MSE Colloquium: Liang Qi, Computational Designs of Materials: Machine Learning vs. Human Learning

Assistant Professor, Department of Materials Science and Engineering, University of Michigan

All dates for this event occur in the past.

264 MQ
105 W. Woodruff Ave
Columbus, OH 43210
United States

Abstract

To apply statistics and data science tools to aid computational designs of materials is under fast development. Different than other fields, such as image recognition, there are two unique aspects for the applications of data science and machine learning in materials science. First, the training sets are usually very small. Second, physical mechanisms of material properties can be applied to facilitate the constructions of descriptors and machine learning methods. In this talk, I will give two examples to address these two problems. The first example is to use machine learning to predict density and elastic moduli of SiO2-based glasses1. Our machine learning approach relies on a training set generated by high-throughput atomistic simulations and a set of elaborately constructed descriptors with the fundamental physics of interatomic bonding. The predictions of our model are comprehensively compared and validated with a large amount of both simulation and experimental data. The second example is the discovery of local electronic descriptors in alloys for predicting their solute-defect interactions2, which play essential roles in determining their mechanical and functional properties. Here we present a general linear correlation between two descriptors of local electronic structures and the solute-defect interaction energies in binary alloys of refractory metals with transitionmetal substitutional solutes. The correlation accuracy can be further improved by a residual-corrected nonparametric regression model. This discovery process was achieved based on “human learning” from classical electronic models. These results provide the possibility for quantitative and efficient predictions on the solute-defect interactions and defect properties in alloys.

References:
1. Y.J. Hu, G. Zhao, B. Bin, T. Del Rose, Q. Zhao, Q. Zu, Y. Chen, X. Sun, M. de Jong
and L. Qi, “Predicting densities and elastic moduli of SiO2-based glasses by machine
learning”, under consideration by npj Computational Materials (2019)
2. Y.J. Hu, G. Zhao, B. Zhang, C.M. Yang, Z.K. Liu, X.F. Qian and L. Qi, “Local
electronic descriptors for solute-defect interactions in bcc refractory metals”, Nature
Communications, 10, Article number: 4484 (2019)

Bio

 

qi_portrait.jpg
Dr. Liang Qi

Dr. Liang Qi is an assistant professor in Department of Materials Science and Engineering at University of Michigan, Ann Arbor starting from 2015. He studied Materials Science and Engineering at Tsinghua University in China and got his bachelor’s degree in 2003. He earned his master's degree in Department of Materials Science and Engineering at the Ohio State University in 2007 and his doctoral degree in materials science and engineering at University of Pennsylvania in 2009. From 2009 to 2012, he worked as a postdoctoral research fellow at UPenn and Massachusetts Institute of Technology. Between 2012 and 2014, he worked as an assistant project scientist at University of California, Berkeley. His research fields are investigations of the mechanical and chemical properties of materials by applying theoretical and computational tools, including first-principles calculations, atomistic simulations, multiscale modeling and machine learning. He has received the NSF CAREER award in 2019.