I'm trying to derive the entropy of a black hole, given the density of states of a bosonic string (the details are not relevant). The density of states is

$$ \omega(E) = E^\alpha e^{\beta E} $$

The entropy is defined as

$$ S = k\ln \Omega, $$

where $\Omega$ is the number of microstates.

I would suppose that the number of microstates in the energy interval $(E,E+\delta E)$ would be $\delta \Omega = \omega \delta E$. So the increase in entropy is

$$\delta S = k \ln{\omega \delta E} = k \ln (E^\alpha e^{\beta E} \delta E),$$

which can't be right because according to equation $11.9.4$ in http://arxiv.org/abs/1506.07798, it should be (in the high energy limit, which ignores the power contribution in $E$)

$$\delta S = k\beta \delta E.$$

So what is the correct way to derive the entropy from the density of states?

It seems that $\delta S = k\ln \omega(\delta E) $ would work, but it involves a logarithm of a quantity with dimensions.

  • $\begingroup$ Isn't $\delta S = k \frac{\Omega}{\delta \Omega }$? $\endgroup$
    – Prahar
    Commented Jun 29, 2016 at 18:49
  • $\begingroup$ @Prahar Did you mean $\delta S = k \frac{\delta \Omega}{\Omega}$ ? $\endgroup$
    – valerio
    Commented Jun 29, 2016 at 22:04
  • 1
    $\begingroup$ @valerio92 - Yes. I absolutely meant that. Tried to type it out from my phone! $\endgroup$
    – Prahar
    Commented Jun 30, 2016 at 2:39

2 Answers 2


$$S=k \ln [\Omega(E)] = k \ln [\omega (E) \delta E] = k \ln [\omega(E)] +k \ln (\delta E)$$

Last term is an arbitrary constant, so that we can set

$$S = k \ln[\omega (E)]$$

from which

$$\delta S = k \frac{\delta \omega}{\omega}$$

If we can ignore the power contribution and set $\omega (E) \simeq e^{\beta E}$, we get

$$\delta S = k \frac{\delta(e^{\beta E})}{e^{\beta E}} = k \frac{\beta \ e^{\beta E} \delta E}{e^{\beta E}} = k \ \beta \delta E$$

More about entropy and density of states: here.

  • $\begingroup$ I know this is what people do all the time, but how should I think of the units in $S = k \ln[\Omega(E)]$ and $S = k \ln[\omega(E)]$? $\Omega(E)$ is the number of microstates which is dimensionless, but $\omega(E)$ is the density of states which has units of inverse energy. What does it mean to take the logarithm of something with dimensions? $\endgroup$
    – nervxxx
    Commented Apr 21, 2017 at 19:48
  • 2
    $\begingroup$ @nervxxx Sorry for the very late reply, I had missed this comment. You can always throw in some energy scale to make the log dimensionless. For example, $\ln[\omega(E)\delta E] = \ln[\omega(E) E_0] + \ln (\delta E/E_0)$. This doesn't change the result since the expression for $\delta S$ is unchanged. $\endgroup$
    – valerio
    Commented May 29, 2020 at 7:31
  • $\begingroup$ @valerio sorry for response to this early post. your reply is important to me. my question is why does $k \ln (\delta E)$ an arbitrary constant. should $k \ln (\delta E)$ goes to infinity as $\delta E$ approaches zero? $\endgroup$
    – FaDA
    Commented Nov 3, 2022 at 12:20
  • $\begingroup$ @FaDA Taking the limit $\delta E \to 0$ is meaningless in this situation because $\delta E$ is just a notation for an "infinitesimal", and if you want to be rigorous you can rewrite all the derivation using derivatives and series expansions. The limit $\delta E/E\to 0$, on the other hand, makes sense and it's typically taken in derivations related to this one. Also note that a logarithmic divergence is a "very slow" one so in any case you shouldn't worry too much about it ;) $\endgroup$
    – valerio
    Commented Nov 4, 2022 at 13:54
  • 1
    $\begingroup$ thanks for your help! i understand what you mean $\endgroup$
    – FaDA
    Commented Nov 5, 2022 at 3:05

As @valerio92 points out, your mistake is that $S = k \ln (\omega\, \delta E)$, not $\delta S$. To get $\delta S$, you differentiate the right-hand expression to get $\delta S = k \frac{\delta \omega}{\omega}$, and the $\delta E$ drops out and you get an expression with the right dimensions. The notation is a bit misleading, because the $\delta$ in the $\delta E$ is not a differential corresponding to the $\delta$ in the $\delta S$ - it just denotes that we should think of $\delta E$ as a small constant quantity. Once you differentiate the expression for $S$, the "differential" $\delta$ actually ends up on the $\omega$, which is the actual variable quantity.


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.