Kevin Zhou Physics Olympiad Handouts

Electromagnetism VIII: Synthesis

Electromagnetism in matter is covered in chapters 10 and 11 of Purcell, or in sections 6.4 and 7.5 of
Wang and Ricardo, volume 2. All other problems combine ideas covered in previous problem sets.
For more on dielectrics, see chapters II-10 and II-11 of the Feynman lectures. Electromagnetism in
matter is covered in greater detail in chapters 4, 6, and 9 of Griffiths, and chapters I-31 and II-32
through II-37 of the Feynman lectures. There is a total of 81 points.

1 Dielectrics
Idea 1
When a dielectric is placed in an electric field, dipoles inside align with the field, reducing
the field value. For the simplest, most symmetrical situations, the field is simply reduced by
a factor of the dielectric constant κ = ϵ/ϵ0 . Hence a capacitor filled with dielectric has its
capacitance enhanced by κ.

The simple fact above, along with physical intuition, will be enough for most problems. However,
it’s also sometimes useful to think about what’s going on inside a dielectric.

Idea 2
Microscopically, a dielectric carries a polarization P with units of electric dipole moment
density, describing the net effect of its dipoles. This results in a “bound” charge

ρbound = −∇ · P

within the dielectric, as well a surface bound charge

σbound = P · n̂.

In a dielectric, the polarization is related to the total electric field by

P = ϵ0 χe E, ϵ = ϵ0 (1 + χe )

where χe is the electric susceptibility. The tricky thing about using P is that it depends on
the total electric field, including the electric field produced by the polarization itself; that is,
in general we have to solve for P in terms of itself.

Example 1

A point charge q is inside a dielectric κ. Find the electric field and charge density.

Solution
By idea 1, the electric field is
q
E= r̂.
4πϵr2

1
 Kevin Zhou Physics Olympiad Handouts

The dielectric simply shields the field by a factor of κ. To find the charge density, note that
q ϵ0 χe
P= r̂.
4πr2 ϵ
The divergence of P is zero everywhere except for the origin, where negative bound charge
piles up to cancel some of the charge q. (The compensating positive charge is at infinity, or
the outer surface of the dielectric if it is finite.) The total charge at the origin is
 
 ϵ0 χe  χe q
q = q − qbound = q 1 − =q 1− =
ϵ 1 + χe κ

which is consistent with Gauss’s law for E.

Example 2

A dielectric sphere of radius R and dielectric constant κ is placed in a field E0 , and as a

result develops a uniform polarization P. Find P and the field everywhere.

Solution
First let’s compute the field due to the sphere. The uniform polarization is equivalent to having
two uniformly charged balls of total charge ±Q displaced by d so that Qd = (4πR3 /3)P. By
the shell theorem, the field inside is uniform,
P
Ep = − ,
3ϵ0
and the field outside is exactly a dipole field. Now we compute the magnitude of P. The
subtlety is that the atoms in the sphere see not to the applied field E0 , but the total field E,

P = χe ϵ0 E, E = E0 + Ep .

Solving the system, we find

3 κ−1
E= E0 , P=3 ϵ0 E0 .
κ+2 κ+2
The polarizability α of each atom is defined as the dipole moment per applied field,

p = αE0

so we have shown above that

3ϵ0 κ − 1
α=
n κ+2
where n is the number density of atoms. This is the Clausius–Mossotti formula; it relates
the macroscopically measurable parameter κ to the microscopic parameter α.

[2] Problem 1 (Purcell 10.10). Assume that the uniform field E0 that causes the electric field in
example 2 is produced by large capacitor plates very far away. The field lines tangent to the sphere
hit each of the distant capacitor plates in a circle of radius r. Find r in terms of R and κ.

2
 Kevin Zhou Physics Olympiad Handouts

Solution. The field lines are tangent at the widest part of the sphere. Consider a Gaussian surface
which is bounded by a distant capacitor plate, a horizontal slice through the middle of the sphere,
and all of these field lines. Using the results of example 2, the charge contained inside is
3(κ − 1)
Q = πR2 ϵ0 E0 − πr2 ϵE0
κ+2
where the first term is from cutting the polarized sphere. The flux through this surface is
3
Φ= πR2 E0 .
κ+2
Applying Gauss’s law, we have r
3κ
r= R.
κ+2
[2] Problem 2 (Purcell 10.38). Using a similar method to example 2, consider an infinite cylindrical
rod of radius R with a fixed, uniform polarization P, where P may have any orientation. (Don’t
worry about where P comes from; just assume it’s “frozen into” the material. That is, the material
is “ferroelectric”.) Qualitatively describe the electric field everywhere.
Solution. First, note that the component of P parallel to the axis of symmetry of the cylinder
doesn’t do anything, because the cylinder is infinite. (You can think of the effect as moving charge
from infinity in one direction to infinity in the other direction.) Thus, without loss of generality we
can take P to be orthogonal to the axis of symmetry.
Now, the result is equivalent to having two uniform cylinders of total linear charge density λ
separated by d where λd = (πR2 )P. This is equivalent to two lines of charge density ±λ separated
by d = d′ where d′ is the length of the projection of d into the plane perpendicular to the axis of
the lines. So outside, it’s the field of two nearby, parallel lines of opposite charge.
Inside a cylinder with charge density ρ, an easy application of Gauss’s law tells us that the field
ρr ρs
is E = 2ϵ 0
. Thus, combining the two cylinders, the field inside is − 2ϵ 0
= − 2ϵP0 , i.e. a constant.
[3] Problem 3 (Purcell 10.2). A rectangular capacitor with side lengths a and b has separation s ≪ a, b.
It is partially filled with a dielectric with dielectric constant κ. The overlap distance is x.

