Lecture Notes on Electromagnetism

Abstract. The contents of this text is based on the class notes on Electro-
magnetism for the PH412 course by Prof. Ananda Dasgupta, IISER Kolkata.
 Contents

Chapter 1. Introduction 1
1.1. Lecture 1 : Covariance & Lorentz Transformation 1
1.2. Lecture 2 : Tensors 5
Chapter 2. Discovering Electromagnetism 9
2.1. Lecture 3: Obtaining Lorentz Force Law 9
2.2. Lecture 4: Manifestly Covariant form of the Lorentz Force Law 13
2.3. Lecture 5: Least Action Principle 15
2.4. List of Exercises 17
2.5. Lecture 6: Electric field of an Uniformly Moving Point Charge 18
2.6. Lecture 7: Maxwell’s Equations from the Field Tensor 20

iii
 CHAPTER 1

Introduction

1.1. Lecture 1 : Covariance & Lorentz Transformation

1.1.1. Meaning of Covariance. Consider the following Maxwell equation

∇.E = −ρ/0
and the transformation of the co-ordinates

0
xµ → x µ
The equation is said to be covariant under the given transformation if both
sides of it vary in such a way, that in the transformed(primed) co-ordinate system,
the equation again holds true i.e.,

0 0 0 0
∇ .E = −ρ /0
Now consider the vector identity

~ BX
AX( ~ C)
~ = B(
~ A.
~ C)
~ − C(
~ A.
~ B)
~

If we are asked to verify this, one of the simple ways is to choose the vectors in
a convenient way viz.

~
C = cî
~
B = b1 î + b2 ĵ
~ =
A a1 î + a2 ĵ + a3 k̂
It is then easy to see that both sides of the equation are equal. Now we can
rotate the co-ordinate system in any way but this equation identity will still hold
true due to it’s covariance under rotation. Some equations are manifestly covariant
eg., Aµ = B µ . Since the components are equal they will change equally under any
transformation.

1.1.2. The Lorentz Transformation. The distance between two points (xµ
and y µ ) in flat space-time is :

(1.1) l2 = ηµν (xµ − y µ )(xν − y ν )

where, ηµν = diag(1, −1, −1, −1)
1
2 1. INTRODUCTION

0 0
A transformation of the co-ordinates xµ → x µ & y µ → y µ such that the
distance l2 is preserved is called a Lorentz transformation i.e., Lorentz transfor-
mation preserves the interval between any two events in space-time. As a special
case, the distance between the origin and any space-time point xµ (= xµ xµ ) is pre-
served by Lorentz transformation. Note that simple translations of the co-ordinate
axes can also preserve space-time intervals. However we are not interested in such
transformations.
An example of a simple Lorentz transformation is:

0 0
ct ≡ x 0 = γ(ct − βx)
0 0
x ≡ x 1 = γ(−βct + x)
0 0
y ≡ x2=y
0 0
z ≡ x3=z
It can be represented in the matrix form as:

0 
x0
   0 
γ −γβ 0 0 x
0
 x1   −γβ γ 0 0   x1 
(1.2)  0 =  
 x2   0 0 1 0   x2 
x3
0
0 0 0 1 x3

A Lorentz transformation can be written as

0
X = LX
Let (L)µν = Λµ ν
0
xµ = Λµ ν xν
where the summation convention has been used. If an index appears twice in
the same expression (even in the same variable), once upstairs and once downwards
then a sum is carried over that index. Also note that the Greek indices (µ, ν, etc)
take on values 0, 1, 2, 3 while the Latin indices (i, j, k, etc) take on the values 1, 2,
3. xµ ≡ (x0 , xi ) ≡ (ct, x). Now,

0 0
xµ xµ = x µ xµ
0 0
=⇒ X T ηX = X T ηX = (LX)T η(LX)
=⇒ X T ηX = X T (LT ηL)X
Does this mean LT ηL = η ? Yes it does and the key lies in the fact that η is
symmetric.
Proof:
If we have,
 1.1. LECTURE 1 : COVARIANCE & LORENTZ TRANSFORMATION 3

