You are on page 1of 34

Perturbations of Black Holes

David Walker

Supervisor: Dr. B. Nolan

15th April 2002

This dissertation is submitted in partial fulfilment of the requirements for


the Mathematical Sciences B. Sc degree. This work is entirely the work of
the author, except where indicated otherwise.

Signature:

1
Contents
1 Mathematical Background 4
1.1 Manifolds and tensors . . . . . . . . . . . . . . . . . . . . . . 4
1.2 The metric . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3 The metric connection . . . . . . . . . . . . . . . . . . . . . . 7
1.4 Covariant differentiation . . . . . . . . . . . . . . . . . . . . . 8
1.5 The Riemann tensor . . . . . . . . . . . . . . . . . . . . . . . 8
1.6 The Ricci tensor . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.7 The Einstein tensor . . . . . . . . . . . . . . . . . . . . . . . . 9

2 The Vacuum Field Equations in a Spherically Symmetric


Space-time 10
2.1 The non-zero Ricci tensor components . . . . . . . . . . . . . 10
2.2 The non-zero Einstein tensor components . . . . . . . . . . . . 13
2.3 The non-zero mixed Einstein tensor components . . . . . . . . 14
2.4 The contracted Bianchi identities . . . . . . . . . . . . . . . . 14
2.5 The line element . . . . . . . . . . . . . . . . . . . . . . . . . 17

3 Waves Propagating in Schwarzschild Space-time 21


3.1 The wave equation . . . . . . . . . . . . . . . . . . . . . . . . 21
3.2 Separation of variables . . . . . . . . . . . . . . . . . . . . . . 23
3.3 Stability analysis of the perturbations . . . . . . . . . . . . . . 26
3.3.1 ω = 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.3.2 ω = −iα . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.3.3 Time dependence . . . . . . . . . . . . . . . . . . . . . 31

2
Summary
Perturbations of stars and black holes has been one of the main topics of
relativistic astrophysics for the last few decades. They are of particular
importance today, because of their relevance to gravitational wave astronomy.
In this project we shall study the perturbations of Schwarzschild black holes.
The motivation for us to study these perturbations will be to assess the
stability of the spherically symmetric Schwarzschild solution for black holes.
The assumption of spherical symmetry on the black hole makes the solution
physically unrealistic. In order to determine if this solution has physical
significance, we need to study its stability. We do this by studying small
perturbations of the solution, that is, gravitational waves which impinge on
the black hole. These are described by the wave equations in the black hole
space-time. We shall then solve these equations to obtain answers as to
whether the black-hole solutions are accidents that arise as a consequence of
the assumption of spherical symmetry only, or will they exist if perturbed?
These studies will indeed show that the black hole solutions are stable under
small perturbations, that all asymmetries are eventually radiated away and
that asymptotically in time, the system settles down to a Schwarzschild black
hole.

3
1 Mathematical Background
General Relativity is Einstein’s theory of the gravitational field. It describes
gravity as the curvature of the space-time manifold. In this section, we will
define some objects which are the basics of the theory.

1.1 Manifolds and tensors


A manifold can be considered as a space which is locally similar to n-
dimensional Euclidian space Rn . A more detailed description of a manifold is
not required for the scope of this paper. We will just take an n-dimensional
manifold to be a set of points such that each point possesses a set of n
coordinates(x1 , x2 , · · · , xn ), where each coordinate ranges over a subset of
the reals, which may, in particular range from −∞ to +∞.

In physics we want to be able to obtain the same physical results for a prob-
lem, no matter what co-ordinate system we use. Therefore, the equations of
physics must not depend on any particular co-ordinate system.
If we take an equation T (xi ) = 0 in a co-ordinate system xi , then the object
T must transform to T  = 0 where T  is the representative of T in another
co-ordinate system xi .
In order to achieve this, we consider objects called tensors whose transfor-
mation laws are now defined.

Note: It should be pointed out at this stage, that from now on, whenever
an index is repeated, it will imply a summation over the index from 1 to n,
which is the dimension of the manifold. Repeated indexes are called dummy
indices and can be replaced by another index that has not already been used.

A contravariant tensor of rank 1 is defined to be a set of quantities, X a


in the xa -coordinate system which are associated with some point P on the
manifold and which transforms under a change of coordinates according to

∂xa b
X a = X
∂xb
∂xa
where the transformation matrix ∂xb
is evaluated at the point P .

4
We can generalize this definition to obtain contravariant tensors of higher
rank
∂xa ∂xb
X ab... = . . . X cd...
∂xc ∂xd

Similarly a covariant tensor of rank 1 is defined to be a set of quantities,


Xa in the xa -coordinate system which are associated with some point P on
the manifold and which transforms under a change of coordinates according
to
∂xb
Xa = Xb
∂xa
∂xb
where the transformation matrix ∂xa
is again evaluated at the point P .

We can again generalize this definition to obtain covariant tensors of higher


rank
 ∂xc ∂xd
Xab... = . . . Xcd...
∂xa ∂xb

We can now define mixed tensors. A mixed tensor of contravariant rank 1


and covariant rank 1 satisfies
∂xa ∂xd c
Xba = X
∂xc ∂xb d

Mixed tensors of higher rank transform in the same way that of covariant
and contravariant tensors of higher rank do. A mixed tensor of contravariant
rank q and covariant rank q is said to have valence (p, q).

