Part 3 General Relativity

Harvey Reall
Part 3 GR December 14, 2016 ii H.S. Reall
Contents

Preface vii

1 Equivalence Principles 1
1.1 Incompatibility of Newtonian gravity and Special Relativity . . . . 1
1.2 The weak equivalence principle . . . . . . . . . . . . . . . . . . . . 2
1.3 The Einstein equivalence principle . . . . . . . . . . . . . . . . . . . 4
1.4 Tidal forces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.5 Bending of light . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.6 Gravitational red shift . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.7 Curved spacetime . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

2 Manifolds and tensors 11

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2 Differentiable manifolds . . . . . . . . . . . . . . . . . . . . . . . . 12
2.3 Smooth functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.4 Curves and vectors . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.5 Covectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.6 Abstract index notation . . . . . . . . . . . . . . . . . . . . . . . . 22
2.7 Tensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.8 Tensor fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.9 The commutator . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.10 Integral curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

3 The metric tensor 31

3.1 Metrics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.2 Lorentzian signature . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.3 Curves of extremal proper time . . . . . . . . . . . . . . . . . . . . 36

4 Covariant derivative 39
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.2 The Levi-Civita connection . . . . . . . . . . . . . . . . . . . . . . . 43

iii
CONTENTS

4.3 Geodesics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.4 Normal coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

5 Physical laws in curved spacetime 49

5.1 Minimal coupling, equivalence principle . . . . . . . . . . . . . . . . 49
5.2 Energy-momentum tensor . . . . . . . . . . . . . . . . . . . . . . . 51

6 Curvature 55
6.1 Parallel transport . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
6.2 The Riemann tensor . . . . . . . . . . . . . . . . . . . . . . . . . . 56
6.3 Parallel transport again . . . . . . . . . . . . . . . . . . . . . . . . 57
6.4 Symmetries of the Riemann tensor . . . . . . . . . . . . . . . . . . 59
6.5 Geodesic deviation . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
6.6 Curvature of the Levi-Civita connection . . . . . . . . . . . . . . . 62
6.7 Einsteins equation . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

7 Diffeomorphisms and Lie derivative 67

7.1 Maps between manifolds . . . . . . . . . . . . . . . . . . . . . . . . 67
7.2 Diffeomorphisms, Lie Derivative . . . . . . . . . . . . . . . . . . . . 69

8 Linearized theory 77
8.1 The linearized Einstein equation . . . . . . . . . . . . . . . . . . . . 77
8.2 The Newtonian limit . . . . . . . . . . . . . . . . . . . . . . . . . . 80
8.3 Gravitational waves . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
8.4 The field far from a source . . . . . . . . . . . . . . . . . . . . . . . 88
8.5 The energy in gravitational waves . . . . . . . . . . . . . . . . . . . 92
8.6 The quadrupole formula . . . . . . . . . . . . . . . . . . . . . . . . 96
8.7 Comparison with electromagnetic radiation . . . . . . . . . . . . . . 98
8.8 Gravitational waves from binary systems . . . . . . . . . . . . . . . 100

9 Differential forms 105

9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
9.2 Connection 1-forms . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
9.3 Spinors in curved spacetime . . . . . . . . . . . . . . . . . . . . . . 109
9.4 Curvature 2-forms . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
9.5 Volume form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
9.6 Integration on manifolds . . . . . . . . . . . . . . . . . . . . . . . . 114
9.7 Submanifolds and Stokes theorem . . . . . . . . . . . . . . . . . . . 115

Part 3 GR December 14, 2016 iv H.S. Reall

 CONTENTS

10 Lagrangian formulation 119

10.1 Scalar field action . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
10.2 The Einstein-Hilbert action . . . . . . . . . . . . . . . . . . . . . . 120
10.3 Energy momentum tensor . . . . . . . . . . . . . . . . . . . . . . . 123

Part 3 GR December 14, 2016 v H.S. Reall

CONTENTS

Part 3 GR December 14, 2016 vi H.S. Reall

Preface

These are lecture notes for the course on General Relativity in Part III of the
Cambridge Mathematical Tripos. There are introductory GR courses in Part II
(Mathematics or Natural Sciences) so, although self-contained, this course does not
cover topics usually covered in a first course, e.g., the Schwarzschild solution, the
solar system tests, and cosmological solutions. You should consult an introductory
book (e.g. Gravity by J.B. Hartle) if you have not studied these topics before.

Acknowledgment
I am very grateful to Andrius Stikonas for producing the figures.

Conventions
Apart from the first lecture, we will use units in which the speed of light is one:
c = 1.
We will use abstract indices a, b, c etc to denote tensors, e.g. V a , gcd . Equa-
tions involving such indices are basis-independent. Greek indices , etc refer to
tensor components in a particular basis. Equations involving such indices are valid
only in that basis.
We will define the metric tensor to have signature ( + ++), which is the most
common convention. Some authors use signature (+ ).
Our convention for the Riemann tensor is such that the Ricci identity takes
the form
a b V c b a V c = Rc dab V d .

Some authors define the Riemann tensor with the opposite sign.

vii
CHAPTER 0. PREFACE

Bibliography
There are many excellent books on General Relativity. The following is an incom-
plete list:

1. General Relativity, R.M. Wald, Chicago UP, 1984.

2. Advanced General Relativity, J.M. Stewart, CUP, 1993.

3. Spacetime and geometry: an introduction to General Relativity, S.M. Carroll,

Addison-Wesley, 2004.

4. Gravitation, C.W. Misner, K.S. Thorne and J.A. Wheeler, Freeman 1973.

5. Gravitation and Cosmology, S. Weinberg, Wiley, 1972.

Our approach will be closest to that of Wald. The first part of Stewarts book
is based on a previous version of this course. Carrolls book is a very readable
introduction. Weinbergs book contains a good discussion of equivalence principles.
Our treatments of the Newtonian approximation and gravitational radiation are
based on Misner, Thorne and Wheeler.

Part 3 GR December 14, 2016 viii H.S. Reall

Chapter 1

Equivalence Principles

1.1 Incompatibility of Newtonian gravity and Spe-

cial Relativity
Special relativity has a preferred class of observers: inertial (non-accelerating)
observers. Associated to any such observer is a set of coordinates (t, x, y, z) called
an inertial frame. Different inertial frames are related by Lorentz transformations.
The Principle of Relativity states that physical laws should take the same form in
any inertial frame.
Newtons law of gravitation is

2 = 4G (1.1)

where is the gravitational potential and the mass density. Lorentz transforma-
tions mix up time and space coordinates. Hence if we transform to another inertial
frame then the resulting equation would involve time derivatives. Therefore the
above equation does not take the same form in every inertial frame. Newtonian
gravity is incompatible with special relativity.
Another way of seeing this is to look at the solution of (1.1):
Z
(t, y)
(t, x) = G d3 y (1.2)
|x y|

From this we see that the value of at point x will respond instantaneously to
a change in at point y. This violates the relativity principle: events which are
simultaneous (and spatially separated) in one inertial frame wont be simultaneous
in all other inertial frames.
The incompatibility of Newtonian gravity with the relativity principle is not
a problem provided all objects are moving non-relativistically (i.e. with speeds

1
CHAPTER 1. EQUIVALENCE PRINCIPLES

much less than the speed of light c). Under such circumstances, e.g. in the Solar
System, Newtonian theory is very accurate.
Newtonian theory also breaks down when the gravitational field becomes strong.
Consider a particle moving in a circular orbit of radius r about a spherical body
of mass M , so = GM/r. Newtons second law gives v 2 /r = GM/r2 hence
v 2 /c2 = ||/c2 . Newtonian theory requires non-relativistic motion, which is the
case only if the gravitational field is weak: ||/c2  1. In the Solar System
||/c2 < 105 .

r
m
M

Figure 1.1: Circular orbit

GR is the theory that replaces both Newtonian gravity and special relativity.

1.2 The weak equivalence principle

The equivalence principle was an important step in the development of GR. There
are several forms of the EP, which are motivated by thought experiments involving
Newtonian gravity. (If we consider only experiments in which all objects move non-
relativistically then the incompatibility of Newtonian gravity with the relativity
principle is not a problem.)
In Newtonian theory, one can distinguish between the notions of inertial mass
mI , which appears in Newtons second law: F = mI a, and gravitational mass,
which governs how a body interacts with a gravitational field: F = mG g. Note that
this equation defines both mG and g hence there is a scaling ambiguity g g and
mG 1 mG (for all bodies). We fix this by defining mI /mG = 1 for a particular
test mass, e.g., one made of platinum. Experimentally it is found that other bodies
made of other materials have mI /mG 1 = O(1012 ) (Eotvos experiment).
The exact equality of mI and mG for all bodies is one form of the weak equiv-
alence principle. Newtonian theory provides no explanation of this equality.
The Newtonian equation of motion of a body in a gravitational field g(x, t) is

mI x = mG g(x(t), t) (1.3)

Part 3 GR December 14, 2016 2 H.S. Reall

 1.2. THE WEAK EQUIVALENCE PRINCIPLE

using the weak EP, this reduces to

x = g(x(t), t) (1.4)

Solutions of this equation are uniquely determined by the initial position and
velocity of the particle. Any two particles with the same initial position and
velocity will follow the same trajectory. This means that the weak EP can be
restated as: The trajectory of a freely falling test body depends only on its initial
position and velocity, and is independent of its composition.
By test body we mean an uncharged object whose gravitational self-interaction
is negligible, and whose size is much less than the length over which external fields
such as g vary.
Consider a new frame of reference moving with constant acceleration a with
respect to the first frame. The origin of the new frame has position X(t) where
X = a. The coordinates of the new frame are t0 = t and x0 = x X(t). Hence the
equation of motion in this frame is

x0 = g a g0 (1.5)

The motion in the accelerating frame is the same as in the first frame but with a
different gravitational field g0 . If g = 0 then the new frame appears to contain a
uniform gravitational field g0 = a: uniform acceleration is indistinguishable from
a uniform gravitational field.
Consider the case in which g is constant and non-zero. We can define an inertial
frame as a reference frame in which the laws of physics take the simplest form. In
the present case, it is clear that this is a frame with a = g, i.e., a freely falling
frame. This gives g0 = 0 so an observer at rest in such a frame, i.e., a freely falling
observer, does not observe any gravitational field. From the perspective of such an
observer, the gravitational field present in the original frame arises because this
latter frame is accelerating with acceleration g relative to him.
Even if the gravitational field is not uniform, it can be approximated as uniform
for experiments performed in a region of space-time sufficiently small that the non-
uniformity is negligible. In the presence of a non-constant gravitational field, we
define a local inertial frame to be a set of coordinates (t, x, y, z) that a freely falling
observer would define in the same way as coordinates are defined in Minkowski
spacetime. The word local emphasizes the restriction to a small region of spacetime,
i.e., t, x, y, z are restricted to sufficiently small values that any variation in the
gravitational field is negligible.
In a local inertial frame, the motion of test bodies is indistinguishable from the
motion of test bodies in an inertial frame in Minkowski spacetime.

Part 3 GR December 14, 2016 3 H.S. Reall

CHAPTER 1. EQUIVALENCE PRINCIPLES

1.3 The Einstein equivalence principle

The weak EP governs the motion of test bodies but it does not tell us anything
about, say, hydrodynamics, or charged particles interacting with an electromag-
netic field. Einstein extended the weak EP as follows:
(i) The weak EP is valid. (ii) In a local inertial frame, the results of all non-
gravitational experiments will be indistinguishable from the results of the same
experiments performed in an inertial frame in Minkowski spacetime.
The weak EP implies that (ii) is valid for test bodies. But any realistic test
body is made from ordinary matter, composed of electrons and nuclei interacting
via the electromagnetic force. Nuclei are composed of protons and neutrons, which
are in turn composed of quarks and gluons, interacting via the strong nuclear force.
A significant fraction of the nuclear mass arises from binding energy. Thus the
fact that the motion of test bodies is consistent with (ii) is evidence that the
electromagnetic and nuclear forces also obey (ii). In fact Schiff s conjecture states
that the weak EP implies the Einstein EP.
Note that we have motivated the Einstein EP by Newtonian arguments. Since
we restricted to velocities much less than the speed of light, the incompatibility of
Newtonian theory with special relativity is not a problem. But the Einstein EP
is supposed to be more general than Newtonian theory. It is a guiding principle
for the construction of a relativistic theory of gravity. In particular, any theory
satisfying the EP should have some notion of local inertial frame.

1.4 Tidal forces

The word local is essential in the above statement of the Einstein EP.
Consider a lab, freely falling radially towards the Earth, that contains two test
particles at the same distance from the Earth but separated horizontally:

Lab frame

Earth

Figure 1.2: Tidal forces

Part 3 GR December 14, 2016 4 H.S. Reall

 1.5. BENDING OF LIGHT

The gravitational attraction of the particles is tiny and can be neglected. Nev-
ertheless, as the lab falls towards Earth, the particles will accelerate towards each
other because the gravitational field has a slightly different direction at the loca-
tion of the two particles. This is an example of a tidal force: a force arising from
non-uniformity of the gravitational field. Such forces are physical: they cannot be
eliminated by free fall.

1.5 Bending of light

The Einstein EP implies that light is bent by a gravitational field.
Consider a uniform gravitational field again, e.g. a small region near the Earths
surface. A freely falling laboratory is a local inertial frame.

lab

Earth

Figure 1.3: Freely falling lab near Earths surface

Inside the lab, the Einstein EP tells us that light rays must move on straight lines.
But a straight line with respect to the lab corresponds to a curved path w.r.t to a
frame at rest relative to the Earth.

d = ct

1 2
2
gt

Figure 1.4: Light ray in the lab frame. Figure 1.5: Light ray in Earth frame.

This shows that light falls in the gravitational field in exactly the same way as a
massive test particle: in time t is falls a distance (1/2)gt2 . (The effect is tiny: if the

Part 3 GR December 14, 2016 5 H.S. Reall

CHAPTER 1. EQUIVALENCE PRINCIPLES

field is vertical then the time taken for the light to travel a horizontal distance d
is t = d/c. In this time, the light falls a distance h = gd2 /(2c2 ). Taking d = 1 km,
g 10ms2 gives h 5 1011 m.)
NB: this is a local effect in which the gravitational field is approximated as
uniform so the result follows from the EP. It cant be used to calculate the bending
of light rays by a non-uniform gravitational field e.g. light bending by the the Sun.

1.6 Gravitational red shift

Alice and Bob are at rest in a uniform gravitational field of strength g in the
negative z-direction. Alice is at height z = h, Bob is at z = 0 (both are on the
z-axis). They have identical clocks. Alice sends light signals to Bob at constant
proper time intervals which she measures to be A . What is the proper time
interval B between the signals received by Bob?

z
h Alice

0 Bob

Figure 1.6: Pound-Rebka experiment

Alice and Bob both have acceleration g with respect to a freely falling frame.
Hence, by the EP, this experiment should give identical results to one in which Alice
and Bob are moving with acceleration g in the positive z-direction in Minkowski
spacetime. We choose our freely falling frame so that Alice and Bob are at rest at
t = 0.
We shall neglect special relativistic effects in this problem, i.e., effects of order
2 2
v /c where v is a typical velocity (the analysis can extended to include such
effects). The trajectories of Alice and Bob are therefore the usual Newtonian
ones:
1 1
zA (t) = h + gt2 , zB (t) = gt2 (1.6)
2 2
Alice and Bob have v = gt so we shall assume that gt/c is small over the time it
takes to perform the experiment. We shall neglect effects of order g 2 t2 /c2 .

Part 3 GR December 14, 2016 6 H.S. Reall

 1.6. GRAVITATIONAL RED SHIFT

Assume Alice emits the first light signal at t = t1 . Its trajectory is z =

zA (t1 ) c(t t1 ) = h + (1/2)gt21 c(t t1 ) so it reaches Bob at time t = T1 where
this equals zB (T1 ), i.e.,
1 1
h + gt21 c(T1 t1 ) = gT12 (1.7)
2 2
The second light signal is emitted at time t = t1 + A (there is no special
relativistic time dilation to the accuracy we are using here so the proper time
interval A is the same as an inertial time interval). Its trajectory is z =
zA (t1 + A ) c(t t1 A ). Let it reach Bob at time t = T1 + B , i.e.,
the proper time intervals between the signals received by Bob is B . Then we
have
1 1
h + g(t1 + A )2 c(T1 + B t1 A ) = g(T1 + B )2 . (1.8)
2 2
Subtracting equation (1.7) gives
g g
c(A B ) + A (2t1 + A ) = B (2T1 + B ) (1.9)
2 2
The terms quadratic in A and B are negligible. This is because we must
assume gA  c, since otherwise Alice would reach relativistic speeds by the
time she emitted the second signal. Similarly for B .
We are now left with a linear equation relating A and B

c(A B ) + gA t1 = gB T1 (1.10)

Rearranging:
 1    
gT1 gt1 g(T1 t1 )
B = 1 + 1+ A 1 A (1.11)
c c c
where we have used the binomial expansion and neglected terms of order g 2 T12 /c2 .
Finally, to leading order we have T1 t1 = h/c (this is the time it takes the light
to travel from A to B) and hence
 
gh
B 1 2 A (1.12)
c
The proper time between the signals received by Bob is less than that between the
signals emitted by Alice. Time appears to run more slowly for Bob. For example,
Bob will see that Alice ages more rapidly than him.
If Alice sends a pulse of light to Bob then we can apply the above argument
to each successive wavecrest, i.e., A is the period of the light waves. Hence

Part 3 GR December 14, 2016 7 H.S. Reall

CHAPTER 1. EQUIVALENCE PRINCIPLES

A = A /c where A is the wavelength of the light emitted by Alice. Bob

receives light with wavelength B where B = B /c. Hence we have
 
gh
B 1 2 A . (1.13)
c
The light received by Bob has shorter wavelength than the light emitted by Alice:
it has undergone a blueshift. Light falling in a gravitational field is blueshifted.
This prediction of the EP was confirmed experimentally by the Pound-Rebka
experiment (1960) in which light was emitted at the top of a tower and absorbed
at the bottom. High accuracy was needed since gh/c2 = O(1015 ).
An identical argument reveals that light climbing out of a gravitational field
undergoes a redshift. We can write the above formula in a form that applies to
both situations:  
B A
B 1 + A (1.14)
c2
where is the gravitational potential. Our derivation of this result, using the
EP, is valid only for uniform gravitational fields. However, we will see that GR
predicts that this result is valid also for weak non-uniform fields.

1.7 Curved spacetime

The weak EP states that if two test bodies initially have the same position and
velocity then they will follow exactly the same trajectory in a gravitational field,
even if they have very different composition. (This is not true of other forces: in an
electromagnetic field, bodies with different charge to mass ratio will follow different
trajectories.) This suggested to Einstein that the trajectories of test bodies in a
gravitational field are determined by the structure of spacetime alone and hence
gravity should be described geometrically.
To see the idea, consider a spacetime in which the proper time between two
infinitesimally nearby events is given not by the Minkowskian formula

c2 d 2 = c2 dt2 dx2 dy 2 dz 2 (1.15)

but instead by
   
2 2 2(x, y, z) 2 2 2(x, y, z)
c d = 1 + c dt 1 (dx2 + dy 2 + dz 2 ), (1.16)
c2 c2

where /c2  1. Let Alice have spatial position xA = (xA , yA , zA ) and Bob have
spatial position xB . Assume that Alice sends a light signal to Bob at time tA and
a second signal at time tA + t. Let Bob receive the first signal at time tB . What

Part 3 GR December 14, 2016 8 H.S. Reall

 1.7. CURVED SPACETIME

time does he receive the second signal? We havent discussed how one determines
the trajectory of the light ray but this doesnt matter. The above geometry does
not depend on t. Hence the trajectory of the second signal must be the same as
the first signal (whatever this is) but simply shifted by a time t:
t
tB + t
tB

on
ot
ph
2 nd n
o to
p h
tA + t st
1
tA
A B x

Figure 1.7: Light ray paths

Hence Bob receives the second signal at time tB + t. The proper time interval
between the signals sent by Alice is given by
 
2 2A
A = 1 + 2 t2 , (1.17)
c
where A (xA ). (Note x = y = z = 0 because her signals are sent from
the same spatial position.) Hence, using /c2  1,
 1/2  
2A A
A = 1 + 2 1 + 2 t. (1.18)
c c
Similarly, the proper time between the signals received by Bob is
 
B
B 1 + 2 t. (1.19)
c
Hence, eliminating t:
  1  
B A B A
B 1 + 2 1+ 2 A 1 + A , (1.20)
c c c2
which is just equation (1.14). The difference in the rates of the two clocks has
been explained by the geometry of spacetime. The geometry (1.16) is actually
the geometry predicted by General Relativity outside a time-independent, non-
rotating distribution of matter, at least when gravity is weak, i.e., ||/c2  1.
(This is true in the Solar System: ||/c2 = GM/(rc2 ) 105 at the surface of the
Sun.)

Part 3 GR December 14, 2016 9 H.S. Reall

CHAPTER 1. EQUIVALENCE PRINCIPLES

Part 3 GR December 14, 2016 10 H.S. Reall

Chapter 2

Manifolds and tensors

2.1 Introduction
In Minkowski spacetime we usually use inertial frame coordinates (t, x, y, z) since
these are adapted to the symmetries of the spacetime so using these coordinates
simplifies the form of physical laws. However, a general spacetime has no sym-
metries and therefore no preferred set of coordinates. In fact, a single set of
coordinates might not be sufficient to describe the spacetime. A simple example
of this is provided by spherical polar coordinates (, ) on the surface of the unit
sphere S 2 in R3 :
z

Figure 2.1: Spherical polar coordinates

These coordinates are not well-defined at = 0, (what is the value of

there?). Furthermore, the coordinate is discontinuous at = 0 or 2.
To describe S 2 so that a pair of coordinates is assigned in a smooth way to
every point, we need to use several overlapping sets of coordinates. Generalizing
this example leads to the idea of a manifold. In GR, we assume that spacetime is
a 4-dimensional differentiable manifold.

11
CHAPTER 2. MANIFOLDS AND TENSORS

2.2 Differentiable manifolds

You know how to do calculus on Rn . How do you do calculus on a curved space,
e.g., S 2 ? Locally, S 2 looks like R2 so one can carry over standard results. However,
one has to confront the fact that it is impossible to use a single coordinate system
on S 2 . In order to do calculus we need our coordinates systems to mesh together
in a smooth way. Mathematically, this is captured by the notion of a differentiable
manifold:
Definition. An n-dimensional differentiable manifold is a set M together with a
collection of subsets O such that
S
1. O = M , i.e., the subsets O cover M

2. For each there is a one-to-one and onto map : O U where U is an

open subset of Rn .

3. If O and O overlap, i.e., O O 6= then 1

maps from (O
n n
O ) U R to (O O ) U R . We require that this map be
smooth (infinitely differentiable).
The maps are called charts or coordinate systems. The set { } is called an
atlas.

O O M

U
U

(O O ) (O O )

Figure 2.2: Overlapping charts

Remarks.
1. Sometimes we shall write (p) = (x1 (p), x2 (p), . . . xn (p)) and refer to xi (p)
as the coordinates of p.

Part 3 GR December 14, 2016 12 H.S. Reall

 2.2. DIFFERENTIABLE MANIFOLDS

2. Strictly speaking, we have defined above the notion of a smooth manifold. If

we replace smooth in the definition by C k (k-times continuously differen-
tiable) then we obtain a C k -manifold. We shall always assume the manifold
is smooth.

Examples.

1. Rn : this is a manifold with atlas consisting of the single chart : (x1 , . . . , xn ) 7

(x1 , . . . , xn ).

2. S 1 : the unit circle, i.e., the subset of R2 given by (cos , sin ) with R.
We cant define a chart by using [0, 2) as a coordinate because [0, 2)
is not open. Instead let P be the point (1, 0) and define one chart by 1 :
S 1 {P } (0, 2), 1 (p) = 1 with 1 defined by Fig. 2.3.

1 P
x

Figure 2.3: Definition of 1

Now let Q be the point (1, 0) and define a second chart by 2 : S 1 {Q}
(, ), 2 (p) = 2 where 2 is defined by Fig. 2.4.

Q 2
x

Figure 2.4: Definition of 2

Part 3 GR December 14, 2016 13 H.S. Reall

CHAPTER 2. MANIFOLDS AND TENSORS

Neither chart covers all of S 1 but together they form an atlas. The charts
overlap on the upper semi-circle and on the lower semi-circle. On the
first of these we have 2 = 2 1 1 (1 ) = 1 . On the second we have
2 = 2 1
1 (1 ) = 1 2. These are obviously smooth functions.

3. S 2 : the two-dimensional sphere defined by the surface x2 + y 2 + z 2 = 1 in

Euclidean space. Introduce spherical polar coordinates in the usual way:

x = sin cos , y = sin sin , z = cos (2.1)

these equations define (0, ) and (0, 2) uniquely. Hence this defines
a chart : O U where O is S 2 with the points (0, 0, 1) and the line of
longitude y = 0, x > 0 removed, see Fig. 2.5, and U is (0, ) (0, 2) R2 .

z
O

Figure 2.5: The subset O S 2 : points with = 0, and = 0, 2 are removed.

We can define a second chart using a different set of spherical polar coordi-
nates defined as follows:

x = sin 0 cos 0 , y = cos 0 , z = sin 0 sin 0 , (2.2)

where 0 (0, ) and 0 (0, 2) are uniquely defined by these equations.

This is a chart 0 : O0 U 0 , where O0 is S 2 with the points (0, 1, 0) and
the line z = 0, x < 0 removed, see Fig. 2.6, and U 0 is (0, ) (0, 2). Clearly
S 2 = O O0 . The functions 0 1 and 0 1 are smooth on O O0 so
these two charts define an atlas for S 2 .

Remark. A given set M may admit many atlases, e.g., one can simply add extra
charts to an atlas. We dont want to regard this as producing a distinct manifold
so we make the following definition:

Part 3 GR December 14, 2016 14 H.S. Reall

 2.3. SMOOTH FUNCTIONS

z
O0

Figure 2.6: The subset O0 S 2 : points with 0 = 0, and 0 = 0, 2 are removed.

Definition. Two atlases are compatible if their union is also an atlas. The union
of all atlases compatible with a given atlas is called a complete atlas: it is an atlas
which is not contained in any other atlas.
Remark. We will always assume that were are dealing with a complete atlas.
(None of the above examples gives a complete atlas; such atlases necessarily contain
infinitely many charts.)

