Many important catalysts are comprised of isolated metal atoms grafted to amorphous silica.  The quenched disordered silica surface imparts unique properties to the individual sites, and results in an ensemble of non-uniform sites with a distribution of activities.  In many cases, a small fraction of atypical sites is responsible for nearly all of the catalytic activity.  Standard computational methods cannot identify the distinguishing features of these rare and unusually active sites, but we developed an "importance learning" algorithm that can.  We briefly outline components of the method, including machine learning, importance sampling, and inpute descriptors for the local environment.  Finally, we present computational results for hydrogenolysis by Zr/SiO2.

Modeling single atom catalysts on amorphous supports: Applications of importance learning

ACS Spring 2022

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.