Chapter 7.

Covariant Formulation of Electrodynamics

Notes:

Most of the material presented in this chapter is taken from Jackson, Chap. 11, and
Rybicki and Lightman, Chap. 4.
Starting with this chapter, we will be using Gaussian units for the Maxwell equations
and other related mathematical expressions.
In this chapter, Latin indices are used for space coordinates only (e.g., i = 1, 2, 3 ,
etc.), while Greek indices are for space-time coordinates (e.g., ! = 0,1, 2, 3 , etc.).

7.1 The Galilean Transformation

Within the framework of Newtonian mechanics, it seems natural to expect that the
velocity of an object as seen by observers at rest in different inertial frames will differ
depending on the relative velocity of their respective frame. For example, if a particle of
mass m has a velocity u! relative to an observer who is at rest an inertial frame K ! ,
while this frame is moving with at a constant velocity v as seen by another observer at
rest in another inertial frame K , then we would expect that the velocity u of the particle
as measured in K to be
u = u! + v.

(7.1)

That is, it would seem reasonable to expect that velocities should be added when
transforming from one inertial frame to another; such a transformation is called a
Galilean transformation. In fact, it is not an exaggeration to say that this fact is at the
heart of Newtons Second Law. Indeed, if we write the mathematical form of the Second
Law in frame K we have

d 2x
du
=m
2
dt
dt
d ( u! + v )
=m
,
dt

F=m

(7.2)

but since v is constant

du!
= F!,
dt

(7.3)

d 2x
d 2 x!
=
m
.
dt 2
dt 2

(7.4)

F=m

or

The result expressed through equation (7.4) is a statement of the covariance of Newtons
Second Law under a Galilean transformation. More precisely, the Second Law retains the
145

same mathematical form no matter which inertial frame is used to express it, as long as
velocities transform according to the simple addition rule stated in equation (7.1).
It is also important to realize that implicit to this derivation was the fact that everywhere
it was assumed that, although velocities can change from one inertial frame to another,
time proceeds independently of which reference frame is used. That is, if t and t ! are the
time in K and K ! , respectively, and if they are synchronized initially such that t = t ! = 0
then

t = t !,

(7.5)

at all times.

Although the concepts of Galilean transformation (i.e., equation (7.1)) and absolute time
(i.e., equation (7.5)), and therefore Newtons Second Law, are valid for a vast domain of
applications, they were eventually found to be inadequate for system where velocities
approach the speed of light or for phenomenon that are electrodynamic in nature (i.e.,
those studied using Maxwells equations). To correctly account for a larger proportion of
physical systems, we must replace the Galilean by the Lorentz transformation, abandon
the notion of absolute time, and replace the formalism of Newtonian mechanics by that of
special relativity.

7.2 The Lorentz Transformation

The special theory of relativity is based on two fundamental postulates:
I. The laws of nature are the same in two frames of reference in uniform relative
motion with no rotation.
II. The speed of light is finite and independent of the motion of its source in any
frame of reference.
We consider two inertial frames K and K ! , as shown in Figure 7-1, with relative uniform
velocity v along their respective x-axis , the origins of which are assumed to coincide at
times t = t ! = 0 . If a light source at the origin and at rest in K emits a short pulse at time
t = 0 , then an observer at rest in any of the two frames of reference will (according to
postulate II) see a shell of radiation centered on the origin expanding at the speed of light
c . That is, a shell of radiation is observed independent of the frame of the inertial
observer. Because this result obviously contradicts our previous (Galilean) assumption
that velocities are additive, it forces us to reconsider our notions of space and time, and
view them as quantities peculiar to each frame of reference and not universal. Therefore,
we have for the equations of the expanding sphere in the frames

)
) = 0,

c 2t 2 ! x 2 + y 2 + z 2 = 0

c 2 t " 2 ! x " 2 + y" 2 + z " 2

(7.6)

or alternatively,

c 2 t 2 ! x 2 + y 2 + z 2 = c 2 t " 2 ! x " 2 + y" 2 + z " 2 .

