Consider a quantum two-level system indexed by states $|l\rangle = |0\rangle,|1\rangle$ and energies $\epsilon_l$, where $\epsilon_0 = 0$,$\epsilon_1 = \epsilon$. I throw in 2 bosons into the system and let it thermalize to temperature $T$. Now, if I want to get the partition function, I can do it two ways

  • Partition function from counting microstates: Here, the system has $\displaystyle\frac{(2+1)!}{2!\;1!}=3$ microstates, $|00\rangle^+,|01\rangle^+$ and ,$|11\rangle^+$ (the '$+$' superscript signifies symmetric combination), with energies $0,\epsilon, 2\epsilon$ respectively. Thus, the partition function is $$ Z \equiv \displaystyle\sum_{\rm{microstates}}e^{-\beta E_{\rm{microstate}}}=e^{-\beta\times 0}+e^{-\beta\times \epsilon} + e^{-\beta\times 2\epsilon}=1+e^{-\beta\epsilon}+e^{-2\beta\epsilon} $$.
  • Partition function from density matrix: Here, $$ Z = \rm{Tr}[e^{-\beta \hat{H}}]=\rm{Tr}[e^{-\beta\sum_l\epsilon_l\hat{n}_l}]=\displaystyle\prod_l\sum_{n\in[0,1,2]} e^{-\beta\epsilon_l n}\\ =\sum_{n\in[0,1,2]} e^{-\beta\epsilon_0 n}\times \sum_{n\in[0,1,2]} e^{-\beta\epsilon_1 n}\\ = \left(e^{-\beta\times 0\times 0 }+e^{-\beta\times 1\times 0 }+e^{-\beta\times 2\times 0 }\right)\times\left(e^{-\beta\times 0\times \epsilon }+e^{-\beta\times 1\times \epsilon }+e^{-\beta\times 2\times \epsilon }\right)\\ = 3\left(1+e^{-\beta\epsilon}+e^{-2\beta\epsilon}\right). $$

Why is there a discrepancy of a factor of $3$ between these two methods? While constant factors in partition functions don't affect any thermodynamics AFAIK, why is the discrepancy there anyways?

  • $\begingroup$ Yeah, sorry about that. I fixed the title. $\endgroup$ Commented Apr 18, 2018 at 19:55

1 Answer 1


In the first case you are imposing a constraint that the total number of particles must be 2. In the second calculation, however, you are requiring that the maximum number of particles in a single state is 2.

In the first calculation once you know the number of particles in the upper state is $n$ you know the number in the lower state is simply $2-n$ and so you can compute view states as being labelled by the number of particles in the upper state. In the second calculation, however, if you have $n$ particles in the upper state you can still have $0$, $1$ or $2$ particles in the lower state. Since the lower state has $0$ energy however, these possibilities have the same total energy and just result in a 3 fold degeneracy factor.

In short you actually considering 2 very different physical systems here, but because you are considering only a 2 level system it happens that their partition functions only differ by a factor of 3.

  • $\begingroup$ "the maximum number of particles in one state is 2"- but can we then speak of these particles as bosons, rather than particles that obey parastatistics? $\endgroup$ Commented Apr 17, 2018 at 17:10
  • $\begingroup$ @By Symmetry : Thanks, that clarifies things rather nicely. So if the # of particles is fixed, then the first answer is basically the correct one. $\endgroup$ Commented Apr 18, 2018 at 9:18
  • $\begingroup$ @AnalabhaRoy Yes. You could still calculate the partition function by writing it as a trace, but you would have to limit the states you trace over to include only states where the total number of particles was 2. $\endgroup$ Commented Apr 18, 2018 at 10:13
  • $\begingroup$ @AlekseyDruggist I suspect that depends on context and how formal you are being. You could also consider something along the lines of a hard core boson model. My point is that, in the second calculation, only occupation numbers $n \le 2$ are included in the sum $\endgroup$ Commented Apr 18, 2018 at 10:17

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.