Getting the most from molecular simulation

Abstract

Advanced molecular simulation methods are surveyed, emphasizing techniques that can be used to improve the quality of the results obtained by a simulation. Consideration is given to three aspects of simulation performance: sampling of configurations, analysis of data and evaluation of free energies. Monte Carlo methods are highlighted. Methods are discussed in the context of the important regions of phase space for different systems. It is stressed that the shape of the important region for a single system, or the relation between important regions for two different systems, can profoundly affect the performance of a simulation applied to them. Categories of relations for the regions are (i) no overlap; (ii) partial overlap; or (iii) a subset. The last case is especially important because it implies that there is an otherwise hidden asymmetry in the relation between the systems—this asymmetry impacts the performance of simulations applied to evaluate property differences between such systems. Effective application of advanced simulation methods sometimes requires awareness of these issues, indicating that the practice of simulation might be advanced by the development of ways to quantify shapes and relationships of phase-space regions.