Chapter 2

Elements of Ensemble Theory

In 1902 Gibbs, one of the founders of Statistical Mechanics, introduced the concept of ensemble
(a French word meaning assembly of systems). An ensemble is a collection of essentially inde-
pendent systems, which are macroscopically identical but microscopically different.
Three types of ensemble are important in statistical mechanics. The classification depends on
the matter in which the systems interact.

2.1 MICROCANONICAL ENSEMBLE

The microcanonical ensemble is a collection of independent systems, having the same number
of particles N, volume V and an energy between E and E + δE. So the individual systems of
a microcanonical ensemble are separated by rigid, impermeable and insulated walls, such that the
values of E,V and N for a particular system are not affected by the presence of other systems. Thus,
in microcanonical ensemble, neither energy nor matter is exchanged.

N,V, E N,V, E

N,V, E N,V, E

Fig. 2.1. Microcanonical ensemble.

57
60 • An Introduction to Equilibrium Statistical Mechanics

Now
N 3
1 N → −p 2 = 1
E= ∑
2m i=1 i ∑ ∑ p2iα
2m i=1 α=1
N 3
⇒ 2mE = ∑ ∑ p2iα (2.1.9)
i=1 α=1

where →−p 2 = p2 + p2 + p2 , denoting x, y, z components by 1, 2, 3 respectively.

i i1 i2 i3
Obviously, the sum contains 3N = f square terms ( f = degrees of freedom). For E = constant,
1
it describes a sphere of radius ς (E) = (2mE) 2 in the f -dimensional space of the momentum coor-
dinates. Hence Ω (E) is proportional to the volume of phase space contained in the spherical shell
lying between the sphere of radius ς (E) and that of slightly larger radius ς (E + δE).
We now calculate the volume of the N-dimensional sphere. From the dimensional analysis, the
volume of a N-dimensional sphere is
WN = CN ςN (2.1.10)
where CN is a constant and is given by
N
π2
CN =  N  (2.1.11)
Γ 2 +1
(see Appendix II).
Using (2.1.11) in (2.1.10), the volume of the N-dimensional sphere becomes
N
π2
WN = N  ςN (2.1.12)
Γ 2 +1
√
Hence, the volume of the 3N-dimensional sphere of radius 2mE is
3N
π2 3N
W3N (E) =  3N  (2mE) 2 (2.1.13)
Γ 2 +1
V N ∂WN (E)
∴ Ω (E) = δE
h3N
0 ∂E
 3N
2mπ 2 V N

3N
⇒ Ω (E) = 2
 3N  E 2 δE (2.1.14)
h0 Γ 2
Neglecting 1 compared to N
 3N2
VN

2mπ 3N
Ω (E) = ΩE δE =  3N  E 2 δE (2.1.15)
h20 Γ 2
Since 3N = f , we see that
Ω (E) ∼ E f (2.1.16)
 Elements of Ensemble Theory • 61

N is of the order of Avogadro’s number and so very large. Thus, Ω (E) is an extremely rapidly
increasing function of the energy E of the system.
The entropy of the ideal gas is
S = k ln ΩE
 3  
2mπE 2 3N
= Nk ln 2
+ Nk lnV − k ln Γ (2.1.17)
h0 2
   
3N 3N 3N
Now Γ = −1 ! = !, since N  1.
2 2 2
Again by Stirling’s approximation
3N 3N 3N 3N
ln != ln −
2 2 2 2
  23
3N 3N
= N ln −
2 2
    23
3N ∼ 3N 3N
∴ ln Γ = N ln − (2.1.18)
2 2 2
Using (2.1.18) in (2.1.17)
 3   23
2mπE 2 3N 3
S (E,V ) = Nk ln + Nk lnV − Nk ln + Nk
h20 2 2
  3 
4mπ E 2 3
⇒ S (E,V ) = Nk ln V 2
+ Nk (2.1.19)
3h0 N 2
Solving for E in terms of S and V , we obtain the internal energy given by
3 h20 N
   
2 S
E (S,V ) = exp −1 (2.1.20)
4π m V 23 3 Nk
Again from thermodynamics
T dS = dE + PdV (2.1.21)
Then the temperature is given by  
∂E 2 E
T= = (2.1.22)
∂S V 3 Nk
Hence, the specific heat at constant volume is given by
 
∂E 3
CV = = Nk (2.1.23)
∂T V 2
From (2.1.21), the pressure P is given by
 
∂E 2E
P=− = (2.1.24)
∂V S 3V
62 • An Introduction to Equilibrium Statistical Mechanics

Using (2.1.22) in (2.1.24), we finally get

NkT
P= ⇒ PV = NkT
V
This is the equation of state of the ideal gas.
P ROBLEM 2.1: Consider an isolated system consisting of a large number N of weakly interacting
localized particles of spin 1/2 in an external magnetic field H. Each particle has a magnetic moment
µ, which can point either parallel or anti-parallel to the field H. Consider the energy range between
E and E + δE where δE is very small compared to E but microscopically large so that δE  µH.
What is the total number of states Ω (E) lying in this energy range? Hence, find the temperature T .
S OLUTION
→
−
Let n be the number of spins aligned parallel to H . So (N − n) is the number of spins aligned
→
−
anti-parallel to H .
∴ The total energy of the system is
E = −nµH + (N − n) µH
= NµH − 2nµH
N E
⇒ n = − (i)
2 2µH
Then, if the energy of the system is exactly E, the number of accessible microstates is
N! N!
ΩE = N Cn = =    (ii)
n! (N − n)! N E N E
− ! + !
2 2µH 2 2µH
where we use equation (i).
δE
From equation (i), δn = (considering magnitude only).
2µH
Hence, the total number of states in the range E to E + δE is
N! δE
Ω (E) = ΩE δn =     · (iii)
N E N E 2µH
− ! + !
2 2µH 2 2µH
From equation (1.22.10) the temperature T is defined through the relation
1 ∂ ln Ω
= (iv)
kT ∂E
Taking logarithms on both sides of equation (ii)
   
N E N E
ln Ω = ln N! − ln − ! − ln + !. (v)
2 2µH 2 2µH
Using Stirling’s approximation
     
N E N E N E
ln Ω = N ln N − N − − ln − + −
2 2µH 2 2µH 2 2µH
     
N E N E N E
− + l1n + + +
2 2µH 2 2µH 2 2µH
 Elements of Ensemble Theory • 63
       
N E N E N E N E
= N ln N − − ln − − + ln +
2 2µH 2 2µH 2 2µH 2 2µH
   
∂ ln Ω 1 N E 1 1 N E 1
∴ = ln − + − ln + −
∂E 2µH 2 2µH 2µH 2µH 2 2µH 2µH
N E
 
1  2 − 2µH 
= ln  
2µH  N E 
+
2 2µH
N E
 
1 1  2 − 2µH 
= ln  
kT 2µH  N E 
+
2 2µH
N E
−
2 2µH 2µH
⇒ = e kT
N E
+
2 2µH
2µH
N e kT +1
⇒ = 2µH
E e kT −1
−
µH
µH
µH −  
NµH e kT +e kT µH
⇒ − = µH = coth
E µH − kT kT
e kT −e
 
µH
∴ E = −NµH tanh
kT
This is the relation between energy E and temperature T .
P ROBLEM 2.2: Consider an extreme relativistic gas, characterized by the single particle energy
hc  2 1 Cp 4
states ε (nx , ny , nz ) = nx + n2y + n2z 2 . Calculate Ω and hence show that the ratio γ = = .
2l Cv 3
S OLUTION
The single particle energy states are given by
hc  2 1
ε (nx , ny , nz ) = nx + n2y + n2z 2 (i)
2l
2
4ε2 L2 4ε2V 3
∴ n2x + n2y + n2z = 2 2 = 2 2 (ii)
h c h c
Then the total energy of the system is given by
hc N  2 1
E = ∑ εr = ∑ nrx + n2ry + n2rz 2
r 2L r=1
(iii)
2 2 2 2 2 2 2 2 2 4L2 E 2 2
⇒ n1x + n1y + n1z + n2x + n2y + n2z + · · · + nNx + nNy + nNz = 2 2 = R (say)
h c
64 • An Introduction to Equilibrium Statistical Mechanics

This is the equation of a sphere of radius R in 3N-dimensional space; the surface of the sphere
corresponds to a constant energy E. The number of microstates Ω (N,V, E) is equal to the number
of lattice points on the positive segment of this spherical surface. If the energy of the system is
specified in between E and E + δE, the number of microstates Ω (N,V, E → E + δE) within this
range is equal to the number of lattice points lying within the positive part of the spherical shell
formed between the surfaces of radii R and R + δR in the 3N-dimensional space.
If Φ (N,V, E) is the volume of the positive part of the 3N-dimensional sphere of radius R, then
Ω (N,V, E → E + δE) = Ω = Φ (N,V, E + δE) − Φ (N,V, E)
By Taylor’s expansion,
∂Φ 1 ∂2 Φ
Φ (N,V, E + δE) = Φ (N,V, E) + δE + (δE)2 + . . .
∂E 2! ∂E 2
∼ Φ (N,V, E) + ∂Φ δE
=
∂E
∂Φ
∴ Ω = Φ (N,V, E + δE) − Φ (N,V, E) ∼ = δE (iv)
∂E
Now from appendix II, the volume of the 3N-dimensional sphere of radius R is
3N
π2
V3N (R) =   R3N (v)
3N
!
2
Hence,
 √ 3N
πL
 3N 3N 
1 π 2 2LE
3N
hc E 3N
Φ (N,V, E) = = (vi)
3N
 
2 hc 3N
! !
2 2
 √ 3N
πL
∂Φ 3N hc E 3N−1 3N
∴ =   = Φ (vii)
∂E 3N E
!
2
Thus,
∂Φ δE
Ω= δE = 3NΦ (N,V, E) (viii)
∂E E
Taking log on both sides
δE
ln Ω = ln Φ + ln 3N + ln (ix)
E
But from (vi), we have
√    √   
πLE 3N ∼ πLE 3N 3N 3N
ln Φ = N ln − ln ! = N ln − ln + (x)
hc 2 hc 2 2 2
 Elements of Ensemble Theory • 65

where we use Stirling’s approximation.

√   
πLE 3N 3N 3N δE
∴ ln Ω = N ln − ln + + ln 3N + ln
hc 2 2 2 E
 √  32 
3 2 πLE 3N δE
= N ln + + ln 3N + ln
2 3N hc 2 E
 
δE
For macroscopic systems, N  1 and so ln(3N)  3N. Again since δE  E, ln is negligible
E
and so  √ 2/3 
3 2 πLE 3N
= N ln
ln Ω ∼ + (xi)
2 3N hc 2
Then the entropy of the system is
 √ 2/3 
3 2 πLE 3Nk
S = k ln Ω = Nk ln + (xii)
2 3N hc 2
Solving for E,
  32
hc 3N s 3
E=√ e( Nk 2 ) (xiii)
πL 2
The temperature of the gas is given by
 
1 ∂S Nk
= = ⇒ E = NkT (xiv)
T ∂E N,V E
The specific heat at constant volume is
 
∂E
CV = = Nk (xv)
∂T N,S

For the pressure of the system, we obtain

 
∂E E
P=− = (xvi)
∂V N,S 3V
The specific heat at constant pressure is
         
∂E ∂V ∂E 1 ∂E 4 ∂E 4
CP = +P = + = = Nk (xvii)
∂T N,P ∂T N,P ∂T N,P 3 ∂T N,P 3 ∂T N,P 3
CP 4
∴ γ= = (Proved) (xviii)
V 3

P ROBLEM 2.3: The quantity Ω (N,V, E) is called the microcanonical partition function. Show that
= ln Ωmax , where Ωmax is the largest Ωi in the series.
ln Ω (N,V, E) = ln (Ω1 + Ω2 + · · · ) ∼
66 • An Introduction to Equilibrium Statistical Mechanics

S OLUTION
Let us assume that there are as many N systems with Ωi ’s comparable to Ωmax , that is, as many
such systems as there are particles in any one system. Thus,
Ω∼ = N × Ωmax ⇒ ln Ω ∼ = ln Ωmax + ln N (i)
But ln N  N.
∴ ln Ω (N,V, E) ∼
= ln Ωmax (Proved) (ii)

P ROBLEM 2.4: A one-dimensional chain is made up of N identical elements of length l. The angle
between successive elements can be either 0◦ or 180◦ , but there is no difference in internal energy
between these two possibilities. For the sake of counting, one can think of each element as either
pointing to the right (+) or to the left (−). Then one has
N = n+ + n−
L = l (n+ − n− ) = l (2n+ − N) .
(a) Use the microcanonical ensemble to find the entropy as a function of N and n+ , S (N, n+ ).
(b) Find an expression for the tension in the chain as a function of T , N and n+ , ℑ (T, N, n+ ).
(c) Rearrange (b) to give the length as a function of N, T and ℑ.

