BLACK HOLES

IN GENERAL RELATIVITY

Leonardo Gualtieri & Valeria Ferrari

Dipartimento di Fisica, Università degli studi di Roma “Sapienza”

(2011)
Contents

1 The Schwarzschild solution 3

1.1 Line element and general properties . . . . . . . . . . 3
1.2 Horizon . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3 Singularities . . . . . . . . . . . . . . . . . . . . . . . 6
1.3.1 Spacetime extension . . . . . . . . . . . . . . 6
1.3.2 Kruskal and Eddington-Finkelstein coordinates 11
1.3.3 The Finkelstein diagram . . . . . . . . . . . . 13
1.4 Geodesics . . . . . . . . . . . . . . . . . . . . . . . . 15

2 The far field limit of an isolated, stationary object 20

2.1 The case of a weakly gravitating source . . . . . . . . 21
2.2 The case of a general source . . . . . . . . . . . . . . 28
2.3 Mass and angular momentum of an isolated object . 34

3 The Kerr solution 39

3.1 The Kerr metric in Boyer-Lindquist coordinates . . . 40
3.2 Symmetries of the metric . . . . . . . . . . . . . . . . 43
3.3 Frame dragging and ZAMO . . . . . . . . . . . . . . 44
3.4 Horizon structure of the Kerr metric . . . . . . . . . 45
3.4.1 Removal of the singularity at ∆ = 0 . . . . . . 45
3.4.2 The horizon . . . . . . . . . . . . . . . . . . . 50
3.5 The infinite redshift surface and the ergosphere . . . 52
3.6 The singularity of the Kerr metric . . . . . . . . . . . 55
3.6.1 The Kerr-Schild coordinates . . . . . . . . . . 56
3.6.2 The metric in Kerr-Schild coordinates . . . . . 58
3.6.3 Some strange features of the inner region of
the Kerr metric . . . . . . . . . . . . . . . . . 60
3.7 General black hole solutions . . . . . . . . . . . . . . 65

1
4 Geodesics of the Kerr metric 67
4.1 Equatorial geodesics . . . . . . . . . . . . . . . . . . 69
4.1.1 Null geodesics . . . . . . . . . . . . . . . . . . 73
4.1.2 Is E < 0 possible? . . . . . . . . . . . . . . . . 76
4.1.3 The Penrose process . . . . . . . . . . . . . . 79
4.1.4 The innermost stable circular orbit for time-
like geodesics . . . . . . . . . . . . . . . . . . 81
4.1.5 The 3rd Kepler law . . . . . . . . . . . . . . . 82
4.2 General geodesic motion: the Carter constant . . . . 84

5 Black hole thermodynamics 90

5.1 Limits for energy extraction from a Kerr black hole . 90
5.2 The laws of black hole thermodynamics . . . . . . . . 94
5.3 The generalized IInd law of thermodynamics . . . . . 96
5.4 The Hawking radiation . . . . . . . . . . . . . . . . . 101
5.4.1 Quantum fields in Minkowski spacetime . . . 102
5.4.2 Quantum fields in a general spacetime . . . . 106
5.4.3 Particle creation in Schwarzschild spacetime . 109

These notes have been written for a graduate course on black

holes in general relativity; it is assumed that the students have al-
ready followed a basic course on general relativity. The notes are
based on the extensive literature existing on the subject, in partic-
ular on R.W. Wald, General relativity, University of Chicago Press,
Chicago, 1984; S.W. Hawking, G.F.R. Ellis, The large scale struc-
ture of spacetime, Cambridge University Press, Cambridge, 1973;
and on the original articles cited in these books. For more basic
topics on general relativity I refer the reader to C.W. Misner, K.S.
Thorne, J.A. Wheeler, Gravitation, W.H. Freeman and co., S. Fran-
cisco, 1973; S. Weinberg, Gravitation and cosmology, Wiley, New
York, 1972; B.F. Schutz, A first course in general relativity, Cam-
bridge University Press, Cambridge, 1985; H. Stephani, Relativity,
Cambridge University Press, Cambridge, 2004; N. Straumann, Gen-
eral Relativity, Springer Verlag, Berlin, 2004; and on the notes of the
undergraduate course “General Relativity” held by Valeria Ferrari.
I use the geometric units G = c = 1.

2
Chapter 1

The Schwarzschild solution

In this chapter we discuss some features of the Schwarzschild so-

lution; some of them are already covered by fundamental general
relativity courses, and will be only mentioned; some others will be
discussed in greater detail.

1.1 Line element and general properties

The Schwarzschild solution can be written as
! "
2 2M 1
ds = − 1 − dt2 + dr 2 + r 2 (dθ2 + sin2 θdφ2 ) (1.1)
r 1 − 2M
r

in the coordinates (t, r, θ, φ). This solution is asymptotically flat,

since as r → ∞ it tends to the Minkowski metric in polar coordi-
nates.
The spacetime metric (1.1) depends on the positive constant M,
which can be interpreted as the BH mass. A reason for this inter-
pretation can be the following: if a body moves in the Schwarzschild
spacetime, and if r if enough large to assume, with good accuracy,
the weak-field approximation, then the motion of the body is de-
scribed by Newton’s laws, with gravitational potential
M
Φ=− (1.2)
r
which coincides with the gravitational potential of a central body
with mass M. As we will see later on, there are other reasons to
give to M the interpretation of BH mass.

3
 The Schwarzschild solution is spherically symmetric; furthermore,
it is static, i.e., it admits a timelike Killing vector (ξ µ = (1, 0, 0, 0))
hypersurface orthogonal (g0i = 0 if i $= 0). As a consequence, the
Schwarzschild metric is invariant for temporal translations (it is sta-
tionary) and for time reversal t ↔ −t. The Schwarzschild solution
admits four independent Killing vectors: one timelike, and three
spacelike, corresponding to the symmetries of the rotation group
SO(3).
It is possible to show that any asymptotically flat and spheri-
cally symmetric solution of Einstein’s equations in vacuum is also
static, and then it is the Schwarzschild solution. This result, named
the Birkhoff theorem, is very important; it implies, for instance,
that the exterior of spherically symmetric stars is described by the
Schwarzschild solution, and that spherically symmetric objects can-
not be sources of gravitational radiation.

1.2 Horizon
The metric, in these coordinates, is singular at r = 0 and at r = 2M.
However, the singularity r = 2M is only an artifact of the coordinate
choice, and can be removed by changing coordinates (and then it is
called “coordinate singularity”); the singularity r = 0, instead, is a
true singularity of the metric (and is called “curvature singularity”)
1
.
Before discussing the removal of the coordinate singularity, let
us consider in greater detail the surface r = 2M, and the other
hypersurfaces r = constant.
The normal of a hypersurface whose equation is Σ ≡ r−constant =
0 is
nµ = Σ,µ = (0, 1, 0, 0) . (1.3)
Let us consider a generic hypersurface. At any point of such hyper-
surface we can introduce a locally inertial frame, and rotate it in
such a way that the components of the normal vector are
nα = (n0 , n1 , 0, 0) and nα nα = (n1 )2 − (n0 )2 . (1.4)
1 Actually, there are other coordinate singularities: θ = 0, π and φ = 0, 2π; the removal

of such singularities is trivial: it is sufficient to define new angular variables by performing a

rotation; therefore, in the following we will not consider these singularities.

