International Journal of Numerical Modelling-Electronic Networks Devices and Fields最新文献

A graphite-inspired frequency switchable/reconfigurable MIMO (Multiple-input multiple-output) antenna is implemented for THz applications. Reconfigurability in the frequency response is achieved by layering the graphite in the slot of Antenna-2. The proposed antenna provides isolation better than 30 dB, high efficiency, and reasonable gain in the resonating band. An impedance bandwidth of 11.5% (2.04–2.29 THz) & 15.6% (4.06–4.74 THz) for Antenna-2 and 12.5% (3.53–3.99 THz) for Antenna-3 is achieved. Both the presented antennas demonstrate excellent diversity performance like Envelope Correlation Coefficient (ECC) < 0.03, Directive Gain (DG) > 9.95 dB; Total Active Reflection Coefficient (TARC) < −10 dB, and Channel Capacity Loss (CCL) < 0.35 bits/s/Hz, that can be incorporated into biomedical applications, high-frequency telecommunications, sensing applications, biological and astrophysical insights.

This paper presents a novel approach to analyze the electromagnetic interference (EMI) shielding effectiveness (SE) of polyurethane (PU) foam geometries, which are built in Blender software and simulated using CST Studio software. Three different batches of geometries were built to investigate the impact of pore diameters, pore randomness, and void fraction of PU foam on the SE, reflection coefficient (S₁₁), and electromagnetic absorption in the 26.5–40 GHz frequency range. The observed resonance frequency decreased with decreasing pore diameters and void fraction. Decreasing the pore diameter, increasing the pore randomness, and decreasing the void fraction enhanced the SE in the frequency range between the resonance frequency and 40 GHz. The EM absorption increased with increasing the pore diameter and randomness but decreased with increasing the void fraction. This study also presents simulations and measurements of Polytetrafluoroethylene (PTFE) and PU foam materials. The simulation results were compared with the measured ones obtained using vector network analyzer measurements to verify CST Studio's ability to accurately calculate the EM parameters. The measured and simulated results were in good agreement, confirming the accuracy of the results obtained using CST Studio. Our new parametric study fills a gap in existing literature since it combines for the first time an open-source 3D software for 3D rendering with an electromagnetic simulator to evaluate the impact of the pore topography (void fraction, diameter, randomness, etc.) on the EMI shielding performance of PU foams.

Fin-shaped Field Effect Transistors (FinFETs) are essential in the world of sub-nanometer technology nodes because of their remarkable scalability and electrostatic control. This work presents a new, optimized, and small-scale Gaussian-doped FinFET design that improves analog performance and minimizes short channel effects over conventional planar MOSFETs. Our unique structure leverages an Artificial Neural Network (ANN) in conjunction with a Genetic Algorithm (GA) for optimization. The dataset for ANN training was meticulously generated by designing and simulating Gaussian-doped FinFETs with varying Fin-width (W_Fin) and Fin-height (H_Fin). Through this process, we identified optimal W_Fin and H_Fin values that significantly improve performance characteristics. The optimized Gaussian-doped FinFET demonstrates superior control over short channel effects, as evidenced by a subthreshold swing (SS) of 66 mV/dec, an off-state current (I_OFF) of 3.54 pA, and an on-state current (I_ON) of 12 μA. The close alignment between the optimized and simulated performance characteristics, with less than a 5% variance, underscores the efficacy of our optimization approach.

In this article, a genetic algorithm based on coarse and fine encoding is proposed to assist the parameter extraction method of the coplanar waveguide model. First, an equivalent circuit model is proposed to accurately characterize the electrical characteristic parameters of the coplanar waveguide. The proposed quasi-empirical equivalent circuit model not only has a certain physical meaning but can also realize the solution of the nonlinear relationship between the device performance parameters and the dimensional structure parameters. Then, a single-step genetic algorithm is proposed to accelerate the parameter extraction based on the proposed semi-empirical model of the coplanar waveguide. On this basis, a coarse and fine encoding genetic algorithm is proposed to accelerate the parameter extraction. The proposed parameter extraction method not only avoids the problem of inaccurate element values that may be caused by artificially determining partial parameter values, but also omits the process of solving simultaneous equations. It can also avoid the problem of insufficient solution accuracy caused by the large order-of-magnitude difference between the values of equivalent circuit elements. Therefore, the quasi-empirical equivalent circuit model and the parameter extraction method accelerated by the coarse and fine encoding genetic algorithm proposed can achieve accurate and efficient modeling of devices.

Compared to traditional technology computer-aided design (TCAD) simulations, using neural networks to predict semiconductor device performance does not face convergence problems. This advantage is particularly significant when simulating devices made of materials like silicon carbide (SiC), which exhibit complex physical behaviors, making them difficult to converge in simulations. In addition, traditional TCAD software lacks the capability to deduce device structural parameters from device performance metrics. This article selects four critical structural parameters of 4H-SiC trench gate MOS devices: trench depth (D_t), gate oxide thickness (T_ox), drift region doping concentration (N_d), and P-region channel P-region length (L) as variables. Firstly, two types of neural network architectures were constructed and trained to serve as a classifier and a value predictor, respectively, among them, the breakdown mechanism classifier achieved an accuracy rate of 97% in the validation process. The average error of breakdown voltage prediction was 5.6%. In order to ensure the accuracy and stability of the prediction, we randomly selected 1000 sets of parameters within the value range for simulation to obtain a new dataset and improve the neural network structure. The improved neural network achieved average errors of 2.9% and 4.9% in the prediction of breakdown voltage and on-resistance, respectively. Subsequently, we built an optimizer based on the improved neural network, achieving an automated design process for device structural parameters according to target breakdown voltage and on-resistance. In the accuracy validation of the optimizer, the average error between target values and actual values of breakdown voltage