Reproduce “Neural Network Path Optimization for Finding Transition States on a Machine Learned Potential” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Kareem Hegazy, Eric C-Y Yuan, Anup Kumar, Dmitriy Morozov, Daniel Barter, Michael W Mahoney, Teresa Head-Gordon, Samuel M Blau

Popcornn finds transition states between reactants and products by optimizing a neural-network-defined continuous path rather than discrete snapshots. Adaptive integration selects where to evaluate a machine-learned interatomic potential, and gradients update the path. On 1,342 out-of-distribution reactions, it recovered both intended endpoint connectivities in 81%, versus 68% for the strongest NEB setup; evidence is limited to small, non-periodic organic reactions.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.