PHY7006 (2014) Page 1

————————————————————————————————————

GENERAL RELATIVITY AND GRAVITATION

Revision Lecture, 2014

————————————————————————————————————

Turn over
Page 2 PHY7006 (2014)

1 The Equivalence, Geometrical and Covariance Principles.

The Equivalence Principle states that an uniform gravitational field is equivalent to a uniform
acceleration of reference frame.
The General Relativity gives very simple and natural explanation of the Principle of Equivalence:
In curved space-time all bodies move along geodesics, that is why their world lines are the
same in given gravitational field. The situation is the same as in flat space-time when free
particles move along straight lines which are geodesics in flat space-time. The fundamental
physical concept of GR is that a gravitational field is identical to geometry of curved space-time.
This idea, called the Geometrical Principle, entirely determines the mathematical structure of
General Relativity. According to the GR gravity is nothing but a manifestation of space-time
4-geometry. The geometry is determined by metric
ds2 = gik (xm )dxi dxk , (1)
where gik (xm ) is called the metric tensor.
The Covariance Principle says: The shape of all physical equations should be the same in an
arbitrary frame of reference, including the most general case of non-inertial frames. If in contrast
to the covariance principle the shape of physical equations were different in local inertial frames
in presence of gravitational field and in non-inertial frames in absence of gravitational field then
these equations would give different solutions, i.e. different predictions for (a) standing on the
Earth, feeling the effects of gravity as a downward pull and (b) standing in a very smooth elevator
that is accelerating upwards with the acceleration g , hence these equations would contradict
to the basic postulate of the General Relativity, the principle of equivalence, which states that a
uniform gravitational field (like that near the Earth) is equivalent to a uniform acceleration. Hence,
the covariance principle is the mathematical formulation of the principle of equivalence.
The important consequence of the equivalence principle is that any gravitational field can be
eliminated in free falling frames of references, which are called local inertial frames or local
galilean frames. From geometrical point of view to eliminate gravitational field locally means to
find such frame of reference in which
gik → ηik ≡ diag(1, −1, −1, −1). (2)
The basic postulate of the GR states that a uniform gravitational field is equivalent to (which
means is not distinguishable from) a uniform acceleration. In practice this means that a person
cannot feel (locally) the difference between standing on the surface of some gravitating body
(for example the Earth) and moving in a rocket with corresponding acceleration. In other words,
there is no experiment to distinguish between being weightless far out from gravitating bodies in
space and being in free-fall in a gravitational field.
All bodies in a given gravitational field will move in the same manner, if initial conditions are the
same. In other words, in given gravitational field all bodies move with the same acceleration.
In absence of gravitational field, all bodies move also with the same acceleration relative to the
non-inertial frame. Thus we can formulate the Principle of Equivalence which says: locally, any
non-inertial frame of reference is equivalent to a certain gravitational field. Globally (not locally),
actual gravitational fields can be distinguished from corresponding non-inertial frame of
reference by its behavior at infinity: Gravitational Fields generated by gravitating bodies
decay with distance.
PHY7006 (2014) Page 3

2 Tensors
The Principle of Covariance predetermines the mathematical structure of General Relativity: all
equations should contain tensors only. By definition, tensors are objects which are transformed
properly in the course of coordinate transformations from one frame of reference to another. Tak-
ing into account that non-inertial frames of reference in the 4-dimensional space-time correspond
to curvilinear coordinates, it is necessary to develop four-dimensional differential geometry in ar-
bitrary curvilinear coordinates.
Let us consider the transformation of coordinates from one frame of reference (x0 , x1 , x2 , x3 ) to
0 0 0 0
another, (x 0 , x 1 , x 2 , x 3 ):

∂xi 0k 0k ∂xi
dxi = dx = Sk
i
dx , i, k = 0, 1, 2, 3, where S i
k = (3)
∂x0k ∂x0k
is a transformation matrix.
Now we can give the definition of the Contravariant four-vector: The Contravariant four-vector
is the combination of four quantities (components) Ai , which are transformed like differentials of
coordinates:

Ai = Ski A0k . (4)

Let ϕ is a scalar field, then

∂ϕ ∂ϕ ∂x0k ∂ϕ
i
= 0k i
= S̃ik 0k , (5)
∂x ∂x ∂x ∂x

where S̃ik is another transformation matrix. S̃ki is inverse or reciprocal with respect to Ski :

