PH4506: Notes for Part 2: Lorentz Covariant Formulation

Dr Leek Meng Lee

Version: 5th February 2015
Abstract
The logic flow for this part is:
1. We derive Lorentz transformations based on Einsteins 2 axioms of Special Relativity.
2. With Lorentz transformations, we can get all the standard consequences of Special Relativity.
3. Then we formalize Special Relativity by setting up the vector space for it (i.e. by defining the
Minkowski metric) and discuss about vectors and scalars in that space.
4. Then we attempt to rewrite electrodynamics into the language of Special Relativity by first
deducing how the fields transform. Turns out they transform like a tensor in Minkowski space.
5. Then all the important equations in electrodynamics are written into a Lorentz covariant form
(i.e. every inertial frame sees the same form of equations) with perfect consistency. Lorentz
force equation has the upgrade where the momentum is now the relativistic version.
6. Using the new language for electrodynamics, we revisit the point charge case and (i) rederive
Lienard-Wiechert potentials in covariant form, (ii) resolve the coincidence why the electric field
points from the present position and not the retarded position, (iii) realise that power is a
Lorentz scalar and deduce a suitable generalization of Larmors formula.
7. Finally, we recast the Lorentz covariant form of electrodynamics into the framework of Hamiltons variational principle and Lagrangian dynamics.

Contents
1 Special Relativity
1.1 Derivation of Lorentz transformation . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.2 Basic consequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.2.1 Time Dilation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.2.2 Length Contraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.2.3 Velocity Addition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.3 Further consequences (Minkowski spacetime diagram, invariant interval and causality)
1.4 Further Developments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.4.1 Minkowski metric, 4-vectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.4.2 Momentum 4-vector and Einstein relation . . . . . . . . . . . . . . . . . . . . .
1.4.3 Derivative 4-vectors and dAlemberts operator . . . . . . . . . . . . . . . . . .
1.4.4 Relativistic Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

.
.
.
.
.
.
.
.
.
.
.

2
. 2
. 4
. 4
. 5
. 5
. 6
. 8
. 8
. 9
. 11
. 12

2 Electrodynamics recast into a manifestly Lorentz covariant form

2.1 Motivation: Magnetism as a relativistic phenomena . . . . . . . . . . . . . . .
~ and B
~ . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2 Transformations of E
~ and B
~ . . . . . . . . . . . . . .
2.3 Rewriting electrodynamics or unification of E
2.3.1 Maxwells equations . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.3.2 Conservation Laws . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.3.3 Lorentz invariants in electrodynamics . . . . . . . . . . . . . . . . . .
2.3.4 Point charge revisited: Lienard-Wiechert Potentials . . . . . . . . . .
2.3.5 Point charge revisited: Fields of a charged particle in constant velocity
2.3.6 Point charge revisited: Lienards generalization of Larmors formula .

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

13
13
15
18
18
21
23
23
25
27

3 Lagrangian description of a classical relativistic U (1) gauge theory

3.1 Review of Lagrangian mechanics and Noethers theorem . . . . . . . . . . . . . . . . . . .
3.2 Relativistic Lagrangian of a charged particle interacting with external EM field . . . . . .
3.3 Lagrangian of the EM field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

31
31
32
35

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

Special Relativity

1.1

Derivation of Lorentz transformation

Einsteins axioms for special relativity are:

1. Laws of nature and the results of all experiments are equivalent in all inertial frames. Inertial
frames are reference frames at relative constant velocities to one another.
2. Speed of light has the same value in all inertial frames and is independent of the motion of the
light source. Speed of light is also the upper limit of physical entities.
Note that from axiom 1, we can say that space is isotropic (all spatial directions are equivalent)
and spacetime is homogeneous (origin of spacetime can be chosen arbitrarily). This will mean that the
transformation is linear.
In fact, without axiom 2, we have the familiar Galilean transformation in Newtonian mechanics. It
is a linear transformation.

Figure 1: Two inertial frames with S 0 -frame at speed u relative to S-frame.

Assuming the origins coincide at t = 0, we have x = x0 + ut

inverse

=
=

x0 = x ut
dx
dx0
=
+u
dt
dt
v = v0 + u

(1)
(2)
(3)

which is the usual relative velocity expression.

Now we start with a more general linear transformation and incorporate axiom 2 at the end.
x0

= (u)(x ut)

(4)

= (u)y

(5)

= (u)z

(6)

= (u)t + (u)x

(7)

t0

The coefficients of y 0 and z 0 are the same due to isotropy. Again due to isotropy, flipping x, x0 and u
will result in no change (we flipped y and y 0 so that we maintain a right handed coordinate system).
x0

(u)(x + ut)

(8)

(u)y

(9)

(u)z

(10)

(u)t (u)x

(11)

We compare to get (u) = (u), (u) = (u), (u) = (u) and (u) = (u) so only  is an odd
function. We write (u) into this form where (u) is even.
(u) =

u
(u)
(u)

(12)

so the set of transformation equations are

x0

= (u)(x ut)
2

(13)

y0
z

t0

= (u)y

(14)

= (u)z


u
x
= (u) t
(u)

(15)
(16)

We can jump into S 0 -frame and by axiom 1, the equations must have the same form. We interchange
primed and unprimed coordinates and reverse the sign of u.
= (u)(x0 + ut0 )

x
y

(18)

= (u)z


u 0
= (u) t0 +
x
(u)

(19)

= (u)2 y so we choose = 1

(21)

= (u)y

z
t
Substitute y 0 = (u)y to get y

(17)

(20)

Then we proceed further with the x and t equations,

x0 = (x ut) x0 = x ut


u2
u
u
0
t = t x ut0 = ut x

2
u
Eliminate t = x0 + ut0 = x x

0
x + ut0
x =
2
u
=

x0


u2

1
Compare with x = (x0 + ut0 ) = x0 + ut0 to get =

1


1
and = q

1
u2
(u)

+

u2

(22)
(23)
(24)
(25)
ut0


 and =

u2

(26)

1

1

u2

 so that =

. The positive root is chosen so that when u 0, 1 and x0 x.

1
The transformation equations are now,
x0

(x ut)

(27)

(28)

z0

(29)



u
t x

(30)

We determine by using axiom 2 where every inertial frame sees speed of light as c. So we need the
velocity addition formula. Take infinitesimal intervals,
dx0
dt0
divide one equation by the other,

dx0
dt0

= (dx udt)


u
= dt dx

dx udt
=
dt u dx
=

v0

dx
dt

1 u dx
dt
vu
1 uv

(31)
(32)
(33)
(34)
(35)

where v 0 is the velocity of the object measured in S 0 -frame and v is the velocity of the object measured
in S-frame. Now consider the object to be light, then v = c and v 0 = c, so

1
Finally with = q
1

c =

cu
1 uc

(36)

c2

(37)

, the Lorentz transformation equations are,

u2
c2

x0

(x ut)

(38)

y0

(39)

z0

u 
t 2x
c

t0

(40)


(41)

We generalise this for arbitrary relative velocities. We denote ~u to be the relative velocity in an arbitrary
direction. We resolve the position vector into a component parallel to ~u and a component perpendicular
to ~u.
~x ~u = ux cos = uxk
~x ~u
xk =
u
~x ~u
~xk =
u

u
~x ~u
~u
=
u2
~x ~u
hence, ~x = ~x ~xk = ~x 2 ~u
u

(42)
(43)
(44)
(45)
(46)

The transformations of the parallel component and the perpendicular component has been derived earlier.
We note it again,
~x0k

(~xk ~ut)

(47)

~x0

= ~x


~u ~xk
= t 2
c

(48)

t0

(49)

and,
~x0

= ~x0k + ~x0

(50)

= (~xk ~ut) + ~x


~x ~u
~x ~u
=
~u ~ut + ~x 2 ~u
u2
u
~x ~u
= ~x + ( 1) 2 ~u ~ut
u

(51)

Since ~x ~u = 0, we can also write ~xk ~u = ~x ~u, so



~u ~x
0
t = t 2
c

1.2
1.2.1

(52)
(53)

(54)

Basic consequences
Time Dilation

We recall the 2 inertial frames setup where S 0 -frame is moving at speed u with respect to the x-axis of
the S-frame. Let a clock be in the S 0 -frame (or S 0 -frame is comoving with a clock that is moving with
speed u along the x-axis of the S-frame).
4

 Let event 1 be the second-hand of the clock at the 12 oclock-mark.

Let event 2 be the second-hand of the clock at the 1 oclock-mark.


Recall the inverse transformation: t = t0 + cu2 x0 and we want to find a relationship between
intervals as measured from each frame.


u 
u 
t1 = t01 + 2 x01 and t2 = t02 + 2 x02
c
c


u 0
0
0
0
Take difference, t2 t1 = t2 t1 + 2 (x2 x1 )
c 

u
0
t = t + 2 x0
c
| since the clock is in the same x0 coordinate, x0 = 0
t =

t0

time
(55)
(56)
(57)
(58)

Thus if t0 = a 5 second interval (measured in S 0 -frame), then t > 5 seconds as measured in S-frame.
In other words, if there is a similar clock in S-frame and both clocks are set to start ticking from the
12 oclock-mark together, when the S 0 -clock is ticking at the 1 oclock-mark, the S-clock is seen to be
ticking at the -mark. 1
1.2.2

Length Contraction

Let a stick be on the x0 -axis in the S 0 -frame.

Let event 1 be the left end of the stick.
Let event 2 be the right end of the stick.
Recall the inverse Lorentz transformation: x = (x0 + ut0 ) and we want to find a relationship between
spatial intervals as measured from each frame.
x1
Take difference, x2 x1

(x01 + ut01 ) and x02 = (x02 + ut02 )

(59)

(x02

(60)

x01

u(t02
0

t01 ))

(x + ut )

seeing the stick means both ends are observed simultaneously so t = 0



u
u
then recall t = t0 + 2 x0 so t0 = 2 x0
c
c


2
u
x0 2 x0
(62)
c
1
recall = q
2
1 uc2

|
=
|

(61)

x0

(63)

So, x0 is the length of the stick measured in S 0 -frame where the stick has no relative motion and x
is the length of the stick measured in S-frame and x < x. This is due to relative simultaneity: we
require both ends of the stick to be observed simultaneously in S-frame but actually these 2 events are
not simultaneous in S 0 -frame. The opposite case is true also. This fact can be illustrated clearly on the
Minkowski spacetime diagram which we will learn later.
1.2.3

Velocity Addition

We have already derived the velocity addition formula which is v 0 =

vu
but lets state the situation
1 uv
c2

more precisely.
Basically, there are 3 inertial frames: S, S 0 and S 00 . S 0 -frame is moving at speed u relative to S-frame
and S 00 -frame is moving at speed v relative to S-frame. Then v 0 is the speed of S 00 -frame relative to
S 0 -frame.
1 Assuming

< 12 for simplicity.

Figure 2: A 3 inertial frame setup.

