Académique Documents
Professionnel Documents
Culture Documents
Relativity
Daniel Erenso
Introduction ix
2 Manifolds 33
2.1 What is a Manifold? . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.2 Curves and surfaces in a Manifold . . . . . . . . . . . . . . . . . 35
2.3 Coordinate transformations and summation convention . . . . . . 38
2.4 The Riemannian geometry . . . . . . . . . . . . . . . . . . . . . . 42
2.5 Intrinsic and extrinsic geometry and the metric . . . . . . . . . . 45
2.6 Length, area, and volume . . . . . . . . . . . . . . . . . . . . . . 52
2.7 Local Cartesian coordinates and tangent space . . . . . . . . . . 60
2.8 The signature of a manifold . . . . . . . . . . . . . . . . . . . . . 65
iii
iv CONTENTS
6 Electromagnetism 147
6.1 The Lorentz force . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
6.2 The charge and the current density . . . . . . . . . . . . . . . . . 149
6.3 The electromagnetic …eld equations . . . . . . . . . . . . . . . . . 151
6.4 Electromagnetism in the Lorentz gauge . . . . . . . . . . . . . . 154
6.5 Electromagnetism in arbitrary coordinates . . . . . . . . . . . . . 158
6.6 Equation of motion for a charged particle . . . . . . . . . . . . . 160
This material is my lecture note for General Relativity course (PHYS 4800) at
Middle Tennessee State University.
vii
viii PREFACE
Introduction
ix
x INTRODUCTION
Chapter 1
In this chapter we review the basic notions underlying in the Newtonian and
special relativistic viewpoints of space and time.
1
2 CHAPTER 1. THE SPACETIME OF SPECIAL RELATIVITY
Figure 1.1: Two inertial reference frames. Inertial frame S 0 is moving with a
constant velocity along the poesitive x relative to inertial frame S.
d2 X d2 Y d2 Z
2
= 2
= 2 =0 (1.1)
dt dt dt
and
d2 X 0 d2 Y 0 d2 Z 0
= = = 0: (1.2)
dt2 dt2 dt2
Consider the inertial reference frame S 0 in Fig (1.1). It is moving with a constant
velocity v in the positive x direction
~v = v^
x
~v = v^
x (1.3)
Taking this velocity into account, we can at least expect the x position of the
event for an observer O0 (x0 ) would depend on both x and t. We make a linear
relationship assumption
x0 = Dt + Ex (1.4)
Let’s make a similar assumption for time t0 too, why not?
t0 = At + Bx (1.5)
x = vt
we have
x0 = Dt + Ex ) 0 = Dt + Evt ) D = Ev (1.7)
For x = 0, we also require
x0 = vt0
which leads to
t0
x0 = Dt + Ex ) vt0 = Dt ) D = v : (1.8)
t
Substituting x = 0 into
t0 = At + Bx (1.9)
we …nd
t0 = At (1.10)
so that
t0
D= = v Av: (1.11)
t
From Eqs. (1.7) and (1.11) we see that
E = A; D = Av
t0 = At + Bx; (1.12)
0
x = A (x vt) ; (1.13)
0
y = y; (1.14)
z0 = z; (1.15)
A = 1; B = 0
which leads to
t0 = t; (1.16)
0
x = x vt; (1.17)
0
y = y; (1.18)
z0 = z; (1.19)
2 2 2 2
) (c t0 ) ( x0 ) ( y0 ) ( z0)
2 2
= c 2 A2 A2 v 2 ( t) + c2 B 2 A2 ( x) + 2A c2 B + Av ( x) ( t)
+2A c2 B + Av ( x) ( t) = 0 (1.32)
Now referring to Eq. (1.29), we have
2 2 2 2
(c t0 ) ( x0 ) ( y0 ) ( z0) =
2 2 2 2
(c t) ( x) ( y) ( z)
so that using the result in Eq. (1.32)
1
c 2 A2 A2 v 2 = c2 ) A = q = (1.33)
v2
1 c2
v
2A c2 B + Av = 0 ) c2 B + Av = 0 ) B = A= A= (1.34)
c2 c c
where
v
= ; (1.35)
c
1
= p : (1.36)
2
1
Substituting Eqs. (1.33) and (1.34) into Eqs. 1.25) and (1.26), the Lorentz
transformation is given by
t0 = t; (1.41)
0
x = x vt; (1.42)
y0 = y; (1.43)
0
z = z: (1.44)
ct = (ct0 + x0 ) ; (1.46)
0 0
x = (x + c t ) ; (1.47)
y = y0 ; (1.48)
0
z = z: (1.49)
or in a Matrix form
y0 cos ( ) sin ( ) y
= : (1.52)
x0 sin ( ) cos ( ) x
1.2. AXES ROTATION VS LORENTZ TRANSFORMATION 7
Figure 1.2: A Cartesian coordinate system rotated about the z-axis in a coun-
terclockwise direction by an angle : We named the y-axis "ct".
Let’s omit the y and z coordinate, we can write the Lorentz transformation as
ct0 ct
= (1.59)
x0 x
At least comparing the diagonal elements in Eqs. (1.53) and (1.59), one may be
tempted to say, = cos ( ) ; is some angle of rotation: However, since
1 1; v = 0
=q =
v2 1; v = c
1 c2
8 CHAPTER 1. THE SPACETIME OF SPECIAL RELATIVITY
trigonometric functions can not be an option to consider. But we know that the
hyperbolic function
e +e 1; =0
cosh ( ) = = (1.60)
2 1; !1
we …nd
1 1; =1
= tanh ( )= : (1.63)
0; =0
Now using
= cosh ( ) ; = sinh ( ) ; (1.64)
we can express the "Boost" as
Now comparing Eqs. (1.52) and (1.65), we can see that the boost has essen-
tially the same structure as coordinate-axis rotation except the trigonometric
functions are replaced by hyperbolic function. Therefore, the physics of rela-
tivity (for an inertial frames of reference) is a coordinate transformation in the
Minkowski spacetime! We can easily see this by considering the S and S 0 frame
of references in the standard con…guration (i.e. S 0 is moving with a constant
velocity v in the positive x-direction). Let’s omit the y and z coordinates so
that an event can be described by the coordinates (ct; x) in an inertial frame
S and by coordinates (ct0 ; x0 ) in the S 0 inertial frame. In Fig. 1.3 S 0 frame
is constructed by rotating ct axis in a clockwise direction and the x axis in a
counterclockwise direction by an angle ; which is given by Eq. (1.63), as shown
in Fig. ??. Now if we decompose ct and x into components along ct0 and x0 ; we
can easily see that
Note that the decomposition of the vectors form a parallelogram not a rectangle
like the Cartesian coordinate axes rotation. Now using y 0 = y and z 0 = z for the
boost, we have
or using matrices
2 3 2 32 3
ct0 cosh ( ) sinh ( ) 0 0 ct
6 x0 7 6 sinh ( ) cosh ( ) 0 0 7 6 x 7
6 0 7=6 76 7: (1.66)
4 y 5 4 0 0 1 0 54 y 5
z0 0 0 0 1 z
Show that the interval squared between the two events in problem 1 is invariant.
in the Minkowski spacetime is invariant under the boost. From the expression
in Eq. (1.68), we can also see that it could be positive, negative or zero. This
sign which is de…ned as
8 2
< ( s) > 0 timelike
2
( s) = 0 lightlike
: 2
( s) < 0 spacelike
Timelike: we can …nd an inertial frame in which the two events occur at the
same spacial coordinate.
Spacelike: we can …nd an inertial reference frame on which the two events
occur at the same time coordinate.
Before we see what is the lightcone is, let’s consider a cone in the Euclidean
geometry shown in Fig.1.5. A point on the surface of a cone with coordinates
1.3. THE INTERVAL AND THE LIGHTCONE 11
Figure 1.5: A cone of radius R and height H with its apex centered at the
origin.
1 R
# = 2 tan : (1.70)
H
The equation of a cone can then be expressed as
2
R
x2 + y 2 = z (1.71)
H
For a cone with apex centered at (x1 ; y1 ; z1 ) instead of (0; 0; 0) as shown in Fig.
1.6, one can write
R R
x x1 = (z z1 ) cos (') ; y y1 = (z z1 ) sin (') ; (1.72)
H H
or
R R
x= z cos (') ; y = z sin (') ; (1.73)
H H
where
x=x x1 ; y = y y1 ; z = z z1 :
The equation of the cone becomes
2
2 2 2 R
( x) + ( y) = ( z) : (1.74)
H
12 CHAPTER 1. THE SPACETIME OF SPECIAL RELATIVITY
Figure 1.6: A cone of radius R and height H with its apex centered at (x1 ; y1 ; z1 ).
becomes
2 2 2
( s) = (c t) ( x) ( y) (1.77)
2
For a lightlike ( s) = 0; we …nd
2 2
( x) + ( y) 2
= ( t) (1.78)
c2
which is equation of a cone and it is known as the lightcone. We note that
8 2
< ( s) > 0; Outside (timelike)
2 2 2
(c t) ( x) ( y) = ( s) = 0; On the lightlike (lightlike) (1.79)
: 2
( s) < 0; Inside (spacelike)
Now let’s omit both y and z axis and consider the four di¤erent events, Event 1,
Event 2, Event 3, and Event 4 described by two coordinates (ct1 ; x1 ), (ct2 ; x2 ),
2
(ct3 ; x3 ), and (ct4 ; x4 )The physical interpretation of spacelike ( s) < 0 and
2
timelike ( s) > 0 is what we stated earlier. When the interval is Timelike,
1.3. THE INTERVAL AND THE LIGHTCONE 13
we can …nd an inertial frame in which the two events occur on the same spacial
coordinate; and when it is Spacelike, we can …nd an inertial reference frame on
which the two events occur at the same time coordinate. In order to understand
this let’s consider the boost S 0 with both the y 0 and z 0 coordinates omitted. In
Fig. 1.8 we can see that in the reference frame S 0 Event 1 and Event 2 which
are timelike and occurred at di¤erent time and di¤erent place in reference frame,
S we can see that on the boost frame both have the same spacial coordinates.
This meant we found an inertial reference frame (the boost) on which the two
events occurred at the same point in space. On the other hand Event 1 and
Event 4 which are spacelike events on the S inertial reference frame, on the S 0
frame both events have the same time coordinates. This means we are able to
…nd an inertial reference frame on which Event 1 and Event 4 occurred at the
same time.
where x01 and x02 are the x-coordinates of his feet and his head as measured by
the woman. Using the Lorentz transformation
x0 = (x c t) ; (1.81)
we have
Suppose the woman took the measurement for x2 and x1 at the same time, we
have
t1 = t2 = t
so that one can write the proper length (height) of the man as
r
l0 v2
l0 = (x2 x1 ) = l ) l = = l0 1 ; (1.83)
c2
where
l = x2 x1 :
is the height of the man measured by an observer on the earth (the S inertial
frame). Obviously, the observer on earth measures the Man’s contracted height
(length) by a factor of 1= :
In their journey in the deep space suppose they got pregnant and the man
had to deliver their baby. Their is a clock aboard on the spacecraft and the
man measured the time interval between the beginning of the labor and the
arrival of their baby. Let this time be T0 = t0 = t02 t01 ; where t01 is the time
at which the labor began (Event 1) and t02 (Event 2)is the time at which the
baby arrived as recorded. Similarly, an observer on earth (frame S) recorded
the time of these two events t2 and t1 ; respectively, using his own Clock. Using
the inverse Lorentz transformation for time, where v is replaced by v in Eq.
(1.46), one can write
0 0
t1 = t1 + vx01 ; t2 = t2 + vx02 ; (1.84)
Suppose the observer on S 0 recorded the two events at the same place
then using
0 0
T0 = t0 = t2 t1 :
16 CHAPTER 1. THE SPACETIME OF SPECIAL RELATIVITY
we …nd r
T0 v2
T = T0 = q ) T0 = 1 T (1.87)
1 v2 c2
c2
The result in Eq. (1.87) showsq that the moving clock (on the S 0 frame) ticks
2
more slowly by a factor of 1 vc2 than the clock on a rest frame (the S frame).
Homework Problem 4 : Suppose the couples on the spacecraft celebrated
their child (a girl) sweet sixteen birthday as measured by a clock on board the
spacecraft (S 0 ). The girl is about 1:6m tall as measured by her parents. Assume
the spacecraft is traveling with constant velocity, v = 0:8c, where c is the speed
of light in vacuum.
(a) What would be the age of the girl as measured by an observer on earth (S
inertial frame).
(b) How tall is the girl as measured by an observer on earth (S inertial frame).
in the S 0 frame as the interval is invariant.We can calibrate the length scale for
the ct and x axis so that one may write for spacelike or timelike in both S and
S 0 inertial frames
2 2
(c t) ( x) = 1: (1.91)
The graph for these equations is shown in Fig. 1.10 (using Mathematica)Let
the point described by the coordinates (ct = 0; x = 0) be O, (ct = 0; x = 1)
be A; and (ct = 1; x = 0) be B respectively. OA represents a unit length
and OB a unit time on the S frame (Fig. 1.11). We know that the interval
is invariant in the Minkowski spacetime. This means if we rotate the ct axis
clockwise and x axis counterclockwise by = sinh 1 [v=c] ; we …nd the boost
(the S 0 inertial reference frame) with coordinates ct0 and x0 on which the interval
remains invariant
2 2
(c t0 ) ( x0 ) = 1 (1.92)
We can easily see the length contraction and time dilation from this invariance
and Fig. 1.11
Consider a proper length on the S 0 inertial reference frame which is OC.
Suppose OC is a meterstick (OC = l0 = 1m which is the length measured by
18 CHAPTER 1. THE SPACETIME OF SPECIAL RELATIVITY
As one can can clearly see from Fig. 1.11, Point D on the hyperbolae has a time
coordinate (in the S frame) that is greater than one, T > 1.
Z2 Z2 p
s= ds = c2 dt2 dx2 dy 2 dz 2 : (1.94)
1 1
Figure 1.12: The worldline of a photon (solid line) and a massive particle (broken
line).
the worldline for a particle must lie within the lightcone as shown in Fig. 1.12.
This is also required by relativistic mechanics as it prohibits the acceleration of
a massive particle to speeds greater than or equal to the speed of light c: Let’s
Figure 1.13: An alien spaceship traveling with a speed v along the positive
x-direction relative to the earth.
consider the inertial reference frame at rest, S; be our planet earth. Suppose a
man saw an alien spaceship passing by him with a velocity, v; in the positive x-
direction, as shown in Fig. 1.13. The man saw the alien’s spaceship at a position
20 CHAPTER 1. THE SPACETIME OF SPECIAL RELATIVITY
(x; y; z) at time t near the earth. In the Minkowski spacetime we describe this
even by (t; x; y; z). The worldline of the alien spaceship can be described by a
curved line (inside a cone). Assuming the velocity of the alien’s spaceship, v,
changes magnitude and direction along the x-axis, the worldline for the alien
could be the wiggly line shown in Fig. 1.14 passing through a light cones in
the Minkowski spacetime. This wiggly worldline of the alien is de…ned by the
Figure 1.14: The worldline of the alien. Note that we have omitted the y and z
coordinates and considered a time dependent velocity.
interval, ds;
ds2 = c2 dt2 dx2 ; (1.97)
that leads to a function, x (t) ; that depends on time, t: Another convenient al-
ternative way of expressing the worldline is to use some invariantly de…ned time
parameter, ; that monotonically changes over the alien (or the particle) world-
line. We can then use this parameter to de…ne the worldline using, (t ( ) ; x ( )).
This parameter, ; is known as the proper time: Generally, it is de…ned as
c2 d 2
= ds2 = c2 dt2 dx2 dy 2 dz 2 : (1.98)
For dy = dz = 0; we have
v2
c2 d 2
= c2 dt2 dx2 ) d 2
= dt2 1 ;
c2
1.6. WORLDLINE AND PROPER TIME 21
Z r r
t+
v2 v2
= 1 dt = = 1 t: (1.100)
t c2 c2
Note that here the alien’s spaceship is considered to have a constant velocity
relative to the observer on earth.
Let’s imagine that the alien turned the spaceship engine o¤ between this two
events so that, v = 0: The proper time becomes
= t; (1.101)
which is the same as the time interval between these two events as measured
by the observer clock on earth (i.e. the frame S; that is at rest). It means the
proper time, ; is just the time coordinate recorded by clocks at rest. Generally,
if at any instant in the history of the particle (or the alien in this case), we
introduce an instantaneous rest frame S 0 such that the particle is momentarily
at rest in the S 0 , then the proper time is simply the time recorded by the
clock that moves along with the particle (or in this case the alien spacecraft).
Therefore, the proper time, ; is an invariantly de…ned quantity.
22 CHAPTER 1. THE SPACETIME OF SPECIAL RELATIVITY
describing a photon emitted from the alien spaceship recorded by the observer
on earth. These events described two successive wavecrests of the photon. The
two successive wavecrests, when the photon is emitted are referred as Event 1
(E1) and Event 2 (E2) and recorded as (t1 ; x1 ) and (t2 ; x2 ), respectively, by the
observer on earth. When these wavecreasts received by the observer on earth
are referred as Event 3 (E3) and Event 4 (E4) and recorded as (t3 ; x) and (t4 ; x),
respectively (See Fig 1.17). Note that the x-coordinates for E3 and E4 are the
same as the man received the photon at the same position. From Eq. (1.100),
the proper time for E1 and E2 as measured by an observer on S 0 (the alien)
moving with a speed v is given by
r r
v2 v2
21 = 1 2
(t2 t1 ) = 1 t21 : (1.102)
c c2
On the other hand the proper time for an observer on the S frame (the man
on earth) for E3 and E4, since, v = 0, Eq. (1.100) gives
43 = t4 t3 = t43 : (1.103)
1.7. THE DOPPLER EFFECT 23
Figure 1.17: Doppler e¤ect. Two adjacent wavecrest when a photon is emitted
(E1 and E2) from the alien spaceship and when it is detected by an observer on
earth (E3 and E4).
Along the worldline joining E1 to E3 and E2 to E4, (see Fig.1.18 )the interval
is zero since we are considering a photon (Lightlike, ds = 0)
c2 d 2
= ds2 = c2 dt2 dx2 dy 2 dz 2 = 0: (1.104)
For dy = dz = 0; we have
c2 dt2 dx2 = 0: (1.105)
so that for the worldline joining E1 and E3, we can write
Z 3 Z 3 Z t3 Z x
cdt = dx ) cdt = dx: (1.106)
1 1 t1 x1
The minus sign, as it can be seen from Fig. 1.17, is because of the photon is
traveling in the negative direction. Similarly, for the worldline connecting E2
and E4, we can write
Z 4 Z 4 Z t4 Z x
cdt = dx ) cdt = dx: (1.107)
2 2 t2 x2
Noting that
Z t4 Z t1 Z t3 Z t4 Z t2 Z t3 Z t4
cdt = cdt + cdt + cdt = cdt + cdt + cdt
t2 t2 t1 t3 t1 t1 t3
Z t3
= c t21 + c t43 + cdt
t1
24 CHAPTER 1. THE SPACETIME OF SPECIAL RELATIVITY
so that using Z Z Z
x x1 x
dx = dx + dx: (1.110)
x2 x2 x1
one …nds
Z x Z x1 Z x
c t21 + c t43 dx = dx dx
x1 x2 x1
Z x1
) c ( t43 t21 ) = dx = x2 x1 = x21 (1.111)
x2
43 t43
=q
21 v2
1 c2 t21
Thus the time interval between the two successive wavecrests as observed by
an observer on earth, S rest frame ( 43 ) and the alien on the spaceship, S 0
moving with speed v, ( 21 ) are related by
s
43 1 + vc 1 + vc
=q = : (1.114)
21 1+ v 1 v 1 vc
c c
Note that this times are the period of the emitted photon that can be related
to the corresponding frequencies for an observer on S (f ) and on S 0 (f 0 ) by
43 = 1=f; 21 = 1=f 0
we …nd s
v
f 1 c
= v : (1.115)
f0 1+ c
Figure 1.19: Two Alien spaceship traveling with di¤erent velocity relative to an
observer on earth.
we can write
dx0
cdt = (cdt0 + dx0 ) = c+ dt0 = (c + u0x ) dt0 ;
dt0
vu0x
) dt = 1+ dt0 (1.118)
c2
dx0
dx = (dx0 + c dt0 ) = +c dt0 = (u0x + v) dt0 ; (1.119)
dt0
dy = dy 0 ; dz = dz 0 ; (1.120)
so that
dx (dx0 + c dt0 )
= ; (1.121)
cdt (cdt0 + dx0 )
dy dy 0
= ; (1.122)
cdt (cdt0 + dx0 )
dz dz
= : (1.123)
cdt (cdt0 + dx0 )
1.8. VELOCITY AND ACCELERATION 27
1 v 1 u0
v = tanh ; u0 = tanh : (1.128)
c c
show that
(a)
(b)
tanh ( v ) + tanh ( u0 ) u0 + v
u = ct tanh ( v + u0 )=c =
1 + tanh ( v ) tanh ( u0 ) 1 + u0 v=c2
we have
v2
ux v dux dux cv2 (ux v) dux 1 c2
du0x =d vux = vux + =
1 1 vux 2 vux 2
c2 c2 1 c2 1 c2
dux
) du0x = ; (1.131)
2 vux 2
v 1 c2
2 q 3 " #
v2
uy 1 uy cv2 dux
du0y = d4
c2
5= 1 duy
vux + ; (1.132)
v vux 2
1 c2 ux v 1 c2 1 c2
2 q 3 " #
v2
uz 1 c2 1 duz uz cv2 dux
du0z = d4 v
5= vux + ; (1.133)
1 vux 2
c2 ux v 1 c2 1 c2
where
1
v =q : (1.134)
v2
1 c2
v v dx
ct0 = v (ct x) ) dt0 = v dt dx = v 1 dt(1.135)
c2 c2 dt
vux
) dt0 = v 1 dt (1.136)
c2
then the acceleration becomes
,
du0x dux h vux i ax
0
ax = 0 = 2 v 1 dt = ; (1.137)
dt 2 1 vu x c 2 3 1 vux 3
v c2 v c2
" #,
du 0
1 du u v
du h vux i
y y y 2 x
a0y = 0 = vux +
c
2 v 1 dt
dt v 1 c2 1 vu x c2
c2
1 uy v
) a0y = 2 ay + ax (1.138)
2 1 vu x 2 c2 1 vux 3
v c2 v c2
" #,
du0z 1 duz uz cv2 dux h vux i
az = 0 = vux + 2 v 1 dt
dt v 1 c2 1 vu x c2
c2
1 uz v
) az = az + ax (1.139)
2 1 vux 2 2 c2 1 vux 3
v c2 v c2
1.8. VELOCITY AND ACCELERATION 29
du
ux = u (t) ; uy = uz = 0 ) ax = a = ; ay = az = 0; (1.140)
dt
so that its velocity and acceleration as observed by the alien in the second
spaceship be
u v
u0x = ; u0 = 0; u0z = 0 (1.141)
1 cv2 ux y
and
a
a0x = a0 = ; a0y = 0; az = 0; (1.142)
3 vu 3
v 1 c2
Figure 1.20: When the velocity of the second alien spaceship equals the …rst
alien spaceship, it becomes an instantaneous reference frame (IRF).
kt ( ) k kx ( ) k
= sinh ; = cosh 1: (1.156)
c c c c
The parametric graph for the worldlines, for Alien 1 in the …rst spaceship that
is uniformly accelerating, f ( ) = k; (the red curve) and for the observer on
earth at, x = 0; that is at rest on the S frame (green). The blue line is the
event horizon for Alien 1. Events that are beyond the event horizon will never
be seen by Alien 1.