∂xi ∂x0n ∂xi

Sni S̃kn = = = δki , (6)
∂x0n ∂xk ∂xk
where δki is so called Kronneker symbol, which actually is nothing but the unit matrix:

1 0 0 0
 
 0 1 0 0 
δki =  . (7)
 
 0 0 1 0 
0 0 0 1

Now we can give the definition of the Covariant four-vector: The Covariant four-vector is the
combination of four quantities (components) Ai , which are transformed like components of the
gradient of a scalar field:

∂x0k 0
Ai = A . (8)
∂xi k
Note, that for contravariant vectors we always use upper indices, which are called contravariant
indices, while for covariant vectors we use low indices, which are called covariant indices. Now
we can generalize the definitions of vectors and introduce tensors entirely in terms of transfor-
mation laws.

Turn over
Page 4 PHY7006 (2014)

Scalar, A, is the tensor of the 0 rank. It has only 40 = 1 component and 0 number of indices.
Transformation law is

A = A0 , (9)

we see that transformation matrices appear in transformation law 0 times.

Contravariant and covariant vectors are tensors of the 1 rank. They have 41 = 4 components
and 1 index. Corresponding transformation laws are

Ai = Sni A0n , (10)

Ai = S̃in A0n , (11)

we see only 1 transformation matrix in each transformation law.

Contravariant tensor of the 2 rank has 42 = 16 components and 2 contravariant indices. Cor-
responding transformation law is
k 0nm
Aik = Sni Sm A , (12)

we see 2 transformation matrices in the transformation law.

Covariant tensor of the 2 rank has 42 = 16 components and 2 covariant indices. Corresponding
transformation law is

Aik = S̃in S̃km A0nm , (13)

we see 2 transformation matrices in the transformation law.

Mixed tensor of the 2 rank has 42 = 16 components and 2 indices, 1 contravariant and 1
covariant. Corresponding transformation law is

Aik = Sni S̃km A0n

m, (14)

we see 2 transformation matrices in the transformation law.

Covariant tensor of the 3 rank has 43 = 64 components and 3 covariant indices. Corresponding
transformation law is...and so on.
The most general definition: Mixed tensor of the N + M rank with N contravariant and M
covariant indices, has 4N +M = 22(N +M ) components and N + M indices. Corresponding trans-
formation law is

Aik11 ik22......iNkM = Sni11 Sni22 ...SniNN S̃km11 S̃km22 ...S̃kmMM A0n1 n2 ... nN
m1 m2 ...mM , (15)

we see N + M transformation matrices in the transformation law.

Two tensors Aik and B ik are called reciprocal to each other if

Aik B kl = δil . (16)

Now we can introduce a contravariant metric tensor g ik which is reciprocal to the covariant metric
tensor gik :

gik g kl = δil . (17)

PHY7006 (2014) Page 5

With the help of the metric tensor and its reciprocal we can form contravariant tensors from
covariant tensors and vice versa, for example:
Ai = g ik Ak , Ai = gik Ak , (18)
in other words we can rise and descend indices as we like, like a kind of juggling with indices.
We can say that contravariant, covariant and mixed tensors can be considered as different rep-
resentations of the same geometrical object.
For the contravariant metric tensor itself we have very important representation in terms of the
transformation matrix from locally inertial frame of reference (galilean frame) to an arbitrary non-
i
inertial frame, let us denote it as S(0)k . We know that in the galilean frame of reference
1 0 0 0
 
 0 −1 0 0 
g ik =  ≡ η ik ≡ diag(1, −1, −1, −1), (19)
 
0 0 −1 0

 
0 0 0 −1
hence
g ik = S(0)n
i k
S(0)m η nm = S(0)0
i k
S(0)0 i
− S(0)1 k
S(0)1 i
− S(0)2 k
S(0)2 i
− S(0)3 k
S(0)3 . (20)
This means that if we know the transformation law from the local galilean frame of reference
to an arbitrary frame of reference, we know the metric at this arbitrary frame of reference and,
hence, we know the gravitational field which is identical to geometry!
Problem: Given that gik is a covariant tensor of the second rank and that
ds2 = gik dxi dxk , (21)
prove that ds is a scalar.
Solution:
0
ds2 = gik dxi dxk = (S̃in S̃km gnm )(Spi dx0p )(Swk dx0w ) = (S̃in Spi )(S̃km Swk )(gnm
0
dx0p dx0w ) =
0
= δpn δwm (gnm dx0p dx0w ) = gpw
0
dx0p dx0w = gik
0
dx0i dx0k = ds02 , (22)
hence ds = ds0 which means that ds is a scalar.
Problem: How many independent components in the metric tensor?
Solution: First, let us prove that the metric tensor is symmetric, i.e.
gik = gki . (23)
Indeed,
1 1 1
ds2 = gik dxi dxk = (gik dxi dxk + gik dxi dxk ) = (gki dxk dxi + gik dxi dxk ) = (gki + gik )dxi dxk =
2 2 2
= g̃ik dxi dxk , (24)
where
1
g̃ik = (gki + gik ), (25)
2
which is obviously a symmetric one. Then we just drop ” ˜ ”. The end of proof. Now the answer
is obvious: altogether we have 4 × 4 components, 4 components on the diagonal, 3 + 2 + 1 = 6
components above the diagonal and 3 + 2 + 1 = 6 components under the diagonal and we know
that these components are equal to components above the diagonal. Thus the final answer is
there are 4 + 6 = 10 independent components.