X T CX = 0, we could choose X µ = δµν to conclude Cνν = 0. We could also

choose X µ = δµν + δµρ to conclude Cρν = −Cνρ (ν 6= ρ). Thus C T = −C i.e., C is
an anti-symmetric matrix.
Thus we can conclude

LT ηL = η + C

where C is any anti-symmetric matrix. If we take the transpose on both sides

and use η T = η, we get C T = C , but C T = −C . Thus C = 0. We get

(1.3) LT ηL = η

If we take the determinant of the matrices on both sides, we get det(L) = ±1.
It can be shown that (Λ0 0 )2 ≥ 1 [Exercise 1]. Based on the last two statements,
there are four classes of Lorentz transformations viz.

(1) |L| = 1, Λ0 0 ≥ 1 denoted L↑+ eg. the transformation matrix at the begin-
ning of this section.
(2) |L| = 1, Λ0 0 ≤ −1 denoted L↓+ eg. P.T where P is the parity transforma-
tion and T is the time reversal transformation.
(3) |L| = −1, Λ0 0 ≥ 1 denoted L↑− eg. P = diag(1, −1, −1, −1).
(4) |L| = −1, Λ0 0 ≤ −1 denoted L↓− eg. T = diag(−1, 1, 1, 1).

It can be shown that Lorentz transformations (L) form a group with the use of
the fact that L is a linear transformation such that LT ηL = η [Exercise 2] It
can also be shown that only L↑+ forms a proper subgroup of the group of Lorentz
transformations [Exercise 3]. This is known as the proper orthochronous Lorentz
transformation.
0
Under Lorentz transformation, the co-ordinates transform as X = LX. Any
four-vector also transforms in the same way. However the product Σµ Aµ B µ is not
conserved under Lorentz transformation. In 3D, however, this does hold true since
the transformation matrix, R (the rotation matrix) is orthogonal. Thus, in 3D

0 0
X TY = XT Y
4 1. INTRODUCTION

1.1.3. Summary. In summary the following are the properties of Lorentz

transformations (L) :

• They preserve the interval between two space-time events.

• LT ηL = η where η = diag(1, −1, −1, −1). The set of Lorentz transforma-
tions form a group.
• det(L) = ±1, Λ0 0 ≥ 1 or Λ0 0 ≤ 1. Thus there are four classes of Lorentz
transformations of which only L↑+ is a proper subgroup of the Lorentz
group.

1.1.4. List of Exercises.

(1) Show that (Λ0 0 )2 ≥ 1

(2) Show that Lorentz transformations (L) form a group with the use of the
fact that L is a linear transformation such that LT ηL = η
(3) Show that only L↑+ forms a proper subgroup of the group of Lorentz
transformations.
 1.2. LECTURE 2 : TENSORS 5

1.2. Lecture 2 : Tensors

1.2.1. Covariant and Contravariant. Quantities which transforms like the
co-ordinate differentials under co-ordinate transformations are called contravariant
vectors. Suppose under a transformation L, the co-ordinate vectors X transform as
(1.4) X 0 = LX
This can also be written as
(1.5) X 0µ = Λµ ν X ν
where
(1.6) Λµ ν = Lµν
then the transformation of the differentials of the coordinates will be given by the
usual relation1,
∂X 0µ
(1.7) dX 0µ = dX ν
∂X ν
Now any quantity which transforms as
∂X 0µ ν
(1.8) A0µ = A
∂X ν
is defined as a contravariant vector or simply a vector.
∂X 0µ
The quantity can be re-written as
∂X ν
∂X 0µ
= ∂ν (Λµ ρ X ρ )
∂X ν
= Λµ ρ ∂ν X ρ
= Λµ ρ δ ρ ν
(1.9) = Λµ ν
3
X
Now let us have a look at quantities like Aµ B µ .
µ=0
3
X 3
X
A0µ B 0µ = (Λµ ν Aν )(Λµ ρ B ρ )
µ=0 µ=0

