Molecular design of electrolyte additives for aqueous zinc-ion batteries via reinforcement learning and quantum chemistry calculations

Abstract

The use of electrolyte additives is regarded as a cost-effective strategy to regulate the components of aqueous zinc-ion batteries (AZIBs) and to improve their overall electrochemical performance. Here, we demonstrate an artificial-intelligence-guided framework that integrates machine learning and quantum chemistry calculations to accelerate the rational design of electrolyte additives for AZIBs. Using a molecular generator based on a recurrent neural network and the Monte Carlo tree search method, we efficiently identified seven top-ranked candidate molecules, including imidazole, heptane, and dihydropyrimidinone derivatives. These molecules bind Zn2+ ions more strongly than H2O does, due to their higher highest-occupied molecular orbital (HOMO) energies. Furthermore, their binding strengths surpass those of established electrolyte additives such as pyridine, 1,2-dimethoxyethane, and tetrahydrofuran. High-level quantum chemistry calculations reveal that the spatial localization of the HOMO-1 orbital plays a critical role in determining the preferred coordination site for Zn2+. Force-field molecular dynamics simulations provide direct evidence that these molecules preferentially appear in the first solvation sheath structure of Zn2+ ions, effectively modifying the conventional hydration structure of [Zn(H2O)6]2+. This study substantially shortens the screening cycle of functional molecules and provides new insights into the molecular design of electrolyte additives for AZIBs.