1
We shall go further and relate the various factors: (v 0 ) = q
1
1
,
(v) = q
2
1 vc2
(v 0 )

=
|
=
=
=
=

1.3

v 02
c2

, (u) = q

1
q

1
1 2
c
1

and
u2
c2

(64)

v 02
c2

vu
1 uv
c2

uv 2
c2
1

2 !1/2
(65)

(vu)2
c2


uv 2
c2

!1/2
(66)

expand the numerator

1/2 

v 2 2uv + u2
uv 
2uv u2 v 2
1

1 2 + 4
c
c
c2
c2


1/2 
u2
u2 v 2
uv 
v2
1 2
1 2 2 + 4
c
c
c
c


1/2 
2
2
u
v
uv 
1 2
1 2
1 2
c
c
c

uv 
(u)(v) 1 2
c

(67)
(68)
(69)
(70)

Further consequences (Minkowski spacetime diagram, invariant interval

and causality)

Minkowski spacetime diagram: The Minkowski spacetime diagram is one way of graphically representing both S and S 0 -frames together so that by referring to the S-frame axes, we get the coordinates
in S-frame and by referring to the S 0 -frame axes, we get the coordinates in S 0 -frame.
We will construct S-frame in the usual way where the ct vs x axis are 90 degrees to each other. The
S 0 -frame shall be constructed based on Lorentz transformations.
By definition, ct0 axis means x0 = 0,
x0 =0

x0 = (x ut) 0

= x ut
c
x
ct =
u

(71)
(72)

By definition, x0 axis means t0 = 0,


ux  t0 =0
t0 = t 2 0
c
6

ux
c2

(73)

ct

u
x
c

(74)

Note that the spacings are not the same on the axes of different frames: take coordinate (x0 , ct0 ) = (1, 0)
and using Lorentz transformations
1 = (x ut) and 0 = t

u
ux
= x = and t =
c2
c

(75)

Figure 3: Summary of the main features in Minkowski spacetime diagram.

Invariant interval:

Consider this strange Pythagoras theorem for length in S 0 -frame

c2 t02 + x02


ux 2
= c2 2 t 2
+ 2 (x ut)2
c


u2 x2
2tuxc2
2
2 2
2
2 2
2 + x 2xut + u t
= c t +
c2
c


2 2
u x
2
2 2
2
2 2
= c t + x 2 + u t
c





1
u2
u2
2 2
2
=
c t 1 2 + x 1 2
2
c
c
1 uc2
= c2 t2 + x2

(76)
(77)
(78)
(79)
(80)

so this length is the same in S-frame and therefore this length is called the invariant interval because
it is the same (numerical value) regardless of the inertial frame.
If we choose a certain inertial frame where a clock is stationary, the time measured (in this comoving
frame) is called the proper time and is denoted by . As the clock is always at the origin in this frame,
x2 = 0 and so the invariant interval in this frame is c2 2 . We can write the following forms:
c2 2 = c2 t2 + x2
Full 4D: c2 2 = c2 t2 + x2 + y 2 + z 2
Define infinitesimal invariant interval ds: (ds)2 = c2 (d )2 = c2 (dt)2 + (dx)2 + (dy)2 + (dz)2
7

 Conventional writing: ds2 = c2 d 2 = c2 dt2 + dx2 + dy 2 + dz 2

Since there are 3 positive terms and 1 negative term in the invariant interval, it can be positive, zero or
negative. We illustrate this on the Minkowski diagram.

Figure 4: 3 separate regions in the Minkowski spacetime diagram.

There should be a z-axis also but I cant draw it. Thus the light cones should really be hypercones.
The 3 regions are named as follows:
For interval < 0, it is called the timelike region. This is the region where particles with mass
move because particles with mass can only move slower than c.
For interval = 0, it is called the lightlike region. This is the region where photons travel
because photons only travel with speed c.
For interval > 0, it is called the spacelike region. This is the region where particles move
faster than c.

1.4

Further Developments

1.4.1

Minkowski metric, 4-vectors

We will now carry out a more formal treatment so that we can handle 4D vectors (in Minkowski space).
This is an upgrade of the treatment of 3D vectors.
X
3D ~x =
xi ei where ei = ex , ey , ez are the basis vectors (previously denoted as x
, y, z) (81)
i

4D

x
=

x e where e = ect , ex , ey , ez are the basis vectors

(82)

Now we introduce the Einstein summation convention: whenever 2 identical indices appear (one in the
superscript, one in the subscript), a summation is implied. Greek indices run from 0 to 3 and Latin
indices run from 1 to 3. Index 0 = ct, 1 = x, 2 = y and 3 = z.
Now consider the dot product of 3D and 4D vectors with itself which gives the squared length:
3D ~x ~x = xi ei xj ej = xi xj (
ei ej )
4D

(83)

x
x
= x e x e = x x (
e e )

(84)

We shall define e e as the Minkowski metric tensor since the basis vectors characterise the space.
Note that because the dot product commutes, therefore = . Note that the 4D dot product is
supposed to give the squared length so this 4D dot product should be equal to the invariant interval.
c2 t2 + x2 + y 2 + z 2

= x
x
= x x (
e e ) = x x

(85)

with x , x

(x0 , x1 , x2 , x3 ) = (ct, x, y, z) we require =

1
1

(86)

1
1

We can take another perspective where the dot product is formed by a vector from a space with another
vector from the conjugate space. This is like in quantum mechanics where the dot product is formed
between a vector in ket-space and another vector in bra-space and the 2 spaces are hermitian conjugates
of each other.
Define the contravariant position 4-vector: x

(x0 , x1 , x2 , x3 )

(87)

(ct, x, y, z)

(88)

Define the covariant position 4-vector: x

= x

(89)

1
1

ct
x

y
z
1

ct
x

=
y
z

(90)

(91)

(ct, x, y, z)

(92)

The dot product: x x

ct

ct

x

z
y
z

= c2 t2 + x2 + y 2 + z 2

(93)

(94)

and the metric tensor maps a contravariant vector to a covariant vector. This is casually described as
the metric tensor lowering an index.
To map from a covariant vector to a contravariant vector (or raising an index), we need the inverse
metric tensor.
x = x = x = x
which means is the

inverse of .
1
1

1
since =
Explicitly, =

1
1
1.4.2

(95)

1
1

Momentum 4-vector and Einstein relation

We want to define a velocity 4-vector and proceed to construct a momentum 4-vector from there. A
quantity is a 4-vector if it transforms like a 4-vector under Lorentz transformations so we can definite
quantities as we like it but we have to check its transformation to make sure that it is a 4-vector.
Define the contravariant velocity 4-vector: u =

dx
d

(96)

This guess is made by requiring a time derivative on the position 4-vector and choosing the time to be
the proper time because the numerator is already a 4-vector and so we do not want the denominator to
transform also. Thus we can
Define the contravariant momentum 4-vector: p = mu = m

dx
d

(97)

where m is called the rest mass which is the mass measured in the comoving frame of the particle.
Now we need to check that the objects are indeed 4-vectors. First we work out the time component
(or 0-component),
p0 = m

dx0
d

= m
|

cdt
d

(98)

from time dilation, dt = (u1 )d , (u1 ) = q

1
1

u21
c2

= mc(u1 )

(99)

where u1 is the relative velocity of the particle with respect to the S-frame. The comoving frame of the
particle is called the S 00 -frame.

Figure 5: 3 inertial frame setup for deriving the Lorentz transformations of the momentum 4-vector.
We expect a transformation of the form for the space component:


u
x0 = (x ut) = x (ct)
c


u
0
(space component) = space component (time component)
c

(100)
(101)

So we check from S 0 -frame which is moving at speed u with respect to S-frame,

p01

=
=
|
=
|
=
=
=

dx01
d
dx0
m
d
recall dx0 = (u)(dx udt)


dx
dt
m(u)
u
d
d
recall time dilation dt = (u1 )d
 1

p
m(u)
u(u1 )
m


u
(u) p1 (mc(u1 ))
c

u 0
1
(u) p p
c

(102)
(103)

(104)

(105)
(106)
(107)

Indeed p1 transforms like the space component. Now we check the time component which is expected to
transform as

u 
t0 = t 2 x
(108)
c 

u
(109)
ct0 = ct x
c


u
(time component)0 = time component (space component)
(110)
c


u
?
So, p00 = (u) p0 p1
(111)
c
The LHS = p00 = mc(u01 )
(112)
with (u1 ) = q

1
0

u12
c2

where u01 is the velocity of the particle (S 00 -frame) with respect to the S 0 -frame.

10

u1 u
.
1 uc12u

u 
= (u) p0 p1
c


u dx
= (u) mc(u1 ) m
c d
dx
| recall dt = (u1 )d and
= u1
dt 

u
= (u) mc(u1 ) mu1 (u1 )
 cuu 
1
= mc(u)(u1 ) 1 2
c


Velocity addition says that u01 =

The RHS

(113)
(114)

(115)
(116)

uu1 
| recall the identity (u)(u1 ) 1 2 = (u01 )
c
= mc(u01 )

(117)

Thus indeed p0 transforms as a time component. We rewrite the contravariant momentum 4-vector as,


E
p = (p0 , p1 , p2 , p3 ) =
, px , py , pz = (mc(u1 ), mu1 (u1 ), 0, 0)
(118)
c
We shall check that E = mc2 (u1 ) is indeed the total energy.

1/2
u2
mc2 1 21
c
| assume the particle moves slowly, so u1 << c


1 u21
2
+
= mc 1 +
2 c2
1
mu21 +
= |{z}
mc2 +

|{z}
2
| {z }

E = mc2 (u1 )

rest mass energy

Newtonian KE

(119)

(120)
(121)

relativisitic corrections

Recall that for the position 4-vector, we carried out the dot product to find its (squared) length and
that quantity is a Lorentz invariant quantity (Lorentz scalar). We do the same for the momentum
4-vector,
covariant momentum 4-vector: p

=
=

(squared) length: p p = p p =

E2
+ p2x + p2y + p2z
c2

p


E
, px , py , pz
c
E2
2 + p~2
c

(122)
(123)
(124)

We can also form p p by using p = (mc(u1 ), mu1 (u1 ), 0, 0)

2 2

p p = m c (u1 ) +

m2 u21 (u1 )2



u21
= m c (u1 ) 1 2 = m2 c2
c
2 2

(125)

and we equate the two expressions of p p ,

m2 c2
E

E2
+ p~2
c2
= p~2 c2 + m2 c4

(126)
(127)

which is the famous Einstein relation.

