Seminar: Modeling Heterogeneous Catalysis with General, Reactive Machine Learning Potentials

ABSTRACT

Decades of efforts in improving computing power and methods have steadily expanded what theoretical catalysis can explain and predict. Yet heterogeneous catalysis remains a complex, operando problem: surfaces restructure, coverages change, and reaction networks evolve across a vast chemical landscape. In practice, first-principles calculations still limit our ability to explore that landscape, especially when realism demands large systems, long timescales, and extensive sampling. This gap is why AI and machine learning are becoming essential in heterogeneous catalysis but using them well requires the collective wisdom of data science and physical science.

In this talk, I will discuss why heterogeneous catalysis is entering a “data revolution,” and why that shift creates new challenges as well as new opportunities. I will introduce my work on making machine learning an indispensable tool for simulation by supporting or perhaps ultimately replacing electronic-structure calculations via general and reactive element-base machine learning potentials. I will conclude by what I see as the most pressing challenges and promising research directions in the field.

BIOGRAPHY

Dr. Wenbo Xie is a Research Assistant Professor in the School of Physical Science and Technology at ShanghaiTech University. He received his Ph.D. from Queen’s University Belfast under the supervision of Prof. Peijun Hu. His research focuses on realistic modeling of heterogeneous catalysis and developing AI/machine learning methods to address the complexities in catalytic systems. His theory works have been published in Nat. Catal., Angew. Chem. Int. Ed., JACS Au., Acc. Chem. Res., etc.