Figure 1: Possible microstate configurations for U = 4hf and two oscillators.

Note that in this graph, the

factor of 12 was dropped.

12 Oscillators and Boltzmann factors

12.1 Model system: quantum oscillators
Consider a molecule like O2 or N2 , which comprise the majority of air. These are diatomic molecules, in
which the atoms are bound together via a chemical bond. In the section on equipartition, we mentioned that
such a gas could have degrees of freedom either 3 (at very low temperatures), 5 (at room temperature), or 7
(at very high temperatures). The 5 were translational and rotational degrees of freedom, and the remaining
two of seven were the kinetic and potential energy of vibration. However, we said that due to quantum
mechanics the vibrations need a lot of energy to activate. We will explore this effect in this chapter.
A pretty good model for vibrations in molecules and solids is the quantum simple harmonic oscillator.
In this system, the energy levels are given by
1
En = (n + ), n = 0, 1, 2, 3, . . . , (1)
2
where  = hf , and r
1 k
f= , (2)
2π m
k being the spring constant and m being the mass of the oscillator. Note that in some references, the factor
of 21 is dropped an energy is written as En = nhf . Since any shift in energy does not change the physics,
either option will give the same results.
For molecules and solids, typically  is of order a few meV to a few tenths of an eV for light molecules.1
Recall that 1 eV is about 1.6×10−19 J, which sounds like a small amount of energy. However, kT at room
temperature is only about 0.025 eV, so the Boltzmann factors for larger values of  can be rather small at
room temperature. This is why the vibrations in N2 /O2 are mostly inactive at room temperature.

12.2 Derivation of Boltzmann factors for collections of oscillators

12.2.1 Two oscillators with fixed energy U
As a thought experiment, consider an isolated system of two identical oscillators with  = hf . This could
be two N2 molecules out in space, not interacting with anything else. The most relevant macrostate variable
1 Light because of the mass in f .

1
is the internal energy U . To compute the entropy, we use

S(U ) = k ln Ω(U ). (3)

Since the system can only take on integer amounts of energy, we can count the total quanta q in the system.
That is, for oscillator A and B, we have nA and nB that represent the microstate. U is given simply by the
sum nA + nB = q.
As illustrated in Fig 1, for q quanta, there are q + 1 possible configurations. Then

U = (q + 1)hf (4)

(the 1 comes from the two 12 hf energies from the lowest energy levels)

S = k ln(q + 1) (5)

We would like to find the remaining thermodynamic quantities by taking the derivative; however, since
q is an integer, we can only take the finite difference approximation to the derivative:
 
q+2
1 ∂S S(q + 1) − S(q) k ln q+1
= ' = . (6)
T ∂U U (q + 1) − U (q) hf
Solving for T ,
hf
kT =  . (7)
q+2
ln q+1

q+2
This is a bit of a strange equation, but one can see that as q increases, then q+1 gets closer to 1, making the
logarithm closer to zero, thus increasing the temperature.

12.2.2 N oscillators with q quanta

Now let’s consider many oscillators. This is a model of a solid originally formulated by Einstein, which has
many atoms that can oscillate.
The claim is that
N
U = (q + )hf (8)
2
and
N −1+q
 
Ω(q) = (9)
q
from this, we can compute  
q+N
S(q + 1) − S(q) = k ln (10)
q+1
We can then get the temperature as a function of q:
 
q+N
1 S(q + 1) − S(q) k ln q+1
= = . (11)
T U (q + 1) − U (q) hf

12.2.3 Boltzmann factors

Now let’s zoom into one of the oscillators. We want to find out what the probability of this oscillator having
a given number n1 of quanta, given that the entire collection has Q quanta. For two total oscillators (one
“system” and one “environment”), then given n1 , we know that n2 = Q − n1 so there is just one microstate
for each value of n1 . Therefore, the probability is equal for all n1 up to a maximum of Q. For three total
oscillators, given n1 , we have q − n1 quanta to distribute among two “environment oscillators. This can be

2
 1 oscillators 5 oscillators 50 oscillators

0.25 Counting
exp(−n1 hf /kT )
0.20
Probability

0.15

0.10

0.05

0.00
0 5 10 0 5 10 0 5 10
Number of quanta in the system n1

Figure 2: Convergence of the probability distribution of a single oscillator in an environment of other

oscillators. The total energy per oscillator is kept constant, and as the environment becomes large compared
to the system, the distribution approaches the Boltzmann distribution.

done in Q − n1 + 1 ways, as noted previously. So we can see that now the probability of the first oscillator
having n1 quanta decreases as n1 increases.
The general formula for N oscillators with q quanta is:

N −1+q
 
Ω(q, N ) = (12)
q

Substituting Q − n1 for q, we get

N − 1 + Q − n1
 
Ω(Q, N, n1 ) = , (13)
Q − n1

for n1 < Q.
Given a fixed value of Q and N , we can plot the probability of a given state n1 as

Ω(Q, N, nq )
P (n1 , Q, N ) = P . (14)
nq Ω(Q, N, nq )

Note that n1 is a property of the system, while Q and N are properties of the environment. Figure 2 does
this plot for increasing N . You can see that as N gets larger, then the probability curve approaches

exp(−n1 hf /kTenv )
P (n1 , Tenv ) = P , (15)
n1 exp(−n1 hf /kTenv )

where Tenv can be computed from Q and N using Eqn 11. Note that here we don’t actually care about the
total number of oscillators in the environment or the total energy of the environment; all that matters is
the temperature of the environment. This turns out to be generally true. This is what allows us to abstract
away the environment to just having a temperature.

3
 25
Heat capacity per mol (J/K mol)
20

15

10

0
0 50 100 150 200 250 300
Temperature (K)

Figure 3: The heat capacity per mol of a collection of quantum oscillators with hf = .01 eV. The vertical
dotted line is the temperature where T = hf /k, and is about where quantum effects are apparent (below
this temperature, quantum effects reduce the heat capacity).

12.3 Heat capacity of oscillators: Einstein’s models of solids

Here you get to find out why equipartition sometimes is not applicable.
First we write down the internal energy as a function of temperature.
X
U= Ei Pi (T ) (16)
i

At constant volume
dU X dPi
CV = = Ei . (17)
dT i
dT
Skipping over quite a few algebraic manipulations and applications of chain rules, etc, we can find that for
one oscillator,
x2 ex
CV = k x , (18)
(e − 1)2
hf
where x = kT . Therefore, for N oscillators in 3 dimensions,

x2 ex
CV = 3kN . (19)
(ex− 1)2

Let’s consider the limits of this function. It is useful to multiply the numerator and denominator by e−x ,
and bring it inside the square, which results in

x2
CV = 3kN , (20)
ex − 2 + e−x
As T → ∞, x → 0, and ex → 1. This would give us 0/0, but using L’Hopital’s rule and differentiating
(twice), we get
2
lim CV = lim 3kN x = 3kN. (21)
T →∞ x→0 e + e−x

4
This is the same result we got from equipartition for a collection of N atoms in 3D! So this model has the
same high temperature limit as equipartition.
Now let’s consider T → 0. x → ∞. This is a little easier since we know that ex increases faster than
anything else, so as x becomes very large then it will dominate the other terms. Therefore,

lim CV = 0. (22)
T →0

This is very different from the equipartition result, and is a direct result of quantum mechanics! The fact
that at low temperatures, the heat capacity goes to zero was one of the first hints that classical mechanics
is not the whole story.
We can see the entire function plotted in Fig 3. Note that at room temperature the heat capacity flattens
out, but as the temperature is decreased, at around the point of T = hf /k, the heat capacity starts to
decrease. This is roughly the temperature at which the ratio of the Boltzmann factor between n1 and n1 + 1
quanta, exp(−hf /kT ), starts to become apparent. Temperatures higher than that don’t “see” the discrete
energy levels, while for lower temperatures, the discrete energy levels result in the heat capacity going to
zero.