S OLUTION
N!
(a) The number of ways of choosing n+ elements from a total of N is . It follows
(N − n+ )!n+ !
that
N!
Ω (N, n+ ) = (i)
(N − n+ )!n+ !
The entropy of the system is
S (N, n+ ) = k ln Ω = k [ln N! − ln (N − n+ )! − ln n+ !]
Using Stirling’s approximation, the above equation becomes
= k [N ln N − N − (N − n+ ) ln (N − n+ ) − n+ ln n+ + n+ ]
S (N, n+ ) ∼
= k [N ln N − (N − n+ ) ln (N − n+ ) − n+ ln n+ ] (ii)
(b)
 
ℑ ∂S ∂S ∂n+ ∂S 1
=− =− =− ×
T ∂L N,E ∂n+ ∂L ∂n+ 2l
 
k N − n+ n+
=− + ln (N − n+ ) − − ln n+
2l N − n+ n+
 
k N − n+
= − ln
2l n
 + 
kT N − n+
∴ (N, T, n+ ) = − ln (iii)
2l n+
 Elements of Ensemble Theory • 67

(c) From (iii), we have

1
n+ = N   (iv)
1 + exp − 2lℑ
kT
We now use this expression for n+ to find L:
 
2/ℑ
  
1 + exp −
 2   2 kT 
L = l (2n+ − N) = Nl    − 1 = Nl    −  
 2/ℑ   2/ℑ 2/ℑ 
1 + exp − 1 + exp − 1 + exp −
kT kT kT
     
2/ℑ 1/ℑ 1/ℑ
1 − exp − exp − − exp −
kT kT kT
= Nl   = Nl    
2/ℑ 1/ℑ 1/ℑ
1 + exp − exp − + exp −
kT kT kT
 
lℑ
∴ L = Nl tanh (v)
kT
For high temperatures, where kT  lℑ, tanh x −→ x for x, so
Nl 2
L= ℑ (vi)
kT
The fact that the length is proportional to the tension ℑ shows that Hooke’s law applies to this
system, at least for high temperatures.

2.1.3 A two-level system

Consider a system of N (non-interacting and distinguishable) particles. Let the only accessible
states for the system be the ground state of zero energy and an excited state of energy ε.

Fig. 2.2. Two non-degenerate energy levels separated by energy.

Let the number of particles in the ground state of zero energy be N0 and that in the excited state
with energy ε be Nε .
∴ N0 + Nε = N (2.1.25)
The total energy of the system is
E
E = (N0 × 0) + (Nε × ε) ⇒ Nε = (2.1.26)
ε
68 • An Introduction to Equilibrium Statistical Mechanics

If the particles are distinguishable, the number of microstates accessible to the system is equal to
the number of ways choosing Nε particles from N particles and is given by
N! N!
Ω = N CNε = =   (2.1.27)
Nε ! (N − Nε )! E E
! N− !
ε ε
Assuming that N, Nε and (N − Nε ) are all large numbers compared to 1, we use Stirling’s approxi-
mation for the factorials of large numbers and then the entropy is
S = k ln Ω = k [N ln N − Nε ln Nε − (N − Nε ) ln (N − Nε )]
    
E E E E
=k − N ln 1 − − ln (2.1.28)
ε Nε ε Nε
The temperature of the system is given by
   
1 ∂S k Nε
= = ln −1 (2.1.29)
T ∂E N,V ε E
from which the energy E of the system is given by
Nε
E= ε (2.1.30)
1 + e kT
E N
∴ Nε = = ε (2.1.31)
ε 1 + e kT
Now for T → 0, i.e., kTε  1, Nε → 0, that is, all particles are frozen in the ground state. As T
increases, both E and Nε increase and when T → ∞, i.e., kTε  1, Nε ; N2 , that is, half the particles are
in the ground state and half in the excited state and the energy of the system attains its maximum
= Nε
value of E ∼ 2 . The specific heat at constant volume is given by
  ε
∂E  ε 2 e kT
CV = = Nk ε (2.1.32)
∂T V kT (1 + e kT )2

CV

kT
ε
0.42

Fig. 2.3. The specific heat at constant volume of a two-state system, showing Schottky hump.
 Elements of Ensemble Theory • 69

The specific heat CV is zero at both very low and very high temperatures and is maximum at T ∼=
0.42 kε . The hump, observed in the CV vs T plot, is called Schottky anomaly. Such a peak, when
observed experimentally, is therefore an indication of a gap in its energy states.

2.2 CANONICAL ENSEMBLE

This ensemble is a collection of independent systems having same temperature T , volume V and
number N of particles. So the systems of this ensemble are separated by rigid, impermeable, but
conducting walls. Since the systems are separated by conducting walls, heat can be exchanged
between the systems till they reach a common temperature T . Thus, in canonical ensemble, the
systems can exchange energy but not particles.

N,V,T N,V,T

N,V,T N,V,T

Fig. 2.4. Canonical ensemble.

2.2.1 Probability distribution and canonical partition function

Consider a small system A in thermal interaction with a heat reservoir A .
What is the probability Pr of finding the system A in any one particular microstate r of energy
Er ?

A A′

Fig. 2.5. A small system ‘A’ in thermal interaction with a heat reservoir ‘A ’.

We assume weak interaction between A and A , so that their energies are additive. The energy
of A is, of course, not fixed. Only the total energy of the combined system, A0 = A + A (which is
isolated) has a constant value in some range between E 0 and E 0 + δE:
Er + E  = E 0 = constant (2.2.1)
70 • An Introduction to Equilibrium Statistical Mechanics

where E  denotes the energy of the reservoir A . Thus, when A has an energy Er , the reservoir A
must have an energy E  = E 0 − Er .
Let Ω (E  ) denote the number of accessible microstates of A , when it has an energy in the
range near E  . If the system A is in the particular
 state r, the number of states accessible to the
combined system A0 is 1 × Ω (E  ) = Ω E 0 − Er .
Then the probability of finding the system A in this state is
 
Ω (E  ) Ω E 0 −Er
Pr = Pr (Er ) = 0 = ,
Ωtotal Ω0total
where Ω0total denotes the total number of states accessible to the combined system A0 .
But Ω0total = constant = c0 (say).
Pr = c0 Ω E 0 − Er
 
∴ (2.2.2)
0
⇒ ln Pr = ln c + ln Ω (E0 − Er )

(2.2.3)
Since A is very small compared to A , Er  E 0 . Expanding ln Ω about E  = E 0 ,
∂ ln Ω 0
 
 0   0
ln Ω E − Er = ln Ω E −
 
Er − . . .
∂E  E
Since A acts as a reservoir, Er  E 0 and so we can neglect higher order terms.
 
 0   0 ∂ ln Ω
∴ ln Ω E − Er = ln Ω E −
 ∼ 
Er (2.2.4)
∂E  E 0
 
∂ ln Ω
The derivative is evaluated at the fixed energy E  = E0 and is thus a constant, indepen-
∂E  E 0
  1
dent of the energy E  and ∂ ln Ω =β= , where T is the temperature of the reservoir.
∂E E
 0 kT
The equation (2.2.4) then becomes
ln Ω E 0 − Er = ln Ω E 0 − βEr
   

Ω E 0 − Er = Ω E 0 e−βEr
   
⇒
 
Since Ω E 0 is just a constant, independent of r, equation (2.2.2) becomes

Pr = ce−βEr (2.2.5)
where c is a constant, independent of r.
The probability (2.2.5) is a very general result and is of fundamental importance in statisti-
cal mechanics. The exponential factor e−βEr is called the ‘Boltzmann factor’; the corresponding
probability distribution (2.2.5) is known as the ‘canonical distribution’.
Now according to the normalization conditions for probabilities,
∑ Pr = 1 ⇒ c−1 = ∑ e−βE r

r r
 Elements of Ensemble Theory • 71

e−βEr e−βEr
∴ Pr = = (2.2.6)
∑ e−βEr Z
r
where
Z = ∑ e−βEr (2.2.6a)
r
‘Z’ is called the canonical partition function. (The letter Z is used because the German name is
Zustandsumme.) It is the sum over all accessible microstates of the system A.
It should be noted that the expression (2.2.6a) is correct, if the energy levels are discrete and
non-degenerate. In the general case, an energy level Er consists of a group of states gr in number all
of the same energy Er and of equal probability. Thus in summing over all accessible microstates, it
is necessary to repeat equal terms (involving Er ) gr times. Instead, it is simpler to use a modified
form of the partition function
Z = ∑ gr e−βEr (2.2.6b)
r
where the label r goes over all energy levels. If the level is non-degenerate, then gr = 1. Equation
(2.2.6b) is the general form of the partition function.
Accordingly,
gr e−βEr
Pr = (2.2.6c)
∑ gr e−βEr
r
Once the probability distribution is known, various mean values can be computed. For example, let
x be any quantity assuming the value xr in state r of the system A. Then the mean value of x is
∑ xr Pr ∑ xr e−βEr
x= r
= ∑ xr Pr = r
(2.2.7)
∑ Pr r ∑ e−βEr
r r

2.2.2 Energy fluctuations in the canonical ensemble: correspondence with

the microcanonical ensemble
According to equation (2.2.7), the mean energy value is
∑ Er e−βEr
r
E= (2.2.8)
∑ e−βEr
r
Now the canonical partition function is
∂Z
Z = ∑ e−βEr ⇒ −
∂β ∑
= Er e−βEr (2.2.9)
r r
Using (2.2.9) in (2.2.8)
1 ∂Z ∂ ln Z
E =− =− . (2.2.10)
Z ∂β ∂β
 Elements of Ensemble Theory • 73

2.2.3 To calculate the canonical partition function

1. Classical ideal gas
We consider a classical monatomic gas at an absolute temperature T , contained in a volume V . Let
us denote the position vector of the ith molecule by −
→
r i , its momentum by →
−p . Then the total energy
i
of the gas is given by
N
p2i
E=∑ +U (→
−
r1 , →
−
r2 . . . →
−
rN ) (2.2.17)
i=1 2m
The first term on the right-hand side of the above equation represents the total kinetic energy of
all the molecules and the second term represents the potential energy of interaction between the
molecules.
The classical canonical partition is given by
 3
d r1 . . . d 3 rN d 3 p1 . . . d 3 pN
 
1  2

2 →
− →
−

Z = exp −β p1 + · · · + pN +U ( r 1 , . . . , r N )
2m h3N
0
1 βp21 βp2N
  
3 3 →
− →
−
= 3N e
− 2m d p1 . . . e
− 2m d pN e−βU( r 1 ... r N ) d 3 r1 . . . d 3 rN
h0
Since the kinetic energy is a sum of terms, one for each molecule, the corresponding part of the par-
tition function breaks into a product of N integrals, each identical except for the irrelevant variable
∞ − βp2 3
of integration and equal to e 2m d p.
−∞
Since U (→−r , . . . , →
1
−r ) is not in the form of a simple sum of terms for individual molecules, the
N
integration over the coordinates → −r , . . . , →
1
−r is very difficult to carry out. This is why the treatment
N
of non-ideal gases is complicated (Chapter 7).
For an ideal gas, we can set U = 0 and so the integral becomes trivial:

d 3 r1 . . . d 3 rN = V N (2.2.18)

since each integration extends over the volume V of the container.

∴ Z = ξN (2.2.19)

where
 ∞
V βp2
ξ= 3 e− 2m d 3 p (2.2.20)
h0 −∞

is the partition function for a single particle.

Equation (2.2.19) is the relation between the single particle canonical partition function and that
for the whole system, if the particles of the system are identical, non-interacting and distinguishable.
74 • An Introduction to Equilibrium Statistical Mechanics

Now
∞ ∞
p2x + p2y + p2z
 
2 β
− βp 3 − 2m
e 2m d p= e d px d py d pz
−∞ −∞
∞ ∞ ∞
 3
βp2x βp2y βp2
z 2πm
= e
− 2m d px e − 2m d py e− 2m d pz = ,
β
−∞ −∞ −∞

∞ 
−αx2 π
where we use the standard integral e dx =
α
−∞
  32
V 2πm V 3
∴ ξ= 3 = 3
(2πmkT ) 2 (2.2.21)
h0 β h0
Using (2.2.21) in (2.2.19)
VN 3N
Z (N,V, T ) = 3N
(2πmkT ) 2 (2.2.22)
h0
This is the canonical partition function of a classical ideal gas.
What is going to happen, if we would use the same formula, i.e., equation (2.2.19) to calculate
the partition function for a system of N indistinguishable, identical, non-interacting particles. To
illustrate that this is wrong, we consider two indistinguishable particles:
Z (N = 2,V, T ) = ξ2 = ∑ e−βεr1 × ∑ e−βεr2
r1 r2

= ∑ e −2βεr
+ ∑ ∑ e−β(ε r1 +εr2 )
r=r1 =r2 r1 =r2
     
particles both particles in different
in the same state states

However, the particles in the second term are counted twice, but since the particles are indistin-
guishable this does not generate a new state, because

e−β(εr1 +εr2 ) = e−β(εr2 +εr1 ) .

Hence, Z (N = 2,V, T ) = ξ2 counts the particles in different states falsely twice!
We can correct for that by
1
Z (N = 2,V, T ) = ∑ e−2βεr +
2 r1∑ ∑ e−β(εr1 +εr2 )
r=r1 =r2 =r2
1
= ∑ e−2βεr +
2! r1∑ ∑ e−β(εr1 +εr2 ) .
r=r1 =r2 =r2
 Elements of Ensemble Theory • 75

Similarly, for a system of 3 particles

1
Z (N = 3,V, T ) = ∑
r=r1 =r2 =r3
e−3βεr + · · ·
some particles
+
6 ∑ e−β(εr1 +εr2 +εr3 )
r1 =r2 =r3
particles in same state in same state all particles in
different states

1
= ∑
r=r1 =r2 =r3
e−3βεr + · · ·
some particles
+
3! ∑ e−β(ε
r1 =r2 =r3
r1 +εr2 +εr3 )
particles in same state in same state all particles in
different states

In general, for a system of N indistinguishable particles,

1 −β(εr +εr +εr3 + ··· +εrN )
Z (N,V, T ) = ∑
r=r1 =r2 =r3 =···=rN
e−Nβεr + · · · +
some particles N! ∑ e all particles
1 2
in (2.2.23)
r1 =r2 =r3 different states
all particles in in same state
same state

We now use semi-classical argument according to which

1. gas particles are identical
2. only very weak or no interaction between the particles exists
3. there are much more available energy levels than particles (classical continuum of quantum
mechanics).
Thus, the probability that any single particle state is occupied by more than one particle is very
small. This argument greatly simplifies (2.2.23) because only the last term counts.
Thus, for a system of N indistinguishable, identical, non-interacting particles
ξN
Z (N,V, T ) = (2.2.23a)
N!
2. Classical non-ideal gas
For a non-ideal monatomic gas, the interaction term U = 0. Therefore, the canonical partition
function for a classical non-ideal gas is given by
 3
d r1 . . . d 3 rN d 3 p1 . . . d 3 pN
 
1  2

2 →
− →
−

Z (N,V, T ) = exp −β p1 + · · · + pN +U ( r 1 , . . . , r N )
2m h3N
0
1 βp21 βp2N
  
3 3 →
− →
−
= 3N e − 2m d p1 . . . e− 2m d pN e−βU( r 1 ... r N ) d 3 r1 . . . d 3 rN
h0
  3N2
1 2πm
∴ Z (N,V, T ) = 3N Q (2.2.24)
h0 β
where 
→
− →
−
Q= e−βU( r 1 ... r N ) d 3 r1 . . . d 3 rN
known as configurational partition function (chapter 7). The calculation of Q is quite difficult. For
the sake of simplicity, we apply mean field approximation. Then equation (2.2.24) becomes
  3N  V N
1 2πm 2 −UM β 3
Z (N,V, T ) = 3N e d r
h0 β 0
76 • An Introduction to Equilibrium Statistical Mechanics

where UM is the mean field. We take it as follows:


∞, r < r0
UM =
−U, r ≥ r0

UM

r0
0 r

Fig. 2.6. UM vs. r plot.