4
Consider a vector tα tangent to the surface at the same point. tα
must be orthogonal to nβ :
t0 n1
nα tα = −n0 t0 + n1 t1 = 0 ⇒ = (1.5)
t1 n0
thus
tα = Λ(n1 , n0 , a, b) with a, b e
Λ constant and arbitrary.
(1.6)
Consequently the norm of the tangent vector is
tα tα = Λ2 [−(n1 )2 + (n0 )2 + (a2 + b2 )] = Λ2 [−nα nα + (a2 + b2 )] . (1.7)
We have that
• If nµ nµ < 0, the hypersurface is called spacelike, and tµ is
necessarily a spacelike vector.
• If nµ nµ > 0, the hypersurface is called timelike, and tµ can be
timelike, spacelike or null.
• If nµ nµ = 0, the hypersurface is called null, and tµ can be
spacelike or null.
Let us consider a point P on a surface Σ = 0. If the surface is
spacelike, the tangent vectors of the surface lie all outside the light-
cone in P . Therefore, a particle passing through P , whose velocity
vector lies inside the cone, can cross the surface only in one direction.
If the surface is null, the situation is nearly the same: the tangent
vectors to the surface lie inside to the light-cone in P , or are tangent
to it, thus a particle can cross the surface in one direction only. It
the surface is timelike, some tangent vectors of the surface are inside
the cone, some others are outside, i.e. the surface cuts the cone,
and then a particle passing through P can cross the surface in both
directions.
From (1.3), (1.1), in the case of an r = constant surface
2M
nµ nν g µν = g rr = 1 − (1.8)
r
thus the surfaces r = constant are spacelike if r < 2M, null if
r = 2M, timelike if r > 2M.
The null hypersurface r = 2M, then, separates regions of space-
time where r = const are timelike hypersurfaces from regions where

5
r = const are spacelike hypersurfaces; therefore, an object crossing
a null hypersurface r = const can never come back; for this rea-
son, the null hypersurface r = 2M is called horizon. As we will see
when studying other BH solutions, this is a general property of null
hypersurfaces. The horizon r = 2M separates the spacetime in:
• the region with r > 2M, where the r = const. hypersurfaces
are timelike; the r → ∞ limit, where the metric becomes flat,
is in this region, so we can consider this region as the exterior
of the black hole;
• the region with r < 2M, where the r = const. hypersurfaces
are spacelike; an object which falls inside the horizon and enter
in this region can only continue falling to decreasing values of
r, until it reaches the curvature singularity r = 0; this region is
then considered the interior of the BH; the name BH is due to
the fact that nothing, neither objects nor signals of any kind,
can escape from this region.

1.3 Singularities
The fact that r = 0 is a curvature singularity can be shown by
computing the curvature scalar
M2
Rαβγδ Rαβγδ = 48 (1.9)
r6
which diverges at r = 0. The fact that r = 2M is a coordinate
singularity is less easy to prove: the finiteness of the polynomials
in the curvature tensor does not exclude, in principle, that there
is a curvature singularity there. Therefore we need to study this
singularity, finding the coordinate change which allows to remove it.

1.3.1 Spacetime extension

To study the structure of singularities, we must first define rigor-
ously the concept of singularity in general relativity. This is not
obvious, since a singularity does not belong to the spacetime man-
ifold; changing coordinate frame, a singularity can be mapped to
infinity, so that at a first look the metric does not appear singular
(there is nothing strange in a singular behaviour at infinity).

6
 The key notion is the length of geodesics. In Schwarzschild space-
time, for instance, an observer falling into the BH reaches the sin-
gularity r = 0 in a finite amount of proper time, thus its (timelike)
geodesic has a finite length and cannot be extended; this means that
r = 0 is a singularity: whatever coordinate frame we choose, that
geodesic has a finite length.
Therefore, to characterize a singular behaviour we need the no-
tion of geodesic completeness: a spacetime is geodesically complete if
every timelike and null geodesic can be extended to values arbitrar-
ily large of the affine parameter. If, instead, the spacetime admits
at least one incomplete (i.e., which cannot be extended) timelike
or null geodesic, we say that it is geodesically incomplete, and this
means that there is a singularity (either a true curvature singular-
ity or a coordinate singularity). We only consider timelike or null
geodesics because they represent worldlines of massive or massless
particles, and then they represent observers or signals, while space-
like geodesics do not correspond to worldlines of physical objects.
Let us consider the Schwarzschild metric in the coordinates (t, r, θ, φ),
with line element (1.1)
! "
2 2M 1
ds = − 1 − dt2 + dr 2 + r 2 (dθ2 + sin2 θdφ2 ) . (1.10)
r 1 − 2M
r

To be rigorous, the spacetime manifold described by (1.10) is not

defined at r = 0 and at r = 2M. We have then two disconnected
manifolds, M1 with 0 < r < 2M (the interior of the BH) and M2
with r > 2M (the exterior of the BH). A geodesic corresponding
to an observer falling into the BH cannot be extended across the
hypersurface r = 2M, since it does not belong to M1 ∪ M2 , and the
geodesic terminates at a finite value of the affine parameter (notice
that the observer arrives at r = 2M with t = +∞). On the other
hand, since this singularity is not a true curvature singularity, it can
be removed with the following two steps:
• we change the coordinate frame, for instance to the Kruskal
frame (U, V, θ, φ), for which (as we shall show in Section 1.3.2)
32M 3 − r
ds2 = − e 2M dUdV + r 2 (dθ2 + sin2 θdφ2 ) (1.11)
r
(note that the r in (1.11) should be considered as a function of
the coordinates, r(U, V )).

7
 U V
r=0

r < 2M

r > 2M

Figure 1.1: Interior and exterior of a Schwarzschild black hole in Kruskal coor-
dinates.

In these coordinates (see Fig. 1.1), the BH exterior r > 2M

corresponds to (U < 0, V > 0) and the BH interior r < 2M
corresponds to (U > 0, V > 0 with UV < 1); the singularity
r = 2M (and t = +∞) corresponds to the semiaxis (U = 0,
V > 0); the curvature singularity r = 0 corresponds to the
upper branch of the hyperbole UV = 1.
• In the new frame the manifold can be extended across the semi-
axis (U = 0, V > 0), separating M1 and M2 , since the line
element (1.11) is not singular there; this means that we consider
a new manifold,
M ⊃ M1 ∪ M2 , (1.12)
defined by
V > 0, UV < 1 . (1.13)
Therefore, in order to eliminate a coordinate singularity we have to
extend the spacetime manifold.
Notice that when we think of a Schwarzschild BH, we are im-
plicitly considering the extended manifold M: indeed, we assume
that an object falling inside the BH crosses the horizon r = 2M.
Even the discussion of the previous section about the hypersurfaces
r = constant (and, among them, r = 2M), assumes that the man-
ifold is M; we stress that the manifold M is not covered by the

8
 U V
r=0

2M
r=
II

+
t=
IV I

III
t=
ï

r=
2M
r=0

Figure 1.2: Maximal extension of Schwarzschild spacetime in Kruskal coordi-

nates. The dashed line represents the worldline of an observer falling into the
black hole.

coordinates (t, r, θ, φ) and the line element (1.10).

