Statistical Physics Vishnu Jejjala

Examples 1

Let’s go over some crucial examples involving the microcanonical ensemble.

1. A process is called adiabatic if there is no heat exchange with the surroundings. These
processes are also called isentropic as entropy is conserved. The first law of thermodynamics
tells us that
δQ = 0 , δW = −p dV =⇒ dE = δW = −p dV . (1)
If the volume is compressed, meaning that we have done work on the system, the energy
increases by dU = −p dV > 0. The heat capacity is

∂E 
CV = , (2)
∂T V

so using the ideal gas law pV = N kB T ,

N kB T
dE = CV dT = −p dV = − dV . (3)
V
Integrating this expression, we see that
Z T Z V
CV dT dV
=− (4)
N kB T0 T V0 V
CV T V
log = − log . (5)
N kB T0 V0

Using the equipartition theorem, E = 23 N kB T , we have CV = 23 N kB , which implies that

 3
T 2 V
= . (6)
T0 V0

Because of the ideal gas law, this is equivalently written as

 5  5
p T 2 V0 3
= = . (7)
p0 T0 V

2. Suppose we have a volume V , which we arbitrarily divide into two components V = V1 + V2 .

With this division, there ate N1 particles in region 1 and N2 particles in region 2, so that
N = N1 + N2 . For convenience, we write V1 = q1 V and V2 = q2 V , where q1 + q2 = 1. The
total number of microstates must be proportional to V N as entropy is extensive:

Ωtot (V, N ) ∝ V N . (8)

The number of ways in which N particles can be spatially distributed in a room is the product
of the numbers of ways in which the individual particles are distributed. Keeping E and N
fixed in the microcanonical ensemble, this is proportional to the volume V . Indeed, according
to the binomial theorem:
∞
!
N
V1N1 V2N −N1 ,
X
N
Ωtot (V, N ) ∝ (V1 + V2 ) = (9)
N1
N1 =0

1
Statistical Physics Vishnu Jejjala

where N2 = N − N1 . We also know that

Ω1 (V1 , N1 ) ∝ V1N1 , Ω2 (V2 , N2 ) ∝ V2N2 . (10)
The probability for there to be exactly N1 particles in the volume V1 is
! !
Ω(V1 , V2 , N, N1 ) 1 N N
pN1 = = N (q1 V )N1 (q2 V )N −N1 = q1N1 q2N −N1 . (11)
Ωtot (V, N ) V N1 N1

The mean number of particles is now computed:

N N
!
N
q1N1 q2N −N1 N1
X X
⟨N1 ⟩ = pN1 N1 = (12)
N1
N1 =0 N1 =0
N
!
∂ X N ∂
= q1 q1N1 q2N −N1 = q1 (q1 + q2 )N = N q1 (q1 + q2 )N −1 .
∂q1 N1 ∂q1
N1 =0

Now, we know that q1 + q2 sums to 1, so ⟨N1 ⟩/N = q1 = V1 /V . This is what we expect in

equilibrium.
Now, let’s compute the fluctuation by calculating the variance:
N
X
2 2
⟨∆N1 ⟩ = ⟨(N1 − ⟨N1 ⟩) ⟩ = pN1 (N1 − ⟨N1 ⟩)2 . (13)
N1 =0

To simplify this, we observe that

N
X N
X
2
2pN1 N1 ⟨N1 ⟩ = 2⟨N1 ⟩ , pN1 = 1 . (14)
N1 =0 N1 =0

This means that

N N
!
N
q1N1 q2N −N1 N12 − ⟨N1 ⟩2 .
X X
⟨∆N1 ⟩2 = pN1 N12 − ⟨N1 ⟩2 = (15)
N1
N1 =0 N1 =0

Now, we simplify this expression using a similar trick:

  
N1 2 ∂ ∂
q1 N1 = q1 q1 q1N1 . (16)
∂q1 ∂q1
Thus,
∂ 2
 
2
⟨∆N1 ⟩ = q1 (q1 + q2 )N − (q1 N )2 (17)
∂q1
∂
= q1 (q1 N (q1 + q2 )N −1 − (q1 N )2
∂q1
= q1 N (q1 + q2 )N −1 + q1 N (N − 1)(q1 + q2 )N −2 − (q1 N )2 .
 

Simplifying, we have
⟨∆N1 ⟩2 = q1 N + q12 N (N − 1) − q12 N 2 = q1 N − q12 N = q(1 − q)N = q1 q2 N . (18)
The ratio compares the standard deviation to the mean:
p √
⟨∆N1 ⟩2
r
q1 q2 N q2 1
= = √ . (19)
⟨N1 ⟩ q1 N q1 N
This is very small for large-N .

2
Statistical Physics Vishnu Jejjala

3. Recall that Euler’s homogeneous function theorem tells us that

E = T S − pV + µN . (20)

This means that

dE = T dS + S dT − p dV − V dp + µ dN + N dµ . (21)

But the first law of thermodynamics tells us that

dE = T dS − p dV + µ dN . (22)

Thus,
0 = S dT − V dp + N dµ . (23)
This is the Gibbs–Duhem relation. Equivalently,

S(T, p) V (T, p)
dµ(T, p) = − dT + dp . (24)
N N
We derived the expression
" 5  #
T 2 p0
S(T, p) = S0 (T0 , p0 ) + N kB log . (25)
T0 p

Using the ideal gas law pV = N kB T , we have

( "  5  #)
S0 T 2 p0 dp
dµ(T, p) = − + kB log dT + kB T . (26)
N T0 p p

Let us integrate. We must hold p fixed in the integral over T and T fixed in the integral over
R
p. Recall that dx log x = x log x − x. Now,
Z T   Z p
S0 5 T p0 dp
µ(T, p) − µ(T0 , p0 ) = − dT + kB log + kB log + kB T (27)
T0 N 2 T0 p p0 p
S0 5 T 5 p
= − (T − T0 ) − kB T log + kB (T − T0 ) + kB T log
N 2 T0 2 p0
"  5  #  
T 2 p0 5 S0
= −kB T log + kB − (T − T0 ) .
T0 p 2 N