In matrix notation this stands as

A0 = LA
0
B = LB
A0T B 0 = AT LT LB
Evidently, it is not an invariant quantity, since LT L need not necessarily be 1.
Let us see, if we can make the quantity A0 B 0 invariant by choosing some other
transformation rule for B.
A0 = LA
0
B = MB
0T 0 T
A B = A LT M B
1The differentials do transform linearly even under any arbitrary non-linear transformation
X 0µ = X 0µ (X ν )
6 1. INTRODUCTION

By choosing
(1.10) M = (LT )−1
we can ensure
A0T B 0 = AT B
The quantity B which transforms like
B 0 = (LT )−1 B
is defined to be a covariant vector or a covector2. A covariant vector (covector)
maps a contravariant vector (vector) linearly to a scalar. They can be thought of
as dual vectors similar to bras and kets.
(1.11) B T A −→ scalar
~
An example of covariant vector is the gradient of a scalar, ∇φ.
~ · d~r = dφ
∇φ
~ maps the vector d~r to a scalar dφ. We can explicitly check how ∇φ
∇φ ~ transforms.

~ = ∂φ µ
∇φ e
∂X µ
(1.12) = ∂µ φeµ

∂φ0
∂µ0 φ0 =
∂X 0µ
∂φ
=
∂X 0µ
∂φ ∂X ν
=
∂X ν ∂X 0µ
∂X ν
(1.13) = ∂ν φ
∂X 0µ
In the equations above, it should be kept in mind that, a scalar remains invariant
under coordinate transformation, but the functional form of the scalar obviously
changes. φ = φ(X µ ) = φ0 (X 0µ ).
∂X ν
We can easily check that the quantity corresponds to ((LT )−1 )µν .
∂X 0µ

X0 = LX
−1 0
⇒ L X =X
⇒ X = L−1 X 0
X
⇒ Xν = (L−1 )νρ Xρ0
ρ
X
⇒ Xν = (L−1 )Tρν Xρ0
ρ

(1.14) ⇒ Xν = Λρ ν Xρ0

2Under a coordinate transformation, it is the vector components which change. Hence if we

write a vector A = Aµ êµ , under a transformation, the basis vectors êµ must transform opposite to
the vector components Aµ in order to keep the physical quantity A invariant. Covectors transform
oppositely to that of vectors, just like the basis vectors
 1.2. LECTURE 2 : TENSORS 7

where
(1.15) Λρ ν = (L−1 )Tρν
∂X ν
As before we can write as,
∂X 0µ
∂X ν
= ∂µ0 (Λρ ν X 0ρ )
∂X 0µ
= Λρ ν ∂µ X 0ρ
= Λρ ν δ ρ µ
(1.16) = Λµ ν
One has to carefully note the position of the indices of Λ.
1.2.2. Tensors of higher rank. We can also have quantities like Aµ B ν .
Their transformation rule will be given by,
A0µ B 0µ = Λ µ ρ Aρ Λ ν σ B σ
= Λµ ρ Λν σ Aρ B σ
These quantities are second rank tensors. We can have three different kinds of
second rank tensors, contravariant, covariant and mixed. They will transform as
follows,
T 0µν = Λµ ρ Λν σ T ρσ
0
Tµν = Λµ ρ Λν σ T ρσ
T 0µ ν = Λµ ρ Λν σ T ρ σ
 CHAPTER 2

Discovering Electromagnetism

2.1. Lecture 3: Obtaining Lorentz Force Law

In this section we shall see, how electromagnetism (well, not the whole of elec-
tromagnetism, but atleast the Lorentz force law) follows almost naturally from
special relativity.

