Multiscale Simulation of Atomic-Scale Friction using Computation Grid

在本项目中，我们开发了原子尺度摩擦（即固流界面现象）的并发混合模拟代码，并对代码进行了优化，使其能够同时使用具有网格计算技术的世界范围内的超级计算机。混合仿真代码将直接应用于MEMS，燃料电池和催化气体转化器等材料中的各种重要现象。利用本课题提出的缓冲簇方法，构建了由量子密度泛函理论（DFT）和经典分子动力学（MD）组成的混合模拟代码。本文采用改进的粗粒粒子法动态给出了MD区域的边界条件。在混合仿真代码中，实现了一种自适应动态选择量子区域的算法，并通过对MEMS中硅基中观元件的原子尺度摩擦的模拟运行，对其性能进行了研究和改进。我们的混合DFT-MD仿真代码已经在世界范围内的计算网格上进行了大规模、长时间的仿真测试。在日本和美国的几台超级计算机上成功地进行了DFT-MD混合模拟。

In the present project, we have developed concurrent, hybrid simulation code for the atomic-scale friction, i.e., solid-fluid interfacial phenomena, and have optimized the code for concurrent usage of the world-wide supercomputers with the grid computation technologies. The hybrid simulation code will be applied directly to various important phenomena in materials such as the MEMS, the fuel cell, and the catalytic gas-converter.The hybrid simulation code that we have constructed is composed of the quantum density-functional theory (DFT) and the classical molecular dynamics (MD) using the buffered-cluster method proposed by us in the present project. The boundary condition for the MD region is given dynamically by the coarse-grained particles method that we have improved in the present project. In the hybrid simulation code, an algorithm for adaptive, dynamic selection of the quantum region has been implemented, whose performance has been investigated and improved through simulation runs about the atomic-scale friction of Si-based meso-scale components in MEMS.Our hybrid DFT-MD simulation code has been tested on a large-scale, long-time simulation on the world-wide computation grid. The hybrid DFT-MD simulation run using several supercomputers in both Japan and USA has been demonstrated successfully.