Deep-learning potential with anisotropic effective charge tensors for ab initio ferro-piezoelectricity of lead-free (K, Na)NbO3 perovskites

X Zhang and B Li and M Wang and B Liu and WJ Lv and HX Liu and K Liu, CERAMICS INTERNATIONAL, 51, 47675-47684 (2025).

DOI: 10.1016/j.ceramint.2025.08.024

Escalating environmental and health concerns regarding lead-containing piezoceramics have spurred the request for environmentally friendly and sustainable alternatives. Among them, lead-free K1-xNaxNbO3 (KNN) perovskites have drawn increasing attention due to their superior ferroelectric and piezoelectric performances. However, atomic understanding of underlying mechanisms governing ferro-piezoelectric behavior of KNNs is still unclear and of vital importance. Therefore, in this work, we perform ab initio accuracy deep-learning molecular dynamics simulations to investigate nanostructures, polarization, and piezoelectric responses of orthorhombic KNN crystals with varying K/Na concentrations (x = 0 similar to 0.75). The proposed computational framework integrates anisotropic Born effective charge (BEC) tensors with deep-learning potentials, enabling quantum-informed characterization of local polarization and explicit electric field interaction with essential electronic structure information. Moreover, polarization-electric field hysteresis loops and piezoelectric strain coefficients for KNN crystals are systematically evaluated, which demonstrates that piezoelectricity of polycrystalline KNN ceramics positively correlates with Na concentrations and exhibits optimal performance at x = 0.5, ascribed to increased intrinsic atomic elastic responses to a macroscopic stress as well as orientation-dependent domain configuration manipulation. It is envisioned that the incorporation of BEC tensors into machine-learning potentials can be efficiently extended to nano-resolved investigation of diverse ferroelectric materials and complex phenomena with improved scalability, flexibility, universality, and ab initio prediction accuracy.

Return to Publications page