Anda di halaman 1dari 43

RELATIVITY & GRAVITATION

Lectured by Timothy Clifton

SECTION 3 - GRAVITATIONAL WAVES

AND PERTURBATION THEORY

3.1 Perturbation theory

Perturbation theory is a widely used tool for considering physical scenar-

ios that are more complicated than a single body. The idea is to determine

which parts of the geometry can be neglected, when gravity is weak, and

then systematically ignoring them. Our starting point is

gµν = ηµν + hµν

where |hµν |  1, which means we will be considering spacetimes that are

close to Minkowski. This approximation is sufficient to describe everything

in the Solar System, and even the gravitational field of black holes (as long

as we stay far away from their horizons).

1
Exercise: show that the condition g µν gνσ = δ µ σ means that

g µν = η µν − hµν + O(h2 )

where hµν ≡ η µρ η νσ hρσ , and where O(h2 ) indicates (small) terms that are

of size h2 (or smaller)

solution: if gµν = ηµν + hµν + O(h2 ), then we should be able to write

g µν = η µν + f µν + O(h2 ), where f µν ∼ h.

⇒ δ µ σ = g µν gνσ = (η µν + f µν )(ηνσ + hνσ ) + O(h2 )

= η µν ηνσ + η µν hµν + f µν ηνσ + O(h2 )

= δ µ σ + η µν hνσ + f µν ηνσ + O(h2 )

⇒ f µν ηνσ = −η µν hνσ

Now multiply through by η σρ :

⇒ f µν ηνσ η σρ = f µν δν ρ = f µρ = −η µν η σρ hνσ

⇒ g µν = η µν − hµν

where hµν = η µσ η νρ hσρ .

2
Substituting these expressions into the definition of the Christoffel sym-

bols gives

Γσ µν = Γ(1)σ µν + Γ(2)σ µν + O(h3 )

where numbers in brackets indicate order-of-smallness in h. It is straight-

forward to find

1
Γ(1)σ µν = η σρ (∂nu hρµ + ∂µ hρν − ∂ρ hµν )
2

and
1
Γ(2)σ µν = − hστ (∂ν hτ µ + ∂µ hτ ν − ∂τ hµν )
2

Similarly, these can be substituted into the definition of Rµν to find

Rµν = R(1) µν + R(2) µν + O(h3 )

where
1
R(1) µν = (∂µ ∂ρ hρ µ + ∂ρ ∂µ hρ ν − ∂µ ∂ν hρ ρ − hµν )
2

3
and

1
R(2) µν = hρσ (∂µ ∂ν hρσ + ∂ρ ∂σ hµν − ∂µ ∂σ hνρ − ∂ν ∂σ hνρ )
2
1 1
+ (∂σ hρσ − ∂ ρ hσ σ )(∂ρ hµν − ∂µ hνρ − ∂ν hµρ )
2 2
1 1
+ ∂µ hρσ ∂ν hρσ + ∂ σ hρ ν (∂σ hρµ − ∂ρ hσµ )
4 2

The indices in these last equations have been raised using η µν (e.g. hρ µ ≡

η ρν hνµ ). We have also used  ≡ ∂ µ ∂µ .

We can now use these expressions to get perturbative approximations

to the field equations. For example, to leading order

R(1) = η µν R(1) µν = ∂µ ∂ν hµν − hµ µ

1
⇒ G(1) µν = R(1) µν − ηµν R(1) = 8πGTµν + O(h2 )
2

⇒ ∂ν ∂ν hρ µ + ∂ρ ∂µ hρ ν − ∂µ ∂ν hρ ρ − hµν + ηµν (hρ ρ − ∂ρ ∂σ hρσ ) = 16πGTµν + O(h2 )

This equation is the one we will now try and solve. It can be simplified by

defining
1
h̄µν ≡ hµν − ηµν hρ ρ
2

⇒ h̄µν + ηµν ∂ρ ∂σ h̄ρσ − ∂ν ∂ρ h̄ρ ν = −16πGTµν + O(h2 )

4
Example: show that h̄µν is the trace-reverse of hµν , so that

1
h̄ρρ = −hρρ and hµν = h̄µν − ηµν h̄ρρ .
2

Solution: if h̄µν = hµν − 21 ηµν hρρ then

h̄ρρ = h̄µν η µν

1
= hµν η µν − ηµν η µν hρρ
2
1
= hρρ − δµ µ hρρ
2
1
= hρρ − × 4 × hρρ
2

= hρρ − 2hρρ

= −hρρ

and

1 1 1
h̄µν − ηµν h̄ρρ = hµν − ηµν hρρ − ηµν (−hρρ )
2 2 2

= hµν

So the proposed equations are true.

