Lecture notes on Cosmology (ns-tp430m)

by Tomislav Prokopec
Part I: An introduction to the Einstein theory of gravitation
Einsteins theory of gravitation is a geometric theory, in the sense that gravitational forces
exerted by masses are mediated by a nontrivial structure of space and time. In particular, in the
presence of matter, physical distances between bodies change, and time lapses at a dierent rate. All
information about the eects of a matter distribution on space and time are elegantly encoded in a
symmetric metric tensor g

with a Lorentzian signature, which means that a local Minkowski metric

around the observer has the signature (the signs of the diagonal elements), sign[g

] = (+, , , ).
(A completely equivalent sign convention, which is often used, is (, +, +, +).) Once given, the met-
ric tensor completely determines the Lorentzian manifold, which in turn provides a representation of
gravitational interactions. In fact, the metric tensor g

= g

(x, t) provides complete information

on how to measure physical distances and time lapses between space-time points. Furthermore, it
fully species the dynamics of gravitating bodies, and thus it is of a fundamental importance for the
Einstein relativistic theory of gravitation. Before we begin discussing the metric tensor, we shall
now briey consider the metric structure of the special theory of relativity.
1 Special theory of relativity
Special relativity is an important special case of general theory of gravitation, where distances are
determined by the Minkowski metric, which can be written in terms of an innitesimal line element
as follows
ds
2
=

dx

dx

=
_
_
_
_
1 0 0 0
0 1 0 0
0 0 1 0
0 0 0 1
_
_
_
_
diag(1, 1, 1, 1) . (1)
Here and throughout these lectures, we are using Einsteins summation convention over the repeated
indices, e.g.

dx

dx

3
=0

3
=0

dx

dx

. Here x

= (ct, x
i
) (i = 1, 2, 3) is a four-vector
denoting a coordinate position of a point in space and time.
1.1 Lorentz symmetry
The line element ds denotes an invariant distance between two innitesimally displaced points in
space and time, and it is invariant under Lorentz transformations. A Lorentz transformation is
any matrix belonging to real orthogonal matrices in 3+1 dimensional space and time. When taken
together, they built up the orthogonal group O(1, 3), or equivalently, SL(2, (), where ( denotes
the set of complex numbers. In general, Lorentz transformations

are the 4 4 matrices, which

leave invariant the scalar product A B of two four-vectors A

and B

, where A B

.
There are two disjoined classes of Lorentz transformations. Proper Lorentz transformations are
the transformations which are by continuous deformations connected with the identity transforma-
tion

= diag(1, 1, 1, 1), and whose determinant equals to unity, det[

] = 1. Improper Lorentz
transformations are all other transformations. For example, space inversions and time inversions
1
are examples of improper Lorentz transformations. Any combination of improper Lorentz transfor-
mations is also an improper transformation. The condition det[

] = 1 is a sucient, but not

necessary, condition for a transformation to be improper. For example,

(combination
of space and time inversions) is an improper Lorentz transformation with det[

] = 1.
One can show that the Lorentz group has six generators. Three generators generate rotations,
the other three generate boosts. To study the representations of the proper Lorentz group, the
following Ansatz is useful,
= e

K
, (2)
where and

are the 3-vectors of rotations and boosts, respectively. The generators of rotations

S = (S
i
) and boosts

K = (K
i
) (i = 1, 2, 3) satisfy the following commutation relations of the
Lorentz algebra,
_
S
i
, S
j
_
=
ijl
S
l
_
S
i
, K
j
_
=
ijl
K
l
_
K
i
, K
j
_
=
ijl
S
l
, (3)
which constitutes both the algebra of SL(2, () and of O(1, 3). Recall that the commutator of two
matrices A and B is dened as [A, B] = ABBA. From Eqs. (3) we see that the order by which we
perform two consecutive rotations or boosts matters, since two consequtive rotations or boosts do
not in general commute. The symbol
ijl
is the totally antisymmetric tensor in 3 space dimensions,
dened by
123
= 1,
321
= 1. The cyclic (even) permutations do not change the value of
ijl
.
Since
ijl
is totally antisymmetric,
ijl
= 0 when any pair of indices i, j or l is identical.
The Poincare group is an inhomogeneous extension of the Lorentz group, and it is obtained by
adding space and time translations to the Lorentz group. The Poincare group has therefore ten
generators in total. The four additional generators are associated with space and time translations.
Lorentz transformations can be thought of as linear coordinate transformations,
x

=
x

(4)
such that L

= x

/x

. With this we see that the innitesimal line element ds

2
in Eq. (1) is
invariant under a Lorentz transformation, provided

. (5)
Equivalently

. Thus

is the inverse of

, and it is obtained from

simply by raising its indices with the Lorentz metric tensor,

. This then implies

that
1
=
T
, which proves the statement that Lorentz transformations belong to the group of
orthogonal matrices O(1, 3), where 1 and 3 refer to the Lorentzian signature, (+, , , ).
1.2 Causal Structure
Next we consider the causal structure of relativistic mechanics, which describes motion of particles
whose laws obey Lorentz symmetry.
Two space time points are said to be light-like separated if the line element vanishes, ds
2
= 0.
Geometrically, a space-time can be divided into the regions within the past and future light-cones
2
SPACELIKE

SEPARATION
x
geodesic
LIGHTCONE
SEPARATION
FUTURE
LIGHTCONE
PAST
TIMELIKE
SEPARATION
TIMELIKE
Figure 1: The past and future light-cones in Minkowski space-time separate time-like from space-like
distances. Time is on the vertical axis, and space (radial distance) on the horizontal axis. Each point on
the diagram corresponds to a two-dimensional sphere S
2
of the spatial section of space-time.
(time-like separations), and the region outside the light-cones (space-like separations). A point
on a light cone is light-like separated. This is illustrated on the space-time diagram in Figure 1.
Time-like separated points are in general causally connected. They can be connected by a geodesic,
which is any curve that represents a motion of a point particle, x
i
= x
i
(t), which is a solution of
the equation of motion. The geodesics of massless particles are a collection of light-like separated
space-time points.
A point x

= (ct, x
i
) lies on the past light-cone of x

0
= (ct
0
, x
i
0
) if
c(t t
0
) = |x x
0
| . (6)
Similarly, a point on the future light-cone is determined by
c(t t
0
) = |x x
0
| . (7)
Finally, two points are time-like or space-like separated when
c[t t
0
[ < |x x
0
| and c[t t
0
[ > |x x
0
| , (8)
respectively. For any two points on a geodesic, c[t t
0
[ |x x
0
|. The equality can hold only
for massless particles, which reects the fact that only massless particles (e.g. photons) can travel
with the speed of light in vacuum.
2 Metric tensor in general relativity
According to general relativity, a space-time reduces to a (locally) Minkowski space-time in the
absence of matter, or when matter is suciently remote, such that its eects on the metric tensor
are unmeasurably small. In addition it is required that no cosmological term is present.
3
2.1 Newtonian metric tensor
In presence of matter (or when matter is not very distant) physical distances between points in
general change. For example, an approximately static distribution of matter, if it is concentrated
in a nite region of space T such that it can be replaced by an equivalent mass M =
_
D
d
3
x(x)
concentrated at a point x
0
= M
1
_
D
d
3
xx(x) (which we can choose to be at the origin, x
0
=

0),
sources outside the region T the following Newton potential at a point x,

N
(x) = G
N
M
r
(9)
where
G
N
= 6.673(10) 10
11
m
3
kg s
2
(10)
is the Newton constant, and r |x |. According to Einsteins theory of gravitation, the physical
distances of objects in the gravitational eld of this mass distribution are described by the line
element,
ds
2
= c
2
_
1 +
2
N
c
2
_
dt
2

dr
2
1 + 2
N
/c
2
r
2
d
2
, (11)
where d
2
= d
2
+sin
2
()d
2
denotes the volume element of the two-dimensional sphere (the sphere
in three spatial dimensions), and [0, ] and [0, 2] are the two angles covering fully the
sphere. The general relativistic form of the line element (1) is
ds
2
= g

(x)dx

dx

. (12)
By comparing (11) and (12) we easily nd the metric tensor of a static mass distribution expressed
in spherical coordinates (r, , ),
g

=
_
_
_
_
1 + 2
N
/c
2
0 0 0
0 (1 + 2
N
/c
2
)
1
0 0
0 0 r
2
0
0 0 0 r
2
sin
2
()
_
_
_
_
. (13)
An important consequence of the change in the physical distance between space-time points is light
bending around massive bodies, and experimentally conrmed by two expeditions lead by Eddington
and Dyson during the solar eclipse on March 29 in 1919. According to the corpuscular theory of
light, the Newton theory predicts by a factor two smaller bending angle.
A second important consequence is that photons in a gravitational eld gain energy, and hence
their frequency is increased according to / =
N
/c
2
. This eect was experimentally measured
in 1960 by Pound and Rebka. Equivalently, this can be thought of as time dilatation: time passes
slower in a system where gravitational elds are stronger, such that the relative time dilatation
equals, t/t =
N
/c
2
.
2.2 Einsteins equivalence principle
The Einsteins equivalence principle states that an observer cannot perform a local experiment, based
on which he or she would be able to conclude whether he or she is placed in an accelerating or a
4
gravitating system. The origins of the equivalence principle can be traced back to the equivalence
of the inertial and gravitational mass,
m
i
= m
g
, m
i
d
2
x
dt
2
= m
g
g (14)
observed rst by Galileo, where g =
N
denotes the gravitational eld (the gravitational force
per unit mass). In a constant gravitational eld all bodies are accelerated at an identical rate. More
formally, the weak equivalence principle ascertains that in any gravitational eld, a freely falling
observer will not experience any gravitational eects. On the other hand, the strong equivalence
principle ascertains that all physical laws take the same form in freely falling frames. We will use
this to derive the geodesic equation in the next section. Both the weak and strong equivalence
principle have been scrutinized by experiments, since any deviation from the equivalence principle
would imply that an alternative theory of gravity is a more accurate description of reality than
Einsteins theory. So far no violation of the equivalence principle has been observed.
The equivalence principle provides an elegant resolution of the twin paradox of special relativity,
according to which the twin which travels to a star would nd his twin brother more aged, when he
returns to the Earth. Indeed, according to the equivalence principle, the time is equally contracted
in an accelerated system, as it is in a gravitational eld, and the twin in an accelerated ship ages
at a slower rate.
2.3 General covariance
From Eqs. (1113) it follows that the line element is invariant under a general coordinate transfor-
mation (dieomorphism),
x

(x) , (15)
provided ds
2
is invariant, ds
2
= d s
2
. Now an innitesimal coordinate transformation
d x

