Accelerating Materials Discovery with Data-Driven Atomistic Computational Tools
Many of the key technological problems associated with alternative energies (e.g., thermoelectrics, advanced batteries, hydrogen storage, etc.) may be traced back to the lack of suitable materials. Both the materials discovery and materials development processes may be greatly aided by the use of computational methods, particular those atomistic methods based on density functional theory (DFT). Here, we present an overview of our recent work utilizing high-throughput computation and data mining approaches to accelerate materials discovery, specifically highlighting three new approaches: (i) We describe our high-throughput DFT database, the Open Quantum Materials Database (OQMD)1, which contains over 280,000 DFT calculations and is freely available for public use at oqmd.org. (ii) We show how computational crystal structure solution may be addressed via a new hybrid approach, the First-Principles Assisted Structure Solution (FPASS) approach2, which combines experimental diffraction data, symmetry information, and first-principles-based evolutionary algorithmic optimization to automatically solve crystal structures. (iii) We also describe a newly-developed machine learning approach3,4 to construct a materials screening model based on an extensive set of thousands of DFT calculations. The resulting model, which has “learned” rules of chemistry from these many examples, can predict the stability of arbitrary compositions without requiring any a priori knowledge of crystal structure, at about six orders of magnitude lower computational expense than the original QM tools. We use this model to scan—in a matter of minutes—roughly 1.6 million candidate compositions for novel ternary compounds (AxByCz), and predict roughly 4,500 new stable materials.
1) J. E. Saal, S. Kirklin, M. Aykol, B. Meredig, and C. Wolverton "Materials Design and Discovery with High-Throughput Density Functional Theory: The Open Quantum Mechanical Database (OQMD)", JOM 65, 1501 (2013).
2) B. Meredig and C. Wolverton, "A Hybrid Computational-Experimental Approach for Crystal Structure Solution" Nature Materials 12, 123 (2013).
3) B. Meredig and C. Wolverton, “Dissolving the Periodic Table in Cubic Zirconia: Data Mining to Discover Chemical Trends” Chem. Mater. 26, 1985 (2014).
4) B. Meredig, A. Agrawal, S. Kirklin, J. E. Saal, J. W. Doak, A. Thompson, K. Zhang, A. Choudhary, and C. Wolverton, "Combinatorial screening for new materials in unconstrained composition space with machine learning", Phys. Rev. B 89, 094104 (2014).
Christopher Wolverton is a Professor of Materials Science and Engineering at Northwestern University. Before joining the faculty, he worked at the Research and Innovation Center at Ford Motor Company, where he was group leader for the Hydrogen Storage and Nanoscale Modeling Group. He received his BS degree in Physics from the University of Texas at Austin and his PhD degree in Physics from the University of California at Berkeley. After completing his PhD degree, Wolverton performed postdoctoral work at the National Renewable Energy Laboratory (NREL). His research interests include computational studies of a variety of energy-efficient and environmentally friendly materials via first-principles atomistic calculations, high-throughput and data mining tools to accelerate materials discovery, and “multiscale” methodologies for linking atomistic and microstructural scales. Wolverton has authored or co-authored more than 180 peer-reviewed publications (h-index=51), holds nine patents (several others pending), and has given more than 150 invited talks. Wolverton is a Fellow of the American Physical Society, has won the Walder Award for Research Excellence, a Ford Motor Company Technical Achievement Award, and gave the John Dorn Memorial Lecture at Northwestern in 2003.