Summary class 12-14: Entropy and the second Law (2nd part)

2.6 Phase transitions [Fermi 15; parts of Schroeder 5.3]

The entropy change along isotherms under the

dome is:

Thus the entropy change is simply related to the latent

heat DH. This gives us formulas for the volume
dependence of energy and entropy under the dome:

a-g: isotherms; dashed line: saturation dome

C… critical point L..liquid; G…gas, LV…liquid
and saturated vapour coexistence; V…non
saturated vapour. [from Fermi]

Note: Fermi distinguishes gas from vapour.

Many others don’t. Vapour here is defined as
saturated gas; For T>Tc V and L cannot be
distinguished anymore and hence is called
‘gas’.

Phase diagram for H2O (not to scale). [Figure 5.11 from

Schroeder]

The liquid-vapor (or solid-vapor) saturation pressure ps(T) varies strongly with temperature -
the plot is log-linear and covers eight orders of magnitude! Note the Fig 5.11 above has strongly
non-linear scales to bring out certain features. The ‘real’ phase diagram for water can be found
on the next page:

Note that the slope of the solid-liquid transition line is negative (at atmospheric pressures),
meaning that, at temperatures slightly below freezing, you can melt solid ice by compressing it
(a lot)! This is unusual among substances, and will turn out to be related to the fact that liquid
water is denser than ice (i.e. ice floats)
Phase diagram of H2O

https://commons.wikimedia.org/wiki/File:P
hase_diagram_of_water.svg
At 100kPa atmospheric pressure:
100% relative humidity @ 20C -> 3% molar fraction of air are H2O;
100% relative humidity @ 10C -> 1% molar fraction of air are H2O;
100% relative humidity @ -10C -> 0.1% molar fraction of air are H2O;

This explains why dew condenses as the temperature cools during the night, why your breathe
instantly condenses in winter, and why the cold air you bring into your house is so dry (even
after being heated up)! And since the temperature decreases rapidly with altitude as we've
seen in lecture 6, this also limits the amount of water (=clouds) that can be found high up.
We see that saturation curves ps(T) are important to explain many phenomena. We will now
learn how to compute their slope. To do this, we will make a slight detour.

2.7 T-S diagrams and slope of saturation curves [Feynman I.45]

The shape of any phase transition curve ps(T) is constrained by the laws of thermodynamics,
specifically, the fact that entropy exists as a state variable. This sounds incredibly abstract, and
the best way to understand this will be to introduce T-S diagrams. These are in fact widely used
when engineering engines and refrigerators so let's start with this more concrete perspective.

There are clear engineering advantages to have a working substance undergo a phase
transition:
• High energy capacity (ie. the latent heat for vaporizing water is 5x the energy needed to
bring from 0 to 100C)
• Fast energy transfer (bubbles of steam in water quickly move away, by buoyancy)
• Temperature is automatically constant at a phase transition (providing the reversible
‘isotherms’ in Carnot's ideal cycle)
Let us thus contemplate an ideal refrigerator cycle that would exploit a phase change, and draw
it on both p-V and T-S diagrams.

Exercise: Again, try to assign best names:

Exercise: The cycle has 2 isotherms and 2 adiabats, but it Compression, Expansion, Condensation, and
looks more like 5 steps. Assign to each! evaporation

The above cycle is ‘ideal’: being made of an isotherm and adiabat, it minimizes waste (Qh for a
refrigerator). However, eye-balling it, it is hard to see what's ideal about it. This motivates a
new set of axes on which to plot the state of a substance. With respect to T and S, an ideal cycle
is (by definition) a rectangle!
Because it is easy to eye-ball whether a T-S cycle is optimal, T-S diagrams are widely used
in engineering applications. They are also quite useful for formal theory:

How are we to compute net work (and ultimately, efficiency) from a T-S diagram? An amazing
fact comes from the first law, in the form relating state functions: dU = TdS - pdV . Since the
internal energy does not change under a closed cycle , this relate the work done and
heat input:
This equality can be applied to any cycle, and this yields interesting relations between various
state variables. These are simplest and most useful for infinitesimal cycles. To make areas easy
to compute, assume that in each diagram, two sides are parallel to one of the axes. For
example a small cycle with two isochores and two isotherms (constant V and constant T):

The parallel edges make both areas trivial to compute. Since they must be equal:

This is known as a Maxwell relation. (By the same argument one could get four variants of it,
that fix either p or V , and either T or S.)

