularity

= r 2 + a2 cos2 = 0 .
55

(3.89)

If we interpret the Boyer-Lindquist coordinates t, r, , as spherical

polar coordinates, like in Schwarzschild spacetime, it is not clear at
all where is the singularity: = 0 at r = 0, = /2, not at r = 0,
%= /2, but this has no meaning in polar coordinates; we need a
coordinate system which has not the coordinate singularity r = 0,
so that we can distinguish and analyze the curvature singularity.
3.6.1

The Kerr-Schild coordinates

In order to understand the singularity structure, we now change

coordinate frame, to the so-called Kerr-Schild coordinates, which
are well defined in r = 0. Let us start with the metric in Kerr
given in eq. (3.58):
coordinates (t, r, , ),
ds2 = dt2 + dr 2 + d2 + (r 2 + a2 ) sin2 d2 2a sin2 drd
2Mr
2.
(3.90)
+
(dt + dr a sin2 d)

The Kerr-Schild coordinates (t, x, y, z) are defined by

3

a4
x =
r 2 + a2 sin cos + arctan
r
3

a4
2
2

y =
r + a sin sin + arctan
r
z = r cos .
(3.91)
In the next section we will derive the form of the metric in KerrSchild coordinates, showing that the coordinate singularity r = 0
can be removed in this frame; therefore, in this frame we only have
the curvature singularity; to understand the structure of the curvature singularity, then, we must consider it in this frame.
We have
x2 + y 2 = (r 2 + a2 ) sin2
z 2 = r 2 cos2

(3.92)

thus

x2 + y 2 z 2
+ 2 = 1,
(3.93)
r 2 + a2
r
then the surfaces with constant r are ellipsoids (Figure 3.2), and
x2 + y 2
z2
= 1,

2
a2 sin a2 cos2
56

(3.94)

then the surfaces with constant are half-hyperboloids (Figure 3.3).

In Figures 3.2, 3.3 we have represented the r = const, = const
Z

r=0
X

Figure 3.2: r = const ellipsoidal surfaces in the Kerr-Schild frame; the thick
line represents the r = 0 disk.

e=0
Z

e=//4

r=0

e=//2

e=3/4/
e=/

Figure 3.3: = const half-hyperboloidal surfaces in the Kerr-Schild frame; the

thick ring represents the r = 0, = /2 singularity.

surfaces in the Kerr-Schild (t, x, y, z) frame. This means that x, y, z

are represented as Euclidean coordinates, and r, are considered as
functions of x, y, z.
Notice that if we look at Kerr spacetime where r is sufficiently
large, the r, coordinates behave like ordinary polar coordinates.
But closer to the black hole, their nature changes: r = 0 is not a

57

single point but a disk,

x2 + y 2 a2 ,

z=0

(3.95)

and this disk is parametrized by the coordinate . In particular,

(3.96)
r=0
=
2
corresponds to the ring
x2 + y 2 = a2 ,

z = 0.

(3.97)

This is the structure of the singularity of the Kerr metric: it is a

ring singularity. Inside the ring, the metric is perfectly regular.
3.6.2

The metric in Kerr-Schild coordinates

By calling = arctan a/r, we have

r 2 sin2 = a2 cos2

(3.98)

r 2 = (r 2 + a2 ) cos2
a2 = (r 2 + a2 ) sin2

(3.99)

thus

and, rewriting (3.91) as

x = sin r 2 + a2 (cos cos sin sin )

y = sin r 2 + a2 (sin cos + cos sin )

z = r cos

(3.100)

and substituting (3.99) we have

x = sin (r cos a sin )

y = sin (r sin + a cos )

z = r cos .

(3.101)

Differentiating,

sin (r sin + a cos )d

dx = cos (r cos a sin )d
+ sin cos dr

+ sin (r cos a sin )d

dy = cos (r sin + a cos )d
+ sin sin dr
dz = r sin d + cos dr
(3.102)
58

thus

#
$
dx2 + dy 2 + dz 2 = dr 2 + r 2 sin2 + (r 2 + a2 ) cos2 d2
+(r 2 + a2 ) sin2 d2 2 sin2 adrd
= dr 2 + d2 + (r 2 + a2 ) sin2 d2 2a sin2 drd .