1.4.3

Derivative 4-vectors and dAlemberts operator

Intuitively, we define the derivative 4-vector as






1
=
,
,
,
and
=
,
,
,
x
c t x y z
x
c t x y z
11

(128)

We need to check that they are really 4-vectors by checking their transformations.
1

=
00
x
c t0

=
=

x
ct0 x
ct
x
+
+0+0
0
ct ct ct0 x

(129)

(130)

u
| recall Lorentz transformations ct = ct0 + x0 and x = (x0 + ut0 )
c

u
=
+
(131)
ct
c x

so if we recall the Lorentz transformation ct0 = ct uc x , we can deduce that ct

0 = x0 transforms
like the time component of a covariant (lower index) 4-vector. To confirm, we check further,


=
x01
x0

ct
x
+
+0+0
x0 ct x0 x

(132)

u
| recall the Lorentz transformations ct = ct0 + x0 and x = (x0 + ut0 )
c
u

=
+
(133)
c ct
x 


u
=
+
(134)
x
c ct


so if we recall the Lorentz transformation x0 = (x ut), we again realise that the time component
behaves with a wrong sign which means that it is the time component of a covariant (lower index)
4-vector, thus x is a covariant 4-vector.
We can immediately check for x ,


1
u

1
=

(135)
x00
c t0
c t
c x




=
=

(136)
x01
x0
x
c
ct
which means

is behaving like a contravariant 4-vector! Hence we denote:


 


1
,
,
,
=
,
,
,
=
(137)
x
x0 x1 x2 x3
c t x y z
 

 


1
,
,
,
= 0 , 1 , 2 , 3 (138)
=
,
,
,
=
x
x0 x1 x2 x3
c t x y z
x x x x

We shall now take the dot product of these 2 vectors. Recall that it is supposed to give us a quantity
that is Lorentz invariant (Lorentz scalar).

 =

1 2
2
2
2
+
+ 2+ 2
2
2
2
c t
x
y
z
1 2
~2
2 2 +
c t

(139)
(140)

~ 2 is the 3D Laplacian
where =  is called the 4D Laplacian or dAlembertian operator and
2
operator.
1.4.4

Relativistic Mechanics

Newtons first law is bulit into the first axiom of special relativity in the sense that in inertial frames,
there is no net force due to relative motion and so different inertial frames would experience the same
physics.
For Newtons second law, the most intuitive way to fit it into special relativity is to modify the
momentum to relativistic momentum.
d
d~
p
=
(m~v (v)) (where v is the speed of the particle)
F~ =
dt
dt
2 In

some EM books, the dAlembertian has the symbol 2 = .

12

(141)

This relation is for S-frame where the particle is seen to be moving at speed v. How about in S 0 -frame
which is moving at speed u with respect to S-frame along the x axis? We have to transform both
numerator and denominator! For simplicity, we shall take S 0 -frame to be the (instantaneous) comoving
frame of the particle, i.e. v = u. For the y component,
Fy0

=
|
=

dpy
dp0y
dpy
dt


=
=
dt0
(u) dt cu2 dx
(u) 1
dx
=u
where
dt
Fy
2
= (u)Fy
(u) 1 uc2

u dx
c2 dt

(142)

(143)

Similarly for the z component

Fz0 = (u)Fz

(144)

Now for the x component,

Fx0

(u) dpx uc dp0
dp0x


= 0 =
dt
(u) dt cu2 dx

=
=
|

0
u dp
c dt
1 cu2 dx
dt
Fx cu2 dE
dt
2
1 uc2

dpx
dt

where

(145)
(146)

2u
dE
d
du
d
c2
= mc2 (u) = mc2
= u mu(u) = uFx


dt
dt
dt
u2 3/2 dt
2 1 c2

= Fx

(147)

Thus the ordinary force transforms awkwardly. We try to define another force quantity which is a 4vector. We could take the clue from the definition of 4-velocity and define the 4-force (or Minkowski
force) as
K =

dp
d

where p is the 4-momentum and is the proper time. The spatial components are
 
p
~ = dt d~
K
= (u)F~ (where F~ is the ordinary force)
d dt

(148)

(149)

The time component is

K0 =

1 dE
dp0
=
d
c d

(150)

where dE
d can be interpreted as the rate of energy increase with respect to proper time or proper power.
p
There is nothing wrong with using F~ = d~
~ is the relativistic momentum except for awkdt where p
wardness in transformation. Therefore we shall use the covariant Minkowski force K for the sake of
convenience.

2
2.1

Electrodynamics recast into a manifestly Lorentz covariant

form
Motivation: Magnetism as a relativistic phenomena

The objective from this point onwards is to make it clear that electrodynamics is meant to be described
~ and B
~ fields are just 2 sides of the same coin. This means that they are 6
in Minkowski space. The E
~ and B
~ fields shouldnt really
components of a single object called the field strength tensor F . The E
be seen as 2 different fields. The rewriting of electrodynamics in the language of special relativity does
~ and B
~ clearly.
not really generalise anything. It shows this unification of E
13

So in this section, we want to deduce magnetism from the knowledge of electrostatics and special
relativity to motivate the intimate connection between electrodynamics and special relativity.
Consider, in S-frame, we see a wire and inside the wire, we see a string of positive charges moving
to the right with speed v. We assume that a linear charge density can be defined for it. In the same
wire, superimposed on the positive string of charges is a negative one, with moving to the left with
speed v. The net current is to the right and given by
I = 2v

(151)

For a charge q, distance r outside the wire and moving to the right with speed u(< v), there is no
electrical force since the 2 line charges has no net charge.
We now go into S 0 -frame which is moving to the right with speed u with respect to S-frame. Using
the velocity addition formula,
vu
1 vu
c2
v+u
0
speed of negative line charge in S -frame: v =
1 + vu
c2
speed of positive line charge in S 0 -frame: v+ =

(152)
(153)
(154)

The linear charge densities are seen to change by different amounts due to length contraction.
For positive (in S 0 -frame): + = (v+ )0 with (v+ ) = q
For negative (in S 0 -frame): = (v )0

1
1
with (v ) = q
1

(155)

2
v+
c2

(156)

2
v
c2

where 0 is the charge density seen in the comoving frame of the charges,
1
then, = (v)0 with (v) = q
2
1 vc2
and using the velocity addition identities (v+ ) = (v)(u) 1
can calculate the net charge seen in S 0 -frame
net = + +

=
=
=

uv
c2

and (v ) = (v)(u) 1 +

0 ((v+ ) (v ))

uv
uv 
0 (v)(u) 1 2 1 2
c
c
2uv
q
2
c2 1 uc2
14

(157)

uv
c2

, we

(158)
(159)
(160)

Thus there is an electric force on charge q in S 0 -frame,

F 0 = qE

net
20 r
v
qu
q

2
0 c r 1

(161)

= q
=

(162)
u2
c2

But both S and S 0 -frames must see the same physical outcome so there must also be a force on q in
S-frame. We transform the force
F

1
F 0 (refer to the previous section on relativistic mechanics)
(u)
v qu
=
0 c2 r
1
| write 2 = 0 0 and 2v = I
c


0 I
= qu
2r
| {z }

(163)
(164)

(165)

magnetic field in S-frame

So this force is the Lorentz force on a charged particle moving in a magnetic field.

2.2

~ and B
~
Transformations of E

~
To begin the rewriting of electrodynamics into Lorentz covariant form 3 , first we must know how the E
~
and B fields transform and realise that they do not form a 4-vector.
Consider a flat plate capacitor moving with speed u00 with respect to S-frame. Let the capacitor be
in S 00 -frame, i.e. this S 00 -frame is the comoving frame of the capacitor.

Figure 6: Capacitor in S 00 -frame and seen from S-frame. Capacitor plates are parallel to x00 axis.
In S 00 -frame, the capacitor has length l00 and width w and surface charge density 00 . Thus the electric
field seen in S 00 -frame:
Ey00 =

00
0

(166)

In S-frame, the length of the capacitor is measured with length contraction

l=

1 00
1
l where (u00 ) = q
(u00 )
1

u00 2
c2

(167)

Since total charge Q is invariant 4 and width w does not undergo length contraction as it is perpendicular
Q
to the direction of motion, the charge density measured in S-frame is = lw
= (u00 ) l00Qw = (u00 ) 00 .
3 Lorentz

covariant means the expression has the same form in all inertial frames.
am taking this to be an empirical fact. It can be shown by Noethers theorem which will be discussed at the end of
Part 2.
4I

15

The electric field measured in S-frame is5

Ey =

00
= (u00 )
= (u00 )Ey00
0
0

(168)

Now consider the capacitor plates to be perpendicular to the x00 axis,

Figure 7: Capacitor in S 00 -frame and seen from S-frame. Capacitor plates are perpendicular to x00 axis.
So the capacitor plate spacing d undergoes length contraction but the electric field (which is now in
the x direction) does not depend on d so,
Ex = Ex00

(169)

We can now talk about the transformation of magnetic fields because in S-frame, 2 moving charged
plates (we are back to the situation where the capacitor plates are parallel to the x00 axis) amounts to 2
surface currents.
In S-frame, the magnetic field points in the +z direction between the plates using the right hand
rule. Using Amperes law, the magnetic field between the plates is
In S-frame: Bz = 0 |K| = 0 u00

(170)

To compare Lorentz transformed fields, we need to bring a 3rd inertial frame and we call it the S 0 -frame
which is moving at speed u0 with respect to S-frame.

Figure 8: Capacitor in S 00 -frame and seen from S-frame and from S 0 -frame. Capacitor plates are parallel
to x00 axis.
From the addition of velocities, the relative velocity of S 00 -frame with respect to S 0 -frame is u1 =
u u0
00 0 .
1 uc2u
00

In S 0 -frame: Ey0 =

0
00
= (u1 )
= (u1 )Ey00
0
0

(171)

5 We may suspect that the electric field may not be perpendicular to the plates but by symmetry considerations, any
parallel component from the + plate will be cancelled by an opposite parallel component from the plate.