2.1.1. Least Action Principle. To start with we shall assume that the only
object we have at hand is a point particle. We shall rely on the Principle of Least
Action to investigate the motion of the particle. Suppose the particle follows a
certain action minimising (extremising to be precise) path in one particular frame.
Under a Lorentz transformation, the physical path followed by the particle shouldn’t
change. Since the action extremising path is a scalar, the simplest1 choice would
be to consider the action a scalar as well.
Given only a point particle and nothing else, the simplest scalar quantity that we
can form is ds2 = dxµ dxµ . Hence the Action of the particle should be of the form.
Z b
(2.1) S = −mc ds
a
The action integral can be represented as an integral of the Lagrangian with respect
to time.
Z t2
(2.2) S= Ldt
t1

The constant mc has been inserted in order to make the action dimensionally equal
to angular momentum. Writing the space-time interval ds, as
1
ds = (c2 dt2 − dx2 − dy 2 − dz 2 ) 2
1
v2 2

= c 1− 2
c
the action integral takes the form
Z t2 1
v2 2

2
(2.3) S= − mc 1 − 2 dt
t1 c
Comparing with
Z t2
S= Ldt
t1

1Occam’s Razor

9
10 2. DISCOVERING ELECTROMAGNETISM

we can write,
1
v2 2

2
(2.4) L = −mc 1 − 2
c

The quantity

∂L m~v
(2.5) p~ = = 1
∂~v 2

1 − vc2 2

gives the momentum of the particle. The equations of motion for the particle can
be obtained from the Euler-Lagrange equations
 
d ∂L ∂L
=
dt ∂~v ∂~x
d~p
(2.6) ⇒ =0
dt
and the total energy of the particle is given by the quantity

H= p~ · ~v − L
2
1/2
v2

mv 2
= 2
1/2
+ mc 1 − 2
1 − vc2 c
mc2
(2.7) = 2
1/2
1 − vc2

2.1.2. Four Potential. That’s all we can do using just a point particle and
Lorentz invariance. To proceed further towards obtaining Lorentz force law, we
now need to bring in another four vector Aµ . We will not treat Aµ as a dynamical
variable, instead we will consider it to be fixed from outside without any time-
evolution.
Once we have Aµ , let’s see what changes we can make to the action integral. The
term should be a scalar, involving both Aµ and Xµ . Moreover we also need to
have a differential, since we would be doing an integration. Hence the most obvious
choice2 for the additional term would be
Z b
−q Aµ dxµ
a

Here a scalar q is a parameter which determines the interaction of the particle with
the field. Thus the new action integral for the particle is
Z b
(2.8) S= −mcds − qAµ dxµ
a

2The form of the additional term cannot be fully justified from general considerations alone,
it is to some extent a consequence of experimental data.
 2.1. LECTURE 3: OBTAINING LORENTZ FORCE LAW 11

Separating the time and space part of A,

 
φ ~
A= ,A
c
φ ~ · d~r
Aµ dxµ = cdt − A
c 
= φ−A ~ · ~v dt

(2.9)

Hence,
t2 1 !
v2 2
Z 
(2.10) S= 2
−mc 1 − 2 ~
+ q A · ~v − qφ dt
t1 c

The new Lagrangian in this case is

1
v2 2

(2.11) 2
−mc 1 − 2 ~ · ~v − qφ
+ qA
c

2.1.3. Euler-Lagrange equation leading to the Lorentz force law. Let

us write down the Euler Lagrange equation for the system using the new La-
grangian.
 
d ∂L ∂L
=
dt ∂~v ∂~r
(2.12)

The generalised momentum in this case is,

∂L
~π =
∂~v
m~v ~
=
1 + qA
v2 2
1 − c2
 
(2.13) = ~
p~ + q A

Hence,

d~
p ~
dA  
+q ~ + q∇
= −q ∇φ ~ A ~ · ~v
dt dt
Writing the individual components3,
dpi ∂Ai
= −q∂i φ + q∂i (Aj ) vj − q(∂j Ai )vj − q
dt   ∂t
∂Ai
(2.14) = q −∂i φ − + q (∂i Aj − ∂j Ai ) vj
∂t

