Temperature Scaling Method for Markov Chains

TitleTemperature Scaling Method for Markov Chains
Publication TypeJournal Article
Year of Publication2009
AuthorsCrosby LD, Windus TL
Journal TitleJournal of Physical Chemistry A
Date Published01/22
ISBN Number1089-5639
Accession NumberISI:000262522300011
Keywordschemistry, clusters, dynamical nucleation, free-energy perturbation, kinetics, liquid, monte-carlo, performance, tip5p water model, walking

The use of ab initio potentials in Monte Carlo simulations aimed at investigating the nucleation kinetics of water clusters is complicated by the computational expense of the potential energy determinations. Furthermore, the common desire to investigate the temperature dependence of kinetic properties leads to an urgent need to reduce the expense of performing simulations at many different temperatures. A method is detailed that allows a Markov chain (obtained via Monte Carlo) at one temperature to be scaled to other temperatures of interest without the need to perform additional large simulations. This Markov chain temperature-scaling (TeS) can be generally applied to simulations geared for numerous applications. This paper shows the quality of results which can be obtained by TeS and the possible quantities which may be extracted from scaled Markov chains. Results are obtained for a 1-D analytical potential for which the exact solutions are known. Also, this method is applied to water clusters consisting of between 2 and 5 monomers, using Dynamical Nucleation Theory to determine the evaporation rate constant for monomer loss. Although ab initio potentials are not utilized in this paper, the benefit of this method is made apparent by using the Dang-Chang polarizable classical potential for water to obtain statistical properties at various temperatures.

URL<Go to ISI>://000262522300011