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 Vibrations and Waves by 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’s Waves. 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 78 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 µ per unit length, 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 y′
Tx = p ≈ T, Ty = p ≈ T y′
1 + y ′2 1 + y ′2
where we’re expanding to first order in y ′ . Therefore the total force is

Fy = ∆Ty = T y ′′ ∆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?
Solution. This wave moves towards the left. To see this, note that at time t = 0 the wave profile is
f (x), while a small time later it looks like f (x + v∆t). This is the graph of f (x) shifted to the left.
[2] 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 is held in place, with zero velocity. If it is suddenly relased, sketch
the subsequent behavior of the string.

Solution. (a) You can figure out the velocity in two different ways. First, since the wave is
proportional to f (x − vt), the vertical velocity is
∂y ∂y
= −vf ′ = −v
∂t ∂x
so it is proportional to the slope of the string. Alternatively, you can think about how the
string has to move so that a moment later, its shape is the same but translated to the right.
The result is shown below:

2
 Kevin Zhou Physics Olympiad Handouts

To derive the acceleration, you can think about how the velocity profile has to change as the
string moves, or you can think about how it comes about from the tension in the string. In
general, the net force depends on the concavity ∂ 2 y/∂x2 of the string. In this case, it’s only
nonzero at the three kinks.

(b) To keep the string in that position, we must hold it at three points. Since information can’t
travel faster than the speed of waves, only the bits of string near those three points can move
right after release, because they’re the only ones that know about the release. The direction
of motion can be found with the wave equation (the middle goes down, the ends go up).
Now, for the general solution, note that a solution to the wave equation with zero initial
velocity may be written in the form f (x + vt) + f (x − vt). Here, the function f has the same
shape as the wave, but half the height. Evidently, two traveling waves split off in opposite
directions.

[2] Problem 3. 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?

Solution. (a) The velocity is p p

v= T /µ = T L/m
and T = xmg/L where x is the distance from the bottom, so
√
v= xg.

At the most naive level, a wave pulse just travels along the string with a varying velocity, and
takes a total time s
Z L Z L Z L
dt dx 1 dx L
t= dx = =√ √ =2 .
0 dx 0 v g 0 x g
This approximation makes sense as long as the wave pulse can’t “see” that the velocity is
actually changing, which means it works if the pulse has a length much smaller than the length
of the rope itself. This idea is discussed further for the case of quantum matter waves in X1.

(b) Because the tension is higher at higher points on the rope, the top part of the pulse is always
traveling a bit faster than the bottom. Thus, the pulse gets longer as it travels up.

[3] Problem 4. [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. Your final answer should contain
an integral involving vy .)

Solution. We write y(x, t) = g(x, t) + h(x, t), where g has the correct initial position y(x) but no
initial velocity, and h has the correct initial velocity vy (x) but zero initial position. The lesson of
problem 2 was precisely that
y(x − vt) + y(x + vt)
g(x, t) = .
2

3
 Kevin Zhou Physics Olympiad Handouts

Now let’s construct h(x, t). We know it must vanish at t = 0, and that when it’s differentiated with
respect to time at t = 0, we get vy (t). So an initial guess might be
Z t
h(x, t) = vy (x − vt) dt.
0

That’s because, by the fundamental theorem of calculus, the only contribution to ∂h/∂t at t = 0 is
from the change of the upper bound of the integral, so

∂h
= vy (x − vt) = vy (x).
∂t t=0 t=0

However, this solution doesn’t work, because it doesn’t quite satisfy the wave equation. In particular,
at t = 0 we have h = 0, which implies ∂ 2 h/∂x2 = 0, which implies ∂ 2 h/∂t2 = 0 by the wave equation.
In other words, the solution for h should have no initial acceleration because the string begins flat.
But taking the second time derivative of the solution above gives something that’s certainly nonzero.
To fix this, we use exactly the same trick that was used to derive g. We superpose a solution
with dependence on x + vt rather than x − vt,

1 t
Z
h(x, t) = (vy (x − vt) + vy (x + vt)) dt.
2 0

This still has the right initial conditions, and it does solve the wave equation.
We therefore conclude that the general solution is
 Z t 
1
y(x, t) = y(x − vt) + y(x + vt) + (vy (x − vt) + vy (x + vt)) dt .
2 0

Using this solution you can show, for example, that if you poke the string and thereby change either
y or vy at one point in space, the influence of that change spreads out at speed v. In other words,
the physics is “local”: there is no way to send a signal to a distant place instantaneously. This
solution gives the basic prototype for many arguments involving locality in more advanced physics.

[4] Problem 5. Flexible strings, ropes, and chains can display some counterintuitive
p behavior. Suppose
a string carries a small traveling wave on it, moving with speed v = T /µ to the right. Then in
the frame that’s also moving with speed v to the right, the string maintains a constant shape, while
moving along this shape with speed v, like a snake.

(a) In fact, this phenomenon is extremely general. Show that if we have a flexible string loop
floating in zero gravity with any shape, then it is possible for that p
string to move along its
length while maintaining a constant shape, if its speed satisfies v = T /µ.

In the popular “string shooter” toy, a loop of string is shot through spinning wheels with high speed
v. As a result, the string seems to levitate in the air while maintaining a constant shape. This is
partially explained by part (a), but the real explanation also involves the weight and drag forces.
Assume the string experiences a drag force f per unit length, directed against its motion.

4
Kevin Zhou Physics Olympiad Handouts

A fit of the string’s profile to data is shown above, where the wheels are at the origin. The string
moves in the clockwise direction.
(b) Qualitatively, how does the tension in the string vary around the loop? In particular, find
the point P on the figure where the string has tension T = µv 2 . Also, find whether the string
has higher tension just before or just after it goes through the wheels.
(c) By taking measurements from the figure, estimate the drag to weight ratio f /(µg). (You may
have to roughly eyeball some numbers. 20% accuracy is good enough.)
(d) One interesting feature, which you can see in the video linked above, is that if you tap the
string right below the wheels, a pulse will smoothly move to the right along the bottom of the
string with a slow speed u. Explain why, and find an estimate for u. Assume the string has
speed v = 15 m/s. You’ll have to eyeball some numbers again, so expect only 20% accuracy.
Solution. (a) There are no external forces acting along the string, so the string carries a uniform
tension, due to its own motion. Now consider a small piece of the string of length dx, and
suppose the radius of curvature there is r. Then the tension on each side of the piece gives
an inward force 2
p T dθ = T dx/r, while the required centripetal force is (µ, dx)v /r. These are
equal if v = T /µ for any value of r. That is, the internal tension force always provides the
right centripetal force no matter what the string’s local shape is.
By the way, this provides the explanation of an alternative solution to the pearl necklace
problem in M3. In p that problem, it turns out that a necklace will jump off a table once it
accelerates to v = T /µ at the corner. That’s precisely because at that point, its own tension
provides all the centripetal force, so the normal force vanishes. For another example of this
trick, see problem 105 of 200 Puzzling Physics Problems.
(b) Consider the force and acceleration of a small piece ds of the string, where ds points along the
string’s direction of motion. The forces are µg ds acting downward, a drag force −f ds, and
tension forces on each side. Because the string is moving with constant speed, its acceleration
is perpendicular to ds.
Therefore, there is no net force parallel to ds, and balancing forces in that direction implies
that the difference in tension forces has to be
dT = f ds + µg dy

