Part II General Relativity

Lecture Notes

Abstract
These notes represent the material covered in the Part II lecture General Relativity
(GR). While the course is largely self-contained and some aspects of Newtonian Gravity
and Special Relativity will be reviewed, it is assumed that readers will already be famil-
iar with these topics. Also, calculus in N dimensions and Linear Algebra will be used
extensively without being introduced.
This set of notes differs from the long version by representing in almost verbatim style
how the material is presented in the lecture room. It is primarily designed to dispense
with the necessity to take notes during the lectures.
A more in-depth discussion of books is given in the long set of notes. Here we merely
discuss a few books seemingly most suitable for an introduction to general relativity.
• S. M. Carroll: “Spacetime and Geometry: An Introduction to General Relativity”
[2] ; cf. also [1] .
• R. d’Inverno: “Introducing Einstein’s Relativity” [3] .
• J. B. Hartle: “Gravity, An Introduction to Einstein’s General Relativity” [4] .
• L. Ryder: “Introduction to General Relativity” [5] .
• B. Schutz, “A first course in general relativity” [6] .
I would not set any of them apart over the others, but recommend each reader to have a
look at them and find where the best chemistry is found.

Example sheets will be pointed to at some later stage, probably on

http://www.damtp.cam.ac.uk/user/examples

Lectures Webpage:
http://www.damtp.cam.ac.uk/user/us248/Lectures/lectures.html

Cambridge, Dec 10 2016

Ulrich Sperhake

1
CONTENTS 2

Contents
A Preliminaries 4
A.1 Units . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
A.2 Newtonian Gravity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
A.3 Special Relativity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

B Differential geometry 16
B.1 Manifolds and tensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
B.2 The metric tensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
B.3 Geodesics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
B.4 Covariant derivative . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
B.5 The Levi-Civita connection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
B.6 Parallel transport . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
B.7 Normal coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
B.8 The Riemann tensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

C Physical laws in curved spacetimes 38

C.1 The covariance principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
C.2 The energy momentum tensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
C.3 The Einstein equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

D The Schwarzschild solution and classic tests of GR 42

D.1 Schwarzschild’s solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
D.2 Geodesics in the Schwarzschild spacetime . . . . . . . . . . . . . . . . . . . . . . 43
D.3 Classic tests of GR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
D.4 The causal structure of the Schwarzschild spacetime . . . . . . . . . . . . . . . . 51
D.5 Hawking radiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

E Cosmology 59
E.1 Homogeneity and isotropy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
E.2 The Friedmann equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
E.3 Cosmological redshift . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
E.4 Cosmological models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

F Singularities and geodesic incompleteness 70

F.1 Coordinate vs. physical singularities . . . . . . . . . . . . . . . . . . . . . . . . 70
F.2 Geodesic incompleteness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

G Linearized theory and gravitational waves 72

G.1 Plane waves and pp metrics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
G.2 Linearized theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
G.3 The Newtonian limit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
G.4 Gravitational waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
CONTENTS 3

G.5 The quadrupole formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76

A PRELIMINARIES 4

A Preliminaries
A.1 Units
SI units: • metres, second, kilogram etc.

• Adapted to “ourselves” → numbers O(1)

Natural units: • Take into account constants of nature.

→ 1) unifies physical dimensions (space, time etc.)

2) indicate possible breakdown of theories

• Better adapted to extreme physics.

Speed of light
!
c = 299 792 458 m/s ≈ 3 × 108 m/s = const

Suggests to measure all speeds in units of c.

!
⇒ c = 3.00 × 108 m/s = 1

⇒ 1 s = 3.00 × 108 m

Familiar from the light year: 1 yr = 9.4607 × 1015 m

vc=1 ⇒ Galileo trafo, Newtonian kinematics accurate

v.1 ⇒ both break down → need special relativity (SR)

Gravitational constant

m3 !
G = 6.67408 × 10−11 = const
kg s2

Set c = 1, G = 1 ⇒ 1 m = 1.3466 × 1027 kg or 1 s = 4.0370 × 1035 kg

Example: Solar mass M = 1.4771 km = 4.9269 µs → ∼ Schwarzschild radius of sun

M c2
 =1 ⇒ Newtonian gravity (NG) accurate
R G
M
≈1 ⇒ NG breaks down → need general relativity (GR)
R
A PRELIMINARIES 5

Comment: If velocities determined by gravity, the regimes v ≈ 1 and M/R ≈ 1 overlap.

v2 GM M
E.g.: (1) Spherical orbit around mass M : 2 = 2 ⇒ v2 =
c c R R
ve2 2G M 2M
(2) Escape velocity from sphere of mass M : 2
= 2 ⇒ ve2 =
c c R R

Planck’s constant

h kg m2
~ ..= = 1.0545718 × 10−34
2π s

1
Set c = 1, ~ = 1 ⇒ 1 kg = 8.5223 × 1050 Hz or 1m=
3.51767288 × 10−43 kg
~ 1
Compton wavelength: λ̄ = =
mc m
Compare Compton wavelength of a body with its size or available volume.
~ λ̄
E.g. : Sun: λ̄ = = 0.177 × 10−72 m ⇒ 1
M c R
~
Proton: λ̄p = = 0.210268 fm ∼ radius of atomic nuclei ∼ 1 . . . 10 fm
mp c

λ̄
1 ⇒ Classical physics accurate
R
λ̄
≈1 ⇒ Need quantum mechanics (QM)
R

Planck mass
GM ~
Consider a system with 2
= 1 (GR!) and = 1 (QM!).
c R M cR
r
~c
⇒ M= = 2.18 × 10−8 kg = 1.22 × 1019 GeV
G
In this regime, quantum and GR effects are important.

The theory of quantum gravity for this regime remains unknown.

A PRELIMINARIES 6

A.2 Newtonian Gravity

A tale of 3 masses

Newtonian 2-body problem m1

~r1 − ~r2 !
F~1on2 = Gm1a m2p = m2i~r¨2
|~r1 − ~r2 |3 F~
ma = active mass: source of grav. field ,
−F~
mp = passive mass: sensitivity to grav. fields ,
mi = inertial mass: resistance to change motion . m2

~r2 − ~r1 ! ~ ~r2 − ~r1

“Actio = Reactio” ⇒ F~2on1 = Gm1p m2a = −F1on2 = Gm1a m2p
|~r2 − ~r1 |3 |~r2 − ~r1 |3
~r1 , ~r2 are arbitrary, so: m1p m2a = m1a m2p
m1p m2p
⇒ =
m1a m2a
Ratio of passive to active mass is the same for all bodys. Without loss of generality (WLoG),

ma = mp

Note: Same holds for electric charge: qa = qp .

How about inertial mass?

∼ 1590: Galileo: Balls of different mass need same time to roll down a slope.

1922: Eötvös: Torque from Sun’s gravity on torsion balance is < 5 × 10−9

+ many more, all compatible with ma = mp = mi . This leads to the...

Equivalence principles

Weak Equivalence Principle (WEP): Freely falling bodies with negligible gravitational self
interaction follow the same path if they have the same initial velocity and position.

Einstein promotes this to a more general level.

Def. : “local inertial frame” ..= coordinate frame (t, x, y, z) defined by a freely falling observer
in the same way as an inertial frame is defined in Minkowski spacetime. “local” means
 length scale of variations in the gravitational field ~g (cf. figure).
A PRELIMINARIES 7

Lab frame

Earth

Einstein Equivalence Principle (EEP): In a local inertial frame, the results of all non-gravitational
experiments are indistinguishable from those of the same experiment performed in an inertial frame
in Minkowski spacetime.

Strong equivalence principle (SEP): The gravitational motion of a small test body (that may
have gravitational self interaction) depends only on its initial velocity and position but not on its
constitution.

Comments: • SEP ⇒ WEP ; in general WEP ; SEP

• Need “small” objects to avoid tidal effects. E.g. Moon drifting away from
Earth.

• SEP is related to equality of active and passive mass. Say Earth

and Moon have different ma /mp . They’d fall differently in the Sun’s
field ⇒ “Nortvedt effect”

• SEP implies G = const everywhere.

• GR satisfies all three EPs. Gravity is a feature of spacetime!

Gravitational redshift
z
Consider: ~g = (0, 0, −g) , Alice at z = h, Bob at z = 0 Alice g
Alice sends light to Bob.

EEP ⇒ equivalent to frame accelerated with (0, 0, +g) in Bob

Minkowski spacetime x, y

Assumption: v of Bob, Alice  c

v2
⇒ ignore 2 and higher-order SR terms
c
A PRELIMINARIES 8

1 1 !
⇒ zA (t) = h + gt2 , zB (t) = gt2 , vA = vB = gt  c .
2 2

• Alice emits first signal at t1

1
⇒ z1 (t) = zA (t1 ) − c(t − t1 ) = h + gt21 − c(t − t1 )
2
1 2 1
• This reaches Bob at T1 , i.e. h + gt1 − c(T1 − t1 ) = gT12 (∗∗)
2 2
• Alice emits second signal at t2 = t1 + ∆τA .
This reaches Bob at T2 = T1 + ∆τB .

1 1 
⇒ h + g(t1 + ∆τA )2 − c(T1 + ∆τB − t1 − ∆τA ) = g(T1 + ∆τB )2  subtract (∗∗)
2 2 
1 1
⇒ c(∆τA − ∆τB ) + g ∆τA (2t1 + ∆τA ) = g ∆τB (2T1 + ∆τB )
2 2
• Assumption: ∆τA  t1 , ∆τB  T1 , e.g. period in light waves
⇒ c(∆τA − ∆τB ) + g ∆τA t1 = g ∆τB T1
⇒ ∆τB (gT1 + c) = ∆τA (gt1 + c)
−1  
g(T1 − t1 )
   
gT1 gt1  we used gt  1

⇒ ∆τB = 1 + 1+ ∆τA ≈ 1 − ∆τA
c c c  c

h 1g  we used gt  1

• (∗∗) ⇒ − (T1 − t1 ) = (T1 + t1 )(T1 − t1 ) ≈ 0
c 2 |c {z }  c
1
h
⇒ T1 − t1 = to leading order.
c
 
gh !
• ⇒ ∆τB ≈ 1 − 2 ∆τA < ∆τA
c
 
gh
⇒ Signal appears blue shifted to Bob: c ∆τB = λB ≈ 1 − 2 λA
c

Confirmed in Pound-Rebka experiment (1960): light falling in tower.

Light climbing out of a gravity well is red shifted.

Redshift in curved spacetime

Recall: Invariant interval in SR: c2 ∆τ 2 = c2 ∆t2 − ∆x2 − ∆y 2 − ∆z 2

For weak, static gravitational field, this generalizes to (cf. later):

A PRELIMINARIES 9

   
2 2 2φ(x, y, z) 2 2 2φ(x, y, z) φ
c dτ = 1 + c dt − 1 − (dx2 + dy 2 + dz 2 ) ; 1
c2 c2 c2

• Alice: ~xA , Bob: ~xB , at fixed positions!

• Alice emits signals at tA , tA + ∆t

Bob receives the first at tB . When does he see the second?

• The spacetime is static: φ does not depend on t

⇒ The two signals travel on identical trajectories, just shifted in time

⇒ Bob receives the second signal at tB + ∆t .

• But what proper times do Alice’s and Bob’s clocks measure?

   
2 2φA 2 2 2φB
∆τA = 1 + 2 ∆t , ∆τB = 1 + 2 ∆t2
c c
   
φA φB
⇒ ∆τA ≈ 1 + 2 ∆t , ⇒ ∆τB ≈ 1 + 2 ∆t ,
c c
−1
φB − φA
   
φB φA
⇒ ∆τB ≈ 1 + 2 1+ 2 ∆τA ≈ 1 + ∆τA
c c c2

Newtonian gravity for matter fields

Index notation:
• Write vectors, matrices as components: xi = (x1 , x2 , x3 ) = (x, y, z); vi = (v1 , v2 , v3 ) etc.
3
X
• Repeated indices in a product are summed over: Aij vj ..= Aij vj
j=1

• No index may appear more than twice: Aii vi is not defined.

• We may rename indices summed over: Aij vj = Aik vk
• In an equation or a sum, free (not repeated) indices must match on both sides:
wi + Aik vk = 0 is correct; wj = Aik vk is not.
∂
• We denote partial derivatives by ∂i = . Sometimes, we also use a comma:
∂xi
∂vk
vk,i ..= ∂i vk =
∂xi
A PRELIMINARIES 10

Example: Motion of point particle in gravitational field ~g .

m~x¨ = m~g (~x, t) ⇒ ẍi = gi (xk , t).

Let x̃i be a non-inertial coordinate system: x̃i = xi − bi (t).

⇒ x̃¨i = g̃( x̃k , t) = gi (x̃k , t) − b̈i (t)

Comments: 1) If gi is uniform (xk independent) ⇒ ∃ bi such that g̃i = 0.

2) gi not uniform ⇒ we can only get g̃i = 0 locally → freely falling frame

Index version of Newtonian gravity

Tidal forces on two particles at xi , xi + δxi :

d2 d2
xi = gi (xj , t) , (xi + δxi ) = gi (xj + δxj , t)
dt2 dt2
d2
⇒ 2
δxi = δxk ∂k gi + O(δx2j ) ,
dt

d2
⇒ δxi + Eij δxj = 0 , Eij ..= −∂j gi .
∂t2

~
~g is curl free ⇒ ~g = −∇φ ⇔ gi = −∂i φ

It follows: Eij = Eji .

~ · ~g = −4πGρ
Poisson equation: ∇ ⇒ ~ 2 φ = ∂i ∂i φ = 4πGρ
∇ ⇒ Eii = 4πGρ .

The definition Eij = −∂j gi implies

1
∂k Eij = −∂k ∂j gi = ∂j Eki ⇒ Ei[j,k] ..= (Eij,k − Eik,j ) = 0
2

The need for GR: not so much from experiment, but the incompatibility of SR with Newto-
nian space and time.
A PRELIMINARIES 11

A.3 Special Relativity

Extend index notation
• Distinguish upstairs and downstairs indices: v i 6= vi .
3
X
• Summation only over one up and one downstairs index: v j uj ..= v j uj .
j=1

• Latin indices i, j, . . . = 1, 2, 3. Greek indices α, β, . . . = 0 . . . 3.

Metric

Pythagoras as matrix equation: ∆s2 = ∆x2 + ∆y 2 + ∆z 2 = δij ∆xi ∆xj (†)

δij = diag(1, 1, 1) = flat Euclidean metric in Cartesian coords.

In polar coordinates: ds2 = dx2 + dy 2 + dz 2

= dr2 + r2 dθ2 + r2 sin2 θ dφ2
= gij dx̃i dx̃j , gij = diag(1, r2 , r2 sin2 θ) .

Note: Unlike (†), this only works for infinitesimal distances!

Lorentz transformations

Consider inertial (non-accelerated) frames with Cartesian coords.

