Implementation of Hybrid Monte Carlo (Molecular Dynamics) Quantum Mechanical Methodology for Modeling of Condensed Phases on High Performance Computing Environment

Abstract

The overall objective of the present work is to develop and implement a novel multi-step general computational methodology for modeling of complex condensed-phase systems on high-performance computing environments. First, molecular dynamics (MD) or Monte Carlo (MC) simulations of the free interacting clusters, as well as of clusters microsolvated by several molecules from the medium (solvent) are performed.MD orMC simulations are carried out applying either classical empirical interaction potentials, or implementing quantum mechanical MD or MC methodologies. Quantum mechanical MD simulations are carried out with the Born-Oppenheimer approach (BOMD), the Car-Parrinello (CPMD) approach, or using the atom-centered density matrix propagation scheme (ADMP). Sequential to this step, a series of suitably chosen configurations from the statistical physics simulations corresponding to the equilibrated system, which are mutually statistically independent, are subjected to further more rigorous quantum mechanical analysis. In this way, a realistic simulation of complex physico-chemical systems is enabled, in real computational time, without loosing, in statistical sense, any relevant information about the system. Due to the complexity of the algorithms which are used for this hybrid approach, it is of crucial importance to be able to implement the computational strategy on high-performance computing environment. Often, the overall CPU time which is required is very high. Therefore, achieving good parallel efficiency for calculations of such type is far from a trivial task without the use of high-performance low-latency MPI interconnect.

This paper is based on the work done in the framework of the SEE-GRID-SCI FP7 EC funded project, with partial support from NSFB grant D002 - 146/2008.