General Relativity: Solutions to exercises in

Lecture I

January 22, 2018

Exercise 1
Consider a binary system of gravitating objects of masses M and m.

• First consider the case in which m  M and where the small-mass object is in quasi-circular
orbit around the more massive object. Draw the trajectory in two-space and the worldline in a
1 + 1- and in a 2 + 1-dimensional spacetime [Hint: use a co-ordinate system centred in M ].

• Now let m = M and the binary be in circular orbit around the Newtonian centre of mass of the
system. Draw the trajectory in two-space and the worldline in a 1 + 1- and in a 2 + 1-dimensional
spacetime [Hint: use a co-ordinate system centred in the Newtonian centre of mass].

Solution 1

Figure 1: Trajectories in two-space for the cases m  M (left) and m = M (right).

1
Figure 2: Worldline in 1 + 1-dimensional spacetime for the case m  M in polar co-ordinates (left
panel) and Cartesian co-ordinates (middle panel), and for the case m = M in Cartesian co-ordinates
(right panel).

Figure 3: Worldline in 2 + 1-dimensional spacetime for the case m  M (left) and m = M (right).

2
Exercise 2
Consider a two-dimensional space and cover it with two co-ordinate maps: a Cartesian map where
0
{xµ } = (x, y) and a polar map where {xµ } = (r, θ).
0
• Find the co-ordinate transformation f : xµ → xµ
0
• Find the inverse co-ordinate transformation f −1 : xµ → xµ
0 0
 
• Find the components of the transformation matrix Λµµ and its determinant J 0 := ∂xµ /∂xµ 
0
• Find the components of the inverse transformation matrix Λµµ0 and its determinant J := ∂xµ /∂xµ 
 

0 0
• Show that Λµµ Λµν 0 = δ µν 0 and that J J 0 = 1

Solution 2
The co-ordinate transformation is given by:
(
1/2
r = (x2 + y 2 )
f : (1)
θ = arctan (y/x)

The inverse co-ordinate transformation is given by:


−1 x = r cos θ
f : (2)
y = r sin θ

The transformation matrix is given by:

0
0 ∂xµ
Λµµ =
∂xµ !
∂r/∂x ∂r/∂y
=
∂θ/∂x ∂θ/∂y
−1/2 −1/2
!
x (x2 + y 2 ) y (x2 + y 2 )
= −1 −1
(3)
−y (x2 + y 2 ) x (x2 + y 2 )
 
cos θ sin θ
≡  1 1  , (4)
− sin θ cos θ
r r
and its determinant is given by:
0
J 0 = ∂xµ /∂xµ 
 

x2 y2
= +
(x2 + y 2 )3/2 (x2 + y 2 )3/2

−1/2
= x2 + y 2 , (5)

3
or alternatively, using equation (4), is given by:

cos2 θ sin2 θ
J0 = +
r r
1
= . (6)
r
It is trivial to confirm that both expressions for J 0 are equivalent.

The inverse transformation matrix is given by:

∂xµ
Λµµ0 =
∂xµ0 !
∂x/∂r ∂x/∂θ
=
∂y/∂r ∂y/∂θ
!
cos θ −r sin θ
= (7)
sin θ r cos θ
−1/2
!
x (x2 + y 2 ) −y
≡ −1
, (8)
y (x2 + y 2 ) x

and its determinant is given by:

 0
J = ∂xµ /∂xµ 
= r cos2 θ + r sin2 θ
= r, (9)

or alternatively, using equation (8), is given by

x2 y2
J = +
(x2 + y 2 )1/2 (x2 + y 2 )1/2

1/2
= x2 + y 2 . (10)

It is again trivial to confirm that both expressions for J are equivalent.

Matrix multiplication of equations (3) and (8) or equations (4) and (7) yields the identity matrix,
0 0
confirming the result Λµµ Λµν 0 = δ µν 0 . It is also straightforward to confirm that J J 0 = 1 in both
co-ordinate systems.

Exercise 3
Consider a three-dimensional space and cover it with two co-ordinate maps: a Cartesian one where
0
{xµ } = (x, y, z) and a polar one where {xµ } = (r, θ, φ). Address all of the questions in Exercise 2.

4
Solution 3
The co-ordinate transformation is given by:

2 2 2 1/2
 r = (x + yh + z )

 i
−1/2
f : θ = arccos z (x2 + y 2 + z 2 ) (11)


φ = arctan (y/x)


The inverse co-ordinate transformation is given by:


 x = r sin θ cos φ
−1
f : y = r sin θ sin φ (12)
z = r cos θ


The transformation matrix is given by:

0
0 ∂xµ
Λµµ =
∂xµ
∂r/∂x ∂r/∂y ∂r/∂z
 

=  ∂θ/∂x ∂θ/∂y ∂θ/∂z 

 

∂φ/∂x ∂φ/∂y ∂φ/∂z

−1/2 −1/2 −1/2
 
x (x2 + y 2 + z 2 ) y (x2 + y 2 + z 2 ) z (x2 + y 2 + z 2 )
 
2 2 1/2
 xz yz (x + y ) 
=  − (13)
 
2 1/2 2 2 2 2 2 1/2
 (x + y ) (x + y + z ) (x + y ) (x2 + y 2 + z 2 )
2 x2 + y 2 + z 2


 
−1 −1
y (x2 + y 2 ) x (x2 + y 2 ) 0
 
sin θ cos φ sin θ sin φ cos θ
1 1 1
 
−

 cos θ cos φ cos θ sin φ sin θ
≡ r  , (14)

 r r 
 sin φ cos φ 
0
r sin θ r sin θ
whereby, upon simplification, its determinant may be found as:
0
J 0 = ∂xµ /∂xµ 
 

−1/2 2
−1/2
= x2 + y 2 x + y2 + z2 , (15)

or alternatively, using equation (14), is given by:

1
J0 = . (16)
r2 sin θ
It is once more trivial to confirm that both expressions for J 0 are equivalent.

5
The inverse transformation matrix is given by:
∂xµ
Λµµ0 =
∂xµ0
 
∂x/∂r ∂x/∂θ ∂x/∂φ
= ∂y/∂r ∂y/∂θ ∂y/∂φ
 
∂z/∂r ∂z/∂θ ∂z/∂φ
 
sin θ cos φ r cos θ cos φ −r sin θ sin φ
=  sin θ sin φ r cos θ sin φ r sin θ cos φ  (17)
 
cos θ −r sin θ 0
 
−1/2 −1/2
x (x2 + y 2 + z 2 ) xz (x2 + y 2 ) −y
2 −1/2 2 −1/2
 
≡ 
 y (x 2
+ y 2
+ z ) yz (x 2
+ y ) x  ,
 (18)
−1/2 1/2
z (x2 + y 2 + z 2 ) − (x2 + y 2 ) 0

whereby, upon simplification, its determinant may be found as:

J = r2 sin θ , (19)

or alternatively, using equation (18), is given by:

1/2
1/2
J = x2 + y 2 x2 + y 2 + z 2 . (20)

As before, it is again trivial to confirm that both expressions for J are equivalent.

As in Exercise 2, Matrix multiplication of equations (13) and (18) or equations (14) and (17) yields
0 0
the identity matrix, confirming the result Λµµ Λµν 0 = δ µν 0 . It is also straightforward to confirm that
J J 0 = 1 in both co-ordinate systems.

6
 General Relativity: Solutions to exercises in
Lecture II

January 22, 2018

Exercise 1
0
Consider two co-ordinate systems in a two dimensional space {xµ } = (x, y) and {xµ } = (r, θ) which
are related through the well-known co-ordinate transformation
(
1/2
r = (x2 + y 2 )
f :
θ = arctan (y/x)

and its inverse 

x = r cos θ
−1
f :
y = r sin θ
Discuss the differences between the transformation matrix employed to transform a covector in this
space    
µ0
d̃x = Λ µ d̃x , (1)
µ µ0
and the one employed in the co-ordinate transformation
0 0
xµ = Λµµ xµ . (2)

Solution 1
   
µ0
The matrix involved in the transformation of the gradient d̃x = Λ µ d̃x is different from the
µ µ0
µ0 µ0
matrix used in the transformation x = Λ µ xµ . The two matrices, although written identically, are in
fact transposes of each other.
0
To illustrate this, consider the co-ordinate systems {xµ } = (x, y) and {xµ } = (r, θ). It follows that
0 0
x1 = x, x2 = y; x1 = r, x2 = θ. One may now calculate the transformation between co-ordinate
systems as:
0 0
x1 = r = Λ1µ xµ
0 0
= Λ11 x1 + Λ22 x2
0 0
∂x1 1 ∂x1 2
= x + x
∂x1 ∂x2
∂r ∂r
= x+ y, (3)
∂x ∂y

1
 0
and similarly x2 = (∂θ/∂x)x + (∂θ/∂y)y. We may now write the transformation matrix as:
!
0 ∂r/∂x ∂r/∂y
Λµµ = . (4)
∂θ/∂x ∂θ/∂y
   
0
On the other hand, for d̃x = Λµµ d̃x , consider the following explicit transformation:
µ µ0
 0

 
= Λµ1 d̃x
d̃x
1 µ0
0
  0
 
= Λ11 d̃x + Λ21 d̃x
10 20
∂r   ∂θ  
= d̃x + d̃x . (5)
∂x 10 ∂x 20
     
Similarly, one finds d̃x = (∂r/∂y) d̃x + (∂θ/∂y) d̃x . The transformation matrix may now
2 10 20
be written as:
!
0 ∂r/∂x ∂θ/∂x
Λµµ = (6)
∂r/∂y ∂θ/∂y
0
= χµµ . (7)
0
T 0
Clearly χµµ = Λµµ from equation (4), i.e. the transformation matrices are transposes of each other,
as required.

Exercise 2
0
Consider two co-ordinate systems in a four-dimensional spacetime xµ = (t, x, y, z) and xµ = (u, v, y, z)
that are related through the co-ordinate transformation

u=t−x

f :
v =t+x

and its inverse (

−1
t = 21 (v + u)
f :
x = 12 (v − u)

• Compute the matrices employed in the transformations

0 0 0
xµ = Λµµ xµ xµ = Λµµ0 xµ . (8)

• Consider a four-vector with components U µ = (1, 0, 0, 0)T in the co-ordinate system xµ and
0 0
compute the new components U µ in the co-ordinate system xµ .

• Repeat the calculation for the new vector V µ = (−1/2, 1/2, 0, 0)T . Interpret the results.

2
Solution 2
0
For the first part of the question, computing the transformation matrices, first consider Λµµ .
0
0 ∂x0
Λ0µ =
∂xµ
∂u
= , (9)
∂xµ
from which one obtains the following non-zero components:
0 ∂u
Λ00 = =1, (10)
∂t
0 ∂u
Λ01 = = −1 . (11)
∂x
Similarly,
0
0 ∂x1
Λ1µ =
∂xµ
∂v
= , (12)
∂xµ
from which one obtains the following non-zero components:
0 ∂v
Λ10 = =1, (13)
∂t
0 ∂v
Λ11 = =1. (14)
∂x
0
Finally, one may also show that the remaining non-zero components of Λµµ are
0
Λ22 = 1 , (15)
0
Λ33 = 1 . (16)

The transformation matrix may now be written as

1 −1 0 0
 

1 1 0 0
 
µ0
Λµ=   . (17)
0 0 1 0


0 0 0 1

For the inverse transformation matrix we follow the same procedure, from which the inverse transfor-
mation matrix is found as 
1/2 1/2 0 0


−1/2 1/2 0 0
 
µ
Λ µ0 = 
 0
 . (18)
 0 1 0 
0 0 0 1

3
 0
The second part of the question asks to calculate U µ in the new co-ordinate system, i.e. U µ . Whilst
it is obvious that one can do this through matrix multiplication, consider instead the following:
0 0
Uµ = Λµµ U µ
0
= Λµ0 U 0
0
= Λµ0 , (19)

where the fact that the only non-zero component of U µ is U 0 has been used. One can then read directly
from equation (17) the solution as
U µ = (1, 1, 0, 0)T . (20)

For the third and final part of this question one can again apply matrix multiplication to obtain the
result, or consider the basis components as follows:
0 0
Vµ = Λµµ V µ
0 0
= Λµ0 V 0 + Λµ1 V 1 . (21)

Considering this term by term yields

0 0 0
V0 = Λ00 V 0 + Λ01 V 1
= (1).(−1/2) + (−1).(1/2)
= −1 , (22)

and
0 0 0
V1 = Λ10 V 0 + Λ11 V 1
= (1).(−1/2) + (1).(1/2)
= 0, (23)

from which it immediately follows that

0
V µ = (−1, 0, 0, 0)T . (24)

The second part may be interpreted as follows. In {xµ } the four-vector U µ represents a particle at
rest, since all spatial components are zero: the particle may be represented as a vertical worldline
0 0
in a 1 + 1-spacetime. However, when transforming to {xµ } one finds that U µ has two non-zero
components, implying that the particle no longer appears stationary and is moving with a constant
velocity. Represented as a worldline in a 1 + 1-spacetime (u, v) the worldline would be a line of constant
positive (and finite) gradient.
For the third part, the vector V µ has non-zero spatial components and so has a velocity of −1 in the
x-direction. Represented as a worldline in a 1 + 1-spacetime (t, x) it would be represented by a line of
0 0
constant, finite and non-zero gradient. However, when transformed into {xµ }, the four-vector V µ has
0
zero spatial components. So in the co-ordinate system {xµ } the four-vector V µ appears stationary.

4
Exercise 3
Consider a 1 + 1 representation of the sub-spaces with two co-ordinate systems (t, x) and (u, v).

• Draw in the two spacetimes the worldline of a particle with velocity ẋ := dx/dt = 0.

• Draw in the two spacetimes the worldline of a particle with velocity ẋ := k (x = kt) with k < 1.

• Interpret the results.

Solution 3
In this question it is assumed we use the co-ordinate transformations as defined in Exercise 2.

For the first part, let us term the first particle as particle A. Since ẋA = 0 this implies xA = const. The
particle is stationary and at rest in the (t, x) co-ordinate system. In the (u, v) co-ordinate system one
may write

u A = t − xA , (25)
v A = t + xA , (26)

from which one may conclude

uA t − xA
= < 1. (27)
vA t + xA
Since ∂uA /∂vA ' (t − xA )/(t + xA ) = (vA − 2xA )/v. Integrating this yields
2xA
u(v) = v − + const. (28)
v2
We may set the integration constant to zero without loss of generality. We may now plot equation
(28) for various values of xA , the case of xA = 0 corresponding to a straight line of constant gradient
1. The worldlines in both co-ordinate systems are illustrated in Figure 1 by the solid blue line.
For the second part of this question let us term the second particle as particle B. For particle B
one has ẋB = k (i.e. xB = kt), where k < 1. The particle is now moving with constant velocity k
and can be represented as a worldline of gradient k < 1 in the (t, x) co-ordinate system. In the (u, v)
co-ordinate system one may write

uB = t − kt = t(1 − k) , (29)
vB = t + kt = t(1 + k) , (30)

from which one may conclude

uB 1−k
= . (31)
vB 1+k
The condition that (1 − k)/(1 + k) > 0 implies that |k| < 1. Considering values of k in this range, the
following condition on the gradient of the worldline may be obtained

uB 1−k < 1 if k > 0 (Case B) ,
=
vB 1+k > 1 if k < 0 (Case B0 ) .

5
Figure 1: Worldliness for particles A and B in the (t, x) co-ordinate system (left) and the (u, v)
co-ordinate system (right).

The worldlines in both co-ordinate systems are illustrated in Figure 1 by the dashed orange line.

For the final part of the question, for particle A, in the (t, x) co-ordinate system it is at rest. However,
in the (u, v) co-ordinate system it is moving with constant velocity. For particle B, consider the limit
k → 1, whereby ∂x/∂t = 1 and ∂uB /∂vB = 0. In the (x, y) co-ordinate system the particle is moving
with constant velocity, but in the limit k → 1, in the (u, v) co-ordinate system this implies that the
particle appears stationary (or the (v, u) co-ordinate system depending on how one labels the axes).

