PH-107

Quantum Physics and Applications

Elements of Statistical Physics-II

Gopal Dixit
gdixit@phy.iitb.ac.in
Recap
Number of particles between E and E+dE
Depends on the product g(E)f(E)dE, which gives dN(E).

We have written the number of available energy states

within the interval E and E+dE in terms of the density of
states, g(E).

And the probability of a particle occupying an available

state in the interval E and E+dE is expressed in terms of
the probability distribution function, f(E).
f(E): Dependence on Particle Characteristics
Equilibrium Configuration

The general problem we are trying to solve is, given a system with
1
X
Ni = N
i=1

1
X
Ei N i = E
i=1

What is the configuration {Ni} ≡ (N1, N2, …Ni…) for which the
multiplicity Q({Ni}) is maximum?

For this, we have to first learn how to calculate Q({Ni}).

Classical Particles (M-B)

How difficult is the problem?

Let’s start by assuming that i = 2, with gi = 1 for both states, and N

total number of particles.

i.e., we are considering a system with two (non-degenerate)

energy states (E1 and E2), in which N particles have to be
distributed, such that N1+N2 = N and E1 N1+E2 N2 = E

This problem is similar to distributing N “numbered billiard

balls” in 2 “numbered” containers.
Classical Particles (M-B)

1 2 3 … i … N
1 2
Let us calculate a few Q(N1, N2) values.
What is Q(N1 = 0 , N2 = N )?
Ans: 1 (There’s only one way of putting no ball in container 1 and all
balls in container 2)

What is Q(N1 = 1 , N2 = N-1 )?

Ans: N (There are N ways of putting 1 ball in container 1 and the
rest N-1 in container 2)
Classical Particles (M-B)
What is Q(N1 = 2 , N2 = N-2 )?
✓ ◆
N N (N 1)
Ans: =
2 2!
(There’s N choices for the 1st ball in the 1st container, and for each
of them, N-1 choices for the 2nd ball in the 1st container)

"
# 1 2 3 … i … N
1 2
$ %
But we don’t want to count % and $ as two
different cases. So, we divide by 2!
Classical Particles (M-B)
Likewise, for Q(N1 = 3 , N2 = N-3)
✓ ◆
N N (N 1)(N 2)
Ans: =
3 3!

So, for Q(N1 = N1 , N2 = N-N1)

✓ ◆
N N (N 1)(N 2)....(N N1 + 1) N!
Ans: = =
N1 N1 ! N1 !N2 !
Classical Particles (M-B)
Now we divide the 2nd container into two halves.

1 2 3

So that N2 = ⌫1 + ⌫2

The number of ways of realizing the configuration (⌫1 , ⌫2 ) within

the sub-compartments of the 2nd container is
N2 !
Q(⌫1 , ⌫2 ) =
⌫1 !⌫2 !
Classical Particles (M-B)
But we could also consider this as 3 containers with N1 , ⌫1 , ⌫2 balls,
such that
N! N2 !
Q(N1 , ⌫1 , ⌫2 ) =
N1 !N2 ! ⌫1 !⌫2 !
Relabeling properly we get

N!
Q(N1 , N2 , N3 ) =
N1 !N2 !N3 !

N!
So Q(N1 , N2 , ....Ni ....) = Q({Ni }) = Q
i Ni !

Recall that we did not consider degeneracy of the states i.

Classical Particles (M-B)

If we consider the state i (with energy Ei) to be gi-fold degenerate,

we need to consider a group of gi containers (all labeled i) to
distribute the Ni billiard balls.

Ni balls can be distributed in gi copies of the container labeled i in

(gi)Ni ways. The first of the Ni balls can be placed in any of the gi
containers, and so can be the 2nd, 3rd, 4th, and so on…

Same holds true for distribution of Nj balls in gj copies of the

container labeled j.
Classical Particles (M-B)
So
N! Y Ni
Q(N1 , N2 , ....Ni ....) = Q({Ni }) = Q (gi )
i Ni ! i

N! Y Ni
So, our job now simplifies to maximizing Q (g i )
i Ni ! i
1
X 1
X
obeying the constraints Ni = N and Ei N i = E
i=1 i=1

This is done by a cumbersome mathematical technique called the

method of Lagrangean multipliers.
Classical Particles (M-B)
It gives the condition for Q({Ni}) to be maximized that

Ni
= e↵ e Ei
gi

where α and β are constants to be determined.

Later, we will see that β = (kB T)-1, so that

Ei
Ni = g i e↵ e kB T

This equation represents the Maxwell-Boltzmann distribution

function fMB (Ei).
Classical Particles (M-B)
1
X 1
X
↵ Ei
Ni = N =) gi e e =N
i=1 i=1

↵ N
=) e = P1 Ei
i=1 gi e

N gi e Ei
So Ni = P 1 Ei
g
i=1 i e

Or, the probability of finding the particle in state i

Ni gi e Ei
Pi = = P1 Ei
N i=1 gi e
Classical Particles (M-B)
Ni gi e Ei
With Pi = = P1 Ei
the mean energy can be
N i=1 gi e

calculated as
X P
i g i Ei e Ei
1 X Ei
Ē = Pi E i = P Ei
= g i Ei e
i i g i e Z i

X
Ei
Here, Z = gi e is called the partition function.
i

@
You can verify Ē = (ln Z)
@
Classical Particles (M-B)
Ni gi e Ei
With Pi = = P1 Ei
the mean energy can be
N i=1 gi e

calculated as
X P
i g i Ei e Ei
1 X Ei
Ē = Pi E i = P Ei
= g i Ei e
i i g i e Z i
X
Ei
Here, Z = gi e is called the partition function.
i

For any other parameter

X P Ei
i g i yi e
ȳ = Pi y i = P Ei
i i g i e
Maxwell-Boltzmann Distribution
↵
Let us write e = A, then we can see that
1
X 1
X Ei
N= Ni = gi A e kB T

i=1 i=1

In the continuum limit,

X Z Z
Ni ! dN (E) = g(E) fMB (E) dE

E
fMB (E) = A e kB T

↵ N
with A=e = R E
g(E) e kB T
dE
Maxwell-Boltzmann Distribution
The energy of point particles of an ideal gas is purely translational
kinetic energy, i.e. the energy of each molecule is of the form

1
E = mv 2
2

Since the speeds of the gas particles vary continuously from 0 to ∞,

the energy must also vary continuously from 0 to ∞.

