Методы и распределенная программная инфраструктура для численного решения задачи поиска молекулярных кластеров с минимальной энергией

The problem is considered of searching for a geometrical structure of a cluster of identical atoms, where the interaction between the atoms is described by pair potentials. Stochastic optimization methods applied to calculations of this problem are highly time-consuming which hampers their use in the analysis of structures with a large number of atoms. A distributed software infrastructure which realizes stochastic methods in a Grid environment is proposed. This infrastructure enables the concentration of globally distributed computational resources in an integrated computing environment with unlimited scalability. Computing experiment results confirming the efficiency of the proposed approach of searching for a minimum energy molecular cluster conformation are presented. The influence of different parameters on calculation speed and accuracy is investigated.