Hierarchical Clustering of Structural and Electronic Characteristics Obtained from Molecular Dynamics Simulation of Catalytic Reaction on Metal Nanoparticle

Unsupervised machine learning (ML) is examined for the result of molecular dynamics (MD) simulation to extract characteristics catalytic reaction. O–H bond dissociation ethanol on Fe–Co nanoparticle in ab initio MD [S. Fukuhara et al., Chem. Phys. Lett. 731 (2019) 136619] employed as an example. Hierarchical clustering radial distribution function successfully classifies coordinates reaction dendrogram. Moreover, receiver operating characteristic curve reveals distance farthest-neighbor atom from target a dominant descriptor clustering. An optimum structure predicted based these automated analyses. This study shows new way post-process results simulations unsupervised technique and it paves possibility ML-based materials design.