Variational Principle Approach

to General Relativity
Chakkrit Kaeonikhom

Submitted in partial fulfilment of

the requirements for the award of the degree of

Bachelor of Science in Physics

B.S.(Physics)

Fundamental Physics & Cosmology Research Unit

The Tah Poe Academia Institute for Theoretical Physics & Cosmology
Department of Physics, Faculty of Science
Naresuan University
March 15, 2006

To Mae

Variational Principle Approach

to General Relativity
Chakkrit Kaeonikhom

The candidate has passed oral examination by members of examination panel.

This report has been accepted by the panel as partially fulfilment of the course 261493
Independent Study.

....................................
Dr. Burin Gumjudpai, BS MSc PhD AMInstP FRAS

Supervisor

....................................
Dr. Thiranee Khumlumlert, BSc(Hons) MSc PhD

Member

....................................
Alongkorn Khudwilat, BS(Hons) MSc

Member

Acknowledgements
I would like to thank Burin Gumjudpai, who gave motivation to me to learn story
that I have never known, thank for his inspiration that lead me to the elegance of
Physics, thank for his explanation of difficult concepts and thank for his training of
LATEX program. I also thank Daris Samart for willingness to spend his time discussing
to me and for help on some difficult calculations. Thanks Chanun Sricheewin and
Alongkorn Khudwilat for some discussions and way out of some physics problems. I
also thank Artit Hootem, Sarayut Pantian and all other members in the new founded
Tah Poe Academia Institute for Theoretical Physics & Cosmology (formerly the Tah
Poe Group of Theoretical Physics: TPTP). Thank for their encouragement that push
me to do my works. Finally, thank for great kindness of my mother who has been
teaching me and giving to me morale and everything.

II

Title:

Variational Principle Approach to General Relativity

Candidate:

Mr.Chakkrit Kaeonikhom

Supervisor:

Dr.Burin Gumjudpai

Degree:

Bachelor of Science Programme in Physics

Academic Year: 2005

Abstract
General relativity theory is a theory for gravity which Galilean relativity fails to
explain. Variational principle is a method which is powerful in physics. All physical
laws is believed that they can be derived from action using variational principle.
Einsteins field equation, which is essential law in general relativity, can also be derived
using this method. In this report we show derivation of the Einsteins field equation
using this method. We also extend the gravitational action to include boundary terms
and to obtain Israel junction condition on hypersurface. The method is powerful and
is applied widely to braneworld gravitational theory.

III

Contents
1 Introduction
1.1 Background .
1.2 Objectives . .
1.3 Frameworks .
1.4 Expected Use
1.5 Tools . . . . .
1.6 Procedure . .
1.7 Outcome . . .

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

2 Failure of classical mechanics and introduction

2.1 Inertial reference frames . . . . . . . . . . . . .
2.2 Failure of Galilean transformation . . . . . . . .
2.3 Introduction to special relativity . . . . . . . . .

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

to special relativity
. . . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . . .

3 Introduction to general relativity

3.1 Tensor and curvature . . . . . . . . . . . . . . . . . . .
3.1.1 Transformations of scalars, vectors and tensors .
3.1.2 Covariant derivative . . . . . . . . . . . . . . .
3.1.3 Parallel transport . . . . . . . . . . . . . . . . .
3.1.4 Curvature tensor . . . . . . . . . . . . . . . . .
3.2 The equivalence principle . . . . . . . . . . . . . . . . .
3.3 Einsteins law of gravitation . . . . . . . . . . . . . . .
3.3.1 The energy-momentum tensor for perfect fluids
3.3.2 Einsteins field equation . . . . . . . . . . . . .

IV

.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

1
1
1
1
2
2
2
3
4
4
4
6
11
11
11
13
15
16
18
19
19
20

4 Variational principle approach to general relativity

4.1 Lagrangian formulation for field equation . . . . . . .
4.1.1 The Einstein-Hilbert action . . . . . . . . . .
4.1.2 Variation of the metrics . . . . . . . . . . . .
4.1.3 The full field equations . . . . . . . . . . . . .
4.2 Geodesic equation from variational principle . . . . .
4.3 Field equation with surface term . . . . . . . . . . . .
4.3.1 The Gibbons-Hawking boundary term . . . .
4.3.2 Israel junction condition . . . . . . . . . . . .

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

24
24
24
26
28
29
30
30
33

5 Conclusion

38

A Proofs of identities
A.1 c gab = 0 . . . . . . . . . . . . . . . . . . .
A.2 c g ab = 0 . . . . . . . . . . . . . . . . . . .
A.3 Covariant derivative for scalar field, a . .
A.4 Ra bcd = Ra bdc . . . . . . . . . . . . . . . .
A.5 Ra bcd + Ra dbc + Ra cdb = 0 . . . . . . . . . .
A.6 Bianchi identities . . . . . . . . . . . . . . .
A.7 Conservation of Einstein tensor: b Gab = 0

40
40
40
42
42
43
43
44

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

B Detail calculation
B.1 Variance of electromagnetic wave equation under Galilean transformation
B.2 Poissons equation for Newtonian gravitational field . . . . . . . . . .
B.3 Variation of Cristoffel symbols : abc . . . . . . . . . . . . . . . . . .

45
45
46
47

Chapter 1
Introduction
1.1

Background

Classical mechanics is useful to explain physical phenomena but it fails to explain

gravity. General relativity (GR) is proposed to be a satisfactory theory for gravity.
Studying GR leads to the Einsteins field equation which can be derived in standard
way. Our interest is to apply variation principle to derive the field equation in GR.

1.2

Objectives

To study general relativity and tensor calculus which is applied to derive, with variational principle, Einsteins field equation and Israel junction condition.

1.3

Frameworks

To explain failure of Newtonian mechanics, special relativity.

To study general relativity.
To use variational principle for Einsteins field equation and for Israel junction
condition.

1.4

Expected Use

To obtain Einsteins field equation from variational method and to obtain junction condition by including of surface term in action.
A derived-in-detailed report for those who interest with thoroughly calculation
from variational method in general relativity.
Attaining understanding of concept of general relativity and having skills on
tensor calculus.

1.5

Tools

Text books in physics and mathematics.

A high efficiency personal computer.
Software e.g. LATEX, WinEdit and Photoshop

1.6

Procedure

Studying special relativity.

Studying tensor analysis and calculation skills.
Studying concepts of general relativity.
Studying Einsteins field equation by evaluating coupling constant.
Studying Einsteins field equation by variational principle.
Including surface term in action and deriving junction condition.
Making conclusion and preparing report and other presentation.

1.7

Outcome

Understanding of basic ideas of classical mechanics, special relativity and general relativity.
Attaining skills of tensor calculation.
Understanding in detail of the variation method in general relativity.

Chapter 2
Failure of classical mechanics and
introduction to special relativity
2.1

Inertial reference frames

Newton introduced his three laws of motion as axioms of classical mechanics. These
laws have successfully explained motion of most objects known to us. The Newtons
first law states that a body remains at rest or in uniform motion. This law introduces
a frame of reference called inertial frame. All proceeded dynamical laws base on this
law. But what and where is it?
To know or to measure velocity of a particle, we need a frame of reference. For
example when we measure speed of a car, a particular spot on ground is inertial
frame. The ground is on the Earth and is not really inertial frame due to gravity.
Furthermore, all stars in the universe possess gravity therefore nowhere is really locate
on true inertial frame! However we will discuss about inertial frame again in chapter 3.

2.2

Failure of Galilean transformation

In Newtonian mechanics, the concept of time and space are completely separable.
Furthermore time is assumed to be absolute quantity and independent of any observers. Consider inertial frames of reference moving with constant velocity to each
other. In classical mechanics there is transformation law between two inertial frames.
4

v
(x, y, z, t),

vt

(x, y, z, t)

Figure 2.1: Relative velocity of two inertial frames

That is so called Galilean transformation:

x0 = x vt
y0 = y
z0 = z
t0 = t.

