Machine learning-guided design of high-performance Nd-doped lead magnesium niobate-lead titanate ceramics for enhanced piezocatalytic applications.
Abstract
Faced with the dual challenges of energy shortages and environmental pollution, piezoelectric catalysis, an emerging technology that efficiently converts mechanical energy into chemical energy, has attracted growing attention. In this study, the lead-based perovskite Pb(Mg1/3Nb2/3)O3-PbTiO3 (PMN-PT) was employed as a model system. A Bayesian optimization-based multi-model comparison framework was developed to systematically optimize the Nd doping level and PT content of Nd-doped PMN-PT ceramics. The optimal composition, 2.5%Nd: PMN-0.29PT, was rapidly identified. The ceramic exhibits a high piezoelectric coefficient d33 of 1253 ± 28 pC//N, which is in excellent agreement with the model prediction. Moreover, the results from the catalytic experiments based on Nd: PMN-PT ceramics indicated that a RhB degradation rate of 26.2 ± 0.92 × 10^-3 min-1 under 100 W was achieved, while retaining more than 88% of its activity after four cycles. This work provides a new strategy for machine-learning-assisted design of piezoelectric functional materials as well as their piezocatalytic applications in environmental remediation.