The manifold M should still be extended: timelike and null
geodesics from V = 0 cannot be extended backwards, unless we
extend the manifold to V ≤ 0. By considering (−∞ < U < +∞,
−∞ < V < +∞ with UV < 1) we have the maximal extension
(i.e., which cannot be further extended) of the Schwarzschild space-
time. This is the usual Kruskal construction, shown in Fig. 1.2; the
dashed line represents the worldline of an observer falling into the
black hole, and the wave-like curves represent the curvature singu-
larity r = 0.
In the Kruskal coordinates (1.11) the null worldlines with θ, φ
constant, are straight lines at 45o , i.e. U = const. and V = const.;
this can be seen easily from the line element (1.11): assuming θ, φ
constant, if either dU = 0 or dV = 0, then ds = 0 and the infinites-
imal tangent vector is null (to say in different words: any worldline
with tangent vector either (1, 0, 0, 0) or (0, 1, 0, 0), is null). There-

9
fore, the light-cone can be drawn as in Minkowski spacetime, and
then it is easy to describe the causal connections among events (see
for instance, in Fig. 1.2, the dashed line representing the possi-
ble worldline of a massive particle falling inside the black hole). In
particular, we see that Sector I can send signals (and matter and
energy) to sector II only, and receive signals from sector III; fur-
thermore, there is another copy of sector I, i.e. sector IV, which is
causally disconnected from I, but can receive signals from III and
send signals to II.
The incomplete (timelike and null) geodesics of the maximally
extended manifold correspond to a true singularity: in this case, to
r = 0, i.e., in Kruskal coordinates, to UV = 1. There are geodesics
reaching the singularity with finite affine parameter, and cannot be
extended through it; for instance, an observer that falls inside the
BH reaches the singularity in a finite amount of proper time.
It should be stressed that the maximal extension of Schwarzschild
spacetime has no meaning if we consider a BH as an astrophysical
object. Therefore, we do not have to worry about the meaning of
the sector IV, and leave the discussion of other universes to science-
fiction writers. Indeed, the above construction describes an eternal
black hole, whereas astrophysical BHs result from stellar collapse,
which happen at finite values of t. In particular, the region III
cannot exist for an astrophysical BH, because the semiaxis (U < 0,
V = 0) corresponds to t = −∞.
A comment on the r = 0 singularity. The fact that we cannot
say what happens to the observer as it reaches the singularity, con-
stitutes a problem for the theory; on the other hand, such problem
is not severe from an operational point of view, since no signal from
the observer reaching the singularity can be sent outside the black
hole: the consistency of the theory, in a certain sense, is preserved
by the existence of the horizon. Roger Penrose has conjectured that
there is a fundamental principle, the cosmic censorship hypothesis,
stating that all singularities in the universe (with the exception of a
possible initial singularity) are covered by horizons. In other words,
there is no naked singularity. There is no definitive proof of this
conjecture, but there are indications supporting it. Therefore, it is
commonly believed that the cosmic censorship hypothesis is likely
to be correct; it is worth noticing that many mathematical proper-
ties of spacetime would be challenged by the existence of a naked

10
singularity.

1.3.2 Kruskal and Eddington-Finkelstein coordinates

The Kruskal coordinates are defined as follows. Let us consider the
r > 2M manifold, and define
u ≡ t − r∗ , v ≡ t + r∗ (1.14)
with # r $
r∗ ≡ r + 2M ln−1 (1.15)
2M
tortoise coordinate, which tends to −∞ as r → 2M. Notice that
r > 2M corresponds to
−∞ < u < +∞ − ∞ < v < +∞ . (1.16)
and the limit r → 2M, t → +∞ corresponds to u → +∞, with
finite v. Since
dr∗ 1
= , (1.17)
dr 1 − 2M
r
the metric is
! "
2 2M
ds = − 1− (dt2 − dr∗2) + r 2 (dθ2 + sin2 θdφ2 )
r
! "
2M
= − 1− dudv + r 2 (dθ2 + sin2 θdφ2 ) . (1.18)
r
We have
r∗ − r # r $ 2M 2M r∗ −r 2M −r v−u
= ln −1 ⇒ 1− = e 2M = e 2M e 4M .
2M 2M r r r
(1.19)
Then, defining u v
U ≡ −e− 4M , V ≡ e 4M (1.20)
we have the Kruskal metric:
32M 3 − r
ds2 = − e 2M dUdV + r 2 (dθ2 + sin2 θdφ2 ) (1.21)
r
with U < 0, V > 0.
Let us now consider the manifold 0 < r < 2M, and define u, v as
in (1.14),
u ≡ t − r∗ , v ≡ t + r∗ (1.22)

11
but with a new tortoise coordinate
# r $
r∗ ≡ r + 2M ln 1 − (1.23)
2M
which is always negative, tends to zero as r → 0, and to −∞ as
r → 2M. Differentiating this new tortoise coordinate we get the
same expression as (1.17),
dr∗ 1
= , (1.24)
dr 1 − 2M
r

then the metric in u, v is still given by

! "
2 2M
ds = − 1 − dudv + r 2 (dθ2 + sin2 θdφ2 ) (1.25)
r
but
2M 2M −r v−u
1− =− e 2M e 4M (1.26)
r r
thus, defining u v
U ≡ +e− 4M , V ≡ e 4M , (1.27)
we have the same metric of eq. (1.21),
32M 3 − r
ds2 = − e 2M dUdV + r 2 (dθ2 + sin2 θdφ2 ) , (1.28)
r
but with U > 0, V > 0.
This metric, with V > 0 and extended to −∞ < U < +∞,
describes then the exterior and the interior of the BH, as anticipated.
Actually, there is a simpler coordinate system that covers the
region I and II: the Eddington-Finkelstein (EF) coordinates

(v, r, θ, φ) − ∞ < v < +∞ 0 < r < +∞ . (1.29)

The two definitions (1.15), (1.23) can be put together as
% r %
% %
r∗ ≡ r + 2M ln % − 1% (1.30)
2M
and v = t + r∗ . Then, since
dr∗ dvdr
dt2 −dr∗2 = dv 2 −2dvdr∗ = dv 2 −2 dvdr = dv 2 −2 , (1.31)
dr 1 − 2M
r

12
the metric in the EF coordinates is
! "
2 2M
ds = − 1 − dv 2 + 2dvdr + r 2 (dθ2 + sin2 θdφ2 ) . (1.32)
r
This metric covers both the interior and the exterior of the BH, i.e.
the sectors I and II of the Kruskal construction, and is not singular
at the horizon, which is simply r = 2M. Notice that on the hori-
zon v is finite because t → +∞ and r∗ → −∞, while u is instead
divergent. All the computations and derivations involving the inte-
rior and the exterior of the BH, like for instance the study of the
r = constant surfaces of Section 1.2, and the study of what happens
to a spaceship falling inside the BH, can be rigorously performed in
the EF coordinates.
In principle, one could also build the maximal extension of the
Schwarzschild metric by patching together the EF chart, which cov-
ers sectors I,II, with other three (similar) charts, one covering sectors
II,III, one covering sectors III,IV, and one covering sectors IV,II;
fortunately, we do not need to discuss this complicate construction,
since there exists a single coordinate frame, the Kruskal frame, which
covers all the four sectors of the maximal extension. But for other
BH solutions there is no equivalent to the Kruskal coordinates, and
the only way to build the maximal extension is to patch together
different coordinate frames (analogue to the EF frame), which elim-
inate the coordinate singularities but do not provide, alone, the
maximal extension.

1.3.3 The Finkelstein diagram

A useful way to visualize the Schwarzschild spacetime is the Finkel-
stein diagram, in which the axes are (t̃, r) where
% r %
% %
t̃ ≡ v − r = t + 2M ln % − 1% . (1.33)
2M
In the Finkelstein diagram the null lines v = const., correspond-
ing to ingoing massless particles, are straight lines at 45o; the null
lines u = const., corresponding to outgoing massless particles, are
hyperbolic curves. These two sets of curves define the light cones
centered in any point of the spacetime, and then allow to establish
the causal relationship among different events. Differently from the
(t, r) diagram, the light-cones behave regularly at the horizon.

