Set 3. 13.02.2015. Problems 4.5, 4.19, 4.20, 4.24, 4.

25

February 14, 2015

4. Basic principles of statistical mechanics

Problem 4.5 Model for rubber elasticity.

The rubber is assumed to consist of a polymer chain of N rod-like monomers, each of length
a and is subjected to a force F in the +X direction. Each monomer can point independently
along any of X ,Y, Z axes, in either the + or − direction. The energy is only X -dependent; ε = aF
for the monomer pointimg along −X , ε = −aF for the monomer pojnting along +X , ε = 0 for
the monomer along ±Y and ±Z.

(a) Calculate the partition function for the N-monomer chain.

(b) Show that the linear thermal expansivity is negative, as for the real rubber. Interpret the
result physically.

Solution to Problem 4.5

(a) Each monomer has the partition function

Z1 = 2 + 2 + e−βFa + eβFa = 4 + 2 cosh(βFa)

The first terms are from e0y + e− 0y and 0z + e− 0z : the total partition function being

Z = Z1N = 2N [2 + cosh(βFa)]N .

(b) If Nξ+ states point on the average to +X , Nξ− states point to −X , etc., then

hLX i = Na(ξ+ − ξ− ) , hLY i = Na(η+ − η− ) , hLZ i = Na(ζ+ − ζ− ) ,

We immedately obtain that hLY i = hLZ i = 0. Furthermore,

sinh βaF
ξ± = e±βaF /Z1 , → hLX i = Na .
2 + cosh βaF

We immediately get ∂hLX i/∂β > 0. Consequently, ∂hLX i/∂T < 0. Rubber contracts upon
heating.

2
 TD Theory of the rods
Consider the rod of length l streching in longitudinal direction with forch f . The work done
at stretching of the rod on the length dl is

dW = − f dl

.
Let us compare the rod with the ideal gas:

dW = PdV

All formulas formulas used for ideal gas is valud for rod with substitution of P → − f and
V → l. So the generilized force in our problem is − f and generalized coordinate is l. Then

dU (S, l) = T dS = f dl

Maxwell rule will look like

∂(T, S)
= −1
∂( f , l)
In the TD of rod the EoS will be f = f (l, T ) instead of P = P(V, T ) for gases and will have a
linear dependence on T in a range of the small elastic deformations:

l(T, 0) = l(T0 , 0) [1 − α(T − T0 )]

where T0 = 273.25K o , α is the coefficient of linear expansion, defined only experimentally.

Hook’s law tells
l(T, 0) − l(T0, 0) 1
= f
l(T, 0) Eσ
where E is Young module, does not depend on T , σ is the rod aria in the transverse plane. Two
last equations gives
 
l
f = Eσ − 1 where l0 = l(T0 , 0)
l0 (1 + α(T − T0 )

Let us consider E and σ constants, α(≈ 10−5 grad− 1 small, then we will obtain the EOS for rod:
 
l
f = Eσ (1 − α(T − T0 )) − 1
l0
Let us find entropy of rod:
∂S ∂S ∂f
     
Cl
dS = dT + dl = dT − dl
∂T l ∂l T T ∂T l
From EOS which gives
∂f αEσl
 
=−
∂T l l0

3
 Cl αEσl
dT +dS = dl
T l0
To obtain S we need to know Cl (T, l). There are similaries between ideal gases, real gases
and perfect rods: in all three cases the generalized force (P for gases and f for rods) is linear
function of temperature, it means Cl does not depend on l as CV does not depend on V for gases:
∂Cl ∂ f
   2 
= =0
∂l T ∂T 2 l
From here for the case Cl = const
Cl (T ) αEσl 2
Z
S= dT + + const
T 2l0
αEσl 2
S = Cl log T + + const
2l0
The internal energy of ideal rod is
 
l l − l0
dU = T dS + f dl = Cl dT + αEσ + f dl ≈ Cl dT + Eσ dl
l0 l0
Eσ(l − l0 )2
U = Cl T + + const
2l0
In contract to ideal gases which energy does not depend on volume, internal energy of ideal
rods is a square of its deformations.
Subproblem on Rubber Elasticity
As a simple model of an elastic string like, e.g., a rubber band, we consider a linear chain of
N building blocks. Each building block can be in two different states a or b. In these states the
building blocks have length la , and lb and energies εa and εb , respectively. The total length of the
chain is L = Na la + Nb lb and the total energy of the string by itself is E0 = Na εa + Nb εb where
Na = N − Nb is the number of building blocks in state a. The string is streched by an external
force f which turns the total energy of a state into E = E0 − L f .
a) Calculate the partition function of this string as a function of temperature T , the number
of building blocks N, and the external force f . Introduce variables ni ∈ {a, b} that describe in
which state building block i is and write the partition function as a sum over these variables ni .
Solution a)

∑ e−β(E0−L f ) = ∑ ∑ e−β(∑i=1 εni ) eβ( f ∑i=1 lni )

N N
Z(T ) = ···
ni ni ∈{a,b} nN ∈{a,b}
!N
N
∑

−β(εni + f lni )
= e = e−βεa +β f la + e−βεb +β f lb
n∈{a,b}

b) Calculate the average internal energy U of this string as a function of temperature T , the
number of building blocks N, and the external force f .

4
 Solution b)

∂ ln Z(T ) (εa − f la )e−βεa +β f la + (εb − f lb )e−βεb +β f lb

 
U =− =N (1)
∂β e−βεa +β f la + e−βεb +β f lb
c) Calculate the expected length < L > of this string as a function of temperature T , the
number of building blocks N, and the external force f . (Hint: The expected length is a similar
quantity as the expected energy. Find an expression for the expected length through a derivative
similar to the derivative which we use to calculate the average internal energy.) What is the
expected length at zero force in the case εa = εb ? Why?
Solution c)
N
1 ∂ ln Z(T )
 
< L > = < ∑ lni >= |β,N =
i=1 β ∂f
la e−βεa +β f la + lb e−βεb +β f lb
= N
e−βεa +β f la + e−βεb +β f lb
At εa = εb ad f = 0:
la e−βεa + lb e−βεa la + lb
< L >= N −βε −βε
=N
e a +e a 2
If (roughly) half of the building blocks are in state a and half in state b the entropy is maximal
and it is the most preferable state.

Problem 4.19 Suppose the expression S = −kB ∑r Pr ln Pr is accepted as a definition of the

(1)
entropy. Imagine that a system A1 has probability Pr of being in a state r and a system A2 has
(2)
probability Ps of being in a state s. Then

S1 = −kB ∑ Pr ln Pr , S2 = −kB ∑ Ps ln Ps .
(1) (1) (2) (2)

r s

Each state of a composite system A consisting of A1 and A2 can then be labeled by the pair of
numbers, r, s. Let the probability of A being found in the state r, s be denoted by Prs , and the
corresponding entropy is −kB ∑r,s Prs ln Prs .
(1) (2)
(a) If A1 and A2 are weakly interacting so they are statistically independent, then Prs = Pr Ps
Show that under such circumstances the entropy is simply additive, i. e. S = S1 + S2 .
(1) (2)
(b) Suppose that A1 and A2 are not weakly so that Prs 6= Pr Ps . One has, of course, the general,
(1) (2)
relations ∑s Prs = Pr , ∑r Prs = Ps , and ∑r,s Prs = 1. Show that
(1) (2)
Pr Ps
S − (S1 + S2 ) = kB ∑ Prs ln .
rs Prs
Moreover, by using the inequality
− ln x ≥ 1 − x,

5
 (1) (2)
show that S ≤ S1 + S2 , where the equality holds only if Prs = Pr Ps for all r and s. This
means that the existence of correlation between the systems leads to a situation less random
that where the systems are completely independent of each other.

Solution to Problem 4.19 Let us start from the part (b), and the 1st part will be a
limiting case. We have

∑ Pr ln Pr + ∑ Ps ln Ps
(1) (1) (2) (2)
−(S1 + S2 )kB =
r r

∑ ∑ Prs ln Ps
(1) (2)
= Prs ln(Pr +
r,s s,r

∑ Prs ln Pr Ps
 
(1) (2)
= .
r,s

(1) (2)
If Prs = Pr Ps then S = S1 + S2 . Now
(1) (2)
Pr Ps
S − (S1 + S2 ) = kB ∑ Prs ln
r,s Prs
(1) (2)
" #
Pr Ps
≤ kB ∑ Prs −1
r,s Prs

= kB ∑ Pr Ps − Prs = 0 .
h i
(1) (2)

r,s

Problem 4.20 0.1. Consider a system distributed over its accessible states r in accordance
with a probability distribution Pr , and let its entropy be defined by the relations

S = −kB ∑ Pr ln Pr , ∑ Pr = 1 .
r r

Compare this distribution with the canonical distribution

(0) e−βEr
Pr =
∑r e−βEr
corresponding to the same mean energy hEi, i. e.

∑ Pr Er = ∑ Pr
(0)
Er = hEi .
r r

The entropy of the canonical distribution is

S0 = −kB ∑ Pr ln Pr .
(0) (0)

6
(a) Show that
(0)
Pr
S − S0 = kB ∑ Pr ln .
r Pr
(0)
(b) Using the inequality ln x ≤ x − 1 show that S0 ≥ S; the equality sign holds only if Pr = Pr
for all states r. This shows that, for a specific value of mean energy, the entropy S is a
maximum for the canonical distribution.

