Improved algorithms for simulating crystalline membranes

The physics of crystalline membranes, i.e. fixed-connectivity surfaces embedded in three dimensions and with an extrinsic curvature term, is very rich and of great theoretical interest. Numerical simulations are commonly used to study this class of models. Unfortunately, traditional Monte Carlo algorithms suffer from very long auto-correlation times, especially near critical points. In this paper we study the performance of improved Monte Carlo algorithms for simulating crystalline membrane, such as hybrid overrelaxation and unigrid methods, and compare their performance to the more traditional Metropolis algorithm. We find that although the overrelaxation algorithm does not reduce the critical slowing down, it gives an overall gain of a factor 15 over the Metropolis algorithm. The unigrid algorithm does, on the other hand, reduce the critical slowing down exponent to z approximate to 1.7. (C) 1998 Elsevier Science B.V.