Accelerated Search for KNN-Based Ceramics with Large Piezoelectric Constants Using Machine Learning Methods

The (K,Na)NbO3(KNN)-based piezoelectric ceramics are one of the most promising lead-free piezoelectric materials to replace toxic lead-based ones for ultrasonic transducer applications owing to their high Curie temperature and excellent piezoelectric properties. However, it is costly to discover multiple doped compositions with enhanced properties based on the traditional trial and error approach. In this study, we proposed an efficient data-driven machine learning(ML) approach to search for KNN-based ceramics with enhanced piezoelectric properties. The designed ML framework efficiently located the potential composition with a high piezoelectric constant d
33
for the experiment procedure. The newly synthesized composition achieves an outstanding d
33
of ~ 407 pC/N. The results reveal the exceptional efficiency of this approach in accelerating the material design and discovery with tailored properties.