Classic Electrodynamics

Lecture Notes 2023/2024, Técnico Lisboa

Prof. Rogério Jorge, Zakary Buras-Stubbs & Tiago Jorge

Contents
Section 1: Introduction - September 12 2023, Part 1 2
Subsection 1.1: Main Set of Equations . . . . . . . . . . . . . . . . . . . . . . . . 2
Subsection 1.2: Validity of Classical Electrodynamics . . . . . . . . . . . . . . . . 2
Subsection 1.3: Course Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

Section 2: Electrostatics - September 12 2023, Part 2 5

Subsection 2.1: Coulomb’s Law . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
Subsection 2.2: Dirac δ−function . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
Subsection 2.3: Fields from Distributions . . . . . . . . . . . . . . . . . . . . . . . 7
Subsection 2.4: Electrostatic Potential . . . . . . . . . . . . . . . . . . . . . . . . 7
Subsection 2.5: Extra: multipolar expansions . . . . . . . . . . . . . . . . . . . . 8
Subsection 2.6: Macroscopic Media, Conductors and dielectrics . . . . . . . . . . . 9
Subsection 2.7: Electrostatic Energy and Force . . . . . . . . . . . . . . . . . . . 10

Section 3: Magnetostatics - September 14 2023, Part 1 11

Subsection 3.1: The Biot-Savart Law . . . . . . . . . . . . . . . . . . . . . . . . . 11
Subsection 3.2: Direct Integration . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
Subsection 3.3: The Vector Potential . . . . . . . . . . . . . . . . . . . . . . . . . 12
Subsection 3.4: The Scalar Potential . . . . . . . . . . . . . . . . . . . . . . . . . 13
Subsection 3.5: Energy and Self-Inductance . . . . . . . . . . . . . . . . . . . . . 13

Section 4: Formalism of Electromagnetism - September 14 2023, Part 2 14

Subsection 4.1: Faraday’s Law of Induction . . . . . . . . . . . . . . . . . . . . . 14
Subsection 4.2: Generalization to Time-Varying Fields . . . . . . . . . . . . . . . 14
Subsection 4.3: Potentials and Gauge Transformations . . . . . . . . . . . . . . . 15
Subsection 4.4: Conservation of Electromagnetic Energy: Poynting’s Theorem . . 16
Subsection 4.5: Electromagnetic Momentum Conservation . . . . . . . . . . . . . 17

1
Introduction

Section 1: Introduction - September 12 2023, Part 1

Welcome, future physicists, to the intriguing world of Classical Electrodynamics. This
is a field that has fascinated scientists for centuries and continues to do so even today.
We’re going to understand the physics behind the blue light of the sky during the day and
why it turns red at the end of the day, the fundamental reason why electric fields seem
to affect magnetic fields and vice versa, how waves propagate in different mediums and
how electromagnetic radiation is created. Here, we will also encounter the foundations of
semiconductor and plasma physics.
We will also see how electric and magnetic fields are part of a single entity that reveals
itself in different forms in different frames of reference using special relativity. This will
require some mathematical formalism but in the end it allows us to understand the world
around us in a completely different way.
................................................................................................................................
Subsection 1.1: Main Set of Equations
The main equations that we will deal with are the Maxwell equations
∂D ∂B
∇ · D = ρ, ∇ × H − = J, ∇ × E = − , ∇ · B = 0, (1.1)
∂t ∂t
where for, external sources in vacuum, D = ϵ0 E and B = µ0 H. Combining the time
derivative of the first equation with the divergence of the second leads to the continuity
equation for charge and current densities
∂ρ
+ ∇ · J = 0. (1.2)
∂t
Furthermore, the force acted by electric and magnetic fields on a charged particle q is given
by the Lorentz equation
F = q(E + v × B). (1.3)
In vacuum, where ρ = J = 0, the electric and magnetic fields obey the wave equation
∂E ∂B
= c2 ∇2 E, = c2 ∇2 B, (1.4)
∂t ∂t
√
where c = µ0 ϵ0 ≃ 3 × 108 m/s is the speed of light in vacuum.
................................................................................................................................
Subsection 1.2: Validity of Classical Electrodynamics
While many of the advances of our technology rely on the validity of classical electrody-
namics, as the name implies, there is another type of electrodynamics that is not classical
- there is quantum electrodynamics. In quantum electrodynamics, instead of waves with
defined position and velocity, we have photons, field fluctuations and uncertainty relations.
Going from one to other involves quantization of the radiation field, creation and anihilation
operators, emission and absorption of photons and relativistic covariance and antiparticles.