Turn over
Page 6 PHY7006 (2014)

3 The proper time and Physical distance

The geometry of space-time is determined entirely by the space-time interval, ds.

The proper time, τ , is defined as
ds
dτ = . (26)
c
For the proper time between any two events occurring at the same point in space (dxα = 0) we
have
√
ds2 = g00 (dx0 )2 , hence τ in terms of coordinate time t = dx0 /c is dτ = g00 dt. (27)
Thus
√
Z t
τ = τ0 + [ g00 ]|dxα =0 dt0 . (28)
t0

The determination of Physical distance: separating the space and time coordinates in ds2 we
have

ds2 = gαβ dxα dxβ + 2g0α dx0 dxα + g00 (dx0 )2 = az 2 + 2bz + d = 0, (29)
where
z = dx0 , a = g00 , b = g0α dxα and d = gαβ dxα dxβ . (30)
The interval between the events which belong to the same world line of light in Special and
General Relativity is always equal to zero:

ds = 0. (31)
Solving this equation with respect to z = dx0 we find two roots:
√ √
0(1) −b − b2 − ad 0(2) −b + b2 − ad
dx = z− = and dx = z+ = . (32)
a a
Then
√ s
2
v !
c c g00 b
u
0(2) 0(1)
u g0α g0β
dl = dτ = (dx − dx ) = −d= t −gαβ + dxα dxβ . (33)
2 2 c a g00

4 Parallel transport and covariant derivatives of the metric

tensor
In Special Relativity if Ai is a vector dAi is also a vector ( the same is valid for any tensor). But in
curvilinear coordinates this is not the case:
∂x0k 0
Ai = A (34)
∂xi k
PHY7006 (2014) Page 7

∂x0k 0 2 0k
0 ∂ x
dAi = dA k + A k dxl , (35)
∂xi ∂xi ∂xl
thus dAi is not a vector unless x0k are linear functions of xk ( like in the case of Lorentz transfor-
mations). Let us introduce the following very useful notation:
∂
,i = (36)
∂xi
According to the principle of covariance only tensors are allowed to appear in any physical equa-
tions, thus we should replace all differentials like
∂Ai
dAi and ≡ Ai,k (37)
∂xk
by some corrected values which we will denote as
DAi and Ai;k (38)
correspondingly. In arbitrary coordinates to obtain a differential of a vector which forms a vector
we should subtract vectors in the same point, not in different as we have done before. Hence,
we need to produce a parallel transport or a parallel translation. Under a parallel translation of a
vector in galilean frame of reference its components don’t change, but in curvilinear coordinates
they do and we should introduce some corrections:
DAi = dAi − δAi . (39)
These corrections obviously should be linear with respect to all components of Ai and indepen-
dently they should be linear with respect to dxk , hence we can write these corrections as
δAi = −Γikl Ak dxl , (40)
where Γikl are called Christoffel Symbols which obviously don’t form any tensor, because DAi is
the tensor while as we know dAi is not a tensor.
All covariant derivatives of metric tensor are equal to zero:

DAi = gik DAk (41)

DAi = D(gik Ak ) = gik DAk + Ak Dgik , (42)

hence
gik DAk = gik DAk + Ak Dgik , (43)
which obviously means that

Ak Dgik = 0.
Taking into account that Ak is arbitrary vector, we conclude that
Dgik = 0. (44)
Then taking into account that
Dgik = gik;m dxm = 0 (45)
for arbitrary infinitesimally small vector dxm we have
gik;m = 0. (46)

Turn over
Page 8 PHY7006 (2014)

