Kevin Zhou Physics Olympiad Handouts

Thermodynamics III
For more about surface tension, which can be quite tricky, see section 9.3 of Wang and Ricardo,
volume 1, or Kalda’s thermodynamics handout, which also covers humidity. Phase transitions
are covered clearly in section 4.5 of Wang and Ricardo, volume 2. For more detail, chapter 17 of
Blundell and Blundell covers various types of thermodynamic work, chapter 26 covers liquid-gas
phase transitions, and chapter 28 covers phase transitions in general. There is a total of 82 points.

1 Surface Tension
Thermodynamics applies to many systems that aren’t ideal gases, or even gases at all; in such
systems the work is not necessarily d̄W = −P dV . The most important example is surface tension,
which we saw in M2 and M7. We begin with the microscopic origin of surface tension.
Idea 1
For a liquid surface in air, there is an associated energy γA where A is the area of the surface.
This leads to a contribution to the work

d̄W = γ dA.

The surface tension γ is also the force per length exerted along the surface.

The energy γA comes from the fact that liquid molecules at the surface are “missing” neigh-
bors, and hence cannot lower their energy as much by forming cohesive bonds. (Technically,
the same is true for the air molecules too, but air is very sparse compared to liquid, so we
just ignore it.)

[3] Problem 1. Here, we use the above idea to very roughly estimate the surface tension of water.

(a) Estimate the spacing between water molecules. (Hint: you could use known atomic distance
scales, or reverse engineer this from the known density of water.)

(b) Estimate the energy of a hydrogen bond. (Hint: the energy of any kind of chemical bond will
be close enough.)

(c) Using these results, estimate the surface tension of water, and compare this to actual value
γ = 0.073 N/m.

(d) Estimate the typical height of a droplet of water on a flat surface.

Next, we consider problems that combine surface tension with ideas in thermodynamics.
[3] Problem 2 (PPP 62). Two soap bubbles of radii R1 and R2 are joined by a straw. Air goes from
one bubble to the other and a single bubble of radius R3 is formed.
(a) If R1 < R2 , which bubble loses air and which bubble gains it?

(b) Assuming that γ is small and the entire process is isothermal, find an expression for R3 in
terms of R1 , R2 , and the atmospheric pressure P .

1
 Kevin Zhou Physics Olympiad Handouts

(c) For typical soap bubbles, is this a practical way to measure surface tension? Why or why not?

[3] Problem 3 (Grad). A tightly closed jar is completely filled with water. At the bottom of the jar
are two small air bubbles. The pressure at the top of the jar is P0 , the radius of each bubble is R0 ,
and the surface tension is γ. The two bubbles then merge isothermally. Calculate the new pressure

01W
at the top of the jar.

[3] Problem 4. USAPhO 2007, problem A3.

[3] Problem 5. 01^‚ IPhO 2014, problem 1B.

Idea 2
One can also have liquid, solid, and air in the same problem, which leads to some complications.
Let Al and As be the surface areas of the liquid and solid exposed to the air, and Asl be the
surface area of the liquid-solid interface. Then there are three terms in the work,

d̄W = γl dAl + γs dAs + γsl dAsl .

In other words, there are three surface tensions, one associated with each kind of interface.

Both γl and γs arise from the fact that cohesive liquid-liquid or solid-solid bonds are broken
to form a surface. However, γsl is determined by the adhesive forces between the liquid and
solid, which may lead to a positive or negative contribution to the energy.

Specifically, let’s define the energy of adhesion Usl to be the work needed, per area, to separate
a liquid from a solid, thereby turning a liquid-solid interface into a liquid-air and solid-air
interface. By the definitions above,

Usl = γs + γl − γsl .

Now, Usl can be computed in terms of microscopic chemical bond energies, like γs and γl ,
so this result can also be thought of as an microscopic definition of γsl . When a liquid is in
contact with a solid, the solid exerts a force per length of Usl on the boundary of the liquid,
along the solid.

Example 1

The surface of a drop of water makes a contact angle θ with a solid, as shown.

When θ is acute, the surface is said to be hydrophilic. If θ is obtuse, it is hydrophobic. Find

an expression for θ in terms of the relevant surface tensions.

2
Kevin Zhou Physics Olympiad Handouts

Solution
If the liquid drop expands outward by δx, the areas of various surfaces change, as shown.

The change in energy is

dU ∝ γsl δx + γl cos θ δx − γs δx
and this must be equal to zero in equilibrium. Thus,
γs − γsl Usl
cos θ = = − 1.
γl γl
This is Young’s equation. The liquid surface tension γl must be positive; otherwise the
liquid could not exist stably at all, but rather would disperse into gas. Thus, the surface is
hydrophilic when Usl > γl and hydrophobic when Usl < γl .