16

In S 0 -frame: Bz0 = 0 0 u1 = 0 (u1 ) 00 u1

(172)

Finally, we need to relate Ey0 to Ey and Bz and relate Bz0 to Ey and Bz . Eliminating the double primed
fields will give us the Lorentz transformation of the fields between S-frame and S 0 -frame.
Ey0

(u1 )Ey00

(173)


recall the identity (u1 ) = (u00 )(u0 ) 1

and recall (u00 )Ey00 = Ey



u00 u0
0
(u )Ey 1 2
c


0
u 00
0
(u ) Ey 2 u Ey
c


u0 00
0
(u ) Ey 2 u
c
0
recall that Bz = 0 u00


u0
(u0 ) Ey 2
Bz
c 0 0
1
recall 2 = 0 0
c
(u0 ) (Ey u0 Bz )

=
=
=
|
=
|
=
Then for
Bz0

00 0

u u
c2

(174)
(175)
(176)

(177)

(178)

Bz0
= 0 (u1 ) 00 u1
|

(179)



u00 u0
u00 u0
00
recall identity (u1 ) = (u00 )(u0 ) 1 2
and u1 =
= Ey00 0
00 0 and
c
1 uc2u

= 0 (u00 )(u0 )Ey00 0 (u00 u0 )

|
=
=

(180)

00

recall (u )Ey00 = Ey
0 0 (u0 )Ey (u00 u0 )
(u0 ) (0 0 Ey u00 0 0 Ey u0 )

(181)
(182)

1
| recall Ey =
and Bz = 0 u00 and 0 0 = 2
0
c


0
u
= (u0 ) Bz 2 Ey
c

(183)

We can repeat with the capacitor plates set parallel to the xy plane and get
Ez0
By0

= (u0 ) (Ez + u0 By )


u0
0
= (u ) By + 2 Ez
c

(184)
(185)

To get the transformation for Bx , we imagine a solenoid in S 00 -frame, with its axis aligned to the x00 axis.
Bx00

= 0 n00 I 00
|
|

(186)

1
in S-frame, time dilation gives I =
I 00
(u00 )
in S-frame, length contraction gives n = (u00 )n00

= 0 nI

(187)

= Bx

(188)

~ and B,
~ the component parallel to motion is unaffected and this is weird because so far we do
So for E
not have any quantity transforming in this way.
Finally, we collect the whole set of transformations,
Ex0 = Ex
Bx0 = Bx

Ey0 = (u0 )(Ey u0 Bz )

By0 = (u0 ) By +

u
c2 Ez

17

Ez0 = (u0 )(Ez + u0 By )

Bz0 = (u0 ) Bz

u0
c2 Ey

(189)

This set of transformation is nowhere similar to a 4-vector transformation. This also gives a strong hint
~ and B
~ are not different fields as they transform into each other.
that E

2.3
2.3.1

~ and B
~
Rewriting electrodynamics or unification of E
Maxwells equations

~ and B
~ are not like that of a 4-vector but they are that of an antisymmetric 2Transformations of E
6
tensor. An antisymmetric (4D) 2-tensor has 6 elements and there are exactly 6 field components in
~ and B.
~ Thus the conclusion is that, there is only one unified field in electrodynamics and it is the
E
2-tensor field strength denoted as F .
We take on a matrix way of writing:
Position 4-vector:

0
ct
ct0 = (u) ct uc x
x0
x0 = (u)(x ut)

y0
y0 = y
0
z0
z =z
Momentum 4-vector:

E0
E
u
c = (u) c c px

p0x = (u) px uc Ec

py = py
0
pz = pz

u
= c
0
0

uc

0
0

ct
0 0
x
0 0

1 0 y
z
0 1

E 0 /c

u
p0x
= c
p0y 0
0
p0z

uc

0
0

= x0 = x

E/c
0 0
px
0 0

1 0 py
pz
0 1

(190)

= p0 = p

(191)

Thus the Lorentz transformation of the field strength 2-tensor F is

F 0

|
=

we write it into matrix multiplication format

F

=
in matrix notation: F

F
T

(192)

(193)
(194)

where is the transpose of

(195)

It turns out that the correct arrangement of Ex , Ey , Ez , Bx , By and Bz in F so that F 0 = F T gives

the same transformation results with the earlier flying capacitor derivation is

0
Ex /c Ey /c Ez /c
Ex /c
0
Bz
By

(196)
F =
Ey /c Bz
0
Bx
Ez /c By
Bx
0
The field strength in S 0 -frame would simply be

0
0

E
x /c
F 0 =
Ey0 /c
Ez0 /c

denoted as
Ex0 /c
0
Bz0
By0

Ey0 /c
Bz0
0
Bx0

Ez0 /c
By0

Bx0
0

(197)

We quickly check 2 examples:

F 010

1 0 F

(198)

Ex0

10 0 F 0 + 11 0 F 1

(199)

10 01 F 01 + 11 00 F 10


 u   u  E 
Ex
x
=

+
c
c
c
c
=

6 When
7 At

I say 4-vector, the 4 indicates dimension and when I say 2-tensor, the 2 indicates number of indices.
worst, you can take it as the recipe that for every index, you need one Lorentz matrix to transform.

18

(200)
(201)

Ex0
Ex0



u2
= Ex 1 2 2
c
= Ex (indeed)

(203)

F 020

(204)

Ey0
c

2 0 F

22 0 F 2

(202)

(205)


21

22 00 F 20 + 01 F


Ey
u
=
(Bz )
c
c
= (Ey uBz ) (indeed)

Ey0

(206)
(207)
(208)

You can also directly carry out matrix multiplication F 0 = F T to check every element.
Before we see Maxwell equations in covariant form, we first need to realise that there is a current
density 4-vector:
Take an infinitesimal volume V which has charge q, the charge density is = Vq . Assume that V is
moving along the x axis with speed u, the current density is Jx = u. We define the rest charge density
1
as 0 = Vq0 . Since only one dimension is contracted, V = (u)
V0 and so
q
(u) = 0 (u)
V0
= u = 0 u(u)

Jx

(209)
(210)

We recall the (contravariant) momentum 4-vector p = Ec , px , py , pz = (mc(u1 ), mu1 (u1 ), 0, 0), we

are inspired to define the (contravariant) current density 4-vector

J = (c, Jx , Jy , Jz ) = (J 0 , J 1 , J 2 , J 3 ) = (0 c(u), 0 u(u), 0, 0)

(211)

It must be noted that the charge continuity equation is actually a Lorentz invariant equation.8
~ ~
+J
t
1 J 0
Jx
Jy
Jz
+
+
+
c t
x
y
z
0
1
2
J
J
J
J 3
+
+
+
0
1
2
x
x
x
x3
J

(212)

(213)

(214)

0 (charge continuity equation in Lorentz scalar form)

(215)

The Lorentz covariant Maxwell equations are just 2 equations:9

F = 0 J and F + F + F = 0

(216)

We check for example = 0,

0 F

00

01

02

F 0

= 0 J 0

(217)

03

= 0 c

(218)

= 0 c

(219)

= 0 c

(220)

+ 1 F + 2 F + 3 F
Ex /c Ey /c Ez /c
0+
+
+
x
y
z
1~ ~
E
c
1
recall that 2 = 0 0
c
~ E
~

8 We

|
=

0

(221)

call such an equation as the 4-divergence of J is zero.

that the charge continuity equation is obtainable from Maxwell equations. In this covariant form, of course the
same can be done. We take 4-divergence of F = 0 J to get F = 0 J . The LHS is zero because is
symmetric and F is antisymmetric, so J = 0.
9 Recall

19

which is Gauss law. The other 3 Maxwell equations shall be checked in the homework.
Next is the Lorentz force law. We already have a force 4-vector (or 4-force) called the Minkowski
force K .
K =

dp
d

(222)

~ + (~v B))
~ we deduce the relativistic version by
In terms of the fields, or RHS of the law F~ = q(E

~
replacing ~v u and B F giving the ansatz
K = qF u

(223)

We check the spatial part,

K1

= qF 1 u

(224)

10

11

12

13

= qF u0 + qF u1 + qF u2 + qF u3
|

note that F

(225)

11

=0

p
1
| recall u =
= (mc(u), mu(u), 0, 0) where u is in x direction
m
m
| now take u to have 3 components, so u = u = (c(u), ux (u), uy (u), uz (u))
 


Ex
= q(u) c
+ uy Bz + uz (By )
(226)
c


~ + (~u B)
~
= q(u) E
(227)
x

then,
p
~ = d~
K
d
d~
p
dt



~ + (~u B)
~
q(u) E

(228)

recall dt = (u)d


~ + (~u B)
~
q E

(229)

where p~ is the relativistic momentum. The details of K 0 shall be worked out in the homework.
~ Turns out that they form the potential 4-vector
Finally, we need to deal with the potentials: and A.
(or 4-potential).



A =
, Ax , Ay , Az
(230)
c
which means it transforms as follows:


 

u
0
u
= (u)
Ax
and A0x = (u) Ax
and A0y = Ay and A0z = Az
c
c
c
c c

(231)

and we check that indeed these transformations fulfil the transformation properties of the fields. Recall
~ =
~ A~ ,
E
t
Ex

Ex0

Ax

x
t

1
= recall
= 1 and
= 0
x
c t
= 1 + c 0 Ax
=

=
=
=
|
=

01 0

+ c

00

A0x

(233)
(234)

 
u 
(( uAx )) +
Ax 2
c
 u


 
u
u 
0
0
1
+ (( uAx )) + c 1 Ax 2
c
c
c
expand and 4 terms cancel pairwise
 2



u
u2
2 0
2 1
( Ax )
c ( ) 1 2
c
c
1

(232)

c0

20

(235)
(236)

(237)

1
2
1 uc2

recall 2 =

1 + c 0 Ax

(238)

Ex (indeed)

(239)

~ =
~ A,
~
We can check By , recall B
By
= By0

Ax
Az

= 3 Ax 1 Az
z
x
= 03 A0x 01 A0z
 
u 
= 3 Ax 2 1 Az
c


u 3   u 0
3
= Ax 2 + 1 Az
c
c
=

(240)
(241)
(242)
(243)

Az
expand out and use By = 3 Ax 1 Az and Ez =

= 3 + c 0 Az
z
t

u 
(indeed)
(244)
= By + 2 Ez
c
|

Note the very important patterns

Ex = 1 + c 0 Ax
cF 01 = c 0 A1 c 1 A0
F 01 = 0 A1 1 A0

and

B y = 3 Ax 1 Az
F 31 = 3 A1 1 A3

In general
F = A A

(245)

is true.
Recall the Lorenz gauge

~ A
~

= 0 0

1 A1 + 2 A2 + 3 A3

= 0 A0

(247)

(248)

(246)

Thus the Lorenz gauge is Lorentz invariant. The general gauge transformations can also have a covariant
form


~0 = A
~ +
~
A0i = Ai + i
A
A0 = A +
(249)
00
0
0
A
=
A
+

0 =
t
and the differential equations for the potentials (in Lorenz gauge).
)


0
0
J0
~ 2 0 0 2 2 =

A
=

J
(cA0 ) = c
0
t
0
0
A = 0 J
~
i
i
Ai = 0 J i
~ 2A
~ 0 0 2 A
~
A
=

J
0
2
0
t
2.3.2

(250)

Conservation Laws

We have already seen the rewriting of the charge continuity equation into a Lorentz invariant (or scalar)
form:
Charge continuity equation: J = 0

(251)

We have 2 other continuity equations: energy and momentum and from the definition of 4-momentum,
we know that energy and momentum are closely related. Thus we expect the 2 continuity equations to
merge into one and has a similar form to the charge continuity equation, i.e. a 4-divergence form.
We recall from Part 1:
~ J~ = uEM
~ S
~
Energy continuity equation: E
t
21

(252)

~ + J~ B
~ =
~ T 1 S
Momentum continuity equation: E
c2 t

(253)

Since T is already a 3 3 matrix, we can go further and top up T into a 2-tensor T . We have 4

objects to insert: uEM , Sx , Sy and Sz . Since T is already symmetric, so T should also be symmetric,
thus we try

uEM Sx /c Sy /c Sz /c
Sx /c

Energy-momentum tensor: T =
(254)
S /c

T
y

Sz /c


~ + (~u B)
~
We recall from Minkowski force: K 1 = qF 1 u = q(u) E
so we try a similar expression
x

~ + J~ B.
~ So finally the ansatz for the combined continuity equation is
F J and it should contain E
Energy-momentum continuity equation: T = F J

(255)

We check this ansatz. For = 0,

T 0
0 T

00

+ 1 T

10

20

30

+ 2 T + 3 T
1 uEM
1~ ~
+
S
c t
c

=
=
=

F 0 J

(256)

01

02

03

F J1 + F J2 + F J3
1~ ~
E
J
c

(257)
(258)

now for = 1,
T 1
0 T

01

+ 1 T

11

21

= F 1 J

31

(259)

10

12

13

= F J0 + F J2 + F J3



Ex
=

(c) + Bz Jy By Jz
c


~ x
= Ex + (J~ B)

+ 2 T + 3 T

1 Sx  ~
+ ( T )
2
c t
x

(260)
(261)
(262)

So indeed T is the relativistic

generalization
of ( T ). Lastly we must ensure that T is really a


1
2
~ + and make a (really intelligent) guess that T should
2-tensor. We take Tij = 0 Ei Ej 2 ij E
have a similar form.