5
Kevin Zhou Physics Olympiad Handouts

where dy is the y-component of ds. In other words, the tension increases as one goes along
the string, and as one goes to higher elevations. For the parts of the string just before and
after the wheels, the heights are about the same, so the tension is higher just before going
through the wheels. This makes sense, as the wheels are pulling the string in.
Next, consider the forces perpendicular to ds. This contains contributions from the gravita-
tional force and the tension forces, and needs to sum up to the centripetal force. We know
from part (a) that at the point where T = µv 2 , the tension alone accounts for the centripetal
force. Then at this point, gravity can’t contribute in this direction, so ds must be vertical. So
point P is at the rightmost point on the loop.
At other points, T is a bit different from µv 2 , so the difference is made up by gravity. Working
through this will give a complicated differential equation for the shape of the string, which
was numerically solved to make the curve in the diagram above.

(c) There’s a net downward gravitational force µLg on the string. The net drag force is
Z
Fd = f ds = 0

since the string is a closed loop. So the wheels must apply a purely vertical force µLg.
Now, we can finish the problem by considering either forces or torques. Let’s consider torques
first, taking the wheels as the origin. The torque from gravity is µLgxCM , and it has to be
balanced by the torque from the drag force, which is
Z
τd = f r × ds = 2f A

where A is the area of the loop. (That is, the drag force is what lets the string stay up, even
though it doesn’t provide any net upward force!) We therefore conclude that
f xCM L
= .
µg 2A
Roughly eyeballing the values gives
f
xCM ≈ 0.8 m, A ≈ 0.6 m2 , ≈ 2.1.
µg

Alternatively, we can consider forces. (I thank Joshua Wang for pointing out this alternative
solution.) Let the tension right after the wheels be T , so that the tension right before the
wheels is T + f L. In addition, let the string exit the wheels at an angle θ1 above the horizontal,
and enter the wheels at an angle θ2 below the horizontal. Now consider a piece of string as it
passes through the wheels. The horizontal component of Newton’s second law is

µv 2 (cos θ1 + cos θ2 ) = T cos θ1 + (T + f L) cos θ2

and the vertical component is

µv 2 (sin θ1 − sin θ2 ) = T sin θ1 − (T + f L) sin θ2 + µLg

where the last term is the vertical force from the wheels. These equations are equivalent to

(µv 2 − T )(cos θ1 + cos θ2 ) = f L cos θ2

6
Kevin Zhou Physics Olympiad Handouts

and
(µv 2 − T )(sin θ1 − sin θ2 ) = µLg − f L sin θ2 .
Dividing these equations removes the unwanted dependence on v and T , and simplifying gives
f cos θ1 + cos θ2
= .
µg sin(θ1 + θ2 )
Thus, we only have to measure these two angles, which can be done relatively accurately,
accounting for the fact that the axis scales are not equal. I find that
f
θ1 ≈ 48◦ , θ2 ≈ 10◦ , ≈ 1.95.
µg
For comparison, the figure above is from this paper, with parameters f /(µg) = 1.82.

(d) At the bottom part of the string, we have T > µv 2 . Here, waves on the string move a bit
faster than the string itself, so that they can travel slowly against the string’s flow. (A similar
conclusion applies to waves made at the top: here the waves are a bit slower than the string,
so waves propagating leftward will smoothly move to the right. However, you have to take
more care at the top because the string will be moving into your finger, so you can easily get
it tangled up. Untangling the string is the most annoying part of using this toy.) Both sets of
waves end up converging at the point identified in part (c).
Let the tension right below the wheels be T = µv 2 + ∆T . By tracking the change in tension
induced by gravity and drag from the point identified in part (c), we have

∆T = f ℓ0 − µgy0

where ℓ0 is the length of string from the bottom of the wheels to the point in part (c), and y0
is that point’s y-coordinate. Eyeballing some more numbers, the fractional shift is
∆T g
= 2 ((2.0)(1.5 m) − (0.5 m)) = 0.11
µv 2 v
where I took f /(µg) = 2.0 as a rough compromise between the two answers to part (c). So
the change in velocity is roughly
1 ∆T
u≈ v ≈ 0.8 m/s.
2 µv 2

For typical speeds, drag is the dominant factor and f ∝ v 2 , so the ratio u/v remains roughly
constant as you increase the speed of the string.
For a more advanced and thorough treatment of this system, published in a top journal, see
this paper. More generally, there are a lot of tricky questions about flexible strings and chains.
People still write papers disagreeing about the explanation of the chain fountain.

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

7
 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 = ℏk 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 (i.e. energy per unit length) 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.

8
Kevin Zhou Physics Olympiad Handouts

Solution. (a) We will assume that the displacements of the string is small, and take the lowest
order terms. Using 12 mv 2 for kinetic energy of a piece moving in the transverse direction gets

1 1 p
∆K = ∆mẏ 2 = (µ∆x 1 + y ′2 )ẏ 2 ,
2 2
dK 1 p 1
= µẏ 2 1 + y ′2 ≈ µẏ 2 .
dx 2 2
For
p the potential energy, the work done on stretching the string is ∆U = T ∆ℓ where ∆ℓ =
1 + y ′2 ∆x − ∆x ≈ 21 y ′2 ∆x since the displacements are small. Thus

dU 1
= T y ′2 .
dx 2

(b) We have
dK 1 dU 1
= µA2 ω 2 sin2 (kx − ωt), = T A2 k 2 sin2 (kx − ωt).
dx 2 dx 2
Here, we can see that the total energy density is not uniform, but rather comes in “lumps”.
This is also true for electromagnetic waves.