(3.103)

Then, the metric (3.90) is the Minkowski metric plus the term
2Mr
2.
(dt + dr a sin2 d)

(3.104)

Being

a2 z 2
,
(3.105)
r2
the factor 2Mr/ is easily expressed in Kerr-Schild coordinates:
= r 2 + a2 cos2 = r 2 +

2Mr
2Mr 3
.
(3.106)
= 4

r + a2 z 2
The one-form dt + dr a sin2 d is more complicate to transform.
We will prove that
r(xdx + ydy) a(xdy ydx) zdz
.
+
dt + dr a sin2 d = dt +
r 2 + a2
r
(3.107)
First of all, let us express the differentials (3.102) as
cos
yd
xd + sin cos dr
sin
cos
+ xd
dy =
yd + sin sin dr
sin
dz = r sin d + cos dr .

dx =

We have

(3.108)

cos 2

(x + y 2 )d + sin (x cos + y sin )dr

sin
= sin cos (r 2 + a2 )d + sin2 rdr
(3.109)

ydx xdy = (x2 + y 2 )d + sin (y cos x sin )dr

= (r 2 + a2 ) sin2 d + sin2 adr
(3.110)
2
2
(3.111)
zdx = r sin cos d + r cos dr
xdx + ydy =

59

then
a
zdz
r
+ (ydx xdy) 2
+
(xdx + ydy) 2
2
2
r +a
r
!
"r + a
2
r
=
r sin cos d + 2
sin2 dr
r + a2
"
!
a2
2
2

sin dr
+ a sin d + 2
r + a2
#
$
+ r sin cos d + cos2 d
(3.112)
= dr a sin2 d

which proves (3.107). The metric in Kerr-Schild coordinates is then

ds2 = dt2 + dx2 + dy 2 + dz 2

,2
+
r(xdx + ydy) a(xdy ydx) zdz
2Mr 3

dt +
+ 4
+
.
r + a2 z 2
r 2 + a2
r
(3.113)
Notice that the metric has the form
g = + Hl l

(3.114)

with

2Mr 3
r 4 + a2 z 2
and, in Kerr-Schild coordinates,
!
"
r(xdx
+
ydy)

a(xdy

ydx)
zdz

l dx = dt +
+
r 2 + a2
r
H

while in Kerr coordinates

#
$
l dx = dt + dr a sin2 d = dv + a sin2 d

(3.115)

(3.116)

(3.117)

thus l is exactly the null vector (3.53), i.e. the generator of the
principal null geodesics which have been used to define the Kerr
coordinates. The form (3.114), called Kerr-Schild form, has been
the starting point for Kerr to derive his solution.
3.6.3

Some strange features of the inner region of the Kerr

metric

If we took seriously the Kerr metric and its ring singularity, we find
some really weird features. We should keep in mind that what we are
60

going to discuss has no direct link with astrophysical observations,

since only the region r > r+ is causally connected to us. Furthermore, we are considering an ethernal black hole, and it is unlikely
that these properties apply also to astrophysical Kerr black holes,
originating (at finite time) from gravitational collapse. It is only for
completeness in our discussion on the Kerr metric that I briefly discuss such features, which, although fascinating, should not be taken
too seriously.
Maximal extension of the Kerr metric

As we have discussed above, the Kerr metric in Kerr-Schild coordinates is

ds2 = dt2 + dx2 + dy 2 + dz 2
+
,2
2Mr 3
zdz
r(xdx
+
ydy)

a(xdy

ydx)
+ 4
+
dt +
r + a2 z 2
r 2 + a2
r
(3.118)
where r is a function of (t, x, y, z), defined implicitly by
r 4 (x2 + y 2 + z 2 a2 )r 2 a2 z 2 = 0 .

(3.119)

This metric is not singular inside the ring, i.e. at r = 0, %= /2,

or, equivalently,
(3.120)
r = 0 , x2 + y 2 < a2 .
It is singular at the ring r = 0, = /2, i.e. r = 0, x2 + y 2 = a2 ;
this ring is a true curvature singularity: indeed the scalar invariant
R R diverges there.
We can then extend the spacetime manifold to r = 0, %= /2,
removing the coordinate singularity at the interior of the ring. As
discussed in the case of Schwarzschild spacetime, we have to extend the spacetime manifold so that the geodesics can be extended
across the ring itself. But this extension cannot simply consist in the
inclusion of the hypersurface corresponding to the coordinate singularity (in this case, the interior of the ring), as we did to remove
the horizon singularity.
To understand this problem, let us consider an observer falling
to the center of the ring through the = 0 axis; along its geodesic,
x = y = 0 and r = z. It arrives at z = r = 0 (which is not
61