=
x

dx

, (16)
and the line element invariance imply that the coordinate transformation (15) induces the following
coordinate transformation of the metric tensor,
g

( x) =
x

(x) , (17)
while the inverse of the metric tensor transforms as,
g

( x) =
x

(x) . (18)
In general relativity one introduces the notion of covariant vectors A

and contravariant vectors

A

, which are related as A

= g

A
1
. The metric tensor is thus used to lower vector indices.
Conversely, the inverse metric tensor, g

is used for raise vector indices, A

= g

. The inverse
metric tensor g

is dened by
g

(19)
1
More rigorously, a vector eld can be dened in terms of one forms as, A = A

dx

. From the transformation

law of the one form (16) and the requirement A =

A, the transformation law for the components of a vector eld
immediately follows,

A

= (x

/ x

)A

.
5
where

= diag(1, 1, 1, 1) denotes the Kronecker delta. Note that the indices of a tensor are also
lowered and raised by the metric tensor and its inverse. For example, we have T

= g

.
The metric tensor is a special tensor in that its indices are also raised and lowered by the metric
tensor, cf. Eq. (19). In particular we have g

.
All metrics related by a general coordinate transformation (15) are physically equivalent, and
any apparent dierences in metrics should not be ascribed to physical eects. In analogy to gauge
theories, the eects induced by coordinate transformations (dieomorphisms) are sometimes called
gauge artifacts. Accordingly, any physical observable in a metric theory of gravitation should be
invariant under general coordinate transformations (15). This principle of general covariance, and
the requirement that in the weak eld nonrelativistic limit one ought to reproduce the Newton
theory, were the main guiding principles that lead Einstein to the discovery of the general theory
of relativity.
Just like in special relativity, the causal structure of space-time is determined by light-cones
shown in gure 1. The main dierence is that the light-cones of general relativity are not represented
by straight lines (67), but instead they are deformed by the nontrivial structure of the metric tensor.
3 Geodesic equation
Let us now consider a freely falling observer O, who erects a special relativistic coordinate system
in its neighbourhood, such that particles move along trajectories

() = (
0
,
i
) specied by
a non-accelerated motion,
d
2

ds
2
= 0 , (20)
where the line element ds = cd is proportional to a time variable, such that ds
2
c
2
d
2
=

. Now assume that the motion of O changes in such a way that it can be described by a
coordinate transformation,
d

dx

, x

= (ct, x
0
) . (21)
This and Eq. (20) then imply that the observer will perceive an accelerated motion of particles
governed by the geodesic equation,
d
2
x

ds
2
+

(x)
dx

ds
dx

ds
= 0 , (22)
where the new line element is given by Eq. (12), and
g

(x) =

and

=
x

(23)
denote the metric tensor and the (ane) Levi-Civit`a connection, respectively. The form for the
Levi-Civit`a connection, also known as the Christoel symbol, can be inferred by imposing that a
covariant derivative (which is dened below in Eq. (33)) of the metric tensor vanishes,

= 0
g

. (24)
This relation denes the unique metric compatible connection, also known as the Levi-Civit`a con-
nection. Now by making use of an appropriate combination of the derivatives of this type, one nds
6
for the Levi-Civit`a connection,

=
1
2
g

_
, (25)
where

/x

, etc.
An important consequence of the geodesic equation is that trajectories of particles (including
massive particles, as well as massless photons) moving in gravitational elds sourced by a distribu-
tion of masses, exhibit an accelerated motion.
More formally, the geodesic equation expresses the conservation of the covariant derivative of a
velocity 4-vector,
Tu

d

du

d
+

(x)u

= 0 , (26)
where u

= dx

/d, d = ds/c. More generally, a 4-vector eld X

is covariantly conserved if
the covariant derivative T = u

with respect to some time parameter vanishes, DX

/d
dX

/d +

(x)X

= 0. One important example of such a vector eld is the velocity eld u

,
which is covariantly conserved in the absence of external forces, as indicated in Eq. (22). In this
spirit, the general relativistic generalisation of Newtons law can be written as
m
Tu

d
= F

ext
, (27)
where here F

ext
denotes a sum over external forces, excluding gravity, and m is particles mass. For
example, for the electromagnetic eld, F

ext
is the generalized Lorentz force, F

ext
F

Lorentz
= qF

,
where F

, u

= dx

/d,

the covariant derivative, and q denotes the electric

charge.
Let us now dene the covariant (vector) derivative,

. If for a scalar function, f = f(x),

f transform as a vector under a general coordinate transformation, then

is the covariant
derivative. One can easily show that the structure of the covariant derivative acting on a scalar is
trivial,

f =

f, by simply showing that

f transforms as a vector under general coordinate

transformations. For a vector eld A

ought to transform as a two-indexed tensor, and

similarly for tensor elds. To be more concrete, note rst that a derivative of a vector eld transforms
as,

A
,
=
x

A
,
+

2
x

, (28)
and from (23) it follows that the connection transforms noncovariantly,

+

2
x

(29)
Taking these two transformation laws together, we easily nd,

+

2
x

(30)
Upon subtracting this from (28), we nd

A
,

=
_
A
,

_
x

. (31)
7
We have thus reached the conclusion that the following quantity transforms as a tensor,

A
;
= A
,

, (32)
which denes the covariant derivative of a vector eld. Similarly, the covariant derivative of a
two-indexed tensor reads

T
;
= T
,

, (33)
where we used a rather standard notation, according to which a semicolon (;) denotes a covariant
derivative (
;

), and a colon (,) denotes an ordinary derivative (

,

).
4 Geodesic deviation
()
2

1
()
Figure 2: The neighboring geodesics
1
(s) and
2
(s) used in the derivation of the equation of geodesic
deviation. The physical distance between the geodesics is denoted by the vector eld

(s).
We shall now show how one obtains an equation which controls the rate of change of the physical
separation of neighboring geodesics of test particles. To that aim consider two geodesic paths
1
()
and
2
() traced by nearby test particles, with the coordinate vectors, x

() and x

() +

(), as
shown in gure 2,
d
2
x

d
2
+

(x)
dx

d
dx

d
= 0
d
2
(x

)
d
2
+

(x + )
d(x

)
d
d(x

)
d
= 0 . (34)
Upon subtracting these two equations, we get to rst order the physical distance

between the
two geodesics, which is assumed to be a small parameter,
d
2

d
2
+ (

dx

d
dx

d
+ 2

d
dx

d
= 0 . (35)
8
On the other hand, after some eort a second covariant derivative of the vector

can be written
as,
T
2

d
2
=
d
2

d
2
+ (

+ 2

d
u

+ (

, (36)
where u

= dx

/d. By comparing this with Eq. (35) we immediately see that (35) can be written
in a covariant form,
T
2

d
2
=

] . (37)
The expression in the square brackets denes the Riemann curvature tensor 1

. With that in
mind, the equation of geodesic deviation can be recast to the simple form,
T
2

d
2
= 1

, (38)
with the Riemann curvature tensor
1

. (39)
Equation (38) may be used as the denition of the Riemann curvature tensor. Alternatively, it may
be dened in terms of the double covariant derivative acting on a covariant vector eld A

= g

as follows. The covariant derivative

acts on A

as indicated in Eq. (32). A second covariant

derivative acts then on A
;
as on a two-indexed covariant tensor eld B

(cf. Eq. (148)),

B
;
= B
,

. (40)
The dierence of two double covariant derivatives then denes the Riemann curvature tensor,
[

]A

A
;;
A
;;
= 1

. (41)
It is not hard to show that this denition results in the expression for the Riemann curvature tensor,
which is identical to Eq. (39). By studying the symmetries of the Riemann curvature tensor, one
can show that 1

has 20 independent components (in 3 + 1 dimensional space-time).

The equation of geodesic deviation (38) controls the congruence of nearby geodesics. In a at
space-time, the curvature tensor vanishes, and hence T
2

/d
2
= d
2

/d
2
= 0. This is just saying
that two initially parallel geodesics remain parallel at all times. In curved space-times however,
the Riemann tensor is nonvanishing, and as a consequence a freely moving observer sees a relative
acceleration of nearby freely moving test particles. One manifestation of this is the tidal eect
(sometimes referred to as the tidal force) of distant masses, which for example, explains the tides
on the Earth as a consequence of the dierence in the attractive gravitational force of the Moon at
dierent places on the Earth.
5 The Einstein eld equation
The Einstein eld equation cannot be derived. It can be obtained by postulating the principle of
general covariance, by requiring that in the weak eld nonrelativistic limit one recovers the Newton
9
theory of gravitation, and by requiring that the equation of motion contains at most two time
derivatives. For simplicity, we shall rst state the Einstein equation, and then show that it reduces
to the Newton theory.
The Einstein eld equation for the classical theory of gravitation is
G

c
2
g

=
8G
N
c
4
T

, (42)
where G

denotes the Einstein curvature tensor, T

is the stress-energy-momentum tensor (or

in short the stress-energy tensor) of all gravitating matter, G
N
= 6.673(10) 10
11
m
3
/kg/s
2
is the Newton constant, c = 299 792 458 km/s is the speed of light in vacuum, and denotes
the cosmological term. Sometimes is considered as a part of the stress-energy tensor. The
corresponding stress-energy tensor is then, T

= [c
2
/(8G
N
)]g

.
To generalise the conservation law for the stress-energy tensor to the relativistic theory of grav-
itation one denes the covariant conservation law of the stress-energy tensor,

= 0 . (43)
Any known form of matter builds up a stress-energy tensor that is covariantly conserved. Hence,
the consistency of the Einstein eld equation (42) implies the following Bianchi identity for the
Einstein curvature tensor,

= 0 . (44)
One can show that this condition denes uniquely (up to an overall multiplicative constant propor-
tional to g

) the Einstein curvature tensor in terms of the Ricci curvature tensor and the Ricci
curvature scalar,
G

= 1

1
2
g

1, 1 = g

. (45)
where the Ricci tensor is dened in terms of a contraction of the Riemann curvature tensor (39) as
follows,
1

= 1

. (46)
In the absence of asymmetric stresses, the stress-energy tensor is symmetric in its two indices, T

=
T

, which leaves a priory ten independent component functions. The Einstein curvature tensor
is in this case also symmetric, G

= G

, such that it as well contains at most ten independent

components. The Bianchi identity (44) and the covariant stress-energy conservation (43) further
restrict the number of independent functions to six. This then implies that the metric tensor is
completely specied by six independent functions (six degrees of freedom). It turns out that only
two degrees correspond to the propagating degrees of freedom (gravitons), while the other four are
excited only when sourced by matter, and do not propagate in the radiation zone, that is far from
the matter distribution. Two out of these four are the gravitational potentials (the (spatial) Newton
potential and the time-like potential), while the other two are vector-like, and are typically sourced
by a matter distribution with a nonvanishing vorticity.
The Einstein eld equation (42) and Eq. (45) dene how matter curves space-time, which is
expressed though a nontrivial dependence of the metric tensor on space and time, g