Proper distance between spacetime events (t, x, y, z) and (t + ∆t, x + ∆x, y + ∆y, z + ∆z):
∆s2 = −∆t2 + ∆x2 + ∆y 2 + ∆z 2

In SR, no inertial frame is prefered over another ⇒ same proper distance in x̃α :
∆s2 = −∆t̃2 + ∆x̃2 + ∆ỹ 2 + ∆z̃ 2

Note: For ∆s = 0, the events are connected by a light ray.

⇒ All inertial frames measure the same speed of light.

Index notation: ∆s2 = ηαβ ∆xα ∆xβ = ηα̃β̃ ∆x̃α̃ ∆x̃β̃

   
−1 0 0 0 −1 0 0 0
 0 1 0 0   0 1 0 0 
ηαβ = ηα̃β̃ = 
 0
 ⇔ η αβ = η α̃β̃ = 
0 1 0   0 0 1 0 
0 0 0 1 0 0 0 1
η αβ is the inverse of ηαβ .
A PRELIMINARIES 12

Inertial frames are related by x̃α̃ = Λα̃ µ xµ + xµ0 , Λα̃ µ = const

WLoG: xµ0 = 0.
z̃

ỹ

x̃

z
~v

x
!
We want: ηα̃β̃ ∆x̃α̃ ∆x̃β̃ = ηα̃β̃ Λα̃ µ ∆xµ Λβ̃ ν ∆xν = ηµν ∆xµ ∆xν

⇒ ηµν = Λα̃ µ Λβ̃ ν ηα̃β̃

One can show that this is satisfied by the Lorentz transformations

   
γ −γv 0 0 γ γv 0 0
 −γv γ 0 0    γv γ 0 0  1
Λα̃ µ = 
 0 ⇔ Λµ α̃ =  , γ=√
0 1 0   0 0 1 0  1 − v2
0 0 0 1 0 0 0 1

Comments • WLoG, coordinates oriented such that the velocity v points in the x direction

• One can show Λα̃ µ Λµ β̃ = δ α̃ β̃ , Λµ α̃ Λα̃ ν = δ µ ν

World lines and 4-velocity

Def.: The interval between two spacetime events xα and xα + ∆xα is called
timelike :⇔ ηµν ∆xµ ∆xν < 0

null :⇔ ηµν ∆xµ ∆xν = 0

spacelike :⇔ ηµν ∆xµ ∆xν > 0 .

Def.: Proper time ∆τ 2 ..= −∆s2 = ∆t2 − ∆x2 − ∆y 2 − ∆z 2

A PRELIMINARIES 13

Postulate: A clock moving on a world line xα (λ) , λ ∈ R, that is in every point timelike or null,
measures the proper time along this world line
Z λ2 r
dxµ dxµ
τ ..= −ηµν dλ . (‡)
λ1 dλ dλ

Comments: • τ is invariant under reparametrizing λ → µ(λ).

• We often parametrize timelike curves with τ

r
dxµ dxν µ ν .. dxµ dxν
• (‡) ⇒ dτ = −ηµν dτ ⇒ ηµν ẋ ẋ = ηµν = −1
dτ dτ dτ dτ

dxα
Def.: The four velocity along a timelike curve is uα ..=
dτ
By def. ηµν uµ uν = −1

Geodesics
Z r
dxα dxβ
Consider the action S[xα (λ)] = −ηαβ dλ
| {zdλ dλ}
=..L

Geodesics are curves that extremize this action.

They follow from the Euler-Lagrange (EL) equations

d ∂L ∂L d 2 xα
µ
= ⇒ ... ⇒ = 0.
dλ ∂ ẋ ∂xµ dτ 2
One can derive the same equation for null and spacelike geodesics (cf. GR case below).

Postulate: Free massive (massless) particles in special relativity move on straight timelike (null)
curves,
d2 xα
= 0. (A.1)
dτ 2
A PRELIMINARIES 14

Time dilation

Let O, Õ be 2 observers with coordinates xµ , x̃α̃ .

Let Õ move with v in the x direction relative to O.

 
α̃ dt̃
Clock at rest in Õ: ũ = , 0, 0, 0
dτ

dt dxi ! µ α̃
   
µ dt̃ dt̃
Viewed from O: u = , = Λ α̃ ũ = γ , γv , 0, 0 (†)
dτ dτ dτ dτ

dt dt̃ dt
ut component: =γ ⇒ =γ
dτ dτ dt̃
dt̃
⇒ dt = √ : O sees the moving Õ age more slowly.
1 − v2

Lorentz contraction

Def.: Length of a rod in O ..= proper distance ∆s between two events A and B, where xiA is the
position of the rod’s tail at a specified time tA = t0 and xiB is the position of the rod’s head
at the same time tB = t0 .
q p
∆s = ηαβ ∆xα ∆xβ = δij ∆xi ∆xj , ∆xi = xiB − xiA (A.2)

Let the rod be at rest in O.

World lines of head and tail: xµ = (ttail , xi0 ) , y µ = (thead , xi0 + ∆xi )

O will pick events with ttail = thead ⇒ `2 = ∆s2 = δij ∆xi ∆xj

World lines in Õ: (t̃tail , x̃i ) = x̃α̃ = Λα̃ µ xµ

(t̃head , ỹ i ) = ỹ α̃ = Λα̃ µ y µ = Λα̃ µ (xµ + ∆xµ )

Õ will pick events Ã, B̃ with t̃tail = t̃head .

Λ0̃ i ∆xi
⇒ . . . ⇒ ttail = thead + 0̃
= thead − vi ∆xi
Λ0
Proper distance ∆s2ÃB̃ : xµÃ = (thead − vi ∆xi , xi0 ) , xµB̃ = (thead , xi0 + ∆xi )

⇒ `˜2 = ∆s2ÃB̃ = ηµν (xµB̃ − xµÃ )(xνB̃ − xνà )

= −(vi ∆xi )2 + δij ∆xi ∆xj

A PRELIMINARIES 15

`˜ =
p
Orient the rod along the x axis ⇒ ` = ∆x , 1 − vx2 ∆x

Comments: • The sign of v does not matter.

• Velocity perpendicular to the rod causes no contraction.

Four momentum and Doppler shift

Def.: Four momentum of a particle with rest mass m and 4-velocity uµ : pα = muα

ηµν uµ uν = −1 ⇒ ηµν pµ pν = −m2

Let Õ move with v in x direction relative to O; consider particle at rest in Õ.

⇒ p̃α̃ = (m, 0, 0, 0)

pµ = Λµ α̃ p̃α̃ = γm(1, v, 0, 0)

γm = total relativistic energy of particle in O

γmv = linear momentum of particle in O

⇒ pµ = (E, p, 0, 0)

ηµν pµ pν = −E 2 + p2 = −m2 ⇒ E 2 = m2 c4 + p2 c2

Comment: Null curves do not have a four velocity (unit tangent vector), but they have a
four momentum.

Recall for massless particles: E = hν, p = h/λ.

⇒ pα = hν(1, 1, 0, 0) e.g. photon moving in x direction

Now consider such a photon and let Õ move with v in x-direction relative to O.

⇒ in O: pα = (E, E, 0, 0) , E = hν



⇒ in Õ: p̃α̃ = Λα̃ µ pµ = γE − γvE, − γvE + γE, 0, 0 =.. Ẽ, Ẽ, 0, 0 , Ẽ = hν̃
r
ν̃ Ẽ 1−v
Redshift = = γ(1 − v) =
ν E 1+v
Note: • Redshift if Õ moves in same direction as photon. Blueshift if v < 0.

• Transverse Doppler shift if v perpendicular to propagation of photon.

More complicated to calculate!
B DIFFERENTIAL GEOMETRY 16

B Differential geometry
Goal: Extend Euclidean geometry to curved spaces.

Motivation: GR generalizes SR like Riemannian geometry generalizes Euclidean geometry.

Conventions: • An upstairs index in a denominator counts as a downstairs index.

∂
E.g.: ∂i ..=
∂xi
• Contravariant indices: upstairs. Covariant indices: downstairs.

B.1 Manifolds and tensors

Strategy: Start with manifold M. Establish structure on M step by step.

Def.: n dimensional manifold M ..= set of points that locally resembles Euclidean space Rn at
each point. For our purposes, this means that there exists a one-to-one and onto map

φ : M → U ⊂ Rn , p ∈ M 7→ xα ∈ U ⊂ Rn , α = 0, . . . , n − 1 , (B.1)

where U is an open subset of Rn . xα are the coordinates on M.

U ⊂R

xα
B DIFFERENTIAL GEOMETRY 17

Comments: • It is sufficient if we can chop up M and map each piece separately to Rn .

Everything we will develop also holds for such subdivisions of M.

• Think of coordinates like house numbers in a street. Houses don’t change if

we change their numbers. Likewise, we will find objects on M to be invariant
under coordinate changes.

• Curves, vectors etc. live on M, not in the coordinate space. But φ is one-to-
one, so this distinction often blurred.

Functions and curves

Def.: Function on M: f :M→R

f is smooth :⇔ ∀ coordinate systems xα : f (xα ) is a smooth function from Rn to R. If f

is invariant under a change of coordinates, it is a scalar.
Def.: Curve ..= a map λ : I ⊂ R → M, I open. λ is smooth :⇔ ∀ coordinate systems xα : the
map xα ◦ λ : I → Rn is smooth

Vectors

Def.: Let C ∞ be the space of all smooth functions f : M → R, λ be a smooth curve and
p ≡ λ(0) ∈ M. The tangent vector to the curve λ at p ∈ M is the map

d
V : C → R,
∞


f 7→ V (f ) = f λ(t)  .
dt t=0

Tp (M) ..= space of all vectors at p.

d
dt

λ(t)
p

Tp(M)
B DIFFERENTIAL GEOMETRY 18

A vector is a derivative operator! It obeys:

(i) Linearity: α, β ∈ R, f, g ∈ C ∞ ⇒ V (αf + βg) = αV (f ) + βV (g)
(ii) Leibniz rule: f, g ∈ C ∞ ⇒ V (f g) = V (f ) g(p) + f (p) V (g)

dxµ  ∂

d µ
f (xα )
α



Consider coordinates x ⇒ V (f ) = f x λ(t) =
dt dt λ ∂xµ
% ↑ -

vector components basis vectors

∂
One can show: Tp (M) has dimension n and eµ ..= ∂µ ..= form a basis.
∂xµ

dxµ dxµ ∂ d
Components: V µ = ⇒ V = V µ eµ = V µ ∂µ = µ
=
dt dt ∂x dt

∂ ∂ ∂xµ ∂ ∂xµ
Coordinate change xµ → x̃α : eµ = → ẽα = = = eµ
∂xµ ∂ x̃α ∂ x̃α ∂xµ ∂ x̃α
dxµ dx̃α ∂ x̃α dxν ∂ x̃α ν
Vµ = → Ṽ α = = = V
dt dt ∂xν dt ∂xν
⇒ V = V µ eµ = Ṽ α ẽα is invariant!

∂µ is a coordinate basis. Non-coordinate bases also exist but we do not consider them.

Covectors / one-forms

Def.: A covector or one-form is a linear map

η : Tp (M) → R , V 7→ η(V )

Tp∗ (M) ..= Cotangent space of all covectors at p ∈ M. Tp∗ (M) is an n dimensional vector
space. Let eµ be a basis of Tp (M). The components of a covector η are ηµ ..= η(eµ ).

Properties: Linearity: α, β ∈ R, V , W ∈ Tp (M) ⇒ η(αV + βW ) = αη(V ) + βη(W )

Components: η(V ) = η(V µ eµ ) = V µ η(eµ ) = V µ ηµ

Transformation rule: We require η(V ) to be a scalar

! ∂ x̃α µ ∂xµ
⇒ η(V ) = ηµ V µ = η̃α Ṽ α = η̃α V ⇒ η̃β = ηµ
∂xµ ∂ x̃β
B DIFFERENTIAL GEOMETRY 19

d df
Def.: Gradient df of a smooth function f : df : Tp (M) → R , 7→
dt dt
d df
Let V = ∈ Tp (M) ⇒ df (V ) = = V (f )
dt dt

∂xα
 
α α α ∂
Basis: Let f = x , α fixed ⇒ dx (eβ ) = dx = = δαβ
∂xβ ∂x β

⇒ ηα dxα (V ) = ηα dxα (V β ∂β ) = ηα V β dxα (∂β ) = ηα V β δ α β = ηα V α = η(V )

⇒ η = ηα dxα

Tensors
r
, r, s ∈ N0 , is a multilinear map


Def. : A tensor T at p ∈ M of rank s

T : Tp∗ (M) × . . . × Tp∗ (M) × Tp (M) × . . . × Tp (M) → R

| {z } | {z }
r factors s factors

Plug in r one-forms and s vectors and out pops a real number.

0


Examples: • Covector η = tensor of rank 1
: input V , output η(V )

• Vector V can be viewed as: V : Tp∗ (M) → R , η 7→ η(V )

⇒ V is a 10 tensor.

Components of V : η(V ) = ηα dxα (V ) = ηα V α ⇒ V α = dxα (V ) = V (dxα )

This holds for all tensors: T α1 ...αr β1 ...βs = T (dxα1 , . . . , dxαr , eβ1 , . . . , eβs )

• δ : Tp∗ (M) × Tp (M) → R , (η, V ) 7→ η(V ) ∀ η ∈ Tp∗ (M) , V ∈ Tp (M)

1 ∂xα
tensor with components δ(dxα , ∂β ) = dxα (∂β ) = = δαβ


is a 1
∂xβ
r


One can show that tensors of rank s
form a vector space of dimension nr+s and transform
according to

∂ x̃α1 ∂ x̃αr ∂xν1 ∂xνs µ1 ...µr

T̃ α1 ...αr β1 ...βs = . . . . . . T ν1 ...νs
∂xµ1 ∂xµr ∂ x̃β1 ∂ x̃βs
B DIFFERENTIAL GEOMETRY 20

Tensor operations

(1) Addition, scalar multiplication: Let c1 , c2 ∈ R, S, T 1

1
tensors

⇒ c1 S + c2 T : Tp∗ (M) × Tp (M) → R , η, V 7→ c1 S(η, V ) + c2 T (η, V )

0


(2) (Anti-) symmetrization: E.g. for 2
tensor T
1
symmetric part: Sαβ ..= (Tαβ + Tβα ) =.. T(αβ)
2
1
antisymm. part: (Tαβ − Tβα ) =.. T[αβ]
2
1
Index subset: T (αβ)γ δ ..= (T αβγ δ + T βαγ δ )
2
1
non-adjacent indices: T(α|βγ|δ) ..= (Tαβγδ − Tδβγα )
2
Over n > 2 indices: • sum over all permutations

• apply sign of permutation for anti-symm.

• divide by n!

1 α
E.g.: T α [βγδ] = T βγδ + T α δβγ + T α γδβ − T α δγβ − T α γβδ − T α βδγ


3!

