Solved Exercises in Electromagnetism

Notes from Exercises portion of the third-year undergraduate

course Elektromagnetno Polje (Electromagnetic Field), led by as-
sistant professor Martin Klanjšek at the Faculty of Mathematics
and Physics at the University of Ljubljana in the academic year
2020-2021. Credit for the material covered in these notes is due
to professor Klanjšek, while the voice, typesetting, and translation
to English in this document are my own. The original exercises in
Slovene (without solutions) can be found on the course website.
Disclaimer: This document will inevitably contain some
mistakes—both simple typos and legitimate errors. Keep in mind
that these are the notes of an undergraduate student in the process
of learning the material himself, so take what you read with a grain
of salt. If you find mistakes and feel like telling me, I will be grateful
and happy to hear from you, even for the most trivial of errors.
You can reach me by email, in English, Slovene, or Spanish, at
ejmastnak@gmail.com.

For more notes like this, visit ejmastnak.com/fmf

Elijan J. Mastnak
Winter semester 2020-2021
Last update: October 10, 2022
Faculty of Mathematics and Physics, University of Ljubljana
Contents
1 First Exercise Set 4
1.1 Electric Field of a Charged Disk . . . . . . . . . . . . . . . . . . . . 4
1.2 Charged Plate with a Slit . . . . . . . . . . . . . . . . . . . . . . . . 5

2 Second Exercise Set 7

2.1 Theory: The Poisson Equation and the Fourier Transform . . . . . . 7
2.2 Poisson Equation for a Point Particle . . . . . . . . . . . . . . . . . . 8
2.3 Theoretical Interlude: Electric Field of a Charge Distribution . . . . 10
2.4 Electric Field of a Hydrogen Atom . . . . . . . . . . . . . . . . . . . 10

3 Third Exercise Set 12

3.1 Conducting Ribbon in a Parallel-Plate Capacitor . . . . . . . . . . . 12
3.2 A Halved Conducting Cylinder . . . . . . . . . . . . . . . . . . . . . 15

4 Fourth Exercise Set 17

4.1 A Halved Conducting Cylinder (continued) . . . . . . . . . . . . . . 17
4.2 Conducting Sphere in a Uniform Electric Field . . . . . . . . . . . . 19

5 Fifth Exercise Set 21

5.1 Conducting Sphere in a Uniform Electric Field (continued) . . . . . . 21
5.2 Electric Dipole in a Conducting Spherical Shell . . . . . . . . . . . . 22
5.3 Point Charge Above a Conducting Plane . . . . . . . . . . . . . . . . 23

6 Sixth Exercise Set 27

6.1 Force on a Conducting Spherical Shell . . . . . . . . . . . . . . . . . 27
6.2 Point Charge Between Two Conducting Plates . . . . . . . . . . . . . 28
6.2.1 Theory: The Multipole Expansion . . . . . . . . . . . . . . . 28
6.2.2 Returning to the original problem . . . . . . . . . . . . . . . . 29

7 Seventh Exercise Set 31

7.1 Theory: Magnetic Vector Potential and the Biot-Savart Law . . . . . 31
7.2 Magnetic Field of a Circular Current Loop . . . . . . . . . . . . . . . 32
7.3 Magnetic Field of a Rotating Charged Disk . . . . . . . . . . . . . . 33

8 Eighth Exercise Set 36

8.1 Magnetic Force in a Coaxial Cable . . . . . . . . . . . . . . . . . . . 36
8.2 Tension in a Toroidal Inductor . . . . . . . . . . . . . . . . . . . . . 37
8.3 Resistance of a Thin Conducting Plate . . . . . . . . . . . . . . . . . 38

9 Ninth Exercise Set 41

9.1 Mutual Inductance . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
9.2 The Cabrera Experiment and Magnetic Monopoles . . . . . . . . . . 43

10 Tenth Exercise Set 46

10.1 Skin Effect in a Ribbon-Like Conductor . . . . . . . . . . . . . . . . 46
10.2 Theory: Conservation of Electromagnetic Energy . . . . . . . . . . . 49
10.3 Power in a Coaxial and Cylindrical Conductor . . . . . . . . . . . . . 49
 10.4 Cylindrical Conductor with a Slit . . . . . . . . . . . . . . . . . . . . 51

11 Eleventh Exercise Set 54

11.1 A Radially Polarized Sphere . . . . . . . . . . . . . . . . . . . . . . . 54
11.2 A Halved Polarized Sphere . . . . . . . . . . . . . . . . . . . . . . . . 55
11.3 Theory: Dielectric and Displacement Field . . . . . . . . . . . . . . . 57
11.4 Parallel-Plate Capacitor with an Anisotropic Dielectric . . . . . . . . 58

12 Twelfth Exercise Set 61

12.1 Point Dipole in a Spherical Dielectric Cavity . . . . . . . . . . . . . . 61
12.2 Dielectric Constant of Cold Plasma . . . . . . . . . . . . . . . . . . . 63

13 Thirteenth Exercise Set 67

13.1 Theory: Electromagnetic Wave Propagation in Waveguides . . . . . . 67
13.2 A Parallel-Plate Waveguide . . . . . . . . . . . . . . . . . . . . . . . 69

14 Fourteenth Exercise Set 76

14.1 Cylindrical Waveguide . . . . . . . . . . . . . . . . . . . . . . . . . . 76
14.2 Overview: Waveguide with a Quarter-Circle Cross Section . . . . . . 79
14.3 Theory: TEM Waves in Waveguides . . . . . . . . . . . . . . . . . . 80
14.4 TEM Waves in a Coaxial Waveguide . . . . . . . . . . . . . . . . . . 82
1.1. Electric Field of a Charged Disk

1 First Exercise Set

Background Theory: Electric Field of a Charge Distribution

The electric field of a spatial charge distribution with volume charge density ρ(r) is
˚
1 ρ(r̃) d3 r̃ r − r̃
E(r) = 2 ,
4πϵ0 V |r − r̃| |r − r̃|

where r̃ is a placeholder variable for integration and ρ(r̃) d3 r̃ is an infinitesimal element

of charge at the position r̃.

1.1 Electric Field of a Charged Disk

A charged disk has surface charge density σ and radius a. Find the disk’s electric
field E(z) along an axis through the disk’s center and normal to the disk. Analyze the
limit behavior of E(z) for small and large z.
We begin by breaking the disk into concentric rings and integrating over the con-
tributions dE of each ring, where r and dr represent the radius and thickness of a
ring, respectively. We then further divide each ring into infinitesimal segments with
area r dϕ dr. Along the perpendicular z axis through the disk’s center, a single such
segment contributes the electric field
1 σr dϕ dr
dE1 = .
4πϵ0 z 2 + r2
Notice the term r dϕ dr must have units of area to produce charge when multiplied
by the surface charge density σ. The term z 2 + r2 is simply the squared distance
from the charge element to the z axis.
Next, we recognize the circular symmetry of each ring: both the x and y components
of the electric field symmetrically cancel along the z axis, so the electric field only has
a z component. We then relate the magnitudes of dE and dE1 along the z axis with
similar triangles to get the contribution dE of a ring of radius r at the point (0, 0, z):

dE z z σr dϕ dr
=√ =⇒ dE = 2 2 )3/2 4πϵ
.
dE1 2
z +r 2 (z + r 0

We find the total electric field E(z) by integrating over the contributions dE:
ˆ ˆ 2π ˆ a ˆ a
σzr σz r dr
E(z) = dE = dϕ dr = .
0 0 4πϵ0 (z 2 + r2 )3/2 2ϵ0 0 (z 2 + r2 )3/2

We solve the integral with the substitution u = z 2 + r2 , du = 2r dr:

ˆ z 2 +a2  
du σz 1 1
E(z) = = − √ − √
z2 u3/2 2ϵ0 z 2 + a2 z2
  !
σ z σ 1
= 1− √ = 1− 2 .
2ϵ0 z 2 + a2 2ϵ0 1 + az 2

4
1.2. Charged Plate with a Slit

a2
For z ≪ a (very close to the disk), we have 1 + z2
→ ∞ and 1
2 → 0, leaving
1+ a2
z
σ
E(z) → (z ≪ a),
2ϵ0
which is the electric field of an infinite charged plane.
2
For z ≫ a (very far from the disk), we have az 2 ≪ 1 and use the Taylor approximation
(1 + x)p ≈ 1 + px for x ≪ 1 to get
" −1/2 #
a2 a2 σa2
   
σ σ
E(z) = 1− 1+ 2 ≈ 1− 1− 2 = .
2ϵ0 z 2ϵ0 2z 4ϵ0 z 2
Finally, we multiply above and below by π to match the above result to the expression
for a point charge:
πa2 σ σS q
E(z) = 2
= 2
= (z ≫ a),
4πϵ0 z 4πϵ0 z 4πϵ0 z 2
where q = σS is the disk’s charge.

1.2 Charged Plate with a Slit

We take a large, rectangular charged plate
with surface charged density σ and remove a
slit of width a from the plate. Determine the
electric field E in the plane perpendicular to
the plate and passing through the center of
the slit as a function of the vertical distance
z from the plate. Analyze the limit behavior
of E(z) for small and large z.
We begin by breaking the plate into thin rib-
bons and integrating over the contributions
dE of each ribbon, where r and dr will represent the orthogonal distance from the
slit and the thickness of each ribbon, respectively.
We find the electric field of a ribbon a distance r from the slit along the z axis using
Gauss’s law with a cylindrical surface. For a cylinder of radius R and length l, Gauss’s
law reads ‹
qenc qenc
E · dS = 2πRlE = =⇒ E(R) = .
S ϵ0 2πϵ0 lR
Applied to the ribbon, the enclosed charge qenc is the ribbon’s infinitesimal charge
dq1 = σ dS1 = σl√dr, while the cylinder’s radius R is the distance from the ribbon to
the z axis: R = z 2 + r2 , so the contribution dE1 of one ribbon is
dq σl dr σ dr
dE1 = √1 = √ = √ .
2
2πϵ0 l z + r 2 2
2πϵ0 l z + r 2 2πϵ0 z 2 + r2

Because of mirror-image symmetry, both the x and y components of the electric field
cancel, leaving only the z component dEz . We then relate dEz and dE1 using similar
triangles
dEz z σz dr
=√ =⇒ dEz = .
dE1 2
z +r 2 2πϵ0 (z 2 + r2 )

5
1.2. Charged Plate with a Slit

We find the total electric field along the z axis by integrating over the contributions
dEz of all the ribbons. Because of mirror symmetry, we need to calculate only the
contribution of e.g. the right plane and multiple the result by two.
ˆ ˆ ∞ ˆ
σz ∞ dr

σz dr
E(z) = dEz = 2 2 2
=
a/2 2πϵ0 (z + r ) πϵ0 a/2 z 2 + r2
 ∞
σz 1 r σ hπ  a i
= arctan = − arctan .
πϵ0 z z a/2 πϵ0 2 2z

In the limit z ≫ a (very far from the slit), we have arctan 2z

a
→ 0 and the electric
field along the z axis simplifies to
σ
E(z) = (z ≫ a),
2ϵ0
which is the field of an infinite sheet of charge.
In the limit z ≪ a (very close to the slit), we have 2z
a
→ ∞. We use the asymptotic
expansion arctan x ≈ 2 − x for large x to get
π 1

  
σ π π 2z 2σ z
E(z) ≈ − − = .
πϵ0 2 2 a πϵ0 a

In this case the electric field scales linearly as E ∼ z.

6
2.1. Theory: The Poisson Equation and the Fourier Transform

2 Second Exercise Set

2.1 Theory: The Poisson Equation and the Fourier Transform
We begin with Gauss’s law in differential form and the relationship between electric
field E and electric potential U
ρ
∇·E= and E = −∇U
ϵ0
We then substitute E = −∇U into Gauss’s law to get
ρ ρ
=⇒ ∇2 U = − .
 
∇ · − ∇U =
ϵ0 ϵ0
The last equality has the form of a Poisson equation,1 and relates charge density ρ
to electric potential U . In other words, if we know a spatial charge distribution ρ,
we can find the corresponding electric potential U and thus the electric field E with
E = −∇U .
As a simple example we start with a point charge. The charge distribution is
ρ q
ρ(r) = qδ(r) =⇒ ∇2 U = − = − δ(r).
ϵ0 ϵ0
The above equation is the Poisson equation for a point charge. We will solve the
equation with a Fourier transform.
First, we briefly review the Fourier transform, which we can think of as an expansion
over a basis of plane waves of the form eik·r . The expression U (k) plays the role of a
weight function and determines how much each wave contributes to the expansion
˚
U (r) = U (k)eik·r d3 k,

where U (k) is the amplitude of the plane wave with wave vector k. To find U (k) we
take the inner product of both sides of the above equation with the basis function
e−ik̃·r , which gives
˚ ˚ ˚
U (r)e−ik̃·r d3 r = d3 k d3 rU (k)ei(k−k̃)·r .

The integral over r on the right-hand side is in fact a delta function, because the
orthogonal plane waves cancel out over all space except at the origin, where they
constructively interfere to infinity. Recognizing the delta function simplifies the
equation to
˚ ˚
−ik̃·r 3 3
U (r)e d r = (2π) U (k)δ(k − k̃) d3 k = (2π)3 U (k̃).

The delta function suppresses the integral everywhere except at (k − k̃), which leads
to the expression for the amplitude U (k̃) of a wave vector k̃:
˚
1
U (k̃) = U (r)e−ik·r d3 r.
(2π)3
1
In general any equation of the form ∇2 f (r) = g(r) is called a Poisson equation.

7
2.2. Poisson Equation for a Point Particle

This expression can also be interpreted as an inverse Fourier transform, used to

recover U (k) from U (r).
As an intermediate step, we analyze the behavior of the gradient operator under the
Fourier transform. Since the gradient acts only on r, applying the gradient to the
earlier expression for U (r) gives
˚
∇U (r) = U (k)∇eik·r d3 k.

Evaluating the gradient over (x, y, z) components gives

 
ikx
∇eik·r = iky  ei(k1 x+k2 y+k3 z) = ikeik·r .
ikz

The expression for ∇U (r) is thus

˚ ˚
∇U (r) = U (k)∇eik·r d3 k = U (k)ikeik·r d3 k,

which is a Fourier transform of the function U (k)ik. In other words, the gradient op-
erator ∇ transforms to multiplication by ik under the Fourier transform. Analogously,
the Laplacian ∇2 transform into multiplication by (ik)2 = −k 2 .
Finally, we consider the behavior of the delta function under the Fourier transform.
Let δ(k) denote the amplitude in the expansion of δ(r), analogous to the relationship
between U (k) and U (r). Using the inverse Fourier transform and the integral
properties of the delta function produces
˚
1 1 −ik·0 1
δ(k) = 3
δ(r)e−ik·r d3 r = 3
e = .
(2π) (2π) (2π)3

In other words, the delta function transforms to 1/(2π)3 in Fourier space.

Recipe: We now state the recipe for solving the Poisson equation in Fourier space.
Take the Fourier transform the Poisson equation from r into k space (where the
Laplacian operator ∇2 simplifies to −ik 2 under the Fourier transform) and solve
for the amplitude U (k) of each plane wave e−ik·r . Substitute U (k) into the Fourier
transform, ˚
U (r) = U (k)eik·r d3 k,

and evaluate the integral—typically in spherical coordinates—to find find U (r).

2.2 Poisson Equation for a Point Particle

The Poisson equation for a point particle is
q
∇2 U (r) = − δ(r).
ϵ0
Solve the equation for U (r).

8
2.2. Poisson Equation for a Point Particle

The plan is to transform into k space, solve for U (k), then transform back to U (r).
First, we take the Fourier transform of both sides—use the Fourier transform identities
∇2 → −k 2 and δ(r) → (2π) 1
3 from the theory section.

q 1 q
−k 2 U (k) = − =⇒ U (k) = .
ϵ0 (2π)3 (2π)3 ϵ0 k 2

Next, we find U (r) using a second Fourier transform

˚ ˚ ik·r
qeik·r q e
U (r) = d3 k 2 3
= 3
d3 k.
k ϵ0 (2π) (2π) ϵ0 k2

To solve the integral, we introduce an angle θ between r and k and transform to

spherical coordinates to get
ˆ 2π ˆ 1 ˆ ∞ ikr cos θ
q 2e
U (r) = dϕ d[cos θ] dkk .
(2π)3 ϵ0 0 −1 0 k2

Integration over the azimuthal angle ϕ is simple, and produces 2π:

ˆ 1ˆ ∞
q
U (r) = ei cos θkr d[cos θ] dk.
(2π)2 ϵ0 −1 0

Next, we integrate first over θ, (to avoid ei cos θ·∞ from the upper k limit) and recognize
the sine function in the difference of exponential functions:
ˆ ∞ 1 ˆ ∞ ikr
q ei cos θkr q e − e−ikr
U (r) = dk = dk
(2π)2 ϵ0 0 ikr θ=−1 (2π)2 ϵ0 0 ikr
ˆ ∞ ˆ ∞
q 2 sin(kr) 2q
= dk = sinc(κr) dk
(2π)2 ϵ0 0 kr (2π)2 ϵ0 0

The integral of the sinc function is

ˆ ∞
sin x π
dx = ,
0 x 2

and applying this integral to the expression for U (r) gives

2q π q
U (r) = 2
= ,
(2π) ϵ0 2r 4πϵ0 r
which is the electric potential of a point charge. Note that we have derived this
expression directly from Maxwell’s equations, rather than taking it for granted as in
introductory electromagnetism courses.
Finally, we substitute the result for U (r) into the Poisson equation for a point charge.
The result is
q q 1
∇2 = − δ(r) =⇒ ∇2 = −4πδ(r),
4πϵ0 r ϵ0 r
which will be useful in the next problems.

9
2.3. Theoretical Interlude: Electric Field of a Charge Distribution

2.3 Theoretical Interlude: Electric Field of a Charge Distribution

We just solved the Poisson equation for the simple case ρ(r) = δ(r). Can we use this
result to solve the general case ρ = ρ(r)? The answer is yes, if we expand ρ(r) over a
basis of delta functions, as follows:
˚
ρ(r) = d3 r̃ρ(r̃)δ(r − r̃).

In this case, the solution of U (r) to the Poisson equation is

˚ ˚ 3
1 1 d r̃ρ(r̃)
U (r) = d3 r̃ρ(r̃) = .
4πϵ0 |r − r̃| 4πϵ0 |r − r̃|

This result is quite powerful—by solving the Poisson equation for a delta function
and then expanding an arbitrary ρ(r) in terms of the delta function, we now have
the solution to the Poisson equation for any ρ(r). We find the corresponding electric
field with ˚ 3
1 d r̃ρ(r̃) r − r̃
E = −∇U = ,
4πϵ0 |r − r̃|2 |r − r̃|
which agrees with the equation quoted in the previous exercise set.

2.4 Electric Field of a Hydrogen Atom

The hydrogen atom has the electric potential

q e−αr  αr  2
U (r) = 1+ , α= .
4πϵ0 r 2 rB

Find the charge density ρ(r) that generates this potential.

We use the Poisson equation, which connects U and ρ via

ρ(r)
∇2 U (r) = − .
ϵ0
We then calculate the Laplacian of our U (r) and work in spherical coordinates, since
the potential is spherically symmetric (depends only on r). As a review, when acting
on a function that depends only on r, ∇2 in spherical coordinates reads
 
1 ∂ ∂
∇2 = 2 r2 .
r ∂r ∂r

Applying ∇2 to U (r), after some straightforward but rather tedious differentiation,

leads to
qα3 −αr
∇2 U (r) = e .
8πϵ0
Rearranging the Poisson equation then gives

qα3 −αr
ρ(r) = − e .
8π

10
2.4. Electric Field of a Hydrogen Atom

Note that the charge density is negative, which corresponds to the negatively charged
electron cloud. Inserting the definition of α = r2B gives

q − r2r
ρ(r) = − 3 e
B .
πrB

− 2r − r
2
Another interpretation: e rB is equivalent to e rB , which is the square of the
hydrogen atom’s ground state wave function. The square of the wave function is
probability, and multiplying the probability by rq3 gives a charge density.
B

Note that—incorrectly—the proton’s charge does not contribute to our expression for
ρ(r). This is because the proton occurs at the origin, which corresponds to a charge
density singularity at the origin. To avoid dealing with this singularity, we simply
ignored it when evaluating the Laplacian ∇2
We can resolve this problem by separately considering the special case
q
lim U (r) = .
r→0 4πϵ0 r
which just says the that potential should approach the ∼ 1/r proton potential at the
origin. We would then have to solve the Poisson equation for this potential. However,
we already now the solution—the potential is the potential for a point charge and
corresponds to a charge density

ρ(r) = qδ(r).