3A refers to the components of the three vector A

~
i
12 2. DISCOVERING ELECTROMAGNETISM

The term (∂i Aj − ∂j Ai ) vj can be written as,

(∂i Aj − ∂j Ai ) vj = (δik δjl − δjk δil ) vj ∂k Al

= mij mkl vj ∂k Al
= ijm vj (mkl ∂k Al )
h  i
= ~v × ∇ ~ ×A ~
i

Therefore,

 
dpi ∂Ai h 
~ ×A
~
i
= q −∂i φ − + q ~v × ∇
dt ∂t i

Writing,

 
∂Ai
−∂i φ − = Bi
∂t
 
∇~ ×A ~ = ~
E

we get,

d~
p 
~ + ~v × B
~

(2.15) =q E
dt

Equation 2.15 is nothing but the Lorentz force law, though it is not in a manifestly
Lorentz covariant form.
 2.2. LECTURE 4: MANIFESTLY COVARIANT FORM OF THE LORENTZ FORCE LAW 13

2.2. Lecture 4: Manifestly Covariant form of the Lorentz Force Law

In the last section we saw how the Lorentz force law arises from Special Rela-
tivity and the Least Action Principle. However the final form was not manifestly
Lorentz covariant. In this section we shall redo the derivation in a slightly different
manner in order to make it so.
Let’s start with the same Lagrangian as before.
Z b
(2.16) S= −mcds − qAµ dxµ
a

At the extremum, the first order variation in S would vanish.

δS = 0

Z b
(2.17) δs = [−mcδ(ds) − qδAµ dxµ − qAµ δ(dxµ )]
a

The term δ(ds) can be written as,

uµ
δ(ds) = d(δxµ )
c
where
dxµ
uµ ≡ , τ being the proper time.
dτ
Hence,
Z b
δS = −d[(muµ + qAµ )]δxµ ]
a
b b
dxν µ
Z Z
d
+ (muµ + qAµ )δxµ dλ − q∂ν Aµ δx dλ
a dλ a dλ
Z b  ν

d dx
= (muµ + qAµ ) − q∂ν Aµ δxµ dλ = 0
a dλ dλ
dpµ dAµ dxν
(2.18) ⇒ +q = q∂µ Aν
dλ dλ dλ
dλ
Since proper time is monotonically increasing, doesn’t diverge and we can mul-
dτ
dλ
tiply by throughout.
dτ
dpµ dAµ
+q = a∂µ Aν uν
dτ dτ
dpµ
+ q∂ν Aµ uν = a∂µ Aν uν
dτ
dpµ
(2.19) = q(∂µ Aν − ∂ν Aµ )uν
dτ
We call the part in the brackets (λµ Aν −λν Aµ ) as Fµν , an antisymmetric tensor
which leads to the manifestly covariant form.
14 2. DISCOVERING ELECTROMAGNETISM

0
Fµν = Λµ ρ Λν σ Fρσ
F0i = ∂0 Ai − ∂i A0
1 ∂ ∂ φ
= (−ai ) − i c
c ∂t
 ∂x 
1 ∂φ ∂ai
= − i−
c ∂x ∂t
Ei
=
c
Fij = ∂i Aj − ∂j Ai
= −(∂i aj − ∂j ai )
F12 = −B3
F13 = B2
F23 = −B1

Thus

 
0 E1 /c E2 /c E3 /c
 −E1 /c 0 −B3 B2 
(2.20) Fµν =
 −E2 /c B3

0 −B1 
−E3 /c −B2 B1 0

and

0
(2.21) Fµν = Λµ ρ Λν σ Fρσ

Also, if we contract Fµν with the corresponding dual we get a scalar.

