Matthew Schwartz

Statistical Mechanics, Spring 2019

Lecture 9: Phase Transitions

1 Introduction
Some phases of matter are familiar from everyday experience: solids, liquids and gases. Solid H2O
(ice) melting into liquid H2O (water) is an example of a phase transition. You may have heard
somewhere that there are 4 phases of matter: solid, liquid, gas and plasma. A plasma is an ionized
gas, like the sun. I don't know why plasmas get special treatment though  perhaps it's the old idea
of the four elements. In fact, there are thousands of phases. For example, ferromagnetic is a phase,
like a permanent iron magnet. When you heat such a magnet to high enough temperature, it under-
goes a phase transition and stops being magnetic. Conductors, insulators, and semi-conductors are
also phases of matter. At very very high temperatures, nuclei break apart into a quark-gluon phase.
Solids generally have lots of phases, determined by crystal structure or topological properties. For
example, diamond and graphite are two phases of carbon with dierent lattice structure:

Figure 1. Two of the phases of solid carbon

It's actually quite hard to come up with a precise denition of a phase. Some textbooks say
 A phase is a uniform state of matter.
This is an intuitive denition, but not very precise. Taken literally, it is too general: a gas at a
dierent temperature is a dierent uniform state of matter. So is it a dierent phase? We don't
want it to be. We want gas to be the phase.
A more technically precise denition is
 A phase is a state of matter whose properties vary smoothly (i.e. it is an analytic function
of P ; V ; T etc).
You might rst think that this denition makes liquid H 2O (water) and gasesous H2O (steam)
in the same phase, since we can boil water and it slowly becomes steam. Although this does
sound smooth, it is not. For example, consider the temperature of water as heat is added. As you
heat it the temperature rises. But when it hits the boiling point, the temperature does not rise
anymore, instead the heat goes into vaporizing the water. Then onces it's all gas, its temperature
changes again. So the density of H2O change discontinuously and non-analytically as a function
of temperature (T = 99.9C) is very dierent from (T = 100.1C).
Connecting phase to smoothness properties allows to shift focus from phases themselves to the
transformations between phases called phase transitions. Phase transitions are an incredibly
important area of physics.

1
2 Section 2

Physicists take two dierent approaches to phase transitions. On the one hand, we can treat
each phase as its own statistical mechanical system. For example, the ensemble we use to describe
ice is very dierent from the one we use to describe water vapor  neglecting interactions is an
excellent approximation for many gases, but a horrible approximation for solids. This makes the
discontinuities in density, entropy, etc, across a phase transition inherent in our description. It lets
us derive very useful formulas for phase transformations, such as how the pressure and temperature
of the phase transformation are related.
The second approach is to construct a statistical mechanical system that describes a substance
on both sides of a phase transition. For example, if we knew the exact partition function for N
molecules of H2O, we should be able to see the gas, liquid and solid states arise from dierent
limits. This partition function would necessarily have non-smooth properties across the values of
temperature and pressure associated with the phase transition. It is these kinks in otherwise smooth
functions that make phase transitions so interesting. Where do they come from? Water itself is too
complicated to write down an exact partition function, but there are plenty of simpler systems that
we can solve to understand phase transitions. Pursuing these simpler systems leads to concepts
you may have heard of like critical phenomena, the renormalization group, mean-eld theories, etc.
Most of these topics, unfortunately, we will not have time to pursue  phase transitions are the focus
of much of modern condensed matter physics and could easily occupy a year-long graduate course.
In this lecture, we will start with discussing the familiar phases of solid, liquid and gas, and
understand transitions between them using statistical thermodynamics. Then we will discuss some
of the broader, more general aspects of phase transitions.

2 Solids, liquids and gases

It is not hard to gure out what we mean by solid, liquid, or gas. Both solids and liquids are
essentially incompressible. The compressibility is dened
 
1 @V
T =¡ (1)
V @P T

So solids and liquids have T  0. This means no matter how much pressure we put on, we cannot
make solids or liquids much denser. Gases have much larger values of T . Indeed, from the ideal
1
gas law PV = NkBT , we see that for a gas T = P = / 0. Compressibility is an example of an order
parameter, something whose value characterizes the phase.
Solids and liquids of the same substance often have approximately the same density, while the
density of gases is much lower. Liquids and solids are called condensed matter. Solids dier from
liquids and gases in that they are rigid. More precisely, they do not deform under a sheer stress,
i.e. they have zero shear modulus Ss. Thus shear stress is an order parmeter for the liquid-solid
phase transition. Liquids and gases are collectively called uids.
It may be helpful to say a few more words about liquids. Liquids generally have around the
same density as solids, so the atoms are all in contact with each other. Instead of having strong
covalent or metallic bonds, like in a solid (cf. Lecture 14), liquids have weaker ionic or hydrogen
bonds that keep the molecules close. Although the attractive force in liquids is weak, typical
thermal velocities in a liquid are not enough to overcome it.
For example, in water, H2O, the H-O bonds are covalent, with the shared electrons localized
closer to the O than the H. This makes the H slightly positively charged and the O negatively
charged. When a two water molecules approach each other, the H from one is then weakly attracted
to the O from the other, forming a hydrogen bond. The O in each water molecule can form 2
covalent bonds and 2 hydrogen bonds, giving liquid water a tetrahedral formation. Molecules on
the surface of water must have fewer than 2 hydrogen bonds per oxygen, on average. Thus there
G
is any energy cost to having a surface. The Gibbs free energy per area of surface = A is called
the surface tension:
Solids, liquids and gases 3

Figure 2. The H and O atoms in water have small charges (left) forming attractive hydrogen bonds
(middle). Surfaces cannot saturate all the bonds, therefore there is any energy cost G = A to having a
surface, with the surface tension.