The correct total result for the hydrogen atom is the sum of the electron cloud result
and the charge density of the proton nucleus:

qα3 −αr
ρ(r) = qδ(r) − e .
8π

Lesson: Be careful when working with the Poisson equation if U (r) has singularities!

11
3.1. Conducting Ribbon in a Parallel-Plate Capacitor

3 Third Exercise Set

Theory: The Laplace Equation As a quick review from the last exercise set: the
Poisson equation used to solve for the electric field potential generated by a charge
density ρ is

ρ(r) ∂2 ∂2 ∂2
∇2 U (r) = − where ∇2 = + + .
ϵ0 ∂x2 ∂y 2 ∂z 2

Often ρ(r) = 0 in places we’re solving for the electric potential. In this case ∇2 U (r) =
0. This equation is called a Laplace equation.

3.1 Conducting Ribbon in a Parallel-Plate Capacitor

We place a long, thin conducting ribbon be-
tween the plates of a large parallel-plate capac-
itor, perpendicularly to the plates; the ribbon
almost touches both plates with a small, in-
sulating layer of air between the ribbon edges
and the plates. The ribbon height and dis-
tance between the plates is a. We ground
both plates and set the ribbon potential to U0 .
What is the resulting electric potential inside
the capacitor?
First, we decide on a coordinate system—we will use a Cartesian coordinates because
of the problem’s rectangular symmetry. We choose the y axis to vertically connect
the two ribbons (along the same line as a in the figure), the x axis to run from left to
right in the plane of the page, and the z axis to point out of the page.
Note that the problem is independent of z (by translation symmetry in the z direction),
so we only need U (x, y). More so, the problem has reflection symmetry, so we can
find the solution on only one side of the ribbon (one half of the x axis) and reflect
the solution about the y axis.
The space between the capacitor plates is empty—there is no charge, and we use the
Laplace equation for the space between the plates.

∇2 U (x, y) = 0.

Charge can occur only along the ribbon or on the capacitor plates.
Next, we determine boundary conditions for U (x, y) so the equation has a unique
solution; the problem’s boundaries are the ribbon and edges of the capacitor plates.
On the bottom plate, U (x, 0) = 0. On the upper plate, U (x, a) = 0; both potentials
are zero because the plates are grounded, while along the ribbon we have U (0, y) = U0 .
We need one more boundary—infinity. At infinity, we require only that U (x → ∞, y)
is bounded, i.e. that U does not diverge at ∞.
First, we attempt solving the problem with separation of variables and write U (x, y) =
X(x)Y (y)—this approach tends to work well with symmetric problems. Substituting

12
3.1. Conducting Ribbon in a Parallel-Plate Capacitor

this ansatz into the Laplace equation and evaluating ∇2 gives

′′ ′′ X ′′ Y ′′
X Y + XY = 0 =⇒ =− .
X Y
The separation of variables is successful—we succeeded in isolating x-dependent and
y-dependent terms on different sides of the equation.
We then set the equation equal to a separation constant κ2 and get two equations,

X ′′ − κ2 X = 0 and Y ′′ + κ2 Y = 0.

Both equations have simple solutions! The equations for X and Y are solved by
exponential and sinusoidal functions, respectively.

X(x) = Aeκx + Be−κx and Y (y) = C sin(κy) + D cos(κy).

We then substitute the expressions for X and Y back into the ansatz U = XY :

U (x, y) = X(x)Y (y) = Aeκx + Be−κx (C sin(κy) + D cos(κy)) .

We find the coefficients A, B, C and D using the boundary conditions.

• We begin with the most powerful condition, that U (x → ∞, y) is bounded.
This condition implies A = 0 to suppress the divergent exponential function eκ .
• Then, we use the next two simplest conditions, the ones requiring U (x, y) = 0.
Beginning with U (x, 0) = 0, we have

0 ≡ U (x, 0) = 1 · (0 + D) =⇒ D = 0.

With both A = D = 0, we’re left at this point with only

U (x, y) = Be−κx · C sin(κy).

• Next, we apply the condition U (x, a) = 0 to get

0 ≡ U (x, a) = Be−κx C sin(κa) ≡ F e−κx sin(κa),

where we’ve joined the product of two constants into a single constant F = BC.
We have two options: either F = 0 or sin(κa) = 0. The option F = 0 gives the
trivial solution U (x, y) = 0. The non-trivial solution comes from

sin(κa) = 0 =⇒ κa = nπ, n = 1, 2, 3, . . . .

Note that n = 0 leads to a trivial solution U (x, y) = 0, which we reject.

With respect to the quantization of κ and F by the index n, the general solution
at this point is the linear superposition
∞
X
U (x, y) = Fn e−κn x sin(κn y).
n=1

13
3.1. Conducting Ribbon in a Parallel-Plate Capacitor

• To find Fn , we use the final boundary condition U (0, y) = U0 .

∞
X ∞
X  nπy 
U0 = Fn sin(κn y) = Fn sin ,
a
n=1 n=1

where we’ve used κn = a .

nπ
This is a Fourier expansion of the constant U0 over
sine functions.
We find the coefficients by taking the inner product of both sides of the equation,
which amounts to multiplying both sides by sin mπy a and integrating both sides
over y from 0 to a. The left side of the equation for U0 becomes
ˆ a   mπy  a U a 
mπy  U0 a 0
1 − (−1)m .

U0 sin dy = − cos =
0 a mπ a 0 mπ

On the right hand side, with the sum, we switch the sum and integral to get
∞ ˆ a  nπy   mπy  ∞ ˆ a  mπy 
X X Fm a
Fn sin sin dy = Fn δmn sin2 dy = .
0 a a 0 a 2
n=1 n=1

Because of the orthogonality of the sine functions, the integral is zero for m =
̸ n,
and only the case m = n gives a non-zero result. Equating the two sides gives
the desired expression for Fm :
U0 a   Fm a 2U0 
1 − (−1)m = 1 − (−1)m .

=⇒ Fm =
mπ 2 mπ

With Fm known, the final result for U (x, y) is then

∞
2U0 X 1 − (−1)n  nπx   nπy 
U (x, y) = exp − sin .
π n a a
n=1

Some limit cases: for x ≫ a, the exponent terms very small, and we can neglect all
πx
terms in the series except the first term e− a with n = 1. The result is
4U0  πx   πy 
U (x, y) = exp − sin .
π a a

A separate case for which we can find a nice analytic solution is at the center of the
capacitor at y = a2 . The solution reads
∞
a 2U0 X 1 − (−1)n  nπx   nπ 
U (x, 2) = exp − sin .
π n a 2
n=1

Instead of finding U (x, a2 ), we’ll find the electric field E(x, a2 ). Because of reflection
symmetry across the line y = a2 , the electric field cannot have a y component—E
only has an x component. We’ll find Ex (x) from the potential:
∞
∂ 2U0 X   nπx   nπ 
U (x, a2 ) = − 1 − (−1)n exp −

Ex (x) = − sin .
∂x a a 2
n=1

14
3.2. A Halved Conducting Cylinder

Next, note that

n even

0
 n
  nπ  
1 − (−1) sin = 2 n = 1, 5, 9, . . .
2 
−2 n = 3, 7, 11, . . . .


The sum simplifies to

 
4U0 h − πx 3πx 5πx
i 4U
0 − πx 2πx
 2πx 2
Ex (x) = e a − e− a + e− a ∓ . . . = e a 1 − e− a + e− a ∓ ... ,
a a
2πx
which is a geometric series in e− a . The result is
πx
4U0 e− a 4U0 1 2U0
Ex (x) = = πx
− πx
=
.
a 1 + e− 2πx
a a e +e
a a a cosh πx
a

3.2 A Halved Conducting Cylinder

Consider a long cylinder of radius a cut in
half along a plane running along the cylin-
der’s longitudinal axis. We separate the two
cylinder halves by an arbitrarily small amount
(so the halves are insulated) and apply a po-
tential difference U0 between the two halves.
The halved cylinder acts as a capacitor. Find
the electric potential inside the cylinder.
We must first decide on a coordinate system,
and will use cylindrical coordinates because
the problem has cylindrical symmetry. Let the z axis run along the cylinder’s
longitudinal axis. Because of translational symmetry along the z axis, U is independent
of z.
There is no charge inside the cylinder, so the electric potential in the cylinder obeys
the Laplace equation
∇2 U (r, ϕ) = 0.
(I have used U for electric potential and ϕ for the cylindrical coordinate system’s
angular coordinate.) In cylindrical coordinates (when acting on a function independent
of z), the Laplacian operator reads
1 ∂2
 
2 1 ∂ ∂
∇ = r + 2 2.
r ∂r ∂r r ∂ϕ
In our case,
1 ∂2U
 
1 ∂ ∂U
∇2 U (r, ϕ) = r + = 0.
r ∂r ∂r r2 ∂ϕ2
Following the usual recipe, we separate variables using the ansatz U (r, ϕ) = R(r)Φ(ϕ).
Substituting this ansatz into the Laplace equation leads to
 ′ 
1 ′ ′ R ′′ R R
Φ (rR ) + 2 Φ = Φ + R + 2 Φ′′ = 0.
′′
r r r r

15
3.2. A Halved Conducting Cylinder

Note that r′ = 1. Dividing through by Φ and rearranging gives

rR′ R′′ Φ′′

+ r2 =− .
R R Φ

Following the usual separation procedure, we set both sides equal to the separation
constant m2 . The equations for Φ and R read

Φ′′ + m2 Φ = 0 and r2 R′′ + rR′ − m2 R = 0.

The solution for Φ is sinusoidal:

Φ(ϕ) = A sin(mϕ) + B cos(mϕ).

Now, our cylindrical problem is periodic in ϕ with period 2π—this just means the
cylinder repeats after one revolution. Periodicity in ϕ is possible only if m takes on
integer values, so we can immediately index the solutions for Φ with

Φm (ϕ) = Am sin(mϕ) + Bm cos(mϕ), m = 1, 2, 3, . . . .

We only use positive integers because the odd/even symmetry of sin and cos means
negative integers give the same result as positive one—solving for negative m would
be redundant. We reject m = 0 because this solution leads to Φ′′ = 0, meaning Φ is
a linear function. But a linear function can’t be periodic in ϕ, so we reject m = 0.
The second equation for R is solved with powers of r. The result is

Rm (r) = Cm rm + Dm r−m .

The general solution is the linear superposition

∞
X ∞
X
(Am sin(mϕ) + Bm cos(mϕ)) Cm rm + Dm r−m .


U (r, ϕ) = Φm (ϕ)Rm (m) =
m=1 m=1

Note: We ended the problem at this point (ran out of time) and continued in the
fourth exercise set.

16
4.1. A Halved Conducting Cylinder (continued)

4 Fourth Exercise Set

4.1 A Halved Conducting Cylinder (continued)
We left off in the previous exercise set with the general solution
∞
X
U (r, ϕ) = Φm (ϕ)Rm (m)
m=1
X∞
(Am sin(mϕ) + Bm cos(mϕ)) Cm rm + Dm r−m


=
m=1

for the potential inside the cylinder. To find a solution specific to our problem, we
apply boundary conditions. We already applied the periodic boundary condition
U (r, ϕ) = U (r, ϕ + 2π), which required m be integer-valued.
A second boundary condition requires the capacitor halves have a potential difference
U0 between them. It is best to write this potential difference in the symmetric form
(
U0
ϕ ∈ (0, π)
U (a, ϕ) = 2 U0
− 2 ϕ ∈ (π, 2π).
There is another condition—that U does not diverge at r = 0. This condition implies
the Dm coefficients are zero, because the Dm r−m term diverges at r = 0.
Observation: the second boundary condition is an odd function of ϕ. This implies
that only odd (sine) terms can appear in the final solution. This allows us to set the
Am coefficients equal to zero to eliminate the cosine terms. We are left with
∞
X
U (r, ϕ) = Fm rm sin(mϕ),
m=1

where we have defined Bm Cm ≡ Fm .

Applying the second boundary condition at r = a gives
∞
X
U (a, ϕ) = Fm am sin(mϕ).
m=1

To solve this, we take the scalar product of the equation with sin(nϕ):
ˆ 2π ˆ 2π X
∞
U (a, ϕ) sin(nϕ) dϕ = Fm am sin(mϕ) sin(nϕ) dϕ.
0 0 m=1

Plugging in the step values of U (a, ϕ) gives

ˆ ˆ ˆ 2π X
∞
U0 π U0 2π
sin(nϕ) dϕ − sin(nϕ) dϕ = Fm am sin(mϕ) sin(nϕ) dϕ.
2 0 2 π 0 m=1

First, we solve the left-hand side

 
U0 1 π 1 2π U0
− cos(nϕ) 0 + cos(nϕ) π
= [− cos(πn) + 1 + 1 − cos(nπ)]
2 n n 2n
U0
= (1 − (−1)n ) ,
n

17
4.1. A Halved Conducting Cylinder (continued)

where we’ve used the identity cos(nπ) = (−1)n .

Next, we solve the right-hand side. Switching the order of integration and summation
gives
∞ ˆ 2π ∞  
m 2π
X X
m
Fm a sin(mϕ) sin(nϕ) dϕ = Fm a δmn = Fn an π.
0 2
m=1 m=1

Combining the left and right sides gives

U0 1 − (−1)n
 
Fn = .
πnan
So, the solution for U (r, ϕ) is
∞ ∞   
"  #
U0 1 − (−1)n n 1 − (−1)n

X U 0
X r
U (r, ϕ) = rn sin(nϕ) = sin(nϕ).
πnan π a n
n=1 n=1

Next, some limiting cases. It will be easier to work in terms of electric field instead of
potential. We will find the electric field in the two planes parallel and perpendicular
to the slit between the capacitor halves.
First, in the perpendicular (vertical) plane. The field points from high to low potential,
so from the top half of the capacitor to the bottom half. In this plane we can work
with just one coordinate r, which represents the vertical distance from the cylinder’s
center. Note that ϕ = π2 . The component Er we’re after is
∞
U0 X  r n−1 1 1 − (−1)n
 
∂
Er = − U (r, ϕ) =− n sin(nϕ)
∂r ϕ= π2 π a a n ϕ= π2
n=1
∞
U0 X  r n−1 1 − (−1)n
   nπ 
=− sin .
πa a n 2
n=1

The sum simplifies considerably when you realize

n even

1 − (−1) n  nπ  0

sin = 2 n = 1, 5, . . .
n 2 
−2 n = 3, 7, . . . .


We can then write the field as a geometric series

 
−2U0  r 2  r 4 −2U0 1
Er = 1− + ∓ ··· =
.
πa a a πa 1 + r 2
a

Note that Er is largest at r = 0, decreases monotonically with increasing r, and falls

to half of its maximum value at r = a.
In a field parallel to the slit, we would set ϕ = 0. This plane is perpendicular to the
vertical plane, which used the radial component Er , so for the parallel plane we work
with the ϕ component Eϕ .
1 ∂
Eϕ = − U (r, ϕ) .
r ∂ϕ ϕ=0

18
4.2. Conducting Sphere in a Uniform Electric Field

As before, the sum simplifies considerably to a geometric series. The result turns out
to be
2U0 1
Eϕ = −
.
πa 1 − r 2
a
Note that Eϕ diverges at r = a. This is a consequence of the very small slit spacing
between the capacitor halves at r = a; schematically have E = Ud0 → ∞ as d → 0.

4.2 Conducting Sphere in a Uniform Electric Field

A conducting sphere of radius a is placed in a uniform electric field E0 pointing in
the z direction. Find the electric potential U inside and outside the sphere.
We will use spherical coordinates to take advantage of spherical symmetry, meaning
we aim to solve for electric potential U in the form U (r, ϕ, θ). Because of the problem’s
rotational symmetry, the solution will be independent of ϕ, and we need only solve
for U (r, θ).
The sphere is at a constant potential because it is a conductor. We’ll set U = 0 inside
the sphere. In the space around the sphere, we solve the Laplace equation

∇2 U (r, θ) = 0.

We then separate variables via U = R(r)Θ(θ), which leads to the general solution
∞ h
X i
U (r, θ) = Al rl + Bl r−(l+1) Pl (cos θ),
l=0

where Pl are the Legendre polynomials.

On to the boundary conditions. On the surface we’ll set U (a, θ) = 0. And at infinity,
we use the boundary condition

U (r → ∞, 0) = −E0 z = −E0 r cos θ.

This is the potential of a uniform electric field (at infinity, the potential from the
sphere is negligible). This potential is chosen so that

∂U
− = E0 ,
∂z
i.e. so that the potential at infinity recovers the uniform electric field E0 .
We’ll start with the second boundary condition at r → ∞. Applying the condition to
the general solution gives
∞
X
Al rl Pl (cos θ) = −E0 r cos θ.
l=0

Note that the r−(l+1) terms vanish as r → ∞. The entire series equals a single
term proportional to cos θ and the equality holds if only the l = 1 term in the series

19
4.2. Conducting Sphere in a Uniform Electric Field

is non-zero, which generates a corresponding cos θ term from P1 (cos θ) = cos θ ie.
P1 (x) = x. The l = 1 term is

A1 r cos θ = −E0 r cos θ =⇒ A1 = −E0 ,

so we have Al = −E0 δl1 . The solution for U (r, θ) simplifies to

∞
X
U (r, θ) = −E0 r cos θ + Bl r−(l+1) Pl (cos θ).
l=1

Next, the second boundary condition: U (a, θ) = 0. Substituting the condition into
the intermediate solution gives
∞
X
E0 a cos θ = Bl a−(l+1) Pl (cos θ).
l=1

Again, the entire series equals only a single term. Again, this will be the l = 1 term
corresponding to P1 (cos θ) = cos θ. For l ̸= 1 we have Bl = 0. The l = 1 term gives

E0 a cos θ = B1 a−2 cos θ =⇒ B1 = E0 a3 .

With Al and Bl known for all l, the final result is

E0 a3
U (r, θ) = −E0 r cos θ + cos θ.
r2
The first term, −E0 r cos θ, is the potential of the uniform external field E0 . The
second term comes from the sphere. In fact, this second term has the same form as
the potential of an electric dipole!
Limiting cases are discussed in the next exercise set.

20
5.1. Conducting Sphere in a Uniform Electric Field (continued)

5 Fifth Exercise Set

5.1 Conducting Sphere in a Uniform Electric Field (continued)
We left off in the previous exercise set with the electric potential due to the conducting
sphere, which as
E0 a3
U (r, θ) = −E0 cos θ + 2 cos θ.
r
3
More so, we had recognized that the sphere’s contribution Er02a cos θ corresponded to
the potential of an electric dipole. Our next step is to explore the sphere’s dipole
behavior.
We consider an infinitesimal element of the sphere’s surface at the angle θ carrying
charge dq, on the upper hemisphere with positive charge and θ. Recall the electric
field, as for any conductor, is perpendicular to the surface.
We write Gauss’s law for the small surface element, which is simple because the
electric field is perpendicular to the surface
dq de
E⊥ dS = =⇒ = σ = ϵ0 E⊥ .
ϵ0 dS
This equality gives us an expression for σ in terms of the electric field E⊥ perpendicular
to the surface. We can find E⊥ from the potential:

∂U
E⊥ = − = E0 cos θ + 2E0 cos θ =⇒ σ = 3ϵ0 E0 cos θ.
∂r r=a

The charge density’s dependence on θ quantitatively demonstrates the sphere’s

dipole-like charge distribution.
With charge density σ known, we can find the sphere’s electric dipole moment via
ˆ
pe = r̃ dq.

We qualitatively expect pe to point upward (from the negative to the positively

charged hemisphere), and confirm this analytically. By spherical symmetry, only the
z component of pe is non-zero; this is
ˆ ¨ ¨
pez = z̃ dq = (a cos θ) · (σ dS) = (a cos θ) · (3ϵ0 E0 cos θ) · dS.

