(142d) Guiding Experiments Towards New Functional Materials with Informatics

In this talk, we will discuss our informatics approach that integrates machine learning, symmetry analysis and density functional theory (DFT) to guide experiments in search of novel BaTiO₃-based ferroelectric materials with high Curie temperature (T_C). BaTiO₃-based materials have attracted significant attention as a potential alternative for Pb-free piezoelectrics. In the development of BaTiO₃-based piezoelectrics, one of the common strategies involves mixing two end member compositions that have Cubic (Pm-3m) to Tetragonal (P4mm) and Cubic to Rhombohedral (R3m) transformations. Our design objective is to identify suitable chemical substitutions at the Ba- and Ti-sites such that the T_C of the end members are high. We perform DFT calculations on undoped ATiO₃ and BaBO₃ perovskites in P4mm and R3m symmetries, respectively; A and B are divalent and tetravalent cations, respectively. From symmetry-mode analysis, we decompose the two ferroelectric phases (P4mm and R3m) in terms of irreducible representations (irreps). These irreps, along with the unit cell volume from DFT, serve as feature sets for understanding ATiO₃ and BaBO₃ crystal chemistries and their role in impacting the ferroelectric distortions. We then link these features with known experimental T_C data using machine learning methods and predict a new CdTiO₃-BaTiO₃ solid solution for experimental synthesis. We performed experiments and found that our predictions do not agree with the measurements. The reason for the discrepancy is attributed to the anomalous behavior of the CdTiO₃-BaTiO₃ solid solution, the phase diagram of which is found to be different from any of the BaTiO₃-based solid solutions studied so far.