Thermal Physic A Sample Material

A Students Handbook of Physics
Part III. Thermal Physics

February 2009
2
Contents
1 The Kinetic Theory of Ideal Gases 5
1.1 Ideal gases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2 The Boltzmann distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3 The Maxwell-Boltzmann distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3.1 Velocity distributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.3.2 Speed distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.4 Properties of gases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.4.1 Particle energies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.4.2 Heat capacity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.4.3 Pressure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.4.4 Particle ux . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.4.5 Mean free path . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.5 Establishing equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
1.5.1 Chemical equilibrium: particle transport . . . . . . . . . . . . . . . . . . . . . . 19
1.5.2 Mechanical equilibrium: momentum transport . . . . . . . . . . . . . . . . . . 23
1.5.3 Thermal equilibrium: heat transport . . . . . . . . . . . . . . . . . . . . . . . . 23
2 Classical Thermodynamics 27
2.1 Thermodynamic systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.2 The Laws of Thermodynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.2.1 The Zeroth Law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.2.2 The First Law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.2.3 The Second Law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.2.4 The Third Law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
2.3 Thermodynamic potentials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
2.3.1 Internal energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
2.3.2 F, H and G . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
2.3.3 The Maxwell relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
2.3.4 Chemical potential . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3 Real gases and phase transitions 47
3.1 Real gas equation of state . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.1.1 The van der Waals equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.1.2 The Boyle temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
3.1.3 The coexistence region . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
3.1.4 Isothermal compressibility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.2 Phase transitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.2.1 Conditions for phase equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.2.2 Maxwell construction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.2.3 Phase diagrams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
3
4 CONTENTS
3.2.4 Clausius Clapeyron equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
3.3 Irreversible gaseous expansions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.3.1 Joule expansion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
3.3.2 Joule-Thomson expansion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
4 Statistical mechanics 63
4.1 Statistical ensembles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
4.2 The postulates of statistical mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
4.2.1 States and energy levels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
4.2.2 Microstates and macrostates . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
4.2.3 Properties of . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.3 The density of states . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
4.3.1 Derivation in k-space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
4.3.2 Converting to energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
4.4 Classical statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
4.4.1 The Boltzmann distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
4.4.2 The number density distribution . . . . . . . . . . . . . . . . . . . . . . . . . . 74
4.4.3 The partition function and thermodynamic properties . . . . . . . . . . . . . . 74
4.4.4 Agreement with kinetic theory . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
4.5 Indistinguishable particles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
4.5.1 Exchange symmetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
4.5.2 Bosons and fermions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
4.6 Quantum statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
4.6.1 Grand canonical partition function . . . . . . . . . . . . . . . . . . . . . . . . . 82
4.6.2 Quantum occupation functions . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
4.6.3 The classical approximation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
4.6.4 Black body radiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
Chapter 1
The Kinetic Theory of Ideal Gases
Kinetic theory describes the gross thermal
properties of ideal gases, from a consideration of
the average behaviour of the individual particles
that make up the gas. For example, by knowing
the probability distribution of particle energies or
velocities in an ideal gas, kinetic theory is able to
predict large-scale properties of the gas, such as
its internal energy, heat capacity, pressure and
ux.
1.1 Ideal gases
Kinetic theory applies only to ideal gases, which
constitute a simple model of real gases. In the
ideal gas model, we treat gas particles as hard
spheres that are moving randomly (Brownian mo-
tion), and constantly colliding with each other
and with the walls of the container. The assump-
tions of ideal gases and kinetic theory are:
1. Molecular size intermolecular separation
(particle volumes are negligible)
2. Intermolecular forces are negligible
3. All collisions are perfectly elastic, which
means that momentum and energy are con-
served between the colliding objects
4. Energy exchange between particles is solely
due to elastic collisions
5. The molecular speed distribution is constant
in time
All gases behave like ideal gases at very low pres-
sures and high temperatures, since under these
conditions the particles relative volumes and in-
termolecular forces are negligible. Inert gases,
such as neon and argon, behave like ideal gases
even at low temperatures, since they have com-
plete outer electron shells and therefore weak in-
termolecular forces.
Ideal gas law
A dening characteristic of ideal gases is that
they obey the ideal gas law. This law relates
the macroscopic properties pressure, p [Pa], tem-
perature, T [K], and volume, V [m
3
], for a gas
with n moles (Eq. 1.1). It applies only to systems
in equilibrium, meaning that their macroscopic
properties are not changing with time.
Ideal gas law:
pV = nRT (1.1)
= Nk
B
T (1.2)
where the molar gas constant is:
R = N
A
k
B
=
N
n
k
B
(1.3)
The ideal gas law comes from an amalgama-
tion of three other laws, which were discovered
empirically:
1. Boyles law: V 1/p at constant T and n
2. Charless law: V T at constant p and n
3. Avogadros law: V n at constant T and p
Therefore, combining these we have
pV
nT
= constant = R
5
6 CHAPTER 1. THE KINETIC THEORY OF IDEAL GASES
The constant of proportionality is the molar gas
constant, R [8.31 JK
1
mol
1
]. This is related
to the Boltzmann constant, k
B
[1.3810
23
JK
1
]
by Eq. 1.3. Avogadros number, N
A
[6.02210
23
mol
1
], gives the number of gas particles, N, per
mole of gas:
N
A
=
N
n
where n is the number of moles. Therefore, by
eliminating R from the ideal gas law, we can ex-
press it in terms of the Boltzmann constant, k
B
,
for an ideal gas with N particles (Eq. 1.2).
1.2 The Boltzmann distribution
A postulate of kinetic theory is that particles in
an ideal gas obey the Boltzmann distribution of
energy,
1
f() (Eq. 1.4). This is a probability dis-
tribution, where f() d gives the probability that
a particle has an energy between and + d.
The Boltzmann distribution is an exponentially
decaying function of energy, and depends only on
the temperature of the gas (Fig. 1.1). The expo-
nential term in Eq. 1.4 is called the Boltzmann
factor, and the coecient, A, is the normalisa-
tion constant. The value of A is determined from
the requirement that the total probability, which
is the Boltzmann distribution integrated over all
energies, equals unity (Eqs. 1.51.6).
The Boltzmann distribution:
f()d = A e
/k
B
T
. .
Boltzmann
factor
d (1.4)
where the normalisation condition is
_
0
f()d = 1 (1.5)
Therefore, the normalisation constant is
A =
1
_
0
e
/k
B
T
d
(1.6)
1
For a derivation of the Boltzmann distribution, see
Section 4.4.
f()
T
hot
T
cold
d
Figure 1.1: Variation of the Boltzmann distri-
bution with temperature and energy (not nor-
malised). The probability of a particle having an
energy between
0
and
0
+d at T
cold
is given by
the shaded area. Mark
0
on gure + plot using
mathematica.
Precision of statistical methods
Its not hard to convince yourself that using sta-
tistical methods and probability distributions for
systems with many particles is very precise. Con-
sider, for example, one mole of gas, which has
N
A
10
24
particles. If we split the Boltzmann
distribution into 10
6
bins of equal width d,
then there are, on average, of order N
b
10
18
particles per bin. Therefore, the standard devia-
tion for each bin is

1
N
b
10
9
This means that statistical uctuations in the
number of particles per bin are of order one part
in a billion. So the dierence between a real dis-
tribution and a statistically-derived mean distri-
bution is essentially negligible.
1.3 The Maxwell-Boltzmann
distribution
The Maxwell-Boltzmann distribution is a small
extension from the Boltzmann distribution;
whereas the latter gives the probability that a
1.3. THE MAXWELL-BOLTZMANN DISTRIBUTION 7
particle has a given energy, the former gives the
probability that a particle has a given velocity.
For an ideal gas, in the absence of external forces
such as gravity, the total energy of a particle
equals its kinetic energy, which means that
=
1
2
mv
2
Therefore, eliminating from the Eq. 1.4, we can
write the probability distribution in terms of par-
ticle velocity:
f
v
(v) = Ae
mv
2
/2k
B
T
In this case, the Boltzmann factor is a Gaussian
function,
2
whereas for the Boltzmann distribu-
tion of energies, it is an exponentially decreasing
function. The subscript in f
v
(v) therefore indi-
cates that this is not the same function as the
Boltzmann distribution, which was denoted sim-
ply by f(). It is important to be consistent with
notation, since there are several ways of express-
ing the Maxwell-Boltzmann distribution:
3D velocity distribution: f
v
(v) d
3
v
1D velocity distribution: f
v
(v
x
) dv
x
speed distribution: F
v
(v) dv
The velocity distributions take into account the
particles direction, and therefore are functions of
the velocity vector (v in 3D or v
x
in 1D), whereas
the speed distribution depends only on the mag-
nitude, v, and is a function dierent from f
v
(thus
denoted by F
v
). Below, we will look at each dis-
tribution in turn.
1.3.1 Velocity distributions
Lets rst look at the 1D velocity distribution,
f
v
(v
x
) dv
x
. Using = mv
2
x
/2 in Eq. 1.4, the 1D
velocity distribution is
f
v
(v
x
)dv
x
= Ae
mv
2
x
/2k
B
T
dv
x
This gives the probability of a particle having
a velocity in the x-direction between v
x
and
v
x
+dv
x
. (Similar distributions exist for velocity
in either the y or the z direction.) The normalisa-
tion constant, A, is determined by integrating the
distribution over all 1D velocities ( to +),
2
A Gaussian is a function of e
x
2
.
dv
x
dv
z
dv
y
v
Figure 1.2: Element of velocity-space, d
3
v =
dv
x
dv
y
dv
z
, for a particle with velocity v.
and setting the integral equal to one, as shown in
Derivation 1.1. This condition ensures that the
total probability equals unity. The normalised
velocity distribution is a Gaussian distribution
(Eq. 1.7).
Velocity distributions: (Derivations 1.1 and
1.2)
In 1D:
f
v
(v
x
) dv
x
=
_
m
2k
B
T
e
mv
2
x
/2k
B
T
. .
Boltzmann
factor
dv
x
(1.7)
In 3D:
f
v
(v) d
3
v =
_
m
2k
B
T
_3
2
e
mv
2
/2k
B
T
. .
Boltzmann
factor
d
3
v (1.8)
It is straightforward to generalise to 3D, by
changing v
x
to the total velocity, v. The mag-
nitude of the total velocity is
v
2
= v
2
x
+v
2
y
+v
2
z
and the 3D element of velocity-space (Fig. 1.2) is
d
3
v = dv
x
dv
y
dv
z
The expression for the 3D distribution, f(v) d
3
v,
is also a Gaussian distribution (Eq. 1.8 and
Derivation 1.2). Eq. 1.8 gives the probability that
a particles velocity lies in an element of velocity-
space d
3
v from velocity v (Fig. 1.2).
Table 1.1: Derivation of 1D velocity distribution (Eq. 1.7)
f
v
(v
x
)dv
x
= A
e
mv
2
x
/2k
B
T
dv
x
= A
2k
B
T
m
Therefore, from the normalisation condition:
A =
_
m
2k
B
T
f
v
(v
x
)dv
x
=
_
m
2k
B
T
e
mv
2
x
/2k
B
T
dv
x
Starting point: The Maxwell-Boltzmann
distribution for velocity in 1D, f
v
(v
x
)dv
x
is given by substituting = mv
2
x
/2 into
Eq. 1.4. The normalisation condition re-
quires that the integral over all velocities
equals one:
f
v
(v
x
)dv
x
= 1
To integrate, use the standard integral:
e
x
2
dx =
_
Table 1.2: Derivation of 3D velocity distribution (Eq. 1.8)

f
v
(v) d
3
v = (f
v
(v
x
) dv
x
) (f
v
(v
y
) dv
y
) (f
v
(v
z
) dv
z
)
=
_
m
2k
B
T
_3
2
e
m(v
2
x
+v
2
y
+v
2
z
)/2k
B
T
d
3
v
=
_
m
2k
B
T
_3
2
e
mv
2
/2k
B
T
d
3
v
The 3D velocity distribution is the
product of the three 1D component
distributions, where:
d
3
v = dv
x
dv
y
dv
z
and v
2
= v
2
x
+v
2
y
+v
2
z
Use the normalised 1D distribu-
tion from Eq. 1.7. (f
v
(v
y
) dv
y
and
f
v
(v
z
) dv
z
are identical to f
v
(v
x
) dv
x
,
but with dierent subscripts).
1.4. PROPERTIES OF GASES 9
f(v)
v 0
= kT/m
Figure 1.3: Maxwell-Boltzmann 1D velocity dis-
tribution. The distribution is a Gaussian cen-
tred about the mean velocity, v
x
), which is zero
since there is no preferred direction. The stan-
dard deviation is
_
k
B
T/m, and increases with
increased temperature. The shaded area repre-
sents the probability of a particle having a veloc-
ity between v
0
and v
0
+dv. The total area under
the curve is equal to unity (total probability = 1).
Plot using mathematica. Show Thot and Tcold
and v
0
and v +dv
0
.
Average values
Both the 1D and 3D velocity distributions are
Gaussian distributions, centred about the origin.
The standard deviation for a Gaussian of the
form e
x
2
is = 1/
2. Therefore, the stan-

dard deviations for the velocity distributions are
temperature-dependent:
1D
=
k
B
T
m
(1.9)
and
3D
= 3
1D
=
3k
B
T
m
(1.10)
The centre of the distribution corresponds to the
average velocity, v
x
), and the standard devia-
tion corresponds to the rms velocity,
_
v
2
x
). In
worked example 1.4, these average values are cal-
culated explicitly using Eq. 1.7. The 1D velocity
distribution is shown in Fig. 1.3.
1.3.2 Speed distribution
The speed distribution, F
v
(v) dv, gives the prob-
ability of a particle having a speed between v and
v + dv, and thus is independent of the particles
direction of travel. If we express the 3D velocity
distribution in spherical polar coordinates, then
the 3D element of velocity-space is:
d
3
v = v
2
sindv d d (1.11)
Therefore, by integrating over angular coordi-
nates and , we can remove the directional-
dependence of Eq. 1.8 (Derivation 1.3). The
speed distribution is therefore given in Eq. 1.12.
This distribution has the same Gaussian Boltz-
mann factor as the velocity distributions, but the
entire distribution is no longer Gaussian since the
Boltzmann factor now is multiplied by the fac-
tor of v
2
, which appeared from Eq. 1.11. This
means that the probability distribution is skewed
towards higher speeds, as shown in Fig. 1.5. The
average speed, v), the rms speed,
_
v
2
), and the
modal speed, v
m
, which are labelled in Fig. 1.5,
are calculated in worked example 1.6.
Speed distribution: (Derivation 1.3)
F
v
(v) dv = 4v
2
_
m
2k
B
T
_3
2
e
mv
2
/2k
B
T
. .
Boltzmann
factor
dv
(1.12)
1.4 Properties of gases
The Maxwell-Boltzmann probability distribu-
tions can be used to determine average velocities
and speeds, as shown in worked examples 1.4 and
1.6. We can then use these averages to explain
the macroscopic properties of systems in equilib-
rium, as shown below.
1.4.1 Particle energies
Using the expression for the average squared
speed,

v
2
_
, from worked example 1.6, the av-
erage translational kinetic energy per particle is:
KE)
3D
=
1
2
m
_
v
2
_
=
3
2
k
B
T (1.13)
Or, for translational motion in one-dimension:
KE)
1D
=
1
2
m
_
v
2
x
_
=
1
2
k
B
T (1.14)
Figure 1.4: Worked example: Using Eq. 1.7 to calculate average values of particle velocity
1. The average velocity, v
x
), equals the aver-
age of the Gaussian distribution,
x
. The
average velocity equals zero since there is
no preferred direction of travel.
2. The root mean square velocity,
_
v
2
x
), does
not depend on direction, therefore we would
expect it to be non-zero. We see that
_
v
2
x
)
equals the standard deviation,
x
, of the
Gaussian distribution.
3. Since the three components v
x
, v
y
and v
z
are independent, the square of the standard
deviation of the total velocity distribution
is equal to the sum of the square of the com-
ponents:
2
=
2
x
+
2
y
+
2
z
Therefore, the root mean square of the total
velocity is:
_
v
2
) =
_
v
2
x
_
+
_
v
2
y
_
+
v
2
z
_
=
3k
B
T
m
1. v
x
) =
v
x
f
v
(v
x
)dv
x
=
_
m
2k
B
T
v
x
e
mv
2
x
/2k
B
T
dv
x
= 0

x
2.
_
v
2
x
_
=
v
2
x
f
v
(v
x
)dv
x
=
_
m
2k
B
T
v
2
x
e
mv
2
x
/2k
B
T
dv
x
=
k
B
T
m

2
x
To integrate, use the standard integrals:
xe
x
2
dx = 0
x
2
e
x
2
dx =
1
2
_
Table 1.3: Derivation of the speed distribution (Eq. 1.12)

F
v
(v)dv =
2
_
0
_
0
f
v
(v)v
2
sindv d d
= v
2
_
m
2k
B
T
_3
2
e
mv
2
/2k
B
T
dv
2
_
0
d
_
0
sin d
= 4v
2
_
m
2k
B
T
_3
2
e
mv
2
/2k
B
T
dv
Starting point: The speed distribu-
tion is found by integrating the 3D
velocity distribution (Eq. 1.8) over all
angles and , to remove the direc-
tional dependence of the velocity.
The element of 3D velocity space in
spherical polar coordinates is
d
3
v = v
2
sin dv d d
Figure 1.6: Worked example: Using Eq. 1.12 to calculate average particle speeds
1. The average speed, v), is given by the aver-
age of the speed distribution, F
v
(v). The in-
tegration limits are 0 to since we are not
taking direction into account, and therefore
cannot have negative speeds.
2. The root mean square speed,
_
v
2
), equals
the root mean square velocity calculated
from the 3D distribution (see worked ex-
ample 1.4). This follows since the square of
the velocity of a particle is identical to the
square of its speed.
3. The modal speed is the turning point of
F
v
(v), and therefore is the speed at which
_
dF
v
(v)
dv
_
v
m
= 0
4. The ratio of average speeds is therefore
v
m
: v) :
_
v
2
) =

2 :
_
8
3
1 : 1.13 : 1.22
(refer to Fig. 1.5)
To integrate, use the standard integrals:
_
0
x
3
e
x
2
dx =
1
2
2
_
0
x
4
e
x
2
dx =
3
8
2
_
1. v) =
_
0
vF
v
(v)dv
= 4
_
m
2k
B
T
_
3/2

_
0
v
3
e
mv
2
/2k
B
T
dv
=
8k
B
T
m
2.
_
v
2
_
=
_
0
v
2
F
v
(v)dv
= 4
_
m
2k
B
T
_
3/2

_
0
v
4
e
mv
2
/2k
B
T
dv
=
3k
B
T
m
3. Dierentiate F
v
(v) and set derivative equal
to zero:
v
m
=
2k
B
T
m
T
cold
T
hot
F(v)
v
0
v
m
Figure 1.5: Maxwell-Boltzmann speed distribu-

