Sie sind auf Seite 1von 57

Special Relativity & Relativistic Electrodynamics

Sourav Sur

Dept. of Physics & Astrophysics, University of Delhi, Delhi - 110 007, India
Lecture Notes for M.Sc. First Semester, University of Delhi.
Contents
1 Early conceptions: Galilean Relativity 3
1.1 Galilean Transformations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Galilean Principle of Relativity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2 Formulation of Spatial Relativity 6
2.1 The Postulates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2 Lorentz Transformations for a Boost along one axis . . . . . . . . . . . . . . . . . . . . . . . . 6
2.3 The invariant Line element and the Metric . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.4 Poinca`re Transformations for a Boost along one axis . . . . . . . . . . . . . . . . . . . . . . . . 10
2.5 Lorentz Transformations for a Boost in an arbitrary direction . . . . . . . . . . . . . . . . . . 11
2.6 Lorentz Transformations for arbitrary Boosts and Rotations . . . . . . . . . . . . . . . . . . . 12
2.7 Successive Lorentz Boosts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.8 Relativistic Velocity Addition Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.9 General Rule of Transformation of Coordinate Dierentials . . . . . . . . . . . . . . . . . . . . 14
3 Lightcone, World-lines, Intervals and the Proper Time 16
3.1 The Light-cone and the World-lines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.2 Time-like, Space-like and Null Intervals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.3 The Proper Time and Time Dilation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
4 Tensor Analysis 19
4.1 Tensors of various Rank . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4.1.1 Tensor of Rank 0 : Scalar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4.1.2 Tensor of Rank 1 : Vector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4.1.3 Tensors of Rank 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4.2 Tensor Algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
4.2.1 Linear Combination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
4.2.2 Multiplication: Direct Product, Contracted Product and Scalar Product . . . . . . . . 20
4.2.3 Index raising/lowering, Norm and Trace . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4.2.4 Relation between Contravariant and Covariant components of a Tensor . . . . . . . . . 21
4.3 (Anti-)Symmetry of Tensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.3.1 Symmetric Tensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.3.2 Antisymmetric Tensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.3.3 Symmetrization and Antisymmetrization of Tensors . . . . . . . . . . . . . . . . . . . . 22

Email: sourav@physics.du.ac.in, sourav.sur@gmail.com, sourav.sur@uleth.ca


1
CONTENTS
4.3.4 Salient features of Symmetric and Antisymmetric Tensors . . . . . . . . . . . . . . . . . 23
4.3.5 The Levi-Civita Tensor and the Generalized Kronecker Delta . . . . . . . . . . . . . . . 24
4.3.6 Tensor Duality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.4 Tensor Calculus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.4.1 Tensor Dierentiation: Grad, Div, Curl and DAlembertian operators . . . . . . . . . . 25
4.4.2 The Four Volume and Tensor Integration . . . . . . . . . . . . . . . . . . . . . . . . . . 26
5 Special Relativistic Particle Dynamics 27
5.1 Four Velocity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
5.2 Four Acceleration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
5.3 Four Momentum and Energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
5.4 Conservation of Energy and Momentum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
5.5 Four Force . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
6 Special Relativistic Electrodynamics 31
6.1 Four Current Density . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
6.2 Covariant Lorentz Force Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
6.3 Covariant formulation of Maxwells Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
6.4 Transformation of Electromagnetic Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
6.5 Electromagnetic Scalar invariants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
6.6 Potential formulation of Electrodynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
6.6.1 Electromagnetic Four Potential . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
6.6.2 Gauge Transformations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
6.6.3 Gauge conditions: Lorentz and Coulomb gauges . . . . . . . . . . . . . . . . . . . . . . 36
7 Wave Solutions in Electrodynamics the Retarded Potentials 38
7.1 Solutions for scalar and three-vector gauge potentials the Non-covariant approach . . . . . 38
7.1.1 Solution technique in the Lorentz gauge . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
7.1.2 Solution technique in the Coulomb gauge . . . . . . . . . . . . . . . . . . . . . . . . . . 39
7.1.3 Solutions of the Propagator equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
7.1.4 General Solutions for the Potentials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
7.2 Solutions for the four vector gauge potential the Covariant approach . . . . . . . . . . . . . 43
7.2.1 Potential expansion in terms of Greens function, and Solutions . . . . . . . . . . . . . 43
7.2.2 Covariant formulation of Retarded and Advarnced Greens functions . . . . . . . . . . 46
7.2.3 Retarded and Advanced Solutions for the Potential . . . . . . . . . . . . . . . . . . . . 47
8 Radiation by Moving Charges 48
8.1 Scalar and vector potentials due to a point charge the Lienard-Wiechart potentials . . . . . 48
8.2 Retarded Electric and Magnetic elds due to a moving charge . . . . . . . . . . . . . . . . . . 50
8.2.1 Time and space derivatives in the frames of the retarded source and the observer . . . 50
8.2.2 Electric and Magnetic elds from the Lienard-Wiechart potentials . . . . . . . . . . . . 51
8.2.3 Electromagnetic eld tensor and components from generic four potential expression . . 53
8.2.4 Acceleration-dependent parts of the elds and Electromagnetic Radiation . . . . . . . . 54
8.3 Power radiated by an accelerated charge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
8.3.1 Non-relativistic radiant power formula the Larmor result . . . . . . . . . . . . . . . 55
8.3.2 Relativistic radiant power formula the Lienard result . . . . . . . . . . . . . . . . . . 56
2
1 Early conceptions: Galilean Relativity
1.1 Galilean Transformations
Consider two systems of reference and

. is characterized by coordinates (x, y, z) and time t, whereas

is characterized by coordinates (x

, y

, z

) and time t

. Let the origins of the two systems coincide at t = t

= 0
and

is moving relative to with constant velocity v in an arbitrary direction (See Fig. 1).
x
z
x

v = constant
y

O
O

Figure 1: Galilean transformations for frame motion in an arbitrary direction, preserving the axes orientation, .
The Galilean transformations are a set of linear coordinate transformations given by
t

= t
x

= x v
x
t
y

= y v
y
t
z

= z v
z
t
or, in vector notation:
t

= t
x

= x vt
, (1.1)
v
x
, v
y
, v
z
being the components of v along x, y and z respectively.
1.2 Galilean Principle of Relativity
All laws of mechanics are invariant under the Galilean transformations.
Example: System of particles interacting via two-body central potentials
Equation of motion of the i
th
particle in the reference frame

:
m
i
dv

i
dt

j
V
ij

i
x

(1.2)
m
i
being the mass of the i
th
particle, and V
ij
is potential through which it interacts with the j
th
particle.
From Eqs. (1.1) it follows:
t

= t , x

i
x

j
= x
i
x
j
, (1.3)
so that
v

i
=
dx

i
dt

=
dx
i
dt
v = v
i
v =
dv

i
dt

=
dv
i
dt
. (1.4)
3
1 EARLY CONCEPTIONS: GALILEAN RELATIVITY
Now, to work out the relations between the coordinate dierentials in the two frames and

, we proceed as
follows
Consider a scalar function [ = (t, x, y, z) in , and = (t

, x

, y

, z

) in

]. Then
d =

t
dt +

x
dx +

y
dy +

z
dz =

t

dt

+

x

dx

+

y

dy

+

z

dz

. (1.5)
Or,

t
dt +

x
dx +

y
dy +

z
dz =

t

dt +

x

(dx v
x
dt) +

y

(dy v
y
dt) +

z

(dz v
z
dt) . (1.6)
Equating separately the coecients of dt, dx, dy and dz on both sides

=

x
,

y

=

y
,

z

=

z
, and

=

t
+ v
x

x
+ v
y

y
+ v
z

z
. (1.7)
Or, in vectorial form

= ,

t

=

t
+ v . (1.8)
Using Eqs. (1.3), (1.4) and the rst relation of Eqs. (1.8), we can write Eq. (1.2) as
m
i
dv
i
dt
=
i

j
V
ij
|x
i
x
j
| (1.9)
which is the equation of motion of the i
th
particle in the reference frame .
So, the Galilean transformation equations preserve the form of Newtonian equations of motion.
Counter-example: Homogeneous wave equation for a scalar eld
Let us again consider a scalar eld which satises the wave equation
_
1
c
2

2
t
2

2
_
= 0 (1.10)
in the reference frame

.
Then, under the Galilean transformations, using Eqs. (1.8), we have in the reference frame
_
1
c
2
_

2
t
2
+ 2v

t
+ (v ) (v )
_

2
_
= 0 , (1.11)
which shows that the form of the wave equation is not preserved under the Galilean transformations.
For sound waves, this isnt a problem, since sound waves are compressions and rarefactions in a given
medium (e.g. air). Therefore, the preferred reference frame

in which Eq. (1.10) is valid, is obviously the


frame in which the medium is at rest. The same can also hold for light waves, if they propagate through a
medium. However, the essential dierence between the sound and the light wave propagation is that: the frame
motion directly inuences the propagation of sound (by virtue of mechanical vibrations described by Newtonian
classical mechanics), whereas for light, the intervening medium, if exists, has no such manifestation, other than
to support the wave propagation.
Now, the light wave propagation could be shown to be due to transverse electric and magnetic elds,
whose dynamics is governed by the Maxwells equations. So there seemed to be an incompatibility between
the Maxwells equations and Galilean relativity, in the sense that the Galilean transformations fail to preserve
the form of the Maxwells equations (or, the wave equations corresponding to every component of the electric
and the magnetic eld vectors) which involve a xed parameter c, that can be interpreted as a xed speed of
propagation c of light signals.
4
1.2 Galilean Principle of Relativity
Possible resolutions
I. Maxwells theory is incorrect and a proper theory of electromagnetism must have its equations invariant
under Galilean transformations.
II. Galilean relativity applies to the laws of Newtonian mechanics, and the Maxwells theory is correct but
bizarre, in the sense that the electromagnetic wave propagation always requires a preferred frame, which
is the one where the intervening medium the so-called luminiferous ther is at rest.
III. There should exist a relativity principle, other than the Galilean one, for both the laws of mechanics and
the equations of electrodynamics.
The rst alternative was hardly viable, given the immense success of the Maxwells theory of electromagnetism,
otherwise.
The no-go establishment of the ther-hypothesis (following the works of Michaelson and Morley in 1886)
ruled out the second alternative as well.
So, what remained is to nd out a set of transformation relations so that the Maxwells equations are of the
same form, with the same speed parameter c, in all frames moving with uniform relative speeds (i.e. inertial
frames).
The rst step towards obtaining such transformation relations, and hence to formulate a new theory of
relativity, was to abolish the Newtonian concept of absolute space, and to keep the time and the space in
equal footing. That is to say, one should now invoke the concept of space-time (instead of treating space
and time separately) represented by all the spatial coordinates, as well as time multiplied by a constant speed
(from dimensional arguments). Such a speed parameter must be a universal constant, which is argued to be
the speed of light c in vacuum, as we see below.
5
2 FORMULATION OF SPATIAL RELATIVITY
2 Formulation of Spatial Relativity
2.1 The Postulates
I. All laws of physics (except that for gravitation) are independent of uniform translational motion of the
system, i.e., the reference frame (RF), where they operate Principle of Relativity:
Physical Laws
Inv.

(RF)
v=const.
=

(RF)
_
. (2.1)
Corollary: The mathematical equations expressing the laws of nature must be covariant, i.e., invariant in form
under the transformations that leave invariant the innitesimal separation between two events in space-time.
II. The speed of light (c) in empty space is nite and independent of the motion of the source:
c
Inv.

(RF)
v=const.
=

(RF)
_
. (2.2)
Corollary: For physical entities in every inertial frame, there is a nite limiting speed, equal to c (as found
experimentally):
v
lim
= c = const. . (2.3)
The most general set of coordinate transformations which preserve the innitesimal separation between two
space-time events, side-by-side maintaining the constancy of the speed of light, are the Poinca`re transfor-
mations. A sub-class of these are the so-called (homogeneous) Lorentz transformations which preserve the
separation between two events which are not necessarily very close to each other, i.e., the separation is not nec-
essarily innitesimal. We discuss below in detail the form and properties of both the Lorentz and the Poinca`re
transformation equations in certain simplied, as well as generic, scenarios.
2.2 Lorentz Transformations for a Boost along one axis
Consider two reference frames and

represented by (ct, x, y, x) and (ct

, x

, y

, z

) respectively, as shown in
Fig. 2. Let us suppose that
I. The axes of the two frames are parallely oriented, i.e, ct

||ct, x

||x, y

||y, z

||z.
II. Frame

moves with constant velocity v relative to the frame along, say, the positive x- (x-) axis.
III. Origins of the two frames coincide at t = t

= 0.
x
O
y y

v = constant
z

Figure 2: Lorentz transformations for frame motion, preserving the axes orientation, and along one axis.
Consider now an event P, whose coordinates are (ct

, x

, y

, z

) in frame

and (ct, x, y, z) in frame . As


the origin of

moves relative to with speed v along positive x-direction


x

= 0 = x = vt . (2.4)
Conversely, as the origin of moves relative to

with speed v along the negative x

-direction
x = 0 = x

= vt . (2.5)
6
2.2 Lorentz Transformations for a Boost along one axis
Now, the transformation equations between the two systems must be linear, so that for an event in there is
a unique associated event in

, and vice versa. The most general relations satisfying Eqs. (2.4) and (2.5) are
x

= (x vt) , x =
_
x

+ vt

_
[ = dimensionless constant] . (2.6)
Or,
dx

= (dx vdt) , dx =
_
dx

+ vdt

_
. (2.7)
Therefore,
dx

dt

=
_
dx
dt

v
dt
dt

_
=
_
dx
dt
v
_
dt
dt

,
dx
dt
=
_
dx

dt
+ v
dt

dt
_
=
_
dx

dt

+ v
_
dt

dt
. (2.8)
In the limit dx

/dt

c and dx/dt c, Eqs. (2.8) reduce to


dt
dt

=
c
(c v)
,
dt

dt
=
c
(c + v)
. (2.9)
Solving for , subject to the requirement that the transformation equations we seek must reduce to the Galilean
transformation equations in the limit v/c 0, we get
=
_
1
2
_
1/2
, where =
v
c
. (2.10)
One can now write down the full set of transformation equations by solving Eqs.(2.6) for x

and t

, and noting
that there is no frame motion in y- or z-directions in the simplied set-up we are considering presently:
(i) ct

= (ct x)
(ii) x

= (ct + x)
(iii) y

= y
(iv) z

= z
(2.11)
These are Lorentz transformation (LT) equations in the simplest possible scenario where there is a frame
motion, or Lorentz boost, along only one spatial direction. Denoting the coordinates as:
x
0
ct , x
1
x , x
2
y , and x
3
z (2.12)
we write the above equations (2.11) in matrix form as
_
_
_
_
x
0
x
1
x
2
x
3
_
_
_
_
=
_
_
_
_
0 0
0 0
0 0 1 0
0 0 0 1
_
_
_
_
_
_
_
_
x
0
x
1
x
2
x
3
_
_
_
_
. (2.13)
Or, in compact form
x

, [, = 0, 1, 2, 3] , (2.14)
where x

and

are the elements of


x {x

} =
_
_
_
_
x
0
x
1
x
2
x
3
_
_
_
_
: 1 4 Position (column) vector in (3 + 1)-dimensional spacetime , (2.15)
{

} =
_
_
_
_
0 0
0 0
0 0 1 0
0 0 0 1
_
_
_
_
:
4 4 (homogeneous) LT matrix in a (3 + 1)-dimensional
space-time for single Lorentz boost along positive x
1
-axis;
denotes the columns and denotes the rows.
(2.16)
In the above Eq. (2.14), Einsteins summation convention is implied, i.e., repeated indices are summed over:

=0

[ = 0, 1, 2, 3] . (2.17)
7
2 FORMULATION OF SPATIAL RELATIVITY
The inverse transformation equations are given by
x

, [, = 0, 1, 2, 3] , (2.18)
where x

and

are the elements of


x
1
{x

} = (x
0
, x
1
, x
2
, x
3
) : 4 1 Position (row) vector in (3 + 1)-dimensional spacetime , (2.19)

_
=
_
[

]
1
_
: 4 4 Inverse LT matrix for a boost along negative x
1
axis . (2.20)
The distinction between x
0
and x
0
, x
1
and x
1
, and so forth, will be explained later on, when we will make a general
study of tensors and their components. For the time being, however, lets stick to the matrix terminology and focus on
the properties of the LT matrix

, where denotes columns and denotes rows.