To find dS, we find the area of a small band of width da around the sphere’s surface.
The band’s area is 2πr da = 2π(a sin θ)(a dθ). The dipole moment pez is then
ˆ π ˆ 1
2 3
pez = (3ϵ0 E0 a cos θ) · 2π(a sin θ)(a dθ) = 6a πϵ0 E0 cos2 θ d[cos θ]
0 −1
 1
3 1
= 6a πϵ0 E0 cos3 θ = 4πϵ0 E0 a3 .
3 −1

21
5.2. Electric Dipole in a Conducting Spherical Shell

5.2 Electric Dipole in a Conducting Spherical Shell

We place an electric dipole with dipole moment pe in the center of a conducting
spherical shell of radius a. What is the electric potential inside the shell?
We will work in spherical coordinates, which are best suited to the problem’s spherical
symmetry. By rotational symmetry, the potential depends only on the coordinates r
and θ, not ϕ. Besides at the center, the charge density inside the sphere is zero, so
we solve the Laplace equation
∇2 U (r, θ) = 0.
The general solution is
∞ 
X 
U (r, θ) = Al rl + Bl r−(l+1) Pl (cos θ).
l=0

where Pl are the Legendre polynomials.

We then apply boundary conditions to find a solution specific to our problem. First,
the potential on the shell’s surface is constant, since the shell is a conductor. For
convenience, we’ll set U (a, θ) = 0. The second boundary condition concerns the
dipole at the sphere’s center. Namely, the potential approaches the potential of an
electric dipole near the sphere’s center. Quantitatively, this condition reads
pe cos θ
U (r → 0, θ) = .
4πϵ0 r2

We begin with the simpler second boundary condition, (the boundary r → 0 eliminates
the rl -dependent term). Inserted into the general solution, the second condition reads
∞
X pe cos θ −2
U (r → 0, θ) = Bl r−(l+1) Pl (cos θ) = r .
4πϵ0
l=0

Note that the entire series sums to only a single term; for this to work, only the l = 1
term in the series can be non-zero, leaving
pe cos θ −2 pe
B1 r−2 cos θ = r =⇒ B1 = and Bl̸=1 = 0.
4πϵ0 4πϵ0
Note the use of P1 (cos θ) = cos θ. The intermediate solution at this stage is
∞
X pe −2
U (r, θ) = Al rl Pl (cos θ) + r cos θ.
4πϵ0
l=0

We apply the second boundary condition U (a, θ) = 0 to get

∞
X pe −2
Al al Pl (cos θ) = − a cos θ.
4πϵ0
l=0

As before, only the l = 1 term can be non-zero to satisfy the equality. The result is
pe −2 pe
A1 a cos θ = − a cos θ =⇒ A1 = − and Al̸=1 = 0.
4πϵ0 4πϵ0 a3

22
5.3. Point Charge Above a Conducting Plane

With the coefficients Al and Bl known, the solution for U (r, θ) is

   
pe pe pe cos θ 1 r
U (r, θ) = − r cos θ = − .
4πϵ0 r2 4πϵ0 a3 4πϵ0 r 2 a3

The r12 term is the dipole’s contribution. The r

a3
comes from the charge induced on
the conducting shell.
The induced term is worth a closer look, noting that r cos θ = z.
pe pe
Uinduced = − r cos θ = − z.
4πϵ0 a3 4πϵ0 a3
In particular, the associated electric field is
∂ pe
Einduced = − Uinduced = .
∂z 4πϵ0 a3
In other words, the electric field generated by the induced charge is constant! The
uniform field also tells us about the charge distribution on the sphere’s surface: to
create a uniform field in the z direction, the shell must have a dipole-like charge
distribution, with positive charge on the lower hemisphere and negative charge on
the upper hemisphere.
Next, we’re interested in the analytic expression for the surface charge density σ.
We consider a small surface element dS, and consider the total electric field at that
surface. The electric field must be perpendicular to the surface, since the shell is a
conductor. Gauss’s law applied to the surface element reads
dq dq
−E⊥ dS = =⇒ σ ≡ = −ϵ0 E⊥ .
ϵ0 dS
Note the minus sign, indicating the field’s electric flux leaving the surface element
from inside the shell. We find an expression for E⊥ from U :
 
∂U pe cos θ 2 1 3pe
E⊥ = − =+ 3
+ 3 = cos θ.
∂r r=a 4πϵ0 r a r=a 4πϵ0 a3

The associated surface charge density is

3pe
σ = −ϵ0 E⊥ = − cos θ.
4πa3

5.3 Point Charge Above a Conducting Plane

Consider a positive point charge q a distance d above a large, grounded conducting
plane. What is the electric potential in space due to the charge-plane system?
We will use a trick called the method of images to solve the problem. Namely, we
imagine a negative point charge −q a distance d below the plane—a mirror image of
the original positive charge. The resulting charge distribution is an electric dipole.
Note: placing an imaginary negative charge a distance d below the plane does not
change the field above the plane due to the positive charge.

23
5.3. Point Charge Above a Conducting Plane

Considering both points, the potential at an arbitrary position r from the origin is
q 1 q 1
U (r) = − ,
4πϵ0 |r − d| 4πϵ0 |r + d|

where the vector d points perpendicularly up from the plane toward the positive
charge. Introducing an angle θ between d and r, we have
p
|r ± d| = r2 + d2 ± 2rd cos θ.

The expression for U (r) for the two charges is then simply
 
q 1 1
U2 (r) = √ −√ .
4πϵ0 r2 + d2 ± 2rd cos θ r2 + d2 ± 2rd cos θ
For the original configuration of a single positive charge a distance d above the plane,
the potential above the plane agrees with U2 , while the potential below the plane,
where there is in reality no charge, is zero. The correct expression for the single
charge +q is then (
U2 (r) above the plane
U (r) =
0 below the plane.

Next, we’re interested in the surface charge density σ(ρ) on the plane where ρ is the
radial distance in the plane from the origin. As usual, we start with Gauss’s law for
a small surface element of the plane:
dq dq
−E⊥ dS = =⇒ σ ≡ = −σE⊥ .
ϵ0 dS
To find E⊥ , we differentiate U with respect to the vertical coordinate z. First, we
introduce z into the expression |r ± d|
p p
|r ± d| = r2 + d2 ± 2rd cos θ = ρ2 + z 2 + d2 ± 2dz.

We then have
" #
q 1 1
U (ρ, z) = p −p .
4πϵ0 ρ2 + z 2 + d2 − 2dz ρ2 + z 2 + d2 + 2dz

We then find E⊥ and then σ with

 
∂U q d d
σ = −ϵ0 E⊥ = −ϵ0 =− +
∂z z=0 4π (ρ2 + d2 )3/2 (ρ2 + d2 )3/2
q d
=− .
2π (ρ + d2 )3/2
2

With surface charge density σ known, we then ask what is the total charge on the
plane. Integrating the plane over rings with area dS = 2πρ dρ, we have
¨ ˆ ∞
qd
qplane = σ dS = − (2πρ dρ).
0 2π(ρ + d2 )3/2
2

24
5.3. Point Charge Above a Conducting Plane

In terms of the new variable u = ρ2 + d2 , the integral evaluates to

ˆ
qd ∞ du 1 ∞
 
qplane = − = qd 1/2 = −q.
2 d2 u3/2 u d2

Summary of what we did: recognize that the field above the plane from the positive
charge looks like half the field of an electric dipole. Since we know the solution
for a dipole, instead of solving the charge-plane system, we solve the (imaginary)
two-charge system, which gives the same field above the plane anyway. We then reuse
the upper half of the dipole solution for the single-charge plane system, and set the
field below the plane equal to zero. The basic idea is: the field above the plane is the
same for both the positive-charge plane system and for a dipole system, so we can
use either approach to solve for the field above the plane.
Next, we ask what is the electrostatic force on the point charge above the plane?
First, some theory: Theory: Electrostatic Force
The total electrostatic force F acting on the charges enclosed in a region of space V
permeated with an electric field E is
‹  
1 2
F = ϵ0 E(E · n̂) − E n̂ dS,
∂V 2

where n̂ is the normal to the surface ∂V enclosing the charges. Like with Gauss’s
law, a good choice of the boundary surface, usually taking advantage of the problem’s
symmetries, tends to simplify the problem. Alternatively, if the electric field vanishes
at infinity, we choose a surface that closes at infinity.
Back to our problem: we choose an infinite surface whose base runs along the plane,
then turns upward and closes at infinity to enclose the upper half of space above the
plane. The field in this case is the same E⊥ calculated above:
q d
E⊥ = .
2πϵ0 (ρ + d2 )3/2
2

For the part of the surface running parallel to the plane, the normal to the surface n̂
points perpendicularly into the plane, parallel to the electric field. The force equation
for the bottom half of the surface reads
¨   ¨
1 ϵ0
Fe = ϵ0 E 2 n̂ − E 2 n̂ dS = n̂ 2
E⊥ dS.
bottom 2 2 bottom

In fact, the contribution from the upper half of the surface is zero—the upper half
extends to infinity, where the electric field vanishes. We only need to integrate over
the bottom of the surface, running parallel to the plane. Writing dS = 2πρ dρ and
substituting in the expression for E⊥ , the force reads
ˆ ˆ ∞
ϵ0 ∞ q 2 d2 q 2 d2 ρ
Fe = n̂ 2 2 2 3
2πρ dρ = n̂ dρ
2
2 0 4π ϵ0 (ρ + d ) 4πϵ0 0 (ρ + d2 )3
2
ˆ ∞  ∞
q 2 d2 du q 2 d2 1 q2
= n̂ 3
= − n̂ = n̂.
8πϵ0 d2 u 16πϵ0 u2 d2 16πϵ0 d2

25
5.3. Point Charge Above a Conducting Plane

The force points in the direction of n̂—downward into the plane. A final note: if we
write
q2
Fe = n̂,
4πϵ0 (2d)2
the force takes the form of the electric force between a positive and negative charge
separated by a distance 2d—the same situation we used in the method of images
above.

26
6.1. Force on a Conducting Spherical Shell

6 Sixth Exercise Set

Theory: Electrostatic Force Recall from the previous exercise set that the total
electrostatic force F acting on the charges enclosed in a region of space V permeated
with an electric field E is
‹  
1 2
Fe = ϵ0 E(E · n̂) − E n̂ dS,
∂V 2
where n̂ is the normal to the surface ∂V enclosing the charges.

6.1 Force on a Conducting Spherical Shell

We place a conducting sphere of radius a in a homogeneous electric field E0 . Find the
electrostatic force on the upper half of the sphere.
Suppose the field points in the z direction. Recall from the previous exercise set that
the potential from the sphere and electric field is
E0 a3
U (r, θ) = −E0 r cos θ + cos θ.
r2
Qualitatively, there are two main contributions to the force on the sphere: an upwards
contribution in the positive z direction from the external electric field, and a downward
contribution in the negative z direction from the negative charge accumulated on the
bottom half of the sphere.
We are interested in the force on the upper half of the sphere—the next step is to
choose a surface around the sphere’s upper half that will simplify the force calculation.
Recall the field points perpendicularly out of the conducting sphere’s surface at all
points.
With this perpendicular field in mind, choose a surface that tightly hugs the sphere’s
upper half—in this case, the field and normal to the surface n̂ are parallel at all
points outside the sphere. In the hemisphere plane inside the sphere, there is no field
at all. These to facts simplify the dot product E · n̂ in the force equation.
We then have E · n̂ = E and E(E · n̂) = E 2 n̂. The contribution to the force on along
the sphere’s outside surface is
¨
1 2
Fe = ϵ 0 E n̂ dS.
outer 2
The contribution from the hemisphere plane through the sphere zero, since E = 0
inside the sphere.
Next, we find the magnitude E on the sphere’s surface from the potential U (r, θ).
The field points radially outwards, so we differentiate U with respect to r to get
∂U
E= = E0 cos θ + 2E0 cos θ = 3E0 cos θ.
∂r r=a

Inserting E into the force equation gives

¨
1
2
Fe = ϵ0 3E0 cos θ n̂ dS.
2

27
6.2. Point Charge Between Two Conducting Plates

In spherical coordinates, the unit normal n̂ to the sphere’s surface is n̂ = (sin θ cos ϕ, sin θ sin ϕ, cos θ).
The surface element dS at the surface r = a is (just like in the previous exercise sets)
dS = a2 dϕ sin θ dθ. The force on the sphere’s upper half is then
ˆ ˆ
 
sin θ cos ϕ
9ϵ0 E02 π/2 2π
Fe = cos2 θ  sin θ sin ϕ  (a2 sin θ dθ dϕ).
2 θ=0 ϕ=0 cos θ
Both the x and y components will be zero—integrating cos ϕ and sin ϕ over a full
period 2π give zero, while the ϕ contribution to the z component is 2π. We make
this explicit with
ˆ
 
0
9ϵ0 E02 π/2
Fe = cos2 θ  0  (a2 sin θ dθ).
2 θ=0 2π cos θ
The non-zero z component Fz is
ˆ ˆ π/2
9ϵ0 E02 π/2 3 2 2 2
Fz = 2π cos θ(a sin θ dθ) = 9πϵ0 E0 a cos3 θ(sin θ dθ)
2 θ=0 θ=0
ˆ 1
9πϵ0 a2 E02
= 9πϵ0 E02 a2 cos3 θ d[cos θ] = .
0 4
The vector force can be written simply as
9πϵ0 a2 E02
F= ẑ.
4
In other words, the force on the upper half points upward in the positive z direction.

6.2 Point Charge Between Two Conducting Plates

We place two large conducting plates at a right angle to each other, so that the plates
come close together but just barely do not touch. We then place a point charge q along
the line bisecting the right angle between the plates, at a perpendicular distance a
from each plate. Both plates are grounded. What is the electric potential in the region
bounded by the plates at large distances from the plates’ intersection?
Assume r = 0 along the line connecting the two plates. For a single plate, we could
solve the problem with the method of images—see the previous exercise set. With
two plates we proceed analogously, with a mirror image for each plate. Because of
the two reflections from the two plates, we end up with three imaginary charges plus
the one original one in a quadrupole arrangement. (This is hard to describe in words,
it is best to see a picture). For large r, the charge arrangement will have the field of
an electric quadrupole. Solving the problem thus reduces to a multipole expansion of
the electric potential to the quadrupole term.

6.2.1 Theory: The Multipole Expansion

The multipole expansion of U to quadrupole order, using the Einstein summation
convention, is  
1 q pi ri Qij ri rj
U (r) = + 3 + ,
4πϵ0 r r r5

28
6.2. Point Charge Between Two Conducting Plates

where p is the electric dipole moment and Q is the quadrupole moment tensor.
Note: we could think of the charge q is a scalar monopole moment, creating a
logical progression from scalar monopole moment to vector dipole moment to tensor
quadrupole moment.
We find the dipole moment with
˚
p= d3 r̃ρ(r̃)r̃.

We find the quadrupole moment by components:

˚
d3 r̃ρ(r̃) 3r̃i r̃j − δij r̃2 .


Qij =

The discrete analog a configuration of N charges reads

N
X
qn 3rni rnj − δij rn2 .


Qij =
n=1

Note that both definitions produces a symmetric tensor. Also important: the tensor’s
trace—the sum of the diagonal elements is zero:
X   X  2
qn 3x2n − rn2 + 3zn2 − rn2 + 3zn2 − rn2 = qn 3rn − 3rn2 = 0.

tr Q =
n n

6.2.2 Returning to the original problem

For our imaginary quadrupole configuration of four charges, the total charge, and
thus the monopole moment, is zero. Analogously, the total dipole moment of the
arrangement, which consists of two positive and two negative charges, is zero—the
two dipoles cancel each other out.
From the three terms in our multipole expansion of U (r), only the quadrupole term
remains. We just have to calculate the quadrupole tensor Qij . We label the four
charges in the imaginary quadrupole configuration as 1, 2, 3, and 4, where 1 is the
original positive charge in the upper right corner, 2 is the negative image charge in
the upper left corner, 3 is the positive image charge in the lower left corner, and 4 is
the negative image charge in the lower right corner.
Using the discrete formula for Qij , the first component Qxx is
N
X
qn 3x2n − δij rn2 = q 3a2 − 2a2 + (−q) 3a2 − 2a2




Qxx =
n=1
= q 3a2 − 2a2 + (−q) 3a2 − 2a2 = 0.

The other diagonal terms Qyy and Qzz will analogously sum to zero.
All off-diagonal terms with a z component are also zero, since the charges lie in a
plane with z = 0. We thus have Qxz = Qzx = Qyz = Qzy = 0. We have just two

29
6.2. Point Charge Between Two Conducting Plates

terms left calculate: Qxy and Qyx . By the tensor’s symmetry, the two are equal, so
we really only have one term:
N
X
qn 3xn yn − 0 · rn2 = 3qa2 + 3(−q)(−a2 ) + 3qa2 + 3(−q)(−a2 )


Qxy =
n=1
= 12qa2 .

The quadrupole tensor is

12qa2 0
 
0
Q = 12qa2 0 0 .
0 0 0

As expected, the tensor is symmetric with trace tr Q = 0.

Recall the quadrupole expansion of U (r):
 
1 q pi ri Qij ri rj
U (r) = + 3 + .
4πϵ0 r r r5

In our case, with q = 0 and p = 0, we have

12qa2 xy 12qa2 yx 6qa2 xy

 
1
U (r) = 5
+ 5
+ 0 + · · · + 0 =
4πϵ0 r r πϵ0 r5
6qa2 cos ϕ sin ϕ sin2 θ
= .
πϵ0 r3
The second line uses the spherical coordinates x = r cos ϕ sin θ and y = r sin ϕ sin θ.
In fact, the expression for U (r) takes the exact same form as the wave function of a
d electron orbital (angular momentum quantum number l = 2) in a hydrogen atom.
The equipotential surfaces of U (r) have the same spatial distribution as the dxy
orbitals for a hydrogen wave function.

30
7.1. Theory: Magnetic Vector Potential and the Biot-Savart Law

7 Seventh Exercise Set

7.1 Theory: Magnetic Vector Potential and the Biot-Savart Law
We will need to use two more Maxwell equations for magnetostatics. The first is

∇ · B = 0.

This equation rules out the possibility of magnetic monopoles and allows B to be
written as the curl of a vector potential A as

B = ∇ × A.

The second Maxwell equation is

 
∂E
∇ × B = µ0 j + ϵ0 .
∂t

For static situations with a constant electric field this simplifies to

∇ × B = µ0 j.

Substituting the expression B = ∇ × A into the static second Maxwell equation

produces
∇ × ∇ × A = ∇ · ∇ · A − ∇2 A = µ0 j.

The magnetic vector potential is defined only up to a constant; we usually choose A

so that ∇ · A = 0. In this convention, we have

∇2 A = −µ0 j,

where j is the current density vector. This equation is a vector analog of the Poisson
equation ∇2 U = − ϵρ0 from electrostatics. Similarly to how the electrostatic potential
U at a point r in region of space with charge density ρ is found with
˚
1 ρ(r̃) d3 r̃
U= ,
4πϵ0 |r − r̃|

the magnetic potential at a point r in region of space with current density j is

˚
µ0 j(r̃) d3 r̃
A(r) = .
4π |r − r̃|

In problems involving one-dimensional conductors, where j is non-zero only along the

conductor, the expression for j(r̃) d3 r̃ simplifies to

j(r̃) d3 r̃ = j(r̃) dS̃ d˜l = I t̂ d˜l,

where t̂ is the unit normal vector tangent to the conductor, I is the current through
the conductor at the point r̃ and d˜l is a small distance element along the conductor’s

31
7.2. Magnetic Field of a Circular Current Loop

length. The magnetic vector potential for a one-dimensional conductor carrying a

current I then simplifies to
ˆ
µ0 I t̂ dl
A(r) = .
4π |r − r̃|

Recall B = ∇ × A. Taking the curl of the general expression for A in terms of

current density j gives general expression for the magnetic field
ˆ
µ0 j(r̃) × (r − r̃) 3
B(r) = d r̃.
4π |r − r̃|3
This is a general form of the Biot-Savart law for the magnetic field B of a current
distribution.

7.2 Magnetic Field of a Circular Current Loop

A closed conducting loop of radius a carries current I. What is the magnetic vector
potential far from the conducting loop?
Our starting point is the vector potential of a one-dimensional conductor from the
theory section, i.e. ˆ
µ0 I t̂ dl
A(r) = .
4π |r − r̃|
We need expressions for r, r̃ and t̂.
Assume the loop lies in the x, y plane with the z axis normal to the loop. For
mathematical convenience, we rotate the x, y plane so r lies in the x, z plane (i.e.
ϕ = 0); this just gives us one less non-zero component to work with, since the sin ϕ
term in the y component of r is zero. In polar coordinates r reads


r = r sin θ, 0, r cos θ
, where θ is the angle between r and the z axis.
The integration variable r̃, which runs over the current loop in the x, y plane, reads


r̃ = a cos ϕ̃, a sin ϕ̃, 0 ,

where ϕ̃ is the azimuthal angle between r̃ and the x axis.