All liquids have a surface tension. The surface tension of water happens to be particularly large
mN mN mN
(e.g. water = 73 m compared to say, CO2 = 17 m ), but not the largest (mercury has Hg = 486 m ).
One consequence of the surface tension is that a liquid will not expand to ll all the available
volume. Liquids form droplets. Even in zero-gravity, water still forms droplets, as does mercury:

Figure 3. (Left) water in zero gravity. (Right) a ball of liquid mercury in zero gravity being hit by a piece
of metal.

Whether a solid turns into a liquid when it is heated depends not only on the types of inter-
actions among the molecules, but also on the pressure. At low pressures, when solids are heated,
the atoms that break free of their covalent bonds y o into gas, and the solid sublimates. Only if
the pressure is suciently high will the molecules stick around close enough to each other for the
weak attractive interactions to dominate and the liquid phase to form.
Two phases can be in equilibrium with each other. Consider an unopened bottle of water
(constant volume and temperature). The liquid water in the bottle and water vapor in the bottle are
in equilibrium. Water molecules are constantly evaporating from the water into the air, and water
molecules are condensing into the water at exactly the same rate. Thus volume and temperature
are not good ways of characterizing phases. What happens if we open the bottle? Eventually, all
the water will evaporate. An open bottle is at constant temperature and pressure. Under these
conditions generally a single phase dominates. Thus, phases of matter can be characterized by
temperature and pressure.
Why are there single phases at constant T and P ? For a single pure substance, the chemical

P
potential is a function of P and T alone:  = (P ; T ). For example,  = kBT ln kBT3
for a
h
montatomic ideal
 gas, where  is the thermal wavelength,  = 2m kB ; for a more general ideal
p
P
gas  = kBT ln k T  with  the single-particle partition function. There is no N dependence in
B

the chemical potential at xed P and T since the Gibbs free energy G(N ; P ; T ) = N is extensive.
When there are two phases present the Gibbs free energy is G = N11(T ; P ) + N22(T ; P ). Note
however one important dierence between the two-phase case and the two-chemicals case: when
two phases are present there is no entropy of mixing. Solids and liquids coagulate because of the
surface tension, so even if multiple phases are present they are always separated and the mixing
entropy is tiny if not completely absent (maybe
S  kB log(a few) for a few chunks of ice in ice
4 Section 2

water, but this is negligible compared to the entropies of the ice and water separately S  NkB ).
Another way to see that the mixing entropy is absent is that the pressures are the same in two
phases; they don't become partial pressures that add up to the total pressure (partial pressure
will be relevant of water vapor is mixed with something else, like air, but not for pure water/ice
equilibrium that we are discussing here).
At constant temperature and pressure, equilibrium is determined by minimizing the Gibbs free
energy G. Recall that
dG = VdP ¡ SdT + 1dN1 + 2dN2 (2)

So if
N particles go from phase 1 to phase 2 at constant T and P then

G = 1(¡
N ) + 2(
N ) = (2 ¡ 1)
N (3)

Thus, to minimize G, particles move from higher chemical potential to lower chemical potential.
This will keep happening until there are no more particles to change phase. Thus at a xed
value of T and P , as long as 1 =/ 2, only one phase is allowed. In the law of mass action, the
chemical potentials had N dependence which lead to equilibrium situations with dierent amounts
of substances. That N dependence all came from the entropy of mixing. Since here entropy of
mixing is absent, one phase completely annihilates the other.
Note that if we express the chemical
  potential in terms of V rather than P , for a single ideal
N
gas, it takes the form  = kBT ln V
 . This expression does depend on N . So as N changes the
chemical potential changes too. Thus, at constant volume, equilibrium can be achieved with nite
amounts of two phases. We'll come back to this situation when discussing vapor pressure below.
Returning to constant pressure, it is only when 1(T ; P ) = 2(T ; P ) that there is no change in
the Gibbs free energy with
N , and so only then can the two phases can be present at once. Since
1(T ; P ) and 2(T ; P ) are functions, setting them equal generates a curve in the T / P plane. This
curve is the phase boundary. On the phase boundary two phases are in equilibrium.
A diagram of the phases as a function of pressure and temperature is called a phase diagram.
Here are some example phase diagrams for carbon dioxide, argon and water.

Figure 4. Phase diagrams for CO2, Ar; and H2O. The point STP in these plots refers to T = 20C = 293K and P = 1 atm.