Following are the two two important properties of the LT matrix for a single boost along one of the axes:
(i)
T
= =
T

1
=
T

1
= I =

, and (ii) det = 1 . (2.21)


To observe physically the meaning of a boost, let us make the following parametrization:
=
v
c
= tanh , =
_
1
2
_
1/2
= cosh , (2.22)
where is a dimensionless constant, called the Boost or Rapidity parameter. Under the above parameteri-
zation, the LT equations (2.11) take the form:
x
0
= x
0
cosh x
1
sinh ,
x
1
= x
0
sinh + x
1
cosh ,
x
2
= x
2
, and x
3
= x
3
, (2.23)
i.e., the LT matrix (due to a boost along the positive x
1
-direction with constant speed v) takes the form:
{

} =
_
_
_
_
cosh sinh 0 0
sinh cosh 0 0
0 0 1 0
0 0 0 1
_
_
_
_
. (2.24)
Under a further transformation, and reparameterization:
x
0
x
0
= ix
0
_
i.e., t

t = it
_
; x
k
x
k
= x
k
[k = 1, 2, 3] , = i , (2.25)
the above LT matrix (2.24) reduces to
{

} =
_
_
_
_
cos sin 0 0
sin cos 0 0
0 0 1 0
0 0 0 1
_
_
_
_
= {R

} , (2.26)
where R

are the elements of the matrix corresponding to the ordinary (orthonormal) rotation of the x
0
x
1
plane in the (3 + 1)-dimensional space-time (see Fig. 3):

x
0
= ix
0
x
1
= x
1
x
1
= x
1
x
0
= ix
0
O
Figure 3: Simple Lorentz transformation as (orthonormal) rotation of the ix
0
x
1
plane.
8
2.3 The invariant Line element and the Metric
A Lorentz Boost along positive x
1
with speed v may thus be viewed as rotation of the complex ix
0
x
1
plane
by a real angle . In principle, there could be three Lorentz boosts along each of the three spatial axes. The
transformations (2.25) imply the Euclidization of space-time (for clarication, refer to the next subsection).
2.3 The invariant Line element and the Metric
Refer back to the LT equations (2.11) for a boost along the positive x-direction with constant speed v. In
dierential form, they are given by
(i) cdt

= (cdt dx)
(ii) dx

= (cdt + dx)
(iii) dy

= dy
(iv) dz

= dz
(2.27)
Now,
Eq. (2.27 i) + Eq. (2.27 ii) = cdt

+ dx

= (1 ) (cdt + dx) ,
Eq. (2.27 i) Eq. (2.27 ii) = cdt

dx

= (1 + ) (cdt dx) .
Multiplying these two equations, and using the expressions (2.10) for and , we get
c
2
dt
2
dx
2
= c
2
dt
2
dx
2
.
Using further, the Eqs. (2.27 iii & iv):
c
2
dt
2
dx
2
dy
2
dz
2
= c
2
dt
2
dx
2
dy
2
dz
2
, (2.28)
i.e., the Lorentz transformations leave invariant the dierential quantity
ds
2
= c
2
dt
2
dx
2
dy
2
dz
2

_
dx
0
_
2

_
dx
1
_
2

_
dx
2
_
2

_
dx
3
_
2
. (2.29)
Performing a similar operation as above on the set of LT equations (2.11) themselves, we could have also found
that the Lorentz transformations leave invariant the algebraic quantity
s
2
= c
2
t
2
x
2
y
2
z
2

_
x
0
_
2

_
x
1
_
2

_
x
2
_
2

_
x
3
_
2
. (2.30)
As in the ordinary Euclidean (three-)space, the elementary distance between two points is given by
|dx| =
_
x
2
+ y
2
+ z
2

_
(x
1
)
2
+ (x
2
)
2
+ (x
3
)
2
, (2.31)
the quantity ds
2
given by Eq. (2.29) is analogically referred to as the Line element, or the (squared) elemen-
tary separation between two events in a (3 + 1)-dimensional space-time. Galilean transformations leave |dx|
2
invariant, whereas Lorentz transformations leave ds
2
invariant.
Navely, the relative minus sign between the temporal and spatial parts of ds
2
= c
2
dt
2
|dx|
2
may be looked
upon as follows: For signals travelling with the speed of light c. Galilean relativity alludes to instantaneous
reception after transmission, i.e., t = 0, which implies
lim
t0
|x|
t
=
|dx|
dt
= c .
However, for nite c (as according to Maxwells theory), the reception and the transmission should be separated
in time by t = 0. So, one imposes (instead of dt = 0)
ds
2
= c
2
dt
2
|dx|
2
= 0 , =
|dx|
dt
= c ,
i.e., the invariant elementary separation between events is so chosen that it automatically gives nite speed of
light under a certain stipulation

This would be claried further when we will generally discuss the types of separation between events connected by signals
travelling with arbitrary speeds in subsequent subsections.
9
2 FORMULATION OF SPATIAL RELATIVITY
Now, the line element ds
2
being quadratic in the dierentials of the coordinates (dx

; = 0, 1, 2, 3), one may


express it (as any other quadratic):
ds
2
= g

dx

dx

, (2.32)
where the coecients g

(, = 0, 1, 2, 3) form the elements of the so-called (covariant) Metric tensor of rank


2, or simply the metric. We will give a detailed account of the denition and properties of tensors of various
types later on. However, for the time being, consider g

forming an object which could in general be a function


of the coordinates, and is symmetric under the interchange of its indices, i.e., g

= g

, so that there are at


most 4 (4 + 1) /2 = 10 independent components of the in 3 + 1 dimensions:
ds
2
= g
00
_
dx
0
_
2

ONE term for


= = 0.

+ 2g
0i
dx
0
dx
i

Sum of THREE terms for


= 0, = i.The factor of
2 is for symmetry g
0i
= g
i0
.

+ g
ij
dx
i
dx
j

Sum of SIX terms for


= i, = j, due to
symmetry g
ij
= g
ji
.

. (2.33)
In the above (and throughout) we use the following notations:
Greek indices (, , , . . . ): run from 0 to 3, i.e., over the entire (3 + 1)-dimensional space-time,
Latin indices (i, j, k, . . . ): run from 1 to 3, i.e., over the 3-dimensional spatial hypersurface.
Special Relativity (SR) alludes to a special form the metric tensor: g

, referred to as the Minkowski


metric, which is diagonal with each (diagonal) element normalized to unity in magnitude:

= diag (1, 1, 1, 1) i.e., |

| =

. (2.34)
Or, explicitly:

00
= 1 ,
0i
= 0 ( i = 1, 2, 3) ,
ij
=
ij
( i, j = 1, 2, 3) . (2.35)
The line element in Special Relativity is thus
ds
2
SR
=

dx

dx

=
_
dx
0
_
2

_
dx
1
_
2

_
dx
2
_
2

_
dx
3
_
2
. (2.36)
2.4 Poinca`re Transformations for a Boost along one axis
The necessary condition for the constancy of the speed of light c, in a relativity theory governed by certain
transformation equations, is that the line element ds
2
must remain invariant under such transformations.
Lorentz transformations (LT) preserve not only ds
2
, but s
2
[Eq. (2.30)] as well, which is not necessarily
required. The invariance of s
2
is due to the third assumption made in the derivation of the LT equations in
subsection 2.2, viz., the frames of reference and

coincide at t = t

= 0. If we relax this assumption,


then the most general linear transformations that leave ds
2
is invariant are the Poinca`re transformations (PT),
which in the simple case of a single Lorentz boost along (say) the x-axis, with constant speed v, are given by
the equations:
(i) ct

= (ct x) + a
t
(ii) x

= (ct + x) + a
x
(iii) y

= y + a
y
(iv) z

= z + a
z
(2.37)
where (a
t
, a
x
, a
y
, a
z
) are constant shifts between the origins of and

at t = 0, along t, x, y and z respectively.


Denoting
a
0
a
t
, a
1
a
x
, a
z
a
y
, a
3
a
z
(2.38)
the above equations are expressed in compact form as
x

+ a

. (2.39)
10
2.5 Lorentz Transformations for a Boost in an arbitrary direction
2.5 Lorentz Transformations for a Boost in an arbitrary direction
Consider again two reference frames and

characterized by coordinates
_
x
0
, x
1
, x
2
, x
3
_
and
_
x
0
, x
1
, x
2
, x
3
_
respectively.
I. The axes of the two frames are parallely oriented, i.e, x
0
||x
0
, x
1
||x
1
, x
2
||x
2
, x
3
||x
3
.
II. Origins of the two frames coincide at x
0
= x
0
= 0.
III. Frame

moves with constant velocity v relative to the frame in an arbitrary direction.


x
2
x
1
x
3
x
2
x
1
x
3
O

O
v = const.

Figure 4: Lorentz transformation for frame motion, preserving the axes orientation, but in an arbitrary direction.
As such, the generic Lorentz boost with velocity v may be thought of as the resultant of three boosts
_
v
1
, v
2
, v
3
_
along
_
x
1
, x
2
, x
3
_
respectively, and in order to derive this generic boost transformation matrix
elements

(, = 0, 1, 2, 3), we resort to the following two equations:


I. Lorentz transformation equations (in dierential form):
dx

dx

0
dx
0
+

i
dx
i
. (2.40)
II. Invariance of the metric under Lorentz transformations:
ds
2
= ds
2
=

dx

dx

dx

dx

dx

_
_

dx

_
[by Eqs. (2.40)]
=

(2.41)
Let us now suppose a particle to be at rest w.r.to the frame , i.e.,
dx
i
= 0 (i = 1, 2, 3) .
Then, since the origin of the frame

moves relative to the frame with constant velocity v, the particle is


seen to move with velocity v in the frame

, i.e.,
dx
i
dt

= v
i
(i = 1, 2, 3)
From Eqs. (2.40) we get
dx
0
=
0
0
dx
0
, dx
i
=
i
0
dx
0
(i = 1, 2, 3) , (2.42)
=
i
0
=
i

0
0
,
_
where
i
=
v
i
c
=
1
c
dx
i
dt

=
dx
i
dx
0
_
. (2.43)
11
2 FORMULATION OF SPATIAL RELATIVITY
Moreover, by setting = = 0 in Eq. (2.41), we have

00
=

0
=
00
_

0
0
_
2
+
ij

i
0

j
0
=
_

0
0
_
2

i
0

j
0

ij
= 1 . (2.44)
Solving Eqs. (2.42) and (2.44) simultaneously, we obtain:

0
0
= ,
i
0
=
i
,
_
where =
_
1
ij

j
_
1/2
=
_
1 |

|
2
_
1/2
_
. (2.45)
The other elements of the LT matrix

are not uniquely determined. However, one convenient choice is

0
i
=
ij

j
,
i
j
=
i
j
+
( 1)
|

|
2

jk

k
. (2.46)
We leave it to the reader to verify that these indeed satisfy Eq. (2.41). It is also easy to see that in the ordinary
three-vectorial notation, the above elements of

imply that the LT equations are given by


x
0
=
_
x
0

x
_
, x

= x +
( 1)
|

|
2
_

x
_



x
0
, (2.47)
with

=
_

1
,
2
,
3
_
=
v
c
=
_
v
1
c
,
v
2
c
,
v
3
c
_
, =
_
1 |

|
2
_
1/2
. (2.48)
So the LT involves three boosts, along all three spatial axes
_
x
1
, x
2
, x
3
_
with speeds
_
v
1
, v
2
, v
3
_
respectively.
2.6 Lorentz Transformations for arbitrary Boosts and Rotations
If, in addition to the scenario in previous subsection, the axes of the two frames ,

are not taken be parallely


oriented (as in Fig. 5), then the corresponding LT matrix is in its most general form, which involves not only
the three boosts, but three spatial rotations as well.
x
2
x
1
x
3
x
2
x
1
x
3
O

O
v = const.

Figure 5: Lorentz transformation for frame motion, not preserving the axes orientation, and in an arbitrary direction.
Even for the most general LT matrix, the rst of the properties (2.21), viz.,

1
=
T

1
= I

, (2.49)
can be shown to hold. As to the second property, taking determinant of both sides of Eq.(2.41), we have
det (det )
2
= det det = 1 . (2.50)
12
2.7 Successive Lorentz Boosts
However, our interest is in the so-called proper Lorenetz transformations for which

0
0
0 , det = 1 (2.51)
and which can always be converted to the identitiy transformation by continuous variation of parameters of

. In fact, the full LT could be decomposed into: an ordinary spatial rotation, followed by a boost, followed
by a further ordinary rotation. The rst rotation lines up one of the spatial axes, say the x
1
-axis (see Fig. 5),
of with the velocity v of

. Then a boost in this direction with speed v = |v| transforms to a frame which
is at rest relative to

. Finally, another rotation lines up the coordinate frame with that of

.
2.7 Successive Lorentz Boosts
Two frames of reference and

, with coordinates
_
x
0
, x
1
, x
2
, x
3
_
and
_
x
0
, x
1
, x
2
, x
3
_
, have their axes
parallely oriented and origins coincident at x
0
= x
0
= 0. Frame

moves with constant velocity v in an


arbitrary direction, relative to the frame .
x
2
x
3
x
2
x
1
x
3
P
v
v

x
1
O
O

Figure 6: Successive Lorentz Boosts


Consider now a particle P which moves with constant velocity v

relative to

. Let v

be the particles velocity


relative to , and

be the particles rest frame, i.e, the coordinate system


_
x
0
, x
1
, x
2
, x
3
_
attached to the
particle. Then
I. For the relative motion between

and :
x

) x

,
_

=
v
c
_
. (2.52)
where

(
k
) are the LT matrix elements for boosts with speeds v
k
= c
k
.
II. For the relative motion between

and

:
x

) x

,
_

=
v

c
_
. (2.53)
where

(
k
) are the LT matrix elements for boosts with speeds v
k
= c
k
.
III. For the relative motion between

and :
x

) x

,
_

=
v

c
_
. (2.54)
where

(
k
) are the LT matrix elements for boosts with speeds v
k
= c
k
.
Substituting for x

from Eq. (2.52) in Eq. (2.53), and comparing with Eq. (2.54), we have
x

) x

) x

(
k
) =

(
k
)

(
k
) . (2.55)
13
2 FORMULATION OF SPATIAL RELATIVITY
Thus a Lorentz boost along the direction of

= v

/c is the product of the Lorentz boosts along


= v

/c
and

= v/c. One may verify that the product of the boosts is non-commutative:

(
k
)

(
k
) =

(
k
)

(
k
) . (2.56)
2.8 Relativistic Velocity Addition Theorem
Refer back to the situation in the previous subsection. If the relative frame velocity v and the particles velocity
v

w.r.to the frame

are known, then what would be the particles velocity v

w.r.to the frame ?


To determine this, let us re-write Eq. (2.55) as

(
k
) =

0
(
k
)
0

(
k
) +

i
(
k
)
i

(
k
) , (2.57)
and recall from 2.5, that the transformation matrix elements

for a generic Lorentz boost, in an arbitrary


direction, are given by

0
0
= ,
i
0
=
i
,
0
i
=
ik

k
,
i
j
=
i
j
+
( 1)
|

|
2

jk

i

k
, (2.58)
where =
_
1
ik

k
_
1/2
=
_
1 |

|
2
_
1/2
.
Using these, the (0 - 0) and (i - 0) components of the above Eq. (2.57) give

0
0
(
k
) =
0
0
(
k
)
0
0
(
k
) +
0
i
(
k
)
i
0
(
k
) =

_
1 +
ik

k
_
, (2.59)

i
0
(
k
) =
i
0
(
k
)
0
0
(
k
) +
i
j
(
k
)
j
0
(
k
)
=
i
=

_
_

i
+
i
_

_
1
1

_
_


jk

k
|

|
2

i
__
. (2.60)
One may however note that:

il

l
_


jk

k
|

|
2

i
_
=

__

_
|

|
2
= 0 . (2.61)
Since
il

l
is arbitrary and not necessarily zero, we have


jk

k
|

|
2

i
= 0 . (2.62)
Substituting this in Eq. (2.60), and using Eq. (2.59), we nally get

i
=

i
+
i
1 +
jk

k
= v

=
v

+ v
1 + (v

v/c
2
)
. (2.63)
This is the relativistic velocity addition formula.
2.9 General Rule of Transformation of Coordinate Dierentials
Consider again a reference frame (RF) with
Coordinates: x
0
ct, x
1
x, x
2
y, x
3
z.
(3 + 1)-D position vector: x

=
_
x
0
, x
i
_
= (ct, x).
14
2.9 General Rule of Transformation of Coordinate Dierentials
3-D (spatial) position vector: x
i
= x.
Let there be another RF

with
Coordinates: x
0
ct

, x
1
x

, x
2
y

, x
3
z

.
(3 + 1)-D position vector: x

=
_
x
0
, x
i
_
= (ct

, x

).
3-D (spatial) position vector: x
i
= x

.
Let the coordinates in

be related to those in by a coordinate transformation:


x

= x

_
x
0
, x
1
, x
2
, x
3
_
, = 0, 1, 2, 3 . (2.64)
Then the necessary and sucient condition that the four components of x

form a set of independent, real


functions of
_
x
0
, x
1
, x
2
, x
3
_
is that their Jacobian:
J(x x

) =

x
0
x
0
. . .
x
3
x
0
.
.
.
.
.
.
x
0
x
3
. . .
x
3
x
3

= 0 . (2.65)
In such case, one can have the inverse coordinate transformations as:
x

= x

_
x
0
, x
1
, x
2
, x
3
_
, = 0, 1, 2, 3 . (2.66)
At any point in a general (curved, Riemannian) spacetime, the direction (of a four-vector) in a coordinate
frame, say , is determined by the coordinate dierentials dx

dened in that frame. In another frame, say

,
the direction is determined by the associated coordinate dierentials dx

. The relation between the coordinate


dierentials in the two frames is given by the transformation rules:
dx

=
x

dx

, and dx

=
x

dx

. (2.67)
While carrying out such a transformation, we have
x

, and
x

. (2.68)
So, if one considers x

/x

as a matrix the transformation matrix, for transformation from the x

coordinate system to the x

coordinate system then x

/x

is its inverse, i.e., the transformation matrix


for the inverse transformation from the x

coordinate system to the x

coordinate system. Consequently, the


Jocobians of the transformation and the inverse transformation satisfy:
J(x x

)J(x

x) =

x
x

= 1 . (2.69)
In Special Relativity (SR), the transformation matrices are the LT matrices, and therefore the Jacobian is
identically equal to unity:
x