2.3 Smooth functions

We will need the notion of a smooth function on a smooth manifold. If : O U
is a chart and f : M R then note that f 1 is a map from U, i.e., a subset
of Rn , to R.
Definition. A function f : M R is smooth if, and only if, for any chart ,
F f 1 : U R is a smooth function.
Remark. In GR, a function f : M R is sometimes called a scalar field.
Examples.
1. Consider the example of S 1 discussed above. Let f : S 1 R be defined by
f (x, y) = x where (x, y) are the Cartesian coordinates in R2 labelling a point
on S 1 . In the first chart 1 we have f 11 (1 ) = f (cos 1 , sin 1 ) = cos 1 ,
which is smooth. Similary f 1 2 (2 ) = cos 2 is also smooth. If is any
other chart then we can write f = (f 1
1 1
i ) (i ), which is smooth
1
because weve just seen that f i are smooth, and i 1 is smooth from
the definition of a manifold. Hence f is a smooth function.
2. Consider a manifold M with a chart : O U Rn . Denote the other
charts in the atlas by . Let : p 7 (x1 (p), x2 (p), . . . xn (p)). Then we can

Part 3 GR December 14, 2016 15 H.S. Reall

CHAPTER 2. MANIFOLDS AND TENSORS

regard x1 (say) as a function on the subset O of M . Is it a smooth function?

Yes: x1 1
is smooth for any chart , because it is the first component
of the map 1
, and the latter is smooth by the definition of a manifold.

3. Often it is convenient to define a function by specifying F instead of f .

More precisely, given an atlas { }, we define f by specifying functions
F : U R and then setting f = F . One has to make sure that the
resulting definition is independent of on chart overlaps. For example, for
S 1 using the atlas discussed above, define F1 : (0, 2) R by 1 7 sin(m1 )
and F2 : (, ) R by 2 7 sin(m2 ), where m is an integer. On the chart
overlaps we have F1 1 = F2 2 because 1 and 2 differ by a multiple of
2 on both overlaps. Hence this defines a function on S 1 .

Remark. After a while we will stop distinguishing between f and F , i.e., we will
say f (x) when we mean F (x).

2.4 Curves and vectors

Rn , or Minkowski spacetime, has the structure of a vector space, e.g., it makes
sense to add the position vectors of points. One can view more general vectors,
e.g., the 4-velocity of a particle, as vectors in the space itself. This structure does
not extend to more general manifolds, e.g., S 2 . So we need to discuss how to define
vectors on manifolds.
For a surface in R3 , the set of all vectors tangent to the surface at some point p
defines the tangent plane to the surface at p (see Fig. 2.7). This has the structure
of a 2d vector space. Note that the tangent planes at two different points p and q
are different. It does not make sense to compare a vector at p with a vector at q.
For example: if one tried to define the sum of a vector at p and a vector at q then
to which tangent plane would the sum belong?

p q

Figure 2.7: Tangent planes.

On a surface, the tangent vector to a curve in the surface is automatically

tangent to the surface. We take this as our starting point for defining vectors
on a general manifold. We start by defining the notion of a curve in a manifold,
and then the notion of a tangent vector to a curve at a point p. We then show

Part 3 GR December 14, 2016 16 H.S. Reall

 2.4. CURVES AND VECTORS

that the set of all such tangent vectors at p forms a vector space Tp (M ). This is
the analogue of the tangent plane to a surface but it makes no reference to any
embedding into a higher-dimensional space.
Definition A smooth curve in a differentiable manifold M is a smooth function
: I M , where I is an open interval in R (e.g. (0, 1) or (1, )). By this we
mean that is a smooth map from I to Rn for all charts .
Let f : M R and : I M be a smooth function and a smooth curve
respectively. Then f is a map from I to R. Hence we can take its derivative
to obtain the rate of change of f along the curve:
d d
[(f )(t)] = [f ((t))] (2.3)
dt dt
In Rn we are used to the idea that the rate of change of f along the curve at a
point p is given by the directional derivative Xp (f )p where Xp is the tangent
to the curve at p. Note that the vector Xp defines a linear map from the space of
smooth functions on Rn to R: f 7 Xp (f )p . This is how we define a tangent
vector to a curve in a general manifold:
Definition. Let : I M be a smooth curve with (wlog) (0) = p. The tangent
vector to at p is the linear map Xp from the space of smooth functions on M to
R defined by  
d
Xp (f ) = [f ((t))] (2.4)
dt t=0

Note that this satisfies two important properties: (i) it is linear, i.e., Xp (f + g) =
Xp (f )+Xp (g) and Xp (f ) = Xp (f ) for any constant ; (ii) it satisfies the Leibniz
rule Xp (f g) = Xp (f )g(p) + f (p)Xp (g), where f and g are smooth functions and
f g is their product.
If = (x1 , x2 , . . . xn ) is a chart defined in a neighbourhood of p and F f 1
then we have f = f 1 = F and hence
   
F (x) dx ((t))
Xp (f ) = (2.5)
x (p) dt t=0

Note that (i) the first term on the RHS depends only on f and , and the second
term on the RHS depends only on and ; (ii) we are using the Einstein summation
convention, i.e., is summed from 1 to n in the above expression.
Proposition. The set of all tangent vectors at p forms a n-dimensional vector
space, the tangent space Tp (M ).
Proof. Consider curves and through p, wlog (0) = (0) = p. Let their
tangent vectors at p be Xp and Yp respectively. We need to define addition of

Part 3 GR December 14, 2016 17 H.S. Reall

CHAPTER 2. MANIFOLDS AND TENSORS

tangent vectors and multiplication by a constant. let and be constants. We

define Xp + Yp to be the linear map f 7 Xp (f ) + Yp (f ). Next we need
to show that this linear map is indeed the tangent vector to a curve through p.
Let = (x1 , . . . , xn ) be a chart defined in a neighbourhood of p. Consider the
following curve:

(t) = 1 [(((t)) (p)) + (((t)) (p)) + (p)] (2.6)

Note that (0) = p. Let Zp denote the tangent vector to this curve at p. From
equation (2.5) we have
   
F (x) d
Zp (f ) = [(x ((t)) x (p)) + (x ((t)) x (p)) + x (p)]
x (p) dt t=0
       
F (x) dx ((t)) dx ((t))
= +
x (p) dt t=0 dt t=0
= Xp (f ) + Yp (f )
= (Xp + Yp )(f ).

Since this is true for any smooth function f , we have Zp = Xp + Yp as required.

Hence Xp + Yp is tangent to the curve at p. It follows that the set of tangent
vectors at p forms a vector space (the zero vector is realized by the curve (t) = p
for all t).
The next step is to show that this vector space is n-dimensional. To do this,
we exhibit a basis. Let 1 n. Consider the curve through p defined by

(t) = 1 (x1 (p), . . . , x1 (p), x (p) + t, x+1 (p), . . . , xn (p)). (2.7)

The tangent vector to this curve at p is denoted x p . To see why, note that,

using equation (2.5)

   
F
(f ) = . (2.8)
x p x (p)

The n tangent vectors x p are linearly independent. To see why, assume that

there exist constants such that x p = 0. Then, for any function f we

must have  
F (x)
= 0. (2.9)
x (p)

Choosing F = x , this reduces to = 0. Letting this run over all values of we

see that all of the constants must vanish, which proves linear independence.

Part 3 GR December 14, 2016 18 H.S. Reall

 2.4. CURVES AND VECTORS

Finally we must prove that these tangent vectors span the vector space. This
follows from equation (2.5), which can be rewritten
   
dx ((t))
Xp (f ) = (f ) (2.10)
dt t=0 x p
this is true for any f hence
dx ((t))
   

Xp = , (2.11)
dt t=0 x p

i.e. Xp can be written as a linear combination of the n tangent vectors x p .

These n vectors therefore form a basis for Tp (M ), which establishes that the tan-
gent space is n-dimensional. QED.
Remark. The basis { x p , = 1, . . . n} is chart-dependent: we had to choose

a chart defined in a neighbourhood of p to define it. Choosing a different chart

would give a different basis for Tp (M ). A basis defined this way is sometimes called
a coordinate basis.
Definition. Let {e , = 1 . . . n} be a basis for Tp (M ) (not necessarily a coordinate
basis). We can expand any vector X Tp (M ) as X = X e . We call the numbers
X the components of X with respect to this basis.
Example. Using the coordinate basis e = (/x )p , equation (2.11) shows that
the tangent vector Xp to a curve (t) at p (where t = 0) has components
 
dx ((t))
Xp = . (2.12)
dt t=0

Remark. Note the placement of indices. We shall sum over repeated indices
if one such index appears upstairs (as a superscript, e.g., X
) and the other
downstairs (as a subscript, e.g., e ). (The index on x p is regarded as
downstairs.) If an equation involves the same index more than twice, or twice but
both times upstairs or both times downstairs (e.g. X Y ) then a mistake has been
made.
Lets consider the relationship between different coordinate bases. Let = (x1 , . . . , xn )
and 0 = (x0 1 , . . . , x0 n ) be two charts defined in a neighbourhood of p. Then, for
any smooth function f , we have
   
1
(f ) = (f )
x p x (p)
 
0 1 0 1
= [(f ) ( )]
x (p)

Part 3 GR December 14, 2016 19 H.S. Reall

CHAPTER 2. MANIFOLDS AND TENSORS

Now let F 0 = f 0 1 . This is a function of the coordinates x0 . Note that the

components of 0 1 are simply the functions x0 (x), i.e., the primed coordinates
expressed in terms of the unprimed coordinates. Hence what we have is easy to
evaluate using the chain rule:
   
0 0
(f ) = (F (x (x)))
x p x (p)
 0   0 0 
x F (x )
=
x (p) x0 0 (p)
 0   
x
= (f )
x (p) x0 p

Hence we have
x0
     

= (2.13)
x p x (p) x0 p

This expresses one set of basis vectors in terms of the other. Let X and X 0
denote the components of a vector with respect to the two bases. Then we have
   0   
x
X=X =X (2.14)
x p x (p) x0 p

and hence
x0
 
0
X =X (2.15)
x (p)

Elementary treatments of GR usually define a vector to be a set of numbers {X }

that transforms according to this rule under a change of coordinates. More pre-
cisely, they usually call this a contravariant vector.

2.5 Covectors
Recall the following from linear algebra:
Definition. Let V be a real vector space. The dual space V of V is the vector
space of linear maps from V to R.
Lemma. If V is n-dimensional then so is V . If {e , = 1, . . . , n} is a basis for
V then V has a basis {f , = 1, . . . , n}, the dual basis defined by f (e ) =
(if X = X e then f (X) = X f (e ) = X ).
Since V and V have the same dimension, they are isomorphic. For example the
linear map defined by e 7 f is an isomorphism. But this is basis-dependent: a

Part 3 GR December 14, 2016 20 H.S. Reall

 2.5. COVECTORS

different choice of basis would give a different isomorphism. In contrast, there is

a natural (basis-independent) isomorphism between V and (V ) :
Theorem. If V is finite dimensional then (V ) is naturally isomorphic to V . The
isomorphism is : V (V ) where (X)() = (X) for all V .
Now we return to manifolds:
Definition. The dual space of Tp (M ) is denoted Tp (M ) and called the cotangent
space at p. An element of this space is called a covector (or 1-form) at p. If {e }
is a basis for Tp (M ) and {f } is the dual basis then we can expand a covector
as f . are called the components of .
Note that (i) (e ) = f (e ) = ; (ii) if X Tp (M ) then (X) = (X e ) =
X (e ) = X (note the placement of indices!)
Definition. Let f : M R be a smooth function. Define a covector (df )p by
(df )p (X) = X(f ) for any vector X Tp (M ). (df )p is the gradient of f at p.
Examples.

1. Let (x1 , . . . , xn ) be a coordinate chart defined in a neighbourhood of p, recall

that x is a smooth function (in this neighbourhood) so we can take f = x
in the above definition to define n covectors (dx )p . Note that
  !  
x
(dx )p
= = (2.16)
x p x p

Hence {(dx )p } is the dual basis of {(/x )p }.

2. To explain why we call (df )p the gradient of f at p, observe that its compo-
nents in a coordinate basis are
 !    
F
[(df )p ] = (df )p = (f ) = (2.17)
x p x p x (p)

where the first equality uses (i) above, the second equality is the definition
of (df )p and the final equality used (2.8).

Exercise. Consider two different charts = (x1 , . . . , xn ) and 0 = (x0 1 , . . . , x0 n )

defined in a neighbourhood of p. Show that
 
x

(dx )p = 0 (dx0 )p , (2.18)
x 0 (p)

Part 3 GR December 14, 2016 21 H.S. Reall

CHAPTER 2. MANIFOLDS AND TENSORS

and hence that, if and 0 are the components of Tp (M ) w.r.t. the two
coordinate bases, then  
0 x
= . (2.19)
x0 0 (p)
Elementary treatements of GR take this as the definition of a covector, which they
usually call a covariant vector.

2.6 Abstract index notation

So far, we have used Greek letters , , . . . to denote components of vectors or
covectors with respect to a basis, and also to label the basis vectors themselves
(e.g. e ). Equations involving such indices are assumed to hold only in that basis.
For example an equation of the form X = 1 says that, in a particular basis,
a vector X has only a single non-vanishing component. This will not be true in
other bases. Furthermore, if we were just presented with this equation, we would
not even know whether or not the quantities {X } are the components of a vector
or just a set of n numbers.
The abstract index notation uses Latin letters a, b, c, . . .. A vector X is denoted
X or X b or X c etc. The letter used in the superscript does not matter. What
a

matters is that there is a superscript Latin letter. This tells us that the object
in question is a vector. We emphasize: X a represents the vector itself, not a
component of the vector. Similarly we denote a covector by a (or b etc).
If we have an equation involving abstract indices then we can obtain an equation
valid in any particular basis simply by replacing the abstract indices by basis
indices (e.g. a , b etc.). For example, consider the quantity a X a in the
abstract index notation. We see that this involves a covector a and a vector X a .
Furthermore, in any basis, this quantity is equal to X = (X). Hence a X a is
the abstract index way of writing (X). Similarly, if f is a smooth function then
X(f ) = X a (df )a .
Conversely, if one has an equation involving Greek indices but one knows that
it is true for an arbitrary basis then one can replace the Greek indices with Latin
letters.
Latin indices must respect the rules of the summation convention so equations
of the form a a = 1 or b = 2 do not make sense.

2.7 Tensors
In Newtonian physics, you are familiar with the idea that certain physical quanti-
ties are described by tensors (e.g. the inertia tensor). You may have encountered

Part 3 GR December 14, 2016 22 H.S. Reall

 2.7. TENSORS

the idea that the Maxwell field in special relativity is described by a tensor. Ten-
sors are very important in GR because the curvature of spacetime is described
with tensors. In this section we shall define tensors at a point p and explain some
of their basic properties.
Definition. A tensor of type (r, s) at p is a multilinear map

T : Tp (M ) . . . Tp (M ) Tp (M ) . . . Tp (M ) R. (2.20)

where there are r factors of Tp (M ) and s factors of Tp (M ). (Multilinear means

that the map is linear in each argument.)
In other words, given r covectors and s vectors, a tensor of type (r, s) produces a
real number.
Examples.

1. A tensor of type (0, 1) is a linear map Tp (M ) R, i.e., it is a covector.

2. A tensor of type (1, 0) is a linear map Tp (M ) R, i.e., it is an element of

(Tp (M )) but this is naturally isomorphic to Tp (M ) hence a tensor of type
(1, 0) is a vector. To see how this works, given a vector X Tp (M ) we define
a linear map Tp (M ) R by 7 (X) for any Tp (M ).

3. We can define a (1, 1) tensor by (, X) = (X) for any covector and

vector X.

Definition. Let T be a tensor of type (r, s) at p. If {e } is a basis for Tp (M ) with

dual basis {f } then the components of T in this basis are the numbers

T 1 2 ...r 1 2 ...s = T (f 1 , f 2 , . . . , f r , e1 , e2 , . . . , es ) (2.21)

In the abstract index notation, we denote T by T a1 a2 ...ar b1 b2 ...bs .

Remark. Tensors of type (r, s) at p can be added together and multiplied by a
constant, hence they form a vector space. Since such a tensor has nr+s components,
it is clear that this vector space has dimension nr+s .
Examples.

1. Consider the tensor defined above. Its components are

= (f , e ) = f (e ) = , (2.22)

where the RHS is a Kronecker delta. This is true in any basis, so in the
abstract index notation we write as ba .

Part 3 GR December 14, 2016 23 H.S. Reall

CHAPTER 2. MANIFOLDS AND TENSORS

2. Consider a (2, 1) tensor. Let and be covectors and X a vector. Then in

our basis we have

T (, , X) = T ( f , f , X e ) = X T (f , f , e ) = T X
(2.23)
Now the basis we chose was arbitrary, hence we can immediately convert this
to a basis-independent equation using the abstract index notation:

T (, , X) = T ab c a b X c . (2.24)

This formula generalizes in the obvious way to a (r, s) tensor.

We have discussed the transformation of vectors and covectors components under

a change of coordinate basis. Lets now examine how tensor components transform
under an arbitrary change of basis. Let {e } and {e0 } be two bases for Tp (M ).
Let {f } and {f 0 } denote the corresponding dual bases. Expanding the primed
bases in terms of the unprimed bases gives

f 0 = A f , e0 = B e (2.25)

for some matrices A and B . These matrices are related because:

= f 0 (e0 ) = A f (B e ) = A B f (e ) = A B = A B . (2.26)

Hence B = (A1 ) . For a change between coordinate bases, our previous results
give  0   
x x
A = , B = (2.27)
x x0
and these matrices are indeed inverses of each other (from the chain rule).
Exercise. Show that under an arbitrary change of basis, the components of a
vector X and a covector transform as

X 0 = A X , 0 = (A1 ) . (2.28)

Show that the components of a (2, 1) tensor T transform as

T0 = A A (A1 ) T . (2.29)

The corresponding result for a (r, s) tensor is an obvious generalization of this

formula.
Given a (r, s) tensor, we can construct a (r 1, s 1) tensor by contraction. This
is easiest to demonstrate with an example.

Part 3 GR December 14, 2016 24 H.S. Reall

 2.7. TENSORS

Example. Let T be a tensor of type (3, 2). Define a new tensor S of type (2, 1)
as follows
S(, , X) = T (f , , , e , X) (2.30)
where {e } is a basis and {f } is the dual basis, and are arbitrary covectors
and X is an arbitrary vector. This definition is basis-independent because

T (f 0 , , , e0 , X) = T (A f , , , (A1 ) e , X)
= (A1 ) A T (f , , , e , X)
= T (f , , , e , X).

The components of S and T are related by S = T in any basis. Since this

is true in any basis, we can write it using the abstract index notation as

S ab c = T dab dc (2.31)

Note that there are other (2, 1) tensors that we can build from T abc de . For ex-
ample, there is T abd cd , which corresponds to replacing the RHS of (2.30) with
T (, , f , X, e ). The abstract index notation makes it clear how many different
tensors can be defined this way: we can define a new tensor by contracting any
upstairs index with any downstairs index.
Another important way of constructing new tensors is by taking the product of
two tensors:
Definition. If S is a tensor of type (p, q) and T is a tensor of type (r, s) then the
outer product of S and T , denoted S T is a tensor of type (p + r, q + s) defined
by

(S T )(1 , . . . , p , 1 , . . . , r , X1 , . . . , Xq , Y1 , . . . , Ys )
= S(1 , . . . , p , X1 , . . . , Xq )T (1 , . . . , r , Y1 , . . . , Ys ) (2.32)

where 1 , . . . , p and 1 , . . . , r are arbitrary covectors and X1 , . . . , Xq and Y1 , . . . , Ys

are arbitrary vectors.
Exercise. Show that this definition is equivalent to

(S T )a1 ...ap b1 ...br c1 ...cq d1 ...ds = S a1 ...ap c1 ...cq T b1 ...br d1 ...ds (2.33)

Exercise. Show that, in a coordinate basis, any (2, 1) tensor T at p can be written
as    

T =T (dx )p (2.34)
x p x p
This generalizes in the obvious way to a (r, s) tensor.

Part 3 GR December 14, 2016 25 H.S. Reall

CHAPTER 2. MANIFOLDS AND TENSORS

Remark. You may be wondering why we write T ab c instead of Tcab . At the

moment there is no reason why we should not adopt the latter notation. However,
it is convenient to generalize our definition of tensors slightly. We have defined a
(r, s) tensor to be a linear map with r + s arguments, where the first r arguments
are covectors and the final s arguments are vectors. We can generalize this by
allowing the covectors and vectors to appear in any order. For example, consider
a (1, 1) tensor. This is a map Tp (M ) Tp (M ) R. But we could just as well
have defined it to be a map Tp (M ) Tp (M ) R. The abstract index notation
allows us to distinguish these possibilities easily: the first would be written as T a b
and the second as Ta b . (2, 1) tensors come in 3 different types: T ab c , T a b c and
Ta bc . Each type of of (r, s) tensor gives a vector space of dimension nr+s but these
vector spaces are naturally isomorphic so often one does not bother to distinguish
between them.
There is a final type of tensor operation that we shall need: symmetrization
and antisymmetrization. Consider a (0, 2) tensor T . We can define two other (0, 2)
tensors S and A as follows:
1 1
S(X, Y ) = (T (X, Y ) + T (Y, X)), A(X, Y ) = (T (X, Y ) T (Y, X)), (2.35)
2 2
where X and Y are vectors at p. In abstract index notation:
1 1
Sab = (Tab + Tba ), Aab = (Tab Tba ). (2.36)
2 2
In a basis, we can regard the components of T as a square matrix. The components
of S and A are just the symmetric and antisymmetric parts of this matrix. It is
convenient to introduce some notation to describe the operations we have just
defined: we write
1 1
T(ab) = (Tab + Tba ), T[ab] = (Tab Tba ). (2.37)
2 2
These operations can be applied to more general tensors. For example,
1
T (ab)c d = (T abc d + T bac d ). (2.38)
2
We can also symmetrize or antisymmetrize on more than 2 indices. To symmetrize
on n indices, we sum over all permutations of these indices and divide the result
by n! (the number of permutations). To antisymmetrize we do the same but we
weight each term in the sum by the sign of the permutation. The indices that we
symmetrize over must be either upstairs or downstairs, they cannot be a mixture.
For example,
1
T (abc)d = T abcd + T bcad + T cabd + T bacd + T cbad + T acbd .


(2.39)
3!
Part 3 GR December 14, 2016 26 H.S. Reall
 2.8. TENSOR FIELDS

1 a
T a [bcd] =(T bcd + T a cdb + T a dbc T a cbd T a dcb T a bdc ) . (2.40)
3!
Sometimes we might wish to (anti)symmetrize over indices which are not ad-
jacent. In this case, we use vertical bars to denote indices excluded from the
(anti)symmetrization. For example,
1
T(a|bc|d) = (Tabcd + Tdbca ) . (2.41)
2

Exercise. Show that T (ab) X[a|cd|b] = 0.

2.8 Tensor fields

So far, we have defined vectors, covectors and tensors at a single point p. However,
in physics we shall need to consider how these objects vary in spacetime. This leads
us to define vector, covector and tensor fields.
Definition. A vector field is a map X which maps any point p M to a vector
Xp at p. Given a vector field X and a function f we can define a new function
X(f ) : M R by X(f ) : p 7 Xp (f ). The vector field X is smooth if this map is
a smooth function for any smooth f .

Example. Given any coordinate chart = (x1 , . . . , xn ), the vector field x
is
defined by p 7 x p . Hence

   
F
(f ) : p 7 , (2.42)
x x (p)

where F f 1 . You should convince yourself that smoothness of f implies

that the above map defines a smooth function. Therefore /x is a smooth vector
field. (Note that (/x ) usually wont be defined on the whole manifold M since
the chart might not cover the whole manifold. So strictly speaking this is not a
vector field on M but only on a subset of M . We shant worry too much about
this distinction.)
Remark. Since the vector fields (/x )p provide a basis for Tp (M ) at any point
p, we can expand an arbitrary vector field as
 

X=X (2.43)
x

Since /x is smooth, it follows that X is smooth if, and only if, its components
X are smooth functions.

Part 3 GR December 14, 2016 27 H.S. Reall

CHAPTER 2. MANIFOLDS AND TENSORS

Definition. A covector field is a map which maps any point p M to a

covector p at p. Given a covector field and a vector field X we can define a
function (X) : M R by (X) : p 7 p (Xp ). The covector field is smooth if
this function is smooth for any smooth vector field X.
Example. Let f be a smooth function. We have defined (df )p above. Now we
simply let p vary to define a covector field df . Let X be a smooth vector field
and f a smooth function. Then df (X) = X(f ). This is a smooth function of p
(because X is smooth). Hence df is a smooth covector field: the gradient of f .
Remark. Taking f = x reveals that dx is a smooth covector field.
Definition. A (r, s) tensor field is a map T which maps any point p M
to a (r, s) tensor Tp at p. Given r covector fields 1 , . . . , r and s vector fields
X1 , . . . , Xs we can define a function T (1 , . . . , r , X1 , . . . , Xs ) : M R by p 7
Tp ((1 )p , . . . , (r )p , (X1 )p , . . . , (Xs )p ). The tensor field T is smooth if this function
is smooth for any smooth covector fields 1 , . . . , r and vector fields X1 , . . . , Xr .
Exercise. Show that a tensor field is smooth if, and only if, its components in a
coordinate chart are smooth functions.
Remark. Henceforth we shall assume that all tensor fields that we encounter are
smooth.

2.9 The commutator

Let X and Y be vector fields and f a smooth function. Since Y (f ) is a smooth
function, we can act on it with X to form a new smooth function X(Y (f )). Does
the map f 7 X(Y (f )) define a vector field? No, because X(Y (f g)) = X(f Y (g) +
gY (f )) = f X(Y (g))+gX(Y (f ))+X(f )Y (g)+X(g)Y (f ) so the Leibniz law is not
satisfied. However, we can also define Y (X(f )) and the combination X(Y (f ))
Y (X(f )) does obey the Leibniz law (check!).
Definition. The commutator of two vector fields X and Y is the vector field
[X, Y ] defined by
[X, Y ](f ) = X(Y (f )) Y (X(f )) (2.44)
for any smooth function f .
To see that this does indeed define a vector field, we can evaluate it in a coordinate
chart:
   
F F
[X, Y ](f ) = X Y Y X
x x
   
F F
= X Y Y X
x x x x

Part 3 GR December 14, 2016 28 H.S. Reall

 2.10. INTEGRAL CURVES

Y F X F
= X Y
 x x x x

Y X F
= X Y
x x x
 

= [X, Y ] (f )
x
where 

Y X
[X, Y ] = X Y . (2.45)
x x
Since f is arbitrary, it follows that
 

[X, Y ] = [X, Y ] . (2.46)
x
The RHS is a vector field hence [X, Y ] is a vector field whose components in a
coordinate basis are given by (2.45). (Note that we cannot write equation (2.45)
in abstract index notation because it is valid only in a coordinate basis.)
Example. Let X = /x1 and Y = x1 /x2 + /x3 . The components of X are
constant so [X, Y ] = Y /x1 = 2 so [X, Y ] = /x2 .
Exercise. Show that (i) [X, Y ] = [Y, X]; (ii) [X, Y + Z] = [X, Y ] + [X, Z]; (iii)
[X, f Y ] = f [X, Y ] + X(f )Y ; (iv) [X, [Y, Z]] + [Y, [Z, X]] + [Z, [X, Y ]] = 0 (the
Jacobi identity). Here X, Y, Z are vector fields and f is a smooth function.
Remark. The components of (/x ) in the coordinate basis are either 1 or 0. It
follows that  

, = 0. (2.47)
x x
Conversely, it can be shown that if X1 , . . . , Xm (m n) are vector fields that are
linearly independent at every point, and whose commutators all vanish, then, in
a neighbourhood of any point p, one can introduce a coordinate chart (x1 , . . . xn )
such that Xi = /xi (i = 1, . . . , m) throughout this neighbourhood.