(3) Contraction of rs tensor ..= Summation over 1 upper and 1 lower index

→ r−1


s−1
tensor

Example: Let T be a 32 tensor

⇒ 21 tensor S(ω, η, V ) ..= T (dxµ , ω, η, ∂µ , V )

This is basis independent:

 ∂  ∂ x̃µ ∂xβ  α ∂ 
T dx̃µ , ω, η, µ , V = α ∂ x̃µ
T dx , ω, η, β
, V = T (dxα , ω, η, ∂α , V )
∂ x̃ ∂x
| {z } ∂x
=δ β α
µν αµν
Components: S ρ =T αρ
B DIFFERENTIAL GEOMETRY 21

p r



(4) Let S be a q
tensor, T a tensor
s

“outer product” S ⊗ T is a p+r

q+s
tensor with


S ⊗ T (ω 1 , . . . , ω p , η 1 , . . . , η r , X 1 , . . . , X q , Y 1 , . . . , Y s )
..= S(ω 1 , . . . , ω p , X 1 , . . . , X q ) T (η 1 , . . . , η r , Y 1 , . . . , Y s )

One straightforwardly shows:

α ...αp β1 ...βr
(i) S ⊗ T 1 µ1 ...µq ν1 ...νs = S
α1 ...αp
µ1 ...µq T
β1 ...βr
ν1 ...νs

2


(ii) In a coord. basis, a 1
tensor can be written as

T = T µν ρ eµ ⊗ eν ⊗ dxρ

likewise (r, s) tensor

Tensor fields

So far: tensors at point p ∈ M

Def.: Tensor field of rank rs ..= collection
of rs tensors at each p ∈ M.

Like a map : p → T p of rank rs . The tensor field is smooth :⇔

its components in a coordinate basis are smooth functions .

Sometimes we write X p = vector, X = field. Often it’s clear from context.

Example: Vector field (VF) X : M → Tp (M) , p 7→ X p

For function f : X(f ) : M → R , p 7→ X p (f ) is a function

Henceforth we assume all tensor fields to be smooth.

Integral curves

Def.: Integral curve λ of a VF V through p ∈ M ..= curve through p whose tangent at every point
q along the curve is V q .


dxµ λ(t)

d 
= V µ (xα )


In coords: =V ⇒ with xµ λ(t0 ) = xµ (p)
dt λ dt
Has a unique solution by ODE theory.
B DIFFERENTIAL GEOMETRY 22

B.2 The metric tensor

Goal: Measure distances, volumes, etc. → need metric!

Def.: A metric at p ∈ M ..= 02 tensor that is:

(i) symmetric: g(V , W ) = g(W , V ) ∀ V , W ∈ Tp (M) ⇔ gαβ = gβα

(ii) non-degenerate: g(V , W ) = 0 ∀ W ∈ Tp (M) ⇔ V =0

Components: g = gαβ dxα ⊗ dxβ , gµν = g(∂µ , ∂ν ) , ds2 = gαβ dxα dxβ

A metric maps vectors to 1-forms:

V 7→ g(V , . ) =: V , i.e. V : Tp (M) → R , W 7→ V (W ) ..= g(V , W ) = V µ W µ = gµν V µ W ν

Components: Vµ ..= V µ = gµν V ν

g non-degenerate ⇒ g invertible
2
Def.: inverse metric g −1 ..= symmetric


0
tensor g αβ with g αβ gβγ = δ α γ

Example: Line element on the unit sphere, x2 + y 2 + z 2 = 1 in R3 : ds2 = dθ2 + sin2 θ dφ2 ,
1 0 1 0
! !
gαβ = , g αβ =
2
0 sin θ 0 sin12 θ

g −1 maps 1-forms to vectors: g −1 (η, . ) (ω) ..= g −1 (η, ω)

The metric mappings between vectors and 1-forms are inverses of each other:

g −1 g(V , . ), . = V , g g −1 (η, . ), . = η

→ natural isomorphism

Signature

g symmetric ⇒ components of g at p ∈ M are a symmetric matrix

⇒ ∃ basis where gµν is diagonal

g non-degenerate ⇒ all diagonal elements are 6= 0

⇒ we can rescale the basis such that the diagonal elements = ±1

“orthonormal basis” ← basis non-unique!

B DIFFERENTIAL GEOMETRY 23

“Sylvester’s law” ⇒ the number of +1 and −1 entries is independent of basis

Def.: Signature ..= sum +1, −1 over all diagonal elements

Riemannian metrics: signature = + + . . . + or +n = # of dims.

Lorentzian metrics: − + + . . . + or n − 2. Some people use + − − . . . −

Note: Equivalence principle

⇒ in a local inertial frame, the laws of SR hold

⇒ ∃ coords: metric gµν = ηµν = diag(−1, 1, 1, 1) “Lorentz invariant”

Only possible locally! At q 6= p, gµν 6= ηµν in general

Def.: A Riemannian (Lorentzian) manifold

..= (M, g) where M is a diff. manifold and g a Riemannian (Lorentzian) metric

spacetime ..= Lorentzian manifold

Example: Minkowski metric in R4 with Cartesian coords. x0 , x1 , x2 , x3 :

η = −(dx0 )2 + (dx1 )2 + (dx2 )2 + (dx3 )2 , (dx0 )2 ≡ dx0 ⊗ dx0 , . . .

Def.: Let (M, g) be a Lorentzian manifold, V ∈ Tp (M), V 6= 0

V is timelike :⇔ g(V , V ) < 0 timelike

null
null :⇔ g(V , V ) = 0
spacelike

spacelike :⇔ g(V , V ) > 0

local inertial frame: gµν = ηµν

⇒ locally we have the light cone structure of SR

p
Def.: Norm of spacelike vector V ∈ Tp (M) is |V | ..= g(V , V ).
 
Angle between spacelike V , W ∈ Tp (M) is θ ..= arccos g(V ,W )
|V | |W |
.
B DIFFERENTIAL GEOMETRY 24

B.3 Geodesics
Curves

Def.: A curve is timelike (null, spacelike) at a point p ∈ M

:⇔ its tangent vector at that point is timelike (null, spacelike).

Note: This can change along a curve λ(t).

Def.: Length along a spacelike curve

r
t1 t1
dxα dxβ d
Z q Z
s ..= g(V , V )|λ(t) dt = gαβ dt , V = .
t0 t0 dt dt dt

Proper time along timelike curve λ(t)

r
t1 t1
dxα dxβ
Z q Z
τ (t1 ) ..= − g(V , V )|λ(t) dt = −gαβ dt .
t0 t0 dt dt

dxµ 

Def.: Four velocity along timelike curves uµ ..=
dτ λ(τ )

⇒ gµν uµ uν = −1

Noether’s theorem
Z
Action S = L(qk , q̇k , λ) dλ is extremized by the curve satisfying the
 
d ∂L ∂L
Euler-Lagrange eqs.: =
dλ ∂ q̇k ∂qk

Noether’s theorem: (i) L not explicitly dependend on qk

∂L
⇒ pk ..= conserved along curve that extremizes S.
∂ q̇k
(ii) L not explicitly dependend on parameter λ
∂L
⇒ I ..= q̇k − L conserved along curve.
∂ q̇k
B DIFFERENTIAL GEOMETRY 25

Geodesics, variation 1
B

Consider timelike curves from A to B.

WLoG: λ = 0 (1) at A (B).

Z 1
p
S= Ldλ , L = −gµν ẋµ ẋν xα(λ)
0
A
dκ
Note: S invariant under reparametrization κ(λ), >0:
dλ

1
r κ(1)
r
dxµ dxν dxµ dxν
Z Z
S= −gµν dλ = −gµν dκ
0 dλ dλ κ(0) dκ dκ

∂L 1 µ ν µ ν gµα ẋµ
EL eqs.: = (−gµν δ α ẋ − gµν ẋ δ α ) = −
∂ ẋα 2L L
∂L 1
α
= (−ẋµ ẋν ∂α gµν )
∂x 2L
gµα ẋµ ẋµ ẋν ∂α gµν
 
d
⇒ − + =0
dλ L 2L
Z λr
dxµ dxν dτ
Change parameter: τ (λ) = −gµν dλ̃ ⇒ =L
0 dλ̃ dλ̃ dλ
dxµ L dxµ dxν
 
d
⇒ −L gµα + ∂α gµν = 0
dτ dτ 2 dτ dτ
d 2 xµ dxν dxµ 1 dxµ dxν 
⇒ g + ∂ g − ∂ g =0  · g βα

µα ν µα α µν
dτ 2 dτ dτ 2 dτ dτ

d2 xβ  β dxµ dxν
⇒ + µν =0 (†)
dτ 2 dτ dτ

 β
. 1 βρ
Def.: Christoffel symbols: µ ν
.= g (∂µ gνρ + ∂ν gρµ − ∂ρ gµν )
2

symmetric in µ, ν !

p ds
For spacelike geodesics: L̄ = gµν ẋµ ẋν , = L̃ ⇒ Eq. (†) with τ → s
dλ
B DIFFERENTIAL GEOMETRY 26

Geodesics, variation 2
B
dxα dxβ
Z
Alternatively: Ŝ = L̂dλ , L̂ = gαβ
A dλ dλ
Differences: (1) No restriction to timelike geodesics.

(2) Not invariant under reparamterization.

α d
ẋν ẋβ = 0 ,

Euler-Lagrange Eqs. ⇒ ... ⇒ ẍα + ν β
˙ = (?)
dλ
So far so good. But now take (†) and let
2
d2 d2 λ d d2

dτ d dλ d dλ
τ = τ (λ), >0 ⇒ = ⇒ = +
dλ dτ dτ dλ dτ 2 dτ 2 dλ dτ dλ2
 −2 2
d2 xα  α dxν dxβ dλ d λ dxα dxα
⇒ + ν β
= − ∝ (??)
dλ2 dλ dλ dτ dτ 2 dλ dλ
That’s not (?) above! What’s going on?

Answer: Ŝ not invariant under parameter change ⇒ Variation gives different curve

d2 λ
Eqs. (†), (?) only agree if =0 ⇔ λ = c1 τ + c2 , c1 , c2 = const ∈ R (††)
dτ 2

Def.: The parameter λ along a timelike (spacelike) curve is affine ⇔ it is related to

proper time τ (proper distance s) by Eq. (††). For a non-affine parameter, the geodesic
equation is (??).

Summary for all curves (incl. null):

Def.: If a curve C : I ⊂ R → M, λ 7→ xα (λ)

(1) satisfies (?) → geodesic, λ is affine.

(2) satisfies Eq. (??) with non-zero right-hand side → geodesic; λ is non-affine.

(3) satisfies neither (?) nor (??) → it is not a geodesic.

ODE theory ⇒ solutions to (?), (??) unique if xα , ẋα fixed at λ = λ0

Geodesic postulate: Test particles with positive (zero) rest mass move on timelike (null)
geodesics.
B DIFFERENTIAL GEOMETRY 27

α

L̂ gives an easy way to calculate ν β
:

Example: Schwarzschild metric in Schwarzschild coords.:

2M
ds2 = −f dt2 + f −1 dr2 + r2 dθ2 + r2 sin2 θ dφ2 , f =1− , M = const
r
⇒ L̂ = f ṫ2 − f −1 ṙ2 − r2 θ̇2 − r2 sin2 θ φ̇2
d d2 t df
+ f −1 ṫ ṙ = 0


EL for t(τ ): 2f ṫ = 0 ⇒ 2
dτ dτ dr
t t df /dr t

⇒ t r
= r t
= , µ ν
= 0 otherwise
2f

B.4 Covariant derivative

Physical laws involve derivatives.

Problem: Cannot take difference between vectors at different points:

U ∈ Tp (M), V ∈ Tq (M)

→ Covariant derivative ∇ on manifold M

Def.: For functions: ∇f : Tp (M) → R , V 7→ ∇V f ..= V (f ) = V α ∂α f

∇f is 01 tensor: ∇α f ..= (∇f )α = ∂α f

Def.: For vectors: ∇V : Tp (M) → Tp (M) , X 7→ ∇X V with

(1) ∇f X+gY V = f ∇X V + g∇Y V ,

(2) ∇X (V + W ) = ∇X V + ∇X W

(3) ∇X (f V ) = f ∇X V + (∇X f )V (Leibnitz rule)

Equivalently: ∇V : Tp∗ (M) × Tp (M) → R , (η, X) 7→ η(∇X V )

∇V is a 11 tensor: V α ;β ..= ∇β V α ..= (∇V )α β

Def.: Let {eµ } be a basis of Tp (M).

Connection coefficients Γρµν : ∇ν eµ ..= ∇eν eµ = Γρµν eρ

B DIFFERENTIAL GEOMETRY 28

We get for V = V µ eµ , W = W µ eµ

⇒ ∇V W = ∇V (W µ eµ ) = V (W µ )eµ + W µ ∇V eµ

= V ν eν (W µ ) eµ + W µ ∇V ν eν eµ

= V ν eν (W µ ) eµ + W µ V ν ∇ν eµ
| {z }
=Γρµν eρ
 
= V ν ∂ν W ρ + W µ Γρµν eρ

ρ ν ρ ν

⇒ (∇V W ) = V ∂ν W + V Γρµν W µ  V arbitrary


⇒ W ρ ;ν ..= ∇ν W ρ ..= (∇W )ρ ν = ∂ν W ρ + Γρµν W µ

Coordinate transformation xµ → x̃α

∂ x̃σ ∂xα ∂xβ ρ ∂ x̃σ ∂ 2 xρ
⇒ ... ⇒ Γ̃σµν = Γ + (†)
∂xρ ∂ x̃ν ∂ x̃µ βα ∂xρ ∂ x̃ν ∂ x̃µ
⇒ Γσµν is not a tensor! But the difference of two connections is. E.g.:

Def.: Torsion tensor: Tµν λ ..= Γλµν − Γλνµ .

Γ is called torsion free :⇔ Tµν λ = 0.

∂W µ
Comments: • ∂ν W µ = is not a tensor either!
∂xν
Second term in (†) just cancels this to make ∇ν W µ a tensor.

• Example: The Christoffel symbols are a connection.

• Convention: Derivative index = 2nd downstairs index in Γρµν .

Some people use the first.

Covariant derivative of tensors

Obtained from Leibniz rule; rs tensor T r

7→ ∇T is s+1
tensor
B DIFFERENTIAL GEOMETRY 29

E.g. 1-form: ∇V η(W ) ..= ∇V η (W ) + η(∇V W )



⇒ ∇V η (W ) = ∇V η(W ) − η(∇V W )

∇η is a 02 tensor since:

∇η (V , W ) = ∇V η (W ) = ∇V (ηµ W µ ) − ηµ (∇V W )µ

= V ρ ∂ρ (ηµ W µ ) − ηµ V ρ ∂ρ W µ + V ρ Γµνρ W ν

= V ρ W µ ∂ρ ηµ − Γµνρ ηµ V ρ W ν

= ∂ρ ηµ − Γνµρ ην V ρ W µ

Components: ηµ;ρ ..= ∇ρ ηµ ..= (∇η)ρµ = ∂ρ ηµ − Γνµρ ην .