(c) The densities are

dK 1 dU 1
= µv 2 f ′2 , = T f ′2
dx 2 dx 2
and for a wave traveling in one direction, these densities are exactly equal because v 2 = T /µ,
so the total kinetic and potential energy are equal. This is also true for ideal simple harmonic
oscillations.

(d) The total energy is Z ∞  

1 2 1 ′2
E= µẏ + T y dx.
−∞ 2 2
Taking the time derivative and applying the wave equation,
Z ∞ Z ∞
dE ′ ′
= µẏ ÿ + T ẏ y dx ∝ ẏy ′′ + ẏ ′ y ′ dx
dt −∞ −∞

where we used the wave equation in the second equality. Integrating the first term by parts,
Z ∞ ∞ Z ∞
′′ ′
ẏy dx = ẏy − ẏ ′ y ′ dx
−∞ −∞ −∞

and the boundary term vanishes by our assumptions. The remaining term is just the opposite
of the other term in dE/dt, so dE/dt = 0 as desired.

(e) Since the string was initially held steady, there is only potential energy. The amount of
potential energy is just T times the total length the string is stretched, so
p 2T h2
U = T ( 4h2 + L2 − L) ≈
L
where we used h ≪ L in the last step.

9
 Kevin Zhou Physics Olympiad Handouts

[2] Problem 7 (French 7.23). One end of a stretched string is moved transversely at constant velocity
u for a time τ , and is moved back to its starting point with velocity −u during the next interval τ .
As a result, a triangular pulse is set up on the string and moves along it with speed v. Show that
the total energy of the pulse is equal to the work done on the string, working to lowest order in u/v.
Solution. First let’s compute the energy in the pulse. Recall that the energy density is
 2 !
µ ∂y 2 T ∂y 2 T 1 ∂y 2
     
dE ∂y
u= = + = 2
+ .
dx 2 ∂t 2 ∂x 2 v ∂t ∂x

The triangular pulse has height uτ , and the two halves of it have length vτ . Thus, |∂y/∂t| = u and
|∂y/∂x| = u/v across the pulse, so the total energy is
2T τ u2
U = (2vτ )u = .
v
When lifting the string to create the pulse, the string was at an angle of u/v to first order in u/v,
so the transverse force that needed to be applied was T u/v. The distance over which this force was
applied was uτ , for a total work of T τ u2 /v. The same work was done when bringing it down with
constant velocity, so the total work done was 2T τ u2 /v, as expected.
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 8. [A] With a little vector calculus, the results above can be generalized to an arbitrary
number of dimensions. 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).

10
Kevin Zhou Physics Olympiad Handouts

(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).

Solution. (a) To do this, we need to simplify the partial derivatives with respect to x, y, and z.
We have
∂ψ ∂ψ ∂r
=
∂x ∂r ∂x
where these partial derivatives are all keeping the other spatial variables y and z constant. To
evaluate ∂r/∂x, we note that
r 2 = x2 + y 2 + z 2
and take ∂/∂x of both sides, giving
∂r
2r = 2x
∂x
from which we conclude
∂r x
= .
∂x r
Plugging this back in, we conclude
∂ψ x ∂ψ
= .
∂x r ∂r
Of course, we actually want the second spatial derivative, which is

∂2ψ 1 ∂ψ x2 ∂ψ x ∂ ∂ψ 1 − x2 /r2 ∂ψ x2 ∂ 2 ψ
= − + = + 2 2
∂x2 r ∂r r3 ∂r r ∂x ∂r r ∂r r ∂r
where we used the product rule. By similar logic for the other components, we conclude

∂2ψ 2 2 2 2 x2 + y 2 + z 2 ∂ 2 ψ ∂2ψ
   
2 3 − (x + y + z )/r ∂ψ 2 2 ∂ψ
=v + =v + .
∂t2 r ∂r r2 ∂r2 r ∂r ∂r2

Of course, a shortcut to this result would be to just look up the formula for the Laplacian in
spherical coordinates, but it’s good to go through this once in your life.

(b) We notice that

∂2 2 ∂
2
(rψ) = v (rψ)
∂t2 ∂r2
by the product rule. That is, the quantity rψ obeys the ordinary, one-dimensional wave
equation, for which we already know the general solution! We thus conclude

f (r − vt) + g(r + vt)

ψ(r, t) = .
r
If we also want the wave not to blow up at r = 0, we additionally need f (−vt) + g(vt) = 0.

2 Standing Waves

11
 Kevin Zhou Physics Olympiad Handouts

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.

[2] Problem 9. 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 angular 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.

Solution. (a) The standing wave equations are y(x, t) = A sin(kx) cos(ωt) and y(0, t) = y(L, t) =
0. Thus kL = πn, giving the angular frequencies
πvn
ωn =
L
for n ≥ 1. The waves will look like this:

(b) We can replace the sine with a cosine in the above solution (so integrating or differentiating
the above solution with respect to x gets the solutions to this problem). Thus the angular
frequencies are the same, ωn = πvn/L, and the waves look like this:

12
 Kevin Zhou Physics Olympiad Handouts

Technically, while the boundary conditions in part (a) required n ≥ 1, here we can actually
take n ≥ 0. The n = 0 solution just corresponds to the whole string being moved up or
down and staying there, with zero frequency. But this trivial solution is not typically called a
“standing wave”, so it’s conventional to say the lowest frequency is at n = 1.

(c) Let x = 0 be fixed and x = L be free. Then for y(x, t) = A sin(kx) cos(ωt), we have
kL = π(n + 1/2), so
πv
ωn = (n + 1/2)
L
for n ≥ 0. Some standing wave solutions are shown below, though note that the diagram
unfortunately doesn’t include n = 0, which is a legitimate standing wave.

[2] Problem 10. 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 9, often the
standing wave frequencies are all multiples of a single, lowest frequency. This frequency f0 is
called the fundamental, or first harmonic, while the multiple nf0 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.

