Estimating the phase diagrams of deep eutectic solvents within an extensive chemical space.

ABSTRACT

Assessing the formation of a deep eutectic solvent (DES) necessitates a solid-liquid equilibrium phase diagram. Yet, many studies focusing on DES applications do not include this diagram because of challenges in measurement, leading to misidentified eutectic points. The present study provides a practical approach for estimating the phase diagram of any binary mixture from the structural information, utilizing machine learning and quantum chemical techniques. The selected machine learning model provides reasonably high accuracy in predicting melting point (R2 = 0.84; RMSE = 40.53 K) and fusion enthalpy (R2 = 0.84; RMSE = 4.96 kJ mol-1) of pure compounds upon evaluation by test data. By pinpointing the eutectic point coordinates within an extensive chemical space, we highlighted the impact of the mole fractions and melting properties on the eutectic temperatures. Molecular dynamics simulations of selected mixtures at the eutectic points emphasized the pivotal role of hydrogen bonds in dictating mixture behavior.