Derivation of Black Hole Solutions

1
Contents

1 The Schwarzschild Metric 6

2 Schwarzschild Black Hole 17

2.1 Eddington-Finkelstein Coordinates . . . . . . . . . . . . . . . . . . . . . . . . 17

2.1.1 Advanced time parameter . . . . . . . . . . . . . . . . . . . . . . . . 18

3 Internal Schwarzschild Solution 20

4 Penrose-Carter diagrams 24

4.1 Geodesics under conformal transformations . . . . . . . . . . . . . . . . . . . 24

4.2 Penrose-Carter Diagram for Minkowski Spacetime. . . . . . . . . . . . . . . . 25

4.3 Maximal Extensions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

4.4 The Kruskal Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

5 Charged Balck Holes 29

5.0.1 Solving the Maxwell equations and the electric charge . . . . . . . . . 31

5.0.2 Calculation of the energy-momentum tensor . . . . . . . . . . . . . . . 36

5.0.3 Solving the field equations . . . . . . . . . . . . . . . . . . . . . . . . 38

5.1 Event Horizons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

5.2 Analogue of Eddington-Finkelstein Coordinates . . . . . . . . . . . . . . . . . 40

5.3 Penrose Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

5.4 Double null coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

2
 5.5 Maximal extension . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

6 Rotating Black Holes 46

6.1 Field Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

6.1.1 Degenerate metric . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

6.1.2 Order m4 equations: . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

6.1.3 Order m3 equations: . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

6.1.4 Equation from linear in m equation . . . . . . . . . . . . . . . . . . . 52

6.1.5 Order m2 equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

6.1.6 Stationary field equations from the m order equation . . . . . . . . . . 57

6.1.7 Laplace and eikonal equations . . . . . . . . . . . . . . . . . . . . . . 67

6.1.8 Introduce the complex function γ . . . . . . . . . . . . . . . . . . . . 70

6.1.9 Expression for λ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72

6.1.10 Expression for l0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74

6.2 The Kerr Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75

6.2.1 A solution to the field equations . . . . . . . . . . . . . . . . . . . . . 75

6.2.2 The Schwarzschild solution . . . . . . . . . . . . . . . . . . . . . . . 76

6.2.3 The Kerr solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77

6.3 Independence of Metric on Angular Variable ϕ . . . . . . . . . . . . . . . . . 81

6.4 Boyer-Lindquist Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . 84

6.5 Interptetation as Rotating Body . . . . . . . . . . . . . . . . . . . . . . . . . . 86

6.6 Basic Propertire of the Kerr Solution . . . . . . . . . . . . . . . . . . . . . . . 87

6.6.1 Boyer-Lindquist cordinates . . . . . . . . . . . . . . . . . . . . . . . . 87

6.6.2 Cartesian Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . 88

6.7 Singularities and Event Horizons . . . . . . . . . . . . . . . . . . . . . . . . . 88

3
A Derivation of the Full Field Equations of General Relativity 90

A.1 Mathematical Preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91

A.1.1 The Metric Connection . . . . . . . . . . . . . . . . . . . . . . . . . . 91

A.1.2 Tensor Densities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92

A.1.3 Matrix Identities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92

A.1.4 Metric Identities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93

A.1.5 Palantini Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95

A.2 The Einstein Lagrangian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96

A.3 The Full Field Equations - Inclusion of Matter and Radiation . . . . . . . . . . 98

A.4 The Energy-Momentum Tensor for Dust, a Perfect Fluid . . . . . . . . . . . . 99

A.4.1 Source term: From Newton’s Theory to Special Relativity . . . . . . . 99

A.4.2 Dust . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100

A.4.3 Dust Lagrangian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103

A.4.4 Perfect Fluid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104

A.5 The Energy-Momentum Tensor for the Electro-Magnetic Field . . . . . . . . . 106

A.5.1 Reminder of Maxwell’s Equations . . . . . . . . . . . . . . . . . . . . 106

A.5.2 Potential Formulation of Maxwell’s Equations . . . . . . . . . . . . . 108

A.5.3 Action for Maxwell Equations and Minimal Coupling to Charged Matter 110

A.5.4 The Maxwell Energy-momentum Tensor . . . . . . . . . . . . . . . . 112

A.6 Einstein-Maxwell Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . 113

4
List of Figures

4.1 Penrose diagram of Minkowski spacetime . . . . . . . . . . . . . . . . . . . . 27

4.2 Penrose diagram of the Kruskal solution . . . . . . . . . . . . . . . . . . . . . 28

4.3 Penrose diagram of a black hole. . . . . . . . . . . . . . . . . . . . . . . . . . 28

5.1 Penrose diagram of a black hole. . . . . . . . . . . . . . . . . . . . . . . . . . 45

6.1 Rotating blackhole . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88

5
Chapter 1

The Schwarzschild Metric

The most general spherically symmetric metric is

dτ2 = A(t, r)dt2 − B(t, r)dr2 − 2C(t, r)dtdr − D(t, r)(dθ2 + sin2 θdφ2 ) (1.1)

We introduce a new radial coordinate

r′ = D1/2 (t, r)

The line element then becomes

′ ′
dτ2 = A′ (t, r′)dt2 − B′(t, r′ )dr 2 − 2C ′ (t, r′ )dtdr − r 2 (dθ2 + sin2 θdφ2 ) (1.2)

Consider the differential

A′ (t, r′)dt − B′ (t, r′)dr′

The theory of ordinary differential equations tells us that there exists a function (so called inte-
gration factor) f (t, r′) such that

∂t′ ∂t′
= A′ (t, r′) f (t, r′), = −B′(t, r′ ) f (t, r′)
∂t ∂r′
for some t′ = t′(t, r′ ), making t′ a perfect differential:

∂t′ ∂t′
dt′ = dt + ′ dr′
∂t ∂r
= A′ (t, r′)dt − B′ (t, r′)dr′ f (t, r′)


(1.3)

6
We need to know something of integrating factors and perfect differentials.

Proof:

Consider the ordinary differential equation of first order in two variables:

dy Y(x, y)
( − ) = g(x, y). (1.4)
dx X(x, y)

Observe that

!
d  − R x Y(x′ ,y)/X(x′ ,y)dx′  dy − R x Y(x′ ,y)/X(x′ ,y)dx′ Y(x, y) − R x Y(x′ )/X(x′ )dx′
ye = e − e
dx dx X(x, y)

Thus we multiply a given differential equation by the factor

R x
Y(x′ ,y)/X(x′ ,y)dx′
e−

which turns the LHS of the equation into a perfect differential

d  − R x Y(x′ ,y)/X(x′ ,y)dx′  Rx

′ ′ ′
ye = g(x, y)e− Y(x ,y)/X(x ,y)dx
dx
or

 Rx
′ ′ ′
 Y(x, y) Rx
′ ′ ′
d y(x, y)e− Y(x ,y)/X(x ,y)dx = (dy − dx)e− Y(x )/X(x )dx
X(x, y)
Rx
Y(x′ ,y)/X(x′ ,y)dx′
= (X(x, y)dy − Y(x, y)dx) e− /X(x, y)

(1.5)

Put

Rx
Y(x′ ,y)/X(x′ ,y)dx′
y′ (x, y) = y(x, y)e−

and

R x
Y(x′ ,y)/X(x′ ,y)dx′
f (x, y) = e− /X(x, y)

then we have

7
 dy′ = (X(x, y)dy − Y(x, y)dx) f (x, y)

We use this result to define a new time coordinate t′ for which

dt′ = A′ (t, r′)dt − B′(t, r′ )dr′ f (t, r′)

(1.6)

Squaring, we obtain

′ ′ ′ ′

dt 2 = A 2 dt2 − 2A′ B′ dtdr′ + B 2 dr 2 f 2 (t, r′)

or

′ ′ ′ ′ ′
A′ dt2 − 2B′ dtdr′ = A −1 f −2dt 2 − A −1 B 2 dr 2

the line element then becomes

Dropping the primes, the line element then becomes

dτ2 = A(t, r)dt2 − B(t, r)dr2 − r2 (dθ2 + sin2 θdφ2 ) (1.7)

Dropping the primes, the line element then becomes

dτ2 = eν dt2 − eλ dr2 − r2 (dθ2 + sin2 θdφ2) (1.8)

where

ν = ν(t, r), λ = λ(t, r) (1.9)

The metric components are

g00 = A, g11 = −B, g22 = −r2 , g33 = −r2 sin2 θ (1.10)

g00 = eν , g11 = −eλ , g22 = −r2 , g33 = −r2 sin2 θ (1.11)

The metric is diagonal so that g00 = 1/g00 . The coordinates are labelled

8
 x0 = t, x1 = r, x2 = θ, x3 = φ.

By definition

1
Γabc = gad (gbd,c + gcd,b − gbc,d ).
2

There are 43 /2 = 32 independent Γabc s. However many will be zero. For example, it is obvious
that if all components are distinct the connection vanishes.

Let us do some examples.

1 0d
Γ001 = g (g0d,1 + g1d,0 − g01,d )
2
1 00
= g (g00,1 + g10,0 − g01,0 )
2
ν′
= (1.12)
2

where we have used that the metric is diagonal and the prime denotes differentiation with respect
to r.

Consider Γ100

1 11
Γ100 = g (g01,0 + g01,0 − g00,1 )
2
1
= − g11 g00,1
2
−λ
e ′ ν
= νe
2
eν−λ ′
= ν (1.13)
2

Consider Γ111

1 11
Γ100 = g (g11,1 + g11,1 − g11,1 )
2
1 11
= g g11,1
2
e−λ ′ λ
= λe
2
λ′
= (1.14)
2
9
Consider Γ122

1 11
Γ122 = g (g21,2 + g21,b − g22,1 )
2
1
= − g11 g22,1
2
e−λ
= (−2r)
2
= −re−λ (1.15)

Consider Γ212

1 22
Γ212 = g (g12,2 + g22,1 − g12,2 )
2
1 22
= g g22,1
2
1
= 2r
2r2
= 1/r (1.16)

Consider Γ313

1 33
Γ313 = g (g13,3 + g33,1 − g13,3 )
2
1 33
= g g33,1
2
1
= − 2
(−2r sin2 θ)
2
2r sin θ
= 1/r. (1.17)

Consider Γ133

1 1d
Γ133 = g (g3d,3 + g3d,3 − g33,d )
2
1 11
= g (g31,3 + g31,3 − g33,1 )
2
= −g11 g33,1
= −r sin2 θ/B. (1.18)

10
Consider Γ332

1 33
Γ332 = g (g33,2 + g23,3 − g32,3 )
2
1 33
= g g33,2
2
1 ∂
= − 2
(−r2 sin2 θ)
2r2 sin θ ∂θ
cos θ
= = cot θ. (1.19)
sin θ

Consider Γ333

1 33
Γ333 = g (g33,3 + g33,3 − g33,3 )
2
1 33
= g g33,3 = 0 (1.20)
2

As none of the metric components depend on φ. There are many other examples that turn out
to be zero.

Consider Γ000

1 00
Γ000 = g (g00,0 + g00,0 − g00,0 )
2
1 00
= g g00,0
2
e−ν ν
= ν̇e
2
ν̇
= (1.21)
2

Consider Γ011

1 00
Γ001 = g (g00,1 + g10,0 − g01,0 )
2
1 00
= g g00,1
2
e−ν ′ ν
= νe
2
ν′
= (1.22)
2
11
Consider Γ011

1 00
Γ011 = g (g10,1 + g10,1 − g11,0 )
2
1
= − g00 g11,0
2
eλ−ν
= λ̇ (1.23)
2

Consider Γ100

1 11
Γ100 = g (g01,0 + g01,0 − g00,1 )
2
1
= − g11 g00,1
2
eν−λ ′
= − ν (1.24)
2

Consider Γ101

1 11
Γ101 = g (g01,1 + g11,0 − g01,1 )
2
1 11
= g g11,0
2
e−λ λ
= λ̇e
2
λ̇
= (1.25)
2

We find altogether:

12
 1 ν−λ ′
Γ100 = e ν
2
1 ′
Γ111 = λ
2
Γ122 = −re−λ ,
Γ133 = −re−λ sin2 θ,
Γ212 = Γ221 = Γ313 = Γ331 = 1/r
Γ233 = − sin θ cos θ,
Γ332 = Γ323 = cot θ
ν̇
Γ000 = ,
2
ν′
Γ001 =
2
eλ−ν
Γ011 = λ̇
2
eν−λ ′
Γ100 = − ν
2
1
Γ101 = λ̇ (1.26)
2

All others being zero.

We can calculate the Riemann tensor using

Ra bcd = ∂c Γabd − ∂d Γabc + Γebd Γaec − Γebc Γaed (1.27)

Ra bcd = −Ra bdc (1.28)

The non-vanishing components of the Riemann tensor are

′ !
λ̈ λ̇2 ν′ λ̇ ν′ λ′ ν′′ ν 2
!
0 λ−ν
R 101 = e + − + − −
2 4 4 4 2 4
−λ ′
R0 202 = −re ν /2
R0 303 = −re−λ sin2 θν′ /2
R0 212 = −re−ν λ̇/2
R0 313 = −re−ν sin2 θλ̇/2
R1 212 = re−λ λ′ /2
R1 313 = re−λ sin2 θλ′ /2
R2 323 = (1 − e−λ ) sin2 θ (1.29)

13
We use these to calculate the components of the Ricci tensor.

Rab = ∂b Γcac − ∂c Γcab + Γdac Γcbd − Γdab Γcdc (1.30)

The non-vanishing components of the Ricci tensor are

′′ ′
λ̈ λ̇2 ν̇λ̇ ν 2 ν′ λ′ 1 ′
! !
λ−ν ν
R00 = + − +e + − + ν (1.31)
2 4 4 2 4 4 r
′′ ′2 ′ ′
! 2
!
ν ν νλ 1 ′ λ−ν λ̈ λ̇ ν̇λ̇
R11 = − + − − λ +e + − (1.32)
2 4 4 r 2 4 4
1
R01 = λ̇ (1.33)
r 
r 
R22 = e−λ (λ′ − ν′ ) − 1 + 1 (1.34)
2
R33 = R22 sin2 θ (1.35)

(1.33) vanishing implies

λ = λ(r).

From R00 and R11 vanishing independently

1
0 = eλ−ν R00 + R11 = (λ′ + ν′ ) (1.36)
r

implying

λ + ν = h(t) (1.37)

where h(t) is an arbitrary function of integration. As λ is purely a function of r, (1.34) is an

ordinary differential equation, which we write

e−λ − re−λ λ′ = 1,

or

(re−λ )′ = 1.

Integrating, we get

14
 re−λ = r + Const.

so

eλ = (1 + Const./r)−1 . (1.38)

Note

eν = eh(t)−λ

The line element becomes

 Const.  2 dr2
2
ds = e h(t)
1+ dt − − r2 (dθ2 + sin2 θdφ2 ) (1.39)
r 1 + Const./r

Write

Z t
′
t = eh(u)/2 du (1.40)

then

′
dt 2 = eh(t) dt2

Dropping primes the most general spherically symmetric solution of the vacuum field equations
is

 Const.  2 dr2
2
ds = 1 + dt − − r2 (dθ2 + sin2 θdφ2 ) (1.41)
r 1 + Const./r

All of the metric components are independent of a time coordinate. We have therefore proven
that any spherically symmetric vacuum metric is stationary.

Now consider the Newtonian limit. A point mass M situated at the origin in Newtonian theory
gives rise to a potential

φ = −GM/r

Using this in the weak field limit gives

15
 g00 ≃ 1 + 2φ/c2 = 1 − 2GM/c2 r.

We see that the constant turns out to be −2GM/c2

dr2
!
2GM
2
ds = 1 − 2 dt2 − 2
− r2 (dθ2 + sin2 θdφ2 ) (1.42)
cr 1 − 2GM/c r

16
Chapter 2

Schwarzschild Black Hole

2.1 Eddington-Finkelstein Coordinates

In the Eddington-Finkelstein coordinates radial null geodesics become straight lines

t = t + 2m ln(r − 2m) (2.1)

2m
dt = dt − dr (2.2)
r − 2m
Squaring gives

! ! !−1 2
2M 2M 2m 2m
dt2 =
 
1− 1− dt − 1− dr
r r r r
!−1
4m2
!
2M 2 2m 2m
= 1− dt − 2 dtdr + 2 1 − dr2 (2.3)
r r r r

Now we substitute it into the Schwarzschild line element.

! !−1
2M 2m
ds 2
= 1− 2
dt − 1 − dr2 − r2 (dθ2 + sin2 θdφ2 )
r r
!−1
4m2
! !
2M 2 4m 2m
= 1− dt − dtdr − 1 − 2 1 − dr2
r r r r
−r2 (dθ2 + sin2 θdφ2 )
! !
2M 2 4m 2M
= 1− dt − dtdr − 1 − dr2 − r2 (dθ2 + sin2 θdφ2 ) (2.4)
r r r

17
and obtain the line element

! !
2M 2 4m 2M
2
ds = 1 − dt − dtdr − 1 − dr2 − r2 (dθ2 + sin2 θdφ2 ) (2.5)
r r r

Notice that in this new coordinate system we no longer have a coordinate signularity at R = 2M.
In fact it is regular for the whole range 0 < r < 2m.

2.1.1 Advanced time parameter

As an external observer we cannot chart the course of a particle upon entering the event horizon.
However, a particle entering would pass through the event horizon unaware of anything strange.
The singularity at R = 2M of the Schwarzschild metric is unphysical, it is simply a coordinate
singularity.