5
3.2 Gauge transformations

The general covariance of Einstein’s equations is broken when we start

approximating them using perturbation theory. What is left is covariance

under “small” coordinate transformations:

x0µ = xµ + ξ µ

where |ξ µ | ∼ |hµν | (where ∼ means they are approximately the same size

and ξ µ is a quantity known as a gauge generator). These are called “gauge

transformations”, and the coordinate freedom that existed in the full the-

ory is replaced by the “gauge freedom” to choose ξ µ

Differentiating this gives the gauge transformation matrix

∂x0µ
ν
= δ µ ν + ∂ν ξ µ
∂x

∂x0µ ∂xν
Requiring ∂xν ∂x0ρ = δ µρ then gives the inverse:

∂xµ

= δ µ ν − ∂ν ξ µ + O(h2 )
∂x

These correspond to infinitesimal coordinate transformations.

Let us now perform a gauge transformation on the metric, using the

6
infinitesimal coordinate transformation above:

0 ∂xρ ∂xσ
gµν = 0µ 0ν gρσ = (δ ρ µ − ∂µ ξ ρ )(δ σ ν − ∂ν ξ σ )(ηρσ + hρσ )
∂x ∂x

= ηµν + hµν − ∂µ ξν − ∂ν ξµ + O(h2 )

where we have written ξµ = ηµν ξ ν . This can be compared to the pertur-

0
bative expression for gµν , in the new coordinate system:

0
gµν = ηµν + h0µν

⇒ h0µν = hµν − ∂µ ξν − ∂ν ξµ + O(h2 )

This equation gives the effect that a gauge transformation has on metric

perturbations. It is extremely useful for simplifying calculations in pertur-

bation theory.

7
3.3 The Lorenz gauge

The trace-reversed metric perturbation gauge transforms as follows:

h̄0µν = h̄µν − ∂ µ ξ ν − ∂ ν ξ µ + η µν ∂ρ ξ ρ

This means that

∂ν h̄0µν = ∂µν h̄µν − ξ µ

If we now choose ξ µ = ∂ν h̄µν then

∂ν h̄0µν = 0

This choice of ξ µ results in the “Lorenz gauge”. It’s extremely useful

because it simplifies the leading-order perturbed field equations to

h̄0µν = −16πGTµν

Note: the Lorentz gauge is preserved by any addition gauge transforma-

tion, provided that ξ µ satisfies ξ µ = 0.

8
Exercise: prove that h̄µν transforms as

h̄0µν = h̄µν − ∂ µ ξ ν − ∂ ν ξ µ + η µν ∂ρ ξ ρ

Solution: under the transformation x0µ = xµ + ξ µ we have

h0µν = hµν − ∂µ ξν − ∂ν ξµ + O(h2 )

now

1 1
h̄0µν = h0µν − ηµν h0ρ ρ = hµν − ∂µ ξν − ∂ν ξµ − ηµν (hρ ρ − 2∂ρ ξ ρ )
2 2

= h̄µν − ∂µ ξν − ∂ν ξµ + ηµν ∂ρ ξ ρ

Raising indices with η µσ and η νρ then gives the required result.

9
Example: we can linearize the Schwarzschild solution:

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

2Gm
If we assume r  1 then

   
2 2Gm 2 2Gm
dr2 + r2 (dθ2 + sin2 θdφ2 ) + O h2

ds ' − 1− dt + 1 +
r r
2Gm 2
= −dt2 + dr2 + r2 (dθ2 + sin2 θdφ2 ) + (dt + dr2 ) + O h2

r

= ηµν dxµ dxν + hµν dxµ dxν + O(h2 ) ,

so that
2Gm 2
⇒ hµν dxµ dxν = (dt + dr2 ) .
r

This perturbation is traceless (η µν hµν = 0), so hµν = h̄µν .

