The interpretability of machine learning reveals associations between input features and predicted physical properties in models, which are essential for discovering new materials. However, previous works were mainly devoted to algorithm improvement, while the essential multi-scale characteristics are not well addressed. This paper introduces distortion modes of oxygen octahedrons as cross-scale structural features to bridge chemical compositions and material properties. Combining model-agnostic interpretation methods, we are able to achieve interpretability even using simple machine learning schemes and develop a predictive model of effective mass for a widely used material type, namely perovskite oxides. With this framework, we reach the interpretability of the model, understanding the trend of the effective mass without any prior background information. Moreover, we obtained the knowledge only available to experts, i.e., the interpretation of effective mass from the s–p orbitals hybridization of B-site cations and O²⁻ in ABO₃ perovskite oxides.

In the current work, the bulk ternary (0.85-x) BiFeO₃-xBaTiO₃-0.15PbTiO₃ (BF-BTx-PT, x=0.08-0.35) system has been studied as a potential high-temperature piezoceramics. Samples with various content of BT were prepared via solid-state route, and pure perovskite phase was confirmed by X-ray diffraction. The temperature dependence of dielectric constants confirmed the decrease of Curie temperature with increasing BT content. It was found that the morphotropic phase boundary (MPB) composition of BF-BTx-PT ceramics was in the vicinity of x=0.15, which exhibits optimal properties with piezoelectric constant d₃₃ of 60 pC/N, high Curie temperature of 550 ℃, and low sintering temperature of 920 ℃. Measurements also showed that the depoling temperature was 300 ℃, about 150 ℃ higher than that of commercialized PZT ceramics, which indicated good temperature stability. BF-BTx-PT ceramics are promising candidates for high temperature applications.