1
1

(263)
T =
F F + F F
0
4
We check
T 00

1
0

1
0 F F 0 + 00 F F
4


(264)

~2
2E
~2
recall that 00 = 1 and F F = 2 + 2B
c
!
~2
~2
1
E
1E
1 ~2
=
2 +
B
0
c
2 c2
2
|

(265)

1 ~2
1 ~2
0 E +
B
2
20
= uEM
=

(266)
(267)

~2

~ 2 , T 0i =
In the homework, you will check F F = 2cE2 +2B

Si
c

~ 2 + .
and T ij = 0 Ei Ej 12 ij E

At the end of Part 2, I will explicitly take 4-divergence of equation (263) show its conservation law.
Finally we just want to quickly mention the angular momentum 3-tensor,
Angular momentum 3-tensor definition: M = T x T x

(268)

Conservation law: M = 0 (without external sources)

(269)

22

2.3.3

Lorentz invariants in electrodynamics

We know that by forming dot products, we can make Lorentz scalars and they are invariant. They
have the same numerical value in any inertial frame. Recall
1. dot product of position 4-vector with itself gives the invariant interval
2. dot product of momentum 4-vector with itself gives the Einstein relation
3. dot product of derivative 4-vector with current density 4-vector gives the charge continuity
equation
There are 2 invariants in electrodynamics. The first one is the obvious dot product of the field strength
tensor with itself,
!
~2
E

2
~
(270)
F F = 2 B 2
c
In the homework, you will check some consequences of it and verify this invariant using the Lorentz
~ and B.
~
transformations of E
The second invariant is much less obvious. It is deduced from the second Maxwell equation F +
F + F = 0 that we can define the so called dual field strength tensor F by
1
F =  F with  being the 4D Levi-Civita symbol
2

(271)

and the Maxwell equation becomes F = 0. We check the 0-component,

F0

=
=
|
|

the odd permutations are 3012 , 1023 and 2031

then use F 12 = F 21 , F 23 = F 32 and F 13 = F 31

1 F 23 + 3 F 12 + 2 F 31

=
0

1
0 F
(272)
2


1
1032 1 F 32 + 1023 1 F 23 + 3012 3 F 12 + 3021 3 F 21 + 2013 2 F 13 + 2031 2 F 31 (273)
2
the even permutations are 3021 , 1032 and 2013

= F

23

+ F

12

+ F

31

(274)
(275)

Other components can be checked similarly. The second invariant is thus the dot product of F with
F
1
4~ ~
F F = F F =  F F = E
B (to be checked in the homework)
2
c

(276)

In the homework, you will also check some consequences of it and verify this invariant using the Lorentz
~ and B.
~
transformations of E
2.3.4

Point charge revisited: Lienard-Wiechert Potentials

Although all these rewriting of electrodynamics into Lorentz covariant form does not really contain any
new physics (except maybe relativistic corrections), it does provide a whole new perspective and a new
machinery to uncover more physics.
We shall rederive the Lienard-Wiechert potentials (i.e. retarded potentials for a moving point charge)
in relativistic language.
First we need to rewrite the (retarded) Greens function into a Lorentz invariant form.


1
R
GR (~r, t; ~r0 , t0 ) = (t t0 )
t t0
(277)
4R
c
| we need an identity which we will prove now






| start with (
xx
0 )2 = (x0 x00 )2 + |~r ~r0 |2 = R2 (x0 x00 )2


1
| then = (R x0 + x00 )(R + x0 x00 ) =
[(R ct + ct0 ) + (R + ct ct0 )]
2R
23

|
|
=
|
|
GR (
x, x
0 )

 



1
R
R

t + t0 +
+ t t0
and since delta function is even
2Rc
c
c
 



1
R
R
then =
t t0
+ t t0 +
2Rc
c
c


0 c
0 2
(t t ) (
xx
)
(278)
2
note that the extra delta function is killed by the step function
then =

note that (t t0 ) has the same effect as (ct ct0 ) = (x0 x00 )


c
(x0 x00 ) (
xx
0 )2
2

(279)

So the step function must now be explicitly written because we got an extra delta function. The step
function is invariant in the sense that if t > t0 in one inertial frame, then the transformed t >
transformed t0 in all other inertial frames.
Next is the 4-current density. We recall the point charge density and the point charge current density
c(~r, t)

~ r, t) = q~v (t) 3 (~r ~r0 (t))

= qc 3 (~r ~r0 (t)) and J(~

(280)

Note that we take all these coordinates and time to be measured in S-frame. In S-frame, we shall
parameterise the charged particles position 4-vector with the proper time ,
(x0 )
velocity 4-vector: u

(ct, ~r0 (t)) = (ct( ), ~r0 (t( )) = (ct( ), ~r0 ( ))

d~r0 (t)
d(x0 )
= (c, ~v ) where ~v =
=
d
dt

~ manifestly covariant, we introduce an extra time delta function,

To make J = (c, J)
Z
0
c(~r, t) = J
= qc d(t( )) 3 (~r ~r0 ( ))(t t( ))
|
=
|
c(
x) = J 0

~ r, t)
similarly, J(~

=
=

so, J (
x)

recall that dt = d and put c into (t t( ))

Z
2
qc
d 3 (~r ~r0 ( ))(ct ct( ))
write c = u0
Z
qc d u0 4 (
xx
0 )
Z
q d(t( ))~v 3 (~r ~r0 ( ))(t t( ))
Z
qc d ~v 4 (
xx
0 )
Z
qc d u 4 (
xx
0 )

Now, the retarded potentials (recall A =

~
c,A

(282)

(283)

(284)

(285)
(286)
(287)
(288)

),

Z
1
GR (~r, t; ~r0 , t0 )(~r0 , t0 )dV 0 dt0
0
| to write into d4 x
, we convert dt into unit of length
Z 
1
(~r, t)
1
c
1
=
(x0 x00 ) ((
xx
0 )2 )
J 0 (
x0 ) d4 x
0
c
0 c
2
c
c
Z
1
A0 (
x) =
(x0 x00 )((
xx
0 )2 )J 0 (
x)d4 x
0
20 c2
Z
~
~ r0 , t0 )dV 0 dt0
then, A(~r, t) = 0 GR (~r, t; ~r0 , t0 )J(~
Z 

c
~ x0 ) 1 d4 x
= 0
(x0 x00 ) ((
xx
0 )2 ) J(
0
2
c
1
| recall that 2 = 0 0
c
(~r, t)

(281)

24

(289)

(290)
(291)
(292)
(293)

1
20 c2

~ x0 )d4 x
(x0 x00 )((
xx
0 )2 )J(
0

So the covariant form is10

Z
1
(x0 x00 )((
A (
x) =
xx
0 )2 )J (
x0 )d4 x
0
20 c2
Z
| insert J (
x0 ) = qc d u 4 (
x0 x
0 )
then do the d4 x
0 integral using 4 (
x0 x
0 )
Z
q
d (x0 (x0 )0 )((
xx
0 )2 )u
=
20 c
| the delta imposes the invariant interval (
xx
0 (r ))2 = 0

(294)

(295)

(296)

which is called the light cone condition as zero invariant interval is on the light cone!

|
|

the step function imposes the retardation condition x0 > (x0 )0

X ( i )
 

to evaluate further, we use the identity (f ( )) =



i  df
 d =i 

then

d
d(x0 )
(
xx
0 ( ))2 = 2(x x0 ( ))
= 2(x x0 ( )) u
d
d
( r )
so (f ( )) =
|2(x x0 ( )) u | =r

(297)

Note that ( ) u is timelike and so is negative, thus | ( ) u | = ( ) u



u
q

=
(298)
40 c (x x0 ( )) u  =r
|

A (
x)

Figure 9: The 4D trajectory intersects the lightcone of the point of observation at 2 points. The retarded
point is picked out by the step function and is the one of physica interest. Any distance (invariant
interval) on the light cone is zero.

2.3.5

Point charge revisited: Fields of a charged particle in constant velocity

We now revisit the calculation in Part 1 where we calculated the electric and magnetic fields of a charge
with constant velocity. In the context of special relativity, we simply need to have the charge in S 0 -frame
and the point of observation to be in S-frame. S 0 -frame is moving with speed v along the x axis relative
to S-frame.
10 Strictly speaking, we need to check that the volume element d4 x
is a Lorentz scalar, i.e. the Jacobian = 1. This will
be checked later when we discusss about the Lagrangian of the EM field.

25

We set the charged particle at the origin of S 0 -frame. The point of observation in S 0 -frame coordinates
are (x0 , y 0 , z 0 ). As the charged particle is not moving with respect to S 0 -frame, there is only electrostatic
electric field and no magnetic field, so Bx0 , By0 and Bz0 = 0. The transformation of the fields become:
Ex0 = Ex
0 = Bx

Ey0 = v (Ey vBz
)

0 = v By + cv2 Ez
so, Ey0
Ey

and so, Ez0

Ez


v

= v Ey v 2 Ey
c
= v Ey0

 v

= v Ez + v 2 Ez
c
= v Ez0

~0 =
The Coulomb field as seen in S 0 -frame is E
~0
Ex0 = x
0 E

q
0 .
40 r 02 r

q
r0
sin 0 cos 0 r0
40 r02
p
q
x02 + y 02
x0
p
=
02
02
02
r0
4
0r
x +y

|
Ex = Ex0

similarly, Ey = v Ey0

=
=
=
=

(299)

(300)
(301)
(302)
(303)

Thus using spherical coordinates in S 0 -frame,

and, Ez = v Ez0

Ez0 = v (Ez + vBy
)

0 = v Bz cv2 Ey

q
x0
40 (x02 + y 02 + z 02 )3/2
insert x0 = v (x vt), y 0 = y and z 0 = z
q
v (x vt)
40 (v2 (x vt)2 + y 2 + z 2 )3/2
~0
v y0 E
v y
q
2
2
40 (v (x vt) + y 2 + z 2 )3/2
~
v z0 E
q
v z
2
2
40 (v (x vt) + y 2 + z 2 )3/2

(304)
(305)
(306)

(307)
(308)
(309)
(310)
(311)

So Ex obtained the gamma factor from coordinate transformation while Ey and Ez obtained the gamma
factor from field transformation!
To compare with the corresponding example in Part 1, we set the present position of the particle and
the point of observation to be in the xy plane (so z = 0).