6
 General Relativity: Solutions to exercises in
Lecture III

January 22, 2018

Exercise 1
Consider T as a contravariant tensor of rank 2 with components T µν . Under what conditions can this
tensor be cast as the product of two contravariant vectors U and V, i.e. such that T µν = U µ V ν ?

Solution 1
In a given basis T is represented by a matrix T µν . In these terms a necessary and sufficient condition
to enable T µν to be written as T µν = U µ V ν is that all columns of the matrix T µν must be proportional
to each other (linearly dependent). As an example, consider the following matrix:
 
1 2 4 8
2 4 8 16
T µν =   .
 
3 6 12 24
4 8 16 32
Since the columns of this matrix are proportional to one another, we may choose U µ = (1, 2, 3, 4) and
V ν = (1, 2, 4, 8), thus satisfying T µν = U µ V ν .
Let us now consider this in a co-ordinate independent (covariant) way. T µν = U µ V ν if and only if
S = T µν xν is in the same direction, for any given xν .
µ

Consider the set of orthonormal basis vectors e0 , e1 , e2 and e3 which by definition must satisfy
eµ eν = δ µν . The direction of S µ is independent of the choice of xν (by linearity) if and only if it
is independent of our basis vectors e0 , e1 , e2 and e3 . As such we may obtain the following condition:
T µν eαν = T µα
= Cα S µ ,
where Cα = (C0 , C1 , C2 , C3 ) are constants. Explicitly:
T µ0 = C0 S µ ,
T µ1 = C1 S µ ,
T µ2 = C2 S µ ,
T µ3 = C3 S µ .
Thus the columns must be proportional to each other.

1
Exercise 2
Consider the following equation:
T µν = U µ + V ν .
Is T a generic tensor?

Solution 2
T is not a generic tensor. If T were a tensor then T µν Aµ Bν would have to be a scalar. Instead, one
obtains
T µν Aµ Bν = (U µ + V ν ) Aµ Bν
= (U µ Aµ ) Bν + (V ν Bν ) Aµ
= αBν + βAµ ,
where α ≡ U µ Aµ and β ≡ V ν Bν are both scalars. It immediately follows that αBν + βAµ is not a
scalar and therefore T is not a generic tensor.

Exercise 3
Consider F as a tensor of rank 2 with covariant components Fµν and that is also antisymmetric in one
co-ordinate system, i.e. Fµν = −Fνµ .
• Show that Fµν is antisymmetric in all co-ordinate systems.
• Does the antisymmetry in the covariant indices also apply to the contravariant indices?
• If so, show that F µν is antisymmetric in all co-ordinate systems.

Solution 3
First consider the transformation of Fµν into another co-ordinate system:
Fµ0 ν 0 = Λµµ0 Λνν 0 Fµν
= −Λµµ0 Λνν 0 Fνµ
= −Λνµ0 Λµν 0 Fµν
= −Fν 0 µ0 .
It immediately follows that Fµν is symmetric in all co-ordinate systems. The antisymmetry in covariant
indices indeed also applies to the contravariant indices since F is a tensor. This can be shown by
considering the following:
0 0
F µν = g µµ g νν Fµ0 ν 0
0 0
= −g µµ g νν Fν 0 µ0
0 0
= −g µν g νµ Fµ0 ν 0
= −F νµ ,
as required.

2
Exercise 4
For the first part of the question, consider the antisymmetric tensor Aµν such that Aµν = −Aνµ and
the symmetric tensor B µν such that B µν = B νµ . Prove the following identities:

Aµν B µν = 0 , (1)
1 µν
V µν Aµν = (V − V νµ ) Aµν , (2)
2
1 µν
V µν Bµν = (V + V νµ ) Bµν . (3)
2

Solution 4
For the first identity consider the following:

Aµν B µν = −Aνµ B µν
= −Aµν B µν
= 0,

where we have used the antisymmetry of Aµν in the first step and relabelling dummy indices and the
symmetry of B µν in the second step, hence the required result. For the second and third parts, recall
that a generic rank 2 tensor may be written in terms of a symmetric and antisymmetric component as
follows:
1 µν 1
V µν = (V + V νµ ) + (V µν − V νµ )
2 2
(µν) [µν]
= V +V .

Now consider the action of the antisymmetric tensor Aµν on the symmetric part of V µν , i.e. V (µν) :
1 µν
V (µν) Aµν = (V Aµν + V νµ Aµν )
2
1 µν
= (V Aµν + V µν Aνµ )
2
1 µν
= (V Aµν − V µν Aµν )
2
= 0,

where in the first step we have relabelled dummy indices in the second term, and in the second step
we have used the antisymmetry of Aµν . In a similar fashion we may also consider the action of the
symmetric tensor Bµν on the antisymmetric part of V µν , i.e. V [µν] :
1 µν
V [µν] Bµν = (V Bµν − V νµ Bµν )
2
1 µν
= (V Bµν − V µν Bνµ )
2
1 µν
= (V Bµν − V µν Bµν )
2
= 0.

3
We are now in a position to tackle the second and third identities. For the second identity, consider
the following:

V µν Aµν = V (µν) Aµν + V [µν] Aµν

= V [µν] Aµν
1 µν
= (V − V νµ ) Aµν ,
2
as required.

Finally, we consider the third identity:

V µν Bµν = V (µν) Bµν + V [µν] Bµν

= V (µν) Bµν
1 µν
= (V + V νµ ) Bµν ,
2
as required.

4
 General Relativity: Solutions to exercises in
Lecture IV

January 22, 2018

Exercise 1
Using a co-ordinate system (t, r, θ, φ), consider the metric line element given by

dr2
 
2 2 2 2 2 2 2


ds = −dt + a (t) + r dθ + sin θ dφ , (1)
1 − κr2

where κ = −1, 0, 1.

• Show that a new co-ordinate system (t, χ, θ, φ) the line element (1) can be rewritten as

ds2 = −dt2 + a2 (t) dχ2 + f (χ)2 dθ2 + sin2 θ dφ2 .



(2)

• Find the form of the function f (χ) for κ = −1, 0 and 1.

• Discuss the properties of the metric in the case of κ = 0. [Hint: two metrics g and g0 are
conformally related if it is possible to express them as g = Ω g0 , where Ω ≡ Ω(xµ ) is a generic
function and is referred to as the conformal factor ].

Solution 1
From the invariance of the line element we may write

dr2
 
2 2 2 2 2 2 2


ds = −dt + a (t) + r dθ + sin θ dφ
1 − κr2
= −dt2 + a2 (t) dχ2 + f (χ)2 dθ2 + sin2 θ dφ2



= ds02 .

Letting dt = dθ = dφ = 0 we obtain the relation

dr
√ = dχ . (3)
1 − κr2
This expression may be integrated directly to yield χ as a function of r, yielding:

1
 
 arcsinh r + c ,
χ= r+c , (4)
arcsin r + c ,


for κ = −1, 0 and 1 respectively, and√where c is a constant of integration. Note the inverse hyperbolic
function identity arcsinh r = ln |r + 1 + r2 |, which is also a solution for κ = −1. Since f (χ) = r we
obtain the result 
 sinh χ ,
f (χ) = χ, (5)
sin χ ,


for κ = −1, 0 and 1 respectively, and where we have assumed c = 0.

For the final part of this question, setting κ = 0 yields the line element as

ds2 = −dt2 + a2 (t) dr2 + r2 dθ2 + sin2 θ dφ2 .



(6)

By factoring out the expansion factor a(t) we obtain

dt2
 
2 2 2 2 2 2 2


ds = a (t) − 2 + dr + r dθ + sin θ dφ . (7)
a (t)

Let us define the “conformal time” t̃, where dt̃2 = dt2 /a(t)2 . We may now re-write the metric (7) as

ds2 = a2 (t) −dt̃2 + dr2 + r2 dθ2 + sin2 θ dφ2



(8)
2 2
= a (t) dsMinkowski ,

where the Minkowski line element is the line element for flat space. Thus in the case κ = 0 the metric
is conformally flat. This metric is in general known as the (Friedmann-Lemaitre) Robertson-Walker
(FL)RW metric and is widely used in cosmology to describe an expanding universe.

Exercise 2
Using a co-ordinate system (η, χ, θ, φ), consider the metric line element given by

ds2 = Ω2 −dη 2 + dχ2 + sin2 χ dθ2 + sin2 θ dφ2 .



(9)

Consider now a new co-ordinate system (τ, ρ, θ, φ) where

2 sin η
τ = (10)
cos χ + cos η
2 sin χ
ρ = , (11)
cos χ + cos η

and find the expression of the metric (9) in this new co-ordinate system. Discuss the properties of this
new metric.

2
Solution 2
To calculate the expression for the new metric we must use the following co-ordinate transformation:
0 0 0 0
g α β = Λαα Λββ g αβ . (12)
We must calculate the contravariant metric components since we only have (τ, ρ) in terms of (η, χ) and
not the inverse relationship. It is much simpler to calculate the contravaraint metric components and
then calculate the matrix inverse of the contravariant metric than to define the inverse transformation.
Since our metric is diagonal we may exploit the fact that g αβ = 1/gαβ . Evaluating the non-zero
components of the transformation matrix, we obtain:
0 ∂τ
Λ00 =
∂η
2(1 + cos η cos χ)
= (13)
(cos χ + cos η)2
∂ρ
=
∂χ
0
= Λ11 ,

0 ∂ρ
Λ10 =
∂η
2 sin χ sin η
= (14)
(cos χ + cos η)2
∂τ
=
∂χ
0
= Λ01 ,
and
0 0
Λ22 = Λ33 = 1 . (15)
Note that the transformation matrix is diagonal. The contravariant metric may now be calculated
coefficient by coefficient, yielding
0 0
 0 2  0 2
g 0 0 = Λ00 g 00 + Λ01 g 11
 
1  00 2  00 2
= − 2 Λ0 − Λ1
Ω
1 4
= − 2 , (16)
Ω (cos χ + cos η)2
 2  2
10 10 0 0
g = Λ10 g 00
+ Λ11 g 11
 2 
1  10 2  1 0
= Λ1 − Λ0
Ω2
 
1  00 2  00 2
= Λ0 − Λ1
Ω2
0 0
= −g 0 0 , (17)

3
 0 0 1
g2 2 =
Ω2 sin2 χ
= g 22
4
=
Ω2 (cos χ + cos η)2 ρ2
1 00
= − 2 g0 0 , (18)
ρ
and
0 0 1
g3 3 =
Ω2 sin θ sin2 χ
2

= g 33
1 20 20
= g
sin2 θ
1 0 0
= − 2 2 g0 0 . (19)
ρ sin θ
Thus we may write the contravariant metric components in the new co-ordinate system as
 
−1 0 0 0
 
1 4  0 1 0 0 
α0 β 0
g = 2  . (20)
 
2  2 −1
Ω (cos χ + cos η)  0 0 (ρ ) 0 
 
2 2

−1
0 0 0 ρ sin θ

Let us now define a new conformal factor

2 2
e 2 ≡ Ω (cos χ + cos η) ,
Ω (21)
4
which immediately enables us to write the covariant components of our metric tensor in the new
co-ordinate system as
−1 0 0
 
0
0 1 0 0 
 
2
gα0 β 0 = Ω 
e
2
 . (22)
0 0 ρ 0  
0 0 0 ρ2 sin2 θ
We may now write our metric in the new co-ordinate system as follows
e 2 −dτ 2 + dρ2 + ρ2 dθ2 + ρ2 sin2 θ dφ2 .
ds2 = Ω


(23)

As can be seen in the above expression, the new metric is conformally flat.

4
Exercise 3
Given the four-vector u such that uα uα = −1 and the tensor hµν ≡ gµν + uµ uν , prove the following
identities
hµν uµ = 0 , hµν hλµ = hλν , hµµ = 3 . (24)

Solution 3
For the first identity, consider the following

hµν uµ = gµν uµ + uµ uν uµ
= uν + uν (uµ uµ )
= uν − uν
= 0. (25)

For the second identity, we must first derive an expression for hµν as follows

hµν = g µα hαν
= δ µν + uµ uν . (26)

Using this we may write the following

hµν hλµ = (δ µν + uµ uν ) δ λµ + uλ uµ

= δ µν δ λµ + δ µν uλ uµ + δ λµ uµ uν + uµ uν uλ uµ
= δνλ + uλ uν + uλ uν + uλ uν (uµ uµ )
= δνλ + uλ uν
= hλν . (27)

For the third and final identity, consider the following

hµµ = g µν hνµ
= g µν gµν + g µν uµ uν
= δ µµ + uµ uµ
= 4−1
= 3. (28)

Note: the tensor hµν defines a projection onto a hypersurface orthogonal to uµ (i.e. hµν uµ uν = 0).
For any non-null vector uµ (i.e. uµ uµ 6= 0), one may define the projection operator orthogonal to uµ as

Pu ≡ hµν
uµ uν
= gµν − . (29)
uµ uµ

5
Exercise 4
Consider the following antisymmetric tensor

Fαβ = −2E[α uβ] + αβ γδ Hγ uδ . (30)

Express the vectors E and H in terms of the tensor F. [Hint: contract Fαβ with uβ & αβγδ respectively.]

Solution 4
First, let us write the expression for the antisymmetric part of Eα uβ out in full, which reads as

1
E[α uβ] = (Eα uβ − Eβ uα ) . (31)
2
We may then substitute this expression into equation (30), yielding

Fαβ = −Eα uβ + Eβ uα + αβ γδ Hγ uδ . (32)

Let us work with the above expression for the remainder of the question. Contracting Fαβ with uβ
yields

Fαβ uβ = −Eα uβ uβ + Eβ uα uβ + αβ γδ Hγ uδ uβ

= Eα + Eβ uα uβ + αβ γδ Hγ uδ uβ . (33)

Before proceeding further, let us turn our attention to the last term in equation (33). The Levi-Civita
tensor may be re-written in a fully contravariant form as

αβ γδ = gαµ gβν µνγδ , (34)

which simplifies the third term in equation (33) as follows

αβ γδ Hγ uδ uβ = gαµ gβν µνγδ uδ uβ Hγ

= gαµ µνγδ uδ uν Hγ (lower index with gβν )
= gαµ µδγν uν uδ Hγ (δ ↔ ν as dummy indices)
= −gαµ µνγδ uν uδ Hγ (permute δ ↔ ν in µδγν )
= −gαµ µνγδ uδ uν Hγ (compare with second line)
= 0. (35)

We thus obtain

Fαβ uβ = Eα + Eβ uα uβ
= hβα Eβ , (36)

as required. This may also be written as

Fαβ uβ = hαβ E β . (37)

6
For the second part of the question, first recall the definition of the dual of a tensor
1
F ∗γδ = Fαβ αβγδ . (38)
2
Contracting Fαβ with αβγδ yields

Fαβ αβγδ = 2F ∗γδ

= −Eα uβ αβγδ + Eβ uα αβγδ + αβ γδ Hγ uδ αβγδ . (39)

Let us attack the third term in the above expression by employing the identity we derived in equation
(34) as
αβ γδ = g µγ g νδ αβµν . (40)
With this expression we may re-write the third term as

αβ γδ Hγ uδ αβγδ = g µγ g νδ Hγ uδ αβµν αβγδ

= H µ uν αβµν αβγδ . (41)

We may then expand the contraction over the Levi-Civita tensors as

αβµν αβγδ = −2! δµν

γδ
 
δ γ δ γ 
µ ν
= −2  δ δ 
 
 δµ δν 

= 2 δνγ δµδ − δµγ δνδ

, (42)

from which we may immediately simplify equation (41), yielding

H µ uν αβµν αβγδ = 2H µ uν δνγ δµδ − δµγ δνδ

= 2 H δ uγ − H γ uδ


(43)

We may now re-write equation (39) as

2F ∗γδ = −Eα uβ αβγδ + Eβ uα αβγδ + 2 H δ uγ − H γ uδ

. (44)

Recall from equation (35) the vanishing of the contraction of the Levi-Civita tensor over two indices
with two 4-vectors. This suggests to us that contracting equation (44) with uδ will allow us to eliminate
the first two terms in (44). With this knowledge in mind, contracting with uδ yields

2F ∗γδ uδ = −Eα uβ uδ αβγδ + Eβ uα uδ αβγδ + 2 H δ uγ uδ − H γ uδ uδ

= 2 H δ uγ uδ − H γ uδ uδ ,


(45)

from which it immediately follows that

F ∗γδ uδ = H δ uγ uδ − H γ uδ uδ
= H γ + H δ uγ uδ . (46)

7
As before, the above expressions may be written more succinctly in terms of the projection tensor as

F ∗γδ uδ = hγδ H δ
= hγδ Hδ . (47)

For a physical interpretation consider an orthonormal comoving frame with uµ = (1, 0, 0, 0) and uµ =
(−1, 0, 0, 0), i.e. uµ uµ = −1. In this frame

Fαβ uβ = Fα0
= Eα + E0 uα . (48)

When α = 0

F00 = E0 − E0
= 0. (49)

Additionally

Fi0 = Ei + E0 ui
= Ei , (50)

where i = 1, 2, 3. If Fαβ is the electromagnetic field tensor then Ei is the 3-vector of the electric field.
Next consider the dual tensor

F ∗γδ uδ = −F ∗γ0
= H γ − H 0 uγ . (51)

When γ = 0 then

F ∗00 = − H 0 − H 0

= 0. (52)

Additionally

F ∗i0 = −(H i − H 0 ui )
= −H i , (53)

where again i = 1, 2, 3. H i can be interpreted as the 3-vector of the magnetic field.