Therefore,
 N
  3N2 V0 V
1 2πm  e−UM β d 3 r + e−UM β d 3 r
Z (N,V, T ) = (2.2.25)
h3N
0 β
0 V0
where V0 is the volume corresponding to r = r0 . Now in the range 0 to V0 ,UM → ∞ and in the range
V0 to V , UM = −U. Therefore, equation (2.2.25) becomes
  3N  V N
1 2πm 2 Uβ 3
Z (N,V, T ) = 3N e d r
h0 β V0

1 3N
 U N

⇒ Z (N,V, T ) = 3N (2πmkT ) 2 (V −V0 ) e kT (2.2.26)
h0
This is the canonical partition function for a classical non-ideal gas in the mean field approximation.
This topic will be further discussed in Chapter 7.

2.2.4 Thermodynamic quantities in terms of partition function

All the physical quantities can be expressed completely in terms of the partition function Z. Some
of them are shown below.
 Elements of Ensemble Theory • 77

1. Entropy
According to Boltzmann’s entropy relation, the entropy S is related to the thermodynamic proba-
bility Ω through the relation
S = k ln Ω (2.2.27)
For a classical system, the total number of molecules is
N = ∑ ni (2.2.28)
i
and
gni i
Ω = N! ∏ (2.2.29)
i ni !
Taking logarithm and then using Stirling’s approximation
ln Ω = N ln N − N + ∑ (ni ln gi − ni ln ni + ni ) (2.2.30)

According to classical statistics, for equilibrium state [see equation (3.1.7)]

ni = gi e−α e−βEi = gi Ae−βEi (2.2.31)
where A is a new constant. Using (2.2.31) in (2.2.30)
 
ln Ω = N ln N − N + ∑ ni ln gi − ni ln gi Ae−βEi + ni
i
= N ln N − N + ∑ ni ln gi − ∑ ni ln gi − ln A ∑ ni + β ∑ ni Ei + ∑ ni
i i i i i
= N ln N − N − N ln A + βE + N
= N ln N − N ln A + βE
N
∴ ln Ω = N ln + βE (2.2.32)
A
where we put ∑ ni = N = total number of particles and ∑ ni Ei = E = total energy of the system.
i i
Again using (2.2.28) in (2.2.31)
N
∑ ni = A ∑ gi e−βE i
⇒ N = AZ ⇒ Z =
A
(2.2.33)
i i

where Z = ∑ gi e−βEi −→ partition function.

i
After using (2.2.33), equation (2.2.32) becomes
ln Ω = N ln Z + βE (2.2.34)
Finally, using (2.2.34) in (2.2.27), the expression for entropy becomes
S = Nk ln Z + kβE
 
1
= Nk ln Z + k E
kT
E
∴ S = Nk ln Z + (2.2.35)
T
80 • An Introduction to Equilibrium Statistical Mechanics

Taking logarithms on both sides

   
3N 2πmk
ln Z = N lnV + ln + ln T .
2 h20
Differentiating partially with respect to V
∂ ln Z N
= (2.2.49)
∂V V
Using (2.2.49) in (2.2.48)
pV = NkT (2.2.50)
This is equation of state for an ideal gas.

2. Reproduction of van der Waals’ equation of state

From equation (2.2.26), the canonical partition function for a non-ideal gas in the mean field ap-
proximation is
1  U N

Z (N,V, T ) = 3N (2πmkT )3N/2 (V −V0 ) e kT
h0
Taking logarithms on both sides
 
3N 2πmkT NU
ln Z = ln 2
+ N ln (V −V0 ) + (2.2.51)
2 h0 kT
N aN
Now, V0 ∝ N ⇒ V0 = bN, where b is a constant and U ∝ ⇒ U = , where a is also a constant.
V V
Then equation (2.2.51) becomes
aN 2
 
3N 2πmkT
ln Z = ln + N ln (V − bN) +
2 h20 kTV
Differentiating partially with respect to V
 2
∂ ln Z N N
= −akT + (2.2.52)
∂V V V − bN
Using (2.2.52) in (2.2.48)
 2
N kT N a kT
p = −a + ⇒ p = −  2 + V
V V − bN V
N −b
   N
a V
⇒ p +  2 − b = kT
V N
N
V
Let = v, known as specific volume.
N
 a
∴ p+ (v − b) = kT (2.2.53)
v2
This is the van der Waals equation of state for a real gas.
 Elements of Ensemble Theory • 81

2.2.6 Generalized equipartition theorem

Let qi be the generalized co-ordinate. We want to calculate the ensemble average of the quantity
 
∂H
qi , where H is the Hamiltonian of the system under consideration. Using dτ ≡ d 3N qd 3N p,
∂q j
we can write in the canonical ensemble  
   q ∂H e−βH dτ
∂H i ∂q j
qi =  (2.2.54)
∂q j e−βH dτ
Let us now consider the integral in the numerator. Integrating over q j by parts, it becomes
     
1 −βH (q j )2 1 ∂qi
− qi e + e −βH
dq j dτ( j) (2.2.55)
β (q j )1 β ∂q j
where (q j )1 and (q j )2 are the extreme values of the coordinate q j . If q j is a space coordinate, then
its extreme values will correspond to positions of the walls of the container and so the potential
enegy would be infinite The infinite potential energy immediately implies  that the Hamiltonian
 H
1 −βH (q j )2
of the system would be infinite. Therefore, the integrated part, i.e., − qi e vanishes.
β (q j )1
The integral part contains the term (∂qi /∂q j ), which is equal to δi j . Finally, (2.2.55) becomes
1

δi j e−βH dτ (2.2.55a)
β
Substituting this in (2.2.54)  
∂H
qi = δi j kT (2.2.56)
∂q j
independent of H.
Similarly  
∂H
pi = δi j kT (2.2.57)
∂p j
which is independent of the form of the Hamiltonian H.
Equations (2.2.55) and (2.2.56) are known as generalized equipartition theorem.
This topic will be further discussed in Chapter 3.

2.2.7 Virial theorem

According to the Hamilton’s equations of motion,
∂H
ṗi = − (2.2.58)
∂qi
Using this in (2.2.57)
qi ṗi  = −kT (2.2.59)
Adding over all i from i = 1 to 3N, we obtain
 
3N
∑ qi ṗi = −3NkT (2.2.60)
i=1
82 • An Introduction to Equilibrium Statistical Mechanics

This is known as the virial theorem, because ∑ qi ṗi is known as virial in classical mechanics.
i

P ROBLEM 2.5: Find the classical canonical partition function of a one-dimensional linear har-
monic oscillator. Also find its average energy E.

S OLUTION
Since this is a one-dimensional case, the classical canonical partition function is
d p dx

Z= e−βE .
h0
p2 1
where E is the total energy of the oscillator and is given by E = + mω2 x2 .
2m 2
 2

 ∞ − βp 1 2 2
1 
 ∞
− mω x 
∴ Z=  e 2m d p e 2 dx .
h0 −∞ −∞

 ∞ 
−αx2 π
Using the standard integral e dx = ,
−∞ α
   
1 2πm 2π 2π 2πkT
Z= = = .
h0 β mβω2 h0 βω h0 ω
This is the canonical partition function.
The average energy of the oscillator is
∂ ln Z 1
E = − = = kT.
∂β β

P ROBLEM 2.6: Find the canonical partition function of a relativistic classical ideal gas with
energy-momentum relationship E = pc.

S OLUTION
For a classical ideal gas, the canonical partition function is
d 3N pd 3N x

Z= e−βE = ξN ,
h3N
0
where ξ is the single particle partition function and is given by
3 pd 3 x  ∞  π  2π
4πV
 ∞
−βpc d V

−βpc 2
ξ= e = 3 d pe p dθ sin θ dφ = 3 d pe−βpc p2
h30 h0 0 0 0 h0 0
 3
4πV 1 4πV
 ∞
2 kT
= 3 d (βpc) e −βpc
(βpc) = 3 Γ (3)
h0 (βc)3 0 h0 c
 Elements of Ensemble Theory • 85

Also ‘Z’ can be written as

Z = ∑ N Cn e−βnw (iii)
n
Now the mean value of n is
∑ nN Cn e−βnw ∑ nN Cn e−βnw
n n 1 ∂ ln Z N ∂   
n= = =− =− ln 1 + e−βw
∑ N Cn e−βnw Z w ∂β w ∂β
n
e−βw
=N
1 + e−βw
n −w
⇒ n + ne−βw = Ne−βw ⇒ = e kT
N −n
When n  N,
−w
n ≈ Ne kT (iv)

P ROBLEM 2.10: Show that the partition function of an extreme relativistic gas consisting of 3N
indistinguishable particles, with energy-momentum relationship ε = pc, c being the speed of light
1
  kT 3N
and moving in one dimension is given by Z (N, L, T ) = (3N)! 2L hc , where L being the ‘length’
of the space available.

S OLUTION
The single particle canonical partition function is
 
1
 ∞
L kT

ξ= e −βε
dqd p = e −βpc
dp = L (i)
h h 0 hc
Therefore, the canonical partition function of the system of N non-interacting indistinguishable
particles is
  3N
ξ3N 1 kT
Z (N, L, T ) = = L (ii)
(3N)! (3N)! hc

P ROBLEM 2.11: Consider a system in thermal equilibrium at temperature T − its two states with
energy difference 4.8 × 10−14 erg occur with relative probability e2 erg / deg. Calculate the temper-
ature. Boltzmann’s constant k = 1.38 × 10−16 erg / deg.

S OLUTION
Let P1 and P2 be respectively the probabilities of the state of energies E1 and E2 . Then
E1 E2
P1 = ce− kT and P2 = ce− kT , where c is a constant of proportionality.
P1 (E2 −E1 ) (E2 − E1 ) 4.8 × 10−14
∴ = e kT ⇒ T =  = = 173.913 K
P2 P1 1.38 × 10−16 × ln e2
k ln
P2
86 • An Introduction to Equilibrium Statistical Mechanics

P ROBLEM 2.12: The first vibrational energy of a diatomic molecule is 600 cm−1 above the ground
state. Calculate the relative population of molecules in these two levels at T = 400 K.

S OLUTION
Let N0 and Eo be respectively the number of molecules and the energy of the ground state and
N1 and E1 be the corresponding values for the first excited state.
(E1 −E0 )
∴ N1 = N0 e− kT (i)
Now
hλ
(E1 − E0 ) = hv = = hcv = 6.6 × 10−27 × 3 × 1010 × 600 = 1.2 × 10−13 erg
c
N1
 
(E1 −E0 ) 1.2×10−13
−
∴ = e− kT = e 1.38×10−16 ×400 ∼ = 0.1 ⇒ N1 = 0.1 × N0
N0

P ROBLEM 2.13: Consider a system which can take only three different energy states E1 = 0,
E2 = 1.38 × 10−14 ergs, E3 = 2.76 × 10−14 ergs. Find the probability that at temperature 100 K the
system may be (a) in one of the microstates of energy E3 , (b) in the ground state E1 .

S OLUTION
Let P1 , P2 and P3 be respectively the probabilities with which the microstates can occur in three
energy values E1 , E2 and E3 .
E1
Then P1 = ce− kT = c
E2 1.38×10−14
−
P2 = ce− kT = ce 1.38×10−16 ×100 = ce−1
and
E3 2.76×10−14
−
P3 = ce− kT = ce 1.38×10−16 ×100 = ce−2
where c is a constant of proportionality.
Again these three states can occur in 2, 5 and 4 different ways respectively. Then P1 = 2c,
P2 = 5ce−1 and P3 = 4ce−2 . But
3
∑ Pi = 1 ⇒ (P1 + P2 + P3 ) = 1 ⇒
 
2c + 5ce−1 + 4ce−2 = 1
i=1
1 1
⇒c= ⇒c=
2 + 1.84 + 0.54 4.38
Since e = 2.72.
(a) The probability for the system to be in one of the microstates of energy E3 is
4
P3 = 4ce−2 = = 0.12.
4.38 × (2.72)2
 Elements of Ensemble Theory • 89

Again
∂ ln Z
E =− (vi)
∂β
Comparing (v) and (vi)
Fβ = − ln Z
(vii)
⇒ F = −kT ln Z
Finally, using (vii) in (iii)
E F 1 1 S
∑ Pr ln Pr = − kT + kT =
kT
(F − E) =
kT
(−T S) = −
k
r
⇒ S = −k ∑ Pr ln Pr [Proved] (viii)
r
In case of microcanonical ensemble, we have a group of Ω states (which are equally likely to
occur) for each member. The value of Pr is then Ω1 for each of these states and 0 for all others.
Consequently,
Ω   
1 1
S = −k ∑ ln = k ln Ω [Proved] (ix)
r=1 Ω Ω

P ROBLEM 2.17: The states of a system are (i) a group of g1 equally likely states with a common
energy ε1 , and (ii) a group of g2 equally likely states with a common energy ε2 = ε1 .
    