13
 Kevin Zhou Physics Olympiad Handouts

[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 angular 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 angular frequencies and compare them to those you found in part (a). You’ll
have to use a calculator or computer to do this.
(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!
Solution. (a) If we assume that the solutions are y(x, t) = cos(kx) cos(ωt) where kL = πn, using
the differential equation will get
√ π 2 n2
−ω 2 = −Ak 4 , ω = A 2 .
L
(b) The fundamental frequency 2
√ is proportional to 1/L , and going up by an octave doubles this,
so we need a length of L/ 2.
(c) This is rather involved; see here for a solution.
(d) Plugging y(x, t) = cos(kx) cos(ωt) into the differential equation and setting kL = πn gives
−ω 2 = −v 2 k 2 − Ak 4
which gives solutions r
p πn Aπ 2 n2
ωn = k v 2 + Akn2 = v2 +
L L2
where the extra spacing comes from the Aπ 2 n2 /L2 term. With household equipment, you can
check that this holds for real strings, such as piano strings and guitar strings.

14
Kevin Zhou Physics Olympiad Handouts

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 ℓ is plucked, giving it a transverse displacement y, then
kinetic energy potential energy T ∆ℓ T y2
∼ λẏ 2 , ∼ ∼ 2
length length ℓ ℓ
where our expression for ∆ℓ comes from the p Pythagorean theorem. As we know from M4,
2
the ratio of the coefficients gives ω , so ωℓ ∼ T /λ. For a general wave with wavenumber k,
we would replace ℓ with 1/k above, giving ω ∝ k as expected.

Now, a rod is characterized by a Young’s modulus Y , mass density ρ, length ℓ, 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/ℓ)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.

The neutral line has radius of curvature R ∼ ℓ2 /y. Thus, the strain at the top and bottom
of the rod is of order h/R ∼ hy/ℓ2 . Plugging this in gives
s
Y h
ω∼ .
ρ ℓ2

Again, for a general wavenumber we would replace ℓ 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

15
 Kevin Zhou Physics Olympiad Handouts

Solving for the deflection gives

ρgℓ4
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
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
wave? Assume the string experiences very little damping.
Solution. (a) In order to avoid exciting a given standing wave, we should strike the piano at a
node of that wave, so that there is zero “overlap” between the wave and the strike. So we can

16
 Kevin Zhou Physics Olympiad Handouts

place it 1/7, 2/7, . . . , 6/7 of the way down the string.

(b) The midpoint of the string should be touched. This removes all harmonics that don’t have a
node at the midpoint, i.e. all odd harmonics. The remaining frequencies are even multiples of
the fundamental f0 , and since these are all multiples of 2f0 , we perceive 2f0 as the pitch.

(c) Suppose a standing wave has a spatial profile f (x) and frequency f , and to excite it we drive
at frequency f and amplitude A at a point x0 . Assuming the damping is small, then in the
steady state the spatial profile will be f (x)(A/f (x0 )), which is largest if we drive near a node
of the harmonic, not an antinode!
This may be somewhat unintuitive. The point is that driving at an antinode maximizes the
rate at which you initially put energy into the harmonic; this is what we cared about in part
(a). But driving at the node maximizes the eventual steady state amplitude, which is what
matters here. A real-world example of this is playing jump rope: your hands driving the rope

01@
are near at the nodes, because they need to move much less than the middle of the rope does.

[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.)

Solution. See the official solutions here.

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 angular
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?

Solution. (a) A piece of mass dm = σ dx dy will experience a net force from differing forces from
the sides. Consider the force on the dy side, where the force from surface tension is γ dy, and
∂z
the vertical component for small displacements is γ dy ∂x (since we are going perpendicularly

17
 Kevin Zhou Physics Olympiad Handouts

away from the dy side). To find the change in this vertical force across dx, we will take the
∂2z
differential again to get dFx = γ dy dx ∂x 2 . Adding the force from the y direction gets

∂2z
dFx + dFy = σ dx dy
∂t2
which gives the wave equation
∂2z ∂2z ∂2z
 
γ
2
= +
∂t σ ∂x2 ∂y 2
and hence a wave speed of r
γ
v= .
σ
(b) Let z(x, y, t) = sin(kx x) sin(ky y) cos(ωt). The boundary conditions give kx = πn/L and
ky = πm/L by the same logic as in problem 9. Plugging this into the wave equation the
simplifying gives
γ p πv
ω 2 = (kx2 + ky2 ), ωn,m = n2 + m2 .
σ L
Note that neither n or m can be zero, because then the standing wave√ solution just becomes
zero. So the lowest frequency corresponds to n = m = 1, where ω = 2 πv/L.

(c) When a drum is hit at the center, it primarily excites the fundamental and some of the lower
modes. When it’s hit near the edge, it doesn’t excite these modes, because it’s close to a node
for them. Higher modes are excited instead, leading to a “higher”, “thinner” sound.
[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 angular 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 angular
frequencies of allowed standing waves.

18
Kevin Zhou Physics Olympiad Handouts

(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.)

Solution. (a) The sand can sit still at the nodes, while it gets bounced away from everywhere
else. So the Chladni pattern shows the nodes. (The true story is a bit more complicated.
Very fine dust is substantially affected by the air currents created by the vibrating plate. It
turns out that this causes dust to accumulate near antinodes instead. To avoid confusion, real
demonstrations are often performed with sifted sand, which does not contain dust.)

(b) This is very similar to the result of problem 14. For concreteness, let’s put the origin at the
bottom-left of the plate. For the boundary condition to be satisfied at the bottom and left
edges of the plate, the standing waves should be proportional to cosines,

z(x, y, t) = cos(ωt) cos(kx x) cos(ky y).

For the boundary conditions to be satisfied at the opposite edges of the plate, we require
πn πm
kx = , ky =
L L
from which we conclude p πv
ωnm = n2 + m 2 .
L
(c) Because of these additional restrictions, both n and m have to be odd.
√ This means the lowest
frequency standing wave corresponds to (n, m) = (1, 1)√and ω = 2 πv/L. The next lowest
corresponds to (n, m) = (1, 3) and (3, 1) and thus ω = 10 πv/L.

(d) In this case, the Chladni pattern is a centered plus sign.

(e) Setting π/L = 1 for convenience, the standing wave profiles are

f (x, y) = cos(3x) cos(y), g(x, y) = cos(x) cos(3y).

Neither of these has 90◦ rotational symmetry, but the combinations

h± (x, y) = f (x, y) ± g(x, y)

either stay the same, or flip sign upon a 90◦ rotation. Thus, their Chladni patterns both have
90◦ rotational symmetry.
The two Chladni patterns, computed with Mathematica with the origin moved to the center
of the plate, are shown below.

