| Abstract Scope |
Theoretical simulations and scanning transmission electron microscopy have opened new avenues to study ferroelectric materials at the atomic and mesoscales. Such methods yield information on atomic coordinates, order parameter fields, functional behavior at interfaces, surfaces, and polarization dynamics. Naturally, extraction of generative physics of ferroelectric materials, either in the form of atomistic descriptors or parameters of mesoscopic Ginzburg-Landau model require adaptation of machine learning (ML) methods. However, the in-built correlative nature of traditional ML techniques fails to capture the causal mechanisms driving any physical phenomena. This presentation will focus on several examples of causal ML studies of perovskite oxides of the form ABO3 and its derivatives in which structural distortions drive functionalities such as ferroelectricity, magnetism, and metal-to-insulator transition. A discussion on designing and deploying physics-informed ML schemes, capable to exploit knowledge from physical models along with observational data underpinning solid structure and functionality will be included in the presentation. |