General Relativity and Cosmology, SS 2014

(from notes of a lecture by Prof. Dr. Herbert Petry)

Bernard Metsch
Helmholtz-Institut für Strahlen- und Kernphysik
der Rheinischen Friedrich-Wilhelms-Universität Bonn
Nußallee 14-16
D-53115 Bonn
metsch@hiskp.uni-bonn.de

May 4, 2015
Contents

1 Special Relativity and Electrodynamics 1

1.1 Basic equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1.1 Poincaré invariance . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Special Lorentz transformations . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Electromagnetic waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

2 Action 5
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.1.1 A two-dimensional analogy . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2 The action of electromagnetism . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2.1 Symmetries of the new action . . . . . . . . . . . . . . . . . . . . . . . 8
2.2.2 Equations of motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.3 The full action . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.3.1 Einstein’s field equations . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.4 Invariance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

3 Solutions of the equations 17

3.1 Distinguished coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.1.1 Comoving coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.1.2 Riemannian coordinates . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.1.3 Consequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.2 Solution of the Einstein equation for a massive body . . . . . . . . . . . . . 25
3.2.1 Christoffel symbols, curvature tensor and scalar . . . . . . . . . . . 25
3.2.2 Differential equations . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.2.3 Schwarzschild exterior solution and particle trajectories . . . . . . 32
3.2.4 Light rays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.2.5 Black holes and other features . . . . . . . . . . . . . . . . . . . . . . 36
3.3 Gravitational collapse . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.4 Gravitational waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.4.1 Gravitational waves as approximate solutions of the Einstein equations 46
3.5 Kerr-solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

4 Cosmology 53
4.1 Space-time metric . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.2 The behaviour of electromagnetic waves . . . . . . . . . . . . . . . . . . . . . 54
4.3 Cosmological evolution, the FRW universe . . . . . . . . . . . . . . . . . . . 59

5 Extensions 63
5.1 Kaluza-Klein theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
5.1.1 Geodesics in M 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
5.2 Tetrads . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
5.2.1 Tetrads and the Dirac equation . . . . . . . . . . . . . . . . . . . . . 68

i
ii CONTENTS

A Tensor calculus I
A.1 Tensors and tensor fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I
A.1.1 Elementary operations on tensorfields . . . . . . . . . . . . . . . . . . II
A.1.2 Transformation rule for tensor fields . . . . . . . . . . . . . . . . . . . III
A.1.3 Derivatives of tensor fields . . . . . . . . . . . . . . . . . . . . . . . . . IV
A.2 The curvature and related tensor fields . . . . . . . . . . . . . . . . . . . . . . VIII

B Note on manifolds IX
B.1 Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . IX
B.2 Construction of a new manifold . . . . . . . . . . . . . . . . . . . . . . . . . . IX
B.3 relation between M c and M . . . . . . . . . . . . . . . . . . . . . . . . . . . . X
Chapter 1

Special Relativity and

Electrodynamics

1.1 Basic equations

The relativistic equation of motion for a particle of mass m and charge q in an electromag-
~ x, t), B(~
netic field E(~ ~ x, t) is given by the Einstein-Lorentz equation:
 i
d ~ 1h ~
Einstein-
p~ = q E + ~v × B , (1.1.1) Lorentz
dt c
equation
where according to Einsteins theory of special relativity the momentum is related to the
velocity as
m ~v
p~(~v ) = q . (1.1.2)
2
1 − |~vc2|
Accordingly, with ~z(t) the trajectory as a function of time and ~z˙ (t) = dt d
~z(t) the equation
of motion reads explicitly showing all dependencies:
 i
d m ~z˙ (t) ~ 1 h˙ ~
q = q E(~z(t), t) + ~z(t) × B(~z(t), t) .
dt ˙ 2 c
1 − |~z(t)|
c 2

Here [~x × ~y ] ∈ R 3 denotes the vector product of ~x, ~y ∈ R 3 , i.e. with respect to a standard
P3 i
(right handed) orthonormal Basis ~e1 , ~e2 , ~e3 with (~ei · ~ej ) = δi j and ~x = i=1 x ~ei , ~y =
P3 i
i=1 y ~
e i , the equations
3 X
X 3
k
[~x × ~y ] = εijk xi y j , k = 1, 2, 3
i=1 j=1

hold, where ε123 = 1 and εσ(1)σ(2)σ(3) = Signum(σ) ε123 for a permutation σ ∈ S3 and ε
vanishes otherwise.
~ x, t) and the magnetic field
On the other hand the equations determining the electric field E(~
~
B(~x, t) for a given charge density ρ(~x, t) and a given current density ~j(~x, t) satisfying the
continuity equation continu-
∂ρ ity
div ~j + = 0, (1.1.3)
∂t equation
are the inhomogeneous Maxwell equations
~ Maxwell
~ − 1 ∂ E = 4π ~j ,
rot B ~ = 4π ρ
div E (1.1.4) equations
c ∂t c
as well as the homogeneous Maxwell equations
~
~ + 1 ∂B = 0 ,
rot E ~ = 0.
div B (1.1.5)
c ∂t

1
 2 CHAPTER 1. SPECIAL RELATIVITY AND ELECTRODYNAMICS

1.1.1 Poincaré invariance

The equations of motion and the Maxwell equations are Poincaré invariant in the fol-
lowing sense:
Put x0 = c t and let x = x0 e0 +x1 e1 +x2 e2 +x3 e3 =: xµ eµ (summation convention!) denote
a vector x ∈ R 4 with coordinates xµ with respect to an orthonormal basis {e0 , e1 , e2 , e3 } (see
field below) and let the (antisymmetric) field strength tensor F with components Fµν = −Fνµ
strength with respect to the same basis be declared as
tensor
3
X
F0i = −Fi0 = E i , Fik = − εikl B l , F00 = 0 , i, k = 1, 2, 3 . (1.1.6)
l=1

Poincaré Furthermore a Poincaré transformation p : R 4 → R 4 is declared as

transfor-
mation p(x) = a + Λ x , p(x)µ = aµ + Λµν xν , µ = 0, 1, 2, 3 , (1.1.7)

space-time for all x ∈ R 4 , where a ∈ R 4 is a space-time translation and where Λ : R 4 → R 4 , the

translation Lorentz transformation , is a linear mapping with the property
Lorentz
transfor- η(Λ x, Λ y) = η(x, y) , ∀ x , y ∈ R4 , (1.1.8)
mation

i.e. Λ is an isometry of the Minkowksi metric η : R 4 × R 4 → R defined by

Minkowksi
metric η(x, y) = ηµν xµ y ν , (1.1.9)

where
ηµν := η(eµ , eν ) = εµ δµν ε0 = 1 , ε1 = ε2 = ε3 = −1 ,
defining an orthonormal basis {e0 , e1 , e2 , e3 } of R 4 .
Then the homogeneous Maxwell equations can be written in the form
∂ ∂ ∂
(dF )κλµ := κ
Fλµ − λ
Fκµ + Fκλ = 0 , κ 6= λ 6= µ . (1.1.10)
∂x ∂x ∂xµ
charge- Likewise upon defining the 4-vector charge-current density with j 0 = c ρ component-wise
current by
density
J = j 0 e0 + j 1 e1 + j 2 e 2 + j 3 e 3 (1.1.11)
one can write the inhomogeneous Maxwell equations as
∂ 4π ν
(div F )ν := F µν = J , (1.1.12)
∂xµ c
where
F µ ν := η µκ Fκν , F µν := η νλ F µ λ = η µκ η νλ Fκλ
and where η µλ denotes the matrix elements of the inverse of the matrix with elements
ηµλ = εµ δµλ (where ε0 = 1, ε1 = ε2 = ε3 = −1) i.e.

µλ µ 1, µ = ν
η η λ ν = δν = .
0 , µ 6= ν

Then the continuity equation reads

∂
div J = Jµ = 0 .
∂xµ
arc length Finally, considering the trajectory z to be a function (curve) parameterised by the arc length
d
s, i.e. η(z ′ (s), z ′ (s)) = 1 , where z ′ (s) = ds z(s) , one can write the equation of motion for
the trajectory describing the motion of a point particle with mass m and charge q concisely
as
µ ν
m c2 z ′′ = q F µ ν z ′ . (1.1.13)
1.2. SPECIAL LORENTZ TRANSFORMATIONS 3

The corresponding current density can be written in the form

Z
µ
J (x) = c q ds z ′ (s)µ δ (4) (x − z(s)) .

For a Poincaré transformation p : R 4 → R 4 with px = x

b = Λ x + a, or

bµ = Λµ ν xν + aµ , with ηµν = ηκλ Λκ µ Λλ ν ,

x

(i.e. the Lorentz-transformation Λ is an isometry of R 4 with respect to η : R 4 × R 4 → R isometry

with η(x, y) = ηµν xµ y ν ) , we declare the Poincaré transformed field strength tensor (p∗ F )
component-wise through
(p∗ F )µν (x) = Λκµ Λλν Fκλ (p(x)) . (1.1.14)

Likewise, the Poincaré transformed current density is given by

(p∗ J)µ (x) = Λκµ Jκ (p(x)) . (1.1.15)

Then Poincaré invariance of the basic equations means that if in the Maxwell equations
and the equation of motion for the particles the field strength tensor F is substituted by
p∗ F and simultaneously z(s) is substituted by p−1 (z) then the equations retain the same
form, i.e.

dF = 0 d(p∗ F ) = 0
4π 4π
(div F )ν = c Jν ⇔ (div (p∗ F ))ν = c (p∗ J)ν
µ ν ν
m c2 z ′′ = q F µ ν z ′ m c2 (p−1 (z)′′ )µ = q (p∗ F )µ ν (p−1 (z))′ .

Accordingly the Poincaré transformations are the symmetry transformations for the basic
equations of electrodynamics.

1.2 Special Lorentz transformations

The Lorentz transformations Λ form a group. Apart from space-time reflections the
Lorentz transformations with det Λ = +1 are of particular interest and form the special
Lorentz group as a subgroup. These contain the spatial rotations, but also the velocity
transformations (boosts) such as e.g. the transformation Λ : x 7→ x
b with boosts

b0
x = cosh α x0 + sinh α x1 ,
b1
x = sinh α x0 + cosh α x1 ,
b2
x = x2 ,
b3
x = x3 ,

which corresponds to a velocity transformation with v/c = tanh α in the e1 direction. For
b0 = cb
small velocities v ≪ c one would then find approximately (x0 = ct, x t)

b
t = t,
1
b =
x x1 + v t ,
b2
x = x2 ,
b3
x = x3 ,

i.e. a Galilei transformation, which, however, as we saw above, is not a symmetry trans-
formation for the basic equations.
 4 CHAPTER 1. SPECIAL RELATIVITY AND ELECTRODYNAMICS

1.3 Electromagnetic waves

A particular solution of the source-free Maxwell equations (i.e. J µ = 0) are plane,
monochromatic electromagnetic waves, for which the components of the field strength tensor
are given by
Fµν (x) = A (kµ nν − kν nµ ) cos (η(k, x) + δ) . (1.3.1)
wave Here A denotes the amplitude , δ denotes the phase, the wave vector k = k µ eµ is light-like ,
vector i.e η(k, k) = 0 , and n = nµ eµ is a space-like unit vector , i.e η(n, n) = −1 , perpendicular
polarisa- to the wave vector , i.e. η(k, n) = 0 . This vector
 n is called the polarisation vector .
tion P3
Note that η(k, x) = k x − i=1 k x = ω t − ~k · ~x , so that this field strength tensor
0 0 i i

angular represents a plane wave propagating in the direction of ~k with the angular frequency
frequency given by ω := c k 0 .
Chapter 2

Action

2.1 Introduction
In order to incorporate gravitational forces Einstein proposed to describe gravitation by a
metric gµν (x) that replaces the Minkowski-metric ηµν , but is space- and time dependent:
ηµν 7→ gµν (x) . In order to elucidate the possible consequences of such a replacement we
consider the following two-dimensional toy model:

2.1.1 A two-dimensional analogy

As an analogy illustrating the effects of having a varying metric consider the following
situation: Let a beetle live on a flat surface (R 2 ). Let the length of a curve be determined
by an euklidian metric hx, yi for x, y ∈ R 2 (i.e. hx, yi = x1 y 1 + x2 y 2 = ηµν xµ y ν with
ηµν = δµν , µ , ν = 1, 2): Z σb p
L(x) = dσ hx′ (σ), x′ (σ)i .
σa
We suppose that the animal measures lengths with a metallic rod. Unfortunately the tem-
perature varies from point to point on the surface and accordingly the poor beetle assigns
lengths of curves according to
Z σb p
L(x) = dσ k(x(σ)) hx′ (σ), x′ (σ)i ,
σa

where k(x) represents the shrinking and expansion of the measuring rod according to the
varying temperature on the surface. The length which the beetle attributes to the curve can
also be written as Z σ q b

L(x) = dσ gµν x(σ) x′ µ (σ)x′ ν (σ) ,

σa

where gµν (x) = (k(x))2 δµν . As an example we now consider the special case, where
2 1
k(x) = k(x1 , x2 ) = (x1 )2 +(x2 )2
.
r0 1 +
r02

If we now define
   
1 2 1 2 1 2 2 (x1 )2 + (x2 )2
~x0 (x , x ) := x e1 + x e2 + 1 − e3 ∈ R 3
1+ (x1 )2 +(x2 )2 r0 r0 r02
r02

then |~x0 |2 = (x10 )2 + (x20 )2 + (x30 )2 = 1 , i.e ~x0 ∈ S 2 ⊂ R 3 . Accordingly the curve can be
parameterised as ~x0 (σ) = ~x0 (x1 (σ), x2 (σ)) and the length of the curve is then, after some
computation, simply given by
Z σb
L(x) = dσ |~x′0 (σ)| ,
σa

5
6 CHAPTER 2. ACTION

which then is the length of a curve on the surface of the unit sphere S 2 . Accordingly an
intelligent beetle will infer that it lives on the surface of a sphere. Here curves of shortest
length are along the great circles on the 2-sphere. The image under projection onto R 2 are
then either (part of) (great) circles or segments of straight lines through the origin.
In this example we observe that the change of the local measure of length, i.e. the re-
placement ηµν → gµν (x) , globally implies the replacement R 2 → S 2 . More generally we
conclude that the local replacement for the space-time metric ηµν → gµν (x) will then corre-
spond globally to the replacement R 4 → M 4 , i.e a transition from Minkowski space to a
general 4-dimensional manifold.
2.2. THE ACTION OF ELECTROMAGNETISM 7

2.2 The action of electromagnetism

The homogeneous Maxwell equations in the form

∂ ∂ ∂
(dF )αβγ = α
Fβγ − β
Fαγ + Fαβ = 0
∂x ∂x ∂xγ
imply, via the Poincaré-lemma, that F = dA , where A is a (co)vector field with compo-
nents Aα , i.e. for the components of F

∂ ∂
Fαβ = Aβ − Aα
∂xα ∂xβ
holds. Then the action of electromagnetism is defined for arbitrary curves xi (τi ) (depending electro-
on arbitrary curve parameters τi ) of N particles with mass mi and charge qi and for arbitrary magnetic
vector fields A as action
N Z
X  q 
β
S[xi , A] := − dτi mi c ηαβ ẋα
i (τi ) ẋi (τi ) + qi Aα (xi (τi )) ẋα
i (τi )
i=1
Z
1
− d4x Fαβ (x) F αβ (x) . (2.2.1)
16π

Here ẋα d α
i (τi ) := dτi xi (τi ) denotes the derivative of the curves w.r.t. the curve parameter τi
and where in the last term Fαβ in terms of Aα is given by

∂ ∂
Fαβ = Aβ − Aα , F αβ = η αµ η βν Fµν .
∂xα ∂xβ
According to the principle of extremal action, both the equation of motion determining action
the particles trajectories and the (inhomogeneous) equation determining the fields then principle
follow. More specifically: If xi + λ δxi are (one-parameter families of) variations of the
curves, then requiring the action to be stationary w.r.t. this variation, i.e. the trajectories
xi are extrema of the action, or

∂ 
δS := S[xi + λ δxi , A] = 0,
∂λ λ=0

for arbitrary δxi non-vanishing in a finite domain for the curve parameter, leads to the equa-
tion of motion for the particle trajectories in the form of the Einstein-Lorentz equation:

d ẋµi (τi )
mi c 2 q = qi F µ ν (xi (τi )) ẋνi (τi ) ; (2.2.2)
dτi β
ηαβ ẋα
i (τi ) ẋi (τi )

likewise stationarity of the action w.r.t. arbitrary variations δA , i.e A is an extremum of

the action, or 
∂ 
δS := S[xi , A + λ δA] = 0,
∂λ λ=0

for arbitrary δA with finite support, leads to the imhomogeneous Maxwell equations in
the form:
∂ 4π ν
µ
F µν (x) = J (x) , (2.2.3)
∂x c
where
N
X Z
ν
J (x) = c qi dτi ẋνi (τi ) δ (4) (x − xi (τi ))
i=1

is the charge-current density corresponding to particles with charge qi on trajectories xi (τi ) .

 8 CHAPTER 2. ACTION

Einstein’s suggestion to describe gravitational forces by the replacement ηαβ 7→ gαβ (x)
implies for the action S , letting Fαβ = ∂α Aβ − ∂β Aα unchanged:
N Z
X  q 
β α
S[xi , A, g] := − dτi mi c gαβ (xi (τi )) ẋα
i (τi ) ẋi (τi ) + qi Aα (xi (τi )) ẋi (τi )
i=1
Z p
1
− d4x |g|(x) Fαβ (x) F αβ (x) (2.2.4)
16π
N Z
X  q  Z
1 p
≡ dτi −mi gαβ ẋα
i ẋβi − qi Aα ẋα
i − d4x |g| Fαβ F αβ ,
16π
i=1 | {z }
=: Li;A (xi ,v i )|
v i =ẋi

where for brevity in the last form we suppressed all arguments of the functions and have
chosen the units such that numerically c = 1 (which we shall do so from now on). Also, in
order not to overload the notation, we shall usually not underline four-vectors ∈ R 4 anymore.
Above we introduced the Lagrange function LA : R 4 × R 4 → R for a point particle of
mass m and charge q in the electromagentic field given by A:
q
LA (x, v) := −m gαβ (x) v α v β − q Aα (x) v α .

In the equation (2.2.4) above it is understood that

F α β = g αµ Fµβ , F αβ = g αµ g βν Fµν , |g| := | det g|

and we impose the restriction on gαβ (x) that gαβ (x) = gβα (x) for any fixed x ∈ R 4 can
always be diagonalised to ηαβ , if needed after appropriate scale transformations. This means
signature that g and η are supposed to have the same signature , i.e. the number of positive and
negative eigenvalues are the same. With the new metric gαβ (x), as before, the trajectories
of the particles and the electromagnetic field can be obtained as the extrema of the action,
as we shall show below.

2.2.1 Symmetries of the new action

But first we list symmetries of the action:
• The action of Eq. (2.2.4) is invariant under reparametrisations of the curves, i.e. under
replacements τi 7→ τi (σi ) . Indeed, for τ = τ (σ) , i.e. x(σ) = x(τ (σ)) , we have
Z r !
dxα (τ ) dxβ (τ ) dxα (τ )
dτ m c gαβ + q Aα (x(τ ))
dτ dτ dτ
Z r !
dτ dxα dxβ dxα
= dσ m c gαβ + q Aα (x(τ (σ)))
dσ dτ dτ dτ
Z r !
dxα dτ dxβ dτ dxα dτ
= dσ m c gαβ + q Aα (x(τ (σ)))
dτ dσ dτ dσ dτ dσ
Z r !
dxα (σ) dxβ (σ) dxα (σ)
= dσ m c gαβ + q Aα (x(σ)) .
dσ dσ dσ

gauge • Furthermore, under gauge transformations , i.e. if

transfor-
mations Aα 7→ A′α = Aα + ∂α χ ,

with χ : R 4 → R an arbitrary function, the field strength tensor is invariant:

Fαβ [A′ ] = ∂α A′β − ∂β A′α = ∂α Aβ + ∂α ∂β χ − ∂β Aα − ∂β ∂α χ = Fαβ [A]

2.2. THE ACTION OF ELECTROMAGNETISM 9

and for the action we find:

N
X Z
d
S[xi , A′ , g] = S[xi , A, g] − qi dτi χ(xi (τi )) ,
dτi
i=1 | {z }
=(∂α χ)(xi (τi )) ẋα
i (τi )

where, upon variation, the last term does not contribute for arbitrary functions χ with
compact support.
• The action is Poincaré invariant as long as gαβ (x) = ηαβ but this property is lost in
general.
However, if for an arbitrary diffeomorphism ϕ : R 4 → R 4 (general coordinate trans-
formation) we declare

Aµ 7→ (ϕ∗ A)µ ; gµν 7→ (ϕ∗ g)µν ; xi (τi ) 7→ ϕ−1 (xi (τi )) , (2.2.5)

where
∂ϕα (x) ∂ϕα (x) ∂ϕβ (x)
(ϕ∗ A)µ (x) = Aα (ϕ(x)) , (ϕ∗ g)µν (x) = gαβ (ϕ(x)) , (2.2.6)
∂xµ ∂xµ ∂xν
noting that
F [ϕ∗ A]µν = (ϕ∗ F )[A]µν ,
the action is invariant under the replacements of Eq.(2.2.5): The argument for this
statement is based on the following considerations: Denoting the matrices symbolically
by their matrix elements,
  α    β 
 ∂ϕ  
∗ 
|det [(ϕ g)µν (x)]| = |det [gαβ (ϕ(x)]| det (x)  det ∂ϕ (x)  ,
∂x µ   ∂x ν 
| {z }| {z }
=|det [Dϕ(x)]| =|det [Dϕ(x)]|

∂ϕµ
where Dϕ is the Jacobian of the transformation with matrix elements Dϕµν = ∂xν ,
and therefore
p p
|det [(ϕ∗ g)(x)]| = |det [g(ϕ(x)]| |det [Dϕ(x)]| .

As a consequence, for x 7→ y = ϕ(x) , we have for the integration of a function f :

Z p Z p
4
d x |det [Dϕ(x)]| |g(y(x))| f (y(x)) = d4y |g(y)| f (y) .

