Decoding Reactive Structures in Catalysts by Machine Learning Analysis of Spectra

Anatoly I. Frenkel

Stony Brook University (Stony Brook, NY, US)

Monday, 20th December 2021,16:00 s.t.

The talk will be given online (due to COVID restrictions).

You can join via Zoom:
Meeting ID: 881 3006 2634    Passcode: 304519

Decoding Reactive Structures in Catalysts by Machine Learning Analysis of Spectra

Detecting active species and active sites in nanocatalysts is a major challenge due to the paucity of experimental techniques that can provide atomic-level information for metal species in harsh reaction conditions. We have recently demonstrated that machine-learning methods can be used to decipher information about the three-dimensional geometry of mono- and bimetallic nanoparticles, clusters and “single-atom” catalysts encoded in their X-ray absorption spectra. In other words, we trained a computer to learn how to “invert” the unknown spectrum of a catalyst and map it onto the underlying structural and electronic descriptors. We have also learned how to estimate the a priori unknown number of these descriptors by limiting that to be less than the number of nodes in the latent space of the autoencoder. These applications are demonstrated by reconstructing the geometric shapes and compositional distributions in nanocatalysts studied under in situ and operando conditions from their spectra. An application of these methods to the determination of catalytic descriptors in operando conditions, such as studies of reactivity of dilute Pd-in-Au catalysts for HD-exchange reaction, will be demonstrated.

Bio of Anatoly I. Frenkel

Anatoly Frenkel is a Professor in the Department of Materials Science and Chemical Engineering at the Stony Brook University and a Senior Chemist (Joint Appointment) at the Division of Chemistry, Brookhaven National Laboratory. He received his M.Sc. degree from St. Petersburg University and Ph. D. degree from Tel Aviv University, all in physics, followed by a postdoctoral appointment at University of Washington (Seattle). His research interests focus on development and applications of in situ and operando synchrotron methods to solve a wide range of materials problems, with most recent emphases on catalysis, ranging from nanoparticles to “single-atom” catalysts, electromechanical materials, filtration materials, quantum dots, as well as machine-learning methods for structural analysis and design of nanomaterials. He is a founding Principal Investigator for the Defense Synchrotron Consortium, and the Spokesperson for the Synchrotron Catalysis Consortium, both at Brookhaven National Laboratory. He is a Fellow of the American Physical Society and the author of over 400 publications.