The capacitor is isolated and has constant charge Q.

(a) What is the energy stored in the system?
(b) Using the result of part (a), what is the force on the dielectric? Which direction does it point?
(c) Is your answer to part (b) affected by the presence of fringe fields near the interface?
Solution. (a) The system consists of two capacitors in parallel, with capacitances C1 = ϵ0 (b −
x)a/s and C2 = κϵ0 xa/s. Thus,
C = ϵ0 (a/s)((κ − 1)x + b)
which gives
Q2 Q2 s
U= = .
2C 2ϵ0 a(b + (κ − 1)x)

3
 Kevin Zhou Physics Olympiad Handouts

(b) Note that

dU Q2 s(κ − 1)
F =− = .
dx 2ϵ0 a(b + (κ − 1)x)2
The sign is positive, so it points in direction of increasing x, so the slab is pulled in.

(c) Fringe fields don’t change the result of part (b). The presence of fringe fields does change
the energy found in part (a), but this has essentially no effect on the derivative of the energy,
because shifting the dielectric just shifts the fringe field over essentially unchanged.
Of course, from a force perspective, all of the force is due to the fringe fields, because those
are the only fields with a horizontal component; this paper gives such a calculation. The fact
that you can get the same answer, by using an energy-based derivation that doesn’t depend
on the fringe fields, or by a force-based derivation that relies entirely on the fringe fields, is
just another example of conservation of energy giving us nontrivial information.

[3] Problem 4 (Griffiths 4.28). Two long coaxial cylindrical metal tubes of inner radius a and outer
radius b stand vertically in a tank of dielectric oil, with susceptibility χe and mass density ρ. The
inner one is maintained at potential V , and the outer one is grounded. To what height h does the
oil rise in the space between the tubes?
λ′
Solution. The field in the region with no oil is E = 2πϵλ0 r , and with the oil is E ′ = 2πϵr where λ′ is
the free charge density. Thus,
λ
V = log(b/a),
2πϵ0
and equating with the oil part, we get that λ′ = κλ, as expected. Now, the total charge on this
effective capacitor is
Q = λ′ h + λ(ℓ − h) = λ(χe h + ℓ),
so
Q χe h + ℓ
C= = 2πϵ0 .
V log(b/a)
We know the net force is 12 V 2 (dC/dh) (note that there is not a minus sign here because of the work
done by the battery, as explained in a problem in E2). The gravitational force is ρπgh(b2 − a2 ), so
equating and solving for h gives
ϵ0 χe V 2
h= .
ρ(b2 − a2 )g log(b/a)
[4] Problem 5 (Cahn). [A] The region z < 0 is filled with a dielectric κ. Find the force on a point
charge q a distance d above the origin.

Solution. First, we note there is only bound charge at the plane z = 0, and there cannot be net
charge within the dielectric. This follows from taking the divergence of both sides of P = ϵ0 χe E,
which gives
−ρb = χe ρ.
Since all the charge in the dielectric is bound, this equation can only be satisfied if ρ = ρb = 0.
Now, letting E0 be the electric field due to the point charge alone, we have
σb σb
E|z=0+ = E0 + ẑ, Ez=0− = E0 − ẑ.
2ϵ0 2ϵ0

4
Kevin Zhou Physics Olympiad Handouts

Moreover, by the definition of χe we know that just inside the material,

 
σb
P = ϵ0 χe E0 − ẑ .
2ϵ0
Taking the dot product of both sides with ẑ we have
 
σb
σb = ϵ0 χe E0 · ẑ −
2ϵ0
which rearranges to
χe κ−1
σb = 2ϵ0 E0 · ẑ = 2ϵ0 E0 · ẑ.
χe + 2 κ+1
On the other hand, in the case of a perfectly conducting plane, we had Ez=0− = 0, giving

σ = 2ϵ0 E0 · ẑ.

That is, the charge density for the dielectric is identical except for a constant of (κ − 1)/(κ + 1).
Hence this bound charge is equivalent to an image charge, and the force is
q2 κ−1
F = 2
16πϵ0 d κ + 1
directed towards the dielectric.