(2.1)

Newtons laws are invariant with respect to Galilean transformation

Fi = m
xi = m
x0i = Fi0

(2.2)

The principle of Galilean transformation yields that the velocity of light in two different inertial frames is measured with different values
u0 = c u and u = c.

(2.3)

However the Galilean transformation fails to explain electromagnetic wave equation;

electromagnetic wave equation is not invariant under Galilean transformation. Consider electromagnetic wave equation:
2 =
or

1 2
c2 t2

1 2
2 2 2
+
+

= 0.
x2 y 2
z 2
c2 t2
5

(2.4)

Figure 2.2: Illustration of four-dimensional spacetime, time axis is orthogonal to all

three-dimensional spatial axes
Using chain rule and equation (2.1) to transform coordinate, the wave equation
becomes
c2 v 2 2 2v 2
2
2
1 2
+ 2 0 0 + 02 + 02 2 02 = 0.
c2 x02
c t x
y
z
c t

(2.5)

This equation contradicts to Einsteins postulates in special relativity that physical

laws should be the same in all inertial frames. Therefore we require new transformation law, Lorentz transformation which we shall discuss in the next section.

2.3

Introduction to special relativity

In framework of Newtonian mechanics we consider only three-dimensional space and

flat geometry while in special relativity we consider space and time as one single entity
called spacetime. We do not separate time from space and we have four dimensions
of spacetime. Considering four-dimensional spacetime, time axis is orthogonal to all
axes of space (x, y, z ).
The mathematical quantity which represents space and time is components of
spacetime coordinates
xa = (x0 , x1 , x2 , x3 )
= (ct, x, y, z)

in Cartesian coordinate

(2.6)

Special Relativity (SR) is a theory for physics in flat spacetime called Minkowski
spacetime. If we are talking about spacetime, we must have events. Any events
6

Figure 2.3: Lightcone and trajectory of a particle in four-dimensional spacetime

possess four coordinates describing where and when the particle is. Trajectories of
particles in four-dimensional spacetime form a set of particles worldline. If we reduce
to one dimensional space, we will have two spatial dimensions left. Plotting these two
spatial coordinate axes versus time, we have spacetime diagram as shown in Fig. 2.2
The electromagnetic wave equation introduces speed of light in vacuum which

is c = 1/ 0 0 , where 0 and 0 are permeability and permittivity of free space

respectively. The value c is independent to all reference frame. This statement is
confirmed by Michelson-Morley experiment. Einsteins principle of special relativity
states that
The law of physical phenomena are the same in all inertial reference frames.
The velocity of light is the same in all inertial reference frames.
The spacetime interval or line-element in four dimensional spacetime is
ds2 = c2 dt2 + dx2 + dy 2 + dz 2 .

(2.7)

We can see that if ds2 = 0, this equation becomes equation of spherical wave of light
with radius cdt. We can classify four-vectors into 3 classes namely spacelike, lightlike
and timelike vectors. Any line elements are called
spacelike

if they lie in region

ds2 < 0

lightlike or null

if they lie in region

ds2 = 0

timelike

if they lie in region

ds2 > 0.

equation (2.7) can be expressed in general form as

2

ds = dx dxa =

3
X

gab dxa dxb = gab dxa dxb .

(2.8)

a,b=0

We use Einsteins summation convention. In this convention we do not need to use

summation symbol in equation (2.8). If upper indies and lower indies of four-vectors
are similar (repeated) in any terms, it implies sum over that indices. The symmetrical
tensor gab is called metric tensor. When considering flat Minkowski spacetime, we
write ab instead of gab ,

1 0 0 0
0 1 0 0

(2.9)
gab = ab =
0 0 1 0 .
0 0 0 1
The line element can be computed in matrix form,

1 0 0 0
dx0
0 1 0 0 dx1

ds2 =
0 0 1 0 dx2
0 0 0 1
dx3

dx0
dx1

dx2
dx3

= (dx2 )2 + (dx1 )2 + (dx2 )2 + (dx3 )2

= cdt2 + dx2 + dy 2 + dz 2 .

(2.10)

Coordinate transformation law in SR is Lorentz transformation, a translation or

boosts in one direction between two inertial frames moving relative to each other
with constant velocity.
In relativity, time is not an absolute quantity. Time measured in each reference
frame are not equal. Time measured in one inertial frame by observer on that frame
is called proper time. Since speed of light is the same in all inertial frames, then the
transformation law between two frames is written as
cdt0 = (dt vdx/c)
dx0 = (dx vdt)
dy 0 = dy
dz 0 = dz.
8

(2.11)

This set of equations is Lorentz transformation. The symbol is Lorentz factor,

=p

1
1 v 2 /c2

(2.12)

Lorentz transformation can also be written in general form,

dxa = a b dxb .
where a b is Lorentz matrix,
a b

0 0

0 0
.
=
0
0
1 0
0
0
0 1

(2.13)

p
Here we define1 v/c and 1/ 1 2 .

cdt0

dx0

dy 0 = 0
0
0
dz
0
0

Therefore equation (2.13) is

0 0
cdt

0 0
dx .

1 0
dy
0 1
dz

(2.14)

(2.15)

Using equation (2.10) and equation (2.11), we obtain

c2 dt2 + dr2 = c2 dt02 + dr02
or

c2 dt2 + dx2 = c2 dt02 + dx02

(2.16)

ds2 = ds02

(2.17)

Therefore

where dr2 = dx2 + dy 2 + dz 2 and dr02 = dx02 + dy 02 + dz 02 . Equation ( 2.16) is

rotational transformation in (x, ct) space. Values of and range in 0 1 and
1 respectively. If we introduce new parameter and write and in term
of this new variables as
= tanh
= cosh ,

(2.18)

where in the angle of rotation in this (x, ct) space, then Lorentz transformation
becomes

0
cosh sinh 0 0
ct
ct
x0 sinh cosh 0 0 x

.
0 =
(2.19)
y
0
0
1 0 y
z0
0
0
0 1
z
1

is speed parameter. For photon = 1.

We can see that the Lorentz boost is actually rotation in (x,ct) space by angle .
Lorentz transformation yields two important phenomena in SR. These are time dilation and length contraction. The dilation of time measured in two inertial reference
frames is described by
t = t0
(2.20)
and length contraction of matter measured in two inertial reference frames is described
by
L0
L=
(2.21)

where t0 and L0 are proper time and proper length respectively.

10

Chapter 3
Introduction to general relativity
3.1

Tensor and curvature

In last chapter, we consider physical phenomena based on flat space which is a spacial case of this chapter. In this chapter is curved space is of interest. A physical
quantity needed here is tensors which are geometrical objects. Tensor is invariant in
all coordinate systems. Vectors and scalars are subsets of tensors indicated by rank
(order) of tensors i.e. vector is tensor of rank 1, scalar is zeroth-rank tensor. Any
tensors are defined on manifold M which is n-dimensional generalized object that
it locally looks like Euclidian space Rn .

3.1.1

Transformations of scalars, vectors and tensors

In general relativity, vectors are expressed in general form, X = X a ea where X a is

a component of vector and ea is basis vector of the component. Here a b is general
transformation metric. Contravariant vector or tangent vector X a transforms
like
x0a b
X = a b X b
(3.1)
X 0a =
xb
or can be written in chain rule,
x0a b
dx .
xb
We define Kronecker delta, a b as a quantity with values 0 or 1 under conditions,

1 if a = b
a
b=
0 if a 6= b.
dx0a =

11

Therefore

x0a
xa
=
= a b.
x0b
xb
Another quantity is scalar which is invariant under transformation,
0 (x0a ) = (xa ).

