Ben Sauerwine

Statistical Mechanics Homework 4

16) The Helmholtz free energy of a monatomic ideal gas having N particles of mass
⎡ V 3 2πβh 2 ⎤
m is F = − NkT ⎢ln − ln + 1⎥
⎣ N 2 m ⎦
a) Assume that N is an even number. Determine the density of states g (E ) for this
system by inverting the Laplace transform of the partition function Z (β ) . You will
need to calculate the contour integral
g (E ) = e βE Z (β )dβ
1
∫
2πi Br
where the Bromwich contour Br starts at − i∞ , ends at i∞ , and lies to the right of
all singularities of Z (β ) .
⎡ V 3 2πβh 2 ⎤ −1
F = − NkT ⎢ln − ln + 1⎥ = ln Z
⎣ N 2 m ⎦ β
3
− N
⎡ V 3 2πβh 2 ⎤ ⎛ V ⎞ ⎛ 2πβh ⎞ 2 N
2 N

N ⎢ln − ln + 1⎥ = ln Z Z = ⎜ ⎟ ⎜⎜ ⎟⎟ e
⎣ N 2 m ⎦ ⎝N⎠ ⎝ m ⎠
It being the case that all singularities lie to the left of the contour and the contour goes
counterclockwise, I need to find the singularities and branch cuts. The only singularity of
3
− N
βE ⎛ V ⎞ ⎛ 2πβh ⎞
N2
g (E ) =
1 2

∫ e ⎜ ⎟ ⎜
2πi Br ⎝ N ⎠ ⎝ m ⎠
⎜ ⎟⎟ e N dβ lies at β = 0 .

At this point, I decide to use a table of Laplace transforms. The inversion of

⎛ 1 ⎞ E k −1
F −1 ⎜⎜ k ⎟⎟ = , where k is any real number greater than zero. This can be seen by
⎝ β ⎠ Γ(k )
3
noticing that differentiating the numerator and denominator is necessary N − 1 times to
2
⎛3 ⎞ ⎛3 ⎞
evaluate the singularity, leaving a ⎜ N − 1⎟! in the denominator and leaving ⎜ N −1⎟
⎝2 ⎠ ⎝2 ⎠
Es in the numerator. In light of the form that one can look up, however, the choice of N
even or odd is not really an issue. This leaves me with
3 3
− N − N
⎛ 2πβh 2 ⎞ ⎛ 2πh 2 ⎞
N N
⎛V ⎞ ⎛V ⎞ 1
3
− N
g (E ) =
1 2 2

∫ e βE ⎜ ⎟ ∫
βE
⎜⎜ ⎟⎟ e dβ = e ⎜⎜
N N
⎟⎟ ⎜ ⎟ e β 2
dβ
2πi Br ⎝ N ⎠ ⎝ m ⎠ ⎝ m ⎠ ⎝ N ⎠ 2πi Br
3 ⎡ 3 ⎤
⎛ 2πh ⎞
− N N⎢ N −1
⎥
⎛V ⎞ ⎢ E2
2 2
= e N ⎜⎜ ⎟⎟ ⎜ ⎟ ⎥
⎝ m ⎠ ⎝ N ⎠ ⎢ Γ⎛ 3 N ⎞ ⎥
⎢ ⎜⎝ 2 ⎟⎠ ⎥
⎣ ⎦
 3
− N
b) Now assume that N is odd. Then Z (β ) is proportional to β 2 , so one needs
a branch cut from the point β = 0 to β = −∞ . Evaluation of the above
contour integral then involves doing some nasty branch cut integrals. Guess
that the answer involves the gamma function (as shown in the result above),
and prove that this is correct by evaluating the Laplace transform of g (E ) ,
but with N odd.
3
∞ ∞ N −1
E2
Z (β ) = ∫ g (E )e − βE dE = ∫ e − βE dE
⎛ 3 ⎞
0 0
Γ⎜ N ⎟
⎝2 ⎠
1 1
∞ ∞
E 2 − βE E 2 − βE
Let the basis case be Z (β ) = ∫ e dE = ∫ e dE . I then see that this function
0 ⎛ 3⎞
Γ⎜ ⎟ 0 π
⎝ 2⎠ 2
has no singularities anywhere, but certainly Re[E ] must be greater than zero. Make the
substitution βE → s 2 for
1 1
E = s2 dE = 2 sds
β β
∞
π π
Z (β ) =
2 2
∫ s e ds =
2 −s 2

3 3
2 4
β 2 0
β 2

Z (β ) ∝
1
3
β 2

as expected. I used Mathematica to evaluate the integral, since I can’t find an appropriate
contour. Now let me invoke integration by parts:
⎛ 1 3N ∞
⎞ ∞ ⎛3 ⎞
⎜ N −1 ⎟ −1
⎜ − E 2 e − βE → 0 ⎟ + 1
∞ 3
N −1 ⎜ β ⎟ ⎛3 ⎞ ∫ E ⎝2 ⎠
e − βE dE =
∫0 E 2 e dE = ⎝
− βE
0 ⎠ β ⎜ N − 1⎟ 0
⎝2 ⎠
− βE − βE
dv = − βe dE v=e etc.
⎛3 ⎞
⎛ ⎞
3 ⎜ N −1 ⎟ −1
1 N −1 1 3
u = − E2 du = − ⎜ N − 1⎟ E ⎝ 2 ⎠ dE
β β ⎝2 ⎠

1
and so I see that each step introduces an additional and brings N down by an integer.
β
Chaining this together and applying the basis case, then, I get
 ⎡3 1⎤ 3 3
k = ⎢ ( N − 1) − ⎥ + = N , just as expected. The contribution in brackets
1
Z∝
β k
⎣2 2⎦ 2 2
on the left is that from the recursion, the contribution on the right is that from the basis
case.

17) a) (Pathria 3.26 Modified) The energy eigenvalues of an s-dimensional

s
⎛ 1⎞
harmonic oscillator can be written as ε {n1 ,n2 ,..., ns } = hω ∑ ⎜ ni + ⎟ (use this to
i =1 ⎝ 2⎠
justify the degeneracy for s dimensions, then solve the problem by
calculating the partition function for each direction and using the
factorization theorem.) Find the multiplicity of the j-th energy level.
Evaluate the partition function and the major thermodynamic properties of
a system of N such oscillators and compare your results with a system of sN
one-dimensional oscillators. Also find the chemical potential.

So, in essence, I have s integers that I must add to get

s s
⎛ 1⎞
ε tot = ∑ ε {n1 ,n2 ,...,ns } = hω ∑ ⎜ ni + ⎟
n =1 n =1 ⎝ 2⎠
ε tot s s
− = N = ∑ ni ni ∈ 0,1,2,3...
hω 2 n =1
The number of ways to choose excited states to reach this energy corresponding to N
⎛ N + s − 1⎞
is then g (ε N ) = ⎜⎜ ⎟⎟ . For each particular dimension, however, I know that the
⎝ s −1 ⎠

(ε ) = 1 . Now Z (β ) = ∑ e
∞
− βε n
degeneracy g ni i , and
n=0
∞
Z i (β ) = ∑ e − βε n =
1
hωβ hωβ , using the geometric series summation rule.
n =0 −
e 2
−e 2

Using the factorization theorem, I have for the overall partition function
s
⎡ ⎤
Z (β ) = ⎢ hωβ
1 ⎥ . Now it is simple enough to find the other
hωβ
⎢ 2 − ⎥
⎣e −e 2 ⎦
thermodynamic properties:
 ⎡ hωβ −
hωβ
⎤
F1 part = ln Z (β ) = − ln ⎢e
1 s
2
−e 2 ⎥
β β ⎣ ⎦
hωβ hωβ
∂ ⎡ 2 − ⎤ 1 βh ω
U 1 part = − ln ⎢e − e 2 ⎥ = hsω coth
∂β ⎣ ⎦ 2 2

∂ ⎛⎜ ⎡ 2hkT ⎤ ⎞ hs ω ⎡ ⎤⎤
ω hω ω ω
∂F −
2 kT ⎟
⎡ 2hkT ⎤ ⎡ 2hkT
S1 part = = − skT ln ⎢e − e ⎥⎟ = coth ⎢ e ⎥ − ksLog ⎢ 2 sinh ⎢ e ⎥⎥
∂T ∂T ⎜⎝ ⎣ ⎦⎠ 2T ⎣ ⎦ ⎣⎢ ⎣ ⎦ ⎦⎥
∂ s ⎡ hωβ −
hωβ
⎤
μ1 = F = F1 = − ln ⎢e 2 − e 2 ⎥ = sF1
part
∂N
part
β ⎣ ⎦
dim

b) Now solve the same problem in s dimensions by doing the partition sum over
j, including the degeneracy factor. Evaluate the sum by showing that both
expressions have the same power series.

⎛ N + s − 1⎞
Verifying g (ε N ) = ⎜⎜ ⎟⎟ from the partition function, I wrote
⎝ s −1 ⎠
s
⎡ ⎤ ⎛ s⎞
⎥ = ∑ g (ε j )e = ∑ g (ε j )e ⎝ 2 ⎠
− βhω ⎜ j + ⎟
Z (β ) = ⎢ hωβ
1 − βE j
hωβ
⎢ 2 − ⎥
⎣e −e 2 ⎦
j j

− βhωs .
= ∑ g (ε j )e 2
e − βhωj

j
This is tantamount to an expansion about e − βhω . Now I have

⎡ ⎤
s
⎛ ⎛ ⎞ ⎞
⎜1 + sx + ⎜ s − s (s − 1)⎟ x 2 +
1
⎟
⎢ ⎥ s
⎜ ⎝ 2 ⎠ ⎟
Z (β ) = ⎢
1
⎥ = x2⎜ ⎟
⎢ 1 − x⎥ ⎡ ⎤
⎜⎜ ⎢ s + (s − 1)s + s (s − 1)(s − 2 )⎥ x + ...⎟⎟
1 3
⎢⎣ x ⎥⎦ ⎝⎣ 6 ⎦ ⎠
where the Nth coefficient indicates the Nth energy level (zero-indexed). This does in
fact give results matching what is expected: e.g.,
Note that the (zero-based) third index of the expansion matches the result for g (ε N ) ,
35.
18) (Pathria 3.29) The potential energy of a one-dimensional anharmonic
oscillator may be written as V (q ) = cq 2 − gq 3 − fq 4 where g and f may be
assumed to be very small in value. Show that the leading contribution of the
anharmonic terms to the heat capacity of the oscillator, assumed classical, is
3 ⎡ f 5 g2 ⎤
given by k 2 ⎢ 2 + ⎥T and to the same order the mean value of the
2 ⎣c 4 c3 ⎦
3 gkT
position coordinate q is given by .
4 c2
Write H (q, p ) = V (q ) +
1 2
p , then,
2m
∞ ∞ ⎛ ⎞ ∞
2πm − β (V (q ))
1
dqdp − β ⎜⎝ V (q )+ 2 m p ⎟⎠
2

Q (β ) = ∫ ∫
βh 2 −∫∞
e = e dq
−∞− ∞
h
∞
2πm − βcq 2 βgq 3 βfq 4
βh 2 −∫∞
= e e e dq

∞
2πm − βcq 2 ⎛ β 2 g 2q6 ⎞
≈ 2 ∫
βh − ∞
e ⎜⎜1 + βgq +
3
(
⎟⎟ 1 + βfq 4 dq )
⎝ 2 ⎠
⎛ ∂2F ⎞
CV = −T ⎜⎜ 2 ⎟⎟ F = −kT ln Q(β )
⎝ ∂T ⎠
using Mathematica, then, I get:
which agrees, and for
∫ qe − βH (q , p ) dqdp ∫ qe − βV (q ) dq
q = − βH ( q , p )
= − βV ( q )
∫ e dqdp ∫ e dq
I again use Mathematica:

19) (Pathria 3.32) Revised:

a) The quantum states available to a given physical system are (i) a group of g 1
equally likely states, with a common energy of ε 1 and (ii) a group of g 2
equally likely states, with a common energy ε 2 > ε 1 . Show that the entropy of
⎡ ⎛p ⎞ ⎛ p ⎞⎤
the system is given by S = − k ⎢ p1 ln⎜⎜ 1 ⎟⎟ + p 2 ln⎜⎜ 2 ⎟⎟⎥ , where p1 and p 2 are,
⎣ ⎝ g1 ⎠ ⎝ g 2 ⎠⎦
respectively, the probabilities of the system being in a state belonging to
group 1 or to group 2: p1 + p 2 = 1 .
 Derive the first expression given for S using the microcanonical ensemble for N
systems. To do this, express the total energy in the form E = Nε 1 + M (ε 2 − ε 1 ) ,
where M is an integer, and calculate Ω(M , N ) . Note that Pathria’s probabilities
are those of energy levels, not states. Then substitute probabilities calculated from
the canonical ensemble in this result to obtain the solution to part (a) of Pathria’s
problem.

ε 2 − ε1
E = Nε 1 + M (ε 2 − ε 1 ) . Assuming ∉ Z , then, I
ε1
⎛ N ⎞ M N −M
have Ω(M , N ) = ⎜⎜ ⎟⎟ g 2 g1 so, I have chosen M of the distributions to be in state
⎝M ⎠
g 2 and then one of g 2 states for each of the M distinguishable systems, and the rest
for one of the states g1 .
⎡ ⎛ N ⎞ M N −M ⎤ ⎡ ⎛ N ⎞⎤
S = k ln Ω = k ⎢ln⎜⎜ ⎟⎟ g 2 g1 ⎥ = k ⎢ M ln g 2 + ( N − M ) ln g1 + ln⎜⎜ ⎟⎟⎥ . Of course,
⎣ ⎝M ⎠ ⎦ ⎣ ⎝ M ⎠⎦
M N −M
speaking in the language of probability, = p2 , = p1 , then
N N
S = k ln Ω
≈ k [M ln g 2 + (N − M ) ln g1 + N ln N − N − M ln M + M − ( N − M ) ln ( N − M ) + ( N − M )]
= k [M ln g 2 + (N − M ) ln g1 + N ln N − M ln M − (N − M ) ln ( N − M )]
⎡ g g1 ⎤
= k ⎢ M ln 2 + ( N − M ) ln + N ln N ⎥
⎣ M (N − M ) ⎦
⎡
= − k ⎢ M ln
M
+ ( N − M ) ln
(N − M ) − N ln N ⎤
⎥
⎣ g2 g1 ⎦
⎡ M NM ( N − M ) N ( N − M ) ⎤ ⎡ Np 2 Np ⎤
= − Nk ⎢ ln + ln − ln N ⎥ = − Nk ⎢ p 2 ln + p1 ln 1 − ln N ⎥
⎣ N Ng 2 N Ng1 ⎦ ⎣ g2 g1 ⎦
⎡ p p ⎤
= − Nk ⎢ p 2 ln 2 + p 2 ln N + p1 ln 1 + p1 ln N − ln N ⎥ use p1 + p 2 = 1
⎣ g2 g1 ⎦
⎡ p p ⎤
= − Nk ⎢ p 2 ln 2 + p1 ln 1 ⎥
⎣ g2 g1 ⎦
and letting N → 1 , I get pure probability.
(a) Assuming that the p i are given by a canonical distribution, show that
⎡ ⎤
⎢ ⎛ g 2 −x ⎞ ⎥
S = −k ln ( g1 ) + ln⎜1 + x = β (ε 2 − ε 1 )
x
⎢ ⎜ ⎟
e ⎟+ ⎥
⎢ ⎝ g 1 ⎠ 1+ g 1
e ⎥⎥
x
⎢
⎣ g2 ⎦
, where x is assumed positive.

g i e − βε i g i e − βε i
pi = →
∑ g i e − βεi g1e − βε1 + g 2 e − βε 2
i

g1 g2
p1 = p2 =
g1 + g 2 e − β (ε 2 −ε1 ) g 1e β (ε 2 −ε1 )
+ g2
for
⎡ ⎛p ⎞ ⎛ p ⎞⎤
S = − k ⎢ p1 ln⎜⎜ 1 ⎟⎟ + p 2 ln⎜⎜ 2 ⎟⎟⎥ = ... subject to those substitutions. This expression
⎣ ⎝ g1 ⎠ ⎝ g 2 ⎠⎦
seems sure to be a frustrating exercise in algebra: I use Mathematica to verify the result.