To show it isn’t a physical singularity, let us make a change of coordinates and derive a new
metric. We will keep r, θ but replace t with

r
t = v − r − 2M ln −1 (2.6)
2M

Let us find dt and substitute it back into the Schwarzschild line element. We begin by taking
the derivative of t,

dr
dt = dv − dr −
r/2M − 1
!
r/2M − 1 + 1
= dv − dr
r/2M − 1
rdr
= dv −
r − 2M
!−1
2M
= dv − 1 − dr (2.7)
r

Squaring this we get

!−1 !−2
2 2 2M 2M
dt = dv − 2 1 − dvdr + 1 − dr2 . (2.8)
r r

Now we substitute it into the Schwarzschild line element.

18
 dr2
!
2M
ds 2
= 1− dt2 − − r2 (dθ2 + sin2 θdφ2 )
r 1 − 2M/r
! !−1 !−2 
2M  2 2M 2M 2

= 1− dv − 2 1 − dvdr + 1 − dr 
r r r
!−1
2M
− 1− dr2 − r2 (dθ2 + sin2 θdφ2 )
r
!
2M
= 1− dv2 − 2dvdr − r2 (dθ2 + sin2 θdφ2) (2.9)
r

We have the line element in the Eddington-Finkelstein coordinates

!
2M
2
ds = 1 − dv2 − 2dvdr − r2 (dθ2 + sin2 θdφ2 ). (2.10)
r

Notice that in this new coordinate system we no longer have a coordinate signularity at R = 2M.
We are now able to explore what happens inside the Schwarzschild black hole. Notice however
we still have a singularity at r = 0. This is a real physical singularity, and not due to the
coordinates used.

19
Chapter 3

Internal Schwarzschild Solution

Perfect Fluid Schwarzschild Solution

The metric has the form

ds2 = eν(r)dt2 − eλ(r) dr2 − r2 (dθ2 + sin2 θdφ2 ) (3.1)

We need to solve the full Einstein equation,

1
Gab = Rab − Rgab = 8πGT ab . (3.2)
2

The Ricci scalar is giveb by

!
−λ ′′1 ′ 1 ′ ′ 2 ′ ′ 2 2
R=e ν + ν 2 − ν λ + (ν − λ ) + 2 − 2 . (3.3)
2 2 r r r

The nonvanishing components of the Einstein tensor Gab are given by

1 ν−λ  ′ λ

G00 = e rλ − 1 + e
r2
1  
G11 = 2 rν′ + 1 − eλ
r
′
′′
ν 2 ν′ λ′
!
2 −λ ν 1 ′ ′
G22 = r e + − + (ν − λ )
2 4 4 2r
2
G33 = sin θG22 (3.4)

The star is modelled by a perfect fluid

20
 T ab = (ρ + p)Ua Ub − pgab . (3.5)

where ρ and p are the energy density and uniform pressure as measured in the rest frame of the
fluid, and U is the fluid four velocity - assumed to be

Ua = (eν/2 , 0, 0, 0) (3.6)

U a Ua = 1

 ν
 e ρ 0

0 0 
 0 eλ p 0 0 
T ab =  2
 (3.7)
 0 0 r p 0 

0 0 0 r sin2 θ p
2

Einstein’s equations read

1 −λ  ′ λ

e rλ − 1 + e = 8πGρ, (3.8)
r2

1 −λ  ′ λ

e rν + 1 − e = 8πGp, (3.9)
r2

′
ν′′ ν 2 ν′ λ′
!
−λ 1 ′ ′
e + − + (ν − λ ) = 8πGp (3.10)
2 4 4 2r

Note that the first of these equations involves only λ(r) and the density ρ(r). Write

1
m(r) = (r − re−λ ) (3.11)
2G

then the first equation becomes

dm
= 4πr2 ρ (3.12)
dr

which that can be integrated

Z r
′
m(r) = 4π ρ(r′ )r 2 dr′ . (3.13)
0

21
Here, ρ is the density and so we interpret m(r) as the total mass of the star enclosed by a
sphere of radius r. Say or star extends to a radius R, after which spacetime is described by the
Schwarzschild solution. We must have

Z R
M = m(R) = 4π ρ(r′ )r2 dr. (3.14)
0

where M is the Schwarzschild mass.

!−1
λ 2Gm(r)
e = 1− (3.15)
r

so that the line element becomes

!−1
ν(r) 2Gm(r)
2
ds = e 2
dt − 1 − dr2 − r2 (dθ2 + sin2 θdφ2 ) (3.16)
r

Equation (3.9) becomes

!
1 2Gm(r)
1
2
1− rν′ + 1 − 2 = 8πGp,
r r r

or

(r − 2Gm(r)) rν′ + 1 = r + 8πGpr3 ,

or

r + 8πGpr3
rν′ + 1 = ,


(r − 2Gm(r))

and so

dν 2Gm(r) + 8πGpr3
= (3.17)
dr r (r − 2Gm(r))

From the energy-momentun ∇a T ab = 0, the component b = 1 gives

dν dp
(ρ + p) =− (3.18)
dr dr

22
we use this to eliminate ν from (3.17):

dp (ρ + p)[2Gm(r) + 8πGpr3 ]
=− (3.19)
dr r (r − 2Gm(r))

This is the equation of hydrostatic equilibrium, describing the balance between compression
due to gravity and the pressure gradiant force in the opposite direction.

23
Chapter 4

Penrose-Carter diagrams

Let M denote physical space-time with metric gab . The idea is to construct another “unphysical”
manifold M̃ with boundary I and metric g̃ab , such that M is conformal to the interior of M̃
with

g̃ab = Ω2 gab , (4.1)

(where Ω is the conformal factor) and so that the “infinity” of M is represented by the finite
hypersurface I. We realise the whole physical spacetime M as a subset of the unphysical
spacetime M̃.

Asymptotic properties of M and of fields in M can be investigated by studying I, and the local
behaviour of the fields at I provided the relavent information is conformally invariant.

We will show that the null geodesics of conformally related metrics are the same, and hence
have the same causal structure.

In chapter 7 we use such diagrams when investigating Hawking radiation.

4.1 Geodesics under conformal transformations

We denote the inverse metric to g̃ab by g̃ab . Obviously,

g̃ab = Ω−2 gab .

˜ a denote the derivative operator associated with g̃ab .

Let ∇

˜ a g̃ab = 0
∇ (4.2)

24
implies

1
Γ̃abc = g̃ad (∇b g̃cd + ∇c g̃bd − ∇d g̃bc ) (4.3)
2

But since ∇a gbc = 0, we have

∇b g̃cd = ∇b (Ω2 gcd ) = 2Ωgcd ∇b Ω (4.4)

Γ̃abc = Ω−1 gad (gcd ∇b Ω + gbd ∇c Ω − gbc ∇d Ω)

= 2δa(b ∇c) ln Ω − gbc ∇d ln Ω (4.5)

The tangent, va , to an affinely parameterised geodesic γ with respect to ∇b satifies

vb ∇b va = 0 (4.6)

Hence

vb ∇˜ b va = vb ∇b va + vb Γ̃abc vc
= va (2vc ∇c ln Ω) − (gbc vb vc )gad ∇d ln Ω (4.7)

Thus, γ fails to be a geodesic with respect to ∇˜ b unless gbc vb vc = 0, in which case it is the
non-affinely parameterised geodesic equation - the RHS is proportional to va . Hence we have
proved the result.

4.2 Penrose-Carter Diagram for Minkowski Spacetime.

As a first example to illustrate the idea of a Penrose diagram we consider the procedure for
Minkowski spacetime.

We introduce coordinates

v = t + r, (4.8)
w = t − r, (4.9)

25
It is obvious that

−∞ < v < ∞, −∞ < w < ∞ (4.10)

v ≥ w, (4.11)

the Minkowski spacetime becomes

1
ds2 = dvdw − (v − w)2 (dθ2 + sin2 θdφ2). (4.12)
4

We define new coordinates p and q by

p = tan−1 v, (4.13)
q = tan−1 w, (4.14)

with coordinate ranges

1 1 1 1
− π < p < π, − π < q < π (4.15)
2 2 2 2
p ≥ q. (4.16)

1
ds2 = sec2 p sec2 q[4d pdq − sin2 (p − q)(dθ2 + sin2 θdφ2)] (4.17)
4

and the line element of the unphysical metric is

d s̃2 = 4d pdq − sin2 (p − q)(dθ2 + sin2 θdφ2) (4.18)

1
Ω= sec2 p sec2 q (4.19)
4

We introduce coordinates

t′ = p + q, (4.20)
r′ = p − q, (4.21)

26
The unphysical metric is now

′ ′
d s̃2 = dt 2 − dr 2 − sin2 r′ (dθ2 + sin2 θdφ2 ) (4.22)

subject to the coordinate range

−π < t′ + r′ < π, (4.23)

−π < t′ − r′ < π, (4.24)
r′ ≥ 0. (4.25)

i+
I+

t = Const.

r=0 i0

r = Const.

I−
i−

Figure 4.1: Penrose diagram of Minkowski spacetime

The whole of Minkowski spacetime has been shrunk into a finite or compact region. The process
is called conformal compactification.

I+ is future null infinity,

I− is past null infinity,

i+ is future timelike infinity,

i− is past timelike infinity,

i0 is past spacelike infinity.

27
4.3 Maximal Extensions

4.4 The Kruskal Solution

The coordinate choice most useful for path integral investigation into black hole radiation.
i+ i+
II
I+ I+

i0 I′ I i0

I− I−
II ′
i− i−

Figure 4.2: Penrose diagram of the Kruskal solution

Singularity i+

I+
H+

Black Hole

i0
r=0

I−

i−

Figure 4.3: Penrose diagram of a black hole.

28
Chapter 5

Charged Balck Holes

We now obtain the Reissener-Nordstrom solution that describes a charged non-rotating black
hole. We look for a static, asymptotically flat, shrerically symmetric solution of the Einstein-
Maxwell field equations. The Einstein-Maxwell equations are

Gab = 8πT ab (5.1)

where T ab is the energy-momentum tensor of electromagnetism

1 1
T ab = (−gcd F ac F bd + gab F cd F cd ) (5.2)
4π 4

and where F ab is the field strength tensor. Note that T ab has zero trace,

1 1
T = gab T ab = (−gab gcd F ac F bd + gab gab F cd F cd ) = 0. (5.3)
4π 4

This implies the vanishing of the Ricci scalar as the trace of the Einstein tensor must vanish

1
R − 4R = T aa = 0 ⇒R=0
2

We can then instead use

Rab = 8πT ab (5.4)

In source-free regions the tensor F ab must satisfy Maxwell’s equations

29
 ∇b F ab = 0, (5.5)
∂[a F bc] = 0. (5.6)

We can assume there are coordinates (t, r, θ, φ) so thay the metric reduces to the form

ds2 = eν dt2 − eλ dr2 − r2 (dθ2 + sin2 θdφ2 ). (5.7)

If we impose the condition that the solution is static, then this requires ν and λ are functions of
r only,

ν = ν(r), λ = λ(r). (5.8)

We also assume the solution to be asymptotically flat.

We read off from (5.7) that

 ν 
 e 0 0 0 
 0 −eλ 0 0 
gab =  2
 (5.9)
 0 0 −r 0 

0 0 0 −r2 sin2 θ

and so

 −ν 
 e 0 0 0 
 0 −e−λ 0 0 
gab =  2
 (5.10)
 0 0 −1/r 0 

0 0 0 −1/r sin2 θ
2

and

p p 1
|g| = − det(gab ) = e 2 (ν+λ) r2 sin θ. (5.11)

The non-zero components of the Ricci tensor Rab are then

30
 ν′′ ν′ 2 ν′ λ′ 1 ′
!
λ−ν
R00 = e + − + ν
2 4 4 r
′′ ′2 ′ ′
!
ν ν νλ 1 ′
R11 = − + − − λ
2 4 4 r
r 
R22 = e−λ (λ′ − ν′ ) − 1 + 1
2
R33 = R22 sin2 θ (5.12)

which follow from (1.31)-(1.35) and the assumption that ν = ν(r) and λ = λ(r).

5.0.1 Solving the Maxwell equations and the electric charge

In spherical polar coordinates the Maxwell tensor takes the form

 
 0 −1 0 0 
 1 0 0 0 
F ab = E(r)   . (5.13)
 0 0 0 0 
0 0 0 0

To confirm this ansatz we will take the definition of the Maxwell tensor in Minkoski coordinates
(t, x, y, z):

 
 0 −E x −E y −E z 
 E 0 Bz −By 
F µν =  x (5.14)
 E y −Bz 0 B x 

E z By −B x 0


and find its components in spherical polar coordinates (t, r, θ, φ)

t = t
x = r sin θ cos φ
y = r sin θ sin φ
z = r cos θ (5.15)

via the tansformation formula

′ ∂xσ ∂xρ
F µν = ′ µ ′ ν F σρ , (5.16)
∂x ∂x

31
where we easily find that,

 
1 0 0 0 
∂x 0 sin θ cos φ r cos θ cos φ −r sin θ sin φ


= (5.17)
∂x′ 0 sin θ sin φ r cos θ sin φ r sin θ cos φ 

0 cos θ −r sin θ 0

while inserting the assumption that the elecric field is radial:

E x = E(r) sin θ cos φ

E y = E(r) sin θ sin φ
E z = E(r) cos θ. (5.18)

We obtain from (5.16)

 
1 0 0 0 
0 sin θ cos φ sin θ sin φ cos θ 

0 r cos θ cos φ r cos θ sin φ −r sin θ
 
0 −r sin θ sin φ r sin θ cos φ 0
−E(r) sin θ cos φ −E(r) sin θ sin φ −E(r) cos θ
 
 0 
 E(r) sin θ cos φ 0 0 0 
 
 E(r) sin θ sin φ 0 0 0 

E(r) cos θ 0 0 0

 
1 0 0 0 
0 sin θ cos φ r cos θ cos φ −r sin θ sin φ
 
0 sin θ sin φ r cos θ sin φ r sin θ cos φ 
 
0 cos θ −r sin θ 0
 
1 0 0 0 
0 sin θ cos φ sin θ sin φ cos θ 
= 
0 r cos θ cos φ r cos θ sin φ −r sin θ

0 −r sin θ sin φ r sin θ cos φ 0
 
 0 −E(r) 0 0 
 E(r) sin θ cos φ 0 0 0 

 E(r) sin θ sin φ 0 0 0 


E(r) cos θ 0 0 0

 
 0 −E(r) 0 0
E(r) 0 0 0
=  

 0 0 0 0

0 0 0 0

(5.19)

32
confirming the ansatz.

Raising the indices using the inverse metric, we obtain

 −ν  
 e 0 0 0  0 −1 0 0
 0 −e−λ 0 0  1 0 0 0
F ab = E(r)  2
 
 0 0 0 0

 0 0 −1/r 0   
0 0 0 −1/r2 sin2 θ 0 0 0 0
 −ν 
 e 0 0 0 
 0 −e−λ 0 0 
 
2
 0
 0 −1/r 0 

0 0 0 −1/r sin2 θ
2
 −ν
  0 e−λ 0 0
 
 e 0 0 0
 0 −e−λ 0 0  e−ν 0 0 0
= E(r)    
 0 0 −1/r2 0   0 0 0 0
−1/r2 sin2 θ 0 0 0 0
 
0 0 0
 
 0 1 0 0
−1 0 0 0
 
= e−ν−λ E(r)  (5.20)
 0 0 0 0

0 0 0 0

We verify that (5.6) is automatically satisfied using (5.13). We can use F ab = −F ba to write

2
∂[a F bc] = (∂a F bc + ∂c F ab + ∂b F ca ) (5.21)
3!

We need only check for

a = 0, b = 1, c = 2
a = 0, b = 1, c = 3
a = 0, b = 2, c = 3
a = 1, b = 2, c = 3. (5.22)

First

2
∂[0 F 12] = (∂0 F 12 + ∂2 F 01 + ∂1 F 20 )
3!
= 0 (5.23)

33
then

2
∂[0 F 13] = (∂0 F 13 + ∂3 F 01 + ∂1 F 30 )
3!
= 0 (5.24)

then

2
∂[0 F 23] = (∂0 F 23 + ∂3 F 02 + ∂2 F 30 )
3!
= 0 (5.25)

and lastly

2
∂[0 F 23] = (∂0 F 23 + ∂3 F 02 + ∂2 F 30 )
3!
= 0. (5.26)

To write out (5.5) we will use the following expression that holds for any antisymmetric rank
two tensor X ab = −X ba

1 p
∇a X ab = p ∂a ( |g|X ab ). (5.27)
|g|

To prove this result we need

1 p
Γaab = p ∂b |g|. (5.28)
|g|

Observe for a rank two tensor X ab

∇a X ab = ∂a X ab + Γaac X cb + Γbac X ac
1 p
= ∂a X ab + p (∂c |g|)X cb + Γbac X ac
|g|
1 p
= p ∂c ( |g|X cb ) + Γbac X ac (5.29)
|g|

34
Since we require the tensor X ab to be antisymmetric the last term vanishes and we have estab-
lished the result.

We now apply it to the tensor F ab :

1 p
∇a F ab = p ∂a ( |g|F ab )
|g|
1
e− 2 (ν+µ) 1
= 2 ∂a (e 2 (ν+µ) r2 sin θF ab ) (5.30)
r sin θ

From the t component of Maxwell’s equation (5.5) and using (5.20) we must have

0 = ∇a F a0
1
e− 2 (ν+µ) 1
= 2 ∂a (e 2 (ν+µ) r2 sin θF a0 )
r sin θ
1
e− 2 (ν+µ) 1
= 2
∂1 (e 2 (ν+µ) r2 F 10 )
r
1
e− 2 (ν+µ) 1
= − 2
∂1 (e− 2 (ν+µ) r2 E). (5.31)
r

Integrating this gives

1
E = e 2 (ν+λ) ǫ/r2 (5.32)

where ǫ is the constant of integration. Assuming the solution is asymptotically flat requires

ν, λ → 0 as r → ∞ (5.33)

and so asymptotically

E ∼ ǫ/r2

We therefore interpret ǫ as the charge of the blackhole.

