Understanding Green Solvents for Metals Recycling using Machine-Learned Molecular Dynamics

A projected two million tons of lithium-ion battery waste in the next decade requires development of novel, low-energy, minimally wasteful metals recycling processes to enable a sustainable and environmentally just energy transition. Recently, type III deep eutectic solvents have been gaining great interest as they can supposedly address all of these practical constraints. In particular, ethaline, a 2:1 molar ratio of ethylene glycol and choline chloride, leaches and electrodeposits critical minerals efficiently [1], albeit at the cost of thermal decomposition into toxic products [2]. Computationally-driven efforts to assess neat solvent and its phase stability over temperature cannot be achieved using refined classical molecular dynamics which do not include bond breaking, nor with accurate ab initio simulations due to prohibitive costs. However, machine-learning assisted molecular simulations can bridge the gap between accurate simulations and thermodynamic sampling of these fully explicit models.<br/><br/>Herein, we systematically benchmark density functional and quantum chemistry calculations and use them to construct a machine-learned force field (MLFF) for ethaline. We verify the fidelity of the MLFF against its existing thermodynamic properties and use large-scale molecular dynamics simulations to evaluate the onset of thermal decomposition, after which, at higher temperatures, we find that toxic byproducts such as chloromethane and dimethylaminoethanol are formed irreversibly. Our molecular-level assessment suggests a temperature upper operating limit of ethaline, and is the first to computationally examine the utility and stability of ethaline as a stable solvent for metals recovery applications. The results of our analysis can rationalize the selection of other solvents which span a vast underexplored, combinatorial design space.<br/><br/>[1] M. K. Tran, M.-T. F. Rodrigues, K. Kato, G. Babu, P. M. Ajayan, Nat. Energy 4, 339-345 (2019). doi.org/10.1038/s41560-019-0368-4<br/>[2] P. G. Schiavi, P. Altimari, E. Sturabotti, A. G. Marrani, G. Simonetti, F. Pagnanelli, ChemSusChem 15, 18, (2022). doi.org/10.1002/cssc.202200966