Kevin Zhou Physics Olympiad Handouts

Waves I: The Wave Equation

The basics of waves are covered in chapter 18 of Halliday and Resnick, and the rest of the material
needed for Olympiad physics can be found in chapter 13 of Wang and Ricardo, volume 1. For more
on Fourier series, see chapter 6 of French, and for waves in general, see chapters 7 and 8. For a
more advanced perspective, see chapter 16 of Taylor. For many physical examples, see chapters 4
and 6 of Crawford. For more about the physics of music, see these lecture notes. For more fun, see
chapters I-47 through I-50 of the Feynman lectures. There is a total of 75 points.

1 Traveling Waves
Waves is a vast subject, ranging from the humble wave on a string to electromagnetic waves,
gravitational waves, and quantum matter waves. The math used to analyze waves will appear in
just about every physics class you’ll ever take. But more importantly, the subject is rich in examples,
because waves are the physics of the everyday world.
Example 1

Consider a string with mass density µ under tension T . The transverse displacement of the
string is given by the wave function y(x, t), and for simplicity we assume the wave is shallow,
i.e. ∂y/∂x  1. What’s the equation of motion for y?

Solution
Consider a segment of length ∆x. At each end of the segment, the tension provides horizontal
and vertical forces
T T y0
Tx = p ≈ T, Ty = p ≈ T y0
1 + y 02 1 + y 02
where we’re expanding to first order in y 0 . Therefore the total force is

Fy = ∆Ty = T y 00 ∆x.

This mass of this segment is µ∆x, again to first order, so by Newton’s Second Law,

∂2y ∂2y
T ∆x = µ∆x
∂x2 ∂t2
Cleaning this up a bit, we have the wave equation

∂2y ∂2y T
2
= v2 2 , v2 = .
∂t ∂x µ

Physically, this simply says the string tries to straighten out curvature (represented by
∂ 2 y/∂x2 ). The wave equation is the simplest possible equation of motion for waves. Even in
more complicated situations, we often start with this equation and treat the extra terms as
perturbations. The wave equation thus occupies a position like that of the simple harmonic
oscillator.

1
 Kevin Zhou Physics Olympiad Handouts

Idea 1
We may factor the wave equation as a difference of squares,

(∂t2 − v 2 ∂x2 )y = (∂t − v∂x )(∂t + v∂x )y = 0.

Therefore, functions that satisfy (∂t ± v∂x )y = 0 solve the wave equation. It is simple to
verify that these are functions of the form

y(x, t) = f (x ± vt).

Since the wave equation is linear, superpositions of solutions to the wave equation are also
solutions to the wave equation. The general solution is of the form f (x − vt) + g(x + vt) for
arbitrary functions f and g.

[1] Problem 1. Waves of the form y(x, t) = f (x ± vt) simply translate with uniform velocity v. Does
a wave of the form y(x, t) = f (x + vt) move to the left or the right?

[3] Problem 2. Consider a string with the following shape.

(a) If this is a traveling wave moving to the right with velocity v, carefully draw the velocity and
acceleration of every point on the string.

(b) Now suppose the string has zero initial velocity. Sketch the subsequent behavior of the string.

[2] Problem 3 (HRK). A uniform circular hoop of string is rotating clockwise in the absence of gravity.
The tangential speed is v. Find the speed of waves on this string.

[2] Problem 4. A uniform rope of mass m and length L hangs from a ceiling.

(a) Show that the time it takes for

pa transverse wave pulse to travel from the bottom of the rope
to the top is approximately 2 L/g. Under what circumstances is this approximation good?

(b) Does the pulse get longer or shorter as it travels?

[3] Problem 5. [A] At time t = 0, the position and transverse velocity of a string obeying the wave
equation are given by y(x) and vy (x). Find an explicit expression for y(x, t) in terms of these
functions; this is called d’Alembert’s solution. (Hint: construct solutions with initial position y(x)
and zero initial velocity, and vice versa, and add them together.)

Idea 2
A sinusoidal wave has the form
ω
y(x, t) = A cos(kx − ωt + φ), v=
k

2
 Kevin Zhou Physics Olympiad Handouts

where k is the wavenumber and ω is the (angular) frequency. They are related to the
wavelength and period by
2π 2π
k= , ω= .
λ T
Sinusoidal waves will be especially useful because the wave equation is linear. Fourier analysis
tells us that any initial condition can be written in terms of a sum of sinusoids, so if we know
what happens to the sinusoids, we know what happens in general by superposition. This is
just a generalization of ideas we’ve seen in M4 and E6. Just as we saw there, it can also be
useful to promote y to a complex number, where the physical value of y is the real part; for
a sinusoidal wave we would have y(x, t) = y0 ei(kx−ωt) .