P1 P2
(a) Show that the entropy of the system is given by S = −k P1 ln + P2 ln
     g 1  g2
g2 −x x ε2 − ε1
(b) Also show that S = k ln g1 + ln 1 + e + , where x = ,
g1 1 + (g1 /g2 )ex kT
assumed positive.
(c) Check that at T → 0, S → k ln g1 .

S OLUTION
(a) The canonical partition function is
Z = ∑ gr e−βEr (i)
r
and
gr e−βEr
Pr = (ii)
Z
 
Pr e−βEr Pr
∴ = ⇒ ln = − ln Z − βEr
g Z gr
 r
Pr
⇒ ∑ Pr ln gr = − ln Z − βE (iii)
r
90 • An Introduction to Equilibrium Statistical Mechanics

The Helmholtz free energy is

F = E −TS
⇒ dF = dE − T dS − SdT = dE − (dE + PdV ) − SdT = −PdV − SdT
  
S = − ∂F
∂T

∴  V (iv)
P = − ∂F
∂V

T
Hence,      
∂F ∂F ∂
E = F +TS = F −T = F +β = (Fβ) (v)
∂T V ∂β V ∂β V
Again
∂ ln Z
E =− (vi)
∂β
Comparing (v) and (vi)
F = −kT ln Z (vii)
Using (vii) in (iii), we obtain
   
Pr F −E S Pr
∑ Pr ln gr = kT = − k ⇒ S = −k ∑ Pr ln gr
r r
    
P1 P2
∴ S = −k P1 ln + P2 ln (Proved) (viii)
g1 g2
(b) We know that 
g1 e−βε1 
P1 = 

Z (ix)
g2 e−βε2 
P2 =


Z
Using (ix) in (viii)
S = −k [P1 (−βε1 − ln Z) + P2 (−βε2 − ln Z)]
= k [β(P1 ε1 + P2 ε2 ) + (P1 + P2 ) ln Z]
= k [β(P1 ε1 + P2 ε2 ) + ln Z] (∵ P1 + P2 = 1)
   
g1 ε1 e−βε1 + g2 ε2 e−βε2
=k β + ln Z
Z
   
g1 ε1 e−βε1 + g2 ε2 e−βε2
=k β + ln(g1 e−βε1 + g2 e−βε2 )
g1 e−βε1 + g2 e−βε2
    
g1 ε1 e−βε1 + g2 ε2 e−βε2 g2 −β(ε2 −ε1 )
=k β + ln(g1 e−βε1
) + ln 1 + e
g1 e−βε1 + g2 e−βε2 g1
92 • An Introduction to Equilibrium Statistical Mechanics

If the temperature T is so high that the thermal energy kT is large compared to the separation hν
hν
between energy levels, i.e., βhν = kT  1,
   
1 1 1 kT kT
E ≈ hν + hν = hν + ≈ hν × = kT −→ classical result.
2 kT 2 hν hν

P ROBLEM 2.19: A classical point particle is moving in a 3-dimensional harmonic oscillator po-
tential well, V (r) = 1/2Kr2 = 1/2K(x2 + y2 + z2 ) at absolute temperature T . Obtain a formula for
the probability that the particle is in between r and r + dr from the centre of attraction. Also obtain
a formula for the mean square distance of the particle from the centre of attraction and check your
result by comparison with the equipartition principle.
S OLUTION
The probability that the particle will be in phase space volume element d 3 rd 3 p
βp2 1 2 βp2 1 2
e− 2m −β 2 Kr d 3 pd 3 r e− 2m −β 2 Kr d 3 pr2 sin θ dθ dφ dr
= βp2 1
=  βp2 1
2 2
e− 2m −β 2 Kr d 3 pd 3 r e− 2m −β 2 Kr d 3 pr2 sin θ dθ dφ dr
Therefore, the probability that the particle will lie in the range r and r + dr is
   − βKr2 2 βp2
e 2 r sin θ dθ dφ dr e− 2m d 3 p
θ φ p
P(r)dr =     βKr2 βp2
e− 2 r2 sin θ dθ dφ dr e− 2m d 3 p
θ φ p r
βKr2 βKr2
e− 2 r2 dr
e− 2 r2 dr
= ∞ =   23
βKr2
e− 2 r2 dr 1√ 2
4 π Kβ
0
  32
4 Kβ 1 2
∴ P(r)dr = √ e− 2 βKr r2 dr (i)
π 2

The mean square distance is

∞ 4 − βKr2   52
r e 2 dr 3√ 2
 2
r P(r)dr 0 8 π βK
 
3 2 3kT
2
r  =  = ∞ = 3 = = (ii)
P(r)dr βKr
2 − 2
2
1√
  2 βK K
r e 2 2
dr π
4 βK
0
Therefore, the average potential energy = 12 Kr2  = 12 K 3kT 3
K = 2 kT , which is consistent with the
equipartition law.

P ROBLEM 2.20: Suppose a system of N non-interacting particles at temperature T in an external

→
−
magnetic field H pointing along the z-direction. Assume that each atom has spin 1/2 and an
intrinsic magnetic moment →
−
µ.
 Elements of Ensemble Theory • 93

1. If the magnetic moment of each atom can point either parallel or anti-parallel to the external
→
− →
−
field H , what is the mean magnetic moment µH  (in the direction of H , i.e., z-direction) of
such atom ?
2. Calculate the partition function of one atom, the mean magnetic moment along the field
direction and hence the susceptibility of the system, first by treating the system classically
and then quantum mechanically.

S OLUTION
Each atom can be in two possible states: the state (+) where its spin points up (i.e., parallel to
→
− →
−
H ) and the state (–) where its spin points down (i.e. anti-parallel to H ). In the (+) state, →
−µ is
→
−
parallel to H and so µH = µ. The corresponding magnetic energy of the atom is E+ = −µH. The
probability of finding the atom in this state is
P+ = Ce−βE+ = CeβµH (i)
1 →
−
where C is a constant of proportionality and β = kT .
Similarly, in the (–) state, µ is anti-parallel to
→
−
H and so µH = −µ. The corresponding magnetic energy of the atom is E− = +µH. The probability
of finding the atom in this state is
P− = Ce−βE− = Ce−βµH (ii)
∴ The mean value of µH is
eβµH − e−βµH
 
P+ (µ) + P− (−µ) µH
µH  = = µ βµH = µtanh (βµH) = µtanh .
P+ + P− e + e−βµH kT
When the system is treated classically
The energy of a magnetic dipole in the presence of the external magnetic field is
→
−
E = −→−µ · H = −µH cos θ (iii)
The canonical partition function of the system is
Z = ξN (iv)
where ξ is the single particle canonical partition function and is given by
 2π  π
sinh (βµH)
ξ = ∑ exp (βµH cos θ) = eβµH cos θ sin θ dθ dφ = 4π (v)
θ φ=0 θ=0 βµH
sinh (βµH) N
 
∴ Z = 4π (vi)
βµH
The average magnetic moment along the field direction is given by
∑ µcos θ exp (βµH cos θ)  
θ 1 ∂ 1
µH = µcos θ = = ln ξ = µ coth (βµH) −
∑ exp (βµH cos θ) β ∂H βµH
θ
= µL(βµH) (vii)
94 • An Introduction to Equilibrium Statistical Mechanics

where L(x) is the so-called Langevin function:

1
L (x) = coth x − (viii)
x
Therefore, the magnetization of the system is given by
M = NµH = NµL (x) (ix)
For magnetic fields so strong (or temperatures so low) that the parameter x  1, L (x) ∼
= 1 and so
M = Nµ −→ saturation. On the other hand, for high temperatures (or the magnetic fields so weak)
∼
x
= and so
that the parameter x  1, L (x) ∼
3
Nµ2
M∼
= H (x)
3kT
Therefore, the high temperature susceptibility of the system is given by
Nµ2 C
 
∂M
χ (T ) = Limit = = (xi)
H→0 ∂H 3kT T
This is Curie law of paramagnetism, the constant C being the Curie constant of the system.

When the system is treated quantum mechanically

In this case, the major modification arises from the fact that the magnetic dipole moment −
→
µ and its
component µH along the field direction cannot have arbitrary values.
→
−
The magnetic moment → −µ is related to the total angular momentum  J through the relation
→
− →
−
µ = gµ J
B (xii)
e
where µB = 2mc , known as Bohr magneton and g is the so-called g factor of the atom, e = charge of
an electron, m = mass of an electron and c = velocity of light in vacuum.
Using (xii) in (iii), we have
→
− →−
E = −gµB J · H = −gµB HJz (xiii)
→
−
since H is along the z-direction. In quantum mechanical description, the values which Jz can
assume are discrete and are given by Jz = m, where m = −J, −(J − 1), . . . , (J − 1), J.
∴ The possible energies of the atom are
Em = −gµB Hm (xiv)

Then the partition function of one atom is

J
Z = ∑ e−βEm = ∑ eβgµB Hm (xv)
m m=−J
 Elements of Ensemble Theory • 95

gµB H
Let x = βgµB H = −→ a dimensionless parameter. Therefore,
kT
J
Z= ∑ exm
m=−J
 
= e−xJ + e−x(J−1) + · · · + exJ
1 1
e−xJ − ex(J+1) e−x(J+ 2 ) − ex(J+ 2 ) sinh J + 12 x
 
= = x x = x
1 − ex e− 2 − e 2 sinh
2
sinh J + 12 βg µB H
 
∴ Z= (xvi)
sinh 21 βg µB H
Now the probability of finding the atom in a state labelled m is given by
Pm = Ce−βEm (xvii)
where C is a constant of proportionality, determined from the normalization condition, ∑ Pm = 1.
m
Therefore, the mean value of the magnetic moment along the field direction (i.e. z-direction) is
J
∑ µH Pm ∑ µH eβgµB Hm
m m=−J
µH  = = (xviii)
∑ Pm J
m ∑ eβgµB Hm
m=−J
→
−
Since the field H is along the z-direction, µH = µz = gµB Jz = gµB m, from (iv).
J J
∑ gµB meβgµB Hm ∑ (βgµB m) eβgµB mH
m=−J 1 m=−J 1 1 ∂Z 1 ∂ ln Z
∴ µH  = = = =
J β J β Z ∂H β ∂H
∑ eβgµB Hm ∑ eβgµB mH
m=−J m=−J
   
1 ∂ ln Z ∂x ∂ ln Z ∂ 1 x
= = gµB = gµB ln sinh J + x − ln sinh
β ∂x ∂H ∂x ∂x 2 2
 
J + 21 cosh J + 12 x 21 cosh 2x
  
= gµB −
sinh J + 12 x sinh 2x
 

⇒ µH  = gµB JBJ (x)

    
1 1 1 1 x
where BJ (x) = J+ coth J + x − coth , known as Brillouin function. Therefore,
J 2 2 2 2
the magnetization of the system is given by
M = N µH  = NgµB JBJ (x) (xix)
= 1 for all J and so
For strong fields and low temperatures (x 1), the function BJ (x) ∼
M = NgµB J −→ magnetic saturation.
∼
96 • An Introduction to Equilibrium Statistical Mechanics
 
1 1
On the other hand, for high temperatures and weak fields (x  1), BJ (x) = ∼ 1+ x, so that
3 J

(gµB J)2 g2 µ2B J (J + 1)

 
1
M=N∼ 1+ H =N H (xx)
3kT J 3kT
Therefore, the susceptibility is given by
Ng2 µ2B J (J + 1) 1
χ (T ) = ∝ , which shows that the Curie law is again valid.
3kT T

P ROBLEM 2.21: Show that entropy increases when two ideal gases at the same temperature and
pressure diffuse into each other. Discuss Gibbs paradox in this connection.

S OLUTION
According to Boltzmann’s relation, the entropy S of a system is related to the thermodynamic
probability Ω and is given by
S = k ln Ω (i)
From Maxwell-Boltzmann statistical count
gni i
Ω = N! ∏
i ni !
⇒ ln Ω = ln N! + ∑ (ni ln gi − ln ni !)
i
= ln N! + ∑ (ni ln gi − ni ln ni + ni )
i
= ln N! + ∑ ni (ln gi − ln ni + 1)
i  
gi
Ω = ln N! + ∑ ni ln + 1 (ii)
i ni
Again for most probable state, Ω or ln Ω is maximum and then we have
Ei
ni = e−α gi e− kT (iii)
[see equation (3.1.7)].
ni gi Ei
 
E
− α+ kTi
∴ =e ⇒ ln = α+ (iv)
gi ni kT
Using (iv) in (ii)
 
Ei
ln Ω = ln N! + ∑ ni +α+1
i kT
1
= ln N! +
kT ∑ ni Ei + α ∑ ni + ∑ ni
i i i
98 • An Introduction to Equilibrium Statistical Mechanics
  3 
3 2πmkT 2
SB = Nk + Nk ln V .
2 h20
Hence, the total entropy is simply
  3 
2πmkT 2
SA + SB = 3Nk + 2Nk ln V (ix)
h20
We now remove the partition very slowly so that the gases diffuse into each other. Therefore, the
total number of molecules 2N of the two gases occupy a volume 2V . Since there is no change in
pressure and temperature, the entropy of the system after the partition is removed is
  3 
2πmkT 2
S = 3Nk + 2Nk ln (2V ) (x)
h20
Now, according to the additive property of entropy
S = SA + SB (xi)
But from (ix) and (x), we see that
S − [SA + SB ] = 2Nk ln 2 > 0. (xii)
which is not equal to zero as required by (xi).
This paradox was first discussed by Gibbs and is commonly referred to as the ‘Gibbs paradox’.
The root of the difficulty embodied in the Gibbs paradox is that we treated the gas molecules as
distinguishable, as though interchanging the positions of two molecules would lead to a physically
distinct state of the gas. This is not so. Indeed, if we treated the gas by quantum mechanics,
the molecules would have to be regarded as indistinguishable (see Chapter 3). In this case, N!
permutations of the molecules among themselves do not lead to physically distinct situations, so
that we should subtract the term k ln N! from (vi), so that
  3 
5 V 2πmkT 2
S = Nk + Nk ln (xiii)
2 N h2
This has been verified experimentally at high temperatures and is known as Sackur-Tetrode
equation.
Equation (xii) can be written as
  3 
V 2πmkT 2 5
S = Nk ln e2 (xiv)
N h2
Using this expression, it can be easily verified that SA + SB − S = 0 and hence the paradox is re-
moved.

