Unit 4, Statistical Physics, Lecture 4

Fermi-Dirac and Bose-Einstein Statistics

The Maxwell-Boltzmann distribution function suffers from two major drawbacks.

1. The particles are assumed to be identifiable although in actual practice electrons or

other elementary particles are indistinguishable.
2. Any number of particles was allowed to occupy the same quantum, state while many
particles, notably electrons, obey Pauli’s exclusion principle which does not allow a
quantum state to accept more than one particle.
These drawbacks are removed in the quantum statistics. There are two types of distribution
functions in quantum statistics, viz. Fermi-Dirac (FD) distribution function and the Bose-
Einstein (BE) distribution function.

FD function gives the statistical behavior of free electrons in metals and in heavily doped
semiconductors. Many electrical and thermal properties of solids which the classical statistics
failed to explain, could be understood on the basis of the FD statistics. The particles that obey
the FD statistics are called fermions. We derive below the Fermi-Dirac distribution function.

Fermi-Dirac (FD) Statistics

Basic features:

1. The particles (fermions) are identical and indistinguishable.

2. They obey Heisenberg’s uncertainty principle and the Pauli’s exclusion principle.
3. There is definite a priori restriction on the number of fermions in a quantum state. There
can be simultaneously no more than one particle in a quantum state (i.e. it would either
remain empty or can at best contain one fermion).

Consider a system of N indistinguishable noninteracting particles obeying the Pauli exclusion

principle. Let N1, N2, …, Nn be the number of particles in the system with energies E1, E2, …,
En, respectively, and let gi represent the multiplicity or the degeneracy of energy level Ei. Since
the particles are all indistinguishable, there is only one of choosing N1, N2, …, Nn particles from
the assembly of N particles.

The number of ways in which Ni particles are arranged in gi quantum states (in other words,
the number of microstates which is having equal a priori probability) is given by
𝑔𝑖 𝑔
𝐶𝑁𝑖 = 𝑁 !(𝑔 𝑖−𝑁
!
)!
(34)
𝑖 𝑖 𝑖

Eqn. 1 results from the fact that the particles are indistinguishable and that each quantum state
can accommodate only one particle in accordance with Pauli’s exclusion principle.

The total number of ways W of distributing N1, N2, …, Nn particles in n energy levels is the
product of the terms given by Eqn. 1 over all the levels, i.e.

1
 𝑛
𝑔𝑖 !
𝑊=∏
𝑁𝑖 ! (𝑔𝑖 − 𝑁𝑖 ) !
𝑖=1

(35)
Where ℿ denotes the product, as ⅀ denotes sum. Taking the logarithm of both sides, we get

𝑙𝑛𝑊 = ∑[𝑙𝑛 𝑔𝑖 ! − 𝑙𝑛 𝑁𝑖 ! − 𝑙𝑛(𝑔𝑖 − 𝑁𝑖 )!]

With Sterling’s formula ln n! = n ln n – n, we can write the above expression as

𝑙𝑛𝑊 = ∑[(𝑔𝑖 𝑙𝑛 𝑔𝑖 − 𝑔𝑖 ) − (𝑁𝑖 𝑙𝑛 𝑁𝑖 − 𝑁𝑖 ) − ((𝑔𝑖 − 𝑁𝑖 ) 𝑙𝑛(𝑔𝑖 − 𝑁𝑖 ) − (𝑔𝑖 − 𝑁𝑖 ))]

= ∑[𝑔𝑖 𝑙𝑛 𝑔𝑖 − 𝑁𝑖 𝑙𝑛 𝑁𝑖 − (𝑔𝑖 − 𝑁𝑖 ) 𝑙𝑛(𝑔𝑖 − 𝑁𝑖 )] (36)

Assuming the system is in equilibrium, to find the most probable distribution, W or ln W must
be maximized subject to the total number of particles N and the total energy E are constants,
i.e.