Remark
Physicists almost universally use k and ω rather than λ, f , and T . A nice way of thinking of
these variables is that they represent how quickly the phase φ changes, in space or time,
dφ dφ
k= , ω= .
dx dt
If we use a little special relativity, we can even combine these into a single equation,

k µ = ∂ µ φ.

The fundamental relation between particle and wave properties in quantum mechanics is

pµ = ~k µ .

These are the de Broglie relations, which we’ll cover in X1.

[4] Problem 6. For a wave on a string, there are two contributions to the energy: potential energy
from stretching, and kinetic energy from transverse motion.
(a) Find the kinetic and potential energy density of the string in terms of T , µ, y, and its
derivatives.

(b) Evaluate the above quantities for y = A cos(kx − ωt). Is the total energy density uniform?

(c) Show that for a general traveling wave of the form y = f (x−vt), the total kinetic and potential
energy are equal.

(d) Show that for any wave function y, total energy is conserved. This will require some integration
by parts, as well as the wave equation itself; you should assume y goes to zero at infinity.

(e) Compute the energy of the static configuration in problem 2(b), assuming the triangle has
height h and base L, where h  L.
One warning: as we saw in E6, energy is quadratic, so it does not obey the superposition principle.
Locally, the amount of energy can be more or less than the sum of the energies of the superposed
waves, due to interference.

3
 Kevin Zhou Physics Olympiad Handouts

Remark
How can we account for damping in the wave equation? The simplest thing would be to add
a force proportional to vy , which e.g. could be due to air drag. Then

∂t2 y = v 2 ∂x2 y + A ∂t y.

But what if the string is in a vacuum? Then the simplest kind of damping would be due to
the energy lost in bending and unbending of the string, which takes the form

∂t2 y = v 2 ∂x2 y + A∂t ∂x2 y

because ∂x2 y describes the bending. This is called Kelvin–Voigt damping.

In both cases, it’s straightforward to handle the damping since the wave equation remains
linear; we just plug in a solution of the form ei(kx−ωt) and find the new relation between ω
and k. If we pick k to be a real number, we will generally find ω to be complex, with its
imaginary part corresponding to exponential decay of the wave over time.

[3] Problem 7. [A] With a little vector calculus, the results above can be generalized to arbitrary
dimension. For example, ideal waves in three dimensions obey

∂2ψ
 2
∂2ψ ∂2ψ

2 ∂ ψ
=v + + = v 2 ∇2 ψ
∂t2 ∂x2 ∂y 2 ∂z 2

where the function ψ(r, t) could stand for a variety of things, such as the pressure, density, or
temperature (for a sound wave) or the electric or magnetic field (for an electromagnetic wave).

(a) For simplicity, let’s restrict to waves which have spherical symmetry, so that ψ only depends
on r and t. Plug such a spherical wave into the wave equation, and simplify until you get an
equation only in terms of the partial derivatives of ψ(r, t).

(b) Because the area of a sphere goes as r2 , we expect the energy density of a spherical wave
to fall as 1/r2 , and therefore expect the amplitude to fall as 1/r. Therefore, it is useful to
consider the quantity rψ, which has this falloff factored out. By considering the differential
equation that rψ obeys, find the general solution for ψ(r, t).

2 Standing Waves
Idea 3
A standing wave is a solution to the wave equation of the form

y(x, t) = f (x) cos(ωt).

Typically, only discrete values of ω are possible, with the allowed values depending on the
boundary conditions. If the setup is translationally symmetric, then f (x) will be sinusoidal.
If you want to get some intuition, try playing with this PhET simulation.

4
 Kevin Zhou Physics Olympiad Handouts

[2] Problem 8. Consider a string of length L and wave speed v.

(a) Suppose the ends of the string are fixed, i.e. y(x, t) = 0 at x = 0 and x = L. Find the standing
wave frequencies and sketch the configurations.

(b) Do the same if the ends of the string are free, i.e. ∂y/∂x = 0 at x = 0 and x = L.

(c) Do the same if one end is fixed and one end is free.

[2] Problem 9. 01W USAPhO 1997, problem A1.

Idea 4
When a musical instrument plays a note, typically multiple standing waves are excited, so
the resulting sound is composed of multiple frequencies. As you saw in problem 8, often the
standing wave frequencies are all multiples of a single, lowest frequency. This frequency ω0 is
called the fundamental, or first harmonic, while the multiple nω0 is called the nth harmonic.
The fundamental frequency determines the pitch we perceive, while the distribution of energy
among the harmonics determines the timbre, or tonal quality, of the instrument.