tion. The shaded area represents the proba-
bility of a particle having a speed between v
0
and v
0
+ dv. The total area under the curve is
therefore equal to unity (total probability = 1).
For hot systems, the distribution is skewed to-
wards higher particle speeds. The ratio of aver-
age speeds is v
m
: v) :
_
v
2
) = 1 : 1.13 : 1.22,
and is independent of both T and m. Plot using
mathematica. Show Thot and Tcold and v
0
and
v +dv
0
.
These expressions show that the kinetic energy
of particles in a gas depends only on its temper-
ature. There are three independent degrees of
freedom
3
for translational kinetic energy, corre-
sponding to motion in the x, y or z directions,
respectively. Therefore, Eqs. 1.13 and 1.14 show
that there is a kinetic energy of k
B
T/2 associ-
ated with each translational degree of freedom.
Since the average kinetic energy per particle de-
pends only on the temperature of the gas, it is
often referred to as the thermal energy of the sys-
tem. Atoms are said to be in thermal motion as a
result of their being at a particular temperature,
and their thermal motion increases with increased
temperature. Owing to collisions, the thermal
motion of atoms appears as a random walk. In
larger particles, the random walk due atomic and
molecular collisions is known as Brownian mo-
tion.
Internal energy
The internal energy, U [J], of a gas is the sum of
the kinetic and potential energies of its individual
3
A degree of freedom in this context is a parameter that
denes an independent direction of movement.
particles. The potential energy includes chemical
energy stored in atomic bonds or intermolecular
forces. Since an ideal gas has by denition no
intermolecular forces, the potential energy of its
particles is assumed to be zero. Therefore the in-
ternal energy of an ideal gas results only from the
kinetic energy of its particles. For a monatomic
ideal gas of N particles, with three translational
degrees of freedom (x, y and z), the internal en-
ergy of the gas is therefore:
U
1D
= N KE)
3D
=
3
2
Nk
B
T
3
2
nRT (1.15)
where we have used Eq. 1.3 to eliminate nk
B
.
Similarly, for translational motion in one-
dimension:
U
1D
= N T)
1D
=
1
2
Nk
B
T
=
1
2
nRT (1.16)
Therefore the internal energy of an ideal gas de-
pends only on its temperature. (For real gases,
intermolecular forces are not negligible, therefore
their internal energy is a function of the gass vol-
ume as well as its temperature.)
Eqs. 1.15 and 1.16 show that there is an inter-
nal energy of nRT/2 associated with each degree
of freedom. Diatomic ideal gases also possess ro-
tational and vibrational kinetic and potential en-
ergies, which means they have additional degrees
of freedom. Linear diatomic gases have two ad-
ditional rotational degrees of freedom, and two
additional vibrational degrees of freedom, which
are illustrated in Fig. 1.7. (Need to draw this bet-
ter.) For an ideal gas with f independent degrees
of freedom, it is found that its internal energy is:
U =
f
2
Nk
B
T
=
f
2
nRT (1.17)
where f = 3 for a monatomic ideal gas (trans-
lational motion in x, y or z directions). For a
linear diatomic gas undergoing rotation with no
Figure 1.7: Degrees of freedom for a linear di-
atomic ideal gas. Rotation and vibration each
introduce two additional degrees of freedom.
vibrations, f = 5, and if the gas also has radial
vibrations, then f = 7.
Using Eq. 1.17 together with the ideal gas law,
we nd that the relationship between pressure,
p [Pa], and internal energy per unit volume, u
[Jm
3
], is:
p =
2
f
u (1.18)
Therefore, for an ideal monatomic gas, where f =
3, the pressure is
p =
2
3
u
The equipartition theorem
Equation 1.17 forms the basis of the equipartition
theorem, which states that the internal energy
of a system is shared equally among its dier-
ent contributions. Said another way, with each
degree of freedom, there is an energy factor of
RT/2 [J/mole] or k
B
T/2 [J/molecule] (Eqs. 1.19
and 1.20).
Equipartition theorem: The average energy
associated with each quadratic degree of free-
dom is
E) =
1
2
RT per mole (1.19)
1
2
k
B
T per molecule (1.20)
The equipartition theorem holds true provided
that the energy has a quadratic dependence on
the particular degree of freedom.
4
For example,
the dependence of kinetic energy on v or p is
quadratic:
KE =
1
2
mv
2
p
2
2m
Alternatively, the dependence of the potential en-
ergy of a harmonic oscillator is quadratic in x:
V
SHO
=
1
2
kx
2
where k [Nm
1
] is the spring constant. Refer to
classical mechanics section?
1.4.2 Heat capacity
We saw above that the temperature of an ideal
gas is a measure of its internal energy. Heat ow
into or out of an ideal gas changes its internal
energy, and hence its temperature. The heat ca-
pacity, C [JK
1
], of a system is the amount of
heat, Q, required to raise its temperature by 1 K.
It is often expressed as a partial derivative, and
is given by Eq. 1.21. The variable denes the
constraint on the system. For example, if heat is
added at constant pressure, then = p, whereas
if heat is added at constant volume, then = V .
The specic heat capacity, c [JK
1
kg
1
], is the
heat capacity per unit mass, and the molar heat
capacity, c
m
[JK
1
mol
1
], is the heat capacity
per unit mole (Eqs. 1.22 and 1.23 respectively).
Heat capacity:
C
=
_
Q
T
_
(1.21)
c =
1
m
_
Q
T
_
(1.22)
c
m
=
1
n
_
Q
T
_
(1.23)
The actual rise in temperature with increasing
energy for a given substance depends on the na-
ture of its particles. For example, for monatomic
4
For a statistical mechanics treatment, see Sec-
tion 4.4.4.
gases, the extra energy only goes into increas-
ing their translational kinetic energy, whereas for
diatomic gases, the energy also increases their
rotational or vibrational energies. Therefore, in
qualitative terms, for the same amount of input
energy, the temperature of a monatomic gas in-
creases more than for a diatomic gas.
C
V
and C
p
The heat capacity depends on which variable is
held constant while heat is added to the system.
The two most common heat capacities are the
heat capacity at constant volume, C
V
, and heat
capacity at constant pressure, C
p
. From the con-
servation of energy, we nd that C
V
is equal to
the change of the systems internal energy with
temperature:
C
V
=
dU
dT
(1.24)
(Derivation 1.4). Using the expression for U from
Eq. 1.17 in this, we nd that the heat capacity at
constant volume is proportional to the number of
degrees of freedom, f, of the system (Eq. 1.25).
Therefore, since a monatomic gas has three trans-
lational degrees of freedom, its heat capacity at
constant volume for one mole of gas (n = 1) is:
C
V
=
3
2
R
C
p
is greater than C
V
since the energy goes into
increasing the volume of the gas as well as its
internal energy (its temperature). Using the con-
servation of energy, it is found that the dier-
ence in heat capacities, C
p
C
V
, is equal to nR
(Eq. 1.27 and Derivation 1.4). Therefore, the heat
capacity at constant pressure is given by Eq. 1.26.
For one mole of monatomic gas, this gives:
C
p
=
5
2
R
The ratio of heat capacities is a constant and is
referred to as the adiabatic index, (Eq. 1.28).
For a monatomic ideal gas, the adiabatic index is
therefore:
=
5
3
C
V
and C
p
for ideal gases:
(Derivation 1.4)
Heat capacities for f degrees of freedom
C
V
=
f
2
nR (1.25)
C
p
=
f + 2
2
nR (1.26)
Dierence in heat capacities:
C
p
C
V
= nR (1.27)
Ratio of heat capacities
C
p
C
V
= (1.28)
=
f + 2
f
(1.29)
Aside on latent heat: In this section, we have
only considered ideal gases. However, when a
substance changes state, for example solid to liq-
uid or liquid to gas, then heat input does not
increase the bodys temperature, but instead is
absorbed by the bonds between particles, which
therefore increases the potential energy of the
particles. Since the kinetic energy of the parti-
cles remains constant, the temperature of the sub-
stance does not change during a change of state.
The amount of heat absorbed or released per unit
mass during such a transition is called the specic
latent heat,
5
L [Jkg
1
] (Eq. 1.30). The latent heat
of fusion is the energy required to melt unit mass
of solid to a liquid, whereas the latent heat of va-
porisation is the energy required to vaporise unit
mass of liquid to a gas. Melting or boiling are
endothermic processes, which means that latent
heat is absorbed, whereas condensation or freez-
ing are exothermic processes, meaning that latent
heat is released.
Latent heat:
L =
Q
m
(1.30)
5
The word latent is derived from the Latin latere, to
lie hidden.
Table 1.4: Derivation of expressions for C
V
and C
p
. (Eqs. 1.24 and 1.27)
dU = dQ+dW
= dQpdV
Constant volume:
_
U
T
_
V
=
_
Q
T
_
V
C
V
C
V
=
dU
dT
Constant pressure:
dQ = dU dW
= C
V
dT +pdV
_
Q
T
_
p
= C
V
+p
_
V
T
_
p
C
p
= C
V
+nR
Starting point: From the conservation of energy,
the element of increase in internal energy of a sys-
tem, dU, equals the element of heat added, dQ,
plus the element of work done by the system, dW:
dU = dQ+dW
(This is the First Law of Thermodynamics; see
Section 2.2.2)
The work done equals the energy needed to ex-
pand the gas against pressure, which results from
the molecules impulse (see Section 2.2.2)
dW = pdV
Constant volume: Divide the expression for dU
by dT, holding V constant. From the deni-
tion of heat capacity (Eq. 1.21), C
V
results from
the change in internal energy with temperature.
Since U is a function of T only, for an ideal gas,
then C
V
can be written as the total derivative,
dU/dT, instead of a partial derivative.
Constant pressure: Divide the expression for dQ
by dT, holding p constant. Use the ideal gas law,
pV = nRT, to determine the volume-temperature
derivative at constant pressure.
1.4.3 Pressure
From worked example 1.6, the average squared
speed is:
_
v
2
_
=
3k
B
T
m
Using this to eliminate k
B
T from the ideal gas
law, we can express the pressure in terms of

v
2
_
:
p =
1
3
mn
p
_
v
2
_
where n
p
[m
3
] is the particle number density:
n
p
=
N
V
We reach the same result if we assume that the
pressure results from the force per unit area ex-
erted by molecular collisions with the container
walls (Eq. 1.31 and Derivation 1.5). This con-
rms that pressure is due to molecular impulse,
and not due to intermolecular repulsion as was
originally thought by Newton.
Ideal gas pressure: (Derivation 1.5)
p =
1
3
mn
p
_
v
2
_
(1.31)
1.4.4 Particle ux
We can use a similar treatment to determine an
expression for the particle ux, [m
2
s
1
], which
is the number of particles striking a surface per
unit area and per unit time. The particle ux is
found to depend on the particle number density,
n
p
, and the average particle speed, v) (Eq. 1.32
and Derivation 1.6). These, in turn, depend on
the gass pressure and temperature, therefore the
particle ux can be expressed in terms of p and
T (Eq. 1.33). This shows that particle ux re-
lates the microscopic motion of particles to the
gass macroscopic properties, pressure and tem-
perature.
Particle ux: (Derivation 1.6)
=
1
4
n
p
v) (1.32)
2mk
B
T
(1.33)
Eusion
Eusion is the process by which particles emerge
through a small hole. The rate of eusing parti-
cles per unit area, R
E
[m
2
s
1
], is equal to the
particle ux, , striking the hole if it were closed
o:
R
E
=
1
4
n
p
v)
Expressing the average velocity in terms of the
Maxwell-Boltzmann speed distribution,
v) =
_
vF
v
(v)dv
where F
v
(v)dv is provided in Eq. 1.12, we nd
that
R
E
=
1
4
n
p
_
vF
v
(v)dv
_
F
E
(v)dv
Therefore, we can dene the eusion rate per
unit area in terms of the velocity distribution
of eusing particles, F
E
(v). Since the Maxwell-
Boltzmann speed distribution, F
v
(v), depends on
v
2
, the eusion distribution, F
E
(v), is found to
depend on v
3
:
F
E
(v)dv =
1
4
n
p
vF
v
(v)dv
v
3
e
mv
2
/2k
B
T
dv (1.34)
Therefore, high velocity particles have a greater
probability of eusing since the probability dis-
tribution is skewed towards higher v. This means
that both the pressure and temperature of a gas
in an eusing vessel decrease, since high energy
particles are leaving. Eusion has practical ap-
plications in isotope separation, since the eusion
rate depends on the mass of the eusing particle:
R
E
v)
1
m
In consequence, the heavier isotope is progres-
sively enriched in the eusing vessel with time.
1.4.5 Mean free path
The mean free path, [m], is the average dis-
tance travelled by a particle before it undergoes
Table 1.5: Derivation of the pressure of an ideal gas (Eq. 1.31)
vdt
dS

Cylinder volume = vdt dScos
v v
v = 2v cos
v cos
dp =
F
dS
dN
=
F
dS
N
C
N
(F
v
(v)dv)
=
2mv cos
dS dt
(n
p
v dt dS cos )(sin d d)(F
v
(v)dv)
4
=
1
2
mn
p
v
2
F
v
(v)dv cos
2
sin d d
p =
1
2
mn
p
v
2
F
v
(v)dv
. .
=v
2
/2
_
0
cos
2
sind
2
_
0
d
= mn
p
_
v
2
_
_
1
3
cos
3
_
/2
0
=
1
3
mn
p
_
v
2
_
Starting point: An element of pres-
sure, dp, is given by the force F, per
unit area dS, multiplied by the num-
ber of particles dN, striking the con-
tainer wall in a time dt.
For a particle density of n
p
[m
3
], dN
is given by:
dN = N
C
N
(F
v
(v)dv)
where N
C
= n
p
v dt dS cos
N
=
d
4
=
sin d d
4
N
C
is the number of particles within
the cylinder (refer to diagram);
F
v
(v)dv is the fraction of those with
the correct velocity (between v and
v + dv); N
is the fraction of those

within the correct solid angle, d.
The force exerted on dS due to a
change in particle momentum is
F =
dp
dt
= m
dv
dt
= m
2v cos
dt
(refer to diagram)
Integrating over all space, we nd the
total pressure exerted by all particles
in a gas, in terms of m, n
p
and

v
2
_
.
Table 1.6: Derivation of particle ux (Eq. 1.32)
vdt
dS
Cylinder volume = vdt dScos

d =
dN
dSdt
=
N
C
N
(F
v
(v)dv)
dSdt
=
(n
p
v dt dS cos )(sindd)(F
v
(v)dv)
4dSdt
=
1
4
n
p
vF
v
(v)dv cos sindd
=
1
4
n
p
vF
v
(v)dv
. .
=v
/2
_
0
cos sind
2
_
0
d
=
1
2
n
p
v)
_
1
2
cos
2
_
/2
0
=
1
4
n
p
v)
=
p
2mk
B
T
Starting point: The particle ux is de-
ned as the number of particles striking
per unit area dS, per unit time dt.
For a particle density of n
p
m
3
, dN is
given by:
dN = N
C
N
(F
v
(v)dv)
where N
C
= n
p
v dt dS cos
N
=
d
4
=
sin d d
4
N
C
is the number of particles within the
cylinder (refer to diagram); F
v
(v)dv is
the fraction of those with the correct ve-
locity (between v and v +dv); N
is the
fraction of those within the correct solid
angle, d.
Integrating over all space, we nd the
expression for the ux, , in terms of n
p
and v).
From the ideal gas law, p = n
p
k
B
T, and
from worked example 1.6,
v) =
8k
B
T
m
Therefore, we can express in terms of
pressure and temperature.
1.5. ESTABLISHING EQUILIBRIUM 19
a collision with another particle. If we dene the
scattering cross-section, [m
2
], as
= d
2
where d is the eective particle diameter, then
the mean free path of an ideal gas with particle
number density n
p
, is given by Eq. 1.35 (Deriva-
tion 1.7).
Mean free path: (Derivation 1.7)
=
1
2 n
p
(1.35)
where the scattering cross-section is
= d
2
1.5 Establishing equilibrium
The ideal gas law holds for systems in equilibrium.
A system in equilibrium is one whose macroscopic
properties are not changing with timefor ex-
ample, constant pressure, constant temperature
and constant volume. Equilibrium is an impor-
tant concept in thermal physics, since the subject
of thermodynamics, which deals with the thermal
properties of matter exclusively on macroscopic
scales, only holds for systems in equilibrium.
However, since kinetic theory treats systems on
microscopic scales, we can use it to determine
the transport properties of systems that arent in
equilibrium.
Thermodynamic equilibrium has three compo-
nents:
1. Chemical equilibrium: constant particle
density. This is achieved by particle diusion
from regions of high particle concentration to
low concentration.
2. Mechanical equilibrium: constant pres-
sure. This is achieved by viscous forces,
which transport particle momentum from re-
gions of high momentum to low momentum.
3. Thermal equilibrium: constant tempera-
ture. This is achieved by thermal conduc-
tivity, which is the transport of heat from a
region of high temperature to a region of low
temperature.
The microscopic behaviour governing these pro-
cesses is very similar in each case. Below, we will
look in turn at how each of these equilibria are
established.
1.5.1 Chemical equilibrium: particle
transport
Chemical equilibrium is established when the par-
ticle concentration, n
p
, is uniform throughout
the system. Therefore, if there is a gradient in
the concentration, dn
p
/dz, particle transport, or
diusion, occurs down the concentration gradi-
ent. We will consider two situations: the steady-
state situation, where the concentration gradient
is constant in time, and the non-steady situation,
where the concentration gradient varies with re-
spect to time.
Steady-state diusion
Ficks First Law of diusion states that for
steady-state particle diusion, the particle ux,
6
J [m
2
s
1
], is proportional to the gradient of the
particle number density, n
p
. In one-dimension,
for example along the z-axis, Ficks law is there-
fore:
J
z
= D
n
p
z
where D [m
2
s
1
] is the diusion coecient. An
expression for the diusion coecient is found
by considering the number of particles crossing a
given surface per unit time, as shown in Deriva-
tion 1.8. We nd that D depends on the average
particle speed, v), and the mean free path,
(Eq. 1.37 and Derivation 1.8).
6
The particle ux is the number of particles crossing
unit area per unit time.
Table 1.7: Derivation of particle mean free path (Eq. 1.35)
v
R
t
Cylinder volume = v
R
t
= d
2
d
particle density = n
p
d
d
Effective
collision area
=
dx
dN
=
vdt
n
p
d
=
vdt
n
p
v
R
) dt
2n
p
Starting point: The mean free path, ,

is the mean distance travelled before a
collision, per particle: = dx/dN.
If the time between collisions is dt, then
the mean distance travelled by a particle
with velocity v is:
dx = v dt
In a time dt, the collision volume is
d = v
R
) dt, where v
R
is the relative
velocity between colliding particles (refer
to diagram). If the particles have veloci-
ties v
1
and v
2
, where v
R
= v
1
v
2
, then
the magnitude of the relative velocity is
[v
R
[ =

v
R
v
R
=
_
(v
1
v
2
) (v
1
v
2
)
=
_
v
2
1
+v
2
2
2v
1
v
2
Therefore, the average relative velocity,
for particles where [v
1
[ [v
2
[ v, and
travelling in opposite directions, is
v
R
) =
v
2
1
_
+
v
2
2
_
2 v
1
v
2
)
. .
0
=
_
v
2
1
_
+
v
2
2
_
_
2 v
2
)
2v
vdt
dS
z
0
z
0
-
z
0
+
Particle flux, J
z
Shear stress, P
zx
or Heat flux, j
z
z
Figure 1.8: (Refer to Derivations 1.81.10.) Assume that particles striking an area dS come from
a sphere of radius , centred at dS. The number of particles striking dS in a time dt is dN =
N
C
N
(F
v
(v)dv), where N
C
= n
p
v dt dS cos is number of particles within the cylinder; N
= d/4 =
sin d d is the fraction of particles within the correct solid angle; and F
v
(v)dv is the fraction of
particles with the correct velocity (between v and v +dv). If we consider a reference plane at z
0
, then,
in a time dt, the particles crossing the plane from below come from the plane at z
= z
0
, whereas
those crossing the plane from above come from z
+
= z
0
+ . Is this right?? and are derivations
1.81.10 correct? I should probably draw a hemisphere/semi-circle instead of straight planes...
Particle transport: (Derivation 1.8)
Ficks First Law:
J
z
= D
n
p
z
(1.36)
Diusion coecient:
D =
1
3
v) (1.37)
Non-steady diusion
The rate of change of particle density between z
and z + dz is given by the dierence in particle
ux entering at z and leaving at z +dz:
n
p
t
dz = J(z) J(z +dz)
Therefore, using a Taylor expansion on J(z +dz)
together with Eq. 1.36, we recover Ficks Second
Law, which gives the rate of change of particle
density (Eq. 1.38):
n
p
t
dz = J(z) J(z +dz)
= J(z) J(z)
J
z
dz
= D
2
n
p
z
2
dz
The solution to Eq. 1.38 is found have the form:
n
p
(z, t) =
n
0
4Dt
e
z
2
/4Dt
which is a Gaussian in z (this can be shown by
substitution). Therefore, since the variance of
a Gaussian of the form e
x
2
is 1/2, n
p
has a
variance that grows linearly with time (Eq. 1.39).
The standard deviation, , which is the square
root of the variance, is called the diusion length,
L
D
[m], and it provides a measure of the parti-
cle propagation distance in a time t (Eq. 1.40 and
Fig. 1.9).
Particle transport:
Ficks Second Law:
n
p
t
= D
2
n
p
z
2
(1.38)
Variance and diusion length:
_
z
2
_
=
2
= 2Dt (1.39)
L
D
= =
2Dt (1.40)
Table 1.8: Derivation of diusion coecient (Eq. 1.37, refer to Fig. 1.8)
dJ
z
=
dN(z)
dSdt
cos
=
1
4
n
p
(z)vF
v
(v)dv cos
2
sindd
J
z
=
1
4
n
p
(z
0
)
vF
v
(v)dv
. .
=v
/2
_
0
cos
2
sind
. .
=1/3
2
_
0
d
. .
=2
=
1
6
n
p
(z
0
) v)
J
z
= J
+
z
J
z
=
1
6
n
p
(z
0
) v)
1
6
n
p
(z
0
+) v)
1
6
v)
_
n(z
0
)
n
p
z

_
n(z
0
) +
n
p
z

__
=
1
3
v)
. .
=D
n
p
z
D =
1
3
v)
Starting point: The z-component
of particle ux, for particles inci-
dent at an angle , is the number
of particles, dN, per unit area dS,
per unit time dt. Since the particle
concentration is not constant, N is
a function of z. (Refer to Fig. 1.8
for expression for dN)
The net upward ux, J
+
, and
downward ux, J
, is found by in-
tegrating over all angles in a hemi-
sphere, and over all velocities. The
upward ux is a result of particles
coming from the plane z
= z
0
,
whereas the downward ux is from
particles in the plane z
+
= z
0
+
(Refer to Fig. 1.8).
The total ux, J
z
, is therefore
the dierence between upward and
downward ux. We can approxi-
mate using a Binomial expansion
about z
0
.
By comparing the expression for
J
z
to Ficks First Law in Eq. 1.36,
we nd an expression for the dif-
fusion coecient, D.
t
1
t
2
> t
1
z
n(z)
1
=2Dt
1
2
=2Dt
2
Figure 1.9: Particle diusion in 1D (Ficks Second
Law, Eq. 1.38). The standard deviation increases
with time. The diusion length equals the stan-
dard deviation. Draw in mathematica.
1.5.2 Mechanical equilibrium: mo-
mentum transport
We can use a similar steady-state approach to
Ficks First Law to determine the how pressure
reaches equilibrium. Pressure gradients are due
to gradients in viscous forces between particles.
These are shearing forces per unit area, and are
referred to as shear stress, P
x
[Pa]. Shear stress
acts parallel to the surface on which it acts, simi-
lar to friction between solids. The perpendicular
shear stress gradient, P
zx
, is found to be propor-
tional to the z-gradient of particle velocity in the
x-direction, u
x
:
P
zx
=
u
x
(z)
z
(Refer to the gure in Derivation 1.9). The co-
ecient of proportionality, [Pa s], is called the
viscosity. Viscosity is often dened in terms of
this relation, as the shear stress per unit velocity
gradient. The value of gives a measure of the
resistance of a substance to uid ow. Highly vis-
cous substances, such as honey, have a slow ow
rate and therefore have a large resistance to shear
stress, whereas less viscous uids, such as water,
have lower resistance to shear stress. Using a
microscopic approach, an expression for viscosity
can be calculated (Eq. 1.42 and Derivation 1.9).
The quantity has a similar form to the diusion
coecient (compare Eq. 1.42 with Eq. 1.37).
Shear stress: (Derivation 1.9)
P
zx
=
u
x
z
(1.41)
Viscosity:
=
1
3
n
p
mv) (1.42)
1.5.3 Thermal equilibrium: heat
transport
The nal equilibrium well consider is thermal
equilibrium, which acts to equalise temperature
dierences. This is achieved by energy transfer
heat owdown temperature gradients. There
are three types of heat transfer: conduction, con-
vection and radiation. (See Section 4.6.4 for ra-
diation.)
Conduction
Conduction is the transfer of heat through mat-
ter, from particle to particle. Higher energy par-
ticles transfer some of their energy to lower ener-
getic particles via collisions, which results in a net
heat ow from hotter to cooler regions. Equilib-
rium is achieved when the temperature is uniform
throughout the substance. Conduction occurs in
solids, liquids and, to a lesser extent, in gases.
The one-dimensional heat ux, j
z
[Jm
2
s
1
],
is dened as the heat ow per unit area per unit
time. This obeys a similar dierential equation
to those for particle ux and shear stress, where
the heat ux is proportional to the temperature
gradient:
j
z
=
T
z
The coecient of proportionality, [Wm
1
K
1
],
is called the thermal conductivity, and it has
a similar form to the diusion coecient from
Eq. 1.37 and viscosity from Eq. 1.42 (Eq. 1.45 and
Derivation 1.10).
Multiplying j
z
by the cross-sectional area, A,
we obtain the heat conduction formula, which
gives the rate of heat ow (Eq. 1.44). Draw a
diagram to show what j
z
and A are? Can maybe
include this in the derivation box?
Table 1.9: Derivation of viscosity coecient (Eq. 1.42, refer to Fig. 1.8)
x-direction drift velocity, u
x
(z)
A molecule carries momentum p
x
(z)=mu
x
(z)
z-direction shear stress, P
zx
dP
zx
=
dp
x
(z)
dSdt
cos
=
dN
dSdt
mu
x
(z) cos
=
1
4
n
p
mu
x
(z)vF
v
(v)dv cos
2
sindd
P
zx
=
1
4
n
p
mu
x
(z
0
)
vF
v
(v)dv
. .
=v
/2
_
0
cos
2
sind
. .
=1/3
2
_
0
d
. .
=2
=
1
6
n
p
mu
x
(z
0
) v)
P
zx
= P
+
zx
P
zx
=
1
6
n
p
mu
x
(z
0
) v)
1
6
n
p
mu
x
(z
0
+) v)
1
6
n
p
mv)
_
u
x
(z
0
)
u
x
z

_
u
x
(z
0
) +
u
x
z

__
=
1
3
n
p
mv)
. .
=
u
x
z
=
1
3
n
p
mv)
Starting point: The z-
component of shear stress,
dP
zx
, for particles incident
at an angle , is given by
the particle impulse in the x
direction, F
x
= dp
x
/dt, per
unit area dS.
The particle momentum in the
x-direction is proportional to
the drift velocity per particle,
u
x
(z):
dp
x
(z) = dNmu
x
(z)
(Refer to Fig. 1.8 for expression
for dN)
The net upward shear stress,
P
+
zx
, and downward shear
stress, P
zx
, is found by in-
tegrating over all angles in
a hemisphere, and over all
velocities.
The total shear stress, P
zx
,
is therefore the dierence be-
tween upward and downward
components. We can approxi-
mate using a Binomial expan-
sion about z
0
.
By comparing the expression
for P
zx
to Eq. 1.41, we nd an
expression for the viscosity co-
ecient, .
Thermal conduction: (Derivation 1.10)
j
z
=
T
z
(1.43)
dQ
dt
= A
dT
dz
(1.44)
Thermal conductivity:
=
1
3
n
p
C
N
V
v) (1.45)
Thermal conductivity is a measure of a sub-
stances ability to conduct heat. Materials with
high thermal conductivity, for example silver and
other metals, are good conductors of heat, mean-
ing that they rapidly transport heat and warm
up throughout when in contact with a hot object,
such as a ame. Conversely, materials with low
thermal conductivity, such as wood or rubber, are
thermal insulators, which means that they are
poor conductors of heat.
Newtons law of cooling
For a constant temperature-gradient over a dis-
tance L, the heat conduction equation (Eq. 1.44)
is
dQ
dt
= A
(T
L
T
0
)
L
where T
0
and T
L
are the temperatures at z =
0 and z = L respectively. From the denition
of specic heat capacity (Eq. 1.22), the rate of
change of heat can be written as
dQ
dt
= mc
dT
dt
Therefore, equating these two expressions for
dQ/dt, we nd that the rate of change of temper-
ature of an object is proportional to the tempera-
ture dierence between it, T(t), and its surround-
ings, T
s
. This statement is known as Newtons
law of cooling (Eq. 1.46). The general solution to
Eq. 1.46 is an exponential decay, with time con-
stant = 1/K. The exact solution depends on
boundary conditions. Worked example?
Newtons law of cooling:
dT
dt
=
A
mcL
(T(t) T
s
)
K(T(t) T
s
) (1.46)
where K is a positive constant.
Convection
Convection occurs only in liquids and gases, and
it involves the bulk transport of parcels of uid
or gas. Natural convection occurs when hot,
less dense parcels rise, while cooler, more dense
parcels sink to take their place, which are then in
turn heated and rise. This repeated circulation
is called a convection current. Forced convection
occurs when a fan, pump or draft, for example,
assists convection.
The dierential equations governing convection
are similar to those for conduction (Eqs. 1.44 and
1.46). The dierence is that the constant of pro-
portionality is not the thermal conductivity, ,
but is called the heat transfer coecient, h. This
is not solely a material property, but depends also
on external properties of the ow, such as geom-
etry, temperature and ow velocity. For this rea-
son, h is usually an experimentally-determined
quantity.
Table 1.10: Derivation of thermal conductivity (Eq. 1.45, refer to Fig. 1.8)
dj
z
=
dQ(z)
dSdt
cos
=
dN
dSdt
q(z) cos
=
1
4
n
p
q(z)vF
v
(v)dv cos
2
sindd
j
z
=
1
4
n
p
q(z
0
)
vF
v
(v)dv
. .
=v
/2
_
0
cos
2
sind
. .
=1/3
2
_
0
d
. .
=2
=
1
6
n
p
q(z
0
) v)
j
z
= j
+
z
j
z
=
1
6
n
p
q(z
0
) v)
1
6
n
p
q(z
0
+) v)
1
6
n
p
v)
_
q(z
0
)
q
z

_
q(z
0
) +
q
z
__
=
1
3
n
p
v)
q
z
=
1
3
n
p
v)
q
T
T
z
=
1
3
n
p
C
N
V
v)
. .
=
T
z
=
1
3
n
p
C
N
V
v)
Starting point: The z-
component of heat ux,
dj
z
, for particles incident at an
angle , is given by the heat
gradient, dQ(z), per unit area
dS, per unit time dt.
Write the heat gradient in
terms of the heat gradient per
particle, q(z):
dQ(z) = dNq(z)
(Refer to Fig. 1.8 for expression
for dN)
The net upward heat ux, j
+
z
,
and downward heat ux, j
z
,
are found by integrating over
all angles in a hemisphere, and
over all particle velocities.
The total heat ux, j
z
, is there-
fore the dierence between up-
ward and downward compo-
nents. We can approximate us-
ing a Binomial expansion about
z
0
.
We can express the heat gradi-
ent as a temperature gradient:
q
z
=
q
T
T
z
= C
N
V
T
z
where C
N
V
is the constant vol-
ume heat capacity per particle.
By comparing the expression
for j
z
to Eq. 1.43, we nd an
expression for the thermal con-
ductivity, .
Chapter 2
Classical Thermodynamics
Classical thermodynamics is concerned with
the interchange of matter and energy between a
system and its surroundings, and how the sys-
tem responds to such interchanges. The birth of
the subject coincided with the start of the indus-
trial revolution around the start of the nineteenth
century, motivated by the interest to develop me-
chanical power and engines from heat, such as the
steam engine. At that time, Dalton was only be-
ginning to propose the atomic structure of mat-
ter. Therefore, whereas kinetic theory uses the
microscopic properties of individual particles to
explain the macroscopic properties of a system,
classical thermodynamics treats systems exclu-
sively on a macroscopic scale. Although thermo-
dynamics is more general, and handles systems
other than ideal gases, there is good agreement
between the two descriptions.
2.1 Thermodynamic systems
Classical thermodynamics is based on four laws,
known collectively as the Laws of Thermodynam-
ics. Before presenting these laws in Section 2.2,
we will rst provide an overview of the terminol-
ogy of thermodynamics.
A thermodynamic system is dened by three
components: the system, the surroundings, and
the boundary between them (Fig. 2.1). The sys-
tem is the part whose properties we are inter-
ested in studying, for example an ideal gas in
a cylinder with a piston; the surroundings in-
volve everything else. The system can interact
with its surroundings by exchanging matter or en-
ergy across the boundary. Thermodynamic sys-
tems fall into one of three categories depending
on what can be exchanged between the system
and its surroundings: (i) in an open system, both
System
Surroundings
Boundary
Interchange of energy and
matter across boundary
Figure 2.1: Schematic of a thermodynamic sys-
tem. In an open system, both matter and en-
ergy can be exchanged between the system and
the surroundings; in a closed system only energy
exchange is possible; and in an isolated system,
neither matter nor energy is exchanged.
matter and energy can be exchanged with the sur-
roundings, (ii) in a closed system, only energy can
be exchanged, therefore the number of particles
within the system remains constant, and (iii) in
an isolated system, the system is totally isolated
from its surroundings, meaning that neither mat-
ter nor energy can be exchanged. In this chapter,
well deal primarily with closed systems. In Sec-
tion 2.3.4, we will look briey at the thermody-
namics of open systems, since this is important
for understanding phase transitions.
Heat, work and internal energy
Recall from kinetic theory that the internal en-
ergy, U [J], of a system is the sum of kinetic
and potential energies of the particles that make
it up. For example, the kinetic energy includes
27
28 CHAPTER 2. CLASSICAL THERMODYNAMICS
translational, rotational and vibrational motion,
whereas the potential energy includes chemical
energy stored in atomic bonds, nuclear energy
stored in nuclear bonds, and latent energy associ-
ated with the state of matter (solid, liquid or gas).
Therefore, the internal energy is the sum of all
energies within the system, but does not include
the energy the system possesses as a function of
its overall motion or position. For example, the
kinetic energy of a ball rolling down a hill is not
considered as internal energy.
The combination of kinetic energy and latent
energy is often called thermal energy. The tem-
perature, T [K], of a system provides a measure
of the average kinetic energy of its particles; the
greater the average molecular kinetic energy, the
higher the systems temperature.
The internal energy of a system changes if there
is a net exchange of energy with its surround-
ings. The two ways in which energy is exchanged
across a boundary are via heat ow, Q [J], or
work done, W [J]. These are both processes of
energy transfer from one body to another. Heat
ow refers specically to the transfer of thermal
energy between two bodies at dierent tempera-
tures, whereas work is the general term to de-
scribe energy transfer of any nature, other than
heat. Since classical thermodynamics was devel-
oped before the microscopic behaviour of parti-
cles within a system was understood, the internal
energy of a system is usually understood in terms
of its change with heat and work inputs and out-
puts, as described by the First Law of Thermo-
dynamics (see Section 2.2.2).
Thermodynamic equilibrium
An important concept in thermodynamics is that
of equilibrium. A system is in thermodynamic
equilibrium with its surroundings when its macro-
scopic properties, such as temperature, pressure,
and volume, are not changing with time. In clas-
sical thermodynamics, we deal exclusively with
systems in equilibrium. To describe systems that
arent in equilibrium, we require a microscopic
approach, such as kinetic theory. If a system is
in thermodynamic equilibrium, then it is simul-
taneously in (i) mechanical equilibrium (constant
pressure) (ii) chemical equilibrium (constant par-
ticle concentration), and (iii) thermal equilibrium
(constant temperature). Thermal equilibrium is
achieved when a system and its surrounding are
in thermal contact, which means that heat can
ow between them. A boundary that allows
two systems to be in thermal contact is called a
diathermal boundary, whereas a boundary that
thermally isolates two systems, such that heat
does not ow between them, is called an adiather-
mal boundary.
Thermodynamic states
The thermodynamic state of a system is a broad
term that encompasses all of the macroscopic
properties that dene it. These properties are
called state variables. For example, volume, tem-
perature and internal energy are state variables,
whereas heat and work are not. State variables
are often related to one another with an equa-
tion of state. For example, the ideal gas law,
pV = nRT, is the equation of state that relates
pressure, volume and temperature. A state vari-
able that depends on the physical size of the sys-
tem, such as volume, is called an extensive vari-
able, whereas one that is independent of the sys-
tems size, such as temperature and pressure, is
an intensive variable. Said another way, an inten-
sive variable has the same value whether we are
looking at the whole system, or only at a part of
the system, but this is not true for an extensive
variable.
Thermodynamic processes
When a thermodynamic system changes state, we
say that it undergoes a thermodynamic process.
Processes are classied as either reversible or ir-
reversible. In irreversible processes, energy is per-
manently lost from the system, due to dissipative
forces such as friction. Irreversible systems ex-
hibit hysteresis, namely the change of state de-
pends on the path taken. Since the dissipated
energy cannot be recovered by changing the sys-
tem back to its original state, then the process
is irreversible. Conversely, reversible processes
do not exhibit hysteresis, and the original state
of the system can be recovered by reversing the
operations on the system. In practice, perfectly
reversible processes are idealisations. However,
they can be achieved by carrying out the process
in innitesimal steps, with each step performed
2.2. THE LAWS OF THERMODYNAMICS 29
over a time period much slower than the response
time between the system and surroundings. This
ensures that they are always in thermal equilib-
rium with each other, and therefore the interme-
diate states of the system correspond to denite
valuesthat is, ones that can be known through
measurementof its macroscopic properties. A
process carried out in this way, through a series
of equilibrium states, is called a quasi-static pro-
cess.
Since the intermediate states are, in theory,
known, reversible and quasi-static processes can
be plotted on p-V diagrams. From this, we can
integrate the curve to determine total changes.
In contrast, only the initial and nal states of ir-
reversible processes may be plotted since they do
not proceed via equilibrium states, and therefore
the integration path between them is unknown
(Fig. 2.2).
In this Chapter, we will deal primarily with re-
versible processes since they are easier to treat
mathematically, and also because they represent
the idealised experimental outcomes since energy
is not irrecoverably lost from the system. The
most common reversible gaseous processes are
summarised in Table 2.1. Should this be a new
subsubsection, with more description?
2.2 The Laws of Thermodynam-
ics
Classical thermodynamics is concerned with the
changes in a systems macroscopic properties, and
its energy inputs and outputs, when it undergoes
a thermodynamic process (a change of state). In
practical terms, it deals with quantifying the ex-
traction of useful work from a system during a
change in state. The fundamental results of clas-
sical thermodynamics are encapsulated in four
laws, known as the Laws of Thermodynamics.
These are summarised in Table 2.2. Below, we
will look at each in turn.
2.2.1 The Zeroth Law
When a hot and a cold system are put in thermal
contact with each other, heat ows from the hot
to the cold system. This decreases the tempera-
ture of the hot system, whilst increasing that of
the cold one. When the two systems are at the
same temperature, then no more heat ows and
the systems are said to be in thermal equilibrium.
This is the basis of the Zeroth Law of Thermo-
dynamics (Table 2.2). The Zeroth Law implies
that systems in thermal equilibrium are at the
same temperature as each other, and that their
temperature remains constant in time. Histori-
cally, the Zeroth Law was introduced after the
other three laws. However, it is considered more
fundamental than the others and was therefore
labelled the Zeroth Law rather than the Fourth
Law. It provides the basis for the denition of
temperature in terms of thermal equilibrium.
2.2.2 The First Law
The First Law of Thermodynamics (Eq. 2.5) is a
statement of the conservation of energy. The law
is often stated in terms of an incremental change
of internal energy, dU (Eq. 2.6), rather than the
net change, U (Eq. 2.5). The incremental form
is more useful because we can nd expressions for
the elements dU, dQ and dW in terms of changes
in the systems state variables, such as volume
and temperature changes, as shown below. How-
ever, the expressions derived below only hold for
reversible processes; they do not hold for irre-
versible processes because we do not know how
much energy is lost to the environment.
Work done, dW
Gases and liquids exert pressurea force per unit
areaon the walls of their container. From a
molecular viewpoint, this pressure results from
the change in momentum of molecules colliding
against the container walls. When the systems
volume decreases, work must therefore be done
against these molecular forces. (Recall that for
non-dissipative processes, the work done over a
distance, L, is related to force by W =
_
L
Fdl.)
By considering the work done, W, on a system by
compressing it with a piston, we nd an expres-
sion for W in terms of the pressure and change
in volume of the system (Eqs. 2.82.9 and Deriva-
tion 2.3). Work is done on the system if its vol-
ume decreases (dV is negative), and work is done
by the system if its volume increases (dV is pos-
itive). Equation 2.9 also shows that (i) no work
is done on or by the system if its volume remains
constant, and (ii) the area under a p-V curve of a
p
V
p
V
p
V
Reversible Quasistatic Irreversible
Figure 2.2: p-V diagrams of reversible, quasi-static and irreversible processes, occurring through two
equilibrium states, A and B. Show A and B on diagram.
Table 2.1: Common reversible processes for ideal gases (where pV = nRT):
Isothermal: constant temperature
T = 0
pV = constant (2.1)
Isochoric: constant volume
V = 0
p
T
= constant (2.2)
Isobaric: constant pressure
p = 0
V
T
= constant (2.3)
Adiabatic: no heat exchange
Q = 0
pV
= constant (2.4)
p
V
Isobaric
Isothermal
Adiabatic
Isochoric
REDO. Needs more explanation. eg adiabatic
stu
Table 2.2: The Laws of Thermodynamics:
Zeroth Law: Thermal equilibrium
If two thermodynamic systems are at the same time in thermal equilibrium with a third system,
then they are also in thermal equilibrium with each other.
First Law: Conservation of internal energy
The change in internal energy of a thermodynamic system is given by the sum of work and heat
exchanged with its surroundings:
U = Q+W (2.5)
dU = dQ+dW (2.6)
Conventions:
: net change; : incremental change
+Q: heat into the system; Q: heat out of the system
+W: work done on the system; W: work done by the system
Second Law: Law of increasing entropy
No process is possible where the total entropy of the universe decreases. The total change in
entropy always greater than or equal to zero:
Reversible process: S = 0
Irreversible process: S > 0
S 0 always (2.7)
Third Law: Absolute zero temperature
The entropy of a system approaches zero as its temperature approaches absolute zero.
reversible process represents the total work done
during the change.
Work done for a reversible process:
(Derivation 2.3)
dW = p dV (2.8)
W =
V
2
_
V
1
p dV (2.9)
Heat exchange, dQ
The heat capacity of a system is dened as
C
=
_
Q
T
_
where is the variable that is kept constant while

heat is added. For example, for an isobaric pro-
cess, p, whereas for an isochoric process,
V . Rearranging, we can nd an expression
for an element of heat ow into or out of a sys-
tem (Eq. 2.10). An increase in temperature corre-
sponds to heat owing into the system, whereas a
decrease in temperature corresponds to heat ow-
ing out of the system. For reversible processes,
the total heat input or output is equal to the inte-
gral of heat capacity over temperature, between
the initial and nal temperature limits (Eq. 2.11).
Heat exchange for a reversible process:
dQ = C
dT (2.10)
Q =
T
2
_
T
1
C
dT (2.11)
Internal energy change, dU
The internal energy change is found by combining
the expressions for dW and dQ using the First
Law of Thermodynamics, as shown in Deriva-
tion 2.4 (Eqs. 2.122.13). Even though these ex-
pressions are derived for an isochoric process,
Eq. 2.12 is true for all ideal gas processes, even
when the volume is not kept constant. It shows
that the internal energy of an ideal gas depends
only on its temperature. (For real gases, the in-
ternal energy depends on its volume as well, see
Section ??).
Internal energy for a reversible process:
(Derivation 2.4)
dU = C
V
dT (2.12)
U =
T
2
_
T
1
C
V
dT (2.13)
Using the First Law
Worked example 2.3 uses the above expressions to
calculate W, Q and U for reversible processes
with ideal gases. The First Law can also be used
to derive the adiabatic relationship, pV
= con-
stant, as shown in Derivation 2.5.
2.2.3 The Second Law
The Second Law of Thermodynamics introduces
another function of state, called the entropy, S
[JK
1
]. It states that, for any process within a
closed system, such as the universe, entropy can-
not decrease. Entropy is often interpreted as the
degree of disorder of a system. For example, it
can be thought of as a measure of the number
of dierent ways that quanta of energy in a sys-
tem can be arranged between the particles that
make it up, without changing the total energy of
the system. A systems entropy changes if heat
is transferred into or out from it. The change
in entropy when a reversible quantity of heat,
dQ
rev
, ows into a system at temperature T, is
dened by Eq. 2.14. This denition follows from
the Clausius inequality, which is described below.
Entropy change for a reversible process:
dS =
dQ
rev
T
(2.14)
Whereas the First Law puts restrictions on
what processes are energetically possible, the Sec-
ond Law tells us the direction in which they can
Table 2.3: Derivation of work done for a reversible process (Eqs. 2.82.9)
dx
F=pA
dW = F dx
= pAdx
= p dV
W =
V
2
_
V
1
p(V ) dV
Starting point: Consider the work done on
the gas in a cylinder by moving a piston
through a length dx.
The opposing force results from the sys-
tems pressure, where F = pA
The element of volume is dV = Adx
For reversible processes, the integration
path is dened, therefore we can determine
the total work done by integrating from V
1
to V
2
.
Table 2.4: Derivation of internal energy change for a reversible process (Eqs. 2.122.13)
dU = dQ+dW
= dQ
= C
V
dT
U =
T
2
_
T
1
C
V
dT
Starting point: From the First Law of Ther-
modynamics, the change in internal energy
is the sum of heat in and work done on a
system.
Consider an isochoric process, where dV =
0 (V = constant). From Eqs. 2.8 and 2.10,
dW = 0
and dQ = C
V
dT
For reversible processes, the integration
path is dened, therefore we can determine
the total internal energy change by integrat-
ing over temperature, between the initial
and nal values
Figure 2.3: Worked example: applying the First Law of Thermodynamics to various re-
versible processes Also include pV diagrams to show the processes, work done, and heat
in or out.
Isothermal (dT = 0):
U = C
V
T = 0
W =
V
2
_
V
1
nRT
1
V
dV
= nRT
1
ln
_
V
1
V
2
_
Q = W
= nRT
1
ln
_
V
1
V
2
_
Isochoric (dV = 0):
U = C
V
(T
2
T
1
)
W = pV = 0
Q = U
= C
V
(T
2
T
1
)
Isobaric (dp = 0):
U = C
V
(T
2
T
1
)
W = p
1
(V
2
V
1
)
Q = U W
= C
V
(T
2
T
1
)
+p
1
(V
2
V
1
)
C
p
(T
2
T
1
)
Adiabatic: (pV
= A)
U = C
V
(T
2
T
1
)
W = A
V
2
_
V
1
dV
V
=
A
( 1)
_
V
2
V
V
1
V
1
_
=
(p
2
V
2
p
1
V
1
)
( 1)
C
V
(T
2
T
1
)
Q = 0
Table 2.5: Derivation of the adiabatic p-V relationship (Eq. 2.4) Include more on the
adiabatic condition in earlier section, and reference equation
dU = dW
C
V
dT = pdV
=
nRT
V
dV
=
(C
p
C
V
)T
V
dV
C
V
_
T
2
T
1
dT
T
= C
V
( 1)
_
V
2
V
1
dV
V
ln(T
2
/T
1
) = ( 1) ln(V
2
/V
1
)
T
2
T
1
=
_
V
2
V
1
_
(1)
TV
(1)
= constant
pV
= constant
Starting point: For an adiabatic process,
the heat exchange is zero, dQ = 0, there-
fore, from the First Law, dU = dW.
Use Eqs. 2.8 and 2.12 for expressions for dW
and dU.
The dierence and ratio of heat capacities
are:
nR = C
p
C
V
=
C
p
C
V
(see Section 1.4.2)
Use the ideal gas law to eliminate T from
TV
(1)
.
occur. For example, we know from experience
that a quantity of heat, Q, can spontaneously
ow from a hot to a cold body. Although the
First Law allows the reverse process to happen,
namely heat Q to ow from a cold body to a
hot body, the Second Law states that it cannot
happen spontaneously since, from Eq. 2.14, the
decrease in entropy of the cool object would be
greater than the increase in entropy of the hot ob-
ject, since T
cold
< T
hot
. This would result in a net
decrease of entropy of the combined system (and
universe), which violates the Second Law. Said
another way, the natural direction for change is
towards disorder.
1
Heat engines and heat pumps
The Second Law of Thermodynamics describes
the operation and eciencies of heat engines and
heat pumps. Heat engines are machines that con-
tinuously consume heat to generate useful work.
For example, a steam engine that drives a train
is a heat engine. Heat pumps are machines that
continuously transfer heat from a cold object to
a hot object, requiring work to be done on the
pump since this is against the direction of spon-
taneous heat ow. Refrigerators and air condi-
tioners are examples of heat pumps. Both heat
engines and heat pumps involve cyclic processes,
and therefore energy conversion is continuous.
Kelvin and Clausius found that a perfect heat
engine, which converts all the consumed heat
into useful work, and a perfect heat pump, which
transfers heat from a cool to hot object without
using work, were impossible to construct. The
reason is that both devices violate the Second
Lawthe total entropy of the universe decreases
in both cases. The Second Law is therefore often
stated in terms of the operation of heat engines
and heat pumps, and makes no mention of en-
tropy. These are known as the Kelvin and Clau-
sius formulations of the Second Law of Thermo-
dynamics:
Kelvins formulation: The complete con-
version of heat into work is impossible; it is
impossible to construct a perfect heat engine.
1
This issue raises some interesting philosophical ques-
tions regarding the ultimate fate of the universe; for a good
discussion, refer to (refer to somewhere).
Hot reservoir
Cold reservoir
T
C
T
H
Q
H
Q
C
W
Heat
Engine
Hot reservoir
Cold reservoir
T
C
T
H
Q
H
Q
C
W
Heat
Pump
Figure 2.4: Schematic representations of heat en-
gines (left) and heat pumps (right) operating be-
tween two reservoirs at constant temperatures,
T
H
and T
C
.
Clausius formulation: Heat cannot spon-
taneously ow from a cold to a hot object;
it is impossible to construct a perfect heat
pump.
These statements are equivalent to one another,
and to the entropy formulation of the Second
Law.
The Second Law is satised for a heat engine if
some of the consumed heat is lost to the environ-
ment. For a heat pump, work must be provided to
transport the heat against its natural direction of
ow. These operations are shown schematically
in Fig. 2.4. We consider a heat engine or pump
operating between two large reservoirs, of con-
stant temperature T
H
and T
C
. The heat ow to
or from each reservoir, Q
H
and Q
C
, is small in
comparison with the size of the reservoirs, there-
fore their temperatures remain constant. From
the conservation of energy, we have
Q
H
= Q
C
+W
both for heat engines and heat pumps. Figure 2.4
illustrates that a heat pump can be thought of as
a heat engine operating in reverse.
Eciency and coecient of performance
The eciency of these machines is a measure of
the useful energy output to the energy input. For
a heat engine, the useful energy output is W for
an input energy, Q
H
. Therefore the eciency is
given by Eq. 2.15, where W = Q
H
Q
C
. For
heat pumps, the useful quantity is Q
H
for a heat-
ing device, and Q
C
for a cooling device, for an
input energy of W (Eqs. 2.16 and 2.17 respec-
tively). Since these ratios can be greater than
one, the term coecient of performance, or CoP,
is used instead of eciency.
Eciency and coecients of performance:
Heat engine eciency:
=
W
Q
H
= 1
Q
C
Q
H
(2.15)
Heat pump coecients of performance (CoP):
Heating: CoP
H
=
Q
H
W
(2.16)
Cooling: CoP
C
=
Q
C
W
(2.17)
Carnots theorem
Reversible engines do not dissipate heat energy
via friction. Carnots theorem states that re-
versible heat engines are the most ecient, as
demonstrated in Proof 2.6. This is intuitively ob-
vious, since if heat is lost via friction in a heat
engine, then less energy is available to do work.
A reversible heat engine follows what is called a
Carnot cycle. The details of its operation are
shown in worked example 2.5. A corollary of
Carnots theorem is that all reversible engines
operating between the same two temperatures
have the same eciency, which depends only on
the temperatures of the reservoirs (Eq. 2.18 and
worked example 2.5). Equation 2.18 shows that
the eciency of the engine is maximised by oper-
ating at the lowest and highest possible tempera-
tures of the cold and hot reservoirs, respectively.
Reversible heat engines (Carnot cycle):
(refer to worked example 2.5)
Q
H
T
H
Q
C
T
C

