1. Limited time only! Sign up for a free 30min personal tutor trial with Chegg Tutors
    Dismiss Notice
Dismiss Notice
Join Physics Forums Today!
The friendliest, high quality science and math community on the planet! Everyone who loves science is here!

Homework Help: Potts Model in statistical physics

  1. Oct 18, 2015 #1


    User Avatar
    Gold Member

    1. The problem statement, all variables and given/known data
    The 3 state Potts model is defined by $$-\beta \mathcal H = J \sum_{r,r'} (3 \delta_{\sigma(r), \sigma(r')} - 1) + h\sum_r \delta_{\sigma(r),1},$$ with J > 0 to encourage neighbouring Potts spins to have same value and h orienting field. The spin like variables can take values 1,2 or 3. The spins are on a d dim hypercubic lattice so that each spin has z=2d nearest neighbours.

    Consider the easier hamiltonian $$-\beta \mathcal H_o = H \sum_r \delta_{\sigma(r),1}$$ where H is an external field.

    a) Compute the quantity ##\langle -\beta(\mathcal H- \mathcal H_o)\rangle## wrt easy hamiltonian. Use the fact that, wrt the easy hamiltonian, $$\langle \delta_{\sigma, \sigma'} = \langle \delta_{\sigma,1}\rangle^2 + \langle \delta_{\sigma,2}\rangle^2 + \langle \delta_{\sigma,3}\rangle^2 = \langle \delta_{\sigma,1}\rangle^2 + \frac{(1-\langle \delta_{\sigma,1}\rangle)^2}{2}$$

    b) Use variational mean field theory to find the best lower bound for the original partition function using the easy hamiltonian above. Show that the resulting mean field equation is $$m = \frac{e^{h+3Jzm} - 1}{e^{h+3Jzm} + 2}$$

    2. Relevant equations
    All in section 1, and ##\sum_{\sigma} e^{-\beta \mathcal H_o}## is partition function associated with system governed by easy hamiltonian

    3. The attempt at a solution
    I found the partition function associated with the easy hamiltonian and the object for part a) is: $$\langle (h-H) \sum_r \delta_{\sigma(r),1} + J \sum_{\langle r,r'\rangle} (3 \delta_{\sigma(r), \sigma(r')} - 1) \rangle_{0,H}$$ There are N nodes with a spin and at each node the average of ##\delta_{\sigma(r),1}## is (1+0+0)/3 = 1/3, so for N spins, the first term average is N/3. For the second term, I get a 2dN multiplied by average of ##\langle \delta_{\sigma(r),\sigma(r')}\rangle##. I can use the suggested formula and I get this to also evaluate to 1/3.

    I am just wondering how this equation they gave comes about. I can write $$\langle \delta_{\sigma,\sigma'}\rangle = \sum_{\sigma, \sigma'} \delta_{\sigma, \sigma'}e^{-\beta \mathcal H_o}$$ which is written like $$\langle \delta_{1,\sigma'}\rangle \langle \delta_{\sigma,1}\rangle + \dots$$ I guess?

  2. jcsd
  3. Oct 19, 2015 #2


    User Avatar
    Gold Member

    Can anyone help me understand the formula in a)? part b) should be fine thereafter. Thanks
Share this great discussion with others via Reddit, Google+, Twitter, or Facebook

Have something to add?
Draft saved Draft deleted