AI-Guided Workflows for the Discovery of Novel Materials

The intricate interplay of several underlying processes governing certain materials' properties and functions prevents the explicit, atomistic modelling and hinders the discovery of novel materials. In this talk, I will discuss an artificial-intelligence (AI) approach to identify the key descriptive parameters, termed "materials genes", correlated with the materials performance and reflecting the physical processes that trigger, facilitate, or hinder the materials' behavior.[1] The symbolic-regression sure-independence-screening-and-sparsifying-operator (SISSO) method leverages the typically small high-quality datasets in computational and experimental materials science and it is applied in active-learning workflows[2,3] to guide the discovery of improved, or even novel materials.