Unveiling rare earth dopant configurations and crystal field analysis in tetragonal BaTiO3via machine learning and superposition modeling

Abstract

Barium titanate (BaTiO₃) is a model ferroelectric perovskite with enduring technological relevance in electronics, catalysis, and energy storage. Rare-earth (RE) doping provides a powerful strategy to engineer its structural, electronic, and dielectric properties, yet a comprehensive understanding of dopant site selectivity and its consequences remains elusive. Here, we present the first systematic exploration of the entire RE series (4f⁰–4f¹⁴) in tetragonal BaTiO₃ using an integrated workflow that couples machine-learning-driven structural optimization (CHGNet) with semi-empirical crystal-field analysis via the Superposition Model. This DFT-free approach enables efficient, large-scale mapping of doping trends, revealing clear correlations between ionic size, electronic configuration, site preference, lattice distortions, and stability. Our results show that Ba-site substitution is preferentially stabilized by intrinsic electronic charge redistribution, whereas Ti-site substitution exhibits opposite electronic polarization tendencies, indicating the predominance of intrinsic electronic compensation under equilibrium-like conditions, while oxygen-vacancy-assisted configurations remain metastable. The study further identifies the tetragonality ratio (c/a) as a sensitive descriptor of RE substitution, with Ba-site dopants generally enhancing and Ti-site dopants suppressing local ferroelectric distortion. In parallel, the derived crystal-field parameters provide a trend-based reference dataset for interpreting XRD, EPR, and luminescence spectra in RE-doped perovskites. By linking atomic-scale structure, local electronic redistribution, and lattice strain to ferroelectric distortion, this work establishes a predictive framework for tailoring RE-doped BaTiO₃ and demonstrates a scalable computational paradigm for accelerating the discovery and design of functional oxide materials.