Dynamics near a first-order phase transition with the Metropolis and Swendsen-Wang algorithms

Summary, in English

We have simulated the 10-state 2d Potts model with the Metropolis and Swendsen-Wang algorithms, in order to compare their performances near a first-order phase transition. We show that a simple Markov model relates the frequencies of the flips between coexisting phases and the auto-correlation times. It leads to a coherent description of the dynamics. The ratio of the algorithm efficiencies has little temperature dependence. A slight increase of the cluster algorithm efficiency with volume, as compared to the Metropolis method, is suggested.