EFFECTIVE WAY FOR DETERMINATION OF MULTICANONICAL WEIGHTS

In the last few years generalized ensemble algorithms have become popular as a way to overcome the exponentially slow convergence in numerical simulations of systems with a rough energy landscape. Prominent examples of such an approach are the multicanonical algorithm @1,2#, 1/k sampling @3#, and simulated tempering @4,5#. These algorithms explore larger parts of the phase space than canonical molecular dynamics or Monte Carlo, which at low temperatures are easily trapped in one of the huge number of local minimas. This is because in the canonical ensemble the probability to cross an energy barrier of heights DE is proportional to e at temperature T . On the other hand, in generalized ensembles the probability to cross an energy barrier is independent of temperature. For instance, in the multicanonical algorithm @1,2#, configurations with energy E are updated with a weight: