Periodic Coulomb Tree Method: An Alternative to Parallel Particle Mesh Ewald
Periodic Coulomb Tree Method: An Alternative to Parallel Particle Mesh Ewald
复制标题
DOI：
10.1021/acs.jctc.9b00648
复制
发表时间：
2020-01-01
影响因子：
5.5
通讯作者：
Boateng, Henry A.
中科院分区：
文献类型：
--
作者：
Boateng, Henry A.
粒子网格Ewald（PME）通常是在具有周期性边界条件的模拟中处理静电的首选方法。PME在低处理器数量下的出色效率主要是由于使用了快速傅立叶变换（FFT）。然而，由于FFT的高通信成本，PME并行扩展性差。我们开发了一个周期性库仑树（PCT）方法的静电相互作用在周期性的边界条件作为一种替代PME并行模拟。我们验证PCT的准确性，通过MD模拟使用PME和PCT获得的三个不同的系统的结构和动力学特性的比较，并提供并行的时间比较的两种方法上多达1024个核心。
Particle mesh Ewald (PME) is generally the method of choice for handling electrostatics in simulations with periodic boundary conditions. The excellent efficiency of PME on low processor counts is largely due to the use of the fast Fourier transform (FFT). However, due to the FFT's high communication cost, PME scales poorly in parallel. We develop a periodic Coulomb tree (PCT) method for electrostatic interactions in periodic boundary conditions as an alternative to PME in parallel simulations. We verify the accuracy of PCT by comparison of structural and dynamical properties of three different systems obtained via MD simulations using PME and PCT and provide parallel timing comparisons of the two methods on up to 1024 cores.