(2.22) Fµν F µν = 2(B 2 − E 2 /c2 )

Since Fµν is an antisymmetric tensor, F0i (the electric field vector) and ijk Ajk
(the magnetic field) transform as vectors.
 2.3. LECTURE 5: LEAST ACTION PRINCIPLE 15

2.3. Lecture 5: Least Action Principle

2.3.1. A note on the Least Action Principle: In any problem involving a
Lagrangian, an action (S) is first written. The least action principle states that for
the actual path, this action is stationary. This means that the 1st order variation
in S is zero i.e., δS = 0. In the last lecture we encountered the following equations
while deriving the equation of motion for the action S
Z
(2.23) S = [−mcds − qAµ dxµ ]
Z λ=1
d dxν
(2.24) δS = δxµ dλ[(muµ + qAµ ) − q∂µ Aν ]=0
λ=0 dλ dλ
Note that the limit of integration i.e., the limits of the parameter λ can always
be chosen to be from 0 to 1. If we were to choose the parameter to be τ (the proper
time), we would run into the hassle of putting different limits for different paths.
So we’ll stick to a general parameter λ and assume that such a parameter always
exists. In order to conclude that the integrand vanishes let us first assume that
the integrand is non-zero at atleast one point. Since the integrand is continuous, it
must be non-zero around a neighbourhood of that point also (call this region R).
We are at freedom to choose δxµ . We can choose it to be such, that its support is
a subset of R and that it is positive in its support (support of a function f(x) is the
set of values of x such that f(x)6=0). But then this would make δS > 0. Thus we
must have integrand=0 ∀xµ (see figure 1).

Figure 1. Argument showing integrand appearing in the equation

δS = 0 vanishes ∀xµ

2.3.2. Field Strength Tensor. The manifestly Lorentz covariant form of the
solution to equation 1.2 is

dpµ
(2.25) = qFµν uν
dτ
16 2. DISCOVERING ELECTROMAGNETISM

where Fµν = ∂µ Aν − ∂ν Aµ . Clearly this is an anti-symmetric covariant tensor

of rank 2 and transforms as in 2.4.

 
0 E1 /c E2 /c E3 /c
 −E1 /c 0 −B3 B2 
(2.26) Fµν = 
 −E2 /c B3

0 −B1 
−E3 /c −B2 B1 0
 
0 −E1 /c −E2 /c −E3 /c
 E1 /c 0 −B3 B2
F µν

(2.27) =  
 E2 /c B3 0 −B1 
E3 /c −B2 B1 0

0
(2.28) Fµν = Λµ ρ Λν σ Fρσ = [(LT )−1 F L−1 ]µν

(2.29) F µν Fµν = 2(B 2 − E 2 /c2 )

2.3.3. A note on the Levi-Civita Tensor. The Levi-Civita tensor in arbi-

trary dimensions in constructed with the following two basic properties viz.,
(1) It is completely anti-symmetric
(2) 0123... = 1
It appears to be a numerical tensor and hence it should not change under any trans-
formation of the co-ordinates. Lets look at how a rank-4 Levi-Civita transforms
under Lorentz transformation

(2.30) Λκ π Λλ θ Λµ ρ Λν σ πθρσ = det(Λ)κλµν

There is an extra factor of det(Λ). In three dimensions this problem doesn’t

0
appear and  ijk = ijk . We conclude that κλµν must transform differently to
maintain its numerical tensor identity. We call it a tensor density. Also this is not
a problem if we are concerned with only proper orthochronous Lorentz transforma-
tions. 3.2 shows how κλµν actually transforms.

0 1
(2.31)  κλµν = Λκ π Λλ θ Λµ ρ Λν σ πθρσ = κλµν
det(Λ)

Now consider the tensor F̃ µν = 12 µνλσ Fλσ . This is a covariant tensor density
of rank two since it is the result of the tensor product of a rank-4 contravariant
tensor density and a rank-2 covariant tensor.

 
0 B1 −B2 B3
 −B1 0 E3 /c −E2 /c 
(2.32) F̃ µν = 
 B2 −E3 /c 0 −E1 /c 
−B3 E2 /c E1 /c 0
 2.4. LIST OF EXERCISES 17

2.4. List of Exercises

1.

 
γ −γβ 0 0
 −γβ γ 0 0 
L=
 0

0 1 0 
0 0 0 1

~ and B
How does F µν transform under this? Use this to find out how E ~ transform?
18 2. DISCOVERING ELECTROMAGNETISM

2.5. Lecture 6: Electric field of an Uniformly Moving Point Charge

2.5.1. Some digression - Electric Field of an Uniformly Moving Point
0
Charge. Let S be a frame in which B = 0. S be another frame which is moving
relative to S along the negative x-axis with a uniform speed βc. We look at an ideal
parallel plate capacitor whose plates are parallel to the xy-plane and is stationary
in the frame S. In S, there is only a z-component of electric field (Ez = σ/0 ). In
0 0
S , this Ez → Ez = γEz . This can be derived from the Fµν tensor and can also be
0
understood intuitively. In S , the lengths along the x-axis appear contracted. So
the charge density goes up by a factor of γ in this frame. Thus the field Ez goes
up by the factor γ. The component of E ~ parallel to the motion of S0 (relative to S)
remains unchanged.
0
(2.33) Ek = Ek
0
(2.34) E⊥ = γE⊥
Now consider a point charge Q placed at the origin of S. In S, the field is

(2.35) E~ = Q ~r
4π0 r3
0
In S (moving with a speed βc along the negative x-axis), the co-ordinates are
0 0
(2.36) x = γ(x − βct )
0
(2.37) y=y
0
(2.38) z=z
and the field looks like
0
(2.39) Ex = Ex
0
(2.40) Ey = γEy
0
(2.41) Ez = γEz
0
0 γQ ~ 0)
(~r − βct
(2.42) ~
E =
4π0 [γ 2 (x0 − βct0 )2 + y 0 2 + z 0 2 ]3/2
0 0
At t =0, the magnitude of the field as seen from the frame S can be written as
0
~ |0 = Q (1 − β 2 )
(2.43) |E t =0 02
4π0 r [1 − β 2 sin2 θ0 ]3/2
0 0
where θ is the angle between β~ and (~r − βct ~ 0 ). Thus there is a distribution of the
0 0
strength of electric field when seen from the frame S . For a given r , it is maximum
0
at the points on the lines perpendicular to the line of motion of Q in S .
The electric field of a charged particle initially at rest and suddenly accelerating
to a speed v within a time ∆t is also very interesting. There are three distinct
regions in the space around the particle when we look at the behaviour of E at
later times. At time t the particle is at the position vt. Region I is a sphere of
radius ct around the origin. In this region, the field is that of a moving charged
particle. The region II is of width c∆t. It consists of tangential electric field lines
 2.5. LECTURE 6: ELECTRIC FIELD OF AN UNIFORMLY MOVING POINT CHARGE 19

(a) (b)

Figure 2. (a) The field of a charge initially at rest and suddenly

accelerating at t=0 and thereafter moving with a constant speed
v. (b) The angular dependence of the field strength of a moving
charged particle. The distance of the curve from the origin is pro-
portional to the electric field strength at that angle(θ) for a given
0
r . The red, blue and green curves are for β = 0.8, 0.85 and 0.9
respectively. The curves slowly flatten to the circle as β → 0 and
similarly grows in strength near the poles for increasing β → 1.

which join the field lines of region I with region III. Region III has field lines of
that of a charge stationed at the origin since the information that the charge has
started moving has not arrived there yet.
20 2. DISCOVERING ELECTROMAGNETISM