Maxwell's relation holds for any system, and this will be interesting when we discuss the free
expansion of non-ideal gases (‘throttling’). For now, we focus under the saturation dome,
where the formula simplifies dramatically. There we have seen that

and that p = ps(T) does not actually depend on volume anyway, so

We thus get the Clausius-Clapeyron equation:

It predicts the slope of ps(T), e.g. the liquid-gas curve in the phase diagram, in terms of
measurable quantities: the latent heat and volume difference.

Clausius-Clapeyron equation applies to any phase transformation, not just liquid-gas.

2.8 Maximum efficiency of engines and refrigerators [Schroeder 4.2-4.4]
Now that we have a form of the second law that applies to any substance (DStot ≥ 0), let us
revisit the efficiency of engines.
We will look at a refrigerators, which are very similar to an engine just with arrows reversed. A
refrigerator uses work W to suck heat from its cold interior at Tc, expelling into a room at Tr:

The ‘efficiency’ in this case is called ‘coefficient of performance COP:

The first and second law will limit the COP, irrespective of the details
of the cycle. Since the working substance operates on a closed cycle,
DUsubstance = 0, energy balance gives Qout = Qin +Win -> Qr = Qc + W

Energy conservation alone does not limit the COP. We could have COP=1000 by taking Qc
huge (say 1000 in some units) and W small (say 1), and then we would still have energy
conserved and Qr positive. Such a device however would mostly move heat from cold to warm,
which ought to be restricted by the second law.

To apply the second law DStot ≥ 0, we must add the entropy changes from all components,
in principle three (interior, room, and working substance). The working substance operates on a
closed cycle, so immediately DSsubstance = 0 (entropy is a state function).

and from this:

Let us finally return to the heat engine considered in lecture 10, which takes heat Qh at a hot
temperature Th and dumps Qc at Tc. The efficiency will be a one-line calculation using state
variables! Whatever the closed cycle is, we have DUsubstance = DSsubstance = 0 since these are
state variables; the first and second law then read:

From these 2 equations we get:

These results apply to any closed cycle, and have nothing to do with the Carnot cycle: all we
used is the 1st and 2nd Law. The Carnot cycle just happens to realize the optimal efficiency.

One cannot beat the Carnot efficiency!

2.9 Intermolecular forces and non-ideal gases [parts of Schroeder 5.4]

Real molecules interact with each other, even from afar. Consider a sample of gas in a fixed
volume and given temperature. For an ideal gas, U and p/T wouldn't depend on volume, but
how about a real gas? Do interactions make p/T larger or smaller?

We will first answer the question about U by thinking about interaction energies; the answer
for p/T will follow using the magic of thermodynamics.

Many interactions can be modeled by a two-body intermolecular potential.

A common model is the Lennard-Jones ‘6-12’ potential:

How does this affect energy and pressure in a real gas? Let's figure
out U first. The important thing is that intermolecular potential
decays very fast: each of N molecules only sees its immediate
neighbors within a distance rmax ~ few times s, where the potential
is of order -e. The number of visible neighbors is thus proportional
to s3 and the density, N/V , and summing over all molecules
gives:
The Lennard-Jones interatomic potential
[from Wikipedia]
The attractive force lowers energy, and the more so as the gas
becomes denser. This remains valid as long as the gas is not too dense to probe the repulsive
part of the potential.
How does this impact pressure? We return to Maxwell's relation:
Viewing all variables S and U as functions of independent
variables T and V , this reads, by definition:

This is valid for any variation. Since the phase

space is two-dimensional, the variations dV and dT are independent so to be fully explicit this
amounts to two independent equations:
We just need the second formula. Plugging Maxwell's relation
into it we get a relation that does not involve entropy:

This formula relates the energetics of the gas to its equation of state. Like Maxwell's relation, it
is valid for any substance, and follows just from the existence of entropy as a state variable.