146

(7.7)

Figure 7-1 - Two inertial frames with a relative velocity v along the x-axis .
After consideration of the homogeneity and isotropic nature of space-time, it can be
shown that equation (7.7) applies in general, i.e., not only to light propagation. The
coordinate transformation that satisfies this condition, and the postulates of special
relativity, is the so-called Lorentz Transformation.
We can provide a mathematical derivation of the Lorentz transformation for the system
shown in Figure 7-1 as follows (please note that a much more thorough and satisfying
derivation will be found, by the more adventurous reader, in the fourth problem list).
Because of the homogeneity of space-time, we will assume that the different components
x and x"! of the two frames are linked by a set of linear relations. For example, we write

x0! = Ax0 + Bx1 + Cx2 + Dx3

x1! = Ex0 + Fx1 + Gx2 + Hx3 ,

(7.8)

and similar equations for x2! and x3! , where we introduced the following commonly used
notation x0 = ct, x1 = x, x2 = y, and x3 = z . However, since the two inertial frames
exhibit a relative motion only along the x-axis , we will further assume that the directions
perpendicular to the direction of motion are the same for both systems with

x2! = x2
x3! = x3 .

(7.9)

Furthermore, because we consider that these perpendicular directions should be

unchanged by the relative motion, and that at low velocity (i.e., when v ! c ) we must
have

x1! = x1 " vt,

(7.10)

we will also assume that the transformations do not mix the parallel and perpendicular
components. That is, we set C = D = G = H = 0 and simplify equations (7.8) to

147

x0! = Ax0 + Bx1

x1! = Ex0 + Fx1 .

(7.11)

Therefore, we only need to solve for the relationship between ( x0! , x1! ) and ( x0 , x1 ) . To do
so, we first consider a particle that is at rest at the origin of the referential K such that
x1 = 0 and its velocity as seen by an observer at rest in K ! is !v . Using equations (7.11)
we find that

x1!
v E
=" = .
x0!
c A

(7.12)

Second, we consider a particle at rest at the origin of K ! such that now x1! = 0 and its
velocity as seen in K is v . This time we find from equations (7.11) that

x1 v
E
= =! ,
x0 c
F

(7.13)

and the combination of equations (7.12) and (7.13) shows that A = F ; we rewrite
equations (7.11) as

B %
"
x0! = A $ x0 + x1 '
#
A &
v %
"
x1! = A $ x1 ( x0 ' .
#
c &

(7.14)

Third, we note that from Postulate II the propagation of a light pulse must happen at the
speed of light in both inertial frames. We then set x0! = x1! and x0 = x1 in equations (7.14)
to find that
B
v
=! ,
A
c

(7.15)

and

v &
#
x0! = A % x0 " x1 (
$
c '
v &
#
x1! = A % x1 " x0 ( .
$
c '

(7.16)

Evidently, we could have instead proceeded by first expressing the x as a function of

the x"! with

148

x0 = A!x0! + B!x1!
x1 = E !x0! + F !x1!,

(7.17)

from which, going through the same process as above, we would have found that

v %
"
x0 = A! $ x0! + x1! '
#
c &
v %
"
x1 = A! $ x1! + x0! ' .
#
c &

(7.18)

Not surprisingly, equations (7.18) are similar in form to equations (7.16) with v replaced
by !v . The first postulate of special relativity tells us, however, that the laws of physics
must be independent of the inertial frame. This implies that
A = A!

(7.19)

(this result can also be verified by inserting equations (7.16) and (7.18) into equation
(7.7), as this will yield A 2 = A! 2 ). If we insert equations (7.18) into equations (7.16) we
find that
( " v% 2 +
A = *1 ! $ ' *) # c & -,

!1 2

(7.20)

we can finally write the Lorentz transformation, in its usual form for the problem at hand,
as

x0! = " ( x0 # $ x1 )
x1! = " ( x1 # $ x0 )

x2! = x2

(7.21)

x3! = x3 ,
with

v
c
!= !

!=

" = (1 # ! 2 )

(7.22)
#1 2

The inverse transformation is easily found by swapping the two sets of coordinates, and
by changing the sign of the velocity. We then get
149

x0 = ! ( x0" + # x1" )
x1 = ! ( x1" + # x0" )

(7.23)

x2 = x2"
x3 = x3" .

Alternatively, it should be noted that equations (7.21) could be expressed with a single
matrix equation relating the coordinates of the two inertial frames
!
!
x! = Lx ( " ) x,

(7.24)

where the arrow is used for space-time vectors and distinguishes them from ordinary
space vectors. More explicitly, this matrix equation is written as
" x0! % " (
$ x ! ' $ )(*
$ 1' = $
$ x2! ' $ 0
$# x ! '& $# 0
3

)(*
(
0
0

0 0 % " x0 %
0 0 ' $ x1 '
'$ '.
1 0 ' $ x2 '
0 1 '& $# x '&

(7.25)

It is easy to verify that the inverse of the matrix present in equation (7.25) is that which
can similarly be obtained from equations (7.23). Although these transformations apply to
frames that have their respective system of axes aligned with each other, Lorentz
transformations between two arbitrarily oriented systems with a general relative velocity
v can be deduced by starting with equations (7.21) (or (7.23)) and apply the needed
spatial rotations.
For example, to find the Lorentz transformation L (! ) applicable when the axes for
K and K ! remain aligned to each other, but the relative velocity v is allowed to take on
an arbitrary orientation, we could first make a rotation R that will bring the x1 -axis
parallel to the orientation of the velocity vector, then follow this with the basic Lorentz
transformation Lx ( ! ) defined by equations (7.25), and finish by applying the inverse
rotation R !1 . That is,
L (! ) = R "1Lx ( ! ) R.