= g

(x), such
that it cannot be removed by an arbitrary coordinate transformation. Conversely, g

(x) species
motion of matter, such that the Einstein equation (42) is a self-contained equation that described
dynamics of matter elds via a geometric theory. Observe that the Riemann curvature tensor (39)
contains terms, which are of the form a single derivative acting on the Levi-Civit`a connection, and
10
terms which are quadratic forms in the connection. The Levi-Civit`a connection (25) is in turn
expressed in terms of single derivatives acting on the metric tensor. This then implies that the
Einstein eld equation (42) contains derivatives of the metric tensor up to second order in space
and time, and it that sense it resembles the Maxwell theory of electromagnetism and the Klein-
Gordon equation for scalar matter. (The dynamics of fermions is specied by the Dirac equation,
which at a rst sight contains rst order derivatives only. Nevertheless, one can show that the
matrix structure of the Dirac equation implies that, provided spin is conserved, fermions also obey
a second order dierential equation.) The principal dierence between the dynamics of matter elds
and the dynamic of gravitational eld are the nonlinear terms, contained in the quadratic forms in
the Levi-Civit`a connection, which makes the Einstein theory of gravitation a much more complex
theory than its matter counterparts. Fortunately, these terms are dynamically relevant only in
strong gravitational elds.
A second important dierence is that in the theory of gravitation the dynamical eld is the
metric tensor, which is a two indexed symmetric tensor eld, while in the matter sector there are
vector elds (photons, gluons, weak bosons), spinor elds (fermions) and scalar elds (the recently
discovered higgs particle). As a consequence, upon quantisation, one nds that the propagating
degrees in gravitation are the spin-two massless gravitons (which propagate at the speed of light
in vacuum), while in the matter sector one nds that the propagating modes in gauge theories
are the spin-one vector particles, the propagating modes in fermionic theories are the spin one-half
fermions, and nally the propagating modes in the scalar sector are the spin-zero scalars.
The strength of the coupling between the gravitational and matter elds is governed by the
coupling constant, 8G
N
/c
4
2 10
43
s
2
/kg/m. This is an extremely small constant, such that
only in the presence of matter under extreme conditions (large energy densities), the matter eects
on space and time can be strong. Such extreme conditions are found, for example, in galactic centers,
many of which are believed to host black holes. However, since gravitation is a long range interaction
(recall that the Newton gravitational potential decreases with distance as 1/r), the eects of a mass
distributed throughout space are cumulative, and a relatively dilute matter distribution, if spread
over large regions, can have a large eect on the structure of space and time. This is precisely the
case with the Universe, where one nds a relatively low matter density (the matter density in the
Universe is on average about 10
29
g/cm
3
), which is to a good approximation homogeneous and
isotropic, when averaged over large volumes. Indeed, today we have experimental means for testing
the large scale structure of space-time. This has been used extensively to test the Einstein theory
of gravitation on cosmological scales, as well as to study the evolution of the Universe.
5.1 The Hilbert-Einstein action
The Einstein eld equation (42) can be derived by varying the Hilbert-Einstein action
S
HE
=
_
d
4
x

g
c
4
16G
N
_
1+ 2

c
2
_
(47)
S = S
HE
+ S
matter
S
matter
=
_
d
4
x

gL
matter
, (48)
where g = det[g

] is the determinant of the metric tensor, 1 is the Ricci curvature scalar,

is the cosmological term, and

gL
matter
is the matter eld Lagrangian. Note that the Hilbert-
Einstein action (47) is the most general action which transforms as a scalar under general coordinate
11
transformations, and which contains terms up to second order in derivatives of the metric tensor.
There are in principle two unspecied constants in the action (47), which are not determined by the
symmetry (general covariance). One constant is the dimensionless constant multiplying (c
4
/G
N
)1,
and it can be determined by requiring that the Einstein theory of gravitation reduces to the Newton
theory in the weak eld limit, and we show how to do that in the next section. The second constant
is proportional to the cosmological term , and it can be determined by considering the dynamics
of gravitating bodies on very large (cosmological) scales. This illustrates how powerful the principle
of general covariance can be when constructing the gravitational action.
For example, for a real scalar eld = (x) we have,

gL
matter
=

g
_
1
2
g

)(

) V ()
_
, (49)
where V () denotes the scalar eld potential, such that d
2
V/d
2
V

= m
2

c
2
/
2
denes the scalar
eld mass-squared, and V

denes the scalar eld quartic self-coupling.

In order to calculate the variation S of the action (47), we rst observe that (see Problem 1.3)
g = gg

= gg

, (50)
which immediately implies

g =
1
2

gg

. (51)
Recalling that 1 = g

yields the following intermediate result for the variation of the Hilbert-
Einstein action,
S
HE
=
_
d
4
x

g
_

c
4
16G
N
g

_
1

1
2
g

1

c
2
g

_
+
c
4
16G
N
g

_
. (52)
The variation of the Ricci tensor 1

can be easily found by transforming to a local Minkowski

frame, in which g

+ O(

), such that we have

1

= 1

. (53)
This then implies
g

_
g

_
g

_
, (54)
where we inserted g

inside the derivatives, which is legitimate in the local Minkowski frame. Since
the left-hand-side of Eq. (54) is a scalar, the right-hand-side must also be a scalar, which implies
that the covariant form of Eq. (54) must read,
g

_
g

_
. (55)
This has the form of a covariant divergence, such that upon integration over an invariant measure
in Eq. (52), the variation of the Ricci curvature tensor does not contribute to the Einstein eld
equation,
_
d
4
x

gg

=
_
d
4
x

_
g

_
= 0 . (56)
The last equality follows from the simple observation that the covariant divergence of the contravari-
ant vector appearing in (56) can be also written as
A

=
1

gA

_
,

=
1

g . (57)
12
It then follows from the Gausss integral theorem that the integral (56) can be replaced by a closed
surface integral, whereby the surface S

is arbitrary. If we take the surface suciently far from any

masses, such that variation the metric tensor can be chosen to vanish everywhere on the surface,
g

[
S
= 0, then the integral in Eq. (56) vanishes.
Next we observe that varying the matter eld action (48) yields,
S
matter
=
_
d
4
x

gg

1
2
T

, (58)
where we dened the stress-energy tensor in terms of the variation of the matter action with respect
to the metric tensor as follows,
T

=
2

g
S
matter
g

. (59)
For example, for the scalar eld matter (49) one nds,
T

= (

)(

) g

L
matter
. (60)
By taking account of the intermediate results (52) and (56), we arrive at the following form for the
variation of the action (4748),
S =
_
d
4
x

gg

c
4
16G
N
_
G

c
2
_
+
1
2
T

_
. (61)
Now requiring that S vanishes for an arbitrary variation of the metric tensor g

yields the
Einstein eld equation (42).
6 Weak eld limit
We shall now study the question of correspondence between the Einstein and Newton theory of
gravitation, which is realised when gravitational elds are weak.
According to the equation of geodesic deviation (38), two freely moving test particles in a
curved space time move along trajectories that appear to experience dierent acceleration. As a
consequence, the respective geodesics that are initially set to be parallel eventually deviate from
being parallel. This eect can be ascribed to the tidal elds of distant masses, or to the gravitational
eld of a smoothly distributed matter in the vicinity.
For freely falling test particles one may choose a locally Minkowski coordinate frame, with
respect to which the observer does not move. In this frame dx

/d = u

0
, and the spatial
components of the equation of geodesic deviation (38) simplify to
T
2

i
d
2
= 1
i
0j0

j
, (62)
where we used u

0
, and the antisymmetry property of 1
i

under the exchange of the last

two indices.
Let us now compare the expression (62) with the corresponding expression one obtains in New-
tons theory, which we desire to correspond to the weak eld limit of the Einstein theory. The
acceleration in the Newton theory along the geodesics
1
() and
2
() (see gure 2) is given by,
d
2
x
i
dt
2
=
i

N
[

1
,
d
2
(x
i
+
i
)
dt
2
=
i

N
[

2
, (63)
13
where
N
denotes the Newton potential. Upon subtracting equations (63), and working to linear
order in
i
, we get,
d
2

i
dt
2
=
j

N
, (64)
where we used
i

N
[

2
=
i

N
[

1
+
j

N
[

1
, and
j
=
j
. The weak eld correspondence of
this equation with Eq. (62) then requires
1
i
0j0

i

N
c
2
, (65)
where we identied the parameter with time t multiplied by the speed of light, ct, and we ap-
proximated the covariant derivative in Eq. (62) by an ordinary derivative, T
2

i
/d
2
c
2
d
2

i
/dt
2
.
Taking the trace of (65) then yields,
1
i
0i0
= 1
00

2
i

N
c
2
=
4G
N
c
2

N
, (66)
where (for later convenience) we added to the right hand side the usual source of the Newtonian
Poisson equation. Here
N
denotes the density of matter, which diers by a factor of c
2
from the
energy density appearing in the stress-energy tensor, =
N
c
2
.
In order to make the desired connection with the Einstein theory, we now take the trace of the
Einstein eld equation (42) by multiplying it by g

,
1 =
8G
N
c
4
T 4

c
2
, T g

, (67)
upon which Eq. (42) can be recast to the form,
1

=
8G
N
c
4
_
T

1
2
Tg

_
g

c
2
. (68)
We now assume that the stress-energy tensor of matter can be well approximated by the ideal uid
form,
T

= ( +T)
u

c
2
g

T , (69)
where denotes the energy density and T the pressure of the uid. Recall that the observer is in
the freely falling frame, in which now dx

/dt u

= c
0

, g

, implying that T
00
= , T
0i
= 0,
T
ij
= T
ij
, T = 3T. From this we nd that Eq. (68) can be rewritten as
1
00
=
4G
N
c
4
_
+ 3T
_

c
2
1
ij
=
4G
N
c
4

ij
_
T
_
+
ij

c
2
. (70)
With this we can rewrite the correspondence relation (66) between the Einstein and Newton theory
in the form
1
00
=
4G
N
c
4
_
+ 3T
_

c
2

1
c
2

2
i

N
=
4G
N
c
2

N
. (71)
We see that the Newton limit is reproduced only when =
N
c
2
, = 0 and T = 0. This is justied
for dust, representing an extremely nonrelativistic matter, for which the pressure contribution is
negligible when compared to that of the energy density, T . This is certainly not a good
14
approximation for relativistic uids (such as neutrinos or photons). In particular for a photon uid
we have T = (1/3), such that the true relativistic source of the Newtonian potential is

active
+ 3T . (72)

active
/c
2
is sometimes referred to as the active gravitational mass density and, as we will see, it is
of a fundamental importance for cosmology, since at early epochs the Universe was predominantly
made up of relativistic matter.
The -term does not have a Newtonian equivalent, although sometimes an energy density is
associated to , which is of the form,