Finally, the expression for t̂, the tangent to the current loop, is


t̂ = − sin ϕ̃, cos ϕ̃, 0 .

The small distance element is d˜l = aϕ̃ .

We can now put the pieces together in the equation for A. First,
q
|r − r̃| = (r sin θ − a cos ϕ̃)2 + a2 sin2 ϕ̃ + r2 cos2 θ
q
= r2 + a2 − 2ra sin θ cos ϕ̃
r
2a
q
2
≈ r − 2ra sin θ cos ϕ̃ = r 1 − sin θ cos ϕ̃,
r

32
7.3. Magnetic Field of a Rotating Charged Disk

where the last lines uses r ≫ a (recall we’re interested in the solution far from the
conducting loop). We now have, again using a ≪ r =⇒ ar ≪ 1,
 −1/2  
1 1 2a 1 1 2a
= 1− sin θ cos ϕ̃ ≈ 1+ sin θ cos ϕ̃ .
|r − r̃| r r r 2 r

Substituting the expressions for |r−r̃| ,

1
t̂ and d˜l into expression for A gives

ˆ ˆ
 
2π − sin ϕ̃  
µ0 I t̂ dl µ0 I a 1 2a
A(r) = = dϕ̃  cos ϕ̃  1 + sin θ cos ϕ̃ .
4π |r − r̃| 4π r 0 2 r
0

The integrals conveniently simplify, since we are integrating sinusoidal terms over an
entire period. Only the integral of cos2 ϕ̃ in the y component is nonzero. We end up
with Ax = Az = 0 and
ˆ
µ0 I a 2π a sin θ µ0 I a2
Ay (r) = cos2 ϕ̃ dϕ̃ = sin θ,
4π r 0 r 4 r2
or, in vector form,
µ0 I a 2 µ0 I a2
A(r) = sin θ ŷ = ẑ × r̂.
4 r2 4 r2
The second expression is preferable: the first, in terms of ŷ, holds only with x, y plane
rotated so ϕ = 0 and r lies in the x, z plane. The second, which uses sin θ ŷ = ẑ × r̂,
holds for any orientation of the x, y plane.
Finally, using Iπa2 = |m| (where m is the loop’s magnetic dipole moment, which
points in the direction of the loop’s normal), the expression for A simplifies to
µ0 m × r̂ µ0 m × r
A(r) = 2
= .
4πr 4πr3
This is the same form of magnetic vector potential as for a magnetic dipole. In other
words, a circular current loop behaves as magnetic dipole at long distances.

7.3 Magnetic Field of a Rotating Charged Disk

Consider a charged, insulating disk with uniform surface charge density σ and radius
a. The disk rotates uniformly about an axis through its center with constant angular
speed ω. Find the magnetic field along the axis of rotation.
We choose a coordinate system so the disk lies in the x, y plane and the rotation axis
coincides with the z axis, so ω = (0, 0, ω). Start with the general Biot-Savart law
˚
µ0 j(r̃) × (r − r̃) 3
B(r) = d r̃,
4π |r − r̃|3
and, like in the previous problem, find expressions for each vector quantity in the
equation. The expression for r along the z axis is simply r = (0, 0, z), while the
expression for r̃, which lies in the disk in the x, y plane, is


r̃ = r̃ cos ϕ̃, r̃ sin ϕ̃, 0 .

33
7.3. Magnetic Field of a Rotating Charged Disk

The difference of r and r̃ and its magnitude is

p
and


r − r̃ = − r̃ cos ϕ̃, −r̃ sin ϕ̃, z |r − r̃| = r̃2 + z 2 .
Finally, the current density j, which points tangent to the disk’s rotation, is
j = j(− sin ϕ̃, cos ϕ̃, 0).

Using the expressions for our vector quantities, the cross product j(r̃) × (r − r̃) is
    
− sin ϕ̃ −r̃ cos ϕ̃ z cos ϕ̃
j(r̃) × (r − r̃) = j  cos ϕ̃   −r̃ sin ϕ̃  = j  z sin ϕ̃  .
0 z r̃
Substituting j(r̃) × (r − r̃) and |r − r̃| into the Biot-Savart law gives
˚
 
3 z cos ϕ̃
µ0 j d r̃
B(z) =
3/2  z sin ϕ̃  .
4π 2
r̃ + z 2
r̃

Next, we write j d3 r̃ = j dS̃ d˜l = j dS̃(r̃ dϕ̃) = dI r̃ dϕ̃. Note that the product j dS̃ is
the current element dI in the surface element dS̃. The current dI at the radius r̃ on
the disk rotating with period t0 = 2π ω is

dq dq (σ2πr̃ dr̃)
dI = = ω= ω = σωr̃ dr̃.
t0 2π 2π
Substituting j d3 r̃ = dI r̃ dϕ̃ = σωr̃2 dr̃ dϕ̃ into the Biot-Savart law gives
ˆ a ˆ 2π
 
2 z cos ϕ̃
µ0 σωr̃
B(z) = dr̃ dϕ̃
3/2  z sin ϕ̃  .
4π 0 0 2
r̃ + z 2
r̃
The first and second components of B contain integrals cos and sin terms over a full
period—the result is zero. After integrating over ϕ̃, the magnetic field simplifies to
ˆ a
 
2 0
µ0 σωr̃
B(z) =
4π 0
dr̃
3/2  0  .
r̃ + z 2
2
2πr̃
Only the z component of B is non-zero; it is
ˆ
µ0 a σωr̃3
Bz (z) = dr̃
3/2 .
2 0 r̃2 + z 2
In terms of the new variable u = r̃2 + z 2 , the integral evaluates to
ˆ 2 2
µ0 σω z +a u − z 2 µ0 σω h 1/2 2 −1/2
iz 2 +a2
Bz (z) = du = 2u + 2z u
2 2 z2 u3/2 4 a2
2
p 
µ0 σω z
= z 2 + a2 − z + √ −z
2 z + a2
2

µ0 σω 2z 2 + a2
 
= √ − 2z .
2 z 2 + a2

34
7.3. Magnetic Field of a Rotating Charged Disk

The magnetic field along the z axis is thus B = (0, 0, Bz ) with Bz as above.
Next, we consider the limit case z ≫ a, far from the rotating disk. Expanding the
square root to fourth order in the small quantity az , multiplying out and simplifying
like terms gives
 
µ0 σω  2z 2 + a2 a2 a2 3 a4
   
µ0 σω
Bz = − 2z ≈
 2z + 1− 2 + − 2z
8 z4
q
2 z 1+ a
2 2 z 2z
z2
a2
a2 1 a4 3 a4 3 a6 µ0 σω 1 a4 3 a6
   
µ0 σω
= 2z + − − + + − 2z = + .
2 z z 2 z3 4 z3 8 z5 2 4 z3 8 z5
a6
Neglecting the highest-order z5
term gives the simple result

µ0 σω a4
Bz ≈ , z ≫ a.
8 z3
In other words, far from the disk, the magnetic field falls off as z −3 , just like the field
of a magnetic dipole.
Next, we will try to write the magnetic field in the form Bz ∝ |m| z3
where m is the
disk’s magnetic dipole moment. Integrating over concentric rings with area S carrying
current dI, the disk’s magnetic dipole moment is
ˆ ˆ a ˆ a
2 π
|m| = S dI = (πr̃ ) · (σωr̃ dr̃) = πσω r̃3 dr̃ = σωa4 .
0 0 4

Comparing this expression for |m| to the similar expression for Bz leads to

µ0 σω a4 µ0 |m|
Bz ≈ 3
= ,
8 z 2πz 3
|m|
which is in the desired form Bz ∝ z3
. The general form for the magnetic field of a
magnetic dipole is
µ0 3(m · r)r − mr2
B(r) = .
4π r5
0 |m|
In fact, this expression is equivalent to our result Bz (z) = µ2πz 3 . Since m and r
both point along the z axis, their dot product is m · r = |m|r. Along the z axis,
r = (0, 0, z) and the general expression for the dipole magnetic field simplifies to

µ0 3|m|z 2 − |m|z 2 µ0 |m|

B(r) = ẑ = ẑ,
4π z5 2πz 3
in agreement with our expression for Bz for z ≫ a.

35
8.1. Magnetic Force in a Coaxial Cable

8 Eighth Exercise Set

Theory: Magnetic Force

The magnetic force F on the matter contained in the region of space enclosed in the
region V and permeated by the magnetic field B is
‹  
1 1
F= B(B · n̂) − B2 n̂ dS,
µ0 ∂V 2

where n̂ is the normal vector to the region’s boundary ∂V .

8.1 Magnetic Force in a Coaxial Cable

A long coaxial cable consists of a thin inner wire and outer sheath with radius a.
The inner wire carries a current I, and we connect the sheath to the inner wire at
the cable’s ends so that the current I returns along the outer sheath in the opposite
direction as the current along the inner wire. Find the magnetic force per unit length
on the outer sheath.
There are two contributions to the magnetic force on the sheath: the repulsive,
radially outward force between the sheath and the inner wire and an attractive
“surface tension” force distributed across the sheath’s surface, which carries uniformly
distributed current I (think of the sheath as a collection of parallel conducting wires,
which attract each other).
Consider one-half of the sheath, which forms a semicircular cross section. There are
downward forces F1 at end of the semicircle and an upward magnetic force Fm acting
on the top of the sheath. In equilibrium, the forces are related by Fm = 2F1 .
First, we find the magnetic field in the conductor. Inside the sheath, Ampere’s law
with a circular path encircling the inner wire reads
˛
µ0 I
B · dl = µ0 I =⇒ B(2πr) = µ0 I =⇒ B = .
2πr
Outside the conductor, the net current enclosed by a circular path encircling both
the outer sheath and inner wire is zero, and the Ampere’s law reads
˛
B · dl = µ0 (+I − I) = 0 =⇒ B = 0.

In other words, there is no magnetic field outside the conductor.

With the coaxial cable’s magnetic field known, we now find the magnetic force Fm
on a semi-circular half of the outer sheath.
‹  
1 1 2
Fm = B(B · n̂) − B n̂ dS.
µ0 ∂V 2

We choose an integration surface tightly hugging the half-sheath and work in cylindrical
coordinates r, ϕ. Outside the sheath, the magnetic field is zero and there is no

36
8.2. Tension in a Toroidal Inductor

contribution to Fm . Inside the sheath, the magnetic field is tangent to the semicircle,
so B · n̂ = 0, and the force integral reads
¨
1 1
0 − B 2 n̂ dS.
 
Fm =
µ0 2
µ0 I
The magnetic field at along the sheath (where r = a) is constant and equal to B = 2πa
and can be moved outside the integral. The normal vector in cylindrical coordinates
reads
n̂ = (− cos ϕ, − sin ϕ, 0),
while the surface element is dS = al dϕ. The force integral reads

ˆ π − cos ϕ
 
1 µ0 I 2  − sin ϕ  la dϕ.
Fm = −
2 4π 2 a2 0
0

The x component with cos ϕ integrates to zero over ϕ ∈ [0, π]. Only the y component
is nonzero, and the vector force reads

1 µ0 I 2 µ0 I 2 l
Fm = + (2al) ŷ = ŷ.
2 4π 2 a2 4π 2 a

The forces F1 at the two ends of the semicircle are then

Fm µ0 I 2 l F µ0 I 2
F1 = = =⇒ = .
2 8π 2 a l 8π 2 a

8.2 Tension in a Toroidal Inductor

A toroidal inductor with N coils radius r1 and cross-sectional radius r2 carries a
current I. Find the tension force on a single coil if r1 ≫ r2 .
First, we find the magnetic field inside the inductor with Ampere’s law using a closed
circular path of radius r ∈ (r1 , r1 + r2 ) in the inductor’s equatorial plane. Ampere’s
law reads
µ0 N I
Bin (2πr) = µ0 (N I) =⇒ Bin = .
2πr
Outside the inductor, with a circular path of radius r > r1 + r2 in the inductor’s
equatorial plane, Ampere’s law reads

Bout (2πr) = µ0 (N I + N (−I)) = 0 =⇒ Bout = 0.

Consider a semicircular half of a single inductor coil. As in the previous problem, the
magnetic force with magnitude Fm acts upwards on the top of the semicircle, while
two “surface tension” forces F1 = F2m act downward at the semicircle’s two ends. As
usual, we find the magnetic force Fm using
‹  
1 1 2
Fm = B(B · n̂) − B n̂ dS.
µ0 ∂V 2

37
8.3. Resistance of a Thin Conducting Plate

For the integration surface (awkward to describe, best to see a picture) we choose a
surface that looks like a coin cut in half, basically a thin three-dimensional extension
of a semicircle enclosing the coil’s semicircular upper half. We split the surface into
four parts: the left and right semicircular faces, the circular ribbon along the surface’s
outer radius, and the thin rectangular plane along the surface’s bottom.
The circular upper ribbon occurs just outside the inductor coil (where B = 0) and
thus does not contribute the magnetic force. The left and right semicircular faces
have equal magnitude and opposite-sign contributions, since the normal to the surface
changes sign for each face.
Only the rectangular plane has a nonzero contribution to Fm . Under the assumption
r1 ≫ r2 , the magnetic field along the plane simplifies to
ˆ r1 +r2
µ0 N I µ0 N I
B= dr̃ ≈ .
r1 2πr̃ 2πr1

The magnetic field B points along the toroid’s longitudinal axis (into the plane of a
cross-sectional coil) and is perpendicular to the normal n̂ to the planar integration
surface, so B̂ · n̂ = 0. The magnetic force simplifies to
 ¨
µ0 N I 2
  
1 1 2 1
Fm = 0 − B n̂ dS = − n̂ dS
µ0 2 2µ0 2πr1
µ0 N I 2 µ0 N I 2 r2
   
1 2πr1
=− n̂ r2 =− n̂.
2µ0 2πr1 N 2π r1
The unit vector n̂ points downward, so − n̂ and thus Fm point upward and pull the
coil apart. The magnitude of the tension on the coil is then

Fm µ0 N I 2 r2
F1 = = .
2 4π r1

8.3 Resistance of a Thin Conducting Plate

Consider a conducting plate consisting of
half of a thin annulus with inner radius r1 ,
outer radius r2 , thickness h and conductivity
σ, with electrodes placed at each end. Find
the plate’s electric resistance R when we es-
tablish a potential difference U0 between the
electrodes at the plate’s ends.
We first find the current density in the con-
ducting plate using Ohm’s law

j = σE,

where j and E are the current density and electric field in the conductor. We write j
in terms of the continuity equation
∂ρ
∇·j+ = 0.
∂t

38
8.3. Resistance of a Thin Conducting Plate

For a stationary charge distribution, we have ∇ · j = 0. We then take the divergence

of both sides of the equation j = σE = ρ and apply E = −∇U to get

∇ · j = σ∇ · E = −σ∇2 U = ∇ · j = 0 =⇒ ∇2 U = 0.

We end up with a Laplace equation for U in the conductor. The plan is to find U (r),
then E(r), then j via j = σE, then I and finally resistance with R = UI .
The general solution of the Laplace equation in cylindrical coordinates (including the
m = 0 term) is
∞
X
Am cos(mϕ) + Bm sin(mϕ) · Cm rm + Dm r−m
   
U (r, ϕ) =
m=1
+ (aϕ + b)(c ln r + d).

To find a unique solution, we need boundary conditions for our particular problem.
At the first electrode at ϕ = 0, the electric potential is constant; we’ll set U (r, 0) = 0
for convenience. At the second electrode, the potential is U (r, π) = −U0 to make a
potential difference U0 between electrodes (we choose −U0 so the current runs in the
direction of increasing ϕ).
Along the annulus’s semicircular boundaries the electric field must be tangent to the
surface to satisfy ∇ · j = 0. Along these surfaces the radial component of both j and
E is zero; Er = 0 gives the boundary condition
∂U
Er = − = 0.
∂r r1 ,r2

The general solution for U (r, ϕ) satisfies the boundary condition ∂U

∂r r1 ,r2 = 0 for
all ϕ only if Cm = Dm = c = 0 (try finding to see for yourself). With
∂U
∂r r1 ,r2
Cm = Dm = c = 0, the expression for U (r, ϕ) simplifies to

U (r, ϕ) = (aϕ + b)d ≡ ãϕ + b̃.

The boundary condition U (r, 0) = 0 gives

U (r, 0) ≡= 0 + b̃ =⇒ b̃ = 0.

The final boundary condition U (r, π) = −U0 gives

U0 U0
U (r, π) ≡= −U0 =⇒ ã = − =⇒ U (r, ϕ) = − ϕ.
π π
In other words, the electric potential is a linear function of ϕ.
With U (r, ϕ) known, we find the tangential electric field Et with
1 ∂U U0
Et = − = ,
r ∂ϕ πr
from which we find the tangential current density jt with
σU0
jt = σEt = .
πr

39
8.3. Resistance of a Thin Conducting Plate

We find the total current I from the current density via

¨ ˆ r2  
σU0 σU0 h r2
I= jt dS = · h dr = ln ,
r1 πr π r1

where h is the plate’s thickness. The electrical resistance is then

U0 π
R= = .
I σh ln rr21

40
9.1. Mutual Inductance

9 Ninth Exercise Set

Theory: Inductance
Inductance is the proportionality constant between magnetic flux Φ and current I.
For a single object, e.g. a current loop, the relationship between Φ and I reads

Φ1 = L11 I1 ,

where the quantity L11 is called self-inductance.

For a system of two current-carrying loops, where the current through one loop
generates a magnetic field and thus magnetic flux through the second loop, and vice
versa, the relationships between Φ and I read

Φ1 = L12 I2 and Φ2 = L21 I1 ,

where the quantities L12 and L21 are the loops’ mutual inductances. Without
derivation, we state that L12 = L21 for reasons of symmetry.

9.1 Mutual Inductance

Consider two long, parallel wires of length l
separated by a distance d ≪ l and connected
at their endpoints to form a long conduct-
ing loop. We√ place a square frame with
side length 2d between the wires and sup-
ply the frame with an alternating current
source I1 = I10 sin ωt. Find the system’s mu-
tual inductivity and the induced current in
the wires. Neglect Ohmic losses, and assume
a ≪ d where a is the wires’ radius.
Finding Mutual Inductance
We label the frame as object 1 and the wires
as object 2 and introduce a coordinate y separating the parallel wires so that the
bottom wire occurs at y = 0 and the top wire at y = d. To find the system’s mutual
inductance L12 , we consider the hypothetical magnetic flux Φ1 through the frame
due to a current I2 in the parallel wires. We then find L12 with
Φ1
L12 = .
I2

If the parallel wires carry a current I2 (in opposite directions, since they form a closed
loop), the corresponding magnetic field is
 
µ0 I 2 1 1 µ0 I 2 d
B= + = .
2π y d−y 2π y(d − y)
´
We find magnetic flux with Φ = B · dS; the dot product drops because B is parallel
to the frame’s cross section dS. The square frame’s surface element is dS = 2y dy

41
9.1. Mutual Inductance

(the frame’s width is 2y), and the magnetic flux through the square frame is
¨ ˆ d/2
µ0 I2 d 2y dy 2µ0 I2 d d/2
Φ1 = B dS = 2 =− ln(d − y) 0
0 2π y(d − y) π
 
2µ0 I2 d d 2µ0 d ln 2
= ln d − ln = I2 .
π 2 π

Note the use of symmetry—we integrate only from 0 to d/2 and multiply by two.
The mutual inductance—the proportionality between Φ and I2 is
Φ1 2 ln 2
L12 = = µ0 d.
I2 π
Note that L12 depends only on system’s geometry.
Induced Current in the Parallel Wires
Recall the square frame carries an alternating current

I1 (t) = I10 sin ωt,

so we expect the induced current I2 (t) in the parallel wires to alternate with the
same frequency ω and a general form

I2 (t) = I20 sin(ωt).

We will find the ratio of current amplitudes I20 /I10 .

We’ll solve the problem as follows: use the frame current I1 to find the magnetic flux
Φ2 through the parallel wires, then use Φ2 to find induced voltage U2 in the wires,
and finally use U2 to find the induced current I2 .
We find Φ2 from Φ2 = L21 I1 and the symmetry relation L12 = L21 , i.e.

