Grain boundary engineering in Nickel-rich cathode: A combination of high-throughput first-principles and interpretable machine learning study

• Published in: Acta materialia 2024-09, Vol.276, p.120144, Article 120144
• Main Authors: Hu, Yixuan, Zhang, Yumeng, Wen, Bo, Dai, Fu-Zhi
• Format: Article
• Language: English
• Subjects: Doping; First-principles calculation; Grain boundary engineering; Machine learning; Ni-rich cathode
• ISSN: 1359-6454
• DOI: 10.1016/j.actamat.2024.120144

One of the primary challenges in commercializing nickel (Ni)-rich layered cathodes for lithium-ion batteries (LIBs) is addressing the problem of crack formation during cell cycling. These cracks quickly spread along grain boundaries (GBs), causing capacity and voltage degradation. GB engineering, a classical technique for strengthening materials that has been largely unexplored in LIBs, could potentially inhibit intergranular fractures and significantly extend battery life. In this study, we employed a synergistic approach that combined high-throughput first-principles calculations and an interpretable machine learning (ML) framework to assess the feasibility of enhancing GBs through segregation-induced strengthening via doping. As a result, we generated a comprehensive database containing 64 elements doped at 4 sites in ∑3 [100](012) GB of LiNiO2, representing the largest database for first-principle-aided GB engineering in LIBs to date. This database further enabled the implementation of an interpretable ML workflow to uncover the fundamental principles governing dopant segregation and GB strengthening. Based on the insights gained from this research, we provide practical and promising recommendations for dopant elements that can effectively anchor and reinforce GBs, such as Mg, Al, Si, Ti, Cr, Mn, Fe, Cu, Zn, Hf, and Ce, with some already demonstrated success in experiments. This work on GB engineering serves as a foundation for the development of advanced, doped cathode materials for next-generation batteries. [Display omitted]