Chemical mechanisms are essential for modeling chemical processes, particularly in combustion and pyrolysis, where high temperatures produce a wide range of species. The complexity of these processes, compounded by the introduction of biomass-derived fuels, presents challenges in applying existing knowledge about reaction rates and pathways from fossil fuel combustion. Reactive Molecular Dynamics (RMD) simulations offer a way to explore reactions with minimal prior knowledge, making this approach ideal for combustion and pyrolysis studies. ChemTraYzer, a tool that automates RMD analysis, identifies the reactions occurring during simulations. However, the complexity of these processes often results in extensive reaction networks, complicating the understanding of the underlying mechanisms. In this thesis, I evaluate two reaction exploration methods using ChemTraYzer and accelerated dynamics to identify reactions of ethyl-2-yl formate, an intermediate in biofuel combustion. Both methods successfully identify key pathways, including decomposition, cyclization, and hydrogen migration. By comparing the results, I find that each method discovers additional reaction pathways that complement the other. This suggests that using both methods together provides broader coverage of the reaction space while keeping computational costs manageable. I also tackle the challenge of analyzing the complex reaction networks generated by RMD simulations. I develop a methodology that integrates ChemTraYzer with the Nudged-Elastic-Band (NEB) method to identify and validate key reaction paths, extending existing chemical mechanisms. Although ReaxFF is used for rapid force calculations, I show that validating these reactions with higher-level quantum mechanical methods is essential due to ReaxFF’s limitations, particularly in computing reaction barriers and addressing spin conservation. Hydrocarbon pyrolysis and soot formation serve as the case study, generating a large and complex reaction network. Finally, I propose a method for classifying reactions based on data obtained from RMD simulations. This classification approach simplifies large reaction networks by abstracting reaction pathways into broader categories, facilitating a clearer understanding of complex chemical processes.
Felix Schmalz (Wed,) studied this question.