2.10 Integral curves

In fluid mechanics, the velocity of a fluid is described by a vector field u(x) in
R3 (we are returning to Cartesian vector notation for a moment). Consider a
particle suspended in the fluid with initial position x0 . It moves with the fluid so
its position x(t) satisfies
dx
= u(x(t)), x(0) = x0 . (2.48)
dt
Part 3 GR December 14, 2016 29 H.S. Reall
CHAPTER 2. MANIFOLDS AND TENSORS

The solution of this differential equation is called the integral curve of the vector
field u through x0 . The definition extends straightforwardly to a vector field on a
general manifold:
Definition. Let X be a vector field on M and p M . An integral curve of X
through p is a curve through p whose tangent at every point is X.
Let denote an integral curve of X with (wlog) (0) = p. In a coordinate chart,
this definition reduces to the initial value problem

dx (t)
= X (x(t)), x (0) = xp . (2.49)
dt
(Here we are using the abbreviation x (t) = x ((t)).) Standard ODE theory
guarantees that there exists a unique solution to this problem. Hence there is a
unique integral curve of X through any point p.
Example. In a chart = (x1 , . . . , xn ), consider X = /x1 + x1 /x2 and take
p to be the point with coordinates (0, . . . , 0). Then dx1 /dt = 1, dx2 /dt = x1 .
Solving the first equation and imposing the initial condition gives x1 = t, then
plugging into the second equation and solving gives x2 = t2 /2. The other coords
are trivial: x = 0 for > 2, so the integral curve is t 7 1 (t, t2 /2, 0, . . . , 0).

Part 3 GR December 14, 2016 30 H.S. Reall

Chapter 3

The metric tensor

3.1 Metrics
A metric captures the notion of distance on a manifold. We can motivate the
required definition by considering the case of R3 . Let x(t), a < t < b be a curve
in R3 (were using Cartesian vector notation). Then the length of the curve is
Z b r
dx dx
dt . (3.1)
a dt dt

Inside the integral we see the norm of the tangent vector dx/dt, in other words the
scalar product of this vector with itself. Therefore to define a notion of distance on
a general manifold, we shall start by introducing a scalar product between vectors.
A scalar product maps a pair of vectors to a number. In other words, at a
point p, it is a map g : Tp (M ) Tp (M ) R. A scalar product should be linear in
each argument. Hence g is a (0, 2) tensor at p. We call g a metric tensor. There
are a couple of other properties that g should also satisfy:
Definition. A metric tensor at p M is a (0, 2) tensor g with the following
properties:

1. It is symmetric: g(X, Y ) = g(Y, X) for all X, Y Tp (M ) (i.e. gab = gba )

2. It is non-degenerate: g(X, Y ) = 0 for all Y Tp (M ) if, and only if, X = 0.

Remark. Sometimes we shall denote g(X, Y ) by hX, Y i or X Y .

Since the components of g form a symmetric matrix, one can introduce a basis
that diagonalizes g. Non-degeneracy implies that none of the diagonal elements is
zero. By rescaling the basis vectors, one can arrange that the diagonal elements
are all 1. In this case, the basis is said to be orthonormal. There are many

31
CHAPTER 3. THE METRIC TENSOR

such bases but a standard algebraic theorem (Sylvesters law of inertia) states
that the number of positive and negative elements is independent of the choice
of orthonormal basis. The number of positive and negative elements is called the
signature of the metric.
In differential geometry, one is usually interested in Riemannian metrics. These
have signature + + + . . . + (i.e. all diagonal elements +1 in an orthonormal basis),
and hence g is positive definite. In GR, we are interested in Lorentzian metrics,
i.e., those with signature + + . . . +. This can be motivated by the equivalence
principle as follows. Let spacetime be a 4d manifold M . Consider a local inertial
frame (LIF) at p, with coordinates {x }. A pair of vectors X a , Y a at p have
components X , Y w.r.t the coordinate basis of the LIF. The Einstein EP implies
that special relativity holds in the LIF. In special relativity, we can define a scalar
product X Y where = diag(1, 1, 1, 1). This is Lorentz invariant and
hence gives the same result for all LIFs at p. So the EP predicts that we can
define a (Lorentzian) scalar product at p, i.e., there exists a Lorentzian metric g
at p which has components in a LIF at p. We want g to be defined over the
whole manifold, so we assume it to be a tensor field.
Definition. A Riemannian (Lorentzian) manifold is a pair (M, g) where M is a
differentiable manifold and g is a Riemannian (Lorentzian) metric tensor field. A
Lorentzian manifold is sometimes called a spacetime.
Remark. On a Riemannian manifold, we can now define the length of a curve in
exactly the same way as above: let : (a, b) M be a smooth curve with tangent
vector X. Then the length of the curve is
Z b q
dt g(X, X)|(t) (3.2)
a

Exercise. Given a curve (t) we can define a new curve simply by changing
the parametrization: let t = t(u) with dt/du > 0 and u (c, d) with t(c) = a
and t(d) = b. Show that: (i) the new curve (u) (t(u)) has tangent vector
Y a = (dt/du)X a ; (ii) the length of these two curves is the same, i.e., our definition
of length is independent of parametrization.
In a coordinate basis, we have (cf equation (2.34))
g = g dx dx (3.3)
Often we use the notation ds2 instead of g and abbreviate this to
ds2 = g dx dx (3.4)
This notation captures the intuitive idea of an infinitesimal distance ds being
determined by infinitesimal coordinate separations dx .
Examples.

Part 3 GR December 14, 2016 32 H.S. Reall

 3.1. METRICS

1. In Rn = {(x1 , . . . , xn )}, the Euclidean metric is

g = dx1 dx1 + . . . + dxn dxn (3.5)

(Rn , g) is callled Euclidean space. A coordinate chart which covers all of

R4 and in which the components of the metric are diag(1, 1, . . . , 1) is called
Cartesian.

2. In R4 = {(x0 , x1 , x2 , x3 )}, the Minkowski metric is

= (dx0 )2 + (dx1 )2 + (dx2 )2 + (dx3 )2 . (3.6)

(R4 , ) is called Minkowski spacetime. A coordinate chart which covers all

of R4 and in which the components of the metric are diag(1, 1, 1, 1)
everywhere is called an inertial frame.

3. On S 2 , let (, ) denote the spherical polar coordinate chart discussed earlier.

The (unit) round metric on S 2 is

ds2 = d2 + sin2 d2 , (3.7)

i.e. in the chart (, ), we have g = diag(1, sin2 ). Note this is positive

definite for (0, ), i.e., on all of this chart. However, this chart does
not cover the whole manifold so the above equation does not determine g
everywhere. We can give a precise definition by adding that, in the chart
(0 , 0 ) discussed earlier, g = d0 2 + sin2 0 d0 2 . One can check that this does
indeed define a smooth tensor field. (This metric is the one induced from the
embedding of S 2 into 3d Euclidean space: it is the pull-back of the metric
on Euclidean space - see later for the definition of pull-back.)

Definition. Since gab is non-degenerate, it must be invertible. The inverse metric

is a symmetric (2, 0) tensor field denoted g ab and obeys

g ab gbc = ca (3.8)

Example. For the metric on S 2 defined above, in the chart (, ) we have g =

diag(1, 1/ sin2 ).

Definition. A metric determines a natural isomorphism between vectors and

covectors. Given a vector X a we can define a covector Xa = gab X b . Given a
covector a we can define a vector a = g ab b . These maps are clearly inverses of
each other.

Part 3 GR December 14, 2016 33 H.S. Reall

CHAPTER 3. THE METRIC TENSOR

Remark. This isomorphism is the reason why covectors are not more familiar:
we are used to working in Euclidean space using Cartesian coordinates, for which
g and g are both the identity matrix, so the isomorphism appears trivial.
Definition. For a general tensor, abstract indices can be lowered by contracting
with gab and raised by contracting with g ab . Raising and lowering preserve the
ordering of indices. The resulting tensor will be denoted by the same letter as the
original tensor.
Example. Let T be a (3, 2) tensor. Then T a b cde = gbf g dh g ej T af c hj .

3.2 Lorentzian signature

Remark. On a Lorentzian manifold, we take basis indices , , . . . to run from 0
to n 1.
At any point p of a Lorentzian manifold, we can choose an orthonormal basis
{e } so that g(e , e ) = diag(1, 1, . . . , 1). Such a basis is far from unique.
If e0 = (A1 ) e is any other such basis then we have

= g(e0 , e0 ) = (A1 ) (A1 ) g(e , e ) = (A1 ) (A1 ) . (3.9)

Hence
A A = . (3.10)
These are the defining equations of a Lorentz transformation in special relativity.
Hence different orthonormal bases at p are related by Lorentz transformations. We
saw earlier that the components of a vector at p transform as X 0 = A X . We
are starting to recover the structure of special relativity locally, as required by the
Equivalence Principle.
Definition. On a Lorentzian manifold (M, g), a non-zero vector X Tp (M )
is timelike if g(X, X) < 0, null (or lightlike) if g(X, X) = 0, and spacelike if
g(X, X) > 0.
Remark. In an orthonormal basis at p, the metric has components so the
tangent space at p has exactly the same structure as Minkowski spacetime, i.e.,
null vectors at p define a light cone that separates timelike vectors at p from
spacelike vectors at p (see Fig. 3.1).
Exercise. Let X a , Y b be non-zero vectors at p that are orthogonal, i.e., gab X a Y b =
0. Show that (i) if X a is timelike then Y a is spacelike; (ii) if X a is null then Y a is
spacelike or null; (iii) if X a is spacelike then Y a can be spacelike, timelike, or null.
(Hint. Choose an orthonormal basis to make the components of X a as simple as
possible.)

Part 3 GR December 14, 2016 34 H.S. Reall

 3.2. LORENTZIAN SIGNATURE

timelike
null

spacelike

Figure 3.1: Light cone structure of Tp (M )

Definition.
p On a Riemannian manifold, the norm of a vector X is |X| =
g(X, X) and the angle between two non-zero vectors X and Y (at the same
point) is where cos = g(X, Y )/(|X| |Y |).
Definition. A curve in a Lorentzian manifold is said to be timelike if its tangent
vector is everywhere timelike. Null and spacelike curves are defined similarly.
(Most curves do not satisfy any of these definitions because e.g. the tangent
vector can change from timelike to null to spacelike along a curve.)
Remark. The length of a spacelike curve can be defined in exactly the same way
as on a Riemannian manifold (equation (3.2)). What about a timelike curve?
Definition. let (u) be a timelike curve with (0) = p. Let X a be the tangent to
the curve. The proper time from p along the curve is defined by
d q
= (gab X a X b )(u) , (0) = 0. (3.11)
du
Remark. In a coordinate chart, X = dx /du so this definition can be rewritten
in the form
d 2 = g dx dx , (3.12)
with the understanding that this is to be evaluated along the curve. Integrating
the above equation along the curve gives the proper time from p to some other
point q = (uq ) as s 
Z uq 
dx dx
= du g (3.13)
0 du du (u)

Definition. If proper time is used to parametrize a timelike curve then the

tangent to the curve is called the 4-velocity of the curve. In a coordinate basis, it
has components u = dx /d .

Part 3 GR December 14, 2016 35 H.S. Reall

CHAPTER 3. THE METRIC TENSOR

Remark. (3.12) implies that 4-velocity is a unit timelike vector:

gab ua ub = 1. (3.14)

3.3 Curves of extremal proper time

Consider the following question. Let p and q be points connected by a timelike
curve. A small deformation of a timelike curve remains timelike hence there exist
infinitely many timelike curves connecting p and q. The proper time between p
and q will be different for different curves. Which curve extremizes the proper
time between p and q?
This is a standard Euler-Lagrange problem. Consider timelike curves from p
to q with parameter u such that (0) = p, (1) = q. Lets use a dot to denote a
derivative with respect to u. The proper time between p and q along such a curve
is given by the functional
Z 1
[] = du G (x(u), x(u)) (3.15)
0

where q
G (x(u), x(u)) g (x(u))x (u)x (u) (3.16)
and we are writing x (u) as a shorthand for x ((u)).
The curve that extremizes the proper time, must satisfy the Euler-Lagrange
equation  
d G G

=0 (3.17)
du x x
Working out the various terms, we have (using the symmetry of the metric)
G 1 1
= 2g x = g x (3.18)
x 2G G
G 1
= g, x x (3.19)
x 2G
where we have relabelled some dummy indices, and introduced the important
notation of a comma to denote partial differentiation:

g, g (3.20)
x
We will be using this notation a lot henceforth.
So far, our parameter u has been arbitrary subject to the conditions u(0) = p
and u(1) = q. At this stage, it is convenient to use a more physical parameter,

Part 3 GR December 14, 2016 36 H.S. Reall

 3.3. CURVES OF EXTREMAL PROPER TIME

namely , the proper time along the curve. (Note that we could not have used
from the outset since the value of at q is different for different curves, which
would make the range of integration different for different curves.) The paramers
are related by
 2
d
= g x x = G2 (3.21)
du
and hence d /du = G. So in our equations above, we can replace d/du with
Gd/d , so the Euler-Lagrange equation becomes (after cancelling a factor of G)

dx dx dx
 
d 1
g g, =0 (3.22)
d d 2 d d

Hence
d 2 x dx dx 1 dx dx
g + g, g , =0 (3.23)
d 2 d d 2 d d
In the second term, we can replace g, with g(,) because it is contracted with
an object symmetrical on and . Finally, contracting the whole expression with
the inverse metric and relabelling indices gives

d 2 x
dx dx

+ =0 (3.24)
d 2 d d
where are known as the Christoffel symbols, and are defined by

1
= g (g, + g, g, ) . (3.25)
2
Remarks. 1. = . 2. The Christoffel symbols are not tensor components.
Neither the first term nor the second term in (3.24) are components of a vector
but the sum of these two terms does give vector components. More about this
soon. 3. Equation 3.24 is called the geodesic equation. Geodesics will be defined
below.
Example. In Minkowski spacetime, the components of the metric in an inertial
frame are constant so = 0. Hence the above equation reduces to d2 x /d 2 = 0.
This is the equation of motion of a free particle! Hence, in Minkowski spacetime,
the free particle trajectory between two (timelike separated) points p and q ex-
tremizes the proper time between p and q.
This motivates the following postulate of General Relativity:
Postulate. Massive test bodies follow curves of extremal proper time, i.e., solu-
tions of equation (3.24).

Part 3 GR December 14, 2016 37 H.S. Reall

CHAPTER 3. THE METRIC TENSOR

Remarks. 1. Massless particles obey a very similar equation which we shall dis-
cuss shortly. 2. In Minkowski spacetime, curves of extremal proper time maximize
the proper time between two points. In a curved spacetime, this is true only locally,
i.e., for any point p there exists a neighbourhood of p within which it is true.
Exercises
1. Show that (3.24) can be obtained more directly as the Euler-Lagrange equa-
tion for the Lagrangian
dx dx
L = g (x( )) (3.26)
d d
This is usually the easiest way to derive (3.24) or to calculate the Christoffel
symbols.
2. Note that L has no explicit dependence, i.e., L/ = 0. Show that this
implies that the following quantity is conserved along curves of extremal
proper time (i.e. that is is annihilated by d/d ):
L dx dx dx
L = g (3.27)
(dx /d ) d d d
This is a check on the consistency of (3.24) because the definition of as
proper time implies that the RHS must be 1.

Example. The Schwarzschild metric in Schwarzschild coordinates (t, r, , ) is

2M
ds2 = f dt2 + f 1 dr2 + r2 d2 + r2 sin2 d2 , f =1 (3.28)
r
where M is a constant. We have
 2  2  2  2
dt 1 dr 2 d 2 2 d
L=f f r r sin (3.29)
d d d d
so the EL equation for t( ) is
d2 t
 
d dt dt dr
2f =0 2
+ f 1 f 0 =0 (3.30)
d d d d d
From this we can read off
f0
001 = 010 = , 0 = 0 otherwise (3.31)
2f
The other Christoffel symbols are obtained in a similar way from the remaining
EL equations (examples sheet 1).

Part 3 GR December 14, 2016 38 H.S. Reall

Chapter 4

Covariant derivative

4.1 Introduction
To formulate physical laws, we need to be able to differentiate tensor fields. For
scalar fields, partial differentiation is fine: f, f /x are the components of the
covector field (df )a . However, for tensor fields, partial differentiation is no good
because the partial derivative of a tensor field does not give another tensor field:
Exercise. Let V a be a vector field. In any coordinate chart, let T = V ,
V /x . Show that T do not transform as tensor components under a change
of chart.
The problem is that differentiation involves comparing a tensor at two infinites-
imally nearby points of the manifold. But we have seen that this does not make
sense: tensors at different points belong to different spaces. The mathematical
structure that overcomes this difficulty is called a covariant derivative or connec-
tion.
Definition. A covariant derivative on a manifold M is a map sending every
pair of smooth vector fields X, Y to a smooth vector field X Y , with the following
properties (where X, Y, Z are vector fields and f, g are functions)

f X+gY Z = f X Z + gY Z, (4.1)

X (Y + Z) = X Y + X Z, (4.2)
X (f Y ) = f X Y + (X f )Y, (Leibniz rule), (4.3)
where the action of on functions is defined by

X f = X(f ). (4.4)

Remark. (4.1) implies that, at any point, the map Y : X 7 X Y is a linear

39
CHAPTER 4. COVARIANT DERIVATIVE

map from Tp (M ) to itself. Hence it defines a (1, 1) tensor (see examples sheet
1). More precisely, if Tp (M ) and X Tp (M ) then we define (Y )(, X)
(X Y ).
Definition. let Y be a vector field. The covariant derivative of Y is the (1, 1)
tensor field Y . In abstract index notation we usually write (Y )a b as b Y a or
Y a ;b
Remarks.

1. Similarly we define f : X 7 X f = X(f ). Hence f = df . We can write

this as either a f or f;a or a f or f,a (i.e. the covariant derivative reduces
to the partial derivative when acting on a function).

2. Does the map : X, Y 7 X Y define a (1, 2) tensor field? No - equation

(4.3) shows that this map is not linear in Y .

Example. Pick a coordinate chart on M . Let be the partial derivative in this

chart. This satisfies all of the above conditions. This is not a very interesting
example of a covariant derivative because it depends on choosing a particular
chart: if we use a different chart then this covariant derivative will not be the
partial derivative in the new chart.
Definition. In a basis {e } the connection components are defined by

e e e = e (4.5)

Example. The Christoffel symbols are the coordinate basis components of a

certain connection, the Levi-Civita connection, which is defined on any manifold
with a metric. More about this soon.
Write X = X e and Y = Y e . Now

X Y = X (Y e ) = X(Y )e + Y X e (Leibniz)

= X e (Y )e + Y X e e
= X e (Y )e + Y X e by (4.1)

= X e (Y )e + Y X e
= X e (Y ) + Y e


(4.6)

and hence
(X Y ) = X e (Y ) + Y X (4.7)
so
Y ; = e (Y ) + Y (4.8)

Part 3 GR December 14, 2016 40 H.S. Reall

 4.1. INTRODUCTION

In a coordinate basis, this reduces to

Y ; = Y , + Y (4.9)

The connection components are not tensor components:

Exercise (examples sheet 2). Consider a change of basis e0 = (A1 ) e . Show
that

0 = A (A1 ) (A1 ) + A (A1 ) e ((A1 ) ) (4.10)
The presence of the second term demonstrates that are not tensor components.
Hence neither term in the RHS of equation (4.9) transforms as a tensor. However,
the sum of these two terms does transform as a tensor.
Exercise. Let and be two different connections on M . Show that is
a (1, 2) tensor field. You can do this either from the definition of a connection, or
from the transformation law for the connection components.
The action of is extended to general tensor fields by the Leibniz property.
If T is a tensor field of type (r, s) then T is a tensor field of type (r, s + 1). For
example, if is a covector field then, for any vector fields X and Y , we define

(X )(Y ) X ((Y )) (X Y ). (4.11)

It is not obvious that this defines a (0, 2) tensor but we can see this as follows:

(X )(Y ) = X ( Y ) (X Y )
= X( )Y + X(Y ) X e (Y ) + Y X , (4.12)

where we used (4.7). Now, the second and third terms cancel (X = X e ) and
hence (renaming dummy indices in the final term)

(X )(Y ) = X( ) X Y ,


(4.13)

which is linear in Y so X is a covector field with components

(X ) = X( ) X
= X e ( )


(4.14)

This is linear in X and hence is a (0, 2) tensor field with components

; = e ( ) (4.15)

In a coordinate basis, this is

; = , (4.16)

Part 3 GR December 14, 2016 41 H.S. Reall

CHAPTER 4. COVARIANT DERIVATIVE

Now the Leibniz rule can be used to obtain the formula for the coordinate basis
components of T where T is a (r, s) tensor:

T 1 ...r 1 ...s ; = T 1 ...r 1 ...s , + 1 T 2 ...r 1 ...s + . . . + r T 1 ...r1 1 ...s

1 T 1 ...r 2 ...s . . . s T 1 ...r 1 ...s1 (4.17)

Exercise. Prove this result for a (1, 1) tensor.

Remark. We are using a comma and semi-colon to denote partial, and covariant,
derivatives respectively. If more than one index appears after a comma or semi-
colon then the derivative is to be taken with respect to all indices. The index
nearest to comma/semi-colon is the first derivative to be taken. For example,
f, = f,, f , and X a ;bc = c b X a (we cannot use abstract indices for the
first example since it is not a tensor). The second partial derivatives of a function
commute: f, = f, but for a covariant derivative this is not true in general. Set
= df in (4.16) to get, in a coordinate basis,

f; = f, f, (4.18)

Antisymmetrizing gives

f;[] = [] f, (coordinate basis) (4.19)

Definition. A connection is torsion-free if a b f = b a f for any function

f . From (4.19), this is equivalent to

[] = 0 (coordinate basis) (4.20)

Lemma. For a torsion-free connection, if X and Y are vector fields then

X Y Y X = [X, Y ] (4.21)

Proof. Use a coordinate basis:

X Y ; Y X ; = X Y , + X Y Y X , Y X
= [X, Y ] + 2[] X Y
= [X, Y ] (4.22)

Hence the equation is true in a coordinate basis and therefore (as it is a tensor
equation) it is true in any basis.
Remark. Even with zero torsion, the second covariant derivatives of a tensor field
do not commute. More soon.

Part 3 GR December 14, 2016 42 H.S. Reall

 4.2. THE LEVI-CIVITA CONNECTION

4.2 The Levi-Civita connection

On a manifold with a metric, the metric singles out a preferred connection:
Theorem. Let M be a manifold with a metric g. There exists a unique torsion-free
connection such that the metric is covariantly constant: g = 0 (i.e. gab;c = 0).
This is called the Levi-Civita (or metric) connection.
Proof. Let X, Y, Z be vector fields then

X(g(Y, Z)) = X (g(Y, Z)) = g(X Y, Z) + g(Y, X Z), (4.23)

where we used the Leibniz rule and X g = 0 in the second equality. Permuting
X, Y, Z leads to two similar identities:

Y (g(Z, X)) = g(Y Z, X) + g(Z, Y X), (4.24)

Z(g(X, Y )) = g(Z X, Y ) + g(X, Z Y ), (4.25)

Add the first two of these equations and subtract the third to get (using the
symmetry of the metric)

X(g(Y, Z)) + Y (g(Z, X)) Z(g(X, Y )) = g(X Y + Y X, Z)

g(Z X X Z, Y )
+ g(Y Z Z Y, X) (4.26)

The torsion-free condition implies

X Y Y X = [X, Y ] (4.27)

Using this and the same identity with X, Y, Z permuted gives

X(g(Y, Z)) + Y (g(Z, X)) Z(g(X, Y )) = 2g(X Y, Z) g([X, Y ], Z)

g([Z, X], Y ) + g([Y, Z], X)
(4.28)

Hence
1
g(X Y, Z) = [X(g(Y, Z)) + Y (g(Z, X)) Z(g(X, Y ))
2
+ g([X, Y ], Z) + g([Z, X], Y ) g([Y, Z], X)] (4.29)

This determines X Y uniquely because the metric is non-degenerate. It remains

to check that it satisfies the properties of a connection. For example:
1
g(f X Y, Z) = [f X(g(Y, Z)) + Y (f g(Z, X)) Z(f g(X, Y ))
2
Part 3 GR December 14, 2016 43 H.S. Reall
CHAPTER 4. COVARIANT DERIVATIVE

+ g([f X, Y ], Z) + g([Z, f X], Y ) f g([Y, Z], X)]

1
= [f X(g(Y, Z)) + f Y (g(Z, X)) + Y (f )g(Z, X)
2
f Z(g(X, Y )) Z(f )g(X, Y ) + f g([X, Y ], Z) Y (f )g(X, Z)
+ f g([Z, X], Y ) + Z(f )g(X, Y ) f g([Y, Z], X)]
f
= [X(g(Y, Z)) + Y (g(Z, X)) Z(g(X, Y ))
2
+ g([X, Y ], Z) + g([Z, X], Y ) g([Y, Z], X)]
= f g(X Y, Z) = g(f X Y, Z) (4.30)

and hence g(f X Y f X Y, Z) = 0 for any vector field Z so, by the non-degeneracy
of the metric, f X Y = f X Y .
Exercise. Show that X Y as defined by (4.29) satisfies the other properties
required of a connection.
Remark. In differential geometry, this theorem is called the fundamental theorem
of Riemannian geometry (although it applies for a metric of any signature).
Lets determine the components of the Levi-Civita connection in a coordinate
basis (for which [e , e ] = 0):

1
g( e , e ) = [e (g ) + e (g ) e (g )] , (4.31)
2
that is
1
g( e , e ) = (g, + g, g, ) (4.32)
2
The LHS is just g . Hence if we multiply the whole equation by the inverse
metric g we obtain

1
= g (g, + g, g, ) (4.33)
2
This is the same equation as we obtained earlier; we have now shown that the
Christoffel symbols are the components of the Levi-Civita connection.
Remark. In GR, we take the connection to be the Levi-Civita connection. This
is not as restrictive as it sounds: we saw above that the difference between two
connections is a tensor field. Hence we can write any connection (even one with
torsion) in terms of the Levi-Civita connection and a (1, 2) tensor field. In GR we
could regard the latter as a particular kind of matter field, rather than as part
of the geometry of spacetime.