2 Magnetic Materials
Idea 3
When a magnetic material is placed in an magnetic field, dipoles inside align with the field
(for a paramagnet) or against the field (for a diamagnet). That is, both dielectrics and
diamagnets reduce the applied field within them (the internal fields of electric and magnetic
dipoles are opposite).

Idea 4
The configuration of a magnetic material is described by its magnetization M, which has
units of dipole moment per unit volume. This results in bound current density

Jbound = ∇ × M

as well as a surface bound current density

Kbound = M × n̂.

In a magnetic material, the magnetization obeys

1 χm
M= B, µ = µ0 (1 + χm )
µ0 1 + χm
where χm is the magnetic susceptibility. (Note that χm is not defined the same way as
χe .) Diamagnets have χm < 0 and paramagnets have χm > 0. Most common materials

5
 Kevin Zhou Physics Olympiad Handouts

are only weakly magnetic, with µ ≈ µ0 . Exceptions include superconductors, which are
perfect diamagnets with χm = −1 and hence µ = 0, and ferromagnets, which have a frozen-in
magnetization even when there’s no external field, and hence no meaningful value of µ at all.

Example 3: Griffiths 6.10

An iron rod of length L and square cross section of side a is given a uniform longitudinal
magnetization M and then bent into a circle with a narrow gap of width w.

Find the magnetic field at the center of the gap, assuming w ≪ a ≪ L.

Solution
First, suppose there was no gap. Before the iron rod was bent, its uniform magneti-
zation M corresponded to a bound current density Kbound = M everywhere along its
surface, directed circumferentially. Since a ≪ L, this remains approximately true after
bending the rod. (A small volume bound current density Jbound appears, but we neglect this.)

Therefore, the current density is the same as that of a toroidal solenoid with current I and
n turns per length, where M = In. The field inside is therefore µ0 M , directed in the θ̂
direction and zero everywhere outside the rod.

Now let’s account for the gap. Adding the gap is equivalent to superposing an opposite
magnetization at the gap. Since w ≪ a, we can treat it as an a × a square current loop, with
current I = mw. By the Biot–Savart law, the field due to such a loop at the center is
√
2 2µ0 M w
Bloop = .
πa
Combining the two gives a total field of
√ !
2 2w
B = µ0 M 1− .
πa

Since magnetization can be a bit mathematically nasty, you’ll rarely be asked to find explicit fields,
as in the above example. What’s more important is the conceptual understanding.

[3] Problem 6 (IPhO 2012). Water is a diamagnetic substance. A powerful cylindrical magnet with
field B is placed below the water surface.

(a) Which of the following shows the resulting shape of the water surface?

6
 Kevin Zhou Physics Olympiad Handouts

The magnet is roughly 2/3 as wide as each of these sketches.

(b) Let ρ be the density of the water. If the maximum change in height of the water surface is h,
find an approximate expression for µ − µ0 .

For a closely related problem, see EuPhO 2018, problem 2. Here, the pressure change due to the
interaction of water with an extreme magnetic field causes water to boil. As covered in T3, this
happens roughly when the water pressure hits zero, since the vapor pressure of water at room
temperature is small.

Solution. (a) The potential energy of a dipole is U = −m · B. In a diamagnet, m and B are

antiparallel, so the potential energy of the water is positive where the magnetic field is strong.
The water is thus repelled from regions of higher magnetic field, leading to option D.

(b) The surface of the water is an equipotential, so when the water surface dips, the higher magnetic
dipole interaction energy must be exactly compensated by lower gravitational potential energy.
Per volume, the former is
B 2 χm B2
MB ≈ = 2 (µ − µ0 )
µ0 µ0
while the latter is ρgh. Equating the two gives

µ20 ρgh
µ − µ0 ≈ .
B2
However, there’s a subtlety here. Our derivation of the potential energy in E4 assumed that
m had constant magnitude, but that’s not true for substance like water. In the absence of any
field, water has no magnetic dipole moment at all. Instead, its magnetic dipole moment grows
linearly with the applied field. That means that the energy of a dipole is not just −m · B,
but rather
m·B
Z
U = − m · dB = −
2
where the 1/2 comes from taking the integral of a linear function. This factor of 2 means the
true final answer is
2µ2 ρgh
µ − µ0 ≈ 0 2 .
B
[3] Problem 7. In E4 we covered USAPhO 2015, problem B2, which shows that the fields inside
magnets differ, depending on whether they are made of “Ampere” or “Gilbert” dipoles. For example,
consider a long, thin cylindrical magnet magnetized along its axis. In the Ampere model, the internal
field B0 points along M, while in the Gilbert model the internal field is approximately zero. We
can try to distinguish between the models by drilling a hole into the magnet, and measuring the
field inside the hole.

(a) Suppose we drill a long, thin cylindrical hole inside the magnet, parallel to its axis. Show that
the field in this hole is the same in both the Ampere and Gilbert models.