On some region of the manifold we can define a tensor field to be some


association of a tensor of the same valence to every point in that region, that
is
P → Tb...
a...
(P ),
a...
where Tb... (P ) is the value of the tensor evaluated at P .

5
1.2 The metric
The metric measures distances on the manifold. If we take coordinates

xa = (x0 , x1 , x2 , x3 ) = (t, r, θ, φ)

then the four-dimensional spherically symmetric line element is

ds2 = eν dt2 − eλ dr2 − r2 dθ2 − r2 sin2 θdφ2

where ν = ν(t, r) and λ = λ(t, r) are arbitrary functions of t and r and ds is


the distance between adjacent space-time points.

Any symmetric covariant tensor field of rank 2, for example gab (x), defines a
metric, which implies that the above metric can be written in the form

ds2 = gab (x)dxa dxb

From this, we get an expression for gab for our metric

gab = diag(eν , −eλ , −r2 , −r2 sin2 θ)

We define the determinant of the metric by

g = det(gab )
= −eν+λ r4 sin2 θ

The metric is non-singular if g = 0, in which case the inverse of gab exists


and is denoted g ab .

Hence its contravariant form is

g ab = diag(e−ν , −e−λ , −r−2 , −r−2 sin−2 θ)

6
1.3 The metric connection
We define an affine connection Γabc to be a quantity which transforms ac-
cording to
∂xa ∂xe ∂xf d ∂xa ∂ 2 xd
Γa = Γ +
bc
∂xd ∂xb ∂xc ef ∂xd ∂xb ∂xc

If our manifold contains an affine connection and a metric, we can define a


metric connection to be a special connection which is a combination of the
metric and its partial derivatives.
1
Γabc = g ad (∂b gdc + ∂c gdb − ∂d gcb )
2

where ∂a = ∂xa
is the partial derivative taken with respect to xa .

Note: The connection is symmetric, (Γabc = Γacb ) .

We need the metric connection because it is an object which will allow us to


differentiate tensors in such a way that our result will be a tensor.

The metric connection is computed as follows:

1 0d
Γ000 = g (∂0 gd0 + ∂0 gd0 − ∂d g00 ) .
2
Setting d = 0 to keep g 0d non-zero, we get
1 00
Γ000 = g (∂0 g00 + ∂0 g00 − ∂0 g00 )
2
1 −ν
= e (∂t (eν ))
2
1
= ν̇
2

7
Computing the other non-zero elements of Γabc , we get

Γ001 = 12 ν  , Γ011 = 12 eλ−ν λ̇,

Γ100 = 12 eν−λ ν  , Γ101 = 12 λ̇ , Γ111 = 12 λ , Γ122 = −re−λ , Γ133 = −re−λ sin2 θ ,

Γ212 = r−1 , Γ233 = − sin θ cos θ ,

Γ313 = r−1 , Γ323 = cot θ

1.4 Covariant differentiation


If we take the partial derivative of a tensor, we find that the result does not
transform as per the usual tensor transformation law. The reason for this is
due to the curvature of space-time. We deal with this problem by defining
a covariant derivative to be a partial derivative plus some correction to
make the result covariant. We therefore define a covariant derivative of a
tensor of rank 1 to be
∇c X a = ∂c X a + Γabc X b
If Γabc is an affine connection then ∇c X a is a tensor of type (1,1).

The covariant derivative also has the property that is reduces to the par-
tial derivative on a scalar field, i.e.
∇a φ = ∂ a φ
The expression for a covariant derivative of a general tensor is given by
∇c Xb...
a... a...
= ∂c Xb... + Γadc Xb...
d...
+ · · · − Γdbc Xd...
a...
− ...

1.5 The Riemann tensor


For a X a , contravariant tensor of rank 1, we define its commutator to be
∇ c ∇ d X a − ∇d ∇ c X a .
Remembering that ∇c X a = ∂c X a + Γabc X b and using the definition for the
covariant derivative of a general tensor, we find that if the connection is
symmetric
∇c ∇d X a − ∇d ∇c X a = Rbcd
a
Xb

8
a
where Rbcd , the Riemann tensor is defined by
a
Rbcd = ∂c Γabd − ∂d Γabc + Γebd Γaec − Γebc Γaed

The fact that the derivatives are non-commuting indicates the presence of a
curvature in the space-time Note that if the Riemann tensor vanishes, the
commutator of the tensor also vanishes and hence the manifold is called affine
flat.

1.6 The Ricci tensor


We define the Ricci tensor to be a contraction of the Riemann tensor, i.e.
c
Rab = Racb
= ∂c Γcab − ∂b Γcac + Γdab Γcdc − Γdac Γcdb

If we impose another contraction, we can define the Ricci Scalar R as

R = g ab Rab

1.7 The Einstein tensor


A combination of the Ricci tensor and the Ricci scalar defines the Einstein
tensor as
1
Gab = Rab − gab R.
2
We shall need this tensor to solve the vacuum field equations

Gab = 0

which shall arise in the next section.