An interesting application is to a substance which satisfies the ideal gas law. Then the right-
hand-side vanishes, so we have proven the following thermodynamics theorem, which we
had first motivated empirically:

Theorem: If a substance satisfies the ideal gas law, then its energy density depends only on
temperature and not on volume.
Throttling and cooling [parts of Schroeder 4.4]
Let us return to free expansion of a gas (‘Joule expansion’). We saw in chapter 1 that for an
ideal gas, the temperature does not change when it expands into a vacuum.
How about a real gas?

The aN2/V term increases with volume, and so U ↓:

an actual gas will cool slightly. Intuitively, as the
molecules get further from each other, they escape an attractive potential well and slow down.

A closely related phenomenon is when we push a gas through a small plug, or constriction that
restricts its flow. This is called throttling or the Joule-Thomson process.

2.10 More realistic cycles [Schroeder 4.4]

Note: Schroeder uses pV diagrams, we use TS diagrams.

How is a real refrigerator cycle different from ideal ones?

Any rectangle in a TS diagram is an ideal (Carnot) cycle: two isotherms and two adiabats. To
practically realize the isotherms, we would like fit the cycle under the saturation dome, e.g. to
stay in the mixed phase, where isotherms are easily realized since they're the same as isobars:

Take a minute to try and place the following four

words on the cycle: (adiabatic) Compression,
Expansion, conDensation, and eVaporization.

The trick to locate the first two is that p and T go up

together. Under (adiabatic) compression,
p must go up, so this is the right-edge. Expansion is
then the left-hand-side. (It may be surprising that
heating a liquid near the left-edge of the saturation
dome creates more liquid! The reason is that liquid is
mostly incompressible, and it expands upon warming (this is true even for liquid water). So if
you heat a liquid-gas mixture that is already mostly liquid, the liquid will grow to take up all the
volume and the gas will disappear. Lowering the pressure under the saturation dome reduces
the temperature, for a liquid basically by evaporation.)
Along horizontal lines, phases change. From left to right, liquid becomes gas, so eVaporation is
bottom and conDensation is top.

The ideal cycle has two problems in practice:

• Real compressors don't like mixed phases. Basically, a compressor moves a substance by
pushing on it, but pushing on the mixed phase might just change its composition instead of
moving it. Also, there are issues of wear due to liquid droplets.
• Doing the expansion adiabatically (in an attempt to recover some work) is not worth the
trouble, because the substance is basically liquid then and liquid volumes change very little
(pdV is not big).

The first issue is usually addressed by doing the compression in the gas phase. The second issue
is addressed by just ‘letting the gas freely expand’ (similar to Joule expansion, e.g. removal of a
barrier). In practice this is achieved by throttling: passing the substance (mostly liquid then)
through a set of small holes, or cotton plug.

This process is not a continuous sequence of equilibria, so (as for Joule expansion) it is not clear
what path to draw. The solution: we will NOT DRAW any path! The system just disappears and
reappears at a lower temperature. S increases in the throttling step. Furthermore, in the actual
cycle the temperature overshoots in the right part of the cycle,
because in practice the pressure is held constant and T ∝ V at
constant pressure. In terms of schematics and pV diagram,
that cycle is:

To numerically compute the COP

or efficiency of real refrigerators,
one needs to look at the changes
in enthalpy H because heat
transfers are done at constant
pressure. (See Schroeder 4.4 for
more.)