Disordered Rock-Salt (DRX) cathode structures are materials where lithium and transition metal atoms are arranged in a disordered solid solution. These structures are currently under intensive study [1] due to their potential higher energy density, bigger capacity than the typical layered cathode structure, and improved flexibility in the selection of raw materials. Driven by this, we have recently compiled a DFT-based database of DRX structures with over 6,000 different metal oxide compositions. Moreover, since oxygen loss is linked to voltage degradation, intergranular cracks, and irreversible phase transformation, developing a descriptor that allows us to understand and reflect the stability of oxygen within the DRX anion lattice is crucial. It represents a significant research challenge that our study aims to address.In this work, we study two things: First, since the generation of oxygen vacancies holds a critical value in elucidating the stability of cathode material oxides[2], our methodology focuses on generating an extensive DFT database of over 1,400 oxygen vacancy formation energies (Evf(O)) from selected DRX structures. We then employ a machine learning (ML) model to predict the Evf(O) and their distribution for a variety of different compositions. The ML model incorporates site-specific features, such as the maximum reduction potential of the cations neighboring the O vacancy, the O p-band center of mass, and information from atomic population analysis (such as the bond order and the bond spin polarization between each O and its nearest neighbors). This approach is necessary because the DRX structure contains different local environments where O vacancies could be created, and recent work from Baldassarri et al. has highlighted the limitations of ML models that only include global features and lack their ability to differentiate the behavior of compounds containing the same elements but in different oxidation states [3]. This allows us to predict the O vacancies energetics with an accuracy comparable to DFT and to develop a descriptor to classify different compositions of DRX structures.Second, given that oxygen destabilizes during delithiation, we analyze the impact and relationship of the Li vacancy formation energies (Evf(Li)) and Evf(O) of each O able to create a vacancy. The results show a strong interaction between these energetics, and in particular, an inverse correlation between Evf(O) of a specific O site and the Evf(Li) of the Li atoms that surround this O, which within these DRX prototypes the number of surrounding Li varies from 1 to 5. In line with what others proposed [4], there is a strong interaction between these two vacancy types. However, we expect that the generation of O vacancies is determined not only by their thermodynamic driving force but also by the displacement/migration of the nearest neighbor lithium (NN-Li) that surrounds each O.The methodology and fundamental understanding developed in this study about DRX O-vacancies and their relationship with NN-Li during delithiation could help investigate and describe compositionally similar transition metal oxide cathodes belonging to the same crystal structure family (layered, spinel-like, γ-LiFeO2). Hence, this research not only enhances our understanding of these materials but also paves the way for the design and development of more flexible and stable energy storage systems.[1] S. Anand, B. Ouyang, T. Chen, G. Ceder, Impact of the energy landscape on the ionic transport of disordered rocksalt cathodes, Physical Review Materials 7(9) (2023) 095801.[2] H. Zhang, B.M. May, F. Omenya, M.S. Whittingham, J. Cabana, G. Zhou, Layered Oxide Cathodes for Li-Ion Batteries: Oxygen Loss and Vacancy Evolution, Chemistry of Materials 31(18) (2019) 7790-7798.[3] B. Baldassarri, J. He, A. Gopakumar, S. Griesemer, A.J.A. Salgado-Casanova, T.-C. Liu, S.B. Torrisi, C. Wolverton, Oxygen Vacancy Formation Energy in Metal Oxides: High-Throughput Computational Studies and Machine-Learning Predictions, Chemistry of Materials 35(24) (2023) 10619-10634.[4] C. James, Y. Wu, B.W. Sheldon, Y. Qi, The impact of oxygen vacancies on lithium vacancy formation and diffusion in Li2-xMnO3-δ, Solid State Ionics 289 (2016) 87-94.
Read full abstract