Introduction to Multicanonical Monte Carlo Simulations

Abstract

Monte Carlo simulation with {\it a-priori} unknown weights have attracted recent attention and progress has been made in understanding (i) the technical feasibility of such simulations and (ii) classes of systems for which such simulations lead to major improvements over conventional Monte Carlo simulations. After briefly sketching the history of multicanonical calculations and their range of application, a general introduction in the context of the statistical physics of the d-dimensional generalized Potts models is given. Multicanonical simulations yield canonical expectation values for a range of temperatures or any other parameter(s) for which appropriate weights can be constructed. We shall address in some details the question how the multicanonical weights are actually obtained. Subsequently miscellaneous topics related to the considered algorithms are reviewed. Then multicanonical studies of first order phase transitions are discussed and finally applications to complex systems such as spin glasses and proteins.