SR
=

, J(x x

) =

SR
= det (

) = 1 . (2.70)
15
3 LIGHTCONE, WORLD-LINES, INTERVALS AND THE PROPER TIME
3 Lightcone, World-lines, Intervals and the Proper Time
3.1 The Light-cone and the World-lines
One of the intriguing concepts of relativity is that of a Light-cone, described as follows:
Consider rst, for simplicity, the (1 + 1)-D spacetime diagram where one axis is of course x
0
= ct and
the other axis is one of the space axes, say x
1
= x, as shown in Fig. 7. The remaining axes x
2
= y and
x
3
= z have been suppressed.
At time t = 0, a particle (or a system of particles) is supposed to be at the origin O.
Elsewhere
Future
B
O
A
x
0
= ct
x
1
= x
Elsewhere
Past
Figure 7: (1 + 1)-D version of the Light-cone that is contained in a (1 + 1)-D spacetime with the axes: x
0
= ct and one
of the space axes, say x
1
= x (the other axes are suppressed). Essentially, the light-cones include the regions above and
below the x-axis and within the angles formed by the lines that make 45
o
with the axes.
Since, according to the Special Relativistic second law, the speed of light c in vacuum is the upper limit
of the magnitude of speeds of all material particles or systems, in the (1+1) spacetime diagram a special
status is given to the lines which make 45
o
angles with the axes, i.e., whose equations are given by:
x
1
= x
0
implying |dx/dt| = c. These lines are called light-lines, as they denote light signals traveling
out at 45
o
after being emitted from the origin O.
The light-lines divide the entire (1 + 1)-D spacetime into three regions:
(I) The (1+1)-D version of the upper and lower light-cones (or, rather light-triangles which are the regions
above and below the x-axis and within the angles formed by the light-lines.
(II) The light-lines themselves, which make 45
o
angles with the axes, and which form the edges of the
light-triangles.
(III) The regions on the left and right outside the light-triangles, called elsewhere.
The (1+1)-D light-cones (or, the light-triangles) are so named not only because their edges are formed by
the light-lines, but also due to the fact that the spacetime evolution trajectories of all material particles
or systems, which have positive denite and nite net mass squared, and whose speeds cannot exceed
the speed of light c (by the Special Relativistic postulate) are conned within them. More specically,
since all material particles have the magnitude of speed |v| = |dx/dt| < c, as time proceeds the events
associated with a material particle or a system would trace out a path, say OB, as shown in Fig. 7, inside
16
3.2 Time-like, Space-like and Null Intervals
the upper half (1 + 1) D cone (or, triangle), called the future cone (t > 0). The locus of all points
on the curve OB precisely satisfy |x| < ct which implies |dx/dt| < c. Similarly, the history of all events
associated with a particle or a system, upto the chosen origin O, could be depicted by another curve, say
AO, in the spacetime diagram in Fig. 7. As the locus of all points on OA which satisfy |x| < ct, for
t < 0 inside the lower half (1 + 1) D cone (or, triangle), the latter is called the past cone.
The entire line AOB is called the particles or the systems World line. In general the world line could be
curved, as in Fig. 7, implying non-uniform evolution of the particle or system in space and time. Uniform
motion with constant speed is denoted by a straight world line. If further, the straight line is parallel to
the x
0
= ct axis, then that implies the particle or the system at rest.
The region outside the (1 + 1)-D light-cones (or, the light-triangles) is called elsewhere. No material
particle or system can have an associated event taking place there.
Future
B
A
O
x
0
= ct
x
1
= x
Past
Figure 8: The Light-cone in a (2 + 1)-D spacetime with the axes: x
0
= ct and two of the space axes, say x
1
= x and
x
2
= y (not shown), the third space axis is suppressed. The edges of the lightcone are formed by the light-surfaces that
make 45
o
with the axes.
In (2 + 1)-D, the light-triangle becomes a (2 + 1)-D light-cone, which is an ordinary three dimensional
double cone, as shown in Fig. 8. The top and base of the light-cone are the circles with equation
(x
1
)
2
+ (x
2
)
2
= (x
0
)
2
, or x
2
+ y
2
= c
2
t
2
. The edges of the light-cone are now the 2-D surfaces, called
light-surfaces, which make 45
o
with the axes. All the above arguments for the (1 + 1)-D line-cone hold
for (2 + 1)-D light-cone as well.
The full (3+1)-D spacetime diagram contains a (3+1)-D light cone, which is a higher double-cone whose
top and base are spheres (x
1
)
2
+(x
2
)
2
+(x
3
)
2
= (x
0
)
2
, or x
2
+y
2
+z
2
= c
2
t
2
. The edges of the light-cone
are 3-D hypersurfaces, which make 45
o
with the axes.
3.2 Time-like, Space-like and Null Intervals
Consider now the separation s
AB
between two events P
A
(t
A
, x
A
) and P
B
(t
B
, x
B
) in spacetime:
s
2
AB
= c
2
(t
A
t
B
)
2
|x
A
x
B
|
2
. (3.1)
For any two such events, there are three possibilities:
1. s
2
AB
> 0 : In which case, the events are said to have a time-like separation.
17
3 LIGHTCONE, WORLD-LINES, INTERVALS AND THE PROPER TIME
Since s
2
AB
is invariant under Lorentz transformation (LT), it is always possible to nd a reference
frame (RF)

in which x

A
= x

B
, whence s
2
AB
= c
2
(t

A
t

B
)
2
> 0 (guaranteed).
In the new frame

, the events occur at the same position, but are separated in time. For e.g., one
event may be at the origin, and the other in the past or future region of the light-cone.
Accordingly, the particles which have their event-trajectories inside the light-cone, as for instance
the line AOB in Fig. 5 or 6, are called time-like particles. All material particles moving with speed
v < c are time-like.
2. s
2
AB
< 0 : In which case, the events are said to have a space-like separation.
It is always possible to nd a RF

in which t

A
= t

B
, whence s
2
AB
= |x

A
x

B
|
2
< 0 (guaranteed).
In the new frame

, the events occur at dierent space points at the same instant of time. For e.g.,
one event may be at the origin, and the other in the elsewhere region (outside the light-cone).
Accordingly, the particles which have their event-trajectories outside the light-cone are called space-
like particles. No material particle could be space-like, only some hypothetical particles moving with
speed v > c are space-like.
3. s
2
AB
= 0 : In which case, the events are said to have a light-like or null separation.
The events occur on the light-cone and can be connected only by light signals.
The corresponding particles are the null particles or light particles (photons).
3.3 The Proper Time and Time Dilation
Consider a particle moving with instantaneous velocity v(t) relative to some inertial frame :
In time dt, the change in the position: dx = v(t) dt.
Corresponding innitesimal invariant interval:
ds
2
= c
2
dt
2
|dx|
2
= c
2
dt
2
_
1

(t)

2
_
, where

(t) =
v(t)
c
. (3.2)
In another frame

, in which the particle is supposed to instantaneously at rest:


ds
2
= c
2
dt
2
= c
2
d
2
, where d = dt
_
1

(t)

2
=
dt
(t)
: Lorentz invariant . (3.3)
The time variable is called the Proper time. It is the time as measured in the rest frame of a particle,
or the time measured in a frame moving synchronously with a particle.
Consider now, two successive events P
A
and P
B
associated with a particle, which are separated by the proper
time interval (
B

A
) in the particles rest frame. In another frame , relative to which the particle is moving
with a velocity v and where the time coordinate is t, the events are temporally separated by:
(t
B
t
A
) =
_

B

A
() d . (3.4)
Since =
_
1

2
_
> 1 always, the above equation implies that (t
B
t
A
) > (
B

A
), i.e., the temporal
separation between the events is elongated in the moving frame a phenomenon called Time dilation.
18
4 Tensor Analysis
Denition: Tensors (of specic rank and type) are physical entities associated with a spacetime point, which
follow linear, homogeneous laws of transformation under the transformation of coordinates.
4.1 Tensors of various Rank
Most generally, a tensor carries a number of superscript and subscript indices, the total number of which
signies the so-called rank of the tensor:
T

2
...m

2
...n
Rank = (m + n) . (4.1)
4.1.1 Tensor of Rank 0 : Scalar
Rank 0 tensors are scalars which remain invariant under all coordinate transformations:

_
t, x
i
_
Inv.
Coordinate transformations x

. (4.2)
In SR: a scalar, which is invariant under the Lorentz transformations, is often called a Lorentz scalar, e.g., the
invariant distance squared s
2
= c
2
t
2
|x|
2
.
4.1.2 Tensor of Rank 1 : Vector
Rank 1 tensors are vectors of two types:
I. Contravariant vector:
A

=
_
A
0
, A
i
_
=
_
A
0
,

A
_
. (4.3)
Under a coordinate transformation, this type of vectors transform similar to the coordinate dierentials:
A

=
x

Similar

to
dx

=
x

dx

. (4.4)
In SR: the above transformation rule becomes
A

. (4.5)
II. Covariant vector:
B

= (B
0
, B
j
) . (4.6)
Under a coordinate transformation, this type of vectors transform similar to the partial derivatives of the
coordinate scalars:
B

=
x

Similar

to

=
x

. (4.7)
In SR: the above transformation rule becomes
B

. (4.8)
4.1.3 Tensors of Rank 2
These are of three types:
I. Contravariant tensor: T

2
...m
, which transforms like a direct product of m contravariant vectors:
T

1
...m
=
x

1
x

1
. . .
x
m
x
m
T

1
...m
Similar

to
A

1
. . . A
m
=
x

1
x

1
. . .
x
m
x
m
A

1
. . . A
m
. (4.9)
In SR: the above transformation rule becomes
T

1
...m
=

1
. . .
m
m
T

1
...m
. (4.10)
19
4 TENSOR ANALYSIS
II. Covariant tensor: T

2
...n
, which transforms like a direct product of n covariant vectors:
T

1
...n
=
x

1
x

1
. . .
x
n
x
n
T

1
...n
Similar

to
B

1
. . . B

n
=
x

1
x

1
. . .
x
n
x
n
B

1
. . . B
n
. (4.11)
In SR: the above transformation rule becomes
T

1
...n
=

1

1
. . .
n
n
T

1
...n
. (4.12)
III. Mixed tensor: T

2
...m

2
...n
, which transforms like a direct product of m contravariant and n
covariant vectors:
T

1
...m

1
...n
=
_
x

1
x

1
. . .
x
m
x
m
__
x

1
x

1
. . .
x
n
x
n
_
T

1
...m

1
...n
(4.13)
Similar

to
_
A

1
. . . A
m
_ _
B

1
. . . B

n
_
=
_
x

1
x

1
. . .
x
m
x
m
__
x

1
x

1
. . .
x
n
x
n
_
(A

1
. . . A
m
) (B

1
. . . B
n
) .
In SR: the above transformation rule becomes
T

1
...m

1
...n
=
_

1
. . .
m
m
_
_

1
. . .
n
n
_
T

1
...m

1
...n
. (4.14)
4.2 Tensor Algebra
4.2.1 Linear Combination
Only tensors of the same rank and type (contravariant, covariant or mixed) could be added, to produce another
tensor of identical rank and type. The same is true, in fact, for any linear combination of tensors of same rank
and type. For e.g., given two covariant tensors of rank 2, A

and B

, the linear combination


T

= A

+ B

, [, : Spacetime scalars] , (4.15)


also transforms like a covariant tensor of rank 2.
4.2.2 Multiplication: Direct Product, Contracted Product and Scalar Product
Unlike addition, the multiplication of tensors of any type and rank, same or dierent, is possible. Accordingly,
one can dene three kinds of tensor products:
I. Direct Product: The aggregate obtained via multiplication of any two tensors of the same or dierent
type, and of ranks m and n say, provides another tensor of rank (m + n). For e.g.,
A

= T
a

, A

= T

, A

= T

, . (4.16)
II. Contracted Product: If two dierent types of tensors of ranks m and n say, have p number of matching
contra- and covariant indices, then the direct product of the tensors reduces to another tensor of rank (m+n2p)
with no matching contra- and covariant indices. For e.g.,
A

= T

, A

= T

, A

= T

, . (4.17)
Contraction implies the reduction of any mixed tensor of rank n (> 1), and with p matching contra- and
covariant indices, to a new tensor of rank (n 2p) after summing over the p matched indices. For e.g.,
T

= T

, T

= T

, T


= T

, . (4.18)
III. Scalar Product: Scalar function which is the outcome of a fully contracted product of two tensors of
the same rank and when the contra(co)variant indices of one is identical to the co(contra)variant indices of the
other. For e.g.,
A

= A B , S

= S T , Q

= Q R , . (4.19)
20
4.2 Tensor Algebra
4.2.3 Index raising/lowering, Norm and Trace
Raising and lowering of tensor indices are done by contracting with the metric tensor g

. For e.g.,
A

= g

, A

= g

;
T

= g

, T

= g

;
T
...
...
=
_
g

. . .
_
T
......
, . (4.20)
Accordingly, the scalar products are expressed as
A B = A

= g

= A
a
B

, S T = S

= g

= S

, . (4.21)
The norm of a tensor is its scalar product with itself. For e.g.,
|A|
2
= A

, |S|
2
= S

, |T|
2
= T

, . (4.22)
The trace of a tensor of even rank ( 2) is its full contraction in the mixed form. For e.g.,
tr
_
S

_
= g

= S

, tr
_
T

_
= g

= T


, . (4.23)
For the metric tensor g

itself, the norm and the trace are the same:


|g|
2
tr (g

) = g

= 4 in (3 + 1)-D, (4.24)
where in the fourth step, we have used a unique property of the metric tensor
g

=
_
1 for =
0 otherwise
: (3 + 1)-D Kronecker delta . (4.25)
4.2.4 Relation between Contravariant and Covariant components of a Tensor
In Special Relativity (SR), g

.
A. Components of a Vector A

:
A
0
=
0
A

=
00
A
0
= A
0
, A
i
=
i
A

=
ij
A
j
=
ij
A
j
. (4.26)
Therefore,
A

:=
_
A
0
, A
i
_

_
A
0
,

A
_
= A

:= (A
0
, A
i
) =
_
A
0
,
ij
A
j
_

_
A
0
,

A
_
. (4.27)
B. Components of a second rank Tensor T

:
T
00
=
0

0
T

= (
00
)
2
T
00
= T
00
,
T
0i
=
0

i
T

=
00

ij
T
0j
=
ij
T
0j
, T
i0
=
ij
T
j0
[Similarly] ,
T
ij
=
i

j
T

=
ik

jl
T
kl
=
ik

jl
T
kl
. (4.28)
Corollary: If we choose, T

, then for the Minkowski metric tensor

00
=
00
= 1 ,
0i
= 0 =
0i
,
ij
=
ij

ij
=
ij
. (4.29)
=

and

have the same set of components.


For higher rank tensors, the relationship between the contra- and covariant components could be found in a
similar manner as above.
21
4 TENSOR ANALYSIS
4.3 (Anti-)Symmetry of Tensors
For contravariant or covariant (but not mixed) tensors of rank 2, the symmetry or the antisymmetry under
the interchange of pair(s) of indices are dened as follows:
4.3.1 Symmetric Tensors
Rank 2 tensors:
S

= S

, S

= S

. (4.30)
Rank 3 tensors:
Partially Symmetric: S

= S

, or S

= S

, or S

= S

; (4.31)
Completely Symmetric: S

= S

= S

= S

= S

= S

. (4.32)
Similarly, for higher rank tensors one can dene partial symmetry (under the exchange of a given pair of indices)
and complete symmetry (under the exchange of any pair of indices). Symmetry for covariant tensors always
implies symmetry for their contravariant counterparts, and vice versa.
4.3.2 Antisymmetric Tensors
Rank 2 tensors:
A

= A

, A

= A

=A

= A

= 0. (4.33)
Rank 3 tensors:
Partially Antisymmetric: A

= A

, or A

= A

, or A

= A

; (4.34)
Completely Antisymmetric: A

= A

= A

= A

= A

= A

. (4.35)
Similarly, for higher rank tensors one can dene partial antisymmetry (under the exchange of a given pair of
indices) and complete antisymmetry (under the exchange of any pair of indices). Antisymmetry for covariant
tensors always implies antisymmetry for their contravariant counterparts, and vice versa.
4.3.3 Symmetrization and Antisymmetrization of Tensors
Rank 2 tensors: Any tensor T

(in general asymmetric) could be split up into a symmetric part S

and
antisymmetric part A

:
T

= S

+ A

, S

= T
()
, A

= T
[]
, (4.36)
where
T
()
=
1
2
(T

+ T

) , T
[]
=
1
2
(T

) , (4.37)
are called the symmetrization and the antisymmetrization of T

. The above splitting procedure can be adopted


for a contravariant tensors of rank 2, T

, as well. But for tensors of rank > 2 such a splitting into completely
symmetric (or, symmetrized) and completely antisymmetric (or, antisymmetrized) parts is not possible.
Rank 3 tensors: Symmetrization and Antisymmetrization of third rank covariant tensor T

can be done
as follows:
T
()
=
1
3!
(T

+ T

+ T

+ T

+ T + T

) ,
T
[]
=
1
3!
(T

+ T

+ T

T T

) . (4.38)
By denition, T
()
and T
[]
are completely symmetric and completely antisymmetric tensors, respectively,
and one may note that T
[]
has non-vanishing components only for = = . However, as mentioned above,
mere addition (or any linear combination) of T
()
and T
[]
does not give the original tensor T

.
The procedure of symmetrization and antisymmetrization can be adopted for higher ranked tensors as well. In
general, for any covariant or contravariant tensor of rank p ( 2), in a D-dimensional spacetime (i.e., 1 time +
(D 1) space), the symmetrization and the antisymmetrization schemes are as follows:
22
4.3 (Anti-)Symmetry of Tensors
Rank p 2 tensors: Completely symmetric and antisymmetric tensors constructed out of tensor T

2
...p
are given respectively by:
T
(
1

2
...p)
=
1
p!