We confirm the other Maxwell’s eqations. From the r component

35
 1
a1 e− 2 (ν+µ) 1
∇a F = 2 ∂a (e 2 (ν+µ) r2 sin θF a1 )
r sin θ
1
e− 2 (ν+µ) 1
= 2 ∂0 (e 2 (ν+µ) r2 sin θF 01 )
r sin θ
= 0 (5.34)

as there is no time-dependence. The other two are zero by spherical symmetry assumption. In
particular as F a2 = 0

1
a2 e− 2 (ν+µ) 1
∇a F = 2 ∂a (e 2 (ν+µ) r2 sin θF a2 )
r sin θ
= 0. (5.35)

Similarly for ∇a F a3 = 0.

5.0.2 Calculation of the energy-momentum tensor

We employ the ansatz (5.13) together with (5.9) and (5.10) to compute the Maxwell energy
momentum tensor:

1 1
T ab = (−gcd F ac F bd + gab F cd F cd ). (5.36)
4π 4

First we compute

F cd F cd = gce gd f F cd F e f
= g0e g1 f F 01 F e f + g1e g0 f F 10 F e f
= 2g0e g1 f F 01 F e f
= 2g00 g11 F 01 F 01
= −2e−ν−λ E 2 . (5.37)

We find the componets of T ab : First T 00

36
 1 1
T 00 = (−gcd F 0c F 0d + g00 F cd F cd )
4π 4
1 1
= (−g11 F 01 F 01 + g00 F cd F cd )
4π 4
1 −λ 2 1 ν −ν−λ 2
= (e E − e e E )
4π 2
1 −λ 2
= e E (5.38)
8π

Next T 11

1 1
T 11 = (−gcd F 1c F 1d + g11 F cd F cd )
4π 4
1 1
= (−g00 F 10 F 10 + g11 F cd F cd )
4π 4
1 −ν 2 1 λ −ν−λ 2
= − (e E − e e E )
4π 2
1
= − e−ν E 2 (5.39)
8π

Then T 22

1 1
T 22 = (−gcd F 2c F 2d + g22 F cd F cd )
4π 4
1
= g22 F cd F cd
16π
1 2 −ν−λ 2
= re E . (5.40)
8π

It is easy to see that T 33 = T 22 sin2 θ. We compute T 01 :

1 1
T 01 = (−gcd F 0c F 1d + g01 F cd F cd )
4π 4
1 10
= − g F 01 F 10
4π
= 0. (5.41)

In acordance with (5.12), (note in the derivation of the Schwarzschild solution that R01 =
8πT 01 = 0 implied λ(t, r) = λ(r) by (1.33)).

The other components of T ab are easily seen to be zero, which are in accordance with (5.12).

37
5.0.3 Solving the field equations

The equations R00 = 8πT 00 and R11 = 8πT 11 can be combined

eλ−ν R00 + R11 = 8π[eλ−ν T 00 + T 11 ] (5.42)

By (5.12), (5.38) and (5.39), from the 00 and 11 equations we get

λ′ + ν′ = 0 (5.43)

which by (5.33) means that

λ = −ν. (5.44)

The 22 equation

R22 = 8πT 22 = r2 e−ν−λ E 2

becomes on using (5.12), (5.32) and (5.44)

eν (−rν′ − 1) + 1 = r2 E 2 = ǫ 2 /r2 (5.45)

or

(reν )′ = 1 − ǫ 2 /r2 (5.46)

integrating

eν = 1 + Const./r + ǫ 2 /r2 (5.47)

(note that because R33 = R22 sin2 θ and T 33 = T 22 sin2 θ the field equation R33 = 8πT 33 gives the
same equation as R22 = 8πT 22 ). We obtain the line element

!−1
Const. ǫ 2 Const. ǫ 2
!
2
ds = 1 + 2
+ 2 dt − 1 + + 2 dr2 − r2 (dθ2 + sin2 θdφ2). (5.48)
r r r r

When ǫ = 0, this reduces to the Schwarzschild line element and so

38
 Const. = −2MG/c2 .

We obtain the Reissner-Nordstrom solution

!−1
2MG ǫ 2 2MG ǫ 2
!
2 2
ds = 1 − 2 + 2 dt − 1 − 2 + 2 dr2 − r2 (dθ2 + sin2 θdφ2 ). (5.49)
cr r cr r

5.1 Event Horizons

Recall that at the event horizon was the point after which the coefficients of dr and dt change
signs.

The horizon function H(r) is given by

2m ǫ 2
!
H(r) = 1 − + 2 (5.50)
r r

is quadratic with two distinct roots. Write

r2 H(r) = r2 − 2mr + ǫ 2 = (r − m)2 + ǫ 2 − M 2 . (5.51)

Notice that we must have ǫ 2 ≤ m2 otherwise the roots would be complex.

The two roots are given by

r+ = m + (m2 − ǫ 2 )1/2
r− = m − (m2 − ǫ 2 )1/2 (5.52)

These two roots correspond to two different event horizons, one at r+ and the other at r− .

The line element has three regular regions

I. r+ < r < ∞,
II. r− < r < r+ ,
III. 0 < r < r− . (5.53)

The outer horizon at r+ is much like the event horizon of 2m for a Schwarzschild black hole.

39
Space and time change roles.

At the inner horizon, the Cauchy horizon, space and time change roles again.

Inside the Cauchy horizon the sigularity is space-like.

As the more charge is entered into the black hole, the inner event horizon gets larger, while the
outer event horizon starts to shrink. When we reach the maximum possible charge, i.e. ǫ 2 = m2 ,
the two horizons merge and only the regions I and III exist.

5.2 Analogue of Eddington-Finkelstein Coordinates

ǫ 2 < m2

r+2 r−2
t=t− ln(r − r+ ) + ln(r − r− ) (5.54)
r+ − r− r+ − r−

r+ − r− = 2(m2 − ǫ 2 )1/2

r+ − r− = 2(m2 − ǫ 2 )1/2
r+ + r− = 2m (5.55)
r+ r− = ǫ 2 (5.56)

r+2 dr r−2 dr
dt = dt − +
r+ − r− r − r+ r+ − r− r − r−
r+2 r−2
!
dr
= dt − + (5.57)
r+ − r− r − r+ r − r−

!2
r+2 r−2 dr2 r+2 r−2
!
2 2 1
dt = dt − 2 − dtdr + −
r+ − r− r − r+ r − r− (r+ − r− )2 r − r+ r − r−
(5.58)

Notice that the horizon function is

r2 H(r) = (r − r+ )(r − r− )

40
 (r − r+ )(r − r− ) 2 r2
ds2 = dt − dr2 − r2 (dθ2 + sin2 θdφ2 )
r2 (r − r+ )(r − r− )
r+2 r−2
!
(r − r+ )(r − r− ) 2 (r − r+ )(r − r− ) 1
= dt − 2 − dtdr
r2 r2 r+ − r− r − r+ r − r−
!2
(r − r+ )(r − r− ) dr2 r+2 r−2
+ −
r2 (r+ − r− )2 r − r+ r − r−
r2
− dr2 − r2 (dθ2 + sin2 θdφ2 ) (5.59)
(r − r+ )(r − r− )

Consider the coefficient of −2dtdr:

r+2 r−2
!
(r − r+ )(r − r− ) 1
−
r2 r+ − r− r − r+ r − r−
1 
2 2

= r (r − r − ) − r − (r − r + )
r2 (r+ − r− ) +
1 
2 2 2 2

= r(r + − r − ) − (r + r − − r − r + )
r2 (r+ − r− )
r+ + r− r+ r−
= − 2
r r
2m ǫ 2
= − 2 (5.60)
r r

Consider the coefficient of −dr2 :

!2
(r − r+ )(r − r− ) 1 r+2 r−2 r2
− − +
r2 (r+ − r− )2 r − r+ r − r− (r − r+ )(r − r− )
2
r2 2m ǫ 2 r2
!
= − − 2 +
(r − r+ )(r − r− ) r r (r − r+ )(r − r− )
2

2
!2 
r 1 − 2m − ǫ 
 
=
(r − r+ )(r − r− ) r r2
2 2
 !
1 
1 − 2m ǫ
− 2 

= 2 2
1 − 2m/r − ǫ /r r r
2m ǫ 2
= 1+ − 2 (5.61)
r r

Introduce the function f by

41
 2m ǫ 2
f = 1 − g00 = − 2 (5.62)
r r

Then the line element can becomes

2
ds2 = (1 − f )dt − 2 f dtdr − (1 + f )dr2 − r2 (dθ2 + sin2 θdφ2) (5.63)

Notice we no longer have any coordinate singularities.

5.3 Penrose Diagram

5.4 Double null coordinates

Introduce double null coordinates

v = t + r, w = 2t − v, (5.64)

this implies

v = t + r∗ , w = t − r∗ , (5.65)

where

r+2 r−2
r∗ = r + ln(r − r+ ) − ln(r − r− ) (5.66)
r+ − r− r+ − r−

Obviously

2m ǫ 2
!
(r − r+ )(r − r− ) 2
1− + 2 dvdw = (dt − dr∗2 ) (5.67)
r r r2

The coefficient of dt2 can seen immediately to be correct. Now

r+2 r2
" !#
1
∗
dr = 1 + − − dr (5.68)
r+ − r− r − r+ r − r−

Squaring

42
  !2 
r+2 r−2 r+2 r−2
!
∗2
 1 1  2
dr = 1 + 2 − + −  dr (5.69)
r+ − r− r − r+ r − r− (r+ − r− )2 r − r+ r − r− 

(r − r+ )(r − r− ) ∗2
dr
r2
h (r − r+ )(r − r− )
= 2
dr∗2
r
r+2 r−2
!
(r − r+ )(r − r− ) 1
+2 −
r2 r+ − r− r − r+ r − r−
!2 i
(r − r+ )(r − r− ) 1 r+2 r−2
+ − dr2
r2 (r+ − r− )2 r − r+ r − r−
!2
r2 h 2m ǫ 2
= 1− + 2 +
(r − r+ )(r − r− ) r r
!2
2m ǫ 2 2m ǫ 2 2m ǫ 2 i 2
! !
+2 1 − + 2 − 2 + − 2 dr
r r r r r r
!2
r2 2m ǫ 2 2m ǫ 2
h !
= 1−2 + 2 + + 2
(r − r+ )(r − r− ) r r r r
2 2 2 2i
! ! !
2m ǫ 2m ǫ 2m ǫ
+2 1 − + 2 − 2 + − 2 dr2
r r r r r r
2
r
= dr2 . (5.70)
(r − r+ )(r − r− )

Therefore

!−1
2m ǫ 2 2m ǫ 2 2m ǫ 2
! !
2
1− + 2 dvdw = 1 − + 2 dt − 1 − + 2 dr2
r r r r r r

Thus the line element can be written in double null coordinates as

2m ǫ 2
!
1− + 2 dvdw − r2 (dθ2 + sin2 θdφ2 ) (5.71)
r r

5.5 Maximal extension

ǫ 2 < m2 , we define new coordinates

43
 ! !
′′ −1 r+ − r− ′′ −1 r+ − r−
v = tan exp v , w = tan − exp w (5.72)
4r+2 4r+2

Inverting we see

4r+2 4r+2
v= ln tan v′′ , w=− ln tan w′′ (5.73)
r+ − r− r+ − r−

As

d tan x 1 2
d ln tan x = = dx = dx = 2cosec2x dx
tan x sin x cos x sin 2x
we get

!2
4r+2
dvdw = −4 × 2 2 2
cosec2v′′ cosec2w′′ dv′′ dw′′
r+ − r−

Consider the product tan v′′ tan w′′

tan v′′ tan w′′ ! !

r+ − r− r− − r+
= − exp v exp w
4r+2 4r+2
!
r+ − r−
= − exp (v − w)
4r+2
!
r+ − r− ∗
= − exp r
2r+2
!
r+ − r−
= − exp r
2r+2
" 2 #! " 2 #!
r+ − r− r+ r+ − r− r−
× exp ln(r − r+ ) exp ln(r − r− )
2r+2 r+ − r− 2r+2 r+ − r−
!
r+ − r− 2 2
= − exp 2
r (r − r+ )1/2 (r − r− )r− 1/2r+ . (5.74)
2r+

Therefore the line element has the form

2m ǫ 2 4r+4
!
ds 2
= −64 1 − + 2 cosec2v′′ cosec2w′′ dv′′ dw′′
r r (r+ − r− )2
= −r2 (dθ2 + sin2 θdφ2 ) (5.75)

44
where r is defined implicitly by

!
r+ − r− 2 2
′′ ′′
tan v tan w = − exp 2
r (r − r+ )1/2 (r − r− )r− /2r+ . (5.76)
2r+

This line element is the analogue of the Kruskel solution and represents the maximal analytic
extension of the Reissner-Nordstrm solution for ǫ 2 < m2 .

The Penrose diagram for this maximal extension is shown in fig (5.5)

Singularity Singularity
r=0 i+ i+ r=0
II
I+ r= I+

r=
r+
r+

r=
r∞

r∞
r=

i0 I I i0
r=
r=

r∞
r+

r+
r∞
r=

r=
I− I−
II i−
i−
r=
r−

r−
r=

r=0 III III r=0

r=
r−

r−
r=

i+ II i+
r=

I+ I+
r=
r+
r+

r=

r∞

i0 I i0
r=

r∞
r+

I− I−
r=

i− i−

r=0 III r=0

Figure 5.1: Penrose diagram of a black hole.

45
Chapter 6

Rotating Black Holes

6.1 Field Equations

6.1.1 Degenerate metric

A metric of the form

gab = ηab − 2mlalb , la lb ηab = 0, m is arbitrary. (6.1)

is called a degenerate metric. The Schwarzschild metric is an example as can be seen from the
Eddington-Finkelstein form (2.5)

! !
2M 2 4m 2M
ds 2
= 1− dt − dtdr − 1 − dr2 − r2 (dθ2 + sin2 θdφ2 )
r r r
2 2m
= dt − dr2 − r2 (dθ2 + sin2 θdφ2 ) − (dt + dr)2
r
!2
2 2 2 2 2m xdx + ydy + zdz
= dt − (dx + dy + dz ) − dt + (6.2)
r r

Therefore the Schwarzschild metric can be written in degenerate form with la given by

1  x y z
la = √ 1, , , (6.3)
r r r r

It turns out that the solution for a rotating black-hole is also of the form of a degenerate metric.
Let us proceed. Define

46
 la := ηab lb . (6.4)

The inverse matrix of gab is

gab = ηab + 2mla lb (6.5)

as is easily seen:

gac gcb = (ηac − 2mla lc )(ηcb + 2mlc lb )

= δa b + 2mlalb − 2mlb la + 4m2 la lb lc lc
= δa b

It follows from (6.5) that the contravariant four vector corresponding to la is the same as ηab lb :

la := gac lc = ηac lc . (6.6)

Therefore its indices han be raised and lowered by either the true metric or the Lorentz metric.
Since la is null it has the property

1
0 = ∂d (ηab la lb ) = lc ∂d lc = lc ∂d lc . (6.7)
2

We can consider, g, the determinat of the metric. At any point la is a flat-space null vector:
la lb ηab = 0. We can perform a proper rotation of coordinates in three-space that leaves ηab
invariant and brings la into the form

 
 a 
 a 
  , (6.8)
 0 
 
0

see

  
 1 0 0 0   a 
 0 −1 0 0   a 
(a, a, 0, 0)     = 0. (6.9)
 0 0 −1 0   0 
  
0 0 0 −1 0

In this system we have

47
 1 − 2ma2 −2ma2 0 0
2 2
−2ma −1 − 2ma 0 0
g =
0 0 −1 0
0 0 0 −1
1 − 2ma2 −2ma2
=
−2ma2 −1 − 2ma2
= −(1 − 2ma2 )(1 + 2ma2 ) − 4m2 a4
= −1 (6.10)

Since a three-dimensional rotation has a unit Jacobian, the metric transforms as a scalar under
this transformation. Thus, g = −1 for any degenerate metric. It follows that

∂ √
Γaca = ln −g = 0 (6.11)
∂xc

This simplifies the field equations Rab = 0,

Rab = −∂c Γcab + Γdac Γcdb = 0 (6.12)

Now Rab involves different powers of m and since m is arbitrary each order must vanish sep-
arately. If we note that the Christoffel symbol of the first kind, Γab,d , is linear in m, power
counting is easy and we isolate terms corresponding to different powers of m,

Rab = −∂c Γcab + Γdac Γcdb

= −∂c (gcd Γab,d ) + gcd Γea,d ge f Γcb, f
−∂c (ηcd + 2mlc ld )Γab,d + (ηcd + 2mlc ld )(ηe f + 2mle l f ) Γea,d Γcb, f


=
= −ηcd ∂c Γab,d
−2m∂c lc ld Γab,d + ηcd ηe f Γea,d Γcb, f

+2m(lc ld ηe f + le l f ηcd )Γea,d Γcb, f

+4m2 lc ld le l f Γea,d Γcb, f (6.13)

We just rewrite one term by swapping the dummy variables d and f :

lc ld ηe f Γea,d Γcb, f = lc l f ηed Γea, f Γcb,d

This gives four sets of ten equations:

48
 −ηcd ∂c Γab,d = 0 O(m)
2m∂c lc ld Γab,d − ηcd ηe f Γea,d Γcb, f = 0 O(m2 )

le l f ηcd Γea,d Γcb, f + lc l f ηed Γea, f Γcb,d = 0 O(m3 )

4m2 lc ld le l f Γea,d Γcb, f = 0 O(m4 ) (6.14)

6.1.2 Order m4 equations:

We need to satisfy

lc ld le l f Γea,d Γcb, f = 0

From the form of the degenerate metric (6.1) we have

∂a gbc = −2m∂a (lb lc ). (6.15)

We will use this a number of times in what follows. Consider the part ld le Γea,d

1
ld le Γea,d = ld le le (∂e gad + ∂a ged − ∂d gea )
2
= −mld le ∂e (ld la ) + ∂a (ld le ) − ∂d (le la )

= −m le la (ld ∂e ld ) + (ld ld )le ∂e la +



+ (le le )ld ∂a ld + (ld ld )le ∂a le

−ld la (le ∂d le ) − (le le )ld ∂d la


= 0 (6.16)

Therefore the m4 equations are automatically stisfied by degenerate metrics.