(3.2)

(3.3)

Consider derivative of scalar field = (xa (x0 )) with respect to x0a , using chain rule,
we obtain
xb

=
.
x0a
x0a xb
Covariant vector or 1-form or dual vector Xa in the xa -coordinate system, transforms according to
xb
Xa0 (xa ) =
Xb (xa ) = b a Xb (xa ).
(3.4)
0a
x
There is also a relation
a b b c = a c .
(3.5)
For higher rank tensor transformation follows
0

0k10 k20 ...kk0

l10 ......ll0

x0k1
xll k1 k2 ......kk
x0kk xl1
=
...... k
l1 ......ll
0 ......
0T
xk1
x k x0l1
x0ll
0
0
= k1 k1 ......kk kk l1 l10 ......ll ll0 T k1 k2 ......kk l1 ......ll .

(3.6)

The metric tensor

Any symmetric covariant tensor field of rank 2 such as gab , defines a metric tensor. Metrics are used to define distance and length of vectors. The square of the
infinitesimal distance or interval between two neighboring points (events) is defined
by
ds2 = gab dxa dxb .
(3.7)
Inverse of gab is g ab . The metric g ab is given by
gab g ac = b c .

(3.8)

We can use the metric tensor to lower and raise tensorial indices,
... ...
= gab T......b...
T...a...
...

(3.9)

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

(3.10)

and

12

3.1.2

Covariant derivative

Consider a contravariant vector field X a defined along path xa (u) on manifold at a

point P where u is free parameter. There is another vector X a (xa + xa ) defined at
xa + xa or point Q. Therefore
X a (x + x) = X a (x) + X a .

(3.11)

We are going to define a tensorial derivative by introducing a vector at Q which is

parallel to X a . We can assume the parallel vector differs from X a by a small amount
a (x). Because covariant derivative is defined on curved geometry,
denoted by X
therefore parallel vector are parallel only on flat geometry which is mapped from
manifold. We will explain about it in next section.
The difference vector between the parallel vector and the vector at point Q illustrated in Fig. 3.1 is
a (x)] = X a (x) X
a.
[X a (x) + X a (x)] [X a (x) + X

(3.12)

We define the covariant derivative of X by the limiting process

a
o
1 n a
a (x)
X
(x
+
x)

X
(x)
+
X
x 0 xc

1 a
a .
= lim
X
(x)

X
c
xc 0 x

c X a =

lim
c

a (x) should be a linear function on manifold. We can write

X
a (x) = a (x)X b (x)xc
X
bc

(3.13)

where the abc (x) are functions of coordinates called Cristoffel symbols of the second
kind. In flat space abc (x) = 0. But in curved space it is impossible to make all the
abc (x) vanish over all space. The Cristoffel symbols of second kind are defined by

1 ad gdc gbd gcb

a
bc = g
+

.
(3.14)
2
xb
xc
xd
Following from the last equation that the Cristoffel symbols are necessarily symmetric
or we often called it that torsion-free, i.e.
abc = acb .

13

(3.15)

Xa

Xa

Xa

Xa

Xa

Xa

Xa
Q

x a (u )

Figure 3.1: Vector along the path xa (u)

We now define derivative of a vector on curved space. Using equation (3.13) and
chain rule, we have

1 X a (x) c
a
a
b
c
c X = lim
x + bc (x)X (x)x .
xc 0 xc
xc
Therefore the covariant derivative is
c X a = c X a + abc X b .

(3.16)

The notation c is introduced by c /xc . We next define the covariant derivative

of a scalar field to be the same as its ordinary derivative (prove of this identity is
in Appendix A),

c = c .
Consider
c = c (Xa Y a )
= (c Xa )Y a + Xa (c Y a )
= (c Xa )Y a + Xa (c Y a + abc Y b )
and
c = (c Xa )Y a + Xa (c Y a ).
From equation (3.17), we equate both equations together:
(c Xa )Y a + Xa (c Y a ) = (c Xa )Y a + Xa (c Y a + abc Y b ).
14

(3.17)

Renaming a to b and b to a for the last term in the right-hand side of the equation
above because they are only dummy indices therefore,
(c Xa )Y a = (c Xa )Y a + (c Xa bac Xb )Y a .
Covariant derivative for covariant vector is then
c Xa = c Xa bac Xb .

(3.18)

The covariant derivative for tensor follows

a....
a....
d....
a....
c Tb....
= c Tb....
+ acd Tb....
+ .... dcb Td....
.....

(3.19)

The last important formula is covariant derivative of the metric tensor,

c gab = 0

(3.20)

c g ab = 0.

(3.21)

and
The proves of these two identities are in Appendix A.

3.1.3

Parallel transport

The concept of parallel transport along a path is in flat space. A parallel vector
transporting from a point to another point maintains its unchange in magnitude and
direction. In curve space, components of a vector are expected to change under
parallel transport in different way from the case of flat space. Consider the parallel
transport along a curve xa ( ) with a tangent vector X a = dxa /d where is a
parameter along the curve. Beginning from the formula of the covariant derivative
X a a X b = 0,
we get

c
dxa
dxb
b dx
+ ac
a
= 0
d
d
d
a
c
dxa dxb
b dx dx
+

= 0
ac
a
d x
d d
d

d dxb
dxa dxc
= 0.
+ bac
d d
d d
15

(3.22)

Figure 3.2: Parallel transport of vectors on curved space

We rename a to b and b to a, then
b
c
d2 xa
a dx dx
+ bc
= 0.
d 2
d d

(3.23)

This equation is known as the geodesic equation. The geodesic distance between
any two points is shortest. The geodesic is a curve space generalization of straight
line in flat space.

3.1.4

Curvature tensor

We now discuss the most important concepts of general relativity, That is concept the
of Riemannian geometry or curved geometry which is described in tensorial form.
We will introduce the Riemann tensor by considering parallel transport along an
infinitesimal loop illustrated in Fig. 3.3. The Riemann curvature tensor Ra bcd is
defined by the commutator of covariant derivatives,
Ra bcd X b = (c d d c )X a .
16

(3.24)

Xa

xa

xa

X' a1

X' a2
xa

xa

xa

x a dx a

dx a

Figure 3.3: Transporting X a around an infinitesimal loop

Consider
(c d d c )X a = c d X a d c X a
= c (d X a + adb X b ) d (c X a + acb X b ).
d X a is a tensor type(1, 1). Using equation (3.19) we get
(c d d c )X a = c (d X a + adb X b ) ecd (e X a + aeb X b ) + ace (d X e + edb X b )
d (c X a + acb X b ) + edc (e X a + aeb X b ) ade (c X e + ecb X b )
= c d X a + adb c X b + c adb X b ecd e X a ecd aeb X b
+ace d X e + ace edb X b d c X a acb d X b d acb X b
+edc e X a + edc aeb X b ade c X e ade ecb X b .
We rename e to b in the terms ace d X e and ade c X e . Assuming torsion-free condition
of Cristoffel symbols and using equation (3.24), we have Riemann tensor expressed
in terms of Cristoffel symbols:
Ra bcd = c abd d abc + ebd aec ebc aed .

(3.25)

Ra bcd depends on the metric and the metrics first and second derivatives. It is
anti-symmetric on its last pair of indices,
Ra bcd = Ra bdc .

(3.26)

The last equation introduces identity:

Ra bcd + Ra dbc + Ra cdb = 0.
17

(3.27)

Lowering the first index with the metric, the lowered tensor is symmetric under
interchanging of the first and last pair of indices. That is
Rabcd = Rcdab .

(3.28)

The tensor is anti-symmetric on its first pair of indices as

Rabcd = Rbacd .

(3.29)

We can see that the lowered curvature tensor satisfies

Rabcd = Rabdc = Rbacd = Rcdab
and

Rabcd + Radbc + Racdb = 0.