Covariant derivative of (r, s) tensor:

∇ρ T µ1 ...µr ν1 ...ν2 = ∂ρ T µ1 ...µr ν1 ...νs + Γµσρ1 T σµ2 ...µr ν1 ...ν2 + . . . + Γµσρr T µ1 ...µr−1 σ ν1 ...νs

− Γσν1 ρ T µ1 ...µr σν2 ...νs − . . . − Γσνs ρ T µ1 ...µr ν1 ...νs−1 σ

B.5 The Levi-Civita connection

Note: We need no metric for the covariant derivative.

But: A metric singles out a special connection.

Theorem: On a manifold M with metric g ∃ unique connection with

(1) Γαµν = Γανµ = µαν


Christoffel symbols, torsion free

(2) ∇g = 0 “metric compatible”

This connection is called the Levi-Civita connection.

B DIFFERENTIAL GEOMETRY 30

Proof: “⇒”: Let Γαβγ be metric compatible, symmetric

⇒ ∇α gβγ = 0 ⇒ ∂α gβγ = Γρβα gργ + Γργα gβρ

By definition the Christoffel symbols are

 µ
1 µν
β γ
= g (∂β gγν + ∂γ gνβ − ∂ν gβγ )
2  
1 µν ρ ρ ρ ρ ρ ρ
= g Γγβ gρν + Γνβ gγρ + Γνγ gρβ + Γβγ gνρ − Γβν gργ − Γγν gβρ
2 :::::: ::::::

= Γµβγ

α
“⇐”: Likewise: Γαβγ =

β γ

⇒ . . . ⇒ Γαµν = Γανµ , ∇α gβγ = 0

In GR we use the Levi-Civita connection.

B.6 Parallel transport

A connection gives us a notion of “a tensor that does not change along a curve”

Def.: Let V be a vector field and C an integral curve of V . A tensor T is parallel transported
along C :⇔ ∇V T = 0 along the curve.

Comments: • The tangent of an affinely parametrized geodesic is parallel transported along

itself:
dxµ ∂ dxα dxµ α dxν
U µ ∇µ U α = U µ ∂µ U α + U µ Γανµ U ν = + Γ =0
dλ ∂xµ dλ dλ νµ dλ
d 2 xα ν
α dx dx
µ
⇒ + Γ νµ =0
dλ2 dλ dλ
with non-affine parameter: ∇U U = f U
B DIFFERENTIAL GEOMETRY 31

• ∇V T = 0 determines T uniquely along the curve:

in coords. (xµ ) the curve is xµ (λ)

⇒ V σ ∇σ T µ ν = V σ ∂σ T µ ν + Γµρσ T ρ ν V σ − Γρνσ T µ ρ V σ
d µ
= T ν + Γµρσ T ρ ν V σ − Γρνσ T µ ρ V σ = 0
dλ
ODE theory ⇒ unique solution for all T µ ν
• parallel transport T along a curve from p to q

→ isomorphism between tensors at p, q

Unlike SR, this is path dependent in GR!

• Parallel transport preserves length of vectors:

d
(Wα W α ) = V µ ∇µ (Wα W α ) = 2Wα V µ ∇µ W α
dλ | {z }
=0

⇒ Geodesics do not change their timelike, spacelike or null character.

Def.: Acceleration along timelike curve: aµ ..= uρ ∇ρ uµ

The curve is a geodesic if aµ = 0 or aα = f uα

B.7 Normal coordinates

Def.: Let M be a manifold, Γ a connection, p ∈ M.

“exponential map” : e : Tp → M , X p 7→ q with

q ..= point a unit affine parameter distance along

geodesic through p with tangent X p

Comments: 1) e can be shown to be one-to-one and onto locally

2) The vector X p fixes the parametrization of the geodesic:

One can show that λ X p , 0 ≤ λ ≤ 1 is mapped to point at

affine par. distance λ along the geodesic of X p . (∗∗)

B DIFFERENTIAL GEOMETRY 32

Def.: Let {eµ } be a basis of Tp (M). “Normal coords. in nbhd. of p ∈ M”:

coords. that assigns to q = e(X) ∈ M the coordinates X µ

Note: The coords. X µ are not fixed by the vector X.

We still have the freedom to choose a basis for Tp (M).

Lemma: In normal coordinates, Γµ(νρ) = 0 at p.

If Γ is torsion free, then Γµνρ = 0.

Proof: From (∗∗) ⇒ affinely parametrized geodesic is given by

xµ (λ) = λ Xpµ in normal coords.

dxν dxρ
⇒ geodesic eq.: 0 + Γµνρ = Γµνρ Xpν Xpρ = 0 at p ∀X ∈ Tp (M)
dλ dλ
⇒ Γµ(νρ) = 0

torsion free ⇒ Γµ[νρ] = 0 ⇒ Γµνρ = 0

Note: in general Γµνρ 6= 0 away from p !

Lemma: With Levi-Civita connection ⇒ in normal coords. at p: ∂ρ gµν = 0

Proof: Γρµν = 0 ⇒ 2gσρ Γρµν = ∂ν gσµ + ∂µ gνσ − ∂σ gµν = 0

symmetrize on σ, µ, add ⇒ ∂ν gσµ = 0

Lemma: Let (M, g) be a spacetime with LC connection

⇒ ∃ coordinates at p with: ∂ρ gµν = 0 , gµν = ηµν = diag(−1, +1, +1, +1)

Proof: Choose an orthonormal basis {eµ } for Tp (M)

⇒ at p: X = X 0 e0 + . . . + X 3 e3 defines normal coords. xµ = X µ

(∗∗) ⇒ point an affine par.distance along geodesic through p with

tangent e0 has coords. λ (e0 )µ = (λ, 0, 0, 0)

⇒ The geodesic is xµ (λ) = (λ, 0, 0, 0)

In any coordinate system, the tangent to the curve (λ, 0, 0, 0) is ∂0 = ∂/∂x0

Likewise ∂µ = eµ ⇒ {∂µ } is orthonormal ⇒ gµν = g(∂µ , ∂ν ) = ηµν at p .

B DIFFERENTIAL GEOMETRY 33

Summary: Locally, we can choose coordinates such that the metric is ηµν = diag(−1, 1, 1, 1)
and its first derivatives vanish.

Def.: “local inertial frame at p ∈ M00 := normal coord. chart with these properties

B.8 The Riemann tensor

The commutator

Def.: Commutator of 2 vectorfields V , W : [V , W ]α ..= V µ ∂µ W α − W µ ∂µ V α

[X, Y ] is indeed a vectorfield:

∂ x̃µ
Ṽ ν ∂˜ν W̃ µ − W̃ ν ∂˜ν Ṽ µ = β γ β γ


One shows ... = V ∂ β W − W ∂ β V
∂xγ

Properties: [V , W ] = −[W , V ]

[V , W + U ] = [V , W ] + [V + U ]

[V , f W ] = f [V , W ] + V (f ) W

[U , [V , W ]] + [V , [W , U ]] + [W , [U , V ]] = 0 “Jacobi identity”
 
∂ ∂
Note: , =0 (coord. basis ⇒ commutators vanish)
∂xµ ∂xν
Conversely, one can show:

If V 0 , . . . , V m−1 , m ≤ dim(M) are vector fields which are

lin. indep. ∀p ∈ M with all [V i , V j ] = 0

∂
⇒ In a nbhd. of p ∃ coords. (xµ ): Vi= , i = 0, . . . , m − 1
∂xi

Second derivatives

For function f : ∇ν ∇µ f = ... = ∇µ ∇ν f − 2Γρ[µν] ∂ρ f = ∇µ ∇ν f for torsion free Γ

For vector V : ∇α ∇β V γ − ∇β ∇α V γ = ...

= ∂α Γγρβ V ρ + Γγρβ ∂α V ρ + Γγρα ∂β V ρ + Γγρα Γρσβ V σ − ( α ↔ β )

B DIFFERENTIAL GEOMETRY 34

Def.: Riemann tensor: Rγ ραβ := ∂α Γγρβ − ∂β Γγρα + Γµρβ Γγµα − Γµρα Γγµβ .

⇒ “Ricci Identity”: ∇α ∇β V γ − ∇β ∇α V γ = Rγ ραβ V ρ

Equivalently, we can define:

Def.: For 3 VFs U , V , W : R(U , V ) (W ) = ∇U ∇V W − ∇V ∇U W − ∇[U ,V ] W (‡)

Proof: One shows for a function f : R(f U , V )W = f R(U , V )W

R(U , f V )W = f R(U , V )W

R(U , V )f W = f R(U , V )W

rhs of (‡) is a vector ⇒ Contraction of R(U , V )W with one-form is linear.

Components: [eα , eβ ] = 0, ∇α eβ = Γµβα eµ

⇒ R(eα , eβ )eρ = ∇α ∇β eρ − ∇β ∇α eρ = ∇α (Γµρβ eµ ) − ∇β (Γµρα eµ )

= . . . = ∂α Γνρβ − ∂β Γνρα + Γµρβ Γνµα − Γµρα Γνµβ eν

 
| {z }
=Rν ραβ !

Symmetries of Riemann

(1) Rα βγδ = −Rα βδγ ⇔ Rα β(γδ) = 0 by def.

Torsion = 0, let p ∈ M, (xµ ) normal coords. Then:

(2) Γµνρ = 0 at p, Γµ[νρ] = 0 everywhere


⇒ Rµ νρσ = ∂ρ Γµνσ − ∂σ Γµνρ  antisymmetrize on νρσ


⇒ Rµ [νρσ] = 0 tensorial equation!


µ µ
(3) ∇τ R = ∂τ R  “∂R = ∂∂Γ − Γ ∂Γ = ∂∂Γ”

νρσ νρσ


= ∂τ ∂ρ Γµνσ − ∂τ ∂σ Γµνρ  antisymmetrize on ρστ


⇒ Rµ ν[ρσ;τ ] = 0 “Bianchi identity”

B DIFFERENTIAL GEOMETRY 35

(4) Levi-Civita connection, p ∈ M ⇒ in normal coords.: ∂µ gνρ = 0


⇒ 0 = ∂µ δ ν ρ = ∂µ (g νσ gσρ ) = gσρ ∂µ g νσ  · g ρτ


⇒ ∂µ g ντ = 0
1
⇒ ∂ρ Γτνσ = g τ µ ∂ρ ∂σ gµν + ∂ρ ∂ν gσµ − ∂ρ ∂µ gνσ


2
1

⇒ Rµνρσ = ∂ρ ∂ν gσµ + ∂σ ∂µ gνρ − ∂σ ∂ν gρµ − ∂ρ ∂µ gνσ + “ΓΓ − ΓΓ”
2 | {z }
=0

⇒ Rµνρσ = Rρσµν

It follows: Rαβγδ = −Rβαγδ

Parallel transport and curvature

Let X, Y be VFs with: lin. indep. everywhere and [X, Y ] = 0; let torsion = 0
∂ ∂
⇒ we can choose coords. (s, t, . . .) such that X = , Y =
∂s ∂t
X r (δs,δt)
Let p, q, r, u ∈ M along integral curves of X, Y with coords. u (0,δt)
(0, . . . , 0), (δs, 0, . . .), (δs, δt, 0, . . .), (0, δt, 0, . . .)

Let Z p ∈ Tp (M) , parallel trapo Z along pqrup

Y Y
to get Z 0p ∈ Tp (M)

(Z 0p − Z p )α p (0,0) X q ( δs,0)
⇒ lim = (Rα βγδ Z β Y γ X δ )p
δs,δt→0 δs δt

Proof: long script.

B DIFFERENTIAL GEOMETRY 36

Geodesic deviation

Goal: quantify relative acceleration of geodesics

Def.: Let (M, Γ) be a manifold with connection.

“1-parameter family of geodesics” := a map

γ : I × I 0 → M with I, I 0 ⊂ R, open and

(i) for fixed s, γ(s, t) is a geodesic with affine par. t

(ii) locally, (s, t) 7→ γ(s, t) is smooth, 1-to-1 has smooth inverse

⇒ the family of geodesics forms a 2-dim. surface Σ ⊂ M

Let T be the tangent vector to γ(s = const, t) and S to γ(s, t = const)

dxµ
In coords. (xµ ): S µ =
ds
s= const
⇒ xµ (s + δs, t) = xµ (s, t) + δs S µ (s, t) + O(δs2 )

⇒ δs S points from one geodesic to a nearby one: “deviation vector” T

⇒ “relative velocity” of nearby geodesics: ∇T (δs S) = δs ∇T S

⇒ “relative acceleration” of nearby geodesics: δs ∇T ∇T S

S t= const

Theorem: ∇T ∇T S = R(T , S)T

⇔ T µ ∇µ (T ν ∇ν S α ) = Ra λµν T λ T µ S ν

Lemma: V µ ∇µ W α − W µ ∇µ V α = V µ ∂µ W α + V µ Γαρµ W ρ − W µ ∂µ V α − W µ Γαρµ V ρ

= V µ ∂µ W α − W µ ∂µ V α = [V , W ]α

⇒ ∇V W − ∇W V = [V , W ]
B DIFFERENTIAL GEOMETRY 37

Proof: Use coords. (s, t) on Σ and extend to (s, t, . . .) in nbhd. of Σ

∂ ∂
⇒S= , T = ⇒ [S, T ] = 0
∂s ∂t
No torsion ⇒ ∇T S − ∇S T = [T , S] = 0

⇒ ∇T ∇T S = ∇T ∇S T = ∇S ∇T T +R(T , S)T 2
| {z }
= 0 geodesic!

Comments: • Rα βγδ measures geodesic deviation; manifestation of curvature.

• Rα βγδ = 0 ⇔ relative acceleration = 0 for all families of geodesics.

• Tidal forces arise from geodesic deviation.

The Ricci tensor

Def.: The “Ricci tensor” is Rαβ ..= Rµ αµβ .

The “Ricci scalar” is R ..= g µν Rµν .

1
The “Einstein tensor” is Gαβ − gαβ R .
2

Recall Bianchi Identity: Rαβ[γδ;µ] = 0  · g αγ g βδ


1
⇒ g αγ g βδ Rαβγδ;µ + Rαβδµ;γ + Rαβµγ;δ − Rαβδγ;µ −Rαβγµ;δ − Rαβµδ;γ = 0
 
6 | {z }
=−Rαβγδ;µ

1
⇒ g αγ g βδ Rαβγδ;µ + Rαβδµ;γ + Rαβµγ;δ = 0
 
3
⇒ R;µ − g αγ Rαµ;γ − g βδ Rβµ;δ = 0
 
γ γ 1
⇒ ∇µ R − 2∇γ R µ = −2∇ Rγµ − gγµ R
2

⇒ ∇µ Gµα = 0 “Contracted Bianchi Id.”