2
 As in classical vs quantum mechanics, the quantum theory of radiation can be ignored
if the noncommutativity of√dynamical variables, [a, a+ ] = 1, can be ignored. Since such
elements are of the order of n with n the number of photons per unit volume, the description
of physical phenomena based on classical electrodynamics is reliable when the number of
photons per unit volume is much greater than one. For example, for an FM radio station
that broadcasts at a frequency of 90 MHz, or a wavelength of λ = c/f = 3.33 m, with a
power of P = 105 W, the number of photons per second that arrive 10 kilometers away from
the antenna is the total amount of energy there divided by the energy of a single photon,
that is P/hf = 105 /(6.6 × 10−34 × 90 × 106 ) = 1.6 × 1030 s−1 . Assuming a spherical wave, we
get the photons per volume n = 1.6 × 1030 /(4πr2 c) = 1.6 × 1030 /(4π × 108 × 3 × 108 ) = 1012
m−3 which is much larger than one. That is, in common applications such as radio waves,
the classical approximation of classical electrodynamics is an extremely good one. Another
way of arriving at this result is by noting that, in quantum mechanics, the square of the
average of the electric field operator has the expectation value on a length L of hc/L4 while
in classical electrodynamics this is the energy density of the electromagnetic wave that can
be estimated as the energy density of each photon times the number of photons nhc/L with
the number of photons per unit volume. Then, the classical description should be valid if
n ≫ L−3 .
Physics today makes use of the unified description of electromagnetic, weak and strong
interactions called standard model. From the point of view of the standard model, classical
electrodynamics is a limit of quantum electrodynamics. Also, quantum electrodynamics is a
consequence of spontaneous symmetry breaking of a more general theory. Even the fields E
and B are a classical notion. They are the classical limit of a quantum mechanical description
in terms of real or virtual photons. Such quantum description is needed, for example, in the
spontaneous emission of radiation by atoms, or by any system that initially lacks photons
and has only a small number of photons present finally. While the average behaviour can
be described classically, and we will do that, an accurate description of the dynamics needs
quantum electrodynamics as opposed to classical electrodynamics.
Returning to classical electrodynamics, while the sources ρ(x, t) and J(x, t) are usually
assumed to be continous distributions in x, we know that point charges are restricted to
discrete values. The basic unit of charge is the magnitude of the charge of the electron

qe = e = 1.602 × 10−19 C. (1.5)

Finally, note the lack of symmetry in the appearance of the source terms Eq. (1.1), that is,
there are no magnetic charges and currents, or no magnetic monopoles.
As small historical note, the greeks already know some electromagnetic properties of
matter (lightning, attraction and repulsion of magnets, compass). However, only in the
XVII century did the electromagnetism as a discipline started. A short timeline could be
traced as Cavendish’s experiments around 1771, Coulomb’s experiments in 1785, Faraday’s
experiments in 1835, Maxwell’s equations in 1864 and Hertz’s experiments in 1888, that is,
EM was developed in a timeframe of 100 years. Later, in the XX century, with Einstein,
Lorentz and relativity, the framework of classical electrodynamics was developed. From
1940’s and 1950’s, the Quantum Electrodynamics (QED) revolution started, being part of
the Standard Model. In here, E and B can be seen has the exchange of photons (quantum
description) or as vectorial force fields (classical description).

3
 An important property of Maxwell’s equations in vacuum are their linearity in E and B
leading to the principle of linear superposition. In the real world, this is exploited in, for ex-
ample, hundreds of different telephone conversations being performed on a single microwave
link. In matter, some nonlinearities might occur, such as in magnetic materials and crystals
responding to intense laser beams.
There is, of course, non-linearity in the quantum world. The uncertainty principle allows
the creation of an electron-positron pair by two photons and later can be recombined again
in two photons with different directions and wave vectors. At first order in h, quantum
electrodynamics leads to the following non-linearity relation between D and E
X e4 h
Di = ϵ0 ϵik Ek , ϵij = δik + [2(E 2 − c2 B 2 )δij + 7c2 Bi Bk ] + ... (1.6)
k
45πm4 c7

These are often called vacuum polarization effects which can cause small shifts in atomic
energy levels. Such effects can be seen, for example, when two charges are put too close to
each other, leading to extremely high fields.
................................................................................................................................
Subsection 1.3: Course Overview
The books we will use in this course are Jorge Loureiro’s books "Elementos de Electrod-
inâmica Clássica" and "Física Relativista". For english speakers, Griffith’s book "Introduc-
tion to Electrodynamics" is very pedagogical and covers most of the content of the course,
but leaves out some parts of relativistic electrodynamics that we will cover. However, books
are not mandatory. Lecture notes will be provided, as well as exercises and their solutions.
There are 4 main evaluation moments, two MAP15, one on September 28 at 09h45 and
another on October 19 at 09h45, and two MAP45, one on October 6 at 15h00 and another
on October 26 at 18h00. The final grade will be determined as follows:

0.5 × (T 1 + T 2),
Final Grade = max 0.6 × (T 1 + T 2) + 0.4 × (M 1 + M 2),
0.8 × (T 1 + T 2) + 0.2 × max(M 1, M 2)