[2] Problem 10 (Feynman). Pinch a single length of rubber band about 5 cm long between the
fingernails of your two hands, and twang it in the middle to observe the pitch. Then stretch it to
several times its original length, observing the pitch as you proceed. Make a simple physical model
to explain the observed results.

[5] Problem 11. Some instruments, such as xylophones and marimbas, are made with rigid rods
instead of strings. The equation that describes transverse vibrations is instead

∂2y ∂4y
= −A
∂t2 ∂x4
for a constant A that depends on the material and cross-sectional area.

(a) For a xylophone bar of length L, find the standing wave solutions and their frequencies. For
simplicity, pretend that the solutions are sinusoidal in space, and that the bar has free ends
just like a string, even though this is not true in reality.

(b) When the bar in part (a) is hit, a certain note is sounded. What is the length of the bar that
makes a note one octave higher?

(c) [A] The actual boundary conditions for a free bar are

∂2y ∂3y
= =0
∂x2 ∂x3
at the endpoints, and the solutions aren’t purely sinusoidal in space. Compute the lowest few
standing wave frequencies and compare them to those you found in part (a). You’ll have to
use a calculator or computer to do this.

5
Kevin Zhou Physics Olympiad Handouts

(d) A guitar or piano string satisfies the wave equation with a small additional fourth-order term,

∂2y 2
2∂ y ∂4y
= v − A .
∂t2 ∂x2 ∂x4
Show that the standing wave frequencies are not linearly spaced, as they would be for an ideal
string, but instead are slightly more spaced out. This effect is called inharmonicity. (Hint:
the spatial profiles of the standing waves are still sinusoidal.)
We perceived two notes to be “in tune” when the component frequencies in the notes line up
with each other. But since the frequencies are more spread out than ideal harmonics, a piano
feels more in tune when the fundamental frequencies are spread out a little bit more. This
“stretch tuning” is significant and adds up to about an entire semitone across the piano!

Remark: Beam Theory

Where does the strange fourth-order equation for transverse vibrations above come from?
Since force is the derivative of energy, it’s easier to think about how the energy stored in a
rigid rod differs from that of a string. When a string with tension T , mass per length λ, and
length ` is plucked, giving it a transverse displacement y, then

kinetic energy potential energy T ∆` T y2

∼ λẏ 2 , ∼ ∼ 2
length length ` `
where our expression for ∆` comes from the p Pythagorean theorem. As we know from M4,
the ratio of the coefficients gives ω 2 , so ω` ∼ T /λ. For a general wave with wavenumber k,
we would replace ` with 1/k above, giving ω ∝ k as expected.

Now, a rod is characterized by a Young’s modulus Y , mass density ρ, length `, width w, and
vertical thickness h. If the transverse displacement is y, then
kinetic energy potential energy
∼ ρẏ 2 , ∼ Y (strain)2 .
volume volume
The tricky part is understanding the strain. If you naively used the same logic as for the
string, then all parts of the rod would be stretched, with typical strain (y/`)2 . This is correct
in the limit of large displacements, y  h, where the rod’s thickness is negligible. But for
small displacements, it’s an overestimate.

As the rod is displaced vertically, it slightly shrinks horizontally. As a result, there is a

“neutral line” in the middle of the rod that is neither stretched or compressed. Bonds above
the line are stretched, and bonds below the line are compressed.

6
Kevin Zhou Physics Olympiad Handouts

The neutral line has radius of curvature R ∼ `2 /y. Thus, the strain at the top and bottom
of the rod is of order h/R ∼ hy/`2 . Plugging this in gives
s
Y h
ω∼ .
ρ `2

Again, for a general wavenumber we would replace ` with 1/k, giving the ω ∝ k 2 scaling. For
a derivation of this result by dimensional analysis, see section 9.2.3 of The Art of Insight.

There’s another neat bit of physics we can get here. Consider a horizontal rod with one end
fixed at a wall. What is the vertical deflection of the other end of the rod, due to its own
weight? The gravitational and internal potential energy densities both have “reasonable”,
power-law dependence on the deflection y. Thus, when their derivatives match, so that forces
balance, their absolute values should match within an order of magnitude,
elastic potential energy gravitational potential energy
∼ ∼ ρgy.
volume volume
Solving for the deflection gives
ρg`4
y∼
Y h2
which is the fundamental result of Euler–Bernoulli beam theory. (For a proper derivation in
terms of force and torque balance, see chapters 9 and 10 of Lautrup.)