As extreme cases, note that there is no solution for θ when Usl > 2γl . In this limit, the
surface is so hydrophilic that the liquid spreads out and coats the entire solid; this is
known as perfect wetting. There is also no solution when Usl < 0, in which case the liquid
disperses into many tiny nearly spherical drops, each with a tiny area of contact with the solid.

This derivation was in terms of energy, which is typically easier for surface tension. The
same result can be derived in terms of forces, but it’s more subtle than it looks; the standard
derivation in textbooks is wrong. For a clear derivation, see section 9.3 of Wang and Ricardo.

Example 2

A very thin, hollow glass tube of radius r is dipped vertically inside a container of water.

Find the height to which water can climb in the tube.

3
Kevin Zhou Physics Olympiad Handouts

Solution
We first encountered this problem in M7, where we solved it by using Pascal’s principle,
giving an answer in terms of the contact angle. The derivation above of the contact angle
completes this solution. However, we can also solve the problem using energy or force.

In terms of energy, if we move the height of the water up by δh, then

dU = ρπr2 gh δh + (γsl − γs )2πr δh = 0

and solving gives

2(γs − γsl ) 2γl cos θ
h= =
ρgr ρgr
using Young’s equation. Alternatively, in terms of force, consider the vertical forces acting
on the column of water inside the tube. There is an upward force of adhesion from the solid
wall of 2πrUsl , and a downward surface tension force from the liquid below of 2πrγl . Then

F = 2πr(Usl − γl ) − ρπr2 gh = 0

which yields precisely the same result.

Example 3

Fill a dish with water, and sprinkle something small over it, such as ground pepper. If you
place a drop of detergent in the middle of the dish, then the pepper will “flee” away to the
edges. Why does this happen?

Solution
Detergent is a surfactant, meaning that it decreases the surface tension of water. When one
places the detergent in the middle of the dish, it diffuses outward, making the surface tension
temporarily higher near edges of the dish. This leads to an unbalanced surface tension force
on the pepper grains, pulling them to the edges.

This phenomenon is called the Marangoni effect. Of course, the force vanishes once the
detergent becomes uniform distributed, and the surface tension is uniform again.

Remark
Here’s a neat fact: the number of atoms that fit into a drop of water is comparable to the
number of drops of water that fit inside the tallest mountains. We can show this using rough
estimates, in the style of P1. Let Eb be the energy of a typical chemical bond, let m be the
mass of an atom, and let d be the typical distance between atoms.

The size ` of a droplet of water, such as one that drips from a leaky ceiling, is the size where
surface tension forces balance gravitational ones. By dimensional analysis, we must have
p
` ∼ γ/ρg

4
Kevin Zhou Physics Olympiad Handouts

as we showed in M7. Now, ρ ∼ m/d3 , and the logic of problem 1 implies γ ∼ Eb /d2 , so
p
` ∼ Eb d/mg.

Now consider the height H of the tallest mountains. The height of mountains is limited by
the rigidity of rock; if the pressure is too great, then the rock underneath the mountain will
deform, causing it to sink into the ground. Let’s consider an atom-thick column of this rock.
If it sunk down by a distance d, then the gravitational potential energy harvested would be
mgH. However, the atom at the bottom would have to break its chemical bonds with its
horizontal neighbors, which takes energy Eb . Balancing these gives a maximum height

H ∼ Eb /mg.

We have therefore shown that √

`∼ Hd
which implies the original statement, within a few orders of magnitude.

2 Melting, Freezing, Boiling, Evaporation, and Condensation

Idea 3
A phase transition is a sudden, dramatic change in a system as thermodynamic variables
such as the temperature are varied. Most of the ones you’ll see have a latent heat

Q = mL.

For example, if ice is heated up, its temperature will gradually increase until it hits 0 ◦ C. At
that point, the temperature will remain constant until all of the ice is melted, i.e. when the
full latent heat has been supplied.

Remark
We can roughly estimate the latent heats of melting and evaporation. In general, the latent
heat can go into either in breaking molecular bonds, or increasing the entropy.

When a solid melts into a liquid, the molecules stay right next to each other, so changing
bond energy isn’t the dominant effect. Instead, it’s the increase in entropy as the liquid
molecules become free to rotate. Let’s suppose that the molecules each gain a few extra
possible quantum states. This corresponds to an entropy increase per molecule ∆S ∼ kB ,
which means a latent heat per mass of
T ∆S kB T RT
L= ∼ =
mmol mmol µ
where µ is the molar mass, or equivalently a latent heat per mole L ∼ RT . For water, we
get L ∼ 2 × 105 J/kg, which is of the same order of magnitude as the true value 3.3 × 105 J/kg.