7
 Kevin Zhou Physics Olympiad Handouts

(b) Suppose we drill a short, flat cylindrical hole inside the magnet. Show that the field in this
hole is the same in both the Ampere and Gilbert models.

(c) Is there any way to tell the two models apart, by drilling holes in magnets and measuring the
field inside the hole? What if you used different magnet shapes?

Solution. (a) The trick to answering this part and the next is that drilling the hole is like
superposing another magnet with opposite magnetization. If we make a long thin cavity with
the same shape as the magnet itself, then in the Ampere model the field is B0 − B0 = 0, while
in the Gilbert model the field is 0 − 0 = 0.

(b) In this case, in the Ampere model the field is about B0 − 0, while in the Gilbert model the
field is about 0 − (−B0 ), so the two coincide again.

(c) No, the fields will always coincide. The only difference between an Ampere dipole and a
Gilbert dipole is its internal field; far from such dipoles the fields are identical. If you drill a
hole, then inside the hole, your detector will be far from all the dipoles in the material (which
are atomic-sized), so it won’t be able to tell the difference. That’s why the Gilbert dipole
description works so well.

[3] Problem 8. EFPhO 2004, problem 6. An elegant, tricky problem on permanent magnets.

Solution. See the official solutions here.

[4] Problem 9 (Cahn). [A] A small dipole m in vacuum points towards the plane surface of a medium
with permeability µ. The distance between the dipole and surface is d.

Find the force acting on the dipole.

Solution. We’ll parallel the solution to problem 5 as closely as possible. First, by the same
reasoning as in that problem, there is only bound current at the plane z = 0.
Now, letting B0 be the magnetic field due to the dipole alone, we have
µ0 µ0
B|z=0+ = B0 + K × ẑ, B|z=0− = B0 − K × ẑ
2 2
by Ampere’s law. Therefore, by the definition of χm , we know that just inside the material,
χm  µ0 
µ0 M = B0 − K × ẑ .
1 + χm 2
Now we take the cross product of both sides with ẑ. We know that K = M×ẑ, and furthermore that
taking the cross product of K with ẑ simply rotates it by 90◦ in the xy plane; hence K × ẑ × ẑ = −K.
Then we have
χm  µ0 
µ0 K = B0 × ẑ + K .
1 + χm 2

8
 Kevin Zhou Physics Olympiad Handouts

Upon solving for K, we have

µ0 χm µr − 1
K= (B0 × ẑ) = (B0 × ẑ).
2 χm + 2 µr + 1

On the other hand, in the case of a superconductor, we have B|z=0− = 0, giving

µ0
K = −B0 × ẑ.
2
In this case, we know there is an image dipole of magnitude m directed opposite to the real dipole,
and by the result of problem 12, the force between the dipoles is

3µ0 m2
F = .
2π (2d)4

Note that unlike the case of a dielectric, this force is repulsive. For general µr , the surface current
density is multiplied by a factor of (µr − 1)/(µr + 1), so we have an image dipole at the same location
but with smaller magnitude. The repulsive force is hence

3µ0 m2 µr − 1
F = .
32π d4 µr + 1

[5] Problem 10. Physics Cup 2012, problem 2.

Solution. See the official solutions here.

3 Multipoles
In this section, we explore some of the physics of dipoles and higher multipoles.

[3] Problem 11 (Purcell 10.27). Two monopoles of opposite sign form a dipole, two dipoles of opposite
sign for a quadrupole, and so on. Hence we can construct arbitrarily high multipoles using the rows
of Pascal’s triangle.

The field of a dipole falls as 1/r3 , a quadrupole as 1/r4 , and an octupole as 1/r5 .

(a) To warm up, verify explicitly that the quadrupole field along the axis of the quadrupole starts
at 1/r4 , i.e. that all lower terms cancel.

(b) Prove that this cancellation occurs for general multipoles along their axis.

9
Kevin Zhou Physics Olympiad Handouts

(c) [A] The magnitude and orientation of a dipole is specified by a vector, with three components.
How many numbers are necessary to specify the magnitude and orientation of a quadrupole?
(The linear quadrupoles here are just a special case of a general quadrupole.) Try to generalize
to arbitrary multipoles.

To learn how to decompose an arbitrary charge distribution into multipoles, see section 3.4 of
Griffiths.

Solution. Since we’re lazy we’ll set the Coulomb constant k = 1, and the unit of charge also to 1,
as well as the unit of distance spacing.

(a) See the solution to (b).