The thick lines in the phase diagram are the phase boundaries, determined by 1 = 2. A
phase transition is the transformation as a phase boundary is crossed. We dene
 Melting: transition from solid to liquid.
 Freezing/Fusion: transition from liquid to solid.
 Boiling/vaporization: transition from liquid to gas.
 Condensation: transition from gas to liquid.
 Sublimation: transition from solid to gas.
 Deposition: transition from gas to solid.
Phase boundaries 5

Note in the CO2 phase diagram that as temperature is increased at P = 1 atm, CO2 goes from solid
directly to gas: it sublimates. This is why smoke comes o dry ice, but there is no liquid. Liquid
CO2 requires at least 5 atmospheres of pressure.
In a pure phase (o the phase boundary) there is one type of substance and so G = N with
G
N xed. Then,  = N and so
   
@ 1 @G S
= =¡ (4)
@T P N @T P N

Since S > 0 this implies that the chemical potential always decreases as the temperature goes up.
Moreover if there are two phases with dierent entropies, then as the temperature is raised, the
chemical potential of the one with the larger molar entropy (entropy per mole of particles) will
go down more. Thus the higher entropy state is preferred at larger temperature. This explains
why solids melt when you heat them and liquids boil: the phase transitions are driven by entropy
and Ssolid < Sliquid < Sgas. To be fair, we haven't shown that Ssolid < Sliquid < Sgas, instead, we can
deduce it from phase diagram since when heated solids melt and liquids vaporize.
The rst derivative of G with
 respect to temperature is discontinuous  across
 a phase boundary:
@G @G
innitesimally below it @T = ¡Sliquid and innitesimally above it @T = ¡Sgas. In a pure
P P
phase, G is a smooth function (it and all its derivatives are continuous). Therefore, G changes non-
smoothly, the phase changes. We'll come back to this in Section 4.
Let's observe some features of the CO2 phase diagram. First note that liquid/solid phase
boundary and the liquid/gas phase boundary intersect. The point where they intersect is a special
value of P ; T called the triple point. At the triple point, three phases are in equilibrium together:
solid = liquid = gas.
Note also that the phase boundaries do not extend up forever. They end at a point in P ; T space
called the critical point. Critical points are super interesting experimentally and theoretically.
An implication of the phase boundary ending is that one can go around the phase transition line.
That is, one can smoothly transform a liquid into a gas, without crossing a phase boundary. This
is one reason why it is hard to give precise denitions of phases. For example, we said that solids
and liquids have small compressibilities T . But we didn't say how small. As you move around the
critical point from the liquid side, the compressibility gets larger. At some point we don't consider
it small and the phase is somewhere between a liquid and a gas  like a very dense gas. So let's
not even try to give precise general denitions to dierent phases. Instead, we'll study transitions
between phases. These transitions are precisely dened since the phase boundary is precise.

3 Phase boundaries
Suppose we are close to a phase boundary, but on one side of it. Then a single phase completely
G
dominantes, and N is xed (d N = 0). Since N = , then d G = N d and from Eq. (2), d G =
VdP ¡ SdT , we nd
V S
d = dP ¡ dT (5)
N N
Note the consistency with Eq. (4) at constant P . This holds in any pure phase region, even
arbitrarily close to the phase boundary. In particular, it holds on both sides of the phase boundary,
as we approach the phase boundary. But on the phase boundary, 1 = 2, so the way pressure and
temperature must change as we move along the phase boundary is determined by setting d1 = d2
which gives
V1 S V S
dP ¡ 1 dT = 2 dP ¡ 2 dT (6)
N1 N1 N2 N2
That is
 
S
dP
N
=   (7)
dT
V
N
6 Section 3

This equation, called the Clapeyron equation, determines the shape of the phase boundary.

3.1 Latent heat

S
The Clapeyron equation involves the change in the molar entropy, N . Recall that G = H ¡ TS so
G H S S H
 = N = N ¡ T N . Since 1 = 2 in equilibrium then T
N =
N at the phase boundary. So we can
can also write the Clapeyron equation as
dP 1 L
= ¡
(8)
dT T
1
n
N
where n = V
is the number density and
 
H
L=
(9)
N
is called the latent heat.
Latent heat is the change in enthalpy per molecule, like a reaction enthalpy,
rH but for a
phase transition at saturation (on the phase transition boundary). It is a heat because as you
Q
H
heat something at saturation, the heat put in changes the entropy by
S = T . Since
S = T we
have simply that Q =
H: the heat put in is the change in enthalpy. The latent heat is the heat
put in per molecule to change the phase.
For example, when you boil water, you put more and more heat in, and more water evaporates,
but the temperature doesn't change. The heat you put in is providing the energy it takes to
break up the hydrogen bonds in the water. The enthalpy of formation of liquid water is
fH =
kJ kJ
¡286 mol and for water vapor is
fH = ¡242 mol . The dierence between these is the latent heat
kJ
of vaporization of water at 1 atm: Lvap = 44 mol . This is positive since it takes heat to boil water.
kJ
Note that 44 mol is a large number: water has a lot of energy stored in its hydrogen bonds and it
J
is hard to separate them. For comparison, consider the heat capacity of liquid water CP = 75 mol K ,
kJ
which implies that to heat water from 0C up to 100C takes only
H = 7.5 mol of energy. That
is, it takes 6 times as much energy to vaporize water as it does to heat it up to its boiling point
from its freezing point: water does not want to evaporate.
That the latent heat of vaporization of water is so large is the reason that sweating is such an
ecient form of cooling. Say it's 105F outside, which is higher than body temperature. If you
didn't sweat, your body would just heat up until it reached equilibrium with the air. Instead, liquid
water forms on your skin, and it draws heat from your body, evaporating into air and cooling your
body at the same time. Air conditioners exploit the latent heat of vaporization as well, as we will
explore on a problem set. Note that latent heat, which exploits the dierent chemical potential for
dierent phases, allows temperature dierences to increase spontaneously. This is not in conict
with the second law of thermodynamics because the total entropy is increasing: the evaporated
water has a much larger entropy increase than the entropy decrease from cooling your body.
It's also worth pointing out, for completeness, that the latent heat of fusion for melting ice is
kJ
Lfuse = 6.0 mol , which is not particularly large. Indeed, it is a much smaller energy than Lvap. Note
that Lfuse > 0 since it takes heat to melt ice.
Using the Clapeyron equation, we can deduce some simple features of phase boundaries. Con-
sider rst the solid to liquid transition (i.e. melting). This involves breaking covalent bonds, so
Lfuse > 0. Actually, there is one element with a negative latent heat of fusion: helium. Helium
remains liquid down to T = 0 and must be pressurized to solidify. In general, the density change in
going from solid to liquid is usually very small and slightly negative: most solids are slightly more
¡1
dP
dense than their liquid forms. So
n & 0. Therefore by Eq. (8), dT is generally very large and
positive. That is, the liquid/solid phase boundary is usually quite steep and goes slightly to the
right in the T /P plane. This can be clearly seen in the phase diagrams in Fig. 4 above.
A well known exception to the density decreasing on melting is water. Water expands when it
freezes due to the unusual importance of hydrogen bonding in the liquid. We can see this in the
phase diagram in Fig. 4, or more clearly if we zoom in with a logarithmic T axis:
Phase boundaries 7