= [c
2
/(8G
N
)], and whose equation of state reads,
T

= w

, with w

= 1. The -term is very small in the Universe, and it becomes dynamically

relevant only on very large (cosmological) scales.
So far we have established a link between the Einstein and Newton theory of gravitation, by
establishing a correspondence between certain components of the Riemann curvature tensor and
spatial derivatives of the Newton potential. We shall now show how to construct the metric tensor
in the weak eld limit.
Since the coupling between matter and gravitation is weak (recall that it is governed by 8G
N
/c
4

2 10
43
s
2
kg
1
m
1
), it is often a very good approximation to linearise around the at Minkowski
space-time, in particular when one is asking questions about the evolution of local structures (galax-
ies, clusters of galaxies, etc.),
g

+ h

= diag(1, 1, 1, 1) , (73)
where h

represents a deviation from the at metric,

.
Working to linear order in h

we easily nd the Levi-Civit`a connection (25),

=
1
2

_
, (74)
and for the Ricci tensor 1

= 1

(39) and scalar 1 = g

,
1

=
1
2

1
2
_
h

h
_
1 =

h,

, h Tr[h

] =

, (75)
plus higher order terms.
Since we are working to rst order in h

, we can raise and lower indices with

. It is quite
straightforward to check that the Bianchi identity is automatically satised by the Ricci tensor (75),

_
1

1
2
1g

_
= 0 . (76)
The metric tensor h

is symmetric in its indices, and thus it has in general 10 component

functions, but not all of them are independent. Four of the functions can be constrained by imposing
the invariance under the general linear coordinate transformations,
x

= x

(x) . (77)
15
The metric tensor transforms as
h

= h

. (78)
One can easily show that, provided

= 0, the metric tensor is invariant under the coordinate

transformation (77) provided it satises the following gauge condition (analogous to Lorentz gauge
in electrodynamics,

= 0),

1
2

h = 0 . (79)
In this gauge the Ricci tensor and scalar (75) simplify to,
1

=
1
2
h

, 1 =
1
2
h. (80)
Assuming that the stress-energy tensor takes on the ideal uid form (69), we nd that the
Einstein equations in the uid rest frame (cf. Eqs. (6870)), in the weak eld limit, and in gauge (79),
reduce to the following simple form,
h
00
=
8G
N
c
4
_
+ 3T
_
+ 2

c
2
(81)
h
0i
= 0 (82)
h
ij
=
8G
N
c
4

ij
_
T
_
2
ij

c
2
. (83)
Note that (83) describes the equation for gravitation waves in the weak eld limit in presence of
matter sources, and we comment on its signicance below.
In order to complete the analysis, we now take the nonrelativistic limit, in which
2
i
, and
the pressure T 0. In addition we assume 0. We then nd

2
l
h
00
=
8G
N
c
4

2
l
h
0i
= 0

2
l
h
ij
=
8G
N
c
4

ij
, (84)
from which we conclude
h
00
=
2
N
c
2
, h
ij
=
ij
2
N
c
2
, h
0i
= 0 , (85)
such that in the weak eld nonrelativistic limit the line element takes the form,
ds
2
Newton
=
_
1 +
2
N
c
2
_
c
2
dt
2

_
1
2
N
c
2
_

ij
dx
i
dx
j
. (86)
When the potential is a spherically symmetric distribution of matter,
N
=
N
(r), than (to linear
order in the potential) the line element (86) simplies to
ds
2
Newton
=
_
1 +
2
N
c
2
_
c
2
dt
2

_
1
2
N
c
2
_
dr
2
r
2
_
d
2
+ sin
2
()d
2
_
. (87)
Note that Eqs. (8687) represent the weak eld limit of the line element (11), in which
N
c
2
,
where
N
denotes the Newton potential (see also Problem 1.4).
16
A careful reader has certainly noticed that we have not yet discussed gravitational waves, which
are also a part of the weak eld analysis. Gravitational waves correspond to homogeneous (rela-
tivistic) solutions of equation (83), such that they can be thought of as a linear superposition of
plane waves which propagate with the speed of light in vacuum. In the light of the discussion in
section 1.2, gravitational waves propagate on the light-cone, and hence are consistent with causality.
In the physical gauge, in which h
0
= 0 and h
ij
is traceless and transverse (h
i
i
= 0,
j
h
ij
= 0), the
two physical degrees of freedom correspond to the two mutually orthogonal deformations of space,
such that a circle placed orthogonally to the wave propagation is deformed to an ellipse. Quantum
mechanically, these two degrees of freedom correspond to the two states of the massless spin-two
graviton. These projections of spin on the direction of motion are known as helicities, and can be
either plus two or minus two (in units of ). We postpone a more detailed analysis of gravitational
waves to Part IV, in which we discuss the production of gravitational waves in an inationary epoch
of the early Universe, during which the Universe expands in an accelerated fashion.
7 Tests of general relativity
General relativity has been tested on many grounds, and up to this moment no deviations have been
found from the theorys predictions. Here we mention several tests, and consider in some detail time
dilatation, reshift and light deection by the gravitational eld. The latter, for example, induces
lensing of cosmic microwave background radiation. An interested reader may consult, for example,
Norbert Straumann, General relativity and relativistic astrophysics (Springer-Verlag, 1984), and
Cliord M. Will, The Confrontation between general relativity and experiment, Living Rev. Rel.
9 (2006) 3 [arXiv:gr-qc/0510072].
Tests of the Einstein theory of gravitation include:
(1) Advance of the perihelion of a planet. A disagreement of advance of the perihelion (the point
of closest approach of a planet to the Sun) of Mercury with the Newton theory prediction
was known since a long time ago. In 1859 Le Verrier suggested that the anomaly could be
explained by an unobserved planet Vulcan in an orbit close to the Sun. A general relativistic
calculation was performed in 1916, and explained the Mercury anomaly with an accuracy of
better than 1%. The prediction of the Einstein theory for advance of the Mercury perihelion
is
Einstein
= 42.98

per century, while the observed value is 43

(the agreement implied by

the Lunar ranging measurements is about 0.3%).
(2) Gravitational bending of light was rst measured during the total solar eclipse in 1919 to an
accuracy of about 10% by two scientic teams lead by Dyson and Eddington. The most
accurate modern measurements are based on about 2 million quasar and galaxy observations
by VLBI (1999) over the whole sky, and yield a conrmation of the Einstein theory to an
accuracy of about 0.02%. An alternative conrmation has been reached by the satellite
Hyparcos at the level of 0.1%.
(3) Gravitational redshift of light was rst observed on an Earth experiment in 1960 by Pound and
Rebka. The measurement, which was based on the Mossbauer eect, and was subsequently
improved by Pound and Snider in 1965 to an accuracy of about 1%. The accuracy was further
improved to 2 10
4
in a rocket experiment by Versot and Levine in 1976.
17
(4) Time delay in gravitational eld. This eect has been observed by radar echoes (radar-
ranging o the retro-reector placed in 1969 by Appolo 11 on the Moon), as well as by passive
reections of radar signals o Mercury and Venus, when they were on the opposite side of
the Sun. The active reection o the Viking satellite mirror resulted in an 0.1% test of the
Einstein theory prediction.
(5) Gravitational lensing has rst been observed in 1979. The rst Einstein ring was observed in
1988 on MG 1131 + 0456 in the constellation Leo by Jacqueline Hewitt. Nowadays, gravita-
tional lensing is routinely observed in deep sky images (e.g. by the Hubble space telescope
and, starting in 2020, by the Large Synoptic Survey Telescope, LSST, whose construction
began in 2012).
(6) Geodetic precession is generated by the warped nature of space-time, and it arises when
one massive body rotates in the gravitational eld of another body. The general relativistic
theory of gravitation predicts a precession of about 2

per year of the Earth rotation axis

due to the gravitational interaction between the Earth and the Moon, and it has been con-
rmed by the Lunar laser ranging experiments to an accuracy of about 0.7%. The Stanford-
NASA gyroscope satellite experiment (Gravity Probe B), http://einstein.stanford.edu/,
http://www.gravityprobeb.com/, launched in April 2004, has the designed accuracy goal of
510
5
. The nal results (C. W. F. Everitt et al, Phys. Rev. Lett. 106, 221101 (2011)) give
for the geodetic precession 6.602 0.018 arcsec/year, which agrees with the predicted value
of 6.606 arcsec/year.
(7) Gravitomagnetic precession (Lense-Thirring eect). The eect was independently predicted
by G. E. Pugh (1959) and by Leonard I. Schi (1960). Space-time is warped due to a non-
vanishing angular momentum of the Earth. The warping causes a precession of the axis of
rotating bodies in the Earth orbit (the spin-orbit coupling in the theory of gravitation). The
eect is tiny (about 2.1 10
2
arcsec/year for the Earth) and the detection has been recently
claimed by Ciufolini et al, based on the precession of the two LAGEOS satellites. The result
represents a 20% accurate conrmation of the prediction of general relativity. Even though
Gravity Probe B has been designed to measure gravitomagnetic precession to an accuracy of
about 2%, the actual result had an accuracy of only 20%. The result reported by C. W. F.
Everitt et al (Phys. Rev. Lett. 106, 221101 (2011)) is 37.2 7.2 10
3
arcsec/year, and it
agrees well with the value predicted by general relativity, 39.2 10
3
arcsec/year. Since the
Gravity Probe B result is more reliable than the LAGEOS satellite results, it is considered
yet another novel test of general relativity.
(8) Gravitational radiation from binary pulsar systems. In 1974 Joseph Taylor and Russell Hulse
(Nobel Prize 1993) discovered a binary pulsar system, consisting of a neutron star orbiting
around as-of-yet unseen companion, which they named PSR 1913+16. The pulsar period is
59 ms, the orbital period about 7.75 hours, and the eccentricity 0.617. In fact, the pulsing
period P
p
and its slow-down rate dP
p
/dt are known to a very high accuracy,
P
p
= 59.029997929613(7) ms ,
dP
p
dt
= 8.62713(8) 10
18
, (88)
while the orbital period decreases with the rate
dP
b
dt
= 2.422(6) 10
12
, (89)
18
which is caused predominantly by gravitational radiation, present in a signicant amount
only in strong gravitational elds. When the eect of galactic rotation is subtracted (89), a
comparison with the general relativistic prediction yields,
(dP
b
/dt)
GR
(dP
b
/dt)
observed
= 1.0023 0.005 , (90)
which tests general relativity to an accuracy of about 0.5%. This is so far the only indirect
conrmation for the existence of gravitational radiation, and at present one of a few tests of
general relativity in strong gravitational elds.
(9) Black holes. The center of our galaxy (Milky Way) harbours a black hole placed in the
constellation Sagitarrius A