32 CHAPTER 1. THE SPACETIME OF SPECIAL RELATIVITY
Figure 1.21: The worldline for the accelerating spaceship (red) and the ob-
server on earth (green). The dotted blue line is the horizon for the accelerating
spaceship.
Figure 1.22: The light cones for the accelerating spaceship (red) and the ob-
server on earth (green). The dotted blue line is the horizon for the accelerating
spaceship.
Chapter 2
Manifolds
begins to rotate. It is free to rotate about the x-axis, y-axis, or z-axis. Let’s say
we want to describe the angular position of the center of mass of the meter stick
over a period of time, = 10s; with a time interval of 2s. How many independent
parameters, that depend on time, do we need to describe the angular position
of the center of mass of the meterstick relative to its initial position at t = 0?
Well the answer is simple. We need three independent parameters, the Euler
33
34 CHAPTER 2. MANIFOLDS
angles, ( 1 (t) ; 2 (t) ; 3 (t)) which describes the rotation about the x, y, and
z axes at a given instant of time. Then over the 10 second interval we have a
set that consist of 5 points
1 2 3 1 2 3 1 2 3
(2) ; (2) ; (2) ; (4) ; (4) ; (4) ; (6) ; (6) ; (6) ;
1 2 3 1 2 3
(8) ; (8) ; (8) ; (10) ; (10) ; (10)
We can make the time interval in…nitesimal to continuously describe the angular
position of the center of mass of the meterstick. The resulting set of points form
a Manifold of dimension three.
Let’s consider another example of a Manifold. In classical mechanics you
may have studied what is called the phase space. In this space you can describe
the state of a particle over a period of time using the three coordinates of
space locating the position of the particle and the three coordinates of speed
(or momentum) describing how fast the particle is moving at a given instant of
time. In Cartesian coordinate system we use the independent parameters (x (t) ;
y (t) ; z (t)) for position and (x_ (t) ; y_ (t) ; z_ (t)) for the speed of the particle. I can
represent these independent parameters by (x1 (t) ; x2 (t) ; x3 (t) ; x4 (t) ; x5 (t) ;
x6 (t)). So when you describe the state of the particle say from t = 0 to t = t0 ;
you can use in…nitesimal time interval so that you will have a set of points that
can be parameterized continuously in terms of (x1 (t) ; x2 (t) ; ::: x6 (t)). These
set of points form a Manifold of dimension six.
Now let’s apply this to the Minkowski spacetime where we have three space
coordinates (x (t) ; y (t) ; z (t)) and time, t. This forms a four dimensional man-
ifold with four coordinates each parametrized by the proper time ; (x1 ( ) ;
x2 ( ) ; x3 ( ) ; x4 ( )). Therefore, a manifold is any set that can be continu-
ously parameterized. Therefore, an N dimensional manifold, M; of points is
one for which N independent real coordinates (x1 ; x2 ; x3 ; :::; xN ) are required
to specify any point completely.
A manifold is Continuos: if you pick any point, p, on the Manifold and you
can …nd another points whose coordinates di¤er in…nitesimally from the point
p.
A manifold is di¤ erentiable: if you pick any point, p, on the Manifold and
you can …nd a scalar function that is di¤erentiable at that point p.
Coordinates of a Manifold M : a point in an N -Dimensional Manifold is
represented by the coordinates (x1 ; x2 ; x3 ; ::: xN ) which we represent by xa
where it is understood that a = 1; 2; 3::: N:
Degeneracy in a Manifold : sometimes it may not be possible to cover the
whole manifold with only one none-degenerate coordinate system. Example is
a plane in polar coordinate system ( ; '). A plane is a two dimension Manifold.
(called R2 ). A plane in polar coordinates has a degeneracy at the origin since
' is indeterminate at the origin. (Fig. 2.2)
Coordinate patches: these are coordinate systems that covers a portion of
the Manifold where we have degeneracy. For example the surface of a sphere is
a two dimensional Manifold (called S 2 ). It can be described by two independent
coordinates (x1 = ; x2 = ') except at two points on the Manifold. These are
2.2. CURVES AND SURFACES IN A MANIFOLD 35
Figure 2.3: The north and south pole on the surface of sphere are degenerate
in polar coordinates.
the north and south pole where ' is indeterminate (or there is degeneracy).
(Fig.2.3) There is no coordinate system that covers the entire sphere without
running into these two degenerate points. In this case the smallest number of
patches we need is two.
Atlas: an atlas is a set of coordinate patches that covers the whole Manifold.
f x1 ; x2 ; x3 ; :::xN = 0: (2.5)
The surface shown in blue in Fig. 2.4 in the 3D manifold needs two parameters
to de…ne it
x1 = 2 cos(u1 ); x2 = sin (u2 ) ; x3 = u2 : (2.6)
Note that this surface is in 3D manifold and it parameterized by two coordinates
(u; v), (M = N 1 = 3 1 = 2; it is a hypersurface).
2.2. CURVES AND SURFACES IN A MANIFOLD 37
Example 2.1 Let’s consider the 3-D Euclidean Manifold. A sphere is a High-
persurface since M = 2; (Fig.2.5). A point on a sphere is de…ned by
x2 + y 2 + z 2 = a 2 ; (2.7)
where a is the radius of the sphere. We note that in this case the surface
of the sphere is a Hypersurface that can be de…ned by the equation
2 2 2
g x1 ; x2 ; x3 = x1 + x2 + x3 a2 = 0; (2.8)
and u3 de…ned by
x1 = u1 sin (u2 ) cos (u3 ) ; x2 = u1 sin (u2 ) sin (u3 ) ; x3 = u1 cos (u2 )
we can write the equation that de…ne the surface of the sphere (M = 2)
that is embedded in a 3D manifold (N = 3) using only one parameter (by
eliminating the N 1 = 2; parameters) as
2 2
g x1 ; x2 ; x3 = (u1 sin (u2 ) cos (u3 )) + (u1 sin (u2 ) sin (u3 ))
2
+ (u1 cos (u2 )) a2 = 0 ) g x1 ; x2 ; x3 = u21 a2 = 0 (2.9)
dimension N (M < N ).
g1 x1 ; x2 ; x3 ; :::xN = 0; (2.10)
1 2 3 N
g2 x ; x ; x ; :::x = 0;
1 2 3 N
g3 x ; x ; x ; :::x = 0
:
:
:
1 2 3 N
gN M x ; x ; x ; :::x = 0
so that 2 3 2 32 3
@g @x01 @x02 @x03 @g
@x1 @x1 @x1 @x1 @x01
4 @g 5=6
4 @x01 @x02 @x03 74
5 @g 5 (2.18)
@x2 @x2 @x2 @x2 @x02
@g @x01 @x02 @x03 @g
@x3 @x3 @x3 @x3 @x03
can be written as
2 @g 3 2 @x01 @x02 @x03
32 @x1 @x2 @x3
32
@g 3
@x1 @x1 @x1 @x1 @x01 @x01 @x01 @x1
4 @g2 5 = 6 76 74
1 2 3
4 @x01 @x02 @x03
54 @x @x @x
5 @g 5: (2.19)
@x @x2 @x2 @x2 @x02 @x02 @x02 @x2
@g @x 01
@x02 @x03 1 2 3 @g
@x @x @x
@x3 @x3 @x3 @x3 @x03 @x03 @x03 @x3
so that
1 1
A A = AA = 1: (2.23)
T
We note that the transpose, A ; is given by
2 1 1
3
@x1 @x @x
6 @x01 @x02
2
@x03
2 7
AT = 6
4
@x2 @x @x 7:
5 (2.24)
@x01 @x02
3
@x03
3
@x3 @x @x
@x01 @x02 @x03
Similarly for the inverse matrix, the transpose matrix which we express as
2 01 01 01
3
@x @x @x
@x0a 6 @x1
@x02
@x2
@x02
@x3
@x02 7
=4 @x1 @x2 @x3 5 (2.25)
@xb @x03 @x03 @x03
@x1 @x2 @x3
40 CHAPTER 2. MANIFOLDS
We note that
0a 0a
X @x @xb 0a N 0a
@x0a @x @x1 @x @x2 @x @xN
01
= 1 01
+ 2 01
+ :: N 01
= (2.30)
@x @x @x @x @x @x @x @xb @x01
b=1
2.3. COORDINATE TRANSFORMATIONS AND SUMMATION CONVENTION41
where the summation is evaluated at P . Similarly for the interval between x0a
and x0a + dx0a ; in the none-primed coordinate system, we can write
X @xa N
@xa 01 @xa 02 @xa
dxa = 01
dx + 02
dx ::: 0N dx0N = dx0b ; (2.34)
@x @x @x @x0b
b=1
and
N
X @xa 0b @xa 0b
dxa = 0b
dx = gb (x0 ) dx0b = dx : (2.36)
@x @x0b
b=1
42 CHAPTER 2. MANIFOLDS
The index a; which is known as the free index, can take any value from 1 to
N . The index b, which is known as the dummy index and it must be summed
up from 1 to N .
Let’s reconsider the worldline of the alien in the spaceship in the previous chap-
ter. Imagine there is a surface de…ned by the two coordinates (x; ct) represented
as x1 ; x2 : Let’s consider two points P and Q; as shown in Fig. 2.7. The coordi-
nates for the points are x1 ; x2 and x1 + dx1 ; x2 + dx2 ; respectively. Suppose
this surface is de…ned by the function s x1 ; x2 The interval between these two
@s 1 @s 2
ds = dx + dx
@x1 @x2
@s 1 @s 2 @s 1 @s 2
) ds2 = 1
dx + 2
dx 1
dx + dx
@x @x @x @x2
@s @s 1 1 @s @s 1 2 @s @s 1 2
) ds2 = dx dx + dx dx + dx dx
@x1 @x1 @x1 @x2 @x2 @x1
@s @s
+ 2 2 dx2 dx2 (2.38)
@x @x
2.4. THE RIEMANNIAN GEOMETRY 43
ds2 = g11 x1 ; x2 dx1 dx1 + g12 x1 ; x2 dx1 dx2 + g21 x1 ; x2 dx1 dx2
2 X
X 2
1 2 2 2
+g22 x ; x dx dx = gab (x) dxa dxb (2.39)
a=1 b=1
where we replaced
@s @s @s @s @s @s @s @s
g11 = ; g12 = ; g21 = ; g22 =
@x1 @x1 @x1 @x2 @x2 @x1 @x2 @x2
Consider a curved surface in 3-D Euclidean space. We know that this surface
can be de…ned by a function g, that depends on (x; y; z) in Cartesian, (r; ; ')
in spherical, or (r; '; z) in cylindrical coordinates. Suppose we represent these
coordinates by x1 ; x2 ; x3 ; then we may write the function that de…nes the
surface as
s (x) = g (x) = g x1 ; x2 ; x3 : (2.40)
Now let’s consider a point P on this surface that has coordinates x1 ; x2 ; x3 :
Suppose we consider another point Q with coordinates x1 + dx1 ; x2 + dx2 ; x3 + dx2 ;
we may de…ne the surface between these two points as ds2 . This displacement
is just the di¤erential of Eq. (2.40)
@g @g @g @g @g @g
ds2 = dx1 + dx2 + dx3 : dx1 + dx2 + dx3
@x1 @x2 @x3 @x1 @x2 @x3
(2.41)
or using Einstein’s summation convention, the geometry of the surface between
the two points
3
! 3
!
X @g X @g b
2 a
ds = (ds) (ds) = dx dx
a=1
@xa @xb
b=1
3 X
X 3
@g @g a b
= dx dx : (2.42)
a=1 b=1
@xa @xb
@g @g @g @g
a b
= :
@x @x @xb @xa
Suppose the coordinates for an orthonormal set like the Cartesian, spherical, or
cylindrical, we note that
2 2
3
@g
6 @x1 0 0 7
6 2 7
G=6 6 0 @g
@x2 0 7
7 (2.46)
4 3 5
@g
0 0 @x3
N X
X N
ds2 = gab (x) dxa dxb ; (2.47)
a=1 b=1
@xa @xb 0c 0d
ds2 = gab (x) dx dx ; (2.50)
@x0c @x0d
or
0
ds2 = gcd (x0 ) dx0c dx0d ; (2.51)
where
a b
0 0 @x @x
gcd (x ) = gab (x) 0c 0d : (2.52)
@x @x
Note that x = x (x0 ) :
2.5. INTRINSIC AND EXTRINSIC GEOMETRY AND THE METRIC 45
(a) Intrinsic: when the geometry remains unchanged as viewed in the higher
dimensional manifold.
(b) Extrinsic: when the geometry is di¤erent as viewed in the higher dimen-
sional manifold.
~r ~r0 = (x x0 ) x
^ + (y y0 ) y^ + (z z0 ) z^ (2.54)
is zero,
~ (~r
N ~r0 ) = a (x x0 ) + b (y y0 ) + c (z z0 ) = 0: (2.55)
ax + by + cz = d; (2.56)
46 CHAPTER 2. MANIFOLDS
where
d = ax0 + by0 + cz0 : (2.57)
Using the notation x1 ; x2 ; x3 for (x; y; z) , we may write
where
d = ax10 + bx20 + cx30 : (2.59)
1 2
For a plane that depends on only (x ; x )we may write the interval between
Now let’s consider the 2-D cylindrical surface which we can construct using
our 2-D plane. Suppose the cylinder has radius, a; with its axis along the
z-axis (which we call it x03 axis): Using cylindrical coordinates a point
on the surface of the cylinder can be described by (a; '; z) or using our
coordinates notations (a; x02 ; x03 ), we can de…ne the surface by the function
we may write the interval between two points on this surface, point P
and Q with coordinates (x1 ; x2 ; x3 ) and point Q with coordinates (x1 +
dx1 ; x2 + dx2 ; x3 + dx3 ) just using geometrical visualization, as
2 2 2
ds2 = dx1 + dx2 + dx3 : (2.62)
we …nd
2 2 2 2 2
ds2 = dx1 + dx2 + dx3 = a2 dx02 + dx03 : (2.63)
2.5. INTRINSIC AND EXTRINSIC GEOMETRY AND THE METRIC 47
x1 = ax02 ; x2 = x03
Such kind of geometry is not intrinsic and it is called extrinsic. Its cur-
vature is extrinsic and is a result of the way it is embedded in the three
dimensional space..
This is a 2-D surface embedded in a 3-D manifold. You can not obtain
this geometry from the plane geometry like the 2-D cylindrical geome-
try. Such kind of geometry is intrinsic. This means the geometry of a
sphere is intrinsically curved because we can not transform Eq.(2.65) to
the Euclidean form
2 2
ds2 = dx1 + dx2 (2.66)
over the whole surface by any coordinate transformation. Note that this
can be done locally but not for the whole spherical surface.
Example 2.4 Find the metric for a two-dimensional sphere of radius, a; em-
bedded in a 3-D Euclidean space both in Cartesian coordinates (x1 ; x2 ; x3 ).
Refer to Fig. 2.5
Solution: We recall that the line element in a 3-D Euclidean space is given by
2 2 2
ds2 = dx1 + dx2 + dx3 ; (2.67)
so that
x1 dx1 + x2 dx2
dx3 = r :
2 2
a2 (x1 ) + (x2 )
Then for a 2-D sphere embedded in a 3-D Euclidean space, the metric is
given by
0 12
2 2 B x1 dx1 + x2 dx2 C
ds2 = dx1 + dx2 +B
@ r C :
A (2.70)
2 2
a2 (x1 ) + (x2 )
x1 = x2 ' 0
Let’s use the coordinates (x01 ; x02 ; x03 ). de…ned by the transformation
Note that the origin is set at the north pole of the sphere at point as shown
in Fig. 2.5,. Then the interval can be written as
2
ds2 = dx01 cos x02 x01 sin x02 dx02
0 12
01 01
2 x dx
+ dx01 sin x02 + x01 cos x02 dx02 + @ q A (2.73)
2
a2 (x01 )
or
2 2
ds2 = g11 dx01 + g22 dx02 (2.76)
where
a2 2
g11 = 2 ; g22 = x01 (2.77)
a2 (x01 )
are the none zero elements of the metric tensor. We will see the use of
these elements of the metric tensor in the next section to determine length
and area of a 2-D sphere in a 3-D Euclidean manifold.
so that
x1 dx1 + x2 dx2 + x3 dx3
dx4 = r :
2 2 2
a2 (x1 ) + (x2 ) + (x3 )
We introduce the coordinates (r; ; ') which we represent (x01 ; x02 ; x03 ).
and de…ned by the transformation
or
x1 = x01 sin x02 cos x03 ; x2 = sin x02 sin x03 ; x3 = cos x02 :
Then the distance between P and Q shown by the red line (hypotenuse
side) can be expressed as
2
ds02 = dr2 + ds2s = dr2 + r2 sin2 ( ) d'2 + r2 d (2.82)
ds02 = dx2 + dy 2 + dz 2
so that
rdr
dx4 = p: (2.85)
a2 r2
Then the metric for a 3-D sphere in a 4-D Euclidean manifold is given by
2 r2 dr2
ds2 = ds02 + dx4 = dr2 + r2 sin2 ( ) d'2 + r2 d + ; (2.86)
a2 r2
52 CHAPTER 2. MANIFOLDS
where we used the notation x01 ; x02 ; x03 for the coordinates (r; ; ') and
identify the none zero elements of the metric tensor
a2
g11 = ; g22 = r2 ; g33 = r2 sin2 ( )
a2 r2
We will see the use of these elements of the metric tensor in the next
section to determine length, area, and volume of a 3-D sphere in a 4-D
Euclidean manifold.
the absolute value is because of the fact that for pseudo-Riemannian manifolds it
can be negative as is the case for spacelike in the Minkawski spacetime manifold.
For xa = xa (u) ; we recall write
dxa dxb
dxa = du; dxb = du (2.90)
du du
so that s
Z Q
dxa dxb
LP Q = gab (u) du: (2.91)
P du du
dxp pq 1; p=q
= = (2.93)
dxq 0; p 6= q
For a hypersurface, like the 2D sphere in the 3D manifold, the surface is de…ned
by the equation
x3 = constant ) dx3 = 0:
and the expression for the area reduces to
ZZ p ZZ p
A= jg22 (x) g11 (x)jdx2 dx1 + jg11 (x) g22 (x)jdx1 dx2
where one can absorb the factor 2 into the metric elements. For such coordinates
the metric is diagonal
2 2 2
ds2 = g11 dx1 + g22 dx2 ::: + gN N dxN : (2.94)
xa = constant; (2.95)
p
dA = jg11 g22 jdx1 dx2 (2.96)
xa = constant (2.97)
(a) The distance, D, from the origin to the perimeter on this surface along
a line of constant ' i:e: x02 = cons The origin is at the north pole as
shown in Fig. 2.6.
(b) The circumference of the circle with radius, = R i:e: x01 = R = cons :
(c) The area of the spherical surface enclosed by the perimeter with radius,
= R; (i.e. the surface shaded green in Fig. 2.6).
