dielectric dispersion modeling

This study addresses the limitations of existing head electrical models in accurately and efficiently capturing the frequency-dependent properties and capacitive effects of brain tissue. The authors propose a lumped-parameter equivalent circuit model based on a three-shell (brain, skull, scalp) geometry, which explicitly incorporates the dispersive electromagnetic characteristics of tissues through radial and tangential RC branches, thereby integrating frequency-dependent conductivity and permittivity. Formulated under the electroquasistatic approximation, the model is validated across multiple geometries and frequencies using semi-analytical solutions based on spherical harmonics. Results demonstrate that neglecting dispersion and capacitive pathways leads to significant overestimation of scalp potentials, whereas the proposed model achieves excellent agreement with reference solutions, offering both high accuracy and broad applicability.

This study addresses the high computational cost of conventional electromagnetic simulations, which hinders rapid prototyping and circuit integration for neural sensing and stimulation systems. The authors propose a minimal lumped RC equivalent circuit model that accurately captures the electrical characteristics of a three-layer spherical head model below 50 kHz. Built upon the electroquasistatic approximation, the model incorporates complex conductivity to simultaneously account for tissue dispersion and displacement current effects—a first in low-frequency neurophysiological modeling. Validation against semi-analytical solutions and scalp potential simulations driven by dipole sources demonstrates high fidelity across varying skull thicknesses and dipole eccentricities. The resulting model achieves an optimal balance of accuracy, simplicity, and compatibility with integrated circuit implementation.

This study addresses the scarcity of experimental data on the dielectric properties of biological tissues in the terahertz (THz) band, a critical gap hindering accurate channel modeling for in-body nanoscale communication. Utilizing a photoconductive antenna-based THz time-domain spectroscopy system, the work presents a comprehensive broadband characterization of porcine skin—used as a human tissue surrogate—across the 0.1–11 THz range. It reports, for the first time, an experimental dataset of complex permittivity, refractive index, and absorption coefficient extending up to 11 THz. The results reveal strong water-induced absorption at lower frequencies, alongside pronounced dispersion and narrowband selective transmission characteristics at higher frequencies. These findings provide essential empirical foundations and critical parameters for the development of realistic THz in-body communication channel models.

This work addresses the challenge of accurately characterizing the layered dielectric properties of human skin in the sub-terahertz to terahertz frequency range—a critical barrier to advancing non-invasive diagnostic and imaging applications in this band. The authors propose an integrated dielectric model that combines multi-pole Debye relaxation theory with effective medium approximation, uniquely incorporating key biophysical parameters such as cellular-scale water content, protein-to-lipid ratios, and ionic conductivity into a stratified skin representation. This approach enables a systematic description of the frequency-dependent dielectric response across distinct skin layers while maintaining both physical interpretability and spectral accuracy. The model achieves high-fidelity predictions of dielectric behavior for different skin layers and cell types, thereby establishing a robust theoretical foundation for next-generation terahertz-based non-invasive diagnostic and imaging systems.

This study addresses the lack of systematic modeling approaches for wearable galvanic-coupled communication channels across varying bandwidths. It presents the first physically consistent digital human twin that unifies anatomical structure, propagation geometry, and electrode–skin interfaces into a complex-valued transfer function. By integrating electro-quasistatic theory with experimental validation, the work quantitatively elucidates how bandwidth, interface conditions, and geometric factors govern signal attenuation, phase response, and group delay. Results demonstrate that within the 10 kHz–1 MHz band, the channel exhibits weak dispersion, with attenuation primarily dictated by propagation geometry. Increasing bandwidth amplifies magnitude ripple and delay fluctuations, whereas optimizing the electrode–skin interface significantly enhances both amplitude and phase stability.

This study addresses the challenge of early cancer diagnosis by leveraging intrinsic differences in the electrical properties of healthy and malignant cells. The authors construct the first integrated dataset of multi-source bioelectrical parameters—including conductivity and permittivity—through a systematic review of 33 published studies. They employ three supervised machine learning algorithms—Random Forest, Support Vector Machine, and K-Nearest Neighbors—with hyperparameter optimization to enhance classification performance. The optimized Random Forest model, configured with 100 estimators and a maximum depth of 4, achieves an accuracy of 90%, while KNN and SVM attain F1 scores of 78% and 76.5%, respectively. This work represents the first systematic integration of diverse bioelectrical features with machine learning, offering a high-accuracy, label-free, and non-invasive approach for cancer screening.

This study addresses the challenge of efficiently predicting electromagnetic wave propagation in heterogeneous media with material interfaces, where high-fidelity numerical simulations are computationally prohibitive. The authors propose a physics-informed, frequency-embedded vision Transformer-based autoregressive surrogate model, which—novelty—incorporates Fourier operators into its latent space to encode one-dimensional solutions of Maxwell’s equations. Trained on simulation data generated via the finite volume method, the model achieves accurate roll-out predictions beyond 75 time steps with relative errors below 10%, even in the presence of material discontinuities and unknown parameters. It faithfully reproduces key wavenumber spectral characteristics and dynamic interfacial responses, thereby substantially improving both long-term prediction accuracy and computational efficiency compared to conventional approaches.

This work addresses the well-known ill-conditioning of the electric field integral equation (EFIE) at low frequencies, high frequencies, and fine discretizations, which hinders efficient numerical solution. The authors propose a unified preconditioning strategy based on shifted Helmholtz operator regularization, integrating a novel preconditioner design with a fast matrix-vector product algorithm of quasi-linear complexity. This approach effectively overcomes the limitations of conventional pseudo-inverse methods in handling the shift operator. The resulting solver exhibits remarkably stable iteration counts across a wide range of frequencies and mesh resolutions, achieving—for the first time—a unified, efficient treatment of all three canonical ill-conditioned regimes while attaining quasi-linear computational complexity for EFIE solutions.

This work addresses the susceptibility of conductivity reconstruction in electrical impedance tomography (EIT) to noise and its reliance on explicit regularization by proposing an implicit parameterization framework that eliminates the need for such regularization. The method generates physically plausible conductivity distributions through low-dimensional latent variables and a composite nonlinear mapping, incorporating truncated graph Laplacian embeddings to inject structural priors and employing bound-preserving mappings to improve the conditioning of the optimization problem. The approach achieves, for the first time, high-quality three-dimensional time-difference EIT reconstructions, demonstrating superior physical consistency, structural fidelity, and robustness across 2D/3D simulations, phantom experiments, and in vivo lung data. Notably, it substantially enhances 3D spatial resolution, thereby strengthening the clinical potential of EIT for respiratory monitoring.

This study addresses the challenges of local optima and non-unique solutions in extracting the complex permittivity of polymers by proposing a gradient-enhanced non-dominated sorting genetic algorithm (G-NSGA-II). For the first time, gradient information is incorporated into the NSGA-II framework, combined with transmission and reflection coefficients from multi-thickness samples to formulate a multidimensional constrained model. A population stagnation detection mechanism is further introduced to adaptively trigger local refinement. Experimental validation on six representative polymers in the 20–40 GHz band demonstrates that the proposed method significantly accelerates convergence—reducing the required generations by approximately 50%—and enhances inversion robustness. The retrieved complex permittivity and thickness values show excellent agreement with literature data and direct measurements, thereby substantially improving the efficiency and reliability of broadband dielectric characterization.