Prof. Wenbo XIE, 謝聞博

ORCID iD: 0000-0001-5321-2494
Scopus Author ID: 57222383166

Metrics for Academic and Research Impact

Dr. Wenbo Xie received his B.S. in Chemistry from the University of Georgia in 2014 and his Ph.D. in Chemistry from Queen’s University Belfast in 2023 under the supervision of Prof. Peijun Hu. Before joining City University of Hong Kong in 2026, he was a Research Assistant Professor at the School of Physical Science and Technology, ShanghaiTech University.

His research lies at the intersection of artificial intelligence, materials science, and theoretical chemistry. He develops machine-learning methods for atomistic simulation, with particular interests in data-centric learning, machine-learning interatomic potentials, and realistic modeling of complex catalytic and materials systems. His work aims to bridge first-principles accuracy with the length and time scales required to describe materials under realistic working conditions, enabling more predictive simulations and accelerated materials discovery.

His theoretical research has been published in Nature Catalysis, Angewandte Chemie International Edition, JACS Au, Accounts of Chemical Research, and ICML.

1. Data-Centric AI for Materials Modeling

Rather than simply scaling up model size, we ask a more fundamental question: what atomistic data are truly needed to capture and reproduce the dynamic behavior of complex materials and catalytic systems? We develop new strategies for data generation, sampling, and selection, enabling machine-learning models to learn complex potential-energy landscapes from substantially less but more informative data.

2. Next-Generation Machine-Learning Potentials

We develop next-generation machine-learning interatomic potentials and materials foundation models that extend near-first-principles accuracy to much larger spatial and temporal scales. Our goal is to build accurate, efficient, and transferable models capable of describing increasingly complex chemical environments.

3. Realistic Materials Simulation

We combine machine-learning potentials with advanced sampling and simulation methods to study complex heterogeneous catalytic systems that remain difficult to access directly with conventional DFT. We aim to understand how catalysts dynamically restructure under realistic reaction conditions and how these structural changes ultimately govern their function and reactivity.

4. AI-Driven Materials Discovery

We seek to connect the full materials-discovery pipeline, from data generation and foundation models to atomistic simulation, property prediction, and automated experimental validation, and ultimately build an AI-powered simulation engine for materials discovery.