High-throughput screening of efficient ionic liquid additives for aqueous zinc-ion batteries assisted by machine learning

Electrolyte additives, owing to their simplicity and high efficiency, have been widely employed to enhance the performance of aqueous zinc-ion batteries (AZIBs). To achieve rapid screening of highly effective electrolyte additives, this study integrates theoretical calculations with machine learning (ML) methods, using the sum of binding energy and adsorption energy (E add ) of 2025 ionic liquids (ILs) ion pairs as the prediction target for model training and ranking. Through systematic model performance comparison and experimental validation, the Gradient Boosting Regressor (GBR) model demonstrated exceptionally high predictive accuracy (test set R 2 = 0.9982) and excellent practical feasibility. Additives screened by the GBR model, namely n-Propylammonium tetrafluoroborate (PrBF 4 ) and 1-Butyl-1-methylpyrrolidinium tetrafluoroborate (BMPyrrBF 4 ), enabled zinc (Zn) symmetric cells to achieve cycling lifetimes of 130 h and 550 h, respectively, both superior to the 100 h observed with the ZnSO 4 electrolyte and consistent with the model's predicted ranking. Further analysis revealed that these electrolyte additives disrupted the intrinsic hydrogen-bond network of the electrolyte, reduced the content of free water, and induced preferential Zn 2+ deposition along the Zn (002) crystal plane. In addition, they increased the nucleation overpotential, promoting uniform Zn nucleation. Consequently, these effects effectively suppressed dendrite growth and the accumulation of detrimental by-products, thereby extending battery lifespan. The ML framework developed in this study provides a feasible and efficient pathway for large-scale screening of electrolyte additives and offers valuable guidance for the optimization and development of next-generation aqueous batteries. • High-throughput screening of ionic liquid electrolyte additives • Combined theoretical, machine learning, and experimental validation • Comparative evaluation of six machine learning models • Automated workflow development. • Experimental validation of theory-supported machine learning predictions