6.1.3 Order m3 equations:

le l f ηcd Γea,d Γcb, f + lc l f ηed Γea, f Γcb,d = 0 (6.17)

First note that

lc l f ηed Γea, f Γcb,d = le l f ηcd Γeb,d Γca, f (6.18)

49
shows the second term in on the LHS (6.17) is the same as the first term with a and b exchanged.
Thus the field equations are

le l f ηcd Γea,d Γcb, f + a ↔ b = 0 (6.19)

Consider

1e
le Γea,d = l (∂e gad + ∂a ged − ∂d gea )
2
= −mle [∂e (la ld ) + ∂a (le ld ) − ∂d (le la )]
= −mle ∂e (la ld ), (6.20)

where again we have used (6.15), and also consider

1 f
l f Γcb, f = l (∂c gb f + ∂b gc f − ∂ f gcb )
2
= ml f ∂ f (lc lb ). (6.21)

Combining them we have

le l f ηcd Γea,d Γcb, f = −m2 le l f ηcd ∂e (la ld )∂ f (lc lb )

= −m2 le l f ηcd [ld ∂e la + la ∂e ld ][lb ∂ f lc + lc ∂ f lb ]
= −m2 le l f ηcd [ld lb (∂e la )(∂ f lc ) + la lb (∂e ld )(∂ f lc ) + la ld ∂e lc ∂ f lb ].
(6.22)

Let us consider the terms separately, first,

−m2 le l f ηcd ld lb (∂e la )(∂ f lc ) = ...ηcd ld ∂ f lc

= ...lc ∂ f lc
= 0. (6.23)

where we have used (6.7). Next

−m2 le l f ηcd la lb (∂e ld )(∂ f lc ) = −m2 la lb (le ∂e ld )(l f ∂ f ld ) (6.24)

50
by a simple rearangement of terms. Next

−m2 le l f ηcd la ld ∂e lc ∂ f lb = ...ηcd ld ∂e lc

= ...lc ∂e lc
= 0. (6.25)

where we have used (6.7) again.

Therefore

le l f ηcd Γea,d Γcb, f = −m2 la lb (le ∂e ld )(l f ∂ f ld ) (6.26)

Using this in (6.19), the m3 equations lead to

−m2 la lb (vc vc ) = 0 (6.27)

where

vc := ld ∂d lc . (6.28)

Therefore vc is a null vector. It is also orthogonal to the null vector lc , as easily seen

vb lb = (lc ∂c lb )lb = lc (lb ∂c lb ) = 0 (6.29)

where we have used (6.7). Notice that

va = gab vb
= (ηab − 2mla lb )vb
= ηab vb (6.30)

so that the indices of vc can be lowered and raised with the Lorentz metric, as with lc . At any
chosen point la and va may be written

lb = (|~l|, ~l), vb = (|~v|, ~v) (6.31)

where ~l an ~v are ordinary three-vectors in Eucldean space. If θ is the angle between ~l an ~v then

51
 ~l · ~v
cos θ = (6.32)
|~l| |~v|

Now because lb and vb are orthogonal (from (6.29)),

lb vb = la vb ηab = l0 v0 − ~l · ~v
= |~l||~v|(1 − cos θ) = 0 (6.33)

Thus cos θ = 1, and so ~v is parallel to ~l at any given point. We may therefore write

vb = lc ∂c lb = −A(xa )lb (6.34)

where A is a scalar field.

We will next consider the linear order in m equations and return to the order m2 equations later,
where we will show that they are identically satisified.

6.1.4 Equation from linear in m equation

Using (6.15) in the linear in m equation

−ηcd ∂c Γab,d = 0 (6.35)

gives

−ηcd ∂c Γab,d = −ηcd ∂c [∂a gbd + ∂b gad − ∂d gab ]

= 2mηcd [−∂c ∂d (la lb ) + ∂c ∂a (lb ld ) + ∂c ∂b (la ld )]
= 2m[−2 (la lb ) + ηcd ∂c ∂a (lb ld ) + ηcd ∂c ∂b (la ld )]
= 0 (6.36)

where we have introduced the D’Alembertian opertator

∂2
ηcd ∂c ∂d = 2 = − ∇2 .
∂x02

Define

52
 L := −∂a la . (6.37)

Expanding ηcd ∂c (lb ld ) gives

ηcd ∂c (lb ld ) = ηcd ld ∂c lb + ηcd lb ∂c ld

= −((−lc ∂c lb ) − (lb ∂c lc ))
= −(L + A)lb (6.38)

implying

ηcd ∂c ∂a (lb ld ) = ∂a ηcd ∂c (lb ld ) = −∂a [(L + A)lb ]

Substituting this into equation (6.36) we obtain the equation for la

−2 (la lb ) = ∂b [(L + A)la ] + ∂a [(L + A)lb ] (6.39)

We use this equation in the following to prove the order m2 equations are satisfied, proving the
order m2 equations are automatically satisfied by any solution to (6.39). The entire content of
the field equations are thus embodied in (6.39). After dealing with the m2 equations we will
specialise (6.39) to the stationary case where all x0 derivatives vanish.

6.1.5 Order m2 equations

Recall the order m2 equations

2m∂c lc ld Γab,d − ηcd ηe f Γea,d Γcb, f = 0.

(6.40)

Consider the first term, again we use (6.15), we have

2m
2m∂c lc ld Γab,d ∂c [lc ld (∂a gbd + ∂b gad − ∂d gab )]


=
2
= −2m2 ∂c [lc ld ∂a (lb ld ) + ∂b (la ld ) − ∂d (la lb ) ]


(6.41)

Looking at the first term in the brackets

ld ∂a (lb ld ) = lb (ld ∂a ld ) + (ld ld )∂a lb = 0

53
and similarly the second

ld ∂b (la ld ) = 0.

So

2m∂c lc ld Γab,d 2m2 ∂c [lc ld (lb ∂d la + la ∂d lb )]

=
= 2m2 ∂c [lc ld lb ∂d la + lc ld la ∂d lb )]
= 2m2 ∂c [(ld ∂d la )lc lb + (ld ∂d lb )lc la ]
= 2m2 ∂c [(−Ala )lc lb + (−Alb )lc la ]
= −4m2 ∂c [(Alc )(la lb )]
= −4m2 [la lb ∂c (Alc ) + Alc ∂c (la lb )]
= −4m2 [la lb ∂c (Alc ) + Alc lb ∂c la + Alc la ∂c lb ]
= −4m2 [la lb ∂c (Alc ) + Alb (lc ∂c la ) + Ala (lc ∂c lb )]
= −4m2 [la lb ∂c (Alc ) − 2A2 la lb ]
= −4m2 la lb [∂c (Alc ) − 2A2 ] (6.42)

So that

2m∂c lc ld Γab,d = −2m2 la lb [2∂c (Alc ) − 4A2 ]

(6.43)

Now lets us look at the second term in (6.40)

−ηcd ηe f Γea,d Γcb, f .

We have

ηcd ηe f 4m2
−ηcd ηe f Γea,d Γcb, f = − [∂e (la ld ) + ∂a (le ld ) − ∂d (le la )] ×
4
×[∂c (lb l f ) + ∂b (lc l f ) − ∂ f (lc lb )]
= −m2 ηcd ηe f [∂e (la ld )∂c (lb l f ) + ∂e (la ld )∂b (lc l f ) − ∂e (la ld )∂ f (lc lb )
∂a (le ld )∂c (lb l f ) + ∂a (le ld )∂b (lc l f ) − ∂a (le ld )∂ f (lc lb )
−∂d (le la )∂c (lb l f ) − ∂d (le la )∂b (lc l f ) + ∂d (le la )∂ f (lc lb )].
(6.44)

Tedious but straightforward calculations give

54
 1. ηcd ηe f ∂e (la ld )∂c (lb l f ) = la lb [3A2 + ∂d lc ∂c ld ]
2. ηcd ηe f ∂e (la ld )∂b (lc l f ) = 0
3. −ηcd ηe f ∂e (la ld )∂ f (lc lb ) = −la lb (∂d lc ∂d lc )
4. ηcd ηe f ∂a (le ld )∂c (lb l f ) = 0
5. ηcd ηe f ∂a (le ld )∂b (lc l f ) = 0
6. −ηcd ηe f ∂a (le ld )∂ f (lc lb ) = 0
7. −ηcd ηe f ∂d (le la )∂c (lb l f ) = −la lb (∂d lc ∂d lc )
8. −ηcd ηe f ∂d (le la )∂b (lc l f ) = 0
9. ηcd ηe f ∂d (le la )∂ f (lc lb ) = la lb [3A2 + ∂d lc ∂c ld ] (6.45)

so that

−ηcd ηe f Γea,d Γcb, f = −m2 [1. + · · · + 9.]

= −2m2 la lb [3A2 + ∂d lc ∂c ld − ∂d lc ∂d lc ] (6.46)

Putting together (6.43) and (6.46)

2m∂c lc ld Γab,d − ηcd ηe f Γea,d Γcb, f

= −2m2 la lb [2∂c (lc A) − 4A2 ]

− 2m2 la lb [3A2 + ∂d lc ∂c ld − ∂d lc ∂d lc ]
= −2m2 la lb [2∂c (lc A) − A2 + ∂d lc ∂c ld − ∂d lc ∂d lc ] (6.47)

we see the m2 order equations vanish when

la lb [2∂c (lc A) − A2 + ∂d lc ∂c ld − ∂d lc ∂d lc ] = 0. (6.48)

which implies the scalar equation

2∂c (lc A) − A2 + ∂d lc ∂c ld − ∂d lc ∂d lc = 0. (6.49)

Let us consider the third term, using ld ∂d lc = −Alc and L = −∂c lc

∂d lc ∂c ld = ∂c (ld ∂d lc ) − ld ∂c ∂d lc
= ∂c (−Alc ) + ld ∂d L
= ∂c [(L − A)lc ] + L2 (6.50)

55
Similarly we find for the fourth term using lc ∂d lc = 0

∂d lc ∂d lc = ∂d (lc ∂d lc ) − lc ∂d ∂d lc
= −lc ∂d ∂d lc (6.51)

Expanding (6.39) gives

−la ∂d ∂d lb − lb ∂d ∂d la − 2(∂d la )(∂d lb )

= lb ∂a (L + A) + la ∂b (L + A) + (L + A)(∂b la + ∂a lb ). (6.52)

Contracting with la gives

−lb la ∂d ∂d la = la lb ∂a (L + A) + (L + A)la ∂a lb
= lb la ∂a (L + A) − (L + A)Alb (6.53)

(where we have used la la = 0 as well as la ∂b la = 0 and la ∂a lb = −Alb ). As lb is a common factor

we get

−la ∂d ∂d la = la ∂a (L + A) − (L + A)A
= ∂a [(L + A)la ] − ∂a la (L + A) − (L + A)A
= ∂a [(L + A)la ] + L2 − A2 (6.54)

(where we have used the definition L := −∂a la ). By using (6.51) the left hand side of this
equation can be replaced by ∂d lc ∂d lc . Thus the fourth term of (6.49) can be written

∂d lc ∂d lc = L2 − A2 + ∂a [(L + A)la ]. (6.55)

We now sustitute (6.50) and (6.55) into the LHS of the scalar equation (6.49), to obtain

2∂c (lc A) − A2 + ∂d lc ∂c ld − ∂d lc ∂d lc =
2∂c (lc A) − A2 + ∂a [(L − A)la ] + L2 − ∂a [(L + A)la ] + L2 − A2

2∂c (lc A) − 2∂c (lc A)

0 (6.56)

Thus any solution of (6.39) makes the m2 equations vanish identically.

56
6.1.6 Stationary field equations from the m order equation

We have found that the full content of the field equations is embodied in

−2 (la lb ) = ∂b [(L + A)la ] + ∂a [(L + A)lb ]. (6.57)

We now consider the stationary, or time-independent, case. We will find a simlpification to the
algebraic manipulations will occur with the introduction of the three-vector λ j via

la = l0 (1, λ1, λ2 , λ3 ) (6.58)

Since la is a flat-space null vector (la lb ηab = 0), λ j is a flat-space unit vector,

~λ2 = 1.

For a = b = 0, in the time-independent case we are now considering, (6.57) reduces to

∇2 (l20 ) = 0 (6.59)

For a = 0, b = j , 0 (6.57) reduces to

∇2 (l20 λ j ) = ∂ j [(L + A)l0 ] (6.60)

For a = i , 0, b = j , 0 (6.57) reduces to

∇2 (l20 λi λ j ) = ∂ j [(L + A)l0 λi ] + ∂i [(L + A)l0 λ j ] (6.61)

We can take (6.61) and simplify it to a first-order diferential equation by using (6.60) and (6.59).
First expand the RHS of (6.61) and then use (6.60),

∇2 (l20 λi λ j ) = ∂ j [(L + A)l0 λi ] + ∂i [(L + A)l0 λ j ]

= λi ∂ j [(L + A)l0 ] + λ j ∂i [(L + A)l0 ] + (L + A)l0 (∂ j λi + ∂i λ j )
= λi ∇2 (l20 λ j ) + λ j ∇2 (l20 λi ) + (L + A)l0 (∂ j λi + ∂i λ j ) (6.62)

Rearanging gives

1
∂ j λi + ∂i λ j = [∇2 (l20 λi λ j ) − λi ∇2 (l20 λ j ) − λ j ∇2 (l20 λi )] (6.63)
(L + A)l0

57
Let us expand the derivatives in ∇2 (l20 λi λ j ):

∇2 (l20 λi λ j ) = ∂k ∂k (l20 λi λ j )
= ∂k [λi λ j ∂k (l20 ) + l20 λ j ∂k λi + l20 λi ∂k λ j ]
= λi λ j ∇2 (l20 ) + λ j ∂k (l20 )(∂k λi ) + ∂k (l20 )λi ∂k λ j
+ ∂k (l20 )(∂k λi )λ j + l20 (∂k ∂k λi )λ j + l20 (∂k λi )(∂k λ j )
+ ∂k (l20 )λi ∂k λ j + l20 (∂k λi )(∂k λ j ) + l20 λi ∂k ∂k λ j

= λi λ j ∇2 (l20 ) + 2(∂k (l20 ))(λ j ∂k λi + λi ∂k λ j )

+ l20 (λ j ∇2 λi + λi ∇2 λ j ) + 2l20 (∂k λi )(∂k λ j ). (6.64)

Let us expand the derivatives in λi ∇2 (l20 λ j ):

λi ∇2 (l20 λ j ) = λi ∂k ∂k (l20 λ j )
= λi ∂k (∂k l20 )λ j + l20 ∂k λ j

= λi λ j ∇2 (l20 ) + 2(∂k (l20 ))(λi ∂k λ j ) + l20 λi ∇2 λ j . (6.65)

With these results, let us expand the content of the square brackets on the RHS of (6.63)

∇2 (l20 λi λ j ) − λi ∇2 (l20 λ j ) − λ j ∇2 (l20 λi )

= (∇2 (l20 ))λi λ j + 2∂k (l20 )(λ j ∂k λi + λi ∂k λ j )
+ l20 (λ j ∇2 λi + λi ∇2 λ j ) + 2l20(∂k λi )(∂k λ j )
− λi λ j ∇2 (l20 ) − 2(∂k l20 )λi ∂k λ j − l20 λi ∇2 λ j
− λ j λi ∇2 (l20 ) − 2(∂k l20 )λ j ∂k λi − l20 λ j ∇2 λi
= −(∇2 (l20 ))λi λ j + 2l20 (∂k λi )(∂k λ j )
= 2l20 (∂k λi )(∂k λ j ) (6.66)

where we used (6.59) in the last line. We get

2l0
∂ j λi + ∂i λ j = (∂k λi )(∂k λ j ) (6.67)
L+A

Define

L+A
p := (6.68)
2l0

58
then

1
∂ j λi + ∂i λ j = (∂k λi )(∂k λ j ) (6.69)
p

The gravitational field is now described by (6.59), (6.60), and (6.69).