Solution:
(a) We recall that the metric for a 2-D sphere in a 3-D Euclidean manifold is
given by
2 2
ds2 = g11 dx01 + g22 dx02 ; (2.99)
where we used the notation x01 ; x02 for the coordinates ( ; ') and identify
the none zero elements of the metric tensor
a2 2
g11 = ; g22 = : (2.100)
a2 2
2.6. LENGTH, AREA, AND VOLUME 55
(c) The surface is de…ned by x03 =constant: Then according to Eq. (2.96), the
in…nitesimal area is given by
p
dA = jg11 g22 jdx01 dx02 : (2.106)
To cover the area (colored Aqua) shown in Fig 2.6, we should have for the
limits of integrations [0; R] for and [0; 2 ] for '
ZRZ2 " r #
a R2
A= p d d = 2 a2 1 1 : (2.108)
a2 2 a2
0 0
(a) the distance from its center to the surface of this sphere along constant
and constant .
becomes Z Q p
LP Q = jg11 (x0 ) dx01 dx01 j (2.114)
P
or
Z s Z
R R
a2 a 1 R
D= dr2 = p dr = a sin ; (2.115)
0 a2 r 2
0 a2 r2 a
which is the same as the result we obtained in the previous example.
(b) Across the equator, we have x01 = r = R; x02 = = =2 and obviously
dx01 = 0 and dx02 = 0: Then the length
Z Qq
LP Q = jgab (x0 ) dx0a dx0b j; (2.116)
P
g22 = r2 r=R
= R2
2
g33 = 2
r sin ( ) r=R
= R2 sin2 ( ) (2.119)
and the limit of integration for x02 is (0; ) and for x03 is (0; 2 ) ; the
surface area would be become
Z Z 2 q Z Z 2
2
A= 4
R sin ( ) d d = R2 sin ( ) d d = 4 R2 :
0 0 0 0
(2.120)
(d) In this case we are considering a 3-D sphere embedded in a 4-D Euclidean
space. For this space the volume is de…ned by x04 =constant. Therefore,
applying the relation in Eq. (2.98), an in…nitessimal volume in this 4-D
space is given by
p
dV = jg11 g22 g33 jdx01 dx02 dx03 : (2.121)
Thus using
a2
g11 = ; g22 = r2 ; g33 = r2 sin2 ( ) (2.122)
a2 r2
the volume bounded by the 2-D spherical surface of radius x01 = r = R
becomes
Z RZ Z 2 s 2 4 2
a r sin ( )
V = drd d
0 0 0 a2 r 2
Z R Z Z 2
r2 dr
= a p sin ( ) d d (2.123)
0 a2 r2 0 0
or
Z R 2 Z Z 2
x01 dx01 sin x02 dx02 dx03
V =a q q ;
0 2 0 0 2
a2 (x01 ) a2 (x01 )
Let’s evaluate the integral
Z
r2 dr
I= p (2.124)
a2 r2
Introducing the transformation de…ned by
r = a sin ( ) ) dr = a cos ( ) d
2.6. LENGTH, AREA, AND VOLUME 59
we have
Z Z 2 2 Z
r2 dr a sin ( ) a cos ( ) d
I= p = q = a2 sin2 ( ) d
a2 r2 a2 a2 sin2 ( )
Z
a2 a2 sin (2 )
= (1 cos (2 )) d =
2 2 2
Z 2 2
r dr a
)I= p = [ sin ( ) cos ( )] (2.125)
a2 r2 2
so that using
R
r = a sin ( ) ) sin ( ) = a; for r = R
(2.126)
0; for r = 0
and
q ( q
R 2
cos ( ) = 1 2
sin ( ) = 1 a ; for r = R (2.127)
0; for r = 0
one …nds
2 s 3
Z R 2 2 2
r dr a 4 1 R R R 5
p = sin 1 (2.128)
0 a2 r2 2 a a a
One must be able to recover the 3-D Euclidean space for a ! 1: This
means R a << 1 the result for the volume of a sphere with radius, R must
be that of the volume of a sphere with radius R in 3D Euclidean space
(V =. 43 R3 ). One can easily …nd from Eq. (2.129) using the approxima-
tions for R
a << 1;
s
3 2 2
R R 1 R R 1 R
sin 1 ' ; 1 =1 (2.130)
a a 3 a a 2 a
that gives
( !)
3 2
3 R 1 R R 1 R
V =2 a 1
a 3 a a 2 a
( )
3 3
3 1 R 1 R 4 3
=2 a + )V = R (2.131)
3 a 2 a 3
60 CHAPTER 2. MANIFOLDS
Homework :
(a) Express the circumference, the area, and the volume in terms of D in the
previous example and show that all have maximum values at
a
D= (2.132)
2
(b) Show that the total volume of this space is …nite and is equal to
2 3
V =2 a : (2.133)
(a) Show that circumference, C, and the area, A are still C = 2 R and A =
4 R2 :
(b) The distance D from the center of the sphere to the surface is
1
D = b sinh (R=b) (2.134)
and in this case show that A and V of the sphere are monotonically in-
creasing functions
for all points on the manifold. However, it is possible to …nd coordinates x0a
0
such that at the point P the new metric functions gab (x) satisfy the conditions
0
gab (xaP ) = ab ; (2.137)
0
@gab (x0 )
= 0: (2.138)
@x0c xa
P
2.7. LOCAL CARTESIAN COORDINATES AND TANGENT SPACE 61
We recall from Mathematical methods for any function, g (x) ; that is di¤ er-
entiable for all values of x in the speci…ed domain,
dn g(x)
exists for all n 0 and x 2 R;
dxn
one can write the series expansion about xp in the domain (Taylor series)
X1
1 dn g(x0 )
g(x) = (x xp ) n : (2.139)
n=0
n! dx0n x=xp
1 2
For a function of two variable g (x) = g x ; x ; this becomes
@g(x01 ; x02 )
g x1 ; x2 = g x1p ; x2p + (x1 x1p )
@x01 0 1 2
x =xp ;xp
8
@g(x01 ; x02 ) 1 < @ 2 g(x01 ; x02 )
+ (x2 x2p )2 + (x1 x1p )2
@x02 x0 =x1p ;x2p 2! : @ (x01 )2 0 x =x1p ;x2p
2 01 02
@ g(x ; x )
+2 (x1 x1p )(x2 x2p )
@x01 @x01 x0 =x1p ;x2p
9
@ 2 g(x01 ; x02 ) =
+ 2 (x2 x2p )2 + ::: (2.141)
@ (x02 ) ;
x0 =x1p ;x2p
From this expression you can imagine how it gets nasty for a function of N
variables, like the metric, gab (x) = gab x1 ; x2 ; x3 ; :::xN :
Example 2.8 Let’s reconsider the metric for 2D sphere embedded in a 3D
manifold in Cartesian coordinates
0 12
2 2 B x1 dx1 + x2 dx2 C
ds2 = dx1 + dx2 +B
@r
C :
A (2.142)
2 2
a2 (x1 ) + (x2 )
We note that for the 2D sphere in the 3D manifold, using the metric we
…nd
g11 x1p ; x2p = g22 x1p ; x2p = 1; g12 x1p ; x2p = 0
and
" 2 #
@g11 x01 ; x02 @ x01 x1p
= 2 2
@x01 @x01 a2 x01 x1p x02 x2p
x1p ;x2p
2 3
01 3
6 2 x x1p 2 x 01
x1p 7
=4 2 2 +h i 5
a2 x01 x1p x02 x2p 2 2 2
a2 x01 x1p x02 x2p
x1p ;x2p
=0
Similarly one can easily show that
" 2 #
@g11 x01 ; x02 @ x01 x1p
= 2 2
@x02 @x02 a2 x01 x1p x02 x2p
x1p ;x2p x1p ;x2p
2 3
2
6 2 x01 x1p x02 x2p 7
= 4h i 5 =0
2 2 2
a2 x01 x1p x02 x2p
x1p ;x2p
2.7. LOCAL CARTESIAN COORDINATES AND TANGENT SPACE 63
" 2 #
@g22 x01 ; x02 @ x02 x2p
= 2 2
@x01 @x01 a2 x01 x1p x02 x2p
x1p ;x2p x1p ;x2p
2 3
2
6 2 x01 x1p x02 x2p 7
= 4h i 5 =0 (2.145)
2 2 2
a2 x01 x1p x02 x2p
x1p ;x2p
" 2 #
@g22 x01 ; x02 @ x02 x2p
= 2 2
@x02 @x02 a2 x01 x1p x02 x2p
x1p ;x2p
2 3
02 3
6 2 x x2p 2 x 02
x2p 7
=4 2 2 +h i 5
a2 x01 x1p x02 x2p 2 2 2
a2 x01 x1p x02 x2p
x1p ;x2p
=0 (2.146)
" 2 #
@g12 x01 ; x02 @ 2 x01 x1p x02 x2p
= 2 2 =0
@x01 @x01 a2 x01 x1p x02 x2p
x1p ;x2p x1p ;x2p
" 2 #
@g12 x01 ; x02 @ 2 x01 x1p x02 x2p
= 2 2 (2.147)
=0
@x02 @x02 a2 x01 x1p x02 x2p
x1p ;x2p x1p ;x2p
as you can see from the rectangular shape of the grids. We can show this
quantitatively using the function that de…nes the surface. The surface shown in
Fig. 2.10 is de…ned by the function
Figure 2.10: Three di¤erent tangent spaces (in this case tangent plane) at three
di¤erent points on the manifold.
so that
@g @g @g
dg = dx1 + dx2 + dx3 = 0
@x1 @x2 @x3
@g @g @g
) dg = x
^+ y^ + z^ dx1 x
^ + dx2 y^ + dx3 z^ = 0
@x1 @x2 @x3
~ x1 ; x2 ; x3 d~r x1 ; x2 ; x3 = 0;
dg = A (2.151)
where
~ x1 ; x2 ; x3 = @g x
A ^+
@g
y^ +
@g
z^
@x 1 @x 2 @x3
@g @g @g
1
= 2 sin(x1 ) cos(x1 ); 2 = sin(x2 ); 3 = 1 (2.152)
@x @x @x
Note Eq. (2.151) shows that the vector A ~ x1 ; x2 ; x3 is normal to the surface
1 2 3
at the point with coordinates x ; x ; x :Now let’s pick a point, P , on the
surface with coordinates, ~rp = x1p ; x2p ; x3p and another neighboring point Q
with coordinates ~r = x1 ; x2 ; x3 that are at the same plane that is tangent to
the surface at point P, then we have for the vector on this plane given by
~r = ~r ~rp = x1 x1p x
^ + x2 x2p y^ + x3 x3p z^ (2.153)
and the vector normal to this tangent plane at point P is given by
~ x1 ; x2 ; x3 = @g @g @g
A p p p x
^+ y^ + z^ = m1 x
^ + m2 y^ z^ (2.154)
@x1 @x2 @x3 x1p ;x2p ;x3p
2.8. THE SIGNATURE OF A MANIFOLD 65
where
m1 = 2 sin(x1p ) cos(x1p ); m2 = sin(x2p ) (2.155)
The equation of the tangent plane at this point is determined by
~ x1 ; x2 ; x3
g=A ~r x1 ; x2 ; x3
= m1 x1 x1p + m2 x2 x2p x3 x3p = 0
) x3 x1 ; x2 = m1 x1 x1p + m2 x2 x2p + sin2 (x1p ) + cos(x2p ) (2.156)
or
a b
0 @x @x
gcd (xp ) = gab (xp ) ; (2.161)
@x0c xp @x0d
xp
G0 = X T GX: (2.166)
XT = X 1
) G0 = X 1
GX: (2.168)
G = X0 1
G0 X 0 = X 0 1
X 1
GXX 0 (2.169)
XX 0 = I ) X 0 = X 1
(2.170)
a a
This can be true if the transformation from x ! x0 is linear. The transforma-
tion matrix, X; with element Xba ; at point P each element must be a constant
and we can write
a
x0 = Xba xb : (2.171)
2.8. THE SIGNATURE OF A MANIFOLD 67
is positive for strictly Riemannian. This means gcd (x) must be positive de…nite
and the eigenvalues, a ; for G are also positive de…nite. On the other hand
for Pseudo Riemannian since the metric can be negative the eigenvalues can
a
be negative.p Now if we scale the coordinates x0 by these eigenvalues (i.e.
0a 0a 0
x ! x = j a j), for the metric tensor, G , we can easily show
2 3
1 0 ::: 0
6 0 1 ::: 0 7
6 7
6 : : 7
0
G =66 7: (2.174)
6 : : : 7 7
4 : : 5
0 0 ::: 1
Thus , at any arbitrary point, P; in a pseudo-Riemannian manifold, it is always
a
possible to …nd a coordinate system x0 such that in the neighborhood of P we
have h i
0 2
gab (x0 ) = ab + O x0 x0p ; (2.175)
where [ ab ] = diag ( 1; 1; :::; 1) : The number of positive entries (N+ ) minus
the number of negative entries (N ) in [ ab ] is called the signature of the man-
ifold. For example for the Minkowski spacetime manifold where the metric is
given by
2
ds2 = d (ct) dx2 dy 2 dz 2 (2.176)
2 3
1 0 0 0
6 0 1 0 0 7
=6 4 0 0
7: (2.177)
1 0 5
0 0 0 1
the signature is 2.
N-dimensional volume without a constraint: In an N-dimensional (pseudo)
Riemannian manifold with orthogonal coordinates system where the metric ten-
sor is diagonal, the full N-dimensional volume element dN V is
p
dN V = jgjdx1 dx2 dx3 :::dxN (2.178)
68 CHAPTER 2. MANIFOLDS
Vector Calculus on
manifolds
~t = lim ~s ~s (u + u) ~s (u)
= lim ; (3.1)
u!0 u u!0 u
where s is the in…nitessimal separation vector between the point P with coor-
dinate (u; s (u)) and some nearby point Q with coordinate (u + u; s (u + u))
on the curve, C, in the manifold corresponding (see Fig.??).
69
70 CHAPTER 3. VECTOR CALCULUS ON MANIFOLDS
where x1p ; x2p ; x3p are the coordinates for point P . We note that a point on the
surface can de…ned by a vector
^ + x2 y^ + sin2 (x1 ) + cos(x2 ) z^
~s = x1 x (3.10)
3.3. THE METRIC FUNCTION AND COORDINATE TRANSFORMATIONS71
so that
~s = x1 x
^ + x2 y^ + 2 sin(x1 ) cos(x1 ) x1 sin(x2 ) x2 z^ (3.11)
The two basis vectors, e^1 and e^2 ; at point P , in Fig. 3.2 are given by
~s x2 1 1 2 x
2
e^1 = lim = lim x
^ + y
^ + 2 sin(x ) cos(x ) sin(x ) (3.12)
z^ ;
1
x !0 x1 x1 !0 x1 x1
~s x1 1 1 x
1
e^2 = lim = lim x
^ + y
^ + 2 sin(x ) cos(x ) sin(x2 ) (3.13)
z^
x2 !0 x2 x2 !0 x2 x2
can be obtained from the inner product of the in…nitessimal vector separation.
We note that
ds2 = d~s d~s = e^a (x) dxa e^b (x) dxb = e^a (x) e^b (x) dxa dxb
) ds2 = gab (x) dxa dxb ; (3.17)
where
gab (x) = e^a (x) e^b (x)
is the metric function
(a) Find the basis vectors e^01 and e^02 in the tangent space at point P.
3.3. THE METRIC FUNCTION AND COORDINATE TRANSFORMATIONS73
(b) Re-derive the metric elements for a 2D sphere from the basis vectors.
Solution:
(a) We note that the tangent vector connecting point P with coordinate ( ; ')
with its neighboring point Q with coordinates ( + ; ' + ') (See Fig.
3.4) is expressible as
p
~s = ( cos (')) x
^ + ( sin (')) y^ + a2 2 z^
= [cos (') sin (') '] x
^ + [sin (') + cos (') '] y^
d
p z^ (3.18)
a2 2
" #
= cos (') x
^ + sin (') y^ p z^ +[ sin (') x
^ + cos (') y^] '
a2 2
(3.19)
(3.20)
and
~s
e^' = lim
'!0 '
" #
= cos (') x
^ + sin (') y^ p z^ +[ sin (') x
^ + cos (') y^] :
a2 2 '
(3.21)
Noting that
= =0 (3.22)
' '
we …nd
e^ = cos (') x
^ + sin (') y^ p z^: (3.23)
a2 2
and
e^' = sin (') x
^ + cos (') y^: (3.24)
74 CHAPTER 3. VECTOR CALCULUS ON MANIFOLDS
(b) The metric elements can be determined using the basis vectors
(a) Find the basis vectors e^r ; e^ ; and e^' in the tangent space at point P .
(b) Re-derive the metric elements for a 3D sphere from the basis vectors.
we have
b
e^b (x) d~s = e^b (x) e^a (x) dxa = a dx
a
= dxb : (3.29)
Expressing the in…nitessimal vector separation using the dual coordinates basis
vectors as
d~s = e^a (x) dxa (3.30)
we also …nd
b
e^b (x) d~s = e^b (x) e^a (x) dxa = a dx
a
= dxb : (3.31)
Thus the metric
ds2 = d~s d~s = e^a (x) dxa e^b (x) dxb = g ab (x) dxa dxb ; (3.32)
Using
@xa 0b
dxa = dx ; (3.37)
@x0b
we have
@xa
d~s = e^a dxa = e^a dx0b = e^0c dx0c : (3.38)
@x0b
Noting that
e^b e^b = 1
76 CHAPTER 3. VECTOR CALCULUS ON MANIFOLDS
we have
@xa
e^a dx0b = e^0b e^0b e^0c dx0c = e^0c e^0b e^0b dx0c = e^0b bc dx0c (3.39)
@x0b
@xa
) e^a dx0b = e^0b dx0b : (3.40)
@x0b
so that
@xa
e^0b =
e^a : (3.41)
@x0b
Homework: Show that for the dual basis vector
@x0a c
e^0a = e^ : (3.42)
@xc
Components of a vector in coordinate transformations: In coordinate
transformations the vector components are di¤erent but the vector itself is un-
changed. Suppose the vector, ~v ; is a vector at point P in the xa coordinate
system and ~v 0 is the vector in the x0a coordinates at the same point on the
manifold. These vectors may be expressed in terms of the basis vectors in the
two coordinates di¤erently
But the vector is the same since it describes a geometrical entity that is inde-
pendent of the coordinate system. Therefore, we must have
@x0b c
e^0b = e^ ; (3.47)
@xc
so that
@x0b c @x0b @x0b a
c
e^ e^a = v a c ca =
v 0b = v a v : (3.48)
@x @x @xa
Homework: Show that for the covariant components of a vector transformed by
the equation
@xa
vb0 = va : (3.49)
@x0b
3.3. THE METRIC FUNCTION AND COORDINATE TRANSFORMATIONS77
~v = va e^a : (3.54)
and
~ = wb e^b :
w (3.55)
so that the inner product becomes
~ = va e^a wb e^b = g ab va wb :
~v w (3.56)
where
g ab = e^a e^b : (3.57)
We can use the covariant and controvariant components of the vectors to deter-
mine the inner products
a a
~ = v a e^a wb e^b = e^a e^b v a wb =
~v w b v wb = v a wa (3.58)
or
a
~ = va e^a wb e^b = e^a e^b va wb =
~v w b va w
b
= va w a : (3.59)
Whichever way we determine the inner products we must get the same values.
Thus from Eqs. (3.52) and (3.58), we …nd
~ = gab v a wb = v a wa ) gab wb = wa :
~v w (3.60)
~ = g ab va wb = va wa ) g ab wb = wa :
~v w (3.61)
From Eq. (3.60), we note that the covariant form of the metric tensor can be
used to lower an index and from Eq. (3.61) we also see that the controvariant
form of the metric tensor can be used to raise an index. Applying Eqs. (3.60)
and (3.61), we can express the basis vectors
and
e^c = g cb e^b : (3.63)
Then noting that the inner product
a a
e^a e^c = c ) gad e^d g cb e^b = gad g cb e^d e^b = c
d a
) gad g cb e^d e^b = gad g cb b = c
a
) gab g cb = g cb gab = c: (3.64)
gab v a wb = va wa = g ab va wb = v a wa (3.65)
gab v a v b = g ab va vb = v a va = va v a ; (3.66)
and it can be zero without the vector being actually be a zero vector. We can
see this if we recall the pseudo-Riemannian manifold metric
~v = v a e^a : (3.69)
and
~ = wb e^b :
w (3.70)
are said to be orthogonal when
gab v a wb = g ab va wb = v a wa = va wa = 0: (3.71)
P~ = pa e^a (3.72)
dP~ d a dpa d^
ea
= (p e^a ) = e^a + pa (3.73)
d d d d
For the coordinates xa = xa ( ) ; we have
d^
ea ea dxb
@^
= (3.74)
d @xb d
so that
dP~ dpa dxb @^
ea
= e^a + pa : (3.75)
d d d @xb
In order to better understand the origin of the a¢ ne connections we shall re-
reconsider the 2D sphere embedded in a 3D manifold. We saw that the tangent
space is a plane with basis vectors de…ned by
e^ = cos (') x
^ + sin (') y^ p z^: (3.76)
a2 2
and
e^' = sin (') x
^ + cos (') y^: (3.77)
There follows that
@^
e @^
e' 1
= = sin (') x
^ + cos (') y^ = e^' = f ' ( ; ') e^' : (3.78)
@' @
80 CHAPTER 3. VECTOR CALCULUS ON MANIFOLDS
where
1
f ' ( ; ') = ;
is a function that connects the change in the basis vectors with respect to the
coordinates to the basis vectors. Here we note that we are still in the tangent
space. However, if we switch the variables for the derivatives, we …nd
" #
2
@^
e 1 a2
= p + z
^ = z^;
@ a2 2
(a2 2 )3=2 (a2 2 )3=2
@^
e'
= cos (') x
^ sin (') y^ = (cos (') x
^ + sin (') y^) :
@'
that we can not tell whether it belongs to the tangent space or not. In order to
…nd out that we introduce a basis vector normal to the tangent space in terms
of the basis vector in the tangent space as
e^? = e^ e^' :
This normal basis vector is found to be
2
e^? = p (cos (') x
^ + sin (') y^) + z^: (3.79)
a2 2
e^ = cos (') x
^ + sin (') y^ p z^: (3.80)
a2 2
one …nd !
a2 2
1
z^ = e^? p e^ (3.81)
a2 a2 2
and r
2 2
1
cos (') x
^ + sin (') y^ = 1 e^? e^ : (3.82)
a a2 a2
Using these relations, we can then write
r
2 3
@^
e 1
@^
e'
= p e^ + e^? ; = 1 e^? + 2 e^ : (3.83)
@ a2 2 @' a a2 a
or
@^
e @^
e'
= [f ( ; ')]jj e^ +[f ( ; ')]? e^? ; = [f'' ( ; ')]jj e^ +[f'' ( ; ')]? e^? :
@ @'
(3.84)
where
1
[f ( ; ')]jj = p [f ( ; ')]? =
a2 2
r
2 3
[f'' ( ; ')]jj = 1 [f'' ( ; ')]? = (3.85)
a a2 a2
3.5. THE AFFINE CONNECTIONS 81
These results show that the derivatives for the basis vectors can lead to compo-
nents that does not belong to the tangent space. Let’s see where these compo-
nents are located for a speci…c point for a 2D sphere with radius,
p a = 2 units.