5 The limit of stationarity and the event horizon of a black

hole
For test particle in rest dr = dθ = dφ = 0 the interval ds is given by

2
ds2 = g00 dx0 , (47)

We can see that if g00 = 0,then ds2 = 0, which means that the world line of particle in rest is the
world line of light. Hence, at the surface g00 = 0 no particle with finite rest mass can be in rest.
For this reason this surface is called the limit of stationarity.
Event Horizon is a spherically symmetric surface

F (r) = const. (48)

Its normal vector is defined as usually as

dF
ni = F,i = δi1 . (49)
dr
If at this surface

g 11 = 0 (50)

then
!2
ik 11 11 dF
g ni nk = g n1 n1 = g = 0, (51)
dr

which means that ni is a null vector and any particle with finite rest mass can not move outward
the surface g 11 = 0, thus this surface is the event horizon.
For the Kerr metric g00 = 0 gives

rg r
1− = 0, thus r2 − rg r + a2 cos2 θ = 0, (52)
ρ2

1 rg rg 2
q r
2 2
∆ = r − rg r + a = 0, and rst = (rg ± rg2 − 4a2 cos2 θ) = ± ( ) − a2 cos2 θ. (53)
2 2 2
The location of horizon in the Kerr metric: g 11 = 0 (g11 = ∞) corresponds to
1 rg rg 2
q r
2 2
∆ = r − rg r + a = 0, and r = (rg ± rg2 − 4a2 cos2 θ) = ± ( ) − a2 cos2 θ. (54)
2 2 2

s
2
rg rg

rhor = ± − a2 . (55)
2 2
PHY7006 (2014) Page 9

One can see easily that

rst ≥ rhor , (56)

for example,

rst = rhor , if θ = 0, or θ = π (at the poles), (57)

and
π
rst = 2rg > rhor , if θ = (at the equator). (58)
2
The region between the limit of stationarity and the event horizon is called the ”ergosphere”.
In the Schwarzschild metric as one can see putting a = 0,

rhor = rst , (59)

which means that in this case the ”ergosphere” does not exist.
By the Penrose mechanism it is possible to extract rotational energy of Kerr black hole. That
extraction is made possible because the rotational energy of the black hole is located not inside
the event horizon, but outside in a curl gravitational field. Such field is also called gravimagnetic
field. All objects in the ergosphere are unavoidably dragged by the rotating spacetime. The
Penrose mechanism: Some body enters into the ”ergosphere” and decays then into two pieces.
The momentum of the two pieces of matter can be arranged so that one piece escapes to infinity,
whilst the other falls past the outer event horizon into the black hole. The escaping piece of matter
can have greater mass-energy than the original infalling piece of matter.

6 The Riemann and Ricci tensors and Scalar Curvature

We know that
m
Ai,k,l − Ai,l,k = 0 and Ai; k; l − Ai; l; k = Am Rikl 6= 0 (60)

We can calculate the curvature Riemann tensor directly:

Ai;k;l − Ai;l;k = Ai;k,l − Ai;l,k − Γm m m m

il Am;k − Γkl Ai;m + Γik Am;l + Γlk Ai;m =

p p
= (Ai,k − Γm m m m
ik Am ),l − (Ai,l − Γil Am ),k − Γil (Am,k − Γmk Ap ) + Γik (Am,l − Γml Ap ) =
m p m p
= Ai,k,l − Γm m m m m m
ik Am,l − Γik,l Am − Ai,l,k + Γil Am,k + Γil,k Am − Γil Am,k + Γil Γmk Ap + Γik Am,l − Γik Γml Ap =
m p m p p m p m
= −Γm m m m
ik,l Am + Γil,k Am + Γil Γmk Ap − Γik Γml Ap = −Γik,l Am + Γil,k Am + Γil Γpk Am − Γik Γpl Am =

 
p m p m
= −Γm m m
ik,l + Γil,k + Γil Γpk − Γik Γpl Am = Rikl Am . (61)

Finally
m p m p m
Rikl = Γm m
il,k − Γik,l + Γil Γpk − Γik Γpl . (62)

Turn over
Page 10 PHY7006 (2014)

n
Riklm = gin Rklm . (63)

By straightforward calculations one can show that

1
Riklm = (gim,k,l + gkl,i,m − gil,k,m − gkm,i,l ) + gnp (Γnkl Γpim − Γnkm Γpil ). (64)
2
There are several symmetry properties of the curvature tensor:
1) The Riemann tensor is antisymmetric with respect to permutations of indices within each pair

