A study on polarization switching in hybrid improper ferroelectric double perovskites oxides from first principles computations and machine learning

ABO₃ perovskite oxides with Pnma symmetry exhibit polarization via hybrid improper ferroelectric mechanism (HIF). Here the primary order parameters are rotation (Q_R+) and tilt (Q_T) distortions along a⁰a⁰c⁺ and a^-a^-c⁰, respectively. Naturally, it has been believed that functionalities such as polarization can also be switched by tuning these primary order parameters. In this study, we have identified a unique low-energy polarization switching path via out-of-phase rotation Q_R- for a series of polar double perovskite oxides (DPOs). Finite temperature simulations using ab initio molecular dynamics (AIMD) are performed to outline the design rules of temperature-dependent ultra-fast polarization switching. Later, we constructed a wide compositional space with various combinations of 3d-3d, 3d-4d, and 3d-5d elements to construct machine learning (ML) models. The ML models use the understandings gained from the representative simulations performed on a handful DPOs. This presentation will focus on discussing these results particularly targeted at designing DPOs with low switching barriers and high polarization.