(7.26)

When these operations are performed, one then finds (see the fourth problem list)
x0! = " ( x0 # $ % x )
x! = x +

(" # 1)
$2

($ % x ) $ # " $x0 .

150

(7.27)

Since the Lorentz transformation mixes space and time coordinates between the two
frames, we cannot arbitrarily dissociate the two types of coordinates. The basic unit in
space-time is now an event, which is specified by a location in space and time given in
relation to any system of reference. This mixture of space and time makes it evident that
we must abandon our cherished and intuitive notion of absolute time.
7.2.1 Four-vectors
Since we saw in the last section that the Lorentz transformation can simply be expressed
as a matrix relating the coordinates of two frames of reference (see equation (7.25)), it
!
!
was natural to define vectors x and x! to represent these coordinates. Because these
vectors have for components, they are called four-vectors. Just like the coordinate
four!
vector contains a four coordinates ( x0 , x1 , x2 , x3 ) , an arbitrary four-vector A has four
!
components ( A0 , A1 , A2 , A3 ) . Moreover, the invariance of the four-vector x expressed
through equation (7.7) is also applicable to any other four-vectors. That is, a four-vector
must obey the following relation

A02 ! (A12 + A22 + A32 ) = A0" 2 ! ( A1" 2 + A2" 2 + A3" 2 ).

(7.28)

7.2.2 Proper Time and Time Dilation

We define the infinitesimal invariant ds associated with the infinitesimal coordinate
!
four-vector dx as (see equation (7.7))

ds 2 = dx02 ! dx12 + dx22 + dx32

= ( cdt ) ! dx 2 + dy 2 + dz 2 .
2

(7.29)

If a inertial frame K ! is moving relative to another one ( K ) with a velocity v such that
!
dx = v dt (with dx the spatial part of dx ), then equation (7.29) can be written as
ds 2 = c 2 dt 2 ! dx

(7.30)

= c 2 dt 2 1 ! " 2 .

For an observer at rest in K ! , however, it must be (by definition) that dx! = 0 , and from
equation (7.30)

ds 2 = c 2 dt ! 2 = c 2 dt 2 1 " # 2 .

(7.31)

We define the proper time ! with

ds = cd!

151

(7.32)

Comparison of equations (7.31) with (7.32) shows that an element of proper time is the
actual time interval measured with a clock at rest in a system. The elements of time
elapsed in each reference frame (and measured by observers at rest in each of them) are
related through

d! = dt 1 " # 2 =

dt
.
$

(7.33)

Since ! > 1 , then to an observer at rest in K time appears to be passing by more slowly
in K ! . More precisely, a proper time interval ! 2 " ! 1 will be seen in K as lasting
t 2 ! t1 =

"2

"1

d"
1 ! # 2 (" )

"2

$ % (" ) d" ,
"1

(7.34)

where we assumed that the velocity could be changing with (proper) time. This
phenomenon is called time dilation.
7.2.3 Length Contraction
Let us suppose that a rod of length L0 is kept at rest in K ! , and laid down in the x !
direction. We now inquire as to what will be the length for this rod when measured by an
observer in K . One important thing to realize is that the length of the rod as measured in
K will be L = x1 ( b ) ! x1 ( a ) , where x1 ( a ) and x1 ( b ) are the position of the ends of the
rod at the same time t when the measurement occurs. More precisely, t is the time
coordinate associated with K and no other frame. So, from equations (7.21) we have
L0 ! x1" ( b ) # x1" ( a )

= $ &' x1 ( b ) # % x0 () # $ &' x1 ( a ) # % x0 ()
= $ &' x1 ( b ) # x1 ( a ) () ,

(7.35)

or alternatively

L=

L0
.
!

(7.36)

Therefore, to an observer in K the rod appears to be smaller than the length it has when
measured at rest (i.e., in K ! ). This apparently peculiar result is just a consequence of the
fact that, in special relativity, events that are simultaneous (i.e., happen at the same time)
in one reference frame will not be in another (if the two frames are moving relative to one
another). The concept of simultaneity must be abandoned in special relativity.