The point we shall consider has coordinates (x = 1; y = 0; z = 3) which is on
the surface of the 2D sphere. At this point the two basis vectors (^ e ; e^' ) and
the vector normal to the tangent space, are shown in Fig. 3.5.
p
@^
e 1 @^
e 3 1
= p e^ + e^? ; ' = e^? + e^ : (3.86)
@ 3 @' 4 4
Therefore, generally for none spherical geometry or none Euclidean manifold,
the change in the basis vectors with respec to the coordinate at a point on a
manifold (derivative of the basis vector at that point) can have both normal and
tangential components to the tangent space. However, since we are con…ned to
the tangent space, for example in the 2D sphere of radius a, we are con…ned to
the tangent space that is a plane and the normal component does not belong
to the tangent space at that particular point. Therefore, we shall consider only
the projection parallel to the tangent space at point P;
@^
ea e^a
= lim : (3.87)
@xc c
x !0 xc jjTp
Suppose we represent the coe¢ cients resulting from the derivative of the basis
vectors that are components that belong to the tangent space at the point
(e.g.[fac ( ; ')]jjTp ) by bac ; then we can write
@^
ea 1 2 3 N b
= ac e
^1 + ac e
^2 + ac e
^3 ::: + ac e
^N = ac e
^b ; (3.88)
@xc
82 CHAPTER 3. VECTOR CALCULUS ON MANIFOLDS
where N is the dimension of the tangent space. The N 3 coe¢ cients bac are
known collectively as the a¢ ne connection or in older textbooks, the Christo¤el
symbol (of the second kind) at point P .
Homework:
(1) Find all the elements for the a¢ ne connection, bac ; for a point on a 2D
sphere embedded in a 3D Euclidean space. Note that in the expressions
@^e ' @^
e '
= e^ + e^' ; = 'e
^ + 'e
^' ;
@ @'
@^
e' ' @^e' '
= ' e^ + ' e^' ; = '' e
^ + '' e
^' ; (3.89)
@ @'
you are going to determine
' ' ' '
; ; '; '; ' ; ' ; '' ; and '' : (3.90)
we …nd
@^
ea @^
ea
e^d= bac db ) d
ac = e^d (3.93)
@xc @xc
Since d is a dummy index, we can write
b @^
ea
ac = e^b = e^b @c e^a (3.94)
@xc
From now on we will use the notation @c e^a ;
@^
ea
@c e^a = : (3.95)
@xc
Di¤erentiating Eq. (3.92) with respect to the coordinate, xc ; and applying the
notation, we …nd
b
e^a @c e^b + e^b @c e^a = @c a = 0 ) e^b @c e^a = e^a @c e^b (3.96)
Using the de…nition of the derivative of the basis vectors and our notation
@^
ea b d
@c e^a = = bac e^b ) e^d @c e^a = b
ac e
^b e^d = ac b = d
ac (3.97)
@xc
d
) ^d @c e^a
ac = e (3.98)
3.5. THE AFFINE CONNECTIONS 83
e^b @c e^a = a
bc e
^b e^b = e^b a b
bc e
^
a a b
) @c e^ = bc e
^ (3.101)
b @^
ea
ac = e^b : (3.102)
@xc
we have
0b e0a
@^
ac = e^0b : (3.103)
@x0c
so that applying the relations
@xf @x0b d
e^0a = 0a
e^f ; e^0b = e^ : (3.104)
@x @xd
we may write
0b @x0b d @ @xf
ac = e^ e^f
@xd @x0c @x0a
@x0b d ef @xf
@^ @ 2 xf
= e^ + e
^ f
@xd @x0c @x0a @x0c @x0a
@x0b @xf d @^
ef @ 2 xf @x0b d
= e
^ + e^ e^f : (3.105)
@xd @x0a @x0c @x0c @x0a @xd
Using the chain rule one can write
@^
ef ef @xg
@^
=
@x0c @xg @x0c
so that Eq. (3.105) becomes
and
d d
@a gbc = ba gdc + ca gbd (3.117)
Now combining Eqs. (3.115)- (3.117), we can write
d d d d
@c gab + @b gca @a gbc = ac gbd + bc gad + ab gcd + cb gad
d d
ba gdc ca gbd (3.118)
b b
Recalling that we will be interested in a torsionless manifold where ac = ca ;
Eq. (3.118)
@c gab + @b gca @a gbc = 2 dbc gad : (3.119)
Multiplying Eq. (3.120) by g pa ;
a g ad
bc = (@b gcd + @c gdb @d gbc ) (3.124)
2
The right hand side of Eq. (3.124) is known as the metric connection and is
a
often represented by bc
Useful Formulae:
d
abc = gad bc (3.125)
Multiplying by g f a
f f
gf a abc = g f a gad d
bc = d
d
bc = bc
f fa
) bc =g abc (3.126)
f gf d
bc = (@b gcd + @c gdb @d gbc ) (3.127)
2
86 CHAPTER 3. VECTOR CALCULUS ON MANIFOLDS
f gf a
bc = (@b gca + @c gab @a gbc ) (3.128)
2
so that equating Eqs. (3.126) and (3.128), we …nd
1
abc = (@b gca + @c gab @a gbc ) (3.129)
2
The quantity abc is traditionally known as a Christo¤ el symbol of the …rst kind.
Noting that
1
bac = (@a gcb + @c gba @b gac ) (3.130)
2
we can write
1
abc + bac = (@b gca + @c gab @a gbc + @a gcb + @c gba @b gac ) (3.131)
2
taking into account the symmetry of the metric tensor, from Eq. (3.131), we
…nd
@c gab = abc + bac (3.132)
Eq. (3.132) allows us to express the partial derivative of the metric components
in terms of the connection coe¢ cients.
We recall that the determinant of the matrix G can be expressed as
a+b
det G = gab ( 1) Mab = gab fcof [G]ab g ; (3.133)
where
a+b
[cof [G]]ab = ( 1) Mab (3.134)
is the the cofactor matrix to G which is determined from the minor, Mab and it
is a constant matrix. We recall that the minor of matrix G denoted by Mab is
the determinant of the matrix formed from matrix G by removing the ath row
and bth column. Let det G = g; so that
we may write
@c g = cof [G]ab @c gab = gg ab @c gab : (3.137)
Now using the relation (3.132), we can write
We recall
d
abc = gad bc (3.139)
3.6. LOCAL GEODESIC AND CARTESIAN COORDINATES 87
so that
Taking into account that a and b are dummy indices, we can replace b by a so
that
@c g = 2g aac (3.141)
This can be rearranged as
a 1 1 p
ac = @c g = @c ln jgj = @c ln jgj: (3.142)
2g 2
Now for the sake of convenience if we replace c by b, we may write the above
equation as p
a
ab = @b ln jgj (3.143)
The modulus is for the case where the manifold is seudo-Riemannian where the
metric elements can be negative.
where we used the fact that abc (P ) ; xaP ; xbP ; and xcP are constant at point P.
The coordinates xa are independent coordinates,
@xa a @xb b @xc c
= d; = d; = d (3.147)
@xd @xd @xd
88 CHAPTER 3. VECTOR CALCULUS ON MANIFOLDS
so that
@x0a 1 n o
a a b c
= d + bc (P ) (xc xcP ) d + xb xbP d (3.148)
@xd 2
which results in
@x0a a 1 a
= d + dc (P ) (xc xcP ) + a
bd (P ) xb xbP (3.149)
@xd 2
The summation indices are dummy indices and therefore we can replace b by c
in the second term so that
@x0a a 1 a
= d + [ dc (P ) (xc xcP ) + a
cd (P ) (xc xcP )] (3.150)
@xd 2
Since for a torsionless manifold
a a
dc (P ) = cd (P ) ;
we can write
@x0a
= ad + adc (P ) (xc xcP ) (3.151)
@xd
We are interested in the a¢ ne connection at point P so that when we are
evaluating Eq. (3.151) at xc = xcp , we …nd
@x0a a
= d (3.152)
@xd P
@xa a
= d (3.153)
@x0d P
@x0a @ a @
= + adc (P ) e (xc xcP )
@xe @xd @xe d @x
@x0a
) = adc (P ) ce = a
de (P ) (3.154)
@xe @xd
Using the results in Eqs. (3.152)-(3.154), the transformation equation for the
a¢ ne connection
and for a geodesic coordinates the connection is zero at point P , and therefore
0
@gab (x0 )
= 0:
@x0c P
It is important to note that for geodesic coordinates the metric does not nec-
essarily satisfy Eq. (3.158). But we can …nd coordinates x00a that satisfy Eq.
(3.158) by making a linear transformation to the x0a
x00a = Xx0a (3.162)
where Xba are constants.
where e^a are the coordinate basis vectors. The derivative of this vector with
respect to the coordinate, xb ; can be expressed as
@~v
= @b~v = @b (v a e^a ) = e^a @b (v a ) + v a @b (^
ea ) (3.164)
@xb
We recall
c
@b (^
ea ) = ab e
^c
so that
@b~v = (@b v a ) e^a + c a
ab v e
^c (3.165)
Switching the places for ther indices c and a one can write
c a a c
ab v e
^c = cb v e
^a (3.166)
so that
@b~v = (@b v a ) e^a + v c a
cb e
^a = (@b v a + a c
cb v ) e
^a (3.167)
The quantity in the bracket which is represented as
rb v a = @b v a + a c
cb v (3.168)
is known as the covariant derivative of the vector components. Thus the deriv-
ative of a vector can be expressed as
rb v a = @ b v a (3.170)
which is just the ordinary derivative that we are very familiar with!
Homework: Suppose the vector is expressed in terms of its covariant compo-
nents
~v = va e^a (3.171)
show that
c
rb va = @b va ab vc (3.172)
The covariant derivative of a scalar function: for a scalar function
the covariant derivative is
rb = @ b (3.173)
The gradient: The gradient of a scalar function is given by
rb v a = @b v a + a c
cb v (3.177)
r ~v = ra v a = @a v a + a b
ba v : (3.178)
r2 = r r : (3.181)
But in order to apply the relation we derive for the divergence in Eq. (3.180),
we need the vector
~v = (@a ) e^a = va e^a : (3.183)
in terms of its controvariant components. We have seen that the index can be
raised or lowered using the metric tensor. In this case we want to raise it, so we
have
g ab vb = v a
and we can express the vector ~v as
d d a d^
ea (u) dv a (u)
~v (u) = [v (u) e^a (u)] = v a (u) + e^a (u) (3.192)
du du du du
d ea (u) dxb
d^ dv a (u)
) ~v (u) = v a (u) b
+ e^a (u) (3.193)
du dx du du
Using the relation
d^
ea d
= ab e
^d (3.194)
dxb
we have
d dxb d dv a
~v = v a ab e
^d + e^a (3.195)
du du du
and replacing the index a by c in the …rst term
d dxb dv a
~v = v c d
cb e
^d + e^a (3.196)
du du du
3.9. PARALLEL TRANSPORT 93
and d by a
d dxb dv a dv a a c dx
b
~v (u) = v c a
cb e
^a + e^a = + cb v e^a ; (3.197)
du du du du du
dv a dv a dxb
= (3.200)
du dxb du
into Eq. (3.199), we …nd
Dv a @v a dxb a c dx
b
dxb a c dx
b
dxb
= + cb v = @b v a + cb v = (@b v a + a c
cb v )
Du @xb du du du du du
Dv a dxb
) = (rb v a ) ; (3.201)
Du du
where we used the relation in Eq. (3.168).
Homework: Suppose the vector, ~v ; depends on the parameter u on a curve
de…ned by xa (u) is expressed in terms of its covariant components
~
dD
~v = (3.204)
dt
and the acceleration by
d~v
~a = : (3.205)
dt
94 CHAPTER 3. VECTOR CALCULUS ON MANIFOLDS
Suppose you plot the displacement of the particle at di¤erent times, then you
would get generally a curve. The particle would have a constant velocity through
this curve provided its acceleration is zero.
d~v
~a = = 0: (3.206)
dt
This means the particle travels along this curve with a constant velocity. The
velocity would have the same magnitude and direction. The velocity vector
remains parallel at each point on the curve describing the displacement of the
particle as a function of time.
On a curve C on a manifold (See Fig.3.5), a parallel transport of a vector
Figure 3.5: Parallel transport of a vector ~v (Shown by the yellow parallel arrows)
on a manifold.
In a given coordinate system xa with basis vectors e^a , we can express the
in…nitessimal separation vector d~s as
dxa
d~s = du^
ea (3.210)
du
so that the tangent vector becomes
a
~t = dx e^a (3.211)
du
Recalling that in pseudo-Riemmanian manifold the length of a vector ~v is given
by q q q
p
jvj = jgab v a v b j = jg ab va vb j = jv a vb j = jva v b j (3.212)
we have s p
q
dxa dxb jg ab dxa dxb j
jtj = jgab ta tb j = g ab = (3.213)
du du du
we recall the metric or the interval (the distance squared along the curve on the
manifold between the two points P and Q) is
~t = d~s = 0 (3.216)
du
at all points on the curve and the distance, s; does not depend on the parameter,
u; and we clearly can not use it as a¢ ne parameter since it does not satisfy the
condition in Eq. (3.215). But it is possible to …nd a privileged family of a¢ ne
parameter.
Let’s consider two points on the x-y plane. Let P1 be (x1 ; y1 ) and P2 be
(x2 ; y2 ) : Then the distance between these points is given by the integral
Z (2)
L= ds; (3.218)
(1)
where
s s
p dy
2
dx
2
ds = dx2 + dy 2 = 1+ dx = 1+ dy: (3.219)
dx dy
Out of the in…nitely many functions that can be used to connect the two points,
we want to determine the one that would give the minimum distance. Let these
function be denoted by Y (x) : From these in…nite number of functions there is
only one function that gives the minimum distance between the two points. If
this function is y (x) ; then we may write Y (x) in terms of y (x) as
where
dY (x; )
0
Y = : (3.225)
dx
We are interested in the path that gives the the minimum distance between the
two points (i.e. the geodesic). The necessary condition for the distance, L ( ) ;
to be minimum is that the length function, L ( ) ; must have a stationary point
at ( = 0; L ( = 0)). This requires
dL ( )
= 0; (3.226)
d =0
which leads to
"Z #
(2)
dL ( ) 1 1 dY 0 (x; )
= p (2Y 0 ) dx = 0: (3.227)
d =0 (1) 2 1 + Y 02 d
=0
Using
Y (x; ) = y (x) + (x) (3.228)
we may write
dY dy d
= + ) Y 0 (x; ) = y 0 (x) + 0
(x) (3.229)
dx dx dx
so that
dY 0 ( ) d 0 0 0
= [y (x) + (x)] = (x) : (3.230)
d d
There follows that
Y 0 (x; )j =0 = y 0 (x) : (3.231)
and
dY 0 ( ) 0
= (x) (3.232)
d =0
In view of these results, one …nds for the stationary point
Z (2)
dL ( ) 1 1 dY 0 ( )
= p 02
(2Y 0 ) dx
d =0 (1) 2 1+Y d
=0
Z (2)
y 0 (x) 0 (x)
= p dx = 0: (3.233)
(1) 1 + y 02 (x)
Using integration by parts
Z Z
udv = uv vdu (3.234)
3.11. THE CALCULUS OF VARIATION-(A REVIEW FROM THEORETICAL PHYSICS I)99
for
0
(x) = dv ) v = (x) ; (3.235)
that gives !
y0 d y0
u= p ) du = p dx; (3.236)
1 + y0 2 dx 1 + y0 2
we may write the integral as
Z x2 Z !
(2) x2
y 0 0 (x) y0 d y0
p dx = p (x) (x) p dx = 0:
(1) 1 + y0 2 1 + y0 2 x1 dx 1 + y0 2
x1
(3.237)
Due to the conditions
(x1 ) = (x2 ) = 0 (3.238)
the …rst term in the above expression becomes zero. Thus one can write
Z (2) Z x1 !
dL ( ) y 0 0 (x) d y0
= p dx = (x) p dx = 0:
d =0 (1) 1 + y0 2 x1 dx 1 + y0 2
(3.239)
Since (x) is an arbitrary function, for the integral to be zero, we must have
!
d y0 y0
p =0) p = c; (3.240)
dx 1+y 02
1 + y0 2
Note that we have introduced another constant in terms of the constant c. There
follows that
dy
= m ) y (x) = mx + b; (3.242)
dx
which is equation of a straight line.
that can connect these two points, there is only one trajectory on this surface
that is the shortest (the Geodesic). Finding the Geodesic is the general problem
that can be solved applying the calculus of variation.
The surface is de…ned by the function F (x; y (x) ; y 0 (x)): The distance be-
tween these two points determined by evaluating the integral
Z x2
I= F (x; y (x) ; y 0 (x)) dx: (3.245)
x1
3.11. THE CALCULUS OF VARIATION-(A REVIEW FROM THEORETICAL PHYSICS I)101
To determine the equation for the function F that gives the shortest length
joining the two points, let the function for any path connecting the two points
be Y (x) : From these in…nite number of functions there is only one function that
gives the minimum distance between the two points. If this function is y (x) ;
then we may write Y (x) in terms of y (x) as
Y (x; ) = y (x) + (x) ; (3.246)
where (x) is an arbitrary function which must satisfy the condition
(x1 ) = (x2 ) = 0 (3.247)
so that at the two points (x = x1 = x2 ), we …nd
Y (x; ) = y (x) : (3.248)
We also have
dY (x; ) dY ( )
= (x) ) = (x) ; (3.249)
d d =0
and
dY dy d
= + or Y 0 (x; ) = y 0 (x)+ 0
(x) ) Y 0 (x; )j =0 = y 0 (x) ; : (3.250)
dx dx dx
There follows
dY 0 ( ) d 0 0 0 dY 0 ( ) 0
= [y (x) + (x)] = (x) ) = (x) (3.251)
d d d =0
Noting that
d @F dY ( ) @F dY 0 ( )
[F (x; Y (x; ) ; Y 0 (x; ))] = +
d =0 @Y d @Y 0 d =0
@F dY ( ) @F dY 0 ( )
= + 0
(3.254)
@Y =0 d =0 @Y =0 d =0
and substituting
dY ( )
Y (x; )j =0 = y (x) ; = (x) ; Y 0 (x; )j =0 = y 0 (x) ;
d =0
dY 0 ( ) 0
= (x) ; (3.255)
d =0
102 CHAPTER 3. VECTOR CALCULUS ON MANIFOLDS
we …nd
d @
[F (x; Y (x; ) ; Y 0 (x; ))] = F (x; y (x) ; y 0 (x)) (x)
d =0 @y
@
+ F (x; y (x) ; y 0 (x)) 0 (x) ; (3.256)
@y 0
recalling that
(x1 ) = (x2 ) = 0 (3.260)
we …nd
Z x2 Z x2
@ @ @F
F (x; y (x) ; y 0 (x)) 0
(x) dx = (x) (x) dx:
x1 @y 0 x1 @x @y 0
(3.261)
Thus the stationary integral can be put in the form
Z x2
dI ( ) @
= F (x; y (x) ; y 0 (x)) (x) dx+
d =0 x1 @y
Z x2
@ @F
(x) (x) dx = 0: (3.262)
x1 @x @y 0
or
Z x2
dI ( ) @ @F @F
= (x) dx = 0: (3.263)
d =0 x1 @x @y 0 @y
The basis vector with respect to the coordinates xc does not change along the
geodesic curve. This means for the tangent vector
a
~t = dx e^a (3.267)
du
to the curve at least the direction remains the same. In the Euclidean space
this makes the curve to be a straight line where the tangent vectors has same
direction along the line. Thus for Euclidean space the geodesic is a straight line.
