Speaker

Guo Hong（郭鸿）

加拿大皇家科学院院士

加拿大麦吉尔大学教授

Date＆Time

2022.07.25（Mon）AM 9:30

Location

Zoom Meeting ID：950 680 6742 Password：2022

https://www.koushare.com/lives/room/177541

Reporter

Guo Hong obtained B.Sc. in physics from Sichuan Normal University in Chengdu, China. In 1981, he was selected by the China-U.S. Physics Examination and Application (CUSPEA) and joined the graduate program in the Department of Physics and Astronomy, University of Pittsburgh, Pennsylvania, where he obtained M.Sc. in experimental atomic physics under Prof. James Bayfield, M.Sc. in computer science under Prof. Lawrence Rose and PhD in theoretical condensed matter physics under Prof. David Jasnow. In January 1989, he joined the Physics Department, McGill University, Montreal, Canada, where he is currently a Distinguished James McGill Chair Professor of Physics. His research is in theoretical condensed matter physics, materials physics, electronic device physics, mathematical and computational physics. He was elected to Fellow of the American Physical Society in 2004, Fellow of the Royal Society of Canada (Academy of Sciences) in 2007. He is a recipient of the Killam Research Fellowship Award from the Canadian Council for the Arts; the Brockhouse Medal for Excellence in Experimental or Theoretical Condensed Matter Physics from the Canadian Association of Physicists (CAP), and the CAP-CRM Prize in Theoretical and Mathematical Physics from CAP. He can be reached by email at hong.guo@mcgill.ca.

Abstract

Materials informatics (MI) may be considered the 4th paradigm of scientific inquiry, in addition to experimental, theoretical, and computational approaches. MI is made possible by the universal access to abundant scientific data, assisted by advances in software and machine learning (ML) to analyze the data. For materials problems with specific designing goals, physics-based descriptors are necessary to help narrowing down the informatics search. In this talk I shall present materials discovery by MI + ML, discussing simple examples of discovering 2D ferromagnets and solid-state electrolytes, high temperature superconductors, ending with an extensive screening of energetic materials from a large database containing over 100 million molecules. We conclude that backed by theory and first principles simulation methods, and eventually by experimental verifications, MI + ML can be a very efficient approach for materials discovery.