R
= 1
T
C
T
H
(2.18)
In practice, no engine is perfectly frictionless,
but it can be realised to a good approximation
using quasi-static processes. By considering the
operation of reversible engines, we can see where
the denition of entropy as dS = dQ/T comes
from, as well as the equivalence between the heat
engine and the entropy formulations of the Sec-
ond Law, as demonstrated below.
Clausius inequality and entropy
Another corollary of Carnots theorem is the
Clausius inequality, which shows that the ratio
of heat entering the system to the temperature
at the point of heat entry, when integrated over
a complete cycle, is less than or equal to zero
(Eq. 2.19 and Derivation 2.7). If we dene the
quantity dQ/T as an element of entropy change,
dS, then using this in the Clausius equality, we
see that entropy change can never be negative,
implying that the entropy of a closed system can-
not decrease (Eq. 2.20 and Derivation 2.8). This
illustrates that the heat engine formulations of
the Second Law are equivalent to the entropy for-
mulation.
Clausius inequality and entropy:
(Derivations 2.7 and 2.8)
_
dQ
T
0 (2.19)
S 0 (2.20)
2.2.4 The Third Law
The Third Law of Thermodynamics (Table 2.2)
provides a reference point for an absolute entropy
scale, rather than considering only changes in en-
tropy. The reference point corresponds to abso-
lute zero temperature, 0 K. An equivalent formu-
lation of the Third Law is that the specic heat
capacity of all materials equals zero at absolute
zero:
C
V
=
dQ
dT
(2.21)
= T
dS
dT
Table 2.6: Proof that reversible engines are the most ecient
Hot reservoir
Cold reservoir
T
C
T
H
Q`
H
Q`
C
W
Irreversible
engine
Q
H
Q
C
Reversible
engine
Composite
engine
Starting point: Consider the composite en-
gine in the adjacent diagram. The variables
W, Q
H,C
and Q
H,C
are positive in the di-
rections indicated. From Clausius formula-
tion of the Second Law, there is a net ow
of heat away from the hot reservoir:
Q
H
Q
H
0
Q
H
Q
H
W
Q
W
Q
H
Using the denition of eciency from
Eq. 2.15:

where
is the eciency of the irreversible

engine and is that of the reversible engine.
Therefore, reversible engines are the most
ecient:

R

I
Figure 2.5: Worked example: Reversible Carnot engine
p
V
1
2
3
4
Q
in
Q
out
Work
V
1
T
H
T
C
V
2
V
3
V
4
The Carnot engine is a cyclic, reversible en-
gine, which means that it returns to its orig-
inal state after one complete cycle. The de-
tails of each process are described in Steps 1
4 opposite, which correspond to the strokes
14 in the diagram above (refer to worked
example 2.3 for details). Other properties of
the cycle, such as the work done per cycle and
the eciency, are calculated below. The total
work done per cycle is given by the enclosed
area. Is this layout clear?
Equating the adiabatic condition in
Steps 2 and 4:
_
V
3
V
2
_
1
=
_
V
4
V
1
_
1
V
2
V
1
=
V
3
V
4
Total work done by the system (enclosed
area):
W = W
1
+W
2
+W
3
+W
4
= nRln
_
V
2
V
1
_
(T
H
T
C
)
Eciency: depends only on the temper-
atures of the reservoirs
= 1
Q
out
Q
in
= 1
T
C
T
H
Step 1. Isothermal expansion at T
H
: (heat
into the system and work done by
the system)
U = 0
Q
in
= W
1
=
V
2
_
V
1
pdV
= nRT
H
ln
_
V
2
V
1
_
Step 2. Adiabatic expansion from T
H
to
T
C
:
Q = 0
W
2
= U
= C
V
(T
H
T
C
)
and
T
H
T
C
=
_
V
3
V
2
_
1
Step 3. Isothermal compression at T
C
:
(heat out of the system and work
done on the system)
U = 0
Q
out
= W
3
=
V
4
_
V
3
pdV
= nRT
C
ln
_
V
4
V
3
_
= nRT
C
ln
_
V
3
V
4
_
Step 4. Adiabatic compression from T
C
to
T
H
:
Q = 0
W
4
= U
= C
V
(T
H
T
C
)
and
T
H
T
C
=
_
V
4
V
1
_
1
Table 2.7: Derivation of the Clausius inequality (Eq. 2.19)
R

I
1
Q
C
Q
H
R
1
Q
C
Q
H
Q
C
Q
H
Q
C
Q
H
T
C
T
H
Q
C
T
C
Q
H
T
H
For positive Q
H
and negative Q
C
:
Q
H
T
H
+
Q
C
T
C
0
cycle
Q
T
0
Therefore, the Clausius inequality is:
_
dQ
T
0
Starting point: From Carnots theorem, re-
versible engines are more ecient than ir-
reversible engines (Proof 2.6).
Use Eq. 2.15 for the eciency of heat en-
gines, and Eq. 2.18 for the eciency of a
reversible engine.
If we take the heat entering the system, Q
H
,
to be positive, then Q
C
, leaving the system,
is negative.
Generalising to a cycle with n processes
during which heat enters or leaves the sys-
tem, we can write the inequality as a sum
over the cycle.
In the limit where an element of heat dQ en-
ters the system at a temperature T, the sum
becomes an integral Clausius inequality.
Table 2.8: Derivation of the entropy formulation of the Second Law from the Clausius
inequality (Eq. 2.20)
A
B
Irreversible
Reversible
p
V
Clausius inequality:
_
dQ
T
0
B
_
A
dQ
T
+
A
_
B
dQ
R
T
0
B
_
A
dQ
T

B
_
A
dQ
R
T
B
_
A
dS
_
dQ
T

_
dS
S
S 0
Starting point: Let a cycle be composed of
one irreversible process between points A
and B, and a reversible process from B back
to A:
_
dQ
T
=
B
_
A
dQ
T
+
A
_
B
dQ
R
T
From the properties of denite integrals:
A
_
B
dQ
R
T
=
B
_
A
dQ
R
T
Dene an element of entropy as
dS
dQ
R
T
(Eq. 2.14)
Change the limits to integrate around a
complete cycle. The total change in entropy
for a complete cycle is S.
Equate the inequality for S to the Clau-
sius inequality to recover the entropy for-
mulation of Second Law of Thermodynam-
ics (Eq. 2.7).
2.3. THERMODYNAMIC POTENTIALS 41
Therefore, C
V
0 as T 0. Equation 2.21
in turn shows that the heat, dQ, that can be ex-
tracted from a cold body, to cool it further, ap-
proaches zero as T approaches zero. It is there-
fore impossible experimentally to reach absolute
zero temperature. Moreover, as temperatures
approach 0 K in a crystal lattice, atoms carry
out quantum-mechanical zero-point oscillations,
which means they possess a nite zero-point en-
ergy. The concept of absolute zero temperature
is therefore a theoretical reference only.
2.3 Thermodynamic potentials
The Laws of Thermodynamics tell us what pro-
cesses are energetically possible (First Law), and
in which direction they occur (Second Law). We
will now look at ways of describing and pre-
dicting thermodynamic states for non-cyclic pro-
cesses (i.e. those in which the nal state is dif-
ferent from the initial state), using equations of
state. Recall that an equation of state, such as
the ideal gas law pV = nRT, describes the state
of the system by relating its state variables to-
gether, in this case p, V and T. By considering
small changes in one parameter when another is
kept constant, we can use the equation of state
to predict how a system changes during a ther-
modynamic processes.
We can derive other equations of state by con-
sidering changes in the thermodynamic potentials
of a system during a given process. These are
scalar potential functions that depend on state
variables p, V , T, and S, and hence represent the
state of a system. We have already come across
a thermodynamic potential: the internal energy,
U. Along with U, three other common thermo-
dynamic potentials are the Helmholtz free energy,
F, the enthalpy, H, and the Gibbs free energy, G.
Neither F, H nor G have an insightful physical
interpretation, as U does, but they are useful in
determining how properties of a system change
when it changes state, as we shall see below. In
particular, thermodynamic potentials are useful
in determining the parameters that describe equi-
librium states, which correspond to energy min-
ima (stable states).
Analogy with potential elds
We are familiar with the gravitational potential
eld, g(r), in classical mechanics, and the elec-
tric potential eld, V (r), in electromagnetism.
Whereas g(r) and V (r) are functions that vary
with position, r, thermodynamic potentials are
functions that vary with state variables, p, V ,
T or S. A mass in a gravitational eld, or a
charge in an electric potential eld store poten-
tial energy as a result of their position, r. If the
mass or charge changes position, then its poten-
tial energy changes. Analogously, a thermody-
namic system stores potential energy as a result
of its thermodynamic state, which is dened by
a thermodynamic potential, for example U or F.
When the system changes state during a thermo-
dynamic process, then its state variables change
and hence its stored energy changes. This energy
might, for example, be released as heat, or used
to do work.
2.3.1 Internal energy
Lets start with the internal energy. For reversible
processes, we have:
dW = pdV and dQ = TdS
Using these in the First Law, we nd:
dU = dQ+dW
= TdS pdV (2.22)
If we consider the total derivative of a function
U = U(S, V ), as shown in Derivation 2.9, then
we recover Eq. 2.22. This means that Eq. 2.22 is
the exact dierential of dU, and that
T =
_
U
S
_
V
and p =
_
U
V
_
S
These constitute two equations of state that allow
us to determine T and p from the function U and
the state variables V and S.
Internal energy, U(S, V ): (Derivation 2.9)
dU = T dS p dV (2.23)
Equations of state from the total derivative:
T =
_
U
S
_
V
(2.24)
p =
_
U
V
_
S
(2.25)
Irreversible processes
Equations 2.232.25 involve state variables,
which are variables whose values are independent
of the path taken to reach that state. Therefore,
the relations in Eqs. 2.232.25 are true for all pro-
cesses, both reversible and irreversible. However,
for irreversible processes, a larger fraction of en-
ergy is lost as heat and a smaller fraction is avail-
able to do work, compared to reversible processes.
This means that, for irreversible processes:
dW
I
< [p dV [
and dQ
I
> T dS
Nonetheless, Eq. 2.23 and the First Law still hold
true for irreversible processes.
2.3.2 F, H and G
The thermodynamic potentials, F [J], H [J] and
G [J], for a system with state variables p, V , T
and S are dened as:
Helmholtz free energy: F = U TS (2.26)
Enthalpy: H = U +pV (2.27)
Gibbs free energy: G = U TS +pV
(2.28)
The energy term TS can be considered as the
fraction of heat provided to the system from the
environment, in bringing it to a temperature T.
The energy term pV can be considered as the
fraction of work done by the system, against the
environment, in taking it to a volume V .
By dierentiating Eqs. 2.262.28 (and using
dU = TdS pdV ), we can determine the exact
dierentials of F, H and G:
dF = dU TdS SdT
= SdT pdV (2.29)
dH = dU +pdV +V dp
= TdS +V dp (2.30)
dG = dU TdS SdT +pdV +V dp
= SdT +V dp (2.31)
If we equate these expressions to their total
derivative, as we did for U, we determine two
equations of state for each potential (column 2 of
Table 2.10). Need to improve format here. These
relate a state variable to the derivative of a ther-
modynamic potential.
2.3.3 The Maxwell relations
The Maxwell relations are a set of dierential
equations that relate the derivatives of the state
variables p, V , T and S to one another, without
making reference to the thermodynamic poten-
tials (column 3 of Table 2.10). Need to improve
format here. They are derived by dierentiating
the equations of state found above. For example,
dierentiating Eqs. 2.24 and 2.25 with respect to
either V or S, we obtain two expressions for the
second derivative of U:
_
T
V
_
S
=
_

2
U
V S
_
S,V
and
_
p
S
_
V
=
_

2
U
SV
_
V,S
Since the order of dierentiation for second
derivatives does not matter, we can equate the
above expressions:
_
T
V
_
S
=
_
p
S
_
V
We can use the same method to determine three
other similar relations from the equations of state
for F, H and G. Together, the set of four
dierential equations are the Maxwell relations
(Eqs. 2.322.35). A mnemonic for remembering
Maxwells relations is shown in Fig. 2.6. Worked
example 2.7 shows how to use these relations to
Table 2.9: Derivation of the total derivative of U(S, V ) (Eqs. 2.232.25)
For an isochoric process:
dU = TdS
T =
_
U
S
_
V
For an isentropic process:
dU = pdV
p =
_
U
V
_
S
Therefore, the total derivative of U is
dU =
_
U
S
_
V
. .
=T
dS +
_
U
V
_
S
. .
=p
dV
= T dS p dV
Starting point: The total derivative of a
function U = U(S, V ) is
dU =
_
U
S
_
V
dS +
_
U
V
_
S
dV
For an isochoric process, dV = 0:
dW = pdV = 0
and dU = dQ = T dS
For an isentropic process, dS = 0
dQ = TdS = 0
and dU = dW = p dV
prove that the internal energy of an ideal gas de-
pends only on its temperature (this statement is
known as Joules Law).
2.3.4 Chemical potential
So far, we have only considered closed systems,
in which energy but not matter is exchanged be-
tween the system and its surroundings. In this
section, we will look briey at the thermodynam-
ics of open systems, where matter (particles) can
be exchanged. The internal energy of the system
changes since each particle carries energy, and the
volume and entropy of the system will generally
change. Therefore, U is a function of the number
of particles, N, as well as S and V :
U = U(S, V, N)
The exact dierential of U for an open system is
dU =
_
U
S
_
V,N
. .
=T
dS +
_
U
V
_
S,N
. .
=p
dV
+
_
U
N
_
S,V
. .
=
dN
TdS pdV +dN (2.36)
-
S
p
T
V
1st derivative
2nd derivative
Figure 2.6: Mnemonic for Maxwells equations:
going anticlockwise the mnemonic is Society for
the Prevention of Teaching Vectors. The rst
derivative is found by going clockwise, starting
with the letter corresponding to its numerator,
and the second by going anticlockwise, starting
with the adjacent letter, as shown. The diagram
shows
_
S
V
_
T
=
_
p
T
_
V
. There is always a neg-
ative sign when S and p appear together in the
partial derivative.
Table 2.10: Thermodynamic potentials and Maxwells relations:
Need a better format: Use three columns. 1st column: exact dierential and total derivative; 2nd
column: equate these to get equations of state; 3rd column: 2nd derivative to get Maxwells relations.
1. Internal energy: U(S, V )
dU = TdS pdV
dU =
_
U
S
_
V
dS +
_
U
V
_
S
dV
_
_
_
T =
_
U
S
_
V
p =
_
U
V
_
S
_
_
_
_
T
V
_
S
=
_
p
S
_
V
(2.32)
2. Helmholtz free energy: F(T, V ) = U TS
dF = SdT pdV
dF =
_
F
T
_
V
dT +
_
F
V
_
T
dV
_
_
_
S =
_
F
T
_
V
p =
_
F
V
_
T
_
_
_
_
S
V
_
T
=
_
p
T
_
V
(2.33)
3. Enthalpy: H(S, p) = U +pV
dH = TdS +V dp
dH =
_
H
S
_
p
dS +
_
H
p
_
S
dp
_
_
_
T =
_
H
S
_
p
V =
_
H
p
_
S
_
_
_
_
T
p
_
S
=
_
V
S
_
p
(2.34)
4. Gibbs free energy: G(T, p) = U TS +pV
dG = SdT +V dp
dG =
_
G
T
_
p
dT +
_
G
p
_
T
dp
_
_
_
S =
_
G
T
_
p
V =
_
G
p
_
T
_
_
_
_
S
p
_
T
=
_
V
T
_
p
(2.35)
Figure 2.7: Worked example: Using Maxwells relations to show that U is a function of T
only, for an ideal gas
1. By dierentiating the exact dierential of U
with respect to volume, at constant tempera-
ture, and using the Maxwell relation in Eq. 2.33
to eliminate the entropy derivative, we see that
U does not depend on V for an ideal gas (use
pV = nRT):
dU = TdS pdV
_
U
V
_
T
= T
_
S
V
_
T
p
= T
_
p
T
_
V
p
=
nRT
V

nRT
V
= 0
2. By dierentiating the exact dierential of U
with respect to pressure, at constant tempera-
ture, and using the Maxwell relation in Eq. 2.35
to eliminate the entropy derivative, we see that
U does not depend on p for an ideal gas (use
pV = nRT):
dU = TdS pdV
_
U
p
_
T
= T
_
S
p
_
T
p
_
V
p
_
T
= T
_
V
T
_
p
p
_
V
p
_
T
=
nRT
p
+
nRT
p
= 0
Therefore, for an ideal gas, U is not a function of V or p, hence it is a function of T only: U U(T).
Here we introduce a variable known as the chem-
ical potential, , which represents the amount of
internal energy added to a system if a single par-
ticle is added at constant entropy and volume
(Eq. 2.37). The chemical potential is therefore
often thought of as the energy per particle. Dif-
ferent types of particles have dierent chemical
potentials. If the system involves k dierent par-
ticle types, then the chemical potential of the i
th
type is given by Eq. 2.38, and the exact dieren-
tial of U involves the sum over particle types:
dU = TdS pdV +
k
i=1
i
dN
i
We can use a similar argument to write the
chemical potential in terms of the thermodynamic
potentials, F and G:
=
_
F
N
_
V,T
or =
_
G
N
_
p,T
The latter allows us to dene the chemical poten-
tial as the Gibbs free energy per particle, g(p, T)
(Eq. 2.39), which is useful for determining the
equilibrium conditions during phase transitions
(see Section 3.2).
Chemical potential:
In terms of U:
Single species: =
_
U
N
_
S,V
(2.37)
Many species:
i
=
_
U
N
i
_
S,V,N
j=i
(2.38)
In terms of G:
=
_
G
N
_
p,T
g(p, T) (2.39)
Chapter 3
Real gases and phase transitions
So far, weve only been dealing with ideal gases,
which obey the ideal gas law, pV = nRT. Most
gases, however, are real gases, and do not obey
the ideal gas law. Real gases (i) have a nite
molecular size relative to the distance between
them, and (ii) have non-negligible intermolecu-
lar forces. All real gases behave like ideal gases
at very low pressures and high temperatures,
since under these conditions the particles rela-
tive volumes and intermolecular forces are negli-
gible. At high pressures and low temperatures, a
real gas eventually condenses into a liquid, which
has much stronger intermolecular forces. In this
chapter, well look at the thermodynamics of real
gases, as well as the properties governing changes
of state from gas to liquid.
3.1 Real gas equation of state
3.1.1 The van der Waals equation
There are several equations of state for real gases,
which relate the parameters p, V and T in such
a way to account for intermolecular forces and -
nite molecular volumes. A common equation of
state is the van der Waals equation,
1
which is the
ideal gas law modied by two empirical parame-
ters, labelled a [Pa m
6
mol
2
] and b [m
3
mol
1
], as
given in Eq. 3.1. These parameters are constants
for a particular gas; the larger their values, the
more the gas deviates from ideal gas behaviour.
The parameter a is the constant of internal pres-
sure, which accounts for attractive intermolecular
forces that act to reduce the pressure for a given
volume and temperature. The pressure correction
1
The van der Waals force is a weak, intermolecular
force that results from dipole-dipole interactions, involving
either induced or permanent dipoles. It is responsible for
the weak bonds between gas particles.
is found to be inversely proportional to the square
of the volume of the container. For n moles of gas,
the internal pressure is given by p
int
= a(n/V )
2
(derivation not shown). Therefore, the measured
pressure of a real gas, p, is corrected to that of
an ideal gas by adding back the internal pressure
correction for intermolecular forces, p
int
:
p p +a
_
n
V
_
2
The parameter b is the constant of internal vol-
ume, which accounts for the nite molecular vol-
ume of a real gas. This reduces the amount of
space within the container in which particles are
free to move. Letting b equal the eective molecu-
lar volume per mole, the internal volume occupied
by n moles of gas is V
int
= nb. Therefore, the ef-
fective volume is the volume of the container, V ,
reduced by the excluded volume of the particles
themselves, V
int
:
V V nb
The excluded volume is eight times the actual vol-
ume of the particles themselves, since two parti-
cles cannot approach within a distance less than
twice the particle radius. Using the corrected
pressure and volume in the ideal gas law, we re-
trieve the van der Waals equation given in Eq. 3.1.
The van der Waals equation is often written in
terms of the volume per mole, or the molar vol-
ume, v [m
3
mol
1
], where v = V/n (Eq. 3.2).
When the density of the gas is low, n/V 0,
and the van der Waals equation approaches the
ideal gas law.
Other common equations of state are the
Redlich-Kwong equation, the Dieterici equation
and the Berthelot equation. These are all two-
parameter equations, where one parameter ac-
counts for the gass internal pressure and the
47
48 CHAPTER 3. REAL GASES AND PHASE TRANSITIONS
other the internal volume. In the following dis-
cussion, I will only refer to the van der Waals
equation, but a similar treatment can be used for
the other equations of state as well.
Van der Waals equation of state for a real
gas:
_
p +a
_
n
V
_
2
_
(V nb) = nRT (3.1)
Or, in terms of the molar volume, v:
_
p +
a
v
2
_
(v b) = RT (3.2)
where v =
V
n
Van der Waals isotherms
Figure 3.1 shows the isotherms of Eq. 3.1 on a p-
V diagram. The behaviour of the gas depends
on the strength of the intermolecular forces, at
dierent combinations of p, V and T. We will
now look at the dierent parts of the diagram in
turn.
3.1.2 The Boyle temperature
The gaseous region indicated in Fig. 3.1, corre-
sponds to large volumes and, for low tempera-
ture isotherms, low pressures. This is because
intermolecular forces are relatively weak under
these conditions. For increasing temperatures,
intermolecular forces become less signicant, and
therefore real gases approach ideal gas behaviour.
In consequence, high temperature isotherms in
the gaseous region resemble rectangular hyper-
bolae characteristic of Boyles law for ideal gases,
p 1/V .
We can express the equation of state of a real
gas as a virial expansion in v:
pv = RT(1 +
A(T)
v
+
B(T)
v
2
+...)
where A(T), B(T) are temperature dependent
virial coecients. From this, we see that the real
gas equation of state approaches the ideal gas law
when the second virial coecient, A(T), equals
T
B
C
Ideal gas
Liquid
Coexistence region
Gas
p
V
Liquid+vapour
T
C
Figure 3.1: p-V diagram showing van der Waals
isotherms (Eq. 3.1). The critical point is marked
C. T
C
corresponds to the critical isotherm, and
T
B
to the Boyle temperature. Show increasing
temperature.
zero. The temperature at which this occurs is
called the Boyle temperature, T
B
[K]:
A(T
B
) = 0
This temperature therefore marks the transition
from real to ideal gas behaviour. It is a constant
for any gas, and depends on the gas parameters
a and b (Eq. 3.3 and Derivation 3.1).
The Boyle temperature: (Derivation 3.1)
T
B
=
a
Rb
(3.3)
3.1.3 The coexistence region
The small volume region in Fig. 3.1 corresponds
to the liquid region, where intermolecular forces
are signicantly stronger than for gases. Liquids
are highly incompressible, therefore a small re-
duction in volume leads to a large increase in
pressure, as shown by the steep isotherms.
In between the liquid and gaseous regions is
the coexistence region, where liquids and their
3.1. REAL GAS EQUATION OF STATE 49
Table 3.1: Derivation of the Boyle temperature, T
B
(Eq. 3.3)
pv =
RT
1 (b/v)