Figure 10: Setup where point of observation and present position of particle are in the xy plane of
S-frame. We will shift the origin of S-frame to the present position of the particle.
We shift the origin to the present position of the particle and label the the new x coordinate as
xp = x vt. The position vector to the point of observation is ~rp = xp x
+ y y with xp = rp cos and
26

y = rp sin .
~
E

=
=
=

Ex x
+ Ey y
"
#
y
q
xp
v
+

3/2 x

3/2 y
40
v2 x2p + y 2
v2 x2p + y 2
v ~rp
q
40 2 x2 + y 2
3/2

(312)
(313)
(314)

v p

x2p

then, v2 x2p + y 2 =

q
v ~rp
40 r3 3 1 v2 sin2
3/2
p v
c2

1 vc2
rp
q
40 1 v2 sin2
3/2 rp2
c2

v2
c2

+ y2 =

x2p + y 2
1

v2 2
c2 y

v2
c2





v2
v2
= v2 rp2 2 rp2 sin2 = rp2 v2 1 2 sin2
c
c
(315)

(which is exactly as before)

(316)

Thus the reason for the E field to be pointing from the present position instead of from the retarded
position is because both Ex and Ey have the gamma factor and makes it a common factor and allows
the expression to become v (xp x
+ y y) v ~rp . The coincidence is that the gamma factors are obtained
in different ways stated earlier.
2.3.6

Point charge revisited: Lienards generalization of Larmors formula

2

0q
a2 and it is derived based on the
In Part 1, the Larmors formula for point charge radiation is P = 6c
assumption that it was at instantaneous rest or the particle was moving slowly (but accelerating).
We want to generalize it (well, it was generalized by Lienard) by finding out how power P transforms
under Lorentz transfomations and deduce a suitable expression that suits the type of Lorentz object P is.
2
0q
Our check is that when the speed of the object v is set to zero, we recover Larmors formula P = 6c
a2 .

P =

dE
dt

1 dE
d

recall that proper power is K 0 =

1
cK 0

(317)
1 dE
c d
(318)

So if we guess P to be a Lorentz scalar

P

then, K 0

=
|
=

0 q 2 2 guess for v 6= 0 0 q 2
du
a
=
(a a ) where we define a =
6c
6c
d
0 q 2
(a a )
6c2
recall u0 = c
0 q 2
(a a ) u0
6c3

(319)
(320)

(321)

which properly makes K 0 a 0-component of a 4-vector. Thus power P indeed is a Lorentz scalar so now
we simply need to work out a a in terms of ordinary velocity and acceleration.
a =

du
d

dt d
(cv , ~v v )
d dt

dv d~v
dv
= v c
, v + ~v
dt dt
dt


dv
d
1
1
2~v d~v
~v ~a
q
| so
=
=
2
= v3 2


3/2
2
2
dt
dt 1 v
c
dt
c
2 1 vc2
c2


~v ~a 2
~v ~a
=
v4
, v ~a + ~v v4 2
c
c
=

27

(322)
(323)

(324)

a a

(a0 )2 + ~a2

(325)

=
=
=
=
=


 

v ~a
v ~a
v ~a)
4
2~
2~
8 (~
+ v ~a + ~v v 2
~a + ~v v 2
v
c2
c
c


2
2
v ~a)
v ~a)
v ~a)2
4
4 (~
2
2 (~
2 4 (~
v v
+ a + 2v
+ v v
c2
c2
c4
"
1
2
v2
2


v4 a2 + (~v ~a)
+
+


2
2
v
2
c2 1 c2
c2 1 vc2
c4 1



2
v2
c2
4
2
2
+ 2
+ 2
v a + (~v ~a) 2
(c v 2 )2
c v2
(c v 2 )2


2
(~v ~a)
v4 a2 + 2
c v2

(326)
(327)
!#


v2 2
c2

So the generalized power is



0 q 2 4 2 (~v ~a)2
P =
v a + 2
6c
c v2
| we need to massage a bit to get to Lienards form


 
0 q 2 6 2
1
v2
2
=
a 1 2 + 2 (~v ~a)
6c v
c
c




0 q 2 6 2
1 2 2
=
v a 2 v a (~v ~a)2
6c
c
write ~v ~a = va cos so v 2 a2 (~v ~a)2 = v 2 a2 (1 cos2 ) = v 2 a2 sin2 = |~v ~a|2


0 q 2 6 2 |~v ~a|2
a
=
6c v
c2

(328)
(329)
(330)

(331)

(332)
(333)

(334)

This is Lienards formula for the power of a radiating point charge. Obviously when we set ~v = 0, we
get Larmors formula.
We shall now deal with the general expression of the power radiated per unit solid angle or dP
d . As we
are now considering a moving charge, essentially the same geometrical effect in retarded potentials also
occurs here, i.e. the power radiated by the charge is not the power that is passing through the sphere of
v
observation. They are related by the same geometrical factor 1 R~
c .
!

~radiated by particle
~passing through sphere = 1 R ~v S
(335)
S
c
~radiated by particle .
For ~v = 0, the Poynting vectors are the same. The expressions in Part 1 are actually S
Now we can continue the calculation for ~v 6= 0 using the equation from Part 1 just before we set ~v = 0,
~passing through sphere
S
!
!2
h
i h
i
~v
R
1
q
R
~ ((Rc
~v ) ~a) R
~ ((Rc
~v ) ~a) R(336)

=
1
R
~ ~v )3
c
0 c 40 (Rc R
~ ~v
~ ~v
~v
R
R
Rc R
=1
=
c
Rc
Rc

2
h
i h
i
~ ((Rc
~v ) ~a) R
~ ((Rc
~v ) ~a) = R2 R
((Rc
~v ) ~a)
| also write R

2
3
~v ) ~a)
R
R

((
Rc
2

1
q

=
R
~ ~v )5
0 c2 16 2 20
(Rc R
|

write 1

dP
d

~passing through sphere RR

2
= S

2
5
~v ) ~a)
R
R

((
Rc
2

q
=
~ ~v )5
16 2 0
(Rc R
28

(337)

(338)

(339)

q
16 2 0

2


~v ) ~a)
R ((Rc
~v )5
(c R

(340)

We shall now consider 2 cases: (i) where ~v and ~a are (instantaneously) colinear and (ii) where ~v and ~a
are (instantaneously) perpendicular to each other.11
Case of ~a k ~v :
dP
d

=
|
|
=

12

q2
16 2 0

~v ) ~a = cR
~a so,
When ~a and ~v are parallel, (Rc

2

~a)
R (cR

~v )5
(c R
~ (B
~ C)
~ = B(
~ A
~ C)
~ C(
~ A
~ B)
~
use vector identity A
(R
~a) = R(
R
~a) ~a(R
R)
~
so that R
2



q 2 c2 R(R ~a) ~a

(341)

(342)

~v )5
16 2 0 (c R
2



~a)2 + a2 2(R
~a)2 = a2 (R
~a)2
R ~a) ~a = (R
and R(

~v = v cos and R
~a = a cos , negative if decelerating
let ~v point along the z axis, R

q 2 c2
a2 sin2


16 2 0 c5 1 v cos 5
c
1
| then use c2 0 =
0
sin2
0 q 2 a2
=


2
16 c 1 v cos 5
=

(343)

(344)

We can integrate this to get the total power but we will just simply apply Lienards formula. Simply
write ~v ~a = 0 and get
P =

0 q 2 a2 6

6c v

(345)

Note that the expressions for dP

a parallel to ~v (accelerating) and ~a
d and P are applicable for both ~
anti-parallel to ~v (decelerating). The lobes (of power) are stretched towards the direction of ~v (here, the
z axis) although there is no radiation exactly in the direction of ~v . A typical example is bremsstrahlung
where an electron hits a metal and decelerates greatly with the emission of radiation.
Case of ~a ~v :

13

We shall set ~v along the z axis, ~a along the x axis so

= sin cos
~v = v
z , ~a = a
x and R
x + sin sin
y + cos
z

(346)

dP
d

and we shall calculate

first.

2


x + c sin sin
y + (c cos v)
z ) a
x)
dP
q 2 R ((c sin cos
=
d
16 2 0
(c v cos )5
11 The

logic for this entire section may seem messy for a reason. The natural logic should be:
1. include geometrical factor
2. calculate

dP
d

3. integrate over sphere to get total power P

4. work out 2 special cases of ~a k ~v and ~a ~v
The integration over sphere to get P is too difficult to carry out so I changed the logic to:
1. deduce total power P is a Lorentz scalar
2. generalize acceleration to a 4-vector so that P becomes a Lorentz scalar
3. then include geometrical factor to get

dP
d

4. work out 2 special cases of ~a k ~v and ~a ~v

12 You
13 You

may take ~a k ~v at an instant where time is tr .

may take ~a ~v at an instant where time is tr .

29

(347)

2



R

(c
sin

sin

z
+
(c
cos

v)
y
)


q a


5
2
16 0
c5 1 vc cos
2 2

(348)

divide c2 in both numerator and denominator





x

y
z



 sin cos sin sin
cos
| the cross product is 

v


0
cos
sin sin 

c



v 
v
| which = sin2 sin2 + cos cos
x
sin2 sin cos
y + sin cos cos
z
c
c





2
2
sin2 sin2 + cos cos vc
+ sin4 sin2 cos2 + sin2 cos2 cos vc
q 2 a2
=


5
16 2 0 c3
1 vc cos
v
| terms without = sin4 sin4 + cos4 + 2 sin2 sin2 cos2 + sin4 sin2 cos2 + sin2 cos2 cos2
c
| |then 1st + 4th = sin4 sin2 (sin2 + cos2 ) = sin4 sin2
|

|
|
|
|
|
|
|
|
|

|then 3rd + 5th = sin2 cos2 (2 sin2 + cos2 ) = sin2 cos2 (sin2 + 1)
v
then terms without = sin4 sin2 + cos4 + sin2 cos2 sin2 + sin2 cos2
c
|then 1st + 3rd = sin2 sin2 (sin2 + cos2 ) = sin2 sin2
|then 2nd + 4th = cos2 (cos2 + sin2 ) = cos2
v
then terms without = sin2 sin2 + cos2 = sin2 (1 cos2 ) + cos2 = 1 sin2 cos2
c
v
v
v
v
terms with = 2 cos3 2 sin2 sin2 cos 2 sin2 cos cos2
c
c
c
c
v
v
v
2
2
3
2
| = 2 (cos + sin cos (sin + cos )) = 2 cos (cos2 + sin2 ) = 2 cos
c
c
c
v2
v2
v2
2
2
2
terms with 2 = 2 cos + 2 sin cos
c
c
c
1
2
then write 0 c =
0
2

0 q 2 a2 1 sin2 cos2 2 vc cos + vc2 cos2 + vc2 sin2 cos2

=

5
16 2 c
1 vc cos

2
v
v2
v
| the 3 terms, 1 2 cos + 2 cos2 = 1 cos
c
c 
c

2 
2
v
v2
2
0 q 2 a2 1 c cos 1 c2 sin cos
=

5
16 2 c
1 v cos

(349)

(350)

We can integrate this to get the total power but we will just simply apply Lienards formula.


