Dismiss Notice
Join Physics Forums Today!
The friendliest, high quality science and math community on the planet! Everyone who loves science is here!

Show how the Boltzmann entropy is derived from the Gibbs entropy for equilibrium

  1. Dec 30, 2010 #1
    1. The problem statement, all variables and given/known data
    Show how the Boltzmann entropy is derived from the Gibbs entropy for systems in equilibrium.


    2. Relevant equations

    Gibbs entropy S= - [tex]\int[/tex] [tex]\rho[/tex](p,q) (ln [tex]\rho[/tex](p,q)) dpdq
    where [tex]\rho[/tex](p,q) is the probability distribution

    Boltzmann entropy S= ln[tex]\Omega[/tex]
    where [tex]\Omega[/tex] is the number of microstates in a given macrostate.


    3. The attempt at a solution

    1. Well, when the system is in equilibrium (ie when the Boltzmann entropy can be used) all microstates have equal probability. So this means that each microstate has a probability of 1/[tex]\Omega[/tex] and the probability distribution [tex]\rho[/tex] will have a constant value regardless of what p and q are.

    2. I tried putting [tex]\rho[/tex]=1/[tex]\Omega[/tex] and subbing it into the Gibb's equation

    S= - [tex]\int[/tex] 1/[tex]\Omega[/tex] (ln [tex]\1/[tex]\Omega[/tex]) d[tex]\Omega[/tex]
    using d[tex]\Omega[/tex] since we want to add up over all the microstates and there are
    [tex]\Omega[/tex] of them. But I can see that this won't give me the Boltzmann entropy.

    Any ideas?
     
  2. jcsd
  3. Dec 30, 2010 #2
    Your problem is in setting [tex]dp dq = d\Omega[/tex], because with omega, you mean a fixed number, not a variable! It is clearer in this manner:

    [tex]- \int \rho \ln \rho dp dq = - \int \frac{1}{\Omega} \ln \frac{1}{\Omega} dp dq = \left( - \frac{1}{\Omega} \ln \frac{1}{\Omega} \right) \int dp dq = \left( \frac{1}{\Omega} \ln \Omega \right) \Omega = \ln \Omega[/tex]
     
  4. Dec 31, 2010 #3
    Thanks for the reply :) But I'm still not sure how you get to the last step.

    That means the the integral of dpdp = - omega, but I can't see why that is.

    Is it something to do with normalising it?
     
  5. Dec 31, 2010 #4
    ln(1/Ω)= ln1-lnΩ lol
     
  6. Dec 31, 2010 #5
    Well, in a discrete system, omega is the number of microstates, but we're working with a continuous system here: then omega is the phase space volume, by which I mean the "volume" in (p,q)-space formed by all the available (p,q)-points. I think it's just defined that way actually.
     
Know someone interested in this topic? Share this thread via Reddit, Google+, Twitter, or Facebook

Have something to add?
Similar Discussions: Show how the Boltzmann entropy is derived from the Gibbs entropy for equilibrium
Loading...