19
Kevin Zhou Physics Olympiad Handouts

Remark: Plate Theory

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 4 2 2 2 2

2 2

∂4z ∂4z ∂4z
− ∝ ∇ z = ∇ ∇ z = ∂x + ∂ y ∂ x + ∂y z = + 2 + .
∂t2 ∂x4 ∂x2 ∂y 2 ∂y 4
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
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 = ℏk. Substituting this in
immediately gives the Heisenberg uncertainty principle; it fundamentally holds because one

20
Kevin Zhou Physics Olympiad Handouts

cannot get a finite wave without superposing different wavenumbers.

Alternatively, if we had worked with angular 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 angular 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
(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.

21
 Kevin Zhou Physics Olympiad Handouts

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 value of a field at the origin, at time t = 0.
Then at time t, the field 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 field 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 angular frequency ω0 .
(a) Find the wave equation for waves on this string.

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

(c) Compute the phase and group velocity for wavepackets of angular 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 = ℏω and p = ℏk.
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.
Solution. (a) There is now an additional acceleration of −ω 2 z due to the springs, so the wave
equation is
∂2z T ∂2z
= − ω02 z.
∂t2 µ ∂x2
(b) Guessing a sinusoidal solution gets
T 2
ω2 = k + ω02
µ
and the minimum possible angular frequency is ω = ω0 .

(c) The phase velocity vp = ω/k is s

ω T
vp = p
ω 2 − ω02 µ
The group velocity is dω/dk, and we have
dω T
2ω = (2k)
dk µ
so therefore s r
T 1 T ω 2 − ω02
vg = p = .
µ T /µ + ω02 /k 2 µ ω2

22
 Kevin Zhou Physics Olympiad Handouts

(d) In this case, you won’t manage to create any propagating waves. The part of the string near
you will just move up and down, following your hand, analogous to how the position of a mass
on a spring simply follows the force if the driving frequency is much lower than the resonant
frequency. Or, to say this more formally, the wave solutions of the frequency corresponding
to your hand’s driving are exponentially decaying, rather than oscillating and propagating,
formally because the solution for k is imaginary.

[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) Find the phase velocity vp of ripples with wavenumber k, up to a dimensionless constant.

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

Solution. (a) By dimensional analysis, the only possible dispersion relation is

s
γk 3
ω(k) ∝ .
ρ

This tells us that s

ω γk
vp = ∝ .
k ρ

(b) We have s
dω 3 γk
vg = ∝
dk 2 ρ
with the same constant of proportionality as in part (a), giving the desired result. So inter-
estingly, this is a case where dimensional analysis can give us a numeric prefactor!

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).

23
 Kevin Zhou Physics Olympiad Handouts

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.

Solution. Let the reflected wave again be a general wave traveling backward,

yr (x, t) = g(−kx − ωt).

The boundary condition is

∂
(f (kx − ωt) + g(−kx − ωt)) |x=0 = kf ′ (−ωt) − kg ′ (−ωt) = 0
∂x
which tells us that f = g up to a constant. Of course, that constant is just the initial height of the
string at x = 0, which we set to zero. Thus, f = g, so the wave is reflected without a sign flip.

[3] Problem 19. Consider the triangular “plucked” shape of problem 2 again, but suppose that the
string starts at rest, and its two outer corners are always held fixed.

(a) Sketch what happens after the string is released. What is the period of the motion?

(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?

(d) To further build your intuition, try to visualize what happens if the string begins with the
“pluck” off-center. Also, consider what happens if the string starts horizontal and at rest, but
it instantaneously receives an impulse at its center.

Solution. (a) The answer is shown in the early stages of this video. This generally shows how
standing waves are produced. From an initial pluck, the reflections from the ends naturally
create the oppositely-moving waves needed to form a standing wave. Since the wave becomes
inverted after bouncing off the ends, it needs
p to travel a distance of 2L before it reaches its
original state, giving a period of 2L/v = 2L µ/T .

24
 Kevin Zhou Physics Olympiad Handouts

(b) Let the string have a height h initially, so that y(x, t) = f (x − vt) + f (x + vt) where f (x) a
triangle of height h/2 and base L. Since f (x) was in the form of f (x) = x(h/2)/(L/2) = hx/L,
then ẏ = hv/L. When the string is straight, each waveform has its peak at the end with
the other half reflected, so the velocity adds up to give a velocity of 2hv/L at every point
uniformly across the string, so K = 12 (µL)(2hv/L)2 . The potential energy can be found with
U/L = 12 T y ′2 = 21 T (2h/L)2 , which gives U = 2T h2 /L = 2µv 2 h2 /L = K.

(c) Of course, the amplitude decreases over time. But also, sharp features in the string are
damped faster, due to both energy losses in the bending string, and drag with the air. As a
result, after a while, the initially sharp features get smoothed out. You can start to see this
happen in the latter part of the video linked in part (a).

(d) You can check your answers here and here.

[4] Problem 20. [A] The general, turn-the-crank method to find the time evolution of an arbitrary
wave on a string of length L is Fourier series. In this method, we write the initial shape y0 (x) of
the wave as a combination of standing waves,
X πnx
y0 (x) = cn sin , 0 ≤ x ≤ L.
n
L

We know how each standing wave oscillates in time, so by linearity, the entire wave evolves as
X πnx
y(x, t) = cn sin cos(ωn t)
n
L

where ωn is the angular frequency of the nth harmonic.

(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) Now let’s consider the plucked string considered in part (a) of problem 19. If the pluck is
centered at the middle of the string and has height h, find the coefficients cn . (If you’re so
inclined, you can use a computer to see how the resulting y(x, t) approaches the answer to
problem 19 as more terms are included.)

(c) Argue that in general, we have

Z L
LX
y02 (x) dx = |cn |2 .
0 2 n

By applying this result to the plucked string, 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).

25
Kevin Zhou Physics Olympiad Handouts

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

Solution. (a) This works because all the other terms in y0 (x) will cancel out, as seen here:
Z L Z L
πnx πmx 1  πx πx 
dx sin sin = dx cos (m − n) − cos (m + n)
0 L L 2 0 L L

which will always equal to 0 when m ̸= n, since the arguments of sine (the anti-derivative of
cosine) will always be an integer multiple of π. To find the constant of proportionality, we
only need to look at the nth term of the expansion of y0 (x):
Z L Z L
πnx πnx
dx y0 (x) sin = cn dx sin2 = cn L/2
0 L 0 L

Thus the constant of proportionality is 2/L.

