Dielectric and machine learning study of chlorobenzene - n-butanol binary mixtures

The complex permittivity spectra ε*(f), complex modulus spectra M*(f), complex conductivity spectra σ*(f) and complex impedance spectra Z*(f) for the binary liquid mixtures of Chlorobenzene (CB) + n-Butanol (n-BA) over the whole compositional range (0.0 → 1.0) have been investigated at various temperature (303.15 to 323.15 K) in the frequency span 100 Hz to 2 MHz. The relaxation times corresponding to the different polarization mechanisms within the investigated frequency range of the applied AC electric field were determined. The analyzed dielectric data were further utilized to explain the contributions of electrode polarization and ionic conduction to the overall relaxation behavior of the mixture. The complex impedance data, Z*(ω), were analyzed by fitting to an equivalent four-element equivalent circuit model, which includes a capacitance (C₁, C₂) and resistance (R₁, R₂). Furthermore, supervised machine learning models including CatBoost, Extra Trees, Random Forest, and KNN_weighted were applied to predict ε' and ε″ from temperature, composition, and frequency. Model performance and generalization were rigorously evaluated using 5-fold and 10-fold cross-validation with statistical metrics such as R², mean absolute error (MAE), and mean squared error (MSE). The best models achieved R² = 0.835 for ε' with CatBoost and R² = 0.904 for ε″ with KNN_weighted, demonstrating strong predictive capability for CB + n-BA mixtures.