9
2 The Vacuum Field Equations in a Spheri-
cally Symmetric Space-time
We begin by assuming spherical symmetry. The vacuum field equations of
general relativity are given by

Rab = 0.

Now the Einstein tensor Gab = 0 if and only if Rab = 0, so we will look for
solutions to
Gab = g ac Gcb = 0,
to determine the unknown functions ν and λ in the spherically symmetric
line element

ds2 = eν(r,t) dt2 − eλ(r,t) dr2 − r2 dθ2 − r2 sin2 θdφ2

The solutions to Gab = 0 are derived in the following way.

2.1 The non-zero Ricci tensor components


We begin with the Riemann Tensor.
a
Rbcd = ∂c Γabd − ∂d Γabc + Γebd Γaec − Γebc Γaed

With this in mind, and the fact that the Ricci Tensor is defined by

c
Rab = Racb

⇒ Rbd = ∂c Γcbd − ∂d Γcbc + Γebd Γcec − Γebc Γced

The elements of the Ricci Tensor are derived in the following way

R00 = ∂c Γc00 − ∂0 Γc0c + Γe00 Γcec − Γe0c Γce0

10
⇒ R00 = ∂0 Γ000 − ∂0 Γ000 + Γe00 Γ0e0 − Γe00 Γ0e0
+ ∂1 Γ100 − ∂0 Γ101 + Γe00 Γ1e1 − Γe01 Γ1e0
+ ∂2 Γ200 − ∂0 Γ202 + Γe00 Γ2e2 − Γe02 Γ2e0
+ ∂3 Γ300 − ∂0 Γ303 + Γe00 Γ3e3 − Γe03 Γ3e0

Seeing which of these con be non-zero, we get


R00 = ∂1 Γ100 − ∂0 Γ101 + Γe00 Γ1e1 − Γe01 Γ1e0 + Γe00 Γ2e2 + Γe00 Γ3e3
   
1 ν−λ  1
= ∂r e ν − ∂t λ̇ + Γe00 Γ1e1 − Γe01 Γ1e0 + Γe00 Γ2e2 + Γe00 Γ3e3
2 2
   
1 ν−λ  1
= ∂r e ν − ∂t λ̇
2 2
+ Γ000 Γ101 − Γ001 Γ100 + Γ000 Γ202 + Γ000 Γ303
+ Γ100 Γ111 − Γ101 Γ110 + Γ100 Γ212 + Γ100 Γ313
+ Γ200 Γ121 − Γ201 Γ120 + Γ200 Γ222 + Γ200 Γ323
+ Γ300 Γ131 − Γ301 Γ130 + Γ300 Γ232 + Γ300 Γ333

again taking the non-zero elements, we are left with

   
1 ν−λ  1
R00 = ∂r e ν − ∂t λ̇
2 2
+ Γ000 Γ101 − Γ001 Γ100 + Γ100 Γ111 − Γ101 Γ110 + Γ100 Γ212 + Γ100 Γ313

1  ν−λ 1 1 1
= (ν e + ν  eν−λ (ν  − λ )) − λ̈ + ν̇ λ̇
2 2 2 2
1 ν−λ  1  1 ν−λ  1  1 1
− e ν ν + e ν λ − λ̇ λ̇
2 2 2 2 2 2
1 ν−λ  −1 1 ν−λ  −1
+ e νr + e νr
2 2

11
Finally, working through this yields
1 1 2 1 1 1 1
R00 = eν−λ ν  + eν−λ ν  − eν−λ ν  λ − λ̈ + ν̇ λ̇ − λ̇2 + r−1 eν−λ ν 
2 4 4 2 4 4

Similarly, the other non-zero elements of the Ricci tensor are

R01 = r−1 λ̇
1 1 1 1 1 2 1
R11 = − ν  + eλ−ν λ̇2 − eλ−ν ν̇ λ̇ + eλ−ν λ̈ − ν  + ν  λ + r−1 λ
2 4 4 2 4 4
1 −λ  1 −λ  −λ
R22 = − re ν + re λ − e + 1
2 2
R33 = sin2 θR22

Now we use the fact that R = g ab Rab and letting a = b to get the non-zero
elements of the Ricci scalar R, we find that

R = g 00 R00 + g 11 R11 + g 22 R22 + g 33 R33

 
= e−ν R00 − e−λ R11 − r−2 R22 − r−2 sin−2 θ sin2 θR22

= e−ν R00 − e−λ R11 − 2r−2 R22


 
−ν 1 ν−λ  1 ν−λ  2 1 ν−λ   1 1 1
= e e ν + e ν − e ν λ − λ̈ + ν̇ λ̇ − λ̇2 + r−1 eν−λ ν 
2 4 4 2 4 4
 
−λ 1  1 λ−ν 2 1 λ−ν 1 λ−ν 1 2 1   −1 
− e − ν + e λ̇ − e ν̇ λ̇ + e λ̈ − ν + ν λ + r λ
2 4 4 2 4 4
 
1 1
− 2r−2 − re−λ ν  + re−λ λ − e−λ + 1
2 2
Simplifying this equation, we see that
1 −λ   2   
 1
−ν
   