Let

Mik := ∂k λi (6.70)

Then (6.69) becomes

1
Mi j + M ji = Mik (M T )k j (6.71)
p

or

1
M + MT = MM T (6.72)
p

The constant length of λ j , implies

1
∂k (λi λi ) = λi ∂k λi = 0 (6.73)
2
or

M T λ = 0. (6.74)

That is ~λ is in the null space of M T . Moreover, lc ∂c lb = −Alb with b = 0 gives

λ j ∂ j l0 = A (6.75)

and with b = k , 0

λ j ∂ j (l0 lk ) = Aλk (6.76)

where we used the definition of λ given in (6.58), but

59
 λ j ∂ j (l0 λk ) = l0 (λ j ∂ j λk ) + λk λ j ∂ j l0
= l0 (λ j ∂ j λk ) + Aλk (6.77)

where we used (6.75). By comparing (6.76) and (6.77) we see that

λ j ∂ j λk = 0 (6.78)

or

Mλ = 0. (6.79)

We will now be able to solve (6.72), (6.74), and (6.79) for M as a function of ~λ.

Consider a rotation such that

Rλ = λ′ (6.80)

where

 
 1 
′
λ =  0  . (6.81)
 
0
 

If ~λ is in the null space of M and M T , then ~λ′ is in the null space of M ′ and M ′T , where

M ′ = RMRT , M ′T = RM T RT . (6.82)

Now a rotation matrix satisfies RRT = I. Let us write this out in component form,

    
 R11 R12 R13   R11 R21 R31   1 0 0 
RRT =  R21 R22 R23   R12 R22 R32  =  0 1 0 
     
R31 R32 R33 R13 R23 R33 0 0 1


Writing out some of the terms

R11 R11 + R12 R12 + R13 R13 = 1

R21 R11 + R22 R12 + R23 R13 = 0
.. .. .. .
. + . + . = ..

60
Generally:

Ri1 Rk1 + Ri2 Rk2 + Ri3 Rk3 = δik (6.83)

In other words the three rows constitue three orthonormal vectors in 3 spacial dimensions.
Define

~ 1 = (R11 , R12 , R13 )

R
~ 2 = (R21 , R22 , R23 )
R
~ 3 = (R31 , R32 , R33 )
R

We are considering rotation such that

    
 R11 R12 R13   λ1   1 
 R21 R22 R23   λ2  =  0  (6.84)
     
R31 R32 R33 λ3 0

Notice that you could exchange the second and third rows of R without changing (6.84). Let us
choose the ordering such that:

~2 × R
R ~3 = R
~ 1. (6.85)

We will use this later. From that

 
 1 
′
λ =  0 
 
0
 

and that M ′ λ′ = 0 and M ′T λ′ = 0 we see that M ′ must have the form

 
 0 a b 
′ ′ ′
M =  c N11 N12 (6.86)
 

′ ′
d N21 N22


The values a, b, c, d can be determined from the property of M ′ ,

1 ′ ′T
M ′ + M ′T = MM (6.87)
p

61
We have first

 
 0 a+c b+d 
M ′ + M ′T =  a + c · · (6.88)
 

b+d · ·


but then

  
0 a b   0 c d 
1 ′ ′T 1 

MM = c · ·   a · · 
  
p p 
d · · b · ·
 

a2 + b2 · · 
 
1 

= · · ·  (6.89)
 
p 
· · ·


which implies a2 + b2 = 0 or a = b = 0. Now consider 1p M ′ M ′T again:

  
0 0 0   0 c d 
1 ′ ′T 1 

MM = c · ·   0 · · 
  
p p 
d · · 0 · ·
 
 
0 . . 
1 

= 0 .  (6.90)
 
p 
0


implying c = d = 0. So M ′ must be of the form

 
 0 0 0 
′ ′ ′
M =  0 N11 N21 (6.91)
 

′ ′
0 N12 N12


The matrix N ′ satisfies the same equation, (6.87), as M ′ :

T 1 ′ ′T
N′ + N′ = NN (6.92)
p

Write

1 ′
U =I− N
p

62
then the relation (6.92) becomes the following relation for U,

! !
T 1 ′ 1 ′T
UU = I− N I− N
p p
!
1 ′ ′T 1 ′ ′T
= I− N +N − N N
p p
= I (6.93)

Put

!
a b
U= (6.94)
c d

Then UU T = I reads

! ! !
a b a c a2 + b2 ac + bd
=
c d b d ac + bd c2 + d 2
!
1 0
= (6.95)
0 1

a2 + b2 = 1
c2 + d 2 = 1
ac + bd = 0 (6.96)

Now we consider what the result is of taking product U T U:

! ! !
a c a b a2 + c2 ab + cd
= (6.97)
b d c d ab + cd b2 + d 2

From

det Udet U T = 1

we have

63
 det U = ∓1

or

ad − bc = ∓1

multiplying this by c

acd − bc2 = ∓c

and using ac = −bd gives

−bd 2 − bc2 = ∓c

or

−b(c2 + d 2 ) = ∓c

and as c2 + d 2 = 1 we therefore have

b = ±c. (6.98)

Using this in (6.97) we have

a2 + c2 = a2 + b2 = 1
b2 + d 2 = c2 + d 2 = 1
ab + cd = a(±c) + (±b)d
= ±(ac + bd) = 0 (6.99)

therefore

U T U = I.

Now UU T = U T U = I is the well known condition for unitary matrices. Summarising

1 ′
U=I− N, UU T = U T U = I (6.100)
p

64
Since N ′ is a 2 × 2 real matrix, it is therefore either a proper rotation or a an improper rotation,
that is, a rotation plus inversion. Thus it may be written

! !
cos θ − sin θ cos θ − sin θ
U= or . (6.101)
sin θ cos θ − sin θ cos θ

The first case corresponds to detU = +1 (a proper rotation) and the second detU = −1 (an
improper rotation). The first case leads to interesting results. For N ′ and M ′ we have

!
′ 1 − cos θ sin θ
N =p (6.102)
− sin θ 1 − cos θ

 
 0 0 0 
M ′ = p  0 1 − cos θ sin θ (6.103)
 

0 − sin θ 1 − cos θ


We now need to rotate back to the originl coordinates to get

M = RT M ′ R.

The simple form of M ′ allows us to write

Mik = Rli R jk Ml′j

′ ′
= R2i R2k M22 + R3i R3k M33
′ ′
+ R2i R3k M23 + R3i R2k M32
= p(1 − cos θ)(R2i R2k + R3i R3k ) + p sin θ(R2i R3k − R3i R2k ) (6.104)

Let us look at properties of the rotation matrix

    
 R11 R21 R31   R11 R12 R13   1 0 0 
RT R =  R12 R22 R32   R21 R22 R23  =  0 1 0 
   
 
R13 R23 R33 R31 R32 R33 0 0 1


Writing out some terms

R11 R11 + R21 R21 + R31 R31 = 1

R12 R11 + R22 R21 + R32 R31 = 0
.. .. .. .
. + . + . = ..

65
Or generaly:

R1i R1k + R2i R2k + R3i R3k = δik . (6.105)

Now take (6.85), which can be written in component form

ǫlmn R2m R3n = R1l ,

contracting with ǫi jl gives

ǫi jl ǫlmn R2m R3n = (δim δ jn − δin δ jm )R2m R3n

= R2i R3 j − R3i R2 j (6.106)

meaning

R2i R3 j − R3i R2 j = ǫi jl R1l (6.107)

Let us write R1i = Ri . Using the above results in (6.104) gives

Mik = p(1 − cos θ)(δik − Ri Rk ) + p sin θǫikl Rl . (6.108)

Now we had

    
 R11 R12 R13   λ1   1 
 · · ·   λ2  =  0  (6.109)
     
· · · λ3 0

implying

R11 λ1 + R12 λ2 + R13 λ3 = 1

or

R1i λi = 1

Using the notation R1i = Ri , this becomes

66
 ~ · ~λ = 1,
R (6.110)

~ denotes the vector with components Ri . As both R

where R ~ and ~λ are unit vectors,

~ · ~λ = cos ϕ = 1.
R (6.111)

implies ϕ = 0 and hence

~ = ~λ
R (6.112)

or Ri = λi . Then we have our end result

Mik = p(1 − cos θ)(δik − λi λk ) + p sin θǫikl λl . (6.113)

We have now replaced the non-linear implicit relation (6.69) by the above simple explicit ex-
pression for ∂k λi .

6.1.7 Laplace and eikonal equations

We rewrite (6.113) in terms of new parameters α and β

∂k λi = α(δik − λi λk ) + βǫikl λl . (6.114)

It will turn out that α and β determine the metric. A number of important three-vector realtion
follow directly from (6.114). Set i = k and sum

∂i λi = α(δii − λi λi ) + βǫiil λl
= α(3 − 1) + 0

giving

∇ · λ = 2α. (6.115)

Multiplying (6.114) by ǫ jki and summing over i and k gives

67
 ǫ jki ∂k λi = α(δik − λi λk )ǫ jki + βǫikl ǫ jki λl
= βǫikl ǫ jki λl
= −2βλ j

or

∇ × λ = −2βλ. (6.116)

The Laplacian of λ can be obtaind in two ways. First differentiating by (6.114) with respect to
xk gives

∇2 λi = ∂k [α(δik − λi λk ) + βǫikl λl ]
= ∂i α − λi (λk ∂k α) − αλk ∂k λi − αλi ∂k λk + ǫikl λl ∂k β + βǫikl ∂k λl
= ∂i α − λi (∇α · λ) − αλi (∇ · λ) + β(∇ × λ)i + (∇β × λ)i
= ∂i α − λi (∇α · λ) − 2(α2 + β2 )λi + (∇β × λ)i

where we used (6.115), (6.116), and

λk ∂k λi = α(λi − λi λ · λ) + βǫikl λk λl
= 0
(6.117)

(which is just Mλ = 0). We have obtained our first equation for ∇2 λ:

∇2 λ = ∇α − λ(∇α · λ) − 2(α2 + β2 )λ + ∇β × λ (6.118)

To obtain the alternative equation we first derive the vector identity

∇ × (∇ × λ) = ∇(∇ · λ) − ∇2 λ

from

[∇ × (∇ × λ)]m = ǫmni ∂n (ǫikl ∂k λl )

= ǫmni ǫikl ∂n ∂k λl
= (δmk δnl − δml δnk )∂n ∂k λl
= ∂m (∂n λn ) − ∂n ∂n λm .

68
We then have

∇2 λ = ∇(∇ · λ) − ∇ × (∇ × λ)
= ∇(2α) − ∇ × (−2βλ)
= 2∇α + 2(∇β × λ) + 2β∇ × λ
= 2∇α + 2(∇β × λ) − 4β2 λ (6.119)

where again we used (6.115) and (6.116). We equate the expressions (6.118) and (6.119)

∇α − λ(∇α · λ) − 2(α2 + β2 )λ + ∇β × λ
= 2∇α + 2(∇β × λ) − 4β2 λ

from which we obtain

∇α = −∇β × λ − λ(∇α · λ) − 2(α2 − β2 )λ. (6.120)

Dotting this with λ gives

∇α · λ = −(∇α · λ) − 2(α2 − β2 )

and rearanging we have

∇α · λ = β2 − α2 (6.121)

and substituting this into (6.120) gives

∇α = (β2 − α2 )λ − ∇β × λ (6.122)

Equations (6.121) and (6.122) are very important.

Equations analogous to (6.121) and (6.122) with β replacing α on the left hand side can be
obtained. The divergence of (6.116) is zero, implying the divergence of βλ is zero,

∇ · (βλ) = β∇ · λ + ∇β · λ = 0 (6.123)

using (6.115) in this gives

69
 ∇β · λ = −2αβ (6.124)

Now cross (6.122) with λ and use (∇β × λ) × λ = −λ · λ∇β + λ(λ · ∇β). We obtain

∇β = λ(λ · ∇β) + (∇α × λ) (6.125)

which usong (6.124) becomes

∇β = −2αβλ + (∇α × λ). (6.126)

6.1.8 Introduce the complex function γ

Eqations (6.121), (6.122), (6.124), and (6.126) are very important and can be expressed in a
more concise way by the introduction of the complex function γ = α + iβ:

∇γ · λ = ∇α · λ + i∇β · λ
= [β2 − α2 ] + i[−2αβ]
= −(α + iβ)2
= −γ2 (6.127)

∇γ = [(β2 − α2 )λ − ∇β × λ] + i[−2αβλ + (∇α × λ)]

= [(β2 − α2 ) + i(−2αβ)]λ + i[∇α + i∇β] × λ
= −γ2 λ + i(∇γ × λ). (6.128)

We will obtain a pair of simple differential equations that determine γ and show in that γ in turn
determines the metric via l0 and λ j .

The first differential equation is found by forming the Laplacian for γ from (6.128) and using
(6.115), (6.127), and (6.116) we obtain

70
 ∇2 γ = ∇ · ∇γ
= ∇ · [−γ2 λ + i(∇γ × λ)]
= −γ2 ∇ · λ − 2γ∇γ · λ + i∇ · (∇γ × λ)
= −2αγ2 + 2γ3 + i∂i [ǫi jk (∂ j γ)λk ]
= −2αγ2 + 2γ3 − i∂i γ(ǫi jk ∂ j λk )
= −2αγ2 + 2γ3 − i∇γ · (∇ × λ)
= −2αγ2 + 2γ3 − i∇γ · (−2βλ)
= −2αγ2 + 2γ3 − 2iβγ2
= −2γ2 (α + iβ − γ)
= 0 (6.129)

Thus γ is a complex harmonic function. The second differential equation is obtained by squaring
(6.128) and using (6.127)

(∇γ)2 = [−γ2 λ + i(∇γ × λ)]2

= γ4 − (∇γ × λ) · (∇γ × λ)
= γ4 + i(∇γ × λ) · (∇γ + γ2 λ)
= γ4 (6.130)

we get

∇2 γ = 0, (∇γ)2 = γ4 (6.131)

Let us introduce ω ≡ 1/γ, then

!
1 1
∇ω = ∇ = − 2 ∇γ (6.132)
γ γ

Then the equation (∇γ)2 = γ4 becomes

1
(∇ω)2 = 4
(∇γ)2 = 1.
γ

where we used (6.130). Repeating (6.129), altogether we have

1
∇2 γ = 0, (∇ω)2 = 1, ω≡ (6.133)
γ

71
These equations determine the function γ completely, dependent on consistent boundary condi-
tions.

As we shall now show, these equations completely replace the field equations since the metric
functions l0 and λ are determined by γ.

6.1.9 Expression for λ

It is more convenient to express (6.127) and (6.128) in terms of ω. For equation (6.127) we
have

!
1 λ · ∇ω 1
∇ ·λ=− 2
=− 2 (6.134)
ω ω ω

where we have used (6.132) and (6.127). For equation (6.128) we have

! !
1 1 1
∇ = − 2 λ + i(∇ × λ) (6.135)
ω ω ω

or

1 1 i
− 2
∇ω = − 2 λ − 2 (∇ω × λ)
ω ω ω

Substituting (6.135) into (6.134) gives

λ · ∇ω = 1.

Altogether we have

λ · ∇ω = λ · ∇ω∗ = 1, ∇ω = λ + i(∇ω × λ) (6.136)

Thus

∇ω × ∇ω∗ = [λ + i(∇ω × λ)] × [λ − i(∇ω∗ × λ)]

= −iλ × (∇ω∗ × λ) + i(∇ω × λ) × λ + (∇ω × λ) × (∇ω∗ × λ)
= −i[∇ω∗ − λ(λ · ∇ω∗ )] − i[∇ω − λ(λ · ∇ω)] + (∇ω × λ) × (∇ω∗ × λ)
= −i(∇ω∗ − λ) − i(∇ω − λ) + (∇ω × λ) × (∇ω∗ × λ)
= −i[∇ω∗ + ∇ω] + 2iλ + (∇ω × λ) × (∇ω∗ × λ) (6.137)

72
where we used the vector identity λ×(∇ f ×λ) = λ·λ∇ f −λ(λ·∇ f ) and that λ·∇ω = λ·∇ω∗ = 1.

Consider the last term,

[(∇ω × λ) × (∇ω∗ × λ)] p = ǫ pil (∇ω × λ)i (∇ω∗ × λ)l

= ǫ pil (ǫi jk λk ∂ j ω)(ǫlmn λn ∂m ω∗ )
= ǫ pil ǫlmn (ǫi jk λk λn ∂ j ω∂m ω∗ )
= (δ pm δin − δ pn δim )(ǫi jk λk λn ∂ j ω∂m ω∗ )
= ǫi jk λk λi ∂ j ω∂ p ω∗ − λ p ǫi jk λk ∂ j ω∂i ω∗
= λ p ǫi jk λi ∂ j ω∂k ω∗
= λ p [λ · (∇ω × ∇ω∗ )] (6.138)

So

∇ω × ∇ω∗ = −i[∇ω∗ + ∇ω] + 2iλ + λ[λ · (∇ω × ∇ω∗ )]

= −i[∇ω∗ + ∇ω] + Bλ (6.139)

Now as

(∂i ω)εi jk ∂ j ω∂k ω∗ = 0

or

∇ω · (∇ω × ∇ω∗ ) = 0

dotting (6.139) with ∇ω and using (∇ω)2 = 1 and λ · ∇ω = 1 we obtain for B

B = i[1 + ∇ω · ∇ω∗ ] (6.140)

So we get for λ

∇ω + ∇ω∗ − i(∇ω × ∇ω∗ )

λ= (6.141)
1 + ∇ω · ∇ω∗

and we have our expression for λ j in terms of ω.