0 q 2 6 2 |~v ~a|2
P =
a
6c v
c2
| where ~v ~a = va
y


0 q 2 a2 6
v2
=
v 1 2
6c
c
2 2
0 q a 4
=

6c v

(351)

(352)
(353)

We see that when we set = 0, dP

v for vc 1. A typical example
d is sharply peaked in the direction of ~
is circular motion where the electron emits (so-called) synchrotron radiation tangentially to the circular
trajectory.

30

Lagrangian description of a classical relativistic U (1) gauge

theory

3.1

Review of Lagrangian mechanics and Noethers theorem

For a classical system described by n-generalized coordinates: q1 (t), . . . , qn (t), the Lagrangian is defined
by
L(q1 (t), . . . , qn (t); q1 (t), . . . , qn (t); t)

= T V

(354)

where T is the KE and V is the PE. Note that in field theory (where we deal with infinite degrees of
freedom), the form L = T V may not really be valid as we will see later.
Hamiltons principle states that the trajectory in configuration space that extremizes (maximum or
Z tf
minimum) the action S =
Ldt is the actual (or physical) trajectory of the system.14
ti

Extremum means: S
Z

(355)

L(q1 (t), . . . , qn (t); q1 (t), . . . , qn (t); t)dt =

(356)

tf

=
ti

The requirement of extremum translates into n Euler-Lagrange equations of motion.



d L
L

=0
qi
dt qi

(357)

The next thing we want to mention is Noethers theorem which is the formal way of deriving conservation
laws in a theory. Noethers theorem states: for each symmetry of the theory (i.e. transformations that
does not change the Lagrangian), there is a conserved quantity associated to it.
Suppose the Lagrangian is invariant under a change of coordinates with a small parameter ,
1st order change: qi0 = qi + Ki (q)

(358)

where Ki (q) may be a function of all the qi s, which we collectively denote as q.

We take the Lagrangian to be invariant to first order
L(q 0 , q0 )

but L(q, q)


L(q, q)

Ki +
Ki
qi
qi
i


L
d L
from Euler Lagrange,
=
qi
dt qi




X d L
L dKi
Ki +
dt qi
qi dt
i
!
d X L
Ki
dt
qi
i
X  L(q, q)

so the conserved quantity is

= L(q + K, q + K)


X  L(q, q)
L(q, q)

Ki +
Ki
L(q, q)
+
qi
qi
i

(359)
(360)

(361)

(362)

(363)

(364)

X L
Ki .15

i
i

14 We

assume fixed end points at ti and tf .

case of cyclic coordinates means its conjugate momenta is conserved, is a special case. Suppose the Lagrangian
does not depend on q3 , so q30 = q3 +  does not change the Lagrangian. Thus K3 = 1


L
dp3
d
K3 = 0 =
=0
(365)
dt q3
dt
15 The

31

3.2

Relativistic Lagrangian of a charged particle interacting with external

EM field

1 d~x
m

2 dt
1
d~x
= mv 2 is obviously not compatible with special relativity. We have to make an intelligent guess of
dt
2
the free relativistic particle Lagrangian. To do that, we need to lay out the features that it should have.
We start with a free (i.e. non-interacting) relativistic particle for simplicity. However L = T =

Following axiom 1 of Special Relativity, the action S should be a Lorentz scalar so that the action
R
d =dt R
gives us the same physics in every inertial frame. Then we can write S = Ldt =
Ld and
since proper time is invariant, L should be a Lorentz scalar.
In the non-relativistic limit v << c, we should have the familar form: L = 21 mv 2 .
The Hamiltonian can be obtained by a Legendre transform of the Lagrangian L. From Einstein
2
2 2
2 4
relation, we know the
qenergy of a relativistic particle is E = p~ c + m c , so the Hamiltonian is
~ m2 c4 . Note that this expression is exact.
expected to be H = p2 c2 +
We take L = where is some scalar and determine by taking v << c,
r


v2
1 2
v2

v
L= = 1 2 1 2 =

c
2c
2 c2

(366)
r

v2
Thus the relativistic Lagrangian is L = mc 1 2 . The
c


L
d L
equations of motion are obtained from the Euler-Lagrange equations:

=0
~x
dt ~v


d L
so,
= 0
(367)
dt ~v
!
r

v2
2 d
= 0
(368)
1 2
mc
dt ~v
c

2~
v

d
2
= 0
mc2 q c
(369)
dt 2 1 v2
We require L =

1
2
2 mv ,

2 16

so = mc .

c2

d
(m~v )
dt
So the conjugate momentum is p~ =
H

L
~
v

(370)

= m~v . The Hamiltonian is derived from the Legendre transform,

= ~x p~ L

(371)
r

= ~v (m~v ) + mc2

v2
c2

(372)

m~v
p2 c2
, write v 2 = 2
from p~ = q
2
p + m2 c2
1 vc2

p 2 c2
mc2
+
p2 + m2 c2

then, = q

1
1

v2
c2

(373)

=r
1

1
c2

1


p
p2 c2
p2 +m2 c2

=

p2 + m2 c2
mc

=
=
16 We

p 2 c2
m2 c3
p2 + m2 c2
p
+
mc
p2 + m2 c2
p2 + m2 c2
p2 c + m2 c3
p
p2 + m2 c2

can ignore the first term , as it does not contribute in the Euler-Lagrange equations.

32

(374)
(375)

=
=

p
c p2 + m2 c2
p
p2 c2 + m2 c4

(376)
(377)

and indeed
we get the expected Hamiltonian. We are confident of the free relativistic Lagrangian L =
r
v2
mc2 1 2 so now we shall include the interaction with an external EM field. We take inspiration
c
R
R
~
~ . For a particle
from the electric potential energy q and the magnetic energy J~ AdV
= ~v AdV
3
~ We let 1 and 2 be possible
with trajectory ~r(t), = q (~r ~r(t)) so the magnetic energy is q~v A.
(dimensionless) constants, so the charged particle interacting with external EM Lagrangian should be
r
v2
2
~
(378)
L = mc 1 2 + 1 q + 2 q~v A
c
d
~ + q(~v B).
~
We shall determine 1 and 2 by requiring the equations of motion become m~v = q E
dt
So,
L
d L

= 0 (379)
~x
dt ~v
!
r
~
v2

~v A
d
d
~ = 0 (380)
1 q
+ 2 q

mc2 1 2 2 q
~v A
~x
~x
dt ~v
c
dt ~v
!
r
2
v

recall that
mc2 1 2 = m~v |
~v
c

~ and ~v A
~=A
~ |
=
~x
~v
~
~ 2 q dA d (m~v ) + 2 q (~
~ v A)
~ =
1 q
dt
dt
~ v A)
~ = ~v (
~ A)
~ +A
~ (
~ ~v ) + (~v )
~ A
~ + (A
~ )~
~ v |
use vector identity (~
note that

since ~v and ~x are treated as independent, 2nd and 4th terms in the identity are zero |
~
~
~ A
~=B
~ and dA(~x, t) = A + (~v )
~ A
~ |
write
dt
t
~
~ A
~ d (m~v ) + 2 q~v B
~ + 2 q(~v )
~ A
~ =
~ 2 q A 2 q(~v )
1 q
t
dt
~
~ 2 A = E,
~ we need 1 = 1 and 2 = +1 |
for 1
t
~ d (m~v ) + q~v B
~ =
qE
dt
r
so, L = mc

v2
~
q + q~v A
c2

and to get the Hamiltonian, we first need the conjugate momentum

!
r
L

v2
2
~ = m~v + q A
~
p~ =
=
mc 1 2 + q~v A
~v
~v
c

0 (381)

0 (382)

0 (383)

(384)

(385)

Then perform the Legendre transform

H

= ~v p~ L
r


2
2
~ + mc 1 v + q q~v A
~
= ~v m~v + q A
2
c
r
v2
= mv 2 + mc2 1 2 + q
c
~ 2 c2
(~
p q A)
~ and replace all v 2 =
| write m~v = p~ q A
~ 2 + m2 c2
(~
p q A)
33

(386)
(387)
(388)

q
=

~ 2 c2 + m2 c4 + q
(~
p q A)

(389)

Since we must have L to be a Lorentz scalar, we need to check if the 2 extra terms fulfil it.
~
L = mc2 q + q~v A
 

~
= mc2 q(c)
+ q~v A
c
|

(390)
(391)

recall 4-velocity u = (c, ~v ) and 4-potential A =

~
,A
c

= mc2 + qu A

(392)

Thus the second term is indeed a Lorentz scalar since (total) charge is Lorentz invariant.
The last thing to do in this section is to put the first term into a (manifestly) Lorentz scalar form

and derive the covariant equation of motion K = dp

u . We look at only the first term (free
d = qF
particle term) and we also think about Hamiltons principle which Ris to extremize the trajectory and ask
Where is the meaning of trajectory in the action expression S = (mc2 )d ?
To answer that question, we have to go back to the invariant interval which is the trajectory (or
worldline) in Minkowski space!17
Recall, ds2

ds =
thus, S

=
|
=
=

c2 d 2 = c2 dt2 + dx2 + dy 2 + dz 2 = dx dx
p

c2 d 2 = 1cd
Z
1
mc ds
1
r
p
p
dx dx

but ds = dx dx =
d = u u d
d d
Z r
u u
mc
d
1
Z
p
mc
u u d

(393)
(394)
(395)
(396)
(397)
(398)

Our final invariant action is

Z
S = mc

u u d +

qu A d =

Ld

We can deduce the (covariant) Euler-Lagrange equations to be

L
d L



x
d dx

(399)

(400)

d
d u

L
d L

= 0
x
d u

mc u u + qu A +
(qu A ) = 0
x
d mcu
dA
A

q
+ qu
= 0

d u u
d
x

note that u u = c2 from beginning of Part 2

dp
dx A

q
+ qu A
d
d x

(401)
(402)
(403)

|
=

(404)

qu ( A A )

qu ( A A )

qu F

dp
d
dp
d
dp
d

(405)
(406)
(407)

17 Actually we are parameterizing the worldline using the proper time as parameter. This is natural especially for
particles with mass, the notion
q of proper time is meaningful. In more general terms, we can use some parameter and
p
dx
write ds = dx dx = dx
d.
d d

34

3.3

Lagrangian of the EM field

The field has infinite degrees of freedom because it has a value (or values) at every spacetime point
and there are infinitely many points. Thus the Lagrangian has to be indicated at every point, i.e. the
Lagrangian density is the relevant quantity.
Z
Lagrangian: L =
Ld3 ~x
(408)
Z tf
Z tf Z
Ld3 ~xdt
Action: S =
Ldt =
(409)
ti

ti

To determine if the Lagrangian density L is a Lorentz scalar or not, we look at how the volume element
d3 ~xdt transforms under Lorentz transformations. To do that, we look at the Jacobian.