152

7.2.4 Relativistic Doppler Shift

We saw in section 7.2 that the space-time length element ds is invariant as one goes
from one inertial frame to the next. Similarly, since the number of crests in a wave train
can be, in principle, counted, it must be a relativistic invariant. Then, the same must be
true of the phase of a plane wave ! . Mathematically, if ! and k are, respectively, the
angular frequency and the wave vector of a plane wave as measured in K , then the
invariance of the phase means that

! = " t # k $ x = " %t % # k% $ x%.

(7.37)

Since ! = ck and with k = kn , we can write

! ( ct " n # x ) = ! $ ( ct $ " n$ # x$ ) ,

(7.38)

and upon using equations (7.27), while breaking all vectors in parts parallel and
perpendicular to the velocity (i.e., x = x! + x ! ), we have

! ( ct " n! # x! " n $ # x $ ) = ! % ()& ( ct " ' # x! ) " & n%! # ( x! "' ct ) " n%$ # x $ *+

= ! % ()& ct 1 + n%! # ' " & x! # n%! + ' " n%$ # x $ *+ .

(7.39)

If this equation is to hold at all times t and position x , then the different coefficients for
t, x! , and x ! on either side must be equal. That is,

! = "! # (1 + n#! $ % )
!#)
&
k! = " ( k!# + % +
'
c*

(7.40)

k , = k#, .
The first of equations (7.40) is that for the Doppler shift. It is important to note that the
term in parentheses is the same as the one that appears in the non-relativistic version of
the formula, but that there will also be a Doppler shift in the relativistic case (although of
second order) even if the direction of propagation of the wave is perpendicular to velocity
vector. This is because of the presence of the ! factor on the right-hand side of this
equation. Equally important is the fact that,
! as evident from our analysis, the frequency
and the wave vector form a four-vector k (with k0 = ! c ), and that the phase ! is an
invariant resulting from the following scalar product

! !
k ! x = ",
from equation (7.37). The second of equations (7.40) can be transformed to give

153

(7.41)

k"
# % cos (! " ) + $ '(
k &
)"
= # %& cos (! " ) + $ '(
)
cos (! " ) + $
=
,
1 + $ cos (! " )

cos (! ) =

(7.42)

where the first of equations (7.40) was used, and ! and ! " are the angles of k and k!
relative to ! . In turn, the last two of equations (7.40) can be combined as follows

k!
k!"
=
k!
# k!" + $ k0"

sin (% " )
tan (% ) =
.
# &' cos (% " ) + $ ()

(7.43)

Finally, inverting equation (7.42) to get cos (" ! ) as a function of cos (! ) and sin (! ) ,
substituting it into the first of equations (7.40) will easily lead to the following set of
equations

" ! = #" (1 $ n! % & )

k!! = # k! $ & k0

(7.44)

k!' = k ' .
This result is equivalent to that of, and could have been obtained in the same manner as
was done for, equations (7.40). It should also be noted that for a plane wave

! !
k ! k = 0.

(7.45)

7.2.5 The Transformation of Velocities

We now endeavor to find out what will the velocity u of a particle as measured in K be
if it has a velocity u! in K ! . From equations (7.27), we can write

! ( dx!" + vdt " )

u" + v
= !
v # u"
v # dx" '
dt
$
! & dt " + 2 ) 1 + 2
%
c
c (
dx
dx"*
u"*
u* = * =
=
,
v # dx" '
v # u" '
dt
$
$
! & dt " + 2 ) ! & 1 + 2 )
%
%
c (
c (
u! =

dx!

154

(7.46)

and it is easily seen that when both u and v are much smaller than c , equations (7.46)
reduces to the usual Galilean addition of velocities u! + v . The orientation of u ,
specified by its spherical coordinate angles ! and " , can be determined as follows

u # sin (! # )
u"
=
u! $ %&u # cos (! # ) + v '(
) = ) #,

tan (! ) =

(7.47)

since u2 u3 = u2! u3! .

7.2.6 The Four-velocity and the Four-momentum
!
We already know that dx is a four-vector. Since we should expect that the division of a
four-vector by an invariant would not change its character (i.e., the result will still be a
four-vector), then an obvious candidate for a four-vector is the four-velocity
! dx!
U!
d"

(7.48)

The components of the four-velocity are easily evaluated with

dx0 dt
= " uc
dt d!
dx dt
U=
= " uu
dt d!

