AIMNet2-rxn: A machine-learned potential for generalized reaction modeling on a millions-of-pathways scale.

AIMNet2-rxn achieves 1-2 kcal/mol accuracy in reaction modeling across millions of pathways, enabling rapid and precise mechanistic predictions for complex chemical reactions.

• Why it matters: Accurate reaction modeling is essential for understanding reactivity and predicting outcomes, but existing machine-learned potentials lack the precision needed for diverse reactions, limiting their practical use.
• What they did: AIMNet2-rxn was developed using approximately 4.7 million density functional theory calculations to create a general machine-learned interatomic potential that accelerates reaction simulations by a million-fold compared to quantum mechanical methods.
• The result: This model allows high-throughput reaction discovery and detailed pathway analysis, exemplified by thermodynamic evaluation of a complex 11-step pathway, opening new avenues for chemical research and process optimization.