Automated dielectric property analysis in composite materials: Toward reproducible application mapping and machine-learning integration

Dielectric spectroscopy provides critical insight into relaxation dynamics and charge-transport mechanisms in functional polymer composites, yet its analysis is often limited by manual fitting, subjective parameter selection, and poor reproducibility. Here, we present an automated, physics-informed dielectric-spectroscopy framework that converts raw broadband spectra ε*( f, T ) (1 Hz-10 MHz) into validated models, uncertainty-aware parameters, and application-relevant performance indicators. The workflow integrates automated quality assurance, physics-based seeding, multi-start bounded Havriliak–Negami fitting, objective model selection, and cross-domain consistency checks across ε*, M *, and Z * representations. The approach is demonstrated on PVDF/zeolite Na–X nanocomposites spanning 0–50 wt.% filler content and temperatures from 30 to 120 ∘ C. The automated pipeline reliably identifies hidden low- and high-frequency relaxation processes, extracts activation energies and equivalent-circuit parameters, and achieves Arrhenius interpolation errors below 5% within the measured domain. By systematically mapping dielectric permittivity, loss, transport parameters, and interfacial polarization to composition, the framework enables quantitative comparison of materials across application targets such as high- k dielectrics, solid-state electrolytes, and sensing layers. By enforcing reproducible analysis, explicit uncertainty propagation, and machine-readable outputs compatible with data-science and graph-based learning models, this work advances dielectric spectroscopy from manual curve fitting toward a transparent, scalable, and data-driven materials-design methodology for sustainable energy applications.