CHAPTER 5

GASES : the distributions

5.1. Distribution In Groups

Under such conditions, the definition of a distribution in states contains far too much
detail for comfort. Instead we use effectively the distribution in levels, {ni} as defined in
section 1.4.
The point is that it does no violence to the physics to group the states. In the form of the
distribution used for computation the ith group is taken to contain gi states of average energy
𝜀𝑖 . The only difference from the true specification of the states is that in the grouped
distribution all gi states are considered to have the same energy 𝜀𝑖 rather than their correct
individual values. This grouped distribution is illustrated in Fig. 5.1. Let us note several
points before proceeding.
1. The grouping is one of convenience, not of necessity. We are choosing to use large
numbers of states gi taken together and also large numbers of particles n occupying these
states. This will enable us below to use the mathematics of large numbers (e.g. Stirling’s
approximation) in discussing ni and gi in order to work out the distribution

2. The grouping used in the distribution is arbitrary, but not capricious! The rule is that the
number of states remains the same in both sides of Fig. 5.1. Or, more precisely, the density
of states (over a sufficiently coarse energy graining) remains unchanged. If we choose the
energy levels 𝜀𝑖 , then the full microscopic details of the true physical problem determine
the values gi.
 3. A test for whether the grouping makes physical sense is to consider the average I number
of particles per state, often called the distribution function or colloquially the filling factor.
This is defined as
𝑓𝑖 = 𝑛𝑖 /𝑔𝑖
4. Finally, if we return to the example of helium gas (section 4.3), we can see that this I
approach makes some numerical sense. We noted previously that the energy level spacing
between the true states (∆𝜀𝑗 ) was about 10-19 time kBT.
5.2. Identical Particles – Fermions and Bosons
A microstate of course relates to a quantum state of all N particles, and here must come,
therefore, the vital recognition that identical gas particles are fundamentally indistinguishable
from each other. They are, therefore, effectively competing for the same one-particle states.
We can never define which particle is in a particular state, the microstate is completely
specified merely by stating how many particles are in each state.
Consider a system of just two identical indistinguishable particles. It is described in
quantum mechanics by a two-particle wave function ψ(1, 2), in which the label 1 is used to
represent all the co-ordinates (i.e. space and spin co-ordinates) of particle 1 and the label 2
represents the co-ordinates of the other particle. For ψ(l, 2) to be a valid wave function, two
conditions must be satisfied. One is obviously that it should be a valid solution of
Schrödinger’s equation. But the second condition is that it should satisfy the correct symmetry
requirement for interchange of the two labels. The symmetry required, that of ‘interchange
parity’, can be outlined as follows.
If the labels of the two particles are interchanged, then any physical observable cannot
change (since the particles are identical). This implies that
𝜓(1,2) = exp(𝑖𝛿)𝜓(2,1)
since a physical observable always involves the product ψ*ψ. The wave functions are related
by the phase factor exp(i𝛿). But if we make two interchanges, then we come to a mathematical
identity, i.e.
𝜓(1,2) = exp(𝑖𝛿)𝜓(2,1) = exp(2𝑖𝛿)𝜓(1,2)
Hence exp(2i𝛿) = 1, and therefore the interchange factor exp(i𝛿) must equal +1 or -1, one of
the two square roots of +1.
 Particles which interchange co-ordinates with the +1 factor are called ‘bosons’, and
those with the -1 factor are ‘fermions’. We shall use the symbol S to describe the symmetric
wave function of bosons, and A for the antisymmetric function of fermions. It can be taken as
an experimental result, that all particles with integral spin (0, 1, 2, etc.) are bosons; whereas
1 3
odd-half integral spin particles (spin 2 , 2 . 𝑒𝑡𝑐) are fermions.

Next, let us consider the case where the particles are weakly interacting, the relevant
case for our statistical physics. This means that a solution to Schrödinger’s equation for t he
two particles can be written as the product of two one-particle wave functions, i.e.
ψ(1, 2) = ψa(1) · ψb(2) (5.1)
where the labels a and b are two values of the state label j for the one-particle states.
However, as a wave function (5.1) suffers from the defect that it has no interchange parity, it
is neither A nor S. Nevertheless it can be combined with its interchanged partner (an alternative
solution to the Schrödinger equation with the same properties) to yield two wave functions,
one S and one A. These are
1
𝜓𝑆 (1,2) = ( )[𝜓𝑎 (1)𝜓𝑏 (2) + 𝜓𝑎 (2)𝜓𝑏 (1)] (5.2)
√2