Φ2 = L21 I1 = L12 I1 ,

where L12 was found the first part of the problem. We then find U2 with

U12 = −Φ̇2 = −L12 I˙1 .

Next, we find I2 from U2 from the general circuit equation

˙
U = RI + Ls I,

which relates the voltage in a loop of resistance R and self-inductance Ls to the

current I through the loop. In our case, neglecting resistance, we have

U12 ≈ L22 I˙2 ,

where L22 is the parallel wire’s self-inductance. Substituting U12 into the earlier
expression U12 = −L12 I˙1 gives

−L12 I˙1 = L22 I˙2 .

42
9.2. The Cabrera Experiment and Magnetic Monopoles

If we assume both I1 and I2 are sinusoidal with amplitudes I10 and I20 , the above
reduces to
I20 L12
= .
I10 L22
What about the minus sign?
To find the current amplitude ratio, we just need to find the parallel wires self-
inductance L22 using Φ2 = L22 I2
If we send a hypothetical current current I2 through the wires, the corresponding
magnetic field is, as before,
 
µ0 I2 1 1
B= + .
2π y d−y

The field and surface are parallel, so B · dS = B dS, and the magnetic flux is
¨ ˆ
µ0 I2 ld d−a 1
 
1
Φ2 = B dS = + dy,
2π a y d−y

where a is the wire’s radius. The integral evaluates to

 
µ0 I2 l d−a a µ0 I2 l d − a
Φ2 = ln − ln = ln .
2π a d−a π a

Applying d ≫ a we have
µ0 I2 l d
Φ2 = ln .
π a
The self-inductance—the proportionality constant between Φ2 and I2 —is thus
µ0 l d
L22 = ln .
π a

With L22 known, we return to the current amplitude ratio to get

2 ln 2
I20 π µ0 d 2 ln 2
= µ0 l d
= l d
.
I10 d ln a
π ln a

To get a better feel for the numbers involved, if we assume l/d = d/a = 10, we have

I20
≈ 0.06.
I10

9.2 The Cabrera Experiment and Magnetic Monopoles

Consider a superconducting current loop with non-zero inductance L, radius a and
a built-in ammeter. Assume a magnetic monopole passes through the loop along the
axis of symmetry. Find the resulting current pulse in the loop.
First, the magnetic field of a hypothetical monopole is
µ0 g r
B(r) = ,
4π r3

43
9.2. The Cabrera Experiment and Magnetic Monopoles

where g is the “magnetic charge”, with units A m. Our plan is to find the magnetic
flux through the loop, use this to find voltage induced in the loop, use the induced
voltage to find current.
Let d(t) be the monopole’s perpendicular distance from the loop’s center, and let ρ
denote radial distance from the current loop’s center (in the plane of the loop).
The magnetic field magnitude a distance r from the monopole is
µ0 g µ0 g 1
B= = .
4π r2 4π d2 + ρ2
We find the magnetic field component B⊥ perpendicular to the current loop with
similar triangles:
d µ0 gd 1
B⊥ = p B= .
2
d +ρ 2 4π (ρ + d2 )3/2
2

The magnetic flux Φ through the loop, using dS = 2πρ dρ, is

¨ ˆ a
µ0 gd a ρ dρ µ0 gd 1
Φ= B⊥ dS = 2 2 3/2
= (−2) p
2 0 (ρ + d ) 4 ρ + d2
2
0
 
µ0 g d
= 1− √ .
2 a + d2
2

We assume the monopole moves with constant speed v and passes through the loop’s
center at t = 0. On the left of the loop (and thus for negative time), the monopole’s
perpendicular distance from the loop is

d(t) = −vt.

The time-dependent flux on the left of the loop is thus

 
µ0 g vt
Φ(t) = 1+ √ .
2 a2 + v 2 t2
On the right of loop (and for positive time), the distance from the loop is d = vt, and
thus the magnetic flux through the loop is
 
µ0 g vt
Φ(t) = − 1− √ .
2 a2 + v 2 t2
Note the additional minus sign, since on the right of the loop, once the particle passes
through, the magnetic flux points in the opposite direction.
Note that Φ(t) increases from 0 at t → −∞ to a maximum value of µ20 g as t → 0 from
the left. As the particle passes through the loop at t = 0, Φ jumps discontinuously to
− µ20 g , and then decreases back to 0 as t → ∞.
Next, we find the induced voltage in the loop. We begin with a modified Maxwell
equation:
∂B
∇×E=− − µ0 jm ,
∂t

44
9.2. The Cabrera Experiment and Magnetic Monopoles

where jm is magnetic current density and accounts for the possible existence of
magnetic monopoles. We then use this modified Maxwell equation to re-derive the
law of induction in the presence of magnetic monopoles. We integrate the equation
over the loop’s surface to get
¨ ¨ ¨
∂
(∇ × E) dS = − B · dS − µ0 jm dS.
S ∂t S S

We then apply Stokes’ theorem, which leads to

˛
E · dl = Ui = −Φ̇ − µ0 Im ,

where Im is magnetic current through the loop’s cross section and Ui is the induced
voltage in the loop. The magnetic current through the loop is nonzero only at the
singular instant when the magnetic monopole passes through, which we model with
the delta function:
Im = gδ(t),
where δ(t) has units s−1 .
If we neglect resistive and capacitive effects, the loop’s circuit equation reads Ui = LI.
˙
Substituting in Ui = −Φ̇ − µ0 Im gives
˙
Ui = Ui = −Φ̇ − µ0 Im = −Φ̇ − µ0 gδ(t) = LI.

We then integrate this equation over time to get

ˆ t ˆ t ˆ t
L I˙ dt̃ = − Φ̇ dt̃ − µ0 g δ(t̃) dt̃
−∞ −∞ −∞
t t
LI(t̃) = −Φ(t̃) − µ0 gH(t),
−∞ −∞

where H(t) is the Heaviside step function. Since I(t → −∞) = Φ(t → −∞) = 0
(when the monopole is infinitely far from the loop), the equation reduces to

LI(t) = −Φ(t) − µ0 gH(t).

Remember that Φ changes from µ20 g to − µ20 g as the monopole passes through the loop
at t = 0. The discontinuity in Φ is exactly balanced by the Heaviside step function
activating with magnitude µ0 g at t = 0, and the effect is that LI(t) and thus the
current through the loop is as a continuous, measurable quantity.
If we use the Dirac quantization of magnetic charge, which reads
µ0 ge0 2h
= h =⇒ µ0 g = ,
2 e0
we get a numerical result for the quantity µ0 g, and a theoretically expected value
value of LI(t).

45
10.1. Skin Effect in a Ribbon-Like Conductor

10 Tenth Exercise Set

10.1 Skin Effect in a Ribbon-Like Conductor
Consider a long conducting ribbon of con-
ductivity σ, width a, height b and length l
where l ≫ b ≫ a. We place ideal electrodes
at each end of the ribbon, and connect the
electrodes to an alternating voltage source
with frequency ω. Find the dependence of
the conductor’s impedance on frequency and
investigate the high- and low-frequency limits.
Our plan is to solve for the electric field E(t)
in the ribbon, use E to find the potential
difference U , and finally use U to find the
ribbon’s impedance. We begin with the rel-
evant Maxwell equations, which are
∂B
∇×E=− ∇·E=0
∂t
∂E
∇ × B = µ0 j + ϵ0 µ0 ∇ · B = 0.
∂t
Note that ∇ · E = 0 because the conductor is neutral—no electric field escapes.
We make a quasi-static approximation to simplify the equations, neglecting the
displacement current ϵ0 µ0 ∂E
∂t to get

∇ × B ≈ µ0 j.

We take the curl of the first Maxwell equation to get

∂
∇ × (∇ × E) = ∇(∇ · E) − ∇2 E = − ∇ × B.
∂t
We then apply ∇ · E = 0 and ∇ × B = µ0 j = µ0 (σE) to get

∂
−∇2 E = − µ0 σE.
∂t

Since the ribbon is attached to an alternating voltage source, the electric field E = E(t)
reads
∂
E(t) = E0 eiωt =⇒ E(t) = iωE(t).
∂t
Finally, we substitute the ∂E
∂t into the earlier equation −∇2 E = − ∂t
∂
µ0 σE to get

∇2 E − iωµ0 σE ≡ ∇2 E − k 2 E = 0,

where we have defined the amplitude

p 1 + i√
k 2 ≡ iωµ0 σ =⇒ k = iωµ0 σ = √ ωµ0 σ.
2

46
10.1. Skin Effect in a Ribbon-Like Conductor

Next, we define coordinate system whose x axis aligns with the ribbon’s width a and
whose y axis aligns with the ribbon’s height. The conductor’s length corresponds to
the z axis. We choose the origin so that x ∈ [−a/2, a/2], meaning the ribbon’s center
occurs at x = 0.
With respect to this coordinate system, we can then simplify the Laplacian ∇2 E.
Since the ribbon’s width a is much smaller than the height and length, the Laplacian’s
derivatives with respect to y and z are negligible, i.e.
 2
∂2 ∂2 ∂2E

2 ∂
∇ E= 2
+ 2 + 2 E≈ .
∂x ∂y ∂z ∂x2
The alternating voltage is applied along the conductor’s length—along the z axis—so
the electric field reads E = Ez (x) ẑ. Plugging all of these simplifications into the
amplitude equation gives
∂2 
∇2 E − k 2 E ≈ − k 2 Ez (x) ẑ = Ez′′ (x) − k 2 Ez (x) = 0.

Ez (x) ẑ
∂x2
The solutions to the equation
Ez′′ − k 2 Ez = 0
can be written either as exponents or hyperbolic functions; we will use hyperbolic
functions, which are best suited to the problem’s reflection symmetry about the y
axis. The general solution is

Ez (x) = A cosh kx + B sinh kx.

However, the problem’s reflection symmetry means Ez will have only the even
component cosh, and the solution simplifies to

Ez (x) = A cosh kx.

We find A from the problem’s boundary conditions. Assuming Ez (±a/2) = E0 , we

have  
ka ka E0
A cosh ± = A cosh = E0 =⇒ A = .
2 2 cosh ka
2
With the constant A known, the solution for Ez (x) is then
cosh kx
Ez (x) = A cosh kx = E0 .
cosh ka
2

Next, we draw some qualitative sketches of E(x) with k as a parameter. The resulting
curves show that as k (and thus frequency ω) increases, E(x) becomes concentrated
near the ribbon’s outer surfaces x = ±a/2. This is a qualitative demonstration of the
skin effect, where electric field and current become concentrated along a conductor’s
surface at high frequencies.
Potential Difference, Current and Impedance
Next, with Ez (x) known, the potential difference across the ribbon’s is simply

Uz (x) = Ez (x)l,

47
10.1. Skin Effect in a Ribbon-Like Conductor

where l is the conductor length.

With potential difference known, we find the current through the conductor with
¨ ¨ ˆ a/2
σE0
I= j dS = (σEz ) dS = cosh(kx)(b dx),
cosh ka
2 −a/2

where we have written the surface element dS = b dx in terms of the conductor’s

height b. The integral evaluates to
ˆ a/2
σE0 b σE0 b 2 a/2
I= cosh(kx) dx = · · sinh kx
cosh ka
2 −a/2 cosh ka
2
k x=0

2σE0 b ka
= tanh .
k 2
Next, we introduce the dimensionless quantity κ = ka 2 and substitute in U0 = E0 l to
get
σU0 ba 2 ka σU0 ba tanh κ U0 tanh κ
I= tanh = = ,
l ka 2 l κ R0 κ
where, in the last equality, we have substituted in the conductor’s static resistance
l
R0 = .
σS
In terms of R0 , the conductor’s impedance is then
U0 κ
Z≡ = R0 .
I tanh κ

High and Low Frequency Limits

In the low frequency limit, k and thus κ are small (κ ≪ 1) and we expand the
tanh x ≈ function to get
κ
Z ≈ R0 = R0 .
κ
In other words, the ribbon’s impedance approaches the static resistance R0 .
Finally, for large frequencies and thus κ ≫ 1, we use the asymptotic expansion
tanh x ≈ 1 to get
r
R0 a R0 a σµ0 ω
Z ≈ R0 κ = k= (1 + i).
2 2 2
The real component Re Z corresponds to resistance R(ω), while the imaginary com-
ponent Im Z corresponds to reactance. Reactance is out of phase with resistance by
π/2. At high frequencies, the resistance is
r
r0 a σµ0 ω
R(ω) = .
2 2
√
Note the relationship R ∝ ω, meaning resistance increases with frequency, since
the electric field is more concentrated at the conductor edges (the “skin”) and the
current has a smaller effective cross section and thus larger resistance.

48
10.2. Theory: Conservation of Electromagnetic Energy

10.2 Theory: Conservation of Electromagnetic Energy

Conservation of electromagnetic energy is written as the energy balance
∂w
+ ∇ · S + j · E = 0,
∂t
where w is electromagnetic energy density
 
1 2 1 2
w= ϵ0 E + B
2 µ0

and S is the Poynting vector, defined as

1
S= E × B = E × H.
µ0

In integral form for a region of space V , the energy equation reads

˚ ‹ ˚
∂
w dV + S · dS + j · E dV = 0,
∂t V ∂V V

where we have used the divergence theorem to convert the Poynting vector term to a
surface integral.
• The w term corresponds to the changing electromagnetic field energy within
the region V
• The S term encodes energy flow (power) through the surface
• The j · E term corresponds to Ohmic energy losses within the region.

10.3 Power in a Coaxial and Cylindrical Conductor

Find the electromagnetic energy flow (power) through:
1. The cross section of a coaxial cable with inner and outer radii a and b, respec-
tively, and carrying current I, of length l, with potential difference U between
the inner and outer conductor. Neglect resistive losses.
2. The lateral surface of a long, straight cylindrical conductor of radius a, length l,
conductivity σ and carrying a current I with a potential difference U between
the cylinder ends.
Lossless Coaxial Conductor
In a coaxial cable, the magnetic field B points tangent to the circular cross section
(perpendicular to the radial direction), while the electric field E points radially
outward. We assume positive charge accumulates on the inner conductor and negative
charge on the outer conductor.
We find the magnetic field with Ampere’s law using a loop around the inner conductor.
The result is
µ0 I
µ0 I = B · 2πr =⇒ B = .
2πr

49
10.3. Power in a Coaxial and Cylindrical Conductor

Meanwhile, we find electric field with Gauss’s law, using a cylinder enclosing the
inner conductor:
Q
Q = ϵ0 E · 2πrl =⇒ E = .
2πϵ0 lr

The induced charge Q and potential difference between the inner and outer conductor
are related by
ˆ b
Q b
U= E dr = ln .
a 2πϵ0 l a
In terms of U , the electric field is thus
Q U
E= = .
2πϵ0 lr r ln ab

The Poynting vector points in the direction E × B, along the conductor’s longitudinal
axis. In our case we have E ⊥ B, so the Poynting vector magnitude S is
  !
1 1 µ0 I U UI
S= EB = b
= .
µ0 µ0 2πr r ln a 2π ln ab r2

The electromagnetic power through the cross section is thus

¨ ¨ ˆ b !
UI
P = S · dS = S dS = (2πr dr)
a 2π ln ab r2
ˆ b
UI dr UI b
= b
= b
ln = U I,
ln a a r ln a a
where S · dS = S dS because S is parallel to coaxial cable’s cross section.
Cylindrical Conductor
As for the coaxial cable, the cylindrical conductor’s magnetic field is tangent to the
conductor’s circular cross section. Meanwhile, the electric field points in the direction
of the potential difference—along the conductor’s longitudinal axis. The Poynting
vector S = E × B thus along the radial direction. We choose I to flow so that S
points radially inward towards the conductor’s center (as opposed to radially outward
for an opposite current).
The electric field in the conductor, which arises from the potential difference U
between the conductor’s ends, is simply
U
E= ,
l
where l is the conductor’s length.
We find the conductor’s magnetic field with Ampere’s law, using a circular loop
centered along the conductor’s longitudinal axis. Assuming current is uniformly
distributed across the cross section, the magnetic field is
 r 2 µ0 I
B · (2πr) = µ0 I =⇒ B = r.
a 2πa2

50
10.4. Cylindrical Conductor with a Slit

Since E and B is perpendicular, the Poynting vector is

  
1 1 U µ0 I UI
S= EB = r = r.
µ0 µ0 L 2πa2 2πa2 l

Note that S increases linearly with r.

We find electromagnetic power by integrating S over the cylinder’s lateral surface
(not the cross section!). The power through a lateral surface at radius r is
¨ ¨
U Ir2
 
UI
P (r) = S · dS = S dS = r (2πrl) = .
Slat Slat 2πa2 l a2

The power through the entire conductor’s lateral surface occurs at r = a, which
produces the familiar result
U Ia2
P = = U I.
a2
Next, we consider the conductor’s energy balance
˚ ‹ ˚ ˚
∂
− w dV = S · dS + j · E dV = P + j · E dV.
∂t V ∂V V V

∂t = 0, and thus
Because the situation is stationary we have ∂w
˚
P+ j · E dV ≡ P + Ploss = 0 =⇒ Ploss = −P = −U I.
V

Note that Ohmic losses amount to Ploss = −U I, which is the familiar expression from
e.g. high school physics. The energy balance reads
˚
P+ j · E dV = P + Ploss = U I − U I = 0,

and energy is conserved, at it must be.

10.4 Cylindrical Conductor with a Slit

Consider a long cylindrical conductor of radius a and
cross-sectional area S carrying a current I. We cut
a narrow slit of width d ≪ a through the conductor
in a plane parallel to the circular cross section. De-
termine the magnitude and direction of the electric
and magnetic fields in the slit, then calculate the en-
ergy flow through the slit’s lateral surface and confirm
the validity of Poynting’s theorem for conservation of
electromagnetic energy.
Because of the current I, positive charge accumulates
on one surface of the slit and negative charge on the
other. Because d ≪ a, we can thus treat the two
surfaces of the slit as parallel-plate capacitor.

51
10.4. Cylindrical Conductor with a Slit

The charge accumulating on the slit is q(t) = It, and, using the model of a parallel-
plate capacitor, the corresponding electric field is

σ(t) 1 q(t) I
E(t) = = = t.
ϵ0 ϵ0 S ϵ0 S

Next, we find the magnetic field in the slit. The time-varying changing electric field
E(t) creates displacement current, and the Maxwell equation for ∇ × B reads

∂E ∂E
∇ × B = µ0 j + ϵ0 µ0 = ϵ0 µ0 .
∂t ∂t
Note that µ0 j = 0 inside the slit, since there’s no free current. We integrate the
equation over the conductor’s cross-sectional surface S to get
¨ ˛ ¨
∂ ∂E
∇ × B · dS = B · dl = ϵ0 µ0 E · dS = ϵ0 µ0 · (πr2 ).
S ∂t S ∂t

We evaluate the line integral over the surface’s boundary and substitute in electric
field to get
˛
∂E(t) µ0 I µ0 I
B · dl = B · (2πr) = ϵ0 µ0 =⇒ B(r) = r= ,
∂t 2S 2πr

where S = πr2 is the cross-sectional surface area. The magnetic field points tangent
to the cylindrical lateral, just like the magnetic field of a conductor, except that free
current is replaced by displacement current.
Since E points along the conductor’s longitudinal axis and B is tangent to the lateral
surface, S = E × B points along the radial direction. Because B ⊥ S, the power
through the lateral surface is simply
¨ ¨ ¨   
1 1 I µ0 I
P = S dS = EB dS = t r dS.
Slat µ0 Slat µ0 Slat ϵ0 S 2S

We set r = a at the lateral surface to get

I 2a I 2d
P = t(2πad) = t,
2ϵ0 S 2 ϵ0 S

where S = πa2 is cross sectional area at r = a.

Finally we consider the conductor’s energy balance. If we neglect Ohmic losses, the
energy balance reads
˚ ‹
∂ ∂WEM
w dV + S · dS = + P = 0,
∂t V ∂V ∂t

where WEM = WE + WB is the sum of magnetic and electric field energy. Note that
B = B(r) and thus magnetic energy WB is independent of time. It follows that

∂WEM ∂WE
= + 0.
∂t ∂t

52
10.4. Cylindrical Conductor with a Slit

Using WE = wE V , the electric field energy in the slit is

ϵ0 I 2 I 2d 2
 
1 2 2
W E = wE V = ϵ0 E · (Sd) = Sdt = t .
2 2 ϵ20 S 2 2ϵ0 S

In terms of WE , the time derivative of total magnetic field energy is thus

∂WEM ∂WE ∂ I 2d 2 I 2d
= = t = t = P.
∂t ∂t ∂t 2ϵ0 S ϵ0 S

Up to a negative sign (depending on the direction of current), ∂WEM

∂t and P are equal,
indicating that energy is conserved, as it must be.