5
 Kevin Zhou Physics Olympiad Handouts

When a liquid becomes a gas, the dominant effect is typically the huge increase in entropy
kB log(Vgas /Vliq ) per molecule because they get much more space to move. The ratio inside
the logarithm is huge, which means that while the volumes per molecule Vgas and Vliq vary by
order-one amounts between phase transitions, the logarithm of their ratio is always around
the same value, which turns out to be about 10. This gives a latent heat per mass of
10kB T 10RT
L∼ = .
mmol µ
This result is called Trouton’s rule, and it is surprisingly accurate for most liquids. However,
the latent heat of vaporization for water is noticeably higher, L = 2.26 × 106 J/kg. This is
because of the extra energy needed to break hydrogen bonds.

[3] Problem 6. The temperature T at which a phase transition happens depends on the pressure P ,
yielding a “coexistence curve” P (T ) where the two phases can be in equilibrium with each other.
The exact relationship is given by the Clausius–Clapeyron equation
dP L
=
dT T (V2 − V1 )

where L is the total latent heat for some amount of material, and V2 and V1 are the corresponding
volumes of that material when it is in each of the phases. (Depending on convention, L could be
the latent heat per mole, in which case the Vi are volumes per mole, or both quantities could be per
unit mass, in which case the Vi become densities.) In this problem, you will derive this equation.

(a) Consider an infinitesimal Carnot cycle operating between temperatures T and T + dT , and
pressures P and P + dP , chosen so that the isothermal heating and cooling steps involve
supplying latent heat. Compute the work done by the cycle.

(b) Argue that we may ignore all heat transfer except for the latent heat.

(c) Derive the Clausius–Clapeyron equation by setting the efficiency equal to the Carnot efficiency.

[3] Problem 7. [A] In this exercise you’ll find a quicker, more advanced derivation of the Clausius–
Clapeyron equation.

(a) The Gibbs free energy is defined as G = U + P V − T S. Show that for reversible processes,

dG = V dP − S dT.

Two phases can only be in thermodynamic equilibrium if they have the same Gibbs free energy
per molecule. Otherwise, turning one phase to the other would reduce the Gibbs free energy,
which turns out to be equivalent to increasing the entropy of the universe. (For more details,
see section 16.5 of Blundell and Blundell.)

(b) Suppose that the Gibbs free energies per molecule G/N for two phases are equal at temperature
T0 and pressure P0 . Derive the Clausius–Clapeyron equation by demanding this is also true
at temperature T0 + dT and P0 + dP .

6
 Kevin Zhou Physics Olympiad Handouts

Remark: Thermodynamic Potentials

You might sometimes see the Clausius–Clapeyron equation written in terms of a difference in
enthalpy ∆H rather than a latent heat. The enthalpy is the state function H = U + P V , so

dH = V dP + d̄Q.

This is useful because many lab experiments happen at constant pressure, dP = 0, leaving
dH = d̄Q. That is, only heat changes the enthalpy, so the latent heat of a phase transition
must be the difference in enthalpies of the two phases, L = ∆H. That in turn is useful
because enthalpy is a state function, so given a new phase transition you can calculate L by
just looking up the enthalpy values for each of the phases in a table.

We’ve now covered all the classic “thermodynamic potentials”. As we just saw, the
enthalpy H is useful for bookkeeping heat. As we saw in T2, the Helmholtz free en-
ergy F is minimized in thermodynamic equilibrium, given constant temperature and
volume. (This is the relative of the statement that the system’s internal energy U is
minimized in equilibrium, given constant entropy and volume, which is just the usual
statement of mechanical equilibrium.) And as we saw in the problem above, the Gibbs free en-
ergy G is minimized in thermodynamic equilibrium, given constant temperature and pressure.

More generally, what’s going on is that the number of possibly useful potentials doubles
every time we add another pair of “thermodynamic conjugate variables”. Before learning
about thermodynamics, we just had U . When we learned about temperature and entropy,
we additionally cared about F . And now upon accounting for pressure and volume, we have
H and G. If we had another pair, such as magnetization and external magnetic field, we
could define 4 more potentials, which would each be useful in different situations.

[2] Problem 8. Ice skaters can move with little friction because they actually glide on a thin layer
of water. Estimate how heavy an ice skater has to be to melt ice by just standing on their skates,
assuming the ice is at temperature −5 ◦ C.
Now we focus on the specifics of liquid-gas phase transitions.
[3] Problem 9. Suppose that at pressure P0 , a liquid-gas phase transition takes place at temperature
T0 . Assume the gas obeys the ideal gas law, and neglect the volume of the liquid.
(a) Assuming the latent heat is temperature-independent, compute the coexistence curve P (T ).