C PHYSICAL LAWS IN CURVED SPACETIMES 38

C Physical laws in curved spacetimes

Goal: How do we use differential geometry to describe physics?

C.1 The covariance principle

What we have so far:
• Equivalence principle: freely falling frame = inertial frame in SR
• Normal coordinates: ∃ coordinates: gαβ = ηαβ , Γαβγ = 0
• laws of SR invariant under Lorentz trafos

this motivates the

Covariance principle: The laws of GR are tensorial, i.e. coordinate invariant, and follow from
SR laws by replacing

(1) the metric: ηµν → gµν .

(2) the derivative: ∂ → ∇ .

Example: Electromagnetic field in SR:

Fµν = F[µν] with F0i = −Ei , Fij = ijk Bk , (i, j, k = 1 . . . 3)

vacuum Maxwell eqs.: η µν ∂µ Fνρ = 0 , ∂[µ Fνρ] = 0

→ in GR: g αβ ∇α Fβγ = 0 , ∇[α Fβγ] = 0

Note: Step SR → GR not unique since Rαβ = 0 in SR

C.2 The energy momentum tensor

Postulate: Mass-energy, stress in GR is described by the “energy momentum tensor” with

(1) T αβ ..= flux of α momentum across a surface of constant xβ

(2) Tαβ = Tβα , ∇µ Tµν = 0

Interpretation: T αβ = T (dxα , dxβ )

T 00 = 0 momentum across t = const = energy density

T i0 = xi momentum density

T 0i = energy flux across surface xi = const

T ij = flux of i momentum across xj = const

C PHYSICAL LAWS IN CURVED SPACETIMES 39

Tαβ often obtained from SR + covariance principle

Particles

1) in SR: Point like object with 4-momentum P µ = muµ = (E, P i )

4-velocity of observer in particle’s rest frame: wµ = (1, 0, 0, 0)

particle energy measured by this observer: E = −ηµν wµ P ν

particle’s rest mass: ηµν P µ P ν = −E 2 + p~ 2 = −m2

2) in GR: Covariance ⇒ P α = muα ⇒ gαβ P α P β = −m2

Particle energy measured by observer: E = −gαβ (p) wα (p) P β (p)

works only if both are at p!

electromagnetic field

1) pre-relativistic notation, Cartesian coordinates:

1
energy density: ρ = (Ei Ei + Bi Bi )
8π
1
momentum density, energy flux: ji = 4π ijk Ej Bk “Poynting vector”
 
1 1
stress tensor: Sij = (Ek Ek + Bk Bk )δij − Ei Ej − Bi Bj
4π 2
∂ρ ∂ji
Maxwell eqs. ⇒ + ∂i ji = 0 , + ∂j Sij = 0
∂t ∂t

2) Special relativity:
 
1 ρ 1 ρσ
Tµν = Fµρ Fν − F Fρσ ηµν = Tνµ
4π 4
T00 = ρ , T0i = −ji , Tij = Sij ; from 1)

Conservation: ∂ µ Tµν = η µσ ∂σ Tµν = 0

 
1 µ 1 µν
3) GR: Tαβ = Fαµ Fβ − F Fµν gαβ with ∇µ Tµα = 0
4π 4
C PHYSICAL LAWS IN CURVED SPACETIMES 40

Dust

Continuum limit of non-interacting particles of rest mass m and number density n

⇒ Particles are freely falling!

In comoving frame: uµ = [1, 0, 0, 0], locally gαβ = ηαβ

⇒ ρ = mn , T i0 = T 0j = 0 , T ij = 0

⇒ T αβ = ρuα uβ = mnuα uβ .

Perfect fluids

Def.: Continuous matter with (1) no heat conduction, (2) no viscosity in locally comoving
frame.

(1) = energy can only flow if the particles flow.

(2) = Force between particles only has radial component.

⇒ For particles on xi axis, only pi momentum can flow.

⇒ T ij 6= 0 only for i = j.

No direction preferred ⇒ T 11 = T 22 = T 33 =.. P

In SR: In comoving frame T αβ = diag(ρ, P, P, P ) ; uα = (1, 0, 0, 0) , g αβ = η αβ

⇒ T αβ = (ρ + P )uαβ + P η αβ

Covariance principle ⇒ T αβ = (ρ + P )uα uβ + P g αβ

Conservation ∇α T αβ = 0

⇒ uα ∇α ρ + (ρ + P )∇α uα = 0

(ρ + P )uα ∇α uβ = −(g αβ + uα uβ )∇α P GR’s version of Euler equation

Comments: • We still need an equation of state describing the matter. E.g. polytrope P ∝ ρΓ

• P =0 ⇒ fluid moves on geodesics

C PHYSICAL LAWS IN CURVED SPACETIMES 41

C.3 The Einstein equations

Postulates of GR

(1) Spacetime is a 4-dim. Lorentzian manifold with metric and Levi-Civita connection.

(2) Free particles follow timelike or null geodesics.

(3) Energy, momentum and stress of matter is described by a symmetric,

conserved tensor Tαβ : ∇α Tαβ = 0 .

(4) Curvature is related to matter by the Einstein eqs.

1 8πG
Gαβ = Rαβ − gαβ R = 4 Tαβ ; G = Newton’s constant
2 c

8πG
Comments: • Factor from Newtonian limit (cf. below)
c4
1 
• Vacuum ⇒ Gαβ = Rαβ − gαβ R = 0  · g αβ

2

⇒R=0 ⇒ Rαβ = 0

• G = 8πT represents 10 coupled, non-linear PDEs. Hard to solve!

Theorem: (Lovelock 1972) Let Hαβ be a symmetric tensor with

(i) in any coords. Hµν = Hµν (gµν , ∂ρ gµν , ∂σ ∂ρ gµν ) at every p ∈ M

(ii) ∇α Hαβ = 0

(iii) Hµν linear in ∂σ ∂ρ gµν

⇒ ∃A,B∈R Hαβ = A Gαβ + B gαβ

8πG
⇒ we can modify Einstein’s eqs.: Gαβ + Λgαβ = Tαβ
c4

Λ = Cosmological constant
Λc4
Λgαβ is equivalent to perfect fluid with ρ = −P =
8πG
D THE SCHWARZSCHILD SOLUTION AND CLASSIC TESTS OF GR 42

D The Schwarzschild solution and classic tests of GR

Einstein was pessimistic about finding solutions to G = 8πT .

But 1915 Schwarzschild found his “black hole” solution. Key simplification: symmetry.

D.1 Schwarzschild’s solution

Symmetric spacetimes

Def.: A spacetime (M, g) is “symmetric in a variable s” if ∃ coordinates xα such that one of the
xα = s and gαβ are independent of s in this coordinate system.

Def.: A spacetime (M, g) is “stationary” if ∃ coordinates xα such that x0 is a timelike coordinate

and gαβ do not depend on x0 .

Def.: A spacetime (M, g) is “static” if it is stationary and in that coordinate system g0i = 0 for
i = 1, 2, 3.

Note: Time reversal t → −t ⇒ g0i dt dxi → −g0i dt dxi

Invariance under time reversal implies g0i = 0: static spacetimes.

Spherically symmetric spacetimes

∃ point O: spacetime is invariant under rotations about O

⇒ . . . ⇒ The angular part of the line element is ∝ 2-sphere metric: dΩ2 ..= dθ2 + sin2 θ dφ2

We assume the metric is invariant under reflection θ → π − θ , φ → −φ: no mixed φ, θ terms

⇒ ds2 = −Ãdt̃2 + 2B̃dt̃ dr̃ + C̃dr̃2 + D̃dΩ2

where Ã, B̃, C̃, D̃ are functions of (t̃, r̃) and D̃ > 0
p
(1) r ..= D̃ ⇒ ds2 = −Â(t̃, r)dt̃2 + 2B̂(t̃, r)dt̃ dr + Ĉ(t̃, r)dr2 + r2 dΩ2

(2) ODE theory ⇒ ∃ integrating factor I(t̃, r) with

 
I(t̃, r) − Â(t̃, r) dt̃ + B̂(t̃, r) dr = dt̂

1 B̂ 2
⇒ . . . ⇒ ds2 = −j(t̂, r)dt̂2 + k(t̂, r)dr2 + r2 dΩ2 with j = , k = Ĉ +
ÂI 2 Â
D THE SCHWARZSCHILD SOLUTION AND CLASSIC TESTS OF GR 43

∂t̂ k
(3) Now Einstein eqs.: Rt̂ r = =0 ⇒ k = k(r)
k2r
r 2 ∂r k + k 2 − k r
Rt̂ t̂ = =0 ⇒ ... ⇒ k= , M = const
k2 r2 r − 2M
−r∂r j + jk − j r − 2M
Rr r = =0 ⇒ j= f (t̂)
−jkr2 r

(4) signature +2 ⇒ f (t̂) > 0

q
Rescale t̂ such that f (t̂)dt̂ = dt

   −1
2 2M 2 2M
⇒ ds = − 1 − dt + 1 − dr2 + r2 (dθ2 + sin2 θ dφ2 ) (†)
r r

Comments: • (†) = unique solution in spherical symmetry and vacuum. It is static!

• lim gαβ = ηαβ : asymptotic flatness

r→∞

• M can be shown to be the mass-energy of the spacetime.

• (†) also describes the spacetime exterior to spherically symmetric stars.

⇒ Birkhoff ’s Theorem: Any spherically symmetric solution of the vacuum Einstein equations is
given by the Schwarzschild metric. It is static and asymptotically flat.

D.2 Geodesics in the Schwarzschild spacetime

   −1
2M 2 2M
L̂ = − 1 − ṫ + 1 − ṙ2 + r2 θ̇2 + r2 sin2 θ φ̇2
r r
!
d ∂ L̂ ∂ L̂
θ component: ⇒ − = 2r2 θ̈ + 4rṙθ̇ − 2r2 sin θ cos θ φ̇2 = 0
dλ ∂ θ̇ ∂θ

ṙθ̇
⇒ θ̈ + 2 − sin θ cos θ φ̇2 = 0
r
Rotate coordinates such that θ = π/2 , θ̇ = 0 at λ0

⇒ θ = π/2 always WLoG!

D THE SCHWARZSCHILD SOLUTION AND CLASSIC TESTS OF GR 44

∂ L̂ 1 ∂ L̂
 
2M
Noether ⇒ (i) =0 ⇒ E=− = 1− ṫ = const
∂t 2 ∂ ṫ r

∂ L̂ 1 ∂ L̂
(ii) =0 ⇒ L= = r2 sin2 θ φ̇ = r2 φ̇ = const
∂φ 2 ∂ φ̇

∂ L̂
 
2M 2

2M
−1 −1 timelike

2 2 2
(iii) =0 ⇒ Q=− 1− ṫ + 1 − ṙ + r φ̇ = 0 null
∂λ r r 
1 spacelike


Meaning of E, L: Consider r  M ⇒ gαβ → ηαβ ⇒ SR

dt
E: for timelike geodesics with λ = τ : E= =γ
dτ
⇒ E m = mγ ⇒ E = energy per rest mass m

dφ
L: Likewise L m = mr2 φ̇ = mγr2 = angular momentum per rest mass
dt
Plug E, L into Eq. for Q
   
2 2 1 2M 2 2M
⇒ −E + ṙ + 2 1 − L = 1− Q
r r r

L2
  
1 2 1 1 2M
⇒ ṙ + V (r) = E 2 , V (r) = 1− −Q (‡)
2 2 2 r r2

Comparison with Newtonian equations

Energy balance for particle in Newtonian gravity

1 1 1 m L2
Ekin = mṙ2 + mr2 φ̇2 = mṙ2 +
2 2 2 2 r2
Mm
Epot = −G , Let’s set G = 1
r
1 1 L2 M m
⇒ mṙ2 + m 2 − = Ekin + Epot = const
2 2 r r
2
Eq. (‡) for Q = −1: (E 2 − 1)m/2 = Ekin + Epot − m MrL3

M L2
⇒ GR has extra term −
r3
D THE SCHWARZSCHILD SOLUTION AND CLASSIC TESTS OF GR 45

1 2 1 L2 M
Summary: ṙ + VN/GR (r) = const . VN (r) = −
2 2 r2 r
1 L2 M M L2
VGR (r) = + Q −
2 r2 r r3
Effective potential V determines trajectories

Newtonian: For L 6= 0: lim V = ∞ , lim V = 0

r→0 r→∞

L2 M L2
Extrema from: VN0 (r) = − + 2 =0 ⇒ r=
r3 r M
3L2 2M
VN00 (r) = − 3 ⇒ VN00 (L2 /M ) = M 4 /L6 > 0
r4 r

VN Newtonian
2
L 4
⇒ circular orbit at r =
M L/M=0
3 L/M=1
VN min. ⇒ orbit stable L/M=2
L/M=4
L/M=8
2

0
0 5 10
r/M
-1

-2
D THE SCHWARZSCHILD SOLUTION AND CLASSIC TESTS OF GR 46

GR null: lim V = −∞ , lim V = 0

r→0 r→∞

0 L2 M L2
Extrema: VGR (r) = − + 3 =0 ⇒ r = 3M
r3 r4
00 3L2 12M L2 00 L2
VGR (r) = − ⇒ VGR (3M ) = − <0
r4 r5 81M 4
GR null geodesics
VGR
⇒ circular orbit at r = 3M 2

VGR max. ⇒ orbit unstable 1

0
0 5 10
r/M
-1

-2 L/M=1
L/M=2
L/M=4
-3 L/M=8

-4

GR timelike: lim V = −∞ , lim V = 0

r→0 r→∞
r
L2 L4
Extrema: r = r± = ± − 3L2 > 0 for L2 > 12 M 2
2M 4M 2
00 00
For L2 > 12 M 2 : VGR (r+ ) > 0 , VGR (r− ) < 0

GR timelike geodesics
VGR
⇒ stable circular orbit at r+ 2

⇒ unstable circular orbit at r− 1

0
0 5 10
r/M
-1
L/M=0 0
L/M=1
L/M=2
-2 2 2
L / M = 12
L/M=4 -0.05
L/M=8
-3

-4 -0.1
0 5 10 15 20
r/M
D THE SCHWARZSCHILD SOLUTION AND CLASSIC TESTS OF GR 47

D.3 Classic tests of GR

Mercury’s perihelion precession

1 2 1 L2 M m .
Newtonian: mṙ + m 2 − =. E (†)
2 2 r r
φ̇ = L/r2 ⇒ φ(t) monotonic → parametrize orbit with φ
1 dr dy .
y ..= ⇒ ... ⇒ ṙ ..=
= −L =. −Ly 0
r dt dφ
 d
(†) ⇒ (y 0 )2 L2 + L2 y 2 − 2M y = 2E

dφ


⇒ 2L2 y 0 y 00 + 2L2 yy 0 − 2M y 0 = 0
M
⇒ y 0 = 0 ∨ y 00 + y =
L2
M
⇒ y= (1 +  cos φ) Keppler ellipse for  < 1: No perihelion precession
L2