Note: The grades of M1 and M2 will only be considered if the minimum grade of (0.5 × T 1 +
0.5 × T 2) ≥ 9.5. Students who choose the continuous assessment option can use one of the
45-minute Midterm Assessment Points (T1 or T2) as a substitute for the exam to retake one
of the tests. Only the test MAP45 with best grade will be counted. Students also have the
option to take an exam, which will be on November 10th at 15h30.
The MAP15 tests (mini-tests) will be performed on the platform

• [Link]

Please register on that platform with your Tecnico ID until September 26. They MAP15
will be done at the end of the theory class of that corresponding day.
The tutors will be

• Zakary Buras-Stubbs, with office Hours Wednesday 14h00-15h30 on Astrophysics Lab

Floor 5, which will take the exercise classes Wednesday 10h-11h30 and Friday 11h-
12h30.

4
 • Tiago Jorge, with office Hours Monday 13h30-15h00 on Physics Building Floor 5, which
will take the exercise classes Tuesday 10h-11h30 and Friday 09h30-11h00.

The exercise classes will be to solve questions put on Fenix the Friday before that week.
I recommend you to form groups in the exercise sessions and solve problems cooperatively.
The exercises will cover topics covered on that same week. Important: Please read the
lecture notes and exercises before the day of the theory and exercise classes.
My email (Rogerio) is [Link] at tecnico, my office hours are on Friday from 14h00
to 15h30, and my office is at IPFN, room 2.03, on Mechanics building 1. Please let me,
Tiago or Zak know if there is any issue with Fenix, the lecture notes, the exercise sessions,
or the course website.

Electrostatics
Here we will compile the theoretical minimum to determine the electrostatic potential ϕ(r)
and the electric field E(r), where r is the position vector, from a given volume, surface and
linear charge distributions.

Section 2: Electrostatics - September 12 2023, Part 2

................................................................................................................................
Subsection 2.1: Coulomb’s Law
As nearly all electrostatics can be derived from Coulomb’s law, we shall start here.
Such law deals with the force F between charged bodies at rest with respect to each other.
However, instead of forces, we shall use the concept of an electric field E, which is the force
per unit charge acting at a given point with that charge q

F = qE, (2.1)

where the electric field due to a point charge q1 at the point r1 can be written as
q1 r − r 1
E= . (2.2)
4πϵ0 |r − r1 |3

The SI unit of charge is the coulomb (C) and the electric field is measured in volts per meter
(V/m). The value of the vacuum permitivity is 8.854 × 10−12 A2 s4 /kg m3 . For a systemP of
n charges, the electric field is found as the vector of each individual field Ei as E = ni Ei .
If the charges are so small and numerous that they can be descibed by a charge density ρ,
the sum is replaced by an integral
ˆ
1 r − r′
E(r) = ρ(r′ ) dV ′ , (2.3)
4πϵ0 |r − r′ |3

where dV ′ = dr′ = d3 r′ = dx′ dy ′ dz ′ is the 3D volume element at r′ .

................................................................................................................................
Subsection 2.2: Dirac δ−function
The distribution of point charges can be described by the volume density ρ(r) = i qi δ(r−
P
ri ) where qi is the magnitude of the ith charge, ri is the position vector for the ith charge

5
and δ(r − ri ) is the Diract delta-function. While the mathematically correct definition of
δ(r − a) requires a generalization of the usual concept of a function, in here, we will just take
advantage of the following properties of the δ-function
(
0, if r ̸= a,
δ(r − a) = δ(x − ax )δ(y − ay )δ(z − az ) = (2.4)
∞, if r = a,
ˆ (
f (a), if a is inside the volume V ,
f (r)δ(r − a)dV = (2.5)
0, if a is outside the volume V ,
ˆ
∂δ(x − a) ∂f (a)
f (x) dx = − , (2.6)
∂x a
where f is a continuous function. Using Fourier analysis, we can obtain a useful representa-
tion for the δ-function
ˆ ∞ ˆ
1 1 ∞
δ(x) = ikx
e dk = cos(kx)dx. (2.7)
2π −∞ π 0

The δ−function can also be taken as the improper limit α → 0 of the Gaussian function
2 2 2 2
e−(x +y +z )/2α
δ(x) = lim . (2.8)
α→0 (2π)3/2 α3

The following identity is used to find convert between an integral and a differential repre-
sentation of the fields
r − r′
   
1
∇ 2 ′
= −4πδ(r − r ) = −∇ · , (2.9)
|r − r′ | |r − r′ |2

where ∇ means the derivative with respect to r and not r′ .