Note: this perturbation is not in Lorenz gauge, as

 
2Gm
∂t h̄tt = ∂t =0
r

but
 
rr 2Gm 2Gm
∂r h̄ = ∂r =− 6= 0 .
r r2

10
3.4 Linearised solutions in vacuum

In vacuum Tµν = 0, and the field equations in the Lorenz gauge reduce

to

h̄µν = 0

with

∂µ h̄µν = 0

A solution to these equations can be written as

h̄µν = Aµν exp(ikρ xρ )

where Aµν is a symmetric matrix of complex, constant values, and where

kµ are the constant, real components of a vector. We must take the real

part of h̄µν if we want a real spacetime, of course.

Exercise: show that the expression for h̄µν is a solution to h̄µν = 0 if

k µ kµ = 0

and that it satisfies the gauge condition if

Aµν kν = 0

11
Solution: if h̄µν = Aµν exp(ikρ xρ ) then

∂ ∂
h̄µν = η αβ α β
Aµν exp(ikρ xρ )
∂x ∂x

= η αβ ikα kβ Aµν exp(ikρ xρ )

= −k α kα h̄µν

= 0 if k µ kµ = 0

and

∂µ h̄µν = ∂µ (Aµν exp(ikρ xρ ))

= ikµ Aµν exp(ikρ xρ )

= 0 if kµ Aµν = 0 .

The linearised Einstein equations and Lorenz gauge condition are therefore

satisfied if k µ kµ = 0 and kµ Aµν = 0.

12
3.5 Transverse-traceless gauge

To investigate gravitational waves it is often useful to specialise to the

“transverse-traceless” gauge. Recall that in Lorenz gauge there existed

the residual gauge freedom

x0µ = xµ + ξ µ

where ξ µ = 0. This freedom can be used to enforce the additional con-

ditions

h̄TT
0i = 0 and h̄TTµ µ = 0

as well as the usual Lorenz condition ∂µ h̄µν = 0, which now becomes

∂t h̄00
TT = 0 and ∂i h̄ij
TT = 0

because of the first of the conditions above. This uses up all of our gauge

freedoms. If we now consider the linearised solution in vacuum, h̄µν =

Aµν exp(ikρ xρ ), then these four equations imply

ij
A0i µ 00
TT = ATT µ = ATT = ATT kj = 0 .

If these conditions are satisfied then we are in T T gauge.

13
As an example, let’s take k µ to point in the z-direction

k µ = (ω, 0, 0, k)

The conditions k µ kµ = 0 and Aµν k ν = 0 then imply ω = −k and Aµ3 =

Aµ0 . This gives


 
A00 A01 A02 A03 
 
 
 
A01 A11 A12 A10 
Aµν =
 

 
A02 A12 A22 A02 
 
 
 
A00 A01 A02 A00

A gauge generator that satisfies ξ µ = 0 is then given by

ξ µ = εµ exp(ikρ xρ )

where εµ is a constant vector, and k µ is the same vector as above. The


components of this transformation are

A000 = A00 − ik(ε0 + ε3 ), A012 = A12

A011 = A11 − ik(ε0 − ε3 ), A001 = A01 − ikε1

A022 = A22 − ik(ε0 − ε3 ), A002 = A02 − ikε2

14
If we therefore make the choices

i i
ε0 = − (2A00 + A11 + A22 ), ε1 = − A01
4k k

i i
ε2 − A02 , ε3 = − (2A00 − A11 − A22 )
k 4k

then we have A000 = A001 = A002 = 0, and

1 1
A011 = −A022 = A11 − A22
2 2

This gives  
0 0 0 0
 
 
A011 A012
 
0 0
⇒ A0µν =
 

 
0
 A012 −A011 0

 
 
0 0 0 0

and the perturbation to the metric is then


 
0 0 0 0
 
 
 
0 h+ hx 0
h̄TµνT = hTµνT =
 

 
0 hx h+ 0
 
 
 
0 0 0 0

where h+ ≡ A11 exp(ikρ xρ ) and h× ≡ A12 exp(ikρ xρ ).

15
Another (often simpler) way to put Aµν into the T T gauge is to introduce

the “projection tensor”

Pij = δij − ni nj

where ni obeys ni ni = 1 and points in the direction of propagation of the

gravitational wave. We will use this result in Section 3.12 to calculate the

energy that gravitational waves remove from binary systems.

