7.

Phase Equilibrium and Phase Transition

We consider systems where particle numbers can change (i.e., open systems), e.g., a glass
of water, which exchange the number water molecules with air by evaporation, etc.
Thermodynamic relation of open systems: in chapter 5, we have slipped in a term 𝜇𝑁 in
thermodynamic potentials like 𝑈 𝑆, 𝑉, 𝑁, … = 𝑇𝑆 − 𝑝𝑉 + 𝜇𝑁. Correspondingly, the
thermodynamic identity is
𝑑𝑈 = 𝑇𝑑𝑆 − 𝑝𝑑𝑉 + 𝜇𝑑𝑁
To see why, recall that for a system with fixed particle number 𝑁, 𝑑𝑈 = 𝑇𝑑𝑆 − 𝑝𝑑𝑉. Define 𝑢 = 𝑈/𝑁,
𝑠 = 𝑆/𝑁, and 𝑣 = 𝑉/𝑁, the above can be rewritten 𝑑𝑢 = 𝑇𝑑𝑠 − 𝑝𝑑𝑣 for fixed 𝑁. Now consider cases
where 𝑁 can vary, then
𝑑𝑈 = 𝑑 𝑁𝑢 = 𝑢𝑑𝑁 + 𝑁𝑑𝑢 = 𝑢𝑑𝑁 + 𝑁 𝑇𝑑𝑠 − 𝑝𝑑𝑣 = 𝑢𝑑𝑁 + 𝑇(𝑁𝑑𝑠) − 𝑝(𝑁𝑑𝑣)
But by 𝑑 𝑥𝑦 = 𝑥𝑑𝑦 + 𝑦𝑑𝑥, we get 𝑁𝑑𝑠 = 𝑑 𝑁𝑠 − 𝑠𝑑𝑁, 𝑁𝑑𝑣 = 𝑑 𝑁𝑣 − 𝑣𝑑𝑁, so the above can be
replaced as
𝑑𝑈 = 𝑢𝑑𝑁 + 𝑇 𝑑 𝑁𝑠 − 𝑠𝑑𝑁 − 𝑝 𝑑 𝑁𝑣 − 𝑣𝑑𝑁 = 𝑢𝑑𝑁 + 𝑇𝑑𝑆 − 𝑇𝑠𝑑𝑁 − 𝑝𝑑𝑉 + 𝑝𝑣𝑑𝑁
= 𝑇𝑑𝑆 − 𝑝𝑑𝑉 + 𝑢 − 𝑇𝑠 + 𝑝𝑣 𝑑𝑁 = 𝑇𝑑𝑆 = 𝑝𝑑𝑉 + 𝜇𝑑𝑁
where we define: 𝜇 = 𝑢 − 𝑇𝑠 + 𝑝𝑣, called chemical potential.

Recall 𝐺 = 𝑈 − 𝑇𝑆 + 𝑝𝑉, obviously for open systems,

𝐺 = 𝜇𝑁 Euler’s relation
and 𝜇 = 𝐺/𝑁
Also, 𝑑𝐺 = 𝑑 𝑈 − 𝑇𝑆 + 𝑝𝑉 = 𝑑𝑈 − 𝑇𝑑𝑆 − 𝑆𝑑𝑇 + 𝑝𝑑𝑉 + 𝑉𝑑𝑝 = 𝑇𝑑𝑆 − 𝑝𝑑𝑉 + 𝜇𝑑𝑁 − 𝑇𝑑𝑆 −
𝑆𝑑𝑇 + 𝑝𝑑𝑉 + 𝑉𝑑𝑝, so we arrive at the fundamental relation for an open system:

𝑑𝐺 = −𝑆𝑑𝑇 + 𝑉𝑑𝑝 + 𝜇𝑑𝑁

𝑑𝐺
Obviously, 𝜇 = , and the natural variables of 𝐺 include 𝑇, 𝑝, and 𝑁, etc.:
𝑑𝑁 𝑇,𝑝
𝐺 = 𝐺(𝑇, 𝑝, 𝑁, … )
For a system containing more than one component, one has
𝐺 = ∑𝜇𝑖 𝑁𝑖
𝑑𝐺 = −𝑆𝑑𝑇 + 𝑉𝑑𝑝 + ∑𝜇𝑖 𝑑𝑁𝑖
𝑑𝐺
𝜇𝑖 = 𝑑𝑁𝑖 𝑇,𝑝,𝑁 ,…𝑁 ,𝑁 ,…
1 𝑖−1 𝑖+1