8
 General Relativity: Solutions to exercises in
Lecture V

January 22, 2018

Exercise 1
Let F be a rank-2 antisymmetric tensor, G a rank-2 symmetric tensor and X and rank-3 antisymmetric
tensor. Provide explicit expressions for the following tensors: Fµν , F[µν] , F(µν) , G[µν] , G(µν) , X[αβγ] ,
X(αβγ) , X[αβ]γ , X(αβ)γ , X[αβ](γ) and X(αβ)[γ] .

Solution 1
• Fµν = −Fνµ

• F[µν] = 21 (Fµν − Fνµ ) = 21 (Fµν + Fµν ) = Fµν

• F(µν) = 12 (Fµν + Fνµ ) = 21 (Fµν − Fµν ) = 0

• G[µν] = 0 (the antisymmetric part of a totally symmetric tensor must be zero)

• G(µν) = Gµν
1
• X[αβγ] = 3!
(Xαβγ − Xβαγ + Xγαβ − Xαγβ + Xβγα − Xγβα ) = 16 (2Xαβγ + 2Xγαβ + 2Xβγα )

= 13 (Xαβγ + Xγαβ + Xβγα )

1
• X(αβγ) = 3!
(Xαβγ + Xβαγ + Xγαβ + Xαγβ + Xβγα + Xγβα )

1
= 3!
(Xαβγ − Xαβγ + Xγαβ − Xγαβ + Xβγα − Xβγα ) = 0

• X[αβ]γ = 12 (Xαβγ − Xβαγ ) = Xαβγ

• X(αβ)γ = 21 (Xαβγ + Xβαγ ) = 0

• X[αβ](γ) = X[αβ]γ = Xαβγ

• X(αβ)[γ] = X(αβ)γ = 0

1
Exercise 2
Prove the following identities:

• X((α1 α2 ...αn )) = X(α1 α2 ...αn )

• X[[α1 α2 ...αn ]] = X[α1 α2 ...αn ]

• X(α1 ...[αl αm ]...αn ) =0

• X[α1 ...[αl αm ]...αn ] = X[α1 ...αl αm ...αn ]

Solution 2
• If Yα1 α2 ...αn is a totally symmetric tensor then we may write

Yαπ1 απ2 ...απn = Yα1 α2 ...αn , (1)

where πi denotes permutation over the index i. We may thus write the symmetric part of Y as
1 X
Y(α1 α2 ...αn ) = Yαπ1 απ2 ...απn
n!
= Yα1 α2 ...αn . (2)

Now, letting X(α1 α2 ...αn ) = Yα1 α2 ...αn we may write

Y(α1 α2 ...αn ) = X((α1 α2 ...αn ))

= X(α1 α2 ...αn ) , (3)

as required.

• Similarly to the previous question, if Yα1 α2 ...αn is a totally antisymmetric tensor then we may
write
(−1)π Yαπ1 απ2 ...απn = Yα1 α2 ...αn , (4)
We may thus write the antisymmetric part of Y as
1 X
Y[α1 α2 ...αn ] = (−1)π Yαπ1 απ2 ...απn
n!
= Yα1 α2 ...αn . (5)

Now, letting X[α1 α2 ...αn ] = Yα1 α2 ...αn we may write

Y[α1 α2 ...αn ] = X[[α1 α2 ...αn ]]

= X[α1 α2 ...αn ] , (6)

as required.

2
 • By symmetry (outer round brackets) we have
X(α1 ...[αl αm ]...αn ) = X(α1 ...[αm αl ]...αn ) , (7)
but by antisymmetry (inner square brackets) we have
X(α1 ...[αl αm ]...αn ) = −X(α1 ...[αm αl ]...αn ) , (8)
and thus we may conclude that
X(α1 ...[αl αm ]...αn ) = 0 , (9)
as required.
• First consider
1
Xα1 ...[αl αm ]...αn =(Xα1 ...αl αm ...αn − Xα1 ...αm αl ...αn ) . (10)
2
Now take the full antisymmetric part of this
1 X
(−1)π Xαπ1 ...απl απm ...απn − Xαπ1 ...απm απl ...απn


X[α1 ...[αl αm ]...αn ] =
2n!
1 X
= (−1)π Xαπ1 ...απl απm ...απn
n!
= X[α1 ...αl αm ...αn ] , (11)
where we have used the fact that Xαπ1 ...απm απl ...απn = −Xαπ1 ...απl απm ...απn , as required.

Exercise 3
Let F be a rank-2 antisymmetric tensor with components F µν . From F construct another rank-2 tensor
antisymmetric tensor ∗ F such that
1∗
F := αβµν Fαβ eµ ⊗ eν . (12)
2
∗
The tensor F is usually referred to as the dual of F. Show that the following is true
∗ ∗
( F) = −F . (13)

Solution 3
We may write the dual of F in contravariant index form as
∗ 1
F µν = αβµν Fαβ . (14)
2
Accordingly, the covariant form may be written, using the relation Fµν = gµγ gνδ F γδ , as
∗ 1
Fµν = gµγ gνδ αβγδ Fαβ
2
1 αβ
=  Fαβ
2 µν
1 αγ βδ
= g g γδµν Fαβ
2
1
= γδµν F γδ . (15)
2

3
We may now write

∗ ∗ 1 αβµν ∗
( F µν ) =  ( Fαβ )
2
1 αβµν
=  γδαβ F γδ
4
1 µν

γδ
= −2! δγδ F
4
1
− δγµ δδν − δδµ δγν F γδ


=
2
1 µν
= − (F − F νµ )
2
= −F µν . (16)

Thus we obtain
∗ ∗
( F) = −F , (17)
as required.

Exercise 4
Let V be a rank-3 tensor with components V αβγ and define

(∗ V )αβγ := Vµ µαβγ . (18)

Show that the following is true

1 ∗ αβγ ∗
V µ Vµ = − ( V ) ( V )αβγ . (19)
3!

Solution 4
In addition to equation (18), for fully covariant V we may write

(∗ V )αβγ = V ν ναβγ . (20)

From this we may immediately calculate the contraction as

(∗ V )αβγ (∗ V )αβγ = V ν Vµ µαβγ ναβγ

= V ν Vµ (−3! δνµ )
= −3! V µ Vµ , (21)

and hence we obtain

1 ∗ αβγ ∗
V µ Vµ = − ( V ) ( V )αβγ , (22)
3!
as required.

4
 General Relativity: Solutions to exercises in
Lecture VI

January 29, 2018

Exercise 1
Define the antisymmetric tensor F as Fµν = ∂µ Aν − ∂ν Aµ . Use the results from the previous exercises
to show that
Fµν := 2 ∂[µ Aν] . (1)
Show that such a definition implies that
Fαβ,ν + Fβν,α + Fνα,β = 0 . (2)

Solution 1
For the first part we may simply write
 
1
Fµν = 2 (∂µ Aν − ∂ν Aµ )
2
= 2 ∂[µ Aν] . (3)
For the second part of the question we must write out each of the three terms explicitly. For the first
term in equation (2) we obtain
Fαβ,ν = ∂ν (Fαβ )
= ∂ν (∂α Aβ − ∂β Aα )
= ∂ν ∂α Aβ − ∂ν ∂β Aα . (4)
For the second term
Fβν,α = ∂α (Fβν )
= ∂α (∂β Aν − ∂ν Aβ )
= ∂α ∂β Aν − ∂α ∂ν Aβ . (5)
Finally, for the third term
Fνα,β = ∂β (Fνα )
= ∂β (∂ν Aα − ∂α Aν )
= ∂β ∂ν Aα − ∂β ∂α Aν . (6)
Since ∂α ∂β F = ∂β ∂α F, summing equations (4)–(6) leads to cancellation of terms, giving the result in
equation (2), as required.

1
Exercise 2
Consider a vector V with components V µ relative to a co-ordinate basis, i.e.
V = V µ ∂µ = V µ eµ . (7)
Define an object given by the partial derivative of the components of V, i.e.
Uν µ := ∂ν V µ . (8)
Show that Uν µ is not a tensor. What are the implications of this result? What can be done to construct
a tensor out of measuring the derivative of a tensor?

Solution 2
From equation (8) we may write
Uν = ∂ν V
= ∂ν (V µ eµ )
= ∂ν V µ eµ + V µ ∂ν eµ . (9)
For the second term in the above expression we may think of it as a vector written in terms of some
basis vectors. Let us re-write this as ∂ν eµ = Γαµν eα . We may now write equation (9) as
Uν = ∂ν V µ eµ + V µ Γαµν eα
= ∂ν V µ eµ + V α Γµαν eµ (α ↔ µ in the second term)
= (∂ν V µ + V α Γµαν ) eµ
= (∇ν V µ ) eµ , (10)
and thus we may write
Uν µ = ∇ν V µ , (11)
where the ∇ν we have introduced is defined as the covariant derivative.

Consider the term ∂ν eµ = Γαµν eα . In flat (Minkowski) spacetime, in Cartesian co-ordinates, ∂ν eµ must
vanish as the eµ are all constant, and thus Γαµν must also be zero. However, in the same Minkowski
spacetime, transforming to (for example) spherical polar co-ordinates one would find the basis vector
components are not constant and are in fact functionally dependent on r and θ. As such, ∂ν eµ would be
non-zero in Minkowski spacetime. Since a tensor quantity is defined independently of any co-ordinate
system, the quantity Uν µ cannot be a tensor.

The partial derivative is not a good differential operator when spacetime is not Euclidean but by
construction the covariant derivative does define the components of a tensor.

Exercise 3
Consider a line element in three-dimensional space
ds2 = dr2 + r2 dθ2 + r2 sin2 θ dφ2 , (12)
with a co-ordinate basis {er , eθ , eφ }.

2
 • Construct the corresponding orthonormal basis {er̂ , eθ̂ , eφ̂ }

• Compute the structure coefficients Cr̂θθ̂ and Crθ

θ
. What is the difference between the two?

• Compute the structure coefficients Cr̂θφ̂ , Cr̂φφ̂ , Cθ̂θφ̂ and Cθ̂φφ̂ . Are there others that are non-zero?

Solution 3
Since our metric is diagonal we can immediately read of the orthonormal basis vector components as
1 1
er̂ = er , eθ̂ = eθ , eφ̂ = eφ , (13)
r r sin θ
from which it is straightforward to show that er̂ · er̂ = 1, eθ̂ · eθ̂ = 1 and eφ̂ · eφ̂ = 1. To convince
ourselves this is correct, consider the transformation between the co-ordinate basis and orthonormal
basis
dxî = Λî j dxj , (14)
where
1 0 0
 

Λî j = 0 r 0  . (15)
 

0 0 r sin θ
Using the co-ordinate transformation we may show that

dr̂ = dx1̂
= Λ1̂j dxj
= Λ1̂1 dx1
= dr . (16)

Similarly, one may show that

dθ̂ = r dθ , (17)
dφ̂ = r sin θ dφ . (18)

Now let us write the line element in terms of the orthonormal basis components and prove equivalence

ds2 = gr̂r̂ dr̂2 + gθ̂θ̂ dθ̂2 + gφ̂φ̂ dφ̂2

= dr̂2 + dθ̂2 + dφ̂2
= dr2 + r2 dθ2 + r2 sin2 θ dφ2 , (19)

hence the orthonormal basis vector components are correct.

For the next part of the question recall the definition of the Lie brackets of any two basis vectors,
which may be written in terms of the same basis as
γ
[eα , eβ ] = Cαβ eγ , (20)

3
 γ
where the components Cαβ are termed the structure coefficients. By definition a set of basis vectors
with all of its structure coefficients vanishing is a co-ordinate basis. We may now write
γ
Cαβ = [eα , eβ ]γ
= eαν ∂ν eβγ − eβν ∂ν eαγ . (21)

For the first structure coefficient, applying the above machinery we find

Cr̂θθ̂ = [er̂ , eθ̂ ]θ

= er̂ν ∂ν eθ̂θ − eθ̂ν ∂ν er̂θ
 
r 1
= er̂ ∂r − eθ̂θ ∂θ er̂θ
r
1
= − 2 . (22)
r
θ
As mentioned previously, Crθ = 0 since {ei } is a co-ordinate basis. For the final four requested structure
components we apply the same procedure for calculation. The results are as follows
h iθ
Cr̂θφ̂ = er̂ , eφ̂
= er̂ν ∂ν eφ̂θ − eφ̂ν ∂ν er̂θ

= 0, (23)

h iφ
Cr̂φφ̂ = er̂ , eφ̂
= er̂ν ∂ν eφ̂φ − eφ̂ν ∂ν er̂φ
= er̂r ∂r eφ̂φ − eφ̂ν ∂ν er̂φ
 
1
= ∂r
r sin θ
1
= − 2 , (24)
r sin θ

h iθ
Cθ̂θφ̂ = eθ̂ , eφ̂
= eθ̂ν ∂ν eφ̂θ − eφ̂ν ∂ν eθ̂θ
= eθ̂ν ∂ν eφ̂θ − eφ̂φ ∂φ eθ̂θ

1
= − ∂φ (r)
r sin θ
= 0, (25)

4
and
h iφ
Cθ̂φφ̂ = eθ̂ , eφ̂
= eθ̂ν ∂ν eφ̂φ − eφ̂ν ∂ν eθ̂φ
= eθ̂θ ∂θ eφ̂φ − eφ̂ν ∂ν eθ̂φ
 
1 1
= ∂θ
r r sin θ
cos θ
= − 2 2 . (26)
r sin θ
There are no other non-zero structure coefficients.

5
 General Relativity: Solutions to exercises in
Lecture VII

January 29, 2018

Exercise 1
Show that if g is the metric tensor, then its covariant derivative is zero, i.e.

∇λ gµν = 0 . (1)

Solution 1
By definition ∇Aµ is a vector. As such we may write

∇λ Aµ = gµν (∇λ Aν ) . (2)

We may also write

∇λ Aµ = ∇λ (gµν Aν )
= (∇λ gµν ) Aν + gµν (∇λ Aν ) . (3)

Using equation (2) we may rewrite the above expression as

∇λ Aµ = (∇λ gµν ) Aν + ∇λ Aµ , (4)

from which it immediately follows that

∇λ gµν = 0 , (5)
as required.