13
 ~
t r = 2M

u=
con st.
v
=
co

st.

t.
n

ns
co
t=

st.
n
co
=
u
r=0

Figure 1.3: Eddington-Finkelstein diagram of a Schwarzschild black hole.

Since t̃ + t for r , 2M and r - 2M, the coordinate t̃ coincides

with t far away from the horizon, but they are very different close to
the horizon (where, anyway, the coordinate t has not a clear physical
meaning). Still, inside the horizon t̃ cannot be considered a “time”,
because the vector ∂/∂ t̃ is spacelike; this is clear from Fig. 1.3,
where ∂/∂ t̃ (directed vertically) falls outside the null cones in the
interior of the BH.
Notice that the r = constant lines are vertical lines, while the
t = constant lines are hyperbolic curves (dashed lines in Fig. 1.3);
the t̃-axis represents the singularity, and for this reason it is drawn
wave-like.
As we mentioned above, astrophysical BHs are the product of
gravitational collapse, and then are not eternal black holes. A qual-
itative view of the spacetime of a realistic black hole is shown in the
Finkelstein diagram in Fig. 1.4, where the shadowed area represents
the interior of the star. The r = 0 axis is a curvature singularity for
t̃ ≥ t̃0 , i.e. after the singularity formation; t̃ < t̃0 , the r = 0 axis
is simply the (trivial) coordinate singularity at the origin of polar
coordinates. Also the horizon, represented by the dashed line, is
formed during the collapse.
It is important to stress that, although we have discussed the
entire Schwarzschild solution, only the r > 2M region is directly

14
 ~
t r = 2M

v
=
co
ns
t.
~
t0

st.
r=0

n
co
u= r

fluid

Figure 1.4: Eddington-Finkelstein diagram of a stellar collapse originating a

Schwarzschild black hole. The shadowed area represents the fluid interior of the
star. The curvature singularity (wave-like line) is formed at t̃ = t̃0 . The horizon
is represented by the dashed line.

relevant for astrophysical observations: no signal can come from the

interior of the BH. Therefore, the most physically relevant properties
of black holes are the properties of the exterior region. On the
other hand, it is impossible to have a general understanding of the
physics of black holes (and then of the behaviour of astrophysical
black holes) without having at least a general idea of what’s going
on inside; for this reason we have briefly discussed the features of
the entire solution.

1.4 Geodesics
We remind that the geodesic equations are equivalent to the Euler-
Lagrange equations with Lagrangian L = 1/2gµν ẋµ ẋν where a dot
indicates differentiation with respect to the affine parameter λ (we
call the four-velocity of the particle ẋµ or, equivalently, uµ ). For the
Schwarzschild metric, the Lagrangian is
& ! " )
1 2M 2 ṙ 2
L= − 1− ṫ + ' ( + r 2 θ̇2 + r 2 sin2 θφ̇2 . (1.34)
2 r 1 − 2M r

15
The equations of motion for ṫ, φ̇ and θ̇ are:
• Equation for ṫ:
*! " +
∂L d ∂L d 2M
− =0 → 1− 2ṫ = 0 (1.35)
∂t dλ ∂(ṫ) dλ r
i.e.
const E
ṫ = ' 2M
(=' (. (1.36)
1− r 1 − 2M r
In the case of a massive particle, the constant E can be inter-
preted as the energy of the particle per unit mass at infinity,
since at r → ∞, where the spacetime becomes flat and special
relativity applies, the energy of the particle is P 0 = mu0 = mE.
It should be reminded that, since the Schwarzschild metric ad-
∂ α
mits a timelike Killing vector ∂t → ξ(t) = (1, 0, 0, 0), then
! "
α β 2M
gαβ ξ(t) u = const → − 1 − ṫ = const = −E (1.37)
r
thus E is the constant of motion associated with invariance for
time translations.
• Equation for φ̇: in a similar way it is easy to show that
const L
φ̇ = 2 = 2 2 . (1.38)
r2sin θ r sin θ
∂ α
Since there is a spacelike Killing vector ∂φ
→ ξ(φ) = (0, 0, 0, 1),
such that
α
gαβ ξ(φ) uβ = const → r 2 sin2 θφ̇ = const → L = const,
(1.39)
thus L is the constant of motion associated with invariance for
φ-rotations. It can be interpreted as the angular momentum of
the particle per unit mass.
• Equation for θ̇:
∂L d ∂L d 2
− =0 → (r θ̇) = r 2 sin θ cos θφ̇2 . (1.40)
∂θ dλ ∂(θ̇) dλ
Therefore the equation for θ is
2
θ̈ = − ṙ θ̇ + sin θ cos θφ̇2 . (1.41)
r

16
 Given the spherical symmetry, polar axes can be chosen in ar-
bitrary way. We choose them such that, for a given value of the
affine parameter, say λ = 0, the particle is on the equatorial
plane θ = π2 and its three-velocity (ṙ, θ̇, φ̇) lays on the same
plane, i.e. θ̇(λ = 0) = 0. Thus, we have to solve the following
Cauchy problem
2
θ̈ = − ṙ θ̇ + sin θ cos θφ̇2
r
θ̇(λ = 0) = 0
π
θ(λ = 0) = (1.42)
2
which admits an unique solution. Since
π
θ(λ) ≡ (1.43)
2
satisfies the differential equation and the initial conditions, it
must be the solution. Thus the orbit is plane and to hereafter
we shall assume θ = π2 and θ̇ = 0.
• Equation for ṙ: it is convenient to derive this equation from
the condition uα uα = −1, or uα uα = 0, respectively valid for
massive and massless particles.
A) massive particles:
! "
α β 2M 2 ṙ 2
gαβ u u = − 1 − ṫ + ' ( +r 2 θ̇2 +r 2 sin2 θφ̇2 = −1
r 1 − 2M r
(1.44)
which becomes, by substituting the equations for ṫ and φ̇
! "! "
2M L2
ṙ 2 + 1 − 1 + 2 = E2 (1.45)
r r
B) massless particles:
! "
α β 2m 2 ṙ 2
gαβ u u = − 1 − ṫ + ' ( + r 2 θ̇2 + r 2 sin2 θφ̇2 = 0
r 1 − 2M r
(1.46)
which becomes
! "
2 L2 2M
ṙ + 2 1 − = E2 . (1.47)
r r

17
Summarizing, the geodesic equations are:
π
θ =
2
E
ṫ = ' (
1 − 2M
r
L
φ̇ = 2
r
ṙ 2 = E 2 − V (r) (1.48)
with ! "! "
2M L2
V (r) = 1 − 1+ 2 (1.49)
r r
for massive particles and
! "
L2 2M
V (r) = 2 1 − (1.50)
r r
for massless particles; it is worth noticing that the potential of mass-
less particles depends on L as a multiplication constant, and then
the location of maxima and minima of V (and, more generally, the
shape of V ) does not depend on L, while for massive particles the
dependence on L is non trivial.
We also point out that for massive particles (for which uµ uµ = −1
and the affine parameter is the proper time) E, L are, as we said
above, the energy (at infinity) and the angular momentum per unit
mass, but this is not the case for massless particles. If the particle
is massless, there is an arbitrariness in the definition of the affine
parameter (λ → aλ preserve the normalization uµ uµ = 0), and E, L
scale with λ; then, we can normalize λ so that E, L are the energy
(at infinity) and the angular momentum of the massless particle.
The radial equations
ṙ 2 = E 2 − V (r) (1.51)
allow a qualitative study of the motion analogue to the study per-
formed in Newtonian mechanics. Some results of this qualitative
study are the following:
• geodesics of massless particles describe either open orbits, cor-
responding to deflection of the ingoing particle, or unstable
circular orbits with r = 3M;

18
 • geodesic of massive particles describe either open orbits, cor-
responding to deflection of the ingoing particle, or closed or-
bits, with r ranging between a minimum and a maximum value
(these are not really closed curves due to periastron advance);
• circular orbits of massive particles are allowed only if r ≥ 3M;
if 3M ≤ r < 6M they are unstable orbits; if r ≥ 6M they
are stable orbits; therefore, the innermost stable circular orbit
(ISCO) has r = 6M.
It is important to stress that the classical tests of general rela-
tivity are based on the properties of Schwarzschild’s solution: the
gravitational redshift, the perihelion advance, the deflection of light
signals. But, as we will discuss later on, astrophysical observations
are starting to test more complex spacetimes, like the Kerr metric
which describes a rotating BH.

