A Highly Parallelizable Long Time Dynamical Simulation Algorithm Applied to Proteins

9309883 Rabitz我们提出了一种高度可并行化的分子动力学(MD)方法来研究凝聚相系统的长时间动力学。我们打算将我们的方法应用于蛋白质动力学，其中长时间行为是非常感兴趣的，所提出的方法可以充分利用它们的优势。该方法约束了完全耦合系统的高频运动，从而大大降低了问题的维度，同时允许使用非常大的积分时间步长。该方法是使用系统的低频实数和虚数简正波的周期性更新子空间作为动力学数值积分的约化基础。

9309883 Rabitz We propose a highly parallelizable molecular dynamics (MD) approach for the study of the long time dynamics of condensed phase systems. We intend to apply our method to protein dynamics, where the long time behavior is of great interest, and the proposed methods can be used to their full advantage. The method constrains high frequency motions of the fully coupled system, thus greatly reducing the dimensionality of the problem while simultaneously permitting the use of a very large integration time step. The approach is to use a periodically updated subspace of the low frequency real and imaginary normal modes of the system as a reduced basis for numerical integration of the dynamics.