Data-Driven Additive Discovery with HOMO-Descriptor Enables Durable Aqueous Zinc Batteries via Interfacial Kinetics Engineering.

Abstract

Dendritic growth and parasitic reactions severely hinder aqueous Zn-ion batteries due to interfacial instability and uncontrolled charge transfer. Here, a machine learning-accelerated strategy for rational additive screening, establishing a predictive framework that links the highest occupied molecular orbital (HOMO) energy level to the adsorption and reduction behavior of Zn²⁺, is reported. An interpretable machine learning model (Adaptive Boosting), trained on a curated molecular dataset, achieves high accuracy (Mean Squared Error = 0.2977, Pearson Correlation Coefficient = 0.8032) in HOMO prediction. Guided by this model, 4-dimethylaminopyridine is identified as a high-performance additive, which can suppress Zn dendrite formation by slowing interfacial charge transfer and mitigating local ion starvation through kinetic matching between mass transport and deposition. Moreover, 4-dimethylaminopyridine effectively excludes interfacial H₂O molecules, significantly inhibiting parasitic reactions. Consequently, Zn anode delivers high reversibility of plating/stripping with an average coulombic efficiency of 99.85% over 1600 cycles. The 0.3-Ah NaV₃O₈·1.5H₂O|Zn pouch cell delivers stable cyclability for 70 days, with a capacity retention of 73% after 250 cycles. This work pioneers the integration of machine learning with interfacial electrochemistry, offering a generalizable approach for additive discovery and electrolyte design, and sets a new paradigm for achieving dendrite-free metallic anodes in aqueous systems.