(b) A simple way to do this is to reason inductively. For example, an octupole field is nothing
more than two quadrupoles whose leading terms cancel, so the leading field of an octupole
has to be at least one power lower in r.
However, we will give an explicit proof. A 2N -pole can

be constructed from N + 1 charges,
with charge j placed at x = −j with charge (−1)j Nj . Then the field at point x is

N   ∞ 
N  X 
X
j N 1 −2
X N −2
E(x) = (−1) =x (j/x)k .
j (x + j)2 j k
j=0 j=0 k=0

We see that this can be split into sums of the form f (k) = N j N k
P

j=0 (−1) j j , and the coefficient
of x−2−k is some nonzero multiple times f (k). So it suffices to show that f (k) = 0 for all
k < N , and f (N ) ̸= 0. This is an exercise in algebraic sums. The key idea is to define
N    XN   
j N j j N j
X
g(k) = (−1) = (−1) .
j k j k
j=0 j=k

We see that j k can be written as a linear combination of 0j , . . . , j

k , so it suffices to show
that g(k) = 0 for all k < N , and that g(N ) ̸= 0. We see that
N   
X
jN j
g(k) = (−1)
j k
j=k
N   
X
jN N −k
= (−1)
k j−k
j=k
 X N  
N j N −k
= (−1)
k j−k
j=k
 
N
= (−1)k · 1k=N
k

where we used the fact that M ℓ M

P

ℓ=0 (−1) ℓ = 1M =0 (here 1S is 1 if and only if S is true, and
is 0 otherwise), which follows from the binomial theorem. This completes the proof.

10
 Kevin Zhou Physics Olympiad Handouts

(c) Let’s think of a general quadrupole as a superposition of two dipoles in opposite directions.
Then there are three things that determine a quadrupole: the strength of the quadrupole
moment (i.e. the prefactor of the 1/r4 field), the orientation of the first dipole, and the direction
the second dipole is displaced from it. This is 1 + 2 + 2 = 5 total parameters.
Similarly, to specify an octupole, we do the same above, then specify the direction the second
quadrupole is displaced, giving 5 + 2 = 7 parameters. In general, a 2N -pole has 2N + 1
parameters.

[3] Problem 12 (Purcell 11.23). Two magnetic dipoles are arranged as shown.

Show that the associated potential energy is

µ0 m1 m2
U= (sin θ1 sin θ2 − 2 cos θ1 cos θ2 ) .
4πr3
For what orientations is this potential energy maximized or minimized?

Solution. The magnetic field from a dipole pointing in the z direction is:
µ0 m
B= (2 cos θr̂ + sin θθ̂).
4πr3
Let n̂ be the unit vector perpendicular to r̂ (θ = −π/2). Then the field of m1 is
µ0 m 1
B12 = (2 cos θ1 r̂ + sin θ1 n̂)
4πr3
The potential of a dipole in a field is U = −m · B. Note that m2 = m2 cos θ2 r̂ − m2 sin θ2 n̂.
µ0 m1 m2
U = −m2 · B12 = (sin θ1 sin θ2 − 2 cos θ1 cos θ2 )
4πr3
as desired. To extremize this expression, we set the partial derivatives with respect to θ1 and θ2
equal to zero. The results are

cos θ1 sin θ2 = −2 sin θ1 cos θ2 , sin θ1 cos θ2 = −2 cos θ1 sin θ2

which implies that

cos θ1 sin θ2 = sin θ1 cos θ2 = 0.
This can only hold if
cos θ1 = cos θ2 = 0 or sin θ1 = sin θ2 = 0.

11
 Kevin Zhou Physics Olympiad Handouts

The first option leads to local maxima or minima, where both the angles are ±π/2, with maxi-
mum/minimum energy occurring when the dipoles are anti-aligned/aligned. The second option
leads to the global maximum and minimum,

maximum: (0, 0) or (π, π), minimum: (0, π) or (π, 0)

where the dipoles are anti-aligned/aligned, along the direction of the separation between them.

[2] Problem 13 (Purcell 11.36). Three magnetic compasses are placed at the corners of a horizon-
tal equilateral triangle. As in any ordinary compass, each compass needle is a magnetic dipole
constrained to rotate in a horizontal plane. The Earth’s magnetic field has been shielded. What
orientation will the compass needles eventually assume? Does your result also hold for regular
N -gons?

Solution. We claim they point in the direction of the tangents to the circumcircle of the triangle.
In this case, the field at any one corner due to the compasses at the other corners points in the
tangential direction, so the compasses are all aligned with the local fields.
We can show this claim by symmetry. Consider the field at a given corner of the triangle.
Flipping about the axis that passes through this corner and the midpoint of the opposite side
negates the dipole moments at the other two corners, so it must negate the field. But physically, the
rotation operation negates the tangential component of the field. So there must only be a tangential
component, i.e. the field at this corner is purely tangential. This argument holds unchanged for
regular N -gons.

[3] Problem 14. Some questions about forces between dipoles and other multipoles.

(a) Above, you’ve shown that the force between permanent magnetic dipoles falls off as 1/r4 .
How about two permanent electric dipoles?

(b) How about a permanent dipole and a permanent quadrupole?

(c) How about two permanent quadrupoles?

(d) Now consider an ion and a neutral atom. The electric field of the ion polarizes the atom; the
field of that induced dipole then reacts on the ion. Show that the resulting force is attractive
and falls as 1/r5 .

Solution. (a) The basic form of the fields and forces is identical, so the answer is the same.