Solution to Problem 4.20

Solution 0.2:
(a)

= − ∑ Pr ln Pr − Pr ln Pr
h i
(0) (0)
(S − S0)/kB
r

= − ∑ Pr ln Pr − Pr (−βEr − ln N )
h i
(0)

r
= − ln N − βhEi − ∑ Pr ln Pr
r
(0)
Pr
= ∑ Pr ln Pr
.
r

(b) We have
(0) (0)
!
Pr Pr
∑ Pr ln ≤ ∑ Pr = ∑ Pr − Pr = 0 .
h i
(0)
−1
r Pr r Pr r

Problem 4.24 Consider non-interacting particles subjected to a harmonic potential. Cal-

culate the canonical partition function
(a) for a single particle
(b) for two distinguishable particles
(c) for two spinless fermions
(d) for two spin-zero bosons
(e) for two spin-1/2 fermions.
Compare the internal energies and entropies in these various cases. Study the limit T → 0,
T → ∞, and h̄ = 0 and interpret the results physically.

Solution to Problem 4.24 Quantum states for a particle in a harmonic potential are
specified by the quantum number n and the energies are εn = h̄ω(n + 1/2). Thus we have:

7
(a)
∞
e−βh̄ω/2 1
Z1 = e−βh̄ω/2 ∑ e−βh̄ωn = =
1 − e−βh̄ω 2 sinh(βh̄ω/2)
n=0
βh̄ω
  
−1 1
F1 = −β ln Z = ln 2 sinh
β 2
∂ ln Z h̄ω h̄ω 1
E1 = − = coth(βh̄ω/2) = + h̄ωN(ω) , N(ω) = βh̄ω ,
∂β 2 2 e −1
∂F βh̄ω βh̄ω βh̄ω
    
S1 /kB = − = coth − ln 2 sinh
∂kB T 2 2 2
Limiting cases
• T → ∞ or β → 0
Z1 ≈ kB T /h̄ω , F1 ≈ −kB T ln(kB T /h̄ω) , E1 ≈ kB T , S1 /kB ≈ 1 + ln(kB T /h̄ω).

• T → 0 or β → ∞
Z1 ≈ e−βh̄ω/2 , F1 = h̄ω/2 , E1 = h̄ω/2 , S1 = 0 .

(b)
∞
e−βh̄ω 1
Z2 = Z12 = e−βh̄ω ( ∑ e−βh̄ωn )2 =  2 = 2
n=0 1 − e−βh̄ω 4 sinh (βh̄ω/2)
F2 = −2kB T ln [24 sinh(βh̄ω/2)]
E2 = 2h̄ω [1/2 + N(ω)]
S2 /kB = 2S1 /kB .

(c) Consider first two identical classical particles. In the product Z12 there are two identical terms
in which one of the particles occupy a state a while the second one occupies the state b. In
the particles are non-distinguishable then we have to divide the product by 2! = 2 to get
1 η
Z2cl = Z12 /2 = , η ≡ e−βh̄ω .
2 (1 − η)2
Then the configuration in which both particles occupy the same state get the weight 1/2.
Fermions are not allowed to occupy this state, thus we have to extract this configuration.
In this way we get
∞
η
Z2FD = Z2cl − D/2 , D = e−βh̄ω ∑ e−2nβh̄ω = .
n=0 1 − η2
Thus
η2
Z2FD = .
(1 − η)(1 − η2)

8
(d) In a similar way,
η
Z2BE = Z2cl + D/2 = .
(1 − η)(1 − η2 )
There is no difference between classical and quantum statistics at high temperatures when
η ≪ 1.

(d) In the case of spin-1/2 fermions each state is doubly degenerate. Thus the proper partition
function is
2η2
 
1 1
(2Z1 ) /2! − 2D/2 = 2Z1 − D = η
2 2
− = .
(1 − η)2 1 − η2 (1 − η)(1 − η2)
The calculation of all thermodynamic quantities is straightforward.

Problem 4.25 Let fn be the average occupation of the n-th single-particle level in an ideal
Fermi gas. Consider the binary scattering process where two fermions in states 1 and 2 get
scattered into states 3 and 4. hen , rate of forward scattering is

f1 f2 (1 − f3 )(1 − f4 )R

whereas the rate of reverse scattering is

f3 f4 (1 − f1 )(1 − f2 )R′

where R = |M|2 is the square of the matrix element M of the scattering operator the states 1 and
2 and R = |M ∗ |2 . The principle of detailed balance states that. in equilibrium,

f1 f2 (1 − f3 )(1 − f4 ) = f3 f4 (1 − f1 )(1 − f2 ) .

Show that the Fermi-Dirac distribution is the non-trivial solution of this equation. (This deriva-
tion shows how Pauli exclusion principle leads to the FD distribution).
h i−1
Solution to Problem 4.25 Since fi = eβ(ε i −µ) +1 we have

eβ(εi −µ)
1 − fi = β(ε −µ)
= eβ(εi −µ) fi .
e i +1
Thus,

f1 f2 (1 − f3 )(1 − f4) = eβ(ε3 +ε4 −2µ) f1 f2 f3 f4 , f3 f4 (1 − f1 )(1 − f2 ) = eβ(ε1 +ε2 −2µ) f1 f2 f3 f4 .

Because of the energy conservation law,

ε1 + ε2 = ε3 + ε4 .