𝑑𝐺 𝑑𝐺
and 𝑆=− , 𝑉=
𝑑𝑇 𝑝,𝑁1 ,… 𝑑𝑝 𝑇,𝑁 ,…
1

Finally, by the fact that 𝑑𝐺 = 𝑑 ∑𝜇𝑖 𝑁𝑖 = ∑𝜇𝑖 𝑑𝑁𝑖 + ∑𝑁𝑖 𝑑𝜇𝑖 = −𝑆𝑑𝑇 + 𝑉𝑑𝑝 + ∑𝜇𝑖 𝑑𝑁𝑖 ,
one immediately has the Gibbs-Duhem relation:
𝑆𝑑𝑇 − 𝑉𝑑𝑝 + ∑𝑁𝑖 𝑑𝜇𝑖 = 0
which indicates that among the 𝑛 + 2 intensive variables, only (𝑛 + 1) are independent.
• Phases, Phase Equilibrium:
- A Phase refers to a homogeneous region of matter in which there is no spatial
variation in average density, energy, composition, or other macroscopic properties.
Note that a phase does NOT necessarily contain only one chemically pure substance, e.g., a gas
of uniformly mixed He and N2 is of a single phase. On the other hand, a mixture of water and ice
cube in a bottle is of two phases coexist.

- Phase equilibrium refers to a state under a certain constraint (say a fix 𝑇 and 𝑝),
where two or more phases of a system (water and ice, say) coexist without any
change on macroscopic properties. It corresponds to a state of maximum entropy
(𝑑 2 𝑆 ≤ 0) or minimum Gibbs energy (𝑑 2 𝐺 ≥ 0).

The 𝑆 𝑈 relation leading to phase separation

- Phase transition refers to a discontinuous change in the property of a substance as its
environment is changed infinitesimally. Examples include ice melting, water boiling. As
elaborated above, phase transition occurs when a change leads to increase in total
entropy (or equivalently decrease of Gibbs energy of the system).
- Phase diagram: conditions (e.g., temperature and pressure) under which equilibrium
phases exist is usually presented in a phase diagram. As an example, the figure show
below is a phase diagram of water, where lines separate water-ice, water-vapor, and
vapor-ice phases. At conditions that lie on the lines, two relevant phases coexist in
equilibrium (e.g., on AB, liquid and valor phases of water coexist). Note there is a special
set of condition of (𝑇, 𝑝), where the three
lines intersect. At this point, three phases
of water, liquid, solid (ice), and gas (vapor)
coexist. This point is referred to the Triple
Point of Water (0.01 °C, 4.56 mmHg), or
(273.16 K, 611.66 Pa) , which we have
introduced and used as the reference point
for temperature scales.
• Gibbs phase rule: the number of independent intensive variables specifying an
equilibrium composite system containing 𝐾 components and coexisted in 𝑃 phases is
𝑓 = 𝐾+2−𝑃

This is seen by the following argument: consider a system composed of P phases (solid, liquid,
gas, …), in each there are K different chemical components (i.e., different substances, H2O,
alcohol, sugar, …), in equilibrium, one demands
𝑇 (1) = 𝑇 (2) = ⋯ = 𝑇 (𝑃)
𝑝(1) = 𝑝(2) = ⋯ = 𝑝(𝑃)
(1) (2) (𝑃)
𝜇𝑖 = 𝜇𝑖 = ⋯ = 𝜇𝑖 , 𝑖 = 1,2, … 𝐾

each contains (𝑃 − 1) equations and thus a total of 𝑃 − 1 𝐾 + 2 equations. Solve these

equations to eliminate some of the 𝐾 + 2 𝑃 variables, so the the number of independent
intensive variables one needs to specify a composite system in equilibrium is 𝑓 = 𝐾 + 1 𝑃 −
𝑃 − 1 𝐾 + 2 = 𝐾 + 2 − 𝑃, known as the Gibbs phase rule.