2
Figure 5. Phase diagram for water.

dP
Note that the solid liquid boundary goes up and to the left, so dT < 0. Since L > 0 and T > 0 this
¡1

must mean
n < 0, as is indeed the case for water. It is the tetrahedral structure of solid ice
that makes it not particularly dense. Some other materials with tetrahedral structure (like silicon
or gallium) also have denser liquids than solids.
For the transition between liquid and gas (vaporization), the enthalpy change is again positive

H > 0, so L > 0. In water this is because hydrogen bonds are broken. In general it's because there
are attractive interactions among molecules in liquids sticking them together, and it takes energy
to separate molecules that are attracted to each other. In addition
(PV) = RT upon vaporization
¡1

which also contributes to the latent heat. Gases are usually much less dense than liquids, so
n
is positive and generally much larger than for the solid-liquid transition. Since the gas density is
much less than the liquid density ngas  nliquid, we can write

1 1 1 1 V RT

= ¡  = gas = (10)
n ngas nliquid ngas Ngas P

where the ideal gas law PV = nRT was used in the last step. Thus,
dP PL
= (11)
dT RT 2
This is known as the Clausius Clapeyron equation. Here L is the latent heat of vaporization.
The pressure at the liquid-gas transition is called the vapor pressure.
Since the latent heat is dominated by the enthalpy change of breaking bonds, we expect it to
be a slowly varying function of temperature. If we assume L is independent of T , then we can
integrate the Clausius-Clapeyron equation. Writing it as

1 L
dP = dT (12)
P RT 2
we can integrate both sides to give
 
L
P = C  exp ¡ (13)
RT

with C an integration constant. Starting at any point P = P  and T = T  as a boundary condition

we then have
  
L 1 1
P = P exp ¡

¡ (14)
R T T
8 Section 3

For example, at sea level (P  = 1 bar) water boils at T  = 373 K. At 1000 m, the atmospheric
kJ J
pressure is P = 0.9 bar. Using the latent heat of vaporization of water is L = 42 mol and R = 8.3 mol ,
we get
 ¡1
1 R P
T= ¡ ln = 370.1K (15)
T L P
which is three degrees lower.
Keep in mind that you cannot extrapolate the Clausius-Clapeyron equation too far. Eventually,
the temperature dependence of the latent heat becomes important. For small changes in T and P
it is usually makes predictions in excellent agreement with observation.

3.2 Vapor pressure

The vapor pressure is the pressure of a pure substance at saturation (on the phase boundary).
Generally, vapor pressure refers to the pressure of a gas (vapor), so we use vapor pressure to
describe liquid-gas and solid-gas phase transitions. At saturation, two phases can exist in equi-
librium. Conversely, if two phases are in equilibrium, the pressure of the gas must be the vapor
pressure at that temperature. When there is a mixture of liquids or gases, the partial pressure of
each must equal the appropriate vapor pressure in equilibrium.
For example, if we have a sealed bottle of water, there will be some water vapor in the bottle,
above the water. If there is only water vapor (no air), then the vapor pressure at room temperature
can be determined from the Clausius-Clapeyron equation. Using Eq. (14) at the boiling point
kJ
T  = 373K, P  = 1 atm and L = 42 mol , we nd that at room temperature, T = 298K that
P = 0.034 atm = 3.142kPa. This is consistent with Fig. 5. Note that our calculation implies that the
vapor pressure of water at room temperature is much lower than atmospheric pressure. This may
seem unintuitive, since it implies that water should not evaporate. Indeed, it would not, if there
were no air. In fact, air is only around 1% water at sea level, so the partial pressure of water in air
is 0.01 atm which is about 3 times smaller than the vapor pressure. So water does evaporate into
air. If you seal a bottle of water at room temperature, the water will start to evaporate and the
partial pressure of water in the bottle will increase. It will go up from 0.01 atm to 0.03 atm and
then stop, since it matches the vapor pressure. At this point, the total pressure inside the bottle
has gone up from 1 atm to 1.02 atm. This small pressure increase is responsible for the pt you
sometimes hear when opening a bottle of water, even if it's not carbonated.
What happens if we mix some solute into the water? For example, how does the vapor pressure
of water change when salt is added, and how does its boiling point Tboil change? To an excellent
approximation, the salt stays in the water, so that only the water is in equilibrium with its vapor.
Let w(P ; T ) be the chemical potential of pure liquid water, gas(P ; T ) be the chemical potential
of the pure water vapor, and mixed
w (P ; T ) be the chemical potential of water in the saltwater mix.
The boiling point Tboil = T0 for pure water at a vapor pressure P0 satises

w(P0; T0) = gas(P0; T0) (16)

