Structural Design Strategies of Organic Cathodes for Sustainable Li-Ion Batteries Using Graph-Based Explainable AI

As the impact of carbon emissions on global warming continues to increase, there is growing interest in sustainable cathode materials for Li-ion batteries consisting of organic materials [1]. Additionally, organic materials have the advantage of flexible structural design, allowing easy modulation of cathodic properties [2]. Due to these advantages, extensive research has been conducted on exploring new materials and analyzing structure-property relationships in a wide structural space. Recently, several machine learning (ML) techniques have been employed to accelerate property prediction and propose new materials [3,4]. These studies have successfully explored new candidate materials in a wide structural space, but the black-box nature of ML models has limited the analysis of the impact of structure on properties. Beyond material exploration, structural design strategies for modulation electrochemical performance are crucial for providing insights into novel cathode designs. In this study, we present structural design strategies for property modulation of organic cathodes using explainable AI (XAI) and statistical techniques. We develop GNN models to predict various cathodic properties (voltage, capacity, specific energy, and voltage descent) and analyze them using graph-based XAI. Through statistical analysis of the results, we identify positive effect motifs (P-motif) and negative effect motifs (N-motif) for each property, proposing structural design strategies to optimize cathodic properties in organic cathodes. Our findings demonstrate the utility of graph-based XAI analysis in understanding complex structure-property relationships and applying them in structural design strategies. Furthermore, this methodology can be easily extended to other material fields, such as catalysis and redox flow batteries, to uncover the unknown structure-property relationships.<br/><br/><b>References</b><br/>[1] L. Zhao, A. E. Lakraychi, Z. Chen, Y. Liang, Y. Yao, ACS Energy Lett. <b>2021</b>, 6, 3287–3306.<br/>[2] Y. Lu, Q. Zhang, L. Li, Z. Niu, J. Chen, Chem. <b>2018</b>, 4, 2786–2813 (2018).<br/>[3] R. P. Carvalho, C. F. N. Marchiori, D. Brandell, C. M. Araujo, Energy Storage Mater. <b>2022</b>, 44, 313–325.<br/>[4] X. Zhou, A. Khetan, J. Zheng, M. Huijben, R. A. J. Janssen, S. Er, Digit. Discov. <b>2023</b>, 2, 1016–1025.