(b) The equation of the plucked string is

y0 (x) = h − 2h|x − L/2|/L, 0 < x < L.

By symmetry, if n is even, cn = 0 since y0 (x) is even about x = L/2 and sin(πnx/L) is odd
about x = L/2, so the integral will be 0. Thus we will consider the case where n is odd,
Z L/2 Z L
2 2hx  πnx  2  πnx 
cn = sin dx + (2h − 2hx/L) sin dx
L 0 L L L L/2 L

Using symmetry again, the two integrals above are equal, so we only need to evaluate the
first. We have
Z L/2 !
4h L 
Z
4h L  πnx  L/2 L  πnx  x  πnx 
cn = 2 − x cos + cos dx + 1− sin dx
L πn L 0 πn 0 L L L/2 L L

Considering only odd n, the first term will vanish since cos(πn/2) = 0 for odd n. Also, by
symmetry the integral that goes from L/2 to L should be equal to the one that goes from 0
to L/2. Thus, for odd n,
8h  πn 
cn = 2 2 sin
π n 2
while cn = 0 for even n.

(c) Again, integrating sinusoids with different values of n will get 0, and the same value will get
L/2. Thus when representing y0 (x) as a sum of sinusoids and having the integral of all the
cross terms go to 0, we get that
Z L Z L Z L
X πnx X 2 πnx LX 2
y0 (x) dx = dx c2n sin2 = cn dx sin2 = c .
0 0 n
L n 0 L 2 n n

Using y0 (x) = h − 2h|x − L/2|/L, and symmetry about x = L/2, we get

L L/2 L/2
4h2 x2
Z Z Z
1
y02 (x) dx =2 y02 (x) dx =2 2
dx = h2 L.
0 0 0 L 3

26
 Kevin Zhou Physics Olympiad Handouts

To find the sum of cn , we use our answer above and consider the odd terms:
LX LX 64h2 32h2 L X 1
|cn |2 = 4 4
= 4
2 n 2 n π (2n + 1) π n
(2n + 1)4

To relate that sum to ζ(4), define the sums for the even and odd numbers as E and O so that
X 1 1 15
ζ(4) = E + O, E= = ζ(4), O = ζ(4) − E = ζ(4)
n
(2n)4 16 16

Now equating our expressions will get

1 2 32h2 L 15
h L= ζ(4)
3 π 4 16
from which we conclude
π4
ζ(4) = .
90
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
′
yt (x, t) = tei(k x−ωt)
where k ′ 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 ̸= 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 ′ 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.

27
Kevin Zhou Physics Olympiad Handouts

(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.)

Solution. (a) By continuity of the frequency, we have

p
′ v1 T1 /µ1
k = k=p k.
v2 T2 /µ2

The boundary conditions are continuity of the string, and continuity of the transverse compo-
nent of the tension at x = 0, so that the forces on the ring balance.

(b) Continuity of the string requires

1 + r = t.
Continuity of the transverse component of tension requires

T1 (k − rk) = T2 tk ′ .

Combining this with the result of part (a) gives

√
T2 µ2 Z2
1−r = √ t= t.
T1 µ1 Z1

Combining this with the continuity condition and solving gives the desired results.

(c) A hard boundary can be modeled by setting T1 = T2 and µ2 → ∞, which is equivalent to

Z2 /Z1 → ∞. In this limit, r = −1 and t = 0 as expected.
A soft boundary can be modeled by setting T1 = T2 and µ2 → 0, which sets Z2 /Z1 → 0. In
this case we have r = 1 as expected. Oddly we also have t = 2, but this isn’t really physical
because in a soft boundary, the string at x > 0 doesn’t exist. (It makes no difference from the
standpoint of the reflection coefficient whether or not the string at x > 0 exists, because the
wave carries no energy in the limit µ2 → 0.)

(d) First, we need to find the energy for a wave of given amplitude. The kinetic and potential
energies in a wave are equal, so we can look at either. The potential energy per unit length is
proportional to T y ′2 ∝ T (Ak)2 where A is the amplitude and k is the wavenumber. Since the
waves in this problem have fixed frequency, and v = ω/k, we have k ∝ 1/v. Finally, the total
duration T of the wave is fixed, meaning the total length is L = vτ ∝ v. Combining these
results,
T
U ∝ LT A2 k 2 ∝ vT A2 /v 2 = A2 = ZA2 .
v
In other words, the impedance determines the energy per unit time in a wave pulse of given
amplitude. That gives some intuition for why the transmission of energy is perfect when
Z1 = Z2 . If you want there to be no reflection, then the amplitudes of the transmitted and
incoming waves have to match by continuity. This is only consistent with energy conservation
if the impedances are matched too.

28
Kevin Zhou Physics Olympiad Handouts

Therefore, the statement of energy conservation is

Z1 = Z1 |r|2 + Z2 |t|2 .

Plugging in the expressions above and doing the algebra confirms this.
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,
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.

29
 Kevin Zhou Physics Olympiad Handouts

[3] Problem 22. Consider a symmetric, thin mirror with air on both sides. When an electromagnetic
wave whose electric field has amplitude A hits the mirror from either side, there is a transmitted
wave of amplitude tA and a reflected wave of amplitude rA. Here, t and r are generally complex
numbers, with their phase determining the phase shift of the transmitted or reflected wave. Assume
that no energy is absorbed in the mirror itself.

(a) Suppose light hits the mirror from one side. Using energy conservation, show that |r|2 +|t|2 = 1.

(b) By considering a situation where light hits the mirror from both sides, show that r∗ t + t∗ r = 0.
Show that this means r and t differ in phase by π/2.

(c) Alternatively, one can argue that the amplitude of the electric field must be continuous at
the mirror. (This isn’t quite right, as the charges in the mirror will make a big electric
field confined to the mirror, but it’s reasonable for other situations, such as problem 25, and
happens to give the right answer here.) Using this assumption, show again that r∗ t + t∗ r = 0.

(d) Why don’t these results apply to the situation in problem 21?

(e) ⋆ What can we say about an asymmetric mirror, which has coefficients r and t from one side,
and r′ and t′ from the other? Continue to assume that no energy is absorbed in the mirror.

Solution. (a) The energy contained in an electromagnetic wave is proportional to its squared
amplitude. If we send in a pulse of some duration with amplitude A, then after some time,
we will get a transmitted pulse with amplitude tA and a reflected pulse with amplitude rA,
and energy conservation implies A2 = (|r|2 + |t|2 )A2 , giving the result.

