Screening of Protic Ionic Liquids with Catalytic Potential in the Transesterification Reaction Using COSMO-RS and Machine Learning

Global population growth has led to the use of fossil fuels and global pollution problems. Biodiesel, a renewable and environmentally friendly alternative to petroleum fuel, is produced from organic oils and animal fats, causing food safety issues. Unprocessed crude oils are inexpensive raw materials with a high content of free fatty acids. Ionic liquids (ILs) are used as catalysts for biodiesel production to solve the problems of traditional catalysts. This manuscript proposes the COSMO-RS model and machine learning as predictive tools for screening ILs as catalysts for fatty acid methyl esters (FAME) synthesis. COSMO-RS activity coefficient model was used to obtain the ILs sigma profile and interaction energies (electrostatic-misfit (Emisfit), hydrogen bond (EHB), and van der Waals (EvdW)) to correlate the yield of reaction. The machine learning models, such as K-nearest neighbor, Random Forest Regressor, Decision Tree Regressor model, Gradient Boosting Regressor, and Multilayer Perceptrons model, were applied to correlate the above-mentioned properties. The Gradient Boosting Regressor model, using the analysis of the anion and cation sigma profiles, proved to be more efficient than the same model, using the approach of the interaction energies. Based on the screening study, the ILs [l-arginine][Acetate], [l-arginine][HSO3], and [l-arginine][NO3] were selected, synthesized, and characterized.