Ken-ichi Nomura1 Aiichiro Nakano1 Priya Vashishta1 Rajiv Kalia1

1, Collaboratory for Advanced Computing and Simulations, University of Southern California, Los Angeles, California, United States

Emerging exascale computing will have a profound impact on materials simulations and machine learning (ML) to enable faster and more targeted material discoveries. This requires a new approach to computational materials science that integrates materials simulations with ML techniques. This interdisciplinary integration, along with the ever-tighter coupling between experiments and simulations, will provide a new platform for ML-enabled materials discovery. Here education and training programs on ML-based methodologies for the materials research community are urgent needs for future scientists and engineers to be competitive in the emerging field.
At the DOE Materials Genome Innovation for Computational Software (MAGICS) Center, we develop open-source materials simulation software, ML tools, training courseware that run on desktops to exascale supercomputers. Center software and databases provide function-property-structure relationships in functional materials to help synthesis and characterization of a wide class of materials. We have provided three hands-on trainings in Center-developed materials software databases so far (106 users from 55 universities, national labs and research institutions), and plan to continue the outreach program at annual MAGICS software workshops. In this talk I will discuss the MAGICS software suite and training courseware, and lessons learned from the software workshops.

This work was supported as part of the Computational Materials Sciences Program funded by the U.S. Department of Energy, Office of Science, Basic Energy Sciences, under Award Number DE-SC0014607