73
6.1.10 Expression for l0

We prove that the choice

l20 = Re (γ) = α. (6.142)

leads to a solution of the first two field equations (6.59) and (6.60) as we will now show. We
have first from the first equation in (6.133) that

0 = ∇2 α = ∇2 (l20 ). (6.143)

which solves (6.59). Next consider

∇2 (αλ j ) = α∇2 λ j + 2(∂k α)(∂k λ j ) (6.144)

By using (6.119), (6.114), and (6.121) we get

∇2 (αλ j ) = α∇2 λ j + 2(∂k α)(∂k λ j )

= α[2∂ j α + 2(∇β × λ) j − 4β2 λ j ]
+ 2(∂k α)[α(δ jk − λ j λk ) + βǫ jkl λl ]
= 4α∂ j α − 4αβ2 λ j − 2λ j α(λ · ∇α) + 2α(∇β × λ) j + 2β(∇α × λ) j
= 4α∂ j α − 4αβ2 λ j − 2λ j α(β2 − α2 ) + 2α(∇β × λ) j + 2β(∇α × λ) j
= 4α∂ j α + 2α(α2 − 3β2 )λ j + 2α(∇β × λ) j + 2β(∇α × λ) j (6.145)

We simplify further using (6.122) and (6.126),

∇2 (αλ j ) = 4α∂ j α + 2α(α2 − 3β2)λ j + 2α[(β2 − α2 )λ j − ∂ j α]

+ 2β[2αβλ j + ∂ j β]
= 2α∂ j α + 2β∂ j β
= ∂ j (α2 + β2 ). (6.146)

Or

∇2 (αλ) = ∇(α2 + β2 ). (6.147)

We calculate the RHS of (6.147). From α = p(1 − cos θ) and β = p sin θ we have

74
 α2 + β2 = 2p2 (1 − cos θ) = 2αp (6.148)

From the definition of p,

L+A
p= . (6.149)
2l0

We then have

l0 2
L+A= (α + β2 ). (6.150)
α

Thus with l20 = α the RHS of (6.147) is

∂ j [(L + A)l0 ] = ∂ j (α2 + β2 ). (6.151)

We see that l20 = α is indeed a solution.

6.2 The Kerr Solution

6.2.1 A solution to the field equations

Consider the function

γ = [(x − a)2 + (y − b)2 + (z − c)2 ]−1/2 (6.152)

We will calculate ∇2 γ and (∇ω)2 , (ω = 1/γ). First

∂
[(x − a)2 + (y − b)2 + (z − c)2 ]−1/2 = −(x − a)[(x − a)2 + (y − b)2 + (z − c)2 ]−3/2
∂x

and

∂2
[(x − a)2 + (y − b)2 + (z − c)2 ]−1/2
∂x2
= −[(x − a)2 + (y − b)2 + (z − c)2 ]−3/2
+ 3(x − a)2 [(x − a)2 + (y − b)2 + (z − c)2 ]−5/2 (6.153)

75
so that

∇2 γ = ∇2 [(x − a)2 + (y − b)2 + (z − c)2 ]−1/2

= −3[(x − a)2 + (y − b)2 + (z − c)2 ]−3/2
+ 3[(x − a)2 + (y − b)2 + (z − c)2 ][(x − a)2 + (y − b)2 + (z − c)2 ]−5/2
= 0. (6.154)

Now consider ∇ω

∇ω = ∇[(x − a)2 + (y − b)2 + (z − c)2 ]1/2

[(x − a)î + (y − b) ĵ + (z − c)k̂]
= (6.155)
[(x − a)2 + (y − b)2 + (z − c)2 ]1/2

so obviously we have

(∇ω)2 = 1. (6.156)

Therefore (6.152) is a solution of the field equations (6.133).

6.2.2 The Schwarzschild solution

Consider the choice

γ = (x2 + y2 + z2 )−1/2 (6.157)

We obviously have

1
l20 = . (6.158)
r

Next as ω = 1/γ, ω = r and

∇ω = ∇(x2 + y2 + z2 )1/2
xî + y ĵ + zî
=
(x + y2 + z2 )1/2
2
x y z
= î + ĵ + î (6.159)
r r r
76
Obviously

∇ω × ∇ω∗ = ∇ω × ∇ω = 0 (6.160)

Next

∇ω · ∇ω∗ = ∇ω · ∇ω = 1 (6.161)

Recall λ is given by

∇ω + ∇ω∗ − i(∇ω × ∇ω∗ )

λ =
1 + ∇ω · ∇ω∗
= ∇ω
x y z
= î + î + î (6.162)
r r r

Reading off the components of λ we get

x y z
λ1 = , λ2 = , λ3 = . (6.163)
r r r

Combining this with (6.158) gives for la

1 x y z
 
la = l0 (1, λ1 , λ2, λ3 ) = √ 1, , , (6.164)
r r r r

in agreement with (6.3).

6.2.3 The Kerr solution

Consider the complex valued choice,

γ = (x2 + y2 + (z − ia)2 )−1/2 (6.165)

Let us calculate the functions l0 and λ. We first split ω into real and imaginary parts

ω = 1/γ = ρ + iσ = (x2 + y2 + (z − ia)2 )1/2 = (r2 − a2 − 2iaz)1/2 (6.166)

Squaring and equating real and imaginary parts gives

77
 az
ρ 2 − σ 2 = r 2 − a2 , σ=− (6.167)
ρ

substituting for σ in the real part gives

a2 z 2
ρ2 − = r 2 − a2
ρ2

resulting in the quadratic equation in ρ2 :

ρ4 − ρ2 (r2 − a2 ) − a2 z2 = 0. (6.168)

Completeing the square we obatain

#1/2
r 2 − a2
" 2
2 (r − a2 )2 2 2
ρ = + +a z (6.169)
2 4

where we have choosen the plus sign so that for r ≫ a, ρ ∼ r. Now l20 = Re (γ). From

1 ρ iσ
γ= = 2 2
− 2
ρ + iσ ρ + σ ρ + σ2

we have

ρ ρ3
l20 = = . (6.170)
ρ 2 + σ 2 ρ 4 + a2 z 2

Now let us move onto the calculation of λ. First

!
1
∇ω = ∇
γ
= ∇(r2 − 2iaz − a2 )1/2
1
= (2xî + 2y ĵ + 2zk̂ − 2iak̂)
2ω
r̂ − iak̂
= (6.171)
ω

then

78
 r̂ − iak̂ r̂ + iak̂
∇ω + ∇ω∗ = +
ω ω∗
ω∗ (r̂ − iak̂) + ω(r̂ + iak̂)
=
|ω|2
(ρ − iσ)(r̂ − iak̂) + (ρ + iσ)(r̂ + iak̂)
=
|ω|2
ρr̂ − aσk̂
= 2 (6.172)
|ω|2

and

r̂ − iak̂ r̂ + iak̂ r2 + a2
∇ω · ∇ω∗ = · =
ω ω∗ |ω|2

and

r̂ − iak̂ r̂ + iak̂ r×k

∇ω × ∇ω∗ = × ∗
= 2ia 2
ω ω |ω|

so λ is

∇ω + ∇ω∗ − i(∇ω × ∇ω∗ )

λ =
1 + ∇ω · ∇ω∗
2[ρr̂ − aσk̂ + a(r × k)]
= . (6.173)
|ω|2 + r2 + a2

We can simplify this by noting

|ω|2 = [(r2 − a2 )2 + 4a2 z2 ]1/2

(r2 − a2 )2
= 2[ + a2 z2 ]1/2
4
= 2ρ2 − (r2 − a2 )

because then

|ω|2 + r2 + a2 = 2(ρ2 + a2 ).

79
Substituting this into (6.173) we then have for λ

[ρr̂ − aσk̂ + a(r × k)]

λ= (6.174)
ρ 2 + a2

As

r × k = (xî + yĵ) × k̂ = yî − xĵ

and as σ = −az/ρ, λ is

a2 z
" #
1
λ = 2 ρr̂ + k̂ + a(yî − xĵ)
ρ + a2 ρ
(ρ2 + a2 )z
" #
1
= 2 (ρx + ay)î + (ρy − ax) ĵ + k̂ . (6.175)
ρ + a2 ρ

We can read off the components

ρx + ay
λ1 = ,
a2 + ρ 2

ρy − ax
λ2 = ,
a2 + ρ 2

and

z
λ3 = .
ρ

Then

" #
µ ρx + ay
0 ρy − ax z
lµ dx = l0 dx + 2 dx + 2 dy + dz
a + ρ2 a + ρ2 ρ
" #
0 ρ a z
= l0 dx + 2 (xdx + ydy) + 2 (ydx − xdy) + dz
a + ρ2 a + ρ2 ρ

Finally, as

80
 ds2 = ηµν dxµ dxν − 2mlµ lν dxµ dxν

we have

2mρ3
"
2 0 2 2 ρx
ds = (dx ) − (dx) − 4 dx0 + 2 (xdx + ydy)
ρ +a z
2 2 a + ρ2
#2
a z
+ 2 2
(ydx − xdy) + dz (6.176)
a +ρ ρ

This is the form obtained by Kerr in 1963.

6.3 Independence of Metric on Angular Variable ϕ

We now make axially symmetry of this solution manifest. Consider

u = x0 + u
x = ρ sin θ cos ϕ − aρ sin θ sin ϕ
y = ρ sin θ cos ϕ + aρ sin θ sin ϕ
z = ρ cos θ. (6.177)

The middle two transformations can be written as a complex expression

(ρ + ia)eiϕ sin θ = x + iy (6.178)

Now let us find the differential dz2

dz2 = (cos θdρ − ρ sin θdθ)2 (6.179)

Now let us find the differential dx2 + dy2

dx2 + dy2 = |d(x + iy)|2

= |d[(ρ + ia)eiϕ sin θ]|2
= |eiϕ sin θdρ + i(ρ + ia)eiϕ sin θdϕ + (ρ + ia)eiϕ cos θdθ|2
= | sin θdρ + i(ρ + ia) sin θdϕ + (ρ + ia) cos θdθ|2
= | sin θdρ − a sin θdϕ + ρ cos θdθ + i[ρ sin θdϕ + a cos θdθ]|2
= (sin θdρ − a sin θdϕ + ρ cos θdθ)2 + (ρ sin θdϕ + a cos θdθ)2 (6.180)

81
Now let us find the differential xdx + ydy

1
xdx + ydy = d|x + iy|2
2
1
= d|(ρ + ia)eiϕ sin θ|2
2
1
= d|(ρ + ia) sin θ|2
2
1
= d[(ρ2 + a2 ) sin2 θ]
2
= (ρ2 + a2 ) sin θ cos θdθ + ρ sin2 θdρ (6.181)

Now let us find the differential xdy − ydx

xdy − ydx = −Im[(x + iy)d(x − iy)]

= −Im{[(ρ + ia)eiϕ sin θ]d[(ρ − ia)e−iϕ sin θ]}
= −Im{[(ρ + ia)eiϕ sin θ][e−iϕ sin θdρ − i(ρ − ia)e−iϕ sin θdϕ +
+ (ρ − ia)eiϕ cos θdθ]}
= −Im{(ρ + ia) sin2 θdρ − i(ρ2 + a2 ) sin2 θdϕ + (ρ2 + a2 ) sin θ cos θdθ}
= (ρ2 + a2 ) sin θ2 dϕ − a sin2 θdρ (6.182)

Now let us find the differential zdz

zdz = ρ cos2 θdρ − ρ2 sin θ cos θdθ (6.183)

2mρ3 2mρ
= (6.184)
ρ4 + a2 z2 ρ2 + a2 cos2 θ

Substituting these results into (6.176)

ds2 = (du − dρ)2 − (sin θdρ − a sin θdϕ + ρ cos θdθ)2 − (ρ sin θdϕ + a cos θdθ)2
−(cos θdρ − ρ sin θdθ)2
2mρ h ρ n 2 2 2
o
− 2 du − dρ + (ρ + a ) sin θ cos θdθ + ρ sin θdρ
ρ + a2 cos2 θ a2 + ρ 2
a n 2 2 2 2
o
2
i2
− 2 (ρ + a ) sin θdϕ − a sin θdρ + cos θdρ − ρ sin θ cos θdθ
a + ρ2
(6.185)

82
We multiply out the first and second line and get

du2 + (1 − sin2 θ − cos2 )θdρ2 − [(ρ2 + a2 ) cos2 θ + ρ2 sin2 θ]dθ2 − (ρ2 + a2 ) sin2 θdϕ2
− 2dudρ − (2ρ sin θ cos θ − 2ρ sin θ cos θ)dρdθ + 2a sin2 θdρdϕ
+ (aρ sin θ cos θ − ρa sin θ cos θ)dθdϕ
= du2 − (ρ2 + a2 cos2 θ)dθ2 − (ρ2 + a2 ) sin2 θdϕ2 − 2dudρ + 2a sin2 θdρdϕ (6.186)

The last two lines can be simplified,

ρ2 a2
!
2mρ h
2 2 2
− 2 du + −1 + 2 sin θ + 2 sin θ + cos θ dρ
ρ + a2 cos2 θ a + ρ2 a + ρ2
  i2
+ (ρ sin θ cos θ − ρ sin θ cos θ) dθ + −a sin2 θ dϕ
ρ2 a2
!
2mρ h
2 2
i2
= − 2 du + + − 1 sin θdρ − a sin θdϕ
ρ + a2 cos2 θ a2 + ρ 2 a2 + ρ 2
2mρ h
2
i2
= − 2 du − a sin θdϕ (6.187)
ρ + a2 cos2 θ

Putting together (6.186) and (6.187) we get

ds2 = du2 − (ρ2 + a2 cos2 θ)dθ2 − (ρ2 + a2 ) sin2 θdϕ2 − 2dudρ + 2a sin2 θdρdϕ
2mρ h i2
− 2 2 2
du − a sin2 θdϕ . (6.188)
ρ + a cos θ

Simplifying, we finally obtain the line element

!
2mρ 2mρ
ds 2
= 1− 2 2 2
du2 − 2dudρ + 2 2 2
(2a sin2 θ)dudϕ
ρ + a cos θ ρ + a cos θ
!
2 2 2 2 2 2 2 2 2mρ 4
+ 2a sin θdρdϕ − (ρ + a cos θ)dθ − (ρ + a ) sin θ + 2 (a sin θ) dϕ2 .
2
ρ + a2 cos2 θ
(6.189)

There is no dependence of the line element on the angular coordinate ϕ, so that the solution
(6.189) is manifestly axially symmetric. This is the advanced Eddington-Finkelstein form of
Kerr’s solution.