Furthermore, with the substition of Eq.(2.2.5) , i.e. with xi (τi ) 7→ yi (τi ) = ϕ−1 (xi (τi )) ,
xi (τi ) = ϕ(yi (τi )) , we find
∂ϕµ (yi (τi )) d −1
Aα (xi (τi )) ẋα
i (τi ) 7→ Aµ (ϕ(yi (τi ))) ϕ (xi (τi ))α
∂yiα dτi
∂xµi ∂yiα β
= Aµ (xi (τi )) ẋ (τi ) = Aµ (xi (τi )) ẋµi (τi ) .
∂yiα ∂xβi i
| {z }
µ
=δβ

Analogously,
∂ϕα −1 ∂ϕβ d −1 d −1
gµν (xi ) ẋµi ẋνi 7→ gαβ (ϕ(yi )) µ (ϕ xi ) ν (ϕ−1 xi ) ϕ (xi )µ ϕ (xi )ν
∂yi ∂yi dτ dτ
∂xα
i ∂xβi ∂yiµ ν
γ ∂yi
= gαβ (xi ) µ (y i ) ν (y i ) γ (yi ) ẋi (yi ) ẋδi
∂yi ∂yi ∂xi ∂xδi
= gαβ (xi ) δγα δδβ ẋγi ẋδi = gαβ (xi ) ẋα β
i ẋi .
 10 CHAPTER 2. ACTION

2.2.2 Equations of motion

The requirement that the trajectories are extrema of the action in Eq. (2.2.4) , i.e. with
arbitrary variations of the curves δxi non-vanishing in a finite domain:

d 
S[xi + λ δxi , A, g] =0
dλ λ=0

holds, will then yield the equations of motion for the trajectories as the Euler-Lagrange
equations:  
d ∂Li;A (xi , vi )  ∂Li;A (xi , vi ) 
−  =−  .
dτi ∂viµ vi =ẋi ∂xµi vi =ẋi

For the left hand side of these equations we obtain:

 
d  mi ẋα g m ẍα g mi
q i αµ + qi Aµ  = q i i αµ + q (∂β gαµ ) ẋα β
i ẋi
dτi γ λ
gγλ ẋi ẋi γ λ
gγλ ẋi ẋi γ λ
gγλ ẋi ẋi 1 | {z }
β
= 2 {(∂β gαµ )+(∂α gβµ )} ẋα
i ẋi
 
d  1  + qi ∂ν Aµ ẋνi ,
+mi ẋα
i gαµ q
dτi g ẋ ẋγ λ
γλ i i

while the right hand side reads

β
mi (∂µ gαβ ) ẋαi ẋi
qi ∂µ Aν ẋνi + q .
2 gγλ ẋγi ẋλi

Rearranging terms we thus find:

1 α β
mi gµα ẍα
i + 2 mi {(∂β gαµ ) + (∂α gβµ ) − (∂µ gαβ )} ẋi ẋi
q
gγλ ẋγi ẋλi
 
d q 1 .
= qi (∂µ Aν − ∂ν Aµ ) ẋνi − mi ẋαi gαµ
| {z } dτi γ λ
gγλ ẋi ẋi
=Fµν

We now invoke the invariance of the action under reparametrisation to impose the additional
β
condition gαβ (xi ) ẋα
i ẋi = 1 . Note that this implies

d  β

gαβ (xi ) ẋα
i ẋi = 0.
dτi
Then with this convention, after “raising the index” with the inverse of the metric g , the
equations of motion for the particle trajectories are given by:
 
mi ẍµi + Γµαβ ẋα ẋ
i i
β
= qi F µ ν ẋνi , (2.2.7)

where
F µ ν := g µα Fαν
and where we defined the Christoffel-symbols by
Christoffel-  
symbol µ 1 µλ ∂ ∂ ∂
Γαβ := g gλβ + gλα − gαβ . (2.2.8)
2 ∂xα ∂xβ ∂xλ

For vanishing electromagnetic fields a particle trajectory x(τ ) is thus determined by the
differential equations:

ẍµ (τ ) + Γµαβ (x(τ )) ẋα (τ ) ẋβ (τ ) = 0 , µ = 0, 1, 2, 3 . (2.2.9)

2.2. THE ACTION OF ELECTROMAGNETISM 11

which is called the geodesic equation. geodesic

Likewise the action principle for a variation of the fields A, noting that equation

d 
Fµν [A + λ (δA)] F µν [A + λ (δA)] = 2 F µν [A] (∂µ (δA)ν − ∂ν (δA)µ ) ,
dλ λ=0

leads to

d 
0 = δS = S[xi , A + λ (δA), g]
dλ λ=0
XN Z Z  
ν 1 4
p µν ∂ ∂
= − dτi qi (δA)ν ẋi − d x |g| F (δA)ν − (δA)µ
8π ∂xµ ∂xν
i=1 | {z }
∂
=2 F µν ∂xµ (δA)ν
N Z
X Z p
(δA)ν (x) δ (4) (x − xi (τi )) ẋνi
= − dτi qi d4x p
|g(x)|
i=1
|g(x)|
Z Z !
1 ∂ p  1 p 1 ∂ p
− d4x µ |g| F µν (δA)ν + d4x |g| p |g| F µν (δA)ν
4π ∂x 4π |g| ∂xµ
| {z }
=0
Z !
4
p ν 1 1 ∂ p µν

= − d x |g(x)| J (x) − p |g(x)| F (x) (δA)ν (x) ,
4π |g(x)| ∂xµ

where the charge-current density is now given by

N Z
X 1
J ν (x) := dτi qi ẋνi (τi ) p δ (4) (x − xi (τi )) . (2.2.10)
i=1
|g(x)|

Since (δA) is arbitrary (with compact support) the inhomogeneous Maxwell equations in
the form
1 ∂ p µν

p |g(x)| F (x) = 4π J ν (x) , ν = 0, 1, 2, 3 (2.2.11)
|g(x)| ∂xµ

follow.
 12 CHAPTER 2. ACTION

2.3 The full action

In order to obtain the differential equations determining the metric g(x) , this action should
be supplemented by an additional term SG ([g] which according to a proposition of Einstein
should retain the invariance under arbitrary diffeomorphisms ϕ : R 4 → R 4 : Accordingly one
writes Z p
SG [g] = d4x |g(x)| L[g](x) ,

where L[ϕ∗ g] = ϕ∗ L[g] and where the Lagrange density L[g] is supposed to be a func-
tion(al) of gµν (x), ∂x∂ κ gµν (x) and ∂x∂ κ ∂x∂ λ gµν (x) only, is linear in the second order partial
derivatives and higher order derivatives do not contribute.
In this case one can exploit a theorem by Hilbert which states:
If L[g] contains the second order derivatives of gµν at most linearly and no
derivatives of still higher order, then L[g] has the most general form
1
L[g] = − R[g] + Λ ,
16π G
if Z p
SG [g] = d4x |g(x)| L[g](x)

is to be invariant under arbitrary coordinate transformations. Here, R(g) is

Ricci the so called Ricci scalar, which is a known, uniquely determined function of
scalar gµν (x), ∂x∂ κ gµν (x) and ∂x∂ κ ∂x∂ λ gµν (x) , see below. Furthermore, G = const. is the
gravita- gravitational constant1 and Λ = const. is the so-called cosmological constant .
tional
constant We shall make a comment on the uniqueness of SG later. For the time being we shall
cosmologi- put the cosmological constant Λ = 0 and postpone the discussion of the evidence for and
cal implications of Λ 6= 0 .
constant
In the above, the Ricci (or curvature) scalar is constructed as follows:
The Christoffel symbols were defined as
 
µ 1 µλ ∂ ∂ ∂
Γαβ := g gλα + gλβ − gαβ .
2 ∂xβ ∂xα ∂xλ
curvature With these we define the components of the so-called curvature tensor (which is a (3, 1)-
tensor tensor field, also see exercises) as
∂ ∂
Rµ ναβ := Γµνβ − Γµ + Γµλα Γλνβ − Γµλβ Γλνα , (2.3.1)
∂x α ∂xβ να
Ricci then via contraction the components of the Ricci tensor (which is a (2, 0)-tensor field) by
tensor
Rνβ := Rα ναβ (2.3.2)
curvature and finally by full contraction with the (inverse of the) metric the curvature scalar (thus
scalar a (0, 0) tensor field or function) as
R[g] := g νβ Rνβ , (2.3.3)
which then indeed is linear in the second order partial derivatives of the metric g and can
be shown to possess the property R[ϕ∗ g] = (ϕ∗ R)[g] .
Accordingly the full action now reads
XN Z  q 
β α
S[xi , A, g] = − dτi mi gαβ ẋα ẋ
i i + q A ẋ
i α i
i=1
Z p Z p
1 4 1
− dx |g| Fαβ F αβ
− d4x |g| R[g] . (2.3.4)
16π 16π G
1 In the S(ystème) I(international d’unités) the numerical value of the gravitational constant is:

G = 6.6742(10) 10−11 m3 kg−1 s−2 .

2.3. THE FULL ACTION 13

2.3.1 Einstein’s field equations

The equations determining the metric are obtained from the action (2.3.4) by considering
the variation of the action with respect to g ,i.e. one requires:

d 
δS := S[xi , A, g + λ (δg)] =0
dλ λ=0

for all (δg)µν with compact support.

We start with the calculation of the variation δ S̃ for the first two terms in S:
N Z
X  q  Z
β 1 p
S̃[xi , A, g] = − α α
dτi mi gαβ ẋi ẋi + qi Aα ẋi − d4x |g| Fαβ F αβ .
i=1
16π

Here
Fαβ = ∂α Aβ − ∂β Aα , F µν = g µα g νβ Fαβ
and we obtain2
N
X Z Z  p 
1 (δg)µν ẋµi ẋνi 1
δ S̃ = − mi dτi q − d4x δ |g| Fαβ F αβ
2 β 16π
i=1 gαβ ẋα
i ẋi
Z p
1
− d4x |g| (δg)µα g νβ Fαβ Fµν
8π
N
X1 Z Z  p 
1
= − mi dτi (δg)µν ẋµi ẋνi − d4x δ |g| Fαβ F αβ
i=1
2 16π
Z p
1
− d4x |g| (δg)µα g νβ Fαβ Fµν ,
8π
where for the second form we again took the arc length as a special parametrisation, i.e.
β
gαβ ẋα
i ẋi = 1 holds . The result can be written in the form
Z p
1
δ S̃ = − d4x |g| (δg)µν T µν ,
2
with
µν
T µν = Tmat µν
+ Tem ,
the total energy-momentum tensor, as we shall argue below.

2 For f [g] the variation δf is defined as


d 
δf := f [g + λ (δg)] .
dλ λ=0
14 CHAPTER 2. ACTION

Indeed the first term in δ S̃ contains

N
X Z
µν ẋµi (τi ) ẋνi (τi ) (4)
Tmat (x) = mi dτi p δ (x − xi (τi )) ,
i=1
|g(x)|

which is the material contribution to the energy-momentum tensor. For the last two terms,
with the relations (see exercises):
p 1p
δ |g| = |g| g µν (δg)µν , (δg)µα = −g µγ g αδ (δg)γδ ,
2
we compute:
Z  p  Z p
1 4 1
− d x δ |g| Fαβ F −αβ
d4x |g| (δg)µα g νβ Fαβ Fµν
16π 8π
Z p Z p
1 4 1
= − d x |g| g (δg)µν Fαβ F αβ +
µν
d4x |g| g µγ g αδ (δg)γδ g νβ Fαβ Fµν
32π 8π | {z }
=−g µγ g αν (δg)µν g δβ Fαβ Fδγ
Z !
1 p 1 1
= − d4x |g|(δg)µν g µγ g να Fαβ g δβ Fδγ + Fαβ F αβ g µν ,
2 4π | {z } 4
=Fγδ g δβ Fβα

where in the second step we interchanged δ ↔ ν , γ ↔ µ . Thus indeed this contribution

contains  
µν 1 µγ να δ 1 αβ µν
Tem (x) = g g Fγδ F α + Fαβ F g ,
4π 4
which is the electromagnetic contribution to the energy-momentum tensor.
The variation of the last term in Eq.(2.3.4) reads, with R(g) = g µν Rµν (g) :

d 
δSG := SG [g + λ(δg)]
dλ λ=0
Z h p  p p i
1
= − d4x δ |g| R + |g| (δg)µν Rµν + |g| g µν (δR)µν .
16π G
Now it can be shown (see exercises) that
1 ∂ p 
g µν (δR)µν = p |g| B µ
=: B µ ;µ with B µ := g αβ δΓµαβ − g µα δΓβαβ ,
|g| ∂xµ

i.e this term is the (covariant) divergence of a vector field and will not contribute to the
variation of SG , provided δgµν has compact support. Therefore it is sufficient to compute
Z h p  p i
1
− d4x δ |g| R + |g| (δg)µν Rµν
16π G
Z  
1 4 1p µν
p µα νβ
= − dx |g| g (δg)µν R − |g| g g (δg)αβ Rµν
16π G 2
Z  
1 4
p 1 αβ µα νβ
= − d x |g| (δg)αβ R g − g g Rµν
16π G 2
Z  
1 4
p 1 αβ αβ
= − d x |g| (δg)αβ Rg − R
16π G 2
Z  
1 4
p 1 µν µν
= − d x |g| (δg)µν Rg − R .
16π G 2
Now collecting all contributions we find that the action principle implies
Z   
1 4
p µν 1 1 µν µν
δS = − d x |g| T + Rg − R (δg)µν = 0 ,
2 8π G 2
 2.3. THE FULL ACTION 15

Einstein for all (δg)µν with compact support. Thus one obtains the Einstein equations
equations
1
Gµν (g(x)) := Rµν (g(x)) − R(g(x)) g µν (x) = 8π G T µν (x) (2.3.5)
2
with Gµν the components of the Einstein-tensor , which in turn fulfils
Einstein-
Gµν ;µ = 0 , tensor

which, see exercises, follows from a Bianchi-identity for the curvature tensor. Bianchi-
identity
16 CHAPTER 2. ACTION

2.4 Invariance
Consider Z p
dnx |g(x)| f (x) ,

where g(x) := det [g(x)] here denotes the determinant of the matrix with elements gαβ (x) .
If one replaces x 7→ y = ϕ(x) with ϕ an arbitrary diffeomorphism then
Z Z   α  Z
p p  ∂y  p
dnx |g(x)| f (x) 7→ dnx |g(y(x))| det β  f (y(x)) = dny |g(y)| f (y) .
∂x

Accordingly Z p
dnx |g(x)| R(x)

is invariant under g 7→ ϕ∗ g .
Above the action was given as S[{xi }, A, g] and the field equations follow from the fact that
the solutions of the field equations and the particle trajectories are extrema of this action.
The invariance of the full action under arbitrary diffeomorphisms ϕ discussed above implies
that if the trajectories {xi } and the fields A and g are extrema of the action, then the
trajectories {ϕ−1 (xi )} and the fields ϕ∗ A and ϕ∗ g are also extrema.
Chapter 3

Solutions of the equations

As shown in the previous chapter the theory is invariant under the change to arbitrary coor-
dinates. Nevertheless there are distinguished sets of coordinates, with particular, sometimes
convenient properties, e.g. when solving the equations.

3.1 Distinguished coordinates

3.1.1 Comoving coordinates
Suppose that the metric with matrix elements gµν in terms some coordinates has the form
g00 = 1 , g0i = gi0 = 0 , i = 1, 2, 3 , while gik i, k = 1, 2, 3 are arbitary. Such coordinates where
this holds are called comoving coordinates. Then the trivial curves x0 (τ ) = α τ + β, α = comoving
const. , β = const. , xi (τ ) = y i = const. , i = 1, 2, 3 are solutions of the geodesic equations: coordi-
nates
ẍµ (τ ) + Γµαβ (x(τ )) ẋα (τ ) ẋβ (τ ) = 0 . (2.2.9)

which upon inserting the expression for the Christoffel symbols can also be written as
1
ẍµ + g µλ (∂β gαλ + ∂α gβλ − ∂λ gαβ ) ẋα ẋβ = 0 ,
2
or
1 µλ
ẍµ + g µλ ∂α gβλ ẋα ẋβ = g (∂λ gαβ ) ẋα ẋβ
2
1
⇒ gκµ ẍµ + ∂α gµκ ẋα ẋµ = (∂κ gαβ ) ẋα ẋβ
2
d 1
⇔ (gκµ ẋµ ) = (∂κ gαβ ) ẋα ẋβ . (3.1.1)
dτ 2

17
 18 CHAPTER 3. SOLUTIONS OF THE EQUATIONS

Indeed, with xi (τ ) = y i = const. and thus ẋi (τ ) = 0 , i = 1, 2, 3 the right-hand side of this
equation always vanishes, because also for the single term with non-vanishing velocities
1
(∂κ g00 ) ẋ0 ẋ0 = 0 , κ = 0, 1, 2, 3 ,
2
holds, since g00 = 1 and hence ∂κ g00 = 0 . For κ = i = 1, 2, 3 also the left-hand side vanishes:
d

gi0 ẋ0 + gik |{z}
ẋk = 0 .
dτ |{z}
=0 =0

The single remaining equation for κ = 0 then reads

d

g00 ẋ0 = ẍ0 = 0
dτ
and has the simple solution x0 (τ ) = α τ + β , α = const. , β = const. . This means that the
particles then do not move and that for these particles only the time changes (linearly).
We shall now argue that for a given space-time metric g, given as a function of arbitrary
coordinates, one can always find a set of coordinates such that g expressed in these new
coordinates takes the comoving form mentioned above. The argument is based on the
Jacobi- Jacobi-method for finding solutions for particle trajectories in classical mechanics and
method which we shall therefore recapitulate.

Intermezzo: Jacobi-method

In classical mechanics the dynamics for the motion of a particle is determined

by the Hamilton-function H : R 3 × R 3 × R → R:

H(x, p, t) , with x, p ∈ R 3 .

The particle trajectory can be obtained as follows: For the momentum set
∂
p = ∇W ,i.e. for the components of the momentum we set pi (t) = ∂x i W (Q; x, t) .

Then determine the function W , hwhich depends

i explicitly on x and t and para-
2
matrically on Q ∈ R 3 with det ∂x∂i ∂QW
j 6= 0 , as a solution of the so-called
Hamilton-Jacobi equation:
∂
H(x, (∇W )(x, t), t) + W (x, t) = 0 ,
∂t
which in general is a non-linear partial differential equation. Once a solution is
found the trajectories x(t) are then obtained algebraically from the equations
∂W 3 3
∂Qj (Q; x, t) = −Pj where P ∈ R and Q ∈ R are constants.

The action for the motion of a single particle of mass m in the gravitational field described
by the metric gµν reads
Z q
S[x, g] = −m dτ gµν (x(τ )) ẋµ (τ ) ẋν (τ ) .

Now choose a time coordinate x0 and use this as a curve parameter ,i.e. τ = x0 = t and thus
ẋ0 = 1 . Then (temporarily setting now x(x0 ) = xi (x0 ) ei ∈ R 3 ) the action can be written
as Z Z
S[x, g] = dx0 L(x(x0 ), ẋ(x0 ), x0 ) = dt L(x(t), ẋ(t), t) ,

where (note: v 0 = 1)
q
L(x, v, x0 ) = −m g00 (x0 , x) + 2 g0i (x0 , x) v i (x0 ) + gik (x0 , x) v i v k
3.1. DISTINGUISHED COORDINATES 19

is the Lagrange function: R 3 × R 3 × R → R and a summation over repeated opposite

Latin indices ∈ {1, 2, 3} is understood. With the components of the canonical momentum
∂
pi = L(x, v, x0 )
∂v i
the Hamilton function is constructed via a Legendre transformation
 
H(x, p, x0 ) = pi v i − L(x, v, x0 ) v=v(x,p,x0 ) ,

i.e. the velocity is thought to be known as function of x, p, x0 .

Specifically we have
g0i + gik v k ∂W
pi = −m p =
j j
g00 + 2 g0j v + gjl v v l ∂xi
Then the Hamilton function is given by
q
g0i v i + gik v i v k
H = −m p +m g00 + 2 g0j v j + gjk v j v l
g00 + 2 g0j v j + gjl v j v l
g00 + g0i v i ∂W
= mp =− 0 ,
j l
g00 + 2 g0j v + gjk v v k ∂x

where in the last step we used that W is supposed to be a solution of the Hamilton-Jacobi
equation. With v 0 = 1 we can thus write
∂W gµν v ν
= −m p
∂xµ gκλ v κ v λ
and we find that the Hamilton-Jacobi equation can be written in the form
  
∂ ∂
g µν W W = m2 . (3.1.2)
∂xµ ∂xν

The comoving coordinates are then constructed as follows: With W (Q; x, x0 ) where Q1 , Q2 , Q3 =
const. the particle trajectories follow from
∂W
(Q; x, x0 ) = −Pj = const.
∂Qj
We can rescale W by defining W =: m W0 , for which then
∂ ∂
g µν W0 W0 = 1
∂xµ ∂xν
∂W0
holds. Now put y 0 = W0 for fixed Q = (Q1 , Q2 , Q3 ) and y i = ∂Qi . Then the metric
expressed in terms of the coordinates y has comoving form:
First build
∂y α ∂y β
g αβ (y) := (ϕ ∗ g)αβ (y) = g µν (x(y)) , where x = ϕ(y) ,
∂xµ ∂xν
which determines the transformation of g associated with x ↔ y . Then

∂y 0 ∂y 0 ∂W0 ∂W0
α=β=0: g 00 = g µν = g µν =1
∂xµ ∂xν ∂xµ ∂xν
and

α = 0, β = k :
∂y 0 ∂y k ∂W0 ∂ 2 W0 ∂  1 µν ∂W0 ∂W0 
g 0k (y) = g µν (x) µ = g µν
(x) = g (x) = 0,
∂x ∂xν ∂xµ ∂Qk ∂xν ∂Qk 2 | ∂x µ ν
{z ∂x }
=1
20 CHAPTER 3. SOLUTIONS OF THE EQUATIONS

furthermore g ik (y) is something. Then for matrix elements gαβ with gαβ g βγ = δαγ we obtain
also
g00 (y) = 1 , g0i (y) = gi0 (y) = 0 , gik (y) something .
Note that in the expression
∂xµ ∂xν
gαβ (y) = gµν (x)
∂y α ∂y β
the coordinates x are thought to be expressed through the coordinates y .
3.1. DISTINGUISHED COORDINATES 21

3.1.2 Riemannian coordinates

Let x0 ∈ R 4 be a point in space-time. There is a set of coordinates, such that in terms of
these the geodesics through x0 take the form

y(τ ) = τ y , y ∈ R 4 , y = const. ,

i.e. in these coordinates the geodesics are simply straight lines passing through x0 . These
are constructed as follows:
The geodesics in terms of arbitrary coordinates x follow from the differential equations

ẍµ (τ ) + Γµαβ (x(τ )) ẋα (τ ) ẋβ (τ ) = 0 . (2.2.9)

The solutions to these second order differential equations are unique if the initial values
x(0) = x0 and ẋ(0) = v are given. For fixed x0 the solution then depends on v and
accordingly we write
x(τ ) = x(v; τ ) ,
which, as stated above, is unique for fixed v . Now put

φτ (v) = x(v; τ ) .

This defines a one-parameter set of diffeomorphisms φτ : R 4 → R 4 which map the initial

values v to the actual value of the trajectory at τ . Now if x(v; τ ) is a solution of Eq.(2.2.9),
then this is also true for x(v; λτ ) , because

d2 µ d α d β
x (v; λτ ) + Γµαβ (x(v; λτ )) x (v; λτ ) x (v; λτ )
dτ 2  dτ dτ 
= λ2 ẍµ (v; λτ ) + Γµαβ (x(v; λτ )) ẋα (v; λτ ) ẋβ (v; λτ ) = 0 .