(3.30)

The curvature tensor satisfies a set of differential identities called the Bianchi identities:
a Rbcde + c Rabde + b Rcade = 0.
(3.31)
We can use the curvature tensor to define Ricci tensor by the contraction,
Rab = Rc acb = g cd Rdacb .

(3.32)

Contraction of Ricci tensor then also defines curvature scalar or Ricci scalar R
by
R = g ab Rab .
(3.33)
These two tensors can be used to define Einstein tensor
1
Gab = Rab gab R,
2

(3.34)

which is also symmetric. By using the equation (3.31), the Einstein tensor can be
shown to satisfy the contracted Bianchi identities
b Gab = 0

3.2

(3.35)

The equivalence principle

In chapter 2, we introduced some concepts about inertial frames of reference. We will

discuss about its nature in this chapter.
An inertial frame is defined as one in which a free particle moves with constant
velocity. However gravity is long-range force and can not be screened out. Hence an
18

purely inertial frame is impossible to be found. We can only imagine about. According
to Newtonian gravity, when gravity acts on a body, it acts on the gravitational mass,
mg . The result of the force is an acceleration of the inertial mass, mi . When all bodies
fall in vacuum with the same acceleration, the ratio of inertial mass and gravitational
mass is independent of the size of bodies. Newtons theory is in principle consistent
with mi = mg and within high experimental accuracies mi = mg to 1 in 1,000.
Therefore the equivalence of gravitational and inertial mass implies:
In a small laboratory falling freely in gravitational field, mechanical phenomena are
the same as those observed in an inertial frame in the absence of gravitational field.
In 1907 Einstein generalized this conclusion by replacing the word mechanical phenomena with the laws of physics. The resulting statement is known as the principle
of equivalence. The freely-falling frames introduce the local inertial frames which
are important in relativity.

3.3

Einsteins law of gravitation

In this section, we use Riemannian formalism to connect matter and metric that leads
to a satisfied gravitational theory.

3.3.1

The energy-momentum tensor for perfect fluids

The energy-momentum tensor contains information about the total energy density
measured by an arbitrary inertial observer. It is defined by the notation T ab . We start
by considering the simplest kind of matter field, that is non-relativistic matter or
dust. We can simply construct the energy-momentum tensor for dust by using fourvelocity ua defined as ua = (c, 0, 0, 0) for rest frame and the proper density :
T ab = ua ub .

(3.36)

Notice that T 00 is the energy density.

In general relativity, the source of gravitational field can be regarded as perfect
fluid. A perfect fluid in relativity is defined as a fluid that has no viscosity and no
heat conduction. No heat conduction implies T 0i = T i0 = 0 in rest frame and energy
can flow only if particles flow. No viscosity implies vanishing of force paralleled to
the interface between particles. The forces should always be perpendicular to the
19

interface, i.e. the spatial component T ij should be zero unless i = j. As a result we

write the energy-momentum tensor for perfect fluids in rest frame as
2

c 0 0 0
0 p 0 0

T ab =
(3.37)
0 0 p 0
0 0 0 p
where p is the pressure and is the energy density. Form equation (3.37), it is easy
to show that

p
T ab = + 2 ua ub + pg ab .
(3.38)
c
We simply conclude from the above equation that the energy-momentum tensor is
symmetric tensor. The metric tensor g ab here is for flat spacetime (we often write
ab instead of g ab for flat spacetime). Notice that in the limit p 0, a perfect fluid
reduces to dust equation (3.36). We can easily show that the energy-momentum
tensor conserved in flat spacetime:
b T ab = 0.

(3.39)

Moreover, if we use non-flat metric, the conservation law is

b T ab = 0.

3.3.2

(3.40)

Einsteins field equation

Einsteins field equation told us that the metric is correspondent to geometry and
geometry is the effect of an amount of matter which is expressed in energy-momentum
tensor. Matters cause spacetime curvature. We shall use Remannian formalism to
connect matter and metric. Since covariant divergence of the Einstein tensor Gab
vanishes in equation (3.35), we therefore write
1
gab R = Tab .
(3.41)
2
If there is gravity in regions of space there must be matter present. The proportional
constant is arbitrary. Contract the Einstein tensor by using the metric tensor, we
obtain
Gab = Rab

G = g ab Gab = g ab Tab
1
g ab Rab g ab gab R = g ab Tab
2
1
R aa R = T.
2
20

aa is 4 which is equal to its dimensions. Therefore equation (3.41) can be rewritten

as

1
Rab = Tab gab T .
(3.42)
2
We want to choose appropriate value of constant . Let us consider motion of particle
which follows a geodesic equation
b
c
d2 xa
a dx dx
+

= 0.
bc
d 2
d d

(3.43)

In Newtonian limit, the particles move slowly with respect to the speed of light.
Four-velocity is ua
= u0 , therefore the geodesic equation reduces to
0
0
d2 xa
a dx dx
=

= c2 a00 .
00
2
d
d d

(3.44)

Since the field is static in time, the time derivative of the metric vanishes. Therefore
the Christoffel symbol is reduced to:

1 ad
g 0 gd0 + 0 g0d d g00
a00 =
2
1
= g ad d g00 .
(3.45)
2
In the limit of weak gravitational field, we can decompose the metric into the Minkowski
metric, the equation (2.9), plus a small perturbation,
gab = ab + hab ,

(3.46)

where hab 1 here. From definition of the inverse metric, we find that to the first
order in h,
1
gab
= (ab + hab )1

= ab hab .

g ab =

Since the metric here is diagonal matrix. Our approximate is to only first order for
small perturbation of inverse matric. Equation (3.45) becomes

1 ad
had d h00
2
1 ad

d h00 .
=
2

a00 =

21

(3.47)

Substituting this equation into the geodesic equation ( 3.44),

d 2 xa
c2 ad
d h00 .
=
d 2
2

(3.48)

Using 0 h00 = 0 and since d2 x0 /d 2 is zero, we are left with spacelike components of
the above equation. Therefore,
c2 ij
d2 xi
=
j h00 .
d 2
2

(3.49)

Recall Newtons equation of motion

d2 xi
= i
2
d

(3.50)

where is Newtonian potential. Comparing both equations above, we obtain

h00 =

2
.
c2

(3.51)

In non relativistic limit (dustlike), the energy-momentum tensor reduce to the equation (3.36). We will work in fluid rest frame. Equation (3.38) gives
T00 = c2

(3.52)

and
T = g ab Tab

= g 00 T00

= 00 T00
= c2 .

(3.53)

We insert this into the 00 component of our gravitational field equation (3.42). We
get

1 1
R00 = c2 c2 = c2 .
(3.54)
2
2
This is an equation relating derivative of the metric to the energy density. We shall
expand Ri 0i0 in term of metric. Since R0 000 = 0 then
R00 = Ri 0i0 = i i00 0 ii0 + iib b00 i0b bi0 .