For example, for water (K = 1), along the coexistence lines, P=2, so f = 1, i.e., there is only one
independent intensive variable (either T or p, indeed, the two are dependent on the coexistence
line). At triple point, P = 3, so f = 0, i.e., no independent variable at triple point (this is why ice,
water and vapor coexist only at one single combination of T and p – no freedom of choice!
• Clausius-Clapeyron equation
Consider two neighboring points A and B on a coexistence line (see figure). On this
line, phase ‘I’ (e.g., vapor) and phase ‘II’ (e.g., liquid water) are in equilibrium, thus
have equal chemical potentials: 𝜇𝐴𝐼 = 𝜇𝐴𝐼𝐼 , and 𝜇𝐵𝐼 = 𝜇𝐵𝐼𝐼 . So, along the line AB,
𝑑𝜇𝐼 (𝑇, 𝑝) = 𝑑𝜇𝐼𝐼 (𝑇, 𝑝). Noting that 𝑑𝜇 = −𝑠𝑑𝑇 + 𝑣𝑑𝑝, therefore on the line,
−𝑠 𝐼 𝑑𝑇 + 𝑣 𝐼 𝑑𝑝 = −𝑠 𝐼𝐼 𝑑𝑇 + 𝑣 𝐼𝐼 𝑑𝑝
or
𝑑𝑝 Δ𝑠 𝑙
= Δ𝑣 = 𝑇Δ𝑣
𝑑𝑇

i.e., a change in temperature 𝑑𝑇 will lead to a change in

pressure 𝑑𝑝, which is called the Clausius-Clapeyron
relation (or simply Clapeyron relation). In the above,
𝑙 = 𝑇(𝑠 𝐼𝐼 − 𝑠 𝐼 ) is the latent heat per molecule of phase
transition. As l and Δ𝑣 are all experimentally accessible,
the slope of line AB is then determined.
Example: Carbon has two known solid phases, diamond and graphite. At
ambient conditions (298K and 1 bar), graphite is more stable, as it has a
lower Gibbs free energy than that of diamond by 2900 J/mol. Show that
as the pressure and temperature increases, diamond can be more stable
due to the fact that their Gibbs energies change with T and p differently
(see figure). Also, find the slope of the phase boundary at 500K.
𝜕𝐺
Answer: by the relation 𝑑𝐺 = −𝑆𝑑𝑇 + 𝑉𝑑𝑝, one recognizes that = 𝑉,
𝜕𝑝 𝑇,𝑁
where the molar volume of diamond and graphite are 5.31  10-6 m3 and
𝜕𝐺
3.42  10-6 m3, respectively. Also, = −𝑆, and graphite is known to
𝜕𝑇 𝑝,𝑁
have a higher entropy than diamond. Therefore, there exists a phase
transition line as shown.

To find the slope of phase boundary between diamond and graphite at 500K, we look for the entropy
changes of 1 mole graphite and diamond as the temperature changes from 298 K to 500 K
𝑇 𝐶𝑝
by Δ𝑆 = ‫ 𝑇׬‬2 𝑑𝑇, one gets
1 𝑇
500𝐾 0.0414𝑇−6.1
Δ𝑆 𝐷𝑖𝑎𝑚𝑜𝑛𝑑 = ‫׬‬298𝐾 𝑑𝑇 ≈ 5.21 J/K
𝑇
500𝐾 1 8.54×105
Δ𝑆 𝐺𝑟𝑎𝑝ℎ𝑖𝑡𝑒 = ‫׬‬298𝐾 (16.86 + 4.77 × 10−3 𝑇 − )𝑑𝑇 ≈ 6.59 J/K
𝑇 𝑇2
Looking at thermodynamic table, one also finds 𝑆 𝐷𝑖𝑎𝑚𝑜𝑛𝑑 298𝐾 ≈ 2.38 J/K; 𝑆 𝐺𝑟𝑎𝑝ℎ𝑖𝑡𝑒 298𝐾 ≈ 5.74 J/K,
so
𝑆 𝐷𝑖𝑎𝑚𝑜𝑛𝑑 500𝐾 ≈ 2.38 + 5.21 = 7.59 J/K; 𝑆 𝐺𝑟𝑎𝑝ℎ𝑖𝑡𝑒 500𝐾 ≈ 5.74 + 6.59 = 12.33 J/K
𝑑𝑝 𝐿 Δ𝑆
Then, according to the Clapeyron relation: = = , at 500 K, one has
𝑑𝑇 𝑇Δ𝑉 Δ𝑉
𝑑𝑝 12.33−7.59
≈ ≈ 2.52 × 106 Pa/K = 25.2 bar/K.
𝑑𝑇 5.31−3.42 ×10−6
Example: (i) Find the triple point pressure and temperature of ammonia; (ii) Find the heat of vaporization,
sublimation, and fusion, respectively.
𝑑𝑝 𝐿 𝐿 𝐿𝑝
Answer: By Clapeyron equation: = ≈ = , as 𝑉𝑔 ≫ 𝑉𝑙,𝑠 , and 𝑝𝑉𝑔 = 𝑅𝑇 for 1 mole. The
𝑑𝑇 𝑇Δ𝑉 𝑇𝑉𝑔 𝑅𝑇 2
𝑑𝑝 𝐿
liquid-vapor boundary is expressed as ‫׬‬ =‫׬‬ 𝑑𝑇, i.e.,
𝑝 𝑅𝑇 2
𝐿
(1) ln 𝑝 = − + const .
𝑅𝑇
Similar form exists for the sublimation line, i.e., solid-vapor boundary.