P
T

P(1)

P(2)
...
P(p)
, (4.39)
T
[
1

2
...p]
=
1
p!

P
(1)
P
T

P(1)

P(2)
...
P(p)
, (4.40)
where the summations are for all permutations P of the indices (
1

2
. . .
p
) and the factor (1)
P
indicates
the parity equal to 1 of odd permutations P, and +1 for even permutations P.
4.3.4 Salient features of Symmetric and Antisymmetric Tensors
A few features of symmetric and antisymmetric tensors, and of the symmetrization and the antisymmetrization
of tensors are in order:
1. The scalar product of a symmetric and antisymmetric tensor vanishes identically. For e.g., consider the
rank 2 symmetric covariant and antisymmetric contravariant tensors S

and A

respectively. Then
S

= S

: Interchanging the dummy indices ,


= S

_
A

_
: Using the symmetry and antisymmetry properties of S

and A

= 0 . (4.41)
2. If the tensor T

1
...p
is completely symmetric, then T

1
...p
= T
(
1
...p)
, whereas if the tensor T

1
...p
is
completely antisymmetric, then T

1
...p
= T
[
1
...p]
.
3. In D-dimensions, the total number of independent components of the rank p tensor T

1
...p
and of its
symmetrized and antisymmetrized forms are given respectively by
N = D
p
, N
sym
=
_
D + p 1
p
_
=
(D + p 1)!
p! (D 1)!
, N
antisym
=
_
D
p
_
=
D!
p! (D p)!
. (4.42)
For e.g., if p = 2, then T

1
...p
= T

2
has N = D
2
components, and the number of independent
components of the symmetric tensor T
(
1

2
)
could be computed as
N
sym
[p = 2] = N(diag) +
1
2
N(o-diag) = D +
_
D
2
D
_
2
=
D(D + 1)
2
, (4.43)
which matches with that obtained from the formula for N
sym
in Eq. (4.42), viz., (D+21)!/[2!(D1)!] =
D(D + 1)/2. Similarly, keeping in mind that no diagonal elements exist for the antisymmetric tensor
T
[
1

2
]
, the number of its independent components can be computed as
N
antisym
[p = 2] =
1
2
N(o-diag) =
_
D
2
D
_
2
=
D(D 1)
2
, (4.44)
which again matches with that one gets using the formula for N
antisym
in Eq. (4.42). One may further
note that: since D(D + 1)/2 + D(D 1)/2 = D
2
, the symmetrized and antisymmetrized forms T
(
1

2
)
and T
[
1

2
]
add up to give the original tensor T

2
. This is a unique feature of the rank-2 tensors in any
general D-dimensional spacetime.
4. The above expression for N
antisym
in Eq. (4.42) is also suggestive from the point of view of its validity
only for tensors of rank p D. This is evident from the reduced expression:
N
antisym
=
D!
p! (D p)!
=
D(D 1) (D 2) . . . (D p + 1)
p!
, (4.45)
which vanishes whenever p = D + 1. Therefore, for any tensor whose rank exceeds the dimensionality,
antisymmetrization identically leads to zero. For e.g., in 3-dimensions (i.e., D = 3), the maximum rank of
a completely antisymmetric tensor is 3 (i.e., the tensor is T
[ijk]
), and the tensor has only one non-vanishing
independent component T
123
. The other non-vanishing components are all equal to it magnitude (e.g.,
23
4 TENSOR ANALYSIS
T
231
, |T
213
|). The reason is that complete antisymmetry implies all the indices of a given tensor must be
dierent. So when the rank equals the dimensionality the number of tensor indices equals the number
of dimensional indices, and there is only one choice for all the tensor indices to be unequal. Thus, in
(3 + 1)-dimensions (i.e., D = 4), the maximum rank of a completely antisymmetric tensor is 4 (i.e., the
tensor is T
[]
), and the only non-vanishing independent component of the tensor is T
0123
.
4.3.5 The Levi-Civita Tensor and the Generalized Kronecker Delta
In D-dimensions, the maximum ranked completely antisymmetric tensor is proportional to the Levi-Civita
tensor dened by

T
[
1

2
...
D
]

2
...
D
=
_

_
1 for
1

2
. . .
D
an even permutation of 01 . . . D
+1 for
1

2
. . .
D
an odd permutation of 01 . . . D
0 otherwise
, (4.46)
T
[
1

2
...
D
]

2
...
D
=
_

_
+1 for
1

2
. . .
D
an even permutation of 01 . . . D
1 for
1

2
. . .
D
an odd permutation of 01 . . . D
0 otherwise
. (4.47)
The following contraction identities exist between

and

in (3 + 1)-dimensions:

= 4! ,

= 3!

= 2!

, (4.48)
where

, . . . , (4.49)
are the Generalized Kronecker Delta tensors.
4.3.6 Tensor Duality
In D-dimensions the total number of independent components of a totally antisymmetric tensor of rank p is
the same as that of a totally antisymmetric tensor of rank (D p) due to the fact that
N(D, p) =
D!
p! (D p)!
=
D!
(D p)! [D (D p)]!
= N(D, D p) . (4.50)
This implies that there exists a duality between a rank p totally antisymmetric tensor and a rank (D p)
totally antisymmetric tensor in D-dimensions. Such a duality relationship is given by
A

2
...p
= A
[
1

2
...p]
= p!

2
...p
p+1
...
D

p+1
...
D
, (4.51)
where

A

p+1
...
D
is called the (Hodge-)dual of A

2
...p
. The inverse duality relation is given by

2
...p
=

A
[
1

2
...p]
=
1
p!

2
...p
p+1
...
D
A

p+1
...
D
. (4.52)

Most appropriately, the denitions hold for the covariant and contravariant Levi-Civita tensor densities
1

2
...
D
and

2
...
D
. Tensor densities are the most general entities which follow linear, homogeneous laws of transformation under the
transformation of coordinates, and tensors form only a sub-class of tensor densities. The dierence between a tensor and tensor
density is that the latter involves in its transformation law an extra factor depending on a certain integral power W (called the
weight) of the Jacobian |x/x

|:
T

x
x

W
x

.
However, since the Jacobian is identically equal to unity for the Lorentz transformations in Special Relativity, the above transfor-
mation law for the tensor density T

reduces to the transformation law for a tensor T

:
T

=
x

.
Accordingly, there is no distinction between the Levi-Civita tensor density
1

2
...
D
of weight W = 1 and the tensor
1

2
...
D
,
and also between the Levi-Civita tensor density

2
...
D
of weight W = +1 and the tensor

2
...
D
, in Special Relativity.
24
4.4 Tensor Calculus
The covariant indices of A
...
and

A
...
could be raised to the contravariant indices in the usual way using
the metric tensor g

. For e.g.,
In 3-D:
A
ijk
= 3!
ijk

A = A
ijk
= 3!
ijk
A ,

A
ijk
=
1
3!

ijk
A =

A
ijk
=
1
3!

ijk
A ,
A
ij
= 2!
ijk

A
k
= A
ij
= 2!
ijk
A
k
,

A
ij
=
1
2!

ijk
A
k
=

A
ij
=
1
2!

ijk
A
k
,
A
i
=
ijk

A
jk
= A
i
=
ijk
A
jk
,

A
i
=
ijk
A
jk
=

A
i
=
ijk
A
jk
,
A =
ijk

A
ijk
=
ijk
A
ijk
,

A =
ijk
A
ijk
=
ijk
A
ijk
. (4.53)
In (3 + 1)-D:
A

= 4!

A = A

= 4!

A,

A

=
1
4!

A =

A

=
1
4!

A,
A

= 3!

= A

= 3!

A

,

A

=
1
3!

=

A

=
1
3!

,
A

= 2!

= A

= 2!

A

,

A

=
1
2!

=

A

=
1
2!

,
A

= A

=

A

,

A

=

A

,
A =

=

A

,

A =

. (4.54)
4.4 Tensor Calculus
4.4.1 Tensor Dierentiation: Grad, Div, Curl and DAlembertian operators
Dierentiation of tensors in (3 + 1)-D is analogous to that in ordinary 3-D. We consider only the Special
Relativistic scenario, where the component structure of the dierential operators are as follows:
Covariant and Contravariant Dierential operators:

=
_

x
0
,

x
i
_
(
0
, ) ,

=
_

x
0
,

x
i
_

_

0
,
_
. (4.55)
where /x
i
is the ordinary three-gradient operator. Note that the above co(contra)variant operators have
component structures just opposite to the co(contra)variant four-vectors, viz., A

=
_
A
0
,

A
_
, A

=
_
A
0
,

A
_
.
Covariant Gradient: For a four-scalar (x
k
, t), the (covariant) gradient is dened by

=
_

x
0
,
_
=
_
1
c

t
,
_
. (4.56)
Covariant Divergence: For a contravariant four-vector A

(x
k
, t) (i.e., a contravariant four-tensor of rank 1),
the (covariant) divergence dened by

=
A
0
x
0
+
A
i
x
i
=
1
c
A
0
t
+

A, (4.57)
is a four-scalar (i.e., a four-tensor of rank 0). Similarly for a contravariant four tensor T

of rank 2, the
(covariant) divergence dened by

=
T
0
x
0
+
T
i
x
i
=
__

0
T
00
+
i
T
i0

,
_

0
T
0j
+
i
T
ij
_
, (4.58)
is a contravariant four-vector (i.e., a contravariant four-tensor of rank 1).
In general, for any contravariant four-tensor of rank m, the operation of (covariant) divergence yields a
contravariant four-tensor of rank (m1).
The contravariant divergence is similar to the covariant divergence, e.g.,

, . . . . (4.59)
25
4 TENSOR ANALYSIS
Covariant Curl: For a covariant four-vector A

(x
k
, t), the (covariant) curl is a contravariant antisymmetric
tensor of rank 2, dened by
C

= C
[]
=
1
2

) =

. (4.60)
Recall this from the analogy of the ordinary three-curl of a three-vector A
k
(x
l
) in 3-D, being a three-vector
itself with the i-th component as (
j
A
k

k
A
j
), i.e.,
_


A
_
i
=
1
2

ijk
(
j
A
k

k
A
j
) =
ijk

j
A
k
. (4.61)
The DAlembertian Operator: This is the four dimensional version of the Laplacian operator, dened as

=
0

0
+
i

i
=
1
c
2
d
2
dt
2

2
. (4.62)
4.4.2 The Four Volume and Tensor Integration
The invariant Four Volume element: The volume element in (3 + 1)-dimensions is given by
dV = d
4
x dt d
3
x . (4.63)
Under a general coordinate transformation: x

= x

(x
0
, x
1
, x
2
, x
3
) [cf. Eq. (2.64)], the volume element
transforms to:
dV

= d
4
x

d
4
x . (4.64)
i.e., d
4
x is actually a scalar density (or, a tensor density of rank 0).
Now, by the transformation rule of the metric tensor g

we have
g

=
x

= g

det
_
g

_
=

x
x

2
det (g

x
x

2
g . (4.65)
Therefore, using this relation, we nd the invariant four-volume element:

g dV =

g d
4
x

g dt d
3
x , (4.66)
which is a scalar (or, a tensor of rank 0).
In Special Relativity, g = det (

) = 1, and the invariant four-volume element is simply d


4
x = dt d
3
x.
Tensor Integration: In principle, integrations of tensors over the invariant four-volume yield tensor-like
quantities. However, one may dene surface and volume integrals, by dening innitesimal surface and volume
elements, in a (3 + 1)-D spacetime, in a similar way as in ordinary 3-D space. One may also have the general-
izations of the three-space Divergence and Stokes theorems in (3 + 1)-D. However, these are beyond the scope
of these lectures.
26
5 Special Relativistic Particle Dynamics
5.1 Four Velocity
In a covariant formalism one can construct a four vector, whose spatial components are proportional to the
components of the ordinary three-velocity of a particle (or a system of particles) and the temporal component
is proportional to the invariant speed of light c. The contravariant projections of this four velocity or world
velocity vector are:
u

=
dx

d
=
dx

dt/
=
_
dx
0
dt
,
dx
i
dt
_
= (c, v) . (5.1)
Since the proper time is invariant under a Lorentz transformation (LT), the four velocity u

transforms like
the coordinate dierentials dx

under LT.
The norm of u

is a universal constant:
u

=
2
_
c
2
|v|
2
_
= c
2
. (5.2)
5.2 Four Acceleration
In a covariant formalism one can dene the four acceleration, or the world acceleration of a particle as the
rate of change of the four velocity w.r.to the proper time along the particles world line. The contravariant
projections of the four acceleration are:
a

=
du

d
=
du

dt
=
_
c ,

v + v
_
. (5.3)
Now,
=
d
dt
_
1
v v
c
2
_
1/2
=
3
v

v
c
2
. (5.4)
So
a

=
2
_
_

2
v

v
c
,

v +
2
_
v

v
_
v
c
2
_
_
. (5.5)
One may explicitly verify that
a

< 0 and a

= 0 . (5.6)
5.3 Four Momentum and Energy
In a covariant formalism one can dene the four momentum of a particle as the product of the particles
rest mass m
0
and its four velocity. The spatial components of the four momentum correspond to those for the
relativistic three momentum of the particle, and the temporal component is proportional to the total relativistic
energy. The contravariant projections of the four momentum are:
P

= m
0
u

= m
0
dx

d
= (m
0
c, m
0
v) = (mc, mv) =
_
E
c
, ppp
_
, (5.7)
where
m m
0
: Relativistic mass ,
ppp = mv = m
0
v : Relativistic three-momentum,
E = mc
2
= m
0
c
2
: Relativistic total (i.e., rest + kinetic) energy .
The interpretations of ppp and E as relativistic momentum and energy come from the following considerations:
27
5 SPECIAL RELATIVISTIC PARTICLE DYNAMICS
Expanding ppp and E in powers of (v/c), one nds that in the non-relativistic limit (v/c 0):
ppp m
0
v : Non-relativistic three momentum,
E m
0
c
2
+
1
2
m
0
v
2
: Rest energy + Non-relativistic kinetic energy . (5.8)
ppp and E can be shown to be the only functions of the velocity v, whose conservations are Lorentz invariant.
The conservation of the ppp under a Lorentz transformation (LT) necessarily implies the conservation of E.
This is shown in the following subsection.
The norm of the four momentum P

is given by
P

=
00
_
P
0
_
2
+
ij
P
i
P
j
=
E
2
c
2
|ppp|
2
. (5.9)
But, since
P

= m
2
0
u

= m
2
0

2
_
c
2
v
2
_
= m
2
0
c
2
, (5.10)
we have
E
2
= |ppp|
2
c
2
+ m
2
0
c
4
. (5.11)
Therefore, for massless particles the total energy is simply E = |ppp| c.
5.4 Conservation of Energy and Momentum
Since the four momentum P

= m
0
dx

/d transforms as a four vector under a LT (due to the fact that m


0
and are invariants and dx

is a four vector), in any reaction the change of the sum of the four momenta P

of all the involved particles (say n in number), also transforms as a four vector:
P

n
P

(n)
=

n
P

(n)
. (5.12)
Now, the conservation of the relativistic three-momentum ppp under a LT means that:

n
ppp
(n)
= 0 in one inertial frame =

n
ppp

(n)
= 0 in all other inertial frames .
i.e.,

n
P
i
(n)
= 0 =

n
P
i
(n)
= 0 =
i

n
P

(n)
=
i
0

n
P
0
(n)
+
i
j

n
P
j
(n)
. (5.13)
Since
i
j
are constants (i, j = 1, 2, 3),

n
P
i
(n)
= 0 implies that the last term on the r.h.s. of the above
equation is automatically zero. So we are left with

i
0

n
P
0
(n)
= 0 =

n
P
0
(n)
= 0 =
1
c

n
E
(n)
, (5.14)
as
i
0
(i = 1, 2, 3) are not necessarily zero.
Thus, the conservation of the relativisitic three-momentum p, implies conservation of the total relativistic
energy E, and both in tern imply conservation of the four momentum P

. In other words,

n
P

(n)
= 0 i.e.

n
P

(n)I
=

n
P

(n)F
follows from:
(i)

n
ppp
(n)
= 0 i.e.

n
ppp
(n)I
=

n
ppp
(n)F
and (ii)

n
E
(n)
= 0 i.e.

n
E
(n)I
=

n
E
(n)F
,
(5.15)
where the subscript I (F) denotes initial (nal) conguration. The conservation of the total relativistic energy,
by denition, implies the conservation of the total relativistic mass:

n
m
(n)
= 0 i.e.

n
m
(n)I
=

n
m
(n)F
[m = m
0
] . (5.16)
28
5.4 Conservation of Energy and Momentum
Illustrative Example: Compton eect
A photon hits a stationary particle of rest mass m
0
as shown in Fig. 9. We need to determine the angle of
deection of the photon in terms of its initial and nal energies (E
I
and E
F
respectively), or its initial and
nal frequencies (
I
and
F
respectively).
Recoil

E
I
E
F
m
0
Figure 9: Illustration of the Compton eect.
Initial and nal four momenta (P

(I)
and P

(F)
respectively) of the photon are given by
P

(I)
=
_
E
I
c
, ppp
I
_
, P

(F)
=
_
E
F
c
, ppp
F
_
, (5.17)
where ppp
I
and ppp
F
are the initial and the nal (relativistic) three-momenta of the photon. As, for the photon,
E
I
= c ppp
I
and E
F
= c ppp
F
, we have
P