Example: show that the following corresponds to a perturbation in TT

gauge:
 
1
Aij
TT = P ik P jl − P ij Pkl Akl
2

where Pij = δij − ni nj is the projection tensor, and ni is a space-like unit

vector that points in the direction of propagation of the gravitational wave.

Solution: we still want h̄µν = Aµν


ρ
ikρ x
TT e to satisfy Einstein’s equations

and the Lorenz gauge condition, so require kµ Aµν


TT = 0.

⇒ A0µ
TT kµ = 0 where k µ = (ω, ωni )

⇒ −A00 0i
TT k0 + ATT ωni = 0

⇒ A00 0i
TT = ATT ni (1)

16
and

Aiµ
TT kµ = 0

ij
⇒ −Ai0
TT ω + ATT ωnj = 0

ij
⇒ Ai0
TT = ATT nj (2)

Equations (1) and (2) together give

ij
A00
TT = ATT ni nj .

Now calculate P ij ni = (δ ij − ni nj )ni = nj − nj = 0, which gives

⇒ Aij
TT nj = 0 and Aij
TT ni nj = 0

so Ai0
TT = 0 and A00
TT = 0 .

Next we need to evaluate ATTµµ and Aij


TT kj .

If

A00
TT = 0 then ATTµµ = ATTii ,

but
1
ATTii = (P ik Pil − P ii Pkl )Akl ,
2

17
where

P ik Pil = (δ ik − ni nk )(δil − ni nl )

= δlk − nk nl − nl nk + nk nl

= δlk − nk nl

= Plk

and

1 1
− P ii Pkl = − (δ ii − ni ni )Pkl
2 2
1
= − (3 − 1)Pkl
2

= −Pkl

so ATTii = 0, which implies ATTµµ = 0. Finally,

1 ij
Aij i j
TT kj = (P k P l − P Pkl )A kj
kl
2
1
= ω(P ik P jl − P ij Pkl )Akl nj
2

⇒ Aij
TT kj = 0 as P ij nj = 0 .

The four boxed equations are the conditions to be in TT gauge, so the

proposition is true.

18
3.6 The effect of gravitational waves

Let’s keep working in TT gauge, and consider the effect of our plane

gravitational wave on a test particle with 4-velocity uµ = (1, 0, 0, 0). The

geodesic equation then gives

duµ 1
= −Γµ ρσ uρ uσ = − η µν (∂t hµt + ∂t htµ − ∂ν htt ) + O(h2 )
dτ 2

Recall that in TT gauge hTT


νt = 0 for all time. This means

duµ
=0

i.e. the particle stays at fixed spatial coordinates as the gravitational wave

passes though. This is an important result, but it does not mean that the

gravitational wave has no effect (as we will now see).

Consider two test particles, separated by coordinate distance ∆x in the

x-direction. The proper distance between them is therefore

Z∆xp Z∆xp Z∆x 


1
Lx ≡ gxx dxdx = ηxx + hxx dx = 1 + hxx dx + O(h2 )
2
0 0 0

Z∆x   
A11 A11
= 1+ cos(kz − ωt) = 1 + cos(kz − ωt) ∆x
2 2
0

19
Likewise, two particles separated by coordinate distance ∆y in the y-

direction have a proper distance between them of

 
A22
Ly = 1 − cos(kz − ωt) ∆y
2

And a proper distance in the z-direction of

Z∆z 
1
Lz = 1 + hzz dz = ∆z
2
0

The effect of the h+ polarisation is therefore to increase/decrease the sepa-

ration between test particles in an oscillatory way, in the directions trans-

verse to propagation direction of the wave.

If you imagine a gravitational wave coming upwards, out of the page, the

consequences of the h+ polarisation on a ring of test particles is therefore

There is stretching and squashing in the x and y-directions, so that the

ring turns into an ellipsoid and back again. The stretching and squashing

continues until the wave has passed.