Exercise 2
Using the results of exercise 1, drive the following definition of the Christoffel symbols
1
Γαβγ = g αδ (∂γ gδβ + ∂β gδγ − ∂δ gβγ ) . (6)
2

1
Solution 2
Consider the following three expressions for the covariant derivative of the (covariant) metric tensor
∇λ gµν = gµν,λ − Γαλµ gαν − Γαλν gµα (= 0) , (7)
∇µ gνλ = gνλ,µ − Γαµν gαλ − Γαµλ gνα (= 0) , (8)
∇ν gλµ = gλµ,ν − Γανλ gαµ − Γανµ gλα (= 0) , (9)
where we have (evenly) permuted the covariant indices, as well as having made use of the result of
exercise 1, namely that ∇λ gµν = 0. We have also written partial derivatives as subscript commas for
the sake of brevity.
To prove the result, subtract the last two expressions from the first, i.e. (7) −[(8) + (9)]. This
yields
gµν,λ − gνλ,µ − gλµ,ν − Γαλµ gαν − Γαλν gµα + Γαµν gαλ + Γαµλ gνα + Γανλ gαµ + Γανµ gλα = 0 . (10)
By the torsion-free condition (i.e. Γαµν = Γανµ ) and symmetry of the metric tensor (i.e. gµν = gνµ ) the
red and blue terms cancel, yielding
gµν,λ − gνλ,µ − gλµ,ν + 2 Γαµν gαλ = 0 , (11)
which upon re-arranging gives
1
gαλ Γαµν = (gλµ,ν + gνλ,µ − gµν,λ ) . (12)
2
Multiplying both sides by g βλ gives
1
δαβ Γαµν = g βλ (gλµ,ν + gνλ,µ − gµν,λ ) , (13)
2
which immediately simplifies to
1
Γβµν = g βλ (gλµ,ν + gνλ,µ − gµν,λ ) . (14)
2
Finally, making the substitutions β → α, µ → β, ν → γ and λ → δ we obtain the result
1
Γαβγ = g αδ (gδβ,γ + gδγ,β − gβγ,δ ) , (15)
2
as required.

Exercise 3
Prove the following identities:
∂γ gαβ = Γαβγ + Γβαγ , (16)
gαµ ∂γ g µβ = −g µβ ∂γ gαµ , (17)
∂γ g αβ = −Γαµγ g µβ − Γβµγ g µα , (18)
(ln |g|),α = g µν gµν,α , (19)
1
∇µ Aµ 1/2 µ


= ∂ µ |g| A in a coordinate basis. (20)
|g|1/2

2
Solution 3
• For the first part consider the action of the covariant derivative on gαβ :

∇γ gαβ = gαβ,γ − Γλγα gλβ − Γλγβ gαλ = 0 . (21)

Rearranging yields

gαβ,γ = Γλγα gλβ + Γλγβ gαλ

= gλβ Γλαγ + gαλ Γλβγ
= Γβαγ + Γαβγ , (22)

as required.

• For the second part consider the following expression:

gαµ g µβ ,γ = δαβ ,γ = 0 ,



(23)

which may also be expanded as

gαµ g µβ = gαµ,γ g µβ + gαµ g µβ,γ

,γ
= 0. (24)

Rearranging the above expression, and writing partial derivatives explicitly, we obtain

gαµ ∂γ g µβ = −g µβ ∂γ gαµ , (25)

as required.

• For the third part, let us consider the action of the covariant derivative on the contravariant
metric tensor:
∇γ g αβ = g αβ,γ + Γαγµ g µβ + Γβγµ g αµ = 0 . (26)
Rearranging and making use of the symmetry conditions of the metric tensor and Christoffel
symbol yields
∂γ g αβ = −Γαµγ g µβ − Γβµγ g µα , (27)
as required.

• For the fourth part consider the metric tensor gαβ , which is a rank-2 tensor and specifically a
matrix. For matrices one may consider the Jacobi matrix formula:

∂gµν (xα )
 
∂ α α
det [gµν (x )] = Tr adj (gαβ (x )) , (28)
∂xα ∂xα

where the adjugate of a matrix may be written as

adj (gαβ ) = det (gαβ ) (gαβ )−1

= g αβ det (gαβ )
= g αβ |g| , (29)

3
 where we have omitted writing the dependence of the metric on co-ordinates for brevity, and
written the determinant of the metric tensor as |g|.

We may now rewrite the Jacobi identity in equation (28) in logarithmic form as
∂ 1 ∂
α
[ln det (gαβ )] = [det (gαβ )]
∂x det (gαβ ) ∂xα
1
Tr g αβ |g|gµν,α
 
=
|g|
= Tr g αβ gµν,α
 

= g µν gµν,α . (30)

This may be written more succinctly as

(ln |g|),α = g µν gµν,α , (31)

as required.
• For the fifth and final part, consider the action of the covariant derivative on Aµ :

∇µ Aµ = Aµ,µ + Γµµν Aν . (32)

Using the definition of the covariant derivative derived in exercise 2, we may write Γµµν as:
1
Γµµν = g µδ (gδµ,ν + gνδ,µ − gµν,δ ) . (33)
2
Whilst it is not immediately obvious, it can be shown that the last two terms in brackets in
equation (33) vanish. Consider the following:

g µδ (gνδ,µ − gµν,δ ) = g µδ ∂µ gνδ − g µδ ∂δ gµν

= ∂ δ gνδ − ∂ µ gµν
= ∂ δ gδν − ∂ µ gµν (gνδ = gδν )
= ∂ µ gµν − ∂ µ gµν (δ is a dummy index)
= 0. (34)

Consequently we may rewrite equation (33) as

1
Γµµν = g µδ gδµ,ν . (35)
2
In the fourth part of this exercise we showed that (ln |g|),ν = g µδ gµδ,ν . Using this we may write
1 µδ
Γµµν = g gδµ,ν
2
1
= (ln |g|),ν
2
= ln |g|1/2 ,ν




|g|1/2 ,ν
= (36)
|g|1/2

4
 Returning to equation (32) we may now re-write the expression as


µ µ
|g|1/2 ,ν ν
∇µ A = A ,µ + A
|g|1/2
1/2


|g| ,µ µ
= Aµ,µ + 1/2
A (relabel dummy index)
|g|
1 h 1/2 µ 1/2

µ
i
= |g| A ,µ + |g| ,µ
A
|g|1/2
1
|g|1/2 Aµ ,µ


= 1/2
|g|
1
∂µ |g|1/2 Aµ ,


≡ 1/2
(37)
|g|

as required.

Exercise 4
Optional: The covariant derivative of a contravariant vector U µ is

∇ν U µ := ∂ν U µ + Γµνλ U λ . (38)

Use this expression to obtain the covariant derivative of the covariant vector Uµ .

Solution 4
There are several ways one can go about proving this. Let us consider two such methods.

• Method 1

Consider the following:

∇ν (V µ Uµ ) = V µ;ν Uµ + V µ Uµ;ν
= V µ,ν Uµ + Γµνλ V λ Uµ + V µ Uµ;ν , (39)

where the subscript ;ν denotes the covariant differentiation with respect to xν and we have used
the definition of V µ;ν . Since the quantity V µ Uµ is a scalar we may also write

∇ν (V µ Uµ ) = ∂ν (V µ Uµ )
= V µ,ν Uµ + V µ Uµ,ν . (40)

Combining the above two equations yields

V µ,ν Uµ + V µ Uµ,ν = V µ,ν Uµ + Γµνλ V λ Uµ + V µ Uµ;ν , (41)

which simplifies to
V µ Uµ,ν = Γµνλ V λ Uµ + V µ Uµ;ν , (42)

5
 from which we may obtain
V µ Uµ;ν = V µ Uµ,ν − Γµνλ V λ Uµ . (43)
Now let us set V =µ
δβµ , which gives
δβµ Uµ;ν = δβµ Uµ,ν − Γµνλ δβλ Uµ , (44)
which simplifies to
Uβ;ν = Uβ,ν − Γµνβ Uµ , (45)
where upon setting µ ↔ α and then β → µ we obtain
Uµ;ν = Uµ,ν − Γαµν Uα , (46)
as required.

• Method 2

∇ν Uµ = ∇ν (gµα U α )
= (∇ν
 gµα

 ) + gµα ∇ν U α
gµα U α,ν + Γανλ U λ


=
= gµα U α,ν + gµα Γανλ U λ .


(47)
Now consider the expression
(gµα U α ),ν = Uµ,ν
= gµα,ν U α + gµα U α,ν


, (48)
which upon rearrangement yields
gµα U α,ν = Uµ,ν − gµα,ν U α .


(49)
Substituting equation (49) into equation (47) yields
∇ν Uµ = Uµ,ν − gµα,ν U α + gµα Γανλ U λ . (50)
From exercise 3, part 1, recall the identity
gµα,ν = gλα Γλνµ + gµλ Γλνα . (51)
This enables us to rewrite the last two terms in equation (50) as:
−gµα,ν U α + gµα Γανλ U λ = −gλα Γλνµ U α − gµλ Γλνα U α + gµα Γανλ U λ
= −gλα Γλνµ − gµλ Γλνα U α + gµλ Γλνα U α (α ↔ λ in last term)
= −gλα Γλνµ uα
= −Γλνµ uλ . (52)
We may now use the above expression to rewrite equation (50) as
∇ν Uµ = Uµ,ν − Γλνµ Uλ , (53)
which may be rewritten as
∇ν Uµ := ∂ν Uµ − Γλνµ Uλ , (54)
as required.

6
 General Relativity: Solutions to exercises in
Lecture VIII

January 29, 2018

Exercise 1
Consider the metric describing, in polar co-ordinates (r, θ), a Euclidean space

ds2 = dr2 + r2 θ2 . (1)

• Calculate the Christoffel symbols and geodesic curves associated with this space, which are given
by the geodesic equation
d2 xµ α
µ dx dx
β
+ Γ αβ =0. (2)
dλ2 dλ dλ
• Combine the two second-order differential equations describing the geodesic curves into a single
first-order differential equation for r = r(θ).

• What is the differential equation for a straight line in this space?

Solution 1
• First let us consider the components of the metric and their partial derivatives:
 
1 0
gµν = , (3)
0 r2
 
1 0
g µν = , (4)
0 r−2
 
0 0
gµν,r = , (5)
0 2r
gµν,θ = 0 . (6)

Next, recall the definition of the Christoffel symbols:

1
Γαβγ = g αδ (gδβ,γ + gγδ,β − gβγ,δ ) . (7)
2

1
 Since α can only be r or θ and the metric is diagonal, we may proceed as follows:
1 rr
Γrβγ = g (
grβ,γ
+  − gβγ,r )
gγr,β

2
1
= − g rr gβγ,r , (8)
2
1 θθ
Γθβγ = g (gθβ,γ + gγθ,β − gβγ,θ
)
2
1 θθ
= g gθβ,γ . (9)
2
It immediately follows that the only non-zero Christoffel symbols are given by:

Γrθθ = −r , (10)
1
Γθrθ = . (11)
r
Substituting these expression into the geodesic equation of motion (2) we obtain:

r̈ = rθ̇2 , (12)
2
θ̈ = − ṙθ̇ , (13)
r
where an overdot denotes differentiation with respect to the affine parameter, λ.

• For the second part of the question we may rewrite equation (13) as:
1 d  2 
r θ̇ = 0 , (14)
r2 dλ
which may be integrated to yield
k
θ̇ =
, (15)
r2
where k is a constant of integration. Next, starting from the line element and dividing both sides
by ds2 and taking s as affine we may write

ṙ2 + r2 θ̇2 = 1 . (16)

Using the chain rule we may write equation (16) as:

 2  2
dr dθ 2 dθ
+r =1. (17)
dθ dλ dλ

Substituting equation (15) into equation (17) we obtain:

 k2
r0 (θ)2 + r2

=1, (18)
r4
where a primed quantity denotes differentiation with respect to θ. This may be simplified to
yield r
0 r2
r (θ) = ±r −1 . (19)
k4

2
 • For the final part of the question, let us integrate equation (19), which describes geodesics in this
spacetime. Rearranging both sides of equation (19) gives:

dr
q = ±dθ . (20)
r2
r k4
−1

Integrating both sides of the above equation then yields:

r !
r 2
arctan − 1 = ± (θ + θ0 ) . (21)
k4

−1/2
Making use of the identity cos [arctan (f (x))] = [1 + f (x)2 ] we may take the cosine of both
sides of the above equation, yielding:
r
k4
= cos(θ + θ0 ) , (22)
r2
which may be finally written as
r cos(θ + θ0 ) = k 2 , (23)
which is precisely the equation of a straight line in polar co-ordinates. Thus the geodesic equations
of motion, which we derived in the first part of the question, are straight lines.

Exercise 2
Consider the metric describing the two-dimensional spacetime covered by co-ordinates (t, x) and with
metric
dx2 − dt2
ds2 = . (24)
t2
• Compute the Christoffel symbols.

• Compute the geodesic curves of this spacetime.

Solution 2
• As in question 1, let us first start by writing down the metric components and their partial
derivatives. First consider the components of the metric and their partial derivatives:
 −2 
−t 0
gµν = , (25)
0 t−2
 2 
µν −t 0
g = , (26)
0 t2
 −3 
2t 0
gµν,t = , (27)
0 −2t−3
gµν,x = 0 . (28)

3
 Next, recall the definition of the Christoffel symbols:
1
Γαβγ = g αδ (gδβ,γ + gγδ,β − gβγ,δ ) . (29)
2
We may now write:
1 tt
Γtβγ = g (gtβ,γ + gγt,β − gβγ,t ) , (30)
2
1 xx
Γxβγ = g (gxβ,γ + gγx,β − gβγ,x
) . (31)
2
For equation (30) only β = γ = t or x yields non-zero terms, and for equation (31) only β = x,
γ = t (or vice-versa) result in non-vanishing terms. It immediately follows that the only non-zero
Christoffel symbols are all identical and are given by:
1
Γttt = Γtxx = Γxtx = − . (32)
t
• For the second part of the question let us first write the geodesic equations of motion for this
spacetime. As in exercise 1, an overdot denotes differentiation with respect to the affine param-
eter. With the Christoffel symbol components we may write the geodesic equations of motion
as:
1 2
ṫ + ẋ2 ,


ẗ = (33)
t
2
ẍ = ṫẋ , (34)
t
and from the line element we may write
ẋ2 − ṫ2 = t2 . (35)
We may write equation (34) as
ẍ ṫ
=2 , (36)
ẋ t
which may be rewritten as
d d
ln t2 .


(ln ẋ) = (37)
dλ dλ
Integrating both sides of this equation then yields
ẋ = kt2 , (38)
where k is a constant of integration. Substituting equation (38) into equation (35) yields
k 2 t4 − ṫ2 = t2 , (39)
which may be solved for ṫ to yield √
ṫ = ±t k 2 t2 − 1 . (40)
We may now obtain a differential equation for x as a function of t by dividing equation (38) by
equation (40), yielding
kt
x0 (t) = ± √ , (41)
k 2 t2 − 1

4
 6

0
x

-2

-4

-6
-6 -4 -2 0 2 4 6
t
Figure 1: Hyperbolic geodesics as described by equation (43) for the case x0 = 0 and k = 1. Note
that the geodesics (orange curves) are asymptotic to the lightcone (dashed black line).

which may be integrated to give

1√ 2 2
x − x0 = ± k t −1
k
√
= ± t2 − k −2 , (42)

where x0 is a constant of integration. Finally, upon squaring both sides and rearranging we
obtain
t2 (x − x0 )2
− =1, (43)
(1/k)2 (1/k)2
which is the equation of a hyperbola. Thus the geodesics curves in this spacetime are described
by hyperbolas. This is illustrated in Figure 1.

5
Exercise 3
Given a scalar function φ ≡ φ (xµ ), prove the following identity in a co-ordinate basis:
1 √
φ := ∇µ ∇µ φ = √ −g g µν ∂ν φ .


∂µ (44)
−g

Solution 3
First, consider the following

∇µ ∇µ φ = (g µν φ;ν );µ
= (g µν φ,ν );µ , (45)

since φ is a scalar quantity. Recall the identity we derived in Problem Sheet 7, question 3, part 5:
1
Aµ;µ = 1/2 µ


|g| A ,µ
. (46)
|g|1/2

Using this identity and substituting Aµ = g µν φ,ν , we may now write

1
∇µ ∇µ φ = 1/2 µν


|g| g φ ,ν ,µ
, (47)
|g|1/2
√
which is the desired result, as required. Note that we use |g|1/2 and −g interchangeably.

Exercise 4
Optional: Derive the geodesic equation from the definition of a curve of extremal length.

Solution 4
The Euler-Lagrange equations of motion are derived by extremising the length of a curve. For a given
metric tensor gµν the Lagrangian may be written as

1
L = gµν ẋµ ẋν , (48)
2
where, as before, an overdot denotes differentiation with respect to the affine parameter. The Euler-
Lagrange equations are by definition written as
 
d ∂L ∂L
α
= . (49)
dλ ∂ ẋ ∂xα

Let us now derive each term. First we calculate the RHS of (49):

∂L 1
= gµν,α ẋµ ẋν . (50)
∂xα 2

6
For the LHS of equation (49) first consider:
 µ  ν
∂L 1 ∂ ẋ ν 1 µ ∂ ẋ
α
= gµν α
ẋ + gµν ẋ
∂ ẋ 2 ∂ ẋ 2 ∂ ẋα
1 1
= gµν δαµ ẋν + gµν δαµ ẋµ
2 2
1 1
= gαν ẋ + gµα ẋµ
ν
2 2
= gαµ ẋµ , (51)

where in the last step with have made use of the fact that µ and ν are dummy indices, as well as the
metric tensor being symmetric. Now we differentiate with respect to the affine parameter, yielding:
 
d ∂L d
α
= (gαµ ) ẋµ + gαµ ẍµ
dλ ∂ ẋ dλ
∂
= ẋβ β (gαµ ) ẋµ + gαµ ẍµ
∂x
= gαµ,β ẋβ ẋµ + gαµ ẍµ . (52)

Note that the dummy indices β and µ in the first term in equation (52) enable us to expand this term
as follows:
1
gαµ,β ẋβ ẋµ = (gαµ,β + gαβ,µ ) ẋβ ẋµ . (53)
2
Using equation (53) we may write the Euler-Lagrange equations as:
1 1
gαµ ẍµ + (gαµ,β + gαβ,µ ) ẋβ ẋµ = gµν,α ẋµ ẋν . (54)
2 2
Bringing all terms to the LHS and relabelling the dummy indices µ and ν in the RHS of equation (54)
as β and µ respectively, we obtain
1
gαµ ẍµ + (gαµ,β + gαβ,µ − gβµ,α ) ẋβ ẋµ = 0. (55)
2
Next, multiply both sides of this expression by g δα , yielding
1
ẍδ + g δα (gαµ,β + gαβ,µ − gβµ,α ) ẋβ ẋµ = 0 , (56)
2
where we have used the fact that g δα gαµ = δµδ . Recalling the definition of the Christoffel symbols this
expression may be written more succinctly as

ẍδ + Γδµβ ẋµ ẋβ = 0 . (57)

Let us now relabel the dummy indices as δ → α, µ → β and β → γ, enabling us to rewrite (57) in the
more familiar form
ẍα + Γαβγ ẋβ ẋγ = 0 , (58)
which is precisely the geodesic equation, as required.

7
 General Relativity: Solutions to exercises in
Lecture IX

January 29, 2018

Exercise 1
Consider a torus in a two-dimensional Euclidean space described by the spherical co-ordinate system
(θ, φ). The line element of the torus is then given by

ds2 = (b + a sin φ)2 dθ2 + a2 dφ2 , (1)

where b and a denote the torus radius and the radius of its section, respectively.

Compute the Christoffel symbol components and the non-vanishing components of the (Riemann)
curvature tensor. (Hint: remember that there is only one linearly independent component of the
Riemann tensor in a spacetime of dimension 2).

Solution 1
First let us consider the components of the metric and their partial derivatives:

(b + a sin φ)2 0
 
gµν = , (2)
0 a2
(b + a sin φ)−2 0
 
µν
g = , (3)
0 a−2
gµν,θ = 0 , (4)
 
2a (b + a sin φ) cos φ 0
gµν,φ = . (5)
0 0

Next, recall the definition of the Christoffel symbols:

1
Γαβγ = g αδ (gδβ,γ + gγδ,β − gβγ,δ ) . (6)
2

1
Since α can only be θ or φ and the metric is diagonal, we may proceed as follows:
1 θθ
Γθβγ = g (gθβ,γ + gγθ,β −  gβγ,θ
)
2
1 θθ
= g gθθ,φ (7)
2
a cos φ
= (8)
b + a sin φ
1 φφ
Γφβγ = g ( gφβ,γ
 +   − gβγ,φ )
gγφ,β
2
1
= − g φφ gθθ,φ (9)
2
(b + a sin φ) cos φ
= − . (10)
a
It immediately follows that the only non-zero Christoffel symbols are given by:
a cos φ
Γθθφ = , (11)
b + a sin φ
(b + a sin φ) cos φ
Γφθθ = − . (12)
a
Next, recall the definition of the Riemann tensor:
Rµναβ = Γµνβ,α + Γµρα Γρνβ − Γµνα,β − Γµρβ Γρνα , (13)
which may also be written more compactly as
Rµναβ = Γµνβ,α + Γµρα Γρνβ − (α ↔ β) ,


(14)
where (α ↔ β) denotes writing down the first term in brackets with α and β exchanged. Looking at
the first term in equation (13), we know that Γµνβ,α is non-zero only if α = φ (partial derivative is
non-zero). Next, we are free to choose (µ, ν, β) such that the Christoffel symbol is also non-zero. This
yields the choices (µ, ν, β) = (θ, θ, φ) or (φ, θ, θ). Let us take (µ, ν, β) = (θ, θ, φ), which yields
Rθθφφ = Γθθφ,φ + Γθρφ Γρθφ − Γθθφ,φ − Γθρφ Γρθφ
= Γθθφ,φ + Γθθφ Γθθφ + Γθφφ Γφθφ − Γθθφ,φ − Γθθφ Γθθφ − Γθφφ φ

Γ
 θφ

2
2
= Γθθφ,φ + Γθθφ − Γθθφ,φ − Γθθφ
= 0. (15)
Instead, let us now try α = θ and β = φ in equation (13). We obtain
Rµνθφ = Γµνφ,θ µ ρ µ µ ρ
 + Γρθ Γνφ − Γνθ,φ − Γρφ Γνθ . (16)
Next, let us ensure the partial derivative of the Christoffel symbol does not vanish by choosing µ = θ
and ν = φ, which yields:
Γρφφ
Rθφθφ = Γθρθ  − Γθ θ ρ
φθ,φ − Γρφ Γφθ

= −Γθφθ,φ − Γθθφ Γθφθ

a (a + b sin φ) a2 cos2 φ
= −
(b + a sin φ)2 (b + a sin φ)2
a sin φ
= . (17)
(b + a sin φ)

2
Consequently we may calculate the (only) non-vanishing component of the fully-covariant Riemann
curvature tensor as:

Rθφθφ = gθθ Rθφθφ

= a sin φ (b + a sin φ) . (18)

Exercise 2
Consider the two-dimensional spacetime with line element

ds2 = dv 2 − v 2 du2 . (19)

Compute the Christoffel symbols and the non-vanishing components of the curvature tensor.

Solution 2
As in question 1, let us first start by writing down the metric components and their partial derivatives.
First consider the components of the metric and their partial derivatives:
 
1 0
gµν = , (20)
0 −v 2
 
µν 1 0
g = , (21)
0 −v 2
 
0 0
gµν,v = , (22)
0 −2v
gµν,u = 0 . (23)

Next, recall the definition of the Christoffel symbols:

1
Γαβγ = g αδ (gδβ,γ + gγδ,β − gβγ,δ ) . (24)
2
We may calculate the Christoffel symbols as before:
1 vv
Γvβγ = g (gvβ,γ
+  − gβγ,v )
gγv,β

2
1
= − g vv gβγ,v
2
1 vv
= − g guu,v
2
1
= − (1)(−2v)
2
= v, (25)

3
 1 uu
Γuβγ = g (guβ,γ + gγu,β − 
gβγ,u
)
2
1 uu
= g (guβ,γ + gγu,β )
2
1 uu
= g guu,v
2 
1 1
= − 2 (−2v)
2 v
1
= . (26)
v
Thus we obtain the only non-zero Christoffel symbols as:

Γvuu = v , (27)
1
Γuuv = . (28)
v
Next recall the Riemann curvature tensor as defined in equation (13). Let us first make the first term
vanish by choosing α = u, yielding

Rµνuβ = 
Γµνβ,u
 + Γµ
ρ µ µ ρ
ρu Γνβ − Γνu,β − Γρβ Γνu . (29)

Now let us expand the sum over the dummy indices ρ:

Rµνuβ = Γµvu Γvνβ + Γµuu Γuνβ − Γµνu,β − Γµvβ Γvνu − Γµuβ Γuνu . (30)

Next, let us focus on ensuring the Γµνu,β term is non-vanishing, which requires us to set β = v:

Rµνuv = Γµvu Γvνv + Γµuu Γuνv − Γµνu,v − 

Γµvv
Γv − Γµ Γu .
νu uv νu (31)

From equation (31) let us first consider µ = u and ν = v:

Ruvuv = Γuvu
Γvvv
+ Γuuu
Γ u − Γu − Γu Γu
vv vu,v uv vu
u u 2
= −Γvu,v − (Γuv )
1 1
= 2− 2
v v
= 0. (32)

Let us next (and finally) consider the case where µ = v and ν = u:

Rvuuv = 
Γvvu
Γ v + Γv Γu − Γv
uv uu uv Γvuv
uu,v −  Γ u
uu
v u v
= Γuu Γuv − Γuu,v
 
1
= v −1
v
= 0. (33)

We may conclude that for this particular spacetime the Riemann tensor vanishes everywhere. As such,
we may say that our spacetime is flat.

4
Exercise 3
Consider a geodesic curve C and its tangent vector V. Compute the expression for the second convective
derivative of a vector field A along C, i.e. an explicit expression in component form of

∇V ∇V A . (34)

Recast the resulting expressions in terms of tensors that you have already encountered and interpret
the results.

Solution 3
First we must calculate the action of the convective derivative on A:

∇V A = V µ ∇µ Aα
= V µ ∂µ Aα + Γαµβ Aβ .


(35)

At this point it is important to remark that equation (35) is actually a rank-1 contravariant tensor
which we may call T α (all other indices are dummy indices). With this in mind, we may write the
second convective derivative of A as

∇V ∇V A = ∇V (∇V A)
= V ν ∇ν T α
= V ν ∂ν T α + Γανρ T ρ .


(36)

From here we must explicitly expand (36), yielding:

h
∇V ∇V A = V ν (∂ν V µ ) (∂µ Aα ) + V µ ∂ν ∂µ Aα + (∂ν V µ ) Γαµβ Aβ
+ V µ ∂ν Γαµβ Aβ + V µ Γαµβ ∂ν Aβ
i
+ Γανρ V µ ∂µ Aρ + Γανρ V µ Γρµβ Aβ . (37)

Let us now write the above expression as:

h
∇V ∇V A = V (∂ν V µ ) (∂µ Aα ) + V µ ∂ν ∂µ Aα + (∂ν V µ ) Γαµβ Aβ
ν

i
+ V µ Γαµβ ∂ν Aβ + V µ ∂ν Γαµβ Aβ + Γανρ V µ ∂µ Aρ
h i
+ V ν V µ ∂ν Γαµβ Aβ + Γανρ V µ Γρµβ Aβ
h
= V (∂ν V µ ) (∂µ Aα ) + V µ ∂ν ∂µ Aα + (∂ν V µ ) Γαµβ Aβ
ν

i
+ V µ Γαµβ ∂ν Aβ + V µ ∂ν Γαµβ Aβ + Γανρ V µ ∂µ Aρ
h i
ν µ α β α µ ρ β
+ V V ∂ν Γµβ A + Γνρ V Γµβ A , (38)

5
which may be further simplified as:
h
V (∂ν V µ ) ∂µ Aα + Γαµβ Aβ + V µ ∂ν ∂µ Aα
ν


∇V ∇V A =
i
+ V µ ∂ν Γαµβ Aβ + V µ Γαµβ ∂ν Aβ + V µ Γανρ ∂µ Aρ
h i
+ V ν V µ ∂ν Γαµβ Aβ + Γανρ V µ Γρµβ Aβ
h
= V (∂ν V µ ) (∇µ Aα ) + V µ ∂ν ∂µ Aα
ν

i
+ V µ ∂ν Γαµβ Aβ + 2V µ Γαµβ ∂ν Aβ (letting ρ → β and µ ↔ ν)
h i
ν µ α β α ρ β
+ V V ∂ν Γµβ A + Γνρ Γµβ A . (39)

At this point we remark that all terms in the first square brackets of equation (39) are unchanged
under interchange of µ and ν indices, whereas the two terms in the second pair of square brackets are
not. As such, if we calculate 2∇[ν ∇µ] Aα we will find that the first set of terms in the square brackets
will vanish. Doing this for the second convective derivative we derived we obtain:

2∇[V ∇V] A = V ν V µ ∂ν Γαµβ Aβ + Γαµρ Γρνβ Aβ − ∂µ Γανβ Aβ − Γαµρ Γρνβ Aβ

= V ν V µ Rαβνµ Aβ , (40)

thus we obtain an expression which depends on the Riemann curvature tensor. The expression ∇[V ∇V]
(or ∇[ν ∇µ] in component form) thus measure differences in a vector which is transported in different
directions around (say) a closed loop but which reach the same point. This equation is known as the
geodesic deviation equation.

6
 General Relativity: Solutions to exercises in
Lecture X

January 29, 2018

Exercise 1
Show that the second covariant derivatives of a scalar field commute, i.e. that
∇α ∇β φ = ∇β ∇α φ . (1)
Obtain the expressions for the following third derivatives: ∇α ∇(β ∇γ) φ and ∇[α ∇β] ∇γ φ. [Hint: remem-
ber that the covariant derivative of a scalar field is a vector.]