19
Chapter 2

The far field limit of an

isolated, stationary object

In this chapter we study the gravitational field generated by an

isolated, stationary object.
We require that the source is an isolated object, thus we are
assuming that the spacetime outside the source is vacuum (Tµν = 0).
Therefore, it is reasonable to assume that the spacetime, far away
from the source, tends to the Minkowski spacetime; this condition
is called asymptotic flatness.
If the spacetime is asymptotically flat, we can define, in an ap-
propriate coordinate frame, a space coordinate r such that1
lim gµν = ηµν . (2.1)
r→∞

We call far field limit the region of spacetime in which r - R, where

R is a lengthscale of the source. In the far field limit,
! "
1
gµν = ηµν + O . (2.2)
r
We also assume that the source is stationary, i.e. it does not depend
on time. In this case, it can be shown that Einstein’s equations imply
that the spacetime is also stationary.
We want to determine the metric in the far field limit of an iso-
lated, stationary object. In other words, we want to determine the
1 Actually a complete, coordinate independent definition of asymptotic flatness would re-

quire the introduction of the concept of conformal infinity, which is beyond the scope of these
lectures. However, for the derivation of the far field limit of isolated stationary objects the
definition (2.1) is sufficient.

20
lowest order terms in 1/r in the expansion (2.2). We will show that
the spacetime metric in the far field limit is
! " ! "
2 2M 2 2M
ds = − 1 − dt + 1 + dr 2
r r
4J
+r 2 (dθ2 + sin2 θdφ2 ) − sin2 θdtdφ
r
+ higher order terms in 1/r . (2.3)
with M mass and J angular momentum of the central object.
Far away from the source the expansion (2.2) applies; it can also
be written in the form
gµν = ηµν + hµν (2.4)
with |hµν | , 1. The perturbation hµν is a solution of the equations
of linearized gravity gµν = ηµν + hµν ,
!h̄µν = 0 (2.5)
h̄µν,µ = 0 (2.6)
where we have defined
1
h̄µν ≡ hµν − ηµν hαα (2.7)
2
and we neglect terms O(|h|2) (which are also higher order terms in
the 1/r expansion). We are assuming that the spacetime is station-
ary, therefore (2.5), (2.6) become
∇2 h̄µν = 0 (2.8)
h̄i ν,i = 0 . (2.9)
We stress that (2.8), (2.9) hold only in the far field limit r - R; we
are not making any assumption on the nature of the source.
Nevertheless, to have more insight on the physical meaning of
(2.3), we will start deriving it in the particular case where the grav-
itational field on the source is weak. Subsequently, we will consider
the general case.

2.1 The case of a weakly gravitating source

If the gravitational field on the source is weak, it is possible to lin-
earize the metric on the source, and the linearized Einstein’s equa-

21
tions give
!h̄µν = −16πTµν
h̄µν,µ = 0 , (2.10)

whose solution is
,
Tµν (t − |x-x% |, x% ) 3 %
h̄µν (t, x) = 4 dx (2.11)
V |x-x% |
where V is the three-volume of the source. We choose V such that
its boundary falls outside the source, i.e. Tµν = 0 on the boundary
∂V .
In the standard derivation of the quadrupole formula for gravi-
tational radiation, one considers the oscillating part of the integral
(2.11), neglecting the terms constant in time. Here, on the contrary,
we are considering a stationary source Tµν = Tµν (x% ); the integral
(2.11) becomes ,
Tµν (x% ) 3 %
h̄µν (x) = 4 %
dx . (2.12)
V |x − x |
We denote by Latin letters i, j the space indexes 1, 2, 3; notice that,
in linearized gravity, the indexes of hµν are raised by using the
Minkowski metric, thus hiµ = hi µ .
Let us consider the Taylor expansion of 1/|x−x% |; x is the position
vector of the point where we compute the gravitational perturbation
hµν (with radial coordinate r = |x|); x% is the position vector of a
generic point inside the source; then, |x% | ≤ R , r = |x|. Such an
expansion is commonly named multipolar expansion; we have
- i %i ! "
1 1 ix x 1
= + + O . (2.13)
|x − x% | r r3 r3
Substituting in (2.12) we find
, - ,
4 3 % 4 i xi
h̄µν (x) = Tµν d x + Tµν x%i d3 x% . (2.14)
r V r3 V

The first integral in (2.14) gives

,
Tµν d3 x% . (2.15)
V

22
Being T00 , in the limit of weak gravitational field, the mass-energy
density of the system, by definition
,
T00 d3 x% = M . (2.16)
V

To compute the components µi of the integral (2.15), we first notice

that the stress-energy tensor conservation, for a stationary source,
reduces to
T µν,ν = T µ0,0 + T µi,i = T µi,i = 0 . (2.17)
∂x i
i
Using (2.17), and the property ∂x j = δj , we find

, , ,
µi 3 % µk i 3 % ∂x%i
T d x = T δk d x = T µk %k d3 x%
V V ∂x
, ! µk " V
∂T
= − %k
x%i d3 x% = 0 (2.18)
V ∂x
where we have integrated by parts; the surface terms do not con-
tribute, because, on the boundary of V , Tµν = 0. Therefore, the
components µi of the integral (2.15) vanish.
Let us now compute the second integral in (2.14),
,
Tµν x%i d3 x% . (2.19)
V

The 00 component gives the position of the center of mass, which

can be set to zero by a proper choice of the origin of the coordinates:
,
T00 x%i d3 x% = Mx%icdm = 0 . (2.20)
V

To compute the µi components of the integral (2.19), we start by

proving that it is antisymmetric in the last two indexes:
, ,
µi %j 3 %
T x d x =− T µj x%i d3 x% . (2.21)
V V

Indeed,
, , ! "
' µi %j µj %i
( 3 % µk ∂x%i %j ∂x%j %i 3 %
T x +T x Tdx = x + %k x d x
V V ∂x%k ∂x
, ,
µk ∂
' %i %j ( 3 %
= T x x d x =− x%i x%j T µk,k d3 x% = 0 (2.22)
V ∂x%k V

23
where we have integrated by parts, using the fact that, on the bound-
ary of V , Tµν = 0.
Let us consider now the components µν = jk of the integral
(2.19) ,
T jk x%i d3 x% . (2.23)
V
It is symmetric in the first two indexes, but, from (2.21), it follows
that it is also antisymmetric in the last two indexes. Therefore, it
is easy to show that it is the opposite of itself, and then it vanishes:
, , ,
ki %j 3 % kj %i 3 %
T x dx = − T x d x =− T jk x%i d3 x%
V , V , V
ji %k 3 %
= T x d x = T ij x%k d3 x% (2.24)
V, V