a
v
= RT
_
1
b
v
_
1
a
v
RT
_
1 +
b
v
+O
_
1
v
2
__
a
v
RT +
bRT a
v
+O
_
1
v
2
_
0 = bRT
B
a
when T
B
=
a
Rb
Starting point: Rearrange the van der
Waals equation of state (Eq. 3.2)
Since v b, we can do a binomial ex-
pansion of the term in brackets, and ignore
higher order terms
The equation of state approaches the ideal
gas law when
bRT a
v
0
gaseous vapour exist simultaneously in equilib-
rium. This region therefore describes the condi-
tions under which condensation and vaporisation
take place. We will look further at the thermo-
dynamic properties of these phase transitions in
Section 3.2.
The critical point
The coexistence region shrinks with increasing
temperature and pressure. The point at which
it vanishes is called the critical point, and is la-
belled as C in Fig. 3.1. This point corresponds
to the saddle point on the critical isotherm, at
T
C
, which means that the rst and second deriva-
tives of Eq. 3.2 are zero at this point. Therefore,
the thermodynamic state at the critical point, de-
ned by the parameters p
C
, V
C
and T
C
, can be de-
duced by determining the conditions under which
the rst and second derivatives of Eq. 3.2 are zero
(Eqs. 3.43.6 and Derivation 3.2).
Critical point parameters: (Derivation 3.2)
T
c
=
8a
27Rb
(3.4)
v
c
= 3b (3.5)
p
c
=
a
27b
2
(3.6)
Ratio of critical parameters:
p
c
v
c
RT
c
=
3
8

1
3
(3.7)
Comparing Eqs. 3.3 and 3.4, we see that the
critical temperature and the Boyle temperature
are related by:
T
c
=
8
27
T
B

1
3
T
B
Eliminating the parameters a and b from
Eqs. 3.43.6, we nd that the ratio of critical
point parameters is approximately 1/3 (Eq. 3.7).
This ratio is approximately the same for all sim-
ple gases, which are dened as gases that have
small dipole moments and weak intermolecular
forces even in the liquid phase, such as CO or
Cl
2
. For an ideal gas, the ratio pv/RT equals 1.
Reduced variables
The reduced variables, p, v and

T, are dimension-
less variables that equal the ratio of the respective
Table 3.2: Derivation of the critical point parameters (Eqs. 3.43.6)
At the saddle point:
RT
c
(v
c
b)
2
+
2a
v
3
c
= 0
and
2RT
c
(v
c
b)
3

6a
v
4
c
= 0
v
c
= 3b
T
c
=
8a
27Rb
and p
c
=
a
27b
2
Starting point: The rst and second derivatives of the
van der Waals equation of state in Eq. 3.2 are
_
p
v
_
T
=
RT
(v b)
2
+
2a
v
3
and
_
2
p
v
2
_
T
=
2RT
c
(v b)
3

6a
v
4
These equal zero at the saddle point, where v = v
c
,
T = T
c
and p = p
c
.
To nd v
c
: eliminate a.
To nd T
c
: substitute v
c
into (p/v)
T
.
To nd p
c
: substitute v
c
and T
c
into Eq. 3.2.
variable to their critical value:
p =
p
p
c
, v =
v
v
c
,

T =
T
T
c
(3.8)
By writing the van der Waals equation in reduced
variables, the parameters a and b cancel, as given
by Eq. 3.9. (To show this, eliminate p, v and T
in Eq. 3.2 using Eqs. 3.8, and eliminate critical
values using Eqs. 3.43.6.) This means that all
simple gasesthose with small dipole moments
and weak intermolecular forcesobey the same
p-V diagram when measured in reduced variables.
This statement is called the the law of correspond-
ing states.
Reduced variable equation of state:
_
p +
3
v
2
_
(3 v 1) = 8
T (3.9)
3.1.4 Isothermal compressibility
The gradient of the the isotherms in Fig. 3.1 de-
scribes how easily the substance can be com-
pressed. If it is highly compressible like a gas,
then for a large change in volume, the correspond-
ing change in pressure is small, and the gradient
is shallow. Conversely, if the material is incom-
pressible like a liquid, then a small change in vol-
ume leads to a large increase in pressure, and the
gradient is steep. The isothermal compressibility
of a substance,
T
[m
2
N
1
], is the relative change
in volume in response to a pressure change, at a
constant temperature (Eq. 3.10). This is related
to the gradient of the isotherms in Fig. 3.1 by
Eq. 3.11.
Isothermal compressibility:
T
=
1
V
_
V
p
_
T
(3.10)
gradient =
1
T
V
(3.11)
3.2 Phase transitions
So far, weve only been looking at homogeneous
systems, which are systems with uniform proper-
ties throughout them. A heterogeneous system
involves dierent components, and the proper-
ties of the components change discontinuously at
their boundaries. An example is a mixture of oil
and water. A heterogeneous system can also in-
volve one substance, but in two dierent states of
3.2. PHASE TRANSITIONS 51
matter, such as water and steam. In this case, the
components are called phases, and the boundary
between them is the phase boundary. A phase
transition corresponds to a change of state, for
example condensation or freezing, and involves
mass transfer between the phases. In this sec-
tion, well look at the thermodynamics of such
phase transitions.
3.2.1 Conditions for phase equilibrium
For a two-phase system to exist in thermody-
namic equilibrium, the two components must sat-
isfy:
1. Thermal equilibrium: T
1
= T
2
2. Mechanical equilibrium: p
1
= p
2
3. Chemical equilibrium:
1
=
2
If the system is not in any one of these equilib-
ria, then either heat exchange, volume redistri-
bution or particle exchange wiil occur across the
phase boundary to equalise T, p or , respec-
tively. From the dierential of Gibbs free energy
(Eq. 2.31),
dG = SdT +V dp
we see that the state corresponding to thermal
and mechanical equilibrium (constant T and p,
respectively) corresponds to a minimum Gibbs
free energy (dG = 0 when dT = 0 and dp = 0).
Therefore, we require dG = 0 for phase equilib-
rium. The consequence of this is that the Gibbs
free energy per particle, g, of either phase are
equal at equilibrium (Derivation 3.3):
g
1
(p, T) = g
2
(p, T) (3.12)
By denition, g equals the chemical potential,
(refer to Eq. 2.39). Therefore, phase equilibrium
implies
1
=
2
, as required.
3.2.2 Maxwell construction
In the coexistence region in Fig. 3.1, a liquid
and its gaseous vapour coexist in equilibrium.
From Fig. 3.1, an isotherm from the van der
Waals equation in this region does not correspond
to an isobar, which means that phase equilib-
rium is not satised along the isotherm. These
p
V
A
B
C
D
E
Figure 3.2: Maxwell construction to determine
the equilibrium pressure for phase coexistence.
The area enclosed above the isobar equals that
enclosed below it.
isotherms therefore do not correspond to stable,
equilibrium states, but instead describe unsta-
ble or metastable states (super-heated liquid or
super-cooled vapour). To address this problem,
Maxwell replaced the van der Waals isotherms
in the coexistence region with horizontal isobars
that correspond to the equilibrium pressure. The
isobars are drawn such that the area enclosed
above the isobar equals that enclosed below it, as
shown schematically in Fig. 3.2 (Derivation 3.4),
so that the work in going from E to A is the same
as if the system had followed the path of the van
der Waals curve. This is known as the Maxwell
construction.
Vapour pressure
The pressure of the isobar that results from the
Maxwell construction is known as the vapour
pressure, p
vap
, at that isothermal temperature.
At this pressure, the system is in equilibrium:
the rate of condensation (vapour to liquid) equals
the rate of vaporisation (liquid to vapour), and
the pressures and temperatures of each compo-
nent are equal.
2
As the vapour is slowly com-
pressed from a large initial volume, it reaches the
point E in Fig. 3.2 which represents a saturated
vapour. Further compression causes condensation
2
The same is true for a vapour in equilibrium with its
solid, but for simplicity, I will refer only to a vapour-liquid
system.
Table 3.3: Derivation of phase equilibrium condition, (Eq. 3.12)
G(p, T, ) = g
1
(p, T) N +g
2
(p, T) (1 ) N
dG(p, T, ) = g
1
(p, T)N d g
2
(p, T)N d
For phase equilibrium:
dG(p, T, ) = 0
g
1
(p, T) = g
2
(p, T)
Starting point: The total Gibbs free energy
of a two-phase system with a fraction par-
ticles in phase 1 and (1) in phase 2 is the
sum of individual energies, where g
1
and g
2
are the Gibbs free energies per particle in
either phase.
The Gibbs free energy is minimized when
dG = 0. This occurs when the Gibbs free
energies per particle are equal.
Table 3.4: Derivation of Maxwell construction (refer to Fig. 3.2)
E
_
A
dG =
E
_
A
V dp
=
C
_
A
V dp +
E
_
C
V dp
= 0 at equilibrium
C
_
A
V dp
. .
Area below isobar
=
E
_
C
V dp
. .
Area above isobar
Starting point: The exact dierential for G
at constant temperature is
dG = V dp
This equals zero at equilibrium
Integrate the van der Waals curve between
points A and E in Fig. 3.2 and apply the
above condition.
3.2. PHASE TRANSITIONS 53
of the vapour at constant pressure (the vapour
pressure), provided the temperature is constant.
When the point A is reached, there exists only liq-
uid, and further compression requires high pres-
sures due to the incompressibility of the liquid
phase.
Violation of the equilibrium condition
The isobars that result from the Maxwell con-
struction represent the equilibrium states for a
system in the coexistence region. Metastable
states correspond to the regions AB and DE
on the curve in Fig. 3.2, whereas unstable states
correspond to the region BD. The metastable
states can sometimes be reached in experiments,
provided the system is carefully expanded (A to
B) or compressed (E to D). In these cases, we
get either a superheated liquid (AB), which does
not boil although its temperature exceeds the
boiling point, or a supercooled gas (DE), which
does not condense although its temperature is be-
low the condensation point. However, since such
systems are metastable, they change in a shock-
like manner back to the equilibrium state under
even very slight disturbances. An example of a
superheated liquid is a bubble chamber, which
is rapidly brought to low pressure with a piston;
boiling then occurs on the ion tracks left by ion-
ising particles. An example of a supercooled gas
is a cloud chamber lled with humid air, which is
then adiabatically expanded.
3.2.3 Phase diagrams
Dierent states of matter and phase transitions
are represented on phase diagrams. These are
plots of pressure against temperature, at con-
stant volume. A typical phase diagram is shown
in Fig. 3.3. The line separating any two phases
is called a transition line or a coexistence curve.
There are three transition lines, and crossing any
of them results in a change of state. Along these
lines, the Gibbs free energy per particle of either
phase are equal, g
1
= g
2
, where the value of g de-
pends on the combination of T and p. The con-
ditions along the transition lines correspond to
the isotherms and isobars described by the coex-
istence region in Fig. 3.1. The point at which the
condensation line ends corresponds to the crit-
ical point, C in Fig. 3.1. Above this point, no
phase transition is possiblethe system exists as
a supercritical uid, which has both liquid and
gas-like properties (further described below). The
point at which the three coexistence curves cross
is called the triple point, T. At this point, the
solid, liquid and vapour forms of a substance ex-
ist simultaneously in equilibrium, and their free
energies per particle are equal, g
s
= g
l
= g
v
.
Supercritical uids
Supercritical uids arise by simultaneously in-
creasing the temperature and pressure of a sys-
tem in the coexistence region, above their critical
values, at constant volume. In so doing, the den-
sity of the vapour increases due to increased pres-
sure whilst the density of the liquid decreases due
to thermal expansion. Eventually, their densities
converge and it is no longer possible to distinguish
the liquid from its vapour. The substance is now
a supercritical uid, which has both gas and liq-
uid properties simultaneously. By slightly chang-
ing the pressure and temperature of the uid, its
density can be tuned to make it behave more
like a liquid or more like a gas. A supercritical
uid becomes a gas if its pressure falls below p
c
;
a liquid if its temperature drops below T
c
; or a
distinguishable mixture of both if conditions drop
below p
c
and T
c
simultaneously. There are many
practical uses for supercritical uids, for example,
in dry cleaning and refrigeration.
Phase diagram for water
Water has an anomalous phase diagram. For con-
ventional systems, increasing the pressure of a
liquid at constant temperature causes freezing.
Water, however, has a negative gradient on the
melting transition line, indicating that increasing
the pressure of ice causes melting. This is well
known since it facilitates ice skating.
3.2.4 Clausius Clapeyron equation
The coexistence curves in Fig. 3.3 describe the
conditions required for phase equilibrium. The
Clausius Clapeyron equation represents the gra-
dient of the coexistent curves, and therefore de-
scribes how the pressure changes in response to
a temperature change, such that the system re-
mains in equilibrium. The Clausius Clapeyron
C
Supercritical
fluid
p
c
T
c
p
T
Gas
Liquid
Solid
T
Melting
Freezing
Vaporisation
Condensation
Sublimation
Figure 3.3: A typical phase diagram. The tran-
sition lines represent equilibrium conditions for
phase coexistence. At the triple point, T, all
three phases coexist in equilibrium. Crossing a
transition line represents a change of phase. The
transition line separating liquid and gas termi-
nates at the critical point, C; the transition line
separating liquid and solid does not terminate.
equation is given in Eq. 3.13 (Derivation 3.5).
Eq. 3.13 shows that there is a change in volume,
V , of the substance undergoing the phase tran-
sition. Worked example 3.6 shows how to cal-
culate the temperature dependence of a liquids
vapour pressure using the Clausius Clapeyron
equation.
Clausius Clapeyron equation:
(Derivation 3.5)
dp
dT
=
L
TV
(3.13)
The order of the transition
Phase transitions are characterised by a step tran-
sition in macroscopic properties of the system.
For example, phase transitions obeying the Clau-
sius Clapeyron equation are characterised by a
step change in volume, V , and also in entropy,
L S. This makes sense intuitively, for ex-
ample when water freezes, its volume increases
and its entropy decreases (it becomes more struc-
tured).
These steps in macroscopic properties corre-
spond to discontinuities in the derivative of the
Gibbs free energy, since, from the exact dieren-
tial of G:
_
G
T
_
p
= S
and
_
G
p
_
T
= V
(refer to Eq. 2.35). The order of the phase tran-
sition is dened as the order of the lowest deriva-
tive of G that shows a discontinuity during the
transition. For example, transitions that obey
the Clausius Clapeyron equation are rst order
transitions, since discontinuities are in V and S
(rst derivatives of G). First order transitions
therefore always take place at constant tempera-
ture and are always associated with a latent heat
corresponding to an entropy step.
Second order transitions have discontinuities in
C
p
and
T
, since these correspond to the second
derivatives of G:
_
2
G
T
2
_
p
=
_
S
T
_
p
=
1
T
_
Q
T
_
p
=
C
p
T
and
_
2
G
p
2
_
T
=
_
V
p
_
T
= V
T
These transitions are not associated with latent
heat.
3.3 Irreversible gaseous expan-
sions
To end the discussion on classical thermodynam-
ics, well take a look at two irreversible processes:
the Joule expansion and the Joule-Thompson ex-
pansion. These experiments were carried out to
determine the temperature change of real and
ideal gases on expansion. We can apply re-
versible thermodynamics to irreversible processes
3.3. IRREVERSIBLE GASEOUS EXPANSIONS 55
Table 3.5: Derivation of the Clausius Clapeyron equation (Eq. 3.13)
dg
1
= dg
2
V
1
dp S
1
dT = V
2
dp S
2
dT
dp
dT
=
S
2
S
1
V
2
V
1
=
S
V
=
L
TV
Starting point: Along a coexistence curve,
the Gibbs free energy per particle of both
phases are equal, g
1
= g
2
.
Use the dierential expression for dg from
Eq. 2.31:
dG = V dp SdT
The latent heat of a phase transition is de-
ned as
L = Q = TS
Table 3.6: Worked example: temperature dependence of vapour pressure using Eq. 3.13
1. For a liquid-vapour phase transition, as-
sume V
gas
V
liquid
:
V = V
gas
V
liquid
V
gas
2. Treat the vapour as an ideal gas:
pV
gas
= nRT
3. Assume that the latent heat is independent
of temperature. The molar latent heat is
L
m
=
L
n
J mol
1
Use steps 1, 2, and 3 in the Clausius Clapeyron
equation and integrate.
dp
dT
=
L
TV
L
TV
gas
=
Lp
nRT
2
=
L
m
p
RT
2
_
dp
p
=
L
m
R
_
dT
T
2
ln(p) =
L
m
RT
+ const.
p(T) = p
0
e
L
m
/RT
gas vacuum
partition
removed
Figure 3.4: Joule expansion. When the parti-
tion is removed, the gas expands against a vac-
uum. The container is thermally isolated, there-
fore heat does not enter or leave the container.
provided the initial and nal states are equilib-
rium states. However, we cannot determine the
intermediate, non-equilibrium properties of the
system.
3.3.1 Joule expansion
Consider a partitioned, thermally isolated con-
tainer, with a gas on one side of the partition
and a vacuum on the other, as shown in Fig. 3.4.
Removing the partition allows the gas to expand
into the vacuum, to ll the container. Since there
is no opposing pressure in a vacuum, no work is
done against the surroundings (dW = 0), and
since the system is thermally isolated, then the
expansion is adiabatic (dQ = 0). Therefore, from
the First Law of Thermodynamics, the internal
energy remains constant during the expansion:
dU = 0 U = constant
Joule coecient
Since Joule was investigating the change in tem-
perature with volume increase, he dened the
Joule coecient,
J
[Km
3
], as:
J
=
_
T
V
_
U
The expression for
J
is given in Eq. 3.14 (Deriva-
tion 3.7).
The Joule coecient: (Derivation 3.7)
J
=
_
T
V
_
U
=
1
C
V
_
T
_
p
T
_
V
p
_
(3.14)
Using the ideal gas law in Eq. 3.14, we nd that
the Joule coecient for an ideal gas is
J
= 0
This agrees with what Joule found experimen-
tally, which was that there was no change in tem-
perature during the expansion. Since C
V
,= 0,
then a zero Joule coecient implies that
_
U
V
_
T
= 0
(refer to Derivation 3.7). Therefore the inter-
nal energy of an ideal gas does not depend on
volumeit is a function of temperature only
which agrees with the results found from kinetic
theory (Section 1.4.1). This means that the en-
ergy of ideal gas particles is independent of their
separation, indicating that there are no inter-
molecular forces, as expected.
For a real gas, however, the internal energy is
a function of both temperature and volume, as
shown in worked example 3.5. From this, we nd
that a free expansion of a real gas always results
in cooling:
_
T
V
_
U
=
a
C
V
v
2

