Accelerated Screening of Deep Eutectic Solvents for Nickel/Cobalt Extraction: Integrating Quantum Chemistry, Machine Learning, and Experimental Validation

This study provides a practical approach for fast-screening DESs with properties suitable for critical metal extraction, integrating quantum chemical calculations, and machine learning (ML) predictions. We proposed 218 combinations of hydrogen-bond acceptors and hydrogen-bond donors, potentially forming DESs. The Random Forest and XGBoost classifier models achieved performance scores of 0.72 and 0.74, indicating robust predictive capabilities. Then, we curated the best three DESs suitable as extractants for critical metals that can also transform into a liquid phase at room temperature. The screening process involves two main steps, namely, (i) estimating the solid–liquid equilibrium phase diagram for each mixture to identify DES formation and (ii) using ML models to predict metal extraction selectivity. This study introduces 2,2-bipyridine:Phenol as a novel DES; the extraction system demonstrating high efficiency for nickel and cobalt over manganese and lithium, achieving 99% for Ni and 97% for Co after 60 min from an aqueous solution containing metals typical of spent lithium-ion batteries. The separation factor (SF) at 5 min results further confirm the system’s strong selectivity for Ni, indicating that Ni is extracted 5.18 times more efficiently than Co. These findings highlight the system’s potential for efficient metal separation.