Machine learning aided accelerated prediction and experimental validation of functional properties of K1-xNaxNbO3-based piezoelectric ceramics

The functional properties of piezoelectric ceramics are vital to design materials for energy harvesting applications. In the present study, to accelerate the design process with prediction of functional prop-erties for a composition, a machine learning (ML) model is developed, which predicts (Er , TC , d33) values for K1-xNaxNbO3(KNN)-based materials. Simultaneously, important features were identified which directly influence the functional properties such as Martynov and Batsanov (MB) electronegativity, ionic radii, atomic volume of elements, and atomic polarization. The ML model with the best performance is implemented to predict the functional properties of KNN-based ceramics. The measured functional properties namely, d33 , TC and Er for hydrothermally synthesized KNN10 composition are found to be 131 pC/N, 415 degrees C and 428.4 respectively, which are consistent with the ML prediction. In addition, the proposed model has extended to predict the properties of KNN-ceramics with series doping of elements at A-site (lithium) and B-site (antimony), and it is in good agreement with the trend of experimental outcomes from literatures. Finally, this work discussed the feature-oriented guidelines for optimizing the synthesis of KNN ceramics.(c) 2023 Elsevier Ltd. All rights reserved.

Automated summary

In this study, a machine learning model is developed to predict the functional properties of KNN-based ceramics. By identifying important features, the design process can be accelerated and synthesis methods can be optimized. The experimental results are consistent with the predictions, indicating that the model has good predictive performance and applicability.