Simulation of Potts models with real q and no critical slowing down

Abstract

A Monte Carlo algorithm is proposed to simulate ferromagnetic q-state Potts model for any real q>0. A single update is a random sequence of disordering and deterministic moves, one for each link of the lattice. A disordering move attributes a random value to the link, regardless of the state of the system, while in a deterministic move this value is a state function. The relative frequency of these moves depends on the two parameters q and beta. The algorithm is not affected by critical slowing down and the dynamical critical exponent z is exactly vanishing. We simulate in this way a 3D Potts model in the range 2<q<3 for estimating the critical value q_c where the thermal transition changes from second-order to first-order, and find q_c=2.620(5).