Workflow for modeling of variable composition in multicomponent alloys: from high-throughput first-principles calculations to CALPHAD modeling
XiaoYu Chong1, Jing Feng1, Shang Shunli2, Wang Yi2, Liu Zikui2
1 Key laboratory of material genetic engineering, Kunming University of Science and Technology, Kunming, Yunnan, China
2 Materials Genome Inc., State College, PA, 16803, USA
EXTENDED ABSTRACT: The key technical characteristics of high-throughput computing differ from traditional computing methods as following: realizing parallel and/or automatic calculation for 102~104 tasks; the integrated computing (ICME) method focuses on breaking through the bottleneck of material design/ computing cross-scales. The CALPHAD modeling and database are the basis of cross-scale modeling. Since the thermodynamic data obtained from DFT alone are not enough to accurately describe the phase boundary, experimental data should be introduced into Calphad modeling at the same time. Based on the above ideas, we developed a method and standard workflow for establishing a thermodynamic database based on raw data from high-throughput first-principles calculations and optimizing thermodynamic parameters through experimental phase boundary data and machine learning techniques. The first-principles software package DFTTK was developed to calculate the thermodynamic properties of 104 crystal structures at finite temperature. Through the self-developed thermodynamic modeling program ESPEI based on Markov chain Monte Carlo, the simultaneous automatic optimization was realized for multiple thermodynamic parameters of the high-order system. A developed tool is used to realize the automatic transfer of thermodynamic data between DFTTK and ESPEI. These workflow and software overcomes the traditional "snowball" effect of the previous thermodynamic modeling and reduce the threshold, which can be used to quickly establish the thermodynamic framework for rational design of new material.
Xiaoyu Chong is Ph.D in Engineering, Youth Talent Promotion Project of China Association for Science and Technology, and Research Assistant Professor at the Pennsylvania State University from 2016 to 2019. His research interest is high-throughput first-principle calculations coupling with CALPHAD modeling and machine learning for the design of noble metal materials, high temperature structural ceramics and coating. He is the member of Youth Editorial Committee of Rare Metals, member of Computational Materials Branch of China Materials Research Society. Up to now, he has published more than 80 papers on JACS, Acta Mater, Scripta Mater, APL and other journals as the first or corresponding author, including 4 cover articles and 1 editor’s choice paper.