What about the hx polarisation? A gravitational wave can either be h+

or hx polarized, or a mixture of both

20
Exercise: By rotating hTT
µν using the rotation matrix below

 
1 0 0 0
 
 
 
∂x µ 0 cos θ − sin θ 0
=
 
∂x0ν 
 

0 sin θ cos θ 0
 
 
 
0 0 0 1

and with θ = π4 , show that the effect of the h× polarisation on a ring of

particles is given by the image in the figure below.

21
The effect of the h+ and h× polarisations on a ring of test particles is

therefore given in the following figure:

22
3.7 Linearised solutions with sources

Let us now return to the general equation

h̄µν = −16πGTµν

where Tµν represents an arbitrary distribution of matter. The solution to

this equation can be written as

Tµν (t − |x̄ − ȳ|, ȳ) 3


Z
h̄µν (t, x̄) = +4G dy
|x̄ − ȳ|

where x̄ is the spatial position where h̄µν is being evaluated, ȳ is a point

within the source of the gravitational field, and |x̄ − ȳ| is the distance

between them.

It can be seen that the gravitational perturbation h̄µν is only sourced

by matter that intersects the light cone at the field point (t, x̄). This is a

very important point: in GR gravitational interactions and disturbances

propagate only at the speed of light.

23
3.8 The quadrupole formula

The quadrupole formula is a very useful result for gravitational wave

physics. to derive it we start by making the compact wave source approx-

imation:

|ȳ|  |x̄| ⇒ |x̄ − ȳ| ≈ |x̄|

Our general linearised solution is then

Z
4G
h̄µν (t, x̄) ≈ + Tµν (t − r, ȳ)d3 y
r

where r ≡ |x̄| is the distance between the source and the observer. If we

further assume that the source of the gravitational field is moving much

slower than light, |v̄|  c, then the integral above corresponds to

Z Z
3
T00 d y ≈ T̄ (ū, ū)d3 y ≡ M

Z Z
3
T0i d y ≈ T̄ (ū, x̄i )d3 y ≡ −Qi
Z Z
3
Tij d y ≈ T̄ (x̄i , x̄j )d3 y ≡ Πij

If we now choose our frame of reference to be the centre-of-momentum

24
frame, where Qi = 0, then we are left with

4GM
h̄00 = +
r

4GΠij
h̄ij = +
r

h̄0i = h̄i0 = 0

To go further requires manipulating the stress-energy tensor. Taking the

leading-order parts of the conservation equations gives

∂t T 00 + ∂i T 0i = 0 (3)

∂t T i0 + ∂j T ij = 0 (4)

Let us now consider the integral

Z Z Z
∂k (T ik y j )d3 y = (∂k T ik )y j d3 y + T ij d3 y

Using Gauss’ divergence result, the integral on the LHS can be trans-

formed to a surface integral over the boundary of the original domain of

integration. If the domain of integration is larger than our compact source

(which it is) then it must vanish, as Tµν is only non-zero inside the source.

25
This means
Z Z
ij 3
T d y=− (∂k T ik y j d3 y)

then using equation (2)

Z Z Z
ij 3 i0 d
j 3
T dy= (∂t T )y d y = T i0 y j d3 y
dt

Similarly, exchanging i and j indices, gives

Z Z
d
ji 3
T dy= T j0 y i d3 y
dt

so

Z Z
1d
ij 3
T dy= (T i0 y j + T j0 y i )d3 y (5)
2 dt

Now, consider a new integral

Z Z Z
0k i j 3 0k i j 3
∂k (T y y )d y = (∂k T )y y d y + (T 0i y j + T 0j y i )d3 y

Z Z
00 i j 3
=− (∂t T )y y d y + (T 0i y j + T 0j y i )d3 y
Z Z
d 00 i j 3
=− T yy d y+ (T 0i y j + T 0j y i )d3 y
dt

26
Once again, the LHS of this equation can be set to zero using Gauss’ result.

This gives

Z Z
0i j 0j i d3
(T y + T y )d y = T 00 y i y j d3 y (6)
dt

Now, substituting equation (4) into equation (3) gives

1 d2
Z Z
ij 3
T dy= T 00 y i y j d3 y
2 dt2

The LHS of this equation is identical to our integral pressure term Πij

(after lowering indices with ηij ). We can therefore write h̄ij as

2G d2 Iij
h̄ij = +
r dt2

R
where Iij = ρy i y j d3 y is the quadrupole moment tensor. The above equa-

tion is the “quadrupole formula”. It forms the basis of much of gravita-

tional wave physics.