J
< 0
3.3.2 Joule-Thomson expansion
The Joule-Thomson experiment was carried out
to further understand the cooling properties of
a real gas on expansion. In this case, instead
of a free expansion against a vacuum, the gas
is allowed to expand through a thermally iso-
lated throttling valve, as shown in Fig. 3.6. The
Table 3.7: Derivation of the Joule coecient (Eq. 3.14)
_
T
V
_
U
=
_
U
V
_
T
_
T
U
_
V
=
_
U
V
_
T
_
U
T
_
1
V
=
_
T
_
S
V
_
T
p
__
T
S
T
_
1
V
=
1
C
V
_
T
_
S
V
_
T
p
_
=
1
C
V
_
T
_
p
T
_
V
p
_
Starting point: The reciprocity theorem
gives
_
x
y
_
z
=
_
z
y
_
x
_
x
z
_
y
Use the exact dierential for U (Eq. 2.22),
dU = TdS pdV
to nd expressions for the partial deriva-
tives of U.
The heat capacity at constant volume is de-
ned as
C
V
=
_
Q
T
_
V
=
_
T
S
T
_
V
Use the Maxwell relation in Eq. 2.33 to
eliminate the entropy derivative.
Figure 3.5: Worked example: Internal energy of a real gas
1. Entropy is a state variable, and therefore
can be written as an exact dierential in
terms of any two of the fundamental vari-
ables p, V and T. For example, for S =
S(V, T), the exact dierential is
dS =
_
S
T
_
V
dT +
_
S
V
_
T
dV
Use this to eliminate dS from the exact dif-
ferential for U.
2. Use the denition of C
V
to eliminate the S-
T derivative (given in Derivation 3.7), and
the Maxwell relation from Eq. 2.33 to elim-
inate the S-V derivative.
3. Use the van der Waals equation of state
(Eq. 3.2) to eliminate the pressure terms.
4. For an ideal gas, a = 0, and therefore we
recover dU = C
V
dT
dU = TdS pdV
= T
__
S
T
_
V
dT +
_
S
V
_
T
dV
_
pdV
= C
V
dT +T
_
p
T
_
V
dV pdV
= C
V
dT +
RT
v b
dV
RT
v b
dV +
a
v
2
dV
= C
V
dT +
a
v
2
dV
purpose of the throttling valve, which could also
be a porous plug or a pinhole, is to provide an
impedance against the ow, which would reduce
the pressure of the gas, from p
1
to p
2
(p
1
> p
2
),
without the gas doing external work. The pistons
are moved so as to keep p
1
and p
2
constant.
In this case, the internal energy is not a con-
stant, as it was for the Joule expansion. We nd
instead that the enthalpy, H, is constant (Deriva-
tion 3.8).
Joule-Thomson coecient
The Joule-Thomson coecient,
J
[KPa
1
], de-
scribes the change in temperature with pressure
change, at constant enthalpy, and is dened as:
J
=
_
T
p
_
H
It has an analogous form to
J
, as given by
Eq. 3.15 (Derivation 3.9).
The Joule-Thomson coecient:
(Derivation 3.9)
J
=
_
T
p
_
H
=
1
C
p
_
T
_
V
T
_
p
V
_
(3.15)
Using the ideal gas law in Eq. 3.15, we nd for
an ideal gas that
J
= 0, which means that its
temperature does not change on throttling. For
a real gas, however,
J
is either positive or neg-
ative, depending on the initial values of T and p.
Since p always decreases (p
1
> p
2
), then:
If
J
> 0 : the gas cools down (dT < 0)
If
J
< 0 : the gas warms up (dT > 0)
The sign of
J
for dierent combinations of T and
p is shown in Fig. 3.7. There is an inversion curve,
which is a function of temperature and pressure,
for which the temperature change on expansion
is zero. The inversion curve is given by the con-
dition that
J
= 0
T
p
< 0: heating
> 0: cooling
= 0: inversion curve
Figure 3.7: Isenthalps (lines of constant enthalpy)
for the Joule-Thomson expansion of a real gas.
The inversion curve indicates the conditions for
which there is no temperature change on expan-
sion. Plot using mathematica.
Since C
p
,= 0, this implies that on the inversion
curve:
T
_
V
T
_
p
V = 0
_
V
T
_
p
=
V
T
(3.16)
Inside the inversion curve,
J
> 0, and Joule-
Thomson throttling can be used to cool gases un-
til they liquify, using a countercurrent heat ex-
changer. Outside of the inversion curve,
J
< 0
and the gas heats up on throttling. Fig. 3.7 shows
that there is a maximum temperature, above
which throttling heats gases for all pressures.
This occurs when the inversion curve crosses the
pressure axis at the high-temperature end, and
is therefore given by the conditions
J
= 0 and
p = 0. The value of the inversion temperature
can be found by using the van der Waals equa-
tion (Eq. 3.1) in Eq. 3.16.
Throttling valve
Piston in Piston out
p
1
, V
1
, T
1
p
2
, V
2
, T
2
Figure 3.6: Joule-Thomson expansion. The gas is expanded through a throttle valve, which provides
an impedance against the ow and therefore reduces the pressure of the gas from p
1
to p
2
(p
1
> p
2
),
where p
1
and p
2
are kept constant during the process. The container is thermally isolated, therefore
heat does not enter or leave the container. Show that on side 1 the nal volume is zero and on side 2
the initial volume is zero.
Table 3.8: Derivation of conservation of enthalpy for Joule-Thomson expansion
U = W
= p
1
V
1
p
2
V
2
= p
1
(0 V
1
) p
2
(V
2
0)
= p
1
V
1
p
2
V
2
U
2
U
1
= p
1
V
1
p
2
V
2
U
2
+p
2
V
2
= U
1
+p
1
V
1
H
2
= H
1
enthalpy is conserved
Starting point: For a thermally isolated sys-
tem, dQ = 0, therefore the First Law gives
dU = dW
U = W, where W takes into account
the change in volume of the gas on ei-
ther side of the throttling valve (refer to
Fig. 3.6).
Enthalpy is dened as H = U +pV
Table 3.9: Derivation of the Joule-Thomson coecient (Eq. 3.15)
_
T
p
_
H
=
_
H
p
_
T
_
T
H
_
p
=
_
H
p
_
T
_
H
T
_
1
p
=
_
T
_
S
p
_
T
+V
__
T
S
T
_
1
p
=
1
C
p
_
T
_
S
p
_
T
+V
_
=
1
C
p
_
T
_
V
T
_
p
V
_
Starting point: The reciprocity theorem
gives
_
x
y
_
z
=
_
z
y
_
x
_
x
z
_
y
Use the exact dierential for H (Eq. 2.30),
dH = TdS +V dp
to nd expressions for the partial deriva-
tives of H.
The heat capacity at constant pressure is
dened as
C
p
=
_
Q
T
_
p
=
_
T
S
T
_
p
Use the Maxwell relation in Eq. 2.35 to
eliminate the entropy derivative.
Chapter 4
Statistical mechanics
We saw in kinetic theory and classical thermo-
dynamics (Chapters 1 and 2) that we can de-
scribe systems by taking either a microscopic
approach (considering the average properties of
individual particles in a system), or by taking
a macroscopic approach (considering solely the
gross properties of systems). The macroscopic
approach, however, is limited: it describes sys-
tems in the classical regime, but not in the sub-
atomic, or quantum regime. Statistical mechanics
thus uses a probabilistic, microscopic approach,
similar to that used in kinetic theory, to describe
the macroscopic properties of any system (it is
not restricted to ideal gases). When we apply
statistical mechanics to classical systems, we re-
cover the same results as those from thermody-
namics, whereas by treating systems in the quan-
tum regime, we discover new phenomena that
could not be anticipated from thermodynamics.
Since statistical mechanics uses a microscopic ap-
proach to explain macroscopic properties, it links
together the familiar classical world that we expe-
rience every day, with the unfamiliar subatomic,
quantum world of individual particles. It demon-
strates that all we see around us is a consequence
of the combined behaviour of individual atoms.
4.1 Statistical ensembles
In classical thermodynamics, we dened three dif-
ferent types of system, depending on whether en-
ergy and matter can be exchanged with their sur-
roundings (Section 2.1):
1. Isolated systems: Neither energy nor matter
exchange is possible, therefore the total en-
ergy, U, and total number of particles in the
system, N, are constant
2. Closed systems: Energy exchange can occur,
but not exchange of matter. Therefore U is
variable and N is constant.
3. Open systems: Both energy and matter ex-
change can occur. Therefore, both U and N
are variable.
In statistical mechanics, we make similar classi-
cations. However, instead of dealing with a sin-
gle system, we now wish to look at a collection of
macroscopically-identical systems. This allows us
to determine the probabilistic microscopic prop-
erties of the system, for example the probability
distribution of particle energies. The collection
of such systems is called a statistical ensemble.
There are three common statistical ensembles,
which correspond to the three types of thermo-
dynamic systems described above:
1. Microcanonical ensemble: Collection of iso-
lated systems (U and N are constant).
2. Canonical ensemble: Collection of closed
systems (U is variable and N is constant).
3. Grand canonical ensemble: Collection of
open systems (U and N are variable).
In theory, an ensemble comprises an innite num-
ber of identical systems. Of course in practice,
this is impossible to prepare, therefore statistical
ensembles are mathematical idealisations. Statis-
tical mechanics is thus based on two fundamental
postulates, which allow us to determine the prob-
ability distributions of the microscopic properties
of a given system. These are outlined below.
4.2 The postulates of statistical
mechanics
63
64 CHAPTER 4. STATISTICAL MECHANICS
4.2.1 States and energy levels
The state of a particle is represented by its
wavefunction, (q
1
, q
2
. . . q
n
), where the variables
q
1
, q
2
. . . q
n
correspond to the variables needed to
dene the state. For example, q
1
may correspond
to the particles position, q
2
its momentum, etc.
The square of the wavefunction represents the
probability that a particle has a given value of a
given variable. For example, the probability that
a particle represented by a wavefunction (q
1
),
has a < q
1
< b is given by:
P
ab
=
b
_
a
[(q
1
)[
2
dq
1
(refer to the Quantum Mechanics section, Chap-
ter ??, for more detail). The state of a parti-
cle is often specied, in part, by its energy, .
From quantum mechanics, we know that energy
is quantised: there exists a discrete set of allowed
energy levels that a particle can occupy, which
arise from boundary conditions imposed on the
wavefunction. At high energies, the dierence
between adjacent energy levels is much smaller
than the total energy of the system, and we can
treat as a continuous variable. In this case, we
bundle several energy levels together, and refer
to a particle having energy between and +d,
rather than it being in the n
th
level, since n
at high energies. Discrete and continuous energy
levels are illustrated in Fig. 4.1.
Degeneracy
The terms state and energy level are sometimes
used interchangeably, but there is an important
distinction between them. The energy level tells
us only the particles energy; the state, however,
may also include the particles angular momen-
tum or its spin, for example. This means that any
given energy level may comprise several states of
diering momentum or spin. When several dif-
ferent states correspond to the same particle en-
ergy, then that energy level is said to be degen-
erate. For example, in the absence of a magnetic
eld, the spin-up and spin-down electron states
in a hydrogen atom are at the same energy, and
therefore there is a 2-fold spin degeneracy at each
energy level. For a discrete energy distribution,
degeneracy is usually denoted by the symbol g
i
,
}
bundle of
energy levels
Energy level, n
3
+d
Continuous
energy levels
Discrete
energy levels
0
n = 3
n = 2
n = 1
n = 0
Figure 4.1: Discrete and continuous energy lev-
els. When the spacing between adjacent energy
levels is negligible (continuous distribution), then
energy levels are bundled into intervals of size
d.
which represents the number of states in the i
th
energy level.
4.2.2 Microstates and macrostates
Distinguishable particles
To begin with, lets assume that the particles in
a system are distinguishable. This means that
we can, in theory, tell the dierence between
them. For example, particles localised in a crystal
are distinguishable because of their position, and
atoms of dierent elements are obviously distin-
guishable by their nature. However, even identi-
cal particles in an ideal gas are distinguishable
they can, in theory, be told apart from one
another. This is because distinguishable par-
ticles are dened as having wavefunctions that
do not overlap in space with one another (the
inter-particle distance is much greater than the
de Broglie wavelength of the particle, see Sec-
tion 4.6.3). Therefore, each particle occupies its
own individual potential well. By denition, the
wavefunctions of particles in classical systems do
not overlap, therefore all classical systems, such
as ideal gases, comprise distinguishable particles.
In Section 4.5, we will look at indistinguishable
particlesones whose de Broglie wavelength is
comparable with their inter-particle separation.
Indistinguishability is a quantum mechanical phe-
nomenon that has no classical analogue.
4.2. THE POSTULATES OF STATISTICAL MECHANICS 65
The states of distinguishable particles
There are two terms used to describe a systems
state in terms of the energy levels occupied by
its particlesthe microstate and the macrostate.
The microstate of a system tells us which particle
is in which energy level whereas the macrostate
tells us the total number of particles in each en-
ergy level, and hence the total energy of the sys-
tem seen on macroscopic scales, without making
reference to individual particles. The distinction
is best understood through example. Consider
four particles, labelled A, B, C and D, each in
their own potential well, occupying energy levels
n = 13, as shown in Fig. 4.2. The microstate
and macrostate are dened in the gure. From
the macrostate, the total energy of the system
corresponds to E = 8 (in arbitrary units).
1
3
2
A B C D
Microstate Macrostate
Particle Energy Energy #particles
A
B
C
D
1
2
2
3
1
2
3
1
2
1
Total energy = 1+2+2+3 = 8
Figure 4.2: Example of a possible microstate and
macrostate of a system of four distinguishable
particles (A, B, C and D), occupying three en-
ergy levels (1, 2 and 3).
Fig. 4.3 illustrates two dierent systems with
the same total energy (E = 8), but with a) a
dierent microstate corresponding to the same
macrostate, and b) both a dierent microstate
and a dierent macrostate. The postulates of
statistical mechanics, shown below, are concerned
with describing the microstates and macrostates
of a given system.
The rst and second postulates
The rst postulate of statistical mechanics de-
scribes the probability that a system is in a par-
ticular microstate of a given macrostate,
1
and is
called the postulate of equal a priori
2
probability:
Postulate 1: For a macrostate in equilibrium,
all microstates corresponding to that macrostate
are equally probable.
This means that the microstate in Fig. 4.3 a)
has the same probability of occurring as that in
Fig. 4.2, since they have identical macrostates.
The consequence of this postulate is that if there
are dierent microstates corresponding to a
given macrostate, then the probability of nding
the system in any one of those microstates is 1/.
The second postulate of statistical mechanics tells
us, for a given total energy, which macrostate the
system is likely to be in:
Postulate 2: The observed macrostate is the
one with the most microstates, .
For example, there is only one way of getting
the macrostate in Fig. 4.3 b), which means that
there is only one microstate (all the particles are
in the second energy level), and therefore
b
= 1.
In contrast, we nd there are 12 dierent ways of
getting the macrostate in Fig. 4.3 a), and there-
fore
a
= 12 (check by writing out the various
dierent combinations). Since
a
is greater than
b
, then the macrostate in a is more likely to be
observed than that in b, for the same total energy.
4.2.3 Properties of
From the discussion above, the parameter
is dened as the number of microstates in a
macrostate. For a system of N distinguishable
particles, with n
1
particles in energy level 1, n
2
in level 2, and n
i
in level i, is calculated using
a probability combination formula:
=
N!
n
1
! n
2
! n
3
! . . .
(4.1)
1
Here, we are dealing with a microcanonical ensemble,
which corresponds to a collection of systems in which the
total number of particles and total energy of the system
are constant.
2
A priori means that the postulate has come from the-
oretical deduction rather than from experimental observa-
tion.
1
3
2
A B C D
A
B
C
D
2
2
3
1
1
2
3
1
2
1
1
3
2
A B C D
A
B
C
D
2
2
2
2
1
2
3
0
4
0
a) Same macrostate, different microstate b) Same total energy, different macrostate
Figure 4.3: Systems with the same total energy as that in Fig. 4.2, but for a) a dierent microstate
and b) a dierent macrostate (and microstate).
For example, for the macrostates in Fig. 4.3, we
nd
a
=
4!
1! 2! 1!
= 12
and
b
=
4!
0! 4! 0!
= 1
as expected. In most systems, N is a very large
number. For example, a mole of gas has of order
N 6 10
23
particles (Avagadros number). The
number of particles in any one state, n
i
, is usu-
ally much less than N. In consequence there is
one macrostate for which is much larger than
those corresponding to dierent macrostates of
the same total energy. As a simple example, con-
sider the third macrostate which gives the same
total energy as those in Fig. 4.3 (E = 8): two par-
ticles in level 1, zero in level 2, and two in level
3. The number of microstates in this macrostate
is therefore:
=
4!
2! 0! 2!
= 6
which is signicantly less than
a
. Therefore,
we nd that is strongly peaked at a single
macrostate, for any given total energy. Not
shown rigorous proofis this sucient? From
Postulate 2, this corresponds to the observed
macrostate.
Another look at entropy
Since the number of microstates in a macrostate
depends on the total energy of the macrostate,
then is a function of energy: (). Lets assume
that the value of at a particular energy, , is
that corresponding to the observed macrostate
the one for which is a maximum (Postulate
2). In this case, we can treat as a continuous
function of energy. Can we??
This takes us to another postulate of statistical
mechanics, which is that the entropy of a system
is proportional to the logarithm of , and is given
by Eq. 4.2. This is a denition of entropy, from a
statistics perspective, and cannot be proved.
Entropy:
S = k
B
ln (4.2)
This denition shows that entropy can be
thought of as a measure of the number of ways
of arranging quanta of energy between the parti-
cles in a system (since this is what the variable
represents), which is consistent with the interpre-
tation of entropy as a measure of disorder. We
recover dU = TdS provided we dene
ln
=
1
k
B
T
Therefore, by dierentiating Eq. 4.2, we nd:
S
=
k
B
= k
B
ln
4.3. THE DENSITY OF STATES 67

=
1
T
Rearranging and replacing the partial derivative
with a total derivative, we have d = TdS, which
is analogous to dU = TdS from classical thermo-
dynamics.
4.3 The density of states
If we consider an energy interval d, then the
value of ( + d) () gives us the number
of microstates within that interval. Treating
as a continuous function of , and performing a
Taylor expansion on ( +d) allows us to dene
a function which represents the total number of
states that lie between an energy and +d:
( +d) () () +
d
d
d ()
=
d
d
d
g()d
The function g() is called the density of states;
it represents the number of states per energy in-
terval d, and is a continuous function of energy.
We usually express the density of states as the
number of states per unit volume, which justies
using the term density.
Boundary conditions
If we consider the simple example of a particle
in a 1D innite square well (see Quantum Me-
chanics, Chapter ??), then we know that only
certain wavefunctions satisfy the boundary con-
ditions imposed on the system. This means that
only certain states, and hence energy levels, are
allowed to exist. Generalising to a system of any
shape and dimension, the same thing occurs
some wavefunctions will satisfy the boundary
conditions, allowing particles to occupy those
states, whereas others wont and therefore no par-
ticle can exist in those states. The density of
states tells us which quantum states satisfy the
boundary conditions of a given system. There-
fore, the function g()d depends on the dimen-
sions of the system in question, and has a dierent
energy dependence for systems in 1D, 2D or 3D,
as we shall see below.
4.3.1 Derivation in k-space
The energy levels of a system depend on the
mass of the particle. This means that the form
of g() d is dierent depending on whether the
system involves a particle of mass m or a mass-
less particle, such as a photon. Therefore, to de-
rive an expression for the density of states, we
work in terms of wavelength rather than energy,
since the allowed wavelengths are the same for
both particles with mass and massless particles.
The derivation for the density of states is more
straightforward if we work in k-space, rather than
in -space, where the wavenumber, k, is related
to the wavelength by
k =
2
The density of states in k-space is then denoted

by g(k)dk. By generalising the boundary con-
ditions for a particle in an innite square well,
we can derive expressions for g(k)dk in 3D, 2D
and 1D. These are given by Eqs. 4.34.5 respec-
tively (Derivation 4.1), and plotted in Fig. 4.4.
These expressions hold both for particles with
mass and for massless particles. However, the
density of states is multiplied by a degeneracy
factor, g, which does depend on the nature of
the particle. This accounts for spin or polar-
ization degeneracy. For electrons, each state of
wavevector k can accommodate two electrons
one in the spin-up state and one in the spin-down
state. Therefore, the spin degeneracy for elec-
trons, and other spin-half particles, is g
s
= 2.
Massless particles have a polarization degener-
acy. For photons, each state of wavevector k
can accommodate two photonsone transverse
and one longitudinal polarizationand therefore
g
p
= 2. For phonons, which are the massless par-
ticles corresponding to acoustic waves in matter,
there is an extra transverse polarization state,
therefore the phonon polarization degeneracy is
g
p
= 3. The total number of states per unit vol-
ume/area/length is found by integrating over all
values of k (Eq. 4.6).
Density of states in k-space:
(Derivation 4.1)
In 3D: g
3
(k)dk = g
k
2
2
2
dk (4.3)
In 2D: g
2
(k)dk = g
k
2
dk (4.4)
In 1D: g
1
(k)dk = g
1
dk (4.5)
where the degeneracy factor, g, is due to spin or
polarization degeneracy:
g =
_
_
2 for electrons
2 for photons
3 for phonons
Total number of states per unit vol-
ume/area/length:
n
s
=
_

0
g(k)dk (4.6)
4.3.2 Converting to energy
Once we have expressions for the density of states
in k-space, its straightforward to convert to -
space using the dispersion relations that relate k
to :
For particles of mass m: =

2
k
2
2m
(4.8)
For massless particles: = kc (4.9)
(refer to Section ??, Eqs. ?? and ??). The den-
sity of states for particles with mass are given
in Eqs. 4.104.12 (Derivation 4.2) and plotted
in Fig. 4.5, and those for massless particles in
Eqs. 4.134.15, (Derivation 4.3) and plotted in
Fig. 4.6. The total number of states per unit vol-
ume/area/length is found by integrating over all
values of (Eq. 4.16).
Density of states in -space:
Particles with mass:
3D: g
3
()d = g
2m
3
2
2
3
d (4.10)
2D: g
2
()d = g
m
2
2
d (4.11)
1D: g
1
()d = g
_
m
2
2
d (4.12)
Massless particles:
3D: g
3
()d = g

2
2
2
(c)
3
d (4.13)
2D: g
2
()d = g

2(c)
2
d (4.14)
1D: g
1
()d = g
1
c
d (4.15)
The degeneracy factor, g, is due to spin or po-
larization degeneracy:
g =
_
_
2 for electrons
2 for photons
3 for phonons
Total number of states per unit vol-
ume/area/length:
n
s
=
_

0
g()d (4.16)
Density of modes
By using the dispersion relations for k and ,
we can express the density of states in frequency
space, g()d. For vibrational systems, instead
of referring to a state at wavevector k, we talk
about a mode at frequency . Said another way,
a mode corresponds to a state of a system that is
undergoing vibration. Therefore, g()d repre-
sents the number of modes per unit volume with
a frequency between and + d. The deriva-
tions for g()d in 3D, 2D and 1D are similar to
those for g()d (derivations. 4.24.3), except we
use the dispersion relation (k) instead:
For particles with mass: =
k
2
2m
(4.23)
For massless particles: = ck (4.24)
Table 4.1: Derivation of the density of states in k-space, g(k) dk (Eqs. 4.34.5)
1
n l
m
r
dr
In 3D:
G
3
(k)dk =
1
8
4r
2
dr
=

2
L
3
k
2
3
dk
=
V k
2
2
2
dk
g
3
(k)dk =
k
2
2
2
dk
In 2D:
G
2
(k)dk =
1
4
2rdr
=

2
L
2
k
2
dk
=
Ak
2
dk
g
2
(k)dk =
k
2
dk
In 1D:
G
1
(k)dk = dr
=
L
dk
g
1
(k)dk =
1
dk
From the solution of the Schrodinger equation for a par-
ticle in a one-dimensional innite square well (refer to
worked example ??), the allowed wavevectors in a con-
nement of length L are given by integer values of /L:
k
x
=
n
L
In three-dimensions, the total wavevector is:
k =
_
k
2
x
+k
2
y
+k
2
z
=

L
_
n
2
+m
2
+l
2
r
L
(4.7)
where r
2
= n
2
+m
2
+l
2
The quantum states dened by integers n, m and l fall
on a grid in the positive x, y, z octant (refer to diagram).
In 3D: the number of states between k and k + dk is
given by the volume of a spherical shell in the positive
octant. Use Eq. 4.7 to eliminate r and dr. The volume
of connement is V = L
3
. Therefore, to nd the density
of states we divide by the volume.
In 2D: the number of states between k and k+dk is given
by the area of a circular shell in the positive quadrant.
Use Eq. 4.7 to eliminate r and dr. The area of conne-
ment is A = L
2
. Therefore, divide by the area to nd
the density of states.
In 1D: the number of states between k and k + dk is
given by the length dr. The connement length is L,
therefore, divide by L for the 1D density of states.
g(k)
k
g(k)k
2
3-D
g(k)
k
g(k)k
2-D
g(k)
k
g(k)=const
1-D
Figure 4.4: Density of states in k-space, for both particles with mass and massless particles, in 3D, 2D
and 1D. The dependence on k is quadratic in 3D, linear in 2D, and constant in 1D. Plot in mathematica
g()
k
g()
3-D
g()
k
g()=const
2-D
g()
k
g()1/
1-D
Figure 4.5: Density of states for particles with mass in -space in 3D, 2D and 1D. The dependence on
in 3D is

