(191l) Integrating Computational Reaction Discovery in the Ab Initio Nanoreactor with Kinetic Modeling and Sensitivity Analysis

Computational reaction discovery in complex chemical systems has been an important research topic. Traditionally, this has been carried out through a hypothesis-driven approach, which requires incorporation of certain chemistry rules to guide the reaction discovery. Recent advances in computational chemistry have led to a new era for unravelling complex reactions in a hypothesis-free manner. In particular, our previous achievements in leveraging graphical processing units (GPUs) to solve the electronic Schrödinger equation led to the development of the ab initio nanoreactor, a computational suite for automatic chemical reaction discovery. In this presentation, we apply the ab initio nanoreactor framework to methane pyrolysis reaction network, from the automatic reaction finding to kinetic modeling. Reactions during methane pyrolysis are initially discovered using GPU-accelerated ab initio molecular dynamics. These reaction paths are refined using the growing string method at a higher level of theory. The reaction rate coefficients are calculated using transition state theory based on these optimized reaction paths. The kinetic model is constructed with 53 species and 134 reactions, and validated against experimental data and simulations from literature kinetic models. Beyond kinetic modeling and validation, we further introduce the local brute force and Monte Carlo sensitivity analysis approaches, both of which show great potential for improving the accuracy of methane pyrolysis kinetic model. We envision that the future ab initio nanoreactor framework is carried out iteratively, through integrating the computational reaction discovery with kinetic modeling and sensitivity analysis.