Solution 1
• For the first part of the question we consider the action of the covariant derivatives in order,
remembering that the covariant derivative of a scalar is simply the partial derivative acting on
that scalar. This yields:
∇α ∇β φ = (φ;β );α
= (φ,β );α
= φ,αβ − φ,δ Γδαβ . (2)
Since the Christoffel symbols are symmetric in their lower indices (torsion-free) and partial deriva-
tives commute, we may conclude that ∇α ∇β φ = ∇β ∇α φ, as required.
• For the second part of the question, let us first define the covariant vector Wγ ≡ ∇γ φ. Now
consider
∇α ∇β ∇γ φ = ∇α ∇β Wγ , (3)
and similarly
∇α ∇γ ∇β φ = ∇α ∇γ Wβ . (4)
Using the result of the first part of the question we may write
∇β Wγ = ∇γ Wβ . (5)
Employing the above we may now write
1
∇α ∇(β ∇γ) φ = ∇α (∇β Wγ + ∇γ Wβ )
2
= ∇α ∇β Wγ
= ∇α ∇β ∇γ φ . (6)

1
 • For the final part of the question let us directly expand the expression in question:
∇[α ∇β] ∇γ φ = ∇[α ∇β] Wγ
1
= (∇α ∇β Wγ − ∇β ∇α Wγ )
2
1 δ
= R Wδ
2 αβγ
1 δ
= R φ;γ
2 αβγ
1 δ
= R φ,γ . (7)
2 αβγ

Exercise 2
Prove that for any second-rank tensor, the covariant derivative commutes, i.e. that
∇α ∇β V αβ = ∇β ∇α V αβ . (8)

Solution 2
First recall the fact that ∇β V αβ is a rank-1 contravariant tensor, thus we may define W µ ≡ ∇β V µβ .
Next, let us write explicitly the expression for W µ as follows:
W µ = ∂β V µβ + Γµβδ V δβ + Γββδ V µδ . (9)
We may now write the covariant derivatives acting on V αβ as:
∇α ∇β V αβ = ∇α W α
= ∂α W α + Γααγ W γ . (10)
Thus we may write the LHS and RHS of equation (8) as:
 
∇α ∇β V αβ = ∂α ∂β V αβ + ∂α Γαβδ V δβ + ∂α Γββδ V αδ + Γααγ ∂β V γβ + Γααγ Γγβδ V δβ

+ Γααγ Γββδ V γδ , (11)

and
 
αβ αβ
Γααδ V δβ Γβαδ V αδ + Γββγ ∂α V αγ + Γββγ Γααδ V δγ


∇β ∇α V = ∂β ∂α V + ∂β + ∂β
+ Γββγ Γγαδ V αδ . (12)
Consider each term in eqns. (11) & (12), and let us refer to these equations as L and R respectively,
along with the indices 1–6 indicating terms 1–6 respectively in each expression. Under α ↔ β:
L1 = R1 ,
L2 = R3 ,
L3 = R2 ,
L4 = R4 ,
L5 = R6 ,
L6 = R5 .
We may thus conclude that ∇α ∇β V αβ = ∇β ∇α V αβ , as required.

2
Exercise 3
Optional : Find the matrix of the Lorentz transformations corresponding to a boost v x in the x-direction
followed by a boost v y in the y-direction. What happens if the order of the boosts is reversed?

Solution 3
Let us write the Lorentz boost matrices in the x- and y-directions respectively as:
 
γx γx vx 0 0
γx vx γx 0 0
Λx =   0
 , (13)
0 1 0
0 0 0 1

and
 
γy 0 γy vy 0
 0 1 0 0
Λy = 
γy vy
 . (14)
0 γy 0
0 0 0 1

For the first combination of boosts we obtain:

 
γx γy γx γy vx γy vy 0
 γx vx γx 0 0
Λy Λx =  γx γy vy γx γy vx vy γy
 . (15)
0
0 0 0 1

Similarly, for the reverse transformation we find:

 
γx γy γx vx γx γy vy 0
γx γy vx γx γx γy vx vy 0
Λx Λy =   γy vy
 . (16)
0 γy 0
0 0 0 1

Clearly Λy Λx 6= Λx Λy and so the transformations do not commute.

3
 General Relativity: Solutions to exercises in
Lecture XI

January 29, 2018

All of the following exercises are to be considered in a special-relativistic context and assuming
Cartesian co-ordinates where necessary.

Exercise 1
Within Special Relativity, consider a four-vector V with components:
√ √
V = 3 et + 2 ex . (1)

Determine if V is timelike, null or spacelike. Compute the angles between V and the unit vectors et
and ex .

Solution 1
First let us consider the inner-producut of V with itself:
 √ 2  √ 2 √ √
V·V = 3 et · et + 2 ex · ex + 2 2 3 et · ex
= −3 + 2 + 0
= −1 < 0 , (2)

therefore V is timelike. For the angles, first let us consider the t-component. From the inner-product
we may calculate the angle between V and et as:
V · et
cos θt =
|V · V|1/2 |et · et |1/2
√
= − 3 < −1 , (3)

which is not satisfied for any real θt . Similarly, for θx we find:

√
cos θx = 2 > 1 , (4)

which is also not satisfied for any real θx .

1
Exercise 2
A particle with rest mass m and four-momentum p = mv is analysed by an observer with
four-velocity u. Compute the following:
• The total energy of the particle
• The kinetic energy of the particle
p
• The magnitude of the spatial momentum p := pi pi
√
• The magnitude of the three-velocity v := vi v i

Solution 2
Let us work in the rest frame of the observer. In this frame:
uα = (1, 0) , (5)
uα = (−1, 0) , (6)
pα = (E, p) , (7)
pα = (−E, p) , (8)
where p is the three-momentum.
• The total energy may be obtained directly as
E = −p0 u0
= −pα uα . (9)

• Starting form the expression for the total energy of a particle as E 2 = p2 c2 + m2 c4 one obtains
the rest mass of the particle as:
m2 = E 2 − |p|2
= −p0 p0 − pi pi
= −pα pα , (10)
from which the kinetic energy of the particle may be directly derived as:
1
K.E. = m|v|2
2
1√
= −pα pα v 2
2
1√ p
= −pα pα −vi v i . (11)
2
• Starting again from the expression for the total energy of the particle, the magnitude of the
three-momentum may be calculated as:

1/2
p = E 2 − m2
1/2
= (pα uα )2 + pβ pβ

. (12)

2
 • Finally, the magnitude of the three-velocity may be calculated directly as:
p
v =
E
1
1/2
= E 2 − m2
E
1/2
m2

= 1− 2
E
" #1/2
p α pα
= 1+ . (13)
(pβ uβ )2

Exercise 3
Define the four-acceleration of a particle with four-velocity u as
duµ
aµ := , (14)
dτ
where τ is the proper time. Show that a · u = 0, i.e. that the acceleration is orthogonal to the
four-velocity. What does this mean in a frame co-moving with the particle?

Solution 3
Let us start with the following identity for a particle in General Relativity:

uµ uµ = −1 , (15)

from which it immediately follows that

d
(uµ uµ ) = 0 . (16)
dτ
Expanding the above expression yields:
d duµ µ
(uµ uµ ) = 2 u
dτ dτ
= 2 aµ uµ
= 0. (17)

We may thus conclude that a · u = 0, as required.

In a frame co-moving with the particle, uµ = (1, 0). In this frame aµ uµ = 0 implies that a0 = 0 by
necessity, but that the spatial components of the four-acceleration, ai , are arbitrary.

3
 General Relativity: Solutions to exercises in
Lecture XII

January 29, 2018

Exercise 1
Consider the stress-energy-momentum tensor of a perfect fluid

T µν = (e + p)uµ uν + p g µν , (1)

and its conservation equation

∇µ T µν = 0 . (2)
Show that the equations (2) lead to the Euler equations, i.e. to the equations of conservation of
momentum
(e + p)∇u u = − [∇p + (∇u p) u] . (3)
[Hint: use the projector h = g + u u]. Do equations (3) bear resemblance with the Newtonian Euler
equations?

Solution 1
First, let us write out the covariant derivative of the stress-energy-momentum tensor:

∇µ T µν = (e + p);µ uµ uν + (e + p) uµ;µ uν + uµ uν;µ + p;µ g µν .

(4)

Let us now us the projection tensor hαν = gαν + uα uν to project orthogonally to u, yielding
(( (( (
hαν ∇µ T µν = (e + p);µ( µ µ (( ν
 µ µ (( ν


[u( u(
α(+(u(
α u (uν u )] + (e + p) u ;µ
(
( ( uα
( +
( u
( (
α u(
;µ (uν u )
(
+ uµ gαν uν;µ + uα uµ uν uν;µ + p,µ δαµ + p,µ uµ uα



= (e + p)uµ uα;µ + p,α + p,µ uµ uα , (5)

where we have used the fact that uν uν = −1 and uν uν;µ = 0. Since hαν ∇µ T µν = 0 we may now write

(e + p)uµ uα;µ = − (p,α + p,µ uµ uα ) , (6)

which is equivalent to
(e + p)∇u u = − [∇p + (∇u p) u] , (7)
as required.

1
 In the Newtonian limit we may adopt the following approximations:
• p  e,
• e ≈ ρ0 ,
• v 2  1,
• g00 = −(1 + 2φ), |φ|  1, where φ is the Newtonian potential.
We may immediately let (e + p) → ρ0 , and through expanding the covariant derivative we obtain
ρ0 uµ uα,µ − Γβµα uβ uµ = −p,α − uα uµ p,µ .
 
(8)
We now take: (i) uβ uµ ∼ O(v 2 ) for β 6= µ and uβ uµ = −1 for β = µ and (ii) uα uµ p,µ ∼ O(v 2 ) → 0.
With these in mind we obtain h i
ρ0 uµ uα,µ + Γββα = −p,α . (9)
Recall from Problem Sheet 7, Exercise 3, part 5 we derived the following expression:
1
Γββα = g βδ gδβ,α . (10)
2
Since spacetime is now flat and g00 = −(1 + 2φ), so g 00 = −(1/g00 ) ≈ −1, since |φ|  1. This implies
Γββα = Γ00α
1
≈ (−1) [−(1 + 2φ)],α
2
= φ,α , (11)
hence we obtain
ρ0 (uµ uα,µ + φ,α ) = −p,α , (12)
which may be rewritten as
1
uµ uα,µ = − p,α − φ,α , (13)
ρ0
which is precisely the (Newtonian) incompressible Euler momentum equation with a constant and uni-
form density. This may be written more succinctly as follows. First define the specific thermodynamic
work, w, where w ≡ p/ρ0 and the gravitational acceleration g ≡ −∇φ. The material derivative is
D D ∂
defined in general relativity as Dτ = uµ ∇µ and in the Newtonian regime as Dt = ∂t + u · ∇. We may
now rewrite equation (13) as
Du
= −∇w + g . (14)
Dt

Exercise 2
The stress-energy-momentum tensor of a scalar field Φ is defined as
 
1 1 α
Tµν = ∂µ Φ ∂ν Φ − gµν ∂α Φ ∂ Φ . (15)
4π 2
Derive the expression for the conservation of energy and momentum (2) in this case. Interpret the
results.

2
Solution 2
We first write out the covariant derivative ∇µ Tµν = 0 as follows

1
∇µ (4πTµν ) = Φ,µ;µ Φ,ν + Φ,µ Φ,ν;µ − gµν (Φ,α Φ,α );µ . (16)
2
Defining the differential portion of the third term as ∆ = (Φ,α Φ,α );µ we may write

∆ = Φ,α;µ Φ,α + Φ,α Φ,α;µ

= g αβ gαγ Φ,γ;µ Φ,β + Φ,α Φ,α;µ
= δγβ Φ,γ;µ Φ,β + Φ,α Φ,α;µ
= Φ,β;µ Φ,β + Φ,α Φ,α;µ
= 2Φ,α Φ,α;µ . (17)

Equation (16) becomes

∇µ (4πTµν ) = Φ,µ;µ Φ,ν + Φ,µ Φ,ν;µ − Φ,α Φ,α;ν

= Φ,µ;µ Φ,ν + Φ,ν;µ − Φ,µ;ν Φ,µ

= Φ,µ;µ Φ,ν + g αµ Φ,ν;α − g βµ Φ,β;ν Φ,µ

= Φ,µ;µ Φ,ν + (Φ,ν;α − Φ,α;ν ) g αµ Φ,µ

= Φ,µ;µ Φ,ν + (Φ,ν;α − Φ,α;ν ) Φ,α . (18)

Since partial derivatives commute, and the covariant derivative of a scalar is simply the partial deriva-
tive, the second term in brackets vanishes and we may write

∇µ (4πTµν ) = Φ,µ;µ Φ,ν . (19)

Since we assume Φ,ν 6= 0 and Φ,µ;µ ≡ Φ;µ;µ we may write the conservation of energy and momentum as

Φ;µ;µ = 0 , (20)

which is equivalent to
Φ = 0 . (21)
Thus Φ satisfies the wave equation for a scalar field in vacuum.

Exercise 3
Show that the Einstein equations in vacuum reduce to

Rµν = 0 . (22)

3
Solution 3
Let us start from the definition of the Einstein Tensor
1
Gµν = Rµν − R gµν . (23)
2
In the presence of matter the Einstein field equations, Gµν = 8π Tµν may be written as

1
Rµν − R gµν = 8πTµν . (24)
2
Multiplying both sides of this equation by g µν yields
1
g µν Rµν − R g µν gµν = 8πg µν Tµν , (25)
2
which simplifies to
R = −4πT , (26)
where we have used the fact that g µν gµν = 4 and defined T ≡ g µν Tµν . Substituting R = −4πT back
into equation (24) yields, upon simplification
 
1
Rµν = 8π Tµν − gµν T . (27)
2

In vacuum Tµν = 0, which implies T = 0, and thus we obtain

Rµν = 0 , (28)

as required.

4
 General Relativity: Solutions to exercises in
Lecture XIII

January 29, 2018

Exercise 1
Consider the spherically symmetric static line element

ds2 = −A(r) dt2 + B(r) dr2 + C(r) dΩ2 , (1)

and compute the expressions for the non-zero Christoffel symbols. Use this result to compute the 00
covariant component of the Einstein equations in vacuum, i.e. Rµν = 0.

Solution 1
• Whilst one may calculate the Christoffel symbol components directly, we will derive them from
the Lagrangian for the metric. First let us write the Lagrangian as
1 
L= −A ṫ2 + B ṙ2 + C θ̇2 + C sin2 θ φ̇2 , (2)
2
where the dependence of A, B and C on r has been omitted for brevity and an overdot denotes
differentiation with respect to the affine parameter, λ. We now systematically derive the Euler-
Lagrange equations of motion for each of the four components of our metric. For the t component:

∂L
= 0, (3)
∂t
∂L
= −A ṫ , (4)
 ∂ ṫ
d ∂L
= −A0 ṙ ṫ − A ẗ , (5)
dλ ∂ ṫ

where primed quantities denote differentiation with respect to r. Thus from the Euler-Lagrange
equations we obtain the geodesic equation of motion for t as
 0
A
ẗ = − ṙ ṫ . (6)
A

1
This may be immediately compared to the geodesic equation of motion for t, yielding the non-zero
Christoffel symbol components as
1 A0
 
t t
Γ tr = Γ rt = . (7)
2 A
Next we consider the r component of the Euler-Lagrange equations, yielding
∂L 1 
= −A0 ṫ2 + B 0 ṙ2 + C 0 θ̇2 + C 0 sin2 θ φ̇2 , (8)
∂r 2
∂L
= B ṙ ṫ , (9)
 ∂ ṙ
d ∂L
= B 0 ṙ2 + B r̈ . (10)
dλ ∂ ṙ
We may now write the geodesic equation of motion for r as
1 A0 2 1 B 0 1 C0 1 C0
       
2 2
r̈ = − ṫ − ṙ + θ̇ + sin2 θ φ̇2 , (11)
2 B 2 B 2 B 2 B
from which we directly obtain the Christoffel symbols as
1 A0
 
r
Γ tt = , (12)
2 B
1 B0
 
r
Γ rr = , (13)
2 B
1 C0
 
r
Γ θθ = − , (14)
2 B
1 C0
 
r
Γ φφ = − sin2 θ . (15)
2 B
Now considering the θ component of the Euler-Lagrange equations we obtain
∂L
= C sin θ cos θ φ̇2 , (16)
∂θ
∂L
= C θ̇ ṫ , (17)
  ∂ θ̇
d ∂L
= C 0 ṙ θ̇ + C θ̈ . (18)
dλ ∂ θ̇
We may now write the geodesic equation of motion for θ as
 0
C
θ̈ = − ṙ θ̇ + sin θ cos θ φ̇2 , (19)
C
from which we directly obtain the Christoffel symbols as
1 C0
 
θ θ
Γ rθ = Γ θr = , (20)
2 C
Γθφφ = − sin θ cos θ . (21)