Part 3 GR December 14, 2016 44 H.S. Reall

 4.3. GEODESICS

4.3 Geodesics
Previously we considered curves that extremize the proper time between two points
of a spacetime, and showed that this gives the equation
d 2 x dx dx
+ (x( )) = 0, (4.34)
d 2 d d
where is the proper time along the curve. The tangent vector to the curve has
components X = dx /d . This is defined only along the curve. However, we
can extend X (in an arbitrary way) to a neighbourhood of the curve, so that X
becomes a vector field, and the curve is an integral curve of this vector field. The
chain rule gives
d2 x dX (x( )) dx X
2
= =
= X X , . (4.35)
d d d x
Note that the LHS is independent of how we extend X hence so must be the
RHS. We can now write (4.34) as
X X , + X = 0


(4.36)
which is the same as
X X ; = 0, or X X = 0. (4.37)
where we are using the Levi-Civita connection. We now extend this to an arbitrary
connection:
Definition. Let M be a manifold with a connection . An affinely parameterized
geodesic is an integral curve of a vector field X satisfying X X = 0.
Remarks.
1. What do we mean by affinely parameterized? Consider a curve with pa-
rameter t whose tangent X satisfies the above definition. Let u be some
other parameter for the curve, so t = t(u) and dt/du > 0. Then the tangent
vector becomes Y = hX where h = dt/du. Hence
Y Y = hX (hX) = hX (hX) = h2 X X + X(h)hX = f Y, (4.38)
where f = X(h) = dh/dt. Hence Y Y = f Y describes the same geodesic.
In this case, the geodesic is not affinely parameterized.
It always is possible to find an affine parameter so there is no loss of gen-
erality in restricting to affinely parameterized geodesics. Note that the new
parameter also is affine iff X(h) = 0, i.e., h is constant. Then u = at+b where
a and b are constants with a > 0 (a = h1 ). Hence there is a 2-parameter
family of affine parameters for any geodesic.

Part 3 GR December 14, 2016 45 H.S. Reall

CHAPTER 4. COVARIANT DERIVATIVE

2. Reversing the above steps shows that, in a coordinate chart, for any con-
nection, the geodesic equation can be written as (4.34) with an arbitrary
affine parameter.

3. In GR, curves of extremal proper time are timelike geodesics (with the
Levi-Civita connection). But one can also consider geodesics which are not
timelike. These satisfy (4.34) with an affine parameter. The easiest way
to obtain this equation is to use the Lagrangian (3.26).

Theorem. Let M be a manifold with a connection . Let p M and Xp

Tp (M ). Then there exists a unique affinely parameterized geodesic through p with
tangent vector Xp at p.
Proof. Choose a coordinate chart x in a neighbourhood of p. Consider a curve
parameterized by . It has tangent vector with components X = dx /d . The
geodesic equation is (4.34). We want the curve to satisfy the initial conditions

dx
 

x (0) = xp , = Xp . (4.39)
d =0

This is a coupled system of n ordinary differential equations for the n functions

x (t). Existence and uniqueness is guaranteed by the standard theory of ordinary
differential equations.

Exercise. Let X be tangent to an affinely parameterized geodesic of the Levi-

Civita connection. Show that X (g(X, X)) = 0 and hence g(X, X) is constant
along the geodesic. Therefore the tangent vector cannot change e.g. from timelike
to null along the geodesic: a geodesic is either timelike, spacelike or null.

Postulate. In GR, free particles move on geodesics (of the Levi-Civita connec-
tion). These are timelike for massive particles, and null for massless particles (e.g.
photons).

Remark. In the timelike case we can use proper time as an affine parameter.
This imposes the additional restriction g(X, X) = 1. If and 0 both are proper
times along a geodesic then 0 = + b (i.e. a = 1 above). In other words, clocks
measuring proper time differ only by their choice of zero. In particular, they
measure equal time intervals. Similarly in the spacelike case (or on a Riemannian
manifold), we use arc length s as affine parameter, which gives g(X, X) = 1 and
s0 = s + b. In the null case, there is no analogue of proper time or arc length and
so there is a 2-parameter ambiguity in affine parameterization.

Part 3 GR December 14, 2016 46 H.S. Reall

 4.4. NORMAL COORDINATES

4.4 Normal coordinates

Definition. Let M be a manifold with a connection . Let p M . The expo-
nential map from Tp (M ) to M is defined as the map which sends Xp to the point
unit affine parameter distance along the geodesic through p with tangent Xp at p.
Remark. It can be shown that this map is one-to-one and onto locally, i.e., for
Xp in a neighbourhood of the origin in Tp (M ).
Exercise. Let 0 t 1. Show that the exponential map sends tXp to the point
affine parameter distance t along the geodesic through p with tangent Xp at p.
Definition. Let {e } be a basis for Tp (M ). Normal coordinates at p are defined
in a neighbourhood of p as follows. Pick q near p. Then the coordinates of q are
X where X a is the element of Tp (M ) that maps to q under the exponential map.
Lemma. () (p) = 0 in normal coordinates at p. For a torsion-free connection,
(p) = 0 in normal coordinates at p.
Proof. From the above exercise, it follows that affinely parameterized geodesics
through p are given in normal coordinates by X (t) = tXp . Hence the geodesic
equation reduces to
(X(t))Xp Xp = 0. (4.40)
Evaluating at t = 0 gives that (p)Xp Xp = 0. But Xp is arbitrary, so the first
result follows. The second result follows using the fact that torsion-free implies
[] = 0 in a coordinate chart.
Remark. The connection components away from p will not vanish in general.
Lemma. On a manifold with a metric, if the Levi-Civita connection is used to
define normal coordinates at p then g, = 0 at p.
Proof. Apply the previous lemma. We then have, at p,

0 = 2g = g, + g, g, (4.41)

Now symmetrize on : the final two terms cancel and the result follows.
Remark. Again, we emphasize, this is valid only at the point p. At any point,
we can introduce normal coordinates to make the first partial derivatives of the
metric vanish at that point. They will not vanish away for that point.
Lemma. On a manifold with metric one can choose normal coordinates at p
so that g, (p) = 0 and also g (p) = (Lorentzian case) or g (p) =
(Riemannian case).
Proof. Weve already shown g, (p) = 0. Consider /X 1 . The integral curve
through p of this vector field is X (t) = (t, 0, 0, . . . , 0) (since X = 0 at p). But,

Part 3 GR December 14, 2016 47 H.S. Reall

CHAPTER 4. COVARIANT DERIVATIVE

from the above, this is the same as the geodesic through p with tangent vector e1
at p. It follows that /X 1 = e1 at p (since both vectors are tangent to the curve
at p). Similarly /X = e at p. But the choice of basis {e } was arbitrary. So
we are free to choose {e } to be an orthonormal basis. /X then defines an
orthonormal basis at p too.
In summary, on a Lorentzian (Riemannian) manifold, we can choose coordi-
nates in the neighbourhood of any point p so that the components of the metric at
p are the same as those of the Minkowski metric in inertial coordinates (Euclidean
metric in Cartesian coordinates), and the first partial derivatives of the metric
vanish at p.
Definition. In a Lorentzian manifold a local inertial frame at p is a set of normal
coordinates at p with the above properties.
Thus the assumption that spacetime is a Lorentzian manifold leads to a precise
mathematical definition of a local inertial frame.

Part 3 GR December 14, 2016 48 H.S. Reall

Chapter 5

Physical laws in curved spacetime

5.1 Minimal coupling, equivalence principle

Physical laws in curved spacetime should exhibit general covariance: they should
be independent of any choice of basis or coordinate chart. In special relativity,
we restrict attention to coordinate systems corresponding to inertial frames. The
laws of physics should exhibit special covariance, i.e, take the same form in any
inertial frame (this is the principle of relativity). The following procedure can be
used to convert such laws of physics into generally covariant laws:
1. Replace the Minkowski metric by a curved spacetime metric.
2. Replace partial derivatives with covariant derivatives (associated to the Levi-
Civita connection). This rule is called minimal coupling in analogy with a
similar rule for charged fields in electrodynamics.
3. Replace coordinate basis indices , etc (referring to an inertial frame) with
abstract indices a, b etc.
Examples. Let x denote the coordinates of an inertial frame, and the inverse
Minkowski metric (which has the same components as ).
1. The simplest Lorentz invariant field equation is the wave equation for a scalar
field
= 0. (5.1)
Follow the rules above to obtain the wave equation in a general spacetime:
g ab a b = 0, or a a = 0 or ;a a = 0. (5.2)
A simple generalization of this equation is the Klein-Gordon equation de-
scribing a scalar field of mass m:
a a m2 = 0. (5.3)

49
CHAPTER 5. PHYSICAL LAWS IN CURVED SPACETIME

2. In special relativity, the electric and magnetic fields are combined into an
antisymmetric tensor F . The electric and magnetic fields in an inertial
frame are obtained by the rule (i, j, k take values from 1 to 3) F0i = Ei and
Fij = ijk Bk . The (source-free) Maxwell equations take the covariant form

F = 0, [ F] = 0. (5.4)

Hence in a curved spacetime, the electromagnetic field is described by an

antisymmetric tensor Fab satisfying

g ab a Fbc = 0, [a Fbc] = 0. (5.5)

The Lorentz force law for a particle of charge q and mass m in Minkowski
spacetime is
d2 x q dx
= F (5.6)
d 2 m d
where is proper time. We saw previously that the LHS can be rewritten as
u u where u = dx /d is the 4-velocity. Now following the rules above
gives the generally covariant equation
q ab q
ub b ua = g Fbc uc = F a b ub . (5.7)
m m
Note that this reduces to the geodesic equation when q = 0.

Remark. The rules above ensure that we obtain generally covariant equations.
But how do we know they are the right equations? The Einstein equivalence
principle states that, in a local inertial frame, the laws of physics should take the
same form as in an inertial frame in Minkowski spacetime. But we saw above, that
in a local inertial frame at p, (p) = 0 and hence (first) covariant derivatives
reduce to partial derivatives at p. For example, = g (in any
chart) and, at p, this reduces to in a local inertial frame at p (since
the metric at p is ). Hence all of our generally covariant equations reduce
to the equations of special relativity in a local inertial frame at any given point.
The Einstein equivalence principle is satisfied automatically if we use the above
rules. Nevertheless, there is still some scope for ambiguity, which arises from the
possibility of including terms in an equation involving the curvature of spacetime
(see later). These vanish identically in Minkowski spacetime. Sometimes, such
terms are fixed by mathematical consistency. However, this is not always possible:
there is no reason why it should be possible to derive laws of physics in curved
spacetime from those in flat spacetime. The ultimate test is comparison with
observations.

Part 3 GR December 14, 2016 50 H.S. Reall

 5.2. ENERGY-MOMENTUM TENSOR

5.2 Energy-momentum tensor

in GR, the curvature of spacetime is related to the energy and momentum of
matter. So we need to discuss how the latter concepts are defined in GR. We shall
start by discussing the energy and momentum of particles.
In special relativity, associated to any particle is a scalar called its rest mass
(or simply its mass) m. If the particle has 4-velocity u (again x denote inertial
frame coordinates) then its 4-momentum is

P = mu (5.8)

The time component of P is the particles energy and the spatial components are
its 3-momentum with respect to the inertial frame.
If an observer at some point p has 4-velocity v (p) then he measures the par-
ticles energy, when the particle is at q, to be

E = v (p)P (q). (5.9)

The way to see this is to choose an inertial frame in which, at p, the observer is at
rest at the origin, so v (p) = (1, 0, 0, 0) so E is just the time component of P (q)
in this inertial frame.
By the equivalence principle, GR should reduce to SR in a local inertial frame.
Hence in GR we also associate a rest mass m to any particle and define the 4-
momentum of a particle with 4-velocity ua as

P a = mua (5.10)

Note that
gab P a P b = m2 (5.11)
The EP implies that when the observer and particle both are at p then (5.9) should
be valid so the observer measures the particles energy to be

E = gab (p)v a (p)P b (p) (5.12)

However, an important difference between GR and SR is that there is no analogue

of equation (5.9) for p 6= q. This is because v a (p) and P a (q) are vectors defined
at different points, so they live in different tangent spaces. There is no way they
can be combined to give a scalar quantity. An observer at p cannot measure the
energy of a particle at q.
Now lets consider the energy and momentum of continuous distributions of
matter.

Part 3 GR December 14, 2016 51 H.S. Reall

CHAPTER 5. PHYSICAL LAWS IN CURVED SPACETIME

Example. Consider Maxwell theory (without sources) in Minkowski spacetime.

Pick an inertial frame and work in pre-relativity notation using Cartesian tensors.
The electromagnetic field has energy density
1
E= (Ei Ei + Bi Bi ) (5.13)
8
and the momentum density (or energy flux density) is given by the Poynting vector:
1
Si = ijk Ej Bk . (5.14)
4
The Maxwell equations imply that these satisfy the conservation law
E
+ i Si = 0. (5.15)
t
The momentum flux density is described by the stress tensor:
 
1 1
tij = (Ek Ek + Bk Bk ) ij Ei Ej Bi Bj , (5.16)
4 2

with the conservation law

Si
+ j tij = 0. (5.17)
t
If a surface element has area dA and normal ni then the force exerted on this
surface by the electromagnetic field is tij nj dA.
In special relativity, these three objects are combined into a single tensor, called
variously the energy-momentum tensor, the stress tensor, the stress-energy-
momentum tensor etc. In an inertial frame it is
 
1 1
T = F F F F (5.18)
4 4

where weve raised indices with . Note that this is a symmetric tensor. It
has components T00 = E, T0i = Si , Tij = tij . The conservation laws above are
equivalent to the single equation

T = 0. (5.19)

The definition of the energy-momentum tensor extends naturally to GR:

Definition. The energy-momentum tensor of a Maxwell field in a general space-
time is  
1 c 1 cd
Tab = Fac Fb F Fcd gab (5.20)
4 4

Part 3 GR December 14, 2016 52 H.S. Reall

 5.2. ENERGY-MOMENTUM TENSOR

Exercise (examples sheet 2). Show that Maxwells equations imply that

a Tab = 0. (5.21)

In GR (and SR) we assume that continuous matter always is described by a con-

served energy-momentum tensor:
Postulate. The energy, momentum, and stresses, of matter are described by an
energy-momentum tensor, a (0, 2) symmetric tensor Tab that is conserved: a Tab =
0.
Remark. Let ua be the 4-velocity of an observer O at p. Consider a local inertial
frame (LIF) at p in which O is at rest. Choose an orthonormal basis at p {e }
aligned with the coordinate axes of this LIF. In such a basis, ea0 = ua . Denote
the spatial basis vectors as eai , i = 1, 2, 3. From the Einstein equivalence principle,
E T00 = Tab ea0 eb0 = Tab ua ub is the energy density of matter at p measured by
O. Similarly, Si T0i is the momentum density and tij Tij the stress tensor
measured by O. The energy-momentum current measured by O is the 4-vector
j a = T a b ub , which has components (E, Si ) in this basis.
Remark. In an inertial frame x in Minkowski spacetime, local conservation of
Tab is equivalent to equations of the form (5.15) and (5.17). If one integrates
these over a fixed volume V in surfaces of constant t = x0 then one obtains global
conservation equations. For example, integrating (5.15) over V gives
Z Z
d
E = S ndA (5.22)
dt V S

where the surface S (with outward unit normal n) bounds V . In words: the rate
of increase of the energy of matter in V is equal to minus the energy flux across
S. In a general curved spacetime, such an interpretation is not possible. This is
because the gravitational field can do work on the matter in the spacetime. One
might think that one could obtain global conservation laws in curved spacetime by
introducing a definition of energy density etc for the gravitational field. This is a
subtle issue. The gravitational field is described by the metric gab . In Newtonian
theory, the energy density of the gravitational field is (1/8)()2 so one might
expect that in GR the energy density of the gravitational field should be some
expression quadratic in first derivatives of gab . But we have seen that we can
choose normal coordinates to make the first partial derivatives of gab vanish at
any given point. Gravitational energy certainly exists but not in a local sense.
For example one can define the total energy (i.e. the energy of matter and the
gravitational field) for certain spacetimes (this will be discussed in the black holes
course).

Part 3 GR December 14, 2016 53 H.S. Reall

CHAPTER 5. PHYSICAL LAWS IN CURVED SPACETIME

Example. A perfect fluid is described by a 4-velocity vector field ua , and two

scalar fields and p. The energy-momentum tensor is

Tab = ( + p)ua ub + pgab (5.23)

and p are the energy density and pressure measured by an observer co-moving
with the fluid, i.e., one with 4-velocity ua (check: Tab ua ub = + p p = ). The
equations of motion of the fluid can be derived by conservation of Tab :
Exercise (examples sheet 2). Show that, for a perfect fluid, a Tab = 0 is equiva-
lent to

ua a + ( + p)a ua = 0, ( + p)ub b ua = (gab + ua ub )b p (5.24)

These are relativistic generalizations of the mass conservation equation and Euler
equation of non-relativistic fluid dynamics. Note that a pressureless fluid moves
on timelike geodesics. This makes sense physically: zero pressure implies that the
fluid particles are non-interacting and hence behave like free particles.

Part 3 GR December 14, 2016 54 H.S. Reall

Chapter 6

Curvature

6.1 Parallel transport

On a general manifold there is no way of comparing tensors at different points. For
example, we cant say whether a vector at p is the same as a vector at q. However,
with a connection we can define a notion of a tensor that doesnt change along a
curve:
Definition. Let X a be the tangent to a curve. A tensor field T is parallelly
transported along the curve if X T = 0.
Remarks.

1. Sometimes we say parallelly propagated instead of parallelly transported.

2. A geodesic is a curve whose tangent vector is parallelly transported along

the curve.

3. Let p be a point on a curve. If we specify T at p then the above equation

determines T uniquely everywhere along the curve. For example, consider
a (1, 1) tensor. Introduce a chart in a neighbourhood of p. Let t be the
parameter along the curve. In a coordinate chart, X = dx /dt so X T = 0
gives

0 = X T ; = X T , + T X T X
dT
= + T X T X (6.1)
dt
Standard ODE theory guarantees a unique solution given initial values for
the components T .

55
CHAPTER 6. CURVATURE

4. If q is some other point on the curve then parallel transport along a curve
from p to q determines an isomorphism between tensors at p and tensors at
q.

Consider Euclidean space or Minkowski spacetime with the Levi-Civita connection,

and use Cartesian/inertial frame coordinates so the Christoffel symbols vanish ev-
erywhere. Then a tensor is parallelly transported along a curve iff its components
are constant along the curve. Hence if we have two different curves from p to q then
the result of parallelly transporting T from p to q is independent of which curve
we choose. However, in a general spacetime this is no longer true: parallel trans-
port is path-dependent. The path-dependence of parallel transport is measured
by the Riemann curvature tensor. For Euclidean space or Minkowski spacetime,
the Riemann tensor (of the Levi-Civita connection) vanishes and we say that the
spacetime is flat.

6.2 The Riemann tensor

We shall return to the path-dependence of parallel transport below. First we define
the Riemann tensor is as follows:
Definition. The Riemann curvature tensor Ra bcd of a connection is defined by
Ra bcd Z b X c Y d = (R(X, Y )Z)a , where X, Y, Z are vector fields and R(X, Y )Z is the
vector field
R(X, Y )Z = X Y Z Y X Z [X,Y ] Z (6.2)
To demonstrate that this defines a tensor, we need to show that it is linear in
X, Y, Z. The symmetry R(X, Y )Z = R(Y, X)Z implies that we need only check
linearity in X and Z. The non-trivial part is to check what happens if we multiply
X or Z by a function f :

R(f X, Y )Z = f X Y Z Y f X Z [f X,Y ] Z
= f X Y Z Y (f X Z) f [X,Y ]Y (f )X Z
= f X Y Z f Y X Z Y (f )X Z f [X,Y ] Z + Y (f )X Z
= f X Y Z f Y X Z Y (f )X Z f [X,Y ] Z + Y (f )X Z
= f R(X, Y )Z (6.3)

R(X, Y )(f Z) = X Y (f Z) Y X (f Z) [X,Y ] (f Z)

= X (f Y Z + Y (f )Z) Y (f X Z + X(f )Z)
f [X,Y ] Z [X, Y ](f )Z
= f X Y Z + X(f )Y Z + Y (f )X Z + X(Y (f ))Z

Part 3 GR December 14, 2016 56 H.S. Reall

 6.3. PARALLEL TRANSPORT AGAIN

f Y X Z Y (f )X Z X(f )Y Z Y (X(f ))Z

f [X,Y ] Z [X, Y ](f )Z
= f R(X, Y )Z (6.4)

It follows that our definition does indeed define a tensor. Lets calculate its com-
ponents in a coordinate basis {e = /x } (so [e , e ] = 0). Use the notation
e ,

R(e , e )e = e e
= ( e ) ( e )
= e + e e e (6.5)

and hence, in a coordinate basis,

R = + (6.6)

Remark. It follows that the Riemann tensor vanishes for the Levi-Civita connec-
tion in Euclidean space or Minkowski spacetime (since one can choose coordinates
for which the Christoffel symbols vanish everywhere).
The following contraction of the Riemann tensor plays an important role in
GR:
Definition. The Ricci curvature tensor is the (0, 2) tensor defined by

Rab = Rc acb (6.7)

We saw earlier that, with vanishing torsion, the second covariant derivatives of
a function commute. The same is not true of covariant derivatives of tensor fields.
The failure to commute arises from the Riemann tensor:
Exercise. Let be a torsion-free connection. Prove the Ricci identity:

c d Z a d c Z a = Ra bcd Z b (6.8)

Hint. Show that the equation is true when multiplied by arbitrary vector fields
X c and Y d .

6.3 Parallel transport again

Now we return to the relation between the Riemann tensor and the path-dependence
of parallel transport. Let X and Y be vector fields that are linearly independent
everywhere, with [X, Y ] = 0. Earlier we saw that we can choose a coordinate chart

Part 3 GR December 14, 2016 57 H.S. Reall

CHAPTER 6. CURVATURE

(s, t, . . .) such that X = /s and Y = /t. Let p M and choose the coor-
dinate chart such that p has coordinates (0, . . . , 0). Let q, r, u be the point with
coordinates (s, 0, 0, . . .), (s, t, 0, . . .), (0, t, 0, . . .) respectively, where s and t
are small. We can connect p and q with a curve along which only s varies, with
tangent X. Similarly, q and r can be connected by a curve with tangent Y . p and
u can be connected by a curve with tangent Y , and u and r can be connected by
a curve with tangent X. The result is a small quadrilateral (Fig. 6.1).

u(0, t, 0, . . . , 0) X r(s, t, 0, . . . , 0)

Y Y

X q(s, 0, . . . , 0)
p(0, 0, . . . , 0)

Figure 6.1: Parallel transport

Now let Zp Tp (M ). Parallel transport Zp along pqr to obtain a vector

Zr Tr (M ). Parallel transport Zp along pur to obtain a vector Zr0 Tr (M ). We
shall calculate the difference Zr0 Zr for a torsion-free connection.
It is convenient to introduce a new coordinate chart: normal coordinates at p.
Henceforth, indices , , . . . will refer to this chart. s and t will now be used as
parameters along the curves with tangent X and Y respectively.
pq is a curve with tangent vector X and parameter s. Along pq, Z is par-
allely transported: X Z = 0 so dZ /ds = Z X and hence d2 Z /ds2 =
( Z X ), X . Now Taylors theorem gives

dZ d2 Z
   
1
Zq = Zp + s + s2 + O(s3 )
ds p 2 ds2 p
1
= Zp , Z X X p s2 + O(s3 )


(6.9)
2
where we have used (p) = 0 in normal coordinates at p (assuming a torsion-free
connection). Now consider parallel transport along qr to obtain
 
1 d2 Z
 
dZ
Zr = Zq + t + t2 + O(t3 )
dt q 2 dt2 q

Part 3 GR December 14, 2016 58 H.S. Reall

 6.4. SYMMETRIES OF THE RIEMANN TENSOR

1
= Zq Z Y q t ( Z Y ), Y q t2 + O(t3 )



h 2 i
2


= Zq , Z Y X p s + O(s ) t
1h i
(, Z Y Y p + O(s) t2 + O(t3 )




2
1

, p Z X X s2 + Y Y t2 + 2Y X st p + O( 3 )


= Zp
2
(6.10)

Here we assume that s and t both are O() (i.e. s = a for some non-zero
constant a and similarly for t). Now consider parallel transport along pur. The
result can be obtained from the above expression simply by interchanging X with
Y and s with t. Hence we have
0
Zr Zr Zr = , Z (Y X X Y ) p st + O( 3 )
 

, , Z X Y p st + O( 3 )


=
= (R Z X Y )p st + O( 3 )
= (R Z X Y )r st + O( 3 ) (6.11)

where we used the expression (6.6) for the Riemann tensor components (remember
that (p) = 0). In the final equality we used that quantities at p and r differ
by O(). We have derived this result in a coordinate basis defined using normal
coordinates at p. But now both sides involve tensors at r. Hence our equation is
basis-independent so we can write
Zra
Ra bcd Z b X c Y d


r
= lim (6.12)
0 st

The Riemann tensor measures the path-dependence of parallel transport.

Remark. We considered parallel transport along two different curves from p to r.
However, we can reinterpret the result as describing the effect of parallel transport
0
of a vector Zra around the closed curve rqpur to give the vector Zra . Hence Zra
measures the change in Zra when parallel transported around a closed curve.

6.4 Symmetries of the Riemann tensor

From its definition, we have the symmetry Ra bcd = Ra bdc , equivalently:

Ra b(cd) = 0. (6.13)

Proposition. If is torsion-free then

Part 3 GR December 14, 2016 59 H.S. Reall

CHAPTER 6. CURVATURE

Ra [bcd] = 0. (6.14)

Proof. Let p M and choose normal coordinates at p. Vanishing torsion implies

(p) = 0 and [] = 0 everywhere. We have R = at p.
Antisymmetrizing on now gives R [] = 0 at p in the coordinate basis defined
using normal coordinates at p. But if the components of a tensor vanish in one
basis then they vanish in any basis. This proves the result at p. However, p is
arbitrary so the result holds everywhere.
Proposition. (Bianchi identity). If is torsion-free then

Ra b[cd;e] = 0 (6.15)

Proof. Use normal coordinate at p again. At p,

R ; = R (6.16)

In normal coordinates at p, R = and the latter terms vanish at p, we

only need to worry about the former:

R ; = at p (6.17)

Antisymmetrizing gives R [; ] = 0 at p in this basis. But again, if this is true

in one basis then it is true in any basis. Furthermore, p is arbitrary. The result
follows.