(b) In reality, the latent heat has a (relatively mild) dependence on temperature, changing the
results. As a crude model, suppose that the latent heat per molecule for a monatomic liquid-
gas phase transition has three components: the energy E0 required to break the bonds with
other molecules in the liquid, the kinetic energy needed to be in equilibrium with the rest
of the gas, and the P dV work that must be done to “push” the rest of the gas away, since
the new gas molecule takes up space. Under these assumptions, what is the latent heat per
molecule, and qualitatively how does the dependence P (T ) change?

(c) A closed container of constant volume contains both liquid and gas in equilibrium, at temper-
ature T . Let the latent heat of vaporization per mole be L, and neglect the volume of the

7
 Kevin Zhou Physics Olympiad Handouts

liquid. If the temperature is increased by a very small amount ∆T , by what factor does the
number of moles in gas form change?

Example 4: APhO 2004.3

A cylinder is divided into an upper and lower part by a mobile partition, which is free to
move and conducts heat well. One compartment contains one mole of water vapor, and the
other contains one mole of nitrogen gas. Initially, the system has temperature T = 373 K, the
pressure in both compartments is 0.5 atm, and the compartments have equal volume V0 . A
piston is then slowly lowered, compressing the system isothermally. Sketch the P (V ) curve.

Solution
The answer is shown below.

Initially, we just have an ordinary isothermal compression. Both the water vapor and nitrogen
gas compartments are compressed at the same rate, since they must have equal pressures and
temperatures. When the total volume is halved, the pressure in both reaches one atmosphere.
Now, water condenses at temperature 373 K at pressure p = 1 atm. Thus, as the volume
continues to decrease, the pressure stays constant, the nitrogen compartment’s volume stays
the same, and the water compartment shrinks, as the vapor gradually condenses to liquid.
This process completes once all the vapor is condensed, which is roughly when the total
volume has halved again. After this point, we again have ordinary isothermal compression,
of the nitrogen gas alone.

01^‚
[3] Problem 10. NBPhO 2016, day 2, problem 2. A problem on phase transitions with data analysis.

[3] Problem 11. USAPhO 2015, problem A4. A heat engine with phase transitions.
In practice, water on Earth is more subtle because there are three substances at play: liquid water,
water vapor, and the rest of the atmosphere.
Idea 4: Humidity

Consider a box at constant temperature T containing only water, and let P (T ) be the liquid
water-water vapor coexistence curve. In equilibrium, if we apply any pressure below P (T ),
all of the water will be in vapor form, and if we apply any pressure above P (T ), all of the
water will be in liquid form. Physically, applying a higher pressure forces the water vapor to
condense into liquid, as it packs the molecules closer together, and applying a lower pressure

8
Kevin Zhou Physics Olympiad Handouts

forces the liquid to evaporate into vapor, as it cannot hold itself together against the thermal
motion of the molecules. Coexistence is impossible, except at exactly the pressure P (T ).

However, in everyday life, the two easily coexist over a wide range of pressures. The subtlety
is that the total air pressure has two contributions,

Ptot = Patm + Pvap

where the first is the pressure due to atmospheric gases, such as nitrogen and oxygen, and the
second is the pressure due to water vapor in the air. In everyday conditions, Patm is about
100 times larger than Pvap , and the liquid water feels the pressure Ptot . But water vapor evap-
orates and condenses independently of the air, so Patm has no effect on it at all. As a result,
in equilibrium liquid water and water vapor can coexist, with the vapor having a pressure
Pvap = P (T ). The coexistence curve P (T ) is thus also called the (equilibrium) vapor pressure.

The relative humidity φ is defined as

φ = Pvap /P (T )

and quantifies how saturated the air is with water vapor. In equilibrium, φ = 1, while for
φ < 1 people can cool down by sweating. It is also possible to have φ > 1, which occurs in
humid air high in the atmosphere just before it condenses into a cloud.

The dew point Td satisfies

P (Td ) = Pvap
and represents the temperature at which water would begin condensing out of a given parcel
of air, cooled at constant pressure.

The boiling temperature Tb satisfies

P (Tb ) = Patm .

This is the temperature at which a bubble of pure water vapor, which forms inside the liquid,
has a high enough pressure to push the liquid away and continue to expand. This is why
pressure cookers are useful: they increase Patm , thereby increasing Tb and allowing food to
cook faster. (Technically, we should have Ptot rather than Patm on the right-hand side, but
in practice whenever we boil things, we let the resulting water vapor fly away. So the actual
vapor pressure Pvap never gets that high.)

At the interface between two liquids, boiling can start at a significantly lower temperature
than the boiling temperature of either liquid, because both of the liquids contribute to the
vapor pressure. This is called border boiling, and is treated in IZhO 2020, problem 2.

Example 5

How does alcohol distillation work?

9
 Kevin Zhou Physics Olympiad Handouts

Solution
Let alcohol and water have coexistence curves/equilibrium vapor pressures of Pa (T ) and
Pw (T ). The vapor pressure of alcohol is higher, with pure alcohol boiling at about 80◦ C.

