A Hybrid Parallelizable Algorithm for Computer Simulation of Rigid Body Molecular Dynamics
Molecular dynamics is effective for a nano-scale phenomenon analysis. This paper presents a hybrid parallelizable algorithm for the computer simulation of the motion behavior of molecular chain and open-tree structure on parallel computing system. The algorithm is developed from an approach of rigid body dynamics, in which interbody constraints are exposed so that a system of largely independent multibody subchains is formed. The increased parallelism is obtainable through bringing interbody constraints to evidence and the explicit determination of the associated constraint forces combined with a sequential O(n) procedure. Each subchain then is assigned to a processor for parallel computing. The algorithm offers a sequential O(n) performance if there is only one processor available. The algorithm has O(log2n) computational efficiency if there are as many processors available as number for molecular bodies. For most common scenario, the algorithm will give a computational complexity between O(n) and O(log2n) if number of available processor is less than number of molecular bodies.