Molecular dynamics in arbitrary geometries: Parallel evaluation of pair forces
Molecular dynamics in arbitrary geometries: Parallel evaluation of pair forces
复制标题
DOI：
10.1080/08927020801930554
复制
发表时间：
2008-01
影响因子：
2.1
通讯作者：
G. Macpherson;J. Reese
中科院分区：
文献类型：
--
作者：
G. Macpherson;J. Reese
提出了一种在分布式并行计算机上计算分子动力学（MD）模拟中分子间作用力的新算法。任意相互作用细胞算法（AICA）的目的是操作上的几何域定义的非结构化，任意多面体网格，已被空间分解成不规则的部分并行化。它旨在通过MD在复杂几何形状中进行纳米尺度流体力学模拟，并提供混合MD/连续模拟的MD组件。在模拟开始时计算网格单元的空间关系，并且仅需要包含在单元中的分子相互作用，所述单元的表面的一部分比分子间对势的截止半径更近。AICA已在开源C++代码OpenFOAM中实现，其准确性已通过已发布的MD代码间接验证。在12和32个处理器上进行串行和并行模拟的相同系统给出了相同的结果。性能测试表明，有一个最佳数量的细胞在一个网格中的最大速度计算分子间力，并在网格中有大量的空细胞不会增加显着的计算开销。
A new algorithm for calculating intermolecular pair forces in molecular dynamics (MD) simulations on a distributed parallel computer is presented. The arbitrary interacting cells algorithm (AICA) is designed to operate on geometrical domains defined by an unstructured, arbitrary polyhedral mesh that has been spatially decomposed into irregular portions for parallelisation. It is intended for nano scale fluid mechanics simulation by MD in complex geometries, and to provide the MD component of a hybrid MD/continuum simulation. The spatial relationship of the cells of the mesh is calculated at the start of the simulation and only the molecules contained in cells that have part of their surface closer than the cut-off radius of the intermolecular pair potential are required to interact. AICA has been implemented in the open source C++ code OpenFOAM, and its accuracy has been indirectly verified against a published MD code. The same system simulated in serial and in parallel on 12 and 32 processors gives the same results. Performance tests show that there is an optimal number of cells in a mesh for maximum speed of calculating intermolecular forces, and that having a large number of empty cells in the mesh does not add a significant computational overhead.