(I)
= (|ppp
I
| , ppp
I
) , P

(F)
= (|ppp
F
| , ppp
F
) . (5.18)
For the massive particle, since its initial energy is equal to its rest energy m
0
c
2
, its initial four momentum (

P

(I)
say) is given as

(I)
=
_
m
0
c ,

0
_
. (5.19)
Let its nal four momentum be

P

(F)
. Then since the norm of the four momentum is a Lorentz invariant
quantity, we have

P
(F)

P

(F)
=

P
(I)

P

(I)
= m
2
0
c
2
. (5.20)
Now, by the conservation of four momenta
P

(I)
+

P

(I)
= P

(F)
+

P

(F)
=

P

(F)
= (m
0
c +|ppp
I
| |ppp
F
| , ppp
I
ppp
F
) , (5.21)
using Eqs. (5.18) and (5.19). Again, by the Eq. (5.20)

P
(F)

P

(F)
= m
2
0
c
2
= (m
0
c +|ppp
I
| |ppp
F
|)
2
+|ppp
I
ppp
F
|
2
= 2m
0
c (|ppp
I
| |ppp
F
|) + (|ppp
I
| |ppp
F
|)
2
= |ppp
I
|
2
+|ppp
F
|
2
+ 2 ppp
I
ppp
F
=
1
|ppp
I
|

1
|ppp
F
|
=
2
m
0
c
sin
2

2
, (5.22)
where is the angle between ppp
I
and ppp
F
.
Since, E
I
= c ppp
I
= h
I
and E
F
= c ppp
F
= h
F
, h being the Plancks constant, we nally obtain
= 2 sin
1
_
_
m
0
c
2
2h
_
1

I
__
1/2
_
. (5.23)
29
5 SPECIAL RELATIVISTIC PARTICLE DYNAMICS
5.5 Four Force
In a covariant formalism, the four force on a particle as the rate of change of the four momemtum w.r.to the
proper time along the particles world line. The contravariant projections of the four force are:
F

=
dP

d
=
dP

dt
=
_

E
c
,

FFF
_
, (5.24)
where

FFF =

ppp
dppp
dt
: Relativistic three-force . (5.25)
By dention
F

= ma

= m
0
a

= m
0
du

dt
, (5.26)
which can be regarded as a relativistic generalization to the Newtons second law of motion in classical me-
chanics.
Now, for a particle moving with instantaneous velocity v, the time derivative of the total energy is equal
to the total power dissipated:

E =

FFF v =

ppp v . (5.27)
Therefore,
F

=
_

E
c
,

FFF
_
=
_

FFF v
c
,

FFF
_
. (5.28)
Moreover, since the components of the four-velocity are given by: u
0
= c and u = v, we have
F

=
1
c
_

FFF u, u
0

FFF
_
. (5.29)
30
6 Special Relativistic Electrodynamics
6.1 Four Current Density
Consider a system of n charged particles, with xed charges q
(n)
at locations x
(n)
(t). The ordinary three-charge
density and three-current density are dened respectively by
(x, t) =

n
q
(n)

3
_
x x
(n)
(t)
_

jjj (x, t) =

n
q
(n)

3
_
x x
(n)
(t)
_ dx
(n)
(t)
dt
(6.1)
where
3
_
x x
(n)
_
is the three-dimensional Dirac delta function. In analogy with the three-current density

jjj (x, t) one may dene the four current density vector J

(x, t) which unites both the densities and

jjj as follows:
J

(x, t) =

n
q
(n)

3
_
x x
(n)
(t)
_
dx

(n)
(t)
dt
. (6.2)
To verify that this is indeed a four vector, we set x
0
(n)
(t) = ct ( n), and use the property of the delta function:
_
dx f(x) (x y) = f(y) , (6.3)
to write J

(x, t) as
J

(x, t) =
_
dx
0
(n)

n
q
(n)

_
x
0
x
0
(n)
_

3
_
x x
(n)
(t

)
_
dx

(n)
(t

)
dx
0
(n)
/c
= c
_
cdt

n
q
(n)

4
_
x

(n)
(t

)
_
dx

(n)
(t

)
cdt

= J

(x
i
, t) = c
_
d

n
q
(n)

4
_
x

(n)
()
_
dx

(n)
()
d
. (6.4)
As the four-dimensional delta function
4
_
x

(n)
()
_
is a scalar, and so is the proper time variable ,
whereas dx

(n)
is a four vector, J

is a four vector. One may easily check from the above expression (6.2) for
J

that its temporal and spatial components are proportional to and



jjj, given by Eqs. (6.1), respectively:
J

(x
i
, t) =
_
c
_
x
i
, t
_
,

jjj
_
x
i
, t
_
_
. (6.5)
One may also verify, by using the continuity equation, that the four current density is covariantly conserved,
i.e., its four divergence is zero:

=

t
+

jjj = 0 . (6.6)
6.2 Covariant Lorentz Force Equation
In three dimensions, the electromagnetic force on a particle moving with instantaneous velocity v, and carrying
a charge q is given in terms of the well-known 3-dimensional Lorentz force equation:

FFF
L
= q
_

EEE +
1
c
v

BBB
_
. (6.7)
In a covariant formalism, since the four force is given by Eq.(5.28), we have the covariant Lorentz force equation
F

L
=
_

FFF
L
v
c
,

FFF
L
_
=
q
c
_

EEE v,
_
c

EEE + v

BBB
__
=
q
c
_

EEE u,
_
u
0

EEE + u

BBB
__
, (6.8)
31
6 SPECIAL RELATIVISTIC ELECTRODYNAMICS
u
0
and u being the temporal and spatial components of the four velocity u

=
_
u
0
, u
_
= (c, v).
One may verify that in compact form the above equation can be expressed as (q/c) times the covariant
dot product of a second rank antisymmetric tensor F

, constructed out of the electric and the magnetic eld


vectors, and the four velocity vector u

:
F

L
=
q
c
F

. (6.9)
Explicitly, the form of F

, which is referred to as the electromagnetic eld tensor, is given by


F

: F
00
= 0 , F
0i
= F
i0
= E
i
, F
ik
= F
ki
=
ijk
B
j
(6.10)
= F

= F

=
_
_
_
_
0 E
1
E
2
E
3
E
1
0 B
3
B
2
E
2
B
3
0 B
1
E
3
B
2
B
1
0
_
_
_
_

_
_
_
_
0 E
x
E
y
E
z
E
x
0 B
z
B
y
E
y
B
z
0 B
x
E
z
B
y
B
x
0
_
_
_
_
. (6.11)
To see that F

is indeed a tensor of rank 2, i.e., Eq. (6.9) is a tensorial equation which encodes Lorentz force
equations in every Lorentz frame, we suppose Eq. (6.9) to hold in a Lorentz frame . In another Lorentz frame

, we have the transformed equation


F

L
=
q
c
F

. (6.12)
Now,
F

L
=
dP

d
=
x

dP

d
=
x

L
. (6.13)
Also
u

=
x

or u

=
x

. (6.14)
Substituting this in Eq. (6.13), and using Eqs. (6.9) and (6.12), we have
F

L
=
q
c
F

=
x

L
=
x

_
q
c
F

_
=
x

_
q
c
F

_
=
_
F

_
u

= 0 = F

=
x

, (6.15)
as u

is arbitrary. Thus F

transforms like a rank-2 tensor.


6.3 Covariant formulation of Maxwells Equations
Consider a homogeneous, isotropic and non-dispersive medium in which the Maxwells equations for the electric
and magnetic elds
_

EEE(x, t),

BBB(x, t)
_
, produced by a charge density (x, t) and a current density

jjj(x, t), are:
(i)

EEE = 4
(ii)

BBB

EEE
c
=
4
c

jjj
(iii)

BBB = 0
(iv)

EEE +

BBB
c
= 0
or, in index notation:
(i)
i
E
i
=
4
c
J
0
(ii)
ijk

i
B
j
+
0
E
k
=
4
c
J
k
(iii)
i
B
i
= 0
(iv)
ijk

i
E
j

0
B
k
= 0
, (6.16)
where the overhead dot {} denotes /t and
0
= (1/c)/t.
32
6.3 Covariant formulation of Maxwells Equations
The rst two of the Maxwells equations (6.16), that involve the sources J
0
= c and J
k
=
_

jjj
_
k
, k = 1, 2, 3,
can be expressed in terms of the components of the electromagnetic eld tensor F

[Eq. (6.10)] as
(i)
i
F
i0
=
4
c
J
0
, (ii)
i
F
ik
+
0
F
0k
=
4
c
J
k
. (6.17)
Similarly, the last two of the Maxwells equations (6.16), that do not involve the sources, can be expressed
in terms of the components of a second rank antisymmetric tensor

F

(which like the electromagnetic eld


tensor F

, is also constructed out of



EEE and

BBB components), as
(i)
i

F
i0
= 0 , (ii)
i

F
ik
+
0

F
0k
= 0 . (6.18)
The components of

F

are given explicitly by

:

F
00
= 0 ,

F
0i
=

F
i0
= B
i
,

F
ik
=

F
ki
=
ijk
E
j
(6.19)
=

F

=

F

=
_
_
_
_
0 B
1
B
2
B
3
B
1
0 E
3
E
2
B
2
E
3
0 E
1
B
3
E
2
E
1
0
_
_
_
_

_
_
_
_
0 B
x
B
y
B
z
B
x
0 E
z
E
y
B
y
E
z
0 E
x
B
z
E
y
E
x
0
_
_
_
_
. (6.20)
The covariant tensors F

and

F

can be obtained respectively from F

and

F

as:
F

= F

: F


EEE

EEE,

BBB

BBB
F

, (6.21)

=

F

:

F


EEE

EEE,

BBB

BBB

. (6.22)
Using these, and the explicit forms of F

and

F

, given by Eqs. (6.11) and (6.20), one may verify that



F

is indeed identical to the dual of F

=
1
2

or F

= 2

= 2

F

. (6.23)
Moreover, the above two Maxwells equations involving the sources, viz., Eqs. (6.17 i, ii), whose l.h.s. contain
the three-divergence or(and) time-derivative of the components of F

, can be encoded in a single tensorial


equation whose l.h.s. is the four-divergence of F

. Similarly, the two source-free Maxwells equations, viz.,


Eqs. (6.18 i, ii), can be encoded in a single tensorial equation whose l.h.s. is the four-divergence of

F

. That
is, in the covariant formalism the Maxwells equations are expressed as:
(i)

=
4
c
J

(ii)

= 0
, (6.24)
where J

=
_
c,

jjj
_
is the four current density. Eq. (6.24 ii) is in fact an identity, known as the Maxwell-Bianchi
identity, and can also be expressed like a Jacobi identity:

=
1
2

= 0 =

= 0 . (6.25)
In absence of sources ( = 0,

jjj = 0), the above two Maxwells equations (6.24) in covariant form, are indistin-
guishable from each other under the transformation:
F

i.e.

EEE

BBB,

BBB

EEE , (6.26)
which is therefore called the electromagnetic duality transformation.
33
6 SPECIAL RELATIVISTIC ELECTRODYNAMICS
6.4 Transformation of Electromagnetic Fields
In Special Relativity, F

and

F

being rank-2 tensors, follow the transformation rules


F

,

F

, (6.27)
where

are the Lorentz transformation matrix elements, which for a generic Lorentz boost, in an arbitrary
direction, are given by (see 2.5):

0
0
= ,
i
0
=
i
,
0
i
=
ik

k
=
i
,

i
j
=
i
j
+
( 1)
|

|
2

jk

i

k
=
i
j
+
( 1)
|

|
2

j
. (6.28)
Using these
F
i0
= E
i
=
i

=
i
0

0
k
F
0k
+
i
j

0
0
F
j0
+
i
j

0
k
F
jk
=
_

EEE +

BBB
_
i

_

1
|

|
2
_
_

EEE
_

i
, (6.29)

F
i0
= B
i
=
i

=
i
0

0
k

F
0k
+
i
j

0
0

F
j0
+
i
j

0
k

F
jk
=
_

BBB

EEE
_
i

_

1
|

|
2
_
_

BBB
_

i
. (6.30)
As

2
=
1
1 |

|
2
= |

|
2
=

2
1

2
, we have
1
|

|
2
=

+ 1
. (6.31)
Hence, the expressions (6.29) for the transformed electric and magnetic elds reduce to

EEE

=
_

EEE +

BBB
_


2
+ 1
_

EEE
_

BBB

=
_

BBB

EEE
_


2
+ 1
_

BBB
_

. (6.32)
For a single Lorentz boost along, say, the x
1
x-axis, i.e.,

= x, we have from the above equations (6.32)
E

x
= E
x
E

y
= (E
y
B
z
)
E

z
= (E
z
+ B
y
)
and
B

x
= B
x
B

y
= (B
y
+ E
z
)
B

z
= (B
z
E
y
)
. (6.33)
So, a purely electric or a purely magnetic eld in one reference frame, is in general a combination of electric
and magnetic elds in another.
6.5 Electromagnetic Scalar invariants
In principle, one can construct three independent four-scalar quantities using the electromagnetic eld tensor
or(and) its dual, viz., F

,

F

. However, only two of these are actually linearly independent:


(i) F

=

F

= 2
_

BBB

EEE

2
_
, (ii) F

=

F

= 4

EEE

BBB . (6.34)
Therefore,
I. If

EEE

BBB

in one Lorentz frame , then

EEE

BBB

in every other Lorentz frame

.
II. If

EEE

BBB in one Lorentz frame , then

EEE


BBB

in every other Lorentz frame

, unless

EEE

or

BBB

vanishes
identically.
34
6.6 Potential formulation of Electrodynamics
6.6 Potential formulation of Electrodynamics
6.6.1 Electromagnetic Four Potential
As pointed out in 6.3, the Maxwell-Bianchi identity, viz. Eq. (6.24 ii), is equivalent to vanishing of the sum
of all the cyclic permutations of

= 0 =

= 0 [cf. Eq. (6.25)] .


This has a simple solution of the form
F

, (6.35)
where A

:=
_
,

A
_
is referred to as the electromagnetic four vector potential. Its temporal part is called the
electromagnetic scalar potential , whereas the spatial part is the electromagnetic three-vector potential

A. The
covariant vector A

has components
_
,

A
_
.
Verication:

= 0 .
In three-vectorial notation, the relations between the electric and magnetic vectors (

E and

B) and the scalar
and three-vector potentials ( and

A), can be obtained from the above equation (6.35) as follows:
F
0i
= E
i
=
0
A
i

i
A
0
=

E =

A
c
, (6.36)

F
0i
= B
i
=
1
2

0ijk
F
jk
=
1
2

0ijk
(
j
A
k

k
A
j
) =
0ijk

j
A
k
=

B =

A . (6.37)
Now, substituting Eq. (6.35) in the Maxwells equation (6.24 i) [see 6.3], one obtains the evolution equation
for the four vector potential:

_
=
4
c
J

=
4
c
J

=
2
A

) =
4
c
J

. (6.38)
6.6.2 Gauge Transformations
Under the transformation
A

= A

+
a
, (x, t) := Arbitrary scalar function of x and t , (6.39)
the r.h.s. of Eq. (6.38) is unaected, whereas the l.h.s. transforms to

2
A

=
2
A

+
2

(A

)
=
2
A

2
=
2
A

,
i.e., the l.h.s., and hence the entire equation (6.38), is invariant. This implies that there is a freedom to choose
the gauge (i.e., to impose a condition on the hitherto arbitrary scalar function ) while solving the equation
(6.38) for the four vector potential A

, and hence to determine the measurable quantities electric and


magnetic elds

E,

B using the Eqs. (6.36) and (6.37). As such, the above transformation (6.39) is referred
to as the Gauge transformation, the potential A

as the Gauge potential, the elds



E,

B as the Gauge elds,
the function as the Gauge function, and the condition on (required to be imposed) as the Gauge condition.
35
6 SPECIAL RELATIVISTIC ELECTRODYNAMICS
6.6.3 Gauge conditions: Lorentz and Coulomb gauges
In the following, we will discuss two popular gauge conditions which are of most importance in describing
physical phenomena associated with the gauge elds.
Lorentz gauge condition:
If we choose to be such that

, (6.40)
then

+
2
= 0 . (6.41)
The above condition (6.40) is therefore equivalent to

= 0 i.e.,

A +

c
= 0 . (6.42)
This is known as the Lorentz gauge condition

. In this gauge, the above equation (6.38) reduces to

2
A

=
4
c
J

. (6.43)
Recalling that J

:=
_
c,

jjj
_
, the temporal and spatial components of the above equation (6.43) are given as
(a)
2

_
1
c
2

2
t
2

2
_
= 4
(b)
2

A
_
1
c
2

2
t
2

2
_

A =
4
c

jjj
. (6.44)
These are the inhomogeneous wave equations satised by the scalar potential and the components of the
three-vector potential

A in the Lorentz gauge.
Coulomb gauge condition:
If we now choose to be such that

2
=

A i.e.,
i

i
=
i
A
i
, (6.45)
then

i
A
i

A
i
A
i
+
i

i
A
i
+
i

i
=

A
2
= 0 . (6.46)
The above condition (6.45) is therefore equivalent to

i
A
i
= 0 i.e.,

A = 0 . (6.47)
This is known as the Coulomb gauge condition. In this gauge, Eq. (6.38) reduces to

2
A

c
_
=
4
c
J

. (6.48)
It is easy to verify that the temporal and spatial components of this equation are given by
(a)
2
= 4
(b)
2

A +

c
=
4
c

jjj
. (6.49)
Eq. (6.49 a) is the time-dependent Poisson equation satised by the scalar potential and Eq. (6.49 b) is a
coupled second order dierential equation in and the components of the three-vector potential

A.