R= e ν − ν λ + 2ν − e λ̇ − ν̇ λ̇ + 2λ̈ +2r−1 e−λ (ν  − λ )+2r−2 e−λ − 1
2
2 2

12
2.2 The non-zero Einstein tensor components
We can now calculate the Einstein tensor Gab in the following way.
1
Gab = Rab − gab R
2
This gives us
1
G00 = R00 − g00 R
2
1 ν−λ  1 ν−λ  2 1 ν−λ   1 1 1
= e ν + e ν − e ν λ − λ̈ + ν̇ λ̇ − λ̇2 + r−1 eν−λ ν 
2 4 4 2 4 4
 
1 ν 1 −λ   2   
− e e ν − ν λ + 2ν
2 2
 
1 ν 1 −ν  2 
−1 −λ   −2
 −λ 
− e − e λ̇ − ν̇ λ̇ + 2λ̈ + 2r e (ν − λ ) + 2r e −1
2 2

Simplifying this equation gives


    
−λ λ 1 1
G00 ν
=e e − 2 + 2
r r r

Similarly, the other non-zero elements are


1
G01 = λ̇,
r
ν 1 eλ
G11 = + 2 − 2,
r r r

2
  
r −λ ν λ ν 2
ν  
λ r 2
λ̇ν̇ λ̇ 2
G22 = e − + ν  + − + e−ν − λ̈ − ,
2 r r 2 2 2 2 2
G33 = sin2 θG22 .

13
2.3 The non-zero mixed Einstein tensor components
We can now evaluate the components of the mixed Einstein tensor Gab

Gab = g ac Gcb

⇒ G00 = g 0c Gc0
= g 00 G00 + g 01 G10 + g 02 G20 + g 03 G30
= g 00 G00
     
−ν ν −λ λ 1 1
= e e e + 2
r r2 r
  
λ 1 1
= e−λ − 2 + 2
r r r
and similarly, the other non-zero components of the mixed Einstein tensor
are
1 −ν
G01 = e λ̇,
r
1
G10 = − e−λ λ̇,
r  
ν 1 1
G11 = −e −λ
+ 2 + 2,
r r r
   

1 −λ ν λ λ ν  ν 2 1 λ̇ 2
λ̇ν̇
G22 = e + − − − ν  + e−ν λ̈ + − ,
2 2 r r 2 2 2 2
G33 = G22 .

2.4 The contracted Bianchi identities


It can be shown that the Riemann tensor satisfies the identity
a a a
Rbcd + Rdbc + Rcdb =0

We can raise and lower indices of tensors according to


... ...
T...a... = gab T......b...
...

and
T......a... ab ... ...
... = g T...b...

14
Note: The metrics gab and g ab also have the property that

gab g bc = δac

where δac is known as the Kronecker delta and has the effect of turning the
index c into a.

Lowering the first index with the metric, we find that


e e e
Rabcd + Rabdc + Racdb = gae Rbcd + gae Rbdc + gae Rcdb
e e e
= gae (Rbcd + Rbdc + Rcdb )
= 0.

It can be shown that the Riemann tensor satisfies a set of identities called
the Bianchi identities

∇a Rdebc + ∇c Rdeab + ∇b Rdeca = 0

This implies that

g db [∇a Rdebc + ∇c Rdeab + ∇b Rdeca ] = 0.

Since ∇a gdb = 0 and ∇a g db = 0, we can take g db in and out of covariant


derivatives at will. We get

∇a Rebc
b
+ ∇c Reab
b
+ ∇b Reac
b
=0

From the definition of the Riemann tensor, it follows directly that the tensor
is anti-symmetric on its last pair of indices, i.e.
a
Rbcd = −Rbdc
a
.

Now, using this anti-symmetry on the indices b and a, we get

∇a Rebc
b
− ∇c Reba
b
+ ∇b Reac
b
=0

and so by a contraction,

∇a Rec − ∇c Rea + ∇b Reac


b
= 0.

15
These equations are called the contracted Bianchi identities.

Let us now contract a second time on the indices e and c,



⇒ g ec ∇a Rec − ∇c Rea + ∇b Reac
b
=0

giving us

∇a Rcc − ∇c Rac + ∇b Rca


bc
= 0
⇒ ∇a g cd Rdc − 2∇b Rab = 0
⇒ 2∇b Rab − ∇a g cd Rdc = 0
⇒ 2∇b Rab − ∇a R = 0

Since ∇a R = δab ∇b R, we get


 
1
∇b Rab − δab R =0
2

Now raising the index a with g ca we get


 
1 cb
∇b R − g R = 0
cb
2

which gives us
∇c Gcb = 0
Now lowering the index b with g ba , we get
 
∇c Gcb = ∇c g ba Gca
= g ba ∇c Gca
= 0

This gives us the contracted Bianchi identities

∇b Gba = 0

16
2.5 The line element
We shall now determine the functions ν(t, r) and λ(t, r) for the spherically
symmetric line element

ds2 = eν dt2 − eλ dr2 − r2 dθ2 − r2 sin2 θdφ2

using the the field equations Gab = 0 and also the contracted Bianchi identi-
ties ∇b Gba = 0.