53
11.1. A Radially Polarized Sphere

11 Eleventh Exercise Set

11.1 A Radially Polarized Sphere
Consider a radially polarized sphere of radius a in which electric polarization points
in the radial direction and grows with increasing radius as P(r) = kr where k is a
constant. Find the volume density of bound charges, the surface density of bound
charges, and the total bound charge. Finally, find the electric field due to the polarized
sphere.
Polarization and bound charge density ρb are related by

ρb = −∇ · P,

while electric field and total charge density ρ are related by Gauss’s law

ρ = ϵ0 ∇ · E.

We find volume charge density with ρb = −∇ · P and the known expression for P:

ρb = −∇ · P = −∇ · (kr) = −k∇ · r = −3k,

where we’ve used ∇ · r = 3. Note the bound volume charge density is negative,
since the electric dipoles in the sphere have their positive poles radially outward and
negative pole radially inward, so negative charge is concentrated toward the sphere’s
center.
To relate surface and volume charge density, we integrate ρb = −∇ · P over the
sphere’s volume and apply the divergence theorem
˚ ˚ ¨
out
ρb dV = qb = − ∇ · P dV = − P · dS = −P · S in ,
V V S

where the last step writes S = n̂S and notes that polarization is zero outside the
sphere. The bound surface charge density is then
qb
qb = P · n̂S =⇒ σB = = P · n̂.
S
At the surface r = a, charge density σb evaluates to

σb = P · n̂ = (kr) · n̂ = kr r=a
= ka.

Note that surface charge density is positive, since the electric dipoles have their
positive pole oriented radially outward.
We find total bound charge from ρb and σb by integrating both volume and surface
charge densities:
˚ ¨
4
qb = ρb dV + σb dS = −3k πa3 + ka4πa2 = 0.
V S 3
We should expect total charge to be zero, since all dipoles within the sphere should
cancel out.

54
11.2. A Halved Polarized Sphere

Finding the Electric Field

By Gauss’s law, the electric field outside the sphere is zero, since the sphere’s total
charge is zero.
Inside the sphere, using the problem’s spherical symmetry, we use Gauss’s law with a
spherical surface to get
4 ρb
ϵ0 E4πr2 = ρb V = ρb πr3 =⇒ E(r) = r.
3 3ϵ0
The electric field points radially outwards:
ρb
E(r) = r.
3ϵ0
We then substitute in the charge density ρb = −3k to get
−k P
E= r =⇒ E = − .
ϵ0 ϵ0
Note the linear relationship between electric field and polarization.
The linear relationship arises because the sphere contains only bound charge, meaning
total charge and bound charge are equal, i.e. ρ = ρb . We could then combine
ρb = ρ = −∇ · P and Gauss’s law ρ = ∇ · (ϵ0 E) to get E = − ϵP0 .

11.2 A Halved Polarized Sphere

Consider a sphere of radius a with homogeneous (uniform) polarization P pointing in
the z direction. Find the electric field inside and outside the sphere. We then cut the
sphere in half in a plane perpendicular to the direction of polarization and slightly
separate the halves to form a capacitor. Find the electric field in the slit between the
halves.
First, we find volume density of bound charges, which is

ρb = −∇ · P = 0,

since the divergence of P, which is homogeneous, is zero by definition. In other words,

there are no bound charges in the sphere’s volume.
Next, we find bound surface charge density σb = P · n̂ where n̂ is the normal to the
surface. In terms of an θ between the normal n̂ and polarization P, surface charge
density is
σb = P · n̂ = P cos θ.

We proceed with the Poisson equation, which simplifies to the Laplace equation
within the sphere were ρb = 0.
ρb
∇2 U (r) = − = 0.
ϵ0
We account for the non-zero surface charge density with boundary conditions.

55
11.2. A Halved Polarized Sphere

The general solution for U (r) is

∞
X
U (r) = U (r, θ) = (Al rl + Bl r−(l+1) )Pl (cos θ).
l=0

Since the surface charge distribution σb depends only on cos θ, our solution can
include only cos θ-dependent terms, which occurs for l = 1. The general solution
simplifies to
U (r) = (A1 r + B1 r−2 ) cos θ.
To avoid divergence at r → 0 and r → ∞, we separate U (r) according to
(
Uin = A1 r cos θ r<a
U (r) ≡ −2
Uout = B1 r cos θ r > a.
We find A1 and B1 with boundary conditions.
First boundary condition: the electric potential must be continuous at the boundary
(a discontinuous potential would imply infinite charge, which is nonphysical). The
continuity condition at r = a requires
B1
A1 a = =⇒ B1 = A1 a3 .
a2

For the second boundary condition we require electric field is perpendicular to the
surface. We then apply Gauss’s law near the sphere’s surface to get
⊥ ⊥ ⊥ ⊥
qb = ϵ0 S(Eout − Ein ) =⇒ σb = ϵ0 (Eout − Ein ).
∂Uin/out
We know σb = P cos θ and use Ein/out
⊥ =− ∂r r=a
to get
 
⊥ ⊥ 2B1
σb = P cos θ = ϵ0 (Eout − Ein ) = ϵ0 + A1 cos θ.
a3
We substitute the first boundary condition B1 = A1 a3 into the second to get
P cos θ = ϵ0 (2A1 + A1 ) cos θ =⇒ P = 3ϵ0 A1
and thus
P P a3
A1 = and B1 = .
3ϵ0 3ϵ0
The electric potential is then
P

Uin = 3ϵ r cos θ r<a


0
U (r) ≡
 P a3 1
Uout =
 cos θ r > a.
3ϵ0 r2

We now discuss the solution. The 3 in the denominators is called the depolarization
factor. Next, we note that r cos θ = z—since Uin depends only on z, the field inside
the sphere is homogeneous:
∂Uin P P
Ez = − =− =⇒ E = − ,
∂z 3ϵ0 3ϵ0

56
11.3. Theory: Dielectric and Displacement Field

where we note that P points in the z direction. The field is negative because with our
choice of polarization, positive charges are in the positive z direction and negative
charges in the negative z direction.
The field outside the sphere is field of an electric dipole; the cosine term is analogous
to the dipole dot product term pe · r. A dipole field is expected, since the sphere is
polarized like an electric dipole.
Halved Sphere; Electric Field in the Slit
We now consider the electric field in the slit when the sphere is cut in half in a plane
perpendicular to the polarization.
Because of the slit through the middle, bound charges accumulates on the cut surfaces
as well as the outer surface. Since polarization points “upwards” in the positive z
direction, the bound charges on cut surface of the upper hemisphere are negative,
while the bound charges on the cut surface of the lower sphere are positive. The
relationship σb = P · n̂ is preserved.
To find electric field because of the slit, we model the slit as a parallel-plate capacitor,
for which the electric field reads
σ̃b (P · n̂) P
Ẽ = n̂ = n̂ = ,
ϵ0 ϵ0 ϵ0
where tilde corresponds to the flat surface and n̂ is the normal to the flat surface.
We must also consider the electric field contribution from the bound charge on the
hemispherical surfaces. Reusing the already derived electric field from the bound
charges on the spherical surfaces, which read
P
Esphere = − ,
3ϵ0
the total field in the slit is
P P 2P
E = Ẽ + Esphere = − = .
ϵ0 3ϵ0 3ϵ0

11.3 Theory: Dielectric and Displacement Field

The previous two problems involved ferro-electric materials in which the polarization
was built in to the material, even in the absence of an external electric field. We now
consider dielectric materials, in which polarization occurs only in the presence of an
external field.
Recall the equations ρb = −∇ · P and Gauss’s law ρ = ϵ0 ∇ · E. We subtract the
equations to get
ρf = ρ − ρb = ∇ · (ϵ0 E + P) ≡ ∇ · D,
where the free charge density ρf = ρ − ρb is difference between the total and bound
charge densities. The electric displacement field is defined as

D = ϵ0 E + P.

57
11.4. Parallel-Plate Capacitor with an Anisotropic Dielectric

The D field arises from free charge and is useful when analyzing dielectrics.
For small fields, we make the approximation D ∝ E where D and E are linearly
dependent, which results in the approximate linear relationship

D = ϵ0 χE,

where χ is the permittivity tensor.

11.4 Parallel-Plate Capacitor with an Anisotropic Dielectric

Consider a parallel-plate capacitor with plate separation d, plate surface area S, and
the intra-plate space filled with a dielectric insulator whose dielectric tensor has
components ϵ1 , ϵ2 and ϵ2 . The principle axes corresponding to ϵ1 and ϵ3 are parallel
to the capacitor plates, but the principle axis corresponding to ϵ2 makes an angle ϕ
with the normal to the plates. Find the capacitor’s capacitance C.
First, recall that capacitance is in general defined as
q
C= ,
U
where q is the charge on the capacitor plates and U is the potential difference between
the plates.
Because χ is angled (equivalently, because χ’s second principle axis does not align
with the normal to the capacitor plates), E and D are not parallel.
From the boundary conditions for Maxwell’s equations, we know E must be perpen-
dicular to the capacitor plates. To show this, for review, we consider the Maxwell
equation
∂B
∇×E=− .
∂t
In our static case we have ∂B
∂t = 0. We integrate the equation over a surface hugging
the boundary and apply Stokes’ theorem to get
¨ ˛
∇ × E · ∇ · S = E · dl = 0,
S

where the line integral runs over a rectangular path thinly hugging a capacitor plate.
Because the path thinly hugs the plate, the integral registers only the component of
E parallel to the plates:
(∥) (∥) (∥) (∥)
Eout · l − Ein · l = 0 =⇒ Eout = Ein .

Because the electric field outside the capacitor is zero, we have

(∥) (∥)
Eout = 0 =⇒ Ein = 0.
(∥)
The result Ein = 0 implies the electric field has only a component perpendicular to
the capacitor plates.

58
11.4. Parallel-Plate Capacitor with an Anisotropic Dielectric

Working in the two-dimensional x, y plane (since the problem is invariant to translation

in the z direction along the plates) the dielectric tensor in the principle axis system,
which we denote χ̃, reads  
ϵ1 0
χ̃ = .
0 ϵ2
The capacitor’s coordinate system is rotated by ϕ relative to the principle axis system.
To transform from χ̃ to χ, we rotate the principle axes tensor by rotation matrices:
   
cos ϕ sin ϕ ϵ1 0 cos ϕ − sin ϕ
χ = Rχ̃R =
− sin ϕ cos ϕ 0 ϵ2 sin ϕ cos ϕ
2 2
 
ϵ1 cos ϕ + ϵ2 sin ϕ (ϵ2 − ϵ1 ) sin ϕ cos ϕ
= .
(ϵ2 − ϵ1 ) sin ϕ cos ϕ ϵ1 sin2 ϕ + ϵ2 cos2 ϕ
Note that χ is symmetric, as is expected for the dielectric tensor.
Because of the boundary condition, E∥ = 0, if follows that E has only a y component
(where the y axis is normal to the plates). The electric field then reads E = (Ex , Ey ) ≡
(0, E), which we combine with D = ϵ0 χE to get

ϵ1 cos2 ϕ + ϵ2 sin2 ϕ (ϵ2 − ϵ1 ) sin ϕ cos ϕ

    
Dx 0
= ϵ0 .
Dy (ϵ2 − ϵ1 ) sin ϕ cos ϕ ϵ1 sin2 ϕ + ϵ2 cos2 ϕ E

In terms of E, the electric field between the capacitor plates is simply

U = Ed,

where d is the distance between the plates.

Next, we use the boundary condition on the D field to find the charge q on the plates.
We derive the relevant boundary condition, we apply Gauss’s law to a thin region
tightly enclosing a capacitor plate. For the D field, which obeys ∇ · D = ρf , Gauss’s
law reads ˚ ¨ h i
(⊥) (⊥)
ρf dV = D · dS =⇒ qf = Dout − Din S.
V S
(⊥) (⊥)
In our case, where Dout= 0 and Dout = Dy the charge on the capacitor plates (up
to a minus sign depending on the definition of the surface normal) is

qf = Dy S.

We find Dy from the earlier matrix equation:

U
Dy = ϵ0 ϵ1 sin2 ϕ + ϵ2 cos2 ϕ E = ϵ0 ϵ1 sin2 ϕ + ϵ2 cos2 ϕ


.
d
The charge on the capacitor plates is then
ϵ0 SU
ϵ1 sin2 ϕ + ϵ2 cos2 ϕ .


qf = D y S =
d

In terms of qf , the capacitor’s capacitance C is

qf ϵ0 S
ϵ1 sin2 ϕ + ϵ2 cos2 ϕ = C0 ϵ1 sin2 ϕ + ϵ2 cos2 ϕ ,



C= =
U d

59
11.4. Parallel-Plate Capacitor with an Anisotropic Dielectric

where C0 = ϵ0 S
d is the capacitance of an empty capacitor.
Note that for ϕ = 0, corresponding to an isotropic dielectric in which the dielectric
tensor’s second principle axes does align with normal to the capacitor plates, the
capacitor’s capacitance reduces to C = C0 ϵ2 .

60
12.1. Point Dipole in a Spherical Dielectric Cavity

12 Twelfth Exercise Set

12.1 Point Dipole in a Spherical Dielectric Cavity
Consider an empty spherical cavity of radius a within a vast dielectric material with
permittivity ϵ. We place an electric dipole with dipole momentum pe in the cavity.
What is the resulting electric potential inside and outside the cavity? Find the resulting
effective dipole moment p′e in the dielectric.
We choose our coordinate system so that the dipole pe within the cavity points in
the z direction.
We find the electric potential due to the dipole by solving the Laplace equation

∇2 U (r) = 0.

In spherical coordinates, which are best suited to the problem’s spherical geometry,
the general solution is
∞
X
U (r, θ) = (Al rl + Bl r−(l+1) )Pl (cos θ),
l=0

where Pl are the Legendre polynomials.

We then consider boundary conditions along the spherical cavity’s surface and at
the center of the cavity, where the dipole pe is placed. The potential at the sphere’s
center should approach the potential of an electric dipole:
pe cos θ
U (r → 0, θ) = Udipole = .
4πϵ0 r2
Because this solution contains only cos θ to the first power, only the l = 1 term
containing cos θ can occur in the general solution for U , simplifying the solution to

U (r, θ) = (A1 r + B1 r−2 ) cos θ.

For the potential to converge at r → ∞, the A1 r term must vanish at large r (outside
the cavity). The general solution further simplifies to
(
Udipole + A1 r cos θ r < a
U (r) =
B1 r−2 cos θ r>a
p cos θ

 e

2
+ A1 r cos θ r < a
= 4πϵ0 r
 B1 cos θ

r > a.
r2

Next, we require U is continuous at the cavity’s boundary r = a, which produces

pe B1
U (r → a+ ) = U (r → a− ) =⇒ 2
+ A1 a = 2 .
4πϵ0 a a
Note: We could derive the same result using the boundary conditions for Maxwell’s
equations in materials, among which is required that the component of electric field

61
12.1. Point Dipole in a Spherical Dielectric Cavity

E∥ tangent to the boundary surface must be equal on both sides of the boundary.
The tangential component of electric field is found with
1 ∂U
E∥ = − ,
r ∂θ r=a
∥ ∥
and the boundary condition Ein = Eout then implies
pe B1
− 2
sin θ − A1 a sin θ = 2 sin θ,
4πϵ0 a a
which is the same result we arrived at from requiring continuity of electric potential
U at the boundary.
Finally, we apply the boundary condition on the D field, which applies to the
components D⊥ perpendicular to the boundary surface and reads
⊥ ⊥
Din − Dout = σf ,
where σf is the surface density of free charges along the boundary. We find D⊥ with
∂U
D⊥ = −ϵϵ0 ,.
∂r r=a

In our case, which involves a dielectric with only bound charges, we have σf , and the
boundary condition on D reduces to

⊥ ⊥ ∂Uin ∂Uout
Din = Dout ⇐⇒ ϵ0 ϵin = ϵ0 ϵout .
∂r r=a ∂r r=a

The dielectric constant inside the empty cavity is ϵin = 1, and we denote ϵout ≡ ϵ
in the dielectric. We then substitute Uin and Uout into the boundary condition and
simplify to get
pe B1
2 3
− A1 = 2ϵ 3 .
4πϵ0 a a

We then add the earlier boundary condition requiring continuity of U , which read
pe B1
3
+ A1 = 3 ,
4πϵ0 a a
to the just-derived boundary condition on D⊥ to get
pe B1 3 pe
3 = 3 (1 + 2ϵ) =⇒ B1 = .
4πϵ0 a3 a 1 + 2ϵ 4πϵ0
With B1 known, we can find A1 according to
 
B1 pe pe 3 2(1 − ϵ) pe
A1 = 3 − 3
= 3
−1 = .
a 4πϵ0 a 4πϵ0 a 1 + 2ϵ 1 + 2ϵ 4πϵ0 a3

With the coefficients A1 and B1 known, the solution for U (r, θ) is then
1 2(ϵ − 1) r

pe  2−
 r<a
U (r, θ) = cos θ r 1 + 2ϵ a3
4πϵ0  3 1

r > a.
1 + 2ϵ r2

62
12.2. Dielectric Constant of Cold Plasma

Inside the cavity where r < a, the first 1/r2 term is the potential of the electric dipole
pe , while the term containing r cos θ ≡ z corresponds to a homogeneous field inside
the cavity.
Outside the cavity where r > a, we have another 1/r2 term—again like a dipole, but
with an additional factor 1+2ϵ
3
, which would reduce to the usual 1 for a dielectric
constant ϵ = 1. This is the so-called effective dipole term in the problem instructions.
Outside the cavity we thus have the effective dipole moment
3
p′e = pe .
1 + 2ϵ

To find the surface density of bound charges, we have two options. We could could
use Gauss’s law in the form

⊥ ⊥
 ∂U
σb = ϵ0 Eout − Ein , E⊥ = − .
∂r r=a

Alternatively, we could use the relationship between surface charge density and
polarization, which reads
σb = P · n̂.
Note that the surface’s normal vector n̂ points from the dielectric material’s surface
into the cavity, while the polarization P points out of the cavity. The product P · n̂
thus evaluates to
P · n̂ ≡ −P⊥ ,
where P⊥ is the component of polarization normal to the cavity’s surface. Assuming
a linear relationship between D and E, we then find polarization with

ϵ0 ϵE ≈ D = ϵ0 E + P =⇒ P = ϵ0 (ϵ − 1)E.

We then use P = ϵ0 (ϵ − 1)E to find surface charge density via

σb = P · n̂ = −P⊥ = −ϵ0 (ϵ − 1)E⊥ r=a

.

12.2 Dielectric Constant of Cold Plasma

Find the relative permittivity ϵ of a cold plasma, then find the dispersion relation for
electromagnetic waves in the plasma.
Plasma is a gas of ions, i.e. a gas of free electrons that have been ionized from their
positive nuclear cores. We assume the plasma is cold to neglect thermal motion.
Since the nuclei are much more massive than electrons, we assume the atoms are at
rest as a first approximation.
We then ask how electromagnetic waves propagates through the plasma. We write
the electromagnetic waves in terms of the electric field, which we model with the
oscillatory ansatz
E(r, t) = E0 ei(kz−ωt) .

63
12.2. Dielectric Constant of Cold Plasma

Assuming the positive ion cores have charge q and electrons have charge −q, the force
on a representative free electron of mass m is

F = mr̈ = −qE = −qE0 ei(kz−ωt) ,

where we have introduced the coordinate r to measure the electron’s displacement.

Since force is oscillatory, the resulting motion is modeled by

r = r0 ei(kz−ωt) .