(b) Suppose that light pulses of the same amplitude A and duration hit the mirror from both
sides at the same time. After some time, we will get two outgoing pulses with amplitude
(r + t)A. Then energy conservation gives |r + t|2 = 1. Expanding this out and using part (a)
gives the result.
As for the phase difference of π/2, it’s easiest to see this with phasors. The facts that |r + t|2
and |r|2 + |t|2 = 1 mean that, when r and t are added up in the complex plane, they form the
sides of a right triangle.

(c) For a single incident wave, the amplitude from the incident side is (1+r)A, while the amplitude
on the other side is tA. So by continuity, we have 1 + r = t, which is equivalent to 1 = t − r.
Taking the squared magnitude of both sides recovers the answer to part (b).

(d) Here we’ve assumed that the light travels in the same medium on both sides of the mirror; if this
weren’t true, then the energy in terms of the amplitude would have a different proportionality
constant. But in problem 21, there are different media (strings of different thicknesses) on each
side. On the other hand, if we threaded a mass through a single string, so that the properties
of the string were the same on each side of it, then the mechanical waves transmitted or
reflected off this mass would have amplitudes obeying the identities derived here.

(e) Clearly, repeating the argument of part (a) will give

|r|2 + |t|2 = |r′ |2 + |t′ |2 = 1.

30
 Kevin Zhou Physics Olympiad Handouts

Now, by repeating the argument of part (b), but allowing the wave from the right side to start
with an arbitrary phase shift θ, we have

|r + eiθ t′ |2 + |eiθ r′ + t|2 = 2.

Expanding this out and using our first result gives

eiθ (r∗ t′ + r′ t∗ ) + e−iθ (r∗ t′ + r′ t∗ )∗ = 0.

Since θ is arbitrary, this can only be true if

r∗ t′ + r′ t∗ = 0.

This is the final answer, but we can interpret it a bit better by rearranging it to
 ′ ∗
r r
=− ′ .
t t
By comparing the magnitudes of each side, we conclude that

|r| = |r′ |, |t| = |t′ |.

This makes perfect sense, as it says that the mirror cannot be “one-way”: the amount of
energy transmitted when passing through each direction has to be the same. (We’ve argued
in T2 that if this wasn’t true, one could violate the second law of thermodynamics.) Finally,
let θ = arg(r/t) be the difference in phase shifts on reflection and transmittion from one side,
and let θ′ = arg(r′ /t′ ) be the corresponding quantity for the other side. Then by comparing
the phases of each side above, we have θ = −θ′ + π, i.e. that

θ + θ′ = π.

The arguments here are very simple, but were missed for a surprisingly long time. If you
figured it out a few decades ago, you could have published it and gotten a hundred citations.

[4] Problem 23. Consider two identical, thin, symmetric mirrors, with reflection and transmission
coefficients r and t, 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
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 L, 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.

For the rest of the problem, assume that L takes one of the special values found in part (c).

31
Kevin Zhou Physics Olympiad Handouts

(d) Using energy conservation, recover the result of problem 22(b).

(e) Suppose that a laser with power P has been fired at the interferometer for a long time. Then,
at a certain moment, the laser is suddenly switched off. Find the total energy of the light
that travels from the interferometer back towards the laser after the laser is switched off. For
simplicity, suppose that |t| ≪ 1, and give your answer in terms of P , L, |t|, and c.

(f) Estimate the duration of the light pulse that travels back towards the laser.

Solution. (a) Light can be immediately reflected from the first mirror. It can also go in between
the mirrors and be reflected any number of times before leaving through either mirror.

(b) For a wave to be transmitted, it must be transmitted through one mirror, reflected 2n times,
then travel a distance of L and get transmitted out the other mirror. For each intermediate
reflection, its amplitude gets a factor of α = reikL . Then we have

t2 eikL
tnet = t2 eikL 1 + α2 + α4 + . . . =


1 − r2 e2ikL
as desired. As for reflection, we have

rt2 e2ikL
rnet = r + t2 eikL α + α3 + α5 + . . . = r +


1 − r2 e2ikL

(c) The fraction of the energy transmitted is

|t|2
T = |tnet |2 = .
|1 − r2 e2ikL |

This is maximized when r2 e2ikL is real and positive, so that it is equal to |r|2 , giving

|t|2
T = = 1.
1 − |r|2

This is a striking result: you can put two nearly perfect mirrors next to each other, and light
of the right color will still go right through. This is because the light that goes go through the
first can bounce around inside many times, eventually completely canceling the zeroth order
reflected wave. This phenomenon is called resonant transmission.

(d) In this case, we have |tnet |2 = 1, so energy conservation implies rnet = 0, which means

(t2 /r)r2 e2ikL t2 |r|2 /r t2 |r|2 /r

0=r+ = r + = r + .
1 − r2 e2ikL 1 − |r|2 |t|2

This is equivalent to
r2 t2
2
=− 2
|r| |t|
so that the phases of r2 and t2 differ by π, so the phases of r and t differ by π/2.

32
 Kevin Zhou Physics Olympiad Handouts

(e) We’ve established that, in the steady state, the total reflected wave has zero amplitude. Let
∆t = 2L/c be the time for one round trip within the interferometer. For the first ∆t, the total
reflected wave is “missing” the direct reflected wave, so it has an amplitude of magnitude r.
For the next ∆t, it is missing both the direct reflected wave and the one involving one trip
within the interferometer, and so on.
To make this more quantitative, note that |r| = reikL , so that

rnet = r + t2 eikL (|r| + |r|3 + |r|5 + . . .) = e−ikL |r| + (|r|3 − |r|) + (|r|5 − |r|3 ) + . . . .

So for the first ∆t, the reflected amplitude has magnitude |r|, and for the next ∆t, it has
magnitude |r|3 , and so on. Then we have

2P L |r|2 PL
Er = (P ∆t)(|r|2 + |r|6 + |r|10 + . . .) = 4
≈ 2
c 1 − |r| |t| c
where we used |t| ≪ 1.

(f) Every time ∆t, the reflected pulse weakens by a factor of |r|4 , where 1 − |r|4 ≈ 2|t|2 . Then
the timescale of decay is roughly T ∼ ∆t/|t|2 ∼ L/(|t|2 c).

01W
This question is essentially the same as EuPhO 2024, problem 3.

[3] Problem 24. USAPhO 2004, problem A3.