6.5 Geodesic deviation

Remark. In Euclidean space, or in Minkowski spacetime, initially parallel geodesics
remain parallel forever. On a general manifold we have no notion of parallel.
However, we can study whether nearby geodesics move together or apart. In par-
ticular, we can quantify their relative acceleration.
Definition. Let M be a manifold with a connection . A 1-parameter family of
geodesics is a map : I I 0 M where I and I 0 both are open intervals in R,
such that (i) for fixed s, (s, t) is a geodesic with affine parameter t (so s is the
parameter that labels the geodesic); (ii) the map (s, t) 7 (s, t) is smooth and
one-to-one with a smooth inverse. This implies that the family of geodesics forms
a 2d surface M .
Let T be the tangent vector to the geodesics and S to be the vector tangent
to the curves of constant t, which are parameterized by s (see Fig. 6.2). In
a chart x , the geodesics are specified by x (s, t) with S = x /s. Hence

Part 3 GR December 14, 2016 60 H.S. Reall

 6.5. GEODESIC DEVIATION

s = const

T
T T

S
o
t = const
S

Figure 6.2: 1-parameter family of geodesics

x (s + s, t) = x (s, t) + sS (s, t) + O(s2 ). Therefore (s)S a points from one

geodesic to an infinitesimally nearby one in the family. We call S a a deviation
vector.
On the surface we can use s and t as coordinates. We can extend these to
coordinates (s, t, . . .) defined in a neighbourhood of . This gives a coordinate
chart in which S = /s and T = /t on . We can now use these equations to
extend S and T to a neighbourhood of the surface. S and T are now vector fields
satisfying
[S, T ] = 0 (6.18)

Remark. If we fix attention on a particular geodesic then T (sS) = sT S

can be regarded as the rate of change of the relative position of an infinitesimally
nearby geodesic in the family i.e., as the relative velocity of an infinitesimally
nearby geodesic. We can define the relative acceleration of an infinitesimally
nearby geodesic in the family as sT T S. The word relative is important: the
acceleration of a curve with tangent T is T T , which vanishes here (as the curves
are geodesics).
Proposition. If has vanishing torsion then

T T S = R(T, S)T (6.19)

Proof. Vanishing torsion gives T S S T = [T, S] = 0. Hence

T T S = T S T = S T T + R(T, S)T, (6.20)

where we used the definition of the Riemann tensor. But T T = 0 because T is

tangent to (affinely parameterized) geodesics.

Part 3 GR December 14, 2016 61 H.S. Reall

CHAPTER 6. CURVATURE

Remark. This result is known as the geodesic deviation equation. In abstract

index notation it is:
T c c (T b b S a ) = Ra bcd T b T c S d (6.21)
This equation shows that curvature results in relative acceleration of geodesics.
It also provides another method of measuring Ra bcd : at any point p we can pick
our 1-parameter family of geodesics such that T and S are arbitrary. Hence by
measuring the LHS above we can determine Ra (bc)d . From this we can determine
Ra bcd :
Exercise. Show that, for a torsion-free connection,
2 a
Ra bcd = R (bc)d Ra (bd)c


(6.22)
3

Remarks.

1. Note that the relative acceleration vanishes for all families of geodesics if,
and only if, Ra bcd = 0.

2. In GR, free particles follow geodesics of the Levi-Civita connection. Geodesic

deviation is the tendency of freely falling particles to move together or apart.
We have already met this phenomenon: it arises from tidal forces. Hence the
Riemann tensor is the quantity that measures tidal forces.

6.6 Curvature of the Levi-Civita connection

Remark. From now on, we shall restrict attention to a manifold with metric,
and use the Levi-Civita connection. The Riemann tensor then enjoys additional
symmetries. Note that we can lower an index with the metric to define Rabcd .
Proposition. The Riemann tensor satisfies

Rabcd = Rcdab , R(ab)cd = 0. (6.23)

Proof. The second identity follows from the first and the antisymmetry of the
Riemann tensor. To prove the first, introduce normal coordinates at p, so g = 0
at p. Then, at p,
0 = = (g g ) = g g . (6.24)
Multiplying by the inverse metric gives g = 0 at p. Using this, we have

1
= g (g, + g, g, ) at p (6.25)
2
Part 3 GR December 14, 2016 62 H.S. Reall
 6.7. EINSTEINS EQUATION

And hence (as = 0 at p)

1
R = (g, + g, g, g, ) at p (6.26)
2
This satisfies R = R at p using the symmetry of the metric and the fact that
partial derivatives commute. This establishes the identity in normal coordinates,
but this is a tensor equation and hence valid in any basis. Furthermore p is
arbitrary so the identity holds everywhere.
Proposition. The Ricci tensor is symmetric:

Rab = Rba (6.27)

Proof. Rab = g cd Rdacb = g cd Rcbda = Rc bca = Rba where we used the first identity
above in the second equality.
Definition. The Ricci scalar is

R = g ab Rab (6.28)

Definition. The Einstein tensor is the symmetric (0, 2) tensor defined by

1
Gab = Rab Rgab (6.29)
2

Proposition. The Einstein tensor satisfies the contracted Bianchi identity:

a Gab = 0 (6.30)

which can also be written as

1
a Rab b R = 0 (6.31)
2
Proof. Examples sheet 2.

6.7 Einsteins equation

Postulates of General Relativity.
1. Spacetime is a 4d Lorentzian manifold equipped with the Levi-Civita con-
nection.

2. Free particles follow timelike or null geodesics.

Part 3 GR December 14, 2016 63 H.S. Reall

CHAPTER 6. CURVATURE

3. The energy, momentum, and stresses of matter are described by a symmetric

tensor Tab that is conserved: a Tab = 0.

4. The curvature of spacetime is related to the energy-momentum tensor of

matter by the Einstein equation (1915)

1
Gab Rab Rgab = 8G Tab (6.32)
2
where G is Newtons constant.

We have discussed points 1-3 above. It remains to discuss the Einstein equation.
We can motivate this as follows. In GR, the gravitational field is described by
the curvature of spacetime. Since the energy of matter should be responsible
for gravitation, we expect some relationship between curvature and the energy-
momentum tensor. The simplest possibility is a linear relationship, i.e., a curvature
tensor is proportional to Tab . Since Tab is symmetric, it is natural to expect the
Ricci tensor to be the relevant curvature tensor.
Einsteins first guess for the field equation of GR was Rab = CTab for some
constant C. This does not work for the following reason. The RHS is conserved
hence this equation implies a Rab = 0. But then from the contracted Bianchi
identity we get a R = 0. Taking the trace of the equation gives R = CT (where
T = T a a ) and hence we must have a T = 0, i.e., T is constant. But, T vanishes
in empty space and is usually non-zero inside matter. Hence this is unsatisfactory.
The solution to this problem is obvious once one knows of the contracted
Bianchi identity. Take Gab , rather than Rab , to be proportional to Tab . The coeffi-
cient of proportionality on the RHS of Einsteins equation is fixed by demanding
that the equation reduces to Newtons law of gravitation when the gravitational
field is weak and the matter is moving non-relativistically. We will show this later.
Remarks.

1. In vacuum, Tab = 0 so Einsteins equation gives Gab = 0. Contracting indices

gives R = 0. Hence the vacuum Einstein equation can be written as

Rab = 0 (6.33)

2. The geodesic postulate of GR is redundant. Using conservation of the

energy-momentum tensor it can be shown that a distribution of matter that
is small (compared to the scale on which the spacetime metric varies), and
sufficiently weak (so that its gravitational effect is small), must follow a
geodesic. (See examples sheet 4 for the case of a point particle.)

Part 3 GR December 14, 2016 64 H.S. Reall

 6.7. EINSTEINS EQUATION

3. The Einstein equation is a set of non-linear, second order, coupled, partial

differential equations for the components of the metric g . Very few phys-
ically interesting explicit solutions are known so one has to develop other
methods to solve the equation, e.g., numerical integration.

4. How unique is the Einstein equation? Is there any tensor, other than Gab
that we could have put on the LHS? The answer is supplied by:

Theorem (Lovelock 1972). Let Hab be a symmetric tensor such that (i) in any
coordinate chart, at any point, H is a function of g , g, and g, at that
point; (ii) a Hab = 0; (iii) either spacetime is four-dimensional or H depends
linearly on g, . Then there exist constants and such that

Hab = Gab + gab (6.34)

Hence (as Einstein realized) there is the freedom to add a constant multiple of gab
to the LHS of Einsteins equation, giving

Gab + gab = 8G Tab (6.35)

is called the cosmological constant. This no longer reduces to Newtonian theory

for slow motion in a weak field but the deviation from Newtonian theory is unob-
servable if is sufficiently small. Note that ||1/2 has the dimensions of length.
The effects of are negligible on lengths or times small compared to this quantity.
Astronomical observations suggest that there is indeed a very small positive cos-
mological constant: 1/2 109 light years, the same order of magnitude as the
size of the observable Universe. Hence the effects of the cosmological constant are
negligible except on cosmological length scales. Therefore we can set = 0 unless
we discuss cosmology.
Note that we can move the cosmological constant term to the RHS of the
Einstein equation, and regard it as the energy-momentum tensor of a perfect fluid
with = p = /(8G). Hence the cosmological constant is sometimes referred
to as dark energy or vacuum energy. It is a great mystery why it is so small because
arguments from quantum field theory suggest that it should be 10120 times larger.
This is the cosmological constant problem. One (controversial) proposed solution
of this problem invokes the anthropic principle, which posits the existence of many
possible universes with different values for constants such as . If was very
different from its observed value then galaxies never would have formed and hence
we would not be here.
Remark. We have explicitly written Newtons constant G throughout this section.
Henceforth we shall choose units so that G = c = 1.

Part 3 GR December 14, 2016 65 H.S. Reall

CHAPTER 6. CURVATURE

Part 3 GR December 14, 2016 66 H.S. Reall

Chapter 7

Diffeomorphisms and Lie

derivative

7.1 Maps between manifolds

Definition. Let M , N be differentiable manifolds of dimension m, n respectively.
A function : M N is smooth if, and only if, A 1 is smooth for all
charts of M and all charts A of N (note that this is a map from a subset of
Rm to a subset of Rn ).
If we have such a map then we can pull-back a function on N to define a
function on M :
Definition. Let : M N and f : N R be smooth functions. The pull-
back of f by is the function (f ) : M R defined by (f ) = f , i.e.,
(f )(p) = f ((p)).
Furthermore, allows us to push-forward a curve in M to a curve in
N . Hence we can push-forward vectors from M to N (Figs. 7.1, 7.2)

X (X)
p (p)

M N

Figure 7.1: A curve in M Figure 7.2: The curve in N

67
CHAPTER 7. DIFFEOMORPHISMS AND LIE DERIVATIVE

Definition. Let : M N be smooth. Let p M and X Tp (M ). The

push-forward of X with respect to is the vector (X) T(p) (N ) defined as
follows. Let be a smooth curve in M passing through p with tangent X at p.
Then (X) is the tangent vector to the curve in N at the point (p).
Lemma. Let f : N R. Then ( (X))(f ) = X( (f )).
Proof. Wlog (0) = p.
 
d
( (X))(f ) = (f ( ))(t)
dt
 t=0
d
= ((f ) )(t)
dt t=0
= X( (f )) (7.1)

Exercise. Let x be coordinates on M and y be coordinates on N (we use

different indices , etc for N because N is a different manifold which might not
have the same dimension as M ). Then we can regard as defining a map y (x ).
Show that the components of (X) are related to those of X by
 
y
( (X)) = X (7.2)
x p

The map on covectors works in the opposite direction:

Definition. Let : M N be smooth. Let p M and T(p) (N ). The pull-

back of with respect to is () Tp (M ) defined by ( ())(X) = ( (X))
for any X Tp (M ).
Lemma. Let f : N R. Then (df ) = d( (f )).
Proof. Let X Tp (M ). Then

( (df ))(X) = (df )( (X)) = ( (X))(f ) = X( (f )) = (d( (f )))(X) (7.3)

The first equality is the definition of , the second is the definition of df , the
third is the previous Lemma and the fourth is the definition of d( (f )). Since X
is arbitrary, the result follows.
Exercise. Use coordinates x and y as before. Show that the components of
() are related to the components of by
 
y
( ()) = (7.4)
x p

Remarks.

Part 3 GR December 14, 2016 68 H.S. Reall

 7.2. DIFFEOMORPHISMS, LIE DERIVATIVE

1. In all of the above, the point p was arbitrary so push-forward and pull-back
can be applied to vector and covector fields, respectively.
2. The pull-back can be extended to a tensor S of type (0, s) by defining
( (S))(X1 , . . . Xs ) = S( (X1 ), . . . (Xs )) where X1 , . . . , Xs Tp (M ). Sim-
ilarly, one can push-forward a tensor of type (r, 0) by defining (T )(1 , . . . , r ) =
T ( (1 ), . . . , (r )) where 1 , . . . , r Tp (N ). The components of these
tensors in a coordinate basis are given by
 1   s 
y y
( (S))1 ...s = ... S ... (7.5)
x 1
p xs p 1 s
 1   r 
1 ...r y y
( (T )) =
. . . T 1 ...r (7.6)
x 1 p xr p

Example. The embedding of S 2 into Euclidean space. Let M = S 2 and N =

R3 . Define : M N as the map which sends the point on S 2 with spherical
polar coordinates x = (, ) to the point y = (sin cos , sin sin , cos ) R3 .
Consider the Euclidean metric g on R3 , whose components are the identity matrix
. Pulling this back to S 2 using (7.5) gives ( g) = diag(1, sin2 ) (check!), the
unit round metric on S 2 .

7.2 Diffeomorphisms, Lie Derivative

Definition. A map : M N is a diffeomorphism iff it 1-1 and onto, smooth,
and has a smooth inverse.
Remark. This implies that M and N have the same dimension. In fact, M and
N have identical manifold structure.
With a diffeomorphism, we can extend our definitions of push-forward and
pull-back so that they apply for any type of tensor:
Definition. Let : M N be a diffeomorphism and T a tensor of type (r, s) on
M . Then the push-forward of T is a tensor (T ) of type (r, s) on N defined by

(for arbitrary i T(p) (N ), Xi T(p) (N ))

(T )(1 , . . . , r , X1 , . . . , Xs ) = T ( (1 ), . . . , (r ), (1 ) (X1 ), . . . , (1 ) (Xs ))

(7.7)
Exercises.
1. Convince yourself that push-forward commutes with the contraction and
outer product operations.

Part 3 GR December 14, 2016 69 H.S. Reall

CHAPTER 7. DIFFEOMORPHISMS AND LIE DERIVATIVE

2. Show that the analogue of equation (7.6) for a (1, 1) tensor field is
   
y x
[( (T )) ](p) =
(T )p (7.8)
x p y p

(We dont need to use indices , etc because now M and N have the same
dimension.) Generalize this result to a (r, s) tensor.

Remarks.

1. Pull-back can be defined in a similar way, with the result = (1 ) .

2. Weve taken an active point of view, regarding a diffeomorphism as a map

taking a point p to a new point (p). However, there is an alternative pas-
sive point of view in which we consider a diffeomorphism simply as a change
of chart at p. Consider a coordinate chart x defined near p and another chart
y defined near (p) (Fig. 7.3). Regarding the coordinates y as functions
on N , we can pull them back to define corresponding coordinates, which we
also call y , on M . So now we have two coordinate systems defined near p.
The components of tensors at p in the new coordinate basis are given by the
tensor transformation law, which is exactly the RHS of (7.8).

x y
(p)
p
M
N

Figure 7.3: Active versus passive diffeomorphism.

Definition. Let : M N be a diffeomorphism. Let be a covariant derivative

on N defined by
on M . The push-forward of is a covariant derivative
X T = (X) ( (T ))


(7.9)

where X is a vector field and T a tensor field on N . (In words: pull-back X and
T to M , evaluate the covariant derivative there and then push-forward the result
to N .)
Exercises (examples sheet 3).

1. Check that this satisfies the properties of a covariant derivative.

Part 3 GR December 14, 2016 70 H.S. Reall

 7.2. DIFFEOMORPHISMS, LIE DERIVATIVE

is the push-forward of the Riemann

2. Show that the Riemann tensor of
tensor of .

3. Let be the Levi-Civita connection defined by a metric g on M . Show that

is the Levi-Civita connection defined by the metric (g) on N .

Remark In GR we describe physics with a manifold M on which certain ten-

sor fields e.g. the metric g, Maxwell field F etc. are defined. If : M N
is a diffeomorphism then there is no way of distinguishing (M, g, F, . . .) from
(N, (g), (F ), . . .); they give equivalent descriptions of physics. For example
if the metric g on M has components g in a basis {e } for Tp (M ) then the met-
ric (g) has the same components g in the basis { (e )} for T(p) (N ). If we
set N = M this reveals that the set of tensor fields ( (g), (F ), . . .) is physically
indistinguishable from (g, F, . . .). If two sets of tensor fields are not related by a
diffeomorphism then they are physically distinguishable. It follows that diffeomor-
phisms are the gauge symmetry (redundancy of description) in GR.
Example. Consider three particles following geodesics of the metric g. Assume
that the worldlines of particles 1 and 2 intersect at p and that the worldlines of
particles 2 and 3 intersect at q. Applying a diffeomorphism : M M maps the
worldlines to geodesics of (g) which intersect at the points (p) and (q). Note
that (p) 6= p so saying particles 1 and 2 coincide at p is not a gauge-invariant
statement. An example of a quantity that is gauge invariant is the proper time
between the two intersections along the worldline of particle 2.
Remark. This gauge freedom raises the following puzzle. The metric tensor
is symmetric and hence has 10 independent components. Consider the vacuum
Einstein equation - this appears to give 10 independent equations, which looks
good. But the Einstein equation should not determine the components of the
metric tensor uniquely, but only up to diffeomorphisms. The resolution is that not
all components of the Einstein equations are truly independent because they are
related by the contracted Bianchi identity.
Note that diffeomorphisms allow us to compare tensors defined at different
points via push-forward or pull-back. This leads to a notion of a tensor field
possessing symmetry:
Definition. A diffeomorphism : M M is a symmetry transformation of a
tensor field T iff (T ) = T everywhere. A symmetry transformation of the metric
tensor is called an isometry.
Definition. Let X be a vector field on a manifold M . Let t be the map which
sends a point p M to the point parameter distance t along the integral curve

Part 3 GR December 14, 2016 71 H.S. Reall

CHAPTER 7. DIFFEOMORPHISMS AND LIE DERIVATIVE

of X through p (this might be defined only for small enough t). It can be shown
that t is a diffeomorphism.
Remarks.

1. Note that 0 is the identity map and s t = s+t . Hence t = (t )1 . If

t is defined for all t R (in which case we say the integral curves of X are
complete) then these diffeomorphisms form a 1-parameter abelian group.

2. Given X weve defined t . Conversely, if one has a 1-parameter abelian

group of diffeomorphisms t (i.e. one satisfying the rules just mentioned)
then through any point p one can consider the curve with parameter t given
by t (p). Define X to be the tangent to this curve at p. Doing this for all
p defines a vector field X. The integral curves of X generate t in the sense
defined above.

3. If we use (t ) to compare tensors at different points then the parameter t

controls how near the points are. In particular, in the limit t 0, we are
comparing tensors at infinitesimally nearby points. This leads to the notion
of a new type of derivative:

Definition. The Lie derivative of a tensor field T with respect to a vector field
X at p is
((t ) T )p Tp
(LX T )p = lim (7.10)
t0 t
Remark. The Lie derivative wrt X is a map from (r, s) tensor fields to (r, s) tensor
fields. It obeys LX (S + T ) = LX S + LX T where and are constants.
The easiest way to demonstrate other properties of the Lie derivative is to
introduce coordinates in which the components of X are simple. Let be a
hypersurface that has the property that X is nowhere tangent to (in particular
X 6= 0 on ). Let xi , i = 1, 2, . . . , n 1 be coordinates on . Now assign
coordinates (t, xi ) to the point parameter distance t along the integral curve of X
that starts at the point with coordinates xi on (Fig. 7.4).
This defines a coordinate chart (t, xi ) at least for small t, i.e., in a neighbour-
hood of . Furthermore, the integral curves of X are the curves (t, xi ) with fixed
xi and parameter t. The tangent to these curves is /t so we have constructed
coordinates such that X = /t. The diffeomorphism t is very simple: it just
sends the point p with coordinates x = (tp , xip ) to the point t (p) with coordinates
y = (tp + t, xip ) hence y /x = . The generalization of (7.8) to a (r, s) tensor
then gives
[((t ) (T ))1 ,...,r 1 ,...s ]t (q) = [T 1 ,...,r 1 ,...s ]q (7.11)

Part 3 GR December 14, 2016 72 H.S. Reall

 7.2. DIFFEOMORPHISMS, LIE DERIVATIVE

(t, xi )

xi

Figure 7.4: Coordinates adapted to a vector field

and setting q = t (p) gives

[((t ) (T ))1 ,...,r 1 ,...s ]p = [T 1 ,...,r 1 ,...s ]t (p) (7.12)

It follows that, if p has coordinates (tp , xip ) in this chart,

1
(LX T )1 ,...,r 1 ,...s = lim T 1 ,...,r 1 ,...s (tp + t, xip ) T 1 ,...,r 1 ,...s (tp , xip )

t0 t
 
1 ,...,r i
= T 1 ,...s (tp , xp ) (7.13)
t

So in this chart, the Lie derivative is simply the partial derivative with respect to
the coordinate t. It follows that the Lie derivative has the following properties:

1. It obeys the Leibniz rule: LX (S T ) = (LX S) T + S LX T .

2. It commutes with contraction.

Now lets derive a basis-independent formula for the Lie derivative. First con-
sider a function f . In the above chart, we have LX f = (/t)(f ). However, in
this chart we also have X(f ) = (/t)(f ). Hence

LX f = X(f ) (7.14)

Both sides of this expression are scalars and hence this equation must be valid in
any basis. Next consider a vector field Y . In our coordinates above we have

Y
(LX Y ) = (7.15)
t
but we also have
Y
[X, Y ] = (7.16)
t
Part 3 GR December 14, 2016 73 H.S. Reall
CHAPTER 7. DIFFEOMORPHISMS AND LIE DERIVATIVE

If two vectors have the same components in one basis then they are equal in all
bases. Hence we have the basis-independent result

LX Y = [X, Y ] (7.17)

Remark. Lets compare the Lie derivative and the covariant derivative. The
former is defined on any manifold whereas the latter requires extra structure (a
connection). Equation (7.17) reveals that the Lie derivative wrt X at p depends
on Xp and the first derivatives of X at p. The covariant derivative wrt X at p
depends only on Xp , which enables us to remove X to define the tensor T , a
covariant generalization of partial differentiation. It is not possible to define a
corresponding tensor LT using the Lie derivative. Only LX T makes sense.
Exercises (examples sheet 3).

1. Derive the formula for the Lie derivative of a covector in a coordinate

basis:
(LX ) = X + X (7.18)
Show that this can be written in the basis-independent form (where is the
Levi-Civita connection)

(LX )a = X b b a + b a X b (7.19)

2. Show that the Lie derivative of the metric in a coordinate basis is

(LX g) = X g + g X + g X (7.20)

and that this can be written in the basis-independent form

(LX g)ab = a Xb + b Xa (7.21)

Remark. If t is a symmetry transformation of T (for all t) then LX T = 0. If t

are a 1-parameter group of isometries then LX g = 0, i.e.,

a Xb + b Xa = 0 (7.22)

This is Killings equation and solutions are called Killing vector fields. Consider
the case in which there exists a chart for which the metric tensor does not de-
pend on some coordinate z. Then equation (7.20) reveals that /z is a Killing
vector field. Conversely, if the metric admits a Killing vector field then equation
(7.13) demonstrates that one can introduce coordinates such that the metric tensor
components are independent of one of the coordinates.

Part 3 GR December 14, 2016 74 H.S. Reall

 7.2. DIFFEOMORPHISMS, LIE DERIVATIVE

Lemma. Let X a be a Killing vector field and let V a be tangent to an affinely

parameterized geodesic. Then Xa V a is constant along the geodesic.
Proof. The derivative of Xa V a along the geodesic is
d
(Xa V a ) = V (Xa V a ) = V (Xa V a ) = V b b (Xa V a )
d
= V a V b b Xa + Xa V b b V a (7.23)

The first term vanishes because Killings equation implies that b Xa is antisym-
metric. The second term vanishes by the geodesic equation.
Exercise. Let J a = T a b X b where Tab is the energy-momentum tensor and X b is
a Killing vector field. Show that a J a = 0, i.e., J a is a conserved current.

Part 3 GR December 14, 2016 75 H.S. Reall

CHAPTER 7. DIFFEOMORPHISMS AND LIE DERIVATIVE

Part 3 GR December 14, 2016 76 H.S. Reall

Chapter 8

Linearized theory

8.1 The linearized Einstein equation

The nonlinearity of the Einstein equation makes it very hard to solve. However,
in many circumstances, gravity is not strong and spacetime can be regarded as
a perturbation of Minkowski spacetime. More precisely, we assume our space-
time manifold is M = R4 and that there exist globally defined almost inertial
coordinates x for which the metric can be written

g = + h , = diag(1, 1, 1, 1) (8.1)

with the weakness of the gravitational field corresponding to the components of

h being small compared to 1. Note that we are dealing with a situation in which
we have two metrics defined on spacetime, namely gab and the Minkowski metric
ab . The former is supposed to be the physical metric, i.e., free particles move on
geodesics of gab .
In linearized theory we regard h as the components of a tensor field in the
sense of special relativity, i.e., it transforms as a tensor under Lorentz transforma-
tions of the coordinates x .
To leading order in the perturbation, the inverse metric is

g = h , (8.2)

where we define
h = h (8.3)
To prove this, just check that g g = to linear order in the perturbation.
Here, and henceforth, we shall raise and lower indices using the Minkowski metric
. To leading order this agrees with raising and lowering with g . We shall
determine the Einstein equation to first order in the perturbation h .