U0 =

(7.49)

"1

with ! u = 1 " u 2 c 2 2 . Thus the time component of the four-velocity is ! u times c ,

while the spatial part is ! u times the ordinary velocity. The transformation of the fourvelocity under a Lorentz transformation is (using as always the inertial frames K and K !
defined earlier; that is to say, we are now considering the case of a particle traveling at a
velocity u in K , which in turn has a velocity !v relative to K ! )
U 0! = " v (U 0 # $ % U )

U!! = " v U! # $U 0

(7.50)

U!& = U & ,

"1

with ! v = 1 " # 2 2 . Using equations (7.49) for the definition of the four-velocity, and
inserting it into equations (7.50), we get

155

! u " c = ! v! u ( c # $ % u )

! u "u!" = ! v! u ( u! # $ c )

(7.51)

! u " u"& = ! u u & .

The first of this set of equations can be rewritten to express the transformation of
velocities in terms of the ! 's

u$v(
%
! u " = ! v! u ' 1 # 2 * ,
&
c )

(7.52)

while dividing the second and third by the first of equations (7.51) yields
u! " v
u#v
1" 2
c
u$
u!$ =
.
u#v)
&
% v (1 " 2 +
'
c *
u!! =

(7.53)

It will be recognized that this last result expresses, just like equations (7.46), the !law for
the addition of relativistic velocities. Another important quantity associated with U is the
following relativistic invariant

! !
U ! U = " u2 c 2 # u ! u = c 2 .

(7.54)

!
The four-momentum P of a particle is another fundamental relativistic quantity. It is
simply related to the four-velocity by
!
!
P = mU

(7.55)

where m is the mass of the particle. Referring to equations (7.49), we find that the
components of the four-momentum are

P0 = ! mc
p = ! mu

"1

(7.56)

with ! = 1 " u 2 c 2 2 , and u the ordinary velocity of the particle. If we consider the
expansion of P0 c for a non-relativistic velocity u ! c , then we find

156

P0 c = ! mc 2 = mc 2 +

1
mu 2 + !
2

(7.57)

Since the second term on the right-hand side of equation (7.57) is the non-relativistic
expression of the kinetic energy of the particle, we interpret E = P0 c as the total energy
of the particle. The first term (i.e., mc 2 ) is independent of the velocity, and is interpreted
as the rest energy of the particle.
!
Finally, since P is a four-vector, we inquire about the Lorentz invariant that can be
calculated with

! !
! !
P ! P = m 2U ! U = m 2c2 ,

(7.58)

from equation (7.54). But since we also have

! ! E2
P ! P = 2 " p ! p,
c

(7.59)

we find Einsteins famous equation linking the energy of a particle to its mass

E 2 = m 2c4 + c2 p2
!
p2 $
= m2c4 # 1 + 2 2 &
mc %
"
!
u2 $
= m c #1 + ' 2 2 &
c %
"

(7.60)

2 4

= ' mc 2 ,
or alternatively

E = ! mc 2 =

mc 2
u2
1" 2
c

(7.61)

7.3 Tensor Analysis

So far, we have discussed physical quantities in the context of special relativity using
four-vectors. But just like in ordinary three-space where vectors can be handled using a
component notation, we would like to do the same in space-time. For example, the
position vector x has three components x i , with i = 1, 2, 3 for x, y, and z , respectively,
!
in ordinary space, and similarly the position four-vector x has four components x! ,
with ! = 0, 1, 2, 3 for ct, x, y, and z . Please note the position of the indices (i.e., they

157

are superscript); we call this representation for the four-position or any other four-vector
the contravariant representation.
In ordinary space, the scalar product between, say, x and y can also be written

" y1 %
x ! y = ( x1 , x2 , x3 ) $ y 2 ' ,
$ '
$# y 3 '&

(7.62)

where the vector x was represented as a row vector, as opposed to a column vector, and
for this reason we used subscript. That is, we differentiate between the row and column
representations of the same quantity by changing the position of the index identifying its
components. We call subscript representation the covariant representation. Using these
definitions, and Einsteins implied summation on repeated indices, we can write the
scalar product of equation (7.62) as

x ! y = xi yi .

(7.63)

It should be noted that if the scalar product is to be invariant in three-space (e.g.,

x 2 = x ! x does not change after the axes of the coordinate system, or x itself, are
rotated), then covariant and contravariant components must transform differently under a
coordinate transformation. For example, if we apply a rotation R to the system of axes,
then for x ! y to be invariant under R we must have that

x ! y = #$ xR "1 %& ! [ Ry ] = x' ! y',

(7.64)

where x! and y! are the transformed vectors. It is clear from equation (7.64) that column
(i.e., contravariant) vectors transform with R , while row (i.e., covariant) vectors
transform with its inverse. Similarly, in space-time we define the covariant
!
representation of the four-vector x (often written as x! ) has having the components x! ,
with ! = 0, 1, 2, 3 . Therefore, the scalar product of two four-vectors is
! !
x ! y = x" y" .