Or, more appropriately the Lorentz-Lorentz gauge condition, after Hendrik Antoon Lorentz who, in the course of demonstrating
the covariance of Maxwells equations in 1903, rediscovered this condition already found in 1867 by Ludvig Valentin Lorentz.
36
6.6 Potential formulation of Electrodynamics
One way to work out the solution for

A is to solve rst the equation (6.49 a) for , and then to substitute that
solution in (6.49 b) and solve for

A. However, this is a tedious and cumbersome procedure. There is in fact a
easier way out:
Split the three-current density

jjj in longitudinal and transverse parts (

jjj

and

jjj

) as

jjj =

jjj

+

jjj

,
_

jjj

= 0 ,

jjj

= 0
_
. (6.50)
Now, by the charge-current continuity equation (6.6)

jjj =

jjj

= =

t
_

1
4

2

_
by Eq. (6.49 a)
=

jjj

=
1
4

2

=
1
4

=
_

jjj


1
4

_
= 0 . (6.51)
Moreover,

jjj


1
4

_
=

0. (6.52)
Hence

= 4

jjj

, (6.53)
substituting which in Eq. (6.49 b) we nally obtain the inhomogeneous wave equation satised by the three-
vector potential

A in the Coulomb gauge:

A
_
1
c
2

2
t
2

2
_

A =
4
c

jjj

. (6.54)
37
7 WAVE SOLUTIONS IN ELECTRODYNAMICS THE RETARDED POTENTIALS
7 Wave Solutions in Electrodynamics the Retarded Potentials
In this section, we will focus on determining the solutions of the generic wave equations for the gauge potentials
and

A in Lorentz and Coulomb gauges. We will rst discuss the non-covariant approach of obtaining the
solutions for and the individual components of

A, in these gauges. The wave equation for in the Coulomb
gauge being simply the time-dependent Poissons equation (6.49 a), it is easier to obtain its solution, than
in the Lorentz gauge in which satises the inhomogeneous wave equation (6.44 a). However, on the whole,
working in the Coulomb gauge is not of much advantage as to working in the Lorentz gauge. The reason being
that the components of

A satisfy similar type of wave equations (6.49 b) in the Coulomb gauge, as do the
components of

A as well as in the Lorentz gauge, viz. Eqs. (6.44). So, anyhow we have to solve for the
inhomogeneous wave equation in order to obtain the complete set of solutions for the gauge potentials and

A,
no matter whether we use the Coulomb or the Lorentz gauge condition. Now, an approximate solution of the
inhomogeneous wave equation can be obtained without much diculty and without resorting to the covariant
formulation of electrodynamics. However, such approximation has a limited domain of validity, beyond which
there is as such no clue as to solving the inhomogeneous wave equation in the non-covariant approach. We
therefore resort to the covariant approach of solving the wave equation (6.43) for the four gauge potential
A

in the Lorentz gauge. In fact, the Lorentz gauge is the only one useful for the determination of A

, as
in the Coulomb gauge the corresponding equation (6.48) for A

requires intrinsically non-covariant way of


getting solution due to the presence of the term involving

. The covariant approach is much more elegant and
exact, compared to the non-covariant approach. However, the latter is still of importance in view of the basic
understanding of the development of the technique of solving for A

in a covariant way. One way to see this


is that the covariant approach (which utilizes the Lorenz gauge condition) alludes to a direct generalization
of the exact but non-covariant method of solving the time-dependent Poisson equation (satised by in the
Coulomb gauge). This would be made clear as we go through what follows.
7.1 Solutions for scalar and three-vector gauge potentials the Non-covariant approach
7.1.1 Solution technique in the Lorentz gauge
In the Lorentz gauge, the generic form of the inhomogeneous wave equation is given by

2
(x, t) = 4 f(x, t) , (7.1)
where and f are either and respectively, or a component of

A and the corresponding component of

jjj/c,
respectively [see Eqs. (6.44)].
Let us expand the wave-functions and the sources in Fourier integrals:
(a) (x, t) =
_

(x) e
it
and (b) f(x, t) =
_

d f

(x) e
it
, (7.2)
where is the angular frequency, and

(x), f

(x) are respectively the Fourier transforms of (x, t), f(x, t):
(a)

(x) =
1
2
_

dt (x, t) e
it
and (b) f

(x) =
1
2
_

dt f(x, t) e
it
. (7.3)
Plugging Eqs. (7.2) in Eq. (7.1), and using the vacuum dispersion relation
= c k , k := wave number , (7.4)
we obtain the three-dimensional time-dependent inhomogeneous wave equation, viz. the Helmholtz equation:
_

2
+ k
2
_

(x) = 4 f

(x) . (7.5)
38
7.1 Solutions for scalar and three-vector gauge potentials the Non-covariant approach
We next apply a well-known principle in wave optics, viz. the Huygens-Fresnel principle, which states that:
Given a wave, each point on its wavefront acts as a source for spherical wavelets of varying amplitude (weight).
A linear superposition of the individual wavelets, from each of these point sources, gives rise to a new wavelet.
According to this principle, the solution of the helmholtz equation (7.5) can be expressed as a weighted super-
position of the solutions of an equation similar to Eq. (7.5) but with the source term (on the r.h.s) replaced
by that of a single point source, i.e.,

(x) = 4
_
d
3
x

(x

) G

(x, x

) , (7.6)
where f

(x

) is the wavelet amplitude at the source point x

, and G

(x, x

) is the Greens function or Propagator


that satises the equation
_

2
+ k
2
_
G

(x, x

) =
3
_
x x

_
. (7.7)
Translational symmetry of space demands
G

(x, x

) = G

(x x

) . (7.8)
Denoting

R x x

, (7.9)
and assuming sperical symmetry
G

(x x

) = G

(x x

) i.e., G

R) = G

(R) , (7.10)
we can express Eq. (7.7) as
_

2
+ k
2
_
G

(R) =
3
(

R) . (7.11)
Thus, the original problem of nding solutions for the gauge potentials in the Lorentz gauge boils down to the
determination of the solution of the above Greens function equation (7.11). Before we work out such solution,
let us resort, in the following subsection, to the corresponding scenario in the Coulomb gauge.
7.1.2 Solution technique in the Coulomb gauge
In the Coulomb gauge, the wave equations are
(a)
2
(x, t) = 4 (x, t) and (b)
2
(x, t) = 4 f(x, t) , (7.12)
where, in the second equation, and f denote a component of

A and the corresponding component of

jjj

/c,
respectively [see Eqs. (6.49 b)].
Expansion of the wave-function (x, t) and the source f(x, t) in Eq. (7.12 b) in Fourier integrals would lead
to the same (Helmholtz) equation (7.5) whose solution could be obtained by determining the solution of the
above Greens function equation (7.11). On the other hand, if we expand (x, t) and (x, t) in Eq. (7.12 a) in
Fourier integrals as
(a) (x, t) =
_

(x) e
it
and (b) (x, t) =
_

(x) e
it
, (7.13)
where

(x),

(x) are respectively the Fourier transforms of (x, t), (x, t):
(a)

(x) =
1
2
_

dt (x, t) e
it
and (b)

(x) =
1
2
_

dt (x, t) e
it
, (7.14)
then proceeding as before, and using the vacuum dispersion relation (7.4) and applying the Huygens-Fresnel
principle, we get

(x) = 4
_
d
3
x

(x

) G

(x, x

) . (7.15)
39
7 WAVE SOLUTIONS IN ELECTRODYNAMICS THE RETARDED POTENTIALS
Utilizing again the conditions of translational symmetry and spherical symmetry, we have the following equation
for the Greens function (or, propagator) G

(x, x

) = G

(R):

2
G

(R) =
3
(

R) . (7.16)
In what follows (in the next subsection) we will rst work out the solution of this equation, and then resort to
the determination of the solution of the more general propagator equation (7.11).
7.1.3 Solutions of the Propagator equations
A. Solving
2
G

(R) =
3
(

R) [cf. Eq. (7.16)]:


Let us expand
G

(R) =
1
(2)
3
_
d
3
k G(

k) e
i

R
, (7.17)
where G(

k) is the Fourier transform of G

(R) in the k-space:


G(

k) =
_
d
3
R G

R) e
i

R
. (7.18)
On the other hand, the Fourier transform of
3
(

R) is unity:

3
(

R) =
1
(2)
3
_
d
3
k e
i

R
. (7.19)
Substituting Eqs. (7.17) and (7.19) in the original Greens function equation (7.16), we have
1
(2)
3
_
d
3
k G(

k)
_
i

k
_

_
i

k
_
e
i

R
=
1
(2)
3
_
d
3
k e
i

R
= G(

k) G(k) =
1
k
2
. (7.20)
Plugging this back in Eq. (7.17), we get
G

(R) =
1
(2)
3
_
d
3
k
1
k
2
e
i

R
. (7.21)
Choosing now

k = kz and using sperical polar coordinates (R, , ), in which
d
3
k = k
2
dk sin d d and

k

R = kRcos , (7.22)
being the angle between the z-direction and

R, the above expression (7.21) becomes
G

(R) =
1
(2)
3
_

0
_

0
_
2
0
k
2
dk sin d d
1
k
2
e
ikRcos
=
1
4
2
_

0
dk
_

0
sin d e
ikRcos
. (7.23)
Let = cos , so that d = sin d. Then
G

(R) =
1
4
2
_

0
dk
_
1
1
d e
ikR
=
1
4
2
_

0
dk
_

2 sin(kR)
kR
_
=
1
2
2
R
_

0
d(kR)
sin(kR)
kR
=
1
2
2
R

2
, (7.24)
i.e.,
G

(R) =
1
4R

1
4 |x x

|
. (7.25)
40
7.1 Solutions for scalar and three-vector gauge potentials the Non-covariant approach
B. Solving
_

2
+ k
2
_
G

(R) =
3
(

R) [cf. Eq. (7.11)]:


Because of spherical symmetry (which implies G

R) = G

(R)), we can write the above equation in spherical


polar coordinates as
1
R
2

R
_
R
2
G

(R)
R
_
+ k
2
G

(R) =
3
(

R) =
_
d
2
dR
2
+ k
2
_
[R G

(R)] = R
3
(

R) . (7.26)
Away from the source point R = |x x

| = 0, the delta function has no eect, and therefore [R G

(R)] satises
the homogeneous equation
_
d
2
dR
2
+ k
2
_
[R G

(R)] = 0 . (7.27)
The general solution of this equation is
G

(R) = C
+
G
+

(R) + C

(R) with G

(R) =
e
ikR
R
, C

:= Arbitrary constants . (7.28)


Inserting this in the original equation (7.11), and integrating over a small volume V

around R = |x x

| = 0,
the volume integrals can be approximated as
(C
+
+ C

)
_
V

d
3
x

2
_
1
R
_
+ k
2
(C
+
+ C

)
_
V

d
3
x

1
R

_
V

d
3
x

3
(

R) . (7.29)
Now, the second integral on the l.h.s., viz.,
_
V

d
3
x

1
R
=
_
V

x x

2
dx

sin

1
|x x

|

R|xx

|0
0 . (7.30)
Moreover, an important result of vector calculus is that

2
_
1
R
_

2
_
1
|x x

|
_
= 4
3
(

R) . (7.31)
Therefore, the rst integral on the l.h.s. of Eq. (7.29), viz.,
_
V

d
3
x

2
_
1
R
_
= 4
_
V

d
3
x

3
(

R) . (7.32)
Substituting Eqs. (7.30) and (7.32) in Eq. (7.29), we have in the limit R 0:
C
+
+ C

=
1
4
. (7.33)
Hence, the solution (7.28), which now takes the form
G

(R) = C
_
G
+

(R) G

(R)

(R)
4
=
K
R
sin(kR)
e
ikR
4R
; with C C
+
, K = 2iC , (7.34)
can be treated as a complete solution of the full equation (7.11) [or, Eq. (7.26)] as long as we rely on the
viability of the above approximation [Eq. (7.29)].
7.1.4 General Solutions for the Potentials
I. In the Lorentz gauge:
Substituting the solution (7.28) in Eq. (7.6) we have

(x) = C
+

(x) + C

(x) with

(x) = 4
_
d
3
x

(x

) G

x x

_
. (7.35)
Inserting this in Eq. (7.2 a):
(x, t) = C
+

+
(x, t) + C

(x, t) with

(x, t) = 4
_
d
3
x

d f

(x

) G

x x

_
e
it
. (7.36)
41
7 WAVE SOLUTIONS IN ELECTRODYNAMICS THE RETARDED POTENTIALS
Using now
G

(R) G

x x

_
=
e
ik|xx

|
|x x

|
, (7.37)
we get

(x, t) = 4
_
d
3
x

d
f

(x

)
|x x

|
e
it

, (7.38)
where
t

= t
k

x x

= t
1
c

x x

= t
R
c
. (7.39)
Finally, using Eq. (7.2 b) we re-express (x, t) in Eq. (7.36) as
(x, t) = C
+

+
(x, t) + C

(x, t) with

(x, t) = 4
_
d
3
x

f(x

, t

)
|x x

|
. (7.40)
This is the general solution of the inhomogeneous wave equation for the potential components , A
i
in the
Lorentz gauge. Note that the solution at time t at the eld point x is dependent on the behaviour of the source
at some other time t

= t

+
or t

at a location x

. The particular solution


+
(x, t) is referred to as the retarded
solution, because as shown by the argument of its source function f, an eect observed at the position x at
time t is caused by the action of the source at a distance R = |x x

| away, at an earlier or retarded time


t

= t

+
= t R/c t

ret
. The time dierence t t

ret
= R/c is the time of propagation of the wave disturbance
from the (retarded) source point to the observation point. The particular solution

(x, t), on the other hand,


is known as the advanced solution as it exhibits an eect observed at x at time t due the source at a distance
R = |x x

| away, at a later or advanced time t

= t

= t + R/c t

adv
.
Both the retarded and the advanced solutions are mathematically plausible. However, under the demand
of causality which requires that the potential at (t, x) is set up by the source at an earlier time t

ret
(at the
location x

), C

= 0, so that C
+
= 1/(4) by Eq. (7.33), whence
(x, t) =
_
d
3
x

f(x

, t

ret
)
|x x

|
, t

ret
= t
|x x

|
c
. (7.41)
That is, the viable solutions for the potentials (,

A) in the Lorentz gauge are
(x, t) =
_
d
3
x

_
(x

, t

)
|x x

|
_
ret
,

A(x, t) =
1
c
_
d
3
x

j(x

, t

)
|x x

|
_
ret
, (7.42)
where [ ]
ret
denotes the quantity within is evaluated at the retarded time t

= t

ret
= t R/c = t |x x

| /c.
II. In the Coulomb gauge:
Substituting the solution (7.25), of Eq. (7.16), in Eq. (7.15) we have

(x) =
_
d
3
x

(x

)
|x x

|
. (7.43)
Inserting this in Eq. (7.13 a), and nally using Eq. (7.13 b), we get
(x, t) =
_
d
3
x

(x

)
|x x

|
e
it
=
_
d
3
x

(x

, t)
|x x

|
. (7.44)
Thus the scalar potential is just the instantaneous Coulomb potential due to the charge density (x, t). Hence
the name Coulomb gauge.
The (three-)vector potential

A, however, satises an equation similar to that satised by the (three-)vector
potential

A in the Lorentz gauge. The only dierence being that in the Coulomb gauge,

A is sourced by the
transverse part

jjj

of the (three-)current density

jjj, rather than the latter as a whole. Hence the Coulomb gauge
is often called the Transverse gauge. In this gauge, therefore, the viable solution for

A is

A(x, t) =
1
c
_
d
3
x

(x

, t

)
|x x

|
_
ret
. (7.45)
42
7.2 Solutions for the four vector gauge potential the Covariant approach
7.2 Solutions for the four vector gauge potential the Covariant approach
Let us now focus on solving for the four vector gauge potential A

(x

) on the whole. As we see from the


evolution equation (6.48) in the Coulomb gauge, that we require an intrinsically non-covariant method of
solving for A

in the sense that we rst need to solve for one component of A

, viz., the scalar potential ,


and then insert this solution back into Eq. (6.48) and solve for A

. So the Coulomb gauge is not suitable


for obtaining the solution for A

in a covariant manner. In the Lorentz gauge, on the othe hand, the wave
equation satised by the four vector potential A

is given by

2
A

(x

) =
4
c
J

(x

) . (7.46)
Solving this doesnt require solving for any of the components of A

in the rst place, and therefore the Lorentz


gauge is suitable for obtaining A

in a covariant way (hence, perhaps the name Lorentz gauge).


