Machine Learning–Assisted Exploration of the Structural, Dielectric, and Ferroelectric Properties of CaTiO3 Nanoceramics

Abstract

Background: The structural and electrical properties of Pb-free calcium titanate (CaTiO3) nanoceramics were investigated using a combined experimental and machine learning approach. Methods: X-ray diffraction analysis with Rietveld refinement revealed the formation of an orthorhombic perovskite phase with the Pnma space group and the influence of annealing temperature on crystallite size and lattice strain. Dielectric measurements revealed strong frequency-dependent dispersion, while polarization–electric field hysteresis loops showed an increase in remanent polarization and coercive field with increasing annealing temperature. Results: Machine learning models were used to predict key structural and electrical properties. Among the evaluated models, XGBoost achieved the highest predictive performance for ferroelectric properties, with a coefficient of determination (R2) of 0.992, while the remaining machine learning models also demonstrated strong predictive capability with R2 values exceeding 0.96. Conclusions: These predictions strongly agreed with experimental observations, which demonstrated the ability of machine learning models to capture complex structure–property relationships and guide the optimization of synthesis conditions for CaTiO3 nanoceramics.