(7.65)

Four-vectors have only one index, and are accordingly called contravariant or covariant
tensors of rank one. If we have two different sets of coordinates x! and x "! related to
each other through a Lorentz transformation (i.e., a coordinate transformation)

x !" = x !" x 0 , x1 , x 2 , x 3 ,
then any contravariant vector A! will transform according to

158

(7.66)

#x !" $
A! = $ A
#x
"

(7.67)

Conversely, a covariant vector A! will transform according to

#x $
A"! =
A$
#x !"

(7.68)

Of course, tensors are not limited to one index and it is possible to define tensors of
different ranks. The simplest case is that of a tensor of rank zero, which corresponds to an
invariant. An invariant has the same value independent of the inertial system (or
coordinate system) it is evaluated in. For example, the infinitesimal space-time length ds
is defined with
ds 2 = dx! dx! .

(7.69)

The invariance of the scalar product can now be verified with

! !
$x %
$x "# &
#
A ! B = A#" B" =
A% & B
$x "#
$x
%
#
$x $x "
$x %
&
=
A
B
=
A% B&
%
#
&
&
$x " $x
$x
%
&
&
= ' & A% B = A& B .

(7.70)

Contravariant, covariant, and mixed tensors of rank two are, respectively, defined by

$x !"
$x %
$x %
G"#
! =
$x !"
$x !"
H !"# = %
$x
F !"# =

$x ! # %&
F
$x&
$x&
G%&
$x ! #
$x& %
H &
$x ! #

(7.71)

These definitions can easily be extended to tensors of higher rank. It should be kept in
mind that different flavors of a given tensor are just different representation of the same
mathematical object. It should, therefore, be possible to move from one representation to
another. This is achieved using the symmetric metric tensor g!" ( g!" = g"! ). That is, the
metric tensor allows for indices to be lowered as follows

159

F! " = g!# F #"

(7.72)

F ! " = g"# F !# .
If we define a contravariant version of the metric tensor such that
g!" g"# = g! # = $! #

(7.73)

then indices of tensors can also be raised

F! " = g "# F!#

(7.74)

F ! " = g!# F#" .

The metric tensor can also be made to appear prominently in the definition of the scalar
product. For example,
ds 2 = g!" dx! dx " .

(7.75)

From this, and the definition of the space-time length element (i.e., equation (7.29)), it is
deduced that the metric tensor is diagonal with

g00 = 1,

and

g11 = g22 = g33 = !1.

(7.76)

Furthermore, it is easy to see that g 00 = g00 , g11 = g11 , etc. Please take note that equation
(7.76) is only true for flat space-time in Cartesian coordinates. It would not apply, for
example, if space-time were curved (in the context of general relativity), or if one used
curvilinear coordinates (e.g., spherical coordinates). In flat space-time the metric tensor is
often written as !"# , and called the Minkowski metric. From equation (7.76), we can
also better see the important difference between the contravariant and covariant flavors of
the same four-vector, since
" A0 %
A = $ ',
# A&
!

and

A! = A 0 , (A .

(7.77)

Finally, we inquire about the nature of the gradient operator in tensor analysis. Since we
know from elementary calculus that

!
!x $ !
=
,
!x "# !x "# !x $

(7.78)

then the components of the gradient operator transform as that of a covariant vector.
Since, from equations (7.74), we can write

160

x! = g!" x " ,

(7.79)

then

! g#$ x $ !
!
!x# !
=
=
,
!x " !x " !x#
!x "
!x#

(7.80)

and, since in flat space-time the metric tensor is constant,

% # !g$#
!
!x # ( !
=
x
+
g
$#
!x " '& !x "
!x " *) !x$
!
= g$#+ " #
!x$
== g$"

(7.81)

!
.
!x$

Alternatively, operating on both sides of equation (7.81) with g! "# (while using the
prime coordinates) yields a gradient operator that transforms as a contravariant tensor
since (using the first of equations (7.71) and equation (7.78))

& % !x "# !x "$ ) & !x , ! )

!
!
= g"#$
=
g
!x#"
!x "$ ('
!x !x% +* (' !x "$ !x , +*
=g

= g ,

#
!x "# & !x "$ !x , ) !
!
% !x "
=g
-% , ,
(
%
$ +
,

!x ' !x !x " * !x
!x
!x

(7.82)

!x "# !
!x "# !
=
.
!x !x , !x !x

In what will follow we will use the notation defined below for partial derivatives
& ! !x 0 )
!
! #
=
!x" (' $% +*
"

!
& !
)
!" # " = ( 0 ,% +
'
*
!x
!x

(7.83)

The four-dimensional Laplacian operator ! is the invariant defined by the scalar vector

161

1 "2
!! "# " = 2 2 $ % 2
c "t
#

(7.84)

We recognize the wave equation operator. It is important to realize that equations (7.80)
and (7.82) are tensors only in flat space-time (i.e., no curvature, and using Cartesian
coordinates). The corresponding equations for the general case will differ.