27
3.9 Static sources and Newtonian limit

In the centre of momentum frame, if the source of the gravitational field

is static (not moving) then


d2 Iij
=0
dt2

The only non-vanishing part of h̄µν is then

4GM
h̄00 = +
r

This means h̄µ µ = − 4GM


r , and therefore

1 2GM
h00 = h̄00 − η00 h̄µ µ = +
2 r

1 2GM
hij = h̄ij − ηij h̄µ µ = + δij
2 r

2
 2GM  2  2GM  2
⇒ ds = − 1 − dt + 1 + (dx + dy 2 + dz 2 )
r r

This line-element captures enough to describe Newtonian gravity, and the

leading-order part of the bending of light. It can also be derived by as-

suming that the gravitational field’s source is dominated by its rest mass,

i.e.

|T00 |  |T0i | and |T00 |  |Tij |

28
3.10 Waves from binary systems

Consider two massive bodies, in a circular orbit around each other of

radius R:

By rotating spatial coordinates we can arrange for these bodies to orbit

in the plane x = 0 (as above). The leading order part of the gravitational

field is Newtonian, so this orbit obeys

M v2 GM 2
=
R (2R)2

If we now write v = ΩR, then we get angular speed

r
GM
Ω=
4R3

29
The positions of the two bodies can also be written

xiA = 0, R cos(Ωt), R sin(Ωt)




and

xiB = 0, −R cos(Ωt), −R sin(Ωt)




The density of this system can therefore be written as

ρ − M δ(x)[δ(y − R cos Ωt)δ(z − R sin Ωt) + δ(y + R cos Ωt)δ(z + R sin Ωt)]

Exercise: substitute this expression for ρ into I ij =


R
ρy i y j d3 y to find

 
0 0 0 
 
I ij = M R2 
 
0 1 + cos 2Ωt sin 2Ωt  
 
 
0 sin 2Ωt 1 − cos 2Ωt

30
Solution: we immediately have

Z
xx
I = ρx2 dxdydz = 0

Z
xy yx
I =I = ρxydxdydz = 0
Z
xz zx
I =I = ρxzdxdydz = 0
Z Z
yy 2
 2
y δ(y − R cos Ωt) + y 2 δ(y + R cos Ωt) dy

I = ρy dxdydz = M

= M (R2 cos2 Ωt + R2 cos2 Ωt) = M R2 (1 + cos 2Ωt)


Z Z
zz 2
 2
z δ(z − R sin Ωt) + z 2 δ(z + R sin Ωt) dz

I = ρz dxdydz = M

= M (R2 sin2 Ωt + R2 sin2 Ωt) = M R2 (1 − cos 2Ωt)


Z
yz zy
I =I = ρyzdxdydz
Z
=M yzδ(y − R cos Ωt)δ(z − R sin Ωt)dydz
Z
+M yzδ(y + R cos Ωt)δ(z + R sin Ωt)dydz


= M (R cos Ωt · R sin Ωt) + M (−R cos Ωt) · (−R sin Ωt)

= 2M R2 cos Ωt sin Ωt = M R2 sin 2Ωt

31
This all can be written as follows:
 
0 0 0 
 
⇒ I ij = M R2 
 
0 1 + cos 2Ωt sin 2Ωt 
 
 
0 sin 2Ωt 1 − cos 2Ωt

Substituting into the quadrupole formula gives


 
0 0 0 
8GM R2 Ω2 
 

h̄ij = − 0 cos 2Ω(t − r) sin 2Ω(t − r) 
r 



 
0 sin 2Ω(t − r) − cos 2Ω(t − r)

3.11 Gravity of gravitational waves

Gravitational waves do not have any local energy or momentum (i.e.

they do not contribute to Tµν ). However, they do have their own gravita-

tional field. This requires some careful thought to be properly understood.