For general curve de…ned by xa = xa (u) ; on a manifold, if the curve is
geodesic then the tangent curve must have the same direction at all points on
the curve. This mean the change in the tangent vector with respect to the
parameter u is then only the magnitude that changes and it is given by
d~t dxa
= (u) ~t = (u) e^a ; (3.268)
du du
where (u) is some function of u: Using our result for intrinsic derivative of a
vector ~v (u)
d Dv a dv a dxb
~v (u) = e^a = + acb v c e^a ; (3.269)
du Du du du
for the tangent vector intrinsic derivative, we have
dta a c dx
b
dxa
+ cb t = (u) (3.271)
du du du
Noting that
a
~t = dx e^a = ta e^a (3.272)
du
Eq. (3.271) can be written as
d2 xa a dx
c
dxb dxa
+ cb = (u) (3.273)
du2 du du du
104 CHAPTER 3. VECTOR CALCULUS ON MANIFOLDS
The result in Eq. (3.273) is valid for both non-null and null geodesics parame-
terized in terms of some general parameter u. For a¢ ne parameter u where it
is related to the distance s on the curve
u = as + b (3.274)
we have
du = ads (3.275)
and
~t = ds = 1 (3.276)
du a
which is a tangent vector with a constant length that is independent of the
parameter u: This means
d ~t
= (u) ~t = 0 ) (u) = 0 (3.277)
du
Therefore in general for a privilege parameter, where (u) = 0; called the a¢ ne
parameter the equation for the geodesic in Eq. (3.273) can be written as
d2 xa a dx
c
dxb
+ cb =0 (3.278)
du2 du du
Eq. (3.278) is a parallel transport for the tangent vector that we discussed in
the previous section.
Dta dta dxb
= + acb tc =0 (3.279)
Du du du
which can be shown by replacing
dxa dxc
= ta ; = tc (3.280)
du du
in Eq. (3.278).
Homework:
If we change the a¢ ne parameter u to u0; the coordinates that de…ne the
geodesic curve would change from xa (u) to xa (u0 ) : Show that in terms of the
new a¢ ne parameter, u0; the geodesic in Eq. (3.278) becomes
!
d2 u
d2 xa c
a dx dx
b
du02 dxa
+ cb = du
(3.281)
du02 du0 du0 du0
du0
length along this curve joining these two points, this length can be determined
from
Z 2 Z 2q Z 2s
dxa dxb
L= ds = jgab dxa dxb j = gab du (3.282)
1 1 1 du du
where q
ds
jgab x_ a x_ b j = s_ =
F = (3.285)
du
Using the principle of variation that you have learned in Theoretical Physics Part
I and revised in the previous section the curve along a surface which marks the
shortest distance between two neighboring points is the geodesic. Using the
principle of variation it can be shown that the integral must be stationary for
the geodesic and the function satis…es the Euler-Lagrange Equation:
d @F @F
= 0: (3.286)
du @ x_ c @xc
d gac x_ a x_ a x_ b @a gab
= 0: (3.288)
du s_ 2s_
106 CHAPTER 3. VECTOR CALCULUS ON MANIFOLDS
1 s• x_ a x_ b @a gab
(@b gac ) x_ a x_ b + gac x
•a gac x_ a = 0: (3.292)
s_ s_ 2s_
x_ a x_ b @a gab s•
•a + (@b gac ) x_ a x_ b
) gac x = gac x_ a : (3.293)
2 s_
Noting that we can interchange the indices and write
so that
1
(@b gac ) x_ a x_ b =
(@b gac ) x_ a x_ b + (@a gbc ) x_ b x_ a (3.295)
2
Substituting Eq. (3.295) into Eq. (3.293), we have
1 a b s•
•a +
gac x x_ x_ @b gac + x_ b x_ a @a gbc x_ a x_ b @a gab = gac x_ a : (3.296)
2 s_
1 s•
•a +
gac x [@b gac + @a gbc @a gab ] x_ a x_ b = gac x_ a : (3.297)
2 s_
and multiplying by g dc ;
1 s• dc
g dc gac x
•a + g dc [@b gac + @a gbc @a gab ] x_ a x_ b = g gac x_ a : (3.298)
2 s_
d
Now applying the relationg dc gac = a; we …nd
d a 1 s• d a
ax
• + g dc [@b gac + @a gbc @a gab ] x_ a x_ b = ax
_ : (3.299)
2 s_
1 s•
•d + g dc [@b gac + @a gbc
x @a gab ] x_ a x_ b = x_ d : (3.300)
2 s_
3.13. STATIONARY PROPERTY OF THE NON-NULL GEODESIC 107
a 1 dc
bc = g [@b gac + @a gbc @a gab ] (3.301)
2
we …nd
s•
•d +
x a a b
bc x
_ x_ = x_ d : (3.302)
s_
Comparing the result in Eq. (3.302) with Eq. (3.281)
!
d2 u
d2 xa c
a dx dx
b
du02 dxa
02
+ cb 0 0
= du
(3.303)
du du du du0
du0
Tensor Calculus on
manifolds
~t (~v ) = ~t ~v : (4.1)
This maps the vector …eld, ~v ; to the tangent space, Tp : In this case one vector (~v )
is linearly mapped into the tangent space by the tangent vector ~t (i.e. ~t ! ~t (~v )).
Therefore the tangent vector ~t forms a …rst rank tensor t.
A tensor : Based on the notion of a vector on a manifold, a tensor is de…ned
by the precise set of operations applied to the a set of vectors to produce a
scalar and the number of vectors in the set determines the rank of the tensor.
If there are N number of vectors in the set, the tensor is N th rank tensor (See
the table below)
Tensor Operation Rank
t (~u; ~v ) ~u ~v 2
t (~u; ~v ; w)
~ ~u ~v w
~ 3
t (~u; ~v ; w;
~ ~x) ~u ~v w
~ ~x 4
t (~u; ~v ; w;
~ ~x; ~y ) ~u ~v w
~ ~x ~y 5
From this table we can easily see that a scalar …eld can be classi…ed as a
zero ranked tensor …eld since it does not depend on a vector …eld.
109
110 CHAPTER 4. TENSOR CALCULUS ON MANIFOLDS
A tensor is a linear map of the vectors into the real and therefore any ranked
tensor is linear. This means, for example, for 1st rank tensor, we must have
~t ( ~u + ~v ) = ~t ( ~u) + ~t ( ~v ) = ~t (~u) + ~t (~v ) (4.2)
Example 4.1 Let’s reconsider the 2D sphere in the 3D manifoldFor the vector
we have
t1 = t (~ea ) ~e1 = t1~e1 + t2~e2 ~e1 ; t2 = t (~ea ) ~e2 = t1~e1 + t2~e2 ~e2
t1
) t (~ea ) =
t2
t (~ea ; ~eb ) = (~e1 ~e1 ) + (~e2 ~e1 ) + (~e1 ~e2 ) + (~e2 ~e2 )
g11 g12
) t (~ea ; ~eb ) =
g21 g22
which forms a 2-nd rank tensor. Using the dual basis vector, we can also
express this 2-nd rank tensor as
t ~ea ; ~eb = ~e1 ~e1 + ~e2 ~e1 + ~e1 ~e2 + ~e2 ~e2
g11 g12
) t ~ea ; ~eb =
g21 g22
or
t ~e;a ~eb = ~e1 ~e1 + ~e2 ~e1 + ~e2 ~e2 + ~e2 ~e2
g 12 g 12
) t ~e;a ~eb =
g 21 g 22
Note that
t (~u; ~v ) = tab ua v b = tba ua vb = tab ua v b (4.14)
112 CHAPTER 4. TENSOR CALCULUS ON MANIFOLDS
1
t(a1 a2 :::aN ) = (addition over all permutations of the indices a1 a2 :::aN )
N!
(4.19)
and
1
t[a1 a2 :::aN ] = (Alternating subtraction and addition over
N!
all permutations of the indices a1 a2 :::aN ) (4.20)
1
t[a1 a2 a3 ] =
t[a1 a2 a3 ] (4.21)
3!
For example for 3rd rank tensor, we have
1
t(a1 a2 a3 ) = (ta a a + ta2 a1 a3 + ta1 a2 a3 + ta3 a2 a1 + ta1 a2 a3 + ta1 a3 a2 ) (4.22)
3! 1 2 3
and
1
t(a1 a2 a3 ) = (ta a a ta2 a1 a3 + ta1 a2 a3 ta3 a2 a1 + ta1 a2 a3 ta1 a3 a2 ) (4.23)
3! 1 2 3
Particular subset of indices permutation: we have a di¤erent notations when
the permutation is to particular subset of indices. This is described using a 4th
4.1. TENSORS FIELDS AND RANK OF A TENSOR 113
rank tensor.
1
t(ab)cd = (tabcd + tbacd ) (4.24)
2
symmetric permutation to indices a and b only
1
t[ab]cd = (tabcd tbacd ) (4.25)
2
antisymmetric permutation to indices a and b only
1
ta[bjcjd] = (tabcd tbdca )
2
antisymmetric permutation to indices b and d only (4.26)
Note that the symbol jj are used to exclude unwanted indices from the sym-
metrization () antisymmetrization [] implied.
Example 4.2 Let’s reconsider the 3D sphere in the 4D manifoldWe recall the
tangent space at a point on this 3D sphere form three basis and dual basis
vectors, ~e1 ; ~e2 ; ~e3 (or ~er ; ~e ; ~e' ) and ~e1 ; ~e2 ; ~e3 (or ~er ; ~e ; ~e' ). Write the
expression for the symmetric, t(123) and antisymmetric t[123] components.
Solution: The expressions for these components are given by
1
t(123) = t(123) + t(132) + t(213) + t(231) + t(312) + t(321)
3!
and
1
t[123] = t(123) t(132) + t(312) t(321) + t(231) t(213)
3!
114 CHAPTER 4. TENSOR CALCULUS ON MANIFOLDS
(a) t(312)
(b) t[ab](cd)
g (~u; ~v ) = ~u ~v (4.27)
We recall that covariant and controvariant components of the metric tensor are
given by
gab = g (~ea ; ~eb ) = ~ea ~eb ; g ab = g ~ea ; ~eb = ~ea ~eb (4.29)
and the mixed components
a
gab = g ~ea ; ~eb = g (~ea ; ~eb ) = b (4.30)
Raising and lowering tensor indices. Consider the third rank tensor, t; expressed
in terms of its covariant components
or mixed components
tcab = t (~ea ~eb ~ec ) : (4.32)
We recall from chapter 3 that the covariant form of the metric tensor can be
used to lower and the controvariant to raise the index of a vector. This can be
extended to raising and lowering indices of a higher ranked tensor. For example,
we can lower the index of the 3-rd ranked tensor in Eq. (4.32) by multiplying
it with the covariant form of the metric tensor, gdc ;
noting that
c
gdc = ~ed ~ec ; and ~eb ~ec = b
we have
c
gdc tcab = ~ed ~ec ~ea b = ~ed ~eb ~ea = tdba (4.34)
so that
tdba = gdc tcab (4.35)
4.3. MAPPING TENSORS INTO TENSORS 115
Since the indices are dummy variables and tensor tabc is symmetric, one can
write
tabc = gcd tdab (4.36)
An alternative way of showing this is to use the linearity of operations involving
tensors
gdc tcab = gdc t (~ea ; ~eb ; ~ec ) = t (~ea ; ~eb ; gdc~ec ) = t (~ea ; ~eb ; ~ed ) = tabd (4.37)
where we used
gdc~ec = ~ed
Noting that Homework: Raise the tensor tabc to tabc :
and the second rank tensor t (~u; ~v ) maps the operation involving the vectors ~u
and ~v into real numbers in the tangent space
For example 2D sphere where the tangent space has two basis vectors t(~e1 ; ~e2 )we
t (~u; ~v ) = tab ua v b
= t11 u1 v 1 + t12 u2 v 1 + t11 u1 v 2 + t12 u2 v 2 + t21 u1 v 1 + t22 u2 v 1
+t21 u1 v 2 + t22 u2 v 2
t11 t12 v1
t (~u; ~v ) = u1 u2
t21 t12 v2
The 2-nd rank tensor t(~ea ; ~eb ) maps the vector operation involving ~u (~ea ) and
~v (~ea ) into a real number in the tangent space. Now the question is, can we map
a tensor into another tensor of a di¤erent rank. Consider a 3-rd rank tensor
t (~u; ~v ; w)
~ which maps the three vectors into real numbers. Let’s replace the
two vectors (~u; ~v ) by the basis vectors (~ea ; ~eb ) in the tangent space,
t (~ea ; ~eb ; w)
~ = ~ea ~eb w:
~ (4.39)
~ = wc~ec ;
w (4.40)
116 CHAPTER 4. TENSOR CALCULUS ON MANIFOLDS
we …nd
which is a 2-nd rank tensor. We see that the 3-rd rank tensor t maps the vector,
w;
~ into a second rank tensor s: As another example let’s consider the third rank
tensor
t (~v ; ~eb ; w)
~ = ~v ~eb w:
~ (4.42)
Suppose we express the vectors, ~v and ~u; in terms of its controvariant compo-
nents
~ = wc~ec ; ~v = v a~ea ;
w (4.43)
then
where we …nd a 1st rank tensor, sb : This means the 3rd rank tensor mapped the
two vectors into 1st rank tensor (i.e. a vector).
sab = s (~ea ; ~eb ) = t (~ea ; ~eb ) r (~ea ; ~eb ) = tab rab : (4.45)
Outer product: Consider two …rst rank tensors (i.e. two vectors) u (~
p) and v (~q) :
The outer product or (tensor product) of these two tensors, which is denoted
by u v is given by
u v (~ p; ~q) = u (~
p) v (~q) ; (4.46)
which is a second rank tensor. Suppose (~
p; ~q) ! (~ea ; ~eb ) ; then
Now let’s consider 2-nd rank tensor, t (~ p; ~q), and 1-st rank tensor (a vector),
s (~r) ; the outer product of these two tensors give another tensor, h, of a di¤erent
rank given by
t s (~p; ~q; ~r) = t (~
p; ~q) s (~r) = h: (4.48)
The tensor h is a 3-rd rank tensor. Using the basis and dual basis vectors we
may express the components as
Example 4.2 Let’s consider two vectors in the tangent space for a point in a
3D sphere embedded in 4D manifold given by
p~ = pa~ea ; ~q = q b~eb
where a; b = 1; 2; 3, and ~e1 ; ~e2 ; and ~e3 (or ~er ; ~e ; and ~e' ). Find the
components of the 2-nd rank tensor for
u v (~
p; ~q) = u (~
p) v (~q) ; (4.53)
u p; ~q) = pa q b~ea~eb
v (~
= p1 q 1~e1~e1 + p1 q 2~e1~e2 + p1 q 3~e1~e3
+p2 q 1~e2~e1 + p2 q 2~e2~e2 + p3 q 3~e3~e3
+p3 q 1~e3~e1 + p3 q 2~e3~e2 + p3 q 3~e3~e3
t s = tab sb (4.54)
and
s t = tab sa (4.55)
are not necessarily the same.
Like vectors (1-st rank tensor) tensors of higher rank are geometrical objects
too: We already know that vectors are geometrical objects that can be made
up from a linear combination of the basis vectors
The vector that de…nes a given geometry on a manifold, does not depend on how
it is represented. The geometry that a vector de…nes remain the same geometry
whatever representation we used to describe the vector. The same is true for
118 CHAPTER 4. TENSOR CALCULUS ON MANIFOLDS
higher ranked tensors. As an example, let’s consider a 2-nd rank tensor with
component tab constructed from the outer product of two basis vectors
t = ~ea ~eb : (4.57)
The controvariant components of t
c d
t = ~ea ~eb ~ec ; ~ed = a b: (4.58)
Now suppose we have some general 2-nd rank tensor, t, with controvariant
components, tab (~ea ~eb ) : The action of this second rank tensor on two basis
vectors
tab (~ea ~eb ) ~ec ; ~ed = tab ca db = tcd (4.59)
Therefore what is true for example for 2-d rank tensor
t = tab (~ea ~eb ) = tba (~ea ~eb ) = tab ~ea ~eb (4.60)
is true for any ranked tensor.
3-nd rank tensor: Suppose we have a mixed 3-rd rank tensor, tcab ! t (~ea ; ~eb ; ~ec )
in the x0a coordinate system is given by
@xc @xd
@x0a @x0b
we have
@xc @xd @xc @xd
0a 0b
tcd = scd (4.71)
@x @x @x0a @x0b
so that using Eq. (4.68), we …nd
Well Eq. (4.72) shows that the equality holds under the coordinate transfor-
mation. However, the question is that these components (set of quantities) are
actually form a tensor. The quotient theorem set’s the condition for a set of
quantities actually represent a tensor component.
The quotient theorem: if a set of quantities when contracted with a tensor
produces another tensor, then the original set of quantities are also a tensor.
Suppose in an N dimensional manifold you are given a 3-rd rank tensor,
t; and 1-st rank tensor, v: The tensor, t; has a set of N 3 quantities tabc and
the tensor v has N quantities of v a : Suppose we form a set of N 2 quantities
by contracting the 4-th rank tensor formed by the outer product of these two
tensors (i.e.sab = tabc v c ). Under coordinate transformation xa ! x0a , these set of
elements, using the transformation relations,
@x0a c
t0a = t (4.73)
@xc
c
@x
t0a = tc (4.74)
@x0a
in the new coordinate system are given by
In the relation
@x0f c
v 0f =
v (4.76)
@xc
by switching the the primes with the none primes, we can write
@xf 0c
vf = v (4.77)
@x0c
Substituting this expression into Eq. (4.75), we …nd
rb v a = @b v a + a c
cb v (4.83)
rv = (rb v a ) ~ea : = @b v a + a c
cb v (4.84)
rb v a forms a mixed 2-nd rank tensor. We denote this second rank tensor by
rv. Noting that
r = ~ea @a ; v = v b~eb
4.7. COVARIANT DERIVATIVES OF A TENSOR 121
we can express
Let’s consider the the covariant derivative of a second rank tensor t with ex-
pressed in terms of its controvariant components tab
@c t = @c tab + a fb
cf t + b af
cf t (~ea ~eb ) = rc tab ~ea ~eb (4.92)
where
rc tab = @c tab + a db
cd t + b ad
cd t ; (4.93)
is the covariant derivative.
Homework: Show that for the covariant derivatives of the mixed and covari-
ant component of a 2-nd rank tensor t are given by
rc tab = @c tab + a d
dc tb
d a
bc td (4.94)
d d
rc tab = @c tab ac tdb bc tad (4.95)
Useful relation
@c~ea = a b
bc~
e (4.96)
Homework: Show that the covariant derivative of the metric tensor is zero
rg = 0
rc g ab = @c g ab + a db
cd g + b ad
cd g =0 (4.97)
122 CHAPTER 4. TENSOR CALCULUS ON MANIFOLDS
Useful relations, for example, the a¢ ne connection and the metric are related
by
f gf d
bc = (@b gcd + @c gdb @d gbc ) (4.98)
2
Application of the property of the metric tensor: Suppose we have a 2-nd rank
tensor, t, for which we want to …nd the covariant derivative from its components,
for example, in mixed form. That means we want to …nd rc tab from tad : We can
express the covariant components of this tensor using contraction as
tab = g bd tad :
tab = g bd tad
d~ea (u) d~ea (u) dxc d~eb (u) d~eb (u) dxb
= ; =
du dxc du du dxc du
and using the a¢ ne connection
d~ea (u) f
@c~eb = = bc e
^f (4.101)
dxc
4.8. INTRINSIC DERIVATIVE 123
we have
d~ea (u) f dxc d~eb (u) f dxb
= ac e
^f (u) ; = bc e
^f (u) (4.102)
du du du du
so that
dt (u) dtab dxc f dxb
= ~ea (u) ~eb (u) + tab f
ac e
^f (u) ~eb (u) + tab~ea (u) bc e
^f
du du du du
(4.103)
dt (u) dtab f dx
c
f dx
b
= ~ea (u) ~eb (u)+tab ac ~ef (u) ~eb (u)+tab bc ~ea (u) ~ef (u) :
du du du du
(4.104)
Replacing the dummy index a by f in the …rst term, we can write
dt (u) dtf b f dx
c
f dx
b
= ~ef (u) ~eb (u)+tab ac ~ef (u) ~eb (u)+tab bc ~ea (u) ~ef (u) :
du du du du
(4.105)
Noting that by making the following dummy index change b ! d followed by
a!b
b d d
f dx f dx f dx
tab bc ~ea (u) ~ef (u) = tad dc ~ea (u) ~ef (u) = tbd dc ~eb (u) ~ef (u)
du du du
we …nd
dt (u) dtf b f dx
c
f dx
d
= + tab ac + tbd dc ~ef (u) ~eb (u) : (4.106)
du du du du
that we expressed as
dt (u) Dtf b
= ~ef (u) ~eb (u) : (4.107)
du Du
where
Dtf b dtf b f dx
c
f dx
d
= + tab ac + tbd dc ; (4.108)
Du du du du
is called the intrinsic (absolute) derivative of the component tf b along the curve
de…ned by xa (u) : For the sake of convenience we make change of dummy indices
(f ! a) in Eq. (4.107) and (a ! d) in Eq. (4.108) as
dt (u) Dtab
= ~ea (u) ~eb (u) : (4.109)
du Du
where
Dtab dtab dxc dxd
= + tdb adc + tbd adc : (4.110)
Du du du du
We can switch the dummy indices c and d in the third term as the a¢ ne is
symmetric for Torsionless
Dtab dtab a dx
c
a dx
c
= + tdb dc + tbc cd :
Du du du du
124 CHAPTER 4. TENSOR CALCULUS ON MANIFOLDS
Dtab dxc a dx
c
a dx
c
dxc
= @c tab + tdb dc + tbc cd = @c tab + tdb a
dc + tbc a
cd
Du du du du du
Using the result in Eq. (4.93) we can write
Dtab dxc
= rc tab (4.113)
Du du
Chapter 5
In this chapter we will see how tensor calculus can be used in special relativity
that introduced in chapter one.
where
x0 = ct; x1 = x; x2 = y; x3 = z:
For the boost, for an inertial frame moving with a velocity, v; along the positive
x-direction,
2 2 2 2
ds2 = c2 d = dx00 dx01 dx02 dx03 ; (5.2)
125
126 CHAPTER 5. SPECIAL RELATIVITY USING TENSORS
Let’s determine the a¢ ne connections, obviously one can expect this to be zero
as we are using Cartesian coordinates. We recall the relationship between the
metric and the a¢ ne connection
a g ad
bc = (@b gcd + @c gdb @d gbc ) ; (5.4)
2
which gives
a
bc =0 (5.5)
This shows that the a¢ ne connections are zero everywhere in the Minkowski
spacetime manifold (in Cartesian coordinates).
where
@xc @xd
0 @x0a @x0b
gcd = gab ; gab = gcd : (5.10)
@x0a @x0b @xc @xd
Using Eqs. (5.6) and (5.7), one …nds
2 0 0 0 3 2 3
@x @x @x @x0
@x00 @x01 @x02 @x03 0 0
6 @x1 @x1 @x1 @x1 7 6
c @xc 6 @x00 @x01 @x02 @x03 7 6 0 0 7
7 ; (5.11)
[ a] = =6 7=4
@x0a 4 @x2 @x2 @x2 @x2 5 0 0 1 0 5
@x00 @x01 @x02 @x03
@x3 @x3 @x3 @x3 0 0 0 1
@x00 @x01 @x02 @x03
and
2 3 2 3
@x00 @x00 @x00 @x00
@x0 @x1 @x2 @x3 0 0
6 @x01 @x01 @x01 @x01 7 6
0b @x0b 6 @x0 @x1 @x2 @x3 7 6 0 0 7
7:
= =6 7=4
d
@xd 4 @x02 @x02 @x02 @x02 5 0 0 1 0 5
@x0 @x1 @x2 @x3
@x03 @x03 @x03 @x03 0 0 0 1
@x0 @x1 @x2 @x3
(5.12)
5.1. THE MINKOWSKI SPACETIME IN CARTESIAN COORDINATES127
1 v 1 v=c
= tanh ( ); = = (5.15)
c 0 v=0
and 2 3
cosh ( ) sinh ( ) 0 0
@x0b 6 sinh ( ) cosh ( ) 0 0 7
0b
= =6
4
7: (5.17)
d
@xd 0 0 1 0 5
0 0 0 1
Using Eqs. (5.16) and (5.17), we note that
2 32 3
cosh ( ) sinh ( ) 0 0 cosh ( ) sinh ( ) 0 0
6 sinh ( ) cosh ( ) 0 0 7 6 sinh ( ) cosh ( ) 0 0 7
c
[ a] 0b
= 6 4
76 7
d 0 0 1 0 54 0 0 1 0 5
0 0 0 1 0 0 0 1
2 3
1 0 0 0
6 0 1 0 0 7
) [ a] c 0b 6 7 c
d = 4 0 0 1 0 5 = [ d] (5.18)
0 0 0 1
Homework: We recall that the metric tensor can be used to lower or raise indices.