The initial velocity for this solution is given by


d 
x(v; λτ ) = λv
dτ τ =0

and thus
x(v; λτ ) = x(λ v, τ ) = φτ (λ v) .
If we now put τ = 1 , we obtain

x(v; λ) = x(λ v, 1) = φ1 (λ v)

or, renaming λ to τ :

x(v; τ ) = x(τ v, 1) = φ1 (τ v) ⇔ φ−1

1 (x) = τ v

which means that, after renaming y = v, the coordinate transformation x = φ1 (y) indeed
yields the statement y(τ ) = τ y made above.
In these coordinates it can be shown, see exercise, that for trajectories passing through the
origin y = 0, gαβ,µ (0) = 0 and therefore the metric has the expansion

1
gαβ (y) = gαβ (0) − Rαλβν y λ y ν + O(y 3 ) . (3.1.3)
3
By applying a linear transformation we can diagonalise the (constant) symmetric matrix
with elements gαβ (0) . By appropriate subsequent scale transformations we can then obtain
the Minkowski-metric ηαβ . The remaining freedom in the choice of coordinates are then
Lorentz-transformations that leave the Minkowski-metric invariant.
22 CHAPTER 3. SOLUTIONS OF THE EQUATIONS

3.1.3 Consequences
Uniqueness of the Ricci-scalar With the Riemannian (also called geodesic or nor-
mal) coordinates discussed above we can now also substantiate the claim, that essentially
there is only a single function f of the metric, its first and second partial derivatives, i.e.
f (gµν (x), gµν,α (x), gµν,αβ (x)) which has the property

f (ϕ(x)) = f ((ϕ∗ g)µν (x), (ϕ∗ g)µν,α (x), (ϕ∗ g)µν,αβ (x))

and that is linear in the second partial derivatives . To this end we take ϕ(x) = y and
accordingly f (y) = f (gµν (y), gµν,α (y), gµν,αβ (y)) . If we now put (without loss of generality)
y = 0 we have f (0) = f (ηµν , Rµν|αβ (0)) , where Rµν|αβ (0) contains the dependence of the
second order partial derivatives at y = 0, the first order partial derivatives gµν,α (0) vanishing
at this point. If the dependence on the second order partial derivatives of the metric is to
be linear, we must have
f (0) = C µναβ Rµν|αβ (0) .
Now invariance of f (0) under Lorentz-transformations
′ ′ ′ ′
Λ : Rµν|αβ (0) 7→ R′ µν|αβ (0) = Λα α Λβ β Λµ µ Λν ν Rµ′ ν ′ |α′ β ′ (0)

implies that
′ ′
α′ β ′ ′ ′ ′ ′
Cµ ν = C µναβ Λα α Λβ β Λµ µ Λν ν .
Since C can only depend on the single invariant tensor under Lorentz-transformations:
ηµν we find
C µναβ = c η µα η νβ
and therefore f (0) = c R(0) with R(0) = η βν Rβν (0) = η βν Rβανα
(0) the Ricci-scalar at
y = 0 . Thus, up to the constant factor c the Ricci-scalar is the only function meeting the
requirements stipulated in the claim and accordingly f (x) = R(x) = R(ϕ−1 (0)) .

Energy-momentum tensor for a massive body As a second application of geodesic

coordinates we consider the energy of a massive body, which acts as a source term for the
gravitational field , i.e. the metric, in the Einstein equation

Gµν (x) = 8π G T µν (x) .

In geodesic coordinates the components of the energy-momentum tensor for a massive body
consisting of N particles mi on trajectories yi (τi ) in the vicinity of the origin y = 0 are given
by
N
X Z
µν ẏiµ (τi ) ẏiν (τi ) (4)
T (y) = mi dτi p δ (y − yi (τi ))
i=1
|g(y)|
N
X Z
≈ mi dτi ẏiµ (τi ) ẏiν (τi ) δ (4) (y − yi (τi )) ,
i=1
p
because in geodesic coordinates gµν (y) = ηµν +O(y 2 ) and therefore |g(y)| ≈ 1 in the vicin-
ity of the origin. Therefore, in terms of the momenta pi (yi0 ) = mi ẏi (τi (yi0 )) and changing
the integration variable from τi to yi0 :

N Z
X  
1 dτi µ 0 ν 0
T µν (y) ≈ dyi0 p (y ) p (y ) δ(y 0
− y 0
) δ (3)
(~
y − ~
y (y
i i
0
))
i=1
mi dyi0 i i i i i

XN
1 XZ 1
= 0 0
µ 0 ν 0 (3) 0
pi (y ) pi (y ) δ (~y − ~yi (y )) = d3k 0 k µ k ν ρmi (~k, y)
p (y )
i=1 i m
k
i
3.1. DISTINGUISHED COORDINATES 23

where we defined
mi fixed
XN
ρmi (~k, y) := δ (3) (~k − p~i (y 0 )) δ (3) (~y − ~yi (y 0 )) ,
i=1
µ ν
which obviously
q is the density of particles with mass mi and 4-momentum k (with ηµν k k =
m2 ⇒ k 0 = m2 + |~k|2 ) at the point y . We now generalise ρm (~k, y) to a smooth den-
i i i

sity function for which we shall assume that it is isotropic and homogeneous: ρmi (~k, y) =
ρmi (|~k|) . Then
XZ 1
T µν = d3k q k µ k ν ρmi (|~k|)
mi m2i + |~k|2
and we obtain:
XZ q
µ=ν=0: T 00 = d3k m2i + |~k|2 ρmi (|~k|)
mi
XZ ∞ q
= 4π dk k 2 m2i + k 2 ρmi (k) = a
mi 0

(where we wrote k := |~k|) as well as 1

XZ
µ = j ,ν = 0 : T j0
= d3k k j ρmi (|~k|) = 0 ,
mi
XZ 
kj kl b = 0, j 6= l
µ = j ,ν = l : T jl
= dkq 3
ρmi (|~k|) =
c, j=l
mi m2i + |~k|2

with
XZ (k j )2 4π X ∞
Z
k4
c= d3k q ρmi (|~k|) = dk p 2 ρmi (k) ,
3 m 0 mi + k 2
mi m2i + |~k|2 i

R 2π R 2π Rπ
since in spherical polar coordinates 0 dϕ cos2 ϕ = 0 dϕ sin2 ϕ = π , 0 dϑ sin3 ϑ = 34 and
Rπ R R
dϑ sin ϑ cos2 ϑ = 32 , or, alternatively, observing that d3k (k 1 )2 g(|~k|) = d3k (k 2 )2 g(|~k|) =
R0 3 R
d k (k 3 )2 g(|~k|) = 13 d3k |~k|2 g(|~k|) .
In normal coordinates the components of the energy-momentum tensor therefore have the
form
XZ ∞ p
T 00 = ε = 4π dk k 2 m2 + k 2 ρm (k) ,
m 0
0j j0
T =T = 0,
Z
jl 4π X ∞ k4
T = p δjl = dk √ ρm (k) δjl , (3.1.4)
3 m 0 m2 + k 2

where ε is the energy density and p is the pressure which in general are related by the energy
so called equation of state p = p(ε) . density
Thus the energy-momentum tensor in normal coordinates can be written as pressure
equation
T µν = (ε + p) j µ j ν − p η µν , with j 0 = 1, j 1 = j 2 = j 3 = 0 ⇒ ηµν j µ j ν = 1 . (3.1.5) of state
In arbitrary coordinates we have
T µν (x) = (ε(x) + p(x)) j µ (x) j ν (x) − p(x) g µν (x) , with gµν (x) j µ (x) j ν (x) = 1 (3.1.6)
1 Note
Z ∞
dx x f (|x|) = 0 ,
−∞
because the integrand is an odd function of the argument.
24 CHAPTER 3. SOLUTIONS OF THE EQUATIONS

and the equation of state then reads p(x) = p(ε(x)) . Moreover, in order that energy-
momentum conservation holds, we must have

T µν ;µ (x) = 0 , (3.1.7)

see the exercises.

3.2. SOLUTION OF THE EINSTEIN EQUATION FOR A MASSIVE BODY 25

3.2 Solution of the Einstein equation for a massive body

For the sake of simplification we consider a system with spherically symmetry and time-
independence. Accordingly we express the metric in terms of polar coordinates: x0 =
t, x1 = r, x2 = ϑ, x3 = ϕ, which means that the nonvanishing components of the metric, see
exercises, are given by

g00 = eν(r) , g11 = −eλ(r) , g22 = −r2 , g33 = −r2 sin2 ϑ .

and p(r) = p(ε(r)) is the equation of state. Furthermore we require for the 4-vector j of the
previous paragraph:
1
j1 = j2 = j3 = 0 , j0 = √ ,
g00

such that gµν j µ j ν = 1 .

3.2.1 Christoffel symbols, curvature tensor and scalar

With
1 1
g 00 = e−ν(r) , g 11 = −e−λ(r) , g 22 = − , g 33 = −
r2 r2 sin2 ϑ
∂
and the only non-vanishing partial derivatives given by, with the notations f ′ (r) = ∂r f (r)
and f,µ (x) = ∂x∂ µ f (x) :

g00,1 = ν ′ (r) eν(r) , g11,1 = −λ′ (r) eλ(r) , g22,1 = −2 r , g33,1 = −2 r sin2 ϑ ,
g33,2 = −2 r2 sin ϑ cos ϑ ,

we find for the Christoffel symbols

1 µλ
Γµαβ = g (gλα,β + gλβ,α − gαβ,λ )
2

explicitly

1 −ν(r) ν ′ (r)
Γ0αβ = e (g0α,β + g0β,α − gαβ,0 ) = (δα0 δβ1 + δβ0 δα1 ) ,
2 2
1  
Γ1αβ = − e−λ(r) −ν ′ (r)eν(r) δα0 δβ0 − λ′ (r)eλ(r) δα1 δβ1 + 2 r δα2 δβ2 + 2 r sin2 ϑ δα3 δβ3
2
1

Γ2αβ = − 2 −2 r δα2 δβ1 − 2 r δα1 δβ2 + 2 r2 sin ϑ cos ϑ δα3 δβ3
2r
1
Γ3αβ = − 2 −2 r sin2 ϑ δα3 δβ1 − 2 r sin2 ϑ δα1 δβ3
2 r sin2 ϑ

−2 r2 sin ϑ cos ϑ δα3 δβ2 − 2 r2 sin ϑ cos ϑ δα2 δβ3 ,

Hence the non-vanishing Christoffel symbols are given by

ν′
Γ001 = Γ010 = ,
2
ν′ λ′
Γ100 = eν−λ , Γ111 = ,
2 2
(3.2.1)
Γ122 = − re−λ , Γ133 = −r sin2 ϑ e−λ ,
1
Γ212 = Γ221 =Γ313 = Γ331 = , Γ233 = − sin ϑ cos ϑ,
r
Γ323 = Γ332 = cot ϑ .
26 CHAPTER 3. SOLUTIONS OF THE EQUATIONS

With the Christoffel symbols of Eq.(3.2.1) one then calculates the Ricci tensor of which
the non-vanishing components after some computation can be shown to be
 ′ 2 !
0 ν ′′ ν λ′ ν ′ ν′
R0 = + − + e−λ ,
2 2 4 r
 ′ 2 !
1 ν ′′ ν λ′ ν ′ λ′
R1 = + − − e−λ ,
2 2 4 r
 ′ 
ν − λ′ 1 1
R22 = R33 = + 2 e−λ − 2 .
2r r r

From these the Ricci scalar, via R = Rββ , is computed as

 2 !
′′ λ′ ν ′ ν′ ν ′ − λ′ 2 2
R= ν − +2 +2 + 2 e−λ − .
2 2 r r r2

1
Finally, the Einstein tensor is given by Gµν = Rνµ − 2 R δνµ and has the non-vanishing
components:
 
0 1 1 λ′
G0 = − − e−λ ,
r2 r2 r
 
1 1 1 ν′
G1 = − + e−λ ,
r2 r2 r
!
2
1 ν′ λ′ ν ′ ν ′ − λ′
G22 = G33 = − ν +′′
− + e−λ .
2 2 2 r

The non-vanishing components of the energy-momemtum tensor are given by

1
T 00 = (ε + p) − p g 00 ⇒ T00 = (ε + p) − p = ε ,
g00
T 11 = −p g 11
T 22 = −p g 22 ⇒ T11 = T22 = T33 = −p .
T 33 = −p g 33

and occur on the right-hand side of the Einstein equations:

Gµν (x) = 8π G Tνµ .

3.2.2 Differential equations

An explicit calculation of the energy-momentum tensor, see exercises, shows that the equa-
tions
T µκ ;µ = 0
2
merely imply the single non-trivial equation

−2p′ (r) = ν ′ (r) (p(r) + ε(r)). (3.2.2)

2 This
also follows from direct computation: The Einstein equations read, abbreviating 8π G =: α and
noting that two equations are equal:
 
1 1 λ′
αε = 2
− 2
− e−λ
r r r
 
1 1 ν′
−α p = 2
− 2
+ e−λ
r r r
!
1 ′′ ν′2 λ′ ν ′ ν ′ − λ′
−α p = − ν + − + e−λ .
2 2 2 r
3.2. SOLUTION OF THE EINSTEIN EQUATION FOR A MASSIVE BODY 27

Then the Einstein-equations reduce to the 2 independent equations:

G00 (r) = 8π G ε(r) , (3.2.3)

G11 (r) = −8π G p(ε(r)) . (3.2.4)

Eq.(3.2.3) can be written as

d  −λ(r) 
8π G ε(r) r2 = 1 − (1 − r λ′ (r)) e−λ(r) = − re −r
dr
from which Z r
8π G dρ ρ2 ε(ρ) = −r(e−λ(r) − 1)
0
follows. Now the left-hand side is proportional to the total mass contained within a ball of
radius r : Z r
M(r) = 4π dρ ρ2 ε(ρ) (3.2.5)
0
and therefore  −1
λ(r) 2 G M(r)
e = 1− = −g11 (r) .
r
Eq.(3.2.4) in the form
 
1 ν ′ (r) 1
−8π G p(r) = − + e−λ(r) +
r2 r r2

can now be solved for ν ′ (r) :

2 G M(r)
8π G r p(r) + r2
ν ′ (r) = 2 G M(r)
1− r

which can again be integrated w.r.t. r to yield

Z 2 G M(ρ)
!
r 8π ρ p(ρ) + ρ2
g00 (r) = exp (ν(r)) = exp dρ 2 G M(ρ)
.
0 1− ρ

Eq.(3.2.2) still has to be dealt with. It has the trivial solution ε(r) = p(r) = 0 which is
indeeed valid everywhere outside (r > R0 ) the massive spherical body of radius R0 . For
r < R0 we expect a non-trivial solution. In order to obtain this we now invoke the equation
of state
p(r) = p(ε(r)) ,
First of all, subtraction of the first two equations yields:
λ′ + ν ′ −λ
α(ε + p) = e . (∗)
r
From the second equation we infer by differentiating w.r.t. r :
 
2 2 ν ′′ λ′ + ν ′ λ′ ν ′
−α p′ = − 3 − − 3 + − − e−λ
r r r r2 r
Eliminating ν ′′ with the third equation and subsequently eliminating − 2 α r
p
with the second equation gives:
  !
′ ′ ′ ′ ′ 2 ′ ′
2 2 λ +ν λ ν 2αp ν λ ν ν ′ − λ′
−α p′ = − 3 + + + e −λ
− + − + e−λ
r r3 r2 r r 2r 2r r2
    !
2 2 λ′ + ν ′ λ′ ν ′ −λ 2 2 2 ν′ −λ ν′2 λ′ ν ′ ν ′ − λ′
= − 3 + + + e + − + e + − + e−λ
r r3 r2 r r3 r3 r2 2r 2r r2
!
λ′ ν ′ ν′2 α
= + e−λ = (ε + p) ν ′ ⇒ −2 p′ = ν ′ (p + ε) . (Eq.(3.2.2)) ,
2r 2r 2

where in the last line we used Eq.(∗) .

 28 CHAPTER 3. SOLUTIONS OF THE EQUATIONS

and thus
dp dp
= p′ (r) = (r) ε′ (r) ,
dr dε
which we insert in Eq.(3.2.2) :
1 1
ε′ (r) = − dp
ν ′ (r)(ε(r) + p(ε(r)))
2 dε (r)

and with the expression for ν ′ (r) given above, we find the first order differential equation
2 G M(r)
1 1 8π G r p(ε(r)) + r2
ε′ (r) = − dp
(ε(r) + p(ε(r))) . (3.2.6)
2 dε (r) 1 − 2 G M(r)
r

For the mass function M(r) we obtain from Eq.(3.2.5) the differential equation

M′ (r) = 4π r2 ε(r) , (3.2.7)

and the problem is thus formulated in terms of the two coupled (non-linear) differential
equations (3.2.6) and (3.2.7) which are subject to the boundary conditions (“initial” values)
M(0) = 0 and ε(0) = ε0 = const. with ε0 > 0 .
Inspection of Eq.(3.2.6) (e.g. for small r ; note that then M(r) ∝ r3 ) reveals that ε will be
a decreasing function of r as r increases and we expect that there will be some radius R0
such that ε(r)|r↑R0 = 0 . For r > R0 we then use the trivial solution ε = p = 0 . Thus for
r > R0 we have Z R0
M(r) = M(R0 ) = 4π dρ ρ2 ε(ρ)
0
and with ε = p = 0 we find for r > R0 :
2 G M(R0 )  
′ r2 d 2 G M(R0 )
ν (r) = 2 G M(R0 )
= log 1 −
1 − dr r
r

and accordingly for r > R0 the metric is determined by

   −1
ν(r) 2 G M(R0 ) λ(r) 2 G M(R0 )
g00 (r) = e = 1− , g11 (r) = −e =− 1− .
r r
(3.2.8)
Note that the radius R0 and the total mass M(R0 ) can not be varied independently: Both
are determined by the coupled differential equations (3.2.6) and (3.2.7) of which the so-
lution (apart from the trivial constraint M(0) = 0) is completely determined by ε0 , the
central central energy density, alone. Hence the equation of state completely determines the
energy relation between the total mass of the object M (R0 ) = M (R0 (ε0 )) and its radius R0 (ε0 ),
density
both parametrically dependent on the central energy density ε0 .
3.2. SOLUTION OF THE EINSTEIN EQUATION FOR A MASSIVE BODY 29

A specific model for the equation of state As a specific model for the material
equation of state one might e.g. consider an ideal Fermi gas at zero temperature, i.e. in its
ground state. Using again normal coordinates one can determine the momentum distribution
function ρ(~k, x), see the first section of this chapter, of non-interacting spin- 21 particles of
mass m confined to e.g. a cube with side length L as sketched below: The single particle
wave functions ψ : R 3 × R → C 2 (units such that ~ = c = 1) are given by

ψk,σ (x) = φk (x) χσ ,

where χσ ∈ C 2 , σ = ± 21 is a spinor
1 µ
xν
φk,σ (x) ∝ 3 e−i ηµν k .
L 2

Here, because of the boundary conditions the momenta only take discrete values according
to q
~k = 2π (n1 ~e1 + n2 ~e2 + n3 ~e3 ) , ni ∈ Z, i = 1, 2, 3 , k 0 = m2 + |~k|2 ,
L
where {~e1 , ~e2 , ~e3 } is a ONB of R 3 and where we assumed that the energy-momentum relation
(dispersion relation) is that of relativistically free particles. Then in the ground state where
3
the
p N fermions confined in the volume V = L occupy all the energy levels upto ε = εF =
m2 + kF2 , where kF is the Fermi momentum related to the Fermi energy, the probability
to find a particle with the momentum k = |~k| is given by
X X X X 1
ρ(k, x) = δ (3) (~κ−~k) |ψκ,σ (x)|2 ∝ 2 δ (3) (~κ−~k) |φκ (x)|2 ∝ 2 δ (3) (~κ−~k) ,
L3
σ=± 12 |~
κ|<kF ~|κ|<kF |~
κ|<kF

which upon changing (for large V ) the sum over |~κ| < kF to an integral yields
2
ρ(~k, x) = Θ(kF − |~k|) ,
(2π)3
in fact independent of x . Then
Z Z kF
2 1 3 N
d3k ρ(~k, ~x) = 4π dk k 2 = k = =: n
(2π)3 0 3 π2 F V
is the particle number density and thus the Fermi-momentum is given in terms of n as

1
kF = 3 π 2 n 3 .

The energy density is then calculated according to

Z kF p Z p
E 2 2 m4 x 2
ε= = 4π dk k m 2 + k 2 = dy y 1 + y 2 = n0 m x3 ǫ(x)
V (2π)3 0 π2 0
  13
m3 k kF n
where we introduced n0 := 3π 2 , y := m and x := m = n0 as well as

3 h p  p i
ǫ(x) = 3
x (1 + 2 x2 ) 1 + x2 − log x + 1 + x2 .
8x
The pressure p is then given by differentiating the energy E = ε N with respect to the
volume V for a constant number of particles N :
∂E dε dx
p=− = −ε − V ,
∂V dx dV
dn
which, with dV = − VN2 = − Vn , the relation

dε p
= 3 n0 m x2 ǫ(x) + x3 ǫ′ (x) = n0 m 3 x2 1 + x2
dx
30 CHAPTER 3. SOLUTIONS OF THE EQUATIONS

as well as 2
dx 1 n− 3 x
= 1 = ,
dn 3 n3 3n
0
and 2 2
dx 1 n− 3 dn 1 n− 3 n 1 x
= 1 =− =−
dV 3 n 3 dV 3 n 13 V 3 V
0 0
leads to
x
1
p = −n0 m x3 ǫ(x) + n0 m 3 x2 ǫ(x) + x3 ǫ′ (x) = n0 m x4 ǫ′ (x) ,
3 3
d
where ǫ′ (x) = dx ǫ(x) .
dp
We also need dε . Since ε and p are now given as functions of x we compute
dp d


dp 1 dx x4 ǫ′ (x) 1 4 x3 ǫ′ (x) + x4 ǫ′′ (x) 1 4 x ǫ′ (x) + x2 ǫ′′ (x)
= dx
dε
= d
= 2 3 ′
= .
dε dx
3 dx (x3 ǫ(x)) 3 3 x ǫ(x) + x ǫ (x) 3 3 ǫ(x) + x ǫ′ (x)

With
√ √
′ 3 3 1 + x2 ′′ 12 12 1 + x2 3
ǫ (x) = − ǫ(x) + , ǫ (x) = 2 ǫ(x) − +√
x x x x2 1 + x2
we find:
dp x2
= .
dε 3(1 + x2 )
We now have all ingredients to formulate the differential equations for the mass-radius
relation for compact stars such as white dwarfs or neutron stars in terms of the particle
number density n(r) and the mass function M(r), as an auxiliarly function we use x(r) =
 1
n(r) 3
n0 : From Eq.(3.2.6) with
dε dε dx dn
= ,
dr dx dn dr
where, as above,
dε p
= 3 n0 m x2 ǫ(x) + x3 ǫ′ (x) = n0 m 3 x2 1 + x2
dx
and 2
dx 1 n− 3 x
= = ,
dn 3 n 13 3n
0
we get, after some algebra,
 ′


(x)
dn G p 1+ 1
3 x ǫǫ(x) M+ 4π
3 m r3 x n ǫ′ (x)
= −3 2 n0 x ǫ(x) 1 + x2 2GM
, (3.2.9)
dr r 1− r

while Eq.(3.2.7) reads

dM
= 4π m r2 n ǫ(x) . (3.2.10)
dr

Note that in these formulas x is a function of the number density n and that the energy
density is related the number density via

ε(r) = n0 m (x(n(r)))3 ǫ(x(n(r))) .

These two coupled differential equations can be integrated numerically from r = 0 (where
M(0) = 0 , ε(0) = ε0 ) to R0 where ε(R0 ) = 0 , which then, as stated above, gives the
mass-radius relation [M(R0 (ε0 )) , R0 (ε0 )]. A typical result is displayed in Fig.3.2.1 .
3.2. SOLUTION OF THE EINSTEIN EQUATION FOR A MASSIVE BODY 31

mass-radius relation for compact stars

10

Oppenheimer-Volkoff-Limit Chandrasekhar-Limit

1
M [M0]

0.1

unstable neutron stars unstable white dwarfs (56Fe)

0.01

0.001
1 10 100 1000 10000 100000
R [km]

Figure 3.2.1: Mass-radius relation for compact stars. The curves are the result of a nu-
merical integration of Eqs.(3.2.9) (3.2.10) on the basis of the assumption that the systems
can be described by equations of state of ideal Fermi gases, i.e. a pure neutron gas or a
neutral assembly of protons, neutrons and electrons taking into account the binding energy
as for 56
28 Fe, whichever yields the lowest energy density. For the latter the energy-density
is dominated by the baryonic mass, while the pressure is due to the electrons. Also indi-
cated are the Oppenheimer-Volkoff limit and the Chandrasekhar limit, as well as
the white-dwarf observational data for Sirius B, Eri B and Stein 2051 from: S.L. Shapiro,
S.A. Teukolsky, Black Holes, White Dwarfs and Neutron Stars (J. Wiley & Sons, Inc.
New York, 1983) Chapter 3 .
 32 CHAPTER 3. SOLUTIONS OF THE EQUATIONS

3.2.3 Schwarzschild exterior solution and particle trajectories

The motion for particles outside the mass distribution (r > R0 ) is determined by the metric,
also see Eq.(3.2.8), with non-vanishing components
2 G M0 1
g00 (r) = 1 − , g11 (r) = − 2 G M0
, g22 (r) = −r2 , g33 (r, ϑ) = −r2 sin2 ϑ ,
r 1− r

where we have set M0 = M(R0 ) . We then define the so-called Schwarzschild radius by
Schwarzschild
radius rS := 2 G M0 (3.2.11)

and note that the metric above is well defined in a domain where r > rS . The equation
of motion for the trajectories x(τ ) of particles in the gravitational field determined by this
metric is the geodesic equation in the form
d 1
(gµν ẋν ) = gαβ,µ ẋα ẋβ , (3.1.1)
dτ 2
where as usual we take the arc length as curve parameter, i.e. gαβ ẋα ẋβ = 1 with ẋα (τ ) =
d α
dτ x (τ ) ,. It is understood that the coordinates are the polar coordinates which we used to
determine the metric, i.e.

x0 = t , x1 = r , x2 = ϑ , x3 = ϕ .