To illustrate the origin of Eq. (2.9), we will integrate Eq. (2.9) over a volume and show
that it is zero except at one particular point. Without p loss of generality, we let r = 0.
′

First, consider the gradient of the modulus r = |r| = x + y + z of the position vector
2 2 2

r = (x, y, z)
1 r
∇ = − 3. (2.10)
|r| |r|
By taking derivatives component to component and summing them up, we find that the
divergence of r/r3 vanishes, which itself is the gradient of 1/r, leading to
1
∇2 = 0. (2.11)
|r|

However, integrating it over a sphere centered at 0 and radius a with normal vector n =
r(a)/a and differential area dA = a2 sin θdϕdθ, and using Gauss theorem, leads to
ˆ   ˆ ˆ π ˆ 2π
1 −r r 2
∇·∇ = · ndA = − r sin θdθdϕ = −4π. (2.12)
|r| r3 0 0 r3

6
This is weird, as the integrand was supposed to be zero from Eq. (2.11). This is because, in
our previous derivation, we didn’t take the point r = 0 into account. Taking it into account
leads the expression zero except at that point, that is
1
∇2 = −4πδ(r), (2.13)
|r|
which is easy generalizable for finite r′ by replacing r with r − r′ .
................................................................................................................................
Subsection 2.3: Fields from Distributions
By integrating the generalized Coulomb law for distributions, Eq. (2.3), over a closed
surface, we find Gauss’ law in integral form
˛ P
qi
E · ndS = i=1 (2.14)
n ϵ0
where S is a closed surface, q = i qi the total charge enclosed by this surface and n the unit
P
normal vector perpendicular to the surface.
¸ The differential
´ and integral forms of Gauss’ law
are related by the divergence theorem E · ndS = ∇ · EdV . This leads to the differential
form of Gauss’ law (one of Maxwell’s equations)
ρ
∇·E= . (2.15)
ϵ0
Another way to derive Eq. (2.15) is to apply the divergence directly to Eq. (2.3). As ∇
only applies derivatives of r, this leads to
ˆ ´
r − r′ δ(r − r′ )ρ(r′ )
 
1 ρ
∇·E= ′
ρ(r )∇ · ′ 3
dV = dV = . (2.16)
4πϵ0 |r − r | ϵ0 ϵ0

................................................................................................................................
Subsection 2.4: Electrostatic Potential
From the generalized Coulomb’s law to distributions Eq. (2.3), and applying the gradient
to the scalar 1/|r − r′ |
1 r − r′
∇ = − , (2.17)
|r − r′ | |r − r′ |3
we find ˆ
ρ(r′ )dV

1
E(r) = − ∇ . (2.18)
4πϵ0 |r − r′ |
Then, the electrostatic potential can be defined as
ˆ
1 ρ(r′ )dV ′
E = −∇ϕ, ϕ = . (2.19)
4πϵ0 |r − r′ |
In the case of surface or line charge distributions, the volume integral in Eq. (2.19) should
be replaced by the corresponding surface or line integral. In electrostatics (and not in time-
varying fields), due to the fact that E can be written as the gradient of a function and that
the curl of a gradient is zero, we find that
∇ × E = 0, (2.20)

7
which is another of Maxwell’s equations in static fields.
The differential equations for ϕ can be found combining ∇ · E = ρ/ϵ0 and E = −∇ϕ,
leading to Poisson’s equation
ρ
∇2 ϕ = − . (2.21)
ϵ0
In vacuum, where ρ = 0, Poisson’s equation reduces to Laplace’s equation ∇2 ϕ = 0. The
relation between Eqs. (2.19) and (2.21) can be found from the identity Eq. (2.9)
On a charged surface separating regions 1 and 2, ϕ1 = ϕ2 . The normal derivatives of ϕ
are discontinuous across a charged surface so that
∂ϕ1 ∂ϕ2 σ
E2n − E1n = − = , (2.22)
∂n ∂n ϵ0
where σ is the surface density of charge.
................................................................................................................................
Subsection 2.5: Extra: multipolar expansions
As the direct evaluation of the integral in Eq. (2.19) is, in general, difficult to do, it is
often convenient to expand Eq. (2.19) in powers of x/r or x′ /r and then integrate term by
term, the so called multipolar expansion.
In cartesian coordinates, when r > a (a is the maximum distance of the charges from the
origin), expanding the denominator of Eq. (2.19) in powers of 1/r, we obtain
 
q 1
ϕ= −p· ∇ + ... (2.23)
r r
´ ´
where q = ρdV is the total charge and p = ρxdV is the dipole moment. In spherical
coordinates, the multipole expansion yields
∞ X l 1/2 (
l+1 , if r > r ,
1 ′

