Accelerated Electrolyte Engineering by Deep Learning: A Quaternary Ammonium Additive Enables Ultra-Stable Zinc Anodes

Abstract

Currently, achieving long-cycling zinc anodes for aqueous zinc-ion energy storage devices remains a challenge owing to uncontrolled dendritic growth and concurrent hydrogen evolution reaction (HER). Electrolyte engineering via molecular additives has emerged as an effective strategy for modulating zinc anode interfacial chemistry, but the vast structural space of candidate molecules renders conventional trial-and-error screening inefficient. Here, we present a deep learning (DL) framework for screening high-performance quaternary ammonium cation (QAC) additives for aqueous zinc electrolytes. We developed GATGLP, a graph attention network-based HOMO–LUMO gap (Gap) predictor, to address the data scarcity inherent in ionic molecular systems, which was first pretrained on a curated subset of the QM9 molecular data set and subsequently fine-tuned via transfer learning on a DFT-calculated QAC data set. Through a dual-descriptor screening strategy based on the predicted Gap and the octanol–water partition coefficient (XlogP), a validated library of 378 QAC compounds was ranked, and three candidates were selected for systematic electrochemical validation. Among them, n-octyltrimethylammonium bromide ([N8111]Br) showed a Zn//Zn cycling duration of more than 1800 h at 1 mA cm–2/1 mAh cm–2 and a cycling lifespan of approximately 180 h at 5 mA cm–2/5 mAh cm–2. Across the ZSO-Br, N6111 and N8111 electrolytes, which share the same bromide concentration, three quantities measured by different methods show the same ordering: the relative (002) diffraction proportion, the maximum −Z″ of the impedance arc, and the Zn//Cu sequence length increase from ZSO-Br to N6111 to N8111. The chronoamperometric transient does not follow this ordering, but N8111 is the only electrolyte in which the current magnitude decreases during the step. Taken together, these observations suggest that the selected organic cations may differ in their interfacial accumulation, which could redistribute the Zn deposition flux and contribute to flatter, more continuous deposits. This work combines DL-assisted molecular screening with experimental validation to identify QAC additives for aqueous zinc electrolytes.