Experimentally, one finds that the liquid /solid ammonia has vapor pressures:
3063
(2) ln 𝑝 = 24.38 − Liquid
𝑇
3754
(3) ln 𝑝 = 27.29 − Solid
𝑇
Comparing (1) and (2); (1) and (3), one gets the latent heats of sublimation and vaporization to be
37548.31(R) = 3.12 x 104 J and 30638.31 = 2.55 x 104 J, respectively.
3,063 3,754
(i) At triple point, (2) = (3), i.e., 24.38 − = 27.29 − , solve for 𝑇, one gets triple point of
𝑇 𝑇
ammonia 𝑇𝑡 = 195.2 K. Plug 𝑇𝑡 back in (2) or (3), one find the triple point pressure of ammonia 𝑝𝑡 =
5934 Pa.

(ii) At triple point, 𝐿𝑠𝑢𝑏𝑙𝑖𝑚𝑎𝑡𝑖𝑜𝑛 = 𝐿𝑓𝑢𝑠𝑖𝑜𝑛 + 𝐿𝑣𝑎𝑝𝑜𝑟𝑖𝑧𝑎𝑡𝑖𝑜𝑛 ,

So, 𝐿𝑓𝑢𝑠𝑖𝑜𝑛 = 𝐿𝑠𝑢𝑏𝑙𝑖𝑚𝑎𝑡𝑖𝑜𝑛 − 𝐿𝑣𝑎𝑝𝑜𝑟𝑖𝑧𝑎𝑡𝑖𝑜𝑛 = 3.12 − 2.55 × 104 J.
• Van der Waals (vdW) Fluid
Let’s examine phase transition (between liquid and gas) of the vdW fluid, whose
equation of state is
𝑎𝑁2
𝑝+ 𝑉 − 𝑁𝑏 = 𝑁𝑘𝐵 𝑇, where 𝑎, 𝑏 constants
𝑉2

This is also an approximate expression of the EoS of real gasses as introduced before.

The plot below shows the isotherms of the vdW fluid in the 𝑝 − 𝑉 diagram. At high
temperature, it is seen to resemble an ideal gas, while as temperature lowers, it deviates
from the ideal gas law. Particularly below a critical temperature 𝑇𝐶 , one finds that there
𝜕𝑉
exist portions of the isotherms where 𝜕𝑝 > 0 , meaning that as one increases
𝑇
1 𝜕𝑉
pressure, the volume of the system increases. But 𝜅 𝑇 = − 𝑉 < 0 means the
𝜕𝑝 𝑇
system is unstable as a uniform phase but would like to separate into two (or more)
phases.
 By 𝑑𝐺 = −𝑆𝑑𝑇 + 𝑉𝑑𝑝 + 𝜇𝑑𝑁, and for fixed 𝑁
and 𝑇, it simply reduces to 𝑑𝐺 = 𝑉𝑑𝑝. Diving
by 𝑑𝑉, one gets
𝜕𝐺 𝜕𝑝 𝑁𝑘𝐵 𝑇𝑉 2𝑎𝑁2
=𝑉 =− +
𝜕𝑉 𝑁,𝑇 𝜕𝑉 𝑁,𝑇 𝑉−𝑁𝑏 2 𝑉2

and integrate over 𝑉, one has:

𝑁𝑘𝐵 𝑇 𝑁𝑏 2𝑎𝑁2
𝐺 = −𝑁𝑘𝐵 𝑇 ln 𝑉 − 𝑁𝑏 + − + const.
𝑉−𝑁𝑏 𝑉
Instead of 𝐺(𝑉), one use the EoS to obtain the relation of 𝐺(𝑝) and plotted on left above for a
particular isotherm at temperature 𝑇 < 𝑇𝐶 (right). One notes that the segment 2 → 6 in the plot does
not correspond to stable state (having higher Gibbs energies). As pressure goes up, at “2”, the system
directly jumps to “6”, which corresponds to an abrupt change in volume – a phase transition from ‘gas’
to ‘liquid’. For the system with volume in between “2” and “6”, the thermodynamically stable state is
the combination of two phases – gas and liquid.

