Facile Critical Evaluation of Extensive Lithium–Oxygen Battery Literature Using In-House Data and the Structured Query Learning–Retrieval-Augmented Generation Method

Lithium–oxygen batteries (LOBs) offer combustion fuel-like energy densities but remain constrained by low efficiency, limited cycle life, and coupled degradation pathways linking the electrochemical growth and decay of the reaction product (Li 2 O 2 ) and associated generation of reactive oxygen species, electrolyte and electrode instabilities, and lithium dendrite growth. Here, we introduce a hybrid materials-informatics framework that integrates structured query learning with retrieval-augmented generation (RAG) to systematically analyze the full-text corpus of 3134 peer-reviewed articles in LOBs. Unlike conventional artificial intelligence (AI) tools, which learn from unstructured literature and risk factual drift, the present approach forms a relational performance-validated database, enabling evidence-traceable comparison of cathode architectures, catalyst types, electrolytes, redox mediators, and lithium protection strategies. The analysis reveals composition-dependent performance hierarchies and exposes interdependencies among Li 2 O 2 morphology, singlet-oxygen formation, overpotentials, and solid electrolyte interface disruption, as reported under their documented experimental conditions. Using this method, we identified the catalyst–electrolyte–anode configurations capable of reducing charge polarization by 0.3–0.6 V and extending cycling stability to 100–200 cycles under standard cycling conditions reported in the source studies. This data-driven roadmap establishes a quantitative foundation for translating LOBs from laboratory demonstrations to deployable high-energy systems and demonstrates how materials informatics can accelerate electrochemical materials synthesis and device design.