22

The second term is time derivative which vanishes for static field. The second order of
Christoffel symbols ()2 can be neglected since we only consider a small perturbation.
From this we get

1
i
ib
R00
= i 00 = i g 0 gb0 + 0 g0b b g00
2
1
1
= i g ib b g00 g ib i b g00
2
2
1
1

= i ij j h00 ij i j h00
2
2
1 ij
= i j h00
2
1
= 2 h00 .
(3.55)
2
Substituting equation (3.51) into the above equation and using equation (3.54), we
finally obtain
1
2 = c4 .
2

(3.56)

The connection to Newtonian theory appears when this equation is compared with
Poissons equation for Newtonian theory of gravity. The theory of general relativity
must be correspondent to Newtons non-relativistic theory in limiting case of weak
field. Poissons equation in a Newtonian gravitational field is
2 = 4G

(3.57)

where G is classical gravitational constant. Comparing this with equation (3.56) gives
the value of ,
=

8G
.
c4

(3.58)

From equation (3.41) then the full field equation is

8G
1
Gab = Rab gab R = 4 Tab .
2
c

23

(3.59)

Chapter 4
Variational principle approach to
general relativity
4.1
4.1.1

Lagrangian formulation for field equation

The Einstein-Hilbert action

All fundamental physical equation of classical field including the Einsteins field equation can be derived from a variational principle. The condition required in order to
get the field equation follows from
Z
Ld4 x = 0.
(4.1)
Of course the quantity above must be an invariant and must be constructed from
the metric gab which is dynamical variable in GR. We shall not include function
which is first the derivative of metric because it vanishes at a point P M. The
Riemann tensor is of course made from second derivative set of the metrics, and the
only independent scalar constructed from the metric is the Ricci scalar R. The well

definition of Lagrangian density is L = gR, therefore

Z

gR d4 x
(4.2)
SEH =

24

is known as the Einstein-Hilbert action. We derive field equation by variation of

action in equation (4.2),
Z

SEH =
gR d4 x
Z

=
d4 x gg ab Rab
Z
Z
Z

4
ab
4
ab
=
d x gg Rab + d x gRab g + d4 xR g.
Now we have three terms of variation that
SEH = SEH(1) + SEH(2) + SEH(3)
The variation of first term is
SEH(1) =

d4 x gg ab Rab .

(4.3)

(4.4)

Considering the variation of Ricci tensor,

Rab = Rc acb = c cab b cac + ccd dba cbd dac

(4.5)

Rab = c cab b cac + dba ccd + ccd dba dac cbd cbd dac

= c cab + ccd dba dac cbd dbc cad

b cac + cbd dac dba ccd dbc cad

and the covariant derivative formula:
c cab = c cab + ccd dba dac cbd dbc cad

(4.6)

b cac = b cac + cbd dac dba ccd dbc cad

(4.7)

and also

we can conclude that

Rab = c cab b cac .

(4.8)

Therefore equation (4.4) becomes

Z

c
c
4
ab
SEH(1) =
d x gg c ab b ac

4
ab
c
c
ab
ab
c
c
ab
=
d x g c g ab ab c g b g b ac + ac b g .
25

Remembering that the covariant derivative of metric is zero. Therefore we get

Z

4
ab
c
c
SEH(1) =
d x gg c ab b ac

Z

ab
c
4
ab
c
=
d x g c g ab b g ac
Z
h
i

=
d4 x g c g ab cab g ac bab
Z

=
d4 x g c J c
where we introduce
J c = g ab cab g ac bab .

(4.9)

If J c is a vector field over a region M with boundary . Stokess theorem for the
vector field is
Z
Z
p
p
c
4
d3 x |h|nc J c
(4.10)
d x |g| c J =

where nc is normal unit vector on hypersurface . The normal unit vector nc can
be normalized by na na = 1. The tensor hab is induced metric associated with
hypersurface defined by
hab = gab + na nb .
(4.11)
Therefore the first term of action becomes
Z
p
SEH(1) =
d3 x |h|nc J c = 0.

(4.12)

This equation is an integral with respect to the volume element of the covariant divergence of a vector. Using Stokess theorem, this is equal to a boundary contribution
at infinity which can be set to zero by vanishing of variation at infinity. Therefore
this term contributes nothing to the total variation.

4.1.2

Variation of the metrics

Firstly we consider metric gab . Since the contravariant and covariant metrics are
symmetric matrices then,
gca g ab = c b .
(4.13)
We now consider inverse of the metric:
1
1
g ab = (Aab )T = Aba
g
g
26

(4.14)

where g is determinant and Aba is the cofactor of the metric gab . Let us fix a, and
expand the determinant g by the ath row. Then
g = gab Aab .

(4.15)

If we perform partial differentiation on both sides with respect to gab , then

g
= Aab .
gab

(4.16)

Let us consider variation of determinant g:

g
gab
gab
= Aab gab

g =

= gg ba gab .
Remembering that g ab is symmetric, we get
g = gg ab gab .
Using relation obtained above, we get

1
g =
g
2 g
1 g

=
g ab gab .
2 g

(4.17)

(4.18)

We shall convert from gab to g ab by considering

a d = (gac g cd ) = 0
g cd gac + gac g cd = 0
g cd gac = gac g cd .
Multiply both side of this equation by gbd we therefore have
gbd g cd gac = gbd gac g cd
bc gac = gbd gac g cd
gab = gac gbd g dc .
Substituting this equation in to equation (4.18) we obtain

1
g =
g g ab gac gbd g dc
2
1
=
g cb gbd g dc
2
1
=
g gcd g dc .
2
27

(4.19)

Renaming indices c to a and d to b, we get

1
g =
g gab g ab .
2
The variation of Einstein-Hilbert action becomes
Z
Z

1
4
ab
SEH =
d x gRab g
d4 xR g gab g ab
2
Z
h
i

1
=
d4 x g Rab gab R g ab
2
The functional derivative of the action satisfies
Z X

S
i
S =

d4 x
i

(4.20)

(4.21)

(4.22)

where {i } is a complete set of field varied. Stationary points are those for which
S/i = 0. We now obtain Einsteins equation in vacuum:
1 SEH
1

= Rab gab R = 0.
ab
g g
2

4.1.3

(4.23)

The full field equations

Previously we derived Einsteins field equation in vacuum due to including only gravitational part of the action but not matter field part. To obtain the full field equations,
we assume that there is other field presented beside the gravitational field. The action
is then
S=

1
SEH + SM
16G

(4.24)

where SM is the action for matter. We normalize the gravitational action so that we
get the right answer. Following the above equation we have

1
1
1 SM
1 S

=
Rab gab R +
= 0.
ab
g g
16G
2
g g ab
We now define the energy-momentum tensor as
1 SM
.
Tab = 2
g g ab

(4.25)

This allows us to recover the complete Einsteins equation,

1
Rab Rgab = 8GTab .
2
28

(4.26)

4.2

Geodesic equation from variational principle

Consider motion of particle along a path xa ( ). We will perform a variation on this

path between two points P and Q. The action is simply
Z
S = d.
(4.27)
In order to perform the variation, it is useful to introduce an arbitrary auxiliary
parameter s. Here ds is displacement on spacetime. We have

dxa dxb 1/2

ds
(4.28)
d = gab
ds ds
We vary the path by using standard procedure:
Z
S = d
Z
dxa dxb 1/2
ds
=
gab
ds ds

1
dxa dxb 1/2
dxa dxb
dxa dxb
=
ds gab
2gab
.
gab
2
ds ds
ds ds
ds ds
Considering the last term,

b
dxa dxb
d
dgab a dxb
d2 xb
a dx
2gab
=
2gab x
+2
x
+ 2gab xa 2 .
ds ds
ds
ds
ds
ds
ds
The two points, P and Q are fixed. We can set first term of above equation to zero.
Therefore we obtain

1
dxa dxb 1/2
dxa dxb
dgab dxb a
d2 xb a
S =
ds gab
gab
+2
x + 2gab 2 x
2
ds ds
ds ds
ds ds
ds

Z
1
ds2
dxa dxb
dgab dxb a
d2 xb
=
d 2 gab
+2
x + 2gab 2 xa
2
d
ds ds
ds ds
ds

Z
a
b
b
2 b
dx dx
dx a
1
dgab dx a
=
d gab
+2
x + 2gab 2 x .
2
d d
d d
d
By using chain rule we get

Z
1
gab dxa dxb c
gab dxc dxb a
d2 xb a
S =
d
x + 2 c
x + 2gab 2 x
2
xc d d
x d d
d

Z
a
c
c
a
dx dx b
dx dx
dxc dxa b
d2 xa b
1
b
d b gac
x + c gba
x + a gbc
x + 2gba 2 x
=
2
d d
d d
d d
d

a
2 a
c
dx dx
dx
1
=
d xb
b gac + c gba + a gbc
+ gba 2 .
2
d d
d
29

We set the variation of action to zero, S = 0 and multiply by g db ,

dxa dxc
1 db
d2 xa
g
b gac + c gba + a gbc
+ ad 2 = 0
2
d d
d
We therefore recover geodesic equation:
a
c
d2 xd
d dx dx
+

= 0
ac
d 2
d d

4.3

(4.29)

Field equation with surface term

In section 4.1 we have derived field equation without boundary term which is set to
be zero at infinity. In this section, we shall generalize field equation to general case
which includes boundary term in action.