dt0 dx0 dy 0 dz 0

=
=

t0
t
x0
t
y 0
t
z 0
t

t0
x
x0
x
y 0
x
z 0
x

t0
y
x0
y
y 0
y
z 0
y

t0
z
x0
z
y 0
z
z 0
z






 dtdxdydz





u 
with t0 = t 2 x , x0 = (x ut), y 0 = y and z 0 = z
c


u


0
0 
2
c

 u

0 0 

dtdxdydz
 0
0
1 0 

 0
0
0 1 




 0 0 
 u 0 0 




u
 dtdxdydz
 0 1 0  +  0
1
0




c2 
 0 0 1 
0
0 1 


u2
2 2 2 dtdxdydz
c
dtdxdydz

(410)

(411)

(412)

(413)
(414)

Thus the volume element is a Lorentz scalar and so for the action to be a Lorentz scalar, the Lagrangian
density L must be a Lorentz scalar also. The dependences in L shall be deduced as follows:
q(t)

A (
x)

q(t)

A (
x) (now A (
x) depends on 4-position, so all derivatives must be considered)

L(q(t), q(t))

L(A (
x), A (
x))

To get the Euler-Lagrange equations of motion, we shall carry out the variational calculation,

S
Z tf Z

=
ti
Z tf

=
|
|
|
=

Variation: A0 (
x) = A (
x) + A (
x)

(415)

Fixed endpoints: A (~x, ti ) = 0 = A (~x, tf )

(416)
(417)

d3 ~xdtL

(418)


L
L
A +
( A )
(419)
A
( A )
ti
the variation and the partial derivative commutes: ( A ) = (A )
1
(one dimensional) proof: (x A ) = lim (A0 (x + ) A0 (x) (A (x + ) A (x)))
0 
1
1
| = lim (A0 (x + ) A (x + ) (A0 (x) A (x))) = lim (A (x + ) A (x)) = x (A )
0 
0 


Z tf Z
L
L
d3 ~xdt
A +
A
(420)
A
( A )
ti
Z

d3 ~xdt

35

integrate by parts on the second term







Z tf Z
L
L
L
=
d3 ~xdt
A +
A
A
A
( A )
( A )
ti
 tf



Z tf Z
Z

L
L
L
3
3
=
d ~xdt

A +
d ~x
A 
A
( A )
( A )
ti
ti

(421)
(422)

the last term is zero due to the fixed endpoints condition. Since A is an arbitrary variation, S = 0
only if


L
L

=0
(423)
A
( A )
which are the desired Euler-Lagrange equations of motion for fields. Since L is a Lorentz scalar, the
equations of motion are Lorentz covariant equations (as they should be).
The Euler-Lagrange equations work also for fields interacting with external sources. In the case of
electrodynamics with external sources, we expect to get F = 0 J . For the free (or non-interacting)
part of the Lagrangian, we only have 2 Lorentz scalars to choose from: F F or F F . We choose
F F . For the interaction part, we choose the Lorentz scalar J A which is the usual vector coupling
term inspired from the previous section.
L = 1 F F + 2 J A (where 1 and 2 are constants to determine)

(2 J A )

A
( A )



L
L

A
( A )
recall that F = A A

1 ( A A )( A A )

(424)

(425)

(426)

|
=

( A )
= A = A
( A )
2 J 1 (2 A 2 A 2 A + 2 A )

(427)

2 J 41 F

(428)

expand and note A

(429)
We want F = 0 J so 1 = 14 and 2 = 0 .
1
L = F F + 0 J A
4

(430)

Note a bonus identity from the above working: (A ) 14 F F = F .
Next, we shall look at the conservation laws or Noethers theorem in field theory. As usual, we only
look at 1st order (symmetry) transformations.
Coordinate transformation: x0 = x + x
A0 (
x0 )

Corresponding change in field:

= A (
x) + A (
x)
(
Define a field variation: A
x) = A0 (
x) A (
x)

(431)
(432)
(433)

which is variation of the field while keeping the coordinates fixed. This variation is useful because it
commutes with the partial derivative (as we have seen earlier). The 2 variations are related by (keeping
everything to 1st order),
(
A
x)

= A0 (
x) A (
x)
=
=
=
=

x) A0 (
x0 ) + A0 (
x0 ) A (
x)
A0 (
0
0
0
(A (
x ) A (
x)) + A (
x)
A (
x) (A0 (
x + x
) A0 (
x))
0
A (
x)
A (
x)
x
x
36

(434)
(435)
(436)
(437)
(438)

= A (
x) ( A0 (
x))x
|

to 1st order, we can write

(439)
A0 (
x)

A (
x)

= A (
x) ( A (
x))x

(440)

We demand that these symmetry transformations leave the action invariant (to 1st order).
0

= S
Z Z
=
Ld3 ~xdt
Z Z
Z Z
3
=
(L)d ~xdt +
L(d3 ~xdt)

(441)
(442)
(443)

the variation of the volume element is found by calculating the Jacobian


 x00 x00


x0
 1 + x00



1
1
x01
x
x

 x010
1
1
x
 x
 d3 ~xdt =  x
1 + x
 d3 ~xdt
| d3 ~x0 dt0 =  x0
x1
x0
x1

 .

..
..
..
.. 
. . 

 ..
.
.
.
.
.



x
| then d3 ~x0 dt0 1 +
d3 ~xdt = (1 + x ) d3 ~xdt so (d3 ~xdt) = ( x )d3 ~xdt
x
Z Z
=
(L + L( x )) d3 ~xdt

(444)

(
+ ( L)x
recall A
x) = A (
x) ( A (
x))x and infer L = L
Z Z


+ L( x ) + ( L)x d3 ~xdt
=
L
Z Z


+ (Lx ) d3 ~xdt
=
L
|

(445)
(446)

L
= L A
+
so, L
( A )
A
( A )




L
L
L
L

A
A +
A
A
| add & subtract a term, L =
A
( A )
( A )
( A )





L
L
L

| then L =
A +
A

A
( A )
( A )
| using Euler-Lagrange equations, the first term is zero


Z Z
L
=

A + Lx d3 ~xdt
(447)
( A )
= A ( A )x
| recall A




Z Z
L
L

A
( A ) L x d3 ~xdt
(448)
( A )
( A )
|

We are thus invited to postulate (not rigorous)18 local conservation laws of the form J = 0. The
so-called Noethers current density is thus defined as19


L
L

J =
A
( A ) L x
(449)
( A )
( A )
Now with regards to the energy-momentum tensor we saw earlier, we shall apply Noethers theorem to
the special case where the action is invariant under 4D translations,
x0 = x +  = x = 

(450)

and the fields do not change under the 4D translations

A0 (
x0 ) = A (
x) = A = 0

(451)

18 I have browsed through classical/quantum field theory books and Goldstein, and I am unable to find any satisfactorily
rigorous derivation of this.
19 Be careful that Noethers current density and the external current density have the same notation but they may mean
different quantities.

37

= J


L
0 =
( A ) L 
( A )
| since  is arbitrary,


L

( A ) L
0 =
( A )
| raise the index by multiplying with


L
( A ) L
0 =
( A )
0

(452)
(453)

(454)

(455)

Thus we can define the (canonical) non-symmetric energy-momentum tensor ,20

=

L
( A ) L
( A )

(456)

Note that since we are discussing 4D translational invariance, we have to use L = Lfree = 14 F F .
Interaction with external sources is not translationally invariant.
The quantity that is conserved is revealed when we integrate 0 = J over 3D space.
Z
0 =
J dV
(457)
Z
Z
~ JdV
~
=
0 J 0 dV +
(458)
|
=
|
=

use divergence theorem on second term

Z
I
d
0
~
J dV + J~ dA
dx0
assume the fields and therefore J~ fall off sufficiently fast
Z
d
J 0 dV
dt

(459)

(460)

R
Thus J 0 dV is the conserved quantity.21 RFor the (canonical) energy-momentum tensor , where
= 0, the 4 conserved quantities are 0 dV . These 4 conserved quantities turn out to be the
energy and the momentum.


Z
1
E ~
0

dV = P =
,P
(461)
c
c
Earlier we have a symmetric energy-momentum tensor T and it is actually related to by a 4divergence.22
T = + (where we require = )

(462)

This relationship does not affect the conservation law.

T

=
|
|
=
|
=

(463)

from Noethers theorem, = 0

1
split = ( + )
2
1
1

+
2
2
in second term, rename indices and
1
1
+
2
2

20 Note

(464)

(465)

that this energy-momentum tensor is actually not gauge invariant!

R
~ where J 0 = c and so J 0 dV = Q, the total charge, is the
quick example is to recall the 4-current J = (c, J)
conserved quantity.
22 If you are interested, you can read pg 112-114 of Electrodynamics and Classical theory of Fields and Particles by
A.O. Barut on how to get T directly from variation.
21 A

38

(466)

partial derivatives commute =

1
( + )
2
recall the requirement =

(467)

which is the conservation law for the non-interacting case. The conserved quantities are also not affected.
Z
Z


T 0 dV =
0 + 0 dV
(468)
Z
Z
Z
=
0 dV + 0 00 dV + i i0 dV
(469)
but 00 = 0 since 00 = 00

|
|

use divergence theorem in the 3rd term and assume falls off sufficiently fast
Z
=
0 dV
(470)
Now we shall take the explicit (free field) expressions of and T and work out .

=
|
=

L
( A ) L
( A )
1
L
use L = F F and we already know
= F
4
( A )
1
F ( A ) + F F
4



1
1

=
F F + F F
0
4
1
| replace F F = + F ( A )
4
1
= [ F F + F ( A ) + ]
0
1
= [F F + F ( A ) + ]
0
1
= [F ( A A + A ) + ]
0
1
= [F ( A ) + ]
0
| use product rule on the first term and F = 0 (for free field)
1
= [ (F A ) + ]
0
| write F = F
1
= [ (F A ) + ]
0

(471)

(472)

(473)

(474)
(475)
(476)
(477)

(478)

(479)

Thus = 10 F A which is indeed antisymmetric in the first 2 indices.

As a nice ending to Part 2, we shall take the interacting case and work out explicitly that T =
F J .


1
1
T =
( F F ) + ( F F )
(480)
0
4


1
1
( F )F + F ( F ) + (( F )F + F ( F ))
(481)
=
0
4
|

because we can write F ( F ) = F ( F ), the last 2 terms add

the first term F = 0 J which are Maxwell equations

39



1
1
0 J F + F ( F ) + F ( F )
0
2
for 1st term: J F = F J , for 2nd term: rename and split it,

=
|

(482)

for 3rd term: rename


1 
= F J
F ( F + F + F )
(483)
20
| use homogenous Maxwell equation F + F + F = 0

1 
= F J
F ( F F )
(484)
20
| write F = F

1 
= F J
F ( F + F )
(485)
20
| terms in round brackets are symmetric in but F is antisymmetric in
= F J (indeed)

(486)

40