P ROBLEM 2.22: Imagine that a system R1 has probability Pr1 of being found in a state of r and
a system R2 has probability Ps2 of being found in a state s. Then one has S1 = −k ∑ Pr1 ln Pr1 and
r
 Elements of Ensemble Theory • 99

S2 = −k ∑ Ps2 ln Ps2 . Each state of the composite system R (= R1 + R2 ) can be labelled by a pair of
r
numbers r, s. Let the probability of R being found in this state be denoted by Prs . Then the entropy
is defined by S = −k ∑ ∑ Prs ln Prs . If R1 and R2 are weakly interacting so that they are statistically
r s
independent, then Prs = Pr1 Ps2 . Show that under these circumstances, the entropy is simply additive,
i.e., S = S1 + S2 .

S OLUTION
We have
S = −k ∑ ∑ Prs ln Prs = −k ∑ ∑ Pr1 Ps2 ln Pr1 Ps2 Prs = Pr1 Ps2
   
∵
r s r s
     
= ∑ Ps1 −k ∑ Pr1 ln Pr1 + ∑ Pr1 −k ∑ Ps2 ln Ps2
s r r s
   
= −k ∑ Pr ln Pr + −k ∑ Ps ln Ps
1 1 2 2
∵ ∑ Pr1 =1=∑ Ps2
r s r s

∴ S = S1 + S2 (Proved)

P ROBLEM 2.23: Suppose that a system R1 has probability Pr1 of being found in a state r and a
system R2 has probability Ps2 of being found in a state s. The entropy of R1 is S1 = −k ∑ Pr1 ln Pr1 and
r
that of R2 is S2 = −k ∑ Ps2 ln Ps2 . Each state of the composite system R(= R1 + R2 ) is then labelled by
s
the pair numbers r, s. Let the probability of R being found in this state be Prs ; its entropy is defined
by s = −k ∑ ∑ Prs ln Prs . Assume that R1 and R2 are not weakly interacting, so that Prs = Pr1 Ps2 . Of
r s
course Pr1 = ∑ Prs and Ps2 = ∑ Prs . Moreover, all the probabilities are properly normalized so that
r r
∑ Pr1 = 1, ∑ Ps2 = 1, ∑ ∑ Prs = 1. Show that
r r r s  1 2
(a) S − (s1 + S2 ) = k ∑ Prs ln PPr rsPs , and (b) S ≤ S1 + S2 .
r,s

S OLUTION
(a) We have
Pr1 = ∑ Prs (i)

p2s = ∑ Prs (ii)

S1 = −k ∑ Pr1 ln Pr1 (iii)

r

S2 = −k ∑ Pr2 ln Ps2 (iv)

s
100 • An Introduction to Equilibrium Statistical Mechanics

Using (i) in (iii)

S1 = −k ∑ ∑ Prs ln Pr1 (v)
r s
Again using (ii) in (iv)
S2 = −k ∑ ∑ Prs ln Ps2 (vi)
r s
Adding (v) and (vi)
S1 + S2 = −k ∑ ∑ ln(Pr1 Ps2 ) (vii)
r s
Again according to the question
S = −k ∑ ∑ Prs ln Prs (viii)
r s
[(viii) - (vii)] gives
Pr1 Ps2
S − (S1 + S2 ) = k ∑ + ∑ Prs ln (Proved) (ix)
1 s Prs
(b) Using the inequality, ln x ≤ x − 1
Pr1 Ps2
 1 2 
Pr Ps
ln ≤ −1
Prs Prs
Pr1 Ps2
 1 2 
Pr Ps
⇒ Prs ln ≤ Prs − 1 = (Pr1 Ps2 − Prs ) ≤ 0
Prs Prs
P 1 P2
⇒ k ∑ ∑ Prs ln r s ≤ 0
r s Prs
⇒ [S − (S1 + S2 )] ≤ 0

∴ S ≤ (S1 + S2 ) (x)

P ROBLEM 2.24: Consider a system of N very weakly interacting particles at a temperature T

sufficiently high so that classical statistical mechanics is applicable. Each particle has mass m and
is free to perform one-dimensional oscillation about its equilibrium position. Calculate the heat
capacity of this system, if the restoring force is proportional to x3 .

S OLUTION
The restoring force is given by
F = −Kx3 (i)
where K is the constant of proportionality.
Hence, the potential energy of a particle is
1
V = Kx3 (ii)
4
Therefore, the total energy of a particle is
p2 1
ε= + kx3 (iii)
2m 4
102 • An Introduction to Equilibrium Statistical Mechanics

L 0 Y
L

Now

L
mgz kT  mgL

e− kT dz = 1 − e− kT
mg
0

and
L  2 
− mgz kT mgL
 kT L mgL
ze kT dz = 1 − e− kT − e− kT
mg mg
0
mgL
 kT L mgL
2 
kT
1 − e− kT − e− kT mgL
mg mg e− kT
∴ mgz = mg × = kT − mgL
kT  − mgL
 mgL
1 − e− kT
1−e kT
mg
 
mgL −1
⇒ mgz = kT + mgL 1 − e kT (Ans.)
(ii)

The average kinetic energy of a molecule is given by

 3 
∂ 2πm 2
p2 p2 p2
    2
− βp−
2m +mgz
d 3 pd 3 r d3 p
−β

p2
 e e 2m
∂β β 3 3
= 2m = 2m 2 = = = kT (iii)
2m p2  23 2β 2
  
βp 
e
−β 2m +mgz 3
d pd r 3 e− 2m d 3 p 2πm
β

P ROBLEM 2.26: Calculate the rotational partition function for heteronuclear and homonuclear
molecules separately.
 Elements of Ensemble Theory • 103

S OLUTION
The rotation of a diatomic molecule is specified by the angles (θ, φ) and the corresponding
momenta (Pθ , Pφ ) and its kinetic energy assumes the form
Pθ2 Pφ2
εrot = + .
2I 2I sin2 θ
For heteronuclear diatomic molecule:
1

ξrot
heteronuclear = e−3/kT ld pθ d pφ dθdφ
h2  
2π ∞ π ∞ p2
1 p20
− φ
= dφ e− 2IkT
d pθ dθ  e 2IkT sin2 θ d pφ
h2
φ=0 pφ=0 θ=0 pφ=0
π
1 1√ 1√
= × 2π × 2πIkT × dθ 2πIkT sin θ
h2 2 2
θ=0
2π2 IkT
∴ ξrot
heteronuclear = (i)
h2
For homonuclear diatomic molecule:
For homonuclear diatomic molecule, φ ranges from 0 to π (see Section 4.12.2).
π2 IkT
∴ ξrot
homonuclear =
h2
P ROBLEM 2.27: Study the thermodynamics of a system of N non-interacting diatomic molecules,
each having an electric dipole moment µ, placed in an external electric field E. Assume that (i) the
system is classical and (ii) | µE | kT , where T is the absolute temperature of the system.

S OLUTION
The energy of a diatomic molecule in the presence of the electric field E is given by
 
P2 Pθ2 Pφ2
ε= + + − µE cos θ (i)
2m 2I 2I sin2 θ
where I is the moment of inertia of the molecule.
Now the single particle canonical partition function is given by
1

ξ= e−βε dr dθ dφ d p d pθ d pφ (ii)
h3  
1
V 3
     
2π ∞
1  2
 2
− βP − βPθ
ξ= dr ×  dφ ×  d pe ∞d pθ e
 2m × 2m
⇒ 
h3

r=0 φ=0 P=0 Pθ =0
104 • An Introduction to Equilibrium Statistical Mechanics
  
∞  ∞ βp2φ 
dθeβµE cos θ
−
× d pφe 2I sin2 θ 
 
θ=0 Pφ=0
       ∞  
1 1 1 2mπ 1 2Iπ 1 2Iπ
= 3 V 3 × 2π × × × dθeβµE cos θ · sin θ
h 2 β 2 β 2 β
θ=0
√ 5 π
2mIπ 21
= 3
· 3 · dθ sin θeβµE cos θ
2h β2 0
√ 5
2mIπ 2 1 1 βµE βµE

= · · · e − e
2h3 β 2 βµE
3

√ 5
2mIπ 2 1
∴ ξ= sinh (βµE) (iii)
µEh3 β 52
Therefore, the canonical partition function of the system is
√ 5

N 
N
2mIπ 2 sinh βµE
Z = ξN = (iv)
µEH 3 β 52
√ 5
 
2mIπ 2 5
⇒ ln Z = N ln + N ln sinh βµE − ln β (v)
µEh3 2

The mean energy of the system is given by

 
∂ ln Z 5 N 5 µE
ε=− = − NµE coth (βµE) = NkT − NµE sinh (vi)
∂β 2β 2 kT
The specific heat at constant volume is given by
   2  
∂ε 5 µE 2 µE
CV = = Nk − Nk cosec h (vii)
∂T V 2 kT kT
 
µE
For T → 0, i.e., x = →∞
kT
 
5 Lt
Cv ∼
= Nk ∵ cosec hx = 0 (viii)
2 x→∞
Now for µE/kT  1, i.e., for high temperature

1 µE 2
   
2 µE 1 1
cosec h ∼
=  − + (ix)
kT µE 3 15 kT
kT
106 • An Introduction to Equilibrium Statistical Mechanics

and
∞ 1
∞
) = e− 2u
∑ 3 = ∑ n2 e−u(n+ 2
∑ n2 e−ν
n=0 n=0
u 
= e− 2 e−u + 22 e−2u + 32 e−3u + · · ·

u
= e− 2 · e−u 1 + 22 e−u + 32 e−2u + · · ·
 

3u (1 + e−u )
= e− 2 · (iv)
(1 − e−u )3
Substituting (ii), (iii) and (iv) in (i)
u
  
e− 2 e−u (1 + e−u ) e−u 1
ξ= 1 + xu + +
1 − e−u (1 − e−u )2 (1 − e−u ) 4
u    (v)
e2 2e−u 1
= 1 + xu +
1 − e−u (1 − e−u ) 4
Therefore, the canonical partition function for the system is given by
Z = ξN
   
u   2e−u 1
⇒ ln Z = N ln ξ = N − − ln 1 − e−u + ln 1 − xu + (vi)
2 (1 − e−u )2 4
Then the internal energy of the system is
∂ ln Z ∂ ln Z ∂ ln Z
E =− = kT 2 = −hv
∂β ∂T ∂u
   2 
2
e−u ue−u 2u(e−u )
x 2 1−e−u − −
 1 e−u (1−e−u )2 (1−e−u )3 + 41 
= −Nhv  − −
 2 1 − e−u +   

2e−u
1 + xu −u 2 1
(1−e ) +4

= E0 + Ecorrection (vii)
where
1 Nhv 1 Nhv
E0 = Nhv + u = Nhv + hv (viii)
2 e −1 2 e kT − 1
The corresponding specific heat at constant volume is
 2 hv
∂E0 hv e kT
(CV )0 = = Nk  hv 2 (ix)
∂T V kT
e kT 1
  2 2

e−u ue−u 2u(e−u )
 2 1−e−u − (1−e−u )2 − (1−e−u )3 + 41 
Ecorrection = −xNhv     (x)
2e−u 1
1 − xu
 
−u 2
+4
(1−e )
 Elements of Ensemble Theory • 107

Upto the first order in x, (x) reduces to

 
2
2e−u 2ue−u 4ue (e−u )
Ecorrection = −xNhv − − (xi)
(1 − e−u )2 (1 − e−u )2 (1 − e−u )3 + 14
The corresponding specific heat is
2 2
2e−u 12u (e−u ) 4 (e−u )
(Cv )correction = xNku2 + − (xii)
(1 − e−u )2 (1 − e−u )4 (1 − e−u )3 − 4e−u
(1−e−u )3
This is the correction to the specific heat upto the first order in x.
hv
Comment: For u = kT  1, i.e., for high temperature,
4Nk2
 
4 4kT
(CV )correction = xNk × = xNk × =x T ∝ T,
u hv hv
which shows that the correction due to the anharmonicity increases with temperature, T .