e=//6

e=0

e=//6

e=//4

e=0

e=//4

e=//2

e=//2

r=0

e=3/4/

r=0

e=3/4/
e=5//6

e=/

e=5//6

e=/
r<0

r>0

Figure 3.4: Removal of the disc singularity. The top of the disc in the spacetime
with r 0 (left panel) is identified with the bottom of the spacetime with r 0
(right panel), and viceversa. Crossing the ring, an observer can pass from the
r > 0 region to the r < 0 region (and, crossing again, come back to the r > 0
region) avoiding discontinuities in dr/d and in .

a singularity of the spacetime) with = 0 and a finite value of

dz/d = dr/d; then, jumps to and dr/d changes sign. One
could object that we should not worry about r, , since x, y, z, which
are the coordinates in this frame, behave regularly; on the other
hand, if we compute the curvature scalars, we find that they are
discontinuous as we pass through the ring. Therefore, we have not
really removed the coordinate singularity inside the ring.
Let us consider, now, the equation (3.119) for r: it admits two real
solutions for r (there are other two, but they are complex conjugate),
one positive and one negative. Therefore, for each set of Kerr-Schild
coordinates there are two different real values of r, with opposite
signs. We have then two different (asymptotically flat) spacetimes
described by the metric (3.118), one with r > 0 and one with r < 0.
The spacetime with r < 0 has no horizon, as can be easily verified
studying the surfaces r = const..
If we identify the top of the disc x2 + y 2 < a2 , z = 0 in the
spacetime with r > 0 with the bottom of the disc x2 + y 2 < a2 ,
z = 0 in the spacetime with r < 0, and viceversa, as in Figure 3.4,
we have really removed the coordinate singularity of the disc. Our
observer, falling in the disc with r positive but decreasing, emerges
on the top of the disc of the space with r < 0, = 0. The observer,
at this point, can escape to the asymptotically flat limit r .
62

Actually, the maximal extension of Kerr spacetime is even larger.

A detailed study of geodesic completeness would go far beyond these
lectures, we only give the final result. Requiring that all (timelike
or null) geodesics which do not hit the curvature singularity can
be extended, forward and backwards, for an infinite amount of the
affine parameter, one finds that it is necessary to patch together
an infinity of copies of Kerr spacetime, both with r > 0 and with
r < 0. A schematic structure of the maximally extended spacetime
(considered, for simplicity, along the = 0 axis) is shown in Figure
3.5, where the regions I, II, III correspond to:
I
II
III

:
:
:

r+ < r < +
r < r < r+
< r < r

(exterior of the black hole)

(where the r = const. surfaces are spacelike)
(ring singularity and r < 0 space) .

The dashed hyperbolic curves corresponds to r = 0.

II

r=+

r=+

r=+

r=+

II

r=

r=

III

III

r=

r=

II
r=+

r=+

I
r=+

I
II

r=+

Figure 3.5: Schematic representation of the maximal extension of the Kerr

metric, along the = 0 axis. The dashed hyperbolic curves correspond to
r = 0. We denote with I the exterior of the black hole, with II the regions
between the inner and outer horizons, with III the inner regions where the
r = 0 disc is located and the asymptotically flat region with r < 0.

The situation, then, is very different from that of the Schwarzschild

spacetime, where the two asymptotically flat regions are causally
disconnected; in Kerr spacetime an observer which falls inside the
63

inner horizon can either cross the ring, escaping to the asymptotically flat region r < 0, or reach another copy of the region II, and
then another copy of the region I, which is asymptotically flat with
r > 0, and so on. Such copies of the region I are causally connected.
On the other hand, we should remind that this scheme only describes an eternal black hole. In the case of a black hole originating
from a gravitational collapse this multiplication of spacetimes disappears; indeed, our region I cannot receive signals from a region
II, because they should come from t , when the black hole
was not yet born.
Causality violations

To conclude the discussion of the ring singularity of Kerr spacetime,

we show another weird feature of the region close to the singularity.

a r<0
r=0

Figure 3.6: Close timelike curve in Kerr spacetime.

