Development of an efficient reaction path exploration method via automated transition state search and machine learning

Abstract

Reaction mechanism is a key concept in chemistry. Chemists elucidate reaction mechanisms to gain new chemical insights, eventually contributing to solving various related problems, including chemistry, industry, biology, environment, and so on. In order to minimize human effort in reaction pathway exploration, automated reaction exploration methods have been developed for decades, which can explore numerous reaction pathways, leaning on their high computational power. Among several methods, graph-theoretic methods have emerged as one of popular ones. There have bee a number of studies using these methods for reaction mechanism exploration. However, there are still many things to be improved, especially in terms of efficiency. In this thesis, we present several works contributing to improve this efficiency, in particular focusing on the automated transition state (TS) search through some development of robust algorithms to facilitate TS search and the application of machine learning to accelerate reaction exploration.