7.4 Covariance of Electrodynamics

Before we discuss the invariance of the equations of electrodynamics under Lorentz
transformations, let us rewrite these equations using Gaussian units. For the Maxwell
equations we have

! " D = 4#$
1 'D 4#
!%H&
=
J
c 't
c
!"B = 0
1 'B
!%E+
=0
c 't

(7.85)

We should note that with these units ! 0 = 0 = 1 , and in free space ! = = 1 such that
D = ! E = E and B = H = H . Also important are the equation for the Lorentz force

dp
v
"
%
= q $ E + ! B'
#
&
dt
c

(7.86)

!"
+#$J = 0
!t

(7.87)

and the continuity equation

which is left unchanged from its form in SI units.

Besides the electromagnetic fields, two other quantities appear in Maxwells equations:
the speed of light c and the charge q . We already know that the speed of light is an
invariant (from the second postulate of special relativity). Experiments show that the
charge is also an invariant (for example, a speed dependency of the charge would imply
changes in the net charge in a piece of material with chemical reaction; there is no
experimental evidence for this), and this has implications for what follows. Consider, for
example, the four-volume element d 4 x defined as

d 4 x = dx 0 dx1dx 2 dx 3 ,

162

(7.88)

which transforms as follows under a Lorentz transformation

d 4 x ! = dx ! 0 dx !1dx ! 2 dx ! 3 = Lx ( " ) dx 0 dx1dx 2 dx 3 ,

(7.89)

where Lx ( ! ) is the (Jacobian) determinant of the Lorentz transformation matrix (see

equation (7.25)). It is straightforward to show that Lx ( ! ) = 1 , and therefore that the fourvolume element is an invariant since d 4 x = d 4 x ! . If the charge element dq is to be an
invariant with

dq = !dx1dx 2 dx 3 ,

(7.90)

then the charge density must transform in the same manner as the time component of !a
four-vector. Thus, if we assume that J 0 = !c is the time component of a four-vector J
(the four-current), and we operate on it with the gradient operator of equation (7.83),
then in consideration of equation (7.87) it must be that the space part is the current
density J , and

" !c %
J = $ '
# J&

(7.91)

The continuity equation is now written as

!" J " = 0.

(7.92)

We next look at the vector and scalar potentials in the Lorentz gauge where (using
Gaussian units) we have
1
c2
1
c2

!2 A
4$
" #2A =
J
2
!t
c
!2 %
" # 2 % = 4$&,
2
!t

(7.93)

1 !"
+ # $ A = 0.
c !t

(7.94)

and

If we define the four-potential A

163

" !%
A = $ '
# A&

(7.95)

then equations (7.93) and (7.94) can be written in a covariant form as

!A =

4!
J ,
c

(7.96)

and
! A = 0.

(7.97)

Since the electromagnetic fields can be obtained from the potentials with (again using
Gaussian units)

1 "A
! #$
c "t
B = # % A,
E=!

(7.98)

it can be shown that the components of E and B can be put together in the antisymmetric
second rank electromagnetic tensor F !"

F !" # $! A " % $ " A!

(7.99)

The components of the electromagnetic tensors are explicitly

F !"

$ 0
& Ex
=& y
&E
& z
%E

#E x
0
Bz
#B y

#E y
#B z
0
Bx

#E z '
By )
),
#B x )
)
0 (

(7.100)

Ey
#B z
0
Bx

Ez '
By )
).
#B x )
)
0 (

(7.101)

or, alternatively, in its fully covariant form

$ 0
& #E x
F!" = & y
& #E
&
% #E z

Ex
0
Bz
#B y

Note that E i , for i = x, y, z , are the Cartesian components of E , etc. That F !" has the
desired form can be verified by attempting to express the Maxwell equations with it. A
little calculation will easily show that the first two of equations (7.85) can be written as
164

!" F "# =

4$ #
J
c

(7.102)

while the last two of equations (7.85) become

!" F#$ + ! # F$" + !$ F"# = 0

(7.103)

Finally, to complete this section, we would like to express the Lorentz force in a
covariant form. To do so, we must first introduce two more four-vectors. First, just as we
!
defined in section 7.2.6 the four-velocity as the proper time derivative of dx , we extend
this process to define the four-acceleration a as

a !

dU
d"

(7.104)

It is easy to see that the space part of the four-acceleration equals the ordinary
acceleration in the non-relativistic limit. Note also that the four-acceleration is orthogonal
to the four-velocity

dU 1 d
a U =
U =
U U
d!
2 d!
2
1d c
=
= 0.
2 d!