Consider heating a mixture of alcohol and water, with mole fractions Xa and Xw , with
Xa + Xw = 1. If a small bubble of gas forms inside, then Raoult’s law states that both
alcohol and water vapor will be present, and contribute independently in proportion to their
mole fractions. Then the boiling point satisfies

Xa Pa (Tb ) + Xw Pw (Tb ) = Patm .

This is in between the boiling points of alcohol and water individually. By the ideal gas law,
the ratio of mole fractions of alcohol and water in the vapor is the ratio of vapor pressures, so

Xa0 Xa Pa (Tb )
= .
Xw0 Xw Pw (Tb )

Since the fraction is greater than one, the alcohol in the distilled vapor is more concentrated
than in the liquid.

By the above logic, we could get completely pure alcohol by just repeating the distillation
procedure several times. Actually, it’s more complicated than that because the alcohol and
water molecules will interact, causing Raoult’s law to break down; our calculation above only
applies for an “ideal mixture”. For more about distillation, see these notes.

[3] Problem 12. Kalda Thermodynamics, problem 22. A problem on practically measuring humidity.

[3] Problem 13. EFPhO 2006, problem 1. (Note that the comma in the density of air in part 4
denotes a decimal point.)

3 General Phase Transitions

In this section we’ll illustrate some of the ideas needed to analyze phase transitions in general.
Idea 5
A nonideal gas can be described by the van der Waals equation of state,

an2
 
P + 2 (V − nb) = nRT.
V

You derived the pressure correction in T1 assuming weak attractive intermolecular forces;
the modification of the volume accounts for the fact that the molecules can’t overlap each
other. Remarkably, this equation of state also contains a liquid-gas phase transition!

To see this, consider a plot of the isotherms on a PV diagram.

10
Kevin Zhou Physics Olympiad Handouts

At low temperatures, the isotherms can have negative compressibility, meaning that the
pressure decreases as the volume decreases. This is unphysical, and means that the fluid is
unstable at these points: if you push on it, it’ll just keep shrinking, until it condenses into a
dense liquid. Therefore, parts of these isotherms should be replaced with horizontal lines;
along these horizontal parts liquid and gas coexist, in varying proportions.

Specifically, everything underneath the dotted line should be replaced with horizontal lines.
As described in more detail in section 26.1 of Blundell, this can be shown by demanding
that the liquid and gas have equal Gibbs free energy. As a result, the total area on the PV
diagram of the isotherm that goes underneath the horizontal line equals the area that goes
above it; this is called Maxwell’s equal area rule.

There is a critical isotherm marked in bold above. Above this temperature, there is no
liquid-gas phase transition at all; instead we just have one phase, called a supercritical fluid.
Specifically, this is the temperature of the first isotherm that no longer has a local minimum
in pressure, which means    2 
∂P ∂ P
= =0
∂V T ∂V 2 T
at the critical temperature T = Tc . As you will see in problem 14, this occurs at
8a a
Vc = 3nb, Tc = , Pc = .
27Rb 27b2
This point, marked above, is called the critical point.

11
 Kevin Zhou Physics Olympiad Handouts

Remark
The van der Waals equation of state is accurate for a sparse gas with weak attractive inter-
actions; you shouldn’t expect it to be accurate for dense gases or the liquid state. However,
it still is extremely important because it is one of the simplest equations of state that gives
a liquid-gas phase transition. What’s more, if you zoom in near the critical point and write
the pressure, volume, and temperature as multiples of the critical pressures, volumes, and
temperatures, it turns out that all equations of state give the same results! This deep phe-
nomenon is known as universality, but unfortunately I can’t explain the reason why without
using statistical field theory.

[3] Problem 14. INPhO 2018, problem 6. A series of exercises on the van der Waals gas. Feel free
to look up definitions for part (a).
[4] Problem 15. [A] Here we’ll introduce a simple model for a ferromagnetic phase transition. Consider
N electrons, which may have spins si = ±1. The energy of a configuration is
X J X
E = −B si − si sj .
2N
i i6=j