If ∇b Gba = 0, then
∂a Gab + Γaac Gcb − Γcba Gac = 0

In the case of spherical symmetry, this shows that

if G00 = G01 = G11 = 0,

then G22 = 0

Note: This is proven by setting a = 1 for the contracted Bianchi identities


∇b Gba = 0, and remembering that the only non-zero elements of Gab are
G00 , G01 , G10 , G11 , G22 and G33 . So we are left with 3 independent equations to
solve, which are
  
0 −λ λ 1 1
G0 = e − 2 + 2 = 0,
r r r
1
G10 = − e−λ λ̇ = 0,
r  
1 −λ ν 1 1
G1 = −e + 2 + 2 = 0.
r r r
or more simply
 
−λ λ 1 1
e − 2 + 2 = 0, (1)
r r r
  
ν 1 1
e−λ + 2 − 2 = 0, (2)
r r r
λ̇ = 0. (3)

17
Adding (1) and (2) together, we get

λ + ν  = 0
Integrating this, we find that

λ + ν = f (t)

where f (t) is a function of integration.

Now, from (3), since

λ̇ = 0,
⇒ λ = g(r)

where g(r) is another function of integration. This implies that (1) is an


ordinary differential equation and can be written in the form

re−λ λ − e−λ + 1 = 0
⇒ e−λ − re−λ λ = 1
⇒ (re−λ ) = 1
⇒ re−λ = r + c
c
⇒ e−λ = 1 +
r
 c −1
⇒ eλ = 1 +
r

Now looking at

ds2 = eν dt2 − eλ dr2 − r2 dθ2 − r2 sin2 θdφ2

From above we have λ + ν = f (t), this implies that ν = −λ + f (t)


Thus,

eν dt2 = e−λ+f (t) dt2


= e−λ ef (t) dt2
= e−λ dT 2

18
where,

dT 2 = ef (t) dt2
f (t)
dT = e 2 dt
f (t)
T = e 2 dt

⇒ ds2 = eν dt2 − eλ dr2 − r2 dθ2 − r2 sin2 θdφ2

becomes
ds2 = e−λ dT 2 − eλ dr2 − r2 dθ2 − r2 sin2 θdφ2 .
Renaming T as t we get

ds2 = e−λ dt2 − eλ dr2 − r2 dθ2 − r2 sin2 θdφ2 .

Now substituting for eλ we get


 c 2  c −1 2
2
ds = 1 + dt − 1 + dr − r2 dθ2 − r2 sin2 θdφ2 .
r r

Now for flat space-time, we know that

eν = 1.

For a small perturbation, i.e. slowly moving bodies in a weak gravitational


field that are a large distance from the source, we can write

eν = 1 + 2φ,

where φ is small and the constant 2 is added for convenience.

It can be shown that in this approximation, φ satisfies the equations of the


Newtonian gravitational potential, and so for a spherical body of mass m,
m
φ=− .
r
This implies that
2m
eν = 1 −
r
19
and comparing this with our current line element, we find that c = −2m.

We can now define the Schwarzschild line element as


   −1
2 2m 2 2m
ds = 1 − dt − 1 − dr2 − r2 dθ2 − r2 sin2 θdφ2
r r

where m may be thought of as the mass of a spherical object which generates


the gravitational field.

20
3 Waves Propagating in Schwarzschild Space-
time
In this section we shall derive the wave equation for curved space-time. We
shall then use the process of separation of variables to find that the equations
governing the perturbations of a spherically symmetric system are separable
in all four of the variables t, r, θ and φ. We shall then analyze the solutions
to these equations to determine whether the system is stable.

3.1 The wave equation


The wave equation in four dimensions is
∂ 2ψ ∂ 2ψ ∂ 2ψ ∂ 2ψ
− − 2 − 2 =0
∂t2 ∂x2 ∂y ∂z
or
∂2ψ
− 2
+ ∇2 ψ = 0.
∂t
Now in Minkowski Space-time

η ab = diag(1, −1, −1, −1)

⇒ η ab ∂a ∂b ψ = 0.
For Special Relativity we can denote this

ψ = g ab ∇a (∇b ψ) = 0

This form of the wave equation allows us to generalize to curved space-time.

Next we denote the contravariant vector field

∇ b ψ = ψb = ∂b ψ

This gives us

∇a (∇b ψ) = ∇a ψb
= ∂a ψb − Γcab ψc
= ∂a ∂b ψ − Γcab ∂c ψ

21
and hence
ψ = g ab (∂a ∂b ψ − Γcab ∂c ψ)
where
 −1

2m   2m
gab = diag 1− ,− 1 − , −r2 , −r2 sin2 θ
r r

and
 −1  

2m 2m
g ab = diag 1− ,− 1 − , −r−2 , −r−2 sin−2 θ
r r

Therefore computing ψ gives


ψ = g 00 (∂02 ψ − Γc00 ∂c ψ) + g 11 (∂12 ψ − Γc11 ∂c ψ)
+ g 22 (∂22 ψ − Γc22 ∂c ψ) + g 33 (∂32 ψ − Γc33 ∂c ψ).