GR: Use again y(φ), Q = −1

2 0 2 2 2 2 3 2
 d
⇒ . . . ⇒ L (y ) + L y − 2M y − 2M L y = E − 1

dφ


M
⇒ y 00 + y = + 3M y 2
L2
M2
Expand solution in α ..= 3 ∼ O(10−7 ) for Mercury
L2
⇒ y = y0 + αy1 + O(α2 )
L2 2
 
00 M 00
⇒ y0 + y0 − 2 + α y1 + y1 − y0 + O(α2 ) = 0
L M
M M
Order α0 : y000 + y0 − =0 ⇒ y0 = (1 +  cos φ) Newtonian case!
L2 L2

Order α1 : Plug in y0 and use cos2 φ = (1 + cos 2φ)/2

2
 
00 M 2M M
⇒ y1 + y1 = 2 1 + + 2  cos φ + 2 2 cos 2φ
L 2 L 2L
Ansatz: y1 = A + Bφ sin φ + C cos 2φ
2 M 2
 
M M
⇒ ... ⇒ A = 2 1 + , B= 2 , C=− .
L 2 L 6L2
D THE SCHWARZSCHILD SOLUTION AND CLASSIC TESTS OF GR 48

  
M M 2 1 1
Solution: y = y0 + αy1 = 2 (1 +  cos φ) + α 2 1 + φ sin φ +  − cos 2φ
L L 2 6
Ignore last term ∝ 2
M
⇒y≈ (1 + α +  cos φ + αφ sin φ)
L2
α1 ⇒ cos(φ − αφ) = cos φ cos αφ + sin φ sin αφ ≈ cos φ + αφ sin φ

M
⇒ y≈ {1 + α +  cos[φ(1 − α)]}
L2

Key result: y returns to the same value as (1 − α)φ increases by 2π

3M 2
 
2π
⇒ Perihelion period: φn+1 − φn = ≈ 2π(1 + α) = 2π 1 + 2
1−α L
Mr
Circular timelike geodesic ⇒ r = r+ ⇒ . . . ⇒ L2 = ≈ Mr
1 − 3M/r
M 4300
⇒ ∆φ ≈ 6π =
r century

Light bending

1) Newtonian
1
Without gravitational field: y 00 + y = 0 ⇒ y= sin φ , straight line
b

r1 r2
b φ1
φ2

light from left (φ = π) to right (φ = 0); b = impact parameter

D THE SCHWARZSCHILD SOLUTION AND CLASSIC TESTS OF GR 49

M
With field: Recall y = (1 +  sin φ) , we shifted Phase: cos → sin
L2
Small deflection ⇒ y = 0 at φ = −∆φ, π + ∆φ with ∆φ  1
1 1
⇒ sin(−∆φ) ≈ −∆φ = − , sin(π + ∆φ) ≈ −∆φ = −
 
1
It follows: 1 ⇒ 1

1 M M
Impact parameter = y(π/2) = 2 (1 + ) ≈ 2 
b L L
Angular momentum: mL = |~r × p~| = bmc = bm ⇒ L=b
2 2M
Total deflection 2∆φ = =
 b

r b
φ
π + ∆φ
−∆φ

2) GR

2 0 2 2 2 2 3 2
 d
Geodesic equation for Q = 0: L (y ) + L y − 2M L y = E

dφ


⇒ y 00 + y = 3M y 2

1
Without field: M = 0 ⇒ y000 + y0 = 0 ⇒ y0 = sin φ (like Newtonian)
b
Small deflection: perturb around straightline y0 = (sin φ)/b
M
⇒ y = y0 + ∆y + O(M 2 /b2 )
b
3 1 − cos 2φ
⇒ . . . ⇒ ∆y 00 + ∆y =
b 2
D THE SCHWARZSCHILD SOLUTION AND CLASSIC TESTS OF GR 50

Homogeneous DE: ∆y 00 + ∆y = 0 solved by ∆ỹ = A cos φ + B sin φ

1
Particular solution: ∆ŷ = (3 + cos 2φ)
2b
2
With A = , B = 0, we get y = 0 for φ → π
b
M 1 M 2M
⇒ y = y0 + ∆y = sin φ + 2 (3 + cos 2φ) + 2 cos φ
b b 2b b
Deflection δφ from y = 0 at φ = 0 + δφ
δφ M 2M
⇒0≈ + 2 (3 + 1) + 2
b 2b b
4M
⇒ δφ ≈ −
b
For sun: M = 1.5 km, b ≈ R = 7 × 105 km: |δφ| ≈ 1.7700

Measured by Eddington expedition in 1919.

Shapiro time delay

Radio signal past sun to Venus and back. Measure time delay.
p p 
2 2
1) Without field: Pythagoras ⇒ T = 2 r1 − b + r2 − b
2 2

Venus r1 b Earth
r2

Sun

2) With field
−1
2M L2
  
2 2M
Null geodesic (Q=0): ṙ + 1 − = E2 , ṫ = 1− E
r r2 r
 s   2
dr 2M 2M L
⇒ ... ⇒ =± 1− 1− 1−
dt r r r2 E 2
D THE SCHWARZSCHILD SOLUTION AND CLASSIC TESTS OF GR 51

At point of closest approach: r = b , dr/dt = 0

 2
L2 b2

2M L
⇒ 1− =1 ⇒ = eliminates L, E
b b2 E 2 E2 1 − 2Mb

r1 r2 s
b2 1 − 2M/r

dr dr 2M
Z Z
⇒T =2 +2 , f (r) = 1 − 1− 2
b f (r) b f (r) r r 1 − 2M/b

Solve integral by Taylor expanding f (r) in M/r using M/b  1

p p !
2 2
− −
q  2 2
r + r b r + r b
q
1 1 2 2
⇒ ... ⇒ T ≈ 2 r12 − b2 + r22 − b2 +4M ln + ln
b b
| {z }
=..TMink

r r !
r1 − b r2 − b
+2M + .
r1 + b r2 + b

For Venus and Earth: ∆T ≈ 77 km = 257 µs

D.4 The causal structure of the Schwarzschild spacetime

   −1
2 2M 2 2M
Recall Schwarzschild: ds = − 1 − dt + 1 − dr2 + r2 (dθ2 + sin2 θ dφ2 )
r r
singular at r = 0, r = 2M ; what’s happening?

Light cones

Timelike (null) curves travel inside (on) light cones

⇒ Lightcones display causal structure of spacetime!

t

E.g. Minkowski

in spherical coords.:

r
D THE SCHWARZSCHILD SOLUTION AND CLASSIC TESTS OF GR 52

Radial null geodesics: dθ = dφ = 0; Let λ be an affine parameter

 
2M
t component: We already know 1 − ṫ = E = const
r
d ∂L ∂L
r component: EL eqs.: =
dλ ∂ ṙ ∂r
2M r2
 
⇒ ... ⇒ 1− r̈ = ṙ2 − E 2
r M

Solved by ṙ = ±E

⇒ r = ±Eλ + r0 is also an affine parameter. Let’s use r.

dt ṫ r
⇒ = =±
dr ṙ r − 2M

⇒ . . . ⇒ t(r) = ±(r + 2M ln |r − 2M |) + k , k = const

t
10M

5M

0
0 2M 4M
r

r > 2M : In/Outgoing null geodesics for the +/− sign.

Note: r = const timelike (inside light cone)

 −1  
2 2M 2 2M
r < 2M : ds = − −1 dr + − 1 dt2
r r
⇒ grr < 0 , gtt > 0 ⇒ r is the timelike coordinate, t spacelike

Light cones tilted horizontally, but In/Outgoing direction unclear.

D THE SCHWARZSCHILD SOLUTION AND CLASSIC TESTS OF GR 53

Infalling observers

Consider timelike geodesic starting at large r:

     −1
2M 2M 2 2M
1− ṫ = E , − 1− ṫ + 1 − ṙ2 = Q = −1
r r r
2M
⇒ −E 2 + ṙ2 = −1 +
r
Observer starting at rest at r → ∞ ⇒ E = 1 (cf. Sec. D.2)
r
2M dτ r
Use proper time: ṙ2 = ⇒ =− < 0 infalling!
r dr 2M
2 3/2
r0 − r3/2


⇒ . . . ⇒ τ − τ0 = √
3 2M
How about coordinate time t?
r  −1
dt ṫ r 2M
= =− 1−
dr ṙ 2M r
√ √  √ √ 
2 h √ √ i r + 2M r0 − 2M
3/2
⇒ . . . ⇒ t−t0 = − √ r3/2 − r0 + 6M ( r − r0 ) +2M ln √ √  √ √ 
3 2M r0 + 2M r − 2M

70

τ(r)
60 t(r)

50

40

30

20

10

0
0 5 10 15 20
r/M

Interpretation: • t = proper time of observer at infinity

• geodesic crosses r = 2M at finite τ but t → ∞

• one falls in finite time, but process infinitely redshifted for outside observer
D THE SCHWARZSCHILD SOLUTION AND CLASSIC TESTS OF GR 54

Ingoing Eddington Finkelstein (IEF) coordinates

Ingoing null geodesics: t + 2M ln |r − 2M | = −r + const

2M
New time coordinate: t̄ = t + 2M ln |r − 2M | ⇒ dt̄ = dt + dr
r − 2M
   
2 2M 2 4M 2M
⇒ ... ⇒ ds = − 1 − dt̄ + dt̄dr + 1 + dr2 + r2 (dθ2 + sin2 θ dφ)
r r r

Ingoing geodesics: t̄ = −r + const

Outgoing geodesics: t̄ = r + 4M ln |r − 2M | + const

t
10M

5M

0
0 2M 10M
r

Light cones tilt over inwards at r = 2M .

Def.: Event horizon: The outermost boundary of a region of spacetime from which no null geodesics
or timelike curves can escape to infinity.

Israel’s theorem: If a spacetime is static, asymptotically flat and contains a regular horizon then
it is a Schwarzschild spacetime.

We can use a null coordinate v = t̄ + r ⇒ dt̄ = dv − dr

 
2 2M
⇒ . . . ⇒ ds = − 1 − dv 2 + 2dr dv + r2 dΩ2
r
Note: ∂r is tangent to curves v = const. Clearly g(∂r , ∂r ) = 0.
D THE SCHWARZSCHILD SOLUTION AND CLASSIC TESTS OF GR 55

Outgoing Eddington Finkelstein (OEF) coordinates

Outgoing null geodesics: t − 2M ln |r − 2M | = r + const

2M
New time coordinate: t̃ = t − 2M ln |r − 2M | ⇒ dt̃ = dt − dr
r − 2M
   
2 2M 2 4M 2M
⇒ ... ⇒ ds = − 1 − dt̃ − dt̃dr + 1 + dr2 + r2 (dθ2 + sin2 θ dφ) (∗)
r r r

We can use a null coordinate u = t̃ − r ⇒ dt̃ = du + dr

 
2 2M
⇒ . . . ⇒ ds = − 1 − du2 − 2dr du + r2 dΩ2
r

Ingoing geodesics: t̃ = −r − 4M ln |r − 2M | + const

Outgoing geodesics: t̃ = r + const

t
10M

5M

0
0 2M 10M
r

Now all light cones point outwards.

But above we showed that light cones point inwards at r < 2M . WTF?!?

On the other hand: Schwarzschild should be time symmetric. What’s going on?
D THE SCHWARZSCHILD SOLUTION AND CLASSIC TESTS OF GR 56

Kruskal-Szekeres coordinates
r − 2M
 
Step 1: Combine IEF and OEF: v = t + r + 2M ln ; r∗ : integration constant
r∗
r − 2M
 
u = t − r − 2M ln
r∗
 
2 2M
⇒ ds = − 1 − du dv + r2 dΩ2
r

Step 2: Use an exponential version of u, v:

v u
ṽ = e 4M , ũ = −e− 4M
v−u r − 2M r
⇒ ũṽ = −e 4M = − e 2M
r∗
16M 2 − r
⇒ . . . ⇒ ds2 = − e 2M dũ dṽ + r2 dΩ2
r/r∗

Step 3: Go back to time and radius:

1 1
t̂ = (ṽ + ũ) , r̂ = (ṽ − ũ)
2 2

16M 2 − r
⇒ ... ⇒ ds2 = e 2M (−dt̂2 + dr̂2 ) + r2 dΩ2 , from now on set r∗ = 1
r/r∗
r
r implicitly given through t̂2 − r̂2 = −e 2M (r − 2M )

Comments: • Metric regular at r = 2M

• Radial null geodesics are: t̂ = ±r̂ + const

• Coordinate range:

a) r = 2M ⇒ t̂2 − r̂2 = 0 ⇒ t̂ = ±r̂

√
b) r = 0 ⇒ t̂2 − r̂2 = 2M ⇒ t̂ = ± r̂2 + 2M

c) t̂2 − r̂2 = −er/(2M ) (r − 2M ) monotonically decreasing in r

⇒ t̂2 − r̂2 < 2M for all r > 0

d) No further restrictions on t̂, r̂: r̂ ∈ (−∞, ∞) , t̂2 ≤ r̂2 + 2M

D THE SCHWARZSCHILD SOLUTION AND CLASSIC TESTS OF GR 57

Kruskal diagram
r0
(1) Curves r = r0 : t̂2 − r̂2 = −e 2M (r0 − 2M ) =.. C
√
⇒ t̂ = ± r̂2 + C

1 √ r t
(2) Curves t = t0 : t̂ = (ṽ + ũ) = r − 2M e 4M sinh
2 4M
1 √ r t
r̂ = (ṽ − ũ) = r − 2M e 4M cosh
2 4M

t t̂
⇒ tanh =
4M r̂

t4

2 t=2M
r=0
r=1.5M
r=1.8M
t=M
r=2M r=2M
0 r=3M r=2.5M r=2.5M r=3M

r=1.8M
r=1.5M t=-M

-2 r=0 t=-2M

-4
-4 -2 0 2 4
r
Comments: • Spacetime extended: white hole, black hole, 2 asymptotically flat regions

• Resolves puzzle of light cones in IEF/OEF coordinates.

D THE SCHWARZSCHILD SOLUTION AND CLASSIC TESTS OF GR 58

D.5 Hawking radiation

GR is a classical theory and we do not yet have a quantum theory of gravity.

Semi-classical calculations: QM on curved background spacetime.

Hawking effect: Pair creation of virtual particles

→ one falls into black hole; the other escapes

~c3
Hawking radiation depends on temperature T =
8πGM kB
1 dM π 2 kB
4
− = σT 4 , σ=
A dt 60~3 c2
dM ~c6 πG2 3
⇒ =− ⇒ t = 5120 M ; for M : t = O(1060 ) yr
dt 15360π G2 M 2 ~c6
Note: Energy loss → higher T → more radiation!
E COSMOLOGY 59

E Cosmology
Goal: Simplified model of the entire universe.