For saltwater, the equilibrium condition is mixed

w (P ; T ) = gas(P ; T ).
Recall from the discussion of osmotic pressure that the saltwater has higher entropy than pure
water, due to entropy of mixing, so it has lower Gibbs free energy, G = G0 ¡ TSmix and therefore
lower chemical potential. For small solute concentrations, we found
Ns
mix
w (P ; T ) = w(P ; T ) ¡ kBT (17)
Nw
where Ns is the number of salt molecules and Nw the number of water molecules.
We rst ask how the vapor pressure changes at xed temperature T = T0. Expanding around
P0 by writing P = P0 +
P we get
 
Ns @w N
mixed
w (P ; T0) =  (P
w 0 +
P ; T0) ¡ k T
B 0 = w (P ;
0 0T ) +
P ¡ kBT0 s (18)
Nw @P T Nw
Phase boundaries 9

Similarly,
 
@gas
gas(P ; T0) = gas(P0; T0) +
P (19)
@P T
 
@
= N so setting mixed
V
Now, @P w (P ; T0) = gas(P ; T0) and using Eq. (16) we get
T
 
Vw V Ns

P ¡ gas = kBT0 (20)
Nw Ngas Nw
N
The molar volume of the liquid is much lower than the gas (the gas density n = V is much larger),
V V V k T
so we can drop Nw compared to Ngas . Using the ideal gas law, Ngas = BP 0 we then have
w gas gas 0

 
kBT0 N
(P ¡ P0) ¡ = kBT0 s (21)
P0 Nw
or
Ns

P = ¡ P0 (22)
Nw

This is known as Raoult's law. It says the vapor pressure decreases when a solute is added
proportional to the molar fraction of the solute.
The decrease in vapor pressure can be understood physically. When there is pure water and
pure water vapor, there is an equilibrium between molecules evaporating from the solution and
condensing into it. When solvent is added, the number density of water on the surface goes down
slightly, so fewer water molecules evaporate per unit time while the same number are condensing.
Since more condense than evaporate, the gas pressure goes down until equilibrium is reestablished,
at a lower vapor pressure.
Knowing how the pressure changes, we can then nd the temperature change from the Clausius-
dP PL dP
Clapeyron equation d T = R T 2 , Eq. (11). Recall that d T is the slope of the phase boundary. For
dP
P
small
P and
T we can use dT
=
T
. We want to move back by ¡
P to restore the original
pressure, so
RT 2 Ns RT02

T = ¡
P = (23)
P0L Nw L

Since
P < 0 at xed T , the boundary shifts down/right, thus
T > 0 at xed P . The signs of
these equations are easiest to undrestand by looking at the liquid-vapor boundary in Fig. 5. A
version of this diagram with salt-water included is shown in Fig. 6.

Figure 6. Black curve is the phase boundary for pure water, pink is for salt water. So at 1 atm, the boiling
point goes up and the freezing point goes down.
10 Section 3

For example, if you add a tablespoon of salt (0.547 mol) to 2 liters of water (111 mol), the
vapor pressure at T = 373K goes down from 1 bar by
P = ¡0.005 bar. Using the latent heat of
kJ
vaporization of water L = 42 mol we get that
T = 0.14K. So the boiling point goes up, but by less
than a degree. Thus raising the temperature is not the reason we add salt to water when cooking!
Adding salt to water also lowers its freezing point. The formula is the same as Eq. (23) with
the opposite sign, since the salt is in the water, not the ice.1 See also Fig. 6.
Say we put 1 cup of salt (8.7 moles) out per square meter of ice that is 1 mm thick (1L total,
kJ
55 moles). The latent heat of fusion for water is L = 6.0 mol , about 1 / 7th of the latent heat of
vaporization. Then
J
8.7 8.3 mol K (273K)
2

T = ¡ = ¡16K (24)
55 6.0
kJ
mol

Thus if you salt your sidewalk, it won't freeze until the temperature drops to T = ¡16C = 3F .

3.3 Chemical potential phase diagrams

As we saw, phase boundaries are determined by the condition that the chemical potentials of the
two phases agree. If we are o a phase boundary, the phase with the lower chemical potential will
dominate. To see this, recall that G = N so we are just minimizing Gibbs free energy to nd the
dominant phase. So at a given pressure, a given phase will dominate over the range of temperatures
for which its chemical potential is lowest. This gives us a dierent perspective on phase transitions
which is sometimes useful.
We know that solids will dominate at low temperature and gases at high temperature. Thus
the solid phase as the lowest chemical potential at T = 0 and gas has the lowest chemical potential
at high T . We also know that
 
@ S
=¡ <0 (25)
@T P N

So the slopes of the chemical potential curves at constant pressure are always negative. Moreover,
since Ssolid > Sliquid > Sgas, the gas has the steepest chemical potential curve, followed by liquid, then
solid. Finally, since since the entropy of a solid at T = 0 is zero or nearly zero by the 3rd law of
thermodynamics, the solid line starts o horizontal. So a  / T diagram will look something like this:

Figure 7. Phase diagram in the chemical potential/temperature plane for water. Left shows the curves at
some pressure P . Right shows the eect of increasing the pressure, whereby the dashed lines move to the
solid lines.