(b) The field of a quadrupole goes like 1/r4 , so energy of the dipole goes like U ∼ mB ∼ 1/r4 .
Thus, the interaction energy in this case goes like 1/r4 , for a force of 1/r5 .

(c) A single quadrupole is two dipoles with moments m and −m separated by dr where mdr is
the quadrupole moment order. We have that the energy of the quadrupole is

m · B(r + dr) − m · B = m · ((some sort of derivative of B) · dr).

The field of one quadrupole is 1/r4 , so its derivative is 1/r5 . Thus the energy of interaction
goes like 1/r5 , for a force of 1/r6 .

12
 Kevin Zhou Physics Olympiad Handouts

(d) The field of the ion falls as 1/r2 , so the dipole moment induced is p ∼ 1/r2 . Furthermore, the
dipole moment points along the field and hence the displacement between the ion and atom,
indicating the force is attractive. The electric field from the dipole (and hence the force) goes
as p/r3 ∼ 1/r5 . (You might wonder if the induced dipole then gives the ion itself a dipole
moment. It does, but the resulting force is much weaker than the one we found here, between
the induced dipole and the ion’s overall charge.)

4 Electromagnetic Waves in Matter

In this section, you will work out some of the theory of electromagnetic waves in matter.

Idea 5
In the absence of any free charge or current, Maxwell’s equations in matter are identical to
Maxwell’s equations in vacuum, except that ϵ0 and µ0 are related by ϵ and µ, so the waves
√
propagate with speed 1/ ϵµ = c/n, with E = (c/n)B.

[5] Problem 15. Suppose the regions x < 0 and x > 0 are filled with material with permittivities ϵ1
and ϵ2 , both with permeability µ0 . (This is typical; if you don’t count permanent magnets, most
objects have permeability about µ0 .) We send in an incident wave from the left with electric field

Ei ei(ki ·x−ωi t) .

The wave will be both transmitted and reflected at the interface, so the total electric field is
(
Ei ei(ki ·x−ωi t) + Er ei(kr ·x−ωr t) x < 0,
E=
Et ei(kt ·x−ωt t) x > 0.

The angles with the normal are θi , θr , and θt as shown.

(a) Argue that by continuity of the field at the boundary,

ωi = ωr = ωt .

(b) Suppose the y-axis is oriented so that ki · ŷ = 0. Argue that

kr · ŷ = kt · ŷ = 0, ki · ẑ = kr · ẑ = kt · ẑ.

13
Kevin Zhou Physics Olympiad Handouts

From these conditions, derive the laws of reflection and refraction,

θ i = θr , n1 sin θi = n2 sin θt .

Note that neither this part nor the previous part require Maxwell’s equations; they hold for
all kinds of waves as long as we define ni ∝ 1/vi .

(c) Argue that at the boundary, E∥ and B⊥ must be continuous in general. In this case, because
both sides have the same permittivity µ0 , there is no bound current, so B∥ is also continuous.

(d) Now suppose the electric fields are polarized along the y axis, so Ei , Er , and Et are all parallel
to the y-axis. Then continuity of E∥ gives

Ei + Er = Et .

Using continuity of B∥ , show that

Er n1 cos θi − n2 cos θt Et 2n1 cos θi
= , = .
Ei n1 cos θi + n2 cos θt Ei n1 cos θi + n2 cos θt
These are the Fresnel equations for normal polarized light. (Hint: this is a bit messy, so you
can warm up with the case θi = 0.)

(e) If n1 > n2 , then total internal reflection occurs when

n2
sin θi >
n1
and the wave is totally reflected. Nonetheless, Et is nonzero in this regime. To make sense
of this, show that kt · x is imaginary in this regime, indicating that the “transmitted” wave
does not propagate in the region x > 0, but rather exponentially decays.
Solution. (a) Continuity of the field at the interface gives

Ei′ eωi t + Er′ eωr t = Et′ eωt t .

If ωi ̸= ωt , then the left side cannot be of the form Eeωt , so ωi = ωr , and for the two sides to
match up, we must have ωi = ωr = ωt .
Of course, the deeper reason behind this is just what we said in M4 and will see again in W1.
The differential equation the field obeys is linear, and has no explicit time dependence. Thus,
it has solutions with uniform frequency everywhere.

(b) Again, look at the boundary and match parallel components of E. By fixing z but varying y,
we get an equation of the form

Ei′ eki ·yŷ + Er′ ekr ·yŷ = Et′ ekt ·yŷ .

As before, for this to work, all the frequency factors of ksomething · ŷ must be identical. We
can do the same argument for ẑ.
The first set gives that all the k’s are in the same plane. The second gives that ki sin θi =
kr sin θr = kt sin θt . We have ω/k = c/n, so k = nω/c. Since all the ωs are the same, this
equation then reads
n1 sin θi = n1 sin θr = n2 sin θt ,
which is exactly what we want.