4.3.1

The Gibbons-Hawking boundary term

We begin by putting boundary term in the first part of the action in section 4.1 of
Einstein-Hilbert action,

Z
Z

1
1
1
ab 4
SEH =
g Rab gab R g d x +
hnc J c d3 x. (4.30)
16G M
2
16G
Considering vector J c in the last term of equation (4.9) we have
J c = g ab cab g ac bab .
Using formula (prove in appendix B)

1
cab = g cd a gbd + b gad d gab ,
2
we get

1 cd
= g
g a gbd + b gad d gab
2

ac 1 bd
g
g a gbd + b gad d gab
2
1 ab cd
1
1
=
g g a gbd + g ab g cd b gad g ab g cd d gab
2
2
2
1 ac bd
1 ac bd
1
g g a gbd g g b gad + g ac g bd d gab .
2
2
2

ab

30

(4.31)

Interchanging dummy indices a and b in the second term, a and d in the fourth term,
and a and d in the last term, we obtain result

(4.32)
J c = g ab g cd a gbd d gab .
Now our discussion is on hypersurface. Lowering index of J c with metric gce ,

Jc = g ab a gbd d gab .
(4.33)
By using equation (4.11), we have

nc Jc = nc hab na nb a gbd d gab

= nc hab a gbd nc hab d gab na nb nc a gbd + na nb nc d gab .
From boundary condition, the first term of this equation vanishes since it is projected
on hypersurface which variation of metric and induced metric vanish at such as
g ab = 0,
h

ab

= 0

(4.34)
(4.35)

and we use definition

na nb nc = 0.

(4.36)

nc Jc = nc hab d gab .

(4.37)

Therefore we obtain
Now let us consider arbitrary tensor T a1 ...ak b1 ...bl at P . It is a tensor on the
tangent space to at P if
T a1 ...ak b1 ...bl = ha1 c1 ...hak ck hd1 b1 ...hdl bl T c1 ...bk d1 ...dl

(4.38)

Defining ha b b as a projected covariant derivative on hypersurface by using notation

Da , it satisfies
Dc T a1 ...ak b1 ...bl = ha1 d1 ...hak dk he1 b1 ...hel bl hf c f T d1 ...dk e1 ...ek .

(4.39)

The covariant derivative for induced metric on hypersurface automatically satisfies

(4.40)
Da hbc = ha d hb e hc f d gaf + ne nf = 0
since d gef = 0 and hab nb = 0 (remembering nb is perpendicular to hypersurface.
Therefore its dot product with metric on hypersurface is zero). Next we introduce
extrinsic curvature in the form
Kab = ha c c nb
31

(4.41)

and we can use contract it with induced metric,

K = hab Kab = hab ha c c nb = hab a nb .

(4.42)

Variation of extrinsic curvature given by

K = hab a nb
= hab a nb + hab a nb hab cab nc hab cab nc
= hab a nb hab cab nc hab cab nc
= hab a nb hab cab nc
= hbc ha c a nb hab cab nc

= Dc hbc nb hab cab nc

h
i

bc
bc
= Dc h nb nb h hab cab nc
= hab cab nc .

(4.43)

Using equation (4.31),

K = hab nc

h1

i
g cd a gbd + b gad d gab

2
1 ab d
h n d gab .
=
2

(4.44)

Notice that when substituting equation (4.37) to this equation, we obtain

Considering boundary term from

Z

hnc J c d3 x =

=
=

1
K = nc Jc .
2
Einstein-Hilbert action
Z

hnc Jc d3 x

2
hKd3 x
Z
Z

3
2
hKd x + 2 hKd3 x

(4.45)

and using equation (4.20),

Z
Z
Z

c 3
3
hnc J d x = 2
hKd x
hKhab hab d3 x.

Since hab vanishes at boundary. Therefore we get

Z
Z

c 3
hnc J d x = 2
hKd3 x.

32

(4.46)

Now we have variation of Einstein-Hilbert action in the form

Z
Z

1
1
1
ab 4
SEH =
g Rab gab R g d x

hKd3 x.
16G M
2
8G

(4.47)

We generalize this equation by naming the first term as the variation of gravitational
action, SGravity , we finally get
Z

1
hKd3 x
SEH = SGravity

8G
Z

1
SGravity = SEH +
hKd3 x

8G
Z
Z

1
1
4

SGravity =
gRd x +
hKd3 x.
(4.48)
16G M
8G
The last term is known as Gibbons-Hawking boundary term. In next section,
we will vary this term using junction condition.

4.3.2

Israel junction condition

We begin from considering the action

SGravity = SEH + SGH
Z
Z

1
1
4
gRd x +
hKd3 x.
=
16G M
8G
Its variation is
SGravity

Z
Z

1
1
ab 4
=
gGab g d x +
hna J a d3 x
16G M
16G
Z
Z

1
1
3
+
hKd x +
hKd3 x.
8G
8G

(4.49)

In this section, we are not interested in boundary when performing variation of gravitational action then there is no boundary condition. Variation of K is

K = hb a a nb
= hb a a nb + hb a a nb hb a bac nc + hb a bac nc
= hb a a nb + hb a bac nc + hb a a nb .
Considering the first term of equation (4.50),

hb a a nb = b a + na nb a nb

= na nb + nb na a nb .
33

(4.50)

Using identity
nb = nb nc nc ,

(4.51)

we get

na nb nc nc + nb na a nb

= na nc nc + c a nc nb a nb

hb a a nb =

= nb nc hc a a nb
= nb nc Kc b
= 0.

(4.52)

Since extrinsic curvature is associated with hypersurface then its dot product with
normal vector on vanishes yielding equation (4.52). We continue to do variation of
the second term of the equation (4.50) by using formula (4.31), we then have

1 a c bd
hb n g a gcd + c gad d gac .
hb a bac nc =
2
Therefore variation of K is

1
K = had nc a gcd + c gad d gac + hb a a nb .
(4.53)
2

Applying similar procedure to the equation (4.20) for h, we obtain

1
h hab hab .
(4.54)
h =
2
Using the equation (4.32), finally the gravitational action is
Z
Z

1
1
ab 4
SGravity =
gGab g d x
h hab Khab d3 x
16G M
16G
Z

1
+
hna g cb g ad c gbd d gcb d3 x
16G

1
ad c
a
b
+
h h n a gcd + c gad d gac + 2hb a n d3 x.
16G
Considering
Therefore
SGravity

na g cb g ad = nd hcb + nc nb = nd hcb .

(4.55)

Z
Z

1
1
ab 4
=
gGab g d x
h hab Khab d3 x
16G M
16G

1
+
h nd hcb c gbd nd hcb d gcb + had nc a gcd
16G

a
b
ad c
ad c
+h n c gad h n d gac + 2hb a n d3 x.
34

Interchanging dummy indices between d and c and rename b to a of the first and the
second term in bracket of the third integral term results that the first term and the
last term in bracket cancel out. So do the second and the forth term in bracket of
the third integral term. Therefore
Z
Z

1
1
ab 4
SGravity =
gGab g d x
h hab Khab d3 x
16G M
16G
Z

ad c
1
+
h h n a gcd + 2hb a a nb d3 x.
(4.56)
16G
Considering the first term in the third integral term
had nc a gcd = had a (nc gcd ) had gcd a nc .
Using equation (4.19), we get
had nc a gcd = had a (nc gcd ) + had gde gcf g ef a nc
= had a (nc gcd ) + had (hde + nd ne )gcf g ef a nc
= had a (nc gcd ) + ha e g ef a nf
= had a (nc gcd ) + Kef g ef .
Considering
g ef = (hef + ne nf )
= hef + nf ne + ne nf
= hef + nf ne na na + ne nf nb nb
= hef .