Example 2

How are the sounds of a violin, a trumpet, and a person different in a room full of helium?

Solution
p
As we saw in T3, the speed of sound in air is γp/ρ. When the air is replaced with helium,
ρ decreases, increasing the speed of sound.

The standing wave frequencies of a violin are determined by properties of the strings, which
aren’t affected by the helium. The only difference will be that the way the violin’s sound
reverberates will be subtly changed.

For the trumpet, the standing wave frequencies are proportional to v/L where L is the
length of the air column inside the trumpet. Thus, the standing wave frequencies go up, and
the trumpet makes higher-pitched notes.

The human voice is more subtle. A wind instrument works by exciting standing waves inside
it. But the source of the human voice is the vibrations of the vocal folds, whose fundamental
vibration frequency is directly controlled by your muscles. The entire rest of your vocal tract
does not affect what frequencies are present, but rather affects how energy is distributed
between those frequencies. (For instance, vowels are characterized by having extra energy

7
 Kevin Zhou Physics Olympiad Handouts

near two particular frequencies, called formants.) Helium changes the resonant frequencies of
the vocal tract and thus changes which frequencies emitted by the vocal folds are emphasized.
It thus changes the timbre, but not the pitch.

[3] Problem 12. Some questions about musical instruments.

(a) A piano makes sound by quickly striking a string with a hammer. The seventh harmonic
doesn’t fit in with the rest that well. If you want to eliminate the seventh harmonic, at what
point(s) can you put the hammer?

(b) A violinist can make the note from an open string sound an octave higher by lightly touching
it at a point while bowing it somewhere else. Which point(s) should be touched?

(c) Suppose a string has its ends attached to walls. A person can set up a standing wave by
holding the string at some point and moving it side to side, sinusoidally with fixed amplitude.
At which point(s) should the string be driven to maximize the amplitude of a given standing

01@
wave? Assume the string experiences very little damping.

[5] Problem 13. EuPhO 2017, problem 1. (Hint: don’t try to use fancy math here, EuPhO
problems are designed to be solved with only elementary math and graph reading.)

Idea 5
Standing wave solutions also exist for waves in more than one spatial dimension. In the special
case where the wave medium is uniform, and shaped like a rectangle (in two dimensions) or
a rectangular prism (in three dimensions), all the standing wave solutions can be found by
separation of variables. That is, they can all be written as

ψ(x, y, z, t) = f (x)g(y)h(z) cos(ωt)

where f , g, and h are sinusoids.

[3] Problem 14. The top of a drum is like a string, in that it has a uniform surface mass density σ
and surface tension γ.

(a) Waves on the drum can be described by its height z(x, y, t). Find the wave equation for a
drum. What is the speed of traveling waves?

(b) Consider a square drum of side length L, where the boundaries are fixed to z = 0. Find the
standing wave solutions and the corresponding ω. What’s the lowest standing wave frequency?
The frequencies will not be multiples of a fundamental frequency, so they are called overtones,
rather than harmonics; that’s why drums don’t sound like they’re playing notes. (Special
examples, such as the timpani, are designed to mostly excite the harmonic frequencies.)

(c) Why does a drum sound different if you hit it near the edge, versus at the center?

8
 Kevin Zhou Physics Olympiad Handouts

[4] Problem 15. When sand is sprinkled on a vibrating metal plate, it forms Chladni patterns. Sup-
pose we (unrealistically) model the plate as a square elastic membrane, as in problem 14, of side
length L obeying the wave equation with wave speed v. Unlike in problem 14, we now assume the
boundaries of the plate are free.

(a) Do Chladni patterns form at the nodes or antinodes of a standing wave?

(b) Find the general standing wave solutions z(x, y, t) and their frequencies.

(c) The plate is also fixed in the middle by the support, so z = ∂z/∂x = ∂z/∂y = 0 there, which
removes many of the standing wave solutions. Find the lowest and second-lowest frequencies
of allowed standing waves.

(d) Sketch the Chladni pattern for the lowest standing wave frequency.

(e) For the second-lowest standing wave frequency, there will be two independent standing waves
with that frequency. What superpositions of them will yield Chladni patterns with 90◦
rotational symmetry? (If you want to see these patterns, you’ll need a computer.)

Remark
The treatment of problem 15 is inaccurate because the restoring force in a metal plate is
rigidity, not tension. The waves actually satisfy the two-dimensional analogue of the fourth-
order equation considered in problem 11, which is called the biharmonic equation,

∂2z ∂4z ∂4z ∂4z

∝ ∇4 z = ∇2 ∇2 z = ∂x2 + ∂y2 ∂x2 + ∂y2 z =



− 2 4
+ 2 2 2 + 4.
∂t ∂x ∂x ∂y ∂y
Of course, I didn’t ask you to consider this, because it would have been quite a slog! But if
you want to learn more about this thrilling subject, see Plates, by Bhaskar and Varadan.

Remark: Wavepackets

Purely sinusoidal traveling waves of the form ei(kx−ωt) are unrealistic, because they have
infinite spatial extent. A realistic alternative is a wavepacket, which looks like a sinusoid

9
Kevin Zhou Physics Olympiad Handouts

with wavenumber k but with a finite envelope, as shown below.

To understand how sinusoids are constructed, consider the superposition of two traveling
waves with wavenumbers k ± ∆k. The wavefunction is

ei((k−∆k)x−(ω−∆ω)t) + ei(k+∆k)x−(ω+∆ω)t) = 2ei(kx−ωt) cos(∆k z − ∆ω t).

This is simply a sinusoid of wavenumber k with a slowly varying envelope, whose character-
istic size is 1/∆k, reflecting how the two component waves slowly move in and out of phase.
The wave is still infinite in size, but this can be remedied by superposing infinitely many
wavenumbers; in this case the component sinusoids never get back in phase again.

If the wavenumbers occupy a region ∆k, then the size of the envelope is of order 1/∆k,
because this is the distance required for the component waves to get out of phase with each
other. This yields an “uncertainty principle” for waves,

∆x ∆k & 1.

In quantum mechanics, particles are described by waves with p = ~k. Substituting this in
immediately gives the Heisenberg uncertainty principle; it fundamentally holds because one
cannot get a finite wave without superposing different wavenumbers.

Alternatively, if we had worked with frequencies instead, we would have had

∆t ∆ω & 1.

This is an “acoustic uncertainty principle”, also important in digital signal processing, where
it is called the Gabor limit. Upon using the de Broglie relations, one finds the energy-time
uncertainty principle.

Idea 6
The dispersion relation of a system is the function ω(k) relating the frequency and wavenumber
of sinusoidal waves. The phase and group velocity
ω dω
vp = , vg =
k dk
describe the velocities of sinusoidal waves of wavenumber k and the envelopes of wavepackets
built from sinusoids near wavenumber k, respectively. We can see the latter result from the
remark above: the peak of the envelope is the point where the components are in phase, and
this point travels at speed ∆ω/∆k ≈ dω/dk.

For ideal waves, the dispersion relation is linear, the group and phase velocities are constant
and equal, and waves travel while maintaining their shape. When the dispersion relation
isn’t linear, the group and phase velocities depend on k, so wavepackets gradually fall apart

10
 Kevin Zhou Physics Olympiad Handouts

(i.e. they disperse). For more discussion of these topics, see chapter 6 of Morin.

Remark
In R1, you learned that nothing can go faster than the speed of light. But the phase velocity
can exceed it; for instance, in problem 16 you will find a phase velocity that can be infinite!
This is compatible with relativity, because the phase velocity isn’t the speed of an actual
object. It’s just a formal quantity, namely the rate of change of the position of points of
constant phase in an infinite plane wave. To reinforce the point, suppose we arranged to
stand at different places and clap at the same time. Then we could say “the clap moved
from me to you at infinite speed”, but clearly nothing about this contradicts relativity.

In some textbooks, you’ll read that while the phase velocity can be faster than light,
the group velocity can’t be, because it’s the speed of an actual pulse. But that’s not
quite true in general either, because that result follows from an approximation. For
instance, in materials with really weird dispersion relations, a single pulse can split
up into two, in which case the speed of “the” peak or “the” envelope isn’t even well-
defined. Accordingly, in these cases the group velocity can be formally faster than light, but
it doesn’t contradict relativity because the group velocity ceases to have its intuitive meaning.

If you’re mathematically minded, you might be bothered by the argument that a superluminal
phase velocity is okay because no “actual object” moves faster than light, since it seems hard
to rigorously define the term “actual object”. Luckily, there’s a simple and perfectly rigorous
definition of the light speed limit: the observable effects of an action must lie in the future
light cone of the action. Suppose you change the wavefunction at the origin, at time t = 0.
Then at time t, the wavefunction at all points r > ct must be the same as if you didn’t make
the change at all. The maximum speed at which changes of the wavefunction propagate is
called the signal velocity, and it can never exceed c.