P ROBLEM 2.29: Calculate the canonical partition function and hence the specific heat at constant
volume for a classical system of N non-interacting diatomic molecules enclosed in a box of volume
V at temperature T .
S OLUTION
The energy of a diatomic molecule can be written as
1  2  1
E (p1 ,p2 ,r1 ,r2 ) = p1 + p22 + K |r1 −r2 |2 (i)
2m 2
where p1 ,p2 ,r1 and r2 are the momenta and position coordinates of the two atoms in a diatomic
molecule.
r1 + r2
Let R = and ρ =r1 −r2 .
2
Then the Jacobian is  
 ∂r1 ∂r2   
 ∂R ∂R   1 1 
J =  ∂r ∂r  =  1 1  = −1 ⇒| J |= 1 (ii)
   
1
 ∂ρ ∂ρ 2
2 − 2

∴ d 3 r1 d 3 r2 = d 3 Rd 3 ρ (iii)
Now the single particle canonical partition function is
 
ξ= e−βH d 3 r1 d 3 r2 d 3 p1 d 3 p2 = e−βH d 3 r1 d 3 r2 d 3 Rd 3 ρ
       
−βp21 3 −βp22 3 3 −βkp22 3
= e d p1 e d p2 d R e d ρ
 3  3   32
2πm 2 2πm 2 2π
= × ×V ×
β β βK
 3
8π3 m2 2 1
=V × 9
K β2
108 • An Introduction to Equilibrium Statistical Mechanics

Therefore, the canonical partition function for the system is

 3 2  3N2
N N 8π m 1
Z =ξ =V × 9N (iv)
K β2
 
3 2
 3N2 
8π m 9N
⇒ ln Z = ln V N − ln β (v)
K 2
The mean energy of the system is
∂ ln Z 9N 9
E =−
= = NkT (vi)
∂β 2β 2
Hence, the specific heat at constant volume is
 
∂E 9
CV = = Nk (vii)
∂T V 2

P ROBLEM 2.30: In Problem 2.31, calculate the mean square molecular diameter, |r1 −r2 |2 .

S OLUTION
The mean square molecular diameter is
|r1 −r2 |2 e−βE d 3 r1 d 3 r2 d 3 p1 d 3 p2

2
|r1 −r2 |  = 
e−βE d 3 r1 d 3 r2 d 3 p1 d 3 p2
 2 −β[ 1 ( p2 +p2 )+ 1 Kρ2 ] 3
ρ e 2m 1 2 2 d r1 d 3 r2 d 3 p1 d 3 p2
= 1 2 2 1 2
e−β[ 2m ( p1 +p2 )+ 2 Kρ ] d 3 r1 d 3 r2 d 3 p1 d 3 p2
∞ 4 − 1 βKρ2 ∞ 3 e−x dx
 2 − 1 βKρ2 3 ρ e 2 dρ x2
2kT Γ 52
 
ρ e 2 d ρ 0 2kT 0 3kT
=  1 2 = ∞ = ∞ = 3 =
e 2
− βKρ d3ρ 1 2 K 3 −x K Γ 2 K
ρ2 e− 2 βKρ dρ x 2 e dx

0 0
2
|r1 −r2 |  ∞ T.

P ROBLEM 2.31: Consider a classical system of N non-interacting diatomic molecules enclosed

in a box of volume V and temperature T . The energy for a single molecule is E (p1 ,p2 ,r1 ,r2 ) =
1 2 2
2m p1 + p2 + ε|r12 − r0 |, where  p1 ,p2 ,r1 andr2 are the momenta and position coordinates of the
two atoms in a molecule. ε and r0 are given positive constants and r12 = |r1 −r2 |. Find the specific
heat of the system at a constant volume as a function of temperature.
 Elements of Ensemble Theory • 109

S OLUTION
The energy of a molecule is
1  2
p1 + p22 + ε|r12 − r0 |

E (p1 ,p2 ,r1 ,r2 ) = (i)
2m
r1 +r2
Let R = and ρ =r1 −r2 .
2
Then the Jacobian is
 ∂r1 ∂r2 
 
 
 ∂R ∂R   
  1 1 
J= = −1 ⇒| J |= 1 (ii)

=
 ∂r1 ∂r2   21 − 21 
 
 ∂ρ ∂ρ 
∴ d 3 r1 d 3 r2 = d 3 Rd 3 ρ (iii)
The single particle canonical partition function is
 
−βH 3
ξ= e d r1 d r2 d p1 d p2 = e−βH d 3 r1 d 3 r2 d 3 Rd 3 ρ
3 3 3

       
βp2 βp2
− 2m1 3 − 2m2 3 3 −βε|ρ−r0 | 3
= e d p1 × e d p2 × d R × ρ d ρ (iv)
 
∞
  3      32  
2m 2 1 3 2m 1 3 4π
= 4π τ × 4π τ ×V ×  y2 e−|y−x| dy
β 2 2 β 2 2 (βε)3
0

where y = βερ and x = βεr0 .

Now   
y2 e−y dy = y2 ey − 2yey + 2 ey dy = y2 ey − 2yey + 2ey (v)
Similarly,

y2 e−y dy = −y2 e−y − 2ye−y 2e−y (vi)
∞
∴ y2 e−y dy = x2 e−x + 2xe−y 2e−x (vii)
x
Again
∞ x ∞ ∞ ∞
2 −|y−x| 2 −(y−x) 2 −(y−x) 2 −y
y e dy = y e dy + y e dy = e
−x
y e dy + e x
y2 e−y dy
0 0 0 x x
= e x2 ex − 2xex + 2ex − 2 + ex x2 e−x + 2xe−x + 2e−x
x
   
(viii)
∞
∴ y2 e−|y−x| dy = 2x2 + 4 − 2e−x (ix)
0
 Elements of Ensemble Theory • 111

Now, when the cylinder is rotated, it becomes energetically favourable for the molecules to move
towards the edge of the cylinder, so we expect the density to increase with radius r.
In a frame of reference rotating with a gas, a molecule of mass m at a distance r from the axis
experiences a centrifugal force
Fc = mω2 r (ii)
away from the axis.
∴ The centrifugal potential energy is
1
Vc = − mω2 r2 (iii)
2
From
 the Boltzmann distribution law, we expect the probability of a molecule to be proportional to
Vc
exp − kT and so we can write
mω2 r
 
σ(r) = A exp (iv)
2kT
where A is a constant of proportionality.
Now the total mass of the gas within the cylinder is given by
a a  2 2
mw r
M = 2πL σ(r)rdr = 2πLA r exp dr (v)
2kT
0 0
The integral can be easily evaluated by noting that
mω2 r2 mω2 r2 mω2
    
d
exp = exp (vi)
dr 2kT 2kT r2
mω2 a2
   
2πLAkT
∴ M= exp −1 (vii)
mω2 2kT
Equating (i) and (vii), we have the expression for A:
σ0 mω2 a2 1
A= (viii)
2kT mω2 a2
   
exp −1
2kT
Using (viii), we get the expression for σ(r) as
mω2 a2
 
exp
σ0 mω2 a2 2kT
σ(r) =     2 2  2   (ix)
2kT mω exp mω
2kT
a
−1 a2
exp 2kT −1

r mω2 a2
Let R = and Ω = .
a 2kT
σ(R) exp(ΩR2 )
∴ =Ω (x)
σ0 [exp(Ω) − 1]
112 • An Introduction to Equilibrium Statistical Mechanics

Writing the function as a series expansion, we have

1 + ΩR2 + 2!1 Ω2 R4 + . . . 1 + Ω + 2!1 Ω2 + . . .
   
σ(R)
=Ω = (xi)
Ω + 2!1 (Ω2 + . . . )
 
σ0 1 + +Ω
2 +...
For small ω (i.e., for small Ω), we can write (xi) approximately as
 
σ(R) ∼ 1 + ΩR2
=   (xii)
σ0 1 + Ω2
2
which shows that σ(R) = σ0 , i.e., the density remains unchanged when R2 = 21 ⇒ ar 2 = 12 ⇒ r = √a2 .
Thus for small ω, the density remains unchanged at a distance √a2 from the axis of the cylinder.

P ROBLEM 2.33: Consider an ideal gas at absolute temperature T in a uniform gravitational field
described by acceleration g. By writing the condition of hydrostatic equilibrium for a slice of the
gas located between heights z and z + dz, derive an expression for n(z), the number of molecules per
cm3 at height z. Also show that this result is identical with that derived from statistical mechanics.

S OLUTION
Let us consider a slice between heights z and z + dz. Let the pressure at z be p and that at z + dz
be p + d p.

z + dz, p + dp

z, p

Then for hydrostatic equilibrium,

pA − (p + d p)A = mngAdz (i)
where, A = cross-sectional area of the slice
n = number density
m = particle mass
g = acceleration due to gravity.
∴ d p = −mngdz (ii)
Again treating the gas as an ideal gas, we can write
p = nkT ⇒ d p = kT dn (iii)
Using (iii) in (ii), we have
dn mg
= − dz
n kT
 Elements of Ensemble Theory • 113

On integration
mg
ln n = − z + constant (iv)
kT
Let at z = 0, n = n(0).
mgz
∴ n = n(0)e− kT (v)
Using ideal gas law: p = nkT, p(0) = n(0)kT
mgz
p = p(0)e− kT (vi)
Now the probability P (r,p) d 3 rd 3 p that the molecule has position lying in the range betweenr and
r + dr and momentum in the range between p and p + dp is given by
d 3 rd 3 p −β 2m
p2
 
3 3 +mgz
P (r,p) d rd p ∝ 3
e (vii)
h
Therefore, the probability P (z) dz that a molecule is located at a height between z and z + dz is
   
∞ π 2π
1 

P2
P (z) dz ∝ dx dy ×  e −β 2m
P2 sin θ d p dθ dφ × e−βmgz dz (viii)

×
h3
(x,y) P=0 θ=0 φ=0

But (x,y) dx dt = A, cross-sectional area of the slice.
3/
2

A 2πm mg
∴ P (z) dz = constant × 3 e− kT z (ix)
h β
Let at z = 0, P(z)|z=0 = P(0).
 
A 2πm
∴ P(0) = constant × 3 (x)
h β
− mg z
⇒ P(z) = P(0)e kT (xi)
Thus, the probability of finding a molecule at height z decreases exponentially with height.
Now mgz
n(z) dz ∝ p(z) dz = constant × P(0)e− kT dz (xii)
At z = 0, n = n(0) ⇒ constant × P(0) = n(0)
mgz
∴ n(z) = n(0)e− kT (xiii)
which is same as (v).

P ROBLEM 2.34: Consider a system consists of N weakly interacting particles, each of which can
be in either of two states with respective energies ε and ε2 (ε2 > ε1 ). Calculate explicitly the mean
energy E and heat capacity CV of this system.
 Elements of Ensemble Theory • 115

S OLUTION
Consider a pair of one +ve and one –ve particle.

+ +
a
+ +
a

a
2
+ +

+ +

The adjoining figure shows that there are 4 possible states of this pair. Since the potential energy
of a charged particle of charge in an external electric field is V (x) = −eεx, the average separation
between a +ve ion and a –ve ion is given by
a a
2 · 2a eβ 2 εe − 2 · 2a e−β 2 εe a βeεa
x = a a = tanh (i)
2e β 2 εe + 2e −β 2 εe 2 2
If there are N number of pairs per unit volume, the electric polarization or average dipole moment
per unit volume is given by
a βeεa
µ = Ne x = Ne tanh (ii)
2 2
P ROBLEM 2.36: Calculate the canonical partition function and hence the specific heat of a classical
ideal two-dimensional gas.
S OLUTION
The energy of a single particle is
p2x + p2y
ε= (i)
2m
Then the single particle canonical partition function is
 ∞   ∞ 
1 A βp2

β βp2
( p2x +p2y ) − 2mx − 2my
ξ= 2 e − 2m
dx dy d px d py = 2 e d px × e d py
h h −∞ −∞
 
A 2πm
= 2 (ii)
h β
where A is the area.
116 • An Introduction to Equilibrium Statistical Mechanics

Hence, the canonical partition function of the system is given by

AN 2mπ N
 
Z = 3N (iii)
h β
The mean energy of the system is
 
∂ ln Z
E =− = NkT
∂β V

∴ The specific heat at constant volume is

 
∂E
CV = = Nk (iv)
∂T V

P ROBLEM 2.37: Suppose that a wire of radius r0 is placed along the axis of a metal cylinder of
radius R and length L. The wire is maintained at a positive potential V with respect to the cylinder
and the whole system is at some high absolute temperature T . As a result, electrons emitted from
the hot metals form a dilute gas filling the cylindrical container and in equilibrium with it. The
density of these electrons is so low that their mutual electrostatic interaction can be neglected.
Usine Gauss’ theorem, obtain an expression for the electrostatic potential at a radial distance
r from the wire (r0 < r < R). [Assume the cylinder to be infinitely long so that end effects are
neglected.] Also find the density of the electron gas filling the space between the wire and cylinder,
as a function of the radial distance in equilibrium.

S OLUTION

V V=0

r0
L
R
 Elements of Ensemble Theory • 119

S OLUTION
In the rotating frame of reference, the molecules feel a centrifugal force mω2 r radially outward.
This force can be obtained from a potential
σV
Fr = mω2 rr̂ = − r̂ (i)
σr
where r̂ is the unit vector in the radial direction.
Therefore, the potential energy of a molecule in the rotating frame is
1
V = − mω2 r2 (ii)
2
so that the total energy of a molecule in the rotating frame is given by
p2
 
1 2 2
ε= + − mω r (iii)
2m 2
The probability of finding the molecule in the energy range ε to ε + dε is then given by
1 p2 mω2 r2
P(ε) = e 2mkT e 2kT (iv)
Z
Then the probability of finding the molecule between r and r + dr is obtained by integrating over
all the other coordinates and is given by
mω2 r2
P(r)dr = P(0)e 2kT dr (v)
where P(0) is a constant which correctly normalizes the distribution.
Equation (v) is the radial distribution of molecules. As expected, the molecules crowd towards
the larger radii. It should be noted that we have taken the rotating speed to be very high so that the
effect of the Coriolis force is ignored.
Now from (v), the concentration at r1 is
mω2 r12
σ(r1 ) ∝ e 2kT (vi)
Similarly,
mω2 r22
σ(r2 ) ∝ e 2kT

σ(r1 ) 2 2 2
1 mω (r1 −r2 )
∴ = e 2 kT (vii)
σ(r2 )  
2kT σ(r1 )
⇒ m = 2 2  ln
ω r1 − r22 σ(r2 )

P ROBLEM 2.40: In order to measure a certain physical quantity, it is necessary to know the
equilibrium θ0 of a torsional pendulum. θ0 is determined by the minimum in the potential energy
of the pendulum V (θ) = 21 K(θ − θ0 )2 . The pendulum, when rotating, has a kinetic energy T = 12 Iθ,
where θ = dθ/dt. The precision with which θ0 can be found is limited because the pendulum is in
thermal equilibrium with its environment at a temperature T .
 Elements of Ensemble Theory • 121

(b) The degeneracy of the nth energy level is n2 . The partition function for a single atom, ne-
glecting the unbound states, is
 ε  ∞ α
i
Z = ∑ exp − = ∑ n2 exp 2 (iii)
states kT n=1 n
where α = A kT . Since α > 0, it follows that exp α n2 > 1 for all n. Using this, we can set
   

the lower bound of Z, but Z diverges since the lower bound diverges.
∞
∴ Z> ∑ n2 , which diverges.
n=1

2.3 GRAND CANONICAL ENSEMBLE

It is a collection of large number of essentially independent systems, having same temperature T ,
volume V and chemical potential µ. Thus, the individual systems of a grand canonical ensemble
are separated by rigid, permeable and conducting walls.

