Exploring Chemical Reactions Through Automation and Machine Learning

Explore the potential of machine learning in simulating chemical reactions through this 22-minute seminar by Prof. Fernanda Duarte from the University of Oxford. Delve into the world of machine learned potentials (MLPs) and their application in efficiently mapping nuclear configurations to energies. Discover why, despite their development over a decade ago, MLPs have not yet become routine in chemical reaction simulations. Learn about innovative strategies to generate MLPs for studying reaction mechanisms in both gas-phase and solution, achieving ab initio accuracy with minimal computational resources. Examine the performance and limitations of various methods for obtaining reactive MLPs across different reaction scales. Explore real-world applications of these strategies in diverse systems, including solution-based reactions and ambimodal surfaces, as well as their potential for predicting dynamical quantities like product ratios and free energies. Gain insights into the future of chemical reaction simulation and the role of machine learning in advancing our understanding of reaction mechanisms.

Exploring chemical reactions through automation and machine learning