1 MIN READ

Machine Learning of Chemical Transformations from Atom-Resolved Experiments

Machine Learning of Chemical Transformations from Atom-Resolved Experiments
(Top) Dynamic STEM experiments capture Si impurities moving at the edge of a hole in graphene. (Bottom) Machine learning analysis of dynamic STEM data allows unsupervised identification of structural building blocks and determination of the individual kinetic constants for their transformations. Markov transition matrix for the combined classes of Si–C edge configurations.

Scientific Achievement

Developed a machine learning method to map the chemical transformation networks from scanning transmission electron microscopy movies.

Significance and Impact

Mapping chemical reactions in solids, including identification of relevant structural units and their transformations, enables fundamental understanding of the chemistry and reaction pathways of systems for energy and information technology applications.

Research Details

– Scanning transmission electron microscopy (STEM) was used to image the dynamics of Si impurities at the edge and in the bulk of a graphene monolayer.

– A machine learning pipeline was developed to learn atomic features from noisy data and categorize them into dissimilar configurations.

– Transition probabilities between these configurations define kinetic constants of individual reactions.

M. Ziatdinov, O. Dyck, S. Jesse, and S. V. Kalinin, “Atomic mechanisms for the Si atom dynamics in graphene: chemical transformations at the edge and in the bulk,” Adv. Funct. Mater. 1904480 (2019). DOI: 10.1002/adfm.201904480

Related Researchers