The first term represents the effect of an external magnetic field B, while the second term represents
an interaction, with strength described by the constant J, which tries to make the spins parallel.
(In this simple model, we suppose all distinct pairs of spins interact equally. We could also make
spins only interact with their neighbors, but this would complicate the analysis.)
P
(a) Define the average magnetization as m = i si /N . Find E(m), the energy in terms of m and
the other constants in the problem.
(b) For a fixed value of m, write down the number of states Ω(m) with that magnetization.
(c) The probability of having a given value of m is proportional to e−βE(m) Ω(m). Argue that this
probability is maximized for the value of m that minimizes the free energy
F = E − T S.
Hence the equilibrium configuration minimizes the free energy. This is the statistical mechan-
ical way to argue that F is minimized; the thermodynamic way was covered in T2.
(d) Assuming that N  1 and applying Stirling’s approximation (introduced in T2), show that
the free energy F (m) is minimized when
1
m = tanh(βB + βJm), β = .
kB T
(e) For a fixed B > 0, plot m(T ). This should match with Curie’s law, which you proved in T1.
(f) Now let B = 0. Show that there exists a critical temperature Tc , above which m(T ) is exactly
zero and below which it is nonzero; also find an approximate expression for the magnetization
just below Tc .
This is a phase transition where the material spontaneously becomes magnetized, and the simplest

01hˆ
example of a phase transition which can be understood analytically.
[5] Problem 16. APhO 2011, problem 3. A nice problem on a real-world mechanical phase
transition. Some of the intuition you gained studying the var der Waals gas will be useful.

12
Kevin Zhou Physics Olympiad Handouts

4 Thermodynamic Systems
Now that we know all about the different methods of heat transfer, as well as phase transitions, we
consider some questions involving a mix of these concepts as well as mechanics.
Example 6: IPhO 1967.3

Consider two identical homogeneous balls with the same initial temperatures. One of them
is at rest on a horizontal plane, while the other hangs on a thread.

The same quantity of heat is supplied to both balls. Which has the higher final temperature?

Solution
This infamous problem, which appeared on the first IPhO, was probably the first ever
“troll” Olympiad question. The balls are different because the one on the plane ther-
mally expands upward, while the one on a thread thermally expands downward. This
tiny change in gravitational potential energy means that the ball on the thread ends up hotter.

This is an incredible solution – in the sense that it is not credible. About fifty years after
it was written, physicists at Oxford showed that it is wrong! Suppose the logic above were
actually right. Then a heat engine can be constructed with these four steps:

1. Heat the ball on the plane, therefore raising its center of mass.

2. Attach the ball to a thread and remove the plane.

3. Cool the ball on the thread, therefore raising its center of mass.

4. Put a plane just under the ball’s new position and remove the thread.

This is a heat engine, where the work goes into raising the ball. If the ball is only heated
and cooled a tiny amount dT , then the heat supplied is proportional to dT , but the distance
through which the ball rises is also proportional to dT . Therefore, the efficiency of the cycle
becomes a constant as dT goes to zero, which exceeds the Carnot efficiency (which instead
goes to zero) and hence violates the second law of thermodynamics.

A real ball doesn’t violate the second law, because it also stretches while hanging on the
thread, and squashes while on the plane. The slick solution neglects these effects and considers
only thermal expansion, but the above argument shows that this assumption is inconsistent:
you can’t have the latter without the former. This is an example of how thermodynamics
alone can, perhaps surprisingly, lead to constraints on mechanical properties.

13
 Kevin Zhou Physics Olympiad Handouts

Example 7

Why does a breeze cool you down, and why do clothes make you warmer?

Solution
Like most gases, air has a very small thermal conductivity, and enough viscosity so that you
carry around a thin layer of warm air with you wherever you go. (The main reason you
cool down is because this warm air rises away from you, by convection.) When a breeze is
blowing, it strips off this cushion of warm air, which is why you feel colder.

When you’re sweaty, the same logic applies. The layer of air you carry around is also moist,
saturated with water vapor from your sweat. Again, a breeze removes this layer, allowing
more evaporation to happen, cooling you down.

Clothing material itself actually has a higher thermal conductivity than air. Its real purpose
is to trap the layer of warm, moist air around you, preventing it from being blown away by
breezes or rising from convection.

Example 8

Can you boil water in a pot by putting it into a bigger pot of boiling water?

Solution
No, because boiling is a phase transition that requires latent heat. The water in the small
pot can get heated up to boiling temperature, but it can’t start boiling, because at that point
it’ll be at the same temperature as the bigger pot, and no more heat can flow.

[1] Problem 17. A greenhouse is a structure with a glass roof, which can be used to grow plants even
in very cold temperatures. Explain the effect that makes these greenhouses work.

[2] Problem 18 (IPhO 1996). A thermally insulated piece of metal is heated under atmospheric
pressure by an electric current so that it receive a constant power P . The temperature is

T (t) = T0 (1 + a(t − t0 ))1/4

where T0 , t0 , and a are constants. Find the heat capacity at constant pressure CP (T ).

[3] Problem 19. EFPhO 2011, problem 8. A tricky data analysis problem.

[3] Problem 20. EFPhO 2014, problem 9. A nice problem reviewing radiation and kinetic theory.

[4] Problem 21. 01@ EuPhO 2019, problem 1.