Substituting the expressions for r and E into Newton’s law mr̈ = −qE gives

−mω 2 r0 ei(kz−ωt) = −qE0 ei(kz−ωt) ,

from which we can find the position’s oscillation amplitude r0 via

q
mω 2 r0 = qE0 =⇒ r0 = E0 .
mω 2

The displacement r0 of the negative electron from the positive ion core creates a
dipole moment with amplitude

q2
pe0 = −qr0 = − E0 .
mω 2
In terms of pe0 , the polarization of the plasma is then

nq 2
P0 = npe0 = − E0 ,
mω 2
where n is the number density of electric dipoles in the material.
We then use the relationship P, E and permittivity, which reads

P0 = ϵ0 (ϵ − 1)E0 ,

to find the plasma’s dielectric constant (relative permittivity)

P0 nq 2
ϵ=1+ =1− .
ϵ0 E0 mϵ0 ω 2

nq 2
Next, we note that mϵ0 has units of frequency, and define the plasma frequency

nq 2
ωp2 = ,
mϵ0
which describes the frequency at which displaced electrons oscillated about the positive
ion cores. In terms of plasma frequency, the plasma’s dielectric constant is

ωp2
ϵ=1− .
ω2
Note that for electromagnetic wave frequencies ω > ωp we have ϵ > 0 and for ω < ωp
we have ϵ < 0. However, electromagnetic waves cannot propagate through a material

64
12.2. Dielectric Constant of Cold Plasma

with dielectric constant ϵ < 0, which we will discuss more shortly. Dispersion
Relation for EM Waves in Plasma
Next, we will find the dispersion relation for the electromagnetic waves in the plasma.
In material with zero electric current density (i.e. j = 0) the Maxwell equations read

∇·E=0 ∇ × E = −µ0 ∂H∂t .

∇·H=0 ∇ × H = ϵ0 ϵ ∂E
∂t

We take the curl of the equations for ∇ · E and ∇ × E to get

 
∂ ∂E
∇ × (∇ × E) = −µ0 ϵ0 ϵ
∂t ∂t
∂ E2
∇(∇ · E) − ∇2 E = −µ0 ϵϵ0 2 .
∂t

The relationship ∇ · E = 0 simplifies the second equation to the wave equation

∂2E
∇2 E − µ0 ϵ0 ϵ = 0.
∂t2
We then substitute in the oscillatory electric field ansatz E = E0 ei(kz−ωt) to get

−k 2 E − µ0 ϵ0 ϵ(−ω 2 )E = 0 =⇒ E(µ0 ϵ0 ϵω 2 − k 2 ) = 0.

The equality E(µ0 ϵ0 ϵω 2 − k 2 ) = 0 holds for all E only if

ω2 1 1 c20
µ0 ϵ0 ϵω 2 − k 2 = 0 =⇒ = = .
k2 ϵ0 µ0 ϵ ϵ
ωp2
Next, we substitute in the plasma’s dielectric constant ϵ = 1 − ω2
to get
!
2
ω 2 ω 2 ωp ω 2 ωp2
c20 = ϵ 2 = 2 1 − 2 = 2 − 2 .
k k ω k k

Finally, we rearrange the relationship between c0 , k and ω to get the dispersion

relation q
k 2 c20 = ω 2 − ωp2 =⇒ ω = ωp2 + c20 k 2 .
This is a dispersion relation for electromagnetic waves in the plasma—it relates the
waves’ frequency ω and wave vector k. Note that the dispersion relation prohibits
ω < ωp since at the minimum possible k, i.e. k = 0, frequency is ω = ωp . The
prohibited region of ω < ωp corresponds to ϵ < 0—this relationship quantifies the
previous claim that EM waves cannot propagate through materials with ϵ < 0.
Note that in a vacuum, with ϵ = 1, EM waves obey the linear dispersion relation
ω2 c20
= = c20 =⇒ ω = c0 k.
k2 ϵ

We now consider the limiting cases for the dispersion relation in plasma, i.e.
q
ω = ωp2 + c20 k 2 .

65
12.2. Dielectric Constant of Cold Plasma

For large k, such that c0 k ≫ ωp , we approach the linear relation ω = c0 k. Meanwhile,

very small k, such that c0 k ≪ ωp , result in the constant dispersion relation ω →
ωp . We get a more accurate approximation for small k with the first-order Taylor
approximation
1/2
c20 2 c20 2
    
wp
ω = ωp 1 + 2 k ≈ ωp 1 + 2 k , k≪ .
ωp 2ωp c0

As a side note, we remark that in this limit of small k, where the dispersion relation
obeys ω ∼ k 2 . A quadratic dispersion relation means the photons making up the
electric field behave like particles with mass, which leads to the concept of effective
mass, like in solid state physics.
Further Discussion
Next, we more thoroughly consider the regime of ϵ < 0. Assuming ϵ < 0, we
rearranging the equation

ω2 c20 ω√ ω √
= to get k= ϵ = i ϵR ,
k2 ϵ c0 c0
where k is imaginary because ϵ < 0. Substituting this expression for k into the
oscillator ansatz for the electric field E gives
√
− cω ϵR z −iωt
E = E0 ei(kz−ωt) = E0 e 0 e .

The position-dependent term is proportional to e−z , which means that an electro-

magnetic wave will exponentially decay with position in a material with ϵ < 0.
Next, we use the dispersion relation to find the group and phase velocity of the
electromagnetic waves in the plasma. The phase velocity is
q s
ω 2 + c2 k 2
ωp p 0 ωp2
vphase = = = c20 + 2 .
k k k

Note that vphase > c0 —the phase velocity is greater than the speed of light in vacuum.
The group velocity—the quantity relevant to the universal speed limit of c0 —is

∂ω 2c2 k c0
vgroup = = q 0 =r .
∂k 2 ωp2 + c20 k 2 ωp2
1 + c2 k 2
0

As expected, vgroup < c0 , in agreement with special relativity. As a final comment,

note that vphase vgroup = c20 .

66
13.1. Theory: Electromagnetic Wave Propagation in Waveguides

13 Thirteenth Exercise Set

13.1 Theory: Electromagnetic Wave Propagation in Waveguides
Consider an empty waveguide (i.e. with relative permittivity ϵ = 1). The wave
equation for EM waves reads

1 ∂2E
∇2 E − = 0,
c20 ∂t2

where c0 is the speed of light in a vacuum. As a simplification, we assume the

waveguide has a bounded, pipe-like waveguide, with the longitudinal axis coinciding
with the ẑ axis, so that EM waves are restricted to propagation in the ẑ direction. In
this case, the electric field takes the form

E(r, t) = E(ρ)ei(kz−ωt) ,

where ρ is the radial position within the waveguide’s planar cross section, i.e. the
(x, y) plane perpendicular to the direction of EM propagation along ẑ, k is the wave
vector corresponding to wave propagation in the ẑ direction, and ω is the wave’s
frequency. With respect to this ansatz, the z and t derivatives produce

∂ ∂2 ∂ ∂2
= ik =⇒ = −k 2 and = −iω =⇒ = −ω 2 .
∂z ∂z 2 ∂t ∂t2
Using the expression for ∂z ,
∂
we then write the Laplacian in the separated form

∂2
∇2 = + ∇2⊥ = −k 2 + ∇2⊥ ,
∂z 2
2 2
where ∇2⊥ = ∂x∂
2 + ∂y 2 corresponds to differentiation in the (x, y) plane, i.e. the
∂

waveguide’s cross section. The wave equation in a waveguide then simplifies to

  2 
2 ω 2
∇⊥ + −k E(ρ, t) = 0.
c20

Although the electric field above is E = E(ρ), we can generalize this to

  2 
2 ω 2
∇⊥ + −k E(r, t) = 0.
c20

We can include the ei(kz−ωt) , whose only variables are z and t, because the above
wave equation doesn’t contain ∂z∂
or ∂t
∂
derivatives (but only ∂x
∂
and ∂y
∂
derivatives
because of ∇⊥ ).
2

Since E ∈ R3 has three components, the above wave equation in vector form really
has 3 equations for each of E’s three components.
Similarly, we could derive an analogous vector wave equation for magnetic field H:
  2 
2 ω 2
∇⊥ + −k H(r, t) = 0.
c20

67
13.1. Theory: Electromagnetic Wave Propagation in Waveguides

This vector equation also contain three equations for each of H’s three components.
As a result, the electromagnetic waves’ propagation in the vacuum is nominally
described by 3 + 3 = 6 equations. However, it turns out that we can express multiple
H in terms of E, which simplifies the equations involved. Our goal in the coming
page will be to express Ex , Ey , Hx and Hy in terms of only Ez and Hz .
Relationship Between E and H
We begin with the Maxwell equations
∂H ∂E
∇ × E = −µ0 and ∇ × H = ϵ0 .
∂t ∂t
In component form, the equation for ∇ × E reads
 ∂
    ∂E   
∂y − ikEy
z
∂x Ex Hx
∂ ∂E
× Ey  = ikEx − ∂xz  = iµ0 ω Hy  ,
     
 ∂y 
∂ ∂Ey ∂Ex
∂z Ez ∂x − ∂y Hz

where we have used ∂

∂z = ik and ∂
∂t = −iω. The second equation for ∇ × H reads
 ∂
    ∂H   
∂y − ikHy
z
∂x Hx Ex
∂ ∂H
× Hy  = ikHx − ∂x  = −iϵ0 ω Ey  .
z
     
 ∂y 
∂ ∂Hy ∂Hx
∂z Hz ∂x − ∂y Ez

The first and fifth equations both contain Hx and Ey . They read

∂Ez ∂Hz
− ikEy = iµ0 ωHx and ikHx − = −iϵ0 ωEy .
∂y ∂x
Our next step is to eliminate Hx : we multiply the first equation by k, the second by
µ0 ω, and subtract the two equations to get
 
∂Ez 2 ∂Hz
k − ik Ey = µ0 ω − iϵ0 ωEy .
∂y ∂x

We then rearrange to get

ω2
 
∂Ez ∂Hz 2 2

2
k − µ0 ω = iEy k − ω µ0 ϵ0 = iEy k − 2 ,
∂y ∂x c

and finally solve for Ey in terms of the z componentsEz and Hz :

k ∂E ∂Hz
∂y − µ0 ω ∂x
z

Ey = i ω2
.
c2
− k2

We could perform an analogous procedure the remaining equations to solve for the
remaining components Ex , Hx and Hy in terms of Ez and Hz . Without derivation,

68
13.2. A Parallel-Plate Waveguide

the results are

k ∂H ∂Ez
∂x − ωϵ0 ∂y
z

Hx = i ω2
c2
− k2
k ∂E ∂Hz
∂x + ωµ0 ∂y
z

Ex = i ω2
c2
− k2
k ∂H ∂Ez
∂y + ωϵ0 ∂x
z

Hy = i ω2
.
c2
− k2

Note that these equations apply only in Cartesian coordinates for a pipe-like waveguide
where waves propagate along the ẑ direction.
Next, we consider boundary conditions at the waveguide’s walls. We begin with the
Maxwell equations
∂H
∇ × E = −µ0 and ∇ · H = 0,
∂t
and encircle a small portion of the waveguide boundary wall with a hypothetical
closed loop. We then send to loop to zero thickness while applying Stokes’ law, which
results in
∇ × E → 0 =⇒ E∥ = 0 and H⊥ → 0.
In other words, the component of electric field parallel to the waveguide boundary
and the component of magnetic field normal to the waveguide boundary are zero.
Finally, we find Ez and Hz from the waveguide wave equations
  2   
w Ez
∇2⊥ + − k 2
= 0.
c2 Hz

By convention, since the wave equation is a second-order order equation, we consider

two linear independent solutions with orthogonal components. These are:
• Transverse electric (TE) wave propagation with Ez = 0 and Hz ̸= 0. This is
propogation mode is called transverse electric because, from Ez = 0, the electric
field is perpendicular (transverse) to the direction of EM wave propagation
along the ẑ axis.
• Transverse magnetic (TM) wave propagation with Hz = 0 and Ez ̸= 0. The first
propogation mode is called transverse magnetic (TM) because the magnetic
field is perpendicular (transverse) to the direction of EM wave propagation
(since Hz = 0).
The general solution for EM waves in the waveguide is then a linear combination of
the two solutions.

13.2 A Parallel-Plate Waveguide

Consider a waveguide consisting to two large parallel conducting plates separated by the
distance a, in which waves propagate along the z axis. For both transverse magnetic
and transverse electric wave propagation, find:

69
13.2. A Parallel-Plate Waveguide

1. The solution for the electric field in the waveguide

2. The dispersion relation relating wave vector and wave frequency
3. The wave impedance Z
4. The ratio of electric field amplitudes Ex to Ez

Transverse Magnetic Solution

First, we define our coordinate system, which is easiest to visualize in terms of two
horizontal, parallel plates drawn on a piece of paper.
We choose the direction of wave propagation z to run from left to right along the
paper; the x axis, which runs from bottom to top in the paper, separates the two
plates, and the y axis points out of the page. Because of translational symmetry
along the y axis (i.e. because the plates are “large”), their is no change in physical
quantities with respect to y, so ∂y
∂
= 0.
In transverse magnetic (TM) wave propagation, we have Hz = 0 and Ez ̸= 0. From
the introductory theory section, recall the equations
k ∂E ∂Hz
∂x + ωµ0 ∂y
z
k ∂E ∂Hz
∂y − µ0 ω ∂x
z

Ex = i ω2
Ey = i ω2
c2
− k2 c2
− k2
k ∂H ∂Ez
∂x − ωϵ0 ∂y
z
k ∂H ∂Ez
∂y + ωϵ0 ∂x
z

Hx = i ω2
Hy = i ω2
,
c2
− k2 c2
− k2
which express Ex , Ey , Hx and Hy in terms of Ez and Hz . Using Hz = 0, which holds
in general for TM waves, and ∂y ∂
= 0, which holds for our translationally-invariant
waveguide geometry, the field components simplify to Ey = Hx = 0 and
ik ∂Ez iωϵ0 ∂Ez w2
Ex = and Hy = , κ2 = + k2 .
κ2 ∂x κ2 ∂x c2
Note that H = (0, Hy , 0) has only an y component, while E = (Ex , 0, Ez ) has both x
and z components.
Next, we find Ez using the wave equation
 2
∂ 2 Ez

 2 2 ∂ 2
= −κ2 Ez ,

∇⊥ + κ Ez = + κ Ez = 0 =⇒
∂x2 ∂x2
2
where we have again used κ2 = wc2 + k 2 for conciseness. The last equality holds
because ∂y
∂
= 0. The equation is solved with the oscillatory ansatz

Ez (x) = A sin κx + B cos κx.

Next, we consider boundary conditions along the waveguide plates. The general
condition E∥ = 0 applies to the parallel components to waveguide surface, which in
our geometry are Ez and Ey . The result is Ez = Ey = 0—note that we knew the
result Ey from earlier. Evaluated at the boundaries surfaces x = 0 and x = a, the
condition reads
E∥ = Ez (0) = Ez (a) = 0.

70
13.2. A Parallel-Plate Waveguide

We substitute the boundary condition Ez (0) = 0 into the solution ansatz to get

Ez (0) ≡ 0 = A sin κ · 0 + B cos κ · 0 =⇒ B = 0.

The solution then simplifies to Ez (x) = A sin κx.

Next, we apply the condition Ez (a) = 0, which gives
nπ
Ez (a) ≡ 0 = A sin κa =⇒ κ = , n = 1, 2, 3, . . . .
a
Note that n = 0 gives a trivial solution. The solution for Ez in TM mode is then
nπx
Ez (x) = A sin κx = A sin .
a

Dispersion Relation
Next, we aim to find the dispersion relation between frequency ω and wave vector k.
2
Combining the definition κ2 = wc2 + k 2 with the just-derived result κ = nπ
a shows the
dispersion relation is
r
2 ω2 2
 nπ 2  nπ 2
κ ≡ 2 −k = =⇒ ω = c0 k 2 + .
c a a
At large k, this approaches the free space relation ω = c0 k
Note that frequencies below c0 nπ
a are unattainable, since k cannot decrease below
2

zero. In other words, there is a minimum possible frequency at which EM waves

can propagate through the waveguide. When k = 0, the frequencies are quantized
according to
nπ
ωn = c0 .
a
We call the frequency band ∆ω ≡ ω2 − ω1 the waveguide bandwidth. In our case,
the bandwidth is
c0 π
∆ω = .
a
Wave Impedance
Wave impedance (not to be confused with electrical impedance), is the ratio of the
transverse components of the electric and magnetic fields with respect to the direction
of wave propagation.
For TM waves, wave impedance is defined in terms of the E and H fields as
E⊥
Z≡ .
H∥

For our parallel-plate waveguide with E⊥ = Ex and Hx = 0 =⇒ H∥ = Hy , the

waveguide impedance is
Ex
Z= .
Hy

71
13.2. A Parallel-Plate Waveguide

Using the earlier equations for the components Ex and Hy , we have

Ex ik ∂E
∂x
z
k
Z= = ∂E
= .
Hy iωϵ0 ∂x z ωϵ0

We then use the dispersion relation between ω and k to get

q 2
2
ω
− nπ
r
c20 a 1 1  nπc0 2
Z= = 1− 2 .
ωϵ0 c0 ϵ0 ω a

The coefficient c01ϵ0 has a special meaning—it is the impedance of free space Z0 , which
is more visible in the form
√ r
1 ϵ0 µ0 µ0
= = ≡ Z0 .
c0 ϵ0 ϵ0 ϵ0

In terms of Z0 and the frequency ωn = a ,

nπc0
the dispersion relation is written concisely
as r  ω 2
n
Z(ω) = Z0 1− .
ω
For large ω, the impedance approaches Z0 .
Ratio of Electric Field Amplitudes
Next, we find the ratio of electric field amplitudes Ex .
Ez
Substituting in the equation
for Ex gives
Ez Ez
= ∂Ez κ2 .
Ex ik ∂x

Next, we substitute in the earlier result Ez = A sin κx to get

Ez Ez A sin κx 2 κ
= ∂Ez κ2 = κ = .
Ex ik ∂x ikκA sin κx k

We can interpret κ as a wave vector in the x direction, while k is the wave vector
along the direction of propagation z.
TE Propagation Mode
In TE mode, we have Ez = 0 and we find Hz ̸= 0. Once again, we start with the
field component equations
   
i ∂Ez ∂Hz i ∂Ez ∂Hz
Ex = k + ωµ0 Ey = k − µ0 ω
κ ∂x ∂y κ ∂y ∂x
   
i ∂Hz ∂Ez i ∂Hz ∂Ez
Hx = k − ωϵ0 Hy = k + ωϵ0 ,
κ ∂x ∂y κ ∂y ∂x

which express Ex , Ey , Hx and Hy in terms of Ez and Hz . Using Ez = 0 and ∂

∂y = 0,
the field components simplify to Ex = Hy = 0 and

iµ0 ω ∂Hz ik ∂Hz

Ey = − and Hx = .
κ ∂x κ ∂x

72
13.2. A Parallel-Plate Waveguide

Note that E = (0, Ey , 0) has only an y component, while H = (Hx , 0, Hz ) has both x
and z components.
Next, as for Ez , we find Hz using the wave equation
 2
∂ 2 Hz

 2 2 ∂ 2
= −κ2 Hz .

∇⊥ + κ Hz = + κ Hz = 0 =⇒
∂x2 ∂x2

The equation is solved with

Hz = A sin κx + B cos κx.

The boundary conditions in TE mode are different than in TM mode. For TE mode,
we use the boundary condition H⊥ = 0. For our coordinate system and parallel-plate
waveguide, the component of H perpendicular to the surface is Hx , which is
ik ∂Hz
Hx = .
κ ∂x

Using the above expression for Hx in terms of ∂x ,

∂Hz
the condition H⊥ = Hx = 0 at
the boundary surfaces reads
∂Hz
Hx (x = 0, a) ∝ = 0.
∂x x=0,a

Thus, in TE mode, the boundary condition is written in terms of a derivative—note

that ∂H
∂x is a “normal” derivative; since ∂x points in the direction x, which is normal
z ∂Hz

to the waveguide’s boundary.

