CBE Colloquia - From Electrons to Materials: Machine Learning for Accurate and Scalable Simulations
Supporting the below United Nations Sustainable Development Goals:支持以下聯合國可持續發展目標：支持以下联合国可持续发展目标：
Accurate and efficient first-principles quantum mechanical simulations of molecules and materials are essential for understanding their properties and accelerating materials discovery. However, conventional first-principles methods face a fundamental trade-off between accuracy and computational cost. Recent advances in application of machine learning in first-principles methods offer promising opportunities to overcome these limitations.
In this seminar, I will introduce three complementary approaches that integrate machine learning with first-principles quantum mechanical methods and materials simulations. First, I will introduce machine-learning interatomic potentials, especially ReaxNet, a reactive MLIP that enables large-scale molecular dynamics simulations with near-DFT accuracy at substantially reduced computational cost. Applications to solid-state batteries will illustrate how such simulations can reveal complex atomic-scale processes. Second, I will present Δ-learning approaches that systematically improve the accuracy of quantum chemical predictions by learning corrections from experimental data, with applications ranging from molecular interactions to battery and optoelectronic materials. Finally, I will discuss the development of machine-learned exchange–correlation functionals, highlighting recent progress toward resolving longstanding challenges in density functional theory, including delocalization errors and bond dissociation.
Professor Guanhua CHEN graduated from the Department of Physics at Fudan University in 1986. He ranked co-1st place in the China-US Physics Examination and Application (CUSPEA) program, also known as the Tsung-Dao Lee Program. He then went to the California Institute of Technology for further studies, where he obtained his PhD in Physics in 1992. Following his PhD, he conducted postdoctoral research at the University of Rochester. In 1996, he joined the Department of Chemistry at the University of Hong Kong as an Assistant Professor. He was promoted to Professor in 2006 and served as Head of the Department from 2010 to 2016. He is currently a Chair Professor of Chemistry at the University of Hong Kong. In 2020, Prof. Chen founded the Hong Kong Quantum AI Lab (HKQAI) and has been serving as the center director since.
He has dedicated his research to the accurate numerical solution of the Schrödinger equation, a fundamental equation in quantum mechanics, which can be applied to predict the physicochemical properties of materials and design innovative optoelectronic devices. In recognition of his achievements, Professor Chen was elected a Fellow of the Royal Society of Chemistry in 2011, a Fellow of the American Physical Society in 2014, and received the Croucher Senior Research Fellowship in 2016.
In 2016, together with Mr. Neil Shen, Founder of Sequoia Capital China, and Professor Li Zexiang, former Chairman of DJI, he co-founded the Hong Kong X-Tech Startup Platform to promote technology entrepreneurship in Hong Kong. The platform has been dedicated to supporting and investing in promising technology startups, with a particular focus on hard-tech startups.