14
 Kevin Zhou Physics Olympiad Handouts

(c) For B⊥ , consider a thin Gaussian pillbox that straddles the boundary. By Gauss’s law for
magnetism, the magnetic flux through it must be zero. In the limit of a very thin pillbox, this
ensures the continuity of B⊥ .
H
For E∥ , consider a thin Amperian loop that straddles the boundary, and consider E · ds. As
the width of the loop goes to zero, the magnetic flux through it goes to zero, so this integral
must be zero. Taking loops of various orientations, this ensures the continuity of E∥ .
Note that E⊥ and B∥ need not be continuous, because we can have surface charges and
currents at the boundary. Since both sides have the same µ0 , there are no surface currents,
so B∥ is continuous.

(d) The continuity of B⊥ gives

Bi cos θi − Br cos θr = Bt cos θt .

Since B = En/c, this means

Ei n1 cos θi − Er n1 cos θr = Et n2 cos θt .

Now with the continuity of E∥ (Ei + Er = Et ), and θi = θr , we have

Ei n1 cos θi − Er n1 cos θi = Ei n2 cos θt + Er n2 cos θt

which yields
Er n1 cos θi − n2 cos θt Et 2n1 cos θi
= , =
Ei n1 cos θi + n2 cos θt Ei n1 cos θi + n2 cos θt
as desired.

(e) As seen in part (b), we have

(ki )y = (kt )y , (ki )z = (kt )z

The magnitudes of the wavenumbers are known:

ω ω n2
kt = n2 , ki = n1 , kt = ki
c c n1
Then using the expression for the magnitude of kt :

n2 2
 
2 2 2 2
(kt )x = kt − (kt )y − (kt )z = ki − (ki )2y − (ki )2z
n1
Since sin θi represents k∥ /k, then for sin θi = k∥ /k > n2 /n1 :

n2 2
 
(kt )2x = ki − (ki )2∥ < (ki )2∥ − (ki )2∥ = 0
n1
from which we conclude that
(kt )2x < 0.
Thus the electric field will exponentially decay as it goes into the material.
[5] Problem 16. In most common materials, µ ≈ µ0 while ϵ depends on frequency. We’ll investigate
the origin of this frequency dependence below.

15
Kevin Zhou Physics Olympiad Handouts

(a) Model an electron in an atom as a mass m with charge q attached to a spring, with natural
frequency ω0 and a damping force −mγv, in an electric field E0 e−iωt . Write down the equation
of motion for the electron.

(b) The atomic polarizability is p = αE. Show that

q 2 /m
α= .
−ω 2 + ω02 − iγω

(c) For a gas with small number density n, the Clausius–Mossotti formula reduces to

ϵ = ϵ0 + nα.

Therefore, the permittivity is generally a complex number. The wavevector and frequency
are related by k 2 = µϵω 2 . Explain why the fact that ϵ is complex indicates that waves can be
absorbed.

(d) What frequency maximizes the absorption rate of the electromagnetic waves? Roughly how
many wavelengths does a wave propagate at this frequency before being absorbed?

(e) What frequency maximizes the speed of the electromagnetic waves, and what is that speed?

(f) Transparent objects such as glass can be modeled as having a very high resonant frequency,
much higher than that of visible light. Does blue light or red light refract more when passing
from air to glass?

The intuitive reason that these electrons can affect the propagation speed of light is because they
emit secondary electromagnetic waves that are out of phase with the original wave; this “pushes”
the phase of the composite wave forward or backward, affecting the phase velocity. A complete
explanation can be found in chapter I.31 of the Feynman lectures.

Solution. (a) We have

mr̈ = −mω02 r − mγv + qE0 e−iωt .

(b) Suppose r = r0 eiωt where r0 is potentially complex. Then, we see that E0 ∥ r0 and

−mω 2 r = −mω02 r − mγiωr + q(E0 /r0 )r.

Thus,
m(ω02 − ω 2 + iγω)
E0 /r0 = .
q
We have that p = −r0 q/E0 which yields the result.

(c) If ϵ is complex, then with µ ≈ µ0 and ω 2 being real, then k 2 = µϵω 2 will also be complex.
Thus with a complex wavevector k, the field of E0 ei(k·x−ωt) will exponentially decay since
there will be a negative term from the complex wavevector k.
√
(d) The absorption occurs exponentially from the complex component of kx. With k = ω µϵ ≈
√ nα
ω µ0 ϵ0 (1 + 2ϵ0
), the absorption rate is maximized when the complex component of k is
maximized.
q 2 /m
 