We thus find (with gαβ,0 = gαβ,3 = 0 and gαβ,2 = −δα3 δβ3 2 r2 sin ϑ cos ϑ):

d  rS  0   rS  0
µ=0: 1− ẋ = 0 , ⇒ 1 − ẋ = a = const. , (3.2.12)
dτ r r
d   π
µ=2: − r2 ϑ̇ = −r2 sin ϑ cos ϑ (ϕ̇)2 , ⇒ ϑ = = const. , (3.2.13)
dτ 2
(is a trivial solution)
d

µ=3: − r2 sin2 ϑ ϕ̇ = 0 , ⇒ r2 ϕ̇ |sin 2
{z ϑ} = ℓ = const. . (3.2.14)
dτ
=1

We use gαβ ẋα ẋβ = 1 to find the remaining equation (note that ϑ̇ = 0) and obtain
 rS  0 2 ṙ2
0 = −1 + gαβ ẋα ẋβ = −1 + 1 − (ẋ ) − − r2 ϕ̇2 . (3.2.15)
r 1 − rrS

Now eliminating ẋ0 and ϕ̇ with Eqs.(3.2.12 , 3.2.14) we then find from Eq.(3.2.15) :

a2 ṙ2 ℓ2
0 = −1 + rS − rS − 2
1− r 1− r r
or  
ℓ2  rS  rS
0 = −1 + a2 − ṙ2 + 2 1 − − .
r r r
If we multiply this by m 2 , where m is the mass of a test particle, we find that for the
trajectory of a particle in the exterior field of a spherically symmetric massive body
!
2  GmM
m m m ℓ rS 0
(a2 − 1) = ṙ2 + 2
1− −
| {z 2} |2 {z2 r } r | {zr }
=E =T V

or
E =T +V
holds, i.e. the sum of the kinetic energy T and the potential energy V , where the replacement
1 7→ 1 − rrS reflects the correction within general relativity which mostly is very small
3.2. SOLUTION OF THE EINSTEIN EQUATION FOR A MASSIVE BODY 33

numerically3 , the r.h.s. of the equation is constant and we thus interpret the constant on
the l.h.s. as the total energy of the test particle in the gravitational field.
In order to study the geometric form of the trajectories we express r through ϕ , denoting
d
r′ (ϕ) = dϕ r(ϕ) :

ℓ
r = r(ϕ) ⇒ ṙ = r′ (ϕ) ϕ̇ = r′ (ϕ) = −ℓ u′ (ϕ) ,
r2 (ϕ)
1
where we introduced u(ϕ) := r(ϕ)with u′ (ϕ) = − r2 1(ϕ) r′ (ϕ) . Then we find
 
0 = a2 − 1 − ℓ2 (u′ )2 + ℓ2 u2 − rS ℓ2 u3 − rS u
r
′ a2 − 1 rS
⇒ u =± + 2 u − u2 + rS u3
ℓ2 ℓ
Z u !− 12
a2 − 1 rS 2 3
⇒ ϕ − ϕ0 = ± dv + 2 v − v + rS v ,
u0 ℓ2 ℓ

which determines ϕ(u) , where the underlined term is the general relativity correction. Dif-
ferentiating u′ from the first formula above again w.r.t. ϕ one gets
rS 3 rS 2
u′′ = −u+ u , (3.2.16)
2 ℓ2 2
where we again underlined the correction. Now if the (underlined) correction were absent
the solution of this differential equation would be
1 rS 1
= u(ϕ) = 2 + A cos (ϕ − ϕ0 ) ⇔ r(ϕ) = rS
r(ϕ) 2ℓ 2ℓ2 + A cos (ϕ − ϕ0 )
which is the parametric representation of an ellipse in polar coordinates.
If we now furthermore suppose, that the motion is almost circular, which is a good approx-
imation for (most of) the planetary motion in our solar system, i.e. u is approximately
constant, and write
rS
u(ϕ) = 2 + v(ϕ) ,
2ℓ
where the underlined term is accordingly supposed to be small, we find approximately
 r 2 rS
S
u2 ≈ 2
+v 2 .
2ℓ ℓ
Inserting this in Eq.(3.2.16) we find approximately the differential equation
 
3  rS 2 3 rS2
v ′′ = rS − v 1 − ,
2 2 ℓ2 2 ℓ2
which has the general solution

r !
rS 2
3 2 ℓ2 3 rS2
v(ϕ) = rS + A cos 1− (ϕ − ϕ0 ) ,
2 1 − 3 rS22 2 ℓ2
2ℓ
q
3 r2
where, without loss of generality, we can put ϕ0 = 0 . The factor 1 − 2 ℓS2 is responsible
for a precession of the ellipse. The precession angle ∆ϕ is determined by the equation
r  
2
3 rS2 q 1  2π ≈ 3π rS .
1− (∆ϕ + 2π) = 2π ⇒ ∆ϕ = − 1
2 ℓ2 3 r2 2 ℓ2
1 − 2 ℓS2
3 For the sun and the earth:
rS ⊙ ≈ 3 · 103 m , R0⊙ ≈ 7 · 108 m , rS ⊕ ≈ 9 · 10−3 m , R0⊕ ≈ 6.4 · 106 m ,
respectively.
34 CHAPTER 3. SOLUTIONS OF THE EQUATIONS

In the case of the solar system this corresponds for the planet Mercury to a precession angle
∆ϕ ≈ 0.1′′ per revolution or ≈ 43 arc seconds per century. This is in accordance with the
observational data, after the correction to the orbital motion of the planet Mercury due to
the presence of the other planets in the solar system 4 (≈ 532′′ per century) is subtracted.
Note that this correction is an order of magnitude larger.

3.2.4 Light rays

Geometrical optics implies that light rays are described by geodesics with
gαβ ẋα ẋβ = 0 , (3.2.17)
which holds together with the geodesic equation
d 1
(gµν ẋν ) = gαβ,µ ẋα ẋβ . (3.1.1)
dτ 2


Now, inserting the special solution of the latter, where ϑ = π2 , r2 ϕ̇ = ℓ and 1 − rrS ẋ0 = a
found before, into Eq.(3.2.17) leads to
 rS  0 2 ṙ2 a2 ṙ2 ℓ2
2 2
gαβ ẋα ẋβ = 1 − (ẋ ) − rS − r ϕ̇ = rS − rS − 2 = 0
r 1− r 1− r 1− r r
or
ℓ2  rS 
a2 = ṙ2 + 1 − .
r2 r
1 d
We again write u(ϕ) = r(ϕ) and with the notation u′ (ϕ) = dϕ u(ϕ) obtain the differential
equation
a2 = ℓ2 (u′ )2 + ℓ2 u2 (1 − rS u)
from which, by differentiating w.r.t. ϕ again we obtain
3
u′′ + u = rS u2 , (3.2.18)
2
where the term on the right-hand side is the correction due to general relativity. If this were
absent then a solution of Eq.(3.2.18) would be
1 R0
u0 (ϕ) = cos ϕ ⇔ r0 (ϕ) = ,
R0 cos ϕ
which thus corresponds to a straight line with a distance R0 to the origin.
Writing u = u0 + v and assuming that v is so small that u2 ≈ u20 we obtain approximately
the differential equation
3 rS
u′′ (ϕ) + u(ϕ) = cos2 ϕ ,
2 R02
which has the solution
1 rS

u(ϕ) = cos ϕ + 2 cos2 ϕ + 2 sin2 ϕ .
R0 2 R0
see also Fig.3.2.2 . The deflection angle 2 ∆ϕ can be obtained from the asymptotes, i.e. the
condition (r → ∞ ⇔ u → 0) as follows: Noting that in this case cos ϕ < 0 we find:
2 R0
cos2 ϕ + 2 − 2 cos2 ϕ = − cos ϕ
rS
 2  2
R0 R0
⇒ cos ϕ − =2+
rS rS
s  2  2 !
R0 R0 rS R0 R0 rS rS
⇒ cos ϕ = − 1+2 ≈ − 1+ ≈− ,
rS rS R0 rS rS R0 R0
4 Le Verrier (1859) was the first to point out that the precession of Mercury could not be fully explained

by the influence of the other planets

3.2. SOLUTION OF THE EINSTEIN EQUATION FOR A MASSIVE BODY 35

x2

∆ϕ
2 ∆ϕ

0 R0 x1

Figure 3.2.2: Deflection of light rays.

rS
from which we conclude that the values ϕ = ± π2 + ∆ϕ with ∆ϕ = R0 correspond to the
asymptotes where r → ∞ . The deflection angle is thus given by
rS
2 ∆ϕ = 2 . (3.2.19)
R0
As a consequence the gravitational field thus acts as a lens for light rays! This gravitational
lensing effect is in modern applications used to determine mass distributions in the universe. gravita-
Also in the application of the global positioning system (GPS) this effect has to be taken tional
into account. lensing
 36 CHAPTER 3. SOLUTIONS OF THE EQUATIONS

3.2.5 Black holes and other features

The domain of the Schwarzschild exterior solution for the metric in terms of the spherical
polar coordinates
x0 = t , x1 = r , x2 = ϑ , x3 = ϕ
with non-vanishing components
 rS   rS −1
g 00 = 1 − , g 11 = − 1 − , g 22 = −r2 , g 33 = −r2 sin2 ϑ
r r
is restricted to r > rS , where rS = 2 G M is the Schwarzschild radius, see also Fig.3.2.3 .

r = const.

rS r

Figure 3.2.3: Domain definition for the Schwarzschild exterior (r > rS ) solution.

Kruskal- There exists an extension of this metric, the so-called Kruskal-metric which uses the
metric coordinates
x0 = v , x1 = u , x2 = ϑ , x3 = ϕ (3.2.20)
with the non-vanishing components
g00 = f (v, u) , g11 = −f (v, u) , g22 = −r2 (v, u) , g33 = −r2 (v, u) sin2 ϑ , (3.2.21)
where the function r(v, u) is defined implicitly by
     
r(v, u) r(v, u) 4 rS3 r(v, u)
− 1 exp = u2 − v 2 and f (v, u) = exp − . (3.2.22)
rS rS r(v, u) rS
2 2
The domain where this metric
√ is defined is for all values of v,√u such that u − v > −1 ,
2 2
i.e. below the curve v = u + 1 and above the curve v = − u + 1 (both corresponding
to r = 0) as sketched in Fig.3.2.4 . There is a diffeomorphism φI : (t, r) 7→ (v, u) , explicitly
given by
r  
r r t
v(t, r) = − 1 e 2 rS sinh ,
rS 2 rS
r  
r r t
u(t, r) = − 1 e 2 rS cosh ,
rS 2 rS
which maps the definition domain of the Schwarzschild metric (grey area in Fig.3.2.3) to
the the darkest grey area in Fig.3.2.4 labeled “I”, where u > |v| , with the property that for
the Kruskal metric g = φ∗I g holds, see also exercises, in this region. Furthermore with φII
defined by
r  
r 2 rr t
v(t, r) = 1− e S cosh ,
rS 2 rS
r  
r 2 rr t
u(t, r) = 1− e S sinh ,
rS 2 rS
3.2. SOLUTION OF THE EINSTEIN EQUATION FOR A MASSIVE BODY 37

v r=0 r = rS

r = const.
II
I
u

III
IV
r = const.

r=0 r = rS

Figure 3.2.4: Domain definition for the Kruskal-Szekeres metric.

the region with r < rS , which we originally excluded for the Schwarzschild metric, is then
mapped to the lightest grey shaded area labeled “II” in Fig.3.2.4 . In addition there appears
a second “world” (areas “III” and “IV”) with the same properties, see also exercises.
We can now discuss the future of particle trajectories and light rays: We adopt the Kruskal
coordinates and metric and recall that both the particle trajectories and the light rays are
described by geodesics and posses the additional property

gαβ ẋα ẋβ = 1(0) ≥ 0 ,

where the value in parentheses refers to the case of light rays. Accordingly,

gαβ ẋα ẋβ = f (v, u) (v̇ 2 − u̇2 ) − r2 (v, u)(ϑ̇2 + sin2 ϑ ϕ̇2 ) ≥ 0 ⇒ v̇ 2 − u̇2 ≥ 0 ,

since f (v, u) ≥ 0 . Thus, if the geodesics point in the future (v̇ > 0) we find

v̇ ≥ |u̇|

and therefore Z τ Z τ
dσ v̇(σ) ≥ dσ |u̇(σ)|
0 0
or
v(τ ) − v(0) ≥ |u(τ ) − u(0)| ,
which implies that the endpoints v(τ ), u(τ ) of a geodesic lie necessarily within the (future)
light cones beginning at the starting point v(0), u(0) .

x(v(τ ), u(τ ))

x(v(0), u(0))

Now we can discuss the future fate of trajectories of geodesics, depending on the starting
point x0 := x(v(0), u(0))
(a) x0 ∈ ”I”: Here the starting point is in a domain corresponding to the exterior do-
main where the Schwarzschild solution is valid, see Fig.3.2.5 (left) . There are two
possibilities:
1. The particles and the light rays remain in domain “I” ;
 38 CHAPTER 3. SOLUTIONS OF THE EQUATIONS

v r=0 r = rS v r=0 r = rS

available
future

available
future
x0

II x0 II
u u
III III
IV I IV I

Figure 3.2.5: Available future for particles or light rays starting at x0 = x(v(0), u(0)) ∈ ”I”
(left) or at x0 = x(v(0), u(0)) ∈ ”II” (right) .

2. The particles or light rays cross the line corresponding to r = rS and hit the
singularity at r = 0 .

(b) x0 ∈ ”II”: This situation is depicted in Fig.3.2.5 (right) . Here the particles or light
black hole rays inevitably hit the singularity at r = 0 and the black hole phenomenon occurs:
particles and light rays can never leave domain “II” and this region can only be entered.
For this reason r = rS is called the “event horizon”.

In any case the only effect on the exterior of the black hole is the gravitational field itself.
3.2. SOLUTION OF THE EINSTEIN EQUATION FOR A MASSIVE BODY 39

The wormhole feature of the Kruskal solution The 3-space connected to the Kruskal
solution, i.e. the subspace with v = 0, can be visualised by putting ϑ = π2 and finding a
map which embeds the resulting 2-space isometrically into R 3 .

This can be achieved as sketched as follows: We recall that (setting r = r̄)

    r
t v r
tanh = , u2 − v 2 = − 1 e rS
2 rS u rS
in regions “I” and “III” . Thus v = 0 corresponds to t = 0 and for fixed ϑ = π2 the space-like 2-dimensional
surface is given in terms of the coordinates (u, ϕ) where u is in the range (−∞ → 0 → +∞) , corresponding
to values of r given as (∞ → rS → ∞) , respectively, u and r being related by
r
r r
u=± − 1 e 2 rS .
rS
In terms of Schwarzschild coordinates x1 = r, x2 = ϕ the metric of this space-like 2-dimensional surface is
given by the non-vanishing components5
1
g11 = , g22 = r 2 .
1 − rrS

We now compare this to the metric of R 3 in terms of cylindrical coordinates y 1 = r, y 2 = ϕ, y 3 = z with the
non-vanishing components
g̃11 = 1 , g̃22 = r 2 , g̃33 = 1 ,
where we specify a 2-dimensional surface by a function z(r, ϕ) = z(r) with a metric then given by the
non-vanishing components
∂z ∂z
g̃ˆ11 = g̃11 + g̃33 = 1 + z ′ (r)2 , g̃ˆ22 = r 2 ,
∂r ∂r
which in turn are then identified with g11 , g22 , respectively. Hence,
1
1 + z ′ (r)2 ≡
1 − rrS
or
dz 1
= ±q ,
dr r
−1
r S

from which √
√
z(r) − z(rs ) = ±2 rS r − rS
follows, i.e. the possible values of (r, ϕ) lie on a paraboloid of revolution.
This results in the following picture:

Here the circles correspond to u = const., the middle of the “neck” of the “wormhole” to
u = 0 ⇔ r̄ = rS . Note that limr→∞ g11 (r) = 1 such that the regions “I” and “III”
(connected by the “neck” of the “wormhole”) are asymptotically flat.
5 Note that we here changed the sign, i.e. the metric here has the signature (+, +) .
 40 CHAPTER 3. SOLUTIONS OF THE EQUATIONS

Field of a charged point particle: Upto now we have assumed that the massive body
is uncharged. For the field of a charged point particle there is a solution for the metric,
which exhibit no singularities for realistic elementary particle masses and charges, because
Reissner- Q2 ≫ rS : The Reissner-Nordström-solution, which is a static solution to the Einstein-
Nordström Maxwell field equations, corresponding to the gravitational field of a charged, non-rotating,
spherically symmetric body of mass M and charge Q . For this system we have the equations:
∂ ∂
Gµν = 8π G T µν [F ] , Fµν [A] = µ
Aν − Aµ , F µν ;µ = 0 ,
∂x ∂xν
valid in the exterior region of the body, with the only non-vanishing components of the
electromagnetic potential and the metric given by
Q
A0 = , (3.2.23)
r
! !−1
2 2
rS rQ rS rQ
g00 = 1− + 2 , g11 =− 1− + 2 , g22 = −r2 , g33 = −r2 sin2 ϑ ,
r r r r

2
where rQ = G Q2 . Indeed rQ ≫ rS circumvents any singularity in the metrical components
for realistic elementary particles at finite r .

3.3 Gravitational collapse

There is evidence for the existence of gravitational fields with black hole features, e.g. in
the centre of our milky way. Obviously we need a description on how such black holes can
be formed, which implies that one should be able to find a solution for a gravitationally
collapsing body.
dust As a simplifying assumption we shall suppose that the body is formed from dust, which in
this context corresponds to the assumption that the constituent particles are non-interacting
and that the pressure vanishes: p = 0 . Moreover in the following we shall (again) assume
that the system is spherically symmetric. As before we take spherical polar coordinates

x0 = t , x1 = r , x2 = ϑ , x3 = ϕ ,

and the non-vanishing components of the metrical tensor are now supposed to be given as

f (t)2
g00 = 1 , g11 = − , g22 = −f (t)2 r2 , g33 = −f (t)2 r2 sin2 ϑ , (3.3.1)
1 − a r2
where a = const. . Obviously we here have a set of comoving coordinates. This metric will
be used to describe the interior of a collapsing body.
The non-vanishing Christoffel symbols are given by (denoting f˙(t) = dt d
f (t)) :

f f˙
Γ011 = , Γ022 = f f˙ r2 , Γ033 = f f˙ r2 sin2 ϑ
1 − a r2
f˙ ar
Γ101 = Γ110 = , Γ111 = , Γ122 = −(1 − a r2 ) r , Γ133 = −(1 − a r2 ) r sin2 ϑ ,
f 1 − a r2
f˙ 1
Γ202 = Γ220 = , Γ212 = Γ221 = , Γ233 = − sin ϑ cos ϑ ,
f r
f˙ 1
Γ303 = Γ330 = , Γ313 = Γ331 = , Γ323 = Γ332 = cot ϑ .
f r
(3.3.2)

Then after some computation one finds for the non-vanishing components of the Ricci
tensor:
f¨ f f¨ + 2 f˙2 + 2 a
R00 = −3 , R11 = R22 = R33 = − .
f f2
3.3. GRAVITATIONAL COLLAPSE 41

Then the Ricci scalar is given by

f f¨ + f˙2 + a
R = −6 .
f2
and the Einstein tensor Gµν = Rνµ − 21 R δνµ has the non-vanishing components:
!
f˙2 a 2 f f¨ + f˙2 + a
G00 = 3 + , G 1
1 = G 2
2 = G 3
3 = .
f2 f2 f2

For the energy momentum tensor we have the single non-vanishing component:
T 00 = ε(t) , (3.3.3)
since we assumed that the pressure vanishes. Then the condition T µν ;µ = 0 implies that

∂ 00 f˙(t)
0 = T µν ;µ = T (t) + T 00 Γµ0µ + T 00 Γ000 = ε̇(t) + 3 ε(t)
∂t f (t)
and thus
ε(0)
ε(t) = ,
f (t)3
where we have put f (0) = 1 , and accordingly the total energy of the dust cloud does not
change in time.
Then the Einstein equations yield
f˙2 + a 8π G ε(0)
= , (3.3.4)
f2 3 f3
¨ ˙2
2f f + f + a = 0. (3.3.5)
Now Eq.(3.3.5) is fulfilled if
  !
˙ 1 f˙ a
2
f =a −1 , Note: ⇒ 2 f˙ f¨ = −a 2 ⇒ 2f f¨ = − (3.3.6)
f f f

which after insertion in Eq.(3.3.4) gives

8π G ε(0)
a= , (3.3.7)
3
relating a to the energy density ε(0) at the time t = 0, the beginning of the collapse.
A parametric solution of Eq.(3.3.6), with the initial values f (0) = 1 , f˙(0) = 0 at the starting
point t = 0 of the collapse, is given by
1 1
f (ψ) = (1 + cos ψ) , t(ψ) = √ (ψ + sin ψ) , (3.3.8)
2 2 a
d d
since, with f ′ (ψ) = ′
dψ f (ψ) , t (ψ) = dψ t(ψ) ,

f ′ (ψ) − 21 sin ψ 1 − cos2 ψ 2 − (1 + cos ψ) a

f˙(ψ) = ′ = 1 ⇒ f˙2 = a 2
=a = − a,
t (ψ) √ (1 + cos ψ)
2 a
(1 + cos ψ) 1 + cos ψ f

and indeed 

t(ψ)|ψ=0 = 0 , f (ψ)|ψ=0 = 1 , f˙(ψ) =0
ψ=0

fulfils the initial conditions. The radius of the dust cloud vanishes for f (ψ0 ) = 0 , i.e. for
π
ψ0 = π, see also Fig. 3.3.1 . Then, however, t(π) = 2√ a
, which implies that the lifetime of
the dust cloud, given by
π π
T := t(π) = √ = q
2 a
2 8π G3ε(0)
is determined by the energy density at the starting point of the collapse: t = 0 .
42 CHAPTER 3. SOLUTIONS OF THE EQUATIONS

f (t)

0 T = π
√
2 a
t

Figure 3.3.1: Scale function in gravitational collapse of a spherical dust cloud.