Someone must have been very proud of their algebra, but I am equally proud of my
sanity. Thank you, Mathematica!

b) By using the canonical ensemble, prove that the first expression that was
given for S can be generalized to a system having any number of degenerate energy
levels, not necessarily equally spaced.
S = −k ∑ Pr ln Pr
r

but I must be cautious that here Pr represents overall system energy probabilities, not
individual element probabilities. Using the factorization theorem, I take
S = − k ∑∑ p g , r ln p g , r
r g

Now I have that any total energy must have a factorization into these energy levels (fixed
as unique by the system probabilities Pi and the hierarchy imposed by the fact that this is
a canonical system). Now the probability that any individual particle is chosen for this
P
state is p n ,i = i , because of the number of degenerate states it must share this
gi
probability across. Now using my factorization theorem, I write
P P P P P
S = −k ∑∑ r ln r = −k ∑ g r r ln r = −k ∑ Pr ln r .
r g gr gr r gr gr r gr
c) Check that as T → 0 S → k ln g1 . Interpret this result physically.
 g i e − βε i g i e − βε i
pi = →
∑g ei
i
− βε i
g 1 e − βε1 + g 2 e − βε 2

g1 g2
p1 = − β (ε 2 − ε 1 )
p2 = β (ε 2 −ε 1 )
g1 + g 2 e g1e + g2
β →∞
g1 g2
p1 → → 1 p2 = →0
g1 + g 2 0 g1∞ + g 2

⎡ ⎛ 1 ⎞ ⎛ 0 ⎞⎤
substituting, S = − k ⎢1 ln ⎜⎜ ⎟⎟ + 0 ln⎜⎜ ⎟⎟⎥ = k ln g1 .
⎣ ⎝ g1 ⎠ ⎝ g 2 ⎠⎦
Physically, each particle is dropping to the lowest possible energy state, but due to the
degeneracy some entropy remains since the particle doesn’t care which of these lowest
energy states it’s in.

20 a) (Pathria 3.17) If we take the function f (q, p ) in

∫ f (q, p )e
− βH ( q , p )
dω
(3.5.3 f = β ( − H q, p )
dω = d 3 N qd 3 N p ) to be U − H (q, p ) , then clearly
∫e dω

∫ [U − H (q, p )]e
− βH ( q , p )
f = 0 ; formally, this would mean dω = 0 . Derive from this
⎛ ∂U ⎞
equation expression (3.6.3 E 2 − E = −⎜⎜ ⎟⎟ ) for the mean-square fluctuation in
2

⎝ ∂β ⎠
the energy of a system embedded in the canonical ensemble.

∫ H (q, p )e
− βH ( q , p ) − βH ( q , p )
First, I note that U ≡ E , and dω = U ∫e dω = 1 so that
essentially the meaning of this equation is that the average of the distance of a result from
the average is zero. Now I will take the derivative of both sides with respect to beta:
∫ [U − H (q, p )]e
− βH ( q , p )
dω = 0
∂
∂β
[∫ [U − H (q, p )]e − βH ( q , p )
dω = 0 ]
⎡ ∂U 2⎤
∫ ⎢⎣ ∂β − UH (q, p ) + H (q, p ) ⎥ e − βH (q , p ) dω = 0
⎦
∂U
− E + E2 = 0
2

∂β
∂U
E2 − E =−
2

∂β
and so the rest of 3.6.3 follows.
 dG
b) (Pathria 3.19) Consider the long-time averaged behavior of the quantity
dt
where G = ∑ qi pi and show that the validity of
i

∂H
(3.7.5 ∑p
i
i
∂pi
= ∑ p q&
i
i i = 3NkT ) implies the validity of

∂H
(3.7.6 ∑ ∂q
i
qi = − ∑ p& q
i
i i = 3NkT ) and vice-versa.
i

Write as such:
∫ [ p q& ]e dω = 3NkT
− βH
∂H
∑i pi ∂p = ∑ p q&
i i
=
∫ e dω
i i β − H
i i

∫ p q& dt = p q − ∫ q p& dt
i i i i i i

∂
∂t
[∫ p q& dt = p q
i i i i ]
− ∫ qi p& i dt → pi q& i = −qi p& i

∂G
where at the end I have used = 0 and so integration by parts establishes the
∂t
equality of these terms, and so the equality of (3.7.5) and (3.7.6).