1. This shouldn't be obvious, since we used ngas  nliquid, while we can't use nsolid  nliquid. We actually used
this limit twice: once to drop a term in Eq. (20) and once in Eq. (10). To derive Eq. (23) directly, you can avoid
both expansions. Try it yourself!
General phase transitions 11

What happens when we change the pressure? At a given temperature, if we change the pressure
then
 
@ V 1
= = >0 (26)
@P T N n

So the less dense the phase, the more its chemical potential changes, and increasing the pressure
always drives the chemical potential up. Thus the gas curve shifts up the most as pressure increases.
For water, as shown in the gure, the liquid is denser than the solid so its chemical potential
changes less. We see therefore that at higher pressure, the melting temperature for water is lower
and the boiling temperature is higher. The increase in temperature of boiling at higher pressure
in qualitative agreement with Eq. (14).
Here's another example

Figure 8. Phase diagram for subatomic matter

This gure shows the phase diagram for subatomic matter: quarks and gluons, as a function of
temperature and baryon chemical potential. Baryon chemical potential is the chemical potential
for quarks; since quark number is conserved, it can be nonzero. Some information about this phase
diagram we know from nuclear physics, some from astrophysics (e.g. neutron stars), some from
collisions of ionized lead and ionized gold at particle accelerators, some from cosmology, some from
theoretical calculations and simulations. We even have some insight into this phase diagram from
string theory. A lot is still unknown. An open question about this phase diagram is whether there
is a critical point between the quark-gluon plasma phase and the hadron gas phase (the red dot).
This could have implications for the earliest moments of the universe, just after the big bang.

4 General phase transitions

 
@G
We saw with the liquid/solid/gas phase transitions that @T
changes discontinuously at the
P
phase boundary. When this happens we say the transition is of the rst order.
 
@G
 First order phase transition: @T changes discontinuously at the phase boundary
P
 
@G
The rst in rst order refers to the rst derivative of G. It is possible for @T to be continuous,
P
but higher derivatives of G to be discontinuous:
 n 
@ G
 nth order phase transition: @T n changes discontinuously at the phase boundary
P
12 Section 4

This classication of phase transitions is known as the Ehrenfest classication. It is common to

call any transition with n > 1 a second order transition. This is because rst order transitions
are special: they have latent heat, which is a barrier to changing phases. Second order transitions
are smooth and have no barrier. They occur at points in phase diagrams where the phases merge
into one, and the latent heat vanishes.
In modern treatments, we use rst and second order a little more loosely. It doesn't have to be
the Gibbs free energy that determines the order, it can be the Helmholtz free energy, or just the
energy, depending on what is appropriate for the problem. And it doesn't have to be derivative with
respect to temperature, but can be some other derivative of the energy. The thing whose continuity
 the order parameter. For example, in the Ehrenfest classication,
we are questioning is called
@G
the entropy S = ¡ @T
is the order parameter. For boiling water, density is a natural order
P
Nm
parameter and a little more
 intuitive than the entropy. At xed N and m the density  = V
is
@F
equivalent to using V = @P .
T

4.1 Paramagnetism

Paramagnetic means that a material is attracted to an applied magnetic eld, like iron lings
to a magnet. Many elements (gold, potassium, calcium,...) are paramagnetic. Most compounds
and stable molecules are weakly paramagnetic. Paramagnetic materials generally have unpaired
electrons that are free to align the external eld, while diamagnetic materials have closed orbits.
The opposite of paramagnetic is diamagnetic, which means something is repelled by an applied
magnetic eld (google levitating frog for example). A third type of magnetism is ferromag-
netism, whereby a material can produce a coherent magnetic eld, i.e. as in an ordinary iron
refrigerator magnet. If you heat an iron magnet above 1043 K it will lose its magnetism and
become paramagnetic. The transition between paramagnetic and ferromagnetic phases of iron is
an example of a phase transition.
The ferromagnetism in iron comes about when all the magnetic spins in iron are aligned. The
spins themselves can be thought of as little magnets that attract each other.

Figure 9. When a magnetic material is cooled, its spins spontaneously align as it enters the ferromagnetic
phase.

In this crude model, each pair has energy ¡" if they are aligned and +" if they are not aligned.
So the dierence between the fully aligned (ferromagnetic) state and disordered (paramagnetic)
state is roughly E = 2N". On the other hand, the disordered state has much higher entropy.
The entropy of the fully-aligned magnetic state is zero. But the disordered state has = 2N
congurations and entropy S = NkB ln 2. Thus the free energy in the magnetic state is

Fmagentic = E ¡ TS = ¡N" (27)

and the free energy of the disordered state is

Fdisordered = E ¡ TS = N (" ¡ TkB ln 2) (28)

General phase transitions 13

Thus we see by minimizing the free energy that the transition from paramagnetic to ferromagnetic
state occurs at
"
kBTc  (29)
ln 2
Tc is known as the Curie temperature.
What is a good order parameter for the transition? Entropy S should work, as it goes from
S = N kB ln2 down to zero. Unlike the liquid/gas transition however, the entropy change as the
magnet is cooled will be smooth: at temperatures near Tc the magnetic will have some spins aligned,
be slightly ferromagnetic and have some intermediate entropy. Thus the ferromagnetic phase
transition is second order.
Another order parameter we could consider is the magnetization M . The direction of the



magnetic eld is a vector M , and we can dene M = M . Above Tc, M = 0 exactly. As T is
lowered, M is nonzero. This function M (T ) is going to be continuous across Tc. However all of
the derivatives of M (T ) cannot be continuous  if they were then the function would have to be
M (T ) = 0 exactly for all T (mathematically, this is a property of analytic functions).


