Exploring molecular interactions and dielectric relaxation in n-octanol/DMF binary mixtures: a machine...

The complex permittivity spectra (CPS) of n-Octanol and N, N-Dimethylformamide (DMF) mixtures were examined over the entire concentration range (0.0 → 1.0) and within the frequency range of 200 MHz to 20 GHz, utilizing a vector network analyzer (VNA) at 303.15 K. The complex permittivity data were fitted to various dielectric relaxation models using a complex nonlinear least squares method. The Cole-Cole model was applied to analyze the permittivity spectra, allowing for the determination of the static dielectric constant (ε₀), relaxation strength (Δε), and relaxation time (τ_d). The excess static dielectric constant (ε₀)^E and excess inverse relaxation time (1/τ_d)^E were also calculated and fitted using the Redlich-Kister polynomial. Various dielectric parameters, such as the Kirkwood correlation factor (g^eff, g^f) and Bruggeman parameter (f_B), were evaluated to explore molecular interactions and structural characteristics within the binary mixtures. The concentration dependence of the dielectric relaxation parameters provided insights into the molecular interactions between the components of the mixtures. In addition to traditional analysis, machine learning models were applied to predict the dielectric properties (ε' and ε″) of the mixtures across the frequency and concentration ranges. Models such as LightGBM, MLP Neural Network, and Gradient Boosting were employed, and their performance was evaluated using cross-validation techniques. LightGBM achieved the best predictive accuracy, closely followed by ensemble averaging methods. These models provided an efficient approach to predicting dielectric properties, reducing the need for extensive experimental measurements. This integration of experimental data and machine learning not only offered accurate predictions but also accelerated the characterization process, making it a valuable approach for studying dielectric behavior in complex liquid systems.