Harnessing Lateral Transfer Learning for Pioneering Solid Electrolyte Interphase Innovation

Abstract

The rapid growth in energy storage demands, particularly in electric vehicles and portable electronics, has positioned solid-state batteries (SSBs) at the forefront of cutting-edge research. Compared to conventional lithium-ion batteries, SSBs offer significant advantages, including enhanced safety, higher energy density, and extended cycle life. However, a major challenge lies in identifying solid electrolyte interphase (SEI) materials with high shear modulus (G_s) and exceptional ion transport properties to prevent lithium dendrite formation and improve overall battery performance. Here, we introduce an innovative approach utilizing lateral transfer learning to accelerate the discovery of high-performance SEI materials. Traditional machine learning models often require large datasets, which are typically unavailable for specialized material properties like G_s. To address this, we applied lateral transfer learning, transferring knowledge from models trained on larger datasets (bandgap data) to predict G_s within smaller datasets. By leveraging Crystal Graph Convolutional Neural Networks (CGCNN), the method effectively captures structural relationships at the atomic level, achieving a G_s prediction accuracy of 90%, ultimately identifying 12 promising SEI candidates. This sophisticated methodology not only accelerates material discovery but also opens new pathways for deploying artificial intelligence in advanced energy materials, driving progress toward safer and more efficient SSBs.