X 4π
ϕ= Qlm Ylm × r l (2.24)
l=0 m=−l
2l + 1 r , if r < r′ ,
where Qlm is the multipole moment and Ylm the spherical harmonics.
For our purposes, the multipole expansion is useful to introduce the concept of an electric
dipole moment ˆ
p(r) = r′ ρ(r′ )dV ′ , (2.25)
´
so that the electrostatic potential due to the total charge q = ρdV and the dipole moment
can be written as
q p·r
ϕ= + (2.26)
4πϵ0 r 4πϵ0 r3
and the electric field due to a dipole p at the point r0 with the unit vector n directed from
r0 to r
3n(p · n) − p 1
E= 3
− pδ(r − r0 ), (2.27)
4πϵ0 |r − r0 | 3ϵ0
´
with a volume integral EdV = −p/3ϵ0 with V ´the volume over a sphere that encloses the
charge density. The electrostatic energy W = ρϕdV (derived in a later section) can be
expanded similarly around an origin E(0) = E(r0 ), yielding
W = qϕ(0) − p · E(0) + ... (2.28)

8
 ................................................................................................................................
Subsection 2.6: Macroscopic Media, Conductors and dielectrics
In electromagnetism, a dielectric is an insulating material that becomes polarized in
an applied electric field, causing internal electric fields that reduce the overall field within
the material. This property, quantified by the permittivity, makes dielectrics crucial in
many electrical devices, including capacitors, where they increase the stored charge, and in
transmission lines and antennas, where they affect the propagation of electromagnetic waves.
We shall use the fact that in most substances, the dominant multipole is the dipole,
which, when average over the medium, leads to an electric polarization P (dipole moment
per unit volume) X
P= Ni ⟨pi ⟩ , (2.29)
i

with pi the dipole moment of the ith type of molecule in the medium. Summing Eq. (2.26)
for the total amount of dipoles, replacing the sum with an integral, using the identity in
Eq. (2.17), and applying an integration by parts, we find
ˆ
1 dV ′
ϕ(r) = [ρ(r′ ) − ∇ · P(r′ )] , (2.30)
4πϵ0 |r − r′ |
leading to the replacement of the first Maxwell equation to
1
∇·E= (ρ − ∇ · P). (2.31)
ϵ0
We then define the electric displacement or electric induction vector D as

D = ϵ0 E + P, ∇ · D = ρ, (2.32)

as well as its corresponding integral formulations. Here, ρ is the density of free charges (as
opposed to bound charges) in the dielectric and q is the total free charge enclosed by the
surface S. The density of bound charges ρb is expressed in terms of the polarization vector P
as ρb = −∇ · P. A relation between D and E, also called constitutive relation, is needed to
obtain a solution for the fields and potentials. Here, we assume that the response is linear.
A further simplification is often pursued where the medium is considered isotropic, so that
P is parallel to E
P = ϵo χe E, (2.33)
with χe the electric susceptibility of the medium. For isotropic dielectrics and sufficiently
small fields (where we have a linear relation), the displacement D is proportional to E

D = ϵE, (2.34)

where ϵ = ϵ0 (1 + χe ). A relative electric permittivity is sometimes defined as

ϵ
ϵr = = 1 + χe , (2.35)
ϵ0
also called dielectric constant. If, furthermore, the dielectric is uniform, that is, ϵ is a constant
independent of position, then ∇ · E = ρ/ϵ.

9
 The boundary conditions at a boundary with surface-charge density σ (not including the
polarization charge) separating two different dielectrics are

E1∥ = E2∥ , D2⊥ − D1⊥ = σ, (2.36)

where ∥ and ⊥ means parallel and perpendicular to the boundary surface, respectively. These
conditions can also be written as
∂V1 ∂V2
V1 = V2 , ϵ1 − ϵ2 = σ. (2.37)
∂n ∂n
For bound charges, their surface density is given by σb = P1n − P2n . In terms of the vector n
normal to the surface, directed from region 1 to region 2, the boundary conditions can also
be written as
(D2 − D1 ) · n = σ, (E2 − E1 ) × n = 0. (2.38)

................................................................................................................................
Subsection 2.7: Electrostatic Energy and Force
The work W done in moving a charge from A to B can be calculated using the force on
the charge F = qE as
ˆ B ˆ B ˆ B
W =− F · dl = −q E · dl = q ∇ϕdl = q(ϕB − ϕA ), (2.39)
A A A

so that qϕ can be interpreted as the potential

¸ energy of the test charge in the electrostatic
field. The¸ fact that,´for a closed path, E · dl = −(ϕB − ϕB ) = 0 is an application of Stoke’s
theorem E · dl = S (∇ × E) · ndS = 0 as, in electrostatics, ∇ × E = 0.
The potential energy U produced by an array of charges qi at positions ri can be written
as the sum of each potential energy
1X
U= qi ϕi (r). (2.40)
2 i

The 1/2 factor comes from the fact that, when summing the contribution from every charge
on another, we don’t want to count the same charge twice. The total potential energy for
the case of a continuous distribution can be written as
ˆ
1
U= ρϕdV. (2.41)
2

Using Poisson’s equation ρ = ∇2 ϕ and integrating by parts (neglecting the boundary term)
leads to the following result for the energy of an electrostatic field
ˆ
ϵ0
U= |E|2 dV, (2.42)
2
´
which, if not in vacuum, generalizes to D · EdV /2.