Computing the relevant connections yields


 
m 2m
Γ00 ∂c ψ = 2 1 −
c
∂r ψ
r r
 −1
m 2m
Γ11 ∂c ψ = − 2 1 −
c
∂r ψ
r r
 
2m
Γ22 ∂c ψ = −r 1 −
c
∂r ψ
r
 
2 2m
Γ33 ∂c ψ = −r sin θ 1 −
c
∂r ψ − sin θ cos θ∂θ ψ
r

and so our wave equation is


 −1    
2m 2 2m 2 2(r − m)
ψ = 1− ∂t ψ − 1 − ∂r ψ − ∂r ψ
r r R2

− r−2 (csc2 θ∂φ2 ψ + cot θ∂θ ψ + ∂θ2 ψ)

= 0

22
3.2 Separation of variables
We begin by letting ψ = P (t, r)Y (θ, φ)

 −1  
2m 2 2m
⇒ ψ = 1− ∂t P (t, r)Y (θ, φ) − 1 − ∂r2 P (t, r)Y (θ, φ)
r r
2(r − m)
− ∂r P (t, r)Y (θ, φ)
r2
1  
− 2 csc2 θP (t, r)∂φ2 Y (θ, φ) + P (t, r)∂θ2 Y (θ, φ) + cot θP (t, r)∂θ Y (θ, φ)
r

= 0

Separating these out and assuming that P (t, r)Y (θ, φ) = 0, we see that
 −1 2  
2 2m ∂t P (t, r) 2 2m ∂r2 P (t, r) ∂r P (t, r)
r 1− −r 1− − 2(r − m) =c
r P (t, r) r P (t, r) P (t, r)

and
2
∂φ2 Y (θ, φ) ∂θ Y (θ, φ) ∂θ2 Y (θ, φ)
csc θ + cot θ + =c
Y (θ, φ) Y (θ, φ) Y (θ, φ)

We begin by looking for solutions of the form Y (θ, φ) = Θ(θ)Φ(φ) and sub-
stituting into the above equation, we find that
  
1 Φ Θ + cot θΘ
2 + =c
sin θ Φ Θ

The solution to this equation is a well known one. The subject is known as
Spherical Harmonics. It has been studied extensively and yields a solution
of the form
(n)
(Aln cos(nφ) + Bln sin(nφ)) sinn θPl (cos θ)
where
n = 0, 1, . . . , l , l = 0, 1, 2, . . . ,
Aln , Bln are constants,

23
(n)
and Pl (cos θ) is the nth derivative of the Legendre polynomial Pl with re-
spect to cos θ.

So we get
 ∞ 
∞  
n (n)
Y (θ, φ) = (Aln cos(nφ) + Bln sin(nφ)) sin θPl (cos θ)
l=0 n=l

We now have to deal with P (t, r). Strictly speaking we should be work-
ing with Pln (r, t) but for convenience we shall drop the subscripts. So
 −1 2  
2 2m ∂t P (t, r) 2 2m ∂r2 P (t, r) ∂r P (t, r)
r 1− −r 1 − −2(r−m) = −n(n+1)
r P (t, r) r P (t, r) P (t, r)

Using another separation of variables by letting P (t, r) = W (t)Q(r) we find


that
  2  
W (t) 2m Q (r) 2(r − m)(r − 2m) Q (r) n(n + 1)(r − 2m)
= 1− + − = −ω 2
W (t) r Q(r) r3 Q(r) r3

This gives us
W (t) = eiωt
and
 2  
2m  2(r − m)(r − 2m)  2 n(n + 1)(r − 2m)
1− Q (r)+ Q (r)+ ω − Q(r) = 0
r r3 r3

R(r)
Letting Q(r) = r
, we find that

R(r) R (r)
Q (r) = − + ,
r2 r
R(r) R (r) R (r)
Q (r) = 2 − 2 +
r3 r2 r

24
Substituting these expressions for Q(r), Q (r), and Q (r) into the above equa-
tion we find that
 2 
(r − 2m)2  2m(r − 2m)  ω (r − 2m)
R (r)+ R (r)+ − (rn(n + 1) + 2m) R(r) = 0.
r3 r4 r r5

Now we make a substitution using ”tortoise” coordinates R(r) = T (r∗ ),


where  r 
r∗ = r + 2m ln −1
2m
we find that
r
R (r) = T  (r∗ ),
(r − 2m)
r2 2m
R (r) = T 
(r ∗ ) − T  (r∗ )
(r − 2m)2 (r − 2m)2

Again substituting these expressions for R(r), R (r), and R (r) into the above
equation, we find that
   
 2 2m rn(n + 1) + 2m
T (r∗ ) + ω − 1 − T (r∗ ) = 0
r r3

The importance of the variable r∗ arises from the fact that it ranges from −∞
to +∞ exhausts the entire part of space-time that is accessible to observers
outside the horizon.

This equation can be written in the form



T (r∗ ) + V (r)T (r∗ ) = 0,

where the potential V (r) is given by


  
2 2m rn(n + 1) + 2m
V (r) = ω − 1 −
r r3

= ω 2 − Ṽ (r)

25
3.3 Stability analysis of the perturbations
For each value of ω ∈ C, we have a solution P (t, r) satisfying

P (t, r) = W (t)Q(r)
Q(r)
= eiωt
r
T (r ∗)
= eiωt
r
where  r 
r∗ = r + 2m ln −1
2m
and T (r∗ ) satisfies the ordinary differential equation
 
T  (r∗ ) + ω 2 − Ṽ (r) T (r∗ ) = 0. (4)

Note: The effective potential Ṽ (r) is defined by


  
2m rn(n + 1) + 2m
Ṽ (r) = 1 − ,
r r3

and crucially for what follows, satisfies Ṽ (r) > 0.