[3] Problem 16. Consider transverse waves on a horizontal string with tension T and mass density
µ. The string is attached to the ceiling by a large number of vertical springs, so that if the entire
string is pulled down, it will oscillate with frequency ω0 .

(a) Find the wave equation for waves on this string.

(b) By guessing sinusoidal solutions, find ω(k) and the minimum possible frequency.

(c) Compute the phase and group velocity for wavepackets of frequency ω.

(d) What actually happens if you grab one end of the string and try to wiggle it at a frequency
below the minimum possible frequency?

If we treat the string as a quantum system, excitations of the string are particles with E(p)
determined by the function ω(k) you found, along with the de Broglie relations E = ~ω and p = ~k.
Therefore, there is a minimum energy for excitations. In a relativistic and quantum context, this
means that all the particles must be massive; the minimum energy is mc2 . This is a toy model for
how the Higgs field gives particles mass.

11
 Kevin Zhou Physics Olympiad Handouts

[2] Problem 17. The motion of ripples of short wavelength (less than 1 cm) on water is controlled by
the surface tension γ and density ρ.
(a) Use dimensional analysis to constrain the phase velocity vp of ripples with wavenumber k.

(b) Show that vg = (3/2)vp .

3 Reflection and Transmission

When we considered standing waves in the previous section, we were only considering “steady state”
behavior. Now we consider the dynamics of a wave hitting an obstacle more explicitly.
Example 3

Suppose a string defined for x < 0 ends at a hard wall at x = 0. Show that any wave directed
towards the wall will be reflected back upside-down.

Solution
We suppose that we send in a wave of the form

yin (x, t) = f (kx − ωt).

Let the reflected wave be a general wave traveling backward,

yr (x, t) = g(−kx − ωt).

Both of these expressions only have physical meaning for x < 0, since the string only exists
there. Now, the boundary condition is y(0, t) = 0, so we have

f (−ωt) + g(−ωt) = 0.

This tells us precisely that g = −f , so the wave is reflected upside-down but otherwise
unchanged.

There’s an easy way to visualize what’s going on here. We can imagine that there really is
string for x > 0, but that the point x = 0 stays fixed for some reason. Then this situation
corresponds to an incoming wave coming from the left, and a flipped wave coming from the
right. The two meet and cancel at x = 0, and the flipped wave continues on going to the
left, where the physical string is. Fundamentally, this story works for the same reason as the
method of images in electromagnetism: as long as you satisfy the boundary conditions, you
can do whatever you want beyond the boundary.

[1] Problem 18. Another type of boundary condition is the “soft” boundary condition, which requires
dy/dx = 0 at x = 0. Show that waves are reflected from this boundary but not flipped.
[3] Problem 19. Consider the configuration of problem 2 again.
(a) Suppose the string starts at rest with the two outer corners held fixed. Sketch what happens
after the string is released. What is the period of the motion?

12
 Kevin Zhou Physics Olympiad Handouts

(b) Confirm explicitly that the initial potential energy of the string is equal to the kinetic energy
of the string when it is purely horizontal.

(c) What would prevent a real string from achieving this ideal motion? What will the string look
like after a few oscillations? You can check your answer here.

(d) Try to sketch what happens if the string begins with the “pluck” off-center. You can check
your answer here.

Part (a) shows how standing waves are formed. From an initial pluck or impulse, the reflections
from the ends naturally create the oppositely-moving waves needed to form a standing wave.

[4] Problem 20. [A] The general, turn-the-crank method to solve problems like problem 2 is Fourier
series. That is, one can write the initial shape y0 (x) of the wave as a combination of standing waves,
X πnx
y0 (x) = cn sin .
n
L

(a) The coefficients cn can be extracted by integrating y0 (x) against another sine,
Z L
πnx
cn ∝ dx y0 (x) sin .
0 L

Explain why this works and find the constant of proportionality.

(b) By linearity, the subsequent time evolution is

X πnx
y(x, t) = cn sin cos(ωn t)
n
L

where ωn is the frequency of the nth harmonic. Find the coefficients cn . (If you’re so inclined,
you can use a computer to see how the approximation improves as more terms are included,
or to see how the wave evolves over time.)

(c) Argue that

Z L
LX
y02 (x) dx = |cn |2 .
0 2 n
Using the known values of both sides, show that the Riemann zeta function has value
∞
X 1 π4
ζ(4) = = .
n4 90
n=1

In fact, this is one of the simplest ways to compute ζ(4).

We’ll use the idea of Fourier series to illustrate some conceptual points in W2.

