Intrinsic dielectric properties of faba bean for radio-frequency heating applications: Characterization and modeling

Dielectric properties influence radio-frequency (RF) heating, but measurements on granular materials reflect the combined response of grains and interstitial air and vary with packing density. This study characterized whole faba bean at 5–40.68 MHz, 10.34–20.14% moisture (wet basis), and 20–100 °C. Bulk permittivity was corrected using complex refractive index (CRI), Landau–Lifshitz–Looyenga (LLL), and Lichtenecker (LE) mixture models to estimate intrinsic dielectric properties. Monte Carlo uncertainty propagation yielded lower median relative uncertainty for CRI-estimated dielectric constant (ε′; 12.3%) than for LLL (13.5%) or LE (21.8%). CRI was selected as the reference approach based on published evidence for granular materials and its comparatively low propagated uncertainty. CRI-estimated dielectric constant (ε′) and loss factor (ε″) decreased with increasing frequency and generally increased with moisture content and temperature. A logarithmic-frequency model for ε′ achieved R 2 = 0.9955 in calibration and 0.9943 under grouped cross-validation. For ε″, a mechanistic power-law model with an optimized frequency exponent (n = 0.3323) was compared with reduced and extended empirical models. The extended model performed best in calibration (R 2 = 0.9834) and grouped cross-validation (R 2 = 0.9803), but its higher-order moisture dependence became unidentifiable when entire moisture levels were withheld. In contrast, the mechanistic model retained greater stability under moisture-level holdout (R 2 = 0.7032). These data and models support RF-heating simulation of faba bean within the characterized experimental domain, while highlighting limitations in extrapolation across moisture conditions.