3.3. GRAVITATIONAL COLLAPSE 43

We now discuss the question how this interior solution can be matched to an exterior solution,
for which we take as a candidate the Schwarzschild solution with the coordinates
y0 = t , y1 = r , y2 = ϑ , y3 = ϕ ,
the metric having the non-vanishing components
 rS   rS −1
g 00 = 1 − , g 11 = − 1 − , g 22 = −r2 , g 33 = −r2 sin2 ϑ
r r
with the domain restricted to r > rS , where rS = 2 G M is the Schwarzschild radius .
We therefore have to find a coordinate transformation e t=et(t, r), re = re(t, r) such that the
transformed metric ge can continuously be matched to g . According to Oppenheimer-
Snyder this is achieved by the map φ : (t, r) 7→ (e
t, re) :
q
r Z 1 ρ
1 − a r02 1−ρ
e
t(t, r) = dρ
a a r2
S(t,r) 1− 0 ρ
re(t, r) = f (t) r , (3.3.9)
where s
1 − a r2
S(t, r) := 1 − (1 − f (t)) . (3.3.10)
1 − a r02
The non-vanishing components of the metric are written as
ge00 = B(t, r) , ge11 = −A(t, r) , ge22 = −r2 f (t)2 , ge33 = −r2 f (t)2 sin2 ϑ , (3.3.11)
where
s  2
 −1 1− a r2
a r2 f (t) 1− a r2 S(t,r)
A(t, r) = 1 − , B(t, r) = . (3.3.12)
f (t) S(t, r) 1 − a r02 1− a r2
f (t)

Then for r = r0 one finds: S(t, r0 ) = f (t) and r̃(t, r0 ) = f (t) r0 and therefore:
 
1 a r02
A(t, r0 ) =   , B(t, r 0 ) = 1 − .
a r2 f (t)
1 − f (t)0

These values indeed match the exterior Schwarzschild metric if

a r02 rS rS
g̃00 = B(t, r0 ) = 1 − = ḡ00 = 1 − =1−
f (t) r̃(t, r0 ) r0 f (t)
or
rS = a r03 ,
8π G ε(0)
consistent with, noting that a = 3 ,
4π
rS = 2 G M , with ε(0) r03
M=
3
being the total mass contained in a sphere of radius r0 .
The map φ is well-defined in the domain D , where 0 ≤ t < tS ; fr(t) S
< r ≤ r0 (where
tS is defined by r0 f (tS ) = rS , i.e tS is the time when the dust cloud metric reaches the
Schwarzschild radius, then, see exercises, t̃(tS , r0 ) = ∞ . The image of D is the domain
De , where 0 ≤ t̃ < ∞ ; rS < re ≤ r0 f (t(e e the metric is
t)) , see also Fig.3.3.2 . In the domain D
given by Eqs.(3.3.11) and (3.3.12) , where it is understood, that (t, r) are expressed through
(t̃, r̃) by inverting ϕ . By computation, see exercises, it can then be shown, that indeed
ϕ∗ g = ge holds in the domain D . The astonishing feature of this model for the gravitational
collapse is that for t = tS we have t̃ = ∞ for r̃ = rS , i.e. the dust cloud just disappears
beyond the Schwarzschild radius when the time goes to infinity. However for an observer
comoving with the cloud only a finite time has passed. The unphysical feature of the model
is that for t > tS the exterior Schwarzschild domain and the interior dust cloud domain
are disconnected.
44 CHAPTER 3. SOLUTIONS OF THE EQUATIONS

∞
t t̃
Schwarzschild
domain

ge = g

π
T = 2
√
a
g αβ
interior
domain

tS
f
D

D geαβ

rS r0 r rS r0 r̃

Figure 3.3.2: Domains for the Oppenheimer-Snyder map.

3.4. GRAVITATIONAL WAVES 45

3.4 Gravitational waves

For regions where the energy-momentum tensor vanishes (T µν = 0) according to the Einstein-
equations the curvature tensor (Ricci-tensor should vanish also: Rµν = 0 . The question
then arises if there are solutions which have the character of wave packets propagating
with the velocity of light (c = 1 in our units). In the following xµ shall be coordinates in
Minkowski space R 4 , i.e. for x ∈ R 4 we have x = xµ eµ . For the metric we write

g00 = 1 + f (x) , g01 = g10 = −f (x) , g11 = −1 + f (x) , g22 = g33 = −1 . (3.4.1)

It is now advantageous to make a coordinate transformation to

x0 − x1 x0 + x1
y0 = √ , y1 = √ , y 2 = x2 , y 3 = x3 .
2 2
Then the non-vanishing components of the metric

∂xα ∂xβ
ĝµν = gαβ
∂y µ ∂y ν
are given by
ĝ00 = 2f , ĝ01 = ĝ10 = 1 , ĝ22 = ĝ33 = −1 ,
which has the inverse with the non-vanishing components

ĝ 01 = ĝ 10 = 1 , ĝ 11 = −2f , ĝ 22 = ĝ 33 = −1 .

With the Ansatz f (y) = f (y 0 , y 1 , y 2 , y 3 ) = h(y 0 , y 2 , y 3 ) the non-vanishing Christoffel

symbols are

Γ100 = ∂0 h , Γ102 = Γ120 = ∂2 h , Γ103 = Γ130 = ∂3 h , Γ200 = ∂2 h , Γ300 = ∂3 h

and the single non-vanishing component of the Ricci tensor is then given by

R00 = ∂22 h + ∂32 h .

Thus a solution to the source-free Einstein equations Rµν = 0 is given, if the function f
has the following properties:
 2 
0 1 2 3 ∂ ∂2
f (x) = f (x − x , x , x ) , ∆2 f := + f = 0.
∂x2 2 ∂x3 2
As we shall see this indeed has the character of wave packets, but we first recall some facts
about:

Electromagnetic wave packets in Minkowski-space: An electromagnetic wave packet

in Minkowski-space is described by a vector potential Aµ which is a solution of the homo-
geneous d’Alembert equation Aµ = 0 in Lorenz-gauge ∂µ Aµ = 0 . A possible solution
is provided by
A0 = 0 , A1 = 0 , A2 (x) = f (x0 − x1 ) , A3 = 0 ,
46 CHAPTER 3. SOLUTIONS OF THE EQUATIONS

where f : R → R e.g. has the form:

f

x0 − x1

Then Fµν = ∂µ Aν − ∂ν Aµ will have the form of a pulse

Fµν

x0 − x1

propagating in the e1 -direction with the velocity of light.

Returning to the gravitational waves: The Minkowski metric is disturbed by the function f
and this disturbance moves in Minkowski space with the light velocity in the e1 direction.
This disturbance is modulated in the space directions perpendicular to e1 : A solution of
the equations for f is e.g. provided by
0
−x1 )
f (x0 , x1 , x2 , x3 ) = eiω(x g(x2 , x3 ) ,

where ∆2 g = 0 .
For solutions of this kind (waves propagating in a single constant direction) the superposition
principle holds, i.e if g(f ) and g(f˜) are solutions of the metric, then also g(αf + β f˜) is a
solution for arbitrary α, β ∈ R . But this is not true for solutions with different directions
of propagation, in contrast to electromagnetic waves, where the superposition principle also
holds in this case. The reason is the non-linearity of the Einstein equations.

3.4.1 Gravitational waves as approximate solutions of the Einstein

equations
We take coordinates xµ and write the metric as

gµν = ηµν + hµν ,

where ηµν are the components of the Minkowski metric. We now approximate the Einstein
equations
Gµν = 8π G T µν
by computing Gµν including only upto terms linear in h . One then finds after some com-
putation:
 
1 ∂2 λ ∂2 λ ∂2 λ
Gµν ≈ − h λ− h µ− h ν + hµν
2 ∂xµ ∂xν ∂xλ ∂xν ∂xµ ∂xλ
 
1 ∂2 1
+ λ
h λ− h λδ
ηµν =: G(0) (3.4.2)
2 λ
∂x ∂x δ 2 µν
3.4. GRAVITATIONAL WAVES 47

and the Einstein equation in this approximation, where “raising” and “lowering” of indices
is understood to be performed by η µν , ηµν consistent with this linear approximation, is given
by
G(0)
µν = 16π G Tµν .

(0)
An important observation is that Gµν is invariant under the replacement

∂ ∂
hµν 7→ h̃µν = hµν − Bν − Bµ (3.4.3)
∂xµ ∂xν
which has the form of a gauge transformation6 . gauge
As a consequence one can always impose a gauge fixing condition on hµν : A choice, analo- transfor-
gous to the Lorenz-gauge is the gauge fixing mation

∂ µ 1 ∂ λ
h̃ν = h̃ . (3.4.4)
∂x µ 2 ∂xν λ
Indeed, if for a given h, Bµ is a solution of the inhomogeneous d’Alembert equation

∂ µ 1 ∂ λ
Bν = h ν− h λ
∂xµ 2 ∂xν
then one computes with Eq.(3.4.3)

∂ ˜µ ∂ µ ∂ ∂ ∂2
h ν = h ν − B ν − Bµ
∂xµ ∂xµ ∂xµ ∂xµ ∂xµ ∂xν
| {z }
=Bν
 
∂ µ ∂ µ 1 ∂ ∂
= µ
h ν− µ
h ν+ λ
h λ − 2 λ Bλ
∂x ∂x 2 ∂xν ∂x
1 ∂ λ
= h̃
2 ∂xν λ
and a solution of the inhomogeneous d’Alembert equation always exists, see below.
Now, if Eq.(3.4.4) holds, then one finds
 
∂2 1 ∂2 1 ∂2
G(0)
µν = − h̃ λ
− h̃ λ
− h̃ λ
+ h̃ µν
∂xµ ∂xν λ 2 ∂xν ∂xµ λ 2 ∂xµ ∂xν λ
 
1 ∂ κδ ∂ λ
+ h̃λλ − η h̃ ηµν
2 ∂xδ ∂xκ λ
1
= − h̃µν + ( h̃λλ ) ηµν = 16π G Tµν .
2

If we now define h̄µν := h̃µν − 21 h̃λλ ηµν we find that h̄ in this gauge fulfils the inhomogeneous
d’Alembert equations
h̄µν = −16π G Tµν (3.4.5)
as does the vector potential in electrodynamics

Aµ = 4π Jµ ,

where here J is the charge-current density. Now Eq.(3.4.5) has the general solution

h̄µν = h̄ret. (0)

µν + h̄µν , (3.4.6)

where (in terms of the cartesian coordinates x0 , ~x) , a special solution is provided by
Z
Tµν (x0 − |~x − ~y |, ~y )
h̄ret. 0
x) = −4 G
µν (x , ~ d3y , (3.4.7)
|~x − ~y |
48 CHAPTER 3. SOLUTIONS OF THE EQUATIONS

(0)
the retarded solution and where h̄µν is the general solution of the homogeneous d’Alembert retarded
equation solution
h̄(0)
µν = 0 . (3.4.8)
Now the gauge condition
∂ µ 1 ∂ λ
h̃ = h̃
∂xµ ν 2 ∂xν λ
implies
∂ µ ∂ µ 1 ∂ λ µ 1 ∂ λ 1 ∂ λ
h̄ = h̃ − h̃ η = h̃ − h̃ = 0 .
∂xµ ν ∂xµ ν 2 ∂xµ λ ν 2 ∂xν λ 2 ∂xν λ
Therefore the solution to the homogeneous d’Alembert equation is subject to the gauge
condition
∂ (0) µ
h ν = 0. (3.4.9)
∂xµ
The homogeneous equation (3.4.8) can be solved with the usual plane wave Ansatz
i(k · x)
h̄(0)
µν (x) = aµν e , (3.4.10)

where aµν = const. , (k · x) := ηµν k µ xν and where the 4-vector k = const. must be light-like

(k · k) = ηµν k µ k ν = 0 ,

in order that Eq.(3.4.10) indeed solves the homogeneous equation (3.4.8) .

6 In electrodynamics if A yields a solution to Maxwell’s equations, so does à = A − ∂

µ µ µ ∂xµ
χ with χ
an arbitrary function.
3.4. GRAVITATIONAL WAVES 49

In order that the Ansatz fulfils the gauge condition (3.4.9) we have the condition

aµν kµ = 0 ⇔ aµν k µ = aµν k ν = 0 .

This condition does not fix the gauge completely: We argued above that we can always
impose the condition
∂ µ 1 ∂ λ
h̃ = h̃ .
∂xµ ν 2 ∂xν λ
Now consider the replacement

˜ ∂ ∂
h̃µν 7→ h̃µν = h̃µν − B (0) − B (0) (3.4.11)
∂xµ ν ∂xν µ

where Bµ (x) = bµ ei(k · x) is a solution to the homogeneous d’Alembert-equation. Then

(0)

∂ ˜µ ∂ µ
h̃ = h̃ − η µα kα kµ bν ei(k · x) − kν bµ kµ ei(k · x) ,
∂xµ ν ∂xµ ν | {z }
=0
∂ ˜λ ∂ λ
h̃ = h̃ − η λα kα kν bλ ei(k · x) − kλ bλ kν ei(k · x) ,
∂xν λ ∂xν λ | {z }
=0

and therefore
∂ µ 1 ∂ λ ∂ ˜µ 1 ∂ ˜λ
h̃ν = h̃λ ⇒ h̃ν = h̃ .
∂x µ 2 ∂x ν ∂x µ 2 ∂xν λ
which implies that one can make further gauge transformations of the form (3.4.11) while
preserving the condition of (3.4.9) . This of course reflects the fact, that the solution of
the inhomogeneous d’Alembert-equation is not unique: A solution of the homogeneous
equation can always be added. Under the transformation (3.4.11) we then have
1
h̄µν := h̃µν − h̃λλ ηµν
2  
∂ ∂ 1 λα ∂ (0) ∂ (0) λ
7→ h̃µν − B (0) − (0)
B − ηµν h̃λ − η λ
B − B
∂xµ ν ∂xν µ 2 ∂xα λ ∂xλ
∂ ∂ ∂ λ
= h̄µν − B (0) − B (0) + ηµν B (0) .
∂xµ ν ∂xν µ ∂xλ
and therefore
i(k · x)
h̄(0)
µν = aµν e
 
∂ (0) ∂ (0) 1 λ λα ∂ (0) ∂ (0) λ
7→ h̃µν − B − B − ηµν h̃λ − η B − B
∂xµ ν ∂xν µ 2 ∂xα λ ∂xλ
∂ ∂ ∂ λ
= h̄µν − B (0) − B (0) + ηµν B (0) .
∂xµ ν ∂xν µ ∂xλ
= aµν ei(k · x) − i kµ bν ei(k · x) − i kν bµ ei(k · x) + i ηµν kλ bλ ei(k · x)
(0)
= ãµν ei(k · x) =: h̄
˜ ,
µν

where
ãµν := aµν − i kµ bν − i kν bµ + i ηµν kλ bλ .
Accordingly we can impose the three (so called “transverse-traceless-gauge”) conditions:

ãµν k ν = 0 , ãλλ = 0 , ãµν uν = 0 , (3.4.12)

where u is a time-like vector , i.e. (u · u) > 0 . The argument goes as follows:

First of all

ãµν k ν = aµν k ν − i kµ (k · b) − i (k · k) bµ + i kµ (k · b) = aµν k ν

50 CHAPTER 3. SOLUTIONS OF THE EQUATIONS

and thus
aµν k ν = 0 ⇔ ãµν k ν = 0 .
Then the vanishing of the trace implies

0 = ãλλ = aλλ − 2i (k · b) + i ηκκ (k · b) = aλλ + 2i (k · b) ,

which is one linear equation determining b :

i
(k · b) = − aλλ .
2
Finally

0 = ãµν uν = aµν uν − i kµ (b · u) − i bµ (k · u) + i uµ (k · b)

imply only three other linear equations for b, since the particular linear combination

k µ ãµν uν = k µ aµν uν − i k µ kµ (b · u) − i (k · b) (k · u) + i (k · u) (k · b) = 0 ,
| {z } | {z }
=0 =0

irrespective of b . Thus by an appropriate choice of b the transverse-traceless gauge conditions

(3.4.12) can be fulfilled.
In particular, by introducing, in addition to k , 3 additional vectors: k̄ with


k̄ 0 = k 0 , k̄ i = −k i ⇒ k · k̄ = 2 |~k|2 6= 0 ,

as well as e1 , e2 with the properties

(ej · k) = ej · k̄ = 0 , (j = 1, 2) , (e1 · e1 ) = −1 , (e1 · e2 ) = 0 , (e2 · e2 ) = −1 .


and setting u = k + k̄, we have (u · u) = 2 k · k̄ > 0, thus u is time-like and furthermore

aµν k̄ ν = aµν uν − aµν k ν = 0 (3.4.13)

Eq.(3.4.13) together with Eq.(3.4.9) are then 8 equations for the 10 components aµν (= aνµ ) .
By direct verification, see exercises, one then finds that with arbitrary constants α, β the
general solution is given by



aµν = α e1 µ e2 ν + e2 µ e1 ν + β e1 µ e1 ν − e2 µ e2 ν ,

corresponding to the case of so called tensor polarisation 7 .

7 Note the analogy to the case of the vector potential in electrodynamics, where the plane wave solution

Aµ (x) = aµ ei(k · x) subject to the Lorenz gauge condition leads to the form aµ = α e1µ + β e2µ (upto
pure gauge terms) , with e1 , e2 two independent polarisation vectors, corresponding to the case of vector
polarisation. Indeed, writing in general


a = α e1 + β e2 + γ k + δ k̄ , k · k̄ 6= 0 ,
the Lorenz gauge condition implies (a · k) = 0 and thus


α (k · e1 ) +β (k · e2 ) +γ (k · k) +δ k · k̄ = 0 ⇒ δ=0
| {z } | {z } | {z } | {z }
=0 =0 =0 6=0

and therefore
∂  
Aµ (x) = α e1µ + β e2µ ei(k · x) + γ kµ ei(k · x) = α e1µ + β e2µ ei(k · x) +



µ
−i γei(k · x) .
∂x
3.5. KERR-SOLUTION 51

At this point we like to make a comment on the origin of the gauge transformations in the
linearised Einstein theory discussed above:
The full Einstein equations are invariant under arbitrary diffeomorphisms. Let B µ be
a vector field. This induces a one-parameter family ϕλ of diffeomorphisms with λ ∈ R ,
ϕ0 = Id , via
ϕλ (x0 ) = x(λ; x0 ) ,
where x(λ, x0 ) fulfils the ordinary differential equation
ẋµ = B µ (x)
with the initial condition x(0; x0 ) = x0 . Then, see exercisies,

d 
(ϕλ gµν )
∗
= Bµ;ν + Bν;µ .
dλ λ=0

Writing
gµν = ηµν + hµν
one finds for small λ :
(ϕ∗λ g)µν = gµν + λ (Bµ;ν + Bν;µ ) + O(λ2 ) = ηµν + hµν + λ (Bµ,ν + Bν,µ ) + O(λ2 ) ,
because terms including the Christoffel symbols are of higher order. In this manner
the general invariance under arbitrary diffeomorphisms corresponds to the invariance under
gauge transformations of the linearised equations.

3.5 Kerr-solution
We again start from Minkowski-space (with gµν = ηµν the Minkowski-metric) and now
introduce the generalised “spherical” coordinates ϑ, ϕ, r via gener-
p p alised
x0 = t x1 = r2 + a2 sin ϑ cos ϕ , x2 = r2 + a2 sin ϑ sin ϕ , x3 = r cos ϑ . “spheri-
cal”
The corresponding metric is diagonal and one finds for ds2 = gµν dxµ ⊗ dxν 8
:
coordi-
ρ2 nates
ds2 = dt2 − dr2 − ρ2 dϑ2 − (r2 + a2 ) sin2 ϑ dϕ2 , ρ2 := r2 + a2 cos2 ϑ .
r 2 + a2
The Kerr-solution modifies this to the form Kerr-
    solution
2 rg r 2 ρ2 2 rg r 2
ds = 1 − 2 dt − dr − ρ dϑ − r + a + 2 a sin ϑ sin2 ϑ dϕ2 ,
2 2 2 2 2
ρ ∆ ρ
2
rg r a sin ϑ
+ (dt ⊗ dϕ + dϕ ⊗ dt) , ∆ := r2 − rg r + a2 . (3.5.1)
ρ2
With this metric the Einstein-tensor vanishes: Gµν = 0 . The metric has singularities at
r r
rg rg2 2 2
rg rg2
ra = + − a cos ϑ , rb = + − a2
2 4 2 4
and accordingly the parameters should be such that rb ≤ ra ≤ rg , which defines the domain
in which this (exterior) solution holds. Concerning the interpretation of the parameters a
and rg one can state that rg = 2 G M , with M the mass of the field-producing object, is
analogous to the Schwarzschild radius rS = 2 G M and that L = M a is the magnitude
of the angular momentum of this object. With the Jacobi method it is possible to find the
trajectories of test particles. Finally, for a = 0 one recovers the Schwarzschild metric and
with additionally rg = 0 one obtains the flat-space metric.
8 It is understood that specifying the (infinitesimal) “line element”
ds2 = gµν dxµ ⊗ dxν
is sufficient to read off the non-vanishing components gµν of the metric.
52 CHAPTER 3. SOLUTIONS OF THE EQUATIONS
Chapter 4

Cosmology

In this chapter we shall discuss some cosmological issues, restricted to those aspects which
are related to general relativity. The main ingredients that we shall need from general
relativity are

1. The space time metric;

2. The behaviour of electromagnetic waves.

4.1 Space-time metric

Concerning the space-time metric we encounter the difficulty to completely specify the neces-
sary input data; in general only measurements at particular space-time points are available.
To proceed nevertheless we shall use comoving coordinates and accordingly write the metric
in the symbolical short-hand notation introduced before as

g ≡ ds2 = dt2 − g3 ,

where g3 represents the metric in ordinary 3-space M3 . We postulate the so called

cosmological principle which states that cosmo-
logical
“points in 3-space transform into each other according to symmetry transforma- principle
tions leaving g3 invariant.”

Candidates for M3 are

(a)  
M3 = R 3 with g3 = (dx1 )2 + (dx2 )2 + (dx3 )2 R(t)2
and the symmetry group is SO(3)×R 3 , the group of spatial rotations and translations;

(b)
 
M3 = S3 (3-sphere) with g3 = (dx1 )2 + (dx2 )2 + (dx3 )2 + (dx4 )2 R(t)2 ,

with
(x1 )2 + (x2 )2 + (x3 )2 + (x4 )2 = 1
and the symmetry group is SO(4) ;

(c)  
M 3 = H3 with g3 = (dx1 )2 + (dx2 )2 + (dx3 )2 − (dx4 )2 R(t)2 ,
with
(x1 )2 + (x2 )2 + (x3 )2 − (x4 )2 = −1
and the symmetry group is the Lorentz-group.

53
54 CHAPTER 4. COSMOLOGY

Because the symmetry groups of all three candidates contain the rotation group SO(3)
as subgroup we shall use spherical polar coordinates x1 = r, x2 = ϑ, x3 = ϕ . Then the
metric takes, see also the discussion of the dust could in comoving coordinates, the generic
(diagonal) form, called Robertson-Walker metric :
  Robertson-
1 Walker
g = dt2 − R(t)2 dr 2
+ r 2
dϑ 2
+ r 2
sin 2
ϑ dϕ 2
,
1 − k r2 metric
where k = 0, 1 and −1 correspond to the candidates (a), (b) and (c), respectively.
For the energy-momentum tensor we take the form:

T µν = (ε + p) γ µ γ ν − p g µν

where γ 0 = 1, γ 1 = γ 2 = γ 3 = 0 , the metric is given in comoving coordinates and the energy

density ε(t) and the pressure p(t) are functions of the time t alone.
For the energy-momentum tensor T µν ;µ = 0 holds, which implies

d  d
ε(t) R(t)3 = −p(t) R(t)3 , ⇒ dE = −p dV (4.1.1)
dt | {z } dt | {z }
∼Energy ∼Volume

and inserting the Ansatz in the Einstein equations, see exercises, gives
2 8π G
Ṙ(t) + k = ε(t) R(t)2 , (4.1.2)
3
i.e. 2 equations for the 3 unknown functions ε(t), p(t), R(t) . In order to determine these
functions one thus also needs an additional relation

p(t) = p(ε(t)) , (4.1.3)

the equation of state. The latter must be provided by additional theoretical considerations
(statistical mechanics) that describe the material content of the universe. In general this can
be formulated only if the material content of the universe is dominated by a simple (single)
species of matter.