13
 Kevin Zhou Physics Olympiad Handouts

Idea 7
More generally, the relation between the incoming and reflected waves may depend on the
exact form of the incoming wave. In this case, it’s useful to consider sinusoidal solution. Let

yin (x, t) = ei(kx−ωt) .

Almost all boundary conditions will state that something at the boundary is constant in time,
which is only possible if the reflected wave has the same frequency. So in general we have

yr (x, t) = rei(−kx−ωt)

where r is the reflection coefficient. If the medium exists for x > 0, there is also a transmitted
wave there, of the form
0
yt (x, t) = tei(k x−ωt)
where k 0 might differ from k, and t is the transmission coefficient. In general, both r and t
may depend on k as well as the boundary conditions. Note that the phases of r and t depend
on the conventions we used to define yr (x, t) and yt (x, t), though the magnitudes don’t.

[4] Problem 21. Suppose the string at x < 0 has a tension T1 and mass density µ1 , while the string
at x > 0 has a tension T2 and mass density µ2 . (If you were doing this at home, it would be difficult
to have T1 6= T2 since the whole setup would accelerate longitudinally. But for the sake of the
problem, suppose the two strings are attached at x = 0 by a massless ring which slides on a vertical
frictionless pole, so that the normal force from the pole balances the longitudinal force T2 − T1 .) As
above, let yin (x, t) = ei(kx−ωt) .
(a) Write down k 0 and the boundary conditions at x = 0.

(b) Show that the reflection and transmission coefficients are

Z1 − Z2 2Z1 p
r= , t= , Zi = µi Ti .
Z1 + Z2 Z1 + Z2
The quantity Zi is called the impedance.

(c) What limiting cases correspond to hard and soft boundary conditions? Verify that the reflection
coefficients match the results above.

(d) Suppose the incoming wave has the exponential form above, but only lasts for a long but finite
time τ . After a long time, the incoming wave is gone, and we have a reflected and transmitted
wave. Verify that energy has been conserved. (Be careful: it’s not simply |r|2 + |t|2 = 1.)
The great thing about the coefficients r and t is that they contain all the information about the
reflection and transmission. For complicated problems with multiple interfaces, it’s best to work
purely in terms of r and t, as solving the wave equation as a whole can get messy.
Remark
In the previous problem, you found that a discontinuity in the wave medium can cause
reflection, if the impedances of the two sides are different. This reflection is often wasteful,

14
 Kevin Zhou Physics Olympiad Handouts

and can be reduced by insertion of “impedance matching” devices which soften the
discontinuity. For example, a conical megaphone helps match the air column of the mouth
and throat to the air outside the mouth.

The language of “impedance matching” comes from circuitry. In E7 you found that trans-
mission lines have a characteristic impedance Z. When two transmission lines are attached,
wave reflection occurs if the impedances mismatch. The point here is that the same ideas
apply to many kinds of waves, as long as one generalizes the notion of impedance.

Remark
You can generalize the methodology of the previous problem to a large variety of similar
problems. For example, suppose the ring at x = 0 wasn’t massless. Then the boundary
conditions would have been changed; instead of the transverse force on the ring vanishing,
the transverse force would have had to be equal to its mass times its transverse acceleration.
(You may recall that setup from the preliminary problem set.) You could even put the ring
on a spring, or give it a damping force (in which case the wave energy is no longer conserved).
In all cases, the technique is just to take exponential solutions on both sides and apply the
relevant boundary conditions. I won’t assign such problems, since they usually involve lots
of messy algebra, but the idea is very important in physics.

4 Interference
Idea 8
The intensity of a wave is proportional to its amplitude squared, so if two waves with
amplitudes A1 and A2 are superposed, the resultant intensity is

I ∝ (A1 + A2 )2 .

This differs from the sum of the intensities by an interference term,

p
I = I1 + I2 + 2 I1 I2 cos θ

where θ is the phase difference between the waves.

[4] Problem 22. Consider two thin imperfect mirrors, each with reflection and transmission coefficients
r and t from both sides, placed a distance L apart, with air in between them and outside them.
This system is called a Fabry–Perot interferometer. A wave with wavenumber k hits the apparatus;
we want to find the reflection and transmission coefficients rnet and tnet of the entire system.
(a) Draw all paths that the light could take to be reflected, and to be transmitted.

(b) By applying the principle of superposition and summing a geometric series, show that
rt2 e2ikL t2 eikL
rnet = r + , tnet = .
1 − r2 e2ikL 1 − r2 e2ikL

15
 Kevin Zhou Physics Olympiad Handouts

Note that your answers may differ by phases, depending on your conventions for rnet and tnet .