= − T ik x%j d3 x% = 0 . (2.25)
V

The last integral we have to compute is

, ,
0i %j 3 %
T x d x =− T 0j x%i d3 x% . (2.26)
V V

To express it differently we introduce the tensor *ijk , which is anti-

symmetric in any exchange of its indexes, and has *123 = 1. Then,
if we define ,
Ji ≡ − *ijk T 0j x%k d3 x% , (2.27)
V
we can write ,
1
T 0i x%j d3 x% = − *ijk J k (2.28)
V 2
It is easy to prove that (2.27) implies (2.28), using the very useful
property2
*ijk *klm = δ il δ jm − δ im δ jl . (2.29)
Then, given an antisymmetric tensor B ij = −B ji , we have
1
J i = *ijk B jk ⇐⇒ B ij = *ijk J k : (2.30)
2
2 The prove of (2.29) is the following. *ijk &= 0 only if its three indexes are all different,

thus i &= k and j &= k; the same for *lmk . Therefore *ijk *lmk &= 0 only if the indexes ij and lm
are the same, with any ordering. If they have the same order, i.e. ij = lm, then *ijk *lmk = 1,
while if they have the opposite order, i.e. ij = ml, then *ijk *lmk = −1. Thus (2.29).

24
 1 ijk k 1 ijk klm lm 1 ' il jm (
* J = * * B = δ δ − δ im δ jl B lm = B ij . (2.31)
2 2 2
Finally, replacing ,
jk
B =− T 0j x%k d3 x% , (2.32)
V
we find that (2.27) implies (2.28).
The physical meaning of J k is simple. The tensor T 0i represents
the momentum density of the source,
T 0i = P i (2.33)
i.e. an element of the matter source in the volume d3 x% has momen-
tum Pd3 x% . Writing (2.27) as a vector product,
,
J = x% × Pd3 x% (2.34)

since (u × v)i = *ijk uj v k .

We remind that the angular momentum of a point particle is the
vector product between its momentum and its position vector. The
angular momentum of an element of matter in the source, which has
momentum Pd3 x% and position x% , is then x% × Pd3 x% . Therefore, J
is the total angular momentum of the source.
Summarizing, the multipolar expansion (2.14) gives
! "
4M 1
h̄00 = +O
r r3
! "
2 ijk j k 1
h̄0i = 3 * x J + O
r r3
! "
1
h̄ij = O (2.35)
r3
because
,
T00 d3 x% = M
V
, ,
4 j %j 3 % 4 j 2
3
x T 0i x d x = − 3
x T 0i x%j d3 x% = 3 *ijk xj J k .
r V r V r
(2.36)
Notice that in flat space h̄00 = h̄00 , h̄0i = −h̄0i , h̄ij = h̄ij .

25
 3
In terms of hµν , given by
1
hµν = h̄µν − ηµν h̄αα , (2.37)
2
we have4
! "
2M 1
h00 = +O
r r3
! "
2 j k 1
h0i = 3 *ijk x J + O
r r3
! "
2M 1
hij = δij + O . (2.38)
r r3
Let us transform the solution (2.38) in polar coordinates
x1 = r sin θ cos φ
x2 = r sin θ sin φ
x3 = r cos θ . (2.39)
We have .
(dxi )2 = dr 2 + r 2 dθ2 + r 2 sin2 θdφ2 , (2.40)
i

thus
2M ' 2 (
hij dxi dxj = dr + r 2 dθ2 + r 2 sin2 θdφ2 . (2.41)
r
The transformation of h0i dx0 dxi is less trivial. If we choose the
orientation of our polar coordinates such that
J = (0, 0, J) , (2.42)
then
! " ! "
0 i 2 0 2 0
h0i dx dx = dx *ijk x J dx = − 3 dx J(x1 dx2 − x2 dx1 )
j k i
r3 r
! "
2 2J
= − 3 dx0 Jr 2 sin2 θdφ = − sin2 θdtdφ (2.43)
r r
3 To invert (2.7) we take the trace of (2.7), finding h̄λ = −hλ , and replace this relation
λ λ
into (2.7). ! "
4 Notice that xj is an O 1
term, because in the far field limit xj ∼ r.
r3 r2

26
where the property x1 dx2 − x2 dx1 = r 2 sin2 θdφ can be proven by
differentiation of (2.39).
The corresponding line element is
! ! ""
2 2M 1
ds = − 1 − +O dt2
r r3
! ! ""
2M 1 / 2 2 2 2 2
0
+ 1+ +O dr + r (dθ + sin θdφ )
r r3
! ! ""
4J 2 1
+ − sin θ + O dtdφ . (2.44)
r r2
This is the solution of linearized Einstein’s equations. If we consider
the complete Einstein’s equations
Rµν = 0 (2.45)
we have terms O(|hµν |2 ), which produce terms quadratic in M and
J in the metric (and also higher order terms). Therefore, there are
terms ∼ M 2 /r 2, ∼ J 2 /r 2 in the metric. Then, we cannot say, for
instance, that ! "
2M 1
g00 = −1 + +O (2.46)
r r3
because there is also a term ∼ M 2 /r 2 arising from nonlinear terms
in Einstein’s equations. If we don’t want to compute these terms,
we truncate our expansion at O(1/r 2):
! ! ""
2 2M 1
ds = − 1 − +O dt2
r r2
! ! ""
2M 1 / 2 2 2 2 2
0
+ 1+ +O dr + r (dθ + sin θdφ )
r r2
! ! ""
4J 1
+ − +O sin2 θdtdφ . (2.47)
r r2
Finally, we make the following redefinition of the radial coordinate:
r →r−M. (2.48)
Making this change in (2.47), and neglecting contributions O(1/r 2),
the only term which produces a change in the metric is
! " ! "
2M 2 2 2 2 2M
1+ r (dθ + sin θdφ ) → 1 + (r − M)2 (dθ2 + sin2 θdφ2 )
r r

27
 ! ! ""
2 1
=r 1+O (dθ2 + sin2 θdφ2 ) . (2.49)
r2
With this coordinate redefinition, we find the expression (2.3):
! " ! "
2 2M 2 2M
ds = − 1 − dt + 1 + dr 2
r r
4J
+r 2 (dθ2 + sin2 θdφ2 ) − sin2 θdtdφ
r
+ higher order terms in 1/r . (2.50)
Notice that in the case of a spherically symmetric spacetime (in
which J = 0), the line element (2.50) corresponds to the lineariza-
tion of the Schwarzschild metric in the standard coordinates (t, r, θ, φ).
In the same limit, (2.38) corresponds to the linearization of the
Schwarzschild metric in isotropic coordinates.

2.2 The case of a general source

Let us now consider a general source. The gravitational field can be
strong near the source, whereas far away, where we want to solve
Einstein’s equations, it is weak. Thus, far away from the source, we
can consider linearized gravity, neglecting terms O(|hµν |2 ), and we
can neglect terms with a large power of 1/r, where r is the distance
from the source.
Therefore, we expand the metric far away from the source as
gµν = ηµν + hµν (2.51)
with ! "
aµν (θ, φ) bµν (θ, φ) 1
h̄µν = + +O . (2.52)
r r2 r3
The coefficients aµν , bµν do not depend by r, but they depend by
the angular variables θ, φ, so that they remain finite for r → ∞. We
can express the angular variables in an alternative way, using the
director cosines ni defined as
xi
ni = = (sin θ cos φ, sin θ sin φ, cos θ) (ni ni = 1). (2.53)
r
Thus, we can say that aµν , bµν are functions of the ni .

