Data-driven discovery of high-entropy rare earth aluminates for high-temperature thermal barrier applications

High-entropy rare earth (RE) aluminates are promising candidates for thermal/environmental barrier coatings (T/EBCs), while the immense compositional space presents significant challenges for traditional experimental discovery. To address this issue, artificial neural network (ANN), support vector machine (SVM), and random forest classification (RFC) are employed as three machine learning models to predict the single-phase formation ability of (RE^Ⅰ_1/4RE^Ⅱ_1/4RE^Ⅲ_1/4RE^Ⅳ_1/4)₄Al₂O₉ materials. Both the ANN and RFC models achieve optimal validation accuracy, demonstrating their outstanding ability to capture complex patterns from the dataset. SHapley Additive exPlanations (SHAP) analysis is utilized to interpret the contribution of feature descriptors, revealing the significant impact of factors such as average ionic radius ( $\overline{r}$) on phase stability. According to the prediction results of machine learning, three representative ceramic samples are selected, and single-phase monoclinic crystal structures and uniform elemental distribution are confirmed by X-ray diffraction (XRD) and scanning electron microscopy (SEM). The synthesized ceramics exhibit quasiductile behavior with enhanced damage tolerance combined with lower thermal conductivity, thus making them promising candidates for next-generation T/EBCs.