Year: 2025
Type: preprint
Abstract: Electrolyte development remains a key challenge in advancing beyond-lithium-ion battery chemistries. Data-driven machine learning (ML) approaches have the potential to accelerate electrolyte discovery, but their impact depends on the availability of reliable, information-rich descriptors. Here we develop ORGA (operando Raman gradient analysis) into a high-throughput technique capable of rapidly quantifying key electrolyte properties—including di!usivity, transference number, conductivity, thermodynamic factor, bulk solvation structure, and interfacial impedance—all within a single experimental setup and in just minutes. We demonstrate these capabilities by characterizing a series of ether-based lithium-metal anode electrolytes. By integrating ML and battery modeling, we link these descriptors to plating and stripping e"ciency as well as rate performance. Our findings demonstrate that greater deviation from ideality in the electrolyte correlates positively with improved cycling stability, and illustrate how traditional approaches focused solely on enhancing ionic conductivity may lead to neither improved rate performance nor cycling stability.
Source: ChemRxiv
Authors: Junyi Zhao, Ben Jagger, Siyi Sun, Yiming Xu, Mauro Pasta
Institution: University of Oxford
Language: English
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: green
Awards: G160408179, G5833005875
OpenAlex ID: W4414366731
Cites:
Cited by:
Related to: