In David Chandler's 'intro to statistical mechanics' he states that for an ideal gas at high-temperature

$$ \langle n_j\rangle=\langle N\rangle\frac{e^{-\beta \epsilon_j}}{\sum e^{-\beta \epsilon_j}} $$

Which I can believe from intuition, but he losses me on the derivation.

Starting with the general form for the occupancy of a boson or Fermion gas:

$$ \langle n_j \rangle=[e^{\beta (\epsilon_j-\mu)} \pm 1]^{-1} $$

Then at high-temperature, $\beta \rightarrow 0$, [ page 101 (b) ]

$$ e^{\beta(\epsilon_j-\mu)}>>1 $$

So the assumption can be made that

$$ \langle n_j \rangle=e^{-\beta (\epsilon_j-\mu)} $$

Which would make sense if $e^{\beta(\epsilon_j-\mu)}>>1$, but it sure seems like that $ e^{\beta(\epsilon_j-\mu)}\approx 1$ in that case. Is there another assumption that's made here?
He does say:

Note that if this eqn is true for all $\epsilon_j$, then $-\beta \mu >>1 $

so, $\mu \rightarrow -\infty$ at high-temperature? That doesn't strike me as something that's obvious.

The rest of the derivation:

$$ \langle N \rangle = \sum \langle n_j \rangle =\sum e^{-\beta(\epsilon_j-\mu)}=e^{\beta \mu} \sum e^{-\beta \epsilon_j} \\ e^{\beta \mu} = \frac{\langle N \rangle}{\sum e^{-\beta \epsilon_j}} $$

using $\langle n_j \rangle = e^{-\beta(\epsilon_j - \mu)}$

$$ \langle n_j\rangle=\langle N\rangle\frac{e^{-\beta \epsilon_j}}{\sum e^{-\beta \epsilon_j}} $$


There are indeed other assumptions in the derivation you quote. Namely, Chandler considers the classical limit of an ideal quantum mechanical gas with average particle number

$ <N> = \sum_j <n_j> = \sum [ e^{\beta (e_j - \mu)} ± 1 ]^{-1} $

(plus for Fermi-Dirac, minus for Bose-Einstein statistics). In the classical limit (low density) there are many more single particle states than particles. Thus $ <n_j>$ is much smaller than 1, which implies

$ e^{\beta ( \epsilon_j - \mu)} >> 1 \ \ ($ i.e. the "$± 1$" in the eqn above becomes irrelevant)

In other words, the distinction between FD and BE statistics vanishes in the classical limit.

The second part of your question concerns the chemical potential $\mu$, which is shown to be just the standard Boltzmann factor in the classical limit. You don't need to consider any limits on $\mu$, just the fact that $<n_i> \ \rightarrow \ \exp(-\beta(\epsilon_i - \mu))$ in the classical limit.

  • $\begingroup$ For sure, I'm okay with the $\pm$ being irrelevant if $e^{\beta(\epsilon_j - \mu)} >> 1$. but how is it much greater than 1 if $\beta \rightarrow 0$ $\endgroup$ – Ben Aug 21 '14 at 23:24
  • $\begingroup$ You're right. The chemical potential is just the standard boltzman factor, cause the gas would still behave classically at high temperatures. I think it does help to consider the limits on $\mu$ though. $\endgroup$ – Ben Aug 21 '14 at 23:49
  • $\begingroup$ @Ben H. at high temperature and low particle density, $<n_i>$ is much smaller than one. Its reciprocal is therefore much larger than one and approximately equal to $\exp(\beta(\epsilon_i-\mu))$ $\endgroup$ – Felix Aug 22 '14 at 19:06
  • $\begingroup$ I agree, intuitively that makes sense. I just wasn't seeing how it came out of the math $\endgroup$ – Ben Aug 23 '14 at 0:45

Actually, it does make sense that $\mu \rightarrow - \infty$

Given the ideal chemical potential for in ideal gas:

$$ \mu = -k_B T\ln \left( \frac{V}{N} \left(\frac{mk_B T}{2 \pi \hbar}\right)^{3/2} \right ) $$


$$ \mu \beta \sim - \ln(T) \\ \:\\ \therefore \lim_{T \rightarrow \infty} -\mu \beta >> 1 $$


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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