E.1 Homogeneity and isotropy

Observations: • Electromagnetic observations of universe out to ∼ 1011 pc.

• Galaxies have R ∼ 105 km → point like on scale 1011 pc.

• On scales ∼ 109 pc, the universe looks the same on average and the same
in every direction.

• Hubble redshift → universe seems to expand.

⇒ Model universe as an isotropic, spatially homogeneous spacetime with fluid matter.

Cosmological principle: At a given time, the universe is spatially homogeneous and isotropic
when viewed on large scales.

Weyl’s postulate: The world lines of the universe’s fluid elements are orthogonal to hypersurfaces
of constant time, Σt , to which the cosmological principle applies.

Comments: • Isotropy is observer dependent: a boosted observer will not see isotropy.

• Which observer sees isotropy? The one comoving with the cosmological fluid.

• We do not require homogeneity in time!

Adapted coordinates

Let xi be spatial coordinates comoving with the cosmological fluid.

Let t be proper time along the world lines of observers comoving with the fluid.

Isotropy ⇒ no spatial metric component has a prefered time dependency

⇒ ds2 = −dt2 + g0i dt dxi + a(t)2 hij (xk )dxi dxj .

Consider basis e0 = ∂t , ei = ∂i

⇒ e0 = four-velocity of comoving observers and ei are tangent to Σt

By Weyl’s postulate, e0 is orthogonal to Σt , i.e. to the ei

⇒ g0i = g(e0 , ei ) = 0
E COSMOLOGY 60

⇒ ds2 = −dt2 + a(t)2 hij (xk )dxi dxj (†)

Note: Observers moving relative to the fluid do not move orthogonally to Σt .

⇒ g̃0i 6= 0 ⇒ this observer does not see isotropy.

Recall: Spatial part of spherically symmetric metric: d`2 = C(t, r)dr2 + D(t, r)dΩ2 (‡)

Spherical symmetry = isotropy around one point

Isotropy around every point = “Maximal symmetry” = Spherical symmetry + more

⇒ we can combine (†) and (‡)

 
⇒ factorize C(t, r) = a(t)2 e2β(r) , D(t, r) = a2 (t)r2 ⇒ d`2 = a2 (t) e2β(r) dr2 + r2 dΩ2

Next use our differential geometry on the hypersurface t = const

3-dimensional Ricci scalar R = Ri i , i = 1, 2, 3

2  −2β

Spatial homogeneity ⇒ R = . . . = 1 − ∂ r (re ) = const =.. k̃
r2
1
⇒ e2β =
1 − 61 k̃r2 − A
r

No conical singularity ⇒ lim d`2 ∝ (dr2 + r2 dΩ2 ) ⇒ A = 0

r→0

dr2
 
2 2 2
⇒ Robertson-Walker metric: ds = −dt + a(t) + r2 (dθ2 + sin2 θ dφ2 )
1 − kr2

Note: We can always rescale r, a so that k = +1, 0 or −1

1) k = 0 : d`2 = dr2 + r2 dΩ2 = dx2 + dy 2 + dz 2

Flat metric on R3 ; k = 0 models are called flat.

2) k = +1: r = sin χ ⇒ d`2 = dχ2 + sin2 χdΩ2

Metric on a 3-sphere (w2 + x2 + y 2 + z 2 = r2 in R4 ); closed models

3) k = −1: r = sinh ψ ⇒ d`2 = dψ 2 + sinh2 ψdΩ2

Metric of a “saddle”; open models

E COSMOLOGY 61

E.2 The Friedmann equations

The matter fields

Recall: Perfect fluids are isotropic by definition!

We therefore set Tµν = (ρ + P )uµ uν + P gµν

Comoving frame: uµ = (1, 0, 0, 0) ⇒ T µ ν = diag(−ρ, P, P, P )

⇒ T = T µ µ = −ρ + 3P coordinate invariant!

ȧ
Conservation: ∇µ T µ ν = 0 ⇒ . . . ⇒ ρ̇ = −3 (ρ + P )
a
Equation of state: Cosmological matter typically has P = wρ , w = const
ρ̇ ȧ
⇒ = −3(1 + w) ⇒ ρ ∝ a−3(1+w)
ρ a
(1) Dust

w = 0 ⇒ P = 0 ⇒ ρ ∝ a−3

Pressure between galaxies ≈ 0.

(2) Radiation

Statistical Physics: photons are a gas with P = ρ/3

1
⇒ w= ⇒ ρ ∝ a−4
3
Fourth power of a comes from redshift (cf. below).

(3) Dark energy

Recall Lovelock’s theorem: We can add Λgαβ to the Einstein equations.

! Λ
Perfect fluid with w = −1: 8πTµν = 8πP gµν = −Λgµν ⇒ −P =ρ=
8π
w = −1 ⇒ ρ ∝ a0

Interpretation: Non-zero of the vacuum. Density = const, independent of volume

Do not confuse with dark matter!

E COSMOLOGY 62

ȧ
Def.: H := is the Hubble parameter.
a
aä
q := − is the deceleration parameter.
ȧ2
3H 2
ρcrit := is the critical density; its significance will be revealed below.
8π
ρ 8π
Ω= = ρ is the density parameter.
ρcrit 3H 2

Note: These are time dependent variables. The “parameters” are their present day values.

The Einstein equations

Plug Robertson-Walker metric into Gαβ + Λgαβ = 8πTαβ

ȧ2 + k 2aä + ȧ2 + k

⇒ ... ⇒ 3 − Λ = 8πρ (I) , − Λ = −8πP (II)
a2 a2

ä 4π Λ
= − (ρ + 3P ) + (III)
a 3 3

Note: Eq. (III) follows from (I), (II) but can be useful.

Differentiate (I) and combined with (I), (II)

ȧ 
⇒ . . . ⇒ ρ̇ + 3 (ρ + P ) = 0  · a3

a
d 3 d
⇒ (a ρ) + P a3 = 0
dt dt
Volume V ∝ a3 , energy E = V ρ, so this can be written as dE + P dV = 0

E.3 Cosmological redshift

a2 dt dr
Radial null curves: ds2 = −dt2 + dr2 = 0 ⇒ = ±√
1 − kr2 a(t) 1 − kr2
A galaxy at r = R emits at times te and te + ∆te .

These reach an observer at r = 0 at times to , to + ∆to .

E COSMOLOGY 63

t
to 0 to +∆to
dt dr dt
Z Z Z
⇒ =− √ =
te a R 1 − kr2 te +∆te a

(‘−’ since ingoing geodesic)

to +∆to te +∆te
dt dt
Z Z
⇒ =
to a te a

r=0 r=R

E.g. crests of light wave: ∆te , ∆to  to − te ⇒ a ≈ const

∆to ∆te λo a(to )

⇒ = ⇒ =
a(to ) a(te ) λe a(te )

Nearby galaxies: Taylor expand a(te ) ≈ a(to ) − (to − te )ȧ(to )

a(to ) ȧ(to )
⇒ ≈ 1 + (to − te ) = 1 + (to − te )H(to )
a(te ) a(to )
Hubble’s law with distance = c(to − te ) , c = 1.

Luminosity distance: Surface of constant radius, time: ds2 = a(t)2 r2 (dθ2 + sin2 θ dφ2 )

Area of sphere of constant r: 4πa2 r2

Intensity of light collected at r = 0 from source at r = R:

energy E
I ..= =
area 4πa R (1 + z)2
2 2

Factors (1 + z) from (i) redshift, (ii) reduced rate of photon hits.

E
Luminosity distance DL2 ..=
4πI(1 + z)2
E COSMOLOGY 64

E.4 Cosmological models

General considerations
k Λ
(1) Eq. (I) ⇒ = Ω − 1 +
ȧ2 3H 2
For Λ = 0: ρ > ρcrit ⇔ Ω>1 ⇔ k = +1 “closed”

ρ = ρcrit ⇔ Ω=1 ⇔ k=0 “flat’

ρ < ρcrit ⇔ Ω<1 ⇔ k = −1 “open”

(2) Consider Λ = 0, ρ > 0, P ≥ 0. Then: Eq. (III) ⇒ ä < 0.

Observations: H0 ≈ 71 km/(s Mpc) ⇒ ȧ > 0

a(t)
If ä = 0, then a = 0 at t = −13.8 Gyr

“Big Bang”!

ä < 0 ⇒ Universe is less old

−13.8 Gyr t> −13.8 Gyr t

0
8π 2
(3) Again Λ = 0, ρ > 0, P ≥ 0. Then Eq. (I) ⇒ ȧ2 = a ρ − k.
3
For k = 0, − 1: ȧ2 > 0 always ⇒ ȧ > 0 always (since ȧ > 0 today)
d 3
d
Recall: a ρ = −P a3 = −3a2 P ȧ ≤ 0
dt dt
But: ρa3 ≥ 0, so lim a2 ρ = 0
t→∞

8π 2
Then: Eq. (I) ⇒ ȧ2 = a2 H 2 = a ρ−k → k ⇒ lim ȧ = |k|
3 t→∞

Expansion never stops.

8π 2
For k = +1: Eq. (I) ⇒ ȧ2 = a ρ − 1. As before lim a2 ρ = 0
3 a→∞
p
⇒ ȧ = 0 at a = amax = 3/(8πρ).
4π
Eq. (III) ⇒ lim ä = − (ρ + P )amax < 0
a→amax 3
⇒ contraction back to a = 0: “Big Crunch”
E COSMOLOGY 65

a open
k=−1
flat
k=0

closed
k=1

now t

Let’s calculate solutions of the Friedmann Eqs.

1) Flat, matter dominated: k = 0, P = 0

Recall: a3 ρ = const for P = 0.

 
3 2 1 3 ! 8π 3
Eq. (I) ⇒ 8πa ρ = 3 aȧ + ka − Λa = const =.. 3C ⇒ C= aρ .
3 3

C 1 2
⇒ ȧ2 = + Λa − k (†)
a 3

Λ > 0: Set k = 0 in (†) and use variable

2Λ 3 2Λ 2
u= a ⇒ u̇ = a ȧ
3C C
4Λ2 4Λ2 3 4Λ3 6
 
C Λ 2
2
⇒ u̇ = 2 a4 + a = a + a = 6Λu + 3Λu2
C a 3 C 3C 2

⇒ u̇2 = 3Λ(2u + u2 )
√
⇒ u̇ = 3Λ(2u + u2 )1/2
Z u
1 √
Initial conditions: a = u = 0 at t = 0 ⇒ √ dũ = 3Λ t
0 2ũ + ũ2
E COSMOLOGY 66

Use u = −1 + cosh w
Z u Z u Z w Z w
dũ dũ sinh w̃ dw̃
⇒ √ = p = p = dw̃ = w
0 ũ2 + 2ũ 0 (ũ + 1)2 − 1 0 cosh2 w̃ − 1 0
√
⇒ u + 1 = cosh w = cosh( 3Λt)

2Λ 3 √ 3C h √ i
⇒ a = cosh( 3Λ t) − 1 ⇒ a3 = cosh( 3Λ t) − 1
3C 2Λ

2Λ 3
Λ < 0: Use u = − a
3C

3 3C h √ i
⇒ ... ⇒ a = 1 − cos( −3Λ t)
2(−Λ)

Z √
C √
Z
2
Λ = 0: Set k = 0 in (†) ⇒ ȧ = ⇒ ada = Cdt
a

2 √ 9C 2
⇒ a3/2 = Ct ⇒ a3 = t
3 4
 −1  
ȧ 2 aä ȧ ä 1
⇒H= = , q=− 2 =− =
a 3t ȧ a ȧ 2
“Einstein-de Sitter model”

P = 0, k = 0
a
Λ>0
Λ=0
Λ<0

Einstein-de Sitter

t
E COSMOLOGY 67

2) Matter dominated, no cosmological constant: Λ = 0, P = 0

C
Eq. (†) on page 65 ⇒ ȧ2 = −k
a
 r r r 
a a a
⇒ . . . ⇒ For k = +1: C arcsin − 1− = ±t + b± , b± = const
C C C

r r r 
a a a
For k = −1: C 1 + − arsinh = ±t + b±
C C C

For k = 0: Einstein-de Sitter model from above.

WLoG b± = 0. Also t ≥ 0 for future oriented models.

P = 0, Λ = 0
a
k = -1
k=0
k = +1

Einstein-de Sitter

3) The static Einstein Universe: ȧ = ä = 0, P = 0

3k k
Eqs. (I), (II) ⇒ = Λ + 8πρ , =Λ ⇒ k = 4πa2 ρ
a2 a2
ρ>0 ⇒ k = +1 necessarily for “sensible” matter.

Eq. (†) on page 65 ⇒ 3a = 3C + Λa3 = 3C + a

3C 4
⇒ a= ∧ Λ=
2 9C 2
E COSMOLOGY 68

Note: Conceptually nice model: homogeneous in time as well!

Problem: unstable to perturbations! Say, a = a0 +  ,   a0

C 4 3
Eq. (III) ⇒ a2 ä = − + a
2 27C 2
C 4 4 9 2
⇒ a20 ¨ ≈ − + 2
(a30 +3a20  + . . .) = C =
| 2 {z
27C } 9C 2 4
=0

Has solutions exp(±αt), i.e. also exponentially growing modes.

4) de Sitter Universe: ρ = P = 0 , Λ > 0

ȧ2 + k
Eq. (I) ⇒ 3 =Λ
a2
r r !
3 Λ
k = −1 ⇒ a(t) = sinh t
Λ 3
√
k=0 ⇒ a(t) ∝ e± Λ/3 t
r r !
3 Λ
k = +1 ⇒ a(t) = cosh t
Λ 3

All three turn out to be the same spacetime, just in different coordinates.

de Sitter: ρ = 0, Λ > 0
a
k=-1

t
E COSMOLOGY 69

5) Λ = 0, radiation dominated: P = ρ/3

d 3 d
From Sec. E.2: in general (a ρ) + P a3 = 0
dt dt
d 3 da 1d 4
P = ρ/3 ⇒ 0= (a ρ) + ρa2 = ... = (a ρ)
dt dt a dt
8π 4
⇒ B ..= a ρ = const
3
√ 1/4 √
We find k = 0 ⇒ a= 2B t
s
√
2

t
k = +1 ⇒ a= B 1− 1− √
B
s
√
2
t
k = −1 ⇒ a= B 1+ √ −1
B

P = ρ/3, Λ = 0
a
k = -1
k=0
k = +1

Summary: • Radiation dominated: ρa4 = const

Matter dominated: ρa3 = const
Dark energy: ρ ∝ Λ = const

• Radiation dominant in early universe, dark energy in late

expanding universe.

• Presently: ∼ 75 % dark energy, 4 % visible matter, ∼ 21 %

dark matter, radiation negligible
F SINGULARITIES AND GEODESIC INCOMPLETENESS 70

F Singularities and geodesic incompleteness

Metric components are sometimes 0 or ∞. What happens there?

