By: António Luis Ferreira
From: Univ Aveiro
At: Instituto de Investigação Interdisciplinar, Anfiteatro
Monte-Carlo algorithms in the multicanonical ensemble and other non-Boltzmannian ensembles are used to get estimates of the density of states of lattice spin systems and fluid systems. Those dynamics suffer from critical slowing down and exhibit tunnelling times with a power-law increase with system size. The slowing down of dynamics with system size is a major problem for applications because it determines increasing statistical errors with system size. Using an approximate projection in the macro-state space of the stochastic dynamics I show that it is possible to obtain accurate information on the full phase space dynamics by doing a static calculation that estimates a projected transition matrix. I also discuss the possibility to fine tune the simulation ensemble in order to reduce the slowing down of the dynamics.