(4.57)

Therefore
had nc a gcd = had a (nc gcd ) + Kef hef .

(4.58)

Substituting this equation in equation (4.56), we obtain

Z
Z

1
1
ab 4
SGravity =
gGab g d x
h hab Khab d3 x
16G M
16G
Z
i
h ad c

1
+
h h a n gcd + Kef hef + 2hb a a nb d3 x
16G
Z
Z

1
1
ab 4
=
gGab g d x +
h Kab hab K hab d3 x
16G M
16G
Z
i
h

1
a
b
c
ad
(4.59)
+
h h a n gcd + 2hb a n d3 x
16G
35

Considering the two terms in bracket of the last integral term

h
i

had a (nc gcd ) + 2hb a a nb = had a nc gcd gcd nc + 2hb a a nb

= had a nd ha c a nc + 2hb a a nb
= had a nd + hb a a nb
= g bd hb a a nd + hb a a nb

= hb a a g bd nd + nb

= hb a a g bd nd nc nc + b c nc
h
i
= hb a a nb nc + b c nc

= Db hb c nc

(4.60)

and substituting the last equation into equation (4.59), we obtain

Z
Z

1
1
ab 4
SGravity =
gGab g d x +
h Kab hab K hab d3 x
16G M
16G
Z

1
+
hDb hb c nc d3 x.
(4.61)
16G
Since the last term in the equation (4.61) is divergence term. It yields vanishing of
this integral with boundary at infinity on . Finally, the variation of gravity is
Z

1
gGab g ab d4 x
SGravity =
16G M
Z

1
+
h Kab hab K hab d3 x.
(4.62)
16G
The action for matter on hypersurface is
Z
Smat =
Lmat d3 x

(4.63)

and its variation is

Smat =

htab hab

(4.64)

where the energy-momentum tensor on hypersurface is

tab = 2

1 Smat
.
h hab

(4.65)

The total action is

SGravity = SEH + SGB + Smat .
36

(4.66)

Since we include energy-momentum tensor on hypersurface, then Einstein tensor in

the bulk vanishes.1 Therefore the total action gives the Israel junction condition
Kab hab K = 8Gtab .

(4.67)

The energy-momentum tensor on hypersurface is not necessarily conserved because

energy can flow from the hypersurface to the bulk. The idea has been recently widely
applied to the research field of braneworld gravity and braneworld cosmology (see e.g.
[8] for review of application of Israel junction condition to braneworld gravity.)

The bulk is the region of one dimension left beyond the hypersurface. The bulk and hypersurface

together form total region of the manifold.

37

Chapter 5
Conclusion
Relativity begins with concept of inertial reference frame which is defined by Newtons
first law. Therefore relative frames in one direction is considered allowing us to do
Galilean transformation for inertial frame moving not so fast. However under this
transformation, the speed of light is no longer invariant and electromagnetic wave
equations are neither invariant. Special relativity came along based on Einsteins
principle of spacial relativity. In this theory, there is a unification of time and space
so called spacetime. It uses Lorentz transformation under which electromagnetic
wave equations are invariant. Moreover it reduces to Galilean transformation in
case of small velocity. Therefore SR implies no absolute inertial reference frame
in the universe. Einstein introduced transformation law when including effects of
gravitational field and also introduced the equivalence principle which suggest that
purely inertial reference frame is not sensible. It introduces us local inertial frame in
gravitational field. Therefore Newtonian mechanics based on inertial frame fails to
explain gravity. SR based on flat space is extended to general relativity using concept
of curved space. Curved space is indicated by Riemann tensor, Ra bcd . This theory
attempts to explain gravity with geometry. It tells us that mass is in fact curvature
of geometry This fact is expressed in the Einsteins field equation,
Gab = 8GTab .
All physical laws are believed to obey principle of least action, S = 0. The dynamical
variable in GR is gab therefore the Lagrangian density in action must be a function of
gab . We used Ricci scalar in term of second derivative of metric. We did not choose
function of first derivative of metric since it vanishes at any point on manifold. With
38

the action

SEH =

gRd4 x

we can derive Einsteins field equation by neglecting boundary term (surface term).
If including boundary term then, we must have dynamical variable on boundary or
hypersurface, such as hab which is an induced metric on the hypersurface. Bounding
the manifold, we obtain extrinsic curvature K which is constructed from hab in the
action on the hypersurface. That is
Z

hKd3 x.
SGH =

This is called Gibbon-Howking boundary term. The result of this variation is Israel
junction condition,
Kab hab K = 8Gtab
which is shown in detail calculations in this report. This result can be applied widely
to braneworld gravity.

39

Appendix A
Proofs of identities
A.1

cgab = 0

The identity c gab = 0 is proved by using covariant derivative to tensor type(0, 2):

c gab = c gab dca gbd dcb gad

(A.1)

Using equation (3.14), we have

1
c gab = c gab gbd g de c gae + a gce e gac gad g de c gbe + b gce e gbc
2
2
1

1 b
= c gab e c gae + a gce e gac ea c gbe + b gce e gbc
2
2
1

1
c gab + a gcb b gac
c gba + b gca a gbc
= c gab
2
2
= 0.
(A.2)

A.2

cg ab = 0

The identity c g ab = 0 is proved by using covariant derivative to tensor type(2, 0):

c g ab = c g ab + acd g bd + bcd g ad

40

(A.3)

Therefore

1
c g ab = c g ab + g bd g ae c gde + d gce e gcd + g ad g be c gde + d gce e gcd
2
2
1
1
1
= c g ab + g ae g bd c gde + g bd g ae d gce g ae g bd e gcd
2
2
2
1 be ad
1 ad be
1 be ad
+ g g c gde + g g d gce g g e gcd
2
2
2

1
1 bd
ab
ae
bd
bd
ae
ae
= c g + g c gde g
gde c g + g d gce g
gce d g
2
2

1 ae
1 be
bd
bd
ad
ad
gcd e g + g c gde g
gde c g
g e gcd g
2
2

1 ad
1 be
be
be
ad
ad
+ g d gce g
gce d g g e gcd g
gcd e g
2
2
1

1
= c g ab + g ae c eb gde c g bd + g bd d ca gce d g ae
2
2
1

1 ae b
bd
be
g e c gcd e g
+ g c ea gde c g ad
2
2
1

1 ad b
be
+ g d c gce d g
g be e ca gcd e g ad
2
2
ca are constant then their partial derivatives vanish. Therefore we get
1
1
1
1
c g ab = c g ab g ae gde c g bd g bd gce d g ae + g ae gcd e g bd g be gde c g ad
2
2
2
2
1 ad
1
g gce d g be + g be gcd e g ad
2
2
1
1
1
1
= c g ab da c g bd g bd gce d g ae + g ae gcd e g bd db c g ad
2
2
2
2
1 ad
1
g gce d g be + g be gcd e g ad
2
2
1
1
1
1
= c g ab c g ba g bd gce d g ae + g ae gcd e g bd c g ab
2
2
2
2
1 ad
1
g gce d g be + g be gcd e g ad .
2
2
Interchanging dummy indices d and e, we get
1
1
1
1
c g ab = g be gcd e g ad + g ad gce d g be g ad gce d g be + g be gcd e g ad
2
2
2
2
= 0.
(A.4)

41

A.3

Covariant derivative for scalar field, a

We will show that covariant derivative for scalar field is just ordinary derivative.
Considering

a = a Xb X b
= Xb a X b + X b a Xb
= 2Xb a X b + 2Xb bac X c

= 2a Xb X b 2X b a Xb + Xb X c g bd a gdc + c gad d gac

= 2a 2X b a Xb + X c X d a gdc + X c X d c gad X c X d d gac .
Interchanging dummy indices c and d yields that the last and the forth term cancel
out. Therefore we get
a = 2a 2X b a Xb + X c X d a gdc

= 2a 2X b a Xb + X c a gdc X d X c gdc a X d
= 2a 2X b a Xb + X c a Xc Xd a X d
= 2a X b a Xb Xd a X d

= 2a a Xb X b
= a

A.4

(A.5)

Ra bcd = Ra bdc

Proving the identity Ra bcd = Ra bdc is as follows

Ra bcd = c abd d abc + ebd aec ebc aed

= c abd + d abc ebd aec + ebc aed

= d abc c abd + ebc aed ebd aec

Ra bcd = Ra bdc

42

(A.6)

A.5

Ra bcd + Ra dbc + Ra cdb = 0

Proving the identity Ra bcd + Ra dbc + Ra cdb = 0 is as follows

1.

Ra bcd = c abd d abc + ebd aec ebc aed

2.

Ra dbc = b adc c adb + edc aeb edb aec

3.

Ra cdb = d acb b acd + ecb aed ecd aeb

Ra bcd + Ra dbc + Ra cdb = c abd d abc + ebd aec ebc aed + b adc
c adb + edc aeb edb aec + d acb
b acd + ecb aed ecd aeb
= 0

A.6

(A.7)

Bianchi identities

Considering covariant derivative of Riemann tensor evaluated in locally inertial coordinate:

a Rbcd
e = a
e
Rb
cd

(A.8)

Hats on their indices represent locally inertial coordinate. Notice that for locally
inertial coordinate, Cristoffel symbols which contain first-order derivatives of metrics
must vanish at any points. But the second derivatives of metrics in Riemann tensor
do not vanish. Therefore
a

Rbcd
e = ga

b R cd
e

= gab dce ed
c

= gab g af dcgef + degcf dfgce edgcf ecgdf + efgd

c

1
dcgeb db gce ecgdb + eb gd
(A.9)

Rbcd
c .
e =
2
We therefore obtain

a Rbcd
e + cRa
e + b Rca
e
bd
d

1
=
a dcgeb a db gce a ecgdb + a eb gd
c
2
+cdb gea cda gbe ceb gd
a + cea
gdb

+b da gec b dcgae b ea gd
c + b ecgd
a
= 0.

(A.10)
43

A.7

Conservation of Einstein tensor: bGab = 0

Considering Bianchi identities,

e Rabcd + d Rabec + c Rabde = 0

g ae g bc e Rabcd + d Rabec + c Rabde = 0

a Rad d R + b Rbd = 0.
Renaming indices b to a, we have
2a Rad d R = 0
1
a Rad d R = 0
2

1
a
Rad gad R = 0.
2
We then see that twice-contracted Bianchi identities is equivalent to
a Gad = 0.

44

(A.11)

Appendix B
Detail calculation
B.1

Variance of electromagnetic wave equation under Galilean transformation

Consider electromagnetic wave equation

2 2 2
1 2
+
+

= 0.
x2 y 2
z 2
c2 t2

(B.1)

Transforming t t0 ,

t0 x0
y 0 z 0
=
+
+
+
.
t
t t0
t x0
t y 0
t z 0
Using x0 = x vt, y 0 = y, z 0 = z, t0 = t, we get

v
0
t
t0 x

2



=
v
t2
t t0
t x0
0 2

t x0 2
y 0 2
z 0 2
=
+
+
+
t t02
t x0 t0
t y 0 t0
t z 0 t0
0 2

t
x0 2 y 0 2
z 0 2
v
+
+
+
t t0 x0
t x02
t y 0 t0
t z 0 t0
2
2
2



=
2v 0 0 + v 2 02 .
02
t
x t
x
45

(B.2)

Transforming x x0

t0 x0
y 0 z 0
=
+
+
+
x
x t0
x x0
x y 0
x z 0

=
x0
2

2
=
x2
x02
The y-axis and z-axis do not change, we have

(B.3)

2
2
=
y 2
y 02

(B.4)

2
2
=
.
z 2
z 02

(B.5)

and

Substituting all equations of transformation into equations (B.1),

2
2
2
1 2 2v 2
v2 2
+
+

,
x02 y 02 z 02 c2 t02
c2 x0 t0
c2 x02
we finally have equation (2.5):
c2 v 2 2 2v 2
2
2
1 2
+
+
+

= 0
c2 x02
c2 t0 x0 y 02 z 02 c2 t02

B.2

(B.6)

Poissons equation for Newtonian gravitational

field

Considering a close surface S enclosing a mass M. The quantity g ndS is defined as

the gravitational flux passing through the surface element dS where n is the outward
unit vector normal to S. The total gravitational flux through S is then given by
Z
Z
er n
g ndS = GM
dS.
(B.7)
r2
S
S
By definition (er n/r2 dS) = cos dS/r2 is the element of solid angle d subtended
at M by the element surface dS. Therefore equation (B.7) becomes
Z
Z
g ndS = GM d = 4GM.
S
Z
= 4G (r)dV
v

46

From application of divergence theorem, the left side of this equation can be rewritten
Z
Z
g ndS = gdV.
(B.8)
S

Therefore we obtain
g = 4G.
Recall that
g = .

(B.9)

We therefore have result as in equation (3.57),

2 = 4G.

B.3

Variation of Cristoffel symbols : abc

From equation (3.14) we have

1 ad
abc =
g b gdc + c gbd d gbc
2

1
abc =
g ad b gdc + c gbd d gbc + g ad b gdc + c gbd d gbc .
2
2
ad
Considering variation of g ,
g ad = f d d f g ad
= f d gde g ef g ad
h
i

d ef
ad
ad
= f g gde g g gde
h
i
= f d g ef (e a ) g ad gde
= g ed g ad gde ,
therefore

1
abc = g ed g ad gde b gdc + c gbd d gbc + g ad b gdc + c gbd d gbc .
2
2

1
= g ad ebc ged + g ad b gdc + c gbd d gbc
2

1 ad
=
g b gdc + c gbd d gbc 2ebc ged
2
1 ad
=
g b gdc ebc ged ebd gec + c gbd ebc ged edc geb
2

d gbc + ebd gec + edc geb

1 ad
abc =
(B.10)
g b gdc + c gbd d gbc .
2
47

Bibliography
[1] R. dInverno, Introducing Einsteins Relativity, Oxford University Press (1996)
[2] T. L. Chow, General Relativity and Cosmology: A First Course, Wuerz (1994).
[3] B. F. Schutz, A First Course in General Relativity, Cambridge University Press
(1990).
[4] S. M. Carroll, An Introduction to General Relativity: Spacetime and Geometry,
Wesley (2004).
[5] R. M. Wald, General Relativity, The University of Chicago Press (1984).
[6] H. A. Chamblin and H. S. Reall, Nucl. Phys. B 562 (1999) 133 [arXiv:hepth/9903225].
[7] Y. V. Shtanov, arXiv:hep-th/0005193.
[8] R. Maartens, Living Rev. Rel. 7 (2004) 7 [arXiv:gr-qc/0312059].
P. Brax, C. van de Bruck and A. C. Davis, Rept. Prog. Phys. 67 (2004) 2183
[arXiv:hep-th/0404011].
E. Papantonopoulos, Lect. Notes Phys. 592 (2002) 458 [arXiv:hep-th/0202044].

48