[3] Problem 25 (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

33
Kevin Zhou Physics Olympiad Handouts

monochromatic light varies from 610 nm to 660 nm, for what wavelengths is all the light
energy directed into fiber 2?

Solution. (a) This is a case where the heuristic argument of problem 22(c) works, because there’s
nothing but vacuum at the splitting point. Let the ingoing electric field amplitude be Ein ,
and let the outgoing field amplitudes be E1 and E2 . Then

Ein = E1 + E2 , |Ein |2 = |E1 |2 + |E2 |2

from continuity of the electric field, and energy conservation. Thus, by the Pythagorean
theorem, E1 and E2 must differ in phase by π/2. For the equal ratio splitter relevant to this
problem, one of them is advanced in phase by π/4, while the other is delayed in phase by π/4.

(b) Consider sending in waves with the same phase and equal amplitude Ein along both fibers 1
and 2 simultaneously. If the wave that exits along fiber 1 is always advanced, then the final
amplitudes are
E0 √ E0 √
E1 = (eiπ/4 + eiπ/4 ) √ = 2eiπ/4 E0 , E2 = (e−iπ/4 + e−iπ/4 ) √ = 2e−iπ/4 E0 .
2 2
On the other hand, if a wave that enters along fiber 2 exits along fiber 1 retarded instead, the
final amplitudes are
E0 E0
E1 = (eiπ/4 + e−iπ/4 ) √ = E0 , E2 = (e−iπ/4 + eiπ/4 ) √ = E0 .
2 2

Only the second option respects energy conservation, |E1 |2 + |E2 |2 = 2|E0 |2 , so that one must
occur. In other words, we have:

1 → 1, 2 → 2 : advanced, 1 → 2, 2 → 1 : retarded

(c) Let’s consider the two components of the waves that eventually exit along fiber 1.

• Part of the incident wave stays in fiber 1 at the first splitter, getting advanced by π/4. It
picks up some phase between the two splitters, then gets advanced by π/4 again at the
second splitter.
• Part of the incident wave goes into fiber 2 at the first splitter, getting retarded by π/4.
It picks up some phase between the two splitters, then (by the result of part (b)) gets
retarded by π/4 again at the second splitter.

For all the light to come out along fiber 2, these two components that come out along fiber 1
have to cancel out. That means they need opposite phases, which implies

π/4 + kℓ + π/4 − (−π/4 + k(ℓ + ∆ℓ) − π/4) = (2n + 1)π.

This simplifies to k∆ℓ = 2πn, or nλ = ∆ℓ = 30 µm, from which we conclude

n ∈ {46, 47, 48, 49}, λ ∈ {612, 625, 638, 652} nm.

34
 Kevin Zhou Physics Olympiad Handouts

Remark
Above, we’ve focused on cases where light can only exit a given optical element in two
ways. But in general, you could have n “ports”, in which case you would need an
entire n × n matrix of coefficients S to relate the n input amplitudes to the n output
amplitudes. Generalizing problem 22 to this case shows that S is a unitary matrix, S † = S −1 .

All of the optical elements we’ve seen so far obey optical reciprocity, i.e. the principle that
“if I can see you, then you can see me”, related to time reversal symmetry. However, we can
violate reciprocity by using special materials, such as permanent magnets. For example, a
circulator is an optical element with 3 ports, so that all light entering port 1 exits from port
2, light in port 2 exits from port 3, and light in port 3 exits from port 1. Though it’s strange,
you can check that there’s no way to use it to violate the second law of thermodynamics.

[3] Problem 26. The Sagnac effect is a phase shift observed when an interferometer rotates. Originally
it was used as a test of special relativity, and today it is used to make sensitive gyroscopes.
To illustrate the effect, consider a thin ring of radius R, rotating uniformly about its axis of
symmetry with a small angular velocity Ω. Light of angular frequency ω is inserted at a point P
on the ring, and travels both clockwise and counterclockwise along it. When the two light beams
arrive at point P again, after a full revolution, their phase is compared.

(a) If the ring is hollow, what is the phase difference between the beams when they meet?

(b) What if the ring contains a material with index of refraction n?

To solve this problem, you will need prior exposure to R1.

Solution. (a) There are several ways to set up the calculation. One could try to work in the
reference frame rotating with the ring, but this generally is a bad idea, since noninertial frames
have very confusing behavior in special relativity. One could also work in a series of reference
frames which momentarily move with a light beam as it traverses a point on the ring, but this
is a bit clunky. It turns out to be easiest just to stay in the lab frame.
In the absence of rotation, each beam needs a time ∆t = 2πR/c to arrive back at point P .
With rotation, the relative speed of the ring and light beams is c ± ΩR in the lab frame, so
the beams need time ∆t± = 2πR/(c ± ΩR). The phase shift is
4πωΩR2
∆ϕ = ω|∆t+ − ∆t− | ≈
c2
where we used ΩR ≪ c. (Note that the result is proportional to the area of the ring; it can
be shown that this remains true for an arbitrary ring shape.)

(b) In the rest frame of such a material, light travels with speed c/n. To find the speed in the lab
frame, we need to perform relativistic velocity addition, which gives
 
(c/n) ∓ ΩR c 1
v± = ≈ ∓ ΩR 1 − 2 .
1 ∓ ΩR/nc n n
We derived this same result back in R1. The beams now need a time
2πR 2πRn 1 2πnR 2πΩR2
∆t± = = ≈ ∓
v± ± ΩR c 1 ± ΩR/nc c c2

35
Kevin Zhou Physics Olympiad Handouts

which gives a phase shift of

4πωΩR2
∆ϕ ≈
c2
exactly as in part (a). That is, the value of n drops out! (This problem also appeared on
the 2003 APhO, but as pointed out by Stefan Ivanov here, the official solutions are incorrect
because they didn’t use relativistic velocity addition.)
Interestingly, you can also get the result of part (b) in nonrelativistic physics, if you make appropriate
assumptions about the ether. This led to a lot of historical confusion; however, it turns out that you
can’t explain both the Sagnac effect and the Michelson–Morley experiment with ether simultaneously,
as they require different assumptions about how the ether gets “dragged”.

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 details.

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 your wave
pulse passing by. 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 reflected pulse of equal energy, so
conservation of energy is still satisfied because E + 0 = E.

We can also try to route around this issue. 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

36
Kevin Zhou Physics Olympiad Handouts

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 this 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.

37