28
 The metric perturbation, which we assume to be stationary, sat-
isfies equation (2.8),
∇2 h̄µν = 0 . (2.54)
5
The Laplace operator in spherical coordinates has the form
1 IL
∇2 = 2 ∂r r 2 ∂r + 2 (2.55)
r r
where IL is an operator acting on the angular variables only:
IL ≡ ∂θ2 + cot θ∂θ + sin−2 θ∂φ2 . (2.56)
Substituting (2.52) in (2.55) we easily find
ILaµν (θ, φ) = 0 (2.57)
ILbµν (θ, φ) = −2bµν (θ, φ) . (2.58)
The eigenfunctions of the operator IL are the spherical harmonics
Ylm (θ, φ), with l = 0, 1, . . . and m = −l, −l + 1, . . . , l − 1, l. They
are defined by the property
ILYlm = −l(l + 1)Ylm . (2.59)
Equation (2.58) tells us that bµν is a linear combination of the spher-
ical harmonics with l = 1, which are
1
3
Y11 = − sin θeiφ
8π
1
3
Y10 = cos θ
4π
1
3
Y1−1 = sin θe−iφ . (2.60)
8π
This is equivalent to say that bµν is a linear combination of ni = xi /r,
because
1
1 x1 8π −Y11 + Y1−1
n = = sin θ cos φ =
r 3 2
1
2
x 8π −Y11 − Y1−1
n2 = = sin θ sin φ =
r 3 2i
1
3
x 4π
n3 = = cos θ = Y10 . (2.61)
r 3
5 The theory of Laplace equation and the properties of spherical harmonics are extensively

discussed in the literature. See for instance Jackson’s book Electromagnetism, Chapter 3.

29
Therefore, aµν does not depend on the angular variables ni , while
bµν , being a combination of the three harmonics Y1m , is linear in ni ,
and can be written as
bµν (ni ) = bµν i
i n . (2.62)
Expansion (2.52) can then be written as
! "
aµν bµν xi 1
µν
h̄ = + i 3 +O , (2.63)
r r r3
with aµν , bµνi constant coefficients.
We now impose on (2.63) the gauge condition (2.6)
h̄µν,ν = 0 (2.64)
which in the case of stationary perturbations becomes
h̄µi,i = 0 . (2.65)
We get (remember that in linearized gravity it is irrelevant if a space
index i is high or low)
aµj xj bµji (δ ij r 2 − 3xi xj )
h̄µj,j = − + =0 (2.66)
r3 r5
which has to be satisfied for all (large) values of r and for all values
of ni = xi /r. Thus,
aµj = 0
' ij i j
(
δ − 3n n bµij = 0 . (2.67)
These equations do not involve a00 , b00i , which are then in general
nonvanishing free constants; to simplify the notation, we rewrite
them as
a ≡ a00
bi ≡ b00i . (2.68)
The first of the conditions (2.67) states that all the constants aµν
different from a00 vanish. The second one is more complicate to
analyze. It can be rewritten as
H ij b0ij = 0 (2.69)
H ij bkij = 0 (2.70)

30
where we have defined
H ij ≡ δ ij − 3ni nj . (2.71)
The general solution of (2.69), (2.70) is
b0ij = bδij + ck *ijk (2.72)
bkij = dk δij + di δkj − dj δki (2.73)
where b, ck , dk are constants. A rigorous proof of (2.72), (2.73) would
require an use of the structures of Group Theory which goes beyond
the scope of these lectures. We only give here an intuitive, non-
rigorous proof of the first solution, (2.72).
Equation (2.69) must be satisfied for every value of ni , i.e. for
every value of the angular variables θ, φ. But, while H ij depends on
the angles, bµij cannot depend on the angles. Thus, equation (2.69)
can be satisfied only because of the symmetry properties of H ij : it
is symmetric and traceless
H ij = H ji δij H ij = 0 . (2.74)
All the quantities considered here are tensors in the euclidean three-
dimensional space. The only constant tensors which vanish when
contracted with H ij are the Kronecker delta δij and the completely
antisymmetric tensor *ijk : the former vanishes because H ij is trace-
less, the latter vanishes because H ij is symmetric. The solution b0ij
must be a combination of them, thus we have (2.72).
Summarizing, by imposing the gauge condition (2.6) on the ex-
pansion (2.63) we get
! "
a bi xi 1
h̄00 = + 3 +O
r r r3
! "
bxi xj ck 1
h̄0i = 3
+ *ijk 3 + O
r r r3
! "
1 ' k j i
( 1
h̄ij = 3 −δij dk x + di x + dj x + O (2.75)
r r3
depending on the constants a, bi , b, ck , dk .
The constants b, dk can be eliminated by a (position dependent)
infinitesimal diffeomorphism xµ → xµ + ξ µ with parameter
! "
µ b di
ξ = − ,− (2.76)
r r

31
(it is infinitesimal in the sense that r is large and ξ ∼ 1/r).
The change in the metric is
gµν = ηµν + hµν → gµν + gµα ξ α,ν + gνα ξ α,µ + gµν,α ξ α
= ηµν + hµν + gµα ξ α,ν + gνα ξ α,µ + gµν,α ξ α
(2.77)
therefore the change in the perturbation δhµν is
δhµν = gµα ξ α,ν + gνα ξ α,µ + gµν,α ξ α (2.78)
and, being ξ µ = O(|hµν |), we have that, neglecting terms quadratic
in hµν ,
δhµν = ηµα ξ α,ν + ηνα ξ α,µ . (2.79)
The change in the variable
1
h̄µν = hµν − ηµν η αβ hαβ (2.80)
2
is
1
δ h̄µν = δhµν − ηµν η αβ δhαβ
2
= ηµα ξ α,ν + ηνα ξ α,µ − ηµν ξ α,α . (2.81)
We have
ξ µ,0 = 0
bxi
ξ 0,i =
r3
dk xi
ξ k,i = (2.82)
r3
then
! " ! "
1 dk xk 1
δ h̄00 = −η00 ξ k,k+O = 3 +O
r3 r r3
! " ! "
0 1 bxi 1
δ h̄0i = η00 ξ,i + O 3
=− 3 +O
r r r3
1 / 0
δ h̄ij = ηik ξ k,j + ηjk ξ k,i − ηij ξ k,k = 3 di xj + dj xj − ηij dk xk
r
(2.83)

32
thus, after the diffeomorphism,
! "
a b̃i xi 1
h̄00 = + 3 +O
r r r3
! "
xj ck 1
h̄0i = *ijk 3 + O
r r3
! "
1
h̄ij = O (2.84)
r3
where we have defined
b̃i ≡ bi + di . (2.85)
Furthermore, we can get rid of b̃i by making the (rigid) translation
b̃i
xi → xi + (2.86)
a
which produces the following change in the a/r term:
4 52 −1/2
a ' ( −1/2 b̃i 
= a (xi )2 → a  xi +
r a
4 4 55−1/2 ! "
2 b̃i xi 1
= a r 1+2 2 +O
r a r3
4 5 ! " ! "
a b̃i xi 1 a b̃i xi 1
= 1− 2 +O = − 3 +O .
r r a r3 r r r3
(2.87)
Therefore,
! "
a 1
h̄00 = +O
r r3
! "
xj ck 1
h̄0i = *ijk 3 + O
r r3
! "
1
h̄ij = O . (2.88)
r3
Finally, we compute
1
hµν = h̄µν − ηµν η αβ h̄αβ . (2.89)
2

33
We have
1 αβ a
η h̄αβ = − (2.90)
2 2r
therefore
! "
a 1
h00 = +O
2r r3
! "
xj ck 1
h0i = *ijk 3 + O
r r3
! "
a 1
hij = δij + O . (2.91)
2r r3
With the identifications
a = 4M ck = 2J k (2.92)
the solution (2.91) coincides with the solution (2.38), which we have
derived in the case of a weak field source, and that we have already
shown to be equivalent to the solution (2.3).