(Sag A

), which is about 26000 light years away from us, and

whose mass is about 4.110
6
M

, M

being the mass of the Sun, which is about 8.210

36
kg
(The Max Planck Institute for Extraterrestrial Physics in Garching, Germany, and the UCLA
Galactic Center Group, Los Angeles).
2
Our black hole will soon become active, and between
2014 and 2018 it will accrete a signicant amount of matter. A feast is soon to come! Finally,
there are also strong indications that other galaxies (active galactic nuclei) and quasars harbor
massive black holes in their centra, whose mass could reach a value as high as 10
10
M

. The
largest black hole so far observed is believed to be in the galaxy NGC 4889. Its mass is
estimated to be in the range from 6 billion to 37 billion solar masses.
(10) Direct detection of gravitational radiation.
In 2014, the advanced LIGO (LIGO stands for the Laser Interferometer Gravitational-Wave
Observatory) will start taking data, and direct detection from a merger of two compact stars
is expected within a year. Compact objects are neutron stars and black holes. If successful,
that will be the rst direct detection of gravitational waves. Moreover, apart from cosmolog-
ical tests, these will represent rst tests of the strong gravity regime (that is beyond linear
regime) of general relativity. The LIGO is physically at two locations: rst is the LIGO
Livingston Observatory in Livingston, Louisiana, and second is the LIGO Hanford Obser-
vatory, near Richland, Washington. For more information see http://ligo.org/. More
to the future, the EU plans the underground Einstein Telescope (http://www.et-gw.eu/,
http://en.wikipedia.org/wiki/Einstein Telescope) and the satellite mission LISA (http://sci.esa.int/science-e/www/area/index.cfm?fareaid=27,
http://lisa.nasa.gov/).
(11) Cosmological tests of general relativity.
Cosmology has been used to test general relativity in the strong regime through gravitational
redshift of photons. Moreover, motion of galaxies and clusters of galaxies on large scales
tell us that dark matter and dark energy must be added in order to make general relativity
consistent with observations. It seems that most of the observations of dark matter can be
well explained by modeling it with a cold (nonrelativistic) uid (or dust), whose equation
of state is w
dm
= T
dm
/
dm
= 0. On the other hand, observations of dark energy can be
eplained by a cosmological constant, whose equation of state is w
de
w

= T

= 1.
Alternatively, it is possible to modify general relativity on large scales, such that its eects
2
For a video experience see
http://en.wikipedia.org/wiki/Supermassive black hole. The video can be seen at
http://en.wikipedia.org/wiki/File:A Black Hole%E2%80%99s Dinner is Fast Approaching - Part 2.ogv.
19
mimick those of dark matter and dark energy. Studies of modied gravity are an active area
of research, and the nal judgement (on whether the right thing is to modify gravity or to
add dark matter and dark energy to general relativity) has not as yet been made.
8 Gravitational time dilatation and gravitational redshift
Perhaps the simplest way of understanding time dilatation and gravitational redshift is to consider
two observers O
1
and O
2
placed in a stationary gravitational eld well approximated by the Newton
potential. We assume further that the observers do not move with respect to the center of mass,
such that the gravitational eld they observe appears to be static. The reader should keep in mind
the spherically symmetric Schwarzschild metric and the corresponding line element, as given in
Eqs. (1113), which for convenience we quote again,
ds
2
= c
2
_
1 +
2
N
c
2
_
dt
2

dr
2
1 + 2
N
/c
2
r
2
_
d
2
+ sin
2
()d
2
_
,
N
= G
N
M
r
. (91)
Let us now consider a high frequency light ray passing by two observers, O
1
and O
2
, and let
us assume that the observers measure a time lapse between two subsequent light crests, which we
denote by t
1
and t
2
, respectively. The time lapses t
1
and t
2
can then be easily related by noting
that an observer at an asymptotic innity, where the metric is Minkowski at, would measure a
time lapse between two subsequent wave crests of the same wave, such that the following simple
relation holds,
=
_
g
00
(r
1
)t
1
=
_
g
00
(r
2
)t
2
. (92)
From this we immediately conclude that the two time lapses are related as
t
1
t
2
=

g
00
(r
2
)
g
00
(r
1
)
, (93)
which in a weak gravitational eld reduces to
t
1
t
2
1 +

N
(r
2
)
N
(r
1
)
c
2
. (94)
This implies that the observer, which is placed deeper in the potential well, such that its potential
is more negative, measures a longer lapse between the wave crests. This eect is known as the
gravitational time dilatation.
As an example, let us assume that the rst observer O
1
O

is located on the surface of the

Sun, where
N
(r
1
)

2.12 10
6
c
2
, and O
2
is on the surface of the Earth, where the
potential is much smaller, and can be to a good approximation neglected,
Earth
0. Eq. (94) then
implies
t
Earth

_
1 +

c
2
_
t

= t

2.12 10
6
t

, (95)
20
such that the time lapse between the two crests measured on the Earth is shorter than what would
be measured on the surface of the Sun. Thus, on the Earth surface time lapses faster than on the
surface of the Sun (time contraction), and conversely, t

= t
Earth
+ 2.12 10
6
t
Earth
.
3
The question we now address is, what is the frequency of light measured by the observers O
1
and O
2
. The phase of the wave crests/throughs in a diagonal metric can be represented by the
simple formula
(x
A
; x
B
) =
1

_
x
B
x
A
g

(x)p

dx

=
1

_
t
B
t
A
g
00
(x

)
E
c
dt

_
x
B
x
A
g
ii
(x

) p dx

, (97)
where p

= (E/c, p
i
) denotes the 4-vector of energy and momentum, and = h/(2) = 1.054
10
34
Js is the reduced Planck constant, h = 6.6262 10
34
Js. Since the (static) observers O
1
and
O
2
measure a short time interval between two wave crests, to leading order in t Eq. (97) simplies
to,
(x
A
; x
B
) = = g
00
(r
1
)
E
1
c
t
1
= g
00
(r
2
)
E
2
c
t
2
. (98)
By making use of Eqs. (9293) and of E = , we easily nd
E
1
E
2
=

1

2
=

g
00
(r
2
)
g
00
(r
1
)
1 +
1
c
2
_

N
(r
2
)
N
(r
1
)
_
, (99)
such that energy redshifts as photons climb out of a gravitational potential. This phenomenon is
known as the gravitational redshift of light, and is has been rst observed on the Earth in 1960 by
Pound and Rebka. When applied to the expanding Universe, the gravitational reshift is responsi-
ble for the redshift of photons from distant sources, and from the cosmic microwave background
radiation.
A simple interpretation of this result is obtained by noting that photons are just a special case
of particles, whose relativistic energy is given by the Einstein relation, E =
_
p
2
c
2
+ m
2
c
4
, with
zero mass m = 0 and a momentum p = /c, where denotes the frequency of the photon. When
placed in a gravitational potential,
N
=
N
(x), the energy of particles with m ,= 0 is modied as,
E =
_
p
2
c
2
+ m
2
c
4
+ m
N
. (100)
Since the gravitational eld is conservative, the energy of particles moving in a gravitational eld
must be conserved, which implies
E(x
1
, p
1
) = E(x
2
, p
2
) . (101)
3
The following consideration is erroneous. I encourage the reader to nd a aw in the following reasoning. The
equivalence principle implies that time dilatation in an accelerated system is the same as time dilatation observed
in a system placed in a gravitational eld of an equal magnitude. A quantitative estimate of time dilatation can be
then easily found by identifying acceleration with gravitational eld. We thus have, a = g =
x

N
, or equivalently,

N
(x) =
_

x
a(x) dx. This then implies the following simple expression for time dilatation in an accelerated
system, measured with respect to an inertial system (both systems are assumed to be placed in a vanishingly small
gravitational eld),
t
dilatation
(x
1
, x
2
)
t
inertial
=
1
c
2
_
x2
x1
a(x) dx, (96)
where the system accelerates from point x
1
to point x
2
, and t
inertial
denotes time lapse in the inertial (nonaccelerated)
system. This expression represents an estimate of the aging dierence of the two twins, in the twin paradox of special
relativity, which however gives an incorrect answer. The correct answer can be obtained by a standard use of the
ligh-cone diagram. I encourage the reader to nd the aw in the reasoning in this footnote.
21
For nonrelativistic particles, this reduces simply to the conservation of the kinetic plus potential
energy. This expression is however meaningless for photons, since their mass is zero. A heuristic
derivation for the photons can be nonetheless obtained by replacing m
N
by (E/c
2
)
N
in the
expression (100), which means that photons behave as if they had a gravitational mass equal to
p/c. This then implies,

1
_
1 +

N
(x
1
)
c
2
_
=
2
_
1 +

N
(x
2
)
c
2
_
, (102)
which agrees with (99).
9 Light deection
A gravitationally induced light deection was rst measured during the Solar eclipse on March
29, 1919 by two expeditions, organized by Frank Dyson and Arthur Eddington, respectively. The
observations took place in the Brazilian city of Sobral (Dyson), and in the Portuguese island of
Principe o the West coast of Africa (Eddington). The observers compared positions of stars at
night with the respective positions during the eclipse, and found for the light deection angle (in
arc seconds),
= (1.98 0.16)

(Sobral, Dyson) , = (1.61 0.40)

(Principe, Eddington) , (103)

in agreement with the prediction of the Einstein theory, = 1.75

/d, where R

denotes the Sun

radius, and d the closest distance of the photon to the Sun center.
The Newtons corpuscular theory of light predicts a bending angle which is by a factor two
smaller. This can be understood as follows. In the corpuscular theory of light the origin of the
eect is in light bending, and would correspond to the bending of light rays with respect to absolute
straight lines, dened for example by rigid rods. But there are no absolute straight lines in the
Einstein theory. The space-time of general relativity is curved around massive bodies, resulting in
an additional eect identical in size to the bending angle of light corpuscles in the Newton theory,
explaining thus the general relativistic result.
Let us start our analysis with the general relativistic action for a point particle,
S = mc
_
ds =
_
Ldt , L = mc
_
g

dx

dt
dx

dt
. (104)
The corresponding canonical 4-momentum p

= (E/c, p
i
) is,
p

=
L
(dx

/dt)
=
dt
ds
mcg

dx

dt
,
ds
dt
=
_
g

dx

dt
dx

dt
. (105)
The corresponding Euler-Lagrange equation is then,
dp

dt
=
1
2
(

)p

dx

dt
(106)
where p

= mcdx

/ds. Note that this is just a convenient rewriting of the geodesic equation (36),
with the identication, p

