Some practical considerations in the use of Metropolis and force-biased Monte Carlo simulations on liquids (Q1072822)

A series of four force-biased and two Metropolis Monte Carlo calculations were run on a dense Lennard-Jones fluid just above its melting point. Our purpose was to test some recent theoretical predictions [(*) see the review above (Zbl 0587.76013)] as to the relative speed and stochastic characteristics of these algorithms, and to learn what the most propitious maximum step-size was, in each case. We found that the principal predictions in (*) were upheld for this system, that the Metropolis algorithm was speeded up by using the larger of the two maximum step-sizes tried, that the force-biased algorithm with the full bias was speeded up by using the smaller of the two step-sizes tried, and that the force-biased algorithm, with the bias partially on, was relatively insensitive to the choice of this parameter.