10
Magnetostatics
Magnetostatics deals with constant magnetic fields. However, these are usually created from
the movement of electric charges and, therefore, the term static in magnetostatics does not
mean that nothing is moving in the system.

Section 3: Magnetostatics - September 14 2023, Part 1

When dealing with magnetic fields that are constant in time, Maxwell’s equations are
written as
∇ × H = J, ∇ · B = 0. (3.1)
where J is the current density. Now a word of caution for notation. In some textbooks, B
is called the magnetic field vector, while other textbooks reserve H for the magnetic field.
While we can argue that the fundamental fields are E and B, the macroscopic fields that
take ρ and J as sources are the fields D and H even though they are more of a convenience
to be able to solve the equations in a more straightforward manner. Here, when talking
only about fields in vacuum, we call B the magnetic field and, when talking about different
media, we call B the magnetic induction (or the magnetic flux density) and H the magnetic
field vector.
In isotropic diamagnetic and paramagnetic media, B = µH where µ is the magnetic
permeability. In a material, the average magnetization due to the density of molecular
currents when placed in a magnetic field is given by the magnetization vector M with JM =
∇ × M, leading to the relation
B = µ0 (H + M). (3.2)
In ferromagnetic media, the relation between B and H is a nonlinear one and may not even
be single-valued. The solution of magnetostatic problems in this case becomes much more
difficult. However, in discussing permanent magnets in the context of idealised ferromagnet-
ics, a linear relationship is usually assumed B = µ(H + M0 ) with M0 a constant.
To calculate the magnetic field in non ferromagnetic media, we mainly use one of four
diferent techniques.
................................................................................................................................
Subsection 3.1: The Biot-Savart Law
The magnetic field due to a current element Idl in a vacuum or a homogeneous medium
is given by
µI dl × r
dB = . (3.3)
4π |r|3
Due to the superposition principle, the total field at a point is the integral of Eq. (3.3) over
all the current elements.
As in electrostatics, the fields are an abstraction we use to calculate the forces and find the
motion of different objects. The force experienced by a current element Idl in the presence
of a magnetic induction B is given by
dF = Idl × B. (3.4)
In turms of a current density J, we find that the total force on a current distribution is
ˆ
F = J × BdV, (3.5)

11
 ´
leadin to the following torque τ = x × (J × B)dV .
................................................................................................................................
Subsection 3.2: Direct Integration
Using the current density vector J, the magnetic field can be found from the Biot-Savart
law in Eq. (3.3) integrating over the volume as
ˆ
µ0 r − r′
B= J(r′ ) × dV ′ . (3.6)
4π |r − r′ |3

Using the identity in Eq. (2.17) we find the following formula for B
ˆ
µ0 J(r′ )
B= ∇× dV ′ . (3.7)
4π |r − r′ |

Due to the fact that the divergence of a curl vanishes, we find the third of Maxwell’s equations

∇ · B = 0. (3.8)

The fourth of Maxwell’s equations can be found by taking the curl of B, applying the
identities we have for |r − r′ | involving δ functions, integrating by parts and recalling that
for static phenomena ∇ · J = 0, leading to

∇ × B = µ0 J. (3.9)

Its corresponding integral formulation (Ampère’s law) can be found using Stoke’s theorem,
leading to ˛ ˆ
B · dl = µ0 J · ndS = µ0 I. (3.10)

The integral of Maxwell’s equations for magnetostatics lead to the following boundary
conditions
n · (B2 − B1 ) = 0, n × (H2 − H1 ) = i, (3.11)
where i is the surface current density and n the normal drawn from the first region to the
second. If the current distribution is axially symmetric, then the integral form of Maxwell’s
equation reads ˛
H · dl = I. (3.12)

................................................................................................................................
Subsection 3.3: The Vector Potential
The vector potential A is defined from Eq. (3.7) as
ˆ
J(r′ )
B = ∇ × A, A(r) = µ dV ′ , ∇2 A = −µJ. (3.13)
|r − r′ |

For a linear current, one makes the substitution JdV ′ → Idl′ .