7.2.1 Potential expansion in terms of Greens function, and Solutions
In order to solve the above equation (7.46), we consider the following ansatz:
A

(x

) =
4
c
_
d
4
x

D(x

, x

) J

(x

) , (7.47)
where D(x

, x

) is the four dimensional Greens function, or propagator, that satises the equation

2
x
D(x

, x

) =
4
(x

) , (7.48)

4
(x

) =
3
(xx

) (t t

) being the four dimensional delta function, and


2
x
is the DAlembertian with
spatial (Laplacian) part taken w.r.to the spatial components x
i
of the eld event position four vector x

.
Now, in absence of boundary surfaces:
D(x

, x

) = D(x

) , (7.49)
whence Eq. (7.48) reduces to

2
R
D(R

) =
4
(R

) , (7.50)
where
R

= x

. (7.51)
Let us expand D(R

) in a Fourier integral
D(R

) =
1
(2)
4
_
d
4
k D(k

) e
ikR

, (7.52)
where D(k

) is the Fourier transform of D(R

) in the k

-space:
D(k

) =
_
d
4
x D(R

) e
ikR

. (7.53)
Moreover, since the Fourier transform of the four dimensional delta function is unity (similar to the three
dimensional delta function), i.e.,

4
(R

) =
1
(2)
4
_
d
4
k e
ikR

, (7.54)
we have from Eq. (7.50)
_
d
4
k D(k

_
e
ikR

_
=
_
d
4
k e
ikR

= D(k

(k

) = 1 = D(k

) = D(k

) =
1
k

. (7.55)
Plugging this in Eq. (7.52):
D(R

) =
1
(2)
4
_
d
4
k
k

e
ikR

, (7.56)
i.e.,
D(R

) =
1
(2)
4
_
d
3
k e
i

R
I(k) , where I(k) =
_

dk
0
e
ik
0
R
0
k
2
0
k
2
. (7.57)
43
7 WAVE SOLUTIONS IN ELECTRODYNAMICS THE RETARDED POTENTIALS
Evaluation of I(k): The integrand having simple poles at k
0
= k, we let k
0
to be complex-valued, i.e.,
k
0
= (k
0
) + i (k
0
) , (7.58)
and perform a contour integration
I

(k) =
_

dk
0
e
ik
0
R
0
k
2
0
k
2
, (7.59)
over a suitably chosen closed contour in the (k
0
) (k
0
) plane.
Since both the poles (at k) lie on the (k
0
) axis, we can choose either the contour : (R, C
R
) or the
contour : (A, C
A
) shown in the gure below.
k
k
C
A
C
R
R
A
: (R, C
R
)
: (A, C
A
)
(k
0
)

(k
0
)
(k
0
)

(k
0
)
k
k
Figure 10: Illustration of the Contours.
R and A = Lines that graze the (k
0
) axis from the top and the bottom respectively.
C
R
and C
A
= Semicircular arcs that enclose the open contours R and A in the lower and the upper
half-planes respectively, at innity.
: (R, C
R
) = Closed contour formed by the line R and the curve C
R
.
: (A, C
A
) = Closed contour formed by the line A and the curve C
A
.
Now, the exponential in the integrand of I

(k), Eq. (7.59), is factorized as


e
ik
0
R
0
= e
i[(k
0
) + i (k
0
)]R
0
= e
i (k
0
)R
0
Oscillatory factor
e
(k
0
)R
0
Growth or Decay factor
. (7.60)
For R
0
> 0 : In the upper half-plane where (k
0
) > 0,
e
ik
0
R
0
as |(k
0
)| ,
i.e., the integrand of I

(k), Eq. (7.59), diverges (or, grows without limit) as we carry on increasing (k
0
) in
magnitude, so as to close the contour : (A, C
A
) at innity, on the upper half-plane. Therefore, the contour
: (A, C
A
) is not permissible for R
0
> 0.
In the lower half-plane where (k
0
) < 0,
e
ik
0
R
0
0 as |(k
0
)| ,
i.e., the integrand of I

(k), Eq. (7.59), is regular, with no contribution from the curve C


R
in the integral, as
we carry on increasing (k
0
) in magnitude, so as to close the contour : (R, C
R
) at innity, on the upper
half-plane. Therefore, the contour : (R, C
R
) is permissible for R
0
> 0.
44
7.2 Solutions for the four vector gauge potential the Covariant approach
Choosing the contour : (R, C
R
) we have
I

(k) =
_

dk
0
e
ik
0
R
0
k
2
0
k
2
+

:
0 _
C
R
dk
0
e
ik
0
R
0
k
2
0
k
2
. (7.61)
So,
I(k)

R
0
>0
= (R
0
)
_

dk
0
e
ik
0
R
0
k
2
0
k
2
= (R
0
)
_
:(R,C
R
)
dk
0
e
ik
0
R
0
k
2
0
k
2
. (7.62)
For R
0
< 0 : By a similar argument as above
I(k)

R
0
<0
= (R
0
)
_

dk
0
e
ik
0
R
0
k
2
0
k
2
= (R
0
)
_
:(A,C
A
)
dk
0
e
ik
0
R
0
k
2
0
k
2
. (7.63)
In the above, is the step function:
(y) =
_
1 for y > 0
0 otherwise
. (7.64)
Using the theorem of residues
I

(k) =
_

dk
0
e
ik
0
R
0
k
2
0
k
2
= 2i
_
Res
(k
0
=k)
_
e
ik
0
R
0
k
2
0
k
2
_
+ Res
(k
0
=k)
_
e
ik
0
R
0
k
2
0
k
2
__
= 2i
_
(k
0
k)
e
ik
0
R
0
k
2
0
k
2

k
0
=k
+ (k
0
+ k)
e
ik
0
R
0
k
2
0
k
2

k
0
=k
_
= 2i
_
e
ikR
0
2k
+
e
ikR
0
2k
_
=
i
k
_
e
ikR
0
e
ikR
0
_
=
i
k
2i sin(kR
0
) .
Or,
I

(k) =
_

dk
0
e
ik
0
R
0
k
2
0
k
2
=
2
k
sin(kR
0
) . (7.65)
Substituting this in Eq. (7.57):
D(R

R
0
>0
D
R
(R

) =
(R
0
)
(2)
3
_
d
3
k
k
e
i

R
sin(kR
0
) , (7.66)
D(R

R
0
<0
D
A
(R

) =
(R
0
)
(2)
3
_
d
3
k
k
e
i

R
sin(kR
0
) . (7.67)
Now,
D
R
(R

) =
(R
0
)
(2)
3
_

0
_

0
_
2
0
k
2
dk sin d d
e
i

R
k
sin(kR
0
)
=
(R
0
)
(2)
3
2
_

0
kdk sin(kR
0
)
_

0
d sin e
ikRcos
=
(R
0
)
4
2
_

0
kdk sin(kR
0
)
_
1
1
d e
ikR
= [Substituting: cos = ]
=
(R
0
)
4
2
_

0
kdk sin(kR
0
)
_
2 sin(kR)
kR
_
=
(R
0
)
2
2
R
_

0
dk sin(kR
0
) sin(kR)
=
(R
0
)
2
2
R

1
2
__

0
dk cos
_
k(R
0
R)


_

0
dk cos
_
k(R
0
+ R)

_
=
(R
0
)
4
2
R

_
_
(R
0
R)

:
Redundant, because of (R
0
)
(R
0
+ R)
_
_
.
45
7 WAVE SOLUTIONS IN ELECTRODYNAMICS THE RETARDED POTENTIALS
Hence,
D
R
(R

) =
(R
0
)
4R
(R
0
R) or, explictly, D
R
(x

) =
(x
0
x
0
)
4R
(x
0
x
0
R) . (7.68)
And, similarly,
D
A
(R

) =
(R
0
)
4R
(R
0
+ R) or, explictly, D
A
(x

) =
(x
0
x
0
)
4R
(x
0
x
0
+ R) . (7.69)
D
R
: Retarded Greens function, as the sources temporal position (t

) is always earlier than that (t) of


the observers, i.e., x
0
= ct

< x
0
= ct.
D
A
: Advanced Greens function, because the eld computed at time t is earlier than the time t

at
which the source is convoluted, i.e., x
0
= ct < x
0
= ct

.
The function D
A
is clearly acausal, and hence physically unacceptable.
7.2.2 Covariant formulation of Retarded and Advarnced Greens functions
In general D
R
and D
A
are not invariant under proper Lorentz transformations, apparently because of the theta
functions. In order to make the Greens functions Lorentz invariant, we use the following identity:
For a function f(x) with zeroes at x = x
(i)
,
(f(x)) =

(i)

_
x x
(i)
_
|f/x|
x=x
(i)
. (7.70)
Now consider,

_
R

2
_
=
_
R

_
=
_
(R
0
)
2
R
2
_
=
_
f(R
0
)
_
, (7.71)
where
f(R
0
) = (R
0
)
2
R
2
=
_
R
0
R
_ _
R
0
+ R
_
= Zeroes at R
0
= R . (7.72)
Applying the identity (7.70) :

_
R

2
_
=
_
f(R
0
)
_
=

_
R
0
R
_
|f/R
0
|
R
0
=R
+

_
R
0
+ R
_
|f/R
0
|
R
0
=R
=

_
R
0
R
_
|2R
0
|
R
0
=R
+

_
R
0
+ R
_
|2R
0
|
R
0
=R
,
i.e.,

_
R

2
_
=

_
R
0
R
_
+
_
R
0
+ R
_
2R
. (7.73)
Therefore, Eq. (7.68) can be expressed as:
D
R
(R

) =
(R
0
)
4R
_
2R
_
R

2
_

_
R
0
+ R
_
_
, (7.74)
and since (R
0
) eliminates (R
0
+ R), we have
D
R
(R

) =
(R
0
)
2

_
R

2
_
. (7.75)
Similarly, Eq. (7.69) is expressed as:
D
A
(R

) =
(R
0
)
2

_
R

2
_
. (7.76)
46
7.2 Solutions for the four vector gauge potential the Covariant approach
7.2.3 Retarded and Advanced Solutions for the Potential
Finally, in terms of the retarded and advanced Greens functions, the solution of the wave equation (7.46) is
expressed as:
A

(x

) = A

in
(x

) +
4
c
_
d
4
x

D
R
_
x

_
J

(x

) , (7.77)
or,
A

(x

) = A

out
(x

) +
4
c
_
d
4
x

D
A
_
x

_
J

(x

) . (7.78)
where A

in
and A

out
are the solutions of the source-free (i.e., homogeneous) wave equation
2
A

(x

) = 0.
The covariant forms of the Greens functions in Eqs. (7.75) and (7.76), as combinations of theta and
delta functions, imply that they are non-zero only on backward or forward lightcones, i.e., when the lightcone
condition, viz.,
R

2
= R

= (R
0
)
2
R
2
= 0 , (7.79)
is satised. Non-vanishing value at the backward lightcone is obviously for the retarded Greens function D
R
,
whereas non-vanishing value at the forward lightcone is for the advanced Greens function D
A
.
Taking the view that whatever is observed at a certain instant of time is due to what has happened on
or(and) in the past lightcone the causality argument and ignoring (for the time being) the solutions of
the homogeneous wave equation, we have the only acceptable solution for the four vector gauge potential:
A

(x

) =
4
c
_
d
4
x

D
R
_
x

_
J

(x

) =
1
c
_
d
4
x

(x
0
x
0
)
R

_
x
0
x
0
R
_
J

(x

) ,
i.e.,
A

(x

) =
1
2c
_
d
4
x

(x
0
x
0
) (x

2
) J

(x

) . (7.80)
47
8 RADIATION BY MOVING CHARGES
8 Radiation by Moving Charges
8.1 Scalar and vector potentials due to a point charge the Lienard-Wiechart potentials
Consider a particle of charge q in motion. If denotes the charges proper time, then let r

() be the charges
position four vector, and u

() = dr

/d be the charges four velocity.


The four vector potential due to the charged particle is given (in Lorentz gauge) by
A

(x

) =
4
c
_
d
4
x

D
R
_
x

_
J

(x

) , (8.1)
where D
R
_
x

_
is the retarded Greens function:
D
R
_
x

_
=
(x
0
x
0
)
4R

_
x
0
x
0
R
_
=
(x
0
x
0
)
2

_
x

2
_
, (8.2)
and J

(x

) is the charges four vector current (dened in 6.1, see Eq. (6.4)):
J

(x

) = q
_
c d u

()
4
_
x

()
_
. (8.3)
Substituting Eqs. (8.2) and (8.3) in Eq. (8.1) we have
A

(x

) = 4q
_
d
4
x

_
d u

()
(x
0
x
0
)
2

_
x

2
_

4
_
x

()
_
,
i.e.,
A

(x

) = 2q
_
d u

()
_
x
0
r
0
()
_

_
x

()
2
_
. (8.4)
The integral contributes only at =
0
, where
0
satises:
(1)
_
_
_x

(
0
)
_
_
_
2
= 0 = Lightcone condition , (8.5)
and (2) x
0
> r
0
(
0
) = Retardation requirement . (8.6)
The signicance of these are shown in the Fig. 11 below: The retarded Greens function is by denition non-zero
A
B
r

(
0
)
x

:= (x
0
, x(x
0
))
x
0
x
j
R
x
k
r

()
Figure 11: Illustration of the Lightcone condition and the Retardation requirement.
only on the backward lightcone of the observer. The charged particles world line r

() intersects the entire


(past and future) lightcone at points A and B, earlier and later than x
0
respectively. The lightcone condition
48
8.1 Scalar and vector potentials due to a point charge the Lienard-Wiechart potentials
implies that the earlier point A is given by r

(
0
), and this point is the only part of the path that contributes
to the electromagnetic eld at x

:=
_
x
0
, x(x
0
)
_
, as per the retardation requirement.
Let us now denote
f() =
_
_
_x

()
_
_
_
2
, (8.7)
which, as mentioned, has a zero of signicance only at =
0
. Applying the identity (7.70) shown in 7.2.2,
viz.,
(f(x)) =

(i)

_
x x
(i)
_
|f/x|
x=x
(i)
, where x
(i)
are the zeroes of f(x) ,
we have
(f()) =
_
x

()
2
_
=
(
0
)
|f/|
=
0
=
(
0
)
2 [u

() (x

())]
=
0
. (8.8)
Substituting in Eq. (8.4):
A

(x

) = q
_
d
u

() (
0
)
[u

() (x

())]
=
0
=
_
q u

()
u

() (x

())
_
=
0
. (8.9)
Again, the lightcone constraint [Eq. (8.5)] implies

x
0
r
0
(
0
)

2
|x r(
0
)|
2
= 0 = x
0
r
0
(
0
) = |x r(
0
)| |

R| = R . (8.10)
Therefore,
_
u

()
_
x

()
__
=
0
= u
0
_
x
0
r
0
(
0
)

u [x r(
0
)] = u
0
R u

R , (8.11)
i.e.,
_
u

()
_
x

()
__
=
0
= (u
0
u n) R , where n =

R
R
. (8.12)
If v be the velocity of the moving charge, then since
u
0
= u
0
= c and u = v = c

, where

=
v
c
and =
_
1 |

|
2
_
1/2
,
Eq. (8.12) reduces to
_
u

()
_
x

()
__
=
0
=
_
c
_
1

n
__
ret
, (8.13)
where the subscript ret implies that the quantity inside the brackets [ ] is to be evaluated at the retarded
time t

r
, given by
c t

r
= r
0
(
0
) = x
0
R = c t R . (8.14)
Plugging Eq. (8.13) back in Eq. (8.9) we get the nal expression for the four potential:
A

(x

) =
_
_
q u

c
_
1

n
_
R
_
_
ret
, (8.15)
the temporal and spatial components of which give respectively the scalar and three-vector potentials, commonly
referred to as the Lienard-Wiechart potentials due to the moving point charge:
(x, t) =
_
_
q u
0
c
_
1

n
_
R
_
_
ret
=
_
q
R
_
ret

A(x, t) =
_
_
q u
c
_
1

n
_
R
_
_
ret
=
_
q

R
_
ret
, where = 1

n . (8.16)
In what follows, we shall consider the retarded brackets [ ]
ret
to be implicit, and express the Lienard-Wiechart
potentials simply as
=
q
s
,

A =
q

s
, where s = R =
_
1

n
_
R = R

R

. (8.17)
49
8 RADIATION BY MOVING CHARGES
8.2 Retarded Electric and Magnetic elds due to a moving charge
In principle, one can work out the electromagnetic elds either from the form of the four potential A

given
by Eq. (8.15) or from the (component) Lienard-Wiechart potentials given by Eqs. (8.16). However, these
would require tedious, albeit straightforward, calculations all the way through. As an alternative, one can go
back to the expression (8.4) for A

, given as an integral over the charges proper time , and compute the
electromagnetic eld tensor F

therefrom, and nally work out its components, viz. the expressions for

E
and

B elds. This approach, although requires a lesser amount of work in getting upto F

, is not of much
help anyhow, because getting the expressions for

E and

B again involves tedious calculations.
In what follows, we shall focus on the rst approach, which is straightforward (but an essentially non-
covariant way) of obtaining

E and

B from the Lienard-Wiechart potentials (8.17). We will also outline briey
the alternative (covariant) approach of getting F

from the generic four potential expression (8.4), and the



E
and

B elds therefrom (skipping however the detailed steps, which would be left for the reader).
Now, it should be noted that, when an observer at a certain location, say x, measures the

E and

B elds at
some instant of time t, the elds are not due to the source charge at the same instant t, but at a former instant
t

r
= t R/c, where R/c is the time taken by the signal to travel from the (retarded) source location, say x

r
,
to the eld location x. As the potentials and

A are expressed in terms of the retarded time t

r
(by virtue
of the retarded bracket [ ]
ret
, that is implicit in the expressions (8.17)), we need to nd the relations between
the time derivatives and the (spatial) gradients in two dierent frames of the time one of the (retarded)
sources, i.e. t

r
, and the other of the observers, i.e. t.
8.2.1 Time and space derivatives in the frames of the retarded source and the observer
(b) (a)

R
vvv
x
vvv dt

r
vvv
O (origin)
S

(t

r
)
x

r
x

S(t)
S(t

r
+ dt

r
)

R+ d

R
P(t)
S

(t

r
)
P(t

r
+ dt

r
)

R(t

r
, x)
Figure 12: Correlation between time and space derivatives with respect to the eld point and the retarded source point.
Refer rst to the Fig. 12 a. Here,
P(t, x) = The eld point.
S(t, x

) = Location of the source when the elds are computed.