14
 Kevin Zhou Physics Olympiad Handouts

5 Thermodynamic Fluids
In this section, we focus on problems combining thermodynamics with fluids, as covered in M7.

01hˆ
We begin with some questions which only require fluid statics.

[5] Problem 22. IPhO 1998, problem 2. A very nice real-world fluids/thermo question.

[5] Problem 23. 01hˆ APhO 2009, problem 3. Another nice, simple model of a real-world phenomenon.

Idea 6
In M7 we considered Bernoulli’s principle for incompressible liquids with no temperature.
However, in general fluids are compressible and carry internal energy. To derive Bernoulli’s
principle in this more general context, we apply conservation of energy to a tube of streamlines,
as one mole of gas flows through it. We neglect gravity, since it typically is unimportant for
such rapid flows. The energy of a mole of gas at the entry of the tube is
1 2
µv + cV T1
2 1
where µ is the molar mass, and cV is the heat capacity per mole. Similarly, the energy of a
mole of gas at the other end is
1 2
µv + cV T2 .
2 2
The difference of these two must be the work done on the tube of fluid as a mole of gas flows
through, which is
p1 δV1 − p2 δV2 = R(T1 − T2 )
where the δVi are the volumes of a mole of gas at the entry and exit, and we used the ideal
gas law. Combining and using cp = cV + R gives

1 2
µv + cp T = const
2
along a streamline in steady flow.

Remark
You might also see Bernoulli’s principle in the form
1 2
v + gh + cp T = const
2
where we’ve added on the contribution of gravitational potential energy. In this case, cp is
the heat capacity at constant pressure per unit mass, not per mole. Unfortunately, people
use the letter c or C to denote many different kinds of (specific) heat capacities. Whenever
doing a problem where a heat capacity is given, check the dimensions!

15
 Kevin Zhou Physics Olympiad Handouts

Example 9: Wang

A rocket propels itself by burning fuel to release diatomic gas of temperature T1 in its
combustion chamber, which has cross-sectional area A1 . The gas then flows adiabatically
and is expelled out of the nozzle, which has a cross-sectional area A2 , at a speed v2 relative
to the rocket, pressure p2 , and temperature T2 < T1 . In the limit of steady flow, determine
the thrust experienced by the rocket.

Solution
Since the flow is adiabatic and the gas is diatomic,
 7/2
T1
p1 = p2 .
T2

As discussed in M7, mass conservation in steady flow means ρAv must be the same on both
sides of the nozzle. The ideal gas law tells us that ρ ∝ p/T , so

p 1 A1 v 1 p 2 A2 v 2
= .
T1 T2
Combining these two gives a relation between the velocities,
 5/2
A1 T2
v1 = v2 .
A2 T1

Bernoulli’s principle gives

1 2 7 1 7
µv1 + RT1 = µv22 + RT2
2 2 2 2
where we neglected the gravitational energy. Combining with our previous relation gives

7R(T1 − T2 )
v22 = .
µ (1 − (A1 /A2 )2 (T2 /T1 )5 )

Finally, the thrust is

dp µp2 A2 v22 7p2 A2 (T1 − T2 )

F = = ρ2 A2 v22 = = .
dt RT2 T2 (1 − (A1 /A2 )2 (T2 /T1 )5 )

Remark
It turns out that in steady flow, the density of gas is automatically approximately uniform
everywhere as long as the flow velocity is much less than the speed of sound. That’s why
the more general form of Bernoulli’s principle above isn’t needed for many situations, even
for gases. All of the problems we’ll consider involve rapidly moving objects. (Of course, it’s
possible to subsonically compress air in nonsteady flow; for instance, the air in my bike tires
is compressed to about 7 times atmospheric pressure.)

[3] Problem 24 (Feynman). Air with density ρ, pressure P , and adiabatic index γ is flowing at

16
 Kevin Zhou Physics Olympiad Handouts

uniform speed v through a smooth pipe of constant cross-sectional area A. It is heated as it passes
a wire grid, which offers negligible resistance to the flow, with a power Q̇. This is a simple model
for a jet engine. For simplicity, suppose the output pressure is also P , though this would not be
true in a high efficiency engine.
(a) Find the speed v 0 with which the air exits the tube, in term of the given parameters.

(b) Find the thrust produced, in terms of v 0 and the other parameters.
[3] Problem 25 (Wang 2.17). Consider an ideal gas with pressure p, density ρ, and adiabatic index
γ. A density pulse is set up in the gas, traveling along the −x direction with speed c. Inside the
pulse, the gas has local velocity v  c in the lab frame, and a local density ρ + ∆ρ, where ∆ρ  ρ.
(a) Transform to the reference frame where the pulse is at rest, and find three constraints, using
mass conservation, energy conservation, and the fact that the gas is compressed and expanded
adiabatically when it enters and leaves the pulse (i.e. heat conduction is negligible).