Using this property of the metric tensor for Lorentz transformation in Cartesian
coordinates for the boost …nd
2 3
cosh ( ) sinh ( ) 0 0
6 sinh ( ) cosh ( ) 0 0 7
0b
d =4
6 7 (5.19)
0 0 1 0 5
0 0 0 1
from 2 3
cosh ( ) sinh ( ) 0 0
6 sinh ( ) cosh ( ) 0 0 7
[ a] = 6
c
4
7: (5.20)
0 0 1 0 5
0 0 0 1
128 CHAPTER 5. SPECIAL RELATIVITY USING TENSORS
Similarly
c c
ub = ~u0 ~eb = u0a~e0a ~eb = u0a c
a~
ec ~eb = u0a a b = b 0b
au : (5.28)
where
2 3 2 3
cosh ( ) sinh ( ) 0 0 0 0
6 sinh ( ) cosh ( ) 0 0 7 6 0 0 7
[ b] = 6
0a
4
7=6
5 4
7 (5.29)
0 0 1 0 0 0 1 0 5
0 0 0 1 0 0 0 1
and
2 3 2 3
cosh ( ) sinh ( ) 0 0 0 0
6 sinh ( ) cosh ( ) 0 0 7 6 0 0 7
[ b] = 6
a
4
7=6
5 4 0
7 (5.30)
0 0 1 0 0 1 0 5
0 0 0 1 0 0 0 1
for a four vector we say the vector is time-like, null vector, or space like, when
8
< > 0; time-like
gab ua ub = = 0; null (5.33)
:
< 0; Space-like
where
1
u =q (5.38)
u2
1 c2
The 3-D velocity, ~u; is the velocity of the particle by an observer in the rest
inertial frame, S. In some other inertial from, S 0 ; the four velocity is given by
~u0 = u0a~e0a : (5.40)
The components can be determined using relation
u0a = u ~e0a = ub~eb ~e0a = 0a b
bu : (5.41)
where we used the transformation
~e0a = 0a c
c~
e : (5.42)
Applying the relation
2 3
v v 0 0
@x0a 6 0 0 7
0a
= =6
4
v v 7 (5.43)
c
@x c 0 0 1 0 5
0 0 0 1
5.1. THE MINKOWSKI SPACETIME IN CARTESIAN COORDINATES131
we have 2 3 2 32 0 3
u00 v v 0 0 u
6 u01 7 6 0 0 7 6 u1 7
6 02 7 = 6 v v 76 7 (5.44)
4 u 5 4 0 0 1 0 5 4 u2 5
u03 0 0 0 1 u3
so that 2 3 2 3
u00 v u0 u1
6 u01 7 6 u0 u1 7
6 02 7 = 6 v 7 (5.45)
4 u 5 4 u2 5
u03 u3
and substituting Eq. (5.39)
2 3 2 3
u0 uc
6 u1 7 6 u ux
7
6 2 7=6 7 (5.46)
4 u 5 4 u
u y
5
u3 u
u z
which means
dx0a dx01 dx02 dx03
[u0a ] = = u0 c; ; ; = u0 [c; ~u0 ] (5.48)
d dt0 dt0 dt0
ux
u0 = v u 1 ;
c
132 CHAPTER 5. SPECIAL RELATIVITY USING TENSORS
one …nds 2 3
2 3 ux c
u0x 6 1 uxc
uy 7
4 u0y 5 = 6 7
4 v (1
ux
c ) 5: (5.50)
u0z uz
ux
v (1 c )
Using
v 1
= ; =q (5.51)
c 1 v2
c2
Hint: the four acceleration can be de…ned in terms of the four velocity as
dua dua d
[aa ] = = u = u [ u c; u~
u] (5.56)
d dt dt
which gives
d d u d d d d~u
[aa ] = u [ u c; u~
u] = u ; ( ~u) = u c u ; ~u u +
c
dt dt dt u dt dt u
dt
d d d~u d d
) [aa ] = u c u ; ~u u + u = u c u ; u~a + ~u u (5.57)
dt dt dt dt dt
I prefer to call it the "proper mass", be m0 . The proper mass is the mass
measured by an observer moving with the same velocity of the particle (the
IIF). The four momentum, P~ , on the S inertial frame is de…ned in terms of the
proper mass (rest mass)
a
dx dx1 dx2 dx3
P~ = pa~ea = m0 ua~ea = m0 = u m0 c; m0 ; m0 ; m0 = u [m0 c; m0 ~u]
d d d d
(5.59)
Note that here ~u represent the three velocity. Suppose we de…ne the rest mass
energy which I also prefer to call it the "Proper energy" as
E0 = m0 c2 ; (5.60)
the four momentum can be expressed as
E0
[pa ] = u ; m0 ~u (5.61)
c
or
E
P~ = [pa ] = ; p~ ; (5.62)
c
where p~ is the three-momentum and
E0 m0 c2
E=q =q (5.63)
u2 2
1 c2 1 uc2
is the relativistic energy of the particle as measured by an observer on the S
frame and
m0 ~u
p~ = u m0 ~u = q : (5.64)
2
1 uc2
134 CHAPTER 5. SPECIAL RELATIVITY USING TENSORS
P~ P~ = pa~ea pb~eb = pa pb ab = pa ab pb
2 32 3
1 0 0 0 p0
6 0 1 0 0 7 6 p1 7
= p0 p1 p2 p3 6 4 0 0
76 7
1 0 54 p2 5
0 0 0 1 p3
2 0 3
p
6 p1 7 0 2 2 2 2
= p0 p1 p2 p3 6 7
4 p2 5 = p p1 p2 p3
p3
2 2
E E
) P~ P~ = p~ p~ = p2 : (5.66)
c c
is never be zero since the particle never travel with a speed c. In the case of a
photon or other particles traveling the a speed of light (if it exists), we can not
use the proper time as our a¢ ne parameter since for, u = c;
r
dt u2
d = = dt 1 = 0: (5.69)
u c2
5.3. FOUR MOMENTUM OF A PHOTON AND THE DOPPLER EFFECT135
If t is the tangent vector de…ning the worldline of a photon to the curve de…ning
the trajectory of a photon in the Minkowski spacetime manifold
ds = td = 0; (5.70)
xa = ua " (5.71)
x0 = ct; x = x1 = ct; y = x2 = 0; z = x3 = 0
) xa = [ct; ct; 0; 0] = ua " = [1; 1; 0; 0] ": (5.72)
Which means the tangent vector, ua ; that de…nes the curve for the photon in
terms of this parameter is given by
dxa
~u = ua~ea = ~ea ; (5.73)
d"
where
ua = [1; 1; 0; 0] ; (5.74)
which is a none null vector. But the curve de…ned by
2
ds
u:u = = 0; (5.75)
d"
which is still a null curve. This tangent vector, which gives the four velocity of
the photon is di¤erent from that of a massive particle, where we saw
2
ds
u:u = = c2 (5.76)
d
We also note that the equation of motion for a photon can also be expressed in
terms of this parameter given by
du d
= [1; 1; 0; 0] = 0: (5.77)
d" d"
In the case of a massive particle to …nd the four momentum, we multiplied the
four velocity by the rest mass ("proper mass"). We follow a similar approach
here also. We multiply the four velocity by some parameter, ; so that
P~ = ~u ) pa = ua : (5.78)
136 CHAPTER 5. SPECIAL RELATIVITY USING TENSORS
E
P~ = pa~ea = ; p~ : (5.79)
c
2 2 2 2
k= ) ~k = ; ; : (5.81)
x y y
~ = k u~eu = 2 2
K ; k cos ( ) ; k sin ( ) ; 0 = (1; cos ( ) ; sin ( ) ; 0) ; (5.86)
where we used
2
k= : (5.87)
~ 0 = k 0u~e0u = 2 0 0 2 0 0
K 0 ; k 0 cos ; k 0 sin ;0 = 0 1; cos ; sin ;0 :
(5.88)
We want to …nd the components, k 0u : These components can easily be de-
termined from the fact that vectors or tensors on a manifold are geometrical
properties and would remain the same. That means the four-wavevector must
be the same in S and S 0 coordinate system. Therefore, we can write
Using
2 3
0 0
6 0 0 7
0a
=6
4
7 (5.90)
b 0 0 1 0 5
0 0 0 1
138 CHAPTER 5. SPECIAL RELATIVITY USING TENSORS
we have
2 32 3 2 3
0 0 1 cos ( )
2 6
6 0 0 7 6 7 6
7 6 cos ( ) 7 = 2 6 + cos ( ) 7 7
[k 0u ] = 4 5 4 5 4 5
0 0 1 0 sin ( ) sin ( )
0 0 0 1 0 0
(5.91)
so that 2 3 2 3
1 cos ( )
2 6 cos 0 7 2 6 + cos ( ) 7
[k ] = 0 6
0u
4 sin 0
7=
5
6
4
7:
5 (5.92)
sin ( )
0 0
From the …rst equation follows that
1
0 = (1 cos ( )) ) 0 = (1 cos ( )) : (5.93)
Dividing Eq. (5.95) by Eq. (5.94), we …nd the angle of emission by the laser
pointer
0
sin sin ( ) 0 tan ( )
0 = ) tan = (5.96)
cos (cos ( ) ) [1 (v=c) sec ( )]
Note: In our derivation of the Doppler e¤ ect even though the emission and
observation of the photon occurs at two di¤ erent events that took place at dif-
ferent time, we took the four momentum to be constant. This is because the
equation of a photon in terms of the a¢ ne parameter, "; we used for the world-
line is a constant
dP~
=0 (5.100)
d"
dP~ dpa
F~ = = ~ea (5.103)
d d
and the components of the four-force can be expressed as
dpb dpb dpa
f a = F~ ~ea = ~eb ~ea = a
b = : (5.104)
d d d
We recall for a particle traveling with a speed, u, the proper time is given by
dt
d =
u
dE dW ~
dD
= = f~ = f~ ~u; (5.107)
dt dt dt
140 CHAPTER 5. SPECIAL RELATIVITY USING TENSORS
d~
p
f~ = ; (5.108)
dt
is the three-force. Thus we can write the four-force as
" #
~ f~ ~u ~
F = u ;f : (5.109)
c
We recall the energy and the three-momentum of the massive particle moving
with a velocity, ~u;
E0 m0 c2 m0 ~u
E=q =q ; p~ = u m0 ~
u =q (5.110)
u2 2 2
1 c2 1 uc2 1 uc2
dP~
~u F~ = u : (5.112)
d
If we write the four-momentum in terms of the mass, m; and the four-velocity
as
P~ = m~u = mua~ea ; (5.113)
where m is the relativistic mass, then we have
dP~ d d~u dm
~u F~ = ~u = ~u (m~u) = m~u + ~u ~u : (5.114)
d d d d
and noting that for the four-velocity
d~u
~u ~u = c2 ) 2~u =0
d
one can write
d~u dm dm
~u F~ = m~u + c2 = c2 : (5.115)
d d d
If the force is pure, the mass does not change and we have
dm
= 0 ) m = m0
d
and we …nd
~u F~ = 0: (5.116)
5.4. RELATIVISTIC MECHANICS FOR A MASSIVE PARTICLE 141
h h h
m0 c + = u m0 c + ) u =1+( ); (5.132)
c c m0 c2
h h h h
= cos ( ) + u m0 u cos ( ) ) u m0 u cos ( ) = cos ( ) (5.133)
c c c c
c c
0 = sin ( ) u m0 u sin ( ) ) m0 u sin ( ) = sin ( ) (5.134)
h h u
Substituting this into the Eqs. (5.133) and (5.134) and also using Eq. (5.132),
we can show that
1
h
= 1+ (1 cos ( )) (5.136)
m0 c2
~u ( ) ~u ( ) = c2
we note that
d~u ( ) 1 d 1 d 2
~a ( ) ~u ( ) = ~u ( ) = [~u ( ) ~u ( )] = c = 0: (5.137)
d 2d 2d
Eq. (5.137) shows that the four acceleration is orthogonal to the four veloc-
ity. The parameter (the proper time) we used to de…ne the worldline in the
Minkowski space time is measured by an observer that is at rest frame at all
the time. An accelerating observer is not at rest at all time. However, we can
…nd for such an observer what is known as instantaneous reference frame (IRF)
S 0 on which the accelerating observer is at rest momentarily. We recall that the
basis vectors on the worldline de…ned by
dxa
d~s = ~ea ( ) dxa = ~ea ( ) d = ua ( ) ~ea ( ) d
d
d~s
) = ua ( ) ~ea ( ) = ~u; (5.138)
d
where ~u is the four velocity on the S frame. Similarly for another frame (ac-
clerating), S 0 ; we can write
dx0a d~s0
d~s0 = ~e0a ( ) dx0a = ~e0a ( ) d = u0a ( ) ~e0a ( ) d ) = u0a ( ) ~e0a ( ) :
d d
(5.139)
If this reference frame is an IRF, where the observer is momentarily at rest, we
must have
d~s0
u01 ( ) = u02 ( ) = u03 ( ) = 0 ) = u00 ( ) ~e00 ( ) : (5.140)
d
We know that vectors are geometrical enteties, whether is accelerating or none
accelerating it must remain the same independent of the reference frame. There-
fore
d~s d~s0
= ) ~u = ua ( ) ~ea ( ) = u00 ( ) ~e00 ( ) (5.141)
d d
5.6. ACCELERATING OBSERVERS AND THE TETRADS 145
This means the timelike basis vector ~e00 in the IRF frame is parallel to the
four velocity of the accelerating observer ~u ( ) : The remaining spacelike basis
vectors (~e01 ; ~e02 ; ~e03 ) are orthogonal to ~e00 and to one another and will depend on
the relative velocity of S and S 0 and the relative orientation of their spacial
axes. Therefore observation made by an observer at event, p; on the worldline
whether accelerating or none accelerating corresponds to the instantaneous rest
frame (IRF) S 0 at p. An observer carries along four orthogonal unit vectors,
~e0a ( ) ; (or tetrad) which vary along his worldline but always satisfy
d~s ua ( )
= ~ea ( ) = u
^: (5.143)
d c
the timelike unit vector which is parallel to the four-velocity of the accelerating
observer can be expressed as
~e00 ( ) = u
^ (5.144)
At any event P along the observer’s worldline, the tetrad comprises the basis
vectors of the Cartesian IRF at the event P and de…nes a time direction and
three space directions to which the observer will refer all measurements. Thus,
the results of any measurement made by the observer at the event P are given
by projections of physical quantities (tensors) onto those tetrad vectors.
Electromagnetism
In this chapter by applying what we have learned about tensor calculus, we will
develop the entire theory of relativistic electromagnetism from one equation, the
Lorentz force equation
f~E = q E
~ (6.1)
dpa dpa d E 1 dE d~
p
[f a ] = = v = v ; p~ = v ; : (6.4)
d dt dt c c dt dt
dE = dW = f~ dD;
~ (6.5)
dE dW ~
dD
= = f~ = f~ ~v ; (6.6)
dt dt dt
147
148 CHAPTER 6. ELECTROMAGNETISM
~v f~ = q ~v E
~ + q~v ~v ~ = q ~v E
B ~ (6.9)
!
If we express the EM …eld tensor, ; in terms of its covariant components
and the 4-velocity in terms of its controvariant components, we may write the
components of the four force as
b
Fa = q ab u : (6.17)
F~ ~u = 0 ) Fa ua = q b a
ab u u =0 (6.18)
or
F~ ~u = 0 ) Fb ub = q a b
ba u u =0 (6.19)
which leads to
b a
q( ab + ba ) u u = 0: (6.20)
There follows that
ab = ba ; (6.21)
which means the EM tensor is antisymmetric tensor. The mixed and con-
trovariant component of the EM tensor can be obtained from the metric tensor
for Minkowski spacetime
b
a = g bc ac ;
ab
= g ad b
d = g ad g bc dc (6.22)
where N is the total number of charges in the cube. Note that the proper
length and the proper number density are measured on a rest frame. Then for
the proper charge density, 0 ; and current density, ~j0 ; on can write
0 = qn0 ; ~j0 = 0:
where we assumed that each particle carries a charge, q: Now let’s consider an
Figure 6.1: A cube with length l0 ; and uniform charge density on an inertial
frame S moving with a velocity u in the positive x-direction. The number of
charges per unit volume in the cube, n0 as measured by an observer on S 0 .
observer on a di¤erent inertial frame, S 0 ; moving with a velocity, ~u; along the
positive x-axis. For an observer on the S frame since the length of the cube
along the x-direction is Lorentz contracted
r
u2 l0
l = l0 1 = (6.24)
c2 u
J~ = c ; ~j (6.29)
Recalling that tensors are geometrical properties to the manifold it would remain
the same in any reference frame. Therefore, noting that in the S 0 frame
2
J~0 = (c 0 ; 0) ) J~0 J~0 = (c 0 ) (6.30)
Note that is the charge density of a volume of charge moving with a velocity
~u in the S frame.
O E
~ = (6.32)
0
where E~ is the electric …eld vector. But let’s pretend we do not know Gauss’s
law. But we do know that EM …elds are a result of charges and/or current. We
have seen that both the charge and current densities can be described in terms
of the 4-current density in the Minkowski spacetime manifold. In EM theory
!
we have 2-nd rank tensor, which is the …eld and 1-st rank tensor which is
the 4-current density. We do not have any other 1-st rank tensor that we can
contracted with the EM …eld tensor to get the 1-st rank tensor (the 4-current
density). But we do have the 4-gradient that we can use to contract the …eld
tensor and get the 4-current density. That means
!
r = k J~ (6.33)
where k is some constant that we do not know yet. If the …eld is expressed in
terms of its controvariant components, we may write Eq. (6.33) as
ab
@a = kJ b : (6.34)
The law of conservation of charge (the continuity equation): Let’s …nd @b of Eq.
(6.34)
@b @a ab = k@b J b : (6.35)
152 CHAPTER 6. ELECTROMAGNETISM
We have shown that the EM …eld tensor is antisymmetric, that means we can
write
@b @a ab = @b @a ab = @a @b ab = @b @a ab (6.36)
Homework: Show that
ab ab
@b @a = @b @a (6.37)
for a,b=1,2.