2
 Finally, we consider the φ component of the Euler-Lagrange equations, obtaining
∂L
= 0, (22)
∂φ
∂L
= C sin2 θ ṫ φ̇ , (23)
∂ φ̇
 
d ∂L
= C 0 sin2 θ ṙ φ̇ + C sin 2θ θ̇ φ̇ + C sin2 θ φ̈ . (24)
dλ ∂ φ̇

We may now write the geodesic equation of motion for φ as

 0
C
φ̈ = − ṙ θ̇ − 2 cotθ θ̇ φ̇ , (25)
C

from which we directly obtain the remaining non-zero Christoffel symbols as

1 C0
 
φ φ
Γ rφ = Γ φr = , (26)
2 C
Γφθφ = Γφφθ = cotθ . (27)

• For the second part of the question, recall the definition of the Riemann curvature tensor

Rαβγδ = Γαβδ,γ − Γαβγ,δ + Γµβδ Γαµγ − Γµβγ Γαµδ . (28)

The Ricci tensor is then defined as

Rβδ = Rαβαδ
= Γαβδ,α − Γαβα,δ + Γµβδ Γαµα − Γµβα Γαµδ . (29)

The covariant 00 component may now be written as

*0 µ µ
Γα
R00 = Γα00,α −  
 α α
0α,0 + Γ 00 Γ µα − Γ 0α Γ µ0
= Γα00,α + Γµ00 Γαµα − Γµ0α Γαµ0
= Γr00,r + Γr00 Γαrα − Γµ0α Γαµ0
= Γr00,r + Γr00 Γαrα − Γµ00 Γ0µ0 − Γµ0r Γrµ0
= Γr00,r + Γr00 Γαrα − Γr00 Γ0r0 − Γ00r Γr00
= Γr00,r + Γr00 Γαrα − 2Γr00 Γ0r0
 
φ
= Γ 00,r + Γ 00 Γ r0 + Γ rr + Γ rθ + Γ rφ − 2Γr00 Γ0r0
r r 0 r θ

 
= Γr00,r + Γr00 Γrrr + Γθrθ + Γφrφ − Γ0r0 . (30)

Substituting the values for the Christoffel symbol components into equation (30) we obtain, upon
simplification
1 A00 1 A0 C 0 A0 B 0
 
R00 = + 2 − − . (31)
2B 4B C A B

3
 For completeness, the remaining non-zero covariant components of the Ricci tensor are
1 A00 C 00 1 A0 A0 B 0 1 C 0 C 0 B0
   
R11 = − − + + + + , (32)
2 A C 4A A B 2C C B
1 C 00 1 C 0 B 0 A0
 
R22 = 1 − + − , (33)
2 B 4B B A
R33 = R22 sin2 θ . (34)

Exercise 2
Using the Lagrangian
2L = gαβ ẋα ẋβ , (35)
where an overdot corresponds to differentiation with respect to the proper time, show that the geodesic
equations
ẍα + Γαβγ ẋβ ẋγ = 0 , (36)
are equivalent to the Euler-Lagrange equations
 
∂L d ∂L
α
− =0. (37)
∂x dτ ∂ ẋα

Solution 2
Let us first calculate the first term in equation (37):
∂L 1
γ
= gαβ,γ ẋα ẋβ . (38)
∂x 2
Now we consider the bracketed second term:
∂L 1
gαβ δγα ẋβ + ẋα δγβ


=
∂ ẋγ 2
1
gγβ ẋβ + ẋα gαγ


=
2
= gαγ ẋα . (39)

Finally, we differentiate equation (39) with respect to proper time, yielding:

 
d ∂L d
γ
= (gαγ ) ẋα + gαγ ẍα
dτ ∂ ẋ dτ
= ẋδ gαγ,δ ẋα + gαγ ẍα
= gαγ,δ ẋδ ẋα + gαγ ẍα . (40)

Now we may write down the Euler-Lagrange equations, and solving for ẍα we obtain
1
gαγ ẍα = gαβ,γ ẋα ẋβ − gαγ,δ ẋδ ẋα
2
1 1
gαβ,γ ẋα ẋβ − gαγ,δ ẋδ ẋα + gδγ,α ẋδ ẋα ,


= (41)
2 2

4
where we have made use of the symmetry under interchange of δ ↔ α in the second term on the right
hand side. Since δ is a dummy index we may relabel it as δ → β, yielding
1
gαγ ẍα = (gαβ,γ − gαγ,β − gβγ,α ) ẋα ẋβ . (42)
2
Multiplying both sides by g γµ , using the identity gαγ g γµ = δαµ and bringing all terms to the left hand
side we obtain
1
ẍµ + g γµ (gαγ,β + gβγ,α − gαβ,γ ) ẋα ẋβ = 0 . (43)
2
It is straightforward to confirm that the term multiplying ẋα ẋβ is precisely Γµβα = Γµαβ and thus we
obtain
ẍµ + Γµαβ ẋα ẋβ = 0 , (44)
which is the geodesic equation of motion, as required.

Exercise 3
Optional : Using the Einstein-Hilbert action
√
Z
S= d4 x −g R , (45)

show that the application of a variational principle δS = 0 yields the Einstein field equations in vacuum,
i.e.
1
Rµν − gµν R = 0 . (46)
2

Solution 3
First we may write
√
Z
δS = 0 ⇐⇒ δ d4 x −g R = 0 . (47)
√
Now let us vary −g , yielding
√
δg
δ −g = − √ . (48)
2 −g
Now recall from Problem Sheet 7, Exercise 3, part 4, we proved the following result:

(ln |g|),α = g µν gµν,α . (49)

This implies that

g,α = g g µν gµν,α , (50)
and thus we may write δg as

δg = g g µν δgµν
= −g gµν δg µν . (51)

5
 √
We may now write δ ( −g ) as:
√
g gµν δg µν
δ −g = √
2 −g
1 (−g)
= − √ gµν δg µν
2 −g
1√
= − −g gµν δg µν . (52)
2
We must next consider the variation of the Ricci scalar R = g µν Rµν . We may write this as

δR = δg µν Rµν + g µν δRµν . (53)

Substituting equations (52) and (53) into equation (47) yields:

√  √
√
Z Z
4
d4 x δ

δ d x −g R = −g R + −g δR
√
Z  
4 1 µν µν µν
= d x −g − gµν δg R + (δg Rµν + g δRµν )
2
√
Z    
4 µν 1 µν
= d x −g δg Rµν − gµν R + g δRµν
2
√
Z
= d4 x −g (δg µν Gµν + g µν δRµν ) = 0 , (54)

where Gµν ≡ Rµν − 12 gµν R is the Einstein tensor. It is now clear that in order for us to obtain the

Einstein field equations in vacuum, the second term in brackets in equation (54) must vanish.
Let us now turn our attention to the variation of the Ricci tensor, δRµν . First, recall the definition
of the Riemann curvature tensor

Rµναβ = Γµνβ,α + Γµρα Γρνβ − Γµνα,β − Γµρβ Γρνα . (55)

Next, consider the variation of the Riemann curvature tensor:

δRµναβ = ∂α δΓµνβ + δΓµρα Γρνβ + Γµρα δΓρνβ

− ∂β (δΓµνα ) − δΓµρβ Γρνα − Γµρβ (δΓρνα ) .

(56)

This expression can be written much more succinctly in terms of covariant derivatives. The first and
fourth terms contains a partial derivative, so we consider the following:

∇α δΓµνβ = ∂α δΓµνβ + Γµαρ δΓρνβ − Γραν δΓµρβ − Γραβ δΓµνρ ,

(57)
µ µ µ ρ ρ µ ρ µ



∇β (δΓ να ) = ∂β (δΓ να ) + Γ βρ (δΓ να ) − Γ βν δΓ ρα − Γαβ δΓ νρ . (58)

It immediately follows that the difference between equations (57) and (58) enables equation (56) to be
written as
δRµναβ = ∇α δΓµνβ − ∇β (δΓµνα ) .


(59)
We may now calculate δRµν as follows:

δRνβ = δRαναβ
= ∇α δΓανβ − ∇β (δΓανα ) ,


(60)

6
and thus we obtain upon relabelling indices (β ↔ ν followed by β → µ):

δRµν = ∇α δΓαµν − ∇ν δΓαµα .

(61)

We may now write the second term in brackets in equation (54) as:

g µν δRµν = ∇α g µν δΓαµν − ∇ν g µν δΓαµα

= ∇α g µν δΓαµν − g µα δΓνµν ,


(62)

where we have let α ↔ ν in the second term. We may now write the second term in equation (54) as
√ √
Z Z
d x −g g δRµν = d4 x −g ∇α g µν δΓαµν − g µα δΓνµν .
4 µν


(63)

To proceed further, recall Problem Sheet 7, Exercise 3, part 5, where we proved the following identity:
1 √
Aα;α = √ −g Aα


,α
. (64)
−g

We may define Aα from equation (63) as

Aα = g µν δΓαµν − g µα δΓνµν , (65)

which enables us to rewrite equation (63) as

√ √
Z Z
4 µν
d4 x ∂α −g Aα


d x −g g δRµν =
= 0, (66)

since this is a surface integral, yielding a constant boundary term, and by Stokes’s Theorem vanishes.
We may finally write
√
Z
δS = d4 x −g δg µν Gµν = 0 , (67)

and so we may conclude that

1
Gµν = Rµν − gµν R = 0 , (68)
2
i.e. the Einstein field equations in vacuum, as required.

7
 General Relativity: Solutions to exercises in
Lecture XIV

January 29, 2018

Exercise 1
Using the Friedmann-Lemaı̂tre-Robertson-Walker (FLRW) metric
dr2
 
2 2 2 2 2 2 2


ds = −dt + a (t) + r dθ + sin θ dφ , (1)
1 − κr2
where κ = −1, 0, 1. Compute:
• the non-zero Christoffel symbols
• the non-zero components of the Ricci tensor
• the expression for the Ricci scalar

Solution 1
• The first part of the question asks us to calculate the non-zero Christoffel symbol components of
the FLRW metric. Let us begin by writing the Lagrangian for the FLRW metric:
a2
 
1 02 02 2 2 02 2 2 2 02
L= −t + r + a r θ + a r sin θ φ , (2)
2 1 − κr2
where primed quantities ( 0 ) denote differentiation with respect to the affine parameter, λ. We
have also written a ≡ a(t) for brevity. Next we employ the Euler-Lagrange equations, which may
be written as:  
∂L d ∂L
− =0. (3)
∂xα dλ ∂x0α
The Euler-Lagrange equations are equivalent to the geodesic equations of motion (see Lecture
XIII, exercise 2) and so we can read off the Christoffel symbol components directly. First, we
consider the t-component:
r02
 
∂L 2 02 2 2 02
= a ȧ + r θ + r sin θ φ , (4)
∂t 1 − κr2
∂L
0
= −t0 , (5)
  ∂t
d ∂L
= −t00 , (6)
dλ ∂t0

1
where an overdot ( ˙ ) denotes differentiation with respect to t. We immediately obtain:
a ȧ
t00 = − 02 2 02 2 2 02


r − a ȧ r θ + r sin θ φ . (7)
1 − κr2
We may now read off the Christoffel symbol components directly, obtaining:
aȧ
Γt rr = , (8)
1 − κr2
Γt θθ = a ȧ r2 , (9)
Γt φφ = a ȧ r2 sin2 θ . (10)

Next, we consider the r-component of the Euler-Lagrange equations:

 
∂L 2 κ 02 2 02
= ar + θ + sin θ φ , (11)
∂r (1 − κr2 )2
∂L a2
0
= 2
r0 , (12)
∂r 1 − κr
a2
 
d ∂L 2a ȧ 0 0 κr 02
= t r + r + r00 . (13)
dλ ∂r0 1 − κr2 (1 − κr2 )2 1 − κr2

We thus obtain the geodesic equation of motion as:

ȧ κr
r00 = −2 t0 r0 − r02 + r(1 − κr2 )θ02 + r2 sin2 θ(1 − κr2 φ02 ) , (14)
a 1 − κr2
from which the Christoffel symbols are immediately obtained as:
ȧ
Γrrt = Γrtr = , (15)
a
κr
Γrrr = , (16)
1 − κr2
Γrθθ = −r(1 − κr2 ) , (17)
Γrφφ = −r2 sin2 θ(1 − κr2 ) . (18)

We now consider the θ-component of the Euler-Lagrange equations:

∂L
= a2 r2 sin θ cos θ φ02 , (19)
∂θ
∂L
0
= a2 r 2 θ 0 , (20)
∂θ
 
d ∂L
= 2 a ȧ r2 t0 θ0 + 2a2 r r0 θ0 + a2 r2 θ00 . (21)
dλ ∂θ0

This gives the geodesic equation of motion for θ as:

ȧ 2
θ00 = −2 t0 θ0 − r0 θ0 + sin θ cos θ φ02 , (22)
a r

2
 from which the Christoffel symbols immediately follow as:
ȧ
Γθtθ = Γθθt = , (23)
a
1
Γθrθ = Γθθr = , (24)
r
Γθφφ = − sin θ cos θ . (25)

Finally, we consider the φ-component of the Euler-Lagrange equations:

∂L
= 0, (26)
∂φ
∂L
= a2 r2 sin2 θ φ0 θ0 , (27)
∂φ0
 
d ∂L
= 2a ȧ r2 sin2 θ t0 φ0 + 2a2 r sin2 θ r0 φ0 + 2a2 r2 sin θ cos θ θ0 φ0 + a2 r2 sin2 θ φ00 ,(28)
dλ ∂φ0

from which we obtain the geodesic equation of motion for φ as:

ȧ 2
φ00 = −2 t0 φ0 − r0 φ0 − 2cotθ θ0 φ0 . (29)
a r
Thus the final non-zero Christoffel symbols read:
ȧ
Γφtφ = Γφφt = , (30)
a
1
Γφrφ = Γφφr = , (31)
r
Γφθφ = Γφφθ = cotθ . (32)

• For the second part of the question, recall the definition of the Ricci tensor, which is the con-
traction of the Riemann curvature tensor over the first and third indices. This may be written
as:

Rµν = Rαµαν
= Γαµν,α + Γρµν Γαρα − Γαµα,ν − Γρµα Γαρν . (33)

The FLRW metric is spherically symmetric and possesses no off-diagonal terms, i.e. gµν = 0 if
µ 6= ν. Let us consider the four terms in the definition of the Ricci tensor for µ 6= ν.
In this case Γαµν,α = 0 ∀ µ 6= ν (see the Christoffel symbol components). The third term Γαµα,ν = 0
also, since Γαµα ∝ f (xµ ) and thus Γαµα,ν = 0 since µ 6= ν.
For a spherically symmetric metric Γαµν = 0 if α 6= µ 6= ν, which follows from the definition of
the Christoffel symbols. Using this, it may be shown that for all six independent combinations of
(µ, ν), with µ 6= ν, that Γρµν Γαρα − Γρµα Γαρν = 0. We may thus conclude that Rµν = 0 for µ 6= ν.
Let us consider the diagonal components of the Ricci tensor term-by-term. First the tt component:
0
* 0 ρ ρ
Γα
Rtt =  
tt,α + Γ >α

Γ α α
 tt ρα − Γ tα,t − Γ tα Γ ρt . (34)

3
For the third and fourth terms we obtain:

Γαtα,t = Γi ti,t
a ä − ȧ2
 
= 3 , (35)
a2
Γρtα Γαρt = Γρti Γi ρt

2
= Γi ti (only ρ = i gives a nonzero result)
ȧ2
= 3 2 , (36)
a
where the index i denotes spatial co-ordinates (r, θ, φ). Thus we immediately find:
ä
Rtt = −3 . (37)
a
We next consider the rr component of the Ricci tensor:

Rrr = Γαrr,α + Γρrr Γαρα − Γαrα,r − Γρrα Γαρr . (38)

For the first term:

Γαrr,α = Γt rr,t + Γrrr,r

ȧ2 + a ä + κ 2κ2 r2
= + . (39)
1 − κr2 (1 − κr2 )2
For the second term:

Γρrr Γαρα = Γt rr Γαtα

= Γt rr Γi ti + Γrrr Γi ri

3ȧ2 + 2κ κ2 r2
= + . (40)
1 − κr2 (1 − κr2 )2
For the third term:

Γαrα,r = Γi ri,r
2κ2 r2 κ 2
= 2 + − . (41)
(1 − κr2 ) 1 − κr2 r2
For the fourth term:

Γρrα Γαρr = Γρrt Γt ρr + Γρri Γi ρr

= Γrrt Γt rr + Γt ri Γi tr + Γi ri Γi ir

2
= 2Γrrt Γt rr + Γi ri
κ2 r2 2ȧ2 2
= 2 + 2
+ 2 . (42)
(1 − κr2 ) 1 − κr r
We thus obtain:
2κ + 2ȧ2 + a ä
Rrr = . (43)
1 − κr2

4
The θθ component of the Ricci tensor yields:
Rθθ = Γαθθ,α + Γρθθ Γαρα − Γαθα,θ − Γρθα Γαρθ . (44)
For the first term we obtain:
Γαθθ,α = Γt θθ,t + Γrθθ,r
= −1 + 3κr2 + r2 (ȧ2 + a ä) . (45)
For the second term we obtain:
Γρθθ Γαρα = Γt θθ Γαtα + Γrθθ Γαrα
= Γt θθ Γi ti + Γrθθ Γi ri
= 3ȧ2 r2 + κr2 − 2 . (46)
For the third term we obtain:
Γαθα,θ = Γφθφ,θ
= −cosec2 θ . (47)
For the fourth term we obtain:
Γρθα Γαρθ = Γρθt Γt ρθ + Γρθi Γi ρθ
= Γρθt Γt ρθ + Γt θi Γi tθ + Γj θi Γi jθ
= Γθθt Γt θθ + Γt θi Γi tθ + Γj θi Γi jθ
= Γθθt Γt θθ + Γt θθ Γθtθ + Γj θi Γi jθ
= 2Γθθt Γt θθ + Γj θi Γi jθ
= 2Γθθt Γt θθ + Γrθθ Γθrθ + Γθθr Γrθθ + Γφθφ Γφφθ
 2
= 2Γθθt Γt θθ + 2Γrθθ Γθrθ + Γφθφ
= 2ȧ2 r2 + 2κr2 − 2 + cot2 θ . (48)
We thus obtain:
Rθθ = r2 2κ + 2ȧ2 + a ä .


(49)
Finally, we consider the φφ component of the Ricci tensor:
*0 ρ
Rφφ = Γαφφ,α + Γρφφ Γαρα − 
Γα α
φα,φ − Γ φα Γ ρφ . (50)


The first term gives:

Γαφφ,α = Γi φφ,i
= − cos2 θ + ȧ2 + a ä + 3κ r2 sin2 θ .


(51)
The second term gives:
Γρφφ Γαρα = Γt φφ Γαtα + Γrφφ Γαrα + Γθφφ Γαθα
= Γt φφ Γi ti + Γrφφ Γi ri + Γθφφ Γφθφ
= 3ȧ2 + κ r2 sin2 θ − 1 − sin2 θ .


(52)

5
 The fourth term gives:
Γρφα Γαρφ = Γρφt Γt ρφ + Γρφr Γrρφ + Γρφθ Γθρφ + Γρφφ Γφρφ
 
= Γφφt Γt φφ + Γφφr Γrφφ + Γφφθ Γθφφ + Γt φφ Γφtφ + Γrφφ Γφrφ + Γθφφ Γφθφ
 
= 2 Γφφt Γt φφ + Γφφr Γrφφ + Γφφθ Γθφφ
= 2ȧ2 + 2κ r2 sin2 θ − 2 .


(53)
Thus we obtain:
Rφφ = r2 sin2 θ 2κ + 2ȧ2 + a ä .


(54)
Defining A ≡ 2κ + 2ȧ2 + a ä we may write the non-zero Ricci tensor components more succinctly
as:
ä
Rtt = −3 , (55)
a
A
Rii = gii 2 . (56)
a
• For the third part of the question we are asked to calculate the Ricci scalar. This follows
straightforwardly from equations (55)–(56):
R = g µν Rµν
= g tt Rtt + g ii Rii
ä A
= 3 + g ii gii 2
a a
ä A
= 3 +3 2
a a
6
= 2 κ + ȧ2 + a ä .


(57)
a

Exercise 2
Exploiting the results of the previous exercise, use the Einstein equations for the FLRW metric to
derive the Friedmann equations. For simplicity set Λ = 0.

Solution 2
Since Exercise 3 requires Λ > 0 we will also assume this in the following solution. The Einstein field
equations may be written as:
1
Rµν − gµν R + Λgµν = 8πTµν , (58)
2
where
Tµν = (e + p)uµ uν + pgµν . (59)
We are in the comoving frame of the fluid, where uα = (1, 0) and uα = (−1, 0), and therefore:
Ttt = e , (60)
Tii = pgii . (61)

6
Considering the tt component of the Einstein field equations we obtain:
1
Rtt − gtt R + Λgtt = 8πTtt
2
ä 3
−3 + 2 κ + ȧ2 + a ä − Λ = 8πe


=⇒
a a
 2
ȧ 1 κ
=⇒ = (8πe + Λ) − 2 , (62)
a 3 a

which is the first Friedmann equation. We now consider the spatial component of the Einstein field
equations:
1
Rii − gii R + Λgii = 8πTii
2  
A 1
=⇒ gii 2 + gii − R + Λ − 8πp = 0
a 2
ä κ ȧ2
=⇒ −2 − 2 − 2 + Λ − 8πp = 0 (use equation (62))
a a a
ä 4π Λ
=⇒ = − (e + 3p) + , (63)
a 3 3
which is the second Friedmann equation.

Exercise 3
Optional : Consider the case of an equation of state where p = −e and Λ > 0. Derive the evolution
equation for the scale factor. What type of universe is this?

Solution 3
In the comoving frame the stress-energy-momentum tensor of a perfect fluid may be written as:

T µν = (e + p)uµ uν + pδνµ
= diag(−e, p, p, p) . (64)

The spatial component of the conservation equation (∇µ T µν = 0) trivially vanishes, implying uniform
pressure. However, it is straightforward to show that the time component yields the fluid conservation
equation:
ȧ
ė + 3 (e + p) = 0 . (65)
a
For the given equation of state, this implies that ė = 0 and hence e = e0 = constant. Substituting this
into the second Friedmann equation we obtain:
ä 8π Λ
= e0 +
a 3 3
=⇒ ä = Ca , (66)

7
where C ≡ (Λ + 8πe0 )/3. Now, since Λ > 0 and e0 ≥ 0, then this implies that C > 0. Integrating
equation (66) directly yields: √ √
a(t) = c1 e C t + c2 e− C t , (67)
where the integration constants c1 and c2 may be calculated from this equation and the first Friedmann
equation as:

c 1 + c 2 = a0 , (68)
Ca20 − κ = ȧ0 , (69)

where a0 and ȧ0 are the initial values of a(t) and ȧ(t) at t = 0. Equation (67) has a minimum at:
r
1 c2
tmin = √ ln , (70)
C c1

and since t ≥ 0 for the universe we know that if c2 > c1 then there exists a minimum value of t > 0.
We assume c1 and c2 are both positive. Thus, if c1 > c2 the universe expands exponentially from t = 0.
If, however, c2 > c1 then the universe contracts between t = 0 and tmin , before expanding exponentially
thereafter.

8
 General Relativity: Solutions to exercises in
Lecture XV

February 5, 2018

Exercise 1
The simplest solution to the linearised Einstein equations is a plane wave of the form:
α
h̄µν = R Aµν eiκα x

, (1)

where R denotes the real part, A is the “amplitude” tensor and κ is a null four-vector which satisfies
κα κα = 0. In such a solution, the plane wave donated by equation (1) travels in the spatial direction
~k = (κx , κy , κz )/κ0 , with frequency ω ≡ κ0 = (κj κj )1/2 . Determine the conditions such that the
amplitude tensor A has only two linearly independent components, corresponding to the two states of
polarisation of the gravitational waves.

Solution 1
Aµν has at most 10 independent components. The solution to the linearised field equations  h̄µν = 0
is given by equation (1). Inserting this solution back into the linearised field equations yields:

 h̄µν = η αβ ∂α ∂β h̄µν
= η αβ ∂α iκβ h̄µν

= −η αβ κα κβ h̄µν
= − (κα κα ) h̄µν = 0 , (2)

and thus we obtain the condition κα κα = 0. Defining κα = (ω, κ0~k) and κα = (−ω, κ0~k) we then
1/2
immediately obtain ω ≡ κ0 = (κj κj ) . Next, imposing the Lorentz (also know as de Donder) gauge
h̄µν,µ = 0 we obtain:
α
α
∂µ Aµν eiκα x = iAµν κµ eiκα x
α
= i (κµ Aµν ) eiκα x = 0 , (3)

from which we obtain the condition:

κµ Aµν = 0 , (4)
i.e. the wave vector is orthogonal to Aµν . This condition constitutes a set of 4 algebraic equations and
thus the number of degrees of freedom of Aµν are reduced from 10 to 6. So far, whilst we have imposed

1
the Lorentz gauge, we still have some remaining co-ordinate freedom. If we consider a co-ordinate
transform of the form:
x0µ = xµ + ξ µ , (5)
then  x0µ = 0 if  ξ µ = 0, which is also a wave equation. This has the solution
α
ξµ = Bµ eiκα x , (6)
where κα is the wave vector and Bµ are constant coefficients. The remaining co-ordinate freedom allows
us to transform from Aµν → A0µν such that:
• A0 0ν = 0 (wave amplitude transverse to its propagation direction)
• A0µ µ = 0 (traceless)
The choice of gauge sets uµ A0 µν = 0 for constant and timelike uµ . The choice of Lorentz frame fixes
uµ to point along the time axis. Starting from the metric perturbation:
gµν = ηµν + hµν + O [hµν ]2 ,


(7)
in the new co-ordinate system we obtain:
0
gµν = ηµν + hµν − ξµ,ν − ξν,µ , (8)
from which we immediately relate the metric perturbations between the two co-ordinate systems as:
h0µν = hµν − ξµ,ν − ξν,µ . (9)
Contracting both sides of this equation with η µν gives:
h0 = h − ξ µµ − ξ νν
= h − 2 ξ αα . (10)
0
Now consider the trace-reversed part of h0µν , namely hµν , which may be simplified as follows:
1
h0µν = hµν − ηµν h0
2
1
= hµν − ξµ,ν − ξν,µ − ηµν (h − 2 ξ αα )
2
1
= hµν − ηµν h − ξµ,ν − ξν,µ + ηµν ξ αα
2
= h̄µν − ξµ,ν − ξν,µ + ηµν ξ αα . (11)
xα xα
Substituting our solution h̄µν = Aµν eiκα for the field equations and our solution ξµ = Bµ eiκα for
the transformation between frames into equation (11) yields:
A0µν = Aµν − iκν Bµ − iκµ Bν + iηµν κα B α . (12)
We may now use the above equation to determine the components of B in terms of Aµν . Imposing the
traceless condition implies contracting equation (12) with η µν , yielding:
A0µ µ = Aµµ − iκν B ν − iκµ B µ + 4iκα B α
= Aµµ + 2iκµ B µ = 0 , (13)
which gives the condition:
i
κµ B µ = Aµµ . (14)
2
We next impose the transverse condition. Let us consider the temporal and spatial parts separately.

2
 • For ν = 0 we obtain:
A000 = A00 − 2iκ0 B0 − iκα B α = 0 . (15)
However, we have previously derived equation (14) which upon substitution and simplification
gives the temporal component of B as:
 
i 1 α
B0 = − A00 + A α . (16)
2κ0 2

• For ν = j we obtain:

A00j = A0j − iκ0 Bj − iκj B0

 
κj 1 α
= A0j − iκ0 Bj − A00 + A α . (17)
2κ0 2

From this we can solve for Bj , which upon simplification yields:

  
i 1 α
Bj = 2 −2κ0 A0,j + κj A00 + A α . (18)
2κ0 2

We now have the four constant coefficients for Bµ . We know that equations (17) and (18) satisfy
the transverse condition and therefore A0µν ↔ Aµν . The traceless condition implies 1 condition on the
number of independent components of the amplitude tensor. For ν = 0 the transverse condition implies
κµ Aµν = 0 which is redundant as we have already considered this. As such A00j (from the transverse
condition) yields 3 conditions for A0µν (and therefore Aµν ). Thus we conclude that the TT gauge gives
us a further 4 constraints and so the number of linearly independent components of Aµν is reduced
from 6 to 2. Therefore the gravitational wave only has 2 independent states of polarisation, as required.

As an example, consider a gravitational wave travelling in the positive z-direction, where k µ =

(ω, 0, 0, κz ) ≡ (ω, 0, 0, ω). For such a null vector the conditions k µ Aµν = 0 and A0ν = 0 imply that
A3ν = 0 also. Thus the only non-zero components are A11 , A12 , A21 and A22 . However, using the
traceless condition Aµµ = 0 we also obtain A22 = −A11 . Finally, we know that by symmetry A12 = A21
and thus we may write the amplitude tensor for such a gravitational wave as:
 
0 0 0 0
0 A11 A12 0
Aµν = 0 A12 −A11 0 ,
 (19)
0 0 0 0

which only possesses 2 linearly independent components. It is important to note the that in the TT
gauge we have h̄TµνT = hTµνT .

Exercise 2
The gauge satisfying the requirement of the first exercise is also refereed to as the TT (or transverse-
traceless) gauge. Compute the non-zero components of the Riemann tensor in this gauge.

3
Solution 2
Recall from Exercise 1 that we defined the transformation x0µ = xµ + ξ µ . From the consideration
of nearby geodesics/particles separated by an infinitesimal distance ξ µ (where ||ξ µ ||  1) one may
calculate the geodesic deviation equation:

D2 ξ µ
= Rµναβ uν uα ξ β . (20)
Dτ 2
Let us calculate the RHS of equation (20) to first order in hµν . Assuming neighbouring geodesics/particles
vary slowly, we may express the four-velocity as a unit vector in the time direction plus corrections of
O (hµν ) and higher. Since the Riemann tensor is already first order in hµν , corrections to uν can be
ignored and we may set uν = (1, 0, 0, 0). With this the non-zero components of the geodesic deviation
equation are found as
D2 ξ µ
2
= Rµ00β ξ β . (21)
Dτ
Since Rµ00β 6= 0 this implies that Rµ00β 6= 0 also. From the symmetries of the Riemann curvature
tensor we obtain:
Rµ00β = R0µ0β = −Rµ0β0 = −R0µ0β , (22)
which are the only non-zero components. Thus there is only one independent component to the
Riemann curvature tensor. We may write the expression for the Riemann curvature tensor as:
1
Rµ00β = (hµβ,00 + h00,µβ − hµ0,0β − hβ0,0µ )TT , (23)
2
where the superscript TT denotes evaluation of that quantity in the TT gauge. However, we know
that hµ0 = 0 in the TT gauge and so the last three terms in equation (23) vanish, yielding:

1
Rµ00β = h̄TT . (24)
2 µβ,00
From the plane wave solution given in equation (1) we obtain:

h̄TT TT
µβ,00 = −κ0 κ0 h̄µβ
= −ω 2 h̄TT
µβ (25)

and therefore:
1
Rµ00β = − ω 2 h̄TTµβ . (26)
2
−iωt
Finally, in the TT gauge, and given our solution for h̄TT TT
µβ,00 , we may assume h̄µβ ∝ e and therefore:

1
Rµ00β ∼ − ω 2 e−iωt . (27)
2