报告人：吕健
报告题目：Data-Driven Methods for Crystal Structure Prediction
报告时间：2026年10月15日（周四）上午10：00
报告地点：物理与光电工程学院428会议室
主办单位：物理与光电工程学院、江苏圣理工学院—中俄学院、科学技术研究院
报告人简介：
吕健, 吉林大学唐敖庆领军教授，国家级青年拔尖人才，2013年在吉林大学获得博士学位，2013年至2016年在北京计算科学研究中心从事博士后研究工作，2016年入职吉林大学物理学院；长期从事计算凝聚态物理研究，专注于发展从理论上预测材料微观结构的计算方法与软件；作为核心开发者之一，发展了以CALYPSO命名的晶体结构预测方法及软件，解决了系列高压和低维等限域条件下的凝聚态物质结构难题；在 Acc. Chem. Res., Phys. Rev. Lett. Nat. Commun.,以及 J. Am. Chem. Soc. 等期刊上发表论文100余篇，引用超过16,000次。
报告摘要：
Crystal structures are key to deeply understanding the various physical and chemical properties of condensed matter, and they also form the basis for theoretical materials modeling, simulations and discovery. Theoretical crystal structure prediction of materials, given only their chemical compositions, is crucially important. However, this task is extremely challenging as it involves identifying a vast number of energy minima on the lattice energy surface.
In this report, I will first provide a brief introduction to the CALYPSO structure prediction method. Following that, I will detail our recent advancements in accelerating structure prediction through data-driven approaches, such as machine learning potentials and generative models. Through several benchmark and practical applications, we demonstrate that data-driven approaches are heralding a new era in crystal structure prediction.