Revealing chemical reactions of coal pyrolysis with GPU-enabled ReaxFF molecular dynamics and cheminformatics analysis

The complex chemistry of coal pyrolysis is difficult to be captured by experimental techniques or simulated with the quantum mechanics computational methods. The emerging of both the large-scale coal models and the promising capability of reactive molecular dynamics (ReaxFF MD) motivated us to develop a new methodology by combining graphics processing unit (GPU)-enabled high performance computing with cheminformatics analysis in order to explore the coal pyrolysis mechanisms using ReaxFF MD. The methodology is rooted in two new software tools, GMD-Reax, the first GPU-enabled ReaxFF MD codes that make it practical to simulate large-scale models (～10,000 atoms) on desktop workstations, and visualisation and analysis of reactive molecular dynamics (VARMD), the first software dedicated to analysis of detailed chemical reactions from the trajectories of ReaxFF MD simulation. With this methodology, reasonable product profiles and gas generation sequences of pyrolysis for bituminous coal models ranging from ～1000 to ～10,000 atoms (including the system with 28,351 atoms, one of the largest systems used in ReaxFF MD) have been obtained. The complex and detailed chemical reactions directly revealed by VARMD can provide further information on radical behaviours and their connection with pyrolysates. The methodology presented here offers a new and promising approach to systematically understand the complex chemical reactions in thermolysis of very complicated molecular systems.