µ,V,T µ,V,T µ,V,T

µ,V,T µ,V,T µ,V,T
µ,V,T µ,V,T µ,V,T
µ,V,T µ,V,T µ,V,T

Fig. 2.7. Grand canonical ensemble.

As the separating walls are conducting and permeable, the exchange of heat energy as well as
that of particles between the systems take place in such a way that all the systems arrive at common
temperature T and the chemical potential µ.

2.3.1 Probability distribution and grand canonical partition function

Let us consider a system R of fixed volume V in contact with a large reservoir R with which the
system can exchange not only energy, but particles also. We wish to find the probability Pr of
finding the system R in any one particular microstate r of energy Er .
Neither the energy Er of R nor the number of particles Nr in R are fixed, but the total energy
E and total number of particles N 0 of the combined system R0 = R + R are fixed, i.e.,
0

Er + E  = E 0 = constant
(2.3.1)
Nr + N  = N 0 = constant
where E  and N  denote the energy and number of particles in the reservoir R respectively.
122 • An Introduction to Equilibrium Statistical Mechanics

Let Ω (E  N  ) denote the number of states accessible to R , when it contains N  particles and has
an energy in the range near E  . Then if R is in a particular
 state r, the number of states accessible
to the combined system R0 is 1 × Ω (E  , N  ) = Ω E 0 − Er , N 0 − Nr . Hence, the probability of
finding the system R in this state is
 
1 × Ω E 0 − Er , N 0 − Nr
Pr = (2.3.2)
Ω0total
where Ω0total is the total number of states accessible to the combined system R0 . It is independent of
r and is just a constant. Writing Ω01 = c = constant,
total

Pr = c Ω E 0 − Er , N 0 − Nr
 
(2.3.3)
Taking logarithms on both sides
ln Pr = ln c + ln Ω E 0 − Er , N 0 − Nr
 
(2.3.4)
Since R  0is very 0small compared to R , Er  E 0 and Nr  N 0 , expanding ln Ω (E  , N  )
= ln Ω E − Er , N − Nr about E = E 0 and N  = N 0 , we have
 
   
 0 0
  0 0 ∂ ln Ω ∂ ln Ω
ln Ω E − Er , N − Nr = ln Ω E , N −
 
Er − Nr . . .
∂E  E 0 ∂N  N 0
Since Er  E 0 and Nr  N 0 , we can neglect the higher order terms and so we have
   
∂ ln Ω ∂ ln Ω
ln Ω E 0 − Er , N 0 − Nr = ln Ω E 0 , N 0 −
   
E r − Nr (2.3.5)
∂E  E 0 ∂N  N 0
The derivatives are evaluated
 for E  = E 0 and 0
 N = N and so they are constants, characterizing the


reservoir R . We write ∂ ln Ω = β and ∂ ln Ω = α.
∂E  E 0 ∂N 
Then equation (2.3.5) becomes
ln Ω E 0 − Er , N 0 − Nr = ln Ω E 0 , N 0 − βEr − αNr
   
(2.3.6)
 0 0
  0 0  −βE −αN
⇒ Ω E − Er , N − Nr = Ω E , N e r
  r

Using equation (2.3.6), equation (2.3.3) becomes

Pr = c Ω E 0 , N 0 e−βEr −αNr
 

⇒ Pr = ce−βEr −αNr (2.3.7)

 
where c = c Ω  E 0, N 0
, constant of proportionality, independent of r.
Equation (2.3.7) is called the grand canonical distribution. An ensemble of systems, dis-
tributed according to this probability distribution, is called a grand canonical ensemble.
The parameter β is the temperature parameter of the reservoir. β = 1/kT , where T is the
absolute temperature of the reservoir and µ = −kT α, the chemical potential of the reservoir.
∴ Pr = ceµβNr −βEr (2.3.8)
 Elements of Ensemble Theory • 123

The normalization condition for probabilities demands ∑ Pr = 1. Applying this to equation (2.3.8),
r
1
we have c = and so
∑ eµβNr −βEr
r

eµβNr −βEr eµβNr −βEr

Pr = = (2.3.9)
∑ eµβNr −βEr Z (µ,V, T )
r

where Z (µ,V, T ) = ∑ eµβNr −βEr −→ grand canonical partition function.

r

2.3.2 Density and energy fluctuations in grand canonical ensemble

In this case, fluctuations in both E and N are present.
Let N be the average number of particles of the system R. Then the mean square deviation in it
is given by
2 2
(∆N)2 = N − N = N 2 − N


Now
∑ Nr Pr ∑ Nr eµβNr −βEr
N= r
= ∑ Nr Pr = r
.
∑ Pr r ∑ eµβNr −βEr
r r
The grand canonical partition function is
1 ∂Z
Z (µ,V, T ) = ∑ eµβNr −βEr ⇒
β ∂µ ∑
= Nr eµβNr −βEr .
r r
1 1 ∂Z 1 ∂ ln Z
∴ N= =
β Z ∂µ β ∂µ
Again
∑ Nr2 Pr ∑ Nr2 eµβNr −βEr ∑ Nr2 eµβNr −βEr
1 1 ∂2 Z
N2 = r
=∑ Nr2 Pr = r
= r
=
∑ Pr r ∑ eµβNr −βEr Z (µ,V, T ) Z β2 ∂µ2
r r

1 1 ∂2 Z 1 1 ∂Z 2 1 ∂ 1 1 ∂Z
   
2 ∂N
Hence, (∆N) = 2 − = = kT .
β Z ∂µ2 β Z ∂µ β ∂µ β Z ∂µ ∂µ
∴ The root mean square fluctuation in N is
 1/2  ∂N 1/2
∆∗ N = (∆N)2 = kT .
∂µ T,V
The relative fluctuation in N is then
 1/2
∆∗ N 1 ∂N
= kT (2.3.10)
N N ∂µ T,V
 Elements of Ensemble Theory • 125

and
∑ Er2 eµβNr −βEr ∑ Er2 eµβNr −βEr
2 r r
E = =
∑ eµβNr −βEr Z (µ,V, T )
r

∂2 Z 1 ∂Z 2
    
1 ∂ 1 ∂Z
= = + 2
Z ∂β2 µ,V ∂β Z ∂β Z ∂β

∂ ln Z 2
   
∂ ∂ ln Z ∂E  2
= + =− + E
∂β ∂β ∂β ∂β
   
2 2 2 ∂E 2 ∂E
∴ (∆E) = E − E = − = kT
∂β µ,V ∂T µ,V
and    
1 ∂ ln Z∂ ln Z
N= =− (2.3.14)
β V,T ∂µ∂ (−µβ) V,T
 
Eliminating µ from (2.3.13) and (2.3.14), we see that E = E N,V, T , so that
           
∂E ∂E ∂E ∂N ∂E ∂N
= + × = CV + ×
∂T µ,V ∂T N,V ∂N T,V ∂T µ,V ∂N T,V ∂T µ,V
 
where CV = ∂E ∂T N,V , the specific heat of the system at constant volume.
     
∂E ∂N
∴ (∆E)2 = kT 2 CV +
∂N T,V ∂T µ,V
Again from equation (2.3.13)
       
∂E ∂N ∂N 1 ∂E
= and hence = .
∂ (−µβ) V,T ∂β µ,V ∂T µ,V T ∂µ T,V
     
2 2 1 ∂E ∂E
∴ (∆E) = kT CV + ×
T ∂N T,V ∂µ T,V
   
∂E ∂E
= kT 2CV + kT
∂N T,V ∂µ T,V
 2  
∂E ∂N
= kT 2CV + kT
∂N T,V ∂µ T,V
 
2 ∂N
But (∆N) = kT .
∂µ T,V
 2  2
2 ∂E 2 2∂E
∴ (∆E) = kT CV +2
(∆N) = (∆E) canonical + (∆N)2 (2.3.15)
∂N T,V ∂N T,V
126 • An Introduction to Equilibrium Statistical Mechanics

This is the desired result, which says that the mean square fluctuation in the energy E of a system
in the grand canonical ensemble is equal to the value it would have in the canonical ensemble plus
a contribution arising due to the fluctuation in the particle number. Under ordinary circumstances,
the relative fluctuation in the energy of the system is negligible. However, in the region of phase
transition, the fluctuation in the energy is large because of the second term.

2.3.3 Thermodynamic quantities in terms of the grand canonical partition function

1. Internal energy
This is given by    
∂ ln Z 2 ∂ ln Z
E =− = kT (2.3.16)
∂β µ,V ∂T µ,V

2. Specific heat
The specific heat at constant volume is given by
 
∂E
CV = (2.3.17)
∂T N,V
where E is given by (2.3.16).

3. Thermodynamic pressure
The grand canonical partition function is Z = Z (µ,V, T ). For the sake of convenience, let us write
Z as a function of α (= −µβ), V and β (= 1/kT ), i.e., Z = Z (α,V, β).
     
∂ ln Z ∂ ln Z ∂ ln Z
∴ d ln Z (µ,V, T ) = d ln Z (α,V, β) = dα + dV + dβ (2.3.18)
∂α ∂V ∂β
Now    
1 ∂ ln Z ∂ ln Z
N= =− (2.3.19)
β ∂µ V,T ∂α V,T
Using (2.3.16) and (2.3.19) in (2.3.18)
 
∂ ln Z
d ln Z = −Ndα + dV − Edβ
∂V  
  ∂ ln Z
⇒ d ln Z + Nα + Eβ = αdN + βdE + dV
∂V
      
Nµ E 1 ∂ ln Z
⇒ d k ln Z − + = dE + dV − µdN
kT kT T ∂V
To interpret the second term on the right-hand side of this equation, we compare the expression
enclosed within the parenthesis with the statement of the first law of thermodynamics, viz.,
δQ = dE + δW − µdN (2.3.20)
 Elements of Ensemble Theory • 127

We immediately identify (∂ ln Z/∂V ) dV as δW , the work done by the system.

  
δQ Nµ E
∴ = d k ln Z − + .
T kT kT
  
Nµ E
The quantity k ln Z − + is identified with the entropy S of the system. Accordingly,
kT kT
  
Nµ E
S = k ln Z − +
kT kT
S Nµ E T S + Nµ− E
⇒ ln Z =
+ − =
k kT kT kT
But
 Nµ = G, the
 Gibbs free energy of the system [see Problem (2.46)]. Again G is equal to
E − T S + PV . So finally
kT
P= ln Z (µ,V, T ) (2.3.21)
V
Thus, the grand canonical partition function directly gives the pressure as a function of µ,V
and T .

4. Helmholtz free energy

The Helmholtz free energy is given by
F ≡ Nµ− PV = NkT ln z − kT ln Z (µ,V, T )
where z = eµβ −→ fugacity.
Z (µ,V, T )
F = −kT ln (2.3.22)
zN
which can be compared with the canonical ensemble formula F = −kT ln Z (N,V, T ).

5. Entropy
The entropy of the system is given by
 
E −F ∂ ln Z (µ,V, T ) Z (µ,V, T )
S≡ = kT + k ln
T ∂T µ,V zN
 
∂ ln Z (µ,V, T )
∴ S = kT + k ln Z (µ,V, T ) − Nkµβ (2.3.23)
∂T µ,V

2.3.4 Equation of state

Eliminating µ between (2.3.19) and (2.3.21), one obtains the equation of state, i.e., the (P,V, T )-
relationship.

P ROBLEM 2.42: Show that

∂ ∂ ∂
(a) N = z ln Z(z,V, T ), (b) N 2  − N2 = z z ln Z(z,V, T )
∂z ∂z ∂z
where z = eµβ , known as fugacity.
130 • An Introduction to Equilibrium Statistical Mechanics

where g(T, P) is independent of N j .

 
∂G G
∴ µj = = g(T, P) = (Proved) (xi)
∂N j T,P Nj
i.e., the chemical potential per molecule is just equal to the Gibbs free energy per molecule.
When several species are present, then G = G (T, p, N1 , . . . Nm ) and in general
 
∂G G
µj = = .
∂N j T,P,{N1 ...Ni−1 ...Ni+1 ...Nm } N j

P ROBLEM 2.44: Show that the entropy of a grand canonical ensemble can be written as
S = −k ∑ Pr ln Pr .
r

S OLUTION
The grand canonical partition function is
Z(µ,V, T ) = ∑ eµβNr −βEr (i)
r

and the probability of finding the system in the rth microstate is

eµβNr −βEr eµβNr −βEr
Pr = = (ii)
∑ eµβNr −βEr Z(µ,V, T )
r
∴ ln Pr = − ln Z + µβNr − βEr
∑ Pr ln Pr ∑ Pr ∑ Nr Pr ∑ Er Pr
⇒ ∑ Pr ln Pr = r
∑ Pr
= − ln Z r
∑ Pr
+ µβ r
∑ Pr
−β r
∑ Pr
r
r r r r
⇒ ∑ Pr ln Pr = µβN − βE − ln Z (iii)
r

Let us now introduce a new variable α = µβ.