Recall that in section 3.1 we wrote down (but did not yet use) Rµν to

order h2 . If we do the same with the Einstein tensor then we can write

Gµν = G(1) (2)


µν + Gµν + . . . = 8πGTµν

32
(2)
If we move Gµν to the RHS we have

G(1) 3
µν = 8πG(Tµν + tµν ) + O(h )

1 (2)
where tµν ≡ − 8πG Gµν . When written in this form we see that tµν acts as

a source term for the leading order part of the gravitational field (albeit a

small one). The second-order part of the Einstein tensor is given explicitly

by
1
G(2) (2) ρσ (2) ρσ (1) ρσ (1)
µν = Rµν − (ηµν η Rρσ + hµν η Rρσ − ηµν h Rρσ )
2

(1) (2)
where Rµν and Rµν are given in section 3.1. This quantity is not by itself a

tensor, as can be verified by trying to perform a coordinate transformation.

However, it can be made into a tensor by integrating (or smoothing) it over

a small region of spacetime. This gives

htµν i = (∂µ h̄ρσ )∂ν h̄ρσ − (∂σ h̄ρσ )∂µ h̄νρ


32πG
1
−(∂σ h̄ρσ )∂ν h̄µρ − (∂µ h̄)(∂ν h̄)

2
1

− 2h̄ρν T ρ ν + 2h̄ρν T ρ ν + ηµν hρσ Tρσ



4

where h. . .i denotes the smoother quantity, and where use has been made

of the result h∂ν =i = 0 for any function = = =(xµ ).

33
3.12 Energy radiated from a binary

We can now use htµν i to work out the rate at which binary systems lose

energy through the emission of gravitational waves. In vacuum, and in TT

gauge, we get
1
htµν i = h(∂µ hTρσT )(∂ν hT T ρσ )i
32πG

The energy flux in gravitational waves is therefore

qiGW = −ht0i i

and the rate of energy loss from the system emitting them is

I
dE
=− r2 qiGW r̂i dΩ
dt

where the integration is over a sphere that contains the system at the

centre, and where r̂i are the spatial components of an outward pointing

radial unit vector (such that r̂i r̂i = 1).

We will now use the TT part of the quadrupole formula

ij
1 ij 2G d2 I kl 2G d2 ITT
hij
TT = h̄ij
TT
i j
= (P j P l − P Pkl ) =
2 r dt2 r dt2

ij
where ITT ≡ (P i k P j l − 12 P ij Pkl )I kl . This gives

34
2G ...ij
∂t hij
TT = I
r TT

and
2G ¨ij 2G ...ij 2G ...ij
∂r hij
TT = − I − I ≈ − I
r2 TT r T r TT

Substituting this all back into the equations above gives

G ...T T ...ij
I
dE
=− h I I idΩ
dt 8π ij T T

Finally, the term in brackets can be expanded as

...TT ...ij ... ...ij ... j ...ik 1 ...ij ...kl


I ij I TT = I ij I − 2 I i I r̂j r̂k + I I r̂i r̂j r̂k r̂l
2

Using the known results

I I

dΩ = 4π, r̂i r̂j dΩ = δij
3
I

r̂i r̂j r̂k r̂i dΩ = (δij δkl + δik δjl + δil δjk )
15

then gives
dE G ... ...ij
= − h I ij I i
dt 5

35
Example: consider the binary system from section 3.10, for which

 
0 0 0 
 
...ij 2 3
 
I = 8M R Ω 0 sin 2Ωt − cos 2Ωt
 
 
0 − cos 2Ωt − sin 2Ωt

This gives
... ...ij
⇒ I ij I = 128M 2 R4 Ω6

dE 128
⇒ =− GM 2 R4 Ω6
dt 5

This result is very important for binary pulsar observations, and the recent

gravitational wave detection by LIGO.

3.13 The Hulse-Taylor binary

In 1993 Russell Hulse and Joseph Taylor were awarded the Nobel Prize

in Physics, for their work on the binary system PSR B1913+16. This

was a system of two neutron stars, one of which was a pulsar, that allowed

evidence for the existence of gravitational waves to be inferred. To consider

why and how, let’s return to our two bodies of mass M in circular orbit.

