Consider the canonical and grand canonical partition functions given by $$Z_C=\sum\limits_{i}g(E_i)e^{-\beta E_i}$$ and $$Z_G=\sum\limits_{i}g(E_i)e^{-\beta (E_i-\mu)}$$ respectively with $\beta=\frac{1}{k_BT}$.


$\bullet$ What happens to these partition functions in the limit $\beta\to\infty$? Does it become a constant (in the sense that independent of $E_i$)?

$\bullet$ What is the physical significance of the limiting result (whatever it turns out)?

Update: The existing answer doesn't include the role of $g(E)$ i.e., the degeneracy of the energy level $E$ which is crucial for taking the limit. It also doesn't mention what happens to the grand partition function in the same limit. It is trickier because $\mu$ itself changes with temperature $T$.

  • $\begingroup$ What problems did you encounter in trying to take the limit yourself? $\mathrm{e}^{-x},x\to \infty$ is not a hard limit to take. $\endgroup$
    – ACuriousMind
    Sep 18, 2017 at 9:47
  • $\begingroup$ @ACuriousMind Of course, it's not hard mathematically. But there is another factor sitting in front of it $g(E)$ which is the degeneracy factor, and it requires to know how that behaves. I have also expressed my doubts in a comment to Senor O's answer. If it were really a mathematical problem of taking limits I wouldn't have asked. :-) There is also a question about grand partition function, and I think it's tricky because $\mu$ also varies with $T$. $\endgroup$
    – SRS
    Sep 18, 2017 at 10:06

2 Answers 2


In the limit that $\beta \to \infty $, all the $e^{-\beta E_i}$'s go to zero FAST. But the slowest one to go to zero is the lowest $E_i$. This is the ground state, $E_0$.

For large $\beta$, $Z$ is very small:

$Z = e^{-\beta E_0} + e^{-\beta E_1} + e^{-\beta E_2} + ... \approx e^{-\beta E_0} $

So as the temperature goes to absolute zero, the probability that the system will go into its ground state approaches 1.

$\rm Prob(E_0) \approx \frac{e^{-\beta E_0}}{e^{-\beta E_0}} = 1$

$\rm Prob(E_1) \approx \frac{e^{-\beta E_1}}{e^{-\beta E_0}} = 0$


The degeneracy plays a subtle role but does not change the physical interpretation much. $g(\epsilon_i)$ is simply going to be an integer that multiplies each $e^{-\beta E_i}$. For example, let's consider a two state system with $E_0; g(\epsilon_0) = 2 $ and $E_1; g(\epsilon_1) = 4$. Our partition function is

$Z = 2e^{-\beta E_0} + 4e^{-\beta E_1}$.

In the $\beta \to 0$ ($T \to \infty$) limit, the degeneracy plays a huge role! The probability of being in state with $E_0$ is $\frac{2}{2 + 4} = 33\%$ and $P(E_1) = 67\%$. (Since $e^{0} \to 1)$

But in the $\beta \to \infty$ low temp limit, the degeneracy (of the system - see fermi gas for why that distinction is important) has basically no effect, since muliplying $E_1$ by a constant will not saving it from going to 0 quickly due to the $e^{-\beta E_1}$ term.

I will want to think a bit more about the $Z_G$ before giving an answer - if anyone wants to chime in feel free.

  • 1
    $\begingroup$ 1. You meant $\rm Prob(E_1)$ in the last line of your answer. Right? 2. Could you include in your answer the role of $g(E)$ i.e., the degeneracy of the energy level $E$? You didn't take them into account in your answer. 3. Also, can you comment on the grand partition function as well? The non-trivial fact is that $\mu$ itself changes with $T$ or $\beta$. @Senor O $\endgroup$
    – SRS
    Sep 18, 2017 at 4:56

I resolved the issue myself.

Indeed as $\beta \to \infty$, $e^{-\beta E_i} \to \infty$ as well since E_i is negative. So the partition function will tend to $\infty$.

However, check this out.

Let $E^m$ be defined as min({ $E_i$ }). Then we can write the partition function as \begin{align} Z = e^{-\beta E^m} \sum_i e^{-\beta \underbrace{(E_i - E^m)}_{\Delta E_i}} \end{align} Notice that $\Delta E_i$ is positive, so when $\beta \to \infty$, $Z = e^{-\beta E^m} \to \infty$. However, the probability of the configuration with energy $E^m$ will be 1 and the rest of configurations will have probability zero, which is an expected fact that at zero temperature, the distribution is a delta function about the minimum energy configuration.




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.