Machine learning-driven discovery of innovative hybrid solid electrolytes for high-performance all-solid-state batteries

Abstract

Research has actively focused on polymer/oxide-based hybrid solid electrolytes (HSEs) for next-generation all-solid-state batteries (ASSBs) with high energy densities and excellent safety. To accelerate the commercialization of ASSBs to replace existing lithium-ion batteries (LIBs), new HSE materials with excellent ionic conductivity, electrochemical stability, and the ability to suppress lithium dendrite growth, must be developed. In this study, 61 candidates from 10,368 dual-doped Li 7 La 3 Zr 2 O 12 (LLZO) compositions were screened using machine learning (ML) and density functional theory, to satisfy the fundamental criteria for solid-state electrolytes (SSEs), including a bandgap (E g ), energy above the convex hull (E hull ), ionic conductivity, and elastic properties. Among these, five promising dual-doped LLZO candidates are used as fillers to create poly(vinylidene fluoride-co-hexafluoropropylene) (PVDF-HFP)-based HSEs, that demonstrated a superior electrochemical performance compared with that of bare HSEs made with undoped LLZO fillers. Notably, Zn 0.25 and Ti 0.25 mol dual-doped LLZO (Zn-Ti HSE), used as a filler, exhibited the best electrochemical performances with an ionic conductivity of 8.7 × 10 −4 S cm −1 at 25 °C and electrochemical stability of ∼4.8 V at 55 °C. Furthermore, LiFePO 4 /HSE/Li ASSBs incorporating the two best-performing Zn-Ti HSE demonstrated a superior initial specific capacity (ISC: 167 mAh g −1 at 0.2C), cycling performance (retention: 91 % at 100 cycles), and rate capability (160 mAh g −1 at 1.0C) compared with those of the bare HSE (ISC: 120 mAh g −1 , retention: 82 %, and 55 mAh g −1 at 0.2C). Our findings suggest that an ML-based screening combined with experimental characterization can accelerate the finding of promising SSE materials.

Topics & Concepts