ML Model Rapidly Predicts Key Chemical Reactions, Enabling New Discoveries
-
Researchers developed a machine learning model that can quickly calculate fleeting transition states of chemical reactions, which determine if a reaction will occur.
-
Model can calculate transition state structures in seconds vs hours/days with traditional quantum chemistry techniques.
-
Model was trained on 9,000 reactions and can accurately predict transition states for new reactions.
-
Potential applications include designing new catalysts and reactions, modeling planetary atmosphere interactions, and early evolution of life chemistry.
-
Model represents major advance in predicting chemical reactivity by automating very difficult transition state calculations.