煤热解模拟新方法—ReaxFF MD的GPU并行与化学信息学分析

This paper overviews the new methodology that combines GPU-based high performance computing and cheminformatics analysis for ReaxFF MD simulation. The methodology was proposed to overcome the obstacles of the expensive computational costs and reaction analysis capability lacking in exploration of chemical reaction mechanisms of coal pyrolysis. With the creation of GMD-Reax and VARxMD, the first GPU-enabled ReaxFF MD codes with significant improved computational performances and the first tool allowing for chemical reaction analysis and visualization of ReaxFF MD trajectory, large scale coal models with about 10000 atoms can be simulated. The observed evolution of pyrolyzates with temperature and time, and the detailed chemical reactions obtained in the simulations can provide additional insights on coal pyrolysis, which is hardly accessible experimentally or with other computational methods. We believe that the methodology is promising for a deep and comprehensive understanding of the complex reaction mechanism in coal pyrolysis.