Machine learning accelerates composition design of lead-free high performance piezoceramics based on potassium sodium niobate

Potassium sodium niobate (KNN) based lead-free piezoelectric ceramics have garnered significant attention as a new generation of environmentally friendly materials for energy storage and conversion. However, achieving the critical coexistence of different phases at room temperature for superior piezoelectric response typically requires complex multicomponent adjustments through a lengthy process of trial and error. The intricate elemental composition poses tremendous challenges for experimental design and high-throughput calculation. In this study, we adopt a stepwise machine learning (ML) method to develop models for KNN-based solid solutions with multiple element joint doping. On the one hand, a critical composition with multiple phase coexistence at room temperature can be obtained on the forecasted phase diagrams calibrated by the shift of polymorphic phase transition temperatures, experimentally, which exhibits rhombohedral-orthorhombic-tetragonal slushlike polar nanoregions on the atomic polarization mapping. On the other hand, optimal experimental piezoelectric properties for the critical composition further confirm the accuracy of the predicted piezoresponse in the ML model. This work offers a perspective for lead-free piezoelectric materials design that differs from the traditional artificial trial and error method and may provide insights for other materials fields.