I have to prove, considering the limits of high and low temperatures, duality equation: $$ Z=q e^{2 N K} F\left(e^{-K}\right)=q^{-N}\left(e^{K}+q-1\right)^{2 N} F\left(\frac{e^{K}-1}{e^{K}+q-1}\right) $$ for square lattice with $$Z=\sum_{\sigma_{1}, \sigma_{N}} \prod_{<i j>} e^{K \delta_{\sigma_{i} \sigma_{j}}}$$ where $<ij>$ are neighboring atoms and $\sigma_{i} \epsilon$ $\{1,2, \ldots, q\}$. Also i have information that there is only one critical point.

I can try to make analog of low and high temperature Kramers–Wannier duality, let's assume we have s horizontally and r vertically pairs of different spins, then low temperature decomposition looks like: $$ \exp \left(K \sum_{<i j>} \delta_{\sigma_{i} \sigma_{j}}\right)=\exp ((s+r)K)$$ But I have problems with high temperature decomposition: $$ e^{K \delta_{\sigma_{i} \sigma_{j}}}=\cosh \left(\frac{1+\sigma \sigma^{\prime}}{2}K\right)+\frac{1+\sigma \sigma^{\prime}}{2}\sinh (K) $$

  • $\begingroup$ "I have to prove..." Why do you have to? $\endgroup$
    – hft
    Apr 22 at 0:14
  • 1
    $\begingroup$ i forgot to add tag homework, sorry $\endgroup$ Apr 22 at 0:14
  • $\begingroup$ What textbook are you using for the class? $\endgroup$
    – hft
    Apr 22 at 0:15
  • $\begingroup$ lecture 2 from my university department slava.itp.ac.ru/intro-to-integrable-lattice-models but you will not understand this language $\endgroup$ Apr 22 at 0:16
  • $\begingroup$ I don't understand your high-temperature decomposition. I would have written the latter as $e^{K\delta_{\sigma_i\sigma_j}} = 1 + (e^K-1) \delta_{\sigma_i\sigma_j}$. $\endgroup$ Apr 22 at 11:57

1 Answer 1


The duality equation can be proved using the Fortuin-Kasteleyn representation of the Potts model. Introduce link variables $b_{ij}\in\{0,1\}$ to rewrite the partition function as $$\eqalign{ Z&=\sum_{\{\sigma\}} \prod_{(i,j)} e^{K\delta_{\sigma_i,\sigma_j}}\cr &=\sum_{\{\sigma\}} \prod_{(i,j)} \big[e^K\delta_{\sigma_i,\sigma_j} +e^0(1-\delta_{\sigma_i,\sigma_j})\big]\cr &=\sum_{\{\sigma\}} \prod_{(i,j)}\Big[\sum_{b_{ij}=0}^1\big( (e^K-1)\delta_{\sigma_i,\sigma_j}\delta_{b_{ij},1}+\delta_{b_{ij},0}\Big] }$$ The variables $b_{ij}$ form a graph on the lattice. The sum over the spins $\sigma$ can now be performed leaving $$Z=\sum_{G} u^{b(G)} q^{C(G)}$$ where $u=e^K-1$, $b(G)=\sum b_{ij}$ is the number of edges in the graph $G$ and $C(G)$ is the number of clusters. The duality transformation consists in associating to each link $b_{ij}$ on the lattice a dual link $b_{ij}^*=1-b_{ij}$ on the dual lattice. Each loops on the dual lattice encloses a cluster on the lattice. Using Euler relations, one finally gets the duality relation. For more details, see Ref. 2.


  1. Fortuin, C.M.; Kasteleyn, P.W. (1972). "On the random-cluster model : I. Introduction and relation to other models". Physica. 57 (4): 536–564. doi:10.1016/0031-8914(72)90045-6
  2. G. Grimmett (2006) The Random Cluster Model, Springer-Verlag

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.