77
CHAPTER 8. LINEARIZED THEORY

To first order, the Christoffel symbols are

1
= (h, + h, h, ) , (8.4)
2
the Riemann tensor is (neglecting terms since they are second order in the
perturbation):

R =

1
= (h, + h, h, h, ) (8.5)
2
and the Ricci tensor is
1 1
R = ( h) h h, (8.6)
2 2
where denotes /x as usual, and

h = h (8.7)

To first order, the Einstein tensor is

1 1 1
G = ( h) h h ( h h) . (8.8)
2 2 2
The Einstein equation equates this to 8T (which must therefore be assumed to
be small, otherwise spacetime would not be nearly flat). It is convenient to define
1
h = h h , (8.9)
2
with inverse
1
h = h h , (h = h = h) (8.10)
2
The linearized Einstein equation is then (exercise)
1 1
h + ( h) h = 8T (8.11)
2 2
We must now discuss the gauge symmetry present in this theory. We argued above
that diffeomorphisms are gauge transformations in GR. A manifold M with metric
g and energy-momentum tensor T is physically equivalent to M with metric (g)
and energy momentum tensor (T ) if is a diffeomorphism. Now we are restrict-
ing attention to metrics of the form (8.1). Hence we must consider which diffeo-
morphisms preserve this form. A general diffeomorphism would lead to ( ())
very different from diag(1, 1, 1, 1) and hence such a diffeomorphism would not

Part 3 GR December 14, 2016 78 H.S. Reall

 8.1. THE LINEARIZED EINSTEIN EQUATION

preserve (8.1). However, if we consider a 1-parameter family of diffeomorphisms

t then 0 is the identity map, so if t is small then t is close to the identity and
hence will have a small effect, i.e., (t ()) will be close to diag(1, 1, 1, 1) and
the form (8.1) will be preserved. For small t, we can use the definition of the Lie
derivative to deduce that, for any tensor T

(t ) (T ) = T + tLX T + O(t2 )
= T + L T + O(t2 ) (8.12)

where X a is the vector field that generates t and a = tX a . Note that a is

small so we treat it as a first order quantity. If we apply this result to the energy-
momentum tensor, evaluating in our chart x , then the first term is small (by
assumption) so the second term is higher order and can be neglected. Hence the
energy-momentum tensor is gauge-invariant to first order. The same is true for
any tensor that vanishes in the unperturbed spacetime, e.g. the Riemann tensor.
However, for the metric we have

(t ) (g) = g + L g + . . . = + h + L + . . . (8.13)

where we have neglected L h because this is a higher order quantity (as and h
both are small). Comparing this with (8.1) we deduce that h and h + L describe
physically equivalent metric perturbations. Therefore linearized GR has the gauge
symmetry h h + L for small . In our chart x , we can use (7.21) to write
(L ) = + and so the gauge symmetry is

h h + + (8.14)

There is a close analogy with electromagnetism in flat spacetime, where we can

introduce an electromagnetic potential A , a 4-vector obeying F = 2[ A] . This
has the gauge symmetry
A A + (8.15)
for some scalar . In this case, one can choose to impose the gauge condition
A = 0. Similarly, in linearized GR we can choose to impose the gauge
condition
h = 0. (8.16)
To see this, note that under the gauge transformation (8.14) we have

h h + (8.17)

so if we choose to satisfy the wave equation = h (which we can

solve using a Green function) then we reach the gauge (8.16). This is called

Part 3 GR December 14, 2016 79 H.S. Reall

CHAPTER 8. LINEARIZED THEORY

variously Lorenz, de Donder, or harmonic gauge. In this gauge, the linearized

Einstein equation reduces to

h = 16T (8.18)

Hence, in this gauge, each component of h satisfies the wave equation with
a source given by the energy-momentum tensor. Given appropriate boundary
conditions, the solution can be determined using a Green function.

8.2 The Newtonian limit

We will now see how GR reduces to Newtonian theory in the limit of non-relativistic
motion and a weak gravitational field. To do this, we could reintroduce factors of
c and try to expand everything in powers of 1/c since we expect Newtonian theory
to be valid as c . We will follow an equivalent approach in which we stick to
our convention c = 1 but introduce a small dimensionless parameter 0 <   1
such that a factor of  appears everywhere that a factor of 1/c would appear.
We assume that, for some choice of almost-inertial coordinates x = (t, xi ), the
3-velocity of any particle v i = dxi /d is O(). Recall that in Newtonian theory we
had v 2 || (Fig. 1.1) and so we expect the gravitational field to be O(2 ). We
assume that
h00 = O(2 ), h0i = O(3 ), hij = O(2 ) (8.19)
Well see below how the additional factor of  in h0i emerges.
Since the matter which generates the gravitational field is assumed to move non-
relativistically, time derivatives of the gravitational field will be small compared to
spatial derivatives. Let L denote the length scale over which h varies, i.e., if X
denotes some component of h then |i X| = O(X/L). Our assumption of small
time derivatives is
0 X = O(X/L) (8.20)
For example, in Newtonian theory, the gravitational field of a body of mass M at
position x(t) is = M/|x x(t)| which obeys these formulae with L = |x x(t)|
and |x| = O().
Consider the equations for a timelike geodesic. The Lagrangian is (adding a
hat to avoid confusion with the length L)

L = (1 h00 )t2 2h0i txi (ij + hij ) xi xj (8.21)

where a dot denotes a derivative with respect to proper time . Our non-relativistic
assumption implies that xi = O(). From the definition of proper time we have
L = 1 and hence
(1 h00 )t2 ij xi xj = 1 + O(4 ) (8.22)

Part 3 GR December 14, 2016 80 H.S. Reall

 8.2. THE NEWTONIAN LIMIT

Rearranging gives
1 1
t = 1 + h00 + ij xi xj + O(4 ) (8.23)
2 2
The Euler-Lagrange equation for xi is

d 
2h0i t 2 (ij + hij ) xj = h00,i t2 2h0j,i txj hjk,i xj xk

d
= h00,i + O(4 /L) (8.24)

The LHS is 2xi plus subleading terms. Retaining only the leading order terms
gives
1
xi = h00,i (8.25)
2
Finally we can convert derivatives on the LHS to t derivatives using the chain
rule and (8.23) to obtain
d 2 xi
= i (8.26)
dt2
where
1
h00 (8.27)
2
We have recovered the equation of motion for a test body moving in a Newtonian
gravitational field . Note that the weak equivalence principle automatically is
satisfied: it follows from the fact that test bodies move on geodesics.
Exercise. Show that the corrections to (8.26) are O(4 /L). You can argue as
follows. (8.25) implies xi = O(2 /L). Now use (8.23) to show t = O(3 /L).
Expand out the derivative on the LHS of (8.24) to show that the corrections to
(8.25) are O(4 /L). Finally convert derivatives to t derivatives using (8.23).
The next thing we need to show is that satisfies the Poisson equation (1.1).
First consider the energy-momentum tensor. Assume that one can ascribe a 4-
velocity ua to the matter. Our non-relativistic assumption implies

ui = O(), u0 = 1 + O(2 ) (8.28)

where the second equality follows from gab ua ub = 1. The energy-density in the
rest-frame of the matter is
ua ub Tab (8.29)
Recall that T0i is the momentum density measured by an observer at rest in these
coordinates so we expect T0i ui = O(). Tij will have a part ui uj =
O(2 ) arising from the motion of the matter. It will also have a contribution
from the stresses in the matter. Under all but the most extreme circumstances,

Part 3 GR December 14, 2016 81 H.S. Reall

CHAPTER 8. LINEARIZED THEORY

these are small compared to . For example, in the rest frame of a perfect fluid,
stresses are determined by the pressure p. The speed of sound in the fluid is C
where C 2 = dp/d p/. Our non-relativistic assumption is that C = O() hence
p = O(2 ). This is true in the Solar system, where p/ || 105 at the centre
of the Sun. Hence we make the following assumptions

T00 = (1 + O(2 )), T0i = O(), Tij = O(2 ) (8.30)

where the first equality follows from (8.29) and other two equalities.
Finally, we consider the linearized Einstein equation. Equation (8.19) implies
that
h00 = O(2 ), h0i = O(3 ), hij = O(2 ) (8.31)
Using our assumption about time derivatives being small compared to spatial
derivatives, equation (8.18) becomes

2 h00 = 16(1 + O(2 )), 2 h0i = O(), 2 hij = O(2 ) (8.32)

If we impose boundary conditions that the metric perturbation (gravitational field)

should decay at large distance then these equations can be solved using a Green
function as in (1.2). The factors of  on the RHS above imply that the resulting
solutions satisfy

h0i = O(h00 ) = O(3 ), hij = O(h00 2 ) = O(4 ) (8.33)

Since h0i = h0i , this explains why we had to assume h0i = O(3 ). From the second
result, we have hii = O(4 ) and hence h = h00 + O(4 ). We can use (8.10) to
recover h . This gives h00 = (1/2)h00 + O(4 ) and so, using (8.27) we obtain
Newtons law of gravitation:

2 = 4(1 + O(2 )) (8.34)

We also obtain hij = (1/2)h00 ij + O(4 ) and so

hij = 2ij + O(4 ) (8.35)

This justifies the metric used in (1.16). The expansion of various quantities in
powers of  can be extended to higher orders. As we have seen, the Newtonian ap-
proximation requires only the O(2 ) term in h00 . The next order, post-Newtonian,
approximation corresponds to including O(4 ) terms in h00 , O(3 ) terms in h0i and
O(2 ) in hij . Equation (8.35) gives hij to O(2 ). The above analysis also lets us
write the O(3 ) term in h0i in terms of T0i and a Green function. However, to
obtain the O(4 ) term in h00 one has to go beyond linearized theory.

Part 3 GR December 14, 2016 82 H.S. Reall

 8.3. GRAVITATIONAL WAVES

8.3 Gravitational waves

In vacuum, the linearized Einstein equation reduces to the source-free wave equa-
tion:
h = 0 (8.36)
so the theory admits gravitational wave solutions. As usual for the wave equation,
we can build a general solution as a superposition of plane wave solutions. So lets
look for a plane wave solution:


h (x) = Re H eik x (8.37)

where H is a constant symmetric complex matrix describing the polarization

of the wave and k is the (real) wavevector. We shall suppress the Re in all
subsequent equations. The wave equation reduces to

k k = 0 (8.38)

so the wavevector k must be null hence these waves propagate at the speed of
light relative to the background Minkowski metric. The gauge condition (8.16)
gives
k H = 0, (8.39)
i.e. the waves are transverse.
Example. Consider the null vector k = (1, 0, 0, 1). Then exp(ik x ) = exp(i(t
z)) so this describes a wave of frequency propagating at the speed of light in the
z-direction. The transverse condition reduces to

H0 + H3 = 0. (8.40)

Returning to the general case, the condition (8.16) does not eliminate all gauge
freedom. Consider a gauge transformation (8.14). From equation (8.17), we see
that this preserves the gauge condition (8.16) if obeys the wave equation:

= 0. (8.41)

Hence there is a residual gauge freedom which we can exploit to simplify the
solution. Take

(x) = X eik x (8.42)
which satisfies (8.41) because k is null. Using

h h + + (8.43)

Part 3 GR December 14, 2016 83 H.S. Reall

CHAPTER 8. LINEARIZED THEORY

we see that the residual gauge freedom in our case is

H H + i (k X + k X k X ) (8.44)

Exercise. Show that the residual gauge freedom can be used to achieve longitu-
dinal gauge:
H0 = 0 (8.45)
but this still does not determine X uniquely, and the freedom remains to impose
the additional trace-free condition

H = 0. (8.46)

In this gauge, we have

h = h . (8.47)

Example. Return to our wave travelling in the z-direction. The longitudinal

gauge condition combined with the transversality condition (8.40) gives H3 = 0.
Using the trace-free condition now gives

0 0 0 0
0 H+ H 0
H = 0 H H+ 0
(8.48)
0 0 0 0

where the solution is specified by the two constants H+ and H corresponding

to two independent polarizations. So gravitational waves are transverse and have
two possible polarizations. This is one way of interpreting the statement that the
gravitational field has two degrees of freedom per spacetime point.
How would one detect a gravitational wave? An observer could set up a fam-
ily of test particles locally. The displacement vector S a from the observer to any
particle is governed by the geodesic deviation equation. (We are taking S a to be
infinitesimal, i.e., what we called sS a previously.) So we can use this equation
to predict what the observer would see. We have to be careful here. It would be
natural to write out the geodesic deviation equation using the almost inertial co-
ordinates and therby determine S . But how would we relate this to observations?
S are the components of S a with respect to a certain basis, so how would we de-
termine whether the variation in S arises from variation of S a or from variation
of the basis? With a bit more thought, one can make this approach work but we
shall take a different approach.
Consider an observer following a geodesic in a general spacetime. Our observer
will be equipped with a set of measuring rods with which to measure distances.

Part 3 GR December 14, 2016 84 H.S. Reall

 8.3. GRAVITATIONAL WAVES

At some point p on his worldline we could introduce a local inertial frame with
spatial coordinates X, Y, Z in which the observer is at rest. Imagine that the
observer sets up measuring rods of unit length pointing in the X, Y, Z directions
at p. Mathematically, this defines an orthonormal basis {e } for Tp (M ) (we use
to label the basis vectors because we are using for our almost inertial coordinates)
where ea0 = ua (his 4-velocity) and eai (i = 1, 2, 3) are spacelike vectors satisfying

ua eai = 0, gab eai ebj = ij (8.49)

In Minkowski spacetime, this basis can be extended to the observers entire world-
line by taking the basis vectors to have constant components (in an inertial frame),
i.e., they do not depend on proper time . In particular, this implies that the or-
thonormal basis is non-rotating. Since the basis vectors have constant components,
they are parallelly transported along the worldline. Hence, in curved spacetime,
the analogue of this is to take the basis vectors to be parallelly transported along
the worldline. For ua , this is automatic (the worldline is a geodesic). But for ei it
gives
ub b eai = 0 (8.50)
As we discussed previously, if the eai are specified at any point p then this equation
determines them uniquely along the whole worldline. Furthermore, the basis re-
mains orthonormal because parallel transport preserves inner products (examples
sheet 2). The basis just constructed is sometimes called a parallelly transported
frame. It is the kind of basis that would be constructed by an observer freely falling
with 3 gyroscopes whose spin axes define the spatial basis vectors. Using such a
basis we can be sure that an increase in a component of S a is really an increase in
the distance to the particle in a particular direction, rather than a basis-dependent
effect.
Now imagine this observer sets up a family of test particles near his worldline.
The deviation vector to any infinitesimally nearby particle satisfies the geodesic
deviation equation
ub b (uc c Sa ) = Rabcd ub uc S d (8.51)
Contract with ea and use the fact that the basis is parallelly transported to obtain

ub b [uc c (ea Sa )] = Rabcd ea ub uc S d (8.52)

Now ea Sa is a scalar hence the equation reduces to

d2 S
= Rabcd ea ub uc ed S (8.53)
d 2
where is the observers proper time and S = ea Sa is one of the components of
Sa in the parallelly transported frame. On the RHS weve used S d = ed S .

Part 3 GR December 14, 2016 85 H.S. Reall

CHAPTER 8. LINEARIZED THEORY

So far, the discussion has been general but now lets specialize to our gravi-
tational plane wave. On the RHS, Rabcd is a first order quantity so we only need
to evaluate the other quantities to leading order, i.e., we can evaluate them as
if spacetime were flat. Assume that the observer is at rest in the almost inertial
coordinates. To leading order, u = (1, 0, 0, 0) hence

d2 S
R00 e e S (8.54)
d 2
Using the formula for the perturbed Riemann tensor (8.5) and h0 = 0 we obtain

d2 S 1 2 h
e e S (8.55)
d 2 2 t2
In Minkowski spacetime we could take eai aligned with the x, y, z axes respectively,
i.e., e1 = (0, 1, 0, 0), e2 = (0, 0, 1, 0) and e3 = (0, 0, 0, 1). We can use the same
results here because we only need to evaluate e to leading order. Using h0 =
h3 = 0 we then have
d2 S 0 d2 S 3
= =0 (8.56)
d 2 d 2
to this order of approximation. Hence the observer sees no relative acceleration of
the test particles in the z-direction, i.e, the direction of propagation of the wave.
Let the observer set up initial conditions so that S0 and its first derivatives vanish
at = 0. Then S0 will vanish for all time. If the derivative of S3 vanishes initially
then S3 will be constant. The same is not true for the other components.
We can choose our almost inertial coordinates so that the observer has coordi-
nates x (, 0, 0, 0) (i.e. t = to leading order along the observers worldline).
For a + polarized wave we then have

d2 S 1 1 d2 S 2 1
2
= 2 |H+ | cos( )S1 , 2
= 2 |H+ | cos( )S2 (8.57)
d 2 d 2
where we have replaced t by in 2 h /t2 and = arg H+ . Since H+ is small
we can solve this perturbatively: the leading order solution is S1 = S1 , a constant
(assuming that we set up initial condition so that the particles are at rest to leading
order). Similarly S2 = S2 . Now we can plug these leading order solutions into the
RHS of the above equations and integrate to determine the solution up to first
order (again choosing constants of integration so that the particles would be at
rest in the absense of the wave)
   
1 1
S1 ( ) S1 1 + |H+ | cos( ) , S2 ( ) S2 1 |H+ | cos( )
2 2
(8.58)

Part 3 GR December 14, 2016 86 H.S. Reall

 8.3. GRAVITATIONAL WAVES

This reveals that particles are displaced outwards in the x-direction whilst being
displaced inwards in the y-direction, and vice-versa. S1 and S2 give the average
displacement. If the particles are arranged in the xy plane with S12 + S22 = R2 then
they form a circle of radius R when = + /2. This will be deformed into an
ellipse, then back to a circle, then an ellipse again (Fig 8.1).

= + 12 = + = + 32 = + 2

Figure 8.1: Geodesic deviation caused by + polarized wave.

Exercise. Show that the corresponding result for a polarized wave is the same,
just rotated through 45 (Fig. 8.2).

= + 12 = + = + 32 = + 2

Figure 8.2: Geodesic deviation caused by polarized wave.

Gravitational wave detectors look for the changes in position of test masses
caused by the above effect. For example, the two LIGO observatories (in the US,
see Fig. 8.3) each have two perpendicular tunnels, each 4 km long. There are test
masses (analogous to the particles above) at the end of each arm (tunnel) and
where the arms meet. A beam splitter is attached to the test mass where the arms
meet. A laser signal is split and sent down each arm, where it reflects off mirrors
attached to the test masses at the ends of the arms. The signals are recombined
and interferometry used to detect whether there has been any change in the length

Part 3 GR December 14, 2016 87 H.S. Reall

CHAPTER 8. LINEARIZED THEORY

Figure 8.3: The LIGO Hanford observatory in Washington state, USA. There is
another LIGO observatory in Louisiana. (Image credit: LIGO.)

difference of the two arms. The effect that is being looked for is tiny: plausible
sources of gravitational waves give H+ , H 1021 (see below) so the relative
length change of each arm is L/L 1021 . The resulting L is a tiny fraction of
the wavelength of the laser light but the resulting tiny phase difference between
the two laser signals is detectable!
Gravitational wave detectors have been operating for several decades, gradually
improving in sensitivity. The first direct detection of gravitational waves was
announced by the LIGO collaboration in February 2016. As we will explain below,
there is also very good indirect evidence for the existence of gravitational waves.
We will discuss all of this in more detail later.

8.4 The field far from a source

Lets return to the linearized Einstein equation with matter:
h = 16T (8.59)
Since each component of h satisfies the inhomogeneous wave equation, the so-
lution can be solved using the same retarded Green function that one uses in
electromagnetism:
T (t |x x0 |, x0 )
Z
h (t, x) = 4 d3 x0 (8.60)
|x x0 |

Part 3 GR December 14, 2016 88 H.S. Reall

 8.4. THE FIELD FAR FROM A SOURCE

where |x x0 | is calculated using the Euclidean metric.

Assume that the matter is confined to a compact region near the origin of size
d (e.g. let d be the radius of the smallest sphere that encloses the matter). Then,
far from the source we have r |x|  |x0 | d so we can expand
2
|x x0 |2 = r2 2x x0 + x0 = r2 1 (2/r)x x0 + O(d2 /r2 )


(8.61)
(where x = x/r) hence
|x x0 | = r x x0 + O(d2 /r) (8.62)
T (t |x x0 |, x0 ) = T (t0 , x0 ) + x x0 (0 T )(t0 , x0 ) + . . . (8.63)
where
t0 = t r (8.64)
Now let denote the time scale on which T is varying so 0 T T / . For
example, if the source is a binary star system, then is the orbital period. The
second term in (8.63) is of order (d/ )T . Note that d is the time it takes light
to cross the region containing the matter. Hence d/  1 will be satisfied if the
matter is moving non-relativistically. We assume this henceforth. This implies
that the second term in (8.63) is negligible compared to the first and so
Z
4
hij (t, x) d3 x0 Tij (t0 , x0 ) t0 = t r (8.65)
r
Here we are considering just the spatial components of h , i.e., hij . Other com-
ponents can be obtained from the gauge condition (8.16), which gives
0 h0i = j hji , 0 h00 = i h0i (8.66)
Given hij , the first equation can be integrated to determine h0i and the second
can then be integrated to determine h00 .
The integral in (8.65) can be evaluated as follows. Since the matter is compactly
supported, we can freely integrate by parts and discard surface terms (note also
that t0 does not depend on x0 ). We can also use energy-momentum conservation,
which to this order is just T = 0. Lets drop the primes on the coordinates in
the integral for now.
Z Z
3 ij
d3 x k (T ik xj ) (k T ik )xj
 
d xT =
Z
= d3 x (k T ik )xj drop surface term
Z
= d3 x(0 T i0 )xj conservation law
Z
= 0 d3 x T 0i xj (8.67)

Part 3 GR December 14, 2016 89 H.S. Reall

CHAPTER 8. LINEARIZED THEORY

We can now symmetrize this equation on ij to get

Z Z
3 ij
d xT = 0 d3 x T 0(i xj)
Z  
3 1 0k i j

1 0k i j
= 0 d x k T x x (k T )x x
2 2
Z
1
= 0 d3 x (k T 0k )xi xj integration by parts
2
Z
1
= 0 d3 x (0 T 00 )xi xj conservation law
2
Z
1
= 0 0 d3 x T 00 xi xj
2
1
= Iij (t) (8.68)
2
where Z
Iij (t) = d3 xT00 (t, x)xi xj (8.69)

(Note that T00 = T 00 and Tij = T ij to leading order.) Hence we have

2
hij (t, x) Iij (t r) (8.70)
r
This result is valid when r  d and  d.
Iij is the second moment of the energy density. It is a tensor in the Cartesian
sense, i.e., it transforms in the usual way under rotations R 3 of the coordinates xi .
(The zeroth moment R 3is the itotal energy in matter d xT00 , the first moment is
the energy dipole d xT00 x .)
The result (8.70) describes a disturbance propagating outwards from the source
at the speed of light. If the source exhibits oscillatory motion (e.g. a binary star
system) then hij will describe waves with the same period as the motion of the
source.
The first equation in (8.66) gives
 
2
0 h0i j Iij (t r) (8.71)
r

so integrating with respect to time gives (using i r = xi /r xi )

 
2 2xj 2xj
h0i j Iij (t r) = 2 Iij (t r) Iij (t r)
r r r
2xj
Iij (t r) (8.72)
r
Part 3 GR December 14, 2016 90 H.S. Reall
 8.4. THE FIELD FAR FROM A SOURCE

In the final line we have assumed that we are in the radiation zone r  . This
allows us to neglect the term proportional to I because it is smaller than the term
we have retained by a factor /r. In the radiation zone, space and time derivatives
are of the same order of magnitude. (Note that, even for a non-relativistic source,
the Newtonian approximation breaks down in the radiation zone because (8.20) is
violated.)
Similarly integrating the second equation in (8.66) gives
 
2xj 2xi xj
h00 i Iij (t r) Iij (t r) (8.73)
r r

These expressions are not quite right because when we integrated (8.66) we should
have included an arbitrary time-independent term in h0i , leading to a term in h00
linear in time, as well as an arbitrary time-independent term in h00 . The latter
term can be determined by returning to (8.60) which to leading order gives
4E
h00 (8.74)
r
where E is the total energy of the matter
Z
E = d3 x0 T00 (t0 , x0 ) (8.75)

Why have we not rediscovered the leading-order time-dependent piece (8.73)? This
piece is smaller than the time-independent part by a factor of d2 / 2 so wed have
to go to higher order in the expansion of (8.60) to find this term (and also the
term linear in t, which appears at higher order). Note that energy-momentum
conservation gives
Z Z
0 E = d x (0 T00 )(t , x ) = d3 x0 (i Ti0 )(t0 , x0 ) = 0
3 0 0 0
(8.76)

and hence E is a constant. Similarly,

4Pi
h0i (8.77)
r
where Pi is the total 3-momentum
Z
Pi = d3 x0 T0i (t0 , x0 ) (8.78)

This is also constant by energy-momentum conservation. Note that the term in

h00 that is linear in t is proportional to Pi .

Part 3 GR December 14, 2016 91 H.S. Reall

CHAPTER 8. LINEARIZED THEORY

Remark. We will show below that gravitational waves carry away energy (they
also carry away momentum). So why is the total energy of matter constant? In
fact, the total energy of matter is not constant but one has to go beyond linearized
theory to see this: one would have to correct the equation for energy-momentum
conservation to take account of the perturbation to the connection. But then we
would have to correct the LHS of the linearized Einstein equation, including second
order terms for consistency with the new conservation law of the RHS. So to see
this effect requires going beyond linearized theory.
A final simplification is possible: we are free to choose our almost inertial
coordinates to correspond to the centre of momentum frame, i.e., Pi = 0. If
we do this then E is just the total mass of the matter, which we shall denote M .
Putting everything together we have
4M 2xi xj 2xj
h00 (t, x) + Iij (t r), h0i (t, x) Iij (t r) (8.79)
r r r
To recap, we have assumed r   d and work in the centre of momentum frame.
In h00 we have retained the second term, even though it is subleading relative to
the first, because this is the leading order time-dependent term.