S

(t

r
, x

r
) = Actual (retarded) location of the source which causes the elds.
Since, the position four vector of the retarded source r

:= (ct

r
, x

r
), with t

r
= t R/c, R being the distance
between the retarded source point S

and the eld point P, we have


R = c
_
t t

r
_
=
R
t
= c
_
1
t

r
t
_
. (8.18)
50
8.2 Retarded Electric and Magnetic elds due to a moving charge
Now, refer to the Fig. 12 b. Here we suppose the interval t t

r
= dt

r
, in which time

R changes to

R+ d

R,
and the source is displaced from its retarded position S

(t

r
, x

r
) to the position S(t

r
+ dt

r
, x

) by an amount
v dt

r
(with velocity v). Then

R = v dt

r
+

R+ d

R =

R
t

x
= v . (8.19)
Also,
R
2
=

R

R = 2R
R
t

r
= 2

R
t

r
=
R
t

r
= n v , (8.20)
using Eq. (8.19). Hence,
R
t
=
R
t

r
t

r
t
= ( n v)
t

r
t
. (8.21)
Comparing Eq. (8.18):
t

r
t
=
1
1 n

=

t
=
1

r
. (8.22)
Again, since R = R(t

r
, x) we have
dR =
R
t

x
dt

r
+
R
x
i

r
dx
i
=
R
x
i

t
=
_
R
t

r
_
t

r
x
i

t
+
R
x
i

r
,
i.e.,
R =
_
R
t

r
_
t

r
+

r
R , where
_

r
_
i


x
i

r
. (8.23)
But
R
2
=

x x

2
=

i
_
x
i
x
i
r
_
2
= 2R
R
x
i

r
= 2
_
x
i
x
i
r
_

r
,
as x
i
r
( i) are held xed at t

r
= constant. So,

r
R =

R
R
= n . (8.24)
Moreover,
R = c
_
t t

r
_
= R = c t

r
. (8.25)
Substituting Eqs. (8.24) and (8.25) in Eq. (8.23), and using Eq. (8.20), we have
c t

r
= n v t

r
+ n = t

r
=
n
c
_
1

n
_ =
n
c
. (8.26)
Plugging this again in Eq. (8.23), we nally obtain
=

r

n
c

r
. (8.27)
8.2.2 Electric and Magnetic elds from the Lienard-Wiechart potentials
The electric and magnetic elds can be worked out straightaway from the Lienard-Wiechart potentials (8.17)
as follows:

E =
1
c
A
t
=
_
q
s
_

1
c

t
_
q

s
_
= q
_
s
s
2

1
c

t
_

s
__

B =

A =
_
q

s
_
= q
_

s
+

s
s
2
_ . (8.28)
where (in the second line, i.e., for

B) we have used the vector identity:
(

U) =

U

U ,
_
: scalar ,

U : vector
_
. (8.29)
51
8 RADIATION BY MOVING CHARGES
Now, the retarded brackets [ ]
ret
being implicit, we need to work out the expressions for

E and

B in terms of
time derivatives and (spatial) gradients dened in the retarded source frame, i.e. /t

r
and

r
. Substituting
therefore the relations (8.22) and (8.27) in the above expression for

E in Eq. (8.28), we have

E
q
=
1
s
2
_

s
n
c
s
t

1
c

s
_
=

s
s
2

( n

) s
c s
2

c s
, (8.30)
where, for brevity, we have omitted the subscript r for retardation, i.e., from now on we will use t

for t

r
,
/t

for /t

r
and

for

r
. The overhead dot {} implies /t

/t

r
. From the expression of s, given in
Eq. (8.17), we evaluate:

s =

_
R

R

_
= n +
1
c

R
t

_ _
as

R = n,

=
v
c
=
1
c

R
t

_
= n +
1
c

_
R
R
t

_
= n +
1
c
_

R
_
R
t

_
as

R

R
t

=
R
t

, and

R =

t

R = 0
_
= n +
1
c
R
t

n = n +
1
c

R
t

= n +
1
c
(v)
i.e.,

s = n

. (8.31)
and
s
s
t

=

t

_
R

R

_
=
R
t


R
t

= n v + v

_
as
R
t

= n v ,

R
t

= v
_
i.e.,
s = c
_
_
n


2
+

c
_
_
. (8.32)
Substituting Eqs. (8.31) and (8.32) in Eq. (8.30), and using the relation = s/R = 1 n

, we have

E
q
=
n

s
2

_
n

s
3
_
R
_
_
n


2
+

c
_
_

c s
2
=
n

s
3
_
s +

R

R
2
_
+
1
cs
3
__

RR

_

R


_
R

R

_
R

_
. (8.33)
Since, s = R

R

, rst term on the r.h.s. reduces to:


n

s
3
_
R R
2
_
,
whereas, using the bac-cab rule of vector algebra, viz.,

A
_

B

C
_
=

B
_

C
_
+

C
_

B
_
, (8.34)
second term on the r.h.s. simplies to: 1
c s
3

R
__

RR

_
.
Therefore, the nal expression for the electric eld is

E =
q R
s
3
_
1
2
_
_
n

_
+
q R
2
c s
3
n
__
n

_
. (8.35)
As to the magnetic eld, the rst term in the expression for

B in (8.28) involves

, which is can be reduced


as follows:
_

_
i
=
ijk

x
j

k
=
ijk
t

x
j

k
_
as
i
=
i
(t

) i

=
ijk
n
j
c

k
=
1
c
_

n
_
i
_
as t

=
n
c

t

x
i
=
n
i
c
_
,
i.e.,

n
c
. (8.36)
52
8.2 Retarded Electric and Magnetic elds due to a moving charge
Substituting now the relations (8.22) and (8.27), and the result (8.36), in the expression for

B in Eq. (8.28),
we have

B
q
=

n
c s
+
1
s
2

s
n
c
s
_
, (8.37)
which when simplied using the expressions of s = R [Eq. (8.17)],

s [Eq. (8.31)], and s [Eq. (8.32), yields


the nal expression for the magnetic eld as

B =
q R
s
3
_
1
2
_

n +
q R
2
c s
3
n
_
n
__
n

__
= n

E . (8.38)
Returning back to the notation of the retarded brackets, the above expressions (8.35) and (8.38) for the electric
and the magnetic elds are given as

E =
_
q

3
R
2
_
1
2
_
_
n

_
+
q
c
3
R
2
n
__
n

__
ret
,

B =
_
n

E
_
ret
, (8.39)
where the overhead dot {} now denotes /t (since the subscript ret already implies that everything is
determined at the retarded time t

r
= t R/c).
8.2.3 Electromagnetic eld tensor and components from generic four potential expression
Let us go back to the generic expression (8.4) for the four potential due to a moving point charge, given as an
integral over the charges proper time. Since the electromagnetic eld tensor is given by F

,
we seek to determine rst the quantity

(x

) = 2q

_
d u

()
_
x
0
r
0
()
_

_
x

()
2
_
= 2q
_
d u

()
_

_
x
0
r
0
()
_

_
x

()
2
_
+
_
x

()
2
_

_
x
0
r
0
()
_
_
. (8.40)
As the derivative of the Theta function w.r.to its argument is equal to the delta function of the same argument

,
i.e., d
dx
(x) = (x) , (8.41)
we can express

_
x
0
r
0
()
_
=

0

_
x
0
r
0
()
_
. (8.42)
Moreover, since

_
x

()
2
_
=
_

x
0
r
0
()

2
|x r()|
2
_
=
_

x
0
r
0
()

2
R
2
_
,
the delta function in Eq. (8.42) reduces
_
x

()
2
_
in the second term on the r.h.s. of Eq. (8.40) to
(R
2
), which is zero for R = 0. So, considering R = 0 we write the above Eq. (8.40) as

(x

) = 2q
_
d u

()
_
x
0
r
0
()
_

_
x

()
2
_
. (8.43)
Denoting again
f() =
_
_
_x

()
_
_
_
2
[cf. Eq. (8.7)] ,
we have

_
x

()
2
_

(f()) =

f
d
df

d
d
(f())
=
x

()
[x

()] u

()
d
d

_
x

()
2
_
, (8.44)
as

f() = 2 [x

()] and
df()
d
= 2
_
x

()
_
dr

()
d
= 2
_
x

()
_
u

() . (8.45)

Proof: We have

a
a

d(x)
dx
(x)

dx = (a) + (a) 1 = 1 + 0 1 = 0 .
Now, the limits a being arbitrary, one can always equate the integrand to zero. Hence the proof.
53
8 RADIATION BY MOVING CHARGES
Substituing Eq. (8.44) in Eq. (8.43), we get

(x

) = 2q
_
d
[x

()] u

()
[x

()] u

()

_
x
0
r
0
()
_

d
d

_
x

()
2
_
. (8.46)
Performing an integration by parts

(x

) = 2q
_
(x

) u

(x

) u

:
0 , at the boundaries

_
x
0
r
0
_

_
x

2
_
+2q
_
d
d
d
__
(x

) u

(x

) u

_
x
0
r
0
_
_

_
x

2
_
= 2q
_
d
d
d
_
(x

) u

(x

) u

_
x
0
r
0
_

_
x

2
_
+2q
_
d
d
_
x
0
r
0
_
d
_
(x

) u

(x

) u

_
x

2
_
. (8.47)
As the derivative of the Theta function again takes out the delta function for R = 0, the second term on the
r.h.s. vanishes, and we have

(x

) = 2q
_
d
d
d
_
[x

()] u

()
[x

()] u

()
_

_
x
0
r
0
()
_

_
x

()
2
_
. (8.48)
The integrand is now similar to that in Eq. (8.4), with u

() being replaced by the derivative term in Eq.


(8.48). The end result should also be similar [to Eq. (8.9)]:

(x

) =
_
q
[x

()] u

()

d
d
_
[x

()] u

()
[x

()] u

()
__
ret
, (8.49)
whence
F

(x

) =
_
q
[x

()] u

()

d
d
_
[x

()] u

()
_
x

()

()
[x

()] u

()
__
ret
. (8.50)
Although this expression is manifestly covariant, it is often useful to reduce this and get the expressions for
the (retarded) electric and magnetic eld vectors

E and

B. This is however a tedious (albeit straightforward)
exercise. One needs to make use of the following relations:
x

:= (R, R n) , u

:= c
_
1,

_
, and
du

d
:=
2
c
_

+
2

__
, (8.51)
to work out the derivative factor in Eq. (8.50), and after a series of simplications nally determine F
0i
and
F
ij
. The result would be given by the equations (8.39), which we leave for the reader to verify.
8.2.4 Acceleration-dependent parts of the elds and Electromagnetic Radiation
Let us now decompose the electric and magnetic elds, given by Eqs. (8.39), into the acceleration-dependent
and the acceleration-independent parts (denoted by subscripts a and v respectively):

E =

E
a
+

E
v
;

B =

B
a
+

B
v
,

E
a
=
q
cR
n
__
n

_
_
1 n

_
3
;

E
v
=
q
R
2
_
1
2
_
_
n

_
_
1 n

_
3
,

B
a
= n

E
a
;

B
v
= n

E
v
. (8.52)
For convenience, we have omitted the retarded brackets [ ]
ret
, considering them to be implicit.
Now, the acceleration-dependent elds satisfy:
(i)

B
a

E
a
= 0 = n

B
a
= n

E
a
and (ii) |

B
a
| |

E
a
|
1
R
, (8.53)
whereas, the acceleration-dependent elds satisfy:
(i)

B
v

E
v
= 0 = n

B
v
but n

E
v
= 0 and (ii) |

B
v
| |

E
v
|
1
R
2
. (8.54)
54
8.3 Power radiated by an accelerated charge
The acceleration-dependent terms are the only ones responsible for electromagnetic radiation. This is evident
from the following:


E
a
,

B
a
form an orthogonal set including the propagation vector, which in this case is n =

R/R.

E
v
,

B
v
and n, however, do not form an orthogonal set.
For the acceleration-dependent elds, the magnitude of the Poynting vector is given by
|

S|
a
=
c
4

E
a

B
a

=
c
4

E
a

1
R
2
, (8.55)
so that the amount of energy crossing a spherical closed surface of radius R is equal to
|

S|
a
4R
2

1
R
2
4R
2
1 (indepedent of R) . (8.56)
Therefore the total energy remains the same as the radius R is varied Conservation of Energy.
For the acceleration-independent elds, the magnitude of the Poynting vector is
|

S|
v
=
c
4

E
v

B
v

=
c
4

E
v

1
R
4
, (8.57)
so that |

S|
v
integrated over a closed spherical surface of radius R is
|

S|
v
4R
2

1
R
4
4R
2

1
R
2
. (8.58)
As R , the total energy goes to zero not in support of Radiation.
8.3 Power radiated by an accelerated charge
8.3.1 Non-relativistic radiant power formula the Larmor result
For non-relativistic motion, = v/c 1, and one can approximate the acceleration-dependent elds [cf Eqs.
(8.52)] as:

E
a

q
cR
n
_
n

_
and

B
a
= n

E
a
. (8.59)
The instantaneous energy ux is given by the Poynting vector

S =
c
4

E
a

B
a
=
c
4

E
a

2
n =
c
4
q
2
c
2
R
2

n
_
n

2
n . (8.60)
(b)
(a)

R
d
P
d
S

P
y
z
x
d
n

Figure 13: Power radiated by an accelerated point charge.


Amount of radiation energy crossing unit area in unit time is therefore
dP
d
= |

S| =
q
2

2
4cR
2
sin
2
, (8.61)
where is the angle between the observation direction n and the acceleration direction

(see Fig. 13).


55
8 RADIATION BY MOVING CHARGES
Now, d = R
2
d, where d is the solid angle subtended by the innitesimal area element d at the retarded
source location S

. Hence, the power radiated per unit solid angle is given by


dP
d
= |

S| R
2
=
q
2

2
4c
sin
2
, (8.62)
Choosing

to be in the z-direction, we have d = sin dd, whence the instantaneous power radiated is
evaluated as
P =
q
2

2
4c
_

0
_
2
0
sin
3
dd =
q
2

2
4c
2
4
3
= P =
2
3
q
2

2
c
=
2
3
q
2
v
2
c
3
. (8.63)
This is the Larmor formula for the total power radiated by a non-relativistic accelerated charge. It is also
convenient to express this as
P =
2
3
q
2
m
2
0
c
3
_
dp
NR
dt

dp
NR
dt
_
, (8.64)
where p
NR
= m
0

v is the non-relativistic three-momentum, m


0
being the rest mass.
8.3.2 Relativistic radiant power formula the Lienard result
For relativistic motion, one may generalize the above Larmor expression (8.64) for the total power radiated P,
taking into consideration the following criteria:
P should be a Lorentz scalar,
P should reduce to Eq. (8.63) [or (8.64)] in the limit of small velocities = v/c 0.
A suggested generalization of Eq. (8.64) is
P =
2
3
q
2
m
2
0
c
3
_

dP

d
dP

d
_
=
2
3
q
2
m
2
0
c
3
_

1
c
2
_
dE
d
_
2
+
_
dp
d
_
2
_
, (8.65)
where is the charges proper time [d = dt/ , = (1
2
)
1/2
], and P

:= (E, p) is the four momentum


[p = m
0
v being the relativistic three-momentum and E = m
0
c
2
is the total relativistic energy]. The above
expression (8.65) is, by construction, Lorentz invariant. To see whether it reduces to Eq. (8.63) [or (8.64)] in
the limit = v/c 0, we proceed as follows:
We have
E
2
= p
2
c
2
+ m
2
0
c
4
= 2E
dE
d
= 2pc
2
dp
d
=
dE
d
=
p
E
c
2
dp
d
= v
dp
d
, (8.66)
as,
p
E
=
m
0
v
m
0
c
2
=
v
c
2
. (8.67)
Therefore, Eq. (8.65) reduces to
P =
2
3
q
2
m
2
0
c
3
_
_
dp
d
_
2

2
_
dp
d
_
2
_
. (8.68)
In the non-relativistic limit ( = v/c 0):
d dt , p p
NR
, whence P
2
3
q
2
m
2
0
c
3
_
dp
NR
dt
_
2
: the Larmor formula . (8.69)
So, Eq. (8.65) [or (8.68)] indeed gives the total power radiated by a relativistic accelerated charge.
Now, since p = m
0
v = m
0
c

, we have
dp
d
=
dp
dt
= m
0
c
d
dt
_

_
= m
0
c
_

_
. (8.70)
Substituting

d
dt
=
d
dt
_
1

_
1/2
=
3

=
3


cos , (8.71)
where is the angle between

and

, and squaring, one obtains (after some algebraic simplications):


_
dp
d
_
2
=
6
m
2
0
c
2

2
_
1
2
sin
2

_
+
2

2
m
2
0
c
2
. (8.72)
56
8.3 Power radiated by an accelerated charge
Moreover,
E = m
0
c
2
=
dE
d
=
dE
dt
= m
0
c
2
. (8.73)
Substituting Eqs. (8.72) and (8.73) in Eq. (8.65) and simplifying, one nally gets
P =
2
3
q
2
c

6
_
_

_
2

_
2
_
. (8.74)
This is the Lienards expression for the total power radiated by an accelerated charge in relativistic motion.
57

Das könnte Ihnen auch gefallen