Also, since  r 
r∗ = r + 2m ln −1
2m
we get that

r→∞ ⇒ r∗ → ∞
r → 2m ⇒ r∗ → −∞

Next we define

lim V (r) = V∞
r→∞

lim V (r) = V−∞


r→2m

26
We can now approximate the equation


T (r∗ ) + V (r)T (r∗ ) = 0,
by

T (r∗ ) + V±∞ T (r∗ ) = 0. (5)

The solutions to these equations describe the asymptotic behavior of the


solution at r = ∞ (spacial infinity) and r = 2m (the event horizon).

3.3.1 ω=0
We want to show that there are no time dependent solutions, i.e. ω = 0,
which are regular at both r = 2m and r = ∞.

If ω = 0, then both V∞ = 0 and V−∞ = 0.


So (5) becomes
T  (r∗ ) = 0
and we get the solutions

T (r∗ ) = Ar∗ + B (at r∗ = −∞)


T (r∗ ) = ar∗ + b (at r∗ = +∞).

Now we can only get regular solutions at r∗ = ±∞ if A = a = 0.


This gives us the solutions

T (r∗ ) = B (at r∗ = −∞)


T (r∗ ) = b (at r∗ = +∞).

Now from (4) with ω = 0, we find that

T  (r∗ ) = Ṽ (r)T (r∗ ).

Suppose that B > 0, then


T  (r∗ ) > 0

27
since both Ṽ (r) > 0 and T (r∗ ) > 0 for sufficiently large negative r∗ .

Now, as r∗ → −∞

T (r∗ ) → B
⇒ T  (r∗ ) → 0.

Hence T (r∗ ) increases away from B and so stays positive. Thus the condition
that T  (r∗ ) > 0 persists.
From this we can say that the function does not have an inflection point and
therefore cannot level off again. This implies that the function has no local
maximum and hence that T  (r∗ ) > 0.

But as r∗ → +∞

T (r∗ ) → b
⇒ T  (r∗ ) → 0.

r*

Figure 1: A function that starts with zero slope, increases, and the ends with
zero slope must have an inflection point.

We know that the function cannot start at B with zero slope, increase, and
then level off again with zero slope at b without having an inflection point, as
in Figure 1 and therefore there is a contradiction and this case can be ruled
out.

28
The where B < 0 can be ruled out in a similar fashion. The case where
B = 0 simply leads to the trivial solution T (r∗ ) ≡ 0 which we are not inter-
ested in.

This proves that no physical static solution exists which is regular at both the
event horizon (r = 2m) and at spatial infinity (r = ∞), i.e. that the black
hole cannot support an external static field which could perhaps destroy the
black hole nature of the space-time.

3.3.2 ω = −iα
We shall now show that equation (4) with ω = −iα, where α ∈ R, cannot
have solutions which are finite at both r∗ = −∞ and r∗ = +∞.

Note: This corresponds to P (t, r) = eiωt T (R


r
∗)
increasing exponentially in
time.

In this case V∞ = −α2 and V−∞ = −α2 , so by approximating



T (r∗ ) + V (r)T (r∗ ) = 0
using (5), we find that the solutions to
T  (r∗ ) − α2 T (r∗ ) = 0
describe the asymptotic behavior of the solution at r∗ = −∞ and r∗ = +∞.

The solutions to these equations are


T (r∗ ) = Aeαr∗ + Be−αr∗ (at r∗ = −∞)
T (r∗ ) = aeαr∗ + be−αr∗ (at r∗ = +∞)
Now we can only get regular solutions at r∗ = ±∞ if B = a = 0.
This gives us the solutions
T (r∗ ) = Aeαr∗ (at r∗ = −∞)
T (r∗ ) = be−αr∗ (at r∗ = +∞)
Now from (4) with ω = −iα, we find that
 
T  (r∗ ) = α2 + Ṽ (r) T (r∗ ).

29
Suppose that A > 0, then
T  (r∗ ) > 0
since α2 > 0, Ṽ (r) > 0 and T (r∗ ) > 0 for sufficiently large negative r∗ .

Now, as r∗ → −∞

T (r∗ ) → 0
⇒ T  (r∗ ) → 0.

Hence T (r∗ ) increases away from 0 and so stays positive. Thus the condition
that T  (r∗ ) > 0 persists.

From this we can say that the function does not have an inflection point
and therefore cannot level off again. This implies that the function has no
local maximum and hence that T  (r∗ ) > 0.

But as r∗ → +∞

T (r∗ ) → 0
⇒ T  (r∗ ) → 0.

r*

Figure 2: A function that starts with zero slope, increases to a maximum,


and then decreases to zero must have inflection points.

30
We know that the function cannot start at 0 with zero slope, increase to
a maximum, and then decrease to 0 with zero slope at 0 without having
inflection points, as in Figure 2 and therefore there is a contradiction and
this case can be ruled out.

The case where A < 0 can be ruled out in a similar fashion and the case
where A = 0 is simply the trivial case T (r∗ ) ≡ 0 which we are not interested
in.