F.1 Coordinate vs. physical singularities

   −1
22M 2 2M
Schwarzschild: ds = − 1 − dt + 1 − dr2 + r2 dΩ2
r r
r = 2M : cured with Kruskal coordinates

Questions: 1) Can we predict when better coordinates exist?

2) Is there a “super-Kruskal” that cures r = 0?

Note: Scalars are invariant under coordinate transformations!

Def. : “Kretschmann scalar” κ ..= Rµνρσ Rµνρσ

48 M 2
In Schwarzschild: κ = . . . = ⇒ r = 0 is a genuine, physical singularity.
r6
80
Einstein-de Sitter Universe: κ = . . . = ⇒ t = 0 is a physical singularity.
27t4

F.2 Geodesic incompleteness

1
Consider geodesics in “Kasner V” spacetime: ds2 = − dt2 + z 2 (dx2 + dy 2 ) + zdz 2 , z>0
z
ṫ d
Noether c0 = , c1 = z 2 ẋ , c2 = z 2 ẏ , ˙= , λ = affine
z dλ
1
gµν ẋµ ẋν = − ṫ2 + z 2 (ẋ2 + ẏ 2 ) + z ż 2 = 
z
ci = const,  = +1, −1, 0 for space, timelike, null geodesic (λ = τ or s)

c21 + c22 
Geodesic Eq. ⇒ ... ⇒ ż 2 + 3
− = c20
z z
Consider null geodesics with ẋ = ẏ = 0 , ż < 0 , z = z0 , t = 0 at λ = 0

⇒ x = y = const , z = −c0 λ + z0

Reaches z = 0 at finite λ.
12
κ= ⇒ singularity is physical.
z6
F SINGULARITIES AND GEODESIC INCOMPLETENESS 71

Def.: A geodesic is incomplete if it cannot be extended to arbitrarily large values of its parameter,
either to the future or the past. The termination point is a singularity.

Can this happen too at coordinate singularities? Yes! Example:

Rindler spacetime: ds2 = −z 2 dt2 + dx2 + dy 2 + dz 2 , t, x, y, z, ∈ R , z > 0

Noether: c0 = z 2 ṫ , c1 = ẋ , c2 = ẏ all const.

−z 2 ṫ2 + ẋ2 + ẏ 2 + ż 2 =  = +1, −1 or 0 as in Kasner V

Consider timelike geodesics with ẋ = ẏ = 0 , ż < 0 , z = z0 , t = 0 at τ = 0

p τ
⇒ . . . ⇒ z(τ ) = z02 − τ 2 , t(τ ) = artanh
z0
geodesic hits singular z = 0 at finite τ
√ T
Now transform coordinates: x = X , y=Y , z= Z2 − T 2 , t = artanh
Z
⇒ ... ⇒ ds2 = −dT 2 + dX 2 + dY 2 + dZ 2

Rindler spacetime is just a wedge of Minkowski spacetime!

Here the singular point z = 0 can be cured with better coordinates.

Is there a systematic way to find such better coordinates?

In 4 dimensions no, but in 2 dimensions (e.g. spherical symm.) there is: Sec. 6.4 in Wald.
G LINEARIZED THEORY AND GRAVITATIONAL WAVES 72

G Linearized theory and gravitational waves

G.1 Plane waves and pp metrics
~ B
Plane waves are common in physics. E.g. electromagnetic waves E, ~ ∝ ei(~k~x−ωt)

Rotate coordinates such that ~k = (0, 0, k) ⇒ ~ B

E, ~ ∝ ei(kz−ωt) = eik(z−vt)
ω
phase velocity: v =
k
~ 2f = 0
The components satisfy the wave equation 2f = −∂t2 f + ∇
ω
ω 2 − ~k 2 = 0 k = |~k| = ±ω
~
With f ∝ ei(k~x−ωt) ⇒ ⇒ ⇒ v= = ±1 = ±c
k
with kα = (−ω, ~k)
α
SR notation: f ∝ eikα x

2f = η αβ ∂α ∂β f = 0 ⇒ kα k α = 0

plane wave in +z direction: kα = (−ω, 0, 0, k)

Plane waves also exist in full GR!

Def.: Spacetimes admitting a covariantly constant vector field V are called pp wave spacetimes.
They admit plane wave solutions.

Example: Brinkmann metrics: ds2 = H(u, x, y)du2 + 2du dv + dx2 + dy 2

V ..= ∂v satisfies V β = ∂α V β + Γβµα V µ = 0 + Γβµα δ µ v = Γβvα = . . . = 0

Furthermore: Ruu = 0 ⇒ ... ⇒ ∂x2 H + ∂y2 H = 0

α
Clearly solved by H = H0 eikα x , H0 = const , kα = (−ω, 0, 0, ω)

pp metrics are important in the construction of analytic solutions in GR.

E.g. Aichelburg-Sexl metric: Schwarzschild BH boosted to the speed of light.

G.2 Linearized theory

Consider a system close to a known solution: Metric = background + perturbation

For us background is Minkowski, but idea works for general backgrounds: perturbation theory

⇒ gµν = ηµν + hµν , ηµν = diag(−1, 1, 1, 1) , hµν = O()  1

We regard hµν as a tensor field on Minkowski background

G LINEARIZED THEORY AND GRAVITATIONAL WAVES 73

inverse metric: g µν = η µν + k µν
!
⇒ g µν gνρ = δ µ ρ + k µν ηνρ + η µν hνρ + k µν hνρ = δ µ ρ
| {z }
=O(2 )→0

⇒ k µν = −η µρ η νσ hρσ =: −hµν = O()

1
To O(): Γµνρ = η µσ (∂ρ hσν + ∂ν hρσ − ∂σ hνρ ) ,
2


τ τ
Rµνρσ = ηµτ ∂ρ Γνσ − ∂σ Γνρ  Γ · Γ = O(2 )


1


= ∂ρ ∂ν hµσ + ∂σ ∂µ hνρ − ∂ρ ∂µ hνσ − ∂σ ∂ν hµρ
2

ρ 1 ρ 1 
Rµν = ∂ ∂(µ hν)ρ − ∂ ∂ρ hµν − ∂µ ∂ν h  h := hµ µ , ∂ µ := g µρ ∂ρ

2 2
1 1 1 !
Gµν = ∂ ρ ∂(µ hν)ρ − ∂ ρ ∂ρ hµν − ∂µ ∂ν h − ηµν (∂ ρ ∂ σ hρσ − ∂ ρ ∂ρ h) = 8πTµν
2 2 2
⇒ Tµν  1

1 1
Def.: “trace-reversed perturbation” h̄µν ..= hµν − hηµν ⇔ hµν = h̄µν − h̄ηµν ,
2 2
h̄ = h̄µ µ = −h

1 1
⇒ . . . ⇒ Gµν = − ∂ ρ ∂ρ h̄µν + ∂ ρ ∂(µ h̄ν)ρ − ηµν ∂ ρ ∂ σ h̄ρσ = 8πTµν
2 2
Coordinate freedom
∂ x̃α ∂xν
x̃α = xα − ξ α , ξ α = O() ⇒ = δ α
µ − ∂µ ξ α
, = δ ν β + ∂˜β ξ ν
∂xµ ∂ x̃β

⇒ g̃µν = ηµν + h̃µν , h̃µν = hµν + ∂µ ξν + ∂ν ξµ + O(2 )

Note: • Background unchanged

• 4 free functions: We choose ξµ such that ∂ ν ∂ν ξµ = −∂ ν h̄µν

¯ = 0 “Lorentz gauge”
⇒ . . . ⇒ ∂ ν h̃µν

⇒ − 2G̃µν = ¯ = ∂ ρ ∂ h̃
¯
2h̃µµ ρ µν = −16πTµν

Wave equation! drop ˜ from now on.

G LINEARIZED THEORY AND GRAVITATIONAL WAVES 74

G.3 The Newtonian limit

2
~ 2 Φ = 4πρ ; Φ ∼ v 2  1 (Sec. A.1) ,  ..= v = v 2
Newtonian gravity: ∇
c2
⇒ matter sources weak: ρ ∼ O()

Newtonian matter: T00 = ρ + O(2 )

T0i ∼ T00 vi ∼ O(3/2 )

Tij ∼ T00 vi vj ∼ O(2 )

E.g. perfect fluid: Tµν = (ρ + P )uµ uν + P gµν , P ∼ ρv 2 ≈ 10−5 ρ in sun

In Newt. gravity temporal changes in Φ are caused by motion of sources

∂ ∂ ∂
⇒ ∼ v i = O(1/2 ) i
∂t ∂x ∂x
~ 2 h̄µν = −16πTµν
⇒ 2h̄µν = ∂ ρ ∂ρ h̄µν = ∂ i ∂i h̄µν = ∇
~ 2 h̄00 = −16πT00 = −16πρ + O(2 ) ,
⇒∇ h̄0i = O(3/2 ) , h̄ij = O(2 )

Newton’s law with h̄00 = −4Φ

⇒ h̄ = η µν h̄µν = 4Φ + O(2 ) = −h
1 1
⇒ h00 = h̄00 − η00 h̄ = −2Φ , hij = h̄ij − ηij h̄ = −2Φδij
2 2

or ds2 = −(1 + 2Φ)dt2 + (1 − 2Φ)(dx2 + dy 2 + dz 2 ) cf. Sec. A.2

Geodesics
!
L = (1 + 2Φ)ṫ2 − δij (1 − 2Φ)ẋi ẋj = 1 ṫ2 = (1+ 2Φ)−1 1 + δij ẋi ẋj + O(2 )
 
⇒
1
⇒ ṫ = 1 − Φ + δij ẋi ẋj + O(2 )
2
d L d j ∂L
= 2∂k Φ (ṫ2 + δij ẋi ẋj )

EL for xk : k
= − 2δ jk (1 − 2Φ)ẋ =
dτ ∂ ẋ dτ ∂xk | {z }
=1+O()

d 2 xk d 2 xk
⇒ ẍk = = = −∂k Φ to O() ; test body in Newt. gravity
dτ 2 dt2
G LINEARIZED THEORY AND GRAVITATIONAL WAVES 75

G.4 Gravitational waves

weak field but now: vacuum; no longer “∂t  ∂x ”
~ 2 )h̄µν = 0
⇒ 2h̄µν = (∂t2 − ∇
ρ
Plane wave solution: h̄µν = Hµν eikρ x ; Hµν = const

(i) 2h̄µν = 0 ⇒ kµ k µ = 0 → speed of light

(ii) Lorentz gauge: ∂ ν h̄µν = 0 ⇒ k µ Hµν = 0 “transverse”

E.g. wave in z-dir.: k µ = ω (1, 0, 0, 1) ⇒ Hµ0 + Hµ3 = 0

ρ
Remaining gauge freedom: take ξµ = Xµ eikρ x ⇒ ∂ ν ∂ν ξµ = 0

⇒ . . . ⇒ Hµν → Hµν + i(kµ Xν + kν Xµ − ηµν k ρ Xρ )

⇒ . . . ⇒ ∃Xµ : H0µ = 0 , H µ µ = 0 “traceless”

In this gauge: 1) h = 0 ⇒ hµν = h̄µν

2) plane wave in z-dir.: H0µ = H3µ = H µ µ = 0

 
0 0 0 0
 0 H+ H× 0 
⇒ Hµν =   0 H× −H+ 0 


0 0 0 0

Effect on particles

Consider particle at rest in background Lorentz frame: uα0 = (1, 0, 0, 0)

d α
geodesic eq.: u + Γαµν uµ uν = u̇α + Γα00 = 0
dτ
1
Γα00 = η αβ (∂0 hβ0 + ∂0 h0β − ∂β h00 ) = 0 since H0µ = 0
2
⇒ uα = (1, 0, 0, 0) always

⇒ particle stays at xi = const in this gauge

G LINEARIZED THEORY AND GRAVITATIONAL WAVES 76

Proper separation: ds2 = −dt2 + (1 + h+ )dx2 + (1 − h+ )dy 2 + 2h× dx dy + dz 2

Case 1: H× = 0 , H+ 6= 0 ⇒ h+ oscillates

2 particles at (−δ, 0, 0), (δ, 0, 0) ⇒ ds2 = (1 + h+ ) 4δ 2

2 particles at (0, −δ, 0), (0, δ, 0) ⇒ ds2 = (1 − h+ ) 4δ 2

Case 2: H+ = 0 , H× 6= 0
√ √
2 particles at (−δ, −δ, 0) / 2 , (δ, δ, 0) / 2 ⇒ ds2 = (1 + h× ) 4δ 2
√ √
2 particles at (δ, −δ, 0) / 2 , (−δ, δ, 0) / 2 ⇒ ds2 = (1 − h× ) 4δ 2

G.5 The quadrupole formula

Consider energy density ρ(t, ~y ) with compact support
R
Def.: Quadrupole tensor Iij ..= ρ(t, ~y )y i y j d3 y
1
Reduced quadrupole tensor Qij ..= Iij − Ikk δij , sum over repeated indices!
3

G ... ...
Averaged energy flux in GWs: hpit = hQ Q it−r sum over i, j
5c5 ij ij

Emitted at t − r, observed at t, h.i = time average.

Example: Two equal point masses on Newtonian circular orbit

ρ(~x) = mδ(~x − ~x1 ) + mδ(~x − ~x2 ) ,

xi1 = r (cos φ, sin φ, 0) , xi2 = −r (cos φ, sin φ, 0)

G LINEARIZED THEORY AND GRAVITATIONAL WAVES 77

r
m
Newtonian orbit: ω = φ̇ = G 3
4r
⇒ Ixx = 2mr2 cos2 ωt = mr2 (1 + cos 2ωt)

Iyy = 2mr2 sin2 ωt = 2mr2 (1 − cos2 ωt) = mr2 (1 − cos 2ωt)

Ixy = Iyx = 2mr2 cos ωt sin ωt = mr2 sin 2ωt

2
Qij = Iij − mr2 = Iij − const
3
...
⇒ Qxx = 8ω 3 mr2 sin 2ωt
...
Qyy = −8ω 3 mr2 sin 2ωt
... ...
Qxy = Qyx = −8ω 3 mr2 cos 2ωt

2 G4 m5
Adding all up: hpit =
5 c5 r 5
Gravitational waves directly detected on 14 Sep 2015: New window to the universe!
REFERENCES 78

References
[1] S. M. Carroll. Lecture notes on general relativity, 1997. gr-qc/9712019.
[2] S. M. Carroll. Spacetime and Geometry: An Introduction to General Relativity. Pearson,
2003.
[3] R. d’Inverno. Introducing Einstein’s Relativity. Oxford: Clarendon Press, 1992. ISBN-
9780198596868.
[4] J. B. Hartle. Gravity: An Introduction to Einstein’s General Relativity. Pearson, 2014.
[5] L. Ryder. Introduction to General Relativity. Cambridge University Press, 2009.
[6] B. F. Schutz. A First Course in General Relativity. Cambridge University Press, 2009. 2nd
Edition.