The following discussions are for isotropic quantum harmonic oscillators which have the energy eigenvalues as follows:

$$E=\left(\sum_{i}^{N}n_i+\frac{N}{2}\right)\hbar \omega$$

where $N= $ number of dimensions.

I have two interrelated doubts regarding the partition function calculation on the above system.

One oscillator in 3D is equivalent to three 1D independent oscillators so that if the partition function (P.F.) for a 1D harmonic oscillator is:

$$Z_{1D}=\sum_{n=0}^{\infty} e^{-\beta E_n}$$

then for the 3-D harmonic oscillator it becomes :

$$Z_{3D}=(Z_{1D})^3 \tag{1}$$

On the other hand,

$$Z_{3D}=\sum_{n=0}^{\infty}g(n) e^{-\beta E_n} \tag{2}$$

where here,




On plugging the value of degeneracy, it is not easy to show whether (1) and (2) are equal. I tried solving it but was not getting (2) same as (1).

Shouldn't (1) and (2) be the same or is there a problem with how did the above steps?

If they indeed are the same, what manipulation is needed to show that?


If after calculating the partition function for $N$ independent 1D harmonic oscillators, we try to take the indistinguishability into account by dividing with a factor of $N!$ $\\ $ we get:


If now we have to do the same for the $N$ number of 3D oscillators, we can instead consider the $3N$ number of independent 1D oscillators so that:


But I have read that instead it is correctly calculated as:


which has a different denominator value.

Which one of the above is correct and how?

  • $\begingroup$ The way I see it, in 3D, there are $N$ different particles, and there are $3N$ modes, where the factor of 3 comes from the three possible polarizations. You have to symmetrize with respect to the different particles, since they must be indistinguishable, which is why you only have a factor of $N!$ on the bottom, but you have a factor of $3N$ because there are $3N$ modes. $\endgroup$
    – march
    Commented Nov 9, 2021 at 17:08

1 Answer 1


The partition function of the 1D harmonic oscillator is $$\eqalign{ {\cal Z}_{\rm 1D}&=\sum_{n=0}^{+\infty} e^{-\beta\hbar\omega\big(n+1/2\big)}\cr &=\lambda^{1/2}\sum_{n=0}^{+\infty}\lambda^n\cr &={\lambda^{1/2}\over 1-\lambda} }$$ where $\lambda=e^{-\beta\hbar\omega}$. Consider now the 3D harmonic oscillator. First, one can note that the system is equivalent to three independent 1D harmonic oscillators: $${\cal Z}_{\rm 3D}=\big({\cal Z}_{\rm 1D}\big)^3 ={\lambda^{3/2}\over (1-\lambda)^3}$$ On the other hand, using your equation (2), we get after some algebra, $$\eqalign{ {\cal Z}_{\rm 3D}&=\sum_{n=0}^{+\infty} g(n) e^{-\beta\hbar\omega\big(n+3/2\big)}\cr &=\lambda^{3/2}\sum_{n=0}^{+\infty} {(n+2)(n+1)\over 2}\lambda^n\cr &={\lambda^{3/2}\over 2}{\partial^2\over\partial\lambda^2}\Big[ \sum_{n=2}^{+\infty} \lambda^{n+2}\Big] \cr &={\lambda^{3/2}\over 2}{\partial^2\over\partial\lambda^2}\Big[ \lambda^2\sum_{n=0}^{+\infty} \lambda^n\Big] \cr &={\lambda^{3/2}\over 2}{\partial^2\over\partial\lambda^2}\Big[ {\lambda^2\over 1-\lambda}\Big] \cr &=\lambda^{3/2}\Big[{1\over 1-\lambda}+{4\lambda\over(1-\lambda)^2} +{\lambda^2\over(1-\lambda)^3}\Big]\cr &={\lambda^{3/2}\over(1-\lambda)^3}\cr }$$ i.e. the same result, as expected!

In the Einstein solid, one considers $N$ atoms oscillating around their equilibrium position. In this simple model, two atoms are not expected to exchange their position so the atoms should be considered as distinguishable. Each atom is reduced to a 3D harmonic oscillator, equivalent to three independant 1D harmonic oscillators associated to the three directions. They also should be considered as distinguishable. As a conclusion, you should not divide the partition function by $(3N)!$, nor $N!$. Note that adding or not a factor $1/N!$ to the partition function is not of great importance: in both cases, the Dulong-Petit law for the specific heat is recovered (which is the goal of this model).

  • $\begingroup$ I had to delete my answer in the face of evidence...I started thinking about probabilities and erroneously transfered those thoughts to the partition functions. $\endgroup$
    – anonymous
    Commented Nov 9, 2021 at 11:53
  • $\begingroup$ @Christopher Thank you for the slick manipulation ..the first part is clear. What about the "Second" one? What do you mean by "The three oscillators are distinguishable: the first one is associated to the x-direction, the second one is y-direction, and the third one to the z-direction. You should not divide the partition function by 3!." $\endgroup$
    – Lost
    Commented Nov 10, 2021 at 7:59
  • $\begingroup$ I can consider a 3D oscillator to be a three 1D oscillators constrained in the x-direction so that now I essentially have to divide by a factor of 3! (or [3N]! for N 3D oscillators). What's wrong with this argument? and why dividing by N! is correct as opposed to 3N!even when physically "one can note that the system is equivalent to three independent 1D harmonic oscillators:" is correct. $\endgroup$
    – Lost
    Commented Nov 10, 2021 at 8:02
  • $\begingroup$ Also, "The three oscillators are distinguishable"....consider them indistinguishable just for the sake of argument (the same problem of different factors is faced for free particles ) then what is "the correct Boltzmann factor" N! or 3N! ? $\endgroup$
    – Lost
    Commented Nov 10, 2021 at 8:05
  • 1
    $\begingroup$ In the case of a gas, you cannot follow individually one particle so the N particles are indistinguishable and you should divide by N! $\endgroup$
    – Christophe
    Commented Nov 11, 2021 at 7:35

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.