It is interesting to think of the direction of M rather than its magnitude as the order parameter.


In the magnetic phase, the direction of M picks up some denite value (there can be domains
inside the magnetic with dierent directions, but let's just focus on one domain for now). One


super interesting thing about M picking up a direction is that we cannot know ahead of time
which direction it would be. At high temperature no direction is preferred. Indeed, all the spins
are constantly ipping around in 3D. So the theory at high temperature is rotationally invariant.
At low temperature, it is not. But fundamentally, the interactions between spins are like ~s1
~s2
involving a rotationally-invariant dot product. So the Hamiltonian is rotationally invariant and
it is only the state that violates the symmetry. When this happens we say that the rotational
symmetry is spontaneously broken.
Symmetries are an extraordinarily powerful tool in physics. In this context, they help specify
phases and phase transitions. For another example, note that a liquid is invariant under trans-
lations. Microscopically, of course the molecules are in particular positions. But knowning the
position of some of the molecules does not tell us anything about where molecules far away are.
On the other hand, a solid is not translationally invariant. Once you know where one atom is, you
can pinpoint all the rest throughout the crystal. We say the solid has long-range order. So when
a liquid freezes, translational symmetry is spontaneously broken and long-range order results.
There are lots of consequences of spontaneous symmetry breaking. One is that it tells us that
there must be arbitrarily low energy excitations of the system. This very general result is known as
Goldstone's theorem. For example, in a solid, we know we can push it and it will move. It can
move arbitrarily slowly, so we can push it with arbitrarily low energy to have an eect. However,
as we push it, what we actually do is push the atoms on the side where our hand is. These atoms
push the next atoms, and so on, all throughout the solid. So really, to move the solid by pushing,
we are setting up a wave of very low energy. As we move an atom, the system works to restore the
lattice to how it was. This doesn't happen in a liquid where the translation symmetry is unbroken.
The excitations of a solid are called phonons and have a massless dispersion relation !(k) ! 0 as
k ! 0. Phonons are covered in Lecture 11. The excitations in a magnet from the spontaneously
broken rotational symmetry also have !(k) ! 0 as k ! 0. They are called spin waves.
To understand spontaneous symmetry breaking, the emergence of long-range order, etc. requires
techniques of condensed matter physics that take us well beyond the course material.

4.2 Critical phenomena

The nal topic I want to mention concerns that dot at the end of the liquid/gas phase boundary
denoted as the critical point. Critical points are super-interesting places with a lot of unusual
properties.
14 Section 4

Recall that as you heat up a liquid, it will eventually vaporize. At the point of the phase
transition, the heat will go into latent heat of vaporization and the temperature will not change.
Eventually, all the liquid is vaporized and the heat will start raising the temperature again. We
can see this fairly clearly in a phase diagram in the pressure-volume plane, as this one for CO2:

V
Figure 10. P-V phase diagram for CO2. The x-axis is the specic volume, v = N
. The green region has
liquid and gas.

The lines in this plot are lines of constant temperature, or isotherms. Say we start on left
in the liquid region, along the T = 13C isotherm. As we decompress the CO2 isothermally, we
will move along the isotherm, lowering pressure and increasing the specic volume (lowering the
density). When we hit the phase boundary, the liquid starts vaporizing. During the phase transition
@P
pressure and temperature are xed, so we move horizontally, @v = 0. After the phase transition is
complete, CO2 is all gas, and the decompression lowers the pressure and density once again. So
the isotherm gives a decreasing function P (v) outside of the coexistence region and is at in the
coexistence region.
A natural order parameter for the transition is the number density n, or equivalently the specic
1
volume, v = n . We see from the gure that the specic volume changes discontinuously from liquid
to gas along the T = 13C isotherm, so the transition at this temperature is 1st order.
Now consider what happens as the temperature is increased. As you can see from the gure,
at a higher temperature, the dierence in specic volume between the liquid and gas phases at
constant temperature is smaller. The latent heat of vaporization is smaller too. Eventually, there
is no dierence between the liquid and gas phases and the latent heat vanishes: it takes no energy
to convert a liquid to a gas. This happens at the critical temperature Tc, which intersects the
liquid/gas coexistence region at the critical specic volume vc and critical pressure Pc, that is, at
the critical point. The specic volume changes smoothly from liquid to gas if we pass through the
critical point, so the phase transition at this point is second order.
At temperatures above the critical temperature, the material is both gas and liquid, or neither
gas nor liquid, depending on how you look at it. We call it a supercritical uid. The super
in this context just means beyond  in contrast to the super in superuids or superconductors
which are truly exotic phases of matter. A supercritical uid is in between a gas or a liquid. For
example, supercritical CO2 is used to decaeinate coee  its viscosity and diusivity are like
those of a gas, so it penetrates the beans easily, and its density is like that of a liquid, so a lot of
it can get in. It happens also to bind well to caeine (this property is much more important that
its supercritical uid properties). SupercriticalCO2 is also used in dry cleaning.
@P
On any isotherm in the liquid/gas region, @v
= 0. At the critical point, the length of the
T
horizontal part of the isotherm has gone to zero, but it is still at.
 Moreover,
 since the isotherms
@ 2P
are decreasing on either side of the critical point, we know that @v 2 = 0 as well: the critical
T
General phase transitions 15

