Predicting Miscibility of Ionic Liquids in Binary Mixtures.

Area of Science:

• Physical Chemistry
• Materials Science
• Computational Chemistry

Background:

• Blending ionic liquids (ILs) creates superior solvents with tailored properties.
• Exploring IL-IL mixtures is computationally expensive due to vast design space.

Purpose of the Study:

• Develop a machine learning (ML) framework to predict IL-IL mixture miscibility.
• Accelerate the discovery of novel IL solvents.

Main Methods:

• Trained an artificial neural network (ANN) on 4.9 million COSMO-RS-generated activity coefficients.
• Generalized predictions across temperatures using excess enthalpy.
• Analyzed molecular descriptors and mixture compositions for miscibility correlation.

Main Results:

• Achieved high accuracy in predicting binary IL miscibility using the ANN classifier.
• Identified key molecular factors influencing IL miscibility.
• Experimental validation confirmed computational predictions.

Conclusions:

• The ML framework efficiently predicts IL miscibility, enabling rapid solvent discovery.
• Provides insights beyond the 'like dissolves like' principle for IL mixtures.
• Facilitates targeted design of ionic liquid solvents.