This shows that a perturbation which is initially finite cannot undergo expo-
nential growth and hence is a good indicator of the stability of the solutions.

3.3.3 Time dependence


When we include a time dependence rather than factoring it out, the equa-
tions governing perturbations of the Schwarzschild black hole can be ex-
pressed in the form
∂2Z ∂2Z
= − Ṽ (r)Z, (6)
∂t2 ∂r∗2
where the function Z characterizes the perturbation, r∗ is the tortoise coor-
dinate and Ṽ (r) is a smooth positive potential which goes to zero at both
spacial infinity (r∗ = ∞) and the event horizon (r∗ = −∞).

We know that the complex conjugate Z̄ of Z must satisfy the same equation.
A useful bound on the growth of Z is attained as follows.

∂ Z̄
Multiplying (6) by ∂t
and rearranging gives us
∂ Z̄ ∂ 2 Z ∂ Z̄ ∂ Z̄ ∂ 2 Z
+ Ṽ Z =
∂t ∂t2 ∂t ∂t ∂r∗2
and integrating both sides
∞  ∞ 
∂ Z̄ ∂ 2 Z ∂ Z̄ ∂ Z̄ ∂ 2 Z
+ Ṽ Z dr∗ = dr∗ . (7)
−∞ ∂t ∂t2 ∂t −∞ ∂t ∂r∗2

An integration by parts of the right hand side of (7) yields


∞    ∞ 
∂ Z̄ ∂ 2 Z  ∂ Z̄ ∂Z r∗ =∞ ∂Z ∂ 2 Z̄

dr∗ =   − dr∗
−∞ ∂t ∂r∗2 ∂t ∂r∗ r∗ =−∞ −∞ ∂r∗ ∂t∂r∗

31
It can be shown that  
 ∂ Z̄ ∂Z r∗ =∞
  =0
 ∂t ∂r∗ 
r∗
.
This leaves us with
∞  ∞  
∂ Z̄ ∂ 2 Z ∂Z ∂ 2 Z̄
dr∗ = − dr∗
−∞ ∂t ∂r∗2 −∞ ∂r∗ ∂t∂r∗
and substituting this into (7) gives us
∞  ∞ 
∂ Z̄ ∂ 2 Z ∂ Z̄ ∂Z ∂ 2 Z̄
+ Ṽ Z dr∗ = − dr∗
−∞ ∂t ∂t2 ∂t −∞ ∂r∗ ∂t∂r∗
or ∞ 
∂ Z̄ ∂ 2 Z ∂Z ∂ 2 Z̄ ∂ Z̄
+ + Ṽ Z dr∗ = 0.
−∞ ∂t ∂t2 ∂r∗ ∂t∂r∗ ∂t

If we add this equation to its complex conjugate we get


∞      
∂ Z̄ ∂ 2 Z ∂Z ∂ 2 Z̄ ∂Z ∂ 2 Z̄ ∂ Z̄ ∂ 2 Z ∂ Z̄ ∂Z
+ + + + Ṽ Z + Z̄ dr∗ = 0
−∞ ∂t ∂t2 ∂t ∂t2 ∂r∗ ∂t∂r∗ ∂r∗ ∂t∂r∗ ∂t ∂t

or      

∂ ∂Z ∂ Z̄ ∂ ∂Z ∂ Z̄ ∂  
+ + Ṽ Z Z̄ = 0.
−∞ ∂t ∂t ∂t ∂t ∂r∗ ∂r∗ ∂t

Noticing that xx̄ = |x|2 , the above equation becomes


∞  2  

∂  ∂Z   ∂Z 2
  2
 ∂t  +  ∂r∗  + Ṽ |Z| dr∗ = 0.
−∞ ∂t

Now assuming that the derivatives are continuous, we can take the derivative
outside of the integral to give
    

∂ ∞  ∂Z 2  ∂Z 2
  +  2
∂t −∞  ∂t   ∂r∗  + Ṽ |Z| dr∗ = 0.

and integrating this with respect to t we have our bound on the integral
∞    

 ∂Z 2  ∂Z 2
    2
 ∂t  +  ∂r∗  + Ṽ |Z| dr∗ = C
−∞

32
where C is some constant.
 
∞  ∂Z 2
⇒  
 ∂t  dr∗ < C, since Ṽ > 0
−∞
 
 ∂Z 
⇒  
 ∂t  is bounded

The bound on the time derivative of Z implies that Z cannot grow faster
than a linear function of time. This is a very good indicator that the field
remains finite for all times t > 0.

33
Bibliography
[1] D’Inverno R. (1992). Introducing Einstein’s Relativity, Oxford Univer-
sity Press.

[2] Chandrasekhar S. (1983). The Mathematical Theory of Black Holes, Ox-


ford University Press.

[3] Hawking, S. W. and Ellis, G. F. R. (1973). The large scale structure


of space-time, Cambridge University Press.

[4] Boyce, W. E. and DiPrima, R. C. (1986). Elementary Differential Equa-


tions and Boundary Value Problems, John Wiley & Sons, Inc.

[5] Wald, R. M. (1978). Note on the stability of the Schwarzschild metric,


Enrico Fermi Institute, University of Chicago.

34

You might also like