In general, the boundary condition in TE mode involves the normal derivative of Hz
with respect to the waveguide boundary surface, although the relevant coordinate
will naturally not always be x for different geometries and coordinate systems.
Taking the derivative of Hz and evaluating the boundary condition gives
nπx
Hz (x) = B cos , n = 1, 2, 3, . . . .
a
Note that n = 0 produces the constant solution Hz = B which is not a trivial solution
in itself, but the other field components, which involve derivatives of the constant
quantity Hz , are zero, so we again reject n = 0.
Since the quantity κ is the same in TE mode as in TM mode, the dispersion relation
in TE is the same, and reads
r  nπ 2
ω = c0 k 2 + .
a

For TE waves, wave impedance is defined in terms of the E and H fields as

E∥
Z≡ .
H⊥

73
13.2. A Parallel-Plate Waveguide

For our parallel-plate waveguide with E∥ = Ey and H⊥ = Hx , the waveguide

impedance is
Ey
Z= .
Hx
Using the earlier equations for the components Ey and Hx (although I’ve ignored the
negative sign in Ey ), we have

Ey iµ0 ω ∂H
∂x
z
µ0 ω
Z= = ∂H
= .
Hx ik ∂x z k

We then substitute in the dispersion relation between w and k to get

µ0 ω µ0 ω µ0 c0
Z= =q 2 =q ,
k ω
− nπ

2 ωn2
2
c0 a 1 − ω2

where we’ve used ωn = a .

nπc0
Finally, we use the relationship c0 = √1
ϵ0 µ0 to get
r
µ0 µ0 Z0
µ0 c0 = √ = = Z0 =⇒ Z(ω) = q
2 .
µ0 ϵ0 ϵ0
1 − ωωn

Generalization: Waveguide With a Rectangular Cross Section

Finally, we consider a waveguide with a rectangular cross section with width and
height a and b, corresponding to the x and y directions, respectively. Waves propagate
along the z axis. As before, the “cross-sectional” Laplacian reads

∂2 ∂2
∇2⊥ = + .
∂x2 ∂y 2

Note both ∂x∂

and ∂y
∂
are non-zero, since the waveguide is not finite-dimensional in
the y direction. The wave equation reads
 2
∂2

∂ 2
+ + κ Ez (x, y) = 0.
∂x2 ∂y 2

We solve the equation with separation of variables: Ez (x, y) = X(x)Y (y). We then
substitute the ansatz for Ez into the wave equation to get

X ′′ Y ′′
X ′′ Y + XY ′′ + κ2 XY = 0 =⇒ + = −κ2 .
X Y
X ′′ Y ′′
Since the X terms depend only on x and the Y terms only on y, but X and Y
must be constant for the equation to hold. We thus define

X ′′ Y ′′
= −κ2x =⇒ X ′′ = −κ2x X and = −κ2y =⇒ Y ′′ = −κ2y Y.
X Y
Both equations are solved with sinusoidal solutions of the form

X = A sin κx x + B cos κx x and Y = C sin κy y + D cos κy y.

74
13.2. A Parallel-Plate Waveguide

The boundary conditions now involve both a and b; the results are
nπ mπ
κx = and κy = , n, m = 1, 2, 3, . . . ,
a b
where κ2 = κ2x + κ2y . The dispersion relation now reads
r  nπ 2  mπ 2
ω = c0 k2 + + .
a b
In TM mode, both indices run over n, m = 1, 2, 3, . . ..
Meanwhile, in TE mode, which has cosine solutions, because of the presence of two
indices n and m one of either n or m can be zero.

75
14.1. Cylindrical Waveguide

14 Fourteenth Exercise Set

14.1 Cylindrical Waveguide
Consider a waveguide with a circular cross section of radius a. Find the time-dependent
solution for the electric and magnetic field in the waveguide, along with the dispersion
relation, for both TM and TE propagation modes.
We begin with the wave equation for either Ez or Hz in a pipe-like waveguide:
 
 2 2
 Ez (x, y)
∇⊥ + κ = 0.
Hz (x, y)

Next, we write the Laplacian in cylindrical coordinates—in cylindrical geometry, the

coordinates perpendicular to the propagation direction z are r and ϕ. The Laplacian
reads:
1 ∂2
 
2 1 ∂ ∂
∇⊥ = r + 2 2.
r ∂r ∂r r ∂ϕ

We begin with TM mode, in which we have Hz = 0 and solve for Ez . We find Ez

with separation of variables, using the ansatz

Ez (r, ϕ) = R(r)Φ(ϕ).

We substitute this ansatz into wave equation and get

1
′ 1
rR′ Φ + 2 RΦ′′ + κ2 RΦ = 0.
r r
We then rearrange and get

rR′ + r2 R′′ Φ′′ r2 R′′ + rR′ + κ2 r2 R Φ′′

+ + κ2 r2 = 0 =⇒ =− = m2 ,
R Φ R Φ
where, following the usual separation of variables recipe, we have set the r and
ϕ-dependent terms equal to the separation constant m2 . The two equations are

Φ′′ + m2 Φ = 0

and
r2 R′′ + rR′ + (κ2 r2 − m2 )R = 0.

The equation for Φ has a sinusoidal solution, which we write in the form

Φ(ϕ) ∝ sin(mϕ + ϕm ).

Note that we’ve written the solution with a phase shift ϕm instead of as a linear
combination of sine and cosine terms.
The radial equation is a Bessel equation and is solved by the Bessel functions Jm and
the Neumann functions Nm . The general solution is a linear combination of the form

R(r) ∝ Jm (κr) + Nm (κr).

76
14.1. Cylindrical Waveguide

The Neumann functions apply in situations with divergence near the origin (e.g. a
coaxial cable), and the Bessel functions for convergence near the origin. Our problem
is free of divergence near the origin, so we will use only the Bessel functions. In this
case, the general solution for Ez = ΦR is
∞
X
Ez (r, ϕ) = Am Jm (κm r) sin(mϕ + ϕm ),
m=0

where m indexes the Bessel functions.

We now consider boundary conditions. The electric field in waveguides must satisfy

E∥ ∂
= 0,

where E∥ is the component of electric field tangent (parallel) to the waveguide’s

surface and ∂ denotes the boundary surface. For our cylindrical geometry both Ez
and Eϕ are tangent, and the boundary corresponds to r = a. We apply the condition
to Ez to get
∞
X
Ez (a, ϕ) = Am Jm (κa) sin(mϕ + ϕm ) ≡ 0.
m=0
This implies κ must satisfy

Jm (κa) = 0 =⇒ κa = ξm,n ,

where m = 0, 1, 2, . . . indexes the Bessel functions and n = 1, 2, 3, . . . indexes the zero

of the m-th Bessel function. The solution for κ is then
ξmn
κ= ,
a
where the zeros ξmn are tabulated in the table below. As such, the general solution
for Ez is more appropriately written with two indices in the form
∞ X
∞
X  r
Ez (r, ϕ) = Amn Jm ξmn sin(mϕ + ϕmn ),
a
n=1 m=0

where m indexes the Bessel functions and n indexes the zeros of a given Bessel
function.
n J0 (x) J1 (x) J2 (x) J3 (x) J4 (x) J5 (x)
1 2.4048 3.8317 5.1356 6.3802 7.5883 8.7715
2 5.5201 7.0156 8.4172 9.7610 11.0647 12.3386
3 8.6537 10.1735 11.6198 13.0152 14.3725 15.7002
4 11.7915 13.3237 14.7960 16.2235 17.6160 18.9801
5 14.9309 16.4706 17.9598 19.4094 20.8269 22.2178

Table 1: The first 5 zeros ξmn of the first 6 Bessel functions Jm (x).

With κ known, we find the dispersion relation via

r
ω 2 ξ2
κ2 = 2 − k 2 =⇒ ω = c0 k 2 + mn .
c0 a2

77
14.1. Cylindrical Waveguide

We then find the waveguide’s bandwidth from the difference between the lowest two
frequency modes, which occur for the smallest two zeros ξmn when k = 0, where the
dispersion relation simplifies to
c0
ω(k = 0) = ξmn .
a
The bandwidth is then the difference between the two smallest possible frequencies
ω, which occur for the two smallest values of ξmn . These are:
c0 c0 c0
∆ω = (ξ11 − ξ10 ) ≈ (3.8317 − 2.4048) = 1.47 · .
a a a

TE Mode
In TE mode we have Ez = 0 and solve for Hz . As in TM, we use separation of
variables with the ansatz
Hz (r, ϕ) = R(r)Φ(ϕ).
Following an analogous separation of variables procedure as in TM mode produces
the general solution
∞
X
Hz (r, ϕ) = Am Jm (κr) sin(mϕ + ϕm ).
m=0

In TE mode, we boundary condition applies to H⊥ and reads

H⊥ ∂
= 0,

meaning the component of H perpendicular to the surface ∂ must be zero. In

cylindrical geometry the radial coordinate is perpendicular to the surface, meaning
H⊥ = Hr . We find Hr with
∂Hz ∂Hz
Hr ∝ and Hr ∂
= 0 =⇒ = 0.
∂r ∂r

Applying this boundary condition to the general solution for Hz gives

′
ξmn
′
Jm (κa) = 0 =⇒ κ = ,
a
where ξmn
′ are the zeros of the Bessel function derivatives, tabulated below.

n J0′ (x) J1′ (x) J2′ (x) J3′ (x) J4′ (x) J5′ (x)
1 3.8317 1.8412 3.0542 4.2012 5.3175 6.4156
2 7.0156 5.3314 6.7061 8.0152 9.2824 10.5199
3 10.1735 8.5363 9.9695 11.3459 12.6819 13.9872
4 13.3237 11.7060 13.1704 14.5858 15.9641 17.3128
5 16.4706 14.8636 16.3475 17.7887 19.1960 20.5755

Table 2: The first 5 zeros ξmn of the first 6 Bessel function derivatives Jm
′ (x).

78
14.2. Overview: Waveguide with a Quarter-Circle Cross Section

The correct general solution for Hz is written with two indices in the form
∞ X
∞
′ r
X  
Hz (r, ϕ) = Amn Jm ξmn sin(mϕ + ϕmn ),
a
n=1 m=0

where m indexes the Bessel functions and n indexes the zeros of the Bessel function
derivatives ξmn
′ .

With κ known, the dispersion relation in TE mode is

r
ω2 (ξ ′ )2
κ = 2 − k =⇒ ω = c0 k 2 + mn
2 2
.
c0 a2

We find the bandwidth analogously to TM mode, we just using the new zeros ξmn
′

instead. The difference between the lowest two frequencies is

c0 ′ c0 c0
∆ω = (ξ21 − ξ11 ) ≈ (3.0542 − 1.8412) ≈ 1.21 · .
a a a

14.2 Overview: Waveguide with a Quarter-Circle Cross Section

We briefly sketch the typical problem solving procedure for a waveguide with a
quarter-circle cross section. As for a circular cross section, we work in cylindrical
coordinates and use separation of variables with the ansatzes

Ez (r, ϕ) = R(r)Φ(ϕ) or, in TE mode, Hz (r, ϕ) = R(r)Φ(ϕ).

Substituting these ansatzes into the waveguide wave equation leads to the familiar
general solutions
  X ∞
Ez (r, ϕ)
= Am Jm (κr) sin(mϕ + ϕm ).
Hz (r, ϕ)
m=0

Boundary Conditions in TM Mode

A waveguide with quarter-circular cross section has two additional flat boundaries, in
addition to its circular boundary. In TM mode, we require Ez (ϕ = 0, r) = 0 along
the bottom flat surface and Ez (ϕ = π/2, r) = 0 along the vertical flat surface.
The boundary conditions along the flat surfaces, which are conditions involving the co-
ordinate ϕ, apply to the angular component of the general solution, i.e. sin(mϕ + ϕm ).
The first condition, Ez (0, r) = 0 requires ϕm = 0.
The second condition, Ez (π/2, r) = 0, leads to
 π
sin m = 0,
2
which means m π2 = nπ where n ∈ N. The result is

m = 2, 4, 6, · · · .

In other words, only even values of m are allowed.

79
14.3. Theory: TEM Waves in Waveguides

Finding the radial component of the general solution is analogous to the procedure
for a circular cross section and involves finding zeros of the Bessel functions.
Boundary Conditions in TE Mode
In TE mode, the relevant boundary condition involves the perpendicular magnetic
field H⊥ . Along the circular surface, the perpendicular magnetic field is Hr ∝ ∂H
∂r .
z

Requiring H⊥ = 0 leads to the condition m = 0, 2, 4, 6, . . ., as in TE mode.

Along the bottom flat surface at ϕ = 0, the perpendicular component of magnetic
field is
∂Hz
H⊥ = .
∂ϕ
Applying the boundary condition H⊥ = 0 leads to cos(ϕm ) = 0, which means the
phase term is ϕm = π2 . In other words, the angular solutions are cosine terms, since
 π
sin mϕ + = cos(mϕ).
2

Along the vertical flat surface at ϕ = π/2, the boundary condition reads
∂Hz  π 
H⊥ = , r = 0.
∂ϕ 2
This condition leads to the requirement m = 0, 2, 4, . . .. In TE mode, the case m = 0
is valid because the angular cosine solutions are constant and thus non-trivial, at
m = 0. Similarly, the radial Bessel functions are nontrivial for m = 0, and we get a
valid solution. In TE mode for a quarter-circle cross section, the valid values of m
are thus m = 0, 2, 4, . . ..

14.3 Theory: TEM Waves in Waveguides

In TEM (transverse electric and transverse magnetic) mode we have both Ez = 0 and
Hz = 0, which is a more strict condition than either TE or TM mode individually. In
TEM mode both E and H are perpendicular to the direction of wave propagation, i.e.
both E and B are perpendicular to the wave vector k. We now E ⊥ k and B ⊥ k is
satisfied for EM waves in empty space, and it turns out this condition is also possible
is specialized waveguides, which we consider in this section.
We begin the analysis of TEM waves in waveguides with the familiar ansatz

E(r) = E(ρ)ei(kz−ωt) ,

which assumes E propagates through the waveguide in the z direction.

First Step
Our first step is to show ∇ × E and ∇ × H obey the relationships

∇ × E = ik × E and ∇ × H = ik × H.

We begin the proof with the intermediate calculation

∇ × E(r) = ∇ × E(ρ)ei(kz−ωt) = ∇ × E(ρ)ei(kz−ωt) + ∇ei(kz−ωt) × E(ρ),

 

80
14.3. Theory: TEM Waves in Waveguides

where the second equation holds because the exponent term is a scalar quantity. Next,
we make the calculation

∇ei(kz−ωt) = ikei(kz−ωt) =⇒ ∇ei(kz−ωt) × E(ρ) = ik × E(r).

To show ∇ × E(r) = 0, we just need to show that the term ∇ × E(ρ)ei(kz−ωt) equals
zero. Noting that ρ = (x, y, 0) i.e. the ρz = 0, and also that Ez ≡ 0 in TEM mode,
we have
∂   ∂  
∂x Ex (ρ) ∂x Ex (ρ)
∂  ∂ 
∇ × E(ρ)ei(kz−ωt) =  ∂y × Ey (ρ) ei(kz−ωt) =  ∂y × Ey (ρ) ei(kz−ωt)
∂ Ez (ρ) 0
0
 ∂z 
∂Ey (ρ) ∂Ex (ρ) i(kz−ωt)  
= − e ẑ = ∇ × E(r) z ẑ.
∂x ∂y
Applying Maxwell’s equations shows that
  ∂Hz
∇ × E(r) z = µ0 ≡ 0,
∂t
since Hz = 0 in TEM mode. The result is the desired equality

∇ × E = ik × E,

where k = (0, 0, k).

Proving ∇ × H = ik × H analogous; left as an exercise to the reader i.e. I’m too
lazy to type it out right now.
Second Step
Next, we aim to prove that TEM mode has the dispersion relation ω = c0 k. We begin
with the just-derived equality ∇ × E = ik × E and proceed with
∂H
ik × E = ∇ × E = −µ0 = +iµ0 ωH =⇒ k × E = µ0 ωH.
∂t
Analogously, using ∇ × H = ik × H, we get
∂E
ik × H = ∇ × H = ϵ0 = −iϵ0 ωE =⇒ k × H = −ϵ0 ωE.
∂t

Next, we take the cross product of the above equations with k to get

k × (k × E) = µ0 ωk × H = −ϵ0 µ0 ω 2 E.

The left hand side k × (k × E) evaluates to

k × (k × E) = k(k · E) − k 2 E = 0 − k 2 E,

since k · E = 0 in TEM mode. Using k × (k × E) = −ϵ0 µ0 ω 2 E leads to

ω2
 
2 2 2
−k E = −ϵ0 µ0 ω E =⇒ E k − 2 = 0 =⇒ ω = c0 k.
c0

81
14.4. TEM Waves in a Coaxial Waveguide

In other words, we have proven that TEM mode has the dispersion relation ω = c0 k
Last Step
We begin with the waveguide wave equation, which reads
  2   
2 ω 2 E
∇⊥ + −k = 0.
c20 H

Because of the dispersion relation ω = ck, the equation simplifies to

 
E
∇2⊥ = 0.
H

This equation corresponds to a solution with ω = 0 and k = 0, which corresponds to

a static electromagnetic wave. Mathematically, this means TEM wave propagation
reduces to solving a static (time-independent) Laplace equation.
As an example, we consider the electric field E. Since ∇2⊥ E = 0, (which applies to
the entire electric field), we can also consider only the component E∥ and write

∇2⊥ E∥ = 0.

Meanwhile, the boundary condition for waveguides reads E∥ ∂ = 0. The Laplace

equation for E∥ requires that the Laplacian of E∥ is zero everywhere along the
waveguide cross section, while the boundary condition requires that E∥ itself is zero
along the boundary. The only possible solution for simple2 surfaces is E∥ = 0.
However, for e.g. a coaxial cable with an annular (not simple) cross section, E∥ ̸= 0
(and thus TEM mode) is a valid solution. Another possible waveguide that allows
for TEM mode is a parallel-plate waveguide in which the plates extend to infinity,
meaning the “infinite” cross section is not a simple surface. In general, only waveguides
with non-simple cross sections can carry TEM waves.

14.4 TEM Waves in a Coaxial Waveguide

Consider a coaxial waveguide with an inner conductor of radius a and an outer sheath
of radius b. To make things more interesting, we fill the interior with a material
with the same dielectric constant as the cold plasma in Subsection 12.2. Find the
dispersion relation and impedance for TEM waves in the coaxial waveguide.
Because the waveguide is filled with a dielectric material, the wave speed is c < c0 ,
and the dispersion relation is
ω = ck.
The wave speeds c and c0 are related by
c0 c0 k
c = √ =⇒ ω = √ .
ϵ ϵ
2
By simple surface, we refer to a surface enclosed by a single boundary. For example, a circle
would be a simple surface, but an annulus, which has two boundaries, is not.

82
14.4. TEM Waves in a Coaxial Waveguide

We then substitute in the dielectric constant of cold plasma to get

c0 k q
ω=q =⇒ ω = c20 k 2 + ωp2 .
ωp2
1 − ω2

Note that this is the same dispersion relation as for cold plasma, even though the
plasma problem was solved in unbounded space, while the coaxial waveguide has
bounded geometry.
Impedance
We find the waveguide’s impedance with the relationship Z = UI where U and I are
the potential difference between the inner and outer sheath and the current through
the conductor, respectively, corresponding to TEM waves in the waveguide. In TEM
mode, the EM field is static, which simplifies finding U and I.
We assume the coaxial cable carries a hypothetical current I. Using Ampere’s law,
the current through the inner conductor leads to a tangential magnetic field Hϕ of
the form
I
I = Hϕ · 2πr =⇒ Hϕ = ,
2πr
where Hϕ is the ϕ component of H (tangential to the waveguide’s circular cross
section).
Next, assume a potential difference U between the inner and outer conductors. We
find the associated radial electric field Er using Gauss’s law:
Q
Q = ϵ0 ϵ · 2πrl · Er =⇒ Er = ,
2πϵϵ0 lr
where q is the charge on the inner conductor. Note that because of the coaxial
conductor’s concentric geometry, E has only a radial component and H has only a
tangential component. With Er known, we find the associated potential difference
with ˆ b ˆ b
q dr q b
U= Er dr = =⇒ U = ln .
a wπϵϵ 0 l a r 2πϵϵ0 l a
We can relate U and Er with
b
U = Er r ln .
a
Next, we substitute U and I into the impedance equation to get
U 1 Er b
Z= = ln .
I 2π Hϕ a
We can relate Er and Hϕ with the earlier theoretical TEM result
k
H= × E.
ωµ0
Since E = (Er , 0, 0) and H = (0, Hϕ , 0) are perpendicular, the above vector equation
simplifies to √
k 1 ϵϵ0 µ0 ϵ
r
Hϕ = Er = Er = 2 Er = Er ,
ωµ0 cµ0 µ0 Z0

83
14.4. TEM Waves in a Coaxial Waveguide

where Z0 = √ϵ10 µ0 is the impedance of free space. We substitute this result for Hϕ in
terms of Er into the impedance equation to get
Z0 b 1 Z b
Z= √ ln = q 0 ln .
2π ϵ a 2π ωp2 a
1− ω2

84