(c) Show that all the light is transmitted for some special values of k, even if r ≈ 1. That is,
nearly ideal mirrors can become perfectly transparent! This is called resonant transmission,
and it occurs because the reflected waves perfectly destructively interfere. (Hint: be careful,
and don’t forget that r and t are complex numbers.)

(d) For this system, the statement of energy conservation is

|rnet |2 + |tnet |2 = 1.

Use this fact to get a relation between r and t. (Hint: evaluating this in general is very messy,
so pick an appropriate special value of L to simplify things.)

[3] Problem 23. 01W USAPhO 2004, problem A3.

[3] Problem 24 (Kalda). In fiber optics, devices called equal ratio splitters are often used; these are
devices where two optical fibers are brought into such a contact so that if an electromagnetic wave
is propagating in one fiber, it splits into two equal amplitude waves traveling in each of the fibers.
Assume that all waves propagate with the same polarization, i.e. that all electric fields are parallel.

(a) Show that whenever a wave enters the splitter, from either fiber, one of the outgoing waves is
advanced in phase by π/4, while the other is retarded by π/4.

(b) From part (a) alone, it’s ambiguous which wave is advanced and which wave is retarded. Let’s
suppose that the fibers are set up so that, when a wave enters along fiber 1, the wave that
exits along fiber 1 is advanced. If a wave enters along fiber 2, is the wave that exits along
fiber 1 advanced or retarded?

(c) Now consider two sequentially positioned, identical equal ratio splitters, as shown.

This is called a Mach–Zehnder interferometer. The optical path difference between the inter-
splitter segments of the two fibers is 30 µm. Assuming the wavelength of the incoming
monochromatic light varies from 610 nm to 660 nm, for what wavelengths is all the light
energy directed into fiber 2?

[4] Problem 25. 01@ APhO 2003, problem 2. A nice, though somewhat clunkily worded question on
an interferometer for measuring rotation.

16
Kevin Zhou Physics Olympiad Handouts

Remark: Interference and Energy Conservation

People sometimes get the impression that interference violates energy conservation, but it
doesn’t. For instance, in the double slit experiment, you get destructive interference in some
places, and constructive interference in other places, so that the total energy stays the same.

A natural followup question is: what if you could engineer waves to have destructive
interference everywhere? Wouldn’t that unambiguously violate energy conservation?
Actually, it still won’t, but the reason is a bit subtle and depends on the specifics.

For simplicity, suppose we start with a long string at rest. You hold one end, and your friend
holds the string some distance away. You wiggle your hand, using energy E, to produce a
wave pulse traveling towards your friend. Then you ask your friend to wiggle their hand in
the exact “opposite” way when the wave passes by them, which should also require energy
E, but which should create a wave which perfectly destructively interferes with yours. So
doesn’t an energy 2E just vanish into nowhere?

The subtlety is that your friend will be trying to move the string at the precise moment that
your wave pulse is passing by them. There are two simple limiting cases we can consider.

• If you created the wave by exerting a vertical force profile F (t), then your friend exerts
a force −F (t). But in this case, your friend will be doing negative work on the string,
because it’ll be moving opposite the force they exert. They’re just absorbing the pulse
you put in, so conservation of energy is satisfied because E − E = 0.

• If you created the wave by displacing the rope vertically by y(t), then your friend displaces
it by −y(t). But in this case, the net displacement of the rope at your friend’s hand
will just be zero, because their displacement cancels with the displacement of the wave
pulse. In this case, your friend is acting like a hard wall boundary condition. They
don’t do any work, since their hand doesn’t move. The forward-moving pulse is indeed
completely destroyed, but it is replaced with a backward-moving pulse of equal energy,
so conservation of energy is still satisfied because E + 0 = E.

We can also try to route around this subtlety. For example, suppose you and your friend
tied together some strings into a Y shape, and you each held one of the prongs of the Y,
and made opposite pulses at the same time. Now there’s no issue like the one above, and
once the pulses meet at the vertex, they’ll perfectly destructively interfere, leaving no energy
in the “neck” of the Y. But the waves will also reflect off the vertex, and transmit from
one prong to the other. If you carry out the analysis, you’ll find that all the energy will get
redirected into waves going back up the prongs. Similar arguments hold for electromagnetic
waves encountering optical elements, like beam splitters.

All of thus is not surprising, because interference comes from wave equations, which in turn
are derived from Newton’s laws or Maxwell’s equations, which obey energy conservation.

17