36
The total energy of the system is

GM 2 GM 2
 
1 2
E =K +U =2× Mv − =−
2 2R 4R

GM
as we know v 2 = 4R from before. Using v = ΩR and differentiating

dE 2Ω̇  G2 M 5 1/3
⇒ =−
dt 3 16Ω

From the previous section we now know that the emission of gravitational

waves from such a system causes

dE 128 2 6 GM
 4/3
=− GM Ω
dt 5 4Ω2

Equating these two expressions gives the change in angular velocity due

to GW emission:
48
Ω̇ = × 22/3 × (GM )5/3 Ω11/3
5

which can be integrated to give

128
⇒ Ω−8/3 = × 22/3 × (GM )5/3 (t0 − t)
5

where t0 is an integration constant (the time when Ω → ∞, when the two

bodies eventually coalesce).

37
Now the period of the orbit is

2π 231/8
τ= = 3/8 π(GM )5/8 (t − t0 )3/8
Ω 5

This result gives the rate at which our systems period decreases due to

energy lost through gravitational radiation, τ ∝ (t0 − t)3/8 . This is a

very good match to the observations made by Hulse and Taylor for PSR

B1913+16, and is widely considered to be the first indirect evidence for

the existence for gravitational waves.

38
3.14 The LIGO detections

The first direct detection of GWs was made by the LIGO experiment,

on the 14th of September 2015. This experiment consists of two interfer-

ometers at two different locations in the USA

The idea is this: when a gravitational wave passes, the two arms of the

detectors change length by a small amount. This causes a change in the

interference pattern at the detector. This sounds simple, but gravitational

waves tend to have very low amplitude (the 2015 detection caused the

arms to change length by about 10−18 m). To make a positive detection

therefore a very careful experimentation, and some knowledge of the signal

that is expected.

39
Let’s return to our two bodies in a circular orbit. At the end of section

3.10 we found an explicit expression for h̄ij , for the emitted gravitational

waves. Now, because we have an explicit expression for Ω = Ω(t) in section

3.13, we can work out what an observer at some position r on the z-axis

should be expected to see with his/her gravitational wave detector.

Example: use the results from Section 3.5 of the notes to show that

waves travelling in the z-direction from this system can be writte in TT

gauge as
 
0 0 0 0
 
 
 
(GM )
5
3 2
0 cos 2Ω(t − r) 0 0
h̄TT = (2Ω) 
3
 
µν
r

 
0
 0 − cos 2Ω(t − r) 0
 
 
0 0 0 0

and hence correspond to a wave with + polarization and

5
(GM ) 3 2
h+ = (2Ω) 3 cos 2Ω(t − r).
r

40
Solution: for the system in question we have
 
0 0 0

8GM R2 Ω2 
 

h̄ij = − 0 cos 2Ω(t − r) sin 2Ω(t − r) 
r 



 
0 sin 2Ω(t − r) − cos 2Ω(t − r)

l 1 j
To put this in TT gauge recall h̄TT k kl
ij = (Pi Pj − 2 Pij P )h̄kl , where Pi =

δi j − ẑi ẑ j and ẑ i = (0, 0, 1).  


0 0 0
8GM R2 Ω2 
 
k l

⇒ Pi Pj h̄kl = − 0 cos 2Ω(t − r) 0
r 



 
0 0 0

2
Ω2
and P kl h̄kl = − 8GMrR cos 2Ω(t − r)
 
1
 2 cos 2Ω(t − r) 0 0
8GM R2 Ω2 
 
1 kl

⇒ − Pij P h̄kl =  0 1
cos 2Ω(t − r) 0
2 r 
 2 

 
0 0 0

 
cos 2Ω(t − r) 0 0
4GM R2 Ω2 
 
h̄TT

⇒ ij =  0 − cos 2Ω(t − r) 0
r 



 
0 0 0

GM GM
 23
This gives the desired result when we use Ω2 = 4R2 , or R2 = 4Ω2 .

41
Now, because we have an explicit expression for Ω = Ω(t) in Section

3.13, we can work out what an observer at some position r on the z-axis

should be expected to see with his/her gravitational wave detector.

Exercise: produce some plots of h+ = h+ (t) to see what a gravitational

wave signal looks like.

You should get something that looks a bit like this:

The frequrency and amplitude of the wave increases as the bodies come

together, at t = t0 . This is exactly what was seen by LIGO: two merging

black holes, each with about thirty times the mass of the Sun, as at a

distance of about 1.4 billion light years. The actual signal from the first

LIGO detection is shown on the next page, for the two detectors at Hanford

and Livingston.

42
43

Anda mungkin juga menyukai