, is constant in 2D, and in 1D is 1/
. Plot in mathematica
g()
k
g()
2
3-D
g()
k
g()
2-D
g()
k
g()=const
1-D
Figure 4.6: Density of states for massless particles in -space in 3D, 2D and 1D. The dependence on
is quadratic in 3D, linear in 2D, and constant in 1D. Plot in mathematica
Table 4.2: Derivation of the density of states in -space for particles with mass (Eq. 4.10
4.12)
In 3D: g
3
() d = g
3
(k)
dk
d
d
=
1
2
2
2m
2
_
m
2
2
d
=
2m
3
2
2
3
. .
=
3
d
g
3
()d =
3
d
In 2D: g
2
() d = g
2
(k)
dk
d
d
=
1
2
2m
_
m
2
2
d
=
m
2
2
. .
=
2
d
g
2
()d =
2
d
In 1D: g
1
() d = g
1
(k)
dk
d
d
=
1
_
m
2
2
d
=
_
m
2
2
2
. .
=
1
1
d
g
1
()d =
1
1
d
Starting point: There is a one-to-one corre-
spondence between g()d and g(k)dk
g() = g(k)
dk
d
The dispersion relation for particles with
mass of mass m is:
=

2
k
2
2m
d =

2
m
kdk
Rearranging the above, we have
k =
2m
and dk =
_
m
2
2
d
Eliminate k and dk from Eqs. 4.34.5
Dene the constants,
3
,
2
and
1
as:
3
=
2m
3
2
2
3
(4.17)
2
=
m
2
2
(4.18)
1
=
_
m
2
2
2
(4.19)
Table 4.3: Derivation of the density of states in -space for massless particles (Eq. 4.13 to
4.15)
In 3D: g
3
() d = g
3
(k)
dk
d
d
=
1
2
2
_

c
_
2
d
c
g
3
()d =
3

2
d
In 2D: g
2
() d = g
2
(k)
dk
d
d
=
1
2
c
d
c
g
2
()d =
2
d
In 1D: g
1
() d = g
1
(k)
dk
d
d
=
1
d
c
g
1
()d =
1
d
Starting point: There is a one-to-one corre-
spondence between g()d and g(k)dk
g() = g(k)
dk
d
The dispersion relation for massless parti-
cles is:
= kc
d = cdk
Rearranging the above, we have
k =

c
and dk =
d
c
Eliminate k and dk from Eqs. 4.34.5
Dene the constants
3
,
2
and
1
as:
3
=
1
2
2
(c)
3
(4.20)
2
=
1
2(c)
2
(4.21)
1
=
1
c
(4.22)
4.4. CLASSICAL STATISTICS 73
(refer to Section ??, Eqs. ?? and ??).
4.4 Classical statistics
Statistical mechanics is split into two regimes:
classical and quantum. We saw above that classi-
cal systems correspond to those in which particles
are distinguishableor, said another way, parti-
cles whose wavefunctions do not overlap (inter-
particle separation de Broglie wavelength).
Classical statistics are based on the Boltzmann
distribution, which we used without proof in ki-
netic theory to determine macroscopic properties
of classical systems from the microscopic distribu-
tion of particle energies (Chapter 1). Now we will
look at where the Boltzmann distribution orig-
inates, by considering how a particle interacts
with its surroundings. The derivation outlined
below is not rigorous, but it provides a basic un-
derstanding of what the Boltzmann distribution
represents.
4.4.1 The Boltzmann distribution
Consider a closed sub-system in thermal contact
with a heat reservoir, as shown in Fig. 4.7. To-
gether, the sub-system plus the reservoir form an
isolated system, with total energy U
0
. The ener-
gies of the reservoir and sub-system are U
R
and
, respectively, where U
0
= U
R
+. From the sec-
ond postulate of statistical mechanics, the proba-
bility of the reservoir being in a given macrostate
is proportional to the number of microstates in
that macrostate: P . Since is a function of
energy, so is P. From Eq. 4.2, the entropy of the
reservoir is
S
R
= k
B
ln
R

R
= e
S
R
/k
B
(4.25)
We can nd an expression for S
R
from the exact
derivative of U (dU = TdS pdV ). The volume
of the reservoir is constant, therefore dV
R
= 0
and so, by rearranging dU, the entropy of the
reservoir is
dS
R
=
dU
R
T
S
R
=
U
R
T
Setting U
R
= U
0
, and letting S
0
= U
0
/T,
which is a constant, we nd
S
R
= S
0

T
(4.26)
Since U
0
is constant, the probability of the sub-
system having an energy , equals the probability
that the reservoir has an energy U
R
. Therefore,
P() is proportional to
R
. Using Eq. 4.26 in
Eq. 4.25, we nd:
P()
R
()
e
S
R
()/k
B
e
/k
B
T
Therefore we have recovered the exponential
Boltzmann factor by considering a sub-system of
energy in thermal equilibrium with a heat reser-
voir, and using the postulates of statistical me-
chanics. If we treat the sub-system as a single
particle, then P()d gives the probability that a
particle has an energy between and +d.
U
0
=U
R
+
U
R
Reservoir
Sub-system
Figure 4.7: Isolated system comprising a closed

sub-system with energy in thermal contact with
a heat reservoir with energy U
R
. Energy can
be exchanged between the reservoir and the sub-
system.
The Boltzmann distribution is given by the
Boltzmann factor multiplied by a normalisation
constant, A (Eq. 4.27). It shows that the proba-
bility distribution of particles decreases exponen-
tially with increasing particle energy.
The Boltzmann distribution:
f
B
()d = Ae
/k
B
T
d (4.27)
Occupation functions
The Boltzmann distribution is known in statis-
tical mechanics as an occupation function, f().
This is a probability distribution which repre-
sents the average number of particles per state.
There exist two other occupation functions: the
Bose-Einstein and the Fermi-Dirac distributions.
These apply to systems in the quantum regime,
and are covered in Section 4.6.
4.4.2 The number density distribution
So far, we have come across the density of states,
g()d, which represents the number of states per
unit volume, and the occupation function, f(),
which represents the average number of particles
per state. The product of these two functions
yields another probability distribution, called the
number density distribution, n()d. This distri-
bution represents the average number of particles
per unit volume:
n() d
. .
particles
/volume
= f()
..
particles
/state
g() d
. .
states
/volume
The normalisation condition is that the total
number density of particles, n
T
, is given by the
integral of n()d over all energies (Eqs. 4.28 and
4.29).
Number density distribution:
Particles with energy between and +d:
n()d = f() g()d (4.28)
Total number density of particles:
n
T
=
_
0
n()d
=
_
0
f() g()d (4.29)
Eq. 4.28 shows that the distribution of parti-
cle energies in a system depends on two quanti-
ties. It depends on the probability that a parti-
cle has an energy , which is given by f(), and
it depends on the number of states available at
that energy, which is given by g()d. The den-
sity of states is therefore sometimes thought of
as a weight functionthere is a greater weight in
the probability of a particle having an energy
if there are more states available at that energy.
Conversely, if there are no states available at a
particular energy, say
0
, for a particular system,
such that g(
0
)d = 0, then no particles will have
that energy, even if there is a high corresponding
probability from the occupation function, f(
0
).
These ideas are shown schematically in Fig. ??
(draw two curves and their product).
4.4.3 The partition function and ther-
modynamic properties
Now lets determine the normalisation factor,
A, of the Boltzmann distribution. Substituting
Eq. 4.27 into Eq. 4.29, we have:
n
T
= A
_
0
g()e
/k
B
T
d
The integral shown here is an important quan-
tity in statistical mechanics, called the partition
function, Z:
Z =
_
0
g()e
k
B
T
d
Therefore A = n
T
/Z, and the normalised Boltz-
mann distribution is given by Eq. 4.30. For a dis-
crete set of energy levels,
i
, the integral becomes
a sum over states i, and the density of states ap-
proximates to the degeneracy, g
i
, of the i
th
energy
level (Eqs. 4.31 and 4.32).
Partition function:
Normalised Boltzmann distribution
f
B
()d =
n
T
Z
e
/k
B
T
d (4.30)
Where the partition function, Z, is:
Discrete energy distribution:
Z =

i
g
i
e
i
/k
B
T
(4.31)
Continuous energy distribution:
Z =
_
0
g()e
/k
B
T
d (4.32)
It is not intuitive what the partition function
represents physically, but it is often described as
the sum over states. It is a function that en-
codes the statistical properties of the microstates
of a system, and it depends on the systems tem-
perature and physical dimensions. The partition
function is an important quantity in statistical
mechanics since it provides the link between a
systems microscopic energy levels and its macro-
scopic properties. We will see below that essen-
tially all the macroscopic, thermodynamic prop-
erties of a system can be calculated from its par-
tition function.
Internal energy and heat capacity
The internal energy, U, of a system is the sum of
the energies of its individual particles. The ex-
pression for U is related to the partition function
by the temperature derivative of its natural log-
arithm (Eq. 4.33 and Derivation 4.4). The heat
capacity of the system is then given by the tem-
perature derivative of U (Eq. 4.34).
Internal energy and heat capacity:
(Derivation 4.4)
U = Nk
B
T
2
lnZ
T
(4.33)
C
V
= Nk
B
T
_
2
lnZ
T
+T

2
lnZ
T
2
_
(4.34)
Helmholtz free energy
Recall from thermodynamics (Section 2.3.2) that
the Helmholtz free energy, F, is a thermodynamic
potential, which is related to the internal energy
by F = U TS. Using the statistical denition
of entropy (Eq. 4.2), we can derive an expression
for F in terms of the natural logarithm of the
partition function (Eq. 4.35 and Derivation 4.5).
This is a very useful quantity, since all other ther-
modynamic properties are easily derived from it.
The expressions for pressure and entropy follow
from the exact derivative of F (refer to Eq. 2.33),
and are therefore equal to single derivatives of
F (Eqs. 4.36 and 4.37 respectively). The other
thermodynamic potentials, U, H and G are com-
binations of F and its derivatives, and are given
in Eqs. 4.38 to 4.40. Using Eq. 4.35 in Eq. 4.38,
we recover the expression for U from Eq. 4.33.
Worked example 4.6 shows how to calculate some
thermodynamic properties of an ideal gas from its
partition function.
Other worked examples? SHO or paramagnet
Helmholtz free energy:
F = Nk
B
T lnZ (4.35)
Quantities derived from F:
p =
_
F
V
_
T
(4.36)
S =
_
F
T
_
V
(4.37)
U = F +TS
= F T
_
F
T
_
V
(4.38)
H = U +pV
= F T
_
F
T
_
V
V
_
F
V
_
T
(4.39)
G = U TS +pV
= F V
_
F
V
_
T
(4.40)
Table 4.4: Derivation of the internal energy from the partition function (Eq. 4.33)
u =

n
i
i
=

Ag
i
i
e
i
/k
B
T
=
n
T
Z
g
i
i
e
i
/k
B
T
=
n
T
Z
Zk
B
T
2

T
lnZ
= n
T
k
B
T
2

T
lnZ
Total internal energy (n
T
= N/V ):
U = uV
= Nk
B
T
2

T
lnZ
Heat capacity:
C
V
=
_
U
T
_
V
= Nk
B
T
_
2
lnZ
T
+T

2
lnZ
T
2
_
Starting point: The internal energy of a system is the
sum of its individual particle energies. If there are n
i
particles per unit volume in the i
th
energy level, which
has energy
i
, then the total energy density is
u =

n
i
i
The discrete number density distribution and partition
function are
n
i
= Ag
i
e
i
/k
B
T
Z =

g
i
e
i
/k
B
T
where g
i
is the degeneracy of the i
th
state, and the nor-
malisation constant is A = n
T
/Z.
Dierentiating the natural logarithm of the partition
function, we have:
T
lnZ =

T
_
ln
_
g
i
e
i
/k
B
T
__
=
1
k
B
T
2
g
i
i
e
i
/k
B
T
g
i
e
i
/k
B
T
=
1
k
B
T
2
g
i
i
e
i
/k
B
T
Z
Rearrange and eliminate the summed terms in the ex-
pression for u.
Table 4.5: Derivation of the Helmholtz free energy (Eq. 4.35)
S = k
B
ln
= k
B
i
n
i
ln
_
n
i
N
_
= k
B
i
n
i
_

i
k
B
T
lnZ
_
=
i
n
i
i
T
+k
B
i
n
i
. .
=N
lnZ
=
U
T
+Nk
B
lnZ
F = U TS
= Nk
B
T lnZ
Aside: Stirlings approximation for large n!
lnn! = ln1 + ln2 +... + lnn
=
n
k=1
lnk
n
_
1
lnxdx
= [xlnx x]
n
1
= nlnn n + 1
nlnn n
Starting point: The statistical denition of en-
tropy is
S = k
B
ln
where =
N!
n
1
!n
2
!n
3
! . . .
(Eqs. 4.1 and 4.2)
Using Stirlings theorem (see derivation below),
and N =

i
n
i
:
ln = lnN!
i
lnn
i
!
N lnN N
i
(n
i
lnn
i
n
i
)
=

i
n
i
lnN N
i
n
i
lnn
i
+
i
n
i
. .
=N
=

i
n
i
(lnN lnn
i
)
=
i
n
i
ln
_
n
i
N
_
Use this in the expression for entropy, along with
n
i
=
N
Z
e
i
/k
B
T
degeneracy??
and U =

i
n
i
i
The Helmholtz free energy is dened as
F = U TS
Using the expression for entropy derived above,
we nd an expression for F in terms of the natural
logarithm of the partition function.
Table 4.6: Worked example: thermodynamic properties of an ideal monatomic gas from
its partition function
Ideal monatomic gas partition function:
The density of states in 3D is g()d =
d,
where the constant is
=
V
2m
3
2
2
3
(Eqs. 4.10 and 4.17). What do I do about vol-
ume? does it go in constant?
Therefore, the partition function is:
Z =
_
0
g () e
/k
B
T
d
=
_
0
e
/k
B
T
d
=
_
(k
B
T)
3
2
= V
2m
3
2
2
3
_
(k
B
T)
3
2
= V
_
mk
B
T
2
2
_
3/2
_
0
xe
x/a
dx =
a
3
2
Thermodynamic properties from Z:
Using the partition function, we recover the
same expressions for the macroscopic properties
of ideal gases as those from classical thermody-
namics and kinetic theory:
1. Internal energy: Let Z(T) = CT
3/2
, where
C is a constant (C = V
_
mk
B
/2
2
_
3/2
)
U = Nk
B
T
2

T
lnZ
= Nk
B
T
2

T
lnCT
3/2
=
3
2
Nk
B
T
=
3
2
nRT
2. Heat capacity:
C
V
=
_
U
T
_
V
=
3
2
R per mole
3. Ideal gas law: Let Z(V ) = DV , where D is
a constant (D =
_
mk
B
T/2
2
_
3/2
)
F = Nk
B
T lnZ
p =
_
F
V
_
T
=
_

V
_
T
Nk
B
T lnDV
=
Nk
B
T
V
4.5. INDISTINGUISHABLE PARTICLES 79
4.4.4 Agreement with kinetic theory
Even though kinetic theory makes no mention of
the density of states or the partition function, we
can recover the Maxwell-Boltzmann speed distri-
bution and the equipartition theorem using a sta-
tistical mechanics approach.
Maxwell-Boltzmann distribution
The Maxwell-Boltzmann speed distribution
(given in Eq. 1.12) can be derived directly from
the number density distribution, n()d, using
the 3D density of states and the monatomic ideal
gas partition function per unit volume, as shown
in worked example 4.7.
Equipartition theorem
In kinetic theory, we found each degree of free-
dom associated a quadratic term contributes an
average energy of
E) =
1
2
k
B
T
per molecule, towards the total internal energy of
the system. This statement is called the equipar-
tition theorem (see Section 1.4.1). The same re-
sult is derived directly from the Boltzmann dis-
tribution, as shown in Derivation 4.8.
4.5 Indistinguishable particles
So far, we have only been looking at systems
with distinguishable particles. From a quantum
description, distinguishable particles have non-
overlapping wavefunctions (their inter-particle
separation is much greater than their de Broglie
wavelengths), and can in theory be individu-
ally distinguished by a physical measurement.
Such systems of particles correspond to those in
the classical regime, and their properties are de-
scribed by Boltzmann statistics. Now we will
turn to look at the statistics of indistinguishable
particles. These particles cannot be individu-
ally distinguished by any physical measurement.
Indistinguishability is a quantum phenomenon,
and has no classical analogue. Indistinguishable
particles can be described as having overlapping
wavefunctions; the combined wavefunction must
therefore account for all particles within the same
potential well. Multi-particle wavefunctions have
an intrinsic symmetry called exchange symmetry,
which is described below.
4.5.1 Exchange symmetry
Indistinguishable particles cannot be distin-
guished by any physical measurement. There-
fore their combined wavefunction squared can-
not depend on which particle is in which state.
Consider a two-particle wavefunction, denoted by
(1, 2), which indicates that particle 1 is in state
a and particle 2 in state b. The exchange princi-
ple states that the square of the wavefunction of
two indistinguishable particles is the same if we
exchange their states (Eq. 4.41).
Exchange principle:
[(1, 2)[
2
= [(2, 1)[
2
(4.41)
Exchange symmetry is an operation that de-
scribes how a wavefunction changes upon parti-
cle exchange, under the requirement that the ex-
change principle holds. It is represented by the
exchange operator,

P
12
, which, when acting on a
two-particle wavefunction, exchanges the states
of the two particles (Eq. 4.42). Therefore, by
operating twice on the same wavefunction, we
can determine the eigenvalue equation for

P
12
,
and hence nd its eigenvalues (Eqs. 4.43 and 4.44,
Derivation 4.9). In doing this, we nd that there
are two eigenvalues: p = +1 and p = 1. The
eigenvalue p = +1 represents a symmetric wave-
function on particle exchange and p = 1 repre-
sents an antisymmetric wavefunction (Eq. 4.44).
Table 4.7: Worked example: recovering the Maxwell-Boltzmann distribution
1. Use the normalised Boltzmann distri-
bution and the 3D density of states
in the number density distribution
(Eqs. 4.30 and 4.10 into Eq. 4.28).
2. From worked example 4.6, the density
of states constant, , and the parti-
tion function for a monatomic ideal
gas are
=
V
2m
3
2
2
3
Z = V
_
mk
B
T
2
2
_
3/2
3. Change variables to determine the
speed distribution:
=
1
2
mv
2
d = mvdv
4. We recover the Maxwell-Boltzmann
speed distribution from Eq. 1.12. In
this case the distribution is nor-
malised to the total number density,
n
T
, rather than 1.
n()d = f()g()d
=
n
T
Z
g()e
/k
B
T
d
=
n
T
Z

e
/k
B
T
d
= n
T
_
2
2
mk
B
T
_
3/2
_
2m
3
2
2
3
_
e
/k
B
T
d
= 2n
T
_
1
(k
B
T)
3
_
1/2
e
/k
B
T
d
n(v)dv = 4n
T
v
2
_
m
2k
B
T
_
3/2
e
mv
2
/2k
B
T
dv
4.5. INDISTINGUISHABLE PARTICLES 81
Table 4.8: Derivation of the equipartition of energy (check again)
E) =
Ef()d
f()d
=
A
Ee
(E(p)+(q
i
))/k
B
T
dp . . . dq
i
A
e
(E(p)+(q
i
))/k
B
T
dp . . . dq
i
=
Ee
E/k
B
T
dp
e
E/k
B
T
dp
=

ln
_
_
e
E
dp
_
_
=

ln
_
_
e
bp
2
dp
_
_
=

ln
_
_
1
e
by
2
dy
_
_
=

ln
_
1
ln
_
_
e
by
2
dy
_
_
=
1
2
=
k
B
T
2
Equipartition theorem
Starting point: Assume that the total en-
ergy can be split additively into the form
= E(p) + (q
i
), where E is a quadratic
function in a variable p (E = bp
2
, where b
is a constant), and the variables q
i
in (q
i
)
are not quadratic functions. For example,
energy is a quadratic function in momen-
tum or velocity: E = mv
2
/2 = p
2
/2m.
The average energy of E is found by inte-
grating over the normalised Boltzmann dis-
tribution:
E) =
Ef()d
f()d
where f() = Ae
/k
B
T
By splitting the integral into its E depen-
dence and its (q
i
) dependence, the latter
cancels in the numerator and denominator.
Let = 1/k
B
T to simplify the expression.
Change variables:
y
2
= p
2
dy =
_
dp
and eliminate p and dp
The exchange operator:
(Derivation 4.9)
Exchange operator:
P
12
(1, 2) = (2, 1) (4.42)
Eigenvalue equation:
P
12
(1, 2) = p (1, 2) (4.43)
Eigenvalues: write this better?
p =
_
_
+1 : (2, 1) = +(1, 2)
symmetric wavefunction
1 : (2, 1) = (1, 2)
antisymmetric wavefunction
(4.44)
4.5.2 Bosons and fermions
Due to the requirement that wavefunctions are
either symmetric or antisymmetric upon particle
exchange, two dierent types of particles exist in
nature: bosons and fermions. Bosons are sym-
metric on particle exchange whereas fermions are
antisymmetric. It is also found that bosons have
integer spin whereas fermions have half-integer
spin. Therefore, electrons, which are spin-1/2
particles, are fermions, whereas photons, which
are spin-1 particles, are bosons.
The Pauli exclusion principle
Consider a particle in state a and another in state
b. The normalised symmetric and antisymmetric
wavefunctions are: Need derivation?
Symmetric:
S
(1, 2) =
1
2
(
a
(1)
b
(2) +
a
(2)
b
(1))
= +
S
(2, 1)
Antisymmetric:
A
(1, 2) =
1
2
(
a
(1)
b
(2)
a
(2)
b
(1))
=
A
(2, 1)
If we try to put both particles in the same state,
a, then the symmetric wavefunction, for bosons,
is:
S
(1, 2) =
2
a
(1)
a
(2)
whereas the antisymmetric wavefunction, for
fermions, vanishes:
A
(1, 2) = 0
This means that no two fermions are allowed in
the same quantum state. This is known as the
Pauli exclusion principle, and it leads to very
dierent behaviour for systems of bosons and
fermions in the quantum regime, as outlined be-
low.
4.6 Quantum statistics
4.6.1 Grand canonical partition func-
tion
In Section 4.4, we found that the partition func-
tion, Z, for a classical system of states s with no
degeneracy is
Z =

s
e
E
s
/k
B
T
This expression was derived using the Boltzmann
factor, which in turn treats particles as a collec-
tion of closed systems: the particles can exchange
energy with other particles (their surroundings)
but cannot exchange matter. Therefore, Z refers
to the canonical partition function (refer to Sec-
tion 4.1 for classications of statistical ensem-
bles). In the quantum regime, however, we can
no longer treat particles as closed systems since
we are now dealing with multi-particle wavefunc-
tions, in which the number of particles is not con-
stant; instead, we treat particles as a collection
of open systems, which can exchange both en-
ergy and matter with their surroundings. The
partition function in this case is called the grand
canonical partition function, . We can use a sim-
ilar treatment to that described in in the deriva-
tion for the Boltzmann factor (Section 4.4.1), to
nd an expression for .
Consider an open sub-system in thermal con-
tact with a heat reservoir, as illustrated in
Fig. 4.8. Together, the sub-system plus the reser-
voir form an isolated system, with total energy
U
0
and N
0
particles. The energy of the reser-
voir and sub-system are U
R
and U
S
, respectively,
4.6. QUANTUM STATISTICS 83
Table 4.9: Derivation of exchange operator eigenvalues (Eq. 4.44)
P
12
(1, 2) = (2, 1)
and