8.5 The energy in gravitational waves

We see that the gravitational waves arise when Iij varies in time. Gravitational
waves carry energy away from the souce. Calculating this is subtle: as discussed
previously, there is no local energy density for the gravitational field. To explain
the calculation we must go to second order in perturbation theory. At second
order, our metric + h will not satisfy the Einstein equation so we have to
add a second order correction, writing

g = + h + h(2)
(8.80)
(2)
The idea is that if the components of h are O() then the components of h are
O(2 ).
Now we calculate the Einstein tensor to second order. The first order term is
(1)
what we calculated before (equation (8.8)). We shall call this G [h]. The second
order terms are either linear in h(2) or quadratic in h. The terms linear in h(2) can
be calculated by setting h to zero. This is exactly the same calculation we did
before but with h replaced by h(2) . Hence the result will be (8.8) with h h(2) ,
(1)
which we denote G [h(2) ]. Therefore to second order we have

G [g] = G(1) (1) (2) (2)

[h] + G [h ] + G [h] (8.81)

Part 3 GR December 14, 2016 92 H.S. Reall

 8.5. THE ENERGY IN GRAVITATIONAL WAVES

(2)
where G [h] is the part of G that is quadratic in h. This is:
1 (1) 1 (2)
G(2) (2)
[h] = R [h] R [h]h R [h] (8.82)
2 2
(2)
where R [h] is the term in the Ricci tensor that is quadratic in h. R(1) and R(2)
are the terms in the Ricci scalar which are linear and quadratic in h respectively.
The latter can be written

R(2) [h] = R
(2)
[h] h R
(1)
[h] (8.83)

Exercise (examples sheet 3). Show that

(2) 1 1
R [h] = h h h ( h) + h h + h [ h]
2 4 
1 1 1
+ (h h ) h h h h ( h) (8.84)
2 4 2
For simplicity, assume that no matter is present. At first order, the linearized
(1)
Einstein equation is G [h] = 0 as before. At second order we have

G(1) (2)
[h ] = 8t [h] (8.85)

where
1 (2)
t [h] G [h] (8.86)
8
Equation (8.85) is the equation of motion for h(2) . If h satisfies the linear Einstein
(1)
equation then we have R [h] = 0 so the above results give
 
1 (2) 1 (2)
t [h] = R [h] R [h] (8.87)
8 2
Consider now the contracted Bianchi identity g G = 0. Expanding this, at
first order we get
G(1)
[h] = 0 (8.88)
for arbitrary first order perturbation h (i.e. not assuming that h satisfies any
equation of motion). At second order we get

G(1) (2) (2) (1)

[h ] + G [h] + hG [h] = 0 (8.89)

where the final term denotes schematically the terms that arise from the first order
change in the inverse metric and the Christoffel symbols in g . Now, since
(1)
(8.88) holds for arbitrary h, it holds if we replace h with h(2) so G [h(2) ] = 0.

Part 3 GR December 14, 2016 93 H.S. Reall

CHAPTER 8. LINEARIZED THEORY

If we now assume that h satisfies its equation of motion G(1) [h] = 0 then the final
term in (8.89) vanishes and this equation reduces to

t = 0. (8.90)

Hence t is a symmetric tensor (in the sense of special relativity) that is (i)
quadratic in the linear perturbation h, (ii) conserved if h satisfies its equation of
motion, and (iii) appears on the RHS of the second order Einstein equation (8.85).
This is a natural candidate for the energy momentum tensor of the linearized
gravitational field.
Unfortunately, t suffers from a major problem: it is not invariant under a
gauge transformation (8.14). This is how the impossibility of localizing gravita-
tional energy arises in linearized theory.
Nevertheless, it can be shown that the integral of t00 over a surface of constant
time t = x0 is gauge invariant provided one considers h that decays at infinity,
and restricts to gauge transformations which preserve this property. This integral
provides a satisfactory notion of the total energy in the linearized gravitational
field. Hence gravitational energy does exist, but it cannot be localized.
One can use the second order Einstein equation (8.85) to convert the integral
defining the energy, which is quadratic in h , into a surface integral at infinity
(2)
which is linear in h . In fact the latter can be made fully nonlinear: these surface
integrals make sense in any spacetime which is asymptotically flat, irrespective
of whether or not the linearized approximation holds in the interior. This notion
of energy is referred to as the ADM energy. This is constant in time but there is
a related quantity called the Bondi energy, a non-increasing function of time. The
rate of change of this can be interpreted as the rate of energy loss in gravitational
waves.
We shall follow a less rigorous, but more intuitive, approach in which we convert
t into a gauge-invariant quantity by averaging. For any point p, consider some
region R of R4 of typical coordinate size a (in all directions) centred on p. Define
the average of a tensor X at p by
Z
hX i = X (x)W (x)d4 x (8.91)
R
R
where the averaging function W (x) is positive, satisfies R W d4 x = 1, and tends
smoothly to zero on R. Note that it makes sense to integrate X because we are
treating it as a tensor in Minkowski spacetime, so we can add tensors at different
points.
We are interested in averaging in the region far from the source, in which we
have gravitational radiation with some typical wavelength (in the notation used
above ). Assume that the components of X have typical size x. Since the

Part 3 GR December 14, 2016 94 H.S. Reall

 8.5. THE ENERGY IN GRAVITATIONAL WAVES

wavelength of the radiation is , X will have components of typical size x/.

But the average is Z
h X i = X (x) W (x)d4 x (8.92)
R
where we have integrated by parts and used W = 0 on R. Now W has com-
ponents of order W/a so the RHS above has components of order x/a. Hence
if we choose a  then averaging has the effect of reducing X by a factor
of /a  1. So if we choose a  then we can neglect total derivatives inside
averages. This implies that we are free to integrate by parts inside averages:

hABi = h(AB)i h(A)Bi h(A)Bi (8.93)

because h(AB)i is a factor /a smaller than hBAi. Henceforth we assume a 

so we can exploit these properties.
Exercises (examples sheet 3).
1. Use the linearized Einstein equation to show that, in vacuum,

h R
(2)
[h]i = 0 (8.94)

Hence the second term in t [h] averages to zero.

2. Show that
1 1
ht i = h h h h h 2 h ( h) i (8.95)
32 2

3. Show that ht i is gauge invariant.

Hence we can obtain a gauge invariant energy momentum tensor as long as
we average over a region much larger than the wavelength of the the gravitational
radiation we are studying. This might be a rather large region: the LIGO detector
looks for waves with frequency around 100Hz, corresponding to a wavelength
3000km.
Exercise. Calculate ht i for the plane gravitational wave solution discussed
above. Show that

1 0 0 1
1 2 2 2
0 0 0 0
|H+ |2 + |H |2


ht i = |H+ | + |H | k k = 0 0 0 0

32 32
1 0 0 1
(8.96)
As one would expect, there is a constant flux of energy and momentum travelling
at the speed of light in the z-direction.

Part 3 GR December 14, 2016 95 H.S. Reall

CHAPTER 8. LINEARIZED THEORY

8.6 The quadrupole formula

Now we are ready to calculate the energy loss from a compact source due to
gravitational radiation. The averaged energy flux 3-vector is ht0i i. Consider a
large sphere r = constant far outside the source. The unit normal to the sphere
(in a surface of constant t) is xi . Hence the average total energy flux across this
sphere, i.e., the average power radiated across the sphere is
Z
hP i = r2 dht0i ixi (8.97)

where d is the usual volume element on a unit S 2 .

Calculating this is just a matter of substituting the results of section 8.4 into
(8.95). Since these results apply in harmonic gauge, we have
1 1
ht0i i = h0 h i h 0 hi hi
32 2
1 1
= h0 hjk i hjk 20 h0j i h0j + 0 h00 i h00 0 hi hi (8.98)
32 2
Since hjk (t, x) = (2/r)Ijk (t r) we have
2 ...
0 hjk = I jk (t r) (8.99)
r
and  
2 ... 2
i hjk = I jk (t r) 2 Ijk (t r) xi (8.100)
r r
The second term is smaller than the first by a factor /r  1 and so negligible for
large enough r. Hence
Z
1 1 ... ...
r2 dh0 hjk i hjk ixi = h I ij I ij itr (8.101)
32 2
On the RHS, the average is a time average, taken over an interval a 
centered on the retarded time t r.
Next we have h0j = (2xk /r)Ijk (t r) hence
2xk ... 2xk ...
0 h0j = I jk (t r) i h0j I jk (t r)xi (8.102)
r r
where in the second expressions we have used /r  1 to neglect the terms arising
from differentiation of xk /r. Hence
Z Z
1 2 1 ... ...
r dh20 h0j i h0j ixi = h I jk I jl itr dxk xl (8.103)
32 4

Part 3 GR December 14, 2016 96 H.S. Reall

 8.6. THE QUADRUPOLE FORMULA

R
Now recall the following from Cartesian tensors: dxk xl is isotropic (rotationally
invariant) and hence must equal kl for some constant . Taking the trace fixes
= 4/3. Hence the RHS above is
1 ... ...
h I ij I ij itr (8.104)
3
Next we use h00 = 4M/r + (2xj xk /r)Ijk (t r) to give
2xj xk ...
0 h00 = I jk (t r) (8.105)
r
and
 
4M 2xj xk ... 2xj xk ...
i h00 2 I jk (t r) xi I jk (t r)xi (8.106)
r r r
where weve neglected terms arising from differentiation of xj xk /r in the first
equality because in the radiation zone ( /r  1) theyre negligible compared to
the second term weve retained. In the second equality weve neglected the ...first
term in brackets because this leads to a term in the integral proportional to h I jk i,
which is the average of a derivative and therefore negligible. Hence we have
Z
1 1 ... ...
r2 dh0 h00 i h00 ixi = h I ij I kl itr Xijkl (8.107)
32 8
where Z
Xijkl = dxi xj xk xl (8.108)

is another isotropic integral which well evaluate below.

Next we use h = hjj h00 and the above results to obtain
2 ... 2xj xk ...
I jj (t r)
0 h = I jk (t r) (8.109)
r r
 
2 ... 2xj xk ...
i h = I jj (t r) + I jk (t r) xi (8.110)
r r
R
and hence (using the result above for dxi xj )
Z
1 1 1 ... ... 1 ... ... 1 ... ...
r2 dh 0 hi hixi = h I jj I kk + I jj I kk I ij I kl Xijkl i
32 2 4 6 16
(8.111)
Putting everything together we have
1 ... ... 1 ... ... 1 ... ...
hP it = h I ij I ij I ii I jj + I ij I kl Xijkl itr (8.112)
6 12 16
Part 3 GR December 14, 2016 97 H.S. Reall
CHAPTER 8. LINEARIZED THEORY

To evaluate Xijkl , we use the fact that any isotropic Cartesian tensor is a product
of factors and  factors. In the present case, Xijkl has rank 4 so we can only
use terms so we must have Xijkl = ij kl + ik jl + il jk for some , , .
The symmetry of Xijkl implies that = = . Taking the trace on ij and on kl
indices then fixes = 4/15. The final term above is therefore
1 ... ... ... ...
h I ii I jj +2 I ij I ij i (8.113)
60
and hence
1 ... ... 1 ... ...
hP it = h I ij I ij I ii I jj itr (8.114)
5 3
Finally we define the energy quadrupole tensor, which is the traceless part of Iij
1
Qij = Iij Ikk ij (8.115)
3
We then have
1 ... ...
hP it = hQij Qij itr (8.116)
5
This is the quadrupole formula for energy loss via gravitational wave emission. It
is valid in the radiation zone far from a non-relativistic source, i.e., for r   d.
We conclude that a body whose quadrupole tensor is varying in time will emit
gravitational radiation. A spherically symmetric body has Qij = 0 and so will not
radiate, in agreement with Birkhoff s theorem, which asserts that the unique spher-
ically symmetric solution of the vacuum Einstein equation is the Schwarzschild
solution. Hence the spacetime outside a spherically symmetric body is time inde-
pendent because it is described by the Schwarzschild solution.

8.7 Comparison with electromagnetic radiation

In electromagnetic theory, given a charge distribution we can define the total
charge Z
Q = d3 x (8.117)

and the electric dipole moment

Z
D= d3 xx (8.118)

Similarly for a matter distribution with energy density T00 we have defined the
total energy Z
E = d3 xT00 (8.119)

Part 3 GR December 14, 2016 98 H.S. Reall

 8.7. COMPARISON WITH ELECTROMAGNETIC RADIATION

and we can define the centre of mass

Z
1
X(t) = E d3 xT00 (t, x)x (8.120)

Electromagnetic radiation is produced by the motion of charge. Of course charge is

conserved so Q does not vary with time, just like E does not vary with time. Hence
there is no monpole radiation in either electromagnetism or gravity. However,
D can change with time, and changing D leads to emission of electromagnetic
radiation with power
1
hP it = h|D|2 itr (8.121)
120
Since the analogue of D is X, one might expect gravitational dipole radiation when
X varies. However, we have
E X = P = 0 (8.122)
where P is the total momentum of the mass distribution, which is conserved. Hence
there is no gravitational analogue of electric dipole radiation: it is forbidden by
linear momentum conservation.
In electromagnetic theory, a varying magnetic dipole moment also produces
radiation, although this is usually much weaker than electric dipole radiation.
The magnetic dipole is Z
= d3 x x j (8.123)

where j is the electric current. The analogue of a magnetic dipole moment for a
mass distribution is Z
J = d3 x x (u) (8.124)

where u is the local velocity of matter (i.e. T0i ui as in section 8.2). But this
is simply the total angular momentum of the matter, which is again conserved. So
there is no gravitational analogue of magnetic dipole radiation: it is forbidden by
conservation of angular momentum.
These arguments explain why there is no monopole or dipole gravitational
radiation. Gravitational quadrupole radiation is analogous to electric quadrupole
radiation in electromagnetic theory, which is the leading order effect when the
electric and magnetic dipoles do not vary.
It is easy to detect electromagnetic radiation but gravitational radiation is very
hard to detect. This is because gravitational waves interact only very weakly with
matter (or with each other). This is equivalent to the familiar statement that
gravity is very weak force - the weakest known force in Nature, and much weaker
than the electromagnetic force.

Part 3 GR December 14, 2016 99 H.S. Reall

CHAPTER 8. LINEARIZED THEORY

One way to see this is to consider the energy flux F of a plane gravitational
wave. For example, take a wave with h 1021 and 100Hz, the kind of
signal the LIGO detectors search for. From the 03 component of (8.96) we have,
reinstating factors of G and c to give quantity with the correct dimensions
2c 2
F h 0.01Wm2 (8.125)
32G
where we are just working to an order of magnitude. This is the same as the
energy flux around 30m from a 100W light bulb. Of course an electromagnetic
flux of this magnitude is easily detectable - your eyes are doing it now. (However,
the light has much higher frequency so maybe a better comparison is with 100Hz
electromagnetic waves, i.e., low frequency radio waves, and these would also be
easy to detect at a flux of 0.01Wm2 .) But to detect the same energy flux in
gravitational waves requires spending over a billion dollars to construct a state
of the art detector! A large energy flux produces only a very small effect on the
detector because gravity interacts with matter so weakly.
The weakness of gravity has some advantanges. Because gravitational waves
do not interact much with matter, they do not suffer much distortion as they prop-
agate through the Universe. Unlike electromagnetic waves, they are not absorbed
or scattered by matter. So the waves received by a detector are essentially the
same as the waves emitted by the source, adjusted for cosmological expansion.

8.8 Gravitational waves from binary systems

A fairly common astrophysical system with a time-varying quadrupole is a binary
system, consisting of a pair of stars orbiting their common centre of mass. Con-
sider the case when the stars both have mass M , their separation is d and the
orbital period is so the angular velocity is 1 . Then Newtons second
law gives M 2 d M 2 /d2 which gives M 1/2 d3/2 . The quadrupole tensor
has components of typical size M d2 so the third derivative is of size M d2 / 3
M d2 3 (M/d)5/2 . Hence we obtain the order of magnitude estimate
P (M/d)5 . (8.126)
The power radiated in gravitational waves is greatest when M/d is as large as
possible. Note the large power (5) on the RHS of this equation: if M/d decreases
by an order of magnitude then P decreases by 5 orders of magnitude. So P is a
rapidly decreasing function of M/d. To get a large P we need the system to have
M/d as small as possible, i.e., it has to be as compact as possible.
To understand the size of P , recall that weve used units G = c = 1 in (8.126).
Reinstating units gives
P (M/d)5 LPlanck (8.127)

Part 3 GR December 14, 2016 100 H.S. Reall

 8.8. GRAVITATIONAL WAVES FROM BINARY SYSTEMS

where LPlanck is the Planck luminosity

c5
LPlanck = 4 1052 W (8.128)
G
This is a mind-bogglingly enormous luminosity. The electromagnetic luminosity of
the Sun is L 4 1026 W 1026 LPlanck . There are roughly 1010 galaxies in the
observable Universe, so if a typical galaxy contains 1011 stars we can estimate the
electromagnetic luminosity of all the stars in the Universe as 1021 L 105 LPlanck .
>
Hence a binary with M/d 101 would emit more power in gravitational radiation
than all the stars in the Universe emit in electromagnetic radiation!
How big can M/d be? Obviously d cannot be smaller than the size R of the
stars themselves. However, a normal star has radius R  M . For example, the
Sun has R 7 105 km and M 1.5km so M/R 2 106 hence a binary made
of Sun-like stars would have M/d  106 as d  R. To obtain a larger amount
of radiation we need to consider binary systems made of much more compact
bodies, i.e., bodies with M/R as large as possible. The most compact objects in
Nature are black holes, whose size is given by the Schwarzschild radius R = 2M
(anything more compact than this would collapse to form a black hole: see the
Black Holes course). Almost as compact are neutron stars: stars made of nuclear
matter held together by gravity, like a giant atomic nucleus. So the binaries which
are expected to emit the most gravitational radiation are tightly bound NS/NS,
NS/BH or BH/BH systems (NS: neutron star, BH: black hole).
The emission of gravitational radiation causes the shape of the orbit to change
over time. To a good approximation, valid when the stars are far apart and moving
non-relativistically, we can calculate this by letting the radius of the Newtonian
orbit vary slowly with time. The energy of a Newtonian orbit is E M 2 /d
so d decreases as the system loses energy via gravitational radiation. Hence the
orbital period d3/2 M 1/2 also decreases. To calculate how d varies with time,
we equate dE/dt with P . (See Examples sheet 4.) This approximation can be
improved by including higher order, post-Newtonian, effects.
This prediction of GR has been confirmed observationally. In 1974, Hulse
and Taylor detected a binary pulsar. This is a NS/NS binary in which one of
the stars is a pulsar, i.e., it emits a beam of radio waves in a certain direction.
This star is rotating very rapidly and the beam (which is not aligned with the
rotation axis) periodically points in our direction. Hence we receive pulses of
radiation from the star. The period between successive pulses (about 0.05s) is
very stable and has been measured to very high accuracy. Therefore it acts like
a clock that we can observe from Earth. Using this clock we can determine the
orbital period (about 7.75h) of the binary system, again with good accuracy. The
emission of gravitational waves causes the period of the orbit to decrease by about
10s per year. This small effect has been measured and the result confirms the

Part 3 GR December 14, 2016 101 H.S. Reall

CHAPTER 8. LINEARIZED THEORY

quadrupole formula to an accuracy of 0.3% (the accuracy increases the longer the
system is observed). This is very strong indirect evidence for the existence of
gravitational waves, for which Hulse and Taylor received the Nobel Prize in 1993.
(The gravitational wave luminosity of the Hulse-Taylor system is about 0.02L .)
As a compact binary system loses energy to radiation, the radius of the orbit
shrinks and eventually the two bodies in the system will collide and merge to form
a single black hole (it is unlikely to be a neutron star because a NS cannot have a
mass greater than about 2M ). As the system approaches merger, the velocities
of the two bodies become very large, a significant fraction of the speed of light.
For such a system, the post-Newtonian expansion is useless and the only way of
predicting what will happen is to solve the Einstein equation numerically on a
supercomputer. As the bodies approach merger, the luminosity P still increases in
rough agreement with (8.126). Hence the strongest sources of gravitational waves
are expected to be compact binaries just before merger.
To discuss the direct detection of gravitational waves from a merging compact
binary, we need to estimate the amplitude of the gravitational waves from such a
source. At a distance r, the above estimates give
M d2 M2
hij (8.129)
2r dr
We can use this to estimate the amplitude of waves arriving at a detector on Earth.
We take r to be the distance within which we expect there to exist sufficiently many
suitable binary system that at least one will merge within a reasonable time, say
1 year (we dont want to have to wait for 100 years to detect anything!). The
process of gradual inspiral to final merger is very slow, taking billions of years for
plausible initial conditions (see Examples sheet 4). This implies that r must be of
cosmological size: of the order of 3 108 light years. Taking M to be about ten
times the mass of the Sun and d to be ten times the Schwarzschild radius gives
h 1021 and waves with a frequency of 100 1000 Hz. This is the kind of signal
that the two LIGO detectors (in the US) and the VIRGO detector (in Italy) are
designed to detect.
On 11 February 2016, the LIGO collaboration announced that it had made the
first direct detection of gravitational waves on 14 September 2015 (Fig. 8.4). By
comparing with the detailed predictions of General Relativity (determined using
the post-Newtonian expansion and numerical simulations), it was deduced that
these waves were emitted in the merger of a compact binary at a distance of
1.4 109 light years. The masses of the objects in the binary were estimated to be
36M and 29M . Since these are well above the upper mass limit for a neutron
star, it was deduced that this was a BH/BH binary. The merger produced a
final BH with mass 62M . The missing mass 3M was emitted as gravitational
radiation. The detected signal lasted for about 0.5s. The gravitational wave

Part 3 GR December 14, 2016 102 H.S. Reall

 8.8. GRAVITATIONAL WAVES FROM BINARY SYSTEMS

luminosity at the merger was greater than the electromagnetic luminosity of all
the stars in the visible Universe!

Figure 8.4: The signals detected by the two LIGO observatories on 14 September
2015. The vertical axis is the change in length of the arms of the detectors. (Image
credit: LIGO.)

Gravitational waves from a second BH/BH merger were detected on 26 Decem-

ber 2015, this time with masses 14M and 8M merging to form a BH of mass
21M , at a similar distance to the first detection. Based on these detections, it
is expected that the LIGO detectors will detect roughly one BH/BH merger per
month at their current sensitivity. Upgrades to the sensitivity are planned, which
could lead to 1 detection per day by 2019!

Part 3 GR December 14, 2016 103 H.S. Reall

CHAPTER 8. LINEARIZED THEORY

Part 3 GR December 14, 2016 104 H.S. Reall

Chapter 9

Differential forms

9.1 Introduction
Definition. Let M be a differentiable manifold. A p-form on M is an antisym-
metric (0, p) tensor field on M .
Remark. A 0-form is a function, a 1-form is a covector field.
Definition. The wedge product of a p-form X and a q-form Y is the (p + q)-form
X Y defined by
(p + q)!
(X Y )a1 ...ap b1 ...bq = X[a1 ...ap Yb1 ...bq ] (9.1)
p!q!

Exercise. Show that

X Y = (1)pq Y X (9.2)
(so X X = 0 if p is odd); and

(X Y ) Z = X (Y Z) (9.3)

i.e. the wedge product is associative so we dont need to include the brackets.
Remark. If we have a dual basis {f } then the set of p-forms of the form f 1
f 2 . . . f p give a basis for the space of all p-forms on M because we can write
1
X= X ... f 1 f 2 . . . f p (9.4)
p! 1 p

Definition. The exterior derivative of a p-form X is the (p + 1)-form dX defined

in a coordinate basis by

(dX)1 ...p+1 = (p + 1)[1 X2 ...p+1 ] (9.5)

105
CHAPTER 9. DIFFERENTIAL FORMS

Exercise. Show that this is independent of the choice of coordinate basis.

Remark. This reduces to our previous definition of d acting on functions when
p = 0.
Lemma. If is a (torsion-free) connection then

(dX)a1 ...ap+1 = (p + 1)[a1 Xa2 ...ap+1 ] (9.6)

Proof. The components of the LHS and RHS are equal in normal coordinates at r
for any point r.
Exercises (examples sheet 4). Show that the exterior derivative enjoys the fol-
lowing properties:
d(dX) = 0 (9.7)
d(X Y ) = (dX) Y + (1)p X dY (9.8)
(where Y is a q-form) and
d( X) = dX (9.9)
(where : N M ), i.e. the exterior derivative commutes with pull-back.
Remark. The last property implies that the exterior derivative commutes with a
Lie derivative:
LV (dX) = d(LV X) (9.10)
where V is a vector field.
Definition. X is closed if dX = 0 everywhere. X is exact if there exists a
(p 1)-form Y such that X = dY everywhere.
Remark. Exact implies closed. The converse is true only locally:
Poincare Lemma. If X is a closed p-form (p 1) then for any r M there
exists a neighbourhood O of r and a (p 1)-form Y such that X = dY in O.

9.2 Connection 1-forms

Remark. Often in GR it is convenient to work with an orthonormal basis of
vector fields {ea } obeying
gab ea eb = (9.11)
where = diag(1, 1, . . . , 1). (In 4d, such a basis is sometimes called a tetrad.)
(On a Riemannian manifold we do the same with replaced by .) Since
is the metric in our basis, Greek tensor indices can be lowered with and raised
with .

Part 3 GR December 14, 2016 106 H.S. Reall

 9.2. CONNECTION 1-FORMS

Exercise. Show that the dual basis {fa } is given by

fa = (e )a ea ea ea = (9.12)

Remark. Here we have defined what it means to raise the Greek index labeling
the basis vector. Henceforth any Greek index can be raised/lowered with the
Minkowski metric.
Remark. We saw earlier (section 3.2) that any two orthonormal bases at are
related by a Lorentz transformation which acts on the indices , :
a

e0 = A1 ea , A A = (9.13)

There is an important difference with special relativity: the Lorentz transformation

A need not be constant, it can depend on position. Working with orthonormal
bases brings the equations of GR to a form in which Lorentz transformations are
a local symmetry.
Lemma.
ea eb = gab , ea eb = ab (9.14)
Proof. Contract the LHS of the first equation with a basis vector eb :

ea eb eb = ea = ea = (e )a = gab eb (9.15)

Hence the first equation is true when contracted with any basis vector so it is true
in general. This equation can be written as ea (e )b = gab . Raise the b index to get
the second equation.
Definition. The connection 1-forms are (using the Levi-Civita connection)

( )a = eb a eb (9.16)

Exercise. Show that

( )a = ea (9.17)
where are the Christoffel symbols.
Remark. The indices , on are not tensor indices: they do not transform
correctly under Lorentz transformations. This is just the fact that the components
of the connection are not tensor components. However, for each pair (, ), we do
have a well-defined 1-form .
Lemma. ( )a = ( )a .
Proof.

( )a = (e )b a eb = a (e )b eb eb a (e )b = a ( )a


(9.18)

Part 3 GR December 14, 2016 107 H.S. Reall

CHAPTER 9. DIFFERENTIAL FORMS

and a = 0 because are just constant scalars.