= mu

. In general a massive particle must propagate on the mass shell,

which implies,
g

= m
2
c
2
g

= c
2
. (107)
22
For the photons however, m 0, and this simplies to
g

= 0 , (108)
which establishes the photon dispersion relation, p
0
= p
0
(p
i
, x

).
We are interested in light propagation in the presence of a quasistationary mass distribution
which produces weak gravitational elds (weak lensing), hence the line element can be to a good
accuracy modeled by the following Newtonian diagonal form,
ds
2

_
1 + 2

N
c
2
_
c
2
dt
2

_
1 2

N
c
2
_

ij
dx
i
dx
j
, (109)
where
N
=
N
(x) is the Newton potential of the quasistationary mass distribution.
First we note that in a quasistationary Newtonian metric, the Euler-Lagrange equation for
light (106) implies the conservation law of the canonical energy p
0
/c,
dp
0
dt
=
d
dt
_
(1 + 2
N
/c
2
)p
0
_
= 0 , (110)
while for the spatial momentum we get,
dp
i
dt
=
d
dt
_
(1 2
N
/c
2
)p
i
_
=
c
2
(
i
g

)
p

p
0
. (111)
Upon dividing this by the conserved quantity, (1 + 2
N
/c
2
)(p
0
/c) = const., we nd
d
dt
_
(1 4
N
/c
2
)
dx
dt
_
= 2
N
, (112)
where we took account of dx

/dt = cp

/p
0
and (dx
i
/dt)
2
= c
2
. This equation describes the lensing
of light in a weak quasistationary gravitational eld in general relativity. Note that for relativistic
bodies, the gravitational eld (the force per unit mass), 2
N
, appears to be by a factor two
larger than what one would expect from the nave Newtonian limit.
In a simple case when light bending is small, we can take a light ray to move from a source S
in the y direction, v
y
= c, and we can integrate (112) once to obtain,
dx(y)
dt
=
2
c
_
y
y
S
dy

N
. (113)
where y
S
denote the source position, and we made use of dt = dy/c. The light bending angle

x
=
x
(y
S
, y
O
) in x direction accumulated between the source at x
S
and the observer at x
O
is then
(
x
= v
x
/v
y
= v
x
/c),

x
=
2
c
2
_
y
O
y
S
dy
x

N
. (114)
This is the main result of this section.
We shall now apply formula (114) to the simple case of a point like mass distribution of a mass
M located at the origin, in which case,
N
= G
N
M/r. Eq. (114) then yields

x
=
2G
N
Mx
c
2
_
y
O
y
S
dy
(x
2
+ y
2
)
3/2

4G
N
M
c
2
d
, (115)
23
where x = d represents the closest distance of the light ray to the mass M. When applied to the
Sun, whose mass and the radius are given by M

= 2 10
30
kg and R

= 7 10
8
m, and taking
y
S
, y
O
+, we get the famous Einsteins result (for the Sun),

(d) = 1.75

d
. (116)
It is of interest to note that bending angles of a similar magnitude are produced by typical elliptical
and spiral galaxies (see Problem 1.5).
10 Coupling of matter elds to gravitation
According to the principle of general covariance, matter elds couple to gravitation such that the
corresponding matter action is generally covariant. We now discuss how to construct generally
covariant actions for the relevant matter elds (scalars, gauge elds, fermions).
10.1 Scalar fields
We start with the simplest case of a real scalar eld with a canonical kinetic term, and whose
potential is given by V = V (). If we restrict ourselves to terms containing up to second order
derivatives, the covariant scalar action is then given by,
S

=
_
d
4
x

gL

gL

=
1
2

gg

)(

gV ()
1
2

g
2
1. (117)
The Euler-Lagrange equation of motion is obtained by varying S

with respect to ,
1

gg

_
+
dV ()
d
+ 1 = 0 . (118)
Note that the derivative in this equation becomes the covariant derivative,

= (g)
1/2

(g)
1/2
,
when it acts on the vector eld,

, such that the derivative operator in (118) is nothing

but the dAlembertian operator g

as it acts on a scalar eld.

The scalar potential V often contains a constant, quadratic and quartic term only,
V () = V
0
+
1
2
m
2

c
2

2

2
+

4!

4
. (119)
Note that V
0
is redundant in the sense that it can be combined with the cosmological term of the
Hilbert-Einstein action (47),

c
2
8G
N

0
+ V
0

c
2
8G
N
, (120)
and has no independent physical meaning. This also explains why some authors like to put the
cosmological term as part of the matter action. Phase transitions in eld theory are often represented
by a potential of the type (119) with V
0
> 0 and m
2

< 0, showing that phase transitions in the early

Universe (for example, electroweak phase transition and quantum-chromodynamic (QCD) phase
transition) are intricately related to the problem of vacuum energy in the theory of gravitation
24
(the cosmological constant problem). This question has gained in importance by the recent (2012)
discovery of the higgs particle by the ATLAS and CMS collaborations at the LHC experiment at
CERN, Geneva.
The last term in the scalar Lagrangian (117) represents the coupling of scalar elds to the
Ricci scalar of gravitation, and it is generally covariant. Up to this moment there are no strong
observational constraints on the size or sign of . An exception are composite scalar elds which
are made up of some more fundamental elds that are not Lorentz scalars. For example, chiral
condensates of quantum-chromodynamics (QCD) can be thought of as eective scalar elds made
up of fundamental fermion elds, whose spinor structure forbids covariant coupling to the Ricci
curvature scalar, implying that in this case we expect = 0. A similar conclusion is reached if
composites are made of gauge elds, e.g. we expect that the composite eld g

does not
couple to gravitation, = 0.
10.2 Abelian gauge fields
Next, we consider Abelian gauge elds. A generalisation to nonabelian gauge elds is straightfor-
ward. The covariant eld strength can be dened in terms of gauge eld as follows,
F

, (121)
where the last equality follows quite trivially from the antisymmetry in the denition of the eld
strength tensor. The Abelian gauge eld action is then simply,
S
gauge
=
_
d
4
x

gL
gauge

gL
gauge
=
1
4

gg

. (122)
The equation of motion for the gauge eld A

is obtained by varying (122) with respect to A

, and
it reads,

gg

_
= 0 . (123)
An important property of gauge elds is that in conformally at space-times, whose metric can
be written in the conformally at form,
g

= a(x)
2

= diag(1, 1, 1, 1) , (124)
Eq. (123) reduces to the simple Maxwell equation,

= 0 . (125)
In deriving this we made use of the metric tensor inverse g

= a(x)
2

= diag(1, 1, 1, 1),
and of

g = a
4
. We have just proved that gauge elds couple conformally to gravitation. An
important consequence of this fact is that in conformal space-times (124), examples of which are
de Sitter and power law inationary space-times, as-well-as Friedmann-Lematre-Robertson-Walker
(FLRW) space-times, at the classical level, gauge elds do not couple to gravitation (they feel no
gravitational pull). One says that in conformal space times gauge elds live in conformal vacuum,
given by the appropriately normalised solution of (125). This lead to the popular belief that there
cannot be much photon production through photon coupling to gravitation from the early universe
epochs, which is in fact incorrect.
25
Eq. (118) implies that scalar elds do not in general couple conformally to gravitation. Indeed,
inserting the conformally at metric (124) into the scalar equation (118) results in

+ 2

a
a

+ a
2
dV ()
d
+ a
2
1 = 0 , (126)
which shows that scalar elds do in general feel gravitational force in conformal space-times. If
space-time is in addition homogeneous, in which case the scale factor a = a() is a function of
conformal time only (dened as dt = ad), because of the time derivative acting on , the second
term in (126) looks like a damping term. The damping coecient, 2H = 2aH, is given in terms
of the Hubble parameter H, which is dened as H(t) = (1/a)da/dt, and therefore it is often called
Hubble damping. Hubble damping has important consequences for cosmology, since it is in the crux
of the mechanism for production of cosmological perturbations during inationary epoch, which in
turn seed structures of the Universe. We shall come back to this question when we discuss scalar
and tensor cosmological perturbations.
10.3 Frame fields and fermionic fields
Due to the spinorial structure of fermion elds , fermions transform nontrivially under general
coordinate transformations, S, where S denotes a matrix in spinor space. For that reason,
getting the covariant form of the Dirac equation requires special care. This is easiest done by
making use of the frame eld (also known as the vierbein or tetrad) formalism. The frame eld can
be dened in terms of the metric tensor as follows,
g

(x) = e
a

(x)e
b

(x)
ab
,
ab
= diag(1, 1, 1, 1) , (a, b = 0, 1, 2, 3) , (127)
and hence can be thought of as the transformation of the metric tensor to a locally at coordinate
system (known as the tangent space) with the metric tensor
ab
. The set of tangent spaces at all
points of the space-time is known as the tangent bundle.
The following generalisation of the anticommutation relation for the Dirac matrices is very
natural (since it is generally covariant)

= 2g

. (128)
This implies that the Dirac matrices acquire space-time dependence, which can be easily disentan-
gled by making use of the vierbein,

(x) = e
a

(x)
a
, (129)
where
a
are the Dirac matrices of the corresponding at (tangent) space erected at point x

, and
they obey the standard at space-time anticommutation relation,

a
,
c
= 2
ac
,
ac
= diag(1, 1, 1, 1) . (130)
In order to obtain the covariant formulation of the Dirac equation, it is necessary to introduce a
spin connection

, which are 4 4 matrices in spinor space. The spin connection (a better name
would be the spinor connection) is used to dene the covariant derivative acting on Dirac spinors,

. (131)
More precisely, the spin connection

and the Levi-Civit`a connection,

can be used to dene the

covariant derivative acting on any object, whose transformation properties under general coordinate
26
transformations are known. For example,

has both one vector and two spinor indices, and hence
the covariant derivative acting on

is given by

= 0 . (132)
The covariant derivative of the Dirac matrices must vanish, since there exists a coordinate trans-
formation which transforms

to a locally space-time independent form, and in which

,
implying (132). Given that

can be constructed from the vierbeins, Eqs. (129) and (132) determine
the spin connection

up to an additive multiple of the unit matrix.

