6-2. Data-driven accelerated materials design with machine learning and high-throughput computing

Wei Chen

Illinois Institute of Technology, Chicago, IL

Abstract: Accelerating the discovery of advanced materials is essential for sustainable development and manufacturing innovation. In the talk, I will discuss how the integration of machine learning and high-throughput computing can help uncover the missing piece for accelerated materials design.

I will talk about the development of a comprehensive database of elastic tensors and its contribution to the discovery of a new class of thermoelectric materials. In combination with a newly developed machine learning framework, the elasticity database has been leveraged to design superhard materials and explore high-entropy alloys. The ability to quantify how microscale parameters affect the macroscale material properties marks an important step towards the ab initio prediction of material design rules. Moving beyond bulk phenomena of materials, understanding defect structure is the new frontier for data-driven materials research. I will introduce an effort in applying high-throughput calculations in understanding point defects in intermetallic compounds and how machine learning can elevate this information to derive rules for predicting point-defect types. Lastly, I will discuss the application of machine learning on analyzing and understanding experimental datasets, such as thermodynamic measurements and electron microscope images, for materials design and discovery.