Riklm = −Rkilm = −Rikml . (65)

2) The Riemann tensor is symmetric with respect to permutations of pairs of indices

Riklm = Rlmik . (66)

3) The cyclic sum formed by permutation of any three indices is equal to zero

Riklm + Rimkl + Rilmk = 0. (67)

The most important property of the Riemann tensor is the Bianchi identity:
n n n
Rikl; m + Rimk; l + Rilm; k = 0. (68)

It is easy to verify this identity in a locally inertial frame of reference, where

Γikl = 0, (69)

hence
n n n n n n
Rikl; m + Rimk; l + Rilm; k = Rikl,m + Rimk,l + Rilm,k = (70)

Γnil,m,k − Γnik,m,l + Γnik,l,m − Γnim,l,k + Γnim,k,l − Γnil,k,m = 0. (71)

Taking into account that the Bianchi identity is of a tensor character, we can conclude that it is
valid in any other frame of reference.
Now we can introduce a second rank curvature tensor, called the Ricci tensor, as follows

Rik = g lm Rlimk = Rilk

l
. (72)

We can also introduce a zero rank curvature tensor, i.e. a scalar, called the scalar curvature:

R = g ik Rik . (73)

After contracting the Biancci identity

i i i
Rklm;n + Rknl;m + Rkmn;l =0 (74)

over indices i and n (taking summation i = n) we obtain

i i i
Rklm;i + Rkil;m + Rkmi;l = 0. (75)
PHY7006 (2014) Page 11

According to the definition of Ricci tensor (72), the second term can be rewritten as
i
Rkil;m = Rkl;m . (76)
Taking into account that the Riemann tensor is antisymmetric with respect to permutations of
indices within the same pair
i i
Rkmi = −Rkim = −Rkm , (77)
the third term can be rewritten as
i
Rkmi;l = −Rkm;l . (78)
The first term can be rewritten as
i
Rklm;i = g ip Rpklm;i , (79)
then taking above mentioned permutation twice we can rewrite the first term as
i
Rklm;i = g ip Rpklm;i = −g ip Rkplm;i = g ip Rkpml;i . (80)
After all these manipulations we have
g ip Rkpml;i + Rkl;m − Rkm;l = 0. (81)
Then multiplying by g km and taking into account that all covariant derivatives of the metric tensor
are equal to zero, we have
     
g km g ip Rkpml;i + g km Rkl;m − g km Rkm;l = g km g ip Rkpml + g km Rkl − g km Rkm = 0. (82)
;i ;m ;l

In the first term expression in brackets can be simplified as

g km g ip Rkpml = g ip Rpl = Rli . (83)
In the second term the expression in brackets can be simplified as
g km Rkl = Rlm . (84)
According to the definition of the scalar curvature (73), the third term can be simplified as
 
g km Rkm = R;l = R,l . (85)
;l

Thus
i m
Rl;i + Rl;m − R,l = 0, (86)
replacing in the second term index of summation m by i we finally obtain
i i 1
2Rl;i − R,l = 0, or Rl;i − R,l = 0. (87)
2
Thus the important consequence of Bianchi identity is
i 1
Rl;i − R,l = 0. (88)
2

∂Γlik ∂Γlil
Rik = − + Γlik Γm m l
lm − Γil Γkm
∂xl ∂xk

Turn over
Page 12 PHY7006 (2014)

7 The EFEs

1 8πG
Rik − gik R = 4 Tik , (89)
2 c
where G is the gravitational constant.
Contracting with g ik , we have the Einstein equations in mixed form

8πG i 1 i
Rki = (Tk − δk T ). (90)
c4 2

8πG 1 8πG 1 8πG 1 8πG

     
ik
R = g Rik = 4 g ik Tik − g ik gik T = 4 Tii − δii T = 4 T − 4 = − 4 T. (91)
c 2 c 2 c 2 c
Thus
c4
T =− R, (92)
8πG
hence,
c4 1
 
Tik = Rik − gik R , (93)
8πG 2

c4 1
 
Tki = Rki − δki R . (94)
8πG 2

8 The principle of the least action

The EFEs can be derived from the principle of the least action. This principle says that

