Dielectric and refractive index analysis of ethylene glycol monomethyl ether-methanol mixtures: Molecular interactions and machine learning based predictions

Abstract

This study investigates the dielectric and optical properties of ethylene glycol monomethyl ether (EGMME), methanol (MeOH), and their binary mixtures at various concentrations and temperatures (293.15 K, 303.15 K, 313.15 K, and 323.15 K). The static permittivity (ε₀) and permittivity at optical frequency (ε_∞, equivalent to the square of the refractive index) were measured to analyze molecular interactions. Key dielectric parameters, such as excess static permittivity (ε₀^E), excess permittivity at optical frequency (ε_∞^E), effective Kirkwood correlation factor (g^eff), and Bruggeman factor (f_B) were evaluated. Negative ε₀^E values indicate reduced effective dipole moments, while positive ε_∞^E suggests enhanced electronic polarization, driven by interactions, particularly hydrogen bonding. Mixing rules were used to predict static permittivity and refractive index, with accuracy assessed via root mean square deviation (RMSD). Machine learning (ML) models: Random Forest, Gradient Boosting, and Extreme Gradient Boosting were trained on experimental data to predict the real (εʹ) and imaginary (ε”) part of the dielectric function across 20 Hz to 2 MHz. The models demonstrated high predictive accuracy, reducing experimental dependence and enabling efficient dielectric characterization. This study provides insights into EGMME-MeOH molecular interactions and presents a robust ML framework for predicting dielectric properties, with potential applications requiring controlled dielectric properties like coatings, pharmaceuticals, and material science.