2.3 Mass and angular momentum of an isolated

object
As we have seen, the metric
! " ! "
2M 2M
ds2 = − 1 − dt2 + 1 + dr 2
r r
4J
+r 2 (dθ2 + sin2 θdφ2 ) − sin2 θdtdφ
r
+ higher order terms in 1/r . (2.93)
describes the far field limit of an isolated, stationary source. If
the source is weakly gravitating, we can apply the definitions of
Newtonian physics on the source, and in this case we have seen that
M and J have a simple interpretation: they are the mass and the
angular momentum of the source, respectively.
In the case of a source which is not weakly gravitating, M and
J are simply integration constants of the general far field solution
(2.93). We could ask: which is their physical interpretation in this
case?

34
 One answer which is often given to this question is the follow-
ing. The motion of a test body in the metric (2.3), far away from
a strongly gravitating source, cannot be distinguished from the mo-
tion it would have if the source were weakly gravitating, with mass
M and angular momentum J. Thus, we can give an operational
definition of the mass and angular momentum of the strongly gravi-
tating source, by looking to the motion of test bodies far away from
the source. The mass will be defined by looking to Kepler’s third
law, and the angular momentum by looking to the precession of
gyroscopes orbiting around the source.
A different answer is based on the stress-energy pseudotensor tµν ,
which describes the energy and momentum carried by the gravita-
tional field, and satisfies, together with the stress-energy tensor Tµν ,
a conservation law:
[(−g)(T µν + tµν )],ν = 0 . (2.94)

It can be expressed as a divergence:

∂ζ µνα
(−g)(T µν + tµν ) = (2.95)
∂xα
where
1 ∂ / 0
ζ µνα = (−g)(g µν αβ
g − g µα νβ
g ) . (2.96)
16π ∂xβ
We can also write the quantity ζ µνα in a different way, which will
be useful later:
∂λµναβ
ζ µνα = (2.97)
∂xβ
with
1 / 0
λµναβ ≡ (−g)(g µν g αβ − g µα g νβ ) . (2.98)
16π
Furthermore, in a stationary spacetime, gµν,0 = 0 therefore
∂λµναi
ζ µνα = . (2.99)
∂xi
Let us now consider a spherical three-dimensional volume V centered
on the source, extending up to a radius r far away from the source.
Notice that it is different from the volume V considered in Section
2.1, which was only covering the source.

35
 The total four-momentum enclosed in the volume V is due both
to the source and to the gravitational field, and is
,
µ
P = d3 x(−g)(T 0µ + t0µ ) . (2.100)
V

As explained in Section 2.1, if the three-momentum in the volume

element d3 x located at the space point with coordinates xi is
P i d3 x , (2.101)
then the angular momentum in the same volume element is
*ijk P j xk d3 x . (2.102)
Therefore, the total angular momentum is
,
i ijk
J =* d3 x(−g)(T 0j + t0j )xk . (2.103)
V

Substituting (2.95) in (2.100) and imposing the stationarity of the

spacetime, , ,
0µα 0µk
µ 3 ∂ζ 3 ∂ζ
P = dx = d x . (2.104)
V ∂xα V ∂xk
Using Gauss’ theorem we can express this integral as an integral on
the spherical surface S, orthogonal, at each point, to the vector n,
which is the unit vector normal to the surface S:
,
Pµ = dSζ 0µk nk . (2.105)
S

Substituting (2.95) in (2.103) we find that the total angular momen-

tum enclosed in the volume V is
, 0jl , * +
i 3 ∂ζ k 3 ∂(ζ 0jl xk ) 0jl ∂x
k
J = *ijk dx x = *ijk dx −ζ
V ∂xl V ∂xl ∂xl
, * + ,
∂(ζ 0jl xk ) ∂ ' 0jl k (
= *ijk d3 x − ζ 0jk
= *ijk d 3
x ζ x − λ 0jkl
∂xl ∂xl
,V V
' (
= *ijk dS ζ 0jlxk − λ0jkl nl . (2.106)
S

Substituting (2.96) and (2.98) in (2.105), (2.106) one finds

P µ = (M, 0, 0, 0) , J i = (0, 0, J) . (2.107)

36
We show this explicitly in the case of M. We have that
,
0
P = dSζ 00i ni (2.108)
S

and dS = r 2 dΩ. The metric perturbation is

! "
2M 1
h00 = +O
r r2
! "
2 j k 1
h0i = 3 *ijk x J + O
r r3
! "
2M 1
hij = δij + O . (2.109)
r r2
Being gµν = ηµν +hµν +O(|hµν |2 ), we have that the property g µν gνρ =
δρµ implies
g µν = η µν − hµν + O(|hµν |2 ) (2.110)
where the indexes of hµν have been raised with the Minkowski met-
ric. Indeed,
(η µν − hµν )(ηνρ + hνρ ) = δρµ + O(|hµν |2 ) . (2.111)
Therefore,
2M
g 00 = −1 − h00 + O(|hµν |2 ) = −1 − + O(|hµν |2 ) (2.112)
! r "
ij ij ij 2 2M
g = δ − h + O(|hµν | ) = 1 − δ ij + O(|hµν |2 ) .
r
(2.113)
Finally, the determinant of gµν is
4M
g = (−1 + h00 )(1 + hii ) = −1 − + O(|hµν |2 ) . (2.114)
r
Putting together (2.112), (2.113), (2.114), and neglecting the terms
O(|hµν |2 ) (like the terms ∼ M 2 , ∼ J 2 ), we find
1 i / ' (0
ζ 00i ni = n ∂j (−g) g 00 g ij − g 0i g 0j
16π *! "! "! " +
1 i 4M 2M 2M
= n ∂j 1+ −1 − 1− δij + O(|hµν |2 )
16π r r r
1 i 4M 1 M i i 1 M
= − n ∂i + O(|hµν |2 ) = nn = (2.115)
16π r 4π r 2 4π r 2

37
thus ,
0
P = ζ 00i ni r 2 dΩ = M . (2.116)
S
The case of angular momentum is conceptually similar, but the com-
putation is more lengthy and we don’t repeat it here.
We can conclude that the integration constants M and J ap-
pearing in the far field limit metric of an isolated source (2.3) can
be correctly interpreted as the mass and the angular momentum, re-
spectively, of the system. In the case of a weakly gravitating source,
the system is the source alone, whereas if the source has a strong
gravitational field, the field itself is part of the system and con-
tributes to the total mass M and angular momentum J appearing
in the line element (2.3).
We stress again that, since the source is isolated, the metric ap-
proaches the Minkowski metric far away from the source; this allows
us to assume, for r sufficiently large, that gµν = ηµν + hµν with hµν
small. Furthermore, the hµν terms can be expanded in powers of
1/r. The dominant contribution in this expansion tells us which are
the total mass and angular momentum of the system.

38