δ(Sg + Sm ) = 0, (95)

where Sg and Sm are the actions of gravitational field and matter respectively. Taking into account
that we are going to derive EFEs, the subject of variations is all components of the metric tensor.
The action Sg should depend on configuration of gravitational field, or geometry, in the whole
space-time, hence it should be expressed in terms of a scalar integral over the all space and
over the time coordinate between two given moments of time
Z
Sg = GdΩ̃, (96)

where dΩ̃ is invariant element of 4-volume (see Lecture 3) and G is some scalar function called
the action density. We know that the final equations should contain derivatives of gik no higher
than the second. Otherwise we could not obtain Newtonian Poisson’s equation (see the previous
lecture). In other words, G must contain only gik and Γlmn , i.e

G = G(gik , Γikl ). (97)

PHY7006 (2014) Page 13

Immediately we confront the following problem: it is impossible to construct the scalar from gik
and Γlmn . The only scalar in gravitational field, the scalar curvature R, contains the second
derivatives of gik . Fortunately, there is a rather simple resolution of this paradox: R is linear with
respect to the second derivatives and for this reason, as we will see later, all terms containing
second derivatives don’t contribute to the variations of the action. Let us write the action function
in the following form
Z
√
Sg = α R −gdΩ, (98)

where α is a constant which will be determined later.

Because of the linearity of R with respect to the second derivatives, the invariant action function
can be transformed in the following way
Z
√ Z
√ Z
Sg = α R −gdΩ = α G −gdΩ + α w,ll dΩ, (99)

where G contains only gik and gik,n , w is a function which can be obtained by straightforward
calculations:
√ √ √ n o
−gR = −gg ik Rik = −g g ik Γlik,l − g ik Γlil,k + g ik Γlik Γm ik m l
lm − g Γil Γkm , (100)

obviously
√ √ √
−gg ik Γlik,l = ( −gg ik Γlik ),l − Γlik ( −gg ik ),l (101)

and
√ √ √ √ √
−gg ik Γlil,k = ( −gg ik Γlil ),k − Γlil ( −gg ik ),k = ( −gg il Γkik ),l − Γkik ( −gg il ),l . (102)

Then we obtain
√ √ √ √ √
−gR = ( −gg ik Γlik − −gg il Γkik ),l + −gG = wl , l + −gG, (103)

where
√
wl = −g(g ik Γlik − g il Γkik ) (104)

and
√ √ l √ m √
−gG = Γm ik ik m l l
im ( −gg ), k − Γik ( −gg ),l − (Γil Γkm − Γik Γlm ) −gg
ik
(105)

1 ∂gim
Γiki = g im k . (106)
2 ∂x
According to the Gauss’ theorem the volume integral of a full derivative is reduced to the integral
over boundary. Taking into account that our objective is to obtain proper equations by apply-
ing the principle of the least action, we should keep all boundary conditions fixed. Hence, w
disappears after variation. As a result
Z
√ Z
√
δ R −gdΩ = δ G −gdΩ. (107)

Turn over
Page 14 PHY7006 (2014)

Thus we don’t need G any more, because we proved that the variation of the integral with R is
the same as the variation of the integral with G, hence we can work with R only.
Z
√ Z
√ Z
√ √ √
δ R −gdΩ = δ g Rik −gdΩ = {Rik −gδg ik + g ik Rik δ( −g) + g ik −gδRik }dΩ. (108)
ik

There are three terms in the variation of the action function. Let us first calculate the second
term.
√ 1 1 ∂g 1
δ( −g) = − √ δg = − √ δgik = − √ M ik δgik , (109)
2 −g 2 −g ∂gik 2 −g

where M ik is the minor of the determinant g corresponding to the component gik . Indeed, the
determinant g depends on all components gik . Calculating g with the help, say the first row, one
can write g = M 1i g1i , where M 1i are minors of the components in the first row. Obviously M 1i do
not contain g1i . Hence
∂g
= M 1i . (110)
∂g1i

This is true for any row in determinant:

∂g
= M ni . (111)
∂gni

Taking into account that g ik is reciprocal to gik , i.e. gik g kn = δin , (g ik is inverse matrix of gik ), one
can write g ik = M ik /g, i.e. M ik = gg ik . Thus

∂g
dg = dgik = M ik dgik = gg ik dgik , (112)
∂gik

hence
dg √
g ik dgik = = d ln |g| = d ln(−g) = 2 ln −g. (113)
g

Then

g ik dgik = d(g ik gik ) − gik dg ik = dδii − gik dg ik = −gik dg ik . (114)

Thus
√ 1 1 1√
δ( −g) = − √ gg ik δgik = √ ggik δg ik = − −ggik δg ik . (115)
2 −g 2 −g 2

