High throughput screening of new piezoelectric materials using graph machine learning and knowledge graph approach

Computational methods, such as the Density Functional Theory (DFT), have long been a reliable tool for predicting material properties. However, their use in high-throughput screening has been limited due to computational costs. In this paper, we present a graph-based machine learning (ML) framework that overcomes these limitations, offering a more efficient approach to material selection and property prediction. Our framework, which includes a knowledge graph (KG) approach, and a graph neural network (GNN) based model, significantly reduces the search space by filtering materials from the Crystallography Open Database (COD) using KGs. We then use a modified Gated Graph ConvNet (GatedGCN) model to predict the maximum longitudinal piezoelectric modulus (${{\Vert e}_{\mathrm{ij}}\Vert }_{\max })$ of the screened materials. Based on the study, a list of new perovskite-based piezoelectric materials is shown with the top candidate reaching a value of ${{\Vert e}_{\mathrm{ij}}\Vert }_{\max }$ as high as ∼ 10.81 C/m².