∴ Z(α,V, β) = ∑ eαNr −βEr (iv)
r

Now
∂ ln Z ∂ ln Z ∂ ln Z
d ln Z = dα + dV + dβ = Ndα + pβdV + βdE
∂α ∂V ∂β
 
⇒ d ln Z − Nα + Eβ = −αdN + pβdV + βdE
 
  α  
⇒ d ln Z − Nα + Eβ = β dE + pdV − dN = β dE + pdV − µdN
β
 Elements of Ensemble Theory • 131

But T dS = dE + pdV − µdN.

dS
∴ d(lnZ − Nα + Eβ) =
k
whence
S
= (ln Z − Nµβ + Eβ)
k (v)
S
⇒ ln Z = + Nµβ − Eβ
k
Using (v) in (iii), we finally get
S = −k ∑ Pr ln Pr (Proved) (vi)
r

P ROBLEM 2.45: Using S = −k ∑ Pr ln Pr , show that for a grand canonical ensemble, T dS = dE +

r
pdV − µdN.

S OLUTION
We know that
S = −k ∑ Pr ln Pr (i)
r
Now
eµβNr −βEr
Pr = ⇒ ln Pr = µβNr − βEr − ln Z (ii)
Z (µ,V, T )
Using (ii) in (i)
 
S = −k µβ ∑ Pr Nr − β ∑ Pr Er − ln Z ∑ Pr
r r r
 
∵ ∑ Pr = 1, ∑ Pr Er = E, ∑ Pr Nr = N
 
= −k µβN − βE − ln Z
r r r
= k ln Z + kβE − kµβN
∂ ln Z ∂ ln Z ∂ ln Z
⇒ dS = k dµ+ k dV + k dT + kβdE + kEdβ − kβµdN − kβNdµ− kµNdβ
∂µ ∂V ∂T
= kβNdµ+ kβpdV + kµNdβ − kEdβ + kβdE + kEdβ − kβµdN − kβNdµ− kµNdβ
 
∂ ln Z ∂ ln Z ∂ ln Z ∂ ln Z ∂ ln Z
∵ dT = dβ, = βN, = βp, = µN − E
∂T ∂β ∂µ ∂V ∂β
= kβdE + kβpdV − kβµdN
 
1
⇒ T dS = dE + pdV − µdN ∵ β=
kT
 Elements of Ensemble Theory • 133

Taking logarithms on both sides

α    
V V
ln Z (µ,V, T ) = ∑ ln Z µ, , T = α · ln Z µ, , T
i=1 α α
 
V 1
⇒ ln Z µ, , T = · ln Z (µV, T )
α α
This shows that ln Z (µ,V, T ) is linear in V .

P ROBLEM 2.47: Show that Z (µ,V, T ) = eN for an ideal gas, if the gas molecules are treated as
indistinguishable.

S OLUTION
The grand canonical partition function is Z (µ,V, T ) = ∑ eµβNr −βEr . We can write this partition
r
function in another form:
∞
Z (µ,V, T ) = ∑ eµβN−βENr = ∑ eµβN e−βENr = ∑ eµβN Z (N,V, T )
Nr Nr N=0

where Z (N,V, T ) = ∑ e−βENr −→ canonical partition function.

r
For an ideal gas, the canonical partition function is Z (N,V, T ) = ξN , where ξ is the single
 3/2
particle canonical partition function and is given by ξ = hV3 2πm
β .
∞
∴ Z (µ,V, T ) = ∑ eµβN ξN (i)
N=0
If the gas molecules are treated as indistinguishable, then N! permutations of the molecules among
themselves do not lead to physically distinct situations and hence equation (i) must be divided by
N!.
∞ ∞  N
∑ eµβN ξN ∑ eµβ ξ  
N=0 N=0
∴ Z (µ,V, T ) = = = exp eµβ ξ (ii)
N! N!
Again, we know that
  
1 ∂ ln Z (µ,V, T ) 1 ∂ eµβ ξ
N= = = eµβ ξ. (iii)
β ∂µ β ∂µ
Using (iii) in (ii)
Z(µ,V, T ) = eN (Proved)

P ROBLEM 2.48: Consider a system of N molecules in a container of volume V . Assuming that

there is no correlation whatsoever between the locations of the various molecules, calculate the
probability that a region of volume v contains exactly n molecules.
134 • An Introduction to Equilibrium Statistical Mechanics

S OLUTION
The probability that the region of volume v exactly contains n molecules is given by
∑ eµβn−βEnr eµβn ∑ e−βEnr
r r eµβn Z (n, v, T )
Pn = = = (i)
Z (µ, v, T ) Z (µ, v, T ) Z (µ, v, T )
where Z(n, v, T ) = canonical partition function = ξn /n!, ξ being the single particle canonical parti-
tion function. n! appears because of the essential indistinguishability of the particles. Again, from
problem 2.50, n = eµβ ξ and Z (µ, v, T ) = en .
(eµβ ξ)n / n! (n)n −n
∴ Pn = en
= n! e . This is known as Poisson’s distribution.

P ROBLEM 2.49: Consider a system of N molecules forming a substance which consists of two
phases. The whole system is isolated so that its total energy E and its total volume V are both fixed.
Find the necessary conditions for equilibrium between the two phases.
S OLUTION
Let the two phases be 1 and 2. The whole system is isolated, so that its total energy E, its total
volume V and its total number of particles N are fixed. Let Ni be the number of molecules of the
substance in phase i and denote the energy of this phase by Ei and its volume by Vi .
Then we must have the conservation conditions:
E1 + E2 = E = constant
⇒ dE1 = −dE2 (i)
V1 +V2 = V = constant
⇒ dV1 = −dV2 (ii)
N1 + N2 = N = constant
⇒ dN1 = −dN2 (iii)
The equilibrium condition corresponds to the most probable situation, i.e., when the thermodynamic
probability Ω is maximum. But S = k ln Ω. Hence, the equilibrium condition is that the entropy is
maximum, i.e.,
S = S (E1 ,V1 , N1 ; E2 ,V2 , N2 ) = maximum (iv)
Again S = S1 (E1 ,V1 , N1 ) + S2 (E2 ,V2 , N2 ).
Thus, the maximum condition (iv) yields
dS = dS1 + dS2 = 0 (v)
Now S1 = S1 (E1 ,V1 , N1 ).
     
∂S1 ∂S1 ∂S1
∴ dS1 = dE1 + dV1 + dN1 (vi)
∂E1 V1 ,N1 ∂V1 E1 ,N1 ∂N1 E1 ,V1
In a simple case, when the number N1 is fixed, the fundamental thermodynamic relation asserts that
1
dS1 = (dE1 + p1 dV1 ) (vii)
T1
 Elements of Ensemble Theory • 135

where T1 and p1 are respectively the temperature and mean pressure of phase 1. Under this circum-
stance, dN1 = 0. A comparison of the coefficients of dE and dV in (vi) and (vii) yields
 
∂S1 1
= (viii)
∂E1 V,N1 T1
 
∂S1 p1
= (ix)
∂V1 E1 ,N T1
 
∂S1
Let µ1 = −T1 ∂N1 . The quantity µ1 is the chemical potential per molecule in phase 1.
E1 ,V1
Then (vi) can be written in the form

1 p1 µ1
dS1 = dE1 + dV1 − dN1 (x)
T1 T1 T1
Similarly,

1 p2 µ2 1 p2 µ2
dS2 = dE2 + dV2 − dN2 = − dE1 − dV1 + dN1 (xi)
T2 T2 T2 T2 T2 T2
Using (x) and (xi) in (v)
     
1 1 p1 p2 µ1 µ2
− dE1 + − dV1 − − dN1 = 0 (xii)
T1 T2 T1 T2 T1 T2

Since (xii) is to be valid for arbitrary variations dE1 , dV1 , dN1 , it follows that the coefficients of all
these differentials must vanish separately. Thus, one obtains
 
1 1
− =0
T1 T2
 
p1 p2
− =0
T1 T2
 
µ1 µ2
− =0
T1 T2

whence
T1 = T2 (xiii)
p1 = p2 (xiv)
µ1 = µ2 (xv)

These are the necessary conditions for equilibrium between two phases. (xiii) is the condition for
thermal equilibrium, (xiv) for mechanical equilibrium and (xv) for diffusive equilibrium.
 Elements of Ensemble Theory • 137

where v is 2f-dimensional velocity of the representative points and n̂ is a unit vector, outward
normal to the point of location of the surface element ds of the surface s, enclosing the volume τ
and dτ = dq1 . . . dq f d p1 . . . d p f .
Applying Gauss’ divergence theorem
 
ρv · n̂ds = ∇ · (ρv) dτ (2.5.3)
s τ
Using (2.5.3) in (2.5.2)
  ∂ρ 
+ ∇ · (ρv) dτ = 0. Since τ is arbitrary,
τ ∂t

∂ρ 
+ ∇ · (ρv) = 0 (2.5.4)
∂t
This is the well-known equation of continuity and is a statement of the conservation of density in
phase space. Here ∇ is a 2f-dimensional del operator.
The second term of (2.5.4) can be written as
f  
∇ · (ρv) = ∑ ∂ (ρq̇i ) + ∂ (ρ ṗi )
i=1 ∂qi ∂pi
(2.5.5)
f   f  
∂q̇i ∂ ṗi ∂ρ ∂ρ
= ρ∑ + +∑ q̇i + ṗi
i=1 ∂qi ∂pi i=1 ∂qi ∂pi
In view of the Hamilton’s equation (2.5.1), the first term is
f  2
∂2 H

∂ H
∑ − =0
i=1 ∂qi ∂pi ∂pi ∂qi
∴ Equation (2.5.4) takes the final form
f  
∂ρ ∂ρ ∂ρ
+∑ q̇i + ṗi = 0 (2.5.6)
∂t i=1 ∂qi ∂pi
Again using Hamilton’s equation (2.5.1), we can put (2.5.6) in another form:
f  
∂ρ ∂ρ ∂H ∂ρ ∂H
+∑ − =0
∂t i=1 ∂qi ∂pi ∂pi ∂qi
∂ρ
⇒ + [ρ, H] = 0 (2.5.7)
∂t
f  
∂ρ ∂H ∂ρ ∂H
where [ρ, H] = ∑ − , the Poisson bracket.
i=1 ∂qi ∂pi ∂pi ∂qi
dρ
Since ρ is a function of q’s, p’s and t, the left-hand side is the total time derivative, .
dt
∂ρ ∂ρ
∴ = + [ρ, H] = 0 (2.5.8)
∂t ∂t
138 • An Introduction to Equilibrium Statistical Mechanics

This is known as Liouville’s equation.

Thus, ρ (q1 , . . . q f , p1 , . . . p f ,t) remains constant along the dynamic trajectories, i.e., the lines
whose parametric equations qi = qi (t), pi = pi (t) are derived from the equations of motion (2.5.1),
in Γ-space. This is known as Liouville’s theorem.

2.5.1 Statistical equilibrium

To get an equilibrium distribution, we need
∂ρ
=0 (2.5.9)
∂t
which immediately implies that
∂ρ ∂ρ
q̇i + ṗi = 0 (2.5.10)
∂qi ∂pi
Using the Hamilton’s equation (2.5.1)
∂ρ ∂H ∂ρ ∂H
− =0 (2.5.11)
∂qi ∂pi ∂pi ∂qi
which is easily satisfied, if we take
ρ(qi , pi ) = ρ(H(qi , pi )) (2.5.12)
because then
∂ρ ∂H ∂ρ ∂H
= 0 and =0
∂qi ∂pi ∂pi ∂qi
Thus, there is a restriction (2.5.12) on the equilibrium distribution. In other words, the equilibrium
distribution corresponds to the situation where ρ is constant on surfaces of constant energy H in
phase space. This is indeed the basic assumption of statistical mechanics.

QUESTIONS
1. Give the concept of ensemble.
2. What do you mean by microcanonical ensemble? For what type of system is it suitable?
3. Show that Ω(E) ∼ E f , i.e., Ω(E) is a rapidly increasing function of the energy E of the
system.
4. Discuss the specific heat behaviour of a two-level system. What is Schottky anomaly?
5. What do you mean by canonical ensemble? For what type of system is it suitable?
6. For a canonical system, derive the Boltzmann’s canonical probability distribution law,
pr = e−βEr 1
where β = kT .
∑ e−βEr
r
7. What is partition function? What is its importance in statistical mechanics?
8. Show that the energy fluctuation in a canonical system is proportional to √1N
9. How can you justify the equivalence between a microcanonical ensemble and a canonical
ensemble?
140 • An Introduction to Equilibrium Statistical Mechanics

3. Consider a system of N particles at a temperature T . Each particle has mass m and free to ex-
ecute one-dimensional simple harmonic motion about its mean position. If the restoring force
is proportional to x, calculate the specific heat of the system. Assume that the temperature T
is high enough so that classical statistics is applicable.
4. Find the canonical distribution function for a body rotating with angular velocity.
5. Suppose that  two atoms interact with the mutual potential energy of the form
a 12  a 6 
U (x) = U0 −2 , where x is the separation between the two particles. The
x x
particles are in contact with a heat reservoir at an absolute temperature T such that kT  U0 ,
but high enough so that classical statistics is applicable. Calculate the mean separation x (T )
1 ∂x
of the particles and use it to compute the expansion coefficient α = .
x ∂T