We can now write the generally covariant form of the fermionic action,
S
fermion
=
_
d
4
x

gL
fermion
(133)

gL
fermion
=

g

i

g
m

, (134)
where

=

0
(x),

= (

), and m

denotes the fermion mass. Note that the La-

grangian (134) is hermitean, as it should be. Upon varying the action (133134) with respect to

,
we easily get the fermion equation of motion for curved space-times,
i

) m

c = 0 . (135)
For example, for the conformally at metric (124), the vierbeins are simply,
e
c

(x) =
c

a(x) , e

c
(x) =

c
a(x)
1
, ( = 0, 1, 2, 3; c = 0, 1, 2, 3) , (136)
such that

= a
1

c
, (137)
where
c
are the at space Dirac matrices. After some algebra, one nds that following the Dirac
equation (see Problem 1.7) holds,
_

a
+ iam

cf
= 0 ,
cf
= a
3/2
, (138)
where here
a
= (
0
,
i
) denote the at space-time Dirac matrices. From Eq. (138) we conclude
that massless fermions couple conformally to gravitation, in the sense that the conformally rescaled
massless fermion eld
cf
a
3/2
does not couple to gravitation at the classical level.
As it is indicated in Eq. (138), the presence of a mass term breaks conformal coupling of fermions
to gravitation. This eect has been used to motivate the study of fermion pair production in
rapidly expanding space-times of the early Universe (inationary epoch and early radiation era).
The production is negligible today, since the rate of fermion pair production is determined by the
rate of change of the eective mass term am, which is in turn proportional to the Hubble expansion
rate today, which is tiny.
Finally, we recall that matter couples to gravitation through the stress-energy tensor, which
can be calculated for any matter eld
matter
, , A

, etc., by varying the appropriate matter

action S
matter
S

, S
fermion
, S
gauge
with respect to the metric tensor (see Eq. (59)),
T

=
2

g
S
matter
g

. (139)
27
11 Alternative theories of gravitation

The Einstein theory of gravitation (or general relativity) has up to now passed all experimental
tests. In the future nevertheless, we may witness emergence of a more accurate theory of gravita-
tion, which reduces to the Einstein theory in a certain limit. Moreover, there are some observations
on large (galactic and cosmological) scales, whose explanation may as well be found by extend-
ing the Einstein theory. One motivation to extend the Einstein theory is to provide alternative
explanation for the missing matter problem of the Universe, which is standardly explained by
adding the appropriate amount of nonbaryonic matter (dark matter), which apart from gravita-
tional interaction interacts very weakly with visible matter. This explanation is now questioned by
the work of Douglas Clowe et al. (A Direct Empirical Proof of the Existence of Dark Matter,
Ap. J. Lett. 648 (2006) L109L113 [arXiv:astro-ph/0608407]) where evidence is presented that the
dark and visible matter of the merging Bullet Cluster 1E0657-558 are widely separated, see also
http://en.wikipedia.org/wiki/Bullet Cluster. This observation alone presents a strong sup-
port in favour of theories of dark matter, and it is very hard to explain within extended theories of
gravitation that provide an alternative explanation for dark matter. Howver, Bullet Cluster seems
to undergo a high-velocity merger (the relative velocity of the two clusters is around 4500 km/s),
evident from the spatial distribution of the hot, X-ray emitting gas, which is dicult to account
for within the standard cosmology with cold dark matter and Gaussian initial seeds for galaxy
formation formed during an inationary epoch.
Furthermore, the Universe appears spatially at on cosmological scales, even though the amount
of visible and dark matter makes up only about one-third of what is required to explain the observed
atness. It is possible to get a at universe by adding a cosmological term of the right magnitude.
The true explanation may as well be more subtle, and may arise from an alternative theory of
gravitation, or from an exotic matter component which does not cluster, and which interacts with
other matter only gravitationally, or very weakly.
Finally, there are good theoretical reasons, based on which one may argue that the Einstein the-
ory cannot be the complete theory of gravitation. Namely, when the Einstein theory is canonically
quantised, one obtains a perturbatively nonrenormalisable quantum eld theory. This suggests that
the true theory of quantum gravitation is more complex than the canonically quantised Einstein
theory.
Here we briey review just a couple of simple extensions of the Einstein theory. One should
keep in mind that none of these examples solves the problem of perturbative nonrenormalisability
of the Einstein theory.
A very simple extension of the Einstein theory is the Jordan-Fierz-Brans-Dicke theory (JFBD),
(Jordan, 1949; Fierz, 1956; Brans and Dicke, 1961), in which the Newton constant is a function of
a gravitational scalar eld , which in turn couples to the trace of the matter stress-energy tensor,
and thus can vary in space and time. The action of the theory has the form (in the Jordan-Fierz
physical frame)
S
JFBD
=
c
4
16G

N
_
d
4
x

g
_
1+

)(

)
_
+ S
matter
[
matter
, g

] , (140)
where
matter
denotes matter elds, is a dimensionless constant, is a dimensionless gravitational
scalar eld, and G

N
is the bare Newton constant, such that G

N
/ reduces to the Newton constant
G
N
when = const. The equations of motion are obtained by varying the action (140),
G

=
8G

N
c
4
1

2
_
(

)(

)
1
2
g

)(

)
_
+

(g

)
_
28
g

gg

_
=
8G

N
c
4
1
2+3
T , (141)
where T

= 2(g)
1/2
S
matter
/g

is the matter stress-energy tensor, and T = g

is its trace.
From equations (141) we see that the JFBD theory reduces to the Einstein theory in the limit
when . Since the measured Newton constant is G
N
= G

N
/, a variation in space or time
in matter density will induce a variation of G
N
. Up to this moment, no space-time variations
of the Newton constant have been observed. Solar system observations place a lower bound of
> 600. On the other hand, the VLBI experiments place a stricter bound, > 3500. Further
tests of the JFBD theory are based on the observed constancy of the Newton constant. Indeed,
since G
1
N
dG
N
/dt =
1
d/dt, a lower limit on G
1
N
dG
N
/dt implies an upper limit on the time
variation of . In the JFBD theory one would naturally expect a variation of the order of the
Hubble parameter today, G
1
N
dG
N
/dt H
0
= 0.74 0.03 10
10
/year. The observed upper bound
is signicantly smaller, G
1
N
dG
N
/dt 5 10
12
/year (Lunar ranging, Viking radar reection).
Based on this bound, one cannot yet rule out the JFBD theory, since the observations which lead to
the limit are performed locally (in the Solar system and now). Since the Solar system is virialised,
one does not sense that the Universe is expanding.
The Jordan-Fierz-Brans-Dicke theory belongs to a more general class of theories, which are
known as the scalar-tensor theories (STe) of gravitation (Bergmann 1968, Nordtvedt 1970, Wagoner
1970), and whose action is obtained by generalising the action (140), such that becomes a function
of the scalar eld, = (), and one adds a potential to , V = V (). In general in STe theories
the cosmological term can be absorbed in V .
Another viable extension of the Einstein theory are the scalar-tensor-vector theories of grav-
itation (SVeTe). One example of SVeTe has recently been proposed by Beckenstein (2004) as a
generally covariant version of the nonrelativistic MOND theory of gravitation (Milgrom), which
was at once believed to provide an explanation for the missing matter problem on the galactic
scales by changing the Newton law for very small accelerations. Another possible extension of the
Einstein theory of gravitation is an old idea considered by Einstein. In this theory the metric tensor
is extended to include the antisymmetric components, g

= g

+ B

, where B

= B

.
While it has been shown that such theories are generically unstable, and hence not viable, stability
may be restored by adding a mass term. Consequently, in such a theory the Newton force law gets
modied on scales given by the inverse mass of the antisymmetric B-eld. If the mass is appropri-
ately chosen, such a theory may become a viable candidate for the explanation of the missing mass
problem in galaxies, as well as for the enhanced photon lensing observed in galactic clusters (Moat,
2004). Other ideas pursued by modern researchers include massive gravity theories and bi-metric
theories of gravity. For a recent review on the subject, we suggest Modied Gravity and Cosmol-
ogy by Timothy Clifton, Pedro G. Ferreira, Antonio Padilla and Constantinos Skordis (Phys.Rept.
513 (2012) 1-189, DOI: 10.1016/j.physrep.2012.01.001 [e-Print: arXiv:1106.2476 [astro-ph.CO]]).
Finally, we mention theories, whose action includes higher order curvature invariants, examples
of which are, 1
2
, 1

, 1

, etc. Since the equations of motions in these theories contain

higher order derivatives, they contain more than two independent solutions. Quite generically, some
of these solutions are unstable. An extra eort is needed to construct a higher order derivative theory
in which unstable modes are absent. This may be done only in very special cases, but most of these
cases can be mapped onto a scalar-tensor theory, and thus they contain no new physics.
29
Problems
1.1. Maxwells theory of electromagnetism. (5 points)
The special relativistic form of the Maxwell action coupled to matter can be written as,
o
Maxwell+matter
=
_
d
4
x
_
L
matter
+L
Maxwell
_
L
Maxwell
=
1
4

L
matter
=

. (142)
where j

= (c,

j ) represents a charged matter current density, and F

is the
antisymmetric eld-strength tensor. The F
0i
components harbour the electric eld strength, E
i
=
F
0i
, while the F
ij
components host the magnetic eld strength, B
i
= (1/2)
ijl
F
jl
;
ijl
(
123
= 1) is
a fully antisymmetric symbol in the indices i, j, l = 1, 2, 3.
By making use of the action principle, derive the following two (inhomogeneous) Maxwells
equations,

E = j
0
c ,

B
1
c

E
t
=

j (143)
How would you obtain the homogeneous Maxwells equations,

B = 0 ,

E +
1
c

B
t
= 0 ? (144)
1.2. The geodesic equation. (5 points)
By making use of the action principle, derive the geodesic equation for the 4-velocity of a point
particle, u

= dx

/d from the following general relativistic action for a point particle,

S
point particle
= m
_
ds = mc
_
d
_
g

(x)
dx

d
dx

d
_
1/2
. (145)
In proving this you may use

= 0 (see Problem 3d below). What is the special relativistic

limit of the action (145)?
1.3. General covariance and tensors. (10 points)
(a) (3 points)
Show that _
d
4
x

g (146)
represents a generally covariant measure. The symbol g = det[g

] denotes the determinant

of the metric tensor. You may nd the following denition of the determinant of a 2-indexed
tensor t

useful,

det[t

] =

, (147)
where

represents a totally antisymmetric Levi-Civit`a -symbol in 3+1 dimensions, such

that
0123
= 1, and it is antisymmetric under exchange of any two indices. The -symbol
vanishes whenever any two indices are identical.
Show that

g

transforms as the components of a four-indexed covariant tensor, while

g transforms as the components of a contravariant tensor.