Thus we can use the MB distribution to write

( 12 mv 2 /kB T )
dN (E) = N (E)dE = g(E) A e dE
What is Lagrangean multipliers and why it is needed?
Quantum Distribution Function
Fermi-Dirac Distribution
Q({Ni }) for Fermions
If the particles become indistinguishable (Quantum particles),
the numbering of the billiard balls is gone!

In case of Fermions, the further restriction of occupancy of each

state by only one particle has to be obeyed.

So, unlike for distinguishable particles where

N!
Q(N1 , N2 , ....Ni ....) = Q({Ni }) = Q
i Ni !

For Fermions: Q(N1 , N2 , ..., Ni ..) = Q({Ni }) = 1

However, the state with energy Ei can still have degeneracy gi

 Fermi-Dirac Distribution
General
Q({Ni }) problem (M-B, B-E, F-D)
for Fermions
State State State State T
Index Energy Degeneracy Occupancy Rwe"L = " to keep in mind
But need
(K) (#L ) (%L ) "L LS;
1 #; %; "; that for Fermions, there is a
T
2 #/ %/ "/ further restriction of
R #L "L = #
⋮ ⋮ ⋮ ⋮
occupancy
LS; of each state by
K #L %L "L only one particle
⋮ ⋮ ⋮ ⋮

Here we are generalizing the problem for a larger

How do we
system. fillare
We gi states
lookingoffor
energy
"L ≡Ei "
with Ni Fermions?
; , "/ , … "L …
✓ ◆
gi gi !
=
Ni Ni !(gi Ni )!
Fermi-Dirac Distribution
Q({Ni }) for Fermions

This leads to
Y gi !
Q(N1 , N2 , ...., Ni ...) = Q({Ni }) =
i
Ni !(gi Ni )!

Compared with distinguishable (classical) particles

N! Y Ni
Q(N1 , N2 , ...., Ni ...) = Q({Ni }) = Q (gi )
i Ni ! i

the multiplicity of a particular configuration of Fermions is given by

Y gi !
Q(N1 , N2 , ...., Ni ...) = Q({Ni }) = 1.
i
Ni !(gi Ni )!
Fermi-Dirac Distribution
In this case:
Y gi !
Q(N1 , N2 , ....Ni ...) = Q({Ni }) =
i
Ni !(gi Ni )!

Ni 1
=
gi 1 + e ↵+ Ei

↵ 1
Writing e = A and noting that = (kB T ) , we get
gi
Ni = Ei
1+A e kB T

1
X
Like in M-B statistics, A (or e-α) can be evaluated from Ni = N
i=1
 Fermi-Dirac Distribution

↵ EF /kB T
However, it is customary to write ↵ = EF . So that e =e

In terms of the Fermi energy, Fermi Energy

gi
Ni =
1 + e(Ei EF )/kB T

Again in the continuum limit,

X Z Z
Ni ! dN (E) = g(E)fFD (E)dE
i
1
fFD (E) =
1 + e(E EF )/kB T

This is the Fermi-Dirac distribution function governing the

equilibrium behavior of Fermions (e.g. electrons).
Fermi-Dirac Distribution
Fermi-Dirac Distribution (T=0 and T > 0)

1
fFD (E) =
1 + e(E EF )/kB T
 Bose-Einstein Distribution
Q({Ni }) for Bosons
Bosons are indistinguishable but the restriction of occupancy of
each state by only one particle is gone. A state i may be occupied
by any number of particles.

ooooo|ooooo|…..|ooo
1 2 #$
%$ particles (in the state with energy &$ )

In how many ways can we arrange Ni identical balls and gi-1 partitions?
✓ ◆
Ni + g i 1 (Ni + gi 1)!
=
gi 1 Ni !(gi 1)!
Bose-Einstein Distribution
Q({Ni }) for Bosons

Bosons are indistinguishable but the restriction of occupancy of

each state by only one particle is gone. A state i may be occupied
by any number of particles.

This leads to
Y (Ni + gi 1)!
Q(N1 , N2 , ...., Ni ...) = Q({Ni }) = 1.
i
Ni !((gi 1)!

Maximization of Q({Ni}) finally gives

Ni 1
= ↵+ Ei
gi e 1
Bose-Einstein Distribution
Again in the continuum limit,
X Z Z
Ni ! N (E)dE = g(E) fBE (E)dE
i

1 1
fBE (E) = ↵ eE/kB T
=
e 1 AeE/kB T 1

This is the Bose-Einstein distribution function governing the

equilibrium behavior of Bosons
Bose-Einstein Distribution
1 1
fBE (E) = ↵ eE/kB T
=
e 1 AeE/kB T 1
1
X
Like in previous cases, A (or e-α) can be evaluated from Ni = N
i=1
or in the continuum limit, from
Z
1
N= g(E) dE
A e(E/kB T ) 1

When the restriction on the number of particles is lifted (photons/

phonons), the Bose-Einstein distribution reads
1
fBE (E) =
e(E/kB T ) 1
Also known as the Planckian Equilibrium Distribution
Comparison
Recommended Readings

Statistical Physics, Chapter 10