Now we can rewrite the variation of action as

Z
√ Z
1 √ √
δ R −gdΩ = [(Rik − gik R) −gδg ik + g ik −gδRik ]dΩ. (116)
2
Let us consider now the last term in the variation. For the calculation of δRik we can use the fact
that although Γikn is not a tensor, its variation, δΓikn , is a tensor.
PHY7006 (2014) Page 15

Proof: Let Ai be an arbitrary vector at the point xi . After the parallel transport From the point xi
to the point xi + dxi , as we know, its components are
Ai (xn + dxn ) = Ai (xn ) + (Ai,m (xn ) + Γimp (xn )Ap (xn ))dxm . (117)
Then
δAi (xn + dxn ) = δΓimp (xn )Ap (xn ))dxm . (118)
The left side is a vector because it is the difference between two vectors in the same point, hence
the right side is also a vector. Thus δΓimp (xn ) is a tensor. In a locally galilean frame of reference
n o
g ik δRik = g ik δΓlik,l − δΓlil,k = g ik δΓlik,l − g il δΓkik,l = W,ll , (119)

where
W l = g ik δΓlik − g il δΓkik , (120)
obviously W l is a vector. Now let us prove that the covariant divergence of an arbitrary vector
can be written as follows
1 √
An; n = √ ( −gAn ),n . (121)
−g
Proof:
1 1
An;n = An,n +Γnni Ai = An,n + g nm (gnm,i +gmi,n −gin,m )Ai = An,n + (g nm gnm,i +g nm gmi,n −g nm gni,m )Ai =
2 2

1
= An,n + g nm gnm,i Ai . (122)
2
Taking into account (112), one obtains
g,n n 1 √ √ 1 √
Ai;i = An,n + A = √ [ −gAn,n + ( −g),n An ] = √ ( −gAi ),i . (123)
2g −g −g
As follows from the proof above, in local galilean frame of reference, where g = −1
Ai;i = Ai,i , (124)
hence, returning back to δRik , in local galilean frame of reference we have
g ik δRik = W,ll = W;ll . (125)
Since this is a relation between two tensors (of 0-rank), once this is valid in one frame of reference
it is valid in an arbitrary frame of reference. Hence
√ √ √
−gg ik δRik = −gW;ll = ( −gW l ),l , (126)
this means that according to the Gauss theorem the contribution of the third term in the variation
of the action function is equal to zero. Finally we obtain
Z
1 √
δSg = α (Rik − gik R)δg ik −gdΩ. (127)
2

Turn over
Page 16 PHY7006 (2014)

9 Motion of particles and Propagation of light around Black

Hole
Any object of a small enough mass is called a test particle. Small mass means that the gravita-
tional field generated by this object is negligible in comparison with the external gravitational field
generated by other, much more massive, objects. The role of such test particle can be played by
a planet around a star or a star around a massive black hole, or by a photon propagating around
a neutron star or black hole.
The derivation of the Hamilton−Jacobi:
From the definition of the four-velocity
dxi
ui = , (128)
ds
we have

ds2 = gik dxi dxk = gik ui uk ds2 = ui ui ds2 , (129)

hence

ui ui = 1. (130)

Four-momentum of the particle is defined as

pi = mcui , hence pi pi = g ik pi pk = m2 c2 . (131)

Taking into account that a covariant vector transforms as the gradient of a scalar, we can intro-
duce such a scalar function that
∂S
pi = − , (132)
∂xi
then we immediately obtain the Hamilton−Jacobi equation for a particle in a gravitational field
∂S ∂S
g ik − m2 c2 = 0. (133)
∂xi ∂xk
The shortest way to obtain the Eikonal equation is just to put m = 0 in the Hamilton−Jacobi
equation and change notations.
Let us consider a spherically symmetric gravitational field and assume that this field does not
depend on time, i.e. it is static field. Taking into account the spherical symmetry we can choose
our spherical coordinates in a such way that the plane of orbit coincides with the equatorial plane
θ = π/2 and dθ = 0. Obviously, all the components of a metric tensor are functions of the
radial coordinate only. Let us denote the radial coordinate as x1 = r. We can write the interval
describing such gravitational field as

ds2 = g00 (r)c2 dt2 + g11 (r)dr2 + g33 dφ2 . (134)

In this case the Hamilton−Jacobi equation can be written as

!2 !2 !2
00 ∂S 11 ∂S 33 ∂S
g (r) + g (r) + g (r) − m2 c2 = 0. (135)
c∂t ∂r ∂φ
PHY7006 (2014) Page 17