ωn ωn
β ≡ Im(k) = Im α = (γω)
2cϵ0 2cϵ0 (ω02 − ω 2 )2 + (ωγ)2

16
 Kevin Zhou Physics Olympiad Handouts

q 2 γn ω2
=
2mcϵ0 (ω02 − ω 2 )2 + γ 2 ω 2
The maximum value of this occurs when
dβ
∝ (ω02 − ω 2 )2 + γ 2 ω 2 − ω 2 (2(ω 2 − ω02 ) + γ 2 ) = 0


dω 2

which simplifies to yield

ω04 − ω 4 = 0.
Thus a frequency of ω = ω0 will maximize the absorption rate of the electromagnetic wave.
q n 2
The electric field will have a factor of e−βx , and at ω = ω0 , β = 2γmcϵ 0
. The value of the real
wavevector Re k will be close to (note that Re(α) = 0 at ω = ω0 ):
  
ω0 nα ω0
Re(k) = 1 + Re =
c 2ϵ0 c
1
Then for the wave to fall off by a factor of e, the wave will need to travel a distance of β,
1 k
which is βλ = 2πβ wavelengths. Thus,
k ω0 γmϵ0
=
2πβ πq 2 n
is the number of wavelengths it will travel before the amplitude gets reduced by a factor of e.
(e) The phase velocity is maximized when Reω k , or Re √1µϵ is maximized.
cq 2 n ω 2 − ω02
 
1 1 nα
vp = Re √ ≈ c 1 − Re =c+
µϵ 2 ϵ0 2mϵ0 (ω 2 − ω02 )2 + (γω)2
Differentiating with respect to ω 2 and finding where it’s 0 yields:
(ω 2 − ω02 )2 + γ 2 ω 2 − (ω 2 − ω02 )(2(ω 2 − ω02 ) + γ 2 ) = 0
(ω 2 − ω02 )2 = ω02 γ 2
ω 2 = ω02 ± ω0 γ
Looking at the original, the smaller solution yields the minimum velocity, and the larger
solution yields the maximum velocity (which happens to be greater than c). The maximum
phase velocity is
cq 2 n ω0 γ
vmax = c +
2mϵ0 (ω0 γ) + γ 2 (ω02 + ω0 γ)
2

(f) From the previous part, we have

cq 2 n ω02 − ω 2
vp = c −
2mϵ0 (ω − ω02 )2 + (γω)2
2

and now we know that ω0 ≫ ω, so

cq 2 n ω02 − ω 2 q2n
vp ≈ c − ≈ c − c (1 + ω 2 /ω02 )
2mϵ0 ω04 − 2ω02 ω 2 + (γω)2 4
2mϵ0 ω0
Thus increasing the frequency would decrease the speed of light and increasing the index of

01hˆ
refraction, so blue light would refract more.

01hˆ
[5] Problem 17. IPhO 2002, problem 1. A neat application of electromagnetic waves in matter.

[5] Problem 18. APhO 2007, problem 2. A problem on an exotic negative index of refraction.

17
 Kevin Zhou Physics Olympiad Handouts

Remark
Above, we considered the response of a medium composed of atoms, obeying p = αE. However,
this relation is just an approximation, like Hooke’s law. For larger electric fields, higher order
terms are necessary,
p = αE + α′ E 2 + . . .
which lead to strange effects, studied in the field of nonlinear optics. For example, suppose
we send in light of frequency ω. Then

1 + cos(2ωt)
E 2 ∝ cos2 (ωt) = .
2
That means that a nonlinear medium can respond to light at frequency ω by oscillating, and
hence emitting light, at frequency 2ω. This phenomenon is called frequency doubling, or
second-harmonic generation, and converts red light to blue. Similarly, for a cubic nonlinearity,
you can use trigonometric identities to show that frequency tripling can occur.

5 Electromagnetic Systems
In this section we’ll consider problems that use everything we’ve covered, with a focus on techno-
logical applications and systems with multiple moving parts.

[2] Problem 19 (Purcell 11.19). A magnetic dipole m oscillates so that m(t) = m0 cos ωt. Some of
its flux links the nearby circuit C1 , inducing an electromotive force E1 sin ωt.

If a current I1 flowed in C1 , then the resulting field at the location of the dipole would be B1 . Show
that E1 = (ω/I1 )B1 · m0 . (Hint: recall the results involving mutual inductance in E5.)

Solution. Let there be a flux Φ1 in circuit 1 and Φ2 = B1 · m0 /I2 in circuit 2. Then because
L12 = L21 , as stated in E5, we have

Φ1 /I2 = Φ2 /I1 =⇒ Φ1 = B1 · m0 /I2 .

Then,
E1 (t) = −dΦ1 /dt = (ω/I1 )B1 · m0 sin ωt,
so E1 = (ω/I1 )B1 · m0 .

[3] Problem 20. EFPhO 2007, problem 3. A problem on focusing particles with electric fields.

01T†
Solution. See the official solutions here.

[4] Problem 21. IPhO 2004, problem 3. A practical problem which also reviews damped/driven
oscillations.

18
 Kevin Zhou Physics Olympiad Handouts

[4] Problem 22. EFPhO 2014, problem 1. A challenging problem about a complex nonlinear circuit.

Solution. See the official solutions here.

[5] Problem 23. Physics Cup 2020, problem 1.

Solution. See the official solutions here.

19