An Integrated Parallelizable Algorithm for Computer Motion Simulation of Large-Sized Bio-Molecular Structures

摘要

In this paper, an integrated parallelizable algorithm is presented for computer simulation of dynamics of multibody molecular structures in polymers and biopolymers. The algorithm is developed according to an integrated O(N) simulation procedure developed by the author for calculating interatomic forces and forming/solving equations of motion for large-sized bio-molecular structures. Specifically, the simulation procedure is created via a proper integration between a parallelizable multibody molecular simulation method (PMMM) produced by the author and a parallelizable fast multipole method (PFMM). PFMM is utilized for calculation of atomic forces such as Van der Waals and Coulomb attractions between the atomics in the molecular structures. The parallelizable multibody molecular method is used for forming/solving equations of motion of large-sized molecular structures in polymers and biopolymers. Currently, the calculation of interatomic forces and formation/ solution of equations of motion are treated separately by various procedures. For instance, Fast Multipole Method (FMM) and Cell Multipole Method (CMM) are applied for calculating interatomic forces only. Cartesian Coordinate Method (CCM) and Internal Coordinate Molecular Dynamics Method (ICMM) have been introduced independently for forming/solving equations of motion. Though formation and solution of equations of motions, and atomic force calculations are needed for same molecular structure, there is no direct conversation between two group methods. The proposed algorithm integrates multibody molecular method with fast multipole method in a parallel fashion so that both calculating atomic forces and forming/solving equations of motion can be carried out concurrently in a combined procedure. Computational loads associated with these two simulation tasks then can be divided among sub-chains, and each sub-chain is allocated to a processor on a parallel computing system via a proper integration between PFMM and PMMM. The algorithm can be used on both shared-memory and distributed-memory parallel computational systems. Compared with its counterpart of the integrated O(N) procedure developed by the author before, this algorithm has a computational complexity of O(logN) theoretically (N is number of subsets). The algorithm may find its applications for force calculation and motion simulation associated with large-sized molecular structures of polymers and biopolymers.