12
 At large distances from the current distribution, similarly to the multipole method for
electrostatics, the solution for A reads
ˆ
m×r
A= , m = r′ × JdV ′ . (3.14)
|r|3
For a current loop, the magnetic dipole moment is defined as

m = πIa2 n, (3.15)

where I is the current on the loop, a is its radius πa2 its area and n the vector directed along
its axis. For an arbitrary loop, the magnetic moment is given by
˛
I
m= x × dl. (3.16)
2
Since |x×dl| = 2dS where dS is the differential area, the magnetic moment has the magnitude
|m| = I× Area. The field B due to a magnetic moment m is given by
µ0 3n(n · m) − m 2µ0
B= + mδ(x), (3.17)
4π |r|3 3
where n is the unit vector in the direction of x. If the current distribution is provided by a
number of charged particles with charges qi and masses Mi in motion with velocities vi , the
current density J and magnetic moment m become
X 1X X qi
J= qi vi δ(x − xi ), m = qi (xi × vi ) = Li . (3.18)
i
2 i i
2Mi

where Li is the particle’s orbital angular momentum. The contribution of m to the force on
a localized current distribution in an external magnetic field B is

F = ∇(m · B), (3.19)

leading to the potential energy U of a dipole (or a permanent magnetic moment) in an

external field as U = −m · B.
................................................................................................................................
Subsection 3.4: The Scalar Potential
In vacuum, where J = 0, we have ∇ × H = ∇ × B = 0, hence B and H can be written as
the gradient of a scalar which, due to the restriction ∇ · B = 0, leads to Laplace’s equation
for the potential
B = ∇ψ, ∇2 ψ = 0. (3.20)
However, we note that ψ may not be a single valued function.
................................................................................................................................
Subsection 3.5: Energy and Self-Inductance
The energy of the magnetic field is given by
ˆ ˆ
1 1
W = H · BdV = J · AdV. (3.21)
2 2

13
In the first case, the volume V is where the magnetic field is localized, while the second case
the integration is carried out over the volume occupied by the currents. In the case of a
conductor carrying a current I it may be expressed in terms of the self-inductance L of the
conductor W = LI 2 /2.
The energy of interaction between two current-carrying conductors is
ˆ ˆ
W12 = H1 · B2 dV = J1 · A2 dV, (W12 = W21 ), (3.22)

where the first integral is evaluated over the whole of space and the second over the volume
of one of the conductors.

Formalism of Electromagnetism

Section 4: Formalism of Electromagnetism - September 14 2023, Part 2

In here, we introduce time-varying electromagnetic fields and the conservation laws for
electromagnetism.
................................................................................................................................
Subsection 4.1: Faraday’s Law of Induction
What Faraday observed in 1831 was that a current is induced in a circuit if:
1. the steady current flowing in an adjacent circuit is turned on or off;
2. the adjacent circuit with a steady current flowing is moved relative to the first circuit;
3. a permanent magnet is thrust into or out of the circuit.
The transient current flow, as Faraday pointed out, can be attributed to a changing magnetic
flux linked by the circuit, inducing an electric field with a non-zero closed line integral. Such
non-zero closed line integral is called electromotive force E. In mathematical terms, defining
the flux Φ of B as ˆ
Φ = B · ndS, (4.1)

and the electromotive force as ˛

E= E · dl, (4.2)

where, in a circuit, E is the electric field at the element dl, Faraday’s observation can be
written as
dΦ
E =− . (4.3)
dt
Such law can be put in a differential form using Stoke’s theorem, leading to
∂B
∇×E=− . (4.4)
∂t
Note that this is a generalization of ∇ × E for electrostatic fields.
................................................................................................................................
Subsection 4.2: Generalization to Time-Varying Fields

14
 Up to now, we have derived four equations for electromagnetism but they are still incom-
plete. This is because Ampère’s law was derived in the steady-state case of ∇·J = −∂ρ/∂t =
0. In 1865, Maxwell modified this equations to make them consistent and applicable to any
phenomena, including time-varying ones. A simple reasoning for the extra term that should
appear in Ampère’s law ∇×H = J is the fact that, from the charge and current conservation
formula, we can write  
∂ρ ∂D
∇·J+ =∇· J+ = 0, (4.5)
∂t ∂t
leading to the identification of J → J + ∂D/∂t. Ampère’s law then became

∂D
∇×H=J+ . (4.6)
∂t
The term ∂D/∂t was called by Maxwell the displacement current.
How do we write the electric field in termos of potentials now that we have time-varying
fields? Here, we take B = ∇ × A and ∇ × E = −∂B/∂t and find that ∇ × (E + ∂A/∂t) = 0
should have vanishing curl. Therefore, it can be written as the gradient of a scalar potential
ϕ, leading to
∂A
E = −∇ϕ − . (4.7)
∂t
This concludes the generalization of the previous sections to time-varying fields. We write
here the four Maxwell’s equations for completeness
∂D ∂B
∇ · D = ρ, ∇ × H − = J, ∇ × E = − , ∇ · B = 0, (4.8)
∂t ∂t

................................................................................................................................
Subsection 4.3: Potentials and Gauge Transformations
At this point, let’s restrict ourselves to the the solution of Maxwell’s equations in vacuum,
where ϵ = ϵ0 and µ = µ0 , but allowing the presence of distribution of charges ρ and J
somewhere in space. The inhomogeneous Maxwell’s equations, Eq. (4.8), in terms of the
potentials ϕ and A, can then be written as
∂ ρ
∇·A=− ,∇2 ϕ + (4.9)
∂t ϵ0
2
 
1 ∂ 1 ∂ϕ
∇2 A − 2 ∂t2 − ∇ ∇ · A + 2 = −µ0 J. (4.10)
c A c ∂t

While we have reduced the four first order Maxwell’s equations to only two second order
equations, Eqs. (4.9) and (4.10) still couple ϕ and A on both equations. The uncoupling can
be done exploiting the freedom in the definition of the potentials. The actual fields B and
E are unchanged by the following gauge transformations

B = ∇ × A, A → A + ∇Λ, (4.11)
∂A ∂Λ
E = −∇ϕ − , ϕ→ϕ− , (4.12)
∂t ∂t

15
meaning that we can choose a set of potentials such that the extra term in Eq. (4.10) involving
ϕ can be set to zero
1 ∂ϕ
∇·A+ 2 = 0, (4.13)
c ∂t
leading to the two following inhomogeneous wave equations for ϕ and for A
1 ∂ 2ϕ ρ
2
∇ ϕ− 2 2 =− , (4.14)
c ∂t ϵ0
2
1 ∂ A
∇2 A − 2 2 = −µ0 J. (4.15)
c ∂t
The set of Eqs. (4.13) to (4.15) form a set of equations equivalent to Maxwell’s equations.
The choice Eq. (4.13) is called the Lorentz gauge. Another option is to consider ∇ · A = 0,
which is called the Coulomb gauge. However, the wave equation for A will not be decoupled
from ϕ, but the wave equation for ϕ will. It is a useful gauge if no sources are present and
ϕ = 0.
The set of Eqs. (4.11) and (4.12) are called gauge transformations, and the invariance of
the fields under such transformations is called gauge invariance. The relation betwen A and
ϕ in Eq. (4.13) is called Lorentz condition. We can also see from Eqs. (4.11) and (4.12) that
the field Λ, in order to comply with Eq. (4.13), must satisfy the condition
1 ∂ 2Λ
∇2 Λ = . (4.16)
c2 ∂t2

................................................................................................................................
Subsection 4.4: Conservation of Electromagnetic Energy: Poynting’s Theo-
rem
We start with the energy conservation equation, also called Poynting’s theorem. In
electrostatics,wWe have seen that the energy in the field is given by the volume integral of
E · D/2, while in magnetostatics it is given by B · H/2. Both of these can be derived from
the fact that the total rate dW/dt of doing work W by the fields in a finite volume is given
by ˆ
dW
= J · EdV. (4.17)
dt
This is a generalization´ of the fact
´ that the force on a charged particle is F = q(E + v × B),
so that the work W = F · dl = F · vdt leads to a work rate dW/dt = qv · E. The magnetic
field does no work as it is always perpendicular to the velocity. Using the Maxwell’s equations
to eliminate J, employing the vector identity
∇ · (E × H) = H · (∇ × E) − E · (∇ × H), (4.18)
and using Faraday’s law, we find that J · E is equal to the sum of the time derivative of the
energy density u, given by
1
u = (E · D + B · H), (4.19)
2
and the divergence of the vector E × H, also called the Poynting vector
S = E × H. (4.20)

16
The energy conservation law, or Poynting’s theorem, can then be written as
∂u
+ ∇ · S = −J · E. (4.21)
∂t
The conservation law in Eq. (4.21) can be cast in integral form by integrating it over a
volume V including the fields and sources. Denoting the total energy of the particles as Emec
we have ˆ
dEmec
= J · EdV, (4.22)
dt
´
and the field energy as Ef ield = udV , and applying the divergence theorem, we find
˛
dE d
= (Emec + Ef ield ) = − S · ndS. (4.23)
dt dt

................................................................................................................................
Subsection 4.5: Electromagnetic Momentum Conservation
Generalizing the total force on a charged particle to a distribution of particles, we can
write Newton’s second law for the momentum Pmec as
ˆ
dPmec
= (ρE + J × B)dV. (4.24)
dt

As before, using Maxwell’s equations, performing algebraic manipulations, adding c2 B(∇ ·

B) = 0, and identifying the total electromagnetic momentum Pf ield in the volume V as
ˆ
S
Pf ield = dV, (4.25)
c2
we find the following momentum conservation law
ˆ
d
(Pmec + Pf ield ) = ϵ0 [E(∇ · E) − E × (∇ × E) + c2 B(∇ · B) − c2 B × (∇ × B]dV
dt
ˆ ˆ
= ∇ · TdV = T · ndS,

where T is called the Maxwell stress tensor Tij with components

E 2 + c2 B 2
 
2
Tij = ϵ0 Ei Ej + c Bi Bj − δij . (4.26)
2

Note: while we call it a tensor, there is no need to deal with tensor calculus to solve this
section’s exercises. For now, we will just take it as a matrix.

17