The astronomical observation is, that all visible matter consists mainly of hydrogen and he-
lium, where the mass fraction of 4 He is ≈ 25% and that the contribution of heavier elements
(in cosmology generically called “metals”) is negligible. In addition there is microwave
background radiation, with a black body radiation spectrum characterised by a very low
temperature (2.7 K), which corresponds to very many photons of very low energy, such that
the contribution to the energy density can also be neglected. Furthermore this photon dis-
tribution is highly isotropic and homogeneous in sharp contrast to the matter distribution,
for which a large scale structures in the universe is found.
At this point we make the conjecture, that in the past the photon and matter distributions
were the same (and hence homogeneous and isotropic) such that the Ansatz for the metric
made above can still be used and where it is supposed that the large scale structure found for
the matter distribution can be derived from small fluctuations of the microwave background.

4.2 The behaviour of electromagnetic waves

The first question we address is how the space-time metric influences photon fields and,
conversely, if and how photons can be considered to be suitable test particles for the metric.
The photon field is a solution of the Maxwell equations for vanishing charge and current
density:

dF = 0, ⇔ Fµν,α − Fαν,µ + Fαµ,ν = 0 , (4.2.1)

µν
F ;µ = 0. (4.2.2)
4.2. THE BEHAVIOUR OF ELECTROMAGNETIC WAVES 55

In order to determine the time dependence of the field strength tensor Fµν we make a change
of coordinates, write the metric as
 
1
g = R(t)2 dt 2
− g 3
R(t)2

and take a new time coordinate τ such that

1
dt = dτ , ⇒ R(t) = R(τ (t)) .
R(t)

Then the equation

1 ∂ p
F µν ;µ = p | det g| g µα g νβ Fαβ = 0
| det g| ∂xµ

implies
∂ p
| det g| g µα g νβ Fαβ = 0
∂xµ
and with the metric given in terms of the coordinate τ no further τ dependence is present
in the Maxwell equations. Accordingly, there are solutions of the Maxwell equations of
the form
(0)
Fµν = ei ω τ Fµν + c.c. ,
(0)
where Fµν is independent of the coordinate τ , which is called the conformal time. confor-
Now, mapping back to the original time coordinate t we have as solutions mal
time
(0) dτ (0)
F0ν = ei ω τ (t) F0ν (t) + c.c. , Fjk = ei ω τ (t) Fjk + c.c. ,
dt
where
dτ 1
(t) = .
dt R(t)
Below we shall now discuss the frequency dependence and the behaviour of the energy
density of the radiation fields.

Frequency: The observed frequency of such radiation fields at time t is given by the
behaviour of Fµν for small time changes ∆t of t at a given (fixed) time t :
(0)
Fjk (t + ∆t) = ei ω τ (t+∆t) Fjk + c.c. ,
ei ω [τ (t)+ dt (t)∆t] Fjk + c.c. ,
dτ (0)
≈
∆t (0)
= ei ω R(t) ei ω τ (t) Fjk +c.c.
| {z }
const.

and therefore the observed frequency at time t is given by

ω
ωobs (t) = .
R(t)
 56 CHAPTER 4. COSMOLOGY

Now suppose that the electromagnetic wave was emitted at t = tem and that the same
wave is observed (absorbed) at t = tabs . Then the deviation between ωem = R(tωem ) and
redshift ωabs = R(tωabs ) is measured in terms of the redshift, defined by
ωem R(tabs )
z := −1 ⇒ z= − 1. (4.2.3)
ωabs R(tem )
If we assume that
tem = tabs − d ,
where d is the distance between the absorber and the emitter and d ≪ tabs then we find,
d
expanding in the small parameter tabs :
R(tabs ) R(tabs )
z = −1= −1
R(tabs − d) R(tabs ) − Ṙ(tabs ) d + 21 R̈(tabs ) d2 + O(d3 )
Ṙ(tabs ) Ṙ(tabs )2 2 1 R̈(tabs ) 2
≈ 1+ d+ d − d −1
R(tabs ) R(tabs )2 2 R(tabs )
Ṙ(tabs ) Ṙ(tabs )2 2 1 R̈(tabs ) 2
= d+ d − d .
R(tabs ) R(tabs )2 2 R(tabs )
Denoting the present day values of R, Ṙ and R̈ by R0 , Ṙ0 and R̈0 , respectively, we find in
linear approximation a linear relationship between the redshift and the distance between the
Hubble’s emitter and absorber which is Hubble’s law:
law
z = H0 d
Ṙ0
Hubble with H0 := R 0
the (present day value of the) Hubble constant, which therefore in cos-
constant mology can be used as a constant of integration in the Einstein equations.
In quadratic approximation one would write for the red shift
 q0  2 2
z = H0 d + 1 + H0 d ,
2
where
R̈(tabs ) 1
q0 := −
R(tabs ) H02
decelera- is called the de-acceleration or deceleration parameter , since, as we shall argue below,
tion the basic equations for cosmology as presented here upto now imply R̈ < 0 :
parameter
We recall the Einstein-equation in the form
2 8π G
Ṙ(t) + k = ε(t) R(t)2 , (4.1.2)
3
as well as the relation (from T µν ;µ = 0)
d
d
ε(t) R(t)3 = −p(t) R(t)3 , (dE = −p dV ) . (4.1.1)
dt dt
Differentiating the first w.r.t. to t and using the latter gives
  d
8π G d R3 8π G dt R3 8π G Ṙ
2 Ṙ R̈ = ε =− p − ε R3 2
3 dt R 3 R 3 R
 
8π
= − 8π G p + G ε R Ṙ
3
and we obtain, for Ṙ 6= 0 :
 
1 8π
R̈ = − 8π G p + Gε R
2 3
and thus conclude (assuming ε > 0 and p > 0) , that R̈ is negative and therefore Einstein’s
equations predict that the deceleration parameter q 0 is positive 1 .
1 Note that this conclusion holds only if we indeed assume that the cosmological constant vanishes, Λ = 0,

we shall discuss the implications of Λ 6= 0 later.

4.2. THE BEHAVIOUR OF ELECTROMAGNETIC WAVES 57

Energy density: From the solutions found above

ei ω τ (t) (0) (0)

F0ν = F + c.c. , Fjk = ei ω τ (t) Fjk + c.c. ,
R(t) 0ν

one can compute the energy-momentum tensor T µν [F ] . In particular one finds for the
energy density T 00 = ε of an electromagnetic wave, see also exercises, the dependence:
1
ε∝ .
R(t)4
 58 CHAPTER 4. COSMOLOGY

We now apply the results found to the observed cosmic microwave background (CMB)
radiation, looking back at R(t) at remote times. As indicated above it is observed that the
spectral density ρ(ω) of the low energy radiation can be very well described by a Planck
distribution for black body radiation
~ ω3
ρ(ω) = , (4.2.4)
π 2 c3 e k~ω
BT − 1

Wien’s law which has a maximum at ~ω̄ ≈ 2.82 kB T according to Wien’s law, where kB is the Boltz-
mann constant, ~ is the Planck constant and T is the effective temperature, which for
the cosmic microwave background (CMB) is approximately given by T = 2.7 K . Here ω
are the frequencies observed at the time tabs and these, as we found above, differ from the
frequencies ωem . Now for black body radiation the total energy density (i.e. integrated over
all frequencies) of the photon field is related to the temperature as

εph = σ T 4 ,
π 2 kB
4
where σ = 60 ~3 c2 is the Stefan-Boltzmann constant. On the other hand we stated above
that
1
ε∝
R(t)4
and we therefore conclude that T ∝ R(t)−1 .
As mentioned above the universe consists mainly of hydrogen (1 H) and helium (4 He) oc-
curring in galaxies for which we can make the reasonable approximation that the pressure
vanishes, p = 0 . We then find
8π G
Ṙ(t)2 + k = ε(t) R(t)2
3
as well as (p = 0)
d

ε(t) R(t)3 = 0
dt
and from the latter we infer ε(t) R(t)3 = ε(t0 ) R(t0 )3 or (t0 being the present time)
a
ε(t) = , a = ε(t0 ) R(t0 )3 .
R(t)3
Inserting this in the first equation we find
r
8π G a
Ṙ(t) = −k,
3 R
where we fixed the sign R(t) > 0 from observation. Accordingly, R(t) is an increasing
function of time and thus at earlier times R(t) was smaller. The observation of a microwave
background radiation indicates in conjunction with the considerations above, that at much
earlier times a photon gas dominated the energy density, since T ∝ R(t)−1 . For a(n) (ideal)
photon gas, however, we have p(ε) = 31 εph and thus

d
1 d
εph (t) R(t)3 = − εph (t) R(t)3 ,
dt 3 dt
from which
a
εph (t) =
R(t)4
follows.
In terms of the wave vector k for electromagnetic radiation with k 0 = |~k| = ω (because
kµ k µ = 0) the spectral density is given by (from now on we again put ~ = c = 1) :

1 |~k|
ρ(~k) = ,
4π 3 |~
k|
e kB T
−1
4.3. COSMOLOGICAL EVOLUTION, THE FRW UNIVERSE 59

which indeed after integration over the direction of ~k, which gives a factor 4π |~k|2 and the
substitution |~k| 7→ ω gives back the Planck distribution of Eq.(4.2.4).
For an observer not at rest with respect to the photon gas this distribution will look different:
Moving with a velocity ~v such an observer will measure a frequency different from ω :

ω̂ = (v · k) ,
p
where we introduced the 4-vector v with v 0 = 1 + |~v |2 , according to the Doppler effect. Doppler
In linear approximation in ~v we have v 0 ≈ 1 and then effect
  
  ~v · ~k
ω̂ = |~k| − ~v · ~k = ω 1 − .
|~k|

Therefore the spectral distribution for an observer moving with a velocity ~v will be charac-
terised by a different, effective temperature

T
T̂ (ϑ) = ,
1 − |~v | cos ϑ

where ϑ = ∠(~v , ~k) . For small velocities ~v ≪ 1 we then have in linear approximation
T̂ = T (1 + |~v | cos ϑ) .
More generally the angular dependence of T can be expanded in terms of spherical harmonics
as
∞ X
X ℓ
T̂ (ϑ, ϕ) = T cℓ m Yℓ m (ϑ, ϕ)
ℓ=0 m=−ℓ

and from the anisotropy of the CMB, measured for all polar angles astronomers claim to
be able to determine the coefficients up to ℓ = 600 from the measurement of T̂ for all polar
angles. The (major) dipole deviation discussed above can be ascribed to the effect that the
observer is not comoving. Indeed this is the case for us in our galaxy, who move with respect
to the CMB. This effect has to be subtracted from the CMB distribution before conclusions
concerning the angular dependence of the fluctuations in the CMB can be inferred.
After all possible fits to data modern cosmology seems to favour a non-vanishing cosmological
constant Λ. Recall that up to now we used for the action for the gravitational field:
Z p
1
Sgrav = − d4x |g|(x) R(x) ,
16π G

where R is the Ricci or curvature scalar. According to Hilbert the most general expression
was given by Z
1 p
Sgrav = − d4x |g|(x) (R(x) + Λ) ,
16π G
and then the occurrence of a cosmological constant seems natural.

4.3 Cosmological evolution, the Friedmann-Robertson-

Walker universe
We recall the basic equations of cosmology: From T µν ;µ = 0 we found

d  d
ε(t) R(t)3 = −p(t) R(t)3 , (4.1.1)
dt dt
and insertion in the Einstein equations yielded

2 8π G
Ṙ(t) + k = ε(t) R(t)2 . (4.1.2)
3
60 CHAPTER 4. COSMOLOGY

Ṙ(t)
With H(t) := R(t) we write these in the form:

ε̇(t) = −3 H(t) (ε(t) + p(t)) , (4.3.1)

k 8π
H(t)2 + = G ε(t) . (4.3.2)
R(t)2 3
In order to determine ε(t), p(t), R(t) one also has to make an assumption about the equation-
of-state, p(t) = p(ε(t)) for which we shall assume the linear relationship:

p(t) = λ ε(t) . (4.3.3)

We briefly discuss the following special cases:

1. λ = 0 : This is the assumption p = 0 we invoked for “dust”. Physically it should de-
scribe weakly interacting (ideally free) non-relativistic particles, where the rest energy
(rest mass) dominates the energy density and the pressure (almost) vanishes;
2. λ = 13 : This corresponds to the equation-of-state for ultra-relativistic particles (e.g.
photons) where the rest energy (rest mass) can be neglected;
3. λ = −1 : This corresponds to the situation with a non-vanishing cosmological constant:
Indeed, inserting ε̂ = −p̂ in the general expression we used for the energy-momentum
tensor: T µν = (ε + p) γ µ γ ν − p g µν yields in this case a contribution

T̂ µν = ε̂ g µν .

Adding such a contribution to the energy-momentum tensor on the right-hand side of

the Einstein-equations gives
1
Rµν (x) − R(x) gµν (x) = 8π G T µν (x) + 8π G ε̂ gµν (x)
2
or
1
Rµν (x) − (R(x) + 16π G ε̂) gµν (x) = 8π G T µν (x) .
2
Thus adding a contribution to the energy-momentum tensor which corresponds to an
equation of state with p̂ = −ε̂ = const. is equivalent to shifting the Ricci-scalar R(x)
by a cosmological constant Λ = 16π G ε̂ as was done in the gravitational contribution
to the action, briefly referred to at the end of the preceeding section;
4. λ = − 31 : Although there is no known form of matter with such an equation-of-state,
this value is decisive for R̈ : We recall what we found in the preceeding paragraph:

R̈ 4π G
=− (ε + 3 p)
R 3
and observe, that if λ < − 13 , corresponding to p < − 13 ε , we find that the expansion
of the universe is accelerating, whereas for λ > − 13 , corresponding to p > − 13 ε , it is
decelerating.
The evolution of the energy density ε(t) is related to the evolution of the scale function R(t)
by Eq.(4.3.1), which, with p = λ ε , takes the form:

Ṙ(t) ε̇(t) Ṙ(t)

ε̇(t) = −3 (1 + λ) ε(t) ⇔ = −3 (1 + λ)
R(t) ε(t) R(t)
from which we conclude:
ε(t) ∝ R(t)−3(1+λ) .
Indeed for some of the special cases, discussed above:
1. λ = 0 : ε(t) ∝ R(t)−3 as already discussed in the preceeding section; this corresponds
to the dilution of particles due to the expansion of volume;
4.3. COSMOLOGICAL EVOLUTION, THE FRW UNIVERSE 61

2. λ = 13 : ε(t) ∝ R(t)−4 , also discussed in the preceeding section; here the extra factor
R(t)−1 accounts for the red-shift of the energy;

3. λ = −1 : This implies that the energy density is constant during the expansion, the
increase in energy being related to the fact that this expansion takes place against a
negative (sic) pressure .

We now study the other basic cosmological equation (4.3.2) and write it in the form

k
= H(t)2 (Ω(t) − 1)
R(t)2

with the definitions

ε(t) 3 H(t)2
Ω(t) := and εc (t) := .
εc (t) 8π G
Note that indeed:
 
k ε(t) 8π G 8π
= H(t)2 − 1 = G ε(t) − H(t)2 .
R(t)2 3 H(t) 2 3

In principle Ω0 := Ω(t0 ) can be evaluated from a determination of the total energy density
ε(t0 ) combined with a determination of εc := εc (t0 ) e.g. by the value of H0 = H(t0 ) . For
the present day value (i.e. t = t0 ) of the so called critical density we quote:

εc := εc (t0 ) ≈ 1.88 h2 10−26 kg m−3 ≈ 11 h2 GeV m−3 , h = 0.673 ± 0.012 .

Present observations indicate that Ω0 ≈ 1 . Note that

• Ω > 1 implies k = 1 ,i.e. M3 ≡ S3 ;

• Ω = 1 corresponds to a flat universe. Apparently our universe is nearly flat.

• Ω < 1 implies k = −1 ,i.e. M3 ≡ H3 ;

We now shall study the evolution of Ω(t) by assuming that to the energy density of the
universe in fact all of the first three kinds of energy density discussed above, i.e. the cases
“R” with λ = 31 (ultra-relativistic matter, e.g. photons), “M ” with λ = 0 (ordinary baryonic
matter, but also (so called “cold” dark matter)) and “Λ” with λ = −1 (corresponding to a
non-vanishing cosmological constant) contribute. Accordingly we write for the total energy
density

ε(t) = εR (t) + εM (t) + εΛ (t)

 4  3 !
R0 R0
= εc ΩR + ΩM + ΩΛ , (4.3.4)
R(t) R(t)

where we inserted the scale function dependence with ΩR , ΩM , ΩΛ representing the present
day values, i.e. for t = t0 . Again we note, that in an expanding universe the early uni-
verse was “radiation dominated”, while the future fate of the universe is determined by the
cosmological constant, or, if this vanishes (Λ = 0 , ΩΛ = 0) by the non-relativistic matter.
 62 CHAPTER 4. COSMOLOGY

We recall the so-called Friedmann-equation in the form

Friedmann-
equation k
= H(t)2 (Ω(t) − 1) , (4.3.5)
R(t)2
which evaluated at the present time t = t0 yields
k
= H02 (Ω0 − 1) , Ω0 = ΩR + ΩM + ΩΛ .
R02

Accordingly
H02 R02
Ω(t) − 1 = (Ω0 − 1) . (4.3.6)
H(t)2 R(t)2
ε(t)
Furthermore, from the definition of the critical density, the definition Ω(t) = εc (t) and the
relation in Eq.(4.3.4) we find

H02 εc ε(t) εc Ω(t)

= = =  4  3 ,
H(t)2 εc (t) εc (t) ε(t) R0 R0
ΩR R(t) + ΩM R(t) + ΩΛ

which upon insertion in Eq.(4.3.6) yields the equation determining Ω(t) :

 2
Ω(t) R0
Ω(t) − 1 =  4  3 (Ω0 − 1)
R0 R0 R(t)
ΩR R(t) + ΩM R(t) + ΩΛ

with the solution

Ω0 − 1
Ω(t) − 1 =  2    2 . (4.3.7)
R0 R0 R(t)
1 − Ω0 + ΩR R(t) + ΩM R(t) + ΩΛ R0

Present observations apparently imply that Ω0 ≈ 1 . Before we discuss the implications of

Ω0 = 1 + δ with |δ| ≪ 1 in some detail below, we first investigate what this means for the
scale function R(t) : According to the Friedmann equation (4.3.5) Ω0 = 1 implies k = 0 ,
i.e. our universe is (nearly) flat. For k = 0 Eq.(4.3.2) reads
8π
H(t)2 = G ε(t) , (4.3.8)
3
Differentiating w.r.t. t we find with Eq.(4.3.1) and the equation-of-state in the form p(t) =
λ ε(t) :
 
8π 8π 8π H(t)2
2 H(t) Ḣ(t) = G ε̇(t) = G [−3 H(t) (1 + λ) ε(t)] = G −3 H(t) (1 + λ) 8π
3 3 3 3 G
= −3 H(t)2 (1 + λ) ,

where in the next-to-last step we used Eq.(4.3.8) again. We thus have found

Ḣ(t) 3
2
= − (1 + λ) .
H(t) 2
For λ 6= −1 this differential equation has the solution

Ṙ(t) 2
H(t) := = , (4.3.9)
R(t) 3(1 + λ) t
where we fixed the constant of integration by the condition
1
lim = 0,
t→0 H(t)
4.3. COSMOLOGICAL EVOLUTION, THE FRW UNIVERSE 63

while for λ = −1 we obtain the solution

Ṙ(t)
H(t) := = β = const. (4.3.10)
R(t)

Solving Eq.(4.3.9) (i.e. λ 6= −1) for the scale function one finds
  3(1+λ)
2
t
R(t) = R̃ ,
t̃
2
(see also the exercises where the cases λ = 0 and λ = 31 corresponding to R(t) ∝ t 3 and
1
R(t) ∝ t 2 , respectively are discussed) , while for λ = −1 the solution to Eq.(4.3.10) is

R(t) = R̃ eβ(t−t̃) ,

where R̃ = R(t̃) . The latter shows that if ΩΛ dominates, that then an exponential growth
of the scale function R(t) is possible. Note that in this case the deceleration parameter
R̈ 1
q := − R H 2 is in fact negative and thus actually corresponds to the case of accelerated
expansion.
To conclude we list some remarks concerning the formula found for the evolution of the
density in terms of the critical density found above:
Ω0 − 1
Ω(t) − 1 =  2    2 . (4.3.7)
R0 R0 R(t)
1 − Ω0 + ΩR R(t) + ΩM R(t) + ΩΛ R0

(i ) If for the present day value Ω0 = 1 holds exactly, then Ω(t) = 1 always;

(ii ) If the present day value Ω0 is close to unity: Ω0 = 1 + δ with some relative precision
R2
|δ| ≪ 1 , then, because for t ≪ t0 the term ΩR R(t)0 2 will dominate the denominator
for t ≪ t0 the value of Ω(t) for t ≪ t0 must have been close to unity to an even much
higher precision. This observation is known as the “flatness problem” . flatness
problem
(iii ) If ΩΛ = 0 and Ω0 = 1 − δ , (δ > 0) , then the denominator in Eq.(4.3.7) for t = t0 (of
course!) is unity (Ω0 = ΩR + ΩM ) and for t > t0 in case of expansion (Ṙ > 0) the
denominator will approach 1 − Ω0 and accordingly for t → ∞ the function Ω(t) will
approach 0 as
ΩM R0
Ω(t) ∼ ;
1 − Ω0 R(t)

(iv ) If ΩΛ = 0 for Ω0 = 1 + δ , (δ > 0) , the denominator starts at unity as in (iii )

 2  
R0 R0
but now, because 1 − Ω0 = −δ < 0 and the term ΩR R(t) + ΩM R(t) is a
decreasing function of t the denominator, although positive for t = t0 will vanish at
some t = tS > t0 . This corresponds to H(tS ) = 0 which means that at this point the
universe stops expanding and will start to collapse.

(v ) If ΩΛ 6= 0 then in the future in an expanding universe Ω(t) will approach unity.

62 CHAPTER 4. COSMOLOGY
Chapter 5

Extensions

As a prototype for unification of gravitation and electromagnetism we shall briefly discuss

the Kaluza-Klein theory of gravitation and electromagnetism, which should serve as an Kaluza-
example for theories in higher-dimensional space-times. Klein
theory

5.1 Kaluza-Klein theory

Let M 5 = M × R or M 5 = M × S 1 , where M is space-time and S 1 is the unit circle.
The coordinates for M are as usual (x0 , x1 , x2 , x3 ) and those for M 5 are (x0 , x1 , x2 , x3 , x4 ) .
Furthermore, let gµν (µ, ν ∈ {0, 1, 2, 3}) be the metrical tensor for M and we define a new
metrical tensor γ for M 5 by the components γ44 = −1 , γµν = gµν − Bµ Bν , γµ4 = γ4µ =
−Bµ . This metric admits additional coordinate transformations leaving this form invariant,
namely:
x4 7→ x4 + λ(x) ,

along with
∂
Bµ 7→ B̃µ = Bµ − λ,
∂xµ
which has the form of a gauge transformation in electrodynamics.
For a physical theory unifying gravitation and electromagnetism this suggests that Bµ ∼ Aµ .