𝑁 = ∑𝑁
𝑖=1 𝑁𝑖 = 𝑐𝑜𝑛𝑠𝑡𝑎𝑛𝑡 𝑜𝑟 ∑𝑖 𝛿𝑁𝑖 = 0 (37a)

and 𝑈 = ∑𝑁
𝑖−1 𝑁𝑖 𝐸𝑖 = 𝑐𝑜𝑛𝑠𝑡𝑎𝑛𝑡 (37b)

For the distribution to represent maximum probability, small changes δNi in any of the
individual Ni’s must not alter W. Hence, from Eqn. 3 we get
1 1
𝛿𝑙𝑛𝑊 = ∑𝑖[(𝑁𝑖 − 𝑔𝑖 ) (𝑔 −𝑁 ) (−𝛿𝑁𝑖 ) + 𝛿𝑁𝑖 ln(𝑔𝑖 − 𝑁𝑖 ) − 𝑁𝑖 𝑁 𝛿𝑁𝑖 − 𝑙𝑛𝑁𝑖 𝛿𝑁𝑖 ]
𝑖 𝑖 𝑖

= ∑𝑖 [ln(𝑔𝑖 − 𝑁𝑖 ) − 𝑙𝑛𝑁𝑖 ]𝛿𝑁𝑖 = 0 (38)

This expression is subject to the conditions specified in Eqns. 37a and b. To incorporate these
and obtain a general expression for most probable distribution, we use Lagrange’s method of
undetermined multipliers. Therefore, multiplying Eqn. 37a by –α and Eqn. 37b by – β and
adding it to Eq. 38, we get
∑𝑖[ − 𝑙𝑛𝑁𝑖 + ln(𝑔𝑖 − 𝑁𝑖 ) − 𝛼 − 𝛽𝐸𝑖 ]𝛿𝑁𝑖 = 0 (39)

Where – α and – β are Lagrange’s undetermined multipliers.

Since δNi’s are independent, the quantity in brackets of Eqn. 39 must vanish for each value of
i. Thus,
𝑁
𝑖 𝑁𝑖
ln (𝑔 −𝑁 ) = − 𝛼 − 𝛽𝐸𝑖 𝑜𝑟 = exp(− 𝛼 − 𝛽𝐸𝑖 )
𝑖 𝑖 𝑔𝑖 −𝑁𝑖

𝑔𝑖
Or − 1 = exp( 𝛼 + 𝛽𝐸𝑖 )
𝑁𝑖

𝑵𝒊 𝟏
Or = (40)
𝒈𝒊 𝟏+𝒆𝜶+𝜷𝑬𝒊

This is the general form of FD distribution function for an assembly of fermions among the
various energy levels of a system.

2
In the previous section we have determined that β = 1/kT, where k is the Boltzmann constant
and T is the absolute temperature. The quantity α is generally expressed in the form
𝐸
𝛼 = − 𝑘𝑇𝐹

Where EF is the chemical potential or the Fermi energy of the system. The FD distribution
then takes the form
𝑵𝒊 𝟏
= 𝒇(𝑬𝒊 ) = 𝑬𝒊 −𝑬𝑭 (41)
𝒈𝒊
𝟏+𝒆 𝒌𝑻

Since only one particle may occupy a quantum state, by exclusion principle, f(Ei) (referred as
occupation index) for FD statistics gives the probability that a quantum state of energy Ei is
occupied.

It is to be noted that at T = 0 (i.e. 1 /kT = ∞), the exponent of Eqn. 8 becomes - ∞ for Ei < EF,
whereas for Ei > EF, the exponent becomes + ∞. Therefore, at T = 0K, Eqn. 8 can be written
as

𝑓(𝐸𝑖 ) = 1 𝑓𝑜𝑟 𝐸𝑖 < 𝐸𝐹

= 0 𝑓𝑜𝑟 𝐸𝑖 > 𝐸𝐹
This means mathematically f(Ei) is a step function. Also at absolute zero, all levels up to a
certain energy, known as Fermi energy or Fermi level, are occupied but all the energy states
above it are empty. A plot of Fermi Dirac distribution function versus energy at different
temperatures is shown below.

Fig. 8. Fermi function vs. Energy plot at different temperatures.

At a finite temperature (T > 0 K), fermions shift to higher energies. However, the width of the
region is of the order of kT.

It may be noted here that if the energy levels of the system are very close together, then the
distribution function given in Eqn. 41 can be written in the form
𝑔(𝐸)𝑑𝐸
𝑁(𝐸)𝑑𝐸 = 𝐸𝑖 −𝐸𝐹 (42)
1+𝑒 𝑘𝑇

Where N(E)dE represents the number of particles having energies between E and E + dE.