1
𝜓𝑆 (1,2) = ( )[𝜓𝑎 (1)𝜓𝑏 (2) − 𝜓𝑎 (2)𝜓𝑏 (1)] (5.3)
√2

Equations (5.2) and (5.3) bring out a fundamental difference between bosons and
fermions. Consider the situation in which the labels a and b are identical, i.e. in which the two
particles are both competing for the same state. For bosons, for which (5.2) applies, there is
no problem. (In fact (5.2) gives an enhanced wave function compared to (5.1).) On the other
hand fermions are quite different. The wave function given by (5.3) vanishes identically when
we set a = b. This is the Pauli exclusion principle, which recognizes that ‘no two identical
fermions can occupy the same state’.
5.3. Counting Microstates For Gases
The microstate (effectively the appropriate wave function either ψS or ψA ) can therefore
be labelled just by the one particle state labels a, b, etc. As pointed out earlier, this is simply
a recognition that we cannot know which particles are in which states; even in principle we
can only know which states are occupied. Therefore the counting of microstates is to specify
the occupation of each one-particle state.
5.3.1. Fermions
Therefore, the one-particle states can only have occupation numbers of 0 or 1.
Incidentally, this implies that in the ith group, the number of particles ni cannot exceed
the number of states gi. The number of different ways the states can be so divided is
𝑔𝑖 !
𝑛𝑖 ! (𝑔𝑖 − 𝑛𝑖 )!
This is the contribution to the number of microstates from the ith group. The total i
number of microstates corresponding to an allowable distribution {ni } is therefore
given by
𝑔!
𝑡𝐹𝐷 ({𝑛𝑖 }) = ∏𝑖 𝑔 !(𝑔 𝑖−𝑛 )! (5.4)
𝑖 𝑖 𝑖

5.3.2. Bosons
A direct, if slightly abstract, way of calculating the contribution from the ith group
of the distribution is as follows. In the group there are gi states containing ni identical
particles with no restrictions on occupation numbers. A typical microstate can be
represented as in Fig. 5.2 by (gi - 1) lines and ni crosses. The lines represent divisions
between the gi states, and the crosses represent an occupying particle. New microstates
representing the same distribution, i.e. the same value of ni for the group, can be
obtained by shuffling the lines and crosses on the picture.

Hence our (slightly approximate) final result for the number of microstates to an
allowed distribution for a boson gas is
(𝑛𝑖 +𝑔𝑖 )!
𝑡𝐵𝐸 ({𝑛𝑖 }) = ∏𝑖 (5.5)
𝑛𝑖 !𝑔𝑖 !
 5.3.3 Dilute Gases
For gases, there are two, and only two, possibilities. The gaseous particles are
either bosons (+1,symmetric) or fermions (-1, antisymmetric). There is no half way
house. However, it is profitable to consider a third form of statistics quite explicitly,
namely that of a dilute gas.
The word ‘dilute’ here has a specific meaning. It is to suppose that for all groups
i, the states are very sparsely occupied, i.e.
𝑛𝑖 ≪ 𝑔𝑖 for all i
We shall see later that this condition holds for real gases in the limit of low density
and high temperature, and that it corresponds to a classical limit to either the FD or
the BE form of quantum statistics.
Therefore in the dilute limit tFD and tBE tend to the same value, one from below
and one from above. This ‘classical’ limit is called the Maxwell–Boltzmann (MB)
case, discussed by these famous scientists before quantum mechanics was conceived.
And we write the answer to our third counting problem as
𝑔𝑖 𝑛𝑖
𝑡𝑀𝐵 = ∏𝑖 (5.6)
𝑛𝑖 !

5.4. The Three Distributions

5.4.1. Fermi-Diract statistics
We require to maximize tFD or more conveniently ln tFD , subject to usual
macrostate condition
∑𝑖 𝑛𝑖 = 𝑁 (5.7)
and
∑𝑖 𝑛𝑖 𝜀𝑖 = 𝑈 (5.8)
The lagrange method writes down for the maximum conition
𝑑(ln 𝑡) + 𝛼𝑑(𝑁) + 𝛽𝑑(𝑈) = 0 (5.9)
As anticipated the grouping of the states has no explicit influence on the
equilibrium occupation per state, and the result may be written in terms of a distribution
function
𝑛𝑖∗
𝑓𝑖 = = 1/[𝑒𝑥𝑝(−𝛼 − 𝛽𝜀𝑖 + 1)] (5.10a)
𝑔𝑖
 The distribution function contains no detail about the states except their energies,
and 𝜀𝑖 so in concord with the whole density of states approximation it can be thought
of as effectively a continuous function of the one-particle energy 𝜀𝑖 or simply 𝜀 .
Hence a useful form of (5.10a) is
𝑓𝐹𝐷 (𝜀) = 1/[𝑒𝑥𝑝(−𝛼 − 𝛽𝜀 + 1)] (5.10b)

This is the Fermi-Diract distribution function.

5.4.2. Bose-Einstein statistics
The Bose-Einstein Distribution is
𝑓𝐹𝐷 (𝜀) = 1/[𝑒𝑥𝑝(−𝛼 − 𝛽𝜀 − 1)] (5.11)
5.4.3. Maxwell-Boltzman statistics
The final result for the Maxwell –Boltzman distribution is therefore
𝑛𝑖∗
𝑓𝑖 = = exp(𝛼 + 𝛽𝜀𝑖 ) (5.12a)
𝑔𝑖

Or
𝑓𝑀𝐵 (𝜀) = exp(𝛼 + 𝛽𝜀) (5.12b)
5.4.4. 𝜶 and 𝜷 revisted
The distribution function, defined by the average number of particles per state of
energy 𝜀𝑖 ,is given for a gaseous assembly in thermal equilibrium at temperature T by