There follows that
@ (c ) @
@b J b = 0 ) + r J~ = 0 ) + r J~ = 0 (6.38)
@ (ct) @t
we can write 8 9
01 02 03
>
> 0 >
>
< 01 12 13 =
ab 0
= 02 12 23 : (6.39)
>
> 0 >
>
: 03 13 23 ;
0
Note that the diagonal elements are zero because aa = aa
only when
aa
= 0: This means that there are 6 independent components that we need to
determine. But so far we have only four …led equations given by
ab
@a = kJ b
00 10 20 20
) @0 + @1 + @2 + @2 = kJ 0
01 11 21 21
@0 + @1 + @2 + @2 = kJ 1
02 12 22 22
@0 + @1 + @2 + @2 = kJ 2
03 13 23 23
@0 + @1 + @2 + @2 = kJ 3 (6.40)
bc = @ b Ac @c Ab : (6.41)
Since the …eld equation involves the derivative of the EM …eld tensor, let’s …nd
@a of this equation
@a bc = @a @b Ac @a @c Ab : (6.42)
Applying the property of the metric tensor gab for Minkowski spacetime in
Cartesian coordinates, we can write
We recall that the metric tensor for Minkowski spacetime in Cartesian coordi-
nates is a constant and therefore, we can write
@a g ab bc = g ab (@a @b Ac @a @c Ab ) ) @a a
c = g ab (@a @b Ac @a @c Ab ) (6.44)
Using the controvariant components, one can also write the
ab
@a = kJ b ) @c gab cb
= kgab J b ) @c c
a = kJa ) @a a
c = kJc : (6.45)
Combining Eqs. (6.44) and (6.45), we can write
g ab (@a @b Ac @a @c Ab ) = kJc : (6.46)
Homework: Show that in terms of the …eld tensor, Eq. (6.46) this can be written
as
@c ab + @a bc + @b ca = 0 (6.47)
Solution:
@a bc = @a @b Ac @a @c Ab (6.48)
switching the indices a and b
@b ac = @b @a Ac @b @c Aa
) @b ca = (@b @a Ac @b @c Aa ) = @b @a Ac + @b @c Aa (6.49)
and the indices a and c
@c ba = @c @b Aa @c @a Ab
) @c ab = (@c @b Aa @c @a Ab ) = @c @b Aa + @c @a Ab (6.50)
Note that in Eqs. (6.49) and (6.50) we used the antisymmetric property of the
…eld tensor. Now adding Eqs. (6.48)-(6.50), we …nd
@c ab + @a bc + @b ca = @c @b Aa + @c @a Ab + @a @b Ac @a @c Ab
+ @b @a Ac + @b @c Aa ) @c ab + @a bc + @b ca =0
Therefore the complete EM …eld equations are
@c ab + @a bc + @b ca =0
ab b
@a = kJ (6.51)
where the constant, k = 0; in SI unit. Using the notations for antisymmetric
permutation
1
t[a1 a2 :::aN ] = (Alternating subtraction and addition over
N!
all permutations of the indices a1 a2 :::aN ) (6.52)
one can write
@[c ab] = 0;
so the EM …eld equations
ab b
@[c ab] = 0; @a = 0J
g ab @a @b Ab = @a @b g ab Ab = @a @b Ab = 0 (6.54)
becomes
g ab @a @b Ac = @a g ab @b Ac = @a @ a Ac = kJc (6.56)
For Minkowski spacetime in Cartesian coordinates, we have
1 @ @ @ @
[@a ] = ; ; ; ;
c @t @x @y @z
2 32 1 @ 3 2 1 @ 3
1 0 0 0 c @t c @t
6 0 1 0 0 7 6 @ 7 6 @ 7
@a = ab
g @b = 46 76 @x 7=6 @x 7
0 0 1 0 5 4 @ 5 4 @ 5
@y @y
0 0 0 1 @ @
@z @z
Ac = 0 Jc ) g ac Ac = 0g
ac
Jc ) Aa = 0g
ac
Jc = 0J
a
(6.57)
@ V ~ 1 @V ~ = 0:
@a Aa = 0 ) ;r A =0) +r A (6.59)
c@t c c2 @t
6.4. ELECTROMAGNETISM IN THE LORENTZ GAUGE 155
1 @ 2 A0 @ 2 A0 @ 2 A0 @ 2 A0 0
= 0J ; (6.60)
c2 @t2 @x2 @y 2 @z 2
1 @ 2 A1 @ 2 A1 @ 2 A1 @ 2 A1 1
= 0J ; (6.61)
c2 @t2 @x2 @y 2 @z 2
1 @ 2 A2 @ 2 A2 @ 2 A2 @ 2 A2 2
= 0J ; (6.62)
c2 @t2 @x2 @y 2 @z 2
1 @ 2 A3 @ 2 A3 @ 2 A3 @ 2 A3 3
= 0J : (6.63)
c2 @t2 @x2 @y 2 @z 2
First let’s consider Eq. (6.60) Using the component J 0 = c ; from the four-
current density,
[J a ] = c ; ~j ; (6.64)
1 @2V
r2 V = c 0J
0
: (6.65)
c2 @t2
~ (electric …eld vector) de…ned by
Introducing a vector …eld, E;
~
@A @
~ =
E rV ~ =
); r E r2 V ~
r A (6.66)
@t @t
one can express the Lorentz gauge, in terms of this vector …eld as
2
~= 1 @V ~ = 1 @ V
r A ) r E r2 V (6.67)
c2 @t c2 @t2
Upon substituting this into Eq. (6.65), we …nd
~ =
r E ; (Gauss’s law!)
0
~ = A1 ; A2 ; A3 ; ~j = J 1 ; J 2 ; J 3
A
one …nds
1 @2A~
~ = 0~j
r2 A (6.68)
2
c @t 2
~ =r
B ~
A: (6.69)
we have
r ~ =r
B r ~ =r r A
A ~ ~
r2 A: (6.70)
156 CHAPTER 6. ELECTROMAGNETISM
~ 1 @V ~= 1 @ (rV ) ~
r B = r r2 A r2 A
c2 @t c2 @t
~= ~ 1 @ (rV )
) r2 A r B : (6.72)
c2 @t
Using this relation, Eq. () can be rewritten
" #
2~ ~
~ + 1 @ A + 1 @ (rV ) =
r B ~ ~ 1
0j ) r B + 2
@
rV +
@ A
= ~ (6.73)
0j
c2 @t2 c2 @t c @t @t
~
since for any vector A
r r ~ =0
A (6.78)
E1 = c @1 A0 + @0 A1 ; E 2 = c @2 A0 + @0 A2 ; E 3 =
c @3 A0 + @0 A3 :
(6.84)
Using the metric tensor one can determine the covariant components for the
four-vector potential
0 1 0 10 0 1
A0 1 0 0 0 A
B A C B 0 1 0 0 C B 1 C
Aa = gab Ab ) B 1 C
= B C B A2 C
@ A2 A @ 0 0 1 0 A@ A A
A3 0 0 0 1 A3
0 1 0 1
A0 A0
B A1 C B A1 C
)B C B
@ A2 A = @ A2 A :
C (6.85)
3
A3 A
We recall that the elements of the …eld tensor is related to the components of
the four-vector potential by
ab = @a Ab @ b Aa : (6.87)
B 1 = @ 2 A3 @3 A2 ; B 2 = @3 A1 @1 A3 ; B 3 = @1 A2 @2 A1 : (6.91)
Noting that for the controvariant components for spacial part of the four-vector
can be determined from the corresponding covariant component by using the
metric tensor
Aa = g ab Ab = Aa (6.92)
one can write
B 1 = @2 A3 @3 A2 ; B 2 = @3 A1 @1 A3 ; B 3 = @1 A2 @2 A1 (6.93)
A0a = a b
bA ;
0ab
= a
c
b
d cd : (6.97)
Homework:
@c ab + @a bc + @b ca = 0 or g ab (@a @b Ac @a @c Ab ) = 0 Jc
ab b
@a = 0J (6.98)
6.5. ELECTROMAGNETISM IN ARBITRARY COORDINATES 159
Figure 6.2: A cube with length l0 ; and uniform charge density on an inertial
frame S moving with a velocity u in the positive x-direction. The number of
charges per unit volume in the cube, n0 as measured by an observer on S 0 .
In any other coordinates the metric tensor @a should also be replaced by the
covariant derivative ra :
Homework # 1: Show that for the covariant derivatives of the mixed and
covariant component of a 2-nd rank tensor t are given by
rc tab = @c tab + a d
dc tb
d a
bc td (6.99)
d d
rc tab = @c tab ac tdb bc tad (6.100)
Useful relation
@c~ea = a b
bc~
e (6.101)
Homework # 2: Show that the covariant derivative of the metric tensor is zero
rg = 0
rc g ab = @c g ab + a db
cd g + b ad
cd g =0 (6.102)
Useful relations, for example, the a¢ ne connection and the metric are related
by
f gf d
bc = (@b gcd + @c gdb @d gbc ) (6.103)
2
Then the …eld equations should be written as
rc ab + ra bc + rb ca = 0 or g ab (ra rb Ac r a r c Ab ) = 0 Jc
ab b
@a = 0J (6.104)
160 CHAPTER 6. ELECTROMAGNETISM
Anew
a = Aa + r a = Aa + @a (6.105)
r a Aa = 0 (6.106)
= g ab ra rb
Charge conservation:
ra J a = 0 (6.108)
The EM …eld tensor and the 4-vector potential : When we transform from an
inertial from S to S 0 for arbitrary coordinates should be determined from
For an arbitrary coordinates, one can then write the controvariant component
of the force in terms of the corresponding components for the momentum
where still EM force is pure. Then the equation of motion in terms of the
controvariant components can be written as
dua a c dx
b
a b
m0 + cb u =q bu (6.115)
d d
dxb
= ub (6.116)
d
we may write
dua a b c a b
m0 + cb u u =q bu : (6.117)
d
Or in terms of only the coordinates
d2 xa a dx
b
dxc a dx
b
m0 + bc =q b (6.118)
d 2 d d d
d2 xa a dx
b
dxc q a dx
b
+ bc = b (6.119)
d 2 d d m0 d
a
In the absence of EM …eld or for none charged particle (i.e. b = 0 or q = 0),
we …nd
d2 xa b
a dx dx
c
+ bc =0 (6.120)
d 2 d d
which means the motion of the particle is Geodesic!
Homework # 3: Problem 6.3
Homework # 3: Problem6.5
162 CHAPTER 6. ELECTROMAGNETISM
Chapter 7
F~ (~r) = q E
~ (~r) : (7.1)
~ is usually determined from the electrostatic potential V (~r)
This electric …eld E
as
~ (~r) = rV (~r)
E (7.2)
~ is due to some volume charge distribution with
Suppose this electric …eld E
charge density, ;, we know from the divergence theorem that
~ (~r) = 1
r E (~r) (7.3)
0
where ~g (~r) is the gravitational …eld determined from the gravitational potential,
;
~g (~r) = r (~r) : (7.6)
163
164CHAPTER 7. THE EQUIVALENCE PRINCIPLE AND SPACETIME CURVATURE
Like the electric …eld, the gravitational …eld is due to some mass distribution
in space that can be described a mass density, (~r). Then in view of Eq. (7.3),
the gravitational …eld, ~g ; can be expressed as
where G is the universal gravitational constant. The result in Eq. (7.8) is not
consistent with special theory of relativity since it does not display any time
dependence. This means the gravitational …eld due to some mass distribution
in space is instantaneously felt by another mass in the universe billion light
years away. This requires a travel speed greater than the speed of light which is
in contradiction with special theory of relativity. There is a second fundamental
di¤erence between electromagnetic and gravitational forces. The equation of
motion for a particle with inertial mass, mI ; is given by the 3-force
d2 ~x d2 ~x mG mG
mI = F~ = mG~g ) = ~g = r : (7.9)
dt dt mI mI
However, it is a well-established experimental fact that m
mI is the same for all
G
d2 ~x
= r ; (7.10)
dt2
which is independent of the nature of the particle. This result is di¤erent from
the equation of motion for charged particle where it depends on the mqI which
varies from particle to particle.
The result in Eq. (7.10) shows there is an equivalence between the inertial
mass, mI ; of a particle that determines the particle resistance to an applied
force and gravitational mass, mG ; the quantity that determines the magnitude
of the gravitational force.
The equivalence principle: In a freely falling (non-rotating) laboratory oc-
cupying a small region of spacetime, the laws of physics are those of special
relativity.
a
(p) ' 0 (7.16)
and the geodesic
d2 X a a dX
b
dX c
+ bc =0
d 2 d d
becomes
d2 X a
' 0;
d 2
where a = 0; 1; 2; 3 and X 0 is replaced by ct. This means X a de…ne a local
Cartesian inertial frame in which the laws of special relativity hold locally. We
can then construct such a system, spacetime is pseudo-Riemannian manifold
where the metric is given by
This means
a
b (p) = 0 (7.18)
and the coordinates basis vectors form an orthonormal set at point, p
e^0 (p) ! u
^ (p) -Time like, e^i (p) for i = 1; 2; 3 -Spacelike
7.3. OBSERVERS IN A CURVED SPACETIME 167
For points near to the event at point p, the metric in a local inertial coordinate
system, X a
1
gab (X) = gab (p) + (@c @d gab )p X c X d + ::: (7.21)
2
or
1
gab (X) = ab (p) + (@c @d gab )p X c X d + ::: (7.22)
2
where we used
gab (p) = ab and (@c gab )p = 0:
where
u
^( )
e^0 ( ) ! -Time like, e^i ( ) for i = 1; 2; 3 -Spacelike, (7.24)
c
which are unrelated to ~ea ( ) that we used to label the curved spacetime. If the
observer has the four acceleration
d~u
~a ( ) = ; (7.25)
d
but not rotating, the tetrad basis vectors are Fermi-Walker-transported along
the observer’s worldline satisfying the equation
d^
ea ( ) 1 d~u
= 2 (~u e^a ) (~a e^a ) ~u : (7.26)
d c d
Figure 7.1: Parallel transport of a vector ~v (Shown by the yellow parallel arrows)
on a manifold.
eab ( )
D^ ea ( )
d^ a
= b + cd ecb ( )) ud = 0
(^ (7.30)
D d
This equation is useful for determining what a freely falling observer would
measure at a given event in spacetime.
where jhab j << 1: Let’s assume that in the coordinates system where Eq. (7.32)
is valid, the metric is stationary. That means
@gab
@0 gab = =0 (7.33)
c@t
In other words this mean that the basis vectors do not change with time. An
example is a Cartesian coordinates on a none-rotating earth. We recall that the
worldline of a freely falling object is geodesic
d2 xa a dx
b
dxc
+ bc = 0: (7.34)
d 2 d d
In a weak gravitational …eld we can say the speed of the particle is much smaller
than the speed of light. This requires each components of the three velocity of
the particle must be smaller compared to the speed of light.
Let’s determine the connection. We recall that the connection is related to the
metric by
a g ad
bc = (@b gcd + @c gdb @d gbc ) (7.37)
2
so that
a g ad
00 = (@0 g0d + @0 gd0 @d g00 ) : (7.38)
2
Since the metric is assumed to be stationary
@0 g0d = @0 gd0 = 0
which leads to
a g ad g a0 g ai g ai
00 = @d g00 = @0 g00 + @i g00 = @i g00 : (7.39)
2 2 2 2
@i 00 = 0;
170CHAPTER 7. THE EQUIVALENCE PRINCIPLE AND SPACETIME CURVATURE
we …nd
1 ai a 1
@i h00 + hai @i h00 '
00 = ai
@i h00 : (7.41)
2 2
where we made the approximation
a 1 ai
00 = @i h00 : (7.42)
2
and for a = 0; 1; 2; 3, we have
0 1 0i 1 1 1i 2 1 2i 3 1 3i
00 = @i h00 ; 00 = @i h00 ; 00 = @i h00 ; 00 = @i h00 ; (7.43)
2 2 2 2
which leads to
0 1 1 2 1 3 1
00 = 0; 00 = @1 h00 ; 00 = @2 h00 ; 00 = @3 h00 : (7.44)
2 2 2
From the result in Eq. (7.44) one can easily see that
a 1
00 = rh00 : (7.45)
2
Substituting this result into Eq. (7.36), we …nd
2 2
d2 xa dt d2 ~x c2 dt
= c2 a
00 ) = rh00
d 2 d d 2 2 d
which can be rewritten as
d2 ~x c2 c2
= rh00 = r h00 : (7.46)
dt2 2 2
We recall in Newtonian gravity
d2 ~x mG mG
= r =r (7.47)
dt mI mI
Now comparing Eqs. (7.46) and (7.47), we can easily see that
c2 mG
h00 = (7.48)
2 mI
which gives
2 mG
h00 =
c2 mI
7.4. WEAK GRAVITATIONAL FIELDS AND THE NEWTONIAN LIMIT171
is given by Z
1 1 Q
V (~r) = (r0 ) dr03 = : (7.52)
4 0 4 0 r
Then for
r2 = (4 G) : (7.53)
one can write
GM
(~r) = (7.54)
r
so that
2 GM 1
= : (7.55)
c2 c2 r
Let’s look at the values of this on a surface of the earth, the sun, and a white
dwarf star.
2
Object radius, r = R in m Mass, M in Kg c2
6 24
Earth 6:37 10 5:97 10 10 9
Sun 6:96 108 1:99 1030 10 6
white dwarf star 10 4
where we used for the universal gravitational constant, G
11 m3
G = 6:67408 10 (7.56)
kg:s2
Homework: …nd the value for 2c2 the white dwarf star.
Consider a freely falling massive particles in a weak gravitational …eld pro-
duced by another particle. In each of these particles there are two clocks at
rest,
dxi
=0 (7.57)
dt
Suppose an observer on the freely falling particle frame measures a proper time
interval, d ; between two clicks (two events), is given by
2 2
(ds0 ) = c2 (d ) : (7.58)
This interval must be the same as the interval measured on the reference frame
172CHAPTER 7. THE EQUIVALENCE PRINCIPLE AND SPACETIME CURVATURE
2 2 2 2
(ds) = g00 dx0 = 1+ c2 (dt) : (7.60)
c2
2 GM 1 2 2
(ds) = 1 c (dt) : (7.62)
c2 R
Eqs. (7.58) and (7.62) must be the same since the interval is independent of the
coordinates system. Therefore
s
GM 1 2 2 2 2 GM 1
1 2
c (dt) = c (d ) ) d = 1 dt
c R c2 R
Time dilation in a weak gravitational …eld! For example for a particle accelerated
by earth graviational …eld, using the result we eastimated
2 GM 1 9 GM
= = 10 ) v2 = = 10 9 2
c (7.65)
c2 c2 R R
we …nd
d ' dt: (7.66)
ab
@c ab + @a bc + @b ca = 0; @a = kJ b : (7.67)
The di¤erence is that since now spacetime is curved and the basis vectors are no
longer constant, instead of partial derivatives, @b ; is replaced by the covariant
derivative, rb such that
rb J a = @b J a + a c
cb J ;
a a a d d a
rc b = @c b + dc b bc d; (7.68)
d d
rc ab = @c ab ac db bc ad ; (7.69)
ab
rc ab + ra bc + rb ca = 0; ra = kJ b : (7.70)
The equation of motion for a charged particle will also be the same
Dua q a b
= bu (7.71)
D m0
Dua dua a c dx
b
= + cb u : (7.72)
D d d
But the particle is moving under the in‡uence of gravitational and EM …elds.
The e¤ect of the gravitational …eld is just to change the curvature of the space-
time. This can also be put in terms of the coordinates variables, xa ; as
d2 xa a dx
b
dxc q a dx
b
+ bc = b : (7.73)
d 2 d d m0 d
174CHAPTER 7. THE EQUIVALENCE PRINCIPLE AND SPACETIME CURVATURE
If this geometry is not curved (if it is has a plane geometry), the metric takes
the form
2 2 2
ds2 = 1 dx1 + 2 dx2 + ::: n dxN ; (7.75)
where n = 1: Next we want to generally quantify the curvature of a Rie-
mannian geometry de…ned by
2
(ds) = gab dxa dxb : (7.76)
It is not always easy to …nd out a coordinate transformation exists that allows
us to write the metric in the form of Eq. (7.75) so that one can …nd out whether
the manifold is curved or not. Therefore, we need to seek a way we can measure
the curvature of a manifold or a region of manifold. The curvature tensor (or
Riemann tensor) is a tensor that we use to measure the curvature of a manifold.
We next derive this tensor. We recall for a vector, ~v ; expressed in terms of its
covariant components
~v = va~ea (7.77)
the covariant derivative is given by
where
e
rb va = @b va ab ve : (7.79)
Then taking the covariant derivative one more time for this vector, one can also
write
e e
rc r b va = rc (@b va ab ve ) = rc (@b va ) rc ( ab ve ) (7.80)
e e
= rc (@b va ) ab rc ve rc ( ab ) ve : (7.81)
Noting that
rb va = tba
is a second rank tensor and for a second rank tensor the covariant derivative is
given by
d d
rc tba = @c tba bc tda ac tbd
+ dab @c vd d e
ab dc ve
d
ac @b vd +
d e
ac db ve
= (@b eac e d e
@c ab ) ve + ac db ve d e
ab dc ve
+ eac @b ve e e
ab @c ve + ab @c ve
e
ac @b ve (7.87)
so that we …nd
e e d e d e d
rc rb va rb rc va = @b ac @c ab + ac db ab dc ve = Rabc vd ;
where
d d d e d e d
Rabc = @b ac @c ab + ac eb ab ec ;
Conversely, if Eq. (7.92) is true, then we can …nd coordinates xa such that the
interval can be expressed as Eq. (7.89).
