Machine Learning-Driven Electrolyte Additive Discovery for Ultrafast Charging of Sodium Metal Batteries

Abstract Sodium metal batteries hold immense promise for large-scale, high-power storage systems, yet their fast charging is severely constrained by sluggish Na+ transport across unstable electrode–electrolyte interphases. Conventional fluorine-donating additives suffer from inefficient self-decomposition, inevitably generating resistive organic byproducts. Herein, we propose an active machine learning framework based on Gaussian process regression to develop an innovative additive-triggered anion decomposition pathway. Guided by iterative screening of key physicochemical descriptors, a trace-level (0.02 M) sodium chlorodifluoroacetate (NaCDFA) additive is successfully identified. Theoretical calculations combined with in situ characterizations reveal that the NaCDFA thermodynamically promotes the targeted and continuous decomposition of the PF6– anion, enabling the construction of robust and highly conductive NaF-rich interphases, which substantially reduces the Na+ migration barrier. Consequently, the Na||Na3V2(PO4)3 cell achieves ultrastable cycling over 50,000 cycles at an extreme rate of 200 C. Furthermore, outstanding performance is retained under rigorous practical conditions, including high-mass-loading cathodes and limited sodium excess or anode-free architectures. This work establishes a cost-effective trace-level electrolyte additive strategy for engineering fast-charging interphases, and demonstrates the transformative potential of machine learning-guided materials discovery for advancing battery technology.