P
12
(1, 2) = p(1, 2)
(2, 1) = p(1, 2)
P
12
(2, 1) = (1, 2)
and

P
12
(2, 1) = p
P
12
(1, 2)
= p
2
(1, 2)
(1, 2) = p
2
(1, 2)
p
2
= 1
p = 1
Starting point: Acting on (1, 2) with the
exchange operator exchanges the states of
the two particles:
P
12
(1, 2) = (2, 1)
The eigenvalue equation for

P
12
is
P
12
(1, 2) = p(1, 2)
Equate these expressions, and repeat with
(2, 1).
where U
0
= U
R
+ U
S
; the number of particles
in the reservoir and sub-system are N
R
and N
S
,
respectively, where N
0
= N
R
+N
S
.
E
S
,N
S
U
0
=U
R
+E
S
N
0
=N
R
+N
S
U
R,
N
R
Reservoir
Sub-system
Figure 4.8: Isolated system comprising an open
sub-system with energy E
S
and N
S
particles, in
thermal contact with a heat reservoir at energy
U
R
, with N
R
particles. Energy and matter can
be exchanged between the reservoir and the sub-
system.
The exact dierential of U for an open system
is
dU = TdS pdV +dN
where is the chemical potential, which repre-
sents the change in internal energy per particle at
constant entropy and volume (Eq. 2.36). There-
fore, assuming the reservoir has a constant vol-
ume, the entropy of the reservoir is
dS
R
=
dU
R
T

dN
R
T
S
R
=
1
T
(U
R
N
R
)
Letting U
R
= U
0
E
S
, N
R
= N
0
N
S
, and
S
0
= (U
0
N
0
)/T, we nd
S
R
= S
0
1
T
(E
S
N
S
)
The statistical denition of entropy (Eq. 4.2)
states that
S
R
= k
B
ln
R

R
= e
S
R
/k
B
From the second postulate of statistical mechan-
ics, the probability of the reservoir being in a
given macrostate is proportional to the number of
microstates in that macrostate: P . There-
fore, using the expression for S
R
above, the prob-
ability of nding the sub-system in a state with
energy E
S
and N
S
particles is
P(E
S
, N
S
)
R
(E
S
, N
S
)
e
S
R
(E
S
,N
S
)/k
B
e
(E
S
N
S
)/k
B
T
(4.45)
Summing over all states, s, and all particles, N,
the grand canonical partition function is dened
as
=

s
N
e
(E
S(N)
N
S
)/k
B
T
(4.46)
We will now use this partition function to derive
the occupation functionsthe Fermi-Dirac and
Bose-Einstein distributionsfor quantum sys-
tems.
4.6.2 Quantum occupation functions
Consider a sub-system in a state S, with n parti-
cles at an energy level . The energy and number
of particles of the sub-system are:
E
S
= n and N
S
= n
Therefore, from Eq. 4.45, the probability that the
sub-system is in state S is
P(E
S
, N
S
) e
n()/k
B
T
and the average number of particles per state, as
a function of energy , is therefore
n) =
n
nP
n
P
=
n
ne
n()/k
B
T
n
e
n()/k
B
T
n
nx
n
n
x
n
(4.47)
where x = e
()/k
B
T
Notice that the denominator resembles the grand
canonical partition function, , from Eq. 4.46,
with E
S(N)
= n and N
S
= n. Using Eq. 4.47
along with the Pauli exclusion principle, we can
determine expressions for the Fermi-Dirac and
Bose-Einstein occupation functions, which are
given in Eqs. 4.48 and 4.49 respectively (Deriva-
tions 4.10 and 4.11).
Quantum occupation functions:
Fermi-Dirac distribution (for fermions):
f
fd
() =
1
e
()/k
B
T
+ 1
(4.48)
Bose-Einstein distribution (for bosons):
f
be
() =
1
e
()/k
B
T
1
(4.49)
Eqs. 4.48 and 4.49 are plotted in Fig. ??, for
dierent values of T and . We see that decreas-
ing T changes the curvature of the plots, such as
to put more particles into the lower energy levels.
By decreasing , the shape of the function re-
mains the same, but it is shifted to lower energy
levels.
The Fermi energy
As T 0 K, fermions fall into the lowest en-
ergy levels. Due to the Pauli exclusion princi-
ple, fermions stack up, such that the highest-
energy particles have an energy . For spin-1/2
fermions, such as electrons, each energy level ac-
commodates two particles: one in the spin-up
state and one in the spin-down state (Fig. 4.9).
At T = 0 K, the highest energy level is called the
Fermi temperature, E
F
. This parameter depends
on the particle number density, n, and mass m
(Eq. 4.50 and Derivation 4.12).
The Fermi energy: (Derivation 4.12)
E
F
=

2
2m
_
3
2
n
_
2/3
(4.50)
Bose-Einstein condensation
As T 0 K for a system of bosons, we get very
dierent behaviour to a system of fermions, since
bosons do not obey the Pauli exclusion princi-
ple. From the Bose-Einstein curve in Fig. ??, we
see that at very low temperatures, if there were
no interactions between the bosons, all particles
in the system would fall into the ground state.
Table 4.10: Derivation of the Fermi-Dirac distribution (Eq. 4.48)
n) =
1
n=0
nx
n
1
n=0
x
n
=
0 +x
1 +x
=
1
x
1
+ 1
=
1
e
()/k
B
T
+ 1
Starting point: Use the denition of n)
from Eq. 4.47, where
x = e
()/k
B
T
From the Pauli exclusion principle, no two
fermions are allowed in the same quantum
state, therefore n = 0 or 1.
Table 4.11: Derivation of the Bose-Einstein distribution (Eq. 4.49)
n) =
n=0
nx
n
n=0
x
n
=
0 +x + 2x
2
+ 3x
3
+...
1 +x +x
2
+x
3
+...
=
x
_
1 + 2x + 3x
2
+...
_
1 +x +x
2
+x
3
+...
=
x
d
dx
(1 x)
1
(1 x)
1
=
x
1 x
=
1
x
1
1
=
1
e
()/k
B
T
1
Starting point: Use the denition of n)
from Eq. 4.47, where
x = e
()/k
B
T
Bosons do not obey the Pauli exclusion
principle, therefore n can take any integer
number from n = 0 to .
Use the power series for (1 x)
1
:
1
1 x
= 1 +x +x
2
+x
3
+...
Table 4.12: Derivation of the Fermi energy (Eq. 4.50)
n = 2
E
F
_
0
g()f()d
= 2
E
F
_
0
d
= 2
2m
3
2
2
3
2
3
(E
F
)
3/2
E
F
=

2
2m
_
3
2
n
_
2/3
Starting point: Use the number density distribution
(Eq. 4.28) to determine the total number density of par-
ticles. There are two electrons allowed per spatial state,
therefore the spin-degeneracy is g = 2. The integral
limits are between zero and the Fermi energy, E
F
.
At T = 0 K, the Fermi-Dirac occupation function is
f() =
_
1 for < E
F
0 for > E
F
The density of states in 3D is g() =
, where the
constant is
=
2m
3
2
2
3
Integrate and rearrange to solve for E
F
.
1 0
f()
=E
F
Figure 4.9: System of fermions at T = 0 K. The
highest occupied energy level is the Fermi energy,
E
F
. All states below the Fermi energy are occu-
pied, and all states above it are vacant.
This phenomenon is called Bose-Einstein conden-
sation. Since all the particles are in a single quan-
tum state, then quantum eects become apparent
on macroscopic scales. For example, it is thought
that the superuidity of
4
He at temperatures be-
low 2 K is related to Bose-Einstein condensa-
tion. At these temperatures, liquid helium has
zero viscosity, which means it can ow through
tiny capillaries without apparent friction. For ex-
ample, superuid helium could ow through mi-
croscopic holes in an apparently leak-tight con-
tainer.
4.6.3 The classical approximation
In the approximation that k
B
T, then the
exponential term in Eqs. 4.48 and 4.49 dominates:
e
()/k
B
T
1

In this approximation, both the Fermi-Dirac and
the Bose-Einstein distributions approximate to
the classical Boltzmann distribution:
f( ) =
1
e
()/k
B
T
1
1
e
()/k
B
T
Ae
/k
B
T
Written like this, the normalisation constant, A,
is
A = e
/k
B
T
Using the condition that in the Boltz-
mann approximation, we nd that the classi-
cal approximation corresponds to systems with
a high temperature, T, and a small number den-
sity, n (Eq. 4.52 and Derivation 4.13).
Classical approximation: (Derivation 4.13)
(4.51)
and
n
T
3/2

_
mk
B
2
2
_
3/2
(4.52)
high T & low n
Therefore, classical systems are hot and
sparsely populated, whereas quantum systems
are cooler and densely populated. In the clas-
sical approximation, the number of states avail-
able between an energy and +d is far greater
than the number of particles within that energy
range, which means that particles do not compete
to be in the same state. Therefore bosons and
fermions behave in the same way under classical
conditions, since the Pauli exclusion principle is
not in eect. A postulate of classical statistics
is that there is no upper limit to the number of
particles that can occupy a given energy interval,
d.
The de Broglie wavelength
For classical systems, we can use the equipartition
theorem to equate the average kinetic energy of
a particle with k
B
T:
p
2
2m
k
B
T
The de Broglie wavelength of the particle is de-
ned as
db
=
h
p
Eliminating p using the equipartition theorem,
the de Broglie wavelength for classical particles
is
db

_
h
2
2mk
B
T
_
1/2
Comparing this expression with the classical ap-
proximation inequality in Eq. 4.52, we nd
db

_
h
2
2mk
B
T
_
1/2
1
n
1/3
_
V
N
_
1/3
d (4.53)
where I have dened the average distance between
particles as:
d
_
V
N
_
1/3
Therefore, Eq. 4.53 shows that the de Broglie
wavelength of classical particles is much less than
their interatomic spacing, which means that their
wavefunctions do not overlap, and therefore clas-
sical particles are distinguishable.
The transition temperature
The transition temperature of a system is the
temperature below which we must use quantum
statistics, and above which we can use the clas-
sical approximation. For fermions, this tempera-
ture is called the Fermi temperature T
F
, and for
bosons, the Bose temperature, T
B
. The transition
between classical and quantum statistics occurs
when the de Broglie wavelength of the particles
is comparable with their interatomic spacing:
db
d
Therefore, from Derivation 4.13, this corresponds
to:
e
/k
B
T
1
n
_
2
2
mk
B
T
_
3/2
1
Rearranging the above equation, we can nd
an order of magnitude for the quantum-classical
transition temperature, T
QC
(Eq. 4.54).
Table 4.13: Derivation of conditions for classical statistics (Eq. 4.52)
n =
_
0
g()f()d
=
_
0
g ()
e
()/k
B
T
d
= e
/k
B
T
_
0
e
/k
B
T
d
=
2m
3
2
2
2
(k
B
T)
3/2
e
/k
B
T
=
_
mk
B
T
2
2
_
3/2
e
/k
B
T
e
/k
B
T
= n
_
2
2
mk
B
T
_
3/2
For classical systems:
e
/k
B
T
1
n
_
2
2
mk
B
T
_
3/2
1
high T & low n
Starting point: Use the Boltzmann distri-
bution to determine the total number den-
sity of particles in a classical system
The density of states in 3D is g() =
,
where
=
2m
3
2
2
3
_
0
xe
x/a
=
2
a
3/2
For classical systems we have:
e
()/k
B
T
1
e
/k
B
T
e
/k
B
T
1
e
/k
B
T
1 for all
Therefore, eliminating the exponential fac-
tor from the above expression, we nd that
classical systems correspond to high T and
low n, which means hot, sparsely populated
systems.
Transition temperature:
T
QC

2
2
mk
B
n
2/3
(4.54)
The factor of 2 varies slightly for T
F
and T
B
,
but Eq. 4.54 gives a good approximation. For ex-
ample, from the denition of the Fermi energy in
Eq. 4.50, the Fermi transition temperature is:
T
F
=
E
F
k
B
=

2
2mk
B
_
3
2
n
_
2/3
Classical temperatures
From the equipartition theory, the average energy
of a particle in a classical system is of order k
B
T:
) k
B
T
This does not hold for quantum systems. There-
fore we can start using the classical approxima-
tion when, for a system at temperature T, the
energy of the particle is of order k
B
T. This oc-
curs when T T
QC
.
Need example: fermi and bose heat capacity
4.6.4 Black body radiation
We will now turn our attention to the proper-
ties of photons (electromagnetic radiation). Pho-
tons have integer spin, and are therefore bosons.
Electromagnetic radiation incident on a body
is partially reected, partially transmitted, and
partially absorbed (Fig. ??). These quantities
are characterised by dimensionless quantities
the reectivity , the transmissivity , and the
absorptivity , respectivelywhich represent the
fraction of the incident radiation that is reected,
transmitted or absorbed, and have values be-
tween zero and one. For radiation incident on
any body, the sum of the these fractions equals
unity:
+ + = 1
A black body is an object that absorbs all ra-
diation incident on it, therefore
bb
= 1 and
bb
=
bb
= 0.
In addition to reection, absorption and trans-
mission, any object at a non-zero temperature
emits electromagnetic radiation. Emission is
quantied by the emissivity, e, of the object,
which is another dimensionless quantity, where
0 < e < 1. Below, we will look at four laws
Kirchos law, Plancks radiation law, Wiens dis-
placement law and the Stefan-Boltzmann law
which characterise the emission spectrum of a
black body.
Kirchos law
Kirchos law states that, for bodies in ther-
mal equilibrium with their surroundings, their
emissivity equals their absorptivity, for all wave-
lengths (Eq. 4.55). Since black bodies have = 1,
then, from Kirchos law, e = 1. Therefore, black
bodies are often dened as perfect absorbers and
emitters of electromagnetic radiation. A black
body can be made from an opaque container of a
cavity, with a small hole in the container through
which radiation can enter and leave the cavity.
Radiation entering the cavity has a negligible
chance of escaping through the small hole again,
and therefore reects o the cavity walls multi-
ple times, until it is absorbed. Therefore, the
cavity absorbs all incident radiation, which, from
Kirchos law, means the emissivity of the cav-
ity is e = 1. Therefore, this object constitutes a
black body, and the radiation leaving the cavity
through the small hole is black body radiation.
The frequency spectrum of radiation emitted de-
pends on the temperature of the black body, and
is described by Plancks radiation law.
Kirchos law:
e() = () (4.55)
Plancks radiation law
Plancks radiation law gives the frequency spec-
trum of radiation emitted from a black body at
temperature T. We can determine the emission
spectrum of a black body by applying the Bose-
Einstein distribution to a gas of photons. Treat-
ing a black body as a cavity in an opaque con-
tainer, as described above, we can imagine that
photons are constantly emitted and absorbed by
the container walls, which means that the num-
ber of photons within the cavity is not constant.
We nd that this condition means that the chem-
ical potential of the system is = 0. Do I have
to justify this further? Using this requirement in
the Bose-Einstein occupation function (Eq. 4.49),
the Bose-Einstein distribution for black body ra-
diation is:
f
be
() =
1
e
/k
B
T
1
The energy of a photon of frequency is = .
Therefore, the Bose-Einstein occupation function
in terms of is:
f
be
() =
1
e
/k
B
T
1
(4.56)
Plancks radiation law is then derived using
this occupation function, as shown in Deriva-
tion 4.14. It is usually written in terms
of the energy density per frequency interval,
u() [Jm
3
Hz
1
] (Eq. 4.57) or per wavelength
interval u() [Jm
3
nm
1
] (Eq. 4.58). Equa-
tion 4.58 is plotted in Fig. 4.10 for black bod-
ies of dierent temperatures. We see that with
decreasing temperature (i) the wavelength cor-
responding to maximum emission (the turning
point) increases, and (ii) the area under the
curve decreases (indicating that there are fewer
photons emitted). These two characteristics are
described by Wiens displacement law and the
Stefan-Boltzmann law, respectively.
Plancks radiation law:
(Derivation 4.14)
u()d =
_
8h
c
3
_

3
d
e
h/k
B
T
1
(4.57)
or u()d =
_
8hc
5
_
d
e
hc/k
B
T
1
(4.58)
Wiens displacement law
From Fig. 4.10, the emission spectrum has a turn-
ing point whose wavelength increases with de-
creasing temperature. By dierentiating Plancks
law, we nd that the wavelength of maximum
emission,
max
, is inversely proportional to tem-
perature of the black body. This statement is
known as Wiens displacement law, and is given
in Eq. 4.60 (Derivation 4.15).
Wiens displacement law:
(Derivation 4.15)
T =
2.898 10
3
max
(4.60)
The colour of a black body indicates the maxi-
mum emitted wavelength. Therefore, Wiens law
implies that a black bodys colour is a function of
its temperature. To a good approximation, stars
are black bodies, and their colour indicates their
surface temperature. For example, white dwarfs
are relatively hot stars, with temperatures up to
10 000 K, whereas red giants are much cooler, and
have temperatures around 3 000 K. The temper-
ature of the sun is T 6 000 K, which provides
peak emission in the visible waveband.
Stefan-Boltzmann law
The Stefan-Boltzmann law describes the total
power ux emitted by a black body. This quan-
tity is represented by the area under the curve
described by Plancks radiation law (refer to
Fig. 4.10). Therefore, by integrating Plancks law
(Eq. 4.57) over all frequencies, we nd that the
power ux of a black body, J [Wm
2
], is pro-
portional to the fourth power of its tempera-
ture, T
4
(Eq. 4.61 and Derivation 4.16). The con-
stant of proportionality in Eq. 4.61 is the Stefan-
Boltzmann constant, [5.67 10
8
Wm
2
K
4
],
and is given by Eq. 4.62.
Stefan-Boltzmann law:
(Derivation 4.16)
J = T
4
(4.61)
where =
2
5
k
4
B
15h
3
c
2
(4.62)
Table 4.14: Derivation of Plancks radiation law (Eqs. 4.57 and 4.58)
u()d = n()d
= g()f
be
()d
=
g()d
e
/k
B
T
1
=
_

2
c
3
_

3
d
e
/k
B
T
1
u()d =
_
8h
c
3
_

3
d
e
h/k
B
T
1
or u() d =
8hc
5
d
e
hc/k
B
T
1
Starting point: The energy density distribution is the
photon number density distribution, n()d, multiplied
by the energy per photon,
Use Eq. 4.28 to write the number density in terms of
the density of states and the Bose-Einstein occupation
function for photons (Eq. 4.56).
For photons, the dispersion relation is (k) = ck.
Therefore, using this in the density of states in k-space
in 3D (Eq. 4.3), the photon density of states in -space
is
g()d = 2

2
2
2
c
3
d (4.59)
The factor of 2 accounts for the photon polarization
degeneracy (one transverse state and one longitudinal
state for each state in k-space).
Plancks law is often written in terms of the photon
frequency , where = /2, or the photon wavelength
, where = c/.
Table 4.15: Derivation of Wiens law (Eq. 4.60)
u()d =
_
8hc
5
_
d
e
hc/k
B
T
1
u()
=
_
8hc
5
_
hce
hc/k
B
T
2
k
B
T
_
1
_
e
hc/k
B
T
1
_
2
40hc
6
1
e
hc/k
B
T
1
_
= 0
5
_
e
hc/k
B
T
1
_
=
_
hc
k
B
T
_
e
hc/k
B
T
5
_
e
/
1
_
=

e
/
_
5

_
= 5e
/
T
hc
4.965 k
B
1
max
0.0029
max
2.9 10
3
max
Starting point: The turning point of
Plancks radiation law corresponds to
u()
= 0
Dierentiating Plancks law (Eq. 4.58)
and setting the derivative equal to zero,
we nd an expression for the variable
/, where
=
hc
k
B
T
We can solve for / graphically, by
plotting the functions 5 / and
5e
/
on the same axes and deter-
mining their point of intersection. The
curves intersect when:
max
4.965
Eliminating from the above equation
and solving for T, we recover Wiens law.
Table 4.16: Derivation of the Stefan-Boltzmann law (Eq. 4.61)
U =
_
0
u() d
=
8h
c
3
_
0
3
e
h/k
B
T
1
d
=
8h
15c
3
_
k
B
T
h
_
4
=
8
5
15h
3
c
3
(k
B
T)
4
J =
1
4
cU
=
2
5
k
4
B
15h
3
c
2
. .
=
T
4
T
4
Starting point: The total energy density, U, is found
by integrating Plancks radiation law (Eq. 4.57) over all
photon frequencies, .
_
0
x
3
e
ax
1
dx =
1
a
4
_
4
15
_
From kinetic theory, we found that the ux density of
particles with average velocity c is:
=
1
4
nc
(see Section 1.4.4, Eq. 1.32). The total power ux is
then given by
J =

n
U =
1
4
cU
The Stefan-Boltzmann constant, , is dened as
=
2
5
k
4
B
15h
3
c
2
Weins law
Area: Stefan-Boltzmann law
Figure 4.10: Emission spectrum of a black body.
Plancks radiation law (Eq. 4.58) gives the inten-
sity of radiation with wavelength; Wiens dis-
placement law describes the position of the max-
ima; Stefan-Boltzmanns law gives the area un-
der the curve (total power per unit area). As
the temperature of the black body decreases, the
wavelength of maximum emission increases, and
the area under the curve (indicating the power
uxor number of emitted photons) decreases.

Thermal Physic A Sample Material

Enviado por

Dados do documento

Título original

Direitos autorais

Formatos disponíveis

Compartilhar este documento

Compartilhar ou incorporar documento

Opções de compartilhamento

Você considera este documento útil?

Este conteúdo é inapropriado?

Direitos autorais:

Formatos disponíveis

Thermal Physic A Sample Material

Enviado por

Direitos autorais:

Formatos disponíveis

A Students Handbook of Physics

Part III. Thermal Physics

Table 1.2: Derivation of 3D velocity distribution (Eq. 1.8)

2. Therefore, the stan-

To integrate, use the standard integrals:

Table 1.3: Derivation of the speed distribution (Eq. 1.12)

To integrate, use the standard integrals:

Figure 1.5: Maxwell-Boltzmann speed distribu-

is the fraction of those

Cylinder volume = vdt dScos

Starting point: The mean free path, ,

where is the variable that is kept constant while

is the eciency of the irreversible

4.3. THE DENSITY OF STATES 67

The density of states in k-space is then denoted

Figure 4.7: Isolated system comprising a closed

Você também pode gostar