The symmetry: In order to …nd out the symmetry of the curvature tensor,
we lower the index. Using the metric tensor
e e e f e f e
Rabcd = gae Rbcd = gae @c bd @d bc + bd f c bc f d : (7.93)
Homework: Applying
a g ad
bc = (@b gcd + @c gdb @d gbc ) (7.94)
2
show that
1
Rabcd = (@d @a gbc @d @b gac + @c @b gad @c @a gbd )
2
g ef ( eac f bd ead f bc ) : (7.95)
which leads to
Rabcd = Rcdab (7.99)
Homework: Show Eqs. (7.98) and (7.99)
The cyclic identity: the curvature tensor satis…es the cyclic identity for the
last three indices
Rabcd + Radbc + Racdb = 0; (7.100)
or using the symmetries in Eqs. (7.98) and (7.99) along with the notation we
introduced to, we can write
Ra[bcd] = 0; (7.101)
7.6. THE CURVATURE TENSOR 177
Taking into account the symmetry and cyclic properties, for N dimensional
manifold, the components of the curvature tensor are N 2 (N 2 1)=12 instead of
N 4 . For example for
N = 2 ! 1 component
N = 3 ! 6 components
N = 4 ! 20 components (7.103)
The di¤ erential identity: the curvature tensor also satis…es the di¤erential
identity at a point, p, where the metric is a constant and diagonal and the a¢ ne
connections vanish
gae = ae a ; fbc = 0 (7.104)
we have
e e e f e f e
Rabcd = gae Rbcd = gae @c bd @d bc + bd f c bc f d
e e f e f e
= @c (gae bd ) @d (gae bc ) + bd gae f c bc gae f d
) Rabcd = @c abd @d abc (7.105)
so that
(re Rabcd )p = (@e Rabcd )p = (@e @c abd @e @d abc )p (7.106)
re Rabcd + rc Rabde + rd Rabec = 0 (7.107)
This is known as the Bianchi identity and using the antisymmetry relation
Rbacd = Rabcd
we have
a
Rcd = Racd = g ab Rabcd : (7.112)
178CHAPTER 7. THE EQUIVALENCE PRINCIPLE AND SPACETIME CURVATURE
This shows that raising the …rst index and then contracting the …rst and second
indices of the curvature tensor gives a zero tensor. However, rasing the …rst
index and then contracting the …rst and the last indices gives a none zero tensor
which is known as the Ricci tensor.
e
Rbcd = g ea Rabcd ) Rbc = Rbcd
d
= g da Rabcd (7.114)
or
e
Rbcd = g ea Rabcd ) Rab = Rabd
d
= g de Reabd (7.115)
The Ricci tensor is symmetric. We can prove this using the cyclic identity in
Eq. (7.100). Raising the index a
so that
d d
Rbcd + Rcdb = 0: (7.120)
Noting that
which shows that the Ricci tensor is symmetric. By raising the …rst indices on
both side of Eq. (7.121), we can write
A further contraction of the curvature tensor leads to the curvature scalar. From
Eq.(7.115), we have
d
Rab = Rabd = g de Reabd (7.123)
upon contracting this, the curvature scalar is given by
we may write
a a a a
Rabde = Rabed ) Rbde = Rbed ) Rbae = Rbea = Rbe (7.128)
R = Rbb
we can write
rb Rcb ab
rc R + ra Rbc = 0: (7.132)
Once more invoking the antisymmetric properties of the curvature tensor
we have
ba ab ab ab ab cd
Rcd = Rcd ; Rcd = Rdc ; Rcd = Rab (7.134)
ba ab
) ra Rcd = ra Rcd ; (7.135)
contracting b and d
ba ab ab ab
ra Rcb = ra Rcb ) ra Rcb = ra Rcb = ra Rca
180CHAPTER 7. THE EQUIVALENCE PRINCIPLE AND SPACETIME CURVATURE
where we contracted d. The indices are dummy indices and we can write
ab
ra Rbc = ra Rca = rb Rcb (7.136)
Noting that
b
rc = c rb (7.138)
we may write Eq. (7.137) as
b b
2rb Rcb c rb R = rb 2Rcb cR =0 (7.139)
which leads to
rb Gab = 0 (7.141)
where
1 cb
Gab = Rbc g R (7.142)
2
is known as the Einstein tensor. We have proved earlier that the Ricci tensor
is symmetric and we already know the metric tensor is symmetric. Therefore,
the Einstein tensor is a symmetric tensor.
parallel transported along this curve is when the intrinsic derivative of this
vector is zero
Dva dva b dxc
= ac vb = 0: (7.144)
Du du du
or in terms of its controvariant components
Dv a dv a a c dx
b
= + cb v =0 (7.145)
Du du du
Thus when a vector is parallel transported, we must have
dva b dxc
= ac v b (7.146)
du du
7.8. CURVATURE AND PARALLEL TRANSPORT 181
or
dv a a b dx
c
= bc v (7.147)
du du
Let’s consider a parallel transported vector along a closed curve C on a manifold.
Before we derive the condition for a parallel transport of a vector …eld on a
curved manifold, it is helpful if we refresh our memory about Stoke’s theorem
from Theoretical Physics I.
Stokes’ Theorem: Stokes theorem states that
I Z
~ d~r =
D r D ~ n^ dA:
Curve Surf ace A
bounding A
where the surface is any surface that has C, as its boundary. It could be an
a cyclic shown in the …gure or a hemisphere If the vector is constant (both its
r ~ =0
D
which leads to
I I
~ d~r = 0 ) D
D ~ d~r = 0
Curve Curve
bounding A bounding A
Now let’s say that this curve C is on some manifold instead of the Euclidean
space in which we know the Stoke’s theorem is valid. Also let’s consider a vector
…eld
~v = va (u) e^a (u) (7.148)
that is being parallel transported in this closed curve. When this vector is being
parallel transported in curve C, it is transported by dividing the area bounded
182CHAPTER 7. THE EQUIVALENCE PRINCIPLE AND SPACETIME CURVATURE
by the curve C into small areas each bounded by a small curve cN : Suppose
we assume there is a change v a in the components of the vector during the
transportation of the vector along this curve. This change can be expressed as
X
va = ( v a )N ;
N
dv a a b dx
c
= bc v (7.149)
du du
at all point on the curve. Suppose the transportation took from some point p
on cN to another point which we may describe by the coordinate xc (u) ; then
we may write
Z u a Z u c
dv a b dx
du = bc v du (7.150)
up du up du
which gives
Z u c
a a a b dx
v (u) = v (up ) bc v du: (7.151)
up du
Noting that we can make cN as small as we want, a series expansion about point
p allows us to make the approximations
a a a
bc (u) ' bc (up ) + (@d bc )up xd (u) xd (up ) (7.153)
v a (u) = v a (up )
Z u Z u
a
bc (up ) v b (up ) dxc + a
bc (up ) b
ef (up ) v e (up ) xf (u) xf (up ) dxc
up up
Z u
a
(@d bc )up xd (u) xd (up ) v b (up ) dxc (7.156)
up
7.8. CURVATURE AND PARALLEL TRANSPORT 183
v a (u) = v a (up )
Z u Z u
a
bc (up ) v b (up ) dxc + a
bc (up ) b
ef (up ) v e (up ) xf (u) xf (up ) dxc
p p
Z u
(@d abc )up b
v (up ) x (u)d
x (up ) dxc
d
(7.157)
p
I I I I
f c f c d c d
x (up ) dx = x (up ) dx = 0; x (up ) dx = x (up ) dxc = 0
cn cn cn cn
we …nd I
a 1 a
v = (R ) v b (up ) xc dxd : (7.166)
2 bcd p
cn
where
a a a a e a e
(Rbcd )p = (@c bd )up (@d bc )up + ec (up ) bd (up ) ed (up ) bc (up ) :
For a parallel transport of a vector in the tiny loop cN the components must
remain unchanged. This requires that for a vector …eld to be parallel transported
on cN we must have
I
a 1 a
v = (R ) v b (up ) xc dxd = 0 (7.167)
2 bcd p
cn
Then for a parallel transport of the vector …eld on a closed curve C on the
manifold, the components must not change over the entire area bounded by the
curve C so that
0 1
X X I
va = ( v a )N = @ 1 (Rbcd
a
)p v b (up ) xc dxd A = 0:
2
N N cn N
7.8. CURVATURE AND PARALLEL TRANSPORT 185
m0 N
0 = m0 n0 = ; (8.1)
l03
187
188 CHAPTER 8. THE GRAVITATIONAL FIELD EQUATIONS
where N is the number of particles in the cube of length l0 : In some other frame
S 0 , moving with speed v relative to S we recall that
r
v2 l0
l = l0 1 2
= ; (8.2)
c v
m0
m = q = v m0 ; (8.3)
v2
1 c2
2 m0 N 2
= mn = v = v 0: (8.4)
l03
So what would this suggest about how we represent the mass responsible for
the gravitational …eld? Well to answer this let’s go back to our discussion about
electromagnetism. In electromagnetism, we saw that the charge and current
densities transformed as
0
= u 0; J
~0 = u 0~
v: (8.5)
which is …rst-order in u : This has lead us to representing the charge and current
densities with 4-current density (a vector-1st rank tensor)
[J a ] = u 0 (c; ~v ) = a
u 0u : (8.6)
in terms of the 4-velocity ua . In the case of mass density as we can see from Eq.
(8.4), the mass density is second order in u : This suggests that we can make
a similar argument and express the mass density as tensor product of the four
velocity ua (x) so that we can at least sure that the 00 component will lead to
the right mass density transformation. Therefore, one can then de…ne a 2-nd
rank tensor (the energy-momentum tensor ) T (x) for the mass responsible for
the existence of the gravitational …eld as
where we denoted is the proper density of the ‡uid 0 by and ~u (x) is its
4-velocity.Then the covariant components of the energy-momentum tensor are
given by
T ab = ua ub : (8.8)
Using
ua = u (c; ~u) = u c; ui (8.9)
Components Interpretation
T 00 = u0 u0 = 2u c2 0 Energy density
ui Energy 1
T 0i = u0 ui = 2 i
u cu 0 = 2 2
uc 0 c = area:time e^i c the Energy ‡ux in
the i-direction 1c
M omentumi
T i0 = u0 ui = 2 i
u cu 0 = 2 i
u 0u c = vol me c the momentum
density in the
i-direction c
M omentumi uj M omentumi
T ij = ui uj = 2 i j
u 0u u = vol me c c = area:time e^j the rate of ‡ow
of the i-component
of momentumin per
unit area the jth di-
rection
We have stated that the random thermal motions of the particles will give rise
190 CHAPTER 8. THE GRAVITATIONAL FIELD EQUATIONS
Sol:
2 3 2 32 3
c2 0 0 0 1 0 0 0 p 0 0 0
6 0 p 0 0 7 6 0 1 0 0 7 6 0 p 0 0 7
T ab =6
4 0
7+6 76 7
0 p 0 5 4 0 0 1 0 54 0 0 p 0 5
0 0 0 p 0 0 0 1 0 0 0 p
(8.14)
2 32 3
1 0 0 0 p 0 0 0
6 0 1 0 0 7 6 0 p 0 0 7
6 76 7 (8.15)
4 0 0 1 0 54 0 0 p 0 5
0 0 0 1 0 0 0 p
2 3 2 3
c2 0 0 0 p 0 0 0
6 0 p 0 0 7 6 0 p 0 0 7
) T ab =6
4 0
7+6 7 (8.16)
0 p 0 5 4 0 0 p 0 5
0 0 0 p 0 0 0 p
2 32 3
1 0 0 0 p 0 0 0
6 0 1 0 0 7 6 0 p 0 0 7
6 76 7 (8.17)
4 0 0 1 0 54 0 0 p 0 5
0 0 0 1 0 0 0 p
2 p
3 2 3
c2 c:c 0 0 0 1 0 0 0
6 0 0 0 0 7 6 0 1 0 0 7
) T ab = 6
4
7 p6 7
0 0 0 0 5 4 0 0 1 0 5
0 0 0 0 0 0 0 1
There follows that
p
T ab = + ua ub pg ab : (8.18)
c2
8.3. CONSERVATION OF ENERGY AND MOMENTUM 191
T = uab (8.20)
@a J a = 0 (8.21)
@a T ab = 0 (8.22)
Instead of deriving this relation, we will prove that this is the right equation for
conservation of energy and momentum by reproducing the conservation laws for
a Newtonian perfect ‡uid. Substituting Eq. (8.13) into (8.22), we have
h p i
@a T ab = @a + 2 ua ub p ab = 0 (8.23)
c
Generally, the density, the pressure, and the velocity are all depend on the
coordinate xa ; thus
p p p
@a + ua ub + + (@a ua ) ub + + ua @a ub @a (p) ab
=0
c2 c2 c2
(8.24)
We recall that for the 4-velocity
ub ub = c2 (8.25)
so that
@a ub ub + ub (@a ub ) = @a ub ub = @a c2 = 0
) 2 @a ub ub = 0 ) @a ub ub = 0: (8.26)
so that for the …rst and second terms applying Eq. (8.25) and for the third term
using Eq. (8.26), one …nds
h p i h p i
@a + 2 ua ub ub = @a + 2 ua c2
c c
p a b p
+ 2 (@a u ) u ub = + 2 (@a ua ) c2
c c
p a b
+ 2 u @a u ub = 0 (8.28)
c
which leads to
h p i a 2 p
@a + u c + + 2 (@a ua ) c2 @a (p) ab
ub = 0
c2 c
) @a c2 + p ua + c2 + p @a (ua ) @a (p) ab
ub = 0
2 a ab
) @a c +p u @a (p) ub = 0 (8.29)
Noting that
ab
ub = ua (8.30)
we …nd
so that upon using the relativistic equation of continuity in Eq. (??), we …nd
1 p
(@a p) ua ub + + ua @a ub @a (p) ab
=0 (8.33)
c2 c2
which we may write as
p ua ub
+ @a ub ua = ab
@a p: (8.34)
c2 c2
would be approximated to be
@a ( ua ) = 0 ) @0 u0 + @i ui = 0
@ @ ( ux ) @ ( uy ) @ ( uz )
) ( c) + + + =0 (8.38)
c@t @x @y @z
which gives
@
+ r ( ~u) = 0 (8.39)
@t
which is the classical equation of continuity for a ‡uid. In low pressure limit the
relativistic equation of motion in Eq. (8.34) reduces to
ua ub
@a ub ua = ab
@a p: (8.40)
c2
u0 ub ui
@0 ub u0 = 0b
@0 p ) @i ui u0 = 0i
@0 p: (8.41)
c2 c
where we used
@c u0 c
@0 u0 = = 0; 00
=1 = 0:
c@t c c
Noting that in the classical limit, ui =c << 1;
0i 0i ui ui
= 1) = 1+ ' 1;
c c
so that
@
+ ~u r ~u = rp; (8.54)
@t
Eq. (8.54) is Euler’s classical equation of motion for a perfect ‡uid. Note that
the conservation of energy and momentum
@a T ab = 0 (8.55)
for an arbitray coordinates is expressed as
ra T ab = 0
Homework: derive the expression for
ra T ab = 0:
(b) Weak gravitational …eld -for weak …eld space-time is "slightly" curved, the
metric can be approximated as
gab ' ab + hab (8.57)
and static
@gab
@0 gab = =0 (8.58)
c@t
and …eld and the metric are related by
2
g00 = 1+ (8.59)
c2
Consider the weak gravitational and low velocity limit. In this limit we can
combine the results (a) to (c) and the …eld equation that relates the …eld to the
source (mass) as
r2 c2 = (4 G) c2 = (4 G) T00 (8.60)
2 8 G 2 8 G
) r2 = T00 ) r2 1+ 2 = T00 (8.61)
c2 c4 c c4
196 CHAPTER 8. THE GRAVITATIONAL FIELD EQUATIONS
and …nd the constants and . To this end, we recall the energy-momentum
conservation
ra T ab = 0
so that
ra Rab + Rg ab = 0: (8.72)
ab
In the previous chapter, we have shown that for the Einstein tensor, G ;
1 ab
ra Gab = ra Rab Rg = 0: (8.73)
2
Multiplying this equation by , we have
ra Rab Rg ab = 0 (8.74)
2
so that upon subtracting Eq. (8.74) from Eq. (8.72), we …nd
1 1
+ ra Rg ab = + Rra g ab + g ab ra R = 0: (8.75)
2 2
We have shown that (in one of the homework problem) that the covariant deriv-
ative of the metric is zero
ra g ab = 0:
Employing this result in Eq. (8.75), we …nd
1
+ g ab ra R = 0: (8.76)
2
1
Rab Rgab = Tab (8.77)
2
198 CHAPTER 8. THE GRAVITATIONAL FIELD EQUATIONS
Tab = 0 ) T = Taa = g a T a =0
Rab = 0: (8.85)
8.5. THE EINSTEIN EQUATIONS 199
The following tables show the …eld and the curvature equations resulting from
N = 2; 3 and 4 sapcetime dimension.
N Rab Rabcd
2 R00 ; R01 ; R11
3 R00 ; R01 ; R02 ; R10 ;
R11 ; R12 ; R20 ; R21 ; R22
4 R00 ; R01 ; R02 ; R03 ; R10 ; R11 ; R12 ;
R13 ; R20 ; R21 ; R22 ; R23 ; R30 ; R31 ; R32 ; R33
N # of Rab Independent R # of # of Rabcd # of Indp. Rabcd
Indp.
Rab
2 3 R00 ; R01 ; R11 3 1
3 9 R00 ; R01 ; R02 ; R11 ; R12 ; 6 6
R22
4 16 R00 ; R01 ; R02 ; R03 ; R11 ; 10 20
R12 ; R13 ; R22 ; R23 ; R33
Homework: …nd Rabcd and the total # of Rabcd
We recall that the Ricci tensor, Rab ; is obtained from the curvature tensor
by rasing the …rst index and then contracting the …rst and the last indices
e
Rbcd = g da Rabcd : (8.86)
As we can see from these tables for N = 2 or 3; in an empty part of space the
number of independent …eld equations (which are zero) is greater or equal to
the corresponding independent number of the curvature equations (i.e 3 > 1 for
2-D and 6 = 6 for 3-D) guarantee that the full curvature tensor must vanish for
2-D and 3-D spacetime. This means spacetime in 2-D and 3-D for an empty
part of space is ‡at. On the other hand for 4-D spacetime since the number
of independent …eld equations (10) is greater than the number of independent
curvature equations (20), spacetime is still curved for an empty part of space
even if it is devoid of matter and charge. Therefore, we can make the conclusion
that gravitational …eld exists only in dimension equal or greater than 4-D.
Question: so can we create a 2-D or 3D spacetime (i.e 1-D and 2-D space) in
a laboratory where the gravitational …eld is zero and objects become weightless?
Taking these into consideration, we can write the Einstein …eld equation as
T
R00 = T00 g00 : (8.90)
2
and Ricci tensor, Rab ; is obtained from the curvature tensor by rasing the …rst
index (Eq. (8.91)) and then contracting the …rst and the last indices, we have
c c c e c e c
Rab = Rabc = @b ac @c ab + ac eb ab ec ; (8.92)
so that
c c e c e c
R00 = @0 0c @c 00 + 0c e0 00 ec : (8.93)
Note that the connections are related to the metric by
a g ad
bc = (@b gcd + @c gdb @d gbc ) (8.94)
2
For the …rst term in Eq. (8.93) we …nd
c g cd g cd
@0 0c = (@0 gcd + @c gd0 @d g0c ) = (@c gd0 @d g0c ) ' 0;
2 2
where we have used the metric is static and the space is slightly curved in the
weak-…eld limit. For the same reason, for the last two terms of Eq. (8.93), we
can make the approximation
e c e c
0c e0 = 00 ec ' 0: (8.95)
Applying Eq. (8.100), we …nd for the …eld equation Eq. (8.90) becomes
1 ji T
@j @i h00 = T00 g00 : (8.101)
2 2
g00 ' 1
so that
1 ji T
@j @i h00 = T00 : (8.102)
2 2
We will consider the gravitational …eld is due to some perfect ‡uid for which
energy-momentum tensor is given by
p h p ua ub p i
Tab = + 2 ua ub p ab = + 2 c2 (8.103)
c c c2 c2 ab
For most classical matter distribution
p
<<
c2
so that
Tab = ua ub : (8.104)
There follows that
T00 = u0 u0 = c2 (8.105)
and also
T = Tbb = g ba Tab = ub ub = c2 (8.106)
Then the …eld equation becomes
ji
@j @i h00 ' c2 ) @12 + @22 + @32 h00 ' c2 (8.107)
2 2
) r h00 ' c (8.108)
2
g00 = 1+ = 00 + h00 = 1 + h00 (8.109)
c2
2
) h00 = 2 (8.110)
c
and
8 G
= :
c4
we …nally …nd that
2 8 G 2
r2 ' c ) r2 '4 G (8.111)
c2 c4
202 CHAPTER 8. THE GRAVITATIONAL FIELD EQUATIONS
where g is the gravitational …eld vector like the electric …eld vector. For a very
spherical surface of radius r enclosing the perfect ‡uid, we may write
I
~g d~a = g4 r2 (8.114)
S
so that
GM
g4 r2 = 4 GM ) g =
r2
Then the gravitational …eld vector can be expressed as
GM
~g = r^
r2
Newton’s law of gravitational force on an object of mass, m at a distance r from
the center of the perfect ‡uid can then be written as
GM m
F~G = m~g = r^
r2
Suppose the perfect ‡uid is our planet earth, ME and the object is you or me
on earth (r = RE ), you will …nd
GM m
~g = r^ = 9:8 2
r2 s
and the force
m
F~G = your (or my) mass 9:8 r^:
s2