5.1.1 Geodesics in M 5
The geodesic equation in this 5-dimensional world reads
 
d
1 ∂
γαβ ẋβ = γβδ ẋβ ẋδ , α, β, δ ∈ {0, 1, 2, 3, 4} .
dτ 2 ∂xα

Since γαβ per definition does not depend on x4 we have

d

α=4: ẋ4 + Bµ ẋµ = 0 ⇒ ẋ4 + Bµ ẋµ = c = const. .
dτ

For α = µ = 0, 1, 2, 3 we have
   
d
1 ∂ 1 ∂
gµν ẋν − Bµ Bν ẋν − Bµ ẋ4 = λ ν
gλν ẋ ẋ − (Bλ Bν ) ẋλ ẋν
dτ 2 ∂xµ 2 ∂xµ
   
1 ∂ λ 4 1 ∂
− Bλ ẋ ẋ − Bν ẋ4 xν
2 ∂xµ 2 ∂xµ

63
 64 CHAPTER 5. EXTENSIONS

or
 
d ∂
d

(gµν ẋν ) − λ
Bµ ẋλ ẋ4 + Bν ẋν −Bµ ẋ4 + Bν ẋν
dτ ∂x | {z } |dτ {z }
=c
=0
   
1 ∂ ∂

= µ
gλν ẋλ ẋν − µ
Bλ ẋλ ẋ4 + Bν ẋν ,
2 ∂x ∂x | {z }
=c

from which  
d 1 ∂
(gµν ẋν ) = gλν ẋλ ẋν − c fµν ẋν
dτ 2 ∂xµ
with
∂ ∂
Bν −
fµν := Bµ
∂xµ ∂xν
follows. Raising one index we then obtain

ẍµ + Γµλν ẋλ ẋν = −c f µ ν ẋν

and if we put
q 1 1
c=− , F µν = f µν
m λ λ
we obtain a differential equation for the trajectories which has the form of the equation of
motion for a point particle with mass m and charge q in an electromagnetic field Fµν :
q µ ν
ẍµ + Γµλν ẋλ ẋν = F ν ẋ .
m
q
Thus, if this interpretation holds, then the ratio m appears as an integration constant in
this theory. We still have to interpret the constant λ . To this end we inspect the Ricci
scalar and the action. First of all

det [γ] = −1 · det [g] = − det [g]

and the volume element for integration thus is

p p
| det [γ] d5 x = |g| d4 x dx4 .

For the Ricci scalar one computes

1
R[γ] = R[g] + f µ ν f ν µ
4
and therefore
Z Z Z  
1 5
p 1 4 4
p λ2
− dx |g| R[γ] = − dx dx |g| R[g] + F µ ν F ν µ ,
16π G 16π G 4

which would correspond to the field terms in the action

 Z Z 
1 4
p 1 4
p αβ
− d x |g| R[g] − d x |g| Fαβ F ,
16π G 16π
√
provided that λ = 2 G (and the integration over x4 is finite) . With this identification we
have that Bµ and the electromagnetic potential Aµ are related by

1 1
Bµ = Aµ = √ Aµ .
λ 2 G

Klein- The quantum mechanical Klein-Gordon equation in M 5 for a particle of mass m :

Gordon
equation
5.1. KALUZA-KLEIN THEORY 65
!
1 ∂ p ∂
p |γ| γ αβ β + m2 Φ=0
|γ| ∂xα ∂x

for µ, ν ∈ {0, 1, 2, 3} reads explicitly

"      2 #
1 ∂ ∂ µν
p ∂ ∂ ∂ 2
p − Bα 4 g |g| − Bβ 4 − + m Φ = 0.
|g| ∂xµ ∂x ∂xν ∂x ∂x4

If we now assume
√ ∂
−2 G 4 Φ = −i q Φ ,
∂x
we obtain
"     #
1 ∂ µν
p ∂ q2 2
p − i q Aµ g |g| − i q Aβ + + m Φ = 0,
|g| ∂xµ ∂xν 4G

which has the form of the equation of motion of a Klein-Gordon particle of charge q
coupled to the electromagnetic field according to the principle of “minimal coupling” or
“minimal substitution” but with a mass changed according to
q2
m2 7→ m2 + .
4G
If we attach to the four-dimensional space-time a circle, i.e. M 5 = M × S 1 then with
4 ∂
Φ(x4 ) ∝ e2πi k x , ⇒ Φ = 2π k Φ , k∈Z
∂x4
we can easily explain why charge is quantised in the Kaluza-Klein theory:
√
q = 4π G k =: k e0 , k ∈ Z .

Unfortunately then, as indicated above the mass is changed according to

k 2 e20
m2 7→ m2 +
4G
and thus is quantised as well!
In order to formulate possible generalisations of the Kaluza-Klein theory we briefly sketch
the general structure of this theory in terms of projections: The 5-dimensional manifold is
given as P = M × S 1 . The group S 1 is acting as a symmetry group on P

g ∈ S1 → Rg (x, z) = (x, g z) ,

where x ∈ M and z ∈ S 1 . Then there is a natural projection π : P → M , where in addition

π ◦ Rg = π and Rg π −1 (x) is generated by the S 1 group, where x ∈ M . This is the simplest
example of a principle bundle with the group S 1 over M . S 1 is a Lie-group, which has one
∂ ∂
infinitesimal generator, namely ∂ϕ = ∂x 4 . A bilinear metric on the space of infinitesimal

generators on S 1 , which is invariant under the group action on S 1 itself is then given by
 
∂ ∂
g̃ , = −1 .
∂ϕ ∂ϕ
A metric of the Kaluza-Klein type arises by requiring that the metrical tensor has the
properties
1. it is invariant under Rg ;
∂
2. the vector field X generated on P by the infinitesimal generators ∂ϕ satisfies
 
µ ν ∂ ∂
γµν X X = g̃ , = −1 .
∂ϕ ∂ϕ
 66 CHAPTER 5. EXTENSIONS

5.2 Tetrads
The calculation of the Christoffel-symbols Γ and the various tensors R, R and G is rather
cumbersome. A slightly more systematic treatment uses the concept of tetrads:
tetrad A tetrad (also called Vierbein (german) or repère mobile (french)) consists of four 1-forms
θi , i = 0, 1, 2, 3, each with components θµi , µ = 0, 1, 2, 3 , with the property

gµν = ηik θµi θνk , (5.2.1)

where the signature of η is (+ − −−) as for the Minkowski-metric. For the covariant
derivative we can write
i i
θµ;ν = −ωkν θµk , (5.2.2)
i
where the coefficients ωkν are uniquely determined, since the tetrads θk form a basis of all
1-forms due to (5.2.1) .
These coefficients have the following properties:

(a) According to the definition of the covariant derivative and Eq.(5.2.2) we have

i ∂ i
θµ;ν = θ − Γα i i k
µν θα = −ωkν θµ . (5.2.3)
∂xν µ
Then, using Γα α
µν = Γνµ we find

∂ i ∂ i i
θ − θ + ωkν θµk − ωkµ
i
θνk = 0 . (5.2.4)
∂xν µ ∂xµ ν

(b) For gµν;α = 0 then, taking the covariant derivative of Eq.(5.2.1) we find
i
ηik θµ;α θνk + ηik θν;α
i
θµk = 0 , (5.2.5)

which, together with the property (5.2.3) then leads to

l
ωik µ + ωki µ = 0 , where ωik µ := ηli ωkµ . (5.2.6)
i
This means that for fixed µ the matrix with elements ωkµ , i, k ∈ {0, 1, 2, 3} is an
element of the Lie-algebra (or: is a generator) of the Lorentz-group.

To simplify the notation we introduce the vector-valued 1-form θ with components

 0 
θµ
 θµ1 
θµ ≡ 
 θµ2 


θµ3
i
as well as the Lie-algebra-valued ωµ with matrix elements ωkµ . Then Eq.(5.2.4) can be
written as
∂ ∂
ν
θµ − θ ν + ων θ µ − ωµ θ ν = 0 ,
∂x ∂xµ
or, even more concisely, as
dθ + ω ∧ θ = 0 , (5.2.7)
fundamen- which is the so-called first structure equation. Here θ is called the fundamental form, ω
tal is called the connection form and
form
connection Θ = dθ + ω ∧ θ
form
torsion is called the torsion form . Obviously Eq.(5.2.7) means that the torsion form vanishes.
form
5.2. TETRADS 67

The curvature form, which is a 2-form with values in the Lie-algebra of the Lorentz- curvature
group, is defined by the equation: form
∂ ∂
Ωµν = µ
ων − ωµ + [ωµ , ων ]− or Ω = dω + ω ∧ ω . (5.2.8)
∂x ∂xν
This name is chosen, because the following identity for the components of the curvature
tensor holds:
Rαβµν = η (θα , Ωµν θβ ) ,
which implies that the curvature tensor R vanishes if and only if the curvature form Ω
vanises.
Now we first assume that ωµ = S −1 ∂x∂ µ S , where S is a Lorentz transformation. This
means that ωµ is an element of the Lie-algebra of the Lorentz group. We compute
    
∂ ∂ −1 ∂ ∂ −1 ∂ −1 ∂2
ω µ = S S = S S + S S.
∂xν ∂xν ∂xµ ∂xν ∂xµ ∂xν ∂xµ
Now
    
∂ ∂ −1 ∂ −1 ∂
ων ωµ = S −1 S S S −1
= S S S S
∂xν ∂xµ ∂xν ∂xµ
  
∂ −1 ∂ ∂2
−S −1 S ν
S µ
S − S −1 S S −1 ν µ S
∂x ∂x ∂x ∂x
  
∂ −1 ∂
= − S S ,
∂xν ∂xµ
therefore
∂ ∂2
ωµ = −ων ωµ + S −1 ν µ S ,
∂xν ∂x ∂x
and for the exterior derivative dω we find
∂ ∂
ν
ωµ − ων = − [ων , ωµ ]− ,
∂x ∂xµ
which implies Ω = 0 .
Reversing the steps in this calculation we conclude that if Ω = 0 , then there exists a
Lorentz transfomation S with elements in the corresponding Lie-algebra given by ωµ =
S −1 ∂x∂ µ S .
We can now also proof the following lemma:
If the curvature tensor vanishes, then the metric gµν is induced from the Minkowski
metric ηµν by a coordinate transformation.
The argument goes as follows: If R = 0 then Ω = 0 and there exists ωµ = S −1 ∂x∂ µ S , where
S is a Lorentz transformation. We also have
∂ ∂
µ
θν − θ µ + ωµ θ ν − ων θ µ = 0 .
∂x ∂xν
Inserting ωµ = S −1 ∂x∂ µ S we thus find
   
∂ ∂ ∂ ∂
θν − θµ + S −1 S θ ν − S −1
S θµ = 0 ,
∂xµ ∂xν ∂xµ ∂xν
from which by multiplying with S from the left
   
∂ ∂ ∂ ∂
S θ ν − S θ µ + S θ ν − S θµ = 0 ,
∂xµ ∂xν ∂xµ ∂xν
i.e.
∂ ∂
(S θν ) − (S θµ ) = 0
∂xµ ∂xν
 68 CHAPTER 5. EXTENSIONS

follows, which means that the exterior derivative of S θ vanishes. Then according to the
Poincaré lemma there exists a Y such that
∂
(S θ)µ = Y,
∂xµ
which, after introducing the components of Y via
 0 
Y
 Y1 
Y =  Y2 


Y3
can be written in components as
∂
Yi.
Ski θµk =
∂xµ
Then with ηik = ηlm Sil Skm (Lorentz transformations are isometries of η) we find with the
defining equation of the tetrads Eq.(5.2.1):
∂Y l ∂Y m
gµν = ηik θµi θνk = ηlm Sil θµi Skm θνk = ηlm ,
∂xµ ∂xν
which indeed shows that the metric gµν is induced from the Minkowski metric ηµν by a
coordinate transformation. Note that this argument works in any dimension.
The tetrads are not uniquely defined by the defining equation
gµν = ηik θµi θνk . (5.2.1)
Indeed, if θµi is a tetrad, then also θ̃µi = Ski θµk , because
ηik θ̃µi θ̃νk = ηik Sli Sm θµ θν = ηlm θµl θνm = gµν .
k l m

It is left as an exercise to show that for θ 7→ θ̃ = S θ the connection form ω then transforms
as
∂
ωµ 7→ ω̃u = S ωµ S −1 + S S −1
∂xµ
and the curvature form transforms as
Ωµν 7→ Ω̃µν = S Ωµν S −1 .
We conclude that in the tetrad formulation the Lorentz group appears as a gauge group.

5.2.1 Tetrads and the Dirac equation

A Dirac spinor is a function Ψ : M → C 4 if M is the Minkowski space R 4 with the usual
metric ηµν then the Dirac operator is given by
∂
D := η µν γµ
,
∂xν
where the matrices γµ : R 4 → R 4 have the property
[γµ , γν ]+ := γµ γν + γν γµ = 2 ηµν 1IR 4 . (5.2.9)
Dirac The wave equation for a free spin- 21 particle of mass m is the free Dirac equation :
equation
(i D + m 1IR 4 ) Ψ = 0 .
A generalisation to General Relativity can be sketched as follows: The partial derivative
∂
∂xµ is to be replaced by a suitable covariant derivative and the Minkowski metric ηµν by
a general metrical tensor gµν everywhere. In particular the generalised Dirac matrices γ̃µ
have to satisfy
[γ̃µ , γ̃ν ]+ = 2 gµν 1IR 4 . (5.2.10)
Now Eq.(5.2.10) is readily fulfilled is we put γ̃µ = γk θµk where γi are constant matrices
fulfilling Eq.(5.2.9) :
[γ̃µ , γ̃ν ]+ = θµi θνk [γi , γk ]+ = 2 ηik θµi θνk 1IR 4 = 2 gµν 1I .
5.2. TETRADS 69

Acknowledgements
Thanks are due to Prof. Dr. Herbert Petry, who kindly provided me with material
to his lecture “Allgemeine Relativitätstheorie und Kosmologie” held by him at the Rhein-
ische Friedrich-Wilhelms Universität Bonn and to Dipl. Phys. Matthias Frink and Dr.
Sebastian König who kindly provided me with their notes to this lecture. This material
was the basis of the lecture notes presented here. I also thank the students Matthias
Kruckow, Suvendu Giri and Martin Ueding for scrutinising the manuscript.
Bonn, in July 2014, Bernard Metsch
Appendix A

Tensor calculus

Tensor calculus in an important calculational tool. The main ingredients stem from linear
and multilinear algebra:

A.1 Tensors and tensor fields

Let V be a n-dimensional real vector space and V ∗ the (also n-dimensional) vector space of
linear functions on V (dual space). If {eν } is any basis of V , the dual basis {eµ∗ } on V ∗ is dual
uniquely defined by space
eµ∗ (eν ) := δνµ ∀µ, ν = 1, . . . , n. (A.1.1)

If A : V → V is a linear map and if h∗ ∈ V ∗ then the adjoint map At : V ∗ → V ∗ is defined adjoint

by map
(At h∗ )(k) = h∗ (Ak) ∀k ∈ V. (A.1.2)

Let
ω : V × ··· × V ×V ∗ × ··· × V ∗ → R
| {z } | {z }
p−times q−times

be a map linear in each argument , i.e ω(h1 , . . . , hp , h∗1 , . . . , h∗q ) is a multilinear function of the
vectors hi ∈ V and h∗i ∈ V ∗ . Then ω is called a p-times covariant and q-times contravariant
tensor or briefly just (p, q) tensor. The tensors form a real vector space V (p,q) . Furhermore tensor
we use the convention, that the (0, 1) tensor ω(h) is identified (canonically) with a vector
hω ∈ V via
ω(h∗ ) = h∗ (hω ) ,
thus identifying V ∗ ∗ ≡ V . Furthermore by definition a (0,0) tensor is simple a real number.
A p-times covariant and q-times contravariant tensorfield, in short a (p, q) tensorfield is a tensor-
(smooth) map field
ω : V → V (p,q)
i.e. for each given x ; ω(x) ∈ V (p,q) and ω(x)(h1 , . . . , hp , h∗1 , . . . , h∗q ) ∈ R .
Any multilinear function is fixed uniquely by evaluating it with given basis vectors, eµ of V
and eµ∗ of V ∗ ,i.e.
ν

ω(x) eµ1 , . . . , eµp , eν∗1 , . . . , e∗q ∈ R
fixes ω(x) : Then
ν ···ν ν ···ν ν
ωµ11 ···µqp (x) ≡ ω(x)µ11 ···µqp := ω(x)(eµ1 , . . . , eµp , eν∗1 , . . . , e∗q ) (A.1.3)

are called the components of ω(x) w.r.t. to the basis eµ , µ = 1, . . . , p of V and eν∗ , ν = tensor
1, . . . , q of V ∗ . A (0, 0) tensorfield is simply a (real valued) function. compo-
nents
I
 II APPENDIX A. TENSOR CALCULUS

A.1.1 Elementary operations on tensorfields

We list the following elementary operations on tensor(field)s:

permuta- 1. The symmetric groups Sp and Sq (the groups of permutations of p and q objects,
tions respectively) operate on tensorfields ∈ V (p,q) , i.e. are maps: V (p,q) → V (p,q) , simply
by permuting the p arguments in V and the q arguments in V ∗ , respectively.
′ ′
2. The multiplication (ω ⊗ ω ′ )(x) ∈ V (p+p ,q+q ) of tensorfields ω ∈ V (p,q) (x) and ω ′ (x) ∈
′ ′
tensor V (p ,q ) (tensor product) is defined by
product
(ω ⊗ ω ′ )(x)(h1 , . . . , hp+p′ , h∗1 , . . . , h∗l+l′ ) (A.1.4)
∗ ∗ ′ ∗ ∗
:= ω(x)(h1 , . . . , hp , h1 , . . . , hq ) ω (x)(hp+1 , . . . , hp+p , hq+1 , . . . , hq+q′ ) .
′

In terms of components we have

µ1 ···µ ′ µ1 ···µ ′µq+1 ···µ ′
(ω ⊗ ω ′ )ν1 ···νp+p
q+q q q+q
′ (x) = ων1 ···νp (x) ω νp+1 ···νp+p′ (x) . (A.1.5)

contraction 3. If ω is a given (p, q) tensorfield with p, q > 0 then the contraction yields a (p−1, q −1)
tensorfield ωb , defined by

b (x)(h1 , . . . , hp−1 , h∗1 , . . . , h∗q−1 ) := ω(eµ , h1 , . . . , hp−1 , eµ , h∗1 , . . . , h∗q−1 ) .

ω (A.1.6)

In terms of components we have

µ ···µ αµ ···µ
bν11···νp−1
ω q−1
(x) = ωαν11···νp−1
q−1
(x). (A.1.7)

Here, and in the following, the summation over repeated (opposite) indices is as usual
implicit. Note that the result is independent of the choice of the basis eµ of V (and of
the basis eν∗ of V ∗ dual to this) .

Special tensor fields

Skew(-symmetric) (or antisymmetric) covariant tensor fields of rank p are also called
differential differential forms or in short p-forms.
forms
Let g(x) be a 2-times covariant tensorfield which is non-degenerate at each point x ∈ V (i. e.
g(x)(h, k) = 0 ∀k ∈ V ⇒ h = 0) and has a fixed signature everywhere, which means that
the number of positive and negative eigenvalues of the matrix with components gµν (x) =
g(x)(eµ , eν ) remains the same, which makes sense, if we assume in addition that g(x) is
symmetric , i.e. g(x)(h, k) = g(x)(k, h), which then indeed implies that the eigenvalues are
pseudo- real. Such a tensorfield of rank 2 is also called a pseudo-euclidian metric1 .
euclidian Then g(x) defines a natural (invertible) map g(x) : V ↔ V ∗ via
metric
(g(x)(h))(k) := g(x)(h, k) , ∀h, k ∈ V . (A.1.8)
| {z }
∈V ∗

With this map we can then define the following additional elementary operations on tensor
fields ω ∈ V (p,q) :

4. Define ωg ∈ V (p−1,q+1) by

ωg (x)(h1 , . . . , hp−1 , h∗1 , . . . , h∗q+1 ) := ω(x)(g −1 (x)h∗1 , h1 , . . . , hp−1 , h∗2 , . . . , h∗q+1 ).

(A.1.9)
In terms of the components
ν ···ν ν1 α ν ···ν
ωµ11 ···µq+1
p−1 (x) = g
2
ωαµ q+1
1 ···µp−1 (x) (A.1.10)

“raising” and accordingly this operation is called “raising” of a tensor index.

of a tensor
index
A.1. TENSORS AND TENSOR FIELDS III

5. Analogously one can also define the “lowering” of a tensor index by defining ωg′ ∈
V (p+1,q−1) via “lower-
ing” of a
ωg′ (x)(h1 , . . . , hp+1 , h∗1 , . . . , h∗q−1 ) := ω(x)(h2 , . . . , hp+1 , g(x)h1 , h∗1 , . . . , h∗q−1 ). tensor
(A.1.11) index
In terms of components
ν ···ν αν ···ν
ωµ11 ···µq−1 1 q−1
p+1 (x) = gµ1 α (x) ωµ2 ···µp+1 (x) (A.1.12)

Remark: In the equations (A.1.10, A.1.12) above for the components we suppressed
the prime ′ as well as the index g , since the fact that the tensorfields have different
covariant and contravariant ranks is evident from the covariant and contravariant in-
dices anyway. Likewise we shall suppress c in Eq.(A.1.7) in denoting the components
after contraction.

Examples
(1) Fµν (x) are the components of the electromagnetic field strength tensor. Then the
components of the derived tensor F µ ν (x) , as they appear on the right-hand side of
the equations of motion for the particle trajectories, follow from Fµν (x) by the “raising”
operation. Likewise the components of the derived tensor F µν (x) , appearing in the
inhomogeneous Maxwell-equation follows by application of the “raising” operation
twice.

(2) Fαν (x) F αµ (x) is the result of tensor multiplication followed by a contraction and

(3) by contracting again one obtains Fαβ (x) F αβ (x) as it occurs in the action.

A.1.2 Transformation rule for tensor fields

Let ϕ : V → V be a diffeomorphism and ω a (p, q)-tensorfield. The Jacobian is by definition
µ
a linear map Dϕ(x) : V → V . The matrix of Dϕ(x) has the components ∂ϕ ∂xν (x). We can
now the define a map: ϕ∗ : V (p,q) → V (p,q) by putting

(ϕ∗ ω)(x)(h1 , . . . , hp , h∗1 , . . . , h∗q ) :=

−1
ω(ϕ(x))(Dϕ(x)h1 , . . . , Dϕ(x)hp , [Dϕ (ϕ(x))]t h∗1 , . . . , [Dϕ−1 (ϕ(x))]t h∗q ). (A.1.13)

1 If g(x)(h, k) is positive definite, then it is called a euclidian metric

 IV APPENDIX A. TENSOR CALCULUS

If φ is another such diffeomorphism then

(ϕ ◦ φ)∗ ω = φ∗ (ϕ∗ ω) (A.1.14)

follows. Moreover,

ϕ∗ (ω ⊗ ω ′ ) = ϕ∗ ω ⊗ ϕ∗ ω ′ , b = ϕd
ϕ∗ ω ∗ ω,

ϕ ∗ ω ϕ ∗ g = ϕ∗ ω g and ϕ∗ ωϕ′ ∗ g = ϕ∗ ωg′ (A.1.15)

hold. In other words: The elementary operations defined above all commute with ϕ∗ . For
the components of (ϕ∗ ω) one finds (put y µ = ϕµ (x)) :

α ...α ∂y β1 ∂y βp ∂xν1 ∂xνq

(ϕ∗ ω)µν11...ν q 1 q
...µp (x) = ωβ1 ...βp (ϕ(x)) . . . . . . . (A.1.16)
∂xµ1 ∂xµp ∂y α1 ∂y αq

A.1.3 Derivatives of tensor fields

Let φ : V → R be a smooth function. Then the expression

d 
φ(x + τ h)
dτ τ =0

is linear in h ∈ V (for fixed x) and thus, if we declare


d 
(dφ(x))(h) := φ(x + τ h)
dτ τ =0

then dφ(x) is a covariant tensorfield of rank 1, also called a one-form. Now

(dφ(x))(h) ≡ (∇h φ)(x) ,

where (∇h (φ)) is the derivative of φ in the direction of h, which is indeed linear in φ for
fixed h .
We shall now generalise this derivative, above defined for functions, to tensor fields ω ∈
covariant V (p,q) . The operation is called the covariant derivative and shall be denoted by ∇ . It
derivative should have the following properties:

1. ∇ω is a (p + 1, q)-tensorfield and the product rule (for differentiation) should hold for
′ ′
arbitrary tensorfields ω ∈ V (p,q) and ω ′ ∈ V (p ,q ) :

∇h (ω ⊗ ω ′ ) = (∇h ω) ⊗ ω ′ + ω ⊗ (∇h ω ′ ) ,

where

(∇h ω)(x)(h1 , . . . , hp , h∗1 , . . . , h∗q ) = (∇ω)(x)(h, h1 , . . . , hp , h∗1 , . . . , h∗q ) .

