Accelerated design of lead-free high-performance piezoelectric ceramics with high accuracy via machine learning

The piezoelectric performance serves as the basis for applications of ceramics. ability to rapidly and accurately predict coefficient (d33) is much practical importance exploring high-performance In this work, a data-driven approach combining feature engineering, statistical learning, machine learning (ML), experimental design, synthesis trialed investigate its accuracy in predicting d33 potassium–sodium–niobate ((K,Na)NbO3, KNN)-based atomic radius (AR), valence electron distance (DV) (Schubert), Martynov–Batsanov electronegativity (EN-MB), absolute (EN) are summarized four most representative features describing out all 27 possible These contribute greatly regression classification distinguishing polymorphic phase boundary (PPB). ML method developed work exhibits high An example KNN combined with 6 mol% LiNbO3 demonstrates 184 pC/N, which highly consistent predicted result. This proposes novel feature-oriented guideline accelerating design ceramic systems large d33, expected be widely used other functional materials.