2
$\begingroup$

This is from Problems in Thermodynamics and Statistical Physics by P.T. Landsberg

A system can be in any one of N states. Using the method of undetermined multipliers to show that for the maximum entropy, $S = -k \sum_i p_i \ln p_i$ where $p_i = 1/N$, $$S = k \ln N\,,$$
the solution is as follows:

Using an undetermined multiplier $\alpha$, write $$f = -k \sum_i(p_i \ln p_i - \alpha p_i)$$ $$\frac{\partial f}{\partial p_j} = -k (\ln p_i + 1 -\alpha ) = 0$$

$\ln p_j = \alpha -1$for all $j$, therefore all $p_i$ are equal. The maximum entropy is $$S_{max} = -k \sum_i(\frac{1}{N}\ln \frac{1}{N}) = k \ln N$$

I don't understand the last part on how to go from the sum to the final expression, given $\sum_i p_i = 1\;.$

$\endgroup$
  • $\begingroup$ This is not a homework problem. Who ever flagged it as such obviously did not read through the entire post and is quite possibly an idoit. This is from a problems and solution book! $\endgroup$ – mcodesmart Oct 12 '13 at 7:32
  • $\begingroup$ Please see meta.physics.stackexchange.com/questions/714/… for details on when the homework tag applies. Although this isn't a homework question, the tag is correct in this case - but it doesn't mean it's a bad question. $\endgroup$ – Nathaniel Oct 12 '13 at 7:38
3
$\begingroup$

It's a very simple mathematical identity, which arises because the sum is over $N$ things that are all the same. So $$ -k \sum_{i=1}^N \frac{1}{N}\ln \frac{1}{N} $$ is the same as $$ -k N \left( \frac{1}{N}\ln \frac{1}{N} \right) $$ and I think you should be able to do the rest from there.

$\endgroup$
2
$\begingroup$

Since $p_j$ is same for all $j$, we have for the normalisation $\sum_j p_j = N*p_j = 1$. This implies $p_j = 1/N$.

Now, $S_{max} = -k \sum_j p_j \ln p_j = -k N * (p_j \ln p_j) $. Now substituting $p_j = 1/N$ should give the expression you are looking for.

$\endgroup$

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.