Experimental and molecular dynamics study of molecular interactions in γ-butyrolactone – dimethyl formamide systems with machine learning based density predictions

Abstract

Binary solvent mixtures are increasingly significant due to their ability to enhance reaction rates, modify solubility, and optimize technological separation processes. Among such mixtures, γ-butyrolactone (GBL) and dimethylformamide (DMF) are noteworthy for their versatility in various applicative fields, including pharmaceuticals, coatings, and adhesives. This study integrates experimental methods with molecular dynamics simulations and machine learning techniques to investigate the physicochemical properties and intermolecular interactions of the GBL-DMF binary system across different temperatures and compositions. The measured densities (ρ), speeds of sound (u), and refractive indices (n_D) of the binary mixtures were used to calculate excess molar volume ($V_m^E$), excess isentropic compressibility (${\kappa }_s^E$), and refractive index deviations (${∆}_{\varnothing }{n}_D$), which together give an overview of how the molecules interact with each other, suggesting the presence of strong molecular interactions such as hydrogen bonding and dipole-dipole forces. These observations are further corroborated by molecular dynamics simulations, which align well with the experimental data. Additionally, machine learning algorithms, including Random Forest, Gradient Boosting, XGBoost, and H2O AutoML, were employed to predict density. Among these, H2O AutoML demonstrated superior precision with an R² value of 0.984. This multifaceted approach, combining experimental, computational, and predictive methodologies, offers valuable insights into the design of solvent systems for industrial applications and supports sustainable development efforts.