Lemma. Regard ea as a 1-form. Then

de = e (9.19)

Proof. From the definition of we have

a eb = ( )a eb (9.20)

hence
a (e )b = ( )a eb = ( )a eb (9.21)
and so
(de )ab = 2[a eb] = 2 ( )[a eb] = ( e )ab (9.22)

Remark. Evaluating (9.19) in our basis gives (cf (9.4))

de = ( ) e e = ( [ )] e e (9.23)

and hence
(de ) = 2( [ )] . (9.24)
So if we calculated de then we can read off ( [ )] . If we do this for all then
the connection 1-forms can be read off because the antisymmetry =
implies ( ) = ([ )] + ([ )] ([ )] . Since calculating de is often quite
easy, this provides a convenient method of calculating the connection 1-forms. In
simple cases, one can read off by inspection. It is always a good idea to check
the result by substituting back into (9.19).
Example. The Schwarzschild spacetime admits the obvious tetrad

e0 = f dt, e1 = f 1 dr, e2 = rd, e3 = r sin d (9.25)

p
where f = 1 2M/r. We then have

de0 = df dt + f d(dt) = f 0 dr dt = f 0 e1 e0 (9.26)

de1 = f 2 f 0 dr dr = 0 (9.27)
de2 = dr d = (f /r)e1 e2 (9.28)
de3 = sin dr d + r cos d d = (f /r)e1 e3 + (1/r) cot e2 e3 (9.29)
The first of these suggests we try 0 1 = f 0 e0 . This would give 01 = f 0 e0
and hence 10 = f 0 e0 so 1 0 = f 0 e0 . This would make a vanishing contribu-
tion to de1 ( 1 0 e0 = 0), which looks promising. The third equation suggests

Part 3 GR December 14, 2016 108 H.S. Reall

 9.3. SPINORS IN CURVED SPACETIME

we try 2 1 = (f /r)e2 , which gives 1 2 = (f /r)e2 which again would make a

vanishing contribution to de1 . The final equation suggests 3 1 = (f /r)e3 and
3 2 = (1/r) cot e3 and again these will not spoil the second and third equations.
So these connection 1-forms (and those related by = ) must be the correct
answer.
Remark. From (4.8), the components of the covariant derivative of a vector field
Y a are
Y = e (Y ) + V = V + Y (9.30)
where e , refers to a coordinate basis, and ( ) .
Exercise. What is the corresponding result for a (1, 1) tensor field?

9.3 Spinors in curved spacetime

Remark. Weve seen how to define tensors in curved spacetime. But what about
spinor fields, i.e., fields of non-integer spin? If we use orthonormal bases, this is
straightforward because the structure of special relativity is manifest locally.
Remark. For now, we work at a single point p. Consider an infinitesimal Lorentz
transformation at p
A = + (9.31)
for infinitesimal . The condition that this be a Lorentz transformation (the
second equation in (9.13) gives the restriction
= (9.32)
So an infinitesimal Lorentz transformation at p is described by an antisymmetric
matrix. We now consider a representation of the Lorentz group at p in which the
Lorentz transformation A is described by a matrix D(A).
Definition. The generators of the Lorentz group in the representation D are
matrices T = T defined by
1
D(A) = 1 + T (9.33)
2
when A is given by (9.31).
Example. Lorentz transformations were defined by looking at transformations
of vectors. Lets work out the generators in this defining, vector representation.
Under a Lorentz transformation, the components of a vector transform as X 0 =
A X = X + X so we must have
1
(T ) X = X = X (9.34)
2
Part 3 GR December 14, 2016 109 H.S. Reall
CHAPTER 9. DIFFERENTIAL FORMS

and hence, remembering the antisymmetry of , the Lorentz generators in the

vector representation, which we denote M , have components

(M ) = (9.35)

From these we deduce the Lorentz algebra (the square brackets denote a matrix
commutator)
[M , M ] = . . . (9.36)

Remark. A finite Lorentz transformation can be obtained by exponentiating:

 
1
A = exp M (9.37)
2

We then have  
1
D(A) = exp T (9.38)
2
In these expressions, = are finite parameters describing the transforma-
tion.
Definition. The Dirac gamma matrices are a set of square matrices { } which
obey
+ = 2 (9.39)

Remark. In 4d spacetime, the smallest representation of the gamma matrices is

given by the Dirac representation in which are 4 4 matrices. It is unique up
to equivalence of representations.
Lemma. The matrices
1
T = [ , ] (9.40)
4
form a representation of the Lorentz algebra (9.36). This is the Dirac representa-
tion describing a particle of spin 1/2.
Remark. So far, weve worked at a single point p. Lets now allow p to vary. The
following definition looks more elegant if one uses the language of vector bundles.
Definition. A field in the Lorentz representation D is a smooth map which, for
any point p and any orthonormal basis {e } defined in a neighbourhood of p, gives
a vector in the carrier space of the representation D, with the property that if
(p, {e }) maps to then (p, {e0 }) maps to D(A) where {e0 } is related to {e }
by the Lorentz transformation A.
Example. Take D to be the vector representation. Then the carrier space is just
Rn so we can denote the resulting vector . Then D(A) = A and so under a

Part 3 GR December 14, 2016 110 H.S. Reall

 9.4. CURVATURE 2-FORMS

change of basis we have 0 = A , which is just the usual transformation law

for the components of a vector.
Remark. Given a field transforming in some representation D, we can take a
partial derivative of its components in a coordinate basis, and then convert the
result to our orthonormal basis, i.e., e where refers to the coordinate
basis. However, since the matrix A can depend on position, the partial derivative of
our field will no longer transform homogeneously under a Lorentz transformation:
the result will involve derivatives of A. For tensor fields, we know how to resolve
this problem: introduce the covariant derivative. So now we need to extend the
definition of covariant derivative to a general representation D:
Definition. The covariant derivative of a field transforming in a representation
of the Lorentz group with generators T is, in an orthonormal basis,
1
= + ( ) T (9.41)
2

Remark. One can show that this does indeed transform correctly, i.e., in a rep-
resentation of the Lorentz group, under Lorentz transformations. The connection
1-forms are sometimes referred to as the spin connection because of their role in
defining the covariant derivative for spinor fields.
Definition. The Dirac equation for a spin 1/2 field of mass m is

m = 0. (9.42)

9.4 Curvature 2-forms

Definition. Consider a spacetime with an orthonormal basis. The curvature
2-forms are
1
= R e e (9.43)
2
Remark. The antisymmetry of the Riemann tensor implies = .

Lemma. The curvature 2-forms are given in terms of the connection 1-forms by

= d + (9.44)

Proof. Optional exercise. Direct calculation of the RHS, using the relation (9.17)
and equation (9.19). Youll need to work out the generalization of (6.6) to a
non-coordinate basis.

Part 3 GR December 14, 2016 111 H.S. Reall

CHAPTER 9. DIFFERENTIAL FORMS

Remark. This provides a convenient way of calculating the Riemann tensor in an

orthonormal basis. One calculates the connection 1-forms using (9.19) and then the
curvature 2-forms using (9.43). The components of are R so one can read
off the Riemann tensor components. The only derivatives one needs to calculate
are exterior derivatives, which are usually fairly easy. In simple situations, this is
much faster than calculating the Riemann tensor in a coordinate basis using (6.6).
Example. We determined the connection 1-forms in the Schwarzschild spacetime
previously. From 0 1 = f 0 e0 we have
2
d 0 1 = f 0 de0 + f 00 dr e0 = f 0 e1 e0 + f f 00 e1 e0 (9.45)

we also have
0 1 = 01 11 = 0 (9.46)
where the first equality arises because 0 is non-zero only for = 1 and the second
equality is because 1 1 = 11 = 0 (by antisymmetry). Combining these results we
have  
2
01 = 0 1 = f f 00 + f 0 e0 e1 (9.47)
and hence the only non-vanishing components of the form R01 are

00 02
 1 2
00 2M
R0101 = R0110 = f f + f = f = 3 (9.48)
2 r

Exercise (examples sheet 4). Determine the remaining curvature 2-forms 02 ,

03 , 12 , 13 , 23 (all others are related to these by (9.44)). Hence determine the
Riemann tensor components. Check that the Ricci tensor vanishes.

9.5 Volume form

Definition. A manifold of dimension n is orientable if it admits an orientation: a
smooth, nowhere vanishing n-form a1 ...an . Two orientations  and 0 are equivalent
if 0 = f  where f is an everywhere positive function.
Remark. Any n-form X is related to  by X = f  for some function f . X will
define an orientation provided f does not vanish anywhere. Hence an orientable
manifold admits precisely two inequivalent orientations, corresponding to the cases
f > 0 and f < 0.
Definition. A coordinate chart x on an orientable manifold is right-handed with
respect to  iff  = f (x)dx1 dx2 . . . dxn with f (x) > 0. It is left-handed if
f (x) < 0.

Part 3 GR December 14, 2016 112 H.S. Reall

 9.5. VOLUME FORM

Definition. On an oriented manifold with a metric, the volume form is defined

by p
12...n = |g| (9.49)
in any right-handed coordinate chart, where g denotes the determinant of the
metric in this chart.
Exercise. 1. Show that this definition is chart-independent. 2. Show that (in a
RH coordinate chart)
1
12...n = p (9.50)
|g|
where the upper (lower) sign holds for Riemannian (Lorentzian) signature.
Lemma.
a
a1 ...ap cp+1 ...cn b1 ...bp cp+1 ...cn = p!(n p)![ba11 . . . bpp] (9.51)
where the upper (lower) sign holds for Riemannian (Lorentzian) signature.
Proof. Exercise.
Definition. On an oriented manifold with metric, the Hodge dual of a p-form X
is the (n p)-form ? X defined by

1
(? X)a1 ...anp = a ...a b ...b X b1 ...bp (9.52)
p! 1 np 1 p

Lemma. For a p-form X,

?(? X) = (1)p(np) X (9.53)

(? d ? X)a1 ...ap1 = (1)p(np) b Xa1 ...ap1 b (9.54)

where the upper (lower) sign holds for Riemannian (Lorentzian) signature.
Proof. Exercise (use (9.51)).
Examples.

1. In 3d Euclidean space, the usual operations of vector calculus can be written

using differential forms as

f = df div X = ? d ? X curl X = ? dX (9.55)

where f is a function and X denotes the 1-form Xa dual to a vector field

X a . The final equation shows that the exterior derivative can be thought of
as a generalization of the curl operator.

Part 3 GR December 14, 2016 113 H.S. Reall

CHAPTER 9. DIFFERENTIAL FORMS

2. Maxwells equations are a Fab = 4jb and [a Fbc] = 0 where j a is the

current density vector. These can be written as

d ? F = 4 ? j, dF = 0 (9.56)

The first of these implies d ? j = 0, which is equivalent to a j a = 0, i.e., j a

is a conserved current. The second of these implies (by the Poincare lemma)
that locally there exists a 1-form A such that F = dA.

9.6 Integration on manifolds

Definition. Let M be an oriented manifold of dimension n. Let : O U be a
RH coordinate chart with coordinates x and let X be a n-form. The integral of
X over O is Z Z
X dx1 . . . dxn X12...n (9.57)
O U

Exercise. Show that this is chart independent, i.e., if one replaces with another
RH coordinate chart 0 : O U 0 then one gets the same result.
Remark. How do we extend our definition to all of M ? The idea is just to chop
M up into regions such that we can use the above definition on each region, then
sum the resulting terms.
Definition. Let the RH charts : O U be an atlas for M . Introduce a
unity, i.e., a set of functions h : M [0, 1] such that h (p) = 0 if
partition of P
p/ O , and h (p) = 1 for all p. We then define
Z XZ
X h X (9.58)
M O

Remark.

1. It can be shown that this definition is independent of the choice of atlas and
partition of unity.

2. A diffeomorphism : M M is orientation preserving if () is equivalent

to  for any orientation . It is not hard to show that the integral is invariant
under orientation preserving diffeomorphisms:
Z Z

(X) = X (9.59)
M M

Part 3 GR December 14, 2016 114 H.S. Reall

 9.7. SUBMANIFOLDS AND STOKES THEOREM

Definition. Let M be an oriented

R manifold with a metric g. Let  be the volume
form. The volume of M is M . If f is a function on M then
Z Z
f f (9.60)
M M

Remark. We shall sometimes use the notation

Z Z p
f= dn x |g|f (9.61)
M M

This is an abuse of notation because the RHS refers to coordinates x but M might
not be covered by a single chart. It has the advantage of making it clear that the
integral depends on the metric tensor.

9.7 Submanifolds and Stokes theorem

Definition. Let S and M be oriented manifolds of dimension m and n respectively,
m < n. A smooth map : S M is an embedding if it is one-to-one ((p) 6= (q)
for p 6= q, i.e. [S] does not intersect itself) and for any p S there exists a
neighbourhood O such that 1 : [O] S is smooth. If is an embedding then
[S] is an embedded submanifold of M . A hypersurface is an embedded submanifold
of dimension n 1.
Remark. The technical details here are included for completeness, we wont need
to refer to them again. Henceforth, we will simply say submanifold instead of
embedded submanifold.
Definition. If [S] is a m-dimensional submanifold of M and X is a m-form on
M then the integral of X over [S] is
Z Z
X (X) (9.62)
[S] S

Remark. If X = dY then the fact that d and commute gives

Z Z
dY = d ( (Y )) (9.63)
[S] S

Definition. A manifold with boundary M is defined in the same way as a manifold

except that charts are maps : O U where now U is an open subset of
1 n
2
R = {(x1 , . . . , xn ) Rn : x1 0}. The boundary of M , denoted M , is the set

Part 3 GR December 14, 2016 115 H.S. Reall

CHAPTER 9. DIFFERENTIAL FORMS

of points for which x1 = 0. This is a manifold of dimension n 1 with coordinate

charts (x2 , . . . xn ). If M is oriented then the orientation of M is fixed by saying
that (x2 , . . . xn ) is a RH chart on M when (x1 , . . . , xn ) is a RH chart on M .
Stokes theorem. Let N be an oriented n-dimensional manifold with boundary
and X a (n 1)-form. Then
Z Z
dX = X (9.64)
N N

Remarks. We define the RHS by regarding N as a hypersurface in N (the map

is just the inclusion map) and using (9.62). We often use this result when N is
some region of a larger manifold M . Then N is a hypersurface in M .
Example. Let be a hypersurface in a spacetime M and consider a solution of
Maxwells equations (9.56)). Assume that has a boundary. Then
Z Z Z
1 1
?F = d ? F = ?j Q[] (9.65)
4 4

The final equality defines the total charge on . Hence we have a formula relating
the charge on to the flux through the boundary of . This is Gauss law.
Definition. X Tp (M ) is tangent to [S] at p if X is the tangent vector at p of
a curve that lies in [S]. n Tp (M ) is normal to a submanifold [S] if n(X) = 0
for any vector X tangent to [S] at p.
Remark. The vector space of tangent vectors to [S] at p has dimension m. The
vector space of normals to [S] at p has dimension n m. Any two normals to a
hypersurface are proportional to each other.
Definition. A hypersurface in a Lorentzian manifold is timelike, spacelike or null
if any normal is everywhere spacelike, timelike or null respectively.
Remark. Let M is a manifold with boundary and consider a curve in M with
parameter t and tangent vector X. Then x1 (t) = 0 so

dx1
dx1 (X) = X(x1 ) = = 0. (9.66)
dt
Hence dx1 (X) vanishes for any X tangent to M so dx1 is normal to M . Any
other normal to M will be proportional to dx1 . If M is timelike or spacelike
then we can construct a unit normal by dividing by the norm of dx1 :

(dx1 )a
na = p g ab na nb = 1 (9.67)
g bc (dx1 )b (dx1 )c

Part 3 GR December 14, 2016 116 H.S. Reall

 9.7. SUBMANIFOLDS AND STOKES THEOREM

One can show that this is chart independent. Here we choose the + sign if dx1 is
spacelike and the sign if dx1 is timelike (+ if the metric is Riemannian). Note
that na points out of M if M is timelike (or the metric is Riemannian) but into
M if M is spacelike. This is to get the correct sign in the divergence theorem:
Divergence theorem. If M is timelike or spacelike then
Z p Z p
n a
d x |g| a X = dn1 x |h| na X a (9.68)
M M

where X a is a vector field on M , is the Levi-Civita connection, and h denotes

the determinant of the metric on M induced by pulling back the metric on M .
na X a is a scalar in M so it can be pulled back to M , this is the integrand on the
RHS.
Proof. Follow through the definitions, usingp(9.54), (9.53) and Stokes theorem.
The volume form of M is  where 2...n = |h| in one of the coordinate charts
occuring in the definition of a manifold with boundary. In such a chart, the
components h are the same as the components g with 2 , n. Using
this, g ab (dx1 )a (dx1 )b = g 11 = h/g.

Part 3 GR December 14, 2016 117 H.S. Reall

CHAPTER 9. DIFFERENTIAL FORMS

Part 3 GR December 14, 2016 118 H.S. Reall

Chapter 10

Lagrangian formulation

10.1 Scalar field action

You are familiar with the idea that the equation of motion of a point particle can
be obtained by extremizing an action. You may also know that the same is true
for fields in Minkowski spacetime. The same is true in GR. To see how this works,
consider first a scalar field, i.e., a function : M R and define the action as
the functional

Z
S[] = d4 x gL (10.1)
M

where L is the Lagrangian:

1
L = g ab a b V () (10.2)
2

and V () is called the scalar potential. Now consider a variation + for

some function that vanishes on M (in an asymptotically flat spacetime, M
will be at infinity). The change in the action is (working to linear order in )

S = S[ + ] S[]

Z
d4 x g g ab a b V 0 ()


=
ZM

= d4 x g [a (a ) + a a V 0 ()]
ZM

p Z
= 3 a
d x |h| na + d4 x g (a a V 0 ())
ZM M

= d4 x g (a a V 0 ()) (10.3)
M

119
CHAPTER 10. LAGRANGIAN FORMULATION

Note that we have used the divergence theorem to integrate by parts. A formal
way of writing the final expression is
Z
S
S = d4 x (10.4)
M
where
S
g (a a V 0 ()) (10.5)


The factor of g here means that this quantity is not a scalar (it is an example

of a scalar density). However (1/ g)S/ is a scalar. Weve written things
in this strange way in order to be consistent with how we treat the gravitational
field.
Demanding that S vanishes for arbitrary gives us the equation of motion
S/ = 0, i.e.,
a a V 0 () = 0. (10.6)
The particular choice V () = 21 m2 2 gives the Klein-Gordon equation.

10.2 The Einstein-Hilbert action

For the gravitational field, we seek an action of the form

Z
S[g] = d4 x gL (10.7)
M

where L is a scalar constructed from the metric. An obvious choice for the La-
grangian is L R. This gives the Einstein-Hilbert action

Z Z
1 4 1
SEH [g] = d x gR = R (10.8)
16 M 16 M
where the prefactor is included for later convenience and  is the volume form. The
idea is that we regard our manifold M as fixed (e.g. R4 ) and gab is determined by
extremizing S[g]. In other words, we consider two metrics gab and gab + gab and
demand that S[g + g] S[g] should vanish to linear order in gab . Note that gab
is the difference of two metrics and hence is a tensor field.
We need to work out what happens to  and R when we vary g . Recall the
formula for the determinant expanding along the th row:
X
g= g (10.9)

where we are suspending the summation convention, is any fixed value, and
is the cofactor matrix, whose element is (1)+ times the determinant of the

Part 3 GR December 14, 2016 120 H.S. Reall

 10.2. THE EINSTEIN-HILBERT ACTION

matrix obtained by deleting row and column from the metric. Note that
is independent of the element of the metric. Hence

g
= = gg (10.10)
g

where on the RHS we recall the formula for the inverse matrix g in terms of the
cofactor matrix. We can use this formula to determine how g varies under a small
change g in g (reinstating the summation convention):

g
g = g = gg g = gg ab gab (10.11)
g

(we can use abstract indices in the final equality since g ab gab is a scalar) and hence
1
g = g g ab gab (10.12)
2
From the definition of the volume form we have
1
 = g ab gab (10.13)
2
Next we need to evaluate R. To this end, consider first the change in the Christof-
fel symbols. is the difference between the components of two connections (i.e.
the Levi-Civita connections associated to gab + gab and gab ). Since the difference
of two connections is a tensor, it follows that are components of a tensor
abc . These components are easy to evaluate if we introduce normal coordinates
at p for the unperturbed connection: at p we have g, = 0 and = 0. For the
perturbed connection we therefore have, at p, (to linear order)

1
= g (g, + g, g, )
2
1
= g (g; + g; g; ) (10.14)
2
In the second equality, the semi-colon denotes a covariant derivative with respect
to the Levi-Civita connection associated to gab . The two expressions are equal
because (p) = 0. The LHS and RHS are tensors so this is a basis independent
result hence we can use abstract indices:
1
abc = g ad (gdb;c + gdc;b gbc;d ) (10.15)
2
p is arbitrary so this result holds everywhere.

Part 3 GR December 14, 2016 121 H.S. Reall

CHAPTER 10. LAGRANGIAN FORMULATION

Now consider the variation of the Riemann tensor. Again it is convenient to

use normal coordinates at p, so at p we have (using () = 0 at p)

R =
= (10.16)

where is the Levi-Civita connection of gab . Once again we can immediately

replace the basis indices by abstract indices:

Ra bcd = c abd d abc (10.17)

and p is arbitrary so the result holds everywhere. Contracting gives the variation
of the Ricci tensor:
Rab = c cab b cac (10.18)
Finally we have
R = (g ab Rab ) = g ab Rab + g ab Rab (10.19)
where g ab is the variation in g ab (not the result of raising indices on gab ). Using
(g g ) = ( ) = 0 it is easy to show (exercise)

g ab = g ac g bd gcd (10.20)

Putting everything together, we have

R = g ac g bd Rab gcd + g ab (c cab b cac )

= Rab gab + c (g ab cab ) b (g ab cac )
= Rab gab + a X a (10.21)

where
X a = g bc abc g ab cbc (10.22)
Hence the variation of the Einstein-Hilbert action is
Z
1
SEH = ( R)
16 M
Z  
1 1 ab ab a
=  Rg gab R gab + a X
16 M 2

Z  
1 4 1 ab ab a
= d x g Rg gab R gab + a X (10.23)
16 M 2

The final term can be converted to a surface term on M using the divergence
theorem. If we assume that gab has support in a compact region that doesnt

Part 3 GR December 14, 2016 122 H.S. Reall

 10.3. ENERGY MOMENTUM TENSOR

intersect M then this term will vanish (because vanishing of gab and its derivative
on M implies that X a will vanish on M ). Hence we have

Z Z
1 4 ab

SEH
SEH = d x g G gab = gab (10.24)
16 M M gab

where Gab is the Einstein tensor and

SEH 1
= gGab (10.25)
gab 16

Hence extremization of SEH reproduces the vacuum Einstein equation.

Exercise. Show that the vacuum Einstein equation with cosmological constant is
obtained by extremizing

Z
1
SEH = d4 x g (R 2) (10.26)
16 M

Remark. The Palatini procedure is a different way of deriving the Einstein equa-
tion from the Einstein-Hilbert action. Instead of using the Levi-Civita connec-
tion, we allow for an arbitrary torsion-free connection. The EH action is then a
functional of both the metric and the connection, which are to be varied inde-
pendently. Varying the metric gives the Einstein equation (but written with an
arbitrary connection). Varying the connection implies that the connection should
be the Levi-Civita connection. When matter is included, this works only if the
matter action is independent of the connection (as is the case for a scalar field or
Maxwell field) or if the Levi-Civita connection is used in the matter action.

10.3 Energy momentum tensor

Next we consider the action for matter. We assume that this is given in terms of
the integral of a scalar Lagrangian:

Z
Smatter = d4 x gLmatter (10.27)

here Lmatter is a function of the matter fields (assumed to be tensor fields), their
derivatives, the metric and its derivatives. An example is given by the scalar field
Lagrangian discussed above. We define the energy momentum tensor by

2 Smatter
T ab = (10.28)
g gab

Part 3 GR December 14, 2016 123 H.S. Reall

CHAPTER 10. LAGRANGIAN FORMULATION

in other words, under a variation in gab we have (after integrating by parts using
the divergence theorem to eliminate derivatives of gab if present)

Z
1
Smatter = d4 x gT ab gab (10.29)
2 M

This definition clearly makes T ab symmetric.

Example. Consider the scalar field action we discussed previously.
Z
S= L (10.30)
M

with L given by (10.2). Using the results for  and g ab derived above we have,
under a variation of gab :

Z    
4 1 a b 1 1 cd
S = d x g + g c d V () g ab gab (10.31)
M 2 2 2

Hence  
ab 1 cd
T = + g c d V () g ab
a b
(10.32)
2
If we define the total action to be SEH + Smatter then under a variation of gab we
have

 
1 ab 1 ab
(SEH + Smatter ) = g G + T (10.33)
gab 16 2
and hence demanding that SEH + Smatter be extremized under variation of the
metric gives the Einstein equation

Gab = 8Tab (10.34)

How do we know that our definition of Tab gives a conserved tensor? It follows from
the fact that Smatter is diffeomorphism invariant. In more detail, diffeomorphisms
are a gauge symmetry so the total action S = SEH + Smatter should be diffeomor-
phism invariant in the sense that S[g, ] = S[ (g), ()] where denotes the
matter fields and is a diffeomorphism. The Einstein-Hilbert action alone is dif-
feomorphism invariant. Hence Smatter also must be diffeomorphism invariant. The
easiest way of ensuring this is to take it to be the integral of a scalar Lagrangian
as we assumed above.
Now consider the effect of an infinitesimal diffeomorphism. As we saw when
discussing linearized theory (eq (8.13)), an infinitesimal diffeomorphism shifts gab
by
gab = L gab = a b + b a (10.35)

Part 3 GR December 14, 2016 124 H.S. Reall

 10.3. ENERGY MOMENTUM TENSOR

Matter fields also transform according to the Lie derivative (eq (8.12)), e.g. for a
scalar field:
= L = a a (10.36)
Lets consider this scalar field case in detail. Assume that the matter Lagrangian
is an arbitrary scalar constructed from , the metric, and arbitrarily many of their
derivatives (e.g. there could be a term of the form a b a b or R2 ). Under
an infinitesimal diffeomorphism, (after integration by parts to remove derivatives
from and gab )
Z  
4 Smatter Smatter
Smatter = dx + gab
M gab
1
Z  
4 Smatter b ab
= dx b + gT gab (10.37)
M 2

The second term can be written


Z Z
4 ab
d4 x g a T ab b a T ab b



d x gT a b =
M M

Z
d4 x g a T ab b


= (10.38)
M

where we assume that b vanishes on M so the total derivative can be discarded.

Now diffeomorphism invariance implies that Smatter must vanish for arbitrary b .
Hence we must have
Smatter b
ga T ab = 0. (10.39)

Hence we see that if the scalar field equation of motion (Smatter / = 0) is satisfied
then
a T ab = 0. (10.40)
This is a special case of a very general result. Diffeomorphism invariance plus
the equations of motion for the matter fields implies energy-momentum tensor
conservation. It applies for a matter Lagrangian constructed from tensor fields of
any type (the matter fields), the metric, and arbitrarily many derivatives of the
matter fields and metric.
An identical argument applied to the Einstein-Hilbert action leads to the con-
tracted Bianchi identity (exercise):

a Gab = 0. (10.41)

Hence the contracted Bianchi identity is a consequence of diffeomorphism invari-

ance of the Einstein-Hilbert action.

Part 3 GR December 14, 2016 125 H.S. Reall