Remark: If ω is a (0, 0)-tensorfield, i.e. ω is a real function, then ∇ω = dω , as above;

2. ∇ commutes with the contraction;

3. ∇ commutes with the “raising” and “lowering” operations;

The first two properties imply that the components of ∇ω, conventionally denoted as
ν ···ν
ωµ12 ···µqp1 ;µ1 , are given by

ν ···ν ∂ ν ···ν
ωµ12 ···µqp+1 ;µ1 (x) := ωµ1 ···µq (x) (A.1.17)
∂xµ1 2 p+1
αν2 ···νq ν1 αν3 ···νq ν1 ···νq−1 α
+Γνµ11 α (x) ωµ2 ···µ ν2 νq
p+1 (x) + Γµ1 α (x) ωµ2 ···µp+1 (x) + · · · + Γµ1 α (x) ωµ2 ···µp+1 (x)
ν ···ν ν ···ν ν ···ν
−Γα 1 q α 1 q α 1 q
µ1 µ2 (x) ωαµ3 ···µp+1 (x) − Γµ1 µ3 (x) ωµ2 αµ4 ···µp+1 (x) − · · · − Γµ1 µp+1 (x) ωµ2 ···µp α (x) .
A.1. TENSORS AND TENSOR FIELDS V

Indeed for ω(x) ∈ V (p,q) the covariant derivative ∇ω(x) should be a (p + 1, q) tensor. Ac-
cordingly we inspect the general linear combination (suppressing the argument x)
ν ···ν ∂ ν ···ν
(∇ω)µ11 µ2 ···µ
q
p+1 = ωµ1 ···µq
∂xµ1 2 p+1
αν2 ···νq ν1 αν3 ···νq ν1 ···νq−1 α
+Aνµ11 α ωµ2 ···µ ν2 νq
p+1 + Aµ1 α ωµ2 ···µp+1 + · · · + Aµ1 α ωµ2 ···µp+1
ν ···ν ν ···ν ν ···ν
+Bµα1 µ2 ωαµ
1 q α 1 q α 1 q
3 ···µp+1 + Bµ1 µ3 ωµ2 αµ3 ···µp+1 + · · · + Bµ1 µp+1 ωµ2 µ3 ···µp α

• Subsequently calculating the contraction for ν1 = µ2 = µ yields

∂ µ···νq µ αν2 ···νq ν2 µαν3 ···νq νq µν2 ···νq−1 α
ωµµ3 ···µ p+1 + Aµ1 α ωµµ3 ···µp+1 + Aµ1 α ωµµ3 ···µp+1 + · · · + Aµ1 α ωµµ3 ···µp+1
∂xµ1
µν ···ν µν2 ···νq µν2 ···νq
+Bµα1 µ ωαµ23 ···µqp+1 + Bµα1 µ3 ωµαµ α
3 ···µp+1 + · · · + Bµ1 µp+1 ωµµ3 ···µp α ;

• On the other hand, first contracting and then calculating the covariant derivative yields
∂ µ···νq ν2 µαν3 ···νq νq µν2 ···νq−1 α
ωµµ3 ···µ p+1 + Aµ1 α ωµµ3 ···µp+1 + · · · + Aµ1 α ωµµ3 ···µp+1
∂xµ1
µν2 ···νq µν2 ···νq
+Bµα1 µ3 ωµαµ α
3 ···µp+1 + · · · + Bµ1 µp+1 ωµµ3 ···µp α ;

• If these two operations should commute then

αν ···ν µν ···ν (α↔µ) αν ···ν
Aµµ1 α ωµµ23 ···µqp+1 = −Bµα1 µ ωαµ23 ···µqp+1 = −Bµµ1 α ωµµ23 ···µqp+1
should hold and consequently indeed
Aµαβ = −Bαβ
µ
=: Γµαβ .

The third property implies that

∂
∇g = 0 ⇔ gαβ (x) − Γλµα (x) gλβ (x) − Γλµβ (x) gαλ (x) = 0 , (A.1.18)
∂xµ
i.e. g is covariantly constant.
At this point we exploit a theorem by Levi-Cività, also see the exercises, which states:
If Eq.(A.1.18) holds and Γµαβ = Γµβα then the functions Γµαβ have the form
 
1 ∂ ∂ ∂
Γµαβ (x) = g µλ g αλ (x) + g βλ (x) − g αβ (x) , (A.1.19)
2 ∂xβ ∂xα ∂xλ
which is then the definition of the Christoffel-symbols .
Christoffel-
As a consequence the covariant derivative then also commutes with ϕ∗ :
symbols
∇(ϕ∗ ω) = ϕ∗ (∇ω) (A.1.20)
where it is understood that g is also transformed. We shall verify this explicitly for the
covariant derivative of a (1, 0)-tensorfield (or covariant vectorfield) A (the generalisation for
arbitrary tensorfields is analogous):
For the components of such a tensorfield we have the transformation rule, writing y = ϕ(x),
∂ϕα ∂y α
(ϕ∗ A)µ (x) = Aα (ϕ(x))
µ
= Aα (y) µ
∂x ∂x
The transformation rule for the Christoffel symbols, see exercises, reads
∂xλ ∂y β ∂y γ ∂xλ ∂ 2 y ρ
Γλµν [ϕ∗ g](x) = Γα
βγ [g](y) + ,
∂y α ∂xµ ∂xν ∂y ρ ∂xµ ∂xν
or equivalently,
∂y κ λ ∗ ∂y β ∂y γ ∂ 2 yκ
λ
Γµν [ϕ g](x) = Γκβγ [g](y) µ ν
+ .
∂x ∂x ∂x ∂xµ ∂xν
 VI APPENDIX A. TENSOR CALCULUS

With this, taking the covariant derivative of ϕ∗ A we indeed find

 
∗ ∗ ∂ ∂y α λ ∗ ∂y κ
(∇(ϕ A))µν (x) = (ϕ A)µ;ν (x) = Aα (y) − Γ µν [ϕ g](x)) Aκ (y)
∂xν ∂xµ ∂xλ
∂y β ∂y α ∂ 2 yα ∂y β ∂y γ ∂ 2 yκ
= (∂β Aα )(y) ν µ
+ Aα (y) µ ν − Γκβγ [g](y) Aκ (y) µ ν
− Aκ (y) µ ν
∂x ∂x ∂x ∂x ∂x ∂x ∂x ∂x
∂y β ∂y α ∂y β
∂y α
= (∂β Aα )(y) ν − Γκβα [g](y) Aκ (y) µ
∂x ∂xµ ∂x ∂xν

∂y α
∂y β
= (∂β Aα )(y) − Γκαβ [g](y) Aκ (y) = (ϕ∗ (∇A))µν (x) .
∂x ∂xν
µ

It is worthwhile to discuss two special cases:

(a) Let ωµ1 ···µp (x) be the components of a skew(-symmetric) (or anti-symmetric) covariant
tensorfield of rank p , i.e. for all σ ∈ Sp for the components

ωµσ(1) ···µσ(p) (x) = ε(σ) ωµ1 ···µp (x)

signum holds, where ε(σ) is the signum of the permutation σ ∈ Sp . Then the antisymmetric
part of ∇ω reads

ωµ2 µ3 ...µp+1 ;µ1 (x) − ωµ1 µ3 ...µp+1 ;µ2 (x) + ωµ1 µ2 µ4 ...µp+1 ;µ3 (x) − · · ·
∂
= ωµ µ ...µ (x) − Γα α
µ1 µ2 ωαµ3 ...µp+1 (x) − Γµ1 µ3 ωµ2 αµ4 ...µp+1 (x) − · · ·
∂xµ1 2 3 p+1
∂
− µ2 ωµ1 µ3 ...µp+1 (x) + Γα α
µ2 µ1 ωαµ3 ...µp+1 (x) + Γµ2 µ3 ωµ1 αµ4 ...µp+1 (x) +
∂x
∂
+ µ3 ωµ1 µ2 ...µp+1 (x) − Γα α
µ3 µ1 ωαµ2 ...µp+1 (x) − Γµ3 µ2 ωµ1 αµ4 ...µp+1 (x) · · ·
∂x
..
.
∂ ∂ ∂
= ωµ2 µ3 ...µp+1 (x) − ωµ1 µ3 ...µp+1 (x) + ωµ µ ...µ (x) − · · ·
∂x µ 1 ∂x µ 2 ∂xµ3 1 2 p+1
=: (dω)µ1 µ2 ···µp+1 (x) ,

since all the terms containing the Christoffel symbols pairwise cancel. Accordingly
upon antisymmetrisation the covariant derivative of a skew(-symmetric) covariant ten-
exterior sor field ω of rank p (also called a p-form) is just the exterior derivative dω of this
derivative tensor field (or this p-form).
The exterior derivative has the properties d2 = 0 and it commutes with the map φ∗ :
d ϕ∗ = ϕ∗ d . This implies that if the field strength tensor F fulfills dF = 0 , that
this is also true for ϕ∗ F : d(ϕ∗ F ) = 0 . Moreover, accoring to the Poincaré lemma,
dF = 0 implies that there exists a vector field A, such that F = dA (on star-shaped
domains).
(b) Let ω now be a skew-symmetric contravariant tensorfield of rank q . From this we form
divergence a covariant tensorfield (div ω) (of rank q − 1), the divergence of the antisymmetic
covariant tensor field ω , componentwise by
1 ∂ p 
(div ω)ν1 ···νq−1 (x) := ω µν1 ···νq−1 ;µ (x) = p |g(x)| ω µν1 ···νq−1
(x) .
|g(x)| ∂xµ
(A.1.21)
The last equality follows, because, due to the antisymmetry of ω
∂ µν1 ···νq−1
ω µν1 ···νq−1 ;µ = ω + Γµνα ω αν1 ···νq−1 + Γνµα
1
ω
|
µαν2 ···νq−1
{z }+···
∂xµ |{z}
sym.µ↔α antisym.µ↔α
µν1 ···νq−2 α
+ Γνµα
q−1
ω
| {z }
| {z }
sym.µ↔α antisym.µ↔α
A.1. TENSORS AND TENSOR FIELDS VII

all but the first two terms vanish. On the other hand,
∂ p 1 1 ∂ 1 1 ∂
µ
|g(x)| = p µ
|g(x)| = p |g(x)| g αβ (x) µ gαβ (x)
∂x 2 |g(x)| ∂x 2 |g(x)| ∂x
p αβ λ λ


= |g(x)| g (x) Γµα (x) gλβ (x) + Γµβ (x) gαλ (x)
p   p
β
= |g(x)| Γα µα (x) + Γ µβ (x) = |g(x)| Γα µα (x)

and thus
1 ∂ p µν1 ···νq−1
 ∂
p µ
|g(x)| ω (x) = ω µν1 ···νq−1 (x) + ω µν1 ···νq−1 Γα
µα (x) .
|g(x)| ∂x ∂xµ | {z }
αν1 ···νq−1
=Γµ
µα ω

The divergence occured in the inhomogeneous Maxwell equations:

1 ∂ p
F µν ;µ = p |g(x)| F µν = 4π J ν .
|g(x)| ∂xµ
Remarks:

Notation We already introduced the convention that in tensorfield components the co-
variant derivative w.r.t. a coordinate xµ is denoted by a semicolon: ;µ . Likewise it is
convenient to denote the partial derivative w.r.t. a coordinate xµ by a comma: ,µ , i.e.
∂ ν1 ···νq ν ···ν
ωµ ···µ (x) =: ωµ11 ···µqp ,µ (x) .
∂xµ 1 p
Then e.g. the definition of the Christoffel symbols reads
1
Γµαβ = g αλ (gαλ,β + gβλ,α − gαβ,λ ) .
2

Coordinate differentials Sometimes coordinate differentials are used: Concerning this

we make the following remarks:
Let V be a (finite dimensional) vector space with a basis eµ , µ = 1, . . . , dimV and let
x ∈ V . Then x = xµ eµ and xµ = eµ∗ (x) are the coordinates of x w.r.t. the basis eµ (recall:
eµ∗ (eλ ) = δλκ ) . The basis eµ∗ of the (to V ) dual space V ∗ are linear functions: V → R.
For the coordinate differentials coordi-
 
d µ  d µ  nate
µ
dx (h) = 
x (x + λ h) = e∗ (x + λ h) = eµ∗ (h) , ∀h ∈ V (A.1.22)
dλ dλ differen-
λ=0 λ=0
tials
holds and thus dxµ = eµ∗ .
For a covariant tensor of rank p we can thus write
µ
ω(x) = ωµ1 ···µp (x) eµ∗ 1 ⊗ eµ∗ 2 ⊗ · · · ⊗ e∗ p = ωµ1 ···µp (x) dxµ1 ⊗ dxµ2 ⊗ · · · ⊗ dxµp .
for an arbitrary (p, q)-tensorfield we had
ν ···ν µ
ω(x) = ωµ11 ···µqp (x) eµ∗ 1 ⊗ eµ∗ 2 ⊗ · · · ⊗ e∗ p ⊗ eν1 ⊗ eν2 ⊗ · · · ⊗ eνq .
We can now, observing that
∂xµ
eµ∗ (eν ) = δνµ = ,
∂xν
∂
symbolically put eν ≡ ∂xν and thus write
ν ···ν ∂ ∂ ∂
ω(x) = ωµ11 ···µqp (x) dxµ1 ⊗ dxµ2 ⊗ · · · ⊗ dxµp ⊗ ⊗ ⊗ ··· ⊗ .
∂xµ1 ∂xµ2 ∂xµq
We close this section with the remark, also of physical importantance, that we have found
that if g is a pseudoeuclidian metric, i.e. g is a non-degenerate covariant tensorfield of
rank 2, then with the Levi-Cività theorem one obtains a unique covariant derivative as
given in Eq.(A.1.17) in terms of the Christoffel symbols.
 VIII APPENDIX A. TENSOR CALCULUS

A.2 The curvature and related tensor fields

Now consider the following skew symmetric operation of second order directional derivatives
(with h, k ∈ V fixed) on a tensor field ω :
R(h, k) ω := ∇h ∇k ω − ∇k ∇h ω
which is linear in h ∈ V and k ∈ V . For the components we have
ν ···ν αν ···ν ν ···ν α
R(h, k) ωµ11 ···µqp = R(h, k)να1 ωµ1 ···µ
2 q νq 1 q−1
p + · · · + R(h, k)α ωµ1 ···µp
ν ···ν ν ···ν
−R(h, k)α 1 q α 1 q
µ1 ωαµ2 ···µp − · · · − R(h, k)µp ωµ1 ···µp−1 α ,

where R(h, k)αβ is a matrix depending linearly on h and k and which is antisymmetric in h
and k . Explicitely one thus finds with h = hµ eµ , k = k ν eν :
R(h, k)α α µ ν
β = Rβµν h k ,

where
α
Rβµν = Γα α α λ α λ
νβ,µ − Γµβ,ν + Γµλ Γνβ − Γνλ Γµβ
curvature and these components define a (3, 1)-tensorfield: the curvature tensor .
tensor Due to its construction this (3, 1)-tensorfield R depends on the metric, i.e. R[g] and we have
R[ϕ∗ g] = ϕ∗ R[g] .
Likewise, for the scalar function
α
R[g] := Rµαν g µν
one finds
R[ϕ∗ g] = ϕ∗ R[g] .
Note that the operation ∇h ∇k − ∇k ∇h will be zero, if g is just a constant. In this case ∇h
will just be given in terms of partial derivatives as usual.
Properties (also see exercises):
(a) The curvature tensor has the symmetry properties:
α α α α α
Rβµν = −Rβνµ , Rβµν + Rνβµ + Rµνβ = 0;

λ
(b) The (4, 0) tensorfield Rαβµν = gαλ Rβµν has the symmetry properties:
Rαβµν = Rµναβ = −Rβαµν = −Rαβν µ ;

Ricci- (c) The Ricci-tensor is defined by

tensor
α
Rβν := Rβαν ;

Ricci- (d) The Ricci-scalar is defined by

scalar R := g βν Rβν
and for this
R((ϕ∗ g)(x)) = R(g(φ(x))) = R(φ(x))
holds;
Einstein- (e) The Einstein-tensor is then defined by
tensor
1
Gµν = Rµν − R g µν
2
and
Bianchi- (f) fulfills, according to a Bianchi-identity
identity
Gµν ;µ = 0 .

Note that all the identities given above follow from Eq.(A.1.18) .
Appendix B

Note on manifolds

B.1 Definitions
Consider a collection Uα of open sets in R n with the following properties:

• For Uα and Uβ in this collection there exist subsets Uβα ⊂ Uα and Uαβ ⊂ Uβ such
that there is a differentiable map (diffeomorphism)

ϕβα : Uβα → Uαβ .

(If Uαβ = Uβα = ∅ then there is no such map).

• If Uγ is a third member of this collection then one requires that for

(ϕγβ ◦ ϕβα ) : Uβα ∩ Uγα → Uαγ ∩ Uβγ

we have ϕγβ ◦ ϕβα = ϕγα if the intersections are not empty.

f . In M
We now consider the disjoint union of such sets ∪α Uα = M f we can define an
equivalence relation:

x ∈ Uα and y ∈ Uβ are equivalent (notation: x ∼ y) if y = ϕβα (x) .

We denote by M the set of all equivalent classes. In this manner we defined the manifold manifold
M , which is the first definition of a manifold.
Between the sets M f and M there is a natural map π : M f → M . M̃ is a topological
space and so is M , furthermore, π is a continuous map: π : Uα → M is continuous and
injective. For Vα = π(Uα ) we have M = ∪α Vα and ϕα = π|Uα is invertible and continuous.
If Uβα ⊂ Uα , Uβα 6= ∅ then ϕ−1 β ◦ ϕα = ϕβα . The triple (Uα , Vα , ϕα ) is called a chart and chart
the collection of all charts, {Uα , Vα , ϕα } is called an atlas of M . atlas
There is also an alternative second definition of a manifold M :

(a) M is a topological Hausdorff space;

(b) M admits an atlas {Uα , Vα , ϕα } , such that ϕ−1

β ◦ ϕα = ϕβα is a local diffeomorphism.

B.2 Construction of a new manifold

Let M be declared as in the first definition given above. We now construct a new manifold
M̂ as follows: Replace Uα by Ûα = Uα × R n and replace Uαβ by Ûαβ = Uαβ × R n .
Then M̂ = ∪α Ûα . Furthermore replace ϕαβ by ϕ bαβ and where
bβα (x, h) , where (x, h) ∈ U
bβα (x, h) = (ϕβα (x), Dϕβα (x)(h)) . Then ϕ
ϕ bαβ can again be used to define an equivalence
relation and this then defines a manifold Mc.

IX
 X APPENDIX B. NOTE ON MANIFOLDS

B.3 c and M
relation between M
The relation between Mc and M is as follows: There is a canonical map π : M c → M . For
−1 n c
x ∈ M we have π (x) ∈ R . The manifold M together with the canonical map π is called
tangent the tangent bundle over M . If σ : M → M c is such that π ◦ σ = Id , then σ is called a
bundle vector field . From the construction it is clear that in Uα the image σ(x) can be described
vector field uniquely by the function value X(x) ∈ R n . Then (x, X(x)) represents an equivalence class
c and under π projects on the point x , i.e. π projects (x, X(x)) onto the first component,
in M
which is x .
tensor In order to also describe a tensor bundle associated to M we first recall some facts from
bundle multilinear algebra: Accordingly, consider V = R n with its dual space V ∗ . A (p, q)-tensor
is a multilinear map ω such that ω(h1 , . . . , hp , h∗1 , . . . , h∗q ) is linear in each argument hi ∈
V, h∗i ∈ V ∗ . If A : R n → R n is linear and invertible, then A∗ ω is defined by

(A∗ ω)(h1 , . . . , hp , h∗1 , . . . , h∗q ) = ω(Ah1 , . . . , Ahp , (A−1 )t h∗1 , . . . (A−1 )t h∗q ) ,

with the property (AB)∗ = B ∗ A∗ for two such maps of tensors. Furthermore the (p, q)-
tensors form a vector space V (p,q) and the map A∗ is a linear isomorphism.
Let M again be a manifold defined by charts. Now replace Uα by Uα × V (p,q) , replace the
bαβ (x, ω) = (ϕαβ (x), (Dϕαβ (x)−1 )∗ ω) . Then indeed ϕ
transition maps by ϕ bαβ ◦ ϕ
bβγ = ϕ
bαγ .
The construction of equivalence classes then yields a new manifold M c and a canonical
c → M with π −1 (x) for x ∈ M yields a vector space isomorphic to V (p,q) .
projection π : M
A map σ : M → M c is then represented by a tensor field ω(x) for x ∈ Uα .
Index

Bianchi-identity, VIII contraction, II

Christoffel-symbols, V coordinate differentials, VII
Dirac equation, 68 cosmological principle, 51
Doppler effect, 57 cosmological constant, 12
Einstein-tensor, VIII covariant derivative, IV
Hubble constant, 54 curvature form, 67
Hubble’s law, 54 curvature scalar, 12
Jacobi-method, 18 curvature tensor, VIII, 12
Kaluza-Klein theory, 63
Kerr-solution, 50 deceleration parameter, 54
Klein-Gordon equation, 64 differential forms, II
Kruskal-metric, 36 divergence, VI
Reissner-Nordström, 40 dual space, I
Ricci-scalar, VIII dust, 40
Ricci-tensor, VIII
Robertson-Walker metric, 52 electromagnetic action, 7
Schwarzschild radius, 32 energy density, 23
Wien’s law, 56 equation of state, 23
“lowering” of a tensor index, III exterior derivative, VI
“raising” of a tensor index, II
field strength tensor, 2
Bianchi-identity, 15
flatness problem, 61
Christoffel-symbol, 10
fundamental form, 66
Einstein-Lorentz equation, 1
Einstein equations, 15 gauge transformation, 47
Einstein-tensor, 15 gauge transformations, 8
Friedmann-equation, 60 generalised “spherical” coordinates, 50
Lorentz transformation, 2 geodesic equation, 11
Maxwell equations, 1 gravitational constant, 12
Minkowksi metric, 2 gravitational lensing, 35
Poincaré transformation, 2
Ricci scalar, 12 isometry, 3
action principle, 7 manifold, IX
adjoint map, I
angular frequency, 4 permutations, II
arc length, 2 polarisation, 4
atlas, IX pressure, 23
pseudo-euclidian metric, II
black hole, 38
boosts, 3 redshift, 54
retarded solution, 48
central energy density, 28 Ricci tensor, 12
charge-current density, 2
chart, IX signature, 8
comoving coordinates, 17 signum, VI
conformal time, 53 space-time translation, 2
connection form, 66
continuity equation, 1 tangent bundle, X

XIII
XIV INDEX

tensor, I
tensor bundle, X
tensor components, I
tensor product, II
tensorfield, I
tetrad, 66
torsion form, 66

vector field, X

wave vector, 4