2.6. Lecture 7: Maxwell’s Equations from the Field Tensor

2.6.1. The Maxwell’s Equation.
2.6.1.1. Gauge Invariance.
 
φ
(2.44) Aµ = , ~a
c

(2.45) Fµν = ∂µ Aν − ∂ν Aµ
Under a gauge transformation,
(2.46) Aµ → A0µ + ∂µ χ

0
(2.47) Fµν → Fµν = Fµν
provided,
(2.48) ∂µ ∂ν χ = ∂ν ∂µ χ

2.6.2. Obtaining the Maxwell’s Equations. The action integral as ob-

tained earlier is,
Z
−mcds − qAµ dxµ − αFµν F µν d4 x


(2.49) S=

For a localised charge δq, δqdxµ ,

δqdxµ = ρdvdxµ
µ
= ρdvdt dx
dt
µ
= ρd4 x dx
dt
µ
= ρ dx 4
dt d x
µ
The term ρ dx µ
dt is a four vector and can be represented by j .
µ
jµ = ρ dx
dt
= (ρc, ρ~v )
So in the action integral, the electromagnetic part can be re-written as,
Z Z
(2.50) Sem = − j Aµ d x − α Fµν F µν d4 x
µ 4

The variation in the electromagnetic part becomes,

Z Z
(2.51) δS = − jδAµ d4 x − α δ(Fµν F µν )d4 x

The variation in Fµν F µν ,

(2.52) δ(Fµν F µν ) = 2(δFµν )F µν

(δFµν )F µν + Fµν δ(F µν )

= (δFµν )F µν + F µν δ(Fµν )
= 2(δFµν )F µν
 2.6. LECTURE 7: MAXWELL’S EQUATIONS FROM THE FIELD TENSOR 21

Now,
δFµν = δ(∂µ Aν − ∂ν Aµ )
= ∂µ (δAν ) − ∂ν (δAµ )

Hence,
∂µ (δAν )F µν − ∂ν (δAµ )F µν
= ∂µ (δAν )F µν + ∂ν (δAµ )F νµ
= 2∂µ (δAν )F µν
In deriving the above expression, we first utilised the antisymmetry of F µν and
then simply flipped the dummy indices µ & ν in the second term.
So finally we have,
2(δFµν F µν ) = 2(2∂µ (δAν )F µν )
= 4∂µ (δAν )F µν
= 4∂m u(δAν F µν )
−4δAν ∂µ F µν
Making the variation vanish,
Z Z Z
− j µ δAµ d4 x − 4α ∂µ (δAν F µν )d4 x + 4α ∂ν f µν δAµ d4 x = 0

The second integral vanishes, as it is a volume integral of a 4-divergence. Hence,

Z
(2.53) (j µ + 4α∂ν F µν ) δAµ d4 x = 0

Since the integral must vanish for all arbitrary variations, the quantity (j µ + 4α∂ν F µν )
must vanish identically.
1
(2.54) ∂ν F µν = − j µ
4α
The above equation contains two of the Maxwell’s equation.
Equating the zeroth component of j,
1 c
∂ν F 0ν = − j 0 = − ρ
4α 4α
Ei
F 0i = −
c

Hence,
~ ~ c
− ∇·cE = −c ρ
4α
~ ·E
~ = c2 ρ
⇒ ∇ ρ=
4α 0
which is the Gauss Law.
Equating the other three components of j,
1 i
(2.55) ∂ν F iν = − j
4α
22 2. DISCOVERING ELECTROMAGNETISM

leads to the Ampere’s law,

~
(2.56) ∇~ ×B~ = µ0~j + µ0 0 ∂ E
∂t
The other two Maxwell’s equation,
(2.57) ~ ·B
∇ ~ =0

(2.58) ~ ×E
∇ ~ + ∂B = 0
∂t
follows from the definition of Fµν
(2.59) Fµν = ∂µ Aν − ∂ν Aµ