• Maxwell construction: The pressure at which phase transition

occurs is the horizontal line that make the shaded areas the same
below and above the line. To see why, we recognize that the net
change of G must be zero when one goes through the loop 2-3-4-5-6
(as point ‘2’ and ‘6’ in the G(p) diagram coincide), i.e.,:
𝜕𝐺
𝐺6 − 𝐺2 = ‫𝑝𝑜𝑜𝑙׬ = 𝐺𝑑 𝑝𝑜𝑜𝑙׬‬ 𝑑𝑝 = ‫ = 𝑝𝑑𝑉 𝑝𝑜𝑜𝑙׬‬0
𝜕𝑝 𝑇

But the integration of ‫ 𝑝𝑑𝑉 𝑝𝑜𝑜𝑙׬‬is precisely the total shaded area in the right Figure above.
Repeating for different 𝑇’s, one obtains the vapor pressure curve.
• Metastability – Binodal and Spinodal Lines
In the G(p) plot, one sees that the segments 2-3 and 5-6 are “concave” with p, which
indicates they are thermodynamically allowed. Indeed, in these two ranges, the 𝑝 − 𝑉
curve satisfies 𝜅 𝑇 > 0, meaning that the states in such ranges are thermodynamically
permeable. In these two segments (between the two red dotted lines in the left figure below
(also the right figure in the 𝑇 − 𝜌 diagram), the outer one is the ‘coexistence or binodal
curve’ (also shown in the middle), while the inner one is called ‘spinodal curve’), where
though globally unstable but acceptable as it satisfies 𝜅 𝑇 > 0. If existed, the system is in the
metastable state – state having local but not global minimum in the free energy.
Examples of metastable states include superheated or super-cooled water – the former is a
water of >100 oC without boiling, while the latter is a water at < 0 oC without freezing.
Johannes Diderik van der Waals (Nov. 1837 – Mar. 1923)
A Dutch theoretical physicist, famous for his pioneering work
on the equation of state for gases and liquids, his 1873 thesis
work, which won him Nobel Prize in physics in 1910. His name
is also associated with Van der Waals forces (forces between
stable molecules). Van der Waals's work of real gasses
affirmed the reality of molecules and allowed an assessment
of their size and attractive strength at the time when the
molecular structure of fluids had not been accepted by most
physicists, and liquid and vapor were also considered as
chemically distinct. This, his formula revolutionized the study
of gasses and their equations of state, and the effect of his
work on molecular physics in the 20th century was direct and
fundamental.
• Second-order (continuous) phase transition*
Phase transition we introduced above, for example, liquid water transforms into vapor
steam, all involve distinct changes in macroscopic properties at the phase boundary, such
as density (molar volume), and indeed the energy, entropy and particularly enthalpy, i.e.,
Δℎ ≠ 0. It is the chemical potential, or Gibbs free energy per particle for single-
component system, that is equal at phase boundary. This type of phase transitions is
referred to as the first-order phase transition, a name stemmed from the fact that the
first-order derivatives, such as 𝜕𝑔/𝜕𝑇 = −𝑠, 𝜕𝑔/𝜕𝑝 = 𝑣, that are discontinuous.
There are also cases, such as water-vapor phase transition at the critical point, in which
Δℎ = Δ𝑔 − 𝑇Δ𝑠 = Δ𝜇 − 𝑇Δ𝑠 = −𝑇Δ𝑠 = 0, and so Δ𝑠 = 0, i.e., there is no latent heat,
nor entropy change at phase transition. In other words, 𝜕𝑔/𝜕𝑇 = −𝑠 remain continuous
(so the name of “continuous” phase transition). For such, there could be discontinuity in
the second-order derivatives of Gibbs free energy, e.g., 𝜕 2 𝑔/𝜕𝑇 2 = −𝜕𝑠/𝜕𝑇 = −𝑐𝑝 /𝑇,
𝜕 2 𝑔/𝜕𝑝2 = 𝜕𝑣/𝜕𝑝 = −𝑣𝜅 𝑇 , and 𝜕 2 𝑔/𝜕𝑇𝜕𝑝 = 𝜕𝑣/𝜕𝑇 = 𝑣𝛼𝑝 (so the name of second-
order (continuous) phase transition). Examples of second-order phase transition,
besides the water-steam transition at critical point, include paramagnetic-ferromagnetic
transition of a magnetic system, normal metal-superconductor transition, etc. The study
of the second order transition is the province of modern research and beyond the scope of
this course, which is not to be elaborated further.