30
(b) (2 points)
Show that the covariant derivative of a two indexed contravariant tensor eld T

reads (cf.
Eq. (40)),

;
= T

,
+

. (148)
(c) (2 points)
Show that the covariant derivative of the metric tensor vanishes,

= 0 . (149)
(d) (3 points)
Show that the Einstein curvature tensor G

= 1

(1/2)1g

satises the following Bianchi

identity

= 0 . (150)
Hint: Show rst the following cyclic derivative property for the Riemann curvature tensor,

= 0 , (151)
then covariantize it, and nally contract the appropriate indices.
1.4. The Schwarzschild metric. (13 points)
Consider the Schwarzschild line element (Karl Schwarzschild, 1916), which denes the metric of
a static distribution of mass, which can be approximated by a mass M located at the origin x =

0,
ds
2
= g

(x)dx

dx

= (1 + 2
N
/c
2
)c
2
dt
2

dr
2
1 + 2
N
/c
2
r
2
_
d
2
+ sin
2
() d
2
_
, (152)
where
N
= G
N
M/r denotes Newtons potential, and G
N
= 6.673(10)10
11
m
3
/kg/s
2
Newtons
constant. (For simplicity we set c = 1.)
(a) (6 points)
This solution can be derived by starting with the general spherically symmetric line element,
ds
2
= e

dt
2
e

dr
2
r
2
_
d
2
+ sin
2
() d
2
_
, (153)
where by spherical symmetry, = (r, ) and = (r, ) are functions of the radial distance
r =
_
x
2
+ y
2
+ z
2
and time . By calculating the Levi-Civit`a connection,

=
1
2
g

_
. (154)
and the Riemann and Ricci curvature tensor,
1

, 1

= 1

, (155)
show that the elements of the Einstein tensor
G

1
2
1g

, (156)
31
(here 1 = g

denotes the Ricci scalar curvature) can be expressed in terms of the

functions and dened in (153) as follows,
G
0
0
= e

_
1
r
d
dr

1
r
2
_
+
1
r
2
G
1
1
= e

_
1
r
d
dr
+
1
r
2
_
+
1
r
2
G
1
0
= e

1
r
d
d
G
0
1
= e

1
r
d
d
(157)
and
G
2
2
= G
3
3
=
1
2
e

_
1
2
d
dr
d
dr
+
1
r
d
dr

1
r
d
dr

1
2
_
d
dr
_
2

d
2

dr
2
_
+
1
2
e

_
d
2

d
2
+
1
2
_
d
d
_
2

1
2
d
d
d
d
_
,
(158)
and other elements of G

vanish.
(b) (3 points)
Upon imposing the sourceless (vacuum) Einstein equation, stating that the Einstein curvature
tensor vanishes in the vacuum,
G

= 0 , (159)
and with a help of the Bianchi identity,

= 0 , (160)
show that Eqs. (157) imply Eqs. (158), such that only the following three equations are
independent,
e

_
1
r
d
dr

1
r
2
_
+
1
r
2
= 0
e

_
1
r
d
dr
+
1
r
2
_

1
r
2
= 0
d
d
= 0 . (161)
(c) (2 points)
Next show that the general solution of these equations has the form,
re

= r + constant
+ = h() , (162)
where h() denotes a general function of time. This solution is valid everywhere in space,
except at the origin, r = 0. By the time reparametrization,
t = t() =
_

e
h(

)/2
d

(163)
this then reduces to the Schwarzschild solution (152).
32
(d) (2 points)
Discuss the physical meaning of the special points r = 0 and r = 2G
N
M/c
2
.
(e

) (extra 2 points)
What is the physical reason that there are no dynamical solutions?
NB: If you nd it too dicult to solve the time dependent problem, make a stationary ansatz
from the beginning, = (r), = (r). For this you will earn up to 10 points, if you solve (a)-(d)
correctly.
1.5. Light deection of a thermal sphere. (7 points)
Newtonian spherically symmetric gravitating systems of many particles satisfy the Poisson equa-
tion for the gravitational Newton potential
N
,

N
= 4G
N

N
, (164)
where
N
denotes the mass density, which in a spherically symmetric system is a function of the
distance r from the center of mass,
N
=
N
(r). In an equilibrated system, the distribution of
particles can be approximated by the thermal distribution function f = f(r, v), which is a function
of v = [v[ and r = [r[ only,
f =

1
(2
2
)
3/2
exp
_

v
2
/2 +
N

2
_
, (165)
where v is particles velocity,
N
=
N
(r) Newtons gravitational potential,
2
= v
2
)/3 k
B
T/m
and =
1
e

N
/
2
the density of particles, and
(r) =
_
d
3
vf . (166)
(a) (2 points)
Show that
N
satises the equation of motion,
d
2
dr
2

N
+
2
r
d
dr

N
= 4G
N

1
exp
_

N
(r)

2
_
. (167)
(b) (2 points)
Show that one analytic solution of this equation can be found, which is known as the thermal
sphere. It reads
(r) =

2
2G
N
r
2

N
(r) =
2
ln
_

2
2G
N

1
r
2
_
(168)
Next show that the mass inside a radius r reads,
M(r) =
2k
B
T
mG
N
r . (169)
Discuss the limits r 0 and r .
33
(c) (3 points)
Calculate the deection angle of light in the presence of a mass distribution of a thermal
sphere, by making use of the formula,
(d) =
2
c
2
_
d

N
(x) = (d) =
2
c
2
_
dy
x

N
(x) (170)
where

is the gradient operator in the lens plane, whose two components are transversal
(perpendicular) to the photon path, is the distance along the light geodesic, and d is the
shortest distance from the center of mass (x = 0) to the geodesic.
Assume that mass distribution of an elliptical galaxy can be well approximated by a thermal
sphere, with a typical dispersion of a velocity component = 100 km/s. Calculate the light
deection angle originating at a distant point source (quasar or galaxy).
1.6. The Friedmann equations. (10+5

points)
Consider the spatially at metric (here we set c = 1),
g

= diag(1, a
2
, a
2
, a
2
) . (171)
(a) (2 points)
Calculate the corresponding Levi-Civit`a connection,

=
1
2
g

_
. (172)
(b) (3 points)
Calculate the Riemann curvature tensor and the Ricci tensor, by making use of the expressions,
1

, 1

= 1

, (173)
and show that the Ricci tensor has the form,
1
00
= 3
a
a
, 1
ij
=
_
a
a
+ 2
a
2
a
2
_
g
ij
, (174)
while the Ricci scalar reads,
1 g

= 6
_
a
a
+
a
2
a
2
_
. (175)
(c) (2 points)
By making use of the Einstein equation (c = 1)
G

= 8G
N
T

, (176)
where
G

= 1

1
2
g

1 (177)
34
denotes the Einstein curvature tensor, denotes the cosmological term, and the stress-energy
tensor of an ideal uid equals in the uid rest frame, in which u

= (1,

0 ),
T

= ( + p)u

p , (178)
derive the Friedmann (Friedmann-Lematre-Robertson-Walker, FLRW) equations,
H
2

a
2
a
2
=
8G
N
3
+

3
a
a
=
4G
N
3
( + 3p) +

3
. (179)
(d) (3 points)
Show that the covariant stress-energy conservation implies,
+ 3H(p + ) = 0 . (180)
Show that this is not an independent constraint, and that it can be derived from (179).
Discuss the solutions of equations (179180) for the cases (1) = p = 0, =
0
= const., (2)
p = w 1/a
4
, = 0 (what is the value of w in this case?), and (3) 1/a
3
, = 0 (what
is the value of w in this case?).
(e) (5

points)
Generalize your treatment to a space-time with curved space sections, whose line element in
spherical coordinates reads,
ds
2
= g

dx

dx

= dt
2
a
2
(t)
dr
2
1 kr
2
a
2
(t)r
2
_
d
2
+ sin
2
() d
2
_
. (181)
When k = +1 (k = 1) this metric describes an expanding universe with positively (neg-
atively) curved spatial sections. For k = 0, the metric (181) reduces to the spatially at
metric (171). Derive the corresponding Friedmann equations.
1.7. Fermions in curved space-times. (5 points)
Consider the following covariant Dirac action for fermions in curved space times,
S
fermion
=
_
d
4
x

g
_

+ m

_
, (182)
where

=

0
(x) and m

denotes the fermion mass. The covariant derivative acting on a fermion

eld is given in terms of the spin connection

as,

= (

) , (183)
which is in turn dened by

= 0 . (184)
35
(a) (3 points)
Calculate the elements of the spin connection

and of the Levi-Civit`a connection in homo-

geneous conformal space-times, whose metric is of the following conformally at form,
g

= a()
2

e
a

(x)e
b

(x)
ab
,
ab
= diag(1, 1, 1, 1) , (a, b = 0, 1, 2, 3) , (185)
where a = a() denotes the scale factor, which is a function of conformal time (e.g. in de
Sitter space-time, a = 1/(H
I
) ( < 0), where H
I
denotes the Hubble parameter of de Sitter
space). Show rst that in conformal space-times the vierbein has the form
e
c

(x) =
c

a(x) , e

c
(x) =

c
a(x)
1
, ( = 0, 1, 2, 3; c = 0, 1, 2, 3) . (186)
such that the Dirac matrices are,

(x) = e

a
= a(t)
1

a
. (187)
(b) (2 points)
By making use of (182) and (183), show that the equation of motion for fermions in homoge-
neous conformal space-times can be written as
_

0
+
i

i
iam

c
= 0 ,
c
= a
3/2
, (188)
where here
0
and
i
denote the at space Dirac matrices. Comment on the physical impli-
cations of this result.
1.8. The Jordan-Fierz-Brans-Dicke (JFBD) theory of gravitation. (3 + 5

points)
(a) (3 points)
The equation of motion for the scalar gravitational eld in the JFBD theory reads,
g

gg

_
=
8G

N
c
4
1
2 + 3
T , (189)
where G

N
denotes the bare Newton constant, is a dimensionless constant, T = g

is the
trace of the matter stress-energy tensor, T

= 2(g)
1/2
S
matter
/g

. Assume that within

a FLRW universe, the dominant matter component is scaling as nonrelativistic particles, and
estimate the variation of the Newton constant with time,

G
N
/G
N
, where G
N
= G

N
/.
(b

) (3

points)
In the Jordan-Fierz (physical) frame, in which the JFBD action reads,
S
JFBD
=
c
4
16G

N
_
d
4
x

g
_
1+

)(

)
_
+ S
matter
[
matter
, g

] , (190)
where
matter
denotes the matter elds. Show that the JFBD action in the Einstein (conformal)
frame, which is dened by the following conformal rescaling of the metric tensor,
g

= A
2
()g
E

, A
2
() =
1
, ()
2

_
d ln(A)
d
_
2
=
1
2 + 3
, (191)
reads
S
E
=
c
4
16G

N
_
d
4
x
_
g
E
_
1
E
+ 2g
E

)(

)
_
+ S
matter
[
matter
, A
2
()g
E

] , (192)
with g
E
= det[g
E

].
36
(c

) (2

points)
What are the equations of motion for g
E

and in the Einstein frame?

37