point is a point of inection. It also

 happens to be true that, but is not so easy to show, that all of
@ nP
the derivatives of P vanish, @v n
= 0. Mathematically, this means that P (v) is a non-analytic
T
function at the critical point. Second order phase transitions are super interesting because somehow
this crazy mathematical behavior, where all the derivatives vanish, arises out of functions like the
entropy S or the partition function Z that depend smoothly on temperature, pressure, volume, etc.
 n 
@ P
It's not just the derivatives @v n that vanish. We could equally well have looked at the phase
T
diagram in the T ¡ v plane:

Figure 11. TV diagram for water

In this case, at subcritical pressure, water increases its specic volume when heated until it boils.
At the critical point, the latent heat vanishes and the water  and  steam become the same. The
@ nT
critical point is also a point of inection for T (v), and in fact, @v n = 0, so T (v) is a non-analytic
P
function.
Away from the critical point, the various thermodynamic quantities that we have discussed,
latent heat, enthalpy of formation, heat capacity, isothermal compressibility, etc., help us dis-
tinguish one material from another. But these  nare all 
related to derivatives of thermodynamic
@ P @ nT
quantities. Because all the derivatives vanish, @v n = @v n = 0, all the dimensionful physical
T P
quantities we use to characterize a material either vanish or are innite at this point. For example,
here is a plot of heat capacity of propane near the critical point:

Figure 12. Heat capacity of propane near the critical point, showing the singularity.
16 Section 4

One way to understand the disappearance of scale more physically is to think about water. The
dierence between liquid water and an ideal gas is that water has hydrogen bonds (see Fig. 2). In
water vapor, the molecules are generally too far apart for hydrogen bonds to matter. The relative
importance of hydrogen bonds in water versus gas determines the surface tension and everything
else that makes water a liquid. However, as the density or pressure on the gas is increased, the
relative importance hydrogen bonds in the vapor phase increases too. Consequently the surface
tension of the liquid/gas boundary goes down. So typical droplet sizes grow. At the critical point,
the surface tension vanishes and droplets of any size can form: the single dimensional scale (the
surface tension ) has vanished. This can be seen through the phenomenon of critical opalescence.
We say that the theory near the critical point is scale-invariant or conformal, since no
dimensionful quantity is available to characterize the material. That is, choosing units relative to
the critical values:
T P v 1 n 
T^ = ; P^ = ; v^ = ; n^ = = = (30)
Tc Pc vc v^ nc c
all thermodynamic quantities, such as the free energy, become independent of any other property
of the material . For example, Pc and Tc for neon, argon, krypton, xenon, N2, O2, CO and CH4 are

Ne Ar Kr Xe N2 O2 CO CH4
Tc (C) ¡228.7 1122.3 ¡63.8 16.6 ¡147 ¡118.4 ¡140 ¡82.1 (31)
Pc (atm) 26.9 48 54.3 58 33.5 50.1 34.5 45.8

These temperatures and pressures are dimensionful quantities, with no apparent relation among
them. Now, we rescale the temperatures, pressures, specic volumes and specic densities by these
critical values and look at T^ and n^ near the critical points for the dierent materials:

Figure 13. Reduced temperature versus reduced number density for a variety of dierent substances at
saturation. Adapted from E.A. Guggenheim, J. Chem. Phys. 13, 253 (1945).

Remarkably, in the vicinity of the critical point, the T^ ¡ n

^ phase boundaries all have exactly the
same shape! This implies that all of the derived quantities, such as heat capacity, isothermal
compressibility, etc, should be related in every material near its critical point. This powerful
observation is known as the law of corresponding states.
The shape in Fig. 13 is t by a functions n ^(T^) on the liquid and gas side of the form
3 7 3 7
n^`(T^) = 1 + (1 ¡ T^) + (1 ¡ T^)1/3; ^g(T^) = 1 + (1 ¡ T^) ¡ (1 ¡ T^)1/3
n (32)
4 4 4 4
The fractional exponent makes these functions non-analytic, so that all the derivatives are singular
at the critical point T^ = 1, n
^(k)(1) = 1. This means all the derivatives of T^(n^) vanish at the critical
point n ^ (k)
^ = 1, T (1) = 0.
General phase transitions 17

The exponent 1 / 3 in Eq. (32) is an example of a critical exponent. It characterizes the

approach of the density towards the critical point. If we used an order parameter other then
density, for example, the heat capacity, it would approach the critical point with a dierent scaling
behavior CV  (1 ¡ T^)¡ . Because of the law of corresponding states, the critical exponents can
be calculated with any material for which the order parameter applies. In fact, the material can
be a made-up theoretical one: the universality is so strong that the material doesn't even have to
exist. An important example is the ising model, which treats a material as a lattice of spins with
interaction energies taking the values 1. Computing the critical exponents of the ising model
agrees with measured values of the critical exponents in water to one part in 1000!