# Decimation of a triangular lattice [closed]

Consider the network of spins shown below. The Hamiltonian is given by $$H = - \sum_{\langle i j k \rangle} [J \sigma_i \sigma_j \sigma_k + J_0]$$ with $J,J_o \geq 0$ and $\langle i j k \rangle$ denoting spin in the same triangle (the triangles under consideration are each highlighted by a looping arrow). Decimate over the crosses and find a recursion relation for the renormalised coupling $J'$.

Attempt: I was going to proceed by the method of transfer matrix but then I am not sure what size my matrix would be in this case. The bond action is $$e^{w(\sigma, \sigma' \sigma'')} = e^{J_o} e^{J \sigma \sigma' \sigma''}$$ and I want to find a new (renormalised) bond action $$e^{w'(\mu \mu' \mu'')} = e^{J_o'} e^{J' \mu \mu' \mu''}$$

I think $$e^{w'(\mu \mu' \mu'')} = \sum_{\sigma \sigma' \sigma''} = e^{\mu \sigma \sigma'} e^{\sigma' \mu'' \sigma''} e^{\sigma'' \sigma \mu'}$$ if I understood the set up correctly. I would proceed by writing down the matrix representation of these quantities but I am not sure what size to make them. There are $2 \times 2 \times 2$ possible combinations of spins. I am trying to understand if the transfer matrix method is appropriate for this question where the interaction is not simply of the form $\sigma_i \sigma_{i+1}$ as in the one dimensional Ising model for example.

Thanks for any tips!

• The physics concept I am trying to understand is if the transfer matrix method is appropriate for this question where the interaction is not simply of the form $\sigma_i \sigma_{i+1}$. I have edited this into my answer. Please now reopen. – CAF Apr 20 '16 at 17:31