MSE Talk: Thomas Senftle, Rice University
- Starts: 3:00 pm on Friday, October 18, 2024
- Ends: 4:00 pm on Friday, October 18, 2024
Title: Understanding Metal-Support Interactions in Catalytic Materials with Symbolic Regression
Abstract: The performance of heterogeneous catalysts composed of metal nano-clusters on oxide supports often is impacted by charge transfer across the metal-support interface. Charge transfer alters the oxidation state of the supported metal, which in turn changes its catalytic properties. Charge transfer also influences the metal’s adhesion to the support surface, affecting sintering rates and cluster size distributions. Thus, understanding interfacial charge transfer is essential to controlling the overall activity, selectivity, and stability of these catalytic materials. Descriptors for predicting the extent of charge transfer between the metal and support can be derived from intuition in certain simple cases. However, this task becomes increasingly difficult for complex systems in which interrelated phenomena act together to influence the nature of the charge transfer. In this seminar, I will discuss how we can tackle these intricacies by applying symbolic regression (SR) in concert with density functional theory (DFT). DFT is used to generate metal adsorption energy data for a range of metal-support pairs in the presence of common adsorbates from the reaction environment. This data then serves as the training set for SR tools that scan a feature-space of descriptor candidates for combinations of properties (e.g., electronegativity, ionization potential, electron affinity, etc.) that correlate with the adhesion strength between the metal and the support. This methodology helps us to better understand charge transfer in these systems, as well as provides predictive models that can be used to design supports that take full advantage of beneficial metal-support interactions. I also will highlight the promising performance of a new SR workflow, which we call iterative Bayesian additive regression trees (iBART), that significantly increases computational efficiency without sacrificing accuracy compared to other state-of-the-art machine learning methods.
Bio: Thomas Senftle is an Associate Professor in the Department of Chemical and Biomolecular Engineering at Rice University. Prof. Senftle’s current research focuses on the development and application of computational modeling tools for assessing multi-component catalytic materials at both electronic and atomistic scales. Emphasis is placed on the rational design of catalytic systems for efficient energy conversion, storage, and utilization. He is a recipient of the ACS-PRF Doctoral New Investigator Award, the ACS-GHS Younger Chemist Award, the NSF CAREER Award, and the George R. Brown School of Engineering Teaching + Research Excellence Award.
- Location:
- EMB 105, 15 St. Mary's St.
- Hosting Professor
- Emily Ryan