(b) Combine these relations to find c. Work to lowest order in the small quantities v/c and ∆ρ/ρ.
The quantity c is the adiabatic speed of sound in a gas, which we’ll derive in a different way in W3.
For more about sound waves, see chapter 31 of Blundell, or section 14.1 of Lautrup.

Example 10

Show that the kinetic and potential energy of a sinusoidal, adiabatic sound wave are equal.

Solution
First, we explicitly define our notation. The sound wave profile is

P = P0 + ∆P cos(kx − ωt), ρ = ρ0 + ∆ρ cos(kx − ωt)

and
v v0
= cos(kx − ωt), T = T0 + ∆T cos(kx − ωt).
c c
In problem 25, you derived relations between the parameters (∆P )/P0 , (∆ρ)/ρ0 , v0 /c, and
(∆T )/T0 , which are all assumed small, and of the same order of magnitude. The total energy
density of the fluid, up to second order, is
1
u = ρ0 v 2 + cv ρT
2
where here cv is the heat capacity per unit mass. The first term is the bulk kinetic energy
density, while the change in the second term is the potential energy density, where we’re
using the usual meaning of potential energy as any energy which isn’t kinetic.

This all looks pretty straightforward, but there’s a reason that most introductory textbooks
never write down this expression. You can see the issue by applying the ideal gas law to the
second term. Since P ∝ ρT , this term is proportional to P , but the average of P is just P0 .
This suggests that sound waves have no potential energy density at all, which is wrong. For
instance, if you instantly take out all the macroscopic kinetic energy, setting v to zero, then

17
Kevin Zhou Physics Olympiad Handouts

there is still energy remaining that can be harvested because the pressure is nonuniform.

Here’s the problem: energy is inherently a second order quantity. If ρ and T were both small
quantities, then it would be good enough to multiply them to get the answer to second
order. But instead, ρ and T are the quantities ρ0 and T0 shifted by small quantities ∆ρ and
∆T . That means that to get the quantity ρT correct to second order, we need to get both ρ
and T individually correct to second order, which is beyond the first order approximations
we started with! This is a conceptual issue that occurs whenever you have a perturbation
that shifts existing properties of a medium. (It doesn’t happen for waves on a string, which
we cover in W1, because those waves are parametrized by y, and y = 0 when there is no wave.)

It’s possible to fix this issue, but we need to be careful. First, let’s use the ideal gas law to
change variables to pressure, so we only have one quantity to deal with,
R 1 P
cv ρT = ρT = .
µ γ−1 γ−1
The key insight is that we can keep our expression for ρ the same. The reason is that for
the wave equation to continue to be satisfied at second order, we can only add second order
terms that are constant, or also sinusoids. Adding a constant to the density Ris not allowed
because we know the total number of particles is conserved, so the mass ρ dV is, and
adding a sinusoid can be absorbed by simply redefining ∆ρ.

Now, we know that P ∝ ργ , and letting δρ = ∆ρ cos(kx − ωt) for convenience, we have
 γ
γ δρ γ(γ − 1) δρ 2
 
P ρ
= =1+ + .
P0 ρ0 ρ0 2 ρ0
When we integrate the first term, we get the internal energy of the fluid at rest. The second
term averages to zero, and so can be disregarded. The third term gives the desired result,
γP0 δρ 2
 
1 2
u = ρ0 v + .
2 2 ρ0
Now that the energy is in terms of small quantities squared, we can relax and use first order
results. Using the results derived in problem 25, this can be simplified to
1 2 δρ 2
 
1 2
u = ρ0 v + ρ0 c
2 2 ρ0
and the two contributions are equal, because v/c = −δρ/ρ0 .

As a check on this result, it is often true that the energy density is equal to the momentum
density times the wave speed, u = pc. (For example, this corresponds to E = pc for photons.)
The momentum density is p = ρv = (ρ0 + δρ)v. The first order term represents the overall
momentum of all of the gas, not the momentum due to the sound wave itself, so it can be
ignored. The second order term is
ρ0 v 2
p = δρ v =
c

18
 Kevin Zhou Physics Olympiad Handouts

which is exactly the expected result.

Ironically, while we began this discussion by noting that the energy of a transverse wave
on a string is more straightforward, the longitudinal momentum of a transverse wave on a
string is far more confusing – how can there be any if the string moves only transversely?
Does the above identity u = pc break down for these waves, or do we just need to evaluate
p more carefully? (For one perspective in this ongoing debate, see this paper.) All of this is
too subtle to be relevant to Olympiads; even string theorists get confused about it. It’s just
a reminder that there are always subtleties lurking in even basic physics.

[3] Problem 26. 01hƒ IPhO 2012, problem 1B. A tricky real-world problem on fluids and condensation.

19