Let me remind you that for photons ds = 0. We can introduce some scalar parameter λ varying
along world line of the light signal and introduce then a vector
dxi
ki = , (136)
dλ
which is tangent to the word line. This vector is called four- dimensional wave vector. Then
ds2 = gik dxi dxk = gik k i k k dλ2 = 0 and we have ki k i = g ik ki kk = 0. (137)
Substituting covariant vector
∂ψ
ki = − , (138)
∂xi
where ψ is a scalar, we obtain the Eikonal Equation in gravitational field
g ik Ψ,i Ψ,k = 0. (139)
The physical meaning of Ψ (called the Eikonal follows from
Z
Ψ=− ki dxi , (140)

which looks like the phase of electromagnetic wave. If the Eikonal equation is solved, one can
obtain the world line of photon:
dxi
≡ k i = g in kn = −g in Ψ,n . (141)
dλ
In the equatorial plane of a Schwarzschild black hole the solution of the Eikonal equation can be
written in the form
bω
Ψ = −ωt + φ + Φr (r), (142)
c
whereω is the frequency of the photon and b is its impact parameter. Substituting this expression
to the Eikonal equation we obtain
!2
1 ω2 1 bω rg
 
rg 2 − 2 − 1− (−p1 )2 = 0, (143)
1− r c r c r
where
v " #
dΦr (r) u 1 1 ω 2 b2 ω 2
u
p1 ≡ pr = −Ψ,1 = − = ±t − 2 2 . (144)
dr 1 − rrg 1 − rrg c2 cr

One can easily show that photons can move along unstable circular orbits given by
dUef f (ph)
Uef f (ph) = 1, and = 0, (145)
dr
where Uef f (ph) plays the role of the effective potential for photons and is given by
b2 rg
 
Uef f (ph) = 2 1− . (146)
r r

Turn over
Page 18 PHY7006 (2014)

10 Gravitational waves
A weak gravitational field is a small perturbation of the Galilean metric:

gik = ηik + hik . (147)

It is easy to show that

g ik = η ik − η in η km hnk . (148)

The gravitational wave is a transverse and traceless part of these perturbations and the plane
wave has two independent states of linear polarization. Using a linear coordinate transformation
0
x i = xi + ξ i , (149)

where ξ i are small functions of xi , we can impose on hik the following four supplementary condi-
tions:
1
η km hmi,k − δik η nm hnm,k = 0. (150)
2
After such transformation the Ricci tensor is reduced to
1 ∂ 2 hik
Rik = − η lm l m . (151)
2 ∂x ∂x
According to the Einstein equations in empty space-time Rik = 0, hence gravitational waves
satisfy the wave equation

1 ∂2
(∇2 − )hik = 0, (152)
c2 ∂t2
where ∇2 is the 3-dimensional Laplacian operator.
In approximation of slow motions and small separations we can use the Quadrupole formula for
gravitational waves:

2G d2 Dαβ
hαβ = − (t − R/c), (153)
3c4 R dt2
where R is the distance to the source of gravitational radiation and
Z
Dαβ = (3xα xβ − r2 δαβ )dM is the quadrupole tensor. (154)

Problem: A white dwarf (or neutron star, or black hole) of mass m moves around a black hole of
mass M  m on a circularAssume that the orbital period is T .
(i) What is the frequency of gravitational radiation?
(ii) Estimate also the amplitude of gravitational wave.
Solution:
(i) Taking into account that

xα = δαβ xβ = eα cos ω0 t, (155)

PHY7006 (2014) Page 19

where eα is some constant unit vector in the orbital plane. Using (154), we have
·· ·· ·· 1 ··
hαβ ∝ D̈αβ ∼ (3xα xβ − r2 δαβ ) ∼ (xα xβ ) ∼ eα eβ (cos2 ω0 t) ∼ eα eβ (1 + cos 2ω0 t) ∼ cos ω,(156)
2
where

ω = 2ω0 = 4π/T. (157)

Thus the frequency of gravitational radiation is

4π
2ω = . (158)
T
(ii) Using (153)we can estimate to an order of magnitude, h0 , the amplitude of the gravitational
wave.Omitting indices we have

2G 2G 3 2 2 4Gmr2 ω02 4Gmr2 GM m rg2 m rg rg

h ∼ 4 D̈ ∼ 4 (2ω0 ) mr = 4
= 4 3
== = . (159)
3c R 3c R 2 cR cR r M rR M r R