Data-Driven Exploration of Weakly Solvating Additives toward the Stabilization of Zn Metal Anodes

Abstract

The introduction of weakly solvating additives (WSAs) into aqueous electrolytes holds significant potential in promoting the desolvation kinetics of Zn2+ and reducing polarization. However, the current strategy of WSA screening primarily depends on trial-and-error experiments and theoretical calculations. This study reveals that, among various molecular features, the electrostatic potential minimum (ESPmin) exhibits the strongest correlation with the desolvation activation energy (Ea), demonstrating that ESPmin could serve as an effective descriptor for screening WSAs. Herein, we propose a data-driven screening strategy based on ESPmin, in which ESPmin can be accurately predicted directly from molecular structure using a graph convolutional neural network (GCN) model. Assisted by this strategy, several promising WSAs, including tetrahydropyran, methanol, and acetone, were successfully identified. As a proof of concept, tetrahydropyran was selected as the additive for systematic studies. Remarkably enhanced cycling stability with a lifespan of over 2400 h and low overpotential was demonstrated for the symmetric cell with tetrahydropyran. The proposed data-driven strategy enables the rapid screening of WSAs directly from molecular structures, offering an efficient pathway for the exploration of high-performance electrolyte additives.