( )

(7.105)

From equation (7.104), it is a small step to define the four-force F as

dP
dU
F !
=m
= ma
d"
d"

(7.106)

The Lorentz four-force should involve the electromagnetic fields through F !" and the
velocity of the charge through U ! . The simplest of such relations is

dP!
dU ! q !#
=m
= F U#
d"
d"
c

(7.107)

It can be verified that this relation indeed satisfies equation (7.86). Furthermore, it is seen
that the time component of equation (7.107) yields
dW
= qE ! u,
dt

165

(7.108)

which is just a statement of the conservation of energy. In other words, the rate of change
of particle energy W is the mechanical work done on the particle by the electric field.

7.5 Transformation of the Electromagnetic Fields

As we as previously seen, the E and B fields themselves are not tensor quantities, but
components of the electromagnetic tensor F !" (see equation (7.100)). Therefore, if we
want to inquire about how the electromagnetic fields transform under a Lorentz
transformation, we need to study the transformation of F !" . Going back to the first of
equations (7.71), we write

F!

"#

$x !" $x ! # %&
= %
F .
$x $x&

(7.109)

We should note that if we define a matrix !" # such that

"

$x %"
= # ,
$x

(7.110)

then equation (7.109) is simply written in a matrix form as

F! = "F"T ,

(7.111)

with !T the transpose of ! . In cases where the axes of the two coordinate systems are
aligned but their relative velocity is arbitrarily oriented, one can start with equations
(7.27) for the contravariant coordinates

x ! 0 = " x 0 + #i x i
x!i = x i $

(" $ 1) # i # x j $ "# i x 0
#2

(7.112)

(please note the differences between equations (7.27) and (7.112)) in order to evaluate the
partial derivatives
!x " 0
= #$ 0 # + #%i$ i #
#
!x
!x " i
= &#% i
!x 0
!x " i
(# & 1) % % i .
= $ij &
j
j
!x
%2

166

(7.113)

Inserting equations (7.113) (twice) in equation (7.109) will yield the components of E!
and B! through
F! j 0 = E! j

(7.114)

F ! ij = "# ijk Bk! ,

where Bk ! "B k , and the Levi-Civita symbol is defined such that

! ijk = "!

ijk

#+1, for any even permutation of i = 1, j = 2, and k = 3

%
= $"1, for any odd permutation of i = 1, j = 2, and k = 3
% 0, if any two indices are equal
&

(7.115)

After all the required manipulations are performed, one finds that

E! = " ( E + # $ B ) %

"2
# (# & E )
" +1

"2
B! = " ( B % # $ E ) %
# (# & B ) .
" +1

(7.116)

[To successfully derive equations (7.116) you will need to use the following

1
! ijk F ij
2
j
ij
"i F = # [" $ B ]
Bk =

"i " j F ij = " % (" $ B ) = 0

(7.117)

! ijk " i F j 0 = [" $ E ]

" m ! ijk " i F jm = &'" $ (" $ B ) () .

k

Note that these equations (as well as equations (7.114)) involve space vectors, not tensors
(i.e., four-vectors), and are therefore not valid tensor equations.]
For the simple case where the relative velocity of the aligned inertial systems is directed
along the x-axis (as was the case for the K and K ! systems defined earlier in the
chapter), the matrix ! is given by equation (7.25), and a straightforward multiplication
of matrices (or the corresponding simplification of equations (7.116)) yields

E1! = E1
B1! = B1
E2! = " ( E2 # $ B3 ) B2! = " ( B2 + $ E3 )
E3! = " ( E3 + $ B2 ) B3! = " ( B3 # $ E2 )

167

(7.118)

It should now be obvious from equations (7.116) and (7.118) that the electromagnetic
fields are not independent entities but are completely interrelated through the
transformation of the electromagnetic tensor. Finally, we note that we can calculate the
following invariants from the electromagnetic tensor

F!" F !" = 2 B2 # E2 ,

(7.119)

and

( )

det F !" = det F #!" = ( E $ B ) .

168

(7.120)