Let us consider a curve consisting in a ring just outside the

singular ring, in the spacetime with r < 0:
5
6

:
t = const., = , 0 2, |r| * a, M, r < 0 . (3.121)
2
The curve belongs to the inner region of the black hole, and can be
reached by an observer that crosses the two horizons, pass through
the r = 0 ring, and goes around it up to the z = 0 plane, just outside
the ring (see Fig. 3.6).
The norm of the tangent vector to this curve is (since = /2
and then = r 2 )
m m g = g = g =

$
1 # 2
2 2
2 2
2
(r
+
a
)

a
(r
+
a

2Mr)
r2
64

1 4 2 2
2Ma2
2
2
2
< 0 (3.122)
(r
+r
a
+2Mra
)
=
r
+a
+
r2
r
since r < 0 and |r| * a, M. Therefore the curve is a timelike
curve, and then can be interpreted as the worldline of an observer
(even if it is not a geodesic), but it is also a closed curve; its existence
is a causality violation: the observer meets itself in its own past.
The occurrence of closed timelike curves (CTC) in some solutions
of Einsteins equations was first found by Kurt Godel, but Godels
solution was considered as unphysical. In the present case, instead,
the CTC appears in a solution in some sense related to a physical
process, i.e. the gravitational collapse.
Actually in a real rotating black hole, born in a gravitational
collapse, the structure of ring singularity (and then the occurrence
of CTCs) could be destroyed by the presence of the fluid and of an
initial time of the singularity, but presently there is no definitive
proof that this is the case. Therefore, while the presence of a collapsing fluid surely eliminates the multiple copies of spacetime in the
maximal extension, it is not clear if it also eliminates the causality
violations inherent to the ring singularity.
A possible point of view could be that of considering the problem
of causality violation, together with the problem of the existence of a
singularity (where some timelike geodesics end, in a finite amount of
proper time), as inconsistencies of the theory of general relativity,
which could disappear once a more fundamental theory (unifying
general relativity with quantum field theory) will take its place.
Indeed, quantum gravity effects are expected to be significant near
the singularities.
In any case, we should not worry about this problem, since these
CTCs occur inside the horizon, and then cannot be observed (at
least, as long as we do not fall into a Kerr black hole); this is a further
motivation for the cosmic censorship conjecture, which then protects
us (and the consistency of the observable universe) not only from
future singularities, but also from causality violations associated to
the ring singularity.
=

3.7

General black hole solutions

In general, we can define a black hole as an asymptotically flat

solution of Einsteins equations in vacuum with an horizon, and a
65

curvature singularity inside the horizon. Black holes form in the

gravitational collapse of stars, if they are sufficiently massive.
When a star has collapsed producing a black hole, we can expect
that, after some time, it settles down to a stationary state as a result
of gravitational waves emission. It is then reasonable to consider
stationary black holes (i.e. black holes admitting a killing vector
field which is timelike at r sufficiently large).
There are some remarkable theorems on stationary black holes,
derived by S. Hawking, W. Israel, B. Carter, which prove the following:
A stationary black hole is axially symmetric.
Any stationary, axially symmetric black hole, without electric
charge, is described by the Kerr solution.
Any stationary, axially symmetric black hole is described by the
so-called Kerr-Newman solution, which is a generalization
of Kerr solution with nonvanishing electric charge and nonvanishing electromagnetic fields, characterized by the mass M, the
angular momentum J, and the charge Q of the black hole.
Furthermore, we remark that any static black hole is spherically
symmetric, and, if it has no electric charge, it is described by the
Schwarzschild solution; in presence of electric charge, it is described
by the Reissner-Nordstrom solution, which is the non-rotating
limit of the Kerr-Newman solution, and is characterized by the mass
M and the charge Q. We have not considered in these lectures the
Reissner-Nordstrom and the Kerr-Newman solutions because it is
widely believed that they are not astrophysically relevant; indeed,
if an astrophysical black hole has an electric charge, it would likely
lose it in a very short timescale, due to the interactions with the
surrounding matter.
We can conclude that a general stationary black hole is characterized by three quantities only: the mass M, the angular momentum
J, and the charge Q. All other features of the star which has collapsed to the black hole are not features of the final black hole. This
result has been summarized with the sentence: A black hole has
no hair. For this reason, the unicity theorems are also called no
hair theorems.

66