83
6.4 Boyer-Lindquist Coordinates

We wish to make a change of coordinates that puts the line element in a form suh that the only
cross-term is dϕdt.

write the element (6.189) as

ds2 = g00 du2 + g22 dθ2 + g33 dϕ2 + 2g03 dudϕ + 2g01 dudρ + 2g13 dρdϕ (6.190)

We guess the form of the desired transformation

t = u − A(ρ) du = dt + A′ dρ
ϕ̃ = ϕ − B(ρ) dϕ = d ϕ̃ + B′dρ (6.191)

(6.190) is then

ds2 = g00 (dt + A′ dρ)2 + g22 dθ2 + g33 (d ϕ̃ + B′ dρ)2 + 2g03 (dt + A′ dρ)(d ϕ̃ + B′dρ)
+ 2g01 (dt + A′ dρ)dρ + 2g13 dρ(d ϕ̃ + B′dρ)
′ ′
= g00 dt2 + (g00 A 2 + g33 B 2 + 2g01 A′ + 2g13 B′ + 2g03 A′ B′ )dρ2
+ g22 dθ2 + g33 d ϕ̃2 + 2g03 dtd ϕ̃ + 2(g03 A′ + g33 B′ + g13 )d ϕ̃dρ
+ 2(g00 A′ + g03 B′ + g01 )dtdρ (6.192)

We now demand that the coefficients of d ϕ̃dρ and dtdρ zanish, we must have

g03 A′ + g33 B′ + g13 = 0 (6.193)

and

g00 A′ + g03 B′ + g01 = 0 (6.194)

respectively. Multiplying (6.193) by g03 and (6.194) by g33 and subtracting implies

g33 g01 − g03 g13

A′ = . (6.195)
g203 − g00 g03

Multiplying (6.193) by g00 and (6.194) by g03 and subtracting implies

84
 g00 g13 − g03 g01
B′ = . (6.196)
g203 − g00 g03

!
2 2 2 2mρ 2 4
g33 g01 − g03 g13 = − (ρ + a ) sin θ + 2 (a sin θ) (−1)
ρ + a2 cos2 θ
2mρa sin2 θ
!
− 2 a sin2 θ
ρ + a2 cos2 θ
= (ρ2 + a2 ) sin2 θ (6.197)

2mρa sin2 θ
! !
2mρ 2
g00 g13 − g03 g01 = 1− 2 a sin θ − 2 (−1)
ρ + a2 cos2 θ ρ + a2 cos2 θ
= a sin2 θ (6.198)

!2
2mρa sin2 θ
!
2mρ
g203 − g00 g33 = − 1− 2 ×
ρ2 + a2 cos2 θ ρ + a2 cos2 θ
!
2 2 2 2mρ 2 4
× − (ρ + a ) sin θ + 2 (a sin θ)
ρ + a2 cos2 θ
sin2 θ 4m2 ρ2 a2 sin2 θ
= [ +
ρ2 + a2 cos2 θ ρ2 + a2 cos2 θ
2mρa2 sin2 θ
+ (ρ2 + a2 cos2 θ − 2mρ)[(ρ2 + a2 ) + 2 ]
ρ + a2 cos2 θ
sin2 θ
= [(ρ2 + a2 cos2 θ)(ρ2 + a2 ) + 2mρa2 sin2 θ − 2mρ(ρ2 + a2 )]
ρ2 + a2 cos2 θ
sin2 θ
= 2 2 2
[(ρ2 + a2 cos2 θ)(ρ2 + a2 ) − 2mρ(ρ2 + a2 cos2 θ)]
ρ + a cos θ
= (ρ2 + a2 − 2mρ) sin2 θ (6.199)

Therefore

ρ 2 + a2
A′ = (6.200)
ρ2 + a2 − 2mρ

and

a
B′ = (6.201)
ρ2 a2
+ − 2mρ

85
The line element is now given by (6.192) with the last two terms vanishing by construction. The
calculation of the coefficient of dρ2 can be simplified

′ ′
g00 A 2 + g33 B 2 + 2g01 A′ + 2g13 B′ + 2g03 A′ B′
= A′ (A′ g00 + B′g03 + g01 ) + B′(A′ g03 + B′ g33 + g13 ) + g01 A′ + g13 B′
= g01 A′ + g13 B′
ρ 2 + a2 2 a
= (−1) 2 + a sin θ
ρ + a2 − 2mρ ρ2 + a2 − 2mρ
2 2 2
ρ + a cos θ
= − 2 (6.202)
ρ + a2 − 2mρ

The line element is then

ds2 = g00 dt2 + (g01 A′ + g13 B′ )dρ2 + g22 dθ2 + g33 d ϕ̃2 + 2g03 dtd ϕ̃ (6.203)

ρ2 + a2 cos2 θ 2
!
2 2mρ 2
ds = 1− 2 dt − dρ − (ρ2 + a2 cos2 θ)dθ2
ρ + a2 cos2 θ ρ2 + a2 − 2mρ
4maρ sin2 θ
!
2 2 2 2mρ 2 4 2
− (ρ + a ) sin θ + 2 (a sin θ) d ϕ̃ + 2 dtd ϕ̃
ρ + a2 cos2 θ ρ + a2 cos2 θ
(6.204)

This is the Kerr solution in Boyer-Lindquist coordinates and is analogous to the Schwarzschild
coordinates for a non-rotating black hole.

6.5 Interptetation as Rotating Body

2mr 2 2mar sin2 θ

!
2
ds = 1 − 2 dt + (dtdφ + dφdt)
ρ ρ2
ρ2 sin2 θ h i
− dr2 − ρ2 dθ2 − 2 (r2 + a2 )2 − a2 ∆ sin2 θ dφ2 (6.205)
∆ ρ

where

∆(r) = r2 − 2mr + a2 (6.206)

86
and

ρ2 = r2 + a2 cos2 θ (6.207)

a = J/M (6.208)

where J is the Komar anagular momentum.

6.6 Basic Propertire of the Kerr Solution

6.6.1 Boyer-Lindquist cordinates

If we let a → 0 we obtain

r2
!
2 2mr 2
ds = 1 − 2 dt − 2 dr2
r r − 2GMr
sin2 θ
−r2 dθ2 − 2 r4 dφ2 (6.209)
r

i.e. as a → 0 Boyer-Lindquist cordinates reduce to Schwarschild coordinates.

The Boyer-Lindquist form is the most useful one for investigating the properties of the Kerr
solution.

• If we set a = 0, the regain the Schwarschild solution in Schwarschild coordinates and so m is

identified with the geometric mass;

• The metric coefficients are independent of t and hence the solution is stationary;

• The metric coefficients are independent of φ and hence the solution is axially symmetric, i.e.
there is an axis such that the solution is invariant under rotation about this axis.

• As for discrete symmetries, the solution is invariant under the symultaneous inversion of t and
φ, that is under the transformation

t → −t, φ → −φ. (6.210)

This suggests that the Kerr solution may correspond to a spinning source, since running time
backwards with negative spin direction is equivalent to unning time forward with positive spin
directin.

87
6.6.2 Cartesian Coordinates

√
x = r2 + a2 sin θ cos φ
√
y = r2 + a2 sin θ sin φ
z = r cos θ (6.211)

6.7 Singularities and Event Horizons

Event horizons are null surfaces beyond which it is impossible to return to a certain region of
space.

The stataionary limit surface is timelike exery where except where it s tangent to the event
horizons at the poles. It represents the place past which it is impossible to remail stationay.

inner event horizon ring singularity

outer event horizon r+

r−

ergosphere

Figure 6.1: Rotating blackhole

∆ = r2 − 2mr + a2 = 0, (6.212)

two null event horizons

r± = m ± (m2 − a2 )1/2 . (6.213)

Then in a similarly way in which the Reissner-Nordstrom solution is regular in three regions:

I. r+ < r < ∞

II. r− < r < r+

88
III. 0 < r < r−

In the limit a → 0

89
Appendix A

Derivation of the Full Field Equations of

General Relativity

We must have the condition that physical laws reduce to those of special relativity in an in-
fintesimal small inertial frame.There is a simply procedure for writing down the equations of
matter or radiation in a grtavitational field. They are obtained from equations of matter or radi-
ation in special relativity by replacing partial derivatives ∂a by covariant derivatives ∇a and the
Minkowski metric ηab with the gravitational metric gab , and replacing the volume element d 4 x
√
by the invariant volume element −gd 4 x, where g is the dertminant of gab (where, however, the
metric no longer flat but is a solution to the full field equations). The resulting equations will
automatically satisfy the principle of general covariance and the equivalence principle. This
procedure is called “minimal coupling”. These do not represent the only choice as there are in
fact infinetly many different generally covariant equations that reduce to the same special rela-
tivistic equations. The principle of “minimal coupling” is a simplicity principle. It asserts that
no terms explicitly containing the curvature tensor (which vanishes in special relativity) should
be added in making the transition from special relativity to general relativity.

Here we first derive Einstein’s vacuum field equations for general relativity. To obtain the full
fields equations, we take the action to be

Z
S = (LG + κLM ) d 4 x (A.1)
M

where LG is the Lagrangian density for the gravitational field, LM is the Lagragian density of
the matter or radiation field, and κ is a coupling constant. Both Lagrangians are functionals of
the metric and its derivatives.

The method we use to obtain the full field equations is precisely analogous to the way a source
term for the Maxwell’s equations, a current J a , arises from a variation of the coupled matter-
Maxwell action with respect to the gauge potential, see section A.5.3.

90
The energy-momentum tensor T ab of matter or radiation content can be computed, in general, as
the variation of the action with respect to the metric. If the matter/radiation action is S M [ϕ, gab],
the energy-momentum tensor can be computed as

1 δS M
T ab = 1
. (A.2)
(−g) δgab (x)
2

We will obtain the full field equations of general relativity:

Gab = κT ab (A.3)

where Gab is the Einstein tensor.

We then consider the most important energy-momentum tensors. We first obtain the energy-
momentum tensor for dust, and then for a perfect fluid in Minkowski spacetime. We do this
not via an action principle, but by physical reasoning (it is possible to write down an a dust
Lagrangian, producing all the required field equations - which we touch upon). We make the
tansition to general relativity using the princple of minimal coupling. We then develop the
tensorial formulation of Maxwell’s theory governing the electromagnetic field. We pass to the
generalivistic theory via the minimal coupling procedure, and obtain the energy momentum ten-
sor in source free regions via (A.2). Using this write down the the full field equations for gravity
coupled to the electromanetic field in source free regions, known as the Einstein-Maxwell equa-
tions.

A.1 Mathematical Preliminaries

A.1.1 The Metric Connection

1
Γabc = gad (∂b gdc + ∂c gdb − ∂d gbc ) (A.4)
2

∇c gab = ∂c gab − Γdac gdb − Γdbc gda

1 1
= ∂c gab − gde (∂a gec + ∂c gea − ∂e gac )gdb − gde (∂b gec + ∂c geb − ∂e gbc )gda
2 2
1 1
= ∂c gab − (∂a gbc + ∂c gba − ∂b gac ) − (∂b gac + ∂c gab − ∂a gbc )
2 2
= 0.

∂c gab = Γdac gdb + Γdbc gad (A.5)

91
A.1.2 Tensor Densities

A tensor density of weighht W, usually denoted by a gothic letter, Ta···

b··· , transforms like an
W
ordinary tensor, excep that J appears as a factor where

∂xa
J= , (A.6)
∂x′ b

that is, tensor density of weighht W transforms as

′
′
a··· ∂x a ∂xd
T b··· = JW . . . . . . Ta···
b··· (A.7)
∂xc ∂x′ b

We adhere to the definition of the covariant derivative given by

∇c Ta··· a··· a d··· d a··· d a···

b··· = ∂c T b··· + Γdc T b··· + · · · − Γbc T d··· − WΓdc T b··· . (A.8)

For example, the covariant derivative of a vector density of weight W is

∇c Ta = ∂c Ta + Γabc Tb − WTa .

In the special case where W = 1 and we contract over a and c, we get the important result

∇a Ta = ∂a Ta . (A.9)

A.1.3 Matrix Identities

Consider the matrix A = (Ai j ). Denote by a the determinant of A and denote by C i j the cofactor
of Ai j . Let us fix i, and expand the determinant a by the ith row. Then

n
X
a= Ai j C i j (A.10)
j=1

where we have included summation for clarity. Note we would have

n
X
0= Ai j C k j (A.11)
j=1

92
if k , i as the resulting determinant would have two rows in common. It follows that the inverse
(Ai j ) is given by

1 1
(Ai j ) = (C i j )T = (C ji ) (A.12)
a a

If we partially differentiate both sides of (A.10) with respect to Ai j , we get

∂a
= C i j, (A.13)
∂Ai j

since Ai j does not occur in any of the cofactors C ji (for fixed i, while j runs from 1 to n). This
works for any i and so the formula (A.13) holds for all i and j. Using (A.12) in (A.13)

∂a
= C ij
∂Ai j
= aA ji . (A.14)

Now let us supose that the Ai j are all functions of the coordinates xk . Then the determinant
depends on Ai j , which in turn are functions of the coordinates xk , that is,

 
a = a Ai j (xk ) .

Differentiating with respect to xk , we obtain

∂a ∂a ∂Ai j
k
=
∂x ∂A ji ∂xk
∂Ai j
= A ji k
∂x
∂Ai j
= aA ji k (A.15)
∂x

by (A.12).

A.1.4 Metric Identities

Applying (A.15) to the metric determinant g and using that gab is stmmetric, we obtain the
equation

93
 ∂c g = ggab ∂c gab . (A.16)

We combine this with (A.5) and find

∂c g = ggab ∂c gab (Γdac gdb + Γdbc gad )

= 2gΓaac . (A.17)

Let us compute the covariant derivative of g using (A.8). Then since g is a scalkar density of
weight +2, we have

∇c g = ∂c g − 2gΓaac

and so by equation (A.17) it follows that

∇c g ≡ 0. (A.18)

Similarly, we find from equation (A.17) that

1 1 1
∂c (−g) 2 = (−g)− 2 ∂c (−g)
2
1
= (−g) 2 Γaac ,

that is, by (A.8)

1
∇c (−g) 2 ≡ 0.. (A.19)

a···
In particular, for the tensor T b··· , this leads to the identity

1 1
h i
a··· a···
∇c (−g) 2 T b··· = (−g) 2 (∇c T b··· ), (A.20)

1
that is, one can move factors of (−g) 2 and g through covariant derivatives in the same way that
one can with factors involving the covariant or contravariant metric.

94
A.1.5 Palantini Equation

Recall the Riemann tensor

Rabcd = ∂c Γabd + Γebd Γaec − (c ↔ d) (A.21)

The variation of the Riemann tensor with respect to the the connection (due to a variation in
the connection due to a variation in the metric). Let us consider arbitrary variations of the
a
connection to a new connection Γbc :

a
Γabc 7→ Γbc = Γabc + δΓabc . (A.22)

The variation δΓabc is the difference of two connections, and is therefore itself a tensor. We can
take its covariant derivative,

∇c (δΓabd ) = ∂c (δΓabd ) + Γaec (δΓebd ) − Γecb (δΓaed ) − Γecd (δΓaeb ) (A.23)

We use this expression to calculate the first order variation in the Riemann tensor

δRabcd = ∂c (δΓabd ) + Γaec (δΓebd ) + Γebd (δΓaec ) − (c ↔ d)

= ∂c (δΓabd ) + Γaec (δΓebd ) − Γebc (δΓaed ) − (c ↔ d)
= ∂c (δΓabd ) + Γaec (δΓebd ) − Γecb (δΓaed ) − Γecd (δΓaeb ) − (c ↔ d)
= ∇c (δΓadb ) − ∇d (δΓacb ).

This result,

δRabcd = ∇c (δΓadb ) − ∇d (δΓacb ). (A.24)

is called the Palantini equation.

Contracting over a and c gives the result

δRbd = ∇a δΓabd − ∇d δΓaba

(A.25)

95
A.2 The Einstein Lagrangian

The Einsten Lagrangian that gives the vacuum field equations for General Relativity are

1
LG = (−g) 2 R (A.26)

where the G denotes that it is the Lagrangian gravitational.

The variation

gab 7→ gab + δgab

induces a variation in gab , which we write

gab 7→ gab + δgab . (A.27)

δac = gab gbc

7 → (gab + δgab )(gbc + δgbc )
= δac + δgab gbc + gab δgbc + O(δ2 ) (A.28)

But since δac is a constant tensor, it remains unaltered by any variations and therefore

δgab gbc + gab δgbc = 0 (A.29)

to first order. Multiplying through by gcd we have

δgad = −gab gcd δgbc . (A.30)

We will need the variation of g under the variation δgab . Consider, for an arbitrary matrix M,
the variation of ln | det M| induced by a variation of M’s elements,

δ ln | det M| = ln | det(M + δM)| − ln | det M|

det(M + δM)
= ln
det M
= ln det M−1 (M + δM)|
= ln det(1 + M−1 δM)| (A.31)

96
Write N = 1 + ǫ, then

det N = ǫ i jkl N0i N1 j N2k N3l

= ǫ 0123 (1 + ε00 )(1 + ε11 )(1 + ε22 )(1 + ε33 ) + ǫ 1023 ε01 ε10 (1 + ε22 )(1 + ε33 ) + · · ·
= 1 + Trε + O(ε2 )

Using this in (A.31)

δ ln |detM| = ln(1 + Tr M−1 δM)

= Tr M−1 δM.

Substituting the metric tensor in place of M gives

δg = ggab δgab .

This is used to obtain the result

1 1
δ(−g) 2 = − 1
δg
2(−g) 2
1 g ab
= − g δgab
2 (−g) 21
1 1
= (−g) 2 gab δgab . (A.32)
2

We write the action as

Z
S = gab Rab d 4 x (A.33)
M

A variation in the metric results in a variation in the Ricci tensor, so that

Z  
δS = δgab Rab + gab δRab d 4 x (A.34)
M

We now use the Palantini equation in the form (A.25), so that the second term on the RHS
becomes

97
 Z Z
ab 4
g δRab d x = gab ∇c δΓcab − ∇b δΓcac d 4 x



M
ZM h    i
= ∇c gab δΓcab − ∇b gab δΓcac d 4 x
ZM h i
= ∂c gab δΓcab − gac δΓbab d 4 x (A.35)
M

Where we have used (A.9). This can be converted into a surface integral using the divergence
theorem, which vanishes as we assume variations vanish there.

Hence (A.35) reduces to

Z
δS = δgab Rab d 4 x
ZM
1
h i
= Rab δ (−g) 2 gab d 4 x
ZM h
1 1
i
= Rab gab δ(−g) 2 + Rab (−g) 2 δgab d 4 x
ZM !
1 1
= (−g) 2 Rg − Rab g g δgcd d 4 x
cd ac bd
M 2
Z !
1 1 cd
= − 2 cd
(−g) R − Rg δgcd d 4 x
M 2
Z h
1
i
= − −(−g) 2 Gab δgab d 4 x (A.36)
M

where we have used (A.30) and (A.32).

A.3 The Full Field Equations - Inclusion of Matter and Ra-

diation

The full action is the sum of the gravitational action plus the matter action

Z
S [gab , φ] = (LG + κLM ) d 4 x (A.37)
M

where φ denotes the matter fields. Variation of this total action with respect to the matter fields
φ is equivalent to variation of the matter action S M alone with respect to the matter fiekds,

98
 S [gab , φ] S M [gab , φ]
=0 ⇔ =0 (A.38)
δφ δφ

and will simply give rise to the equations of motion of the matter fields in the gravitational field.

Variation of the gravitational action with respect to the gravitational field will give rise to the
gravitational part of the field equations, we obtain

δS G 1
= −(−g) 2 Gab . (A.39)
δgab

Variation of the matter action with respect to the gravitational field will give us the source term
for the graviational field equations provided by the matter fields

δS M 1
= (−g) 2 T ab , (A.40)
δgab

where the latter equation defines the energy-momentum tensor for the fields present. Dividing
1
through by (−g) 2 , we have the full field equations

Gab = κT ab (A.41)

The above is precisely analogous to the way a source term for the Maxwell’s equations, a
current J a , arises from a variation of the coupled matter-Maxwell action with respect to the
gauge potential, see section A.5.3.

A.4 The Energy-Momentum Tensor for Dust, a Perfect Fluid

A.4.1 Source term: From Newton’s Theory to Special Relativity

Let us first recall the non-relativist Newtonian theory for gravity in the pressence of matter.
In terms of the gravitational potential defined by F = −∇ϕ ~ Newton’s law of gravity in its
differential form is

~ 2 ϕ = 4πGρ
∇ (A.42)

where ρ is the rest local density of matter and ϕ is the local gravitational potential (gravitational
potential energy per unit mass). In order to see how ρ can be generalised in special relativity
consider a dust cloud. A cloud in its rest frame has energy density

99
 ρ 0 = m0 n 0

where m0 is the rest mass of a dust particle and n0 the number of dust particles per unit volume.
In special relativity, the mass of a body is greater than its rest mass by a factor of γ and the
volume decreases by a factor of γ because of Lorentz contraction. Hence, the energy density as
viewed by an observer with velocity v = β with respect to the cloud is

ρ = ρ0 γ 2 .

Obviously ρ is neither a scalar nor the component of a four-vector. The behaviour of ρ is exactly
that of the time-time component of a second-rank tensor T ab :

T ab = ρ0 ua ub , (A.43)

where

ua = γ(1, u)

is the four-vector velocity of the cloud. In the rest frame of the cloud only the time-time com-
ponent of the tensor is non-zero.

A.4.2 Dust

The simplest form of matter field, that of dust. Let us formulate the again the energy-momentum
tensor, and interpret all its componenets. A dust field may be characterised two quantities, the
4-velocity vector field

dxa
ua = , (A.44)
dτ

where τ is the proper time along the world-line of the dust particle and a scalar field

ρ0 = ρ0 (x) (A.45)

describing the proper density of the flow, that is, the density which would be measured by an
observer moving with the flow. The simplest second-rank tensor we can construct from these
two quantities is

100
 T ab = ρ0 ua ub , (A.46)

where ua is the 4-velocity

ua = γ(1, u), (A.47)

1
where γ = (1 − u2 )− 2 . Recall that the proper time is defined by

dτ2 = ηab dxa dxb

= dt2 − dx2 − dy2 − dz2
= dt2(1 − u2 )
= γ2 dt2 (A.48)

The zero-zero component of T ab

dx0 dx0
T 00 = ρ0
dτ dτ
dt2
= ρo 2
dτ
= γ 2 ρo , (A.49)

by (A.48). Therefore, the relativistic energy density of the matter field is the component T 00 of
the second rank tensor T ab .

ρ = γ 2 ρ0 (A.50)

Using (A.47) and (A.50), the components of T ab can be written in the form

 
 1 u x uy uz 
u u2x u x uy u x uz 
T ab = ρ  x 2 (A.51)
uy u x uy uy uy uz 

uz u x uz uy uz u2z

This tensor has the interpretation of the flow of the a component of the four-momentum along
the b direction. Note the following

(i) T 00 is the energy density.

101
(ii) T 0i is the energy flux in the i direction.

(iii) T ii is the flux of the i component of momentum in the i direction.

(iv) T i j is the flux of the i component of momentum in the j direction.

(v) T i0 is the density of the i component of momentum.

This is summarised by:

 Energy Energy flux

 

 density 
 
T ab =  Density of 
 (A.52)
 momentum ”stress”
 


Equations of motion

We now show that the equations that govern the force free motion of a matter field of dust can
be written in the following compat way:

∂b T ab = 0. (A.53)

Using (A.51), in the case of a = 0, this equation becomes

∂ρ ∂ ∂ ∂
+ (ρu x ) + (ρuy ) + (ρuz ) = 0.
∂t ∂x ∂y ∂z

This is the equation of continuity

∂ρ
+ div (ρu) = 0. (A.54)
∂t

In fluid dynamics, this expresses the conservation of matter with density ρ moving with velocity
u. Since matter is the same as energy in special relativity, it follows that the conservation of
energy equation for dust is ∂b T 0b = 0. The equation corresponding to a = α (α = 1, 2, 3) are

∂ ∂ ∂ ∂
(ρu) + (ρu x u) + (ρuy u) + (ρuz u) = 0.
∂t ∂x ∂y ∂z

or

102
 ∂ρ ∂u ∂u ∂u ∂u
+ div (ρu) + ρ + ρu x + ρuy + ρuz = 0.
∂t ∂t ∂x ∂y ∂z

Combining this with (A.54), the above equation can be written

" #
∂u
ρ + (u · grad)u = 0. (A.55)
∂t

Comparing this with the Navier-Stokes equation of motion for a perfect fluid in fluid dynamics,
namely,

" #
∂u
ρ + (u · grad)u = −grad p + ρ f, (A.56)
∂t

where p is the pressure in the fluid and f is the body force per unit mass, we see that (A.55) is
simply the Navier-Stokes equation in the absence of pressure or external forces.

The covariant counterpart of (A.53) is

∇b T ab = 0. (A.57)

A.4.3 Dust Lagrangian

1
Ldust = (−g) 2 ρ(gab ua ub − 1) (A.58)

Variang with respect to ρ yields

1 ∂Ldust
1
= gab ua ub − 1 = 0 (A.59)
(−g) ∂ρ
2

so that gab ua ub = 1. Recall

1
∂(−g) 2 1 1

ab
= − (−g) 2 gab
∂g 2

so that

103
 1∂Ldust
T ab = 1
(−g) ∂gab 2

1
= − gab ρ(gcd uc ud − 1) + ρua ub
2
= ρua ub (A.60)

where we have used gab ua ub = 1.

A.4.4 Perfect Fluid

A perfect fluid is characterised by three quantities:

i) a 4-velocity ua = dxa /dτ ,

ii) a proper density ρ0 = ρ0 (x) ,

iii) a scalar pressure field p = p(x) .

In the limit as p vanishes, a perfect fluid reduces to dust. This suggests that one takes the
energy-momentum tensor for a perfect fluid to be of the form

T ab = ρ0 ua ub + pS ab (A.61)

for some symmetric tensor S ab . The only second-rank tensors which are associated with the
fluid are ua ub and the metric gab , and so the simplest assumption is that

S ab = λua ub + µgab (A.62)

where λ and µ are constants.

We investigate the conservation law ∂b T ab = 0 in special relativity in Minkowski coordinates

and require that it reduces in an appropriate limit to the continuity equation (A.54) and the
Navier-Stokes equation (A.56) in the ansence of body forces.

It follows directly from the definition of the 4-velocity that

ua ua = 1

This implies

104
 1
ua ∂b ua = ∂b (ua ua ) = 0. (A.63)
2

In the non-relativistic limit,

p ≪ ρ,
ua = (1, u),
dp
|u| ≪ |gradp|, (A.64)
dt

these equations become

0 = ∇b T ab
= ∂b T ab
= ∂b [ρua ub + p(λua ub + µηab )]
= ua ub ∂b ρ + (ρ0 + λp)∂b (ua ub ) + (µηab + λua ub )∂b p (A.65)

Projecting parallel to ua :

ua ∂b T ab = ua ∂a ρ0 + (ρ0 + λp)ua ∂b (ua ub ) + ua (µηab + λua ub )∂b p

= ua ∂a ρ0 + (µ + λ)ua ∂a p + (ρ0 + λp)(∂a ua + ub ua ∂b ua )
= ua ∂a ρ0 + (µ + λ)ua ∂a p + (ρ0 + λp)∂a ua
= 0 (A.66)

This becomes

!
∂ρ ∂p
+ (u · grad)ρ + (µ + λ) + (u · grad)p + (ρ0 + λp)div u = 0. (A.67)
∂t ∂t

Projecting perpendicular to ua :

(ηac − ua uc )∂b T cb = (ρ0 + λp)(ηac − ua uc )∂b (uc ub ) + µ(ηac − ua uc )ηcb ∂b p

= (ρ0 + λp)(ηac − ua uc )ub ∂b uc + (δba − ua ub )∂b p
= (ρ0 + λp)ub ∂b ua + µ∂a p − µua ub ∂b p
= 0 (A.68)

105
For a , 0 equation (A.68) becomes

! !
∂u ∂p
−(ρ0 + λp) + (u · grad)u + µgradp + µu + u · gradp = 0 (A.69)
∂t ∂t

In the non-relativistic limit given by (A.64), we have that (A.67) and (A.69) become, respec-
tively,

!
∂ρ ∂p
+ div(ρu) + (µ + λ) + (u · grad)p = 0, (A.70)
∂t ∂t
" #
∂u
ρ + (u · grad)u = µgrad p. (A.71)
∂t

These reduce to the continuity equation and the Navier-Stokes equation in the absence of body
forces if we take µ + λ = 0 and µ = −1. Using this in (A.62), then substituting this into (A.61)
we have

T ab = (ρ0 + p)ua ub − pgab (A.72)

A.5 The Energy-Momentum Tensor for the Electro-Magnetic

Field

A.5.1 Reminder of Maxwell’s Equations

The source equations are

div E = ρ (A.73)
∂E
curl B − = j, (A.74)
∂t

and the internal equations

div B = ρ (A.75)
∂B
curl E − = 0 (A.76)
∂t

106
where E is the electric field, B is the magnetic field. ρ is the charge density, and j. The quantities
ρ and j cannot be prescribed independently because, differentiating (A.73) with respect to t, we
get

∂E ∂ρ
div = , (A.77)
∂t ∂t

and taking the divergence of (A.74) gives

∂E
div = div j. (A.78)
∂t

Thus, ρ and j must satisfy the continuity equation

∂ρ
+ div j = 0. (A.79)
∂t

In order to write these equations in tensorial form, one defines the anti-symmetric tensor F ab ,
called the electromagnetic field strength tensor, by

 
 0 E x Ey E z 
−E Bz −By 
 x 0 (A.80)
−E y −Bz 0 B x 


−E z By −B x 0


and the current density or source 4-vector by

ja = (ρ, j). (A.81)

Then the source equations and internal equations can be written in the form

∂b F ab = ja , (A.82)
∂a F bc + ∂c F ab + ∂b F ca = 0. (A.83)

The anti-symmetry of F ab means that (A.83) can be written in the more compact form as

∂[a F bc] = 0. (A.84)

The continuity equation (A.79) becomes

107
 ∂a ja = 0. (A.85)

This obviously follows from (A.82) as ∂a ∂b F ab = ∂a ja = 0

A.5.2 Potential Formulation of Maxwell’s Equations

∂A
E = −grad φ − (A.86)
∂t
B = curl A. (A.87)

The internal equation are autometically solved.

If we define the 4-potential by

Aa = (φ, A), (A.88)

then (A.86) and (A.87) are equivalent to

F ab = ∂b Aa − ∂a Ab (A.89)

as we now verify. First

∂A0 ∂A1
F 01 = + = −E x
∂x ∂t
∂A0 ∂A2
F 02 = + = −E y
∂y ∂t
∂A0 ∂A3
F 03 = + = −E z
∂y ∂t

then

∂(−A2 ) ∂(−A3 )
F 23 = − = Bx
∂z ∂y
∂(−A1 ) ∂(−A3 )
F 13 = − = −By
∂z ∂x
∂(−A1 ) ∂(−A2 )
F 12 = − = Bz .
∂y ∂x

108
The 4-potential is not uniquely defined since one can perform a gauge transformation

Aa 7→ Ãa = Aa + ∂a ψ (A.90)

where ψ is an arbitrary scalar field. The gauge transformations change the potentials, it leaves
F ab , and hence E and B, unaltered.

Transition to Covariant Formulism

To obtain the covariant formulation, we simply replace ordinary derivatives by covariant deriva-
tives. However (A.84) and (A.89)

∇[a F bc] = ∂[a F bc] (A.91)

and

∇[b Aa] = ∂[b Aa] (A.92)

as

∇[a F bc] = ∇a F bc + ∇c F ab + ∇b F ca
= ∂a F bc − Γdba F dc − Γdca F bd +
+ ∂c F ab − Γdac F db − Γdbc F ad
+ ∂b F ca − Γdcb F da − Γdab F cd
= ∂a F bc + ∂c F ab + ∂b F ca
= ∂[a F bc] ,

and

2∇[b Aa] = ∂b Aa − Γcab Ac − (∂a Ab − Γcba Ac )

= 2∂[b Aa] . (A.93)

The covariant formulation of Maxwell’s equations in special relativity are

∇b F ab = ja (A.94)
∂[a F bc] = 0 (A.95)

109
subject to

∇a ja = 0. (A.96)

In terms of the 4-potetial, we still have

F ab = ∂b Aa − ∂a Ab . (A.97)

By the principle of minimal gravitational coupling, one adopts equations (A.94) and (A.95) in
general relativity, where, however, the metric is no longer flat but a solution of the full field
equations Gab = κT ab where T ab is the energy-momentum tensor of the electromagnetic field.

A.5.3 Action for Maxwell Equations and Minimal Coupling to Charged

Matter

Consider the Lagragian for the electrodynamical field defined by

" #
1 1 ab ab
LE (Aa , F ab ) = − F ab F + (Aa,b − Ab,a )F . (A.98)
4π 2

Then

!
δLE ∂LE ∂LE
= − (A.99)
δAa ∂Aa ∂Aa,b ,b
1
= 0 − (F ab − F ba ),b
4π

and the field equations corresponding to a variation with respect to φa become

(F ab − F ba ),b = 0. (A.100)

Similarly,

110
 δLE ∂LE
= (A.101)
δF ab ∂F ab
" #
∂ 1 1 ce d f ce d f
= − η η F cd F e f + η η (Ac,d − Ad,c )F e f
∂F ab 4π 2
" #
1 1 ae b f 1 ca db ca db
= − η η F e f − η η F cd + η η (Ac,d − Ad,c )
4π 2 2
ac bd
η η  
= −F cd + (Ac,d − Ad,c )
4π

and so the field equations corresponding to a variation with respect to F ab are

F ab = ∂b Aa − ∂a Ab . (A.102)

This equation defines F ab in terms of the 4-potential and implies that F ab is anti-symmetric. The
definition also means that the internal equations are automatically satisfied and (A.100) reduces
to

∂b F ab = 0, (A.103)

namely, the source equations (for source-free regions). Equation (A.102) also allows one to
re-express the Lagrangian as

1 ac bd
LE = η η F ab F cd (A.104)
8π
1
R
In order to add sources, one can add (−g) 2 Aa ja d 4 x to the Maxwell action thus coupling the
matter currect to the Maxwell gauge field

S M [φ] → S M [φ, Aa ]. (A.105)

The combined Maxwell and matter action will then give rise to the Maxwell equations with
source provided that we define the current ja as the variation of the matter action with respect
to the gauge field,

δS M [φ, Aa ]
ja ∝ . (A.106)
δAa

This is the promised analogue to the procedure we followed when obtaining the full field equa-
tions of general relativity in section A.3.

111
A.5.4 The Maxwell Energy-momentum Tensor

To make the transition to the full theory we ussume that

1
(−g) 2 ac bd
LE = g g F ab F cd (A.107)
8π
1
together with the definition (A.102) of F ab in terms of φa . The factor (−g) 2 is included to ensure
that LE is a scalar density.

We will need

1
∂g ∂(−g) 2 1 1
= −ggab ⇒ = − (−g) 2 gab
∂gab ∂g ab 2

Using this we obtain

 1 
∂LE ∂  (−g) 2 ce d f 
=  g g F cd F e f 
∂gab ∂gab 8π

" #
1 1 1
ce d f 1
cd 1
cd
= (−g) 2 gab g g F cd F e f + (−g) 2 g F ac F bd + (−g) 2 g F ca F db
8π 2
1 " #
(−g) 2 cd 1 cd
= − −g F ac F bd + gab F cd F
4π 4

so that

1 " #
∂LE (−g) 2 cd 1 cd
=− −g F ac F bd + gab F cd F (A.108)
∂gab 4π 4

The analogue of (A.40) for the contravariant metric is

δLE 1

ab
= −(−g) 2 T ab . (A.109)
δg

These last two equations lead to the definition of the Maxwell energy-momentum tensor T ab in
source-free regions

1 1
T ab = (−gcd F ac F bd + gab F cd F cd ) (A.110)
4π 4

112
Let us look at the interpretation of the components of T ab in special relativity. Note that the T 00
componentis

1 2
T ab = (E + B2 ), (A.111)
8π

which is the usual expressionfor the energt density. The components

1
(T 01 , T 0,2 , T 0,3 ) = − E × B, (A.112)
4π

where the vector E × B is the Poynting vector of electrodynamics and represents the momentum
density of the electrodynamic field.

We have that

 Energy Energy flux

 

 density 
 
T ab =  Energy 
 (A.113)
 flux ”stress”
 


We don’t compute the ”stress” part of the tensor. We will just mention that it can be thouhgt of
as the pressure of the electrodynamical field.

A.6 Einstein-Maxwell Equations

The full field equations in source-free regions are called the Einstein-Maxwell equations which
read

1
Gab = −2gcd F ac F bd + gab F cd F cd . (A.114)
2

113