New numerical method to study phase transitions

Abstract

We present a new numerical method to identify the nature of a phase transition. When combined with finite-size scaling, the method can identify unambiguously a weak first-order transition even when accessible system sizes are L/ξ<0.05 as in the five-state Potts model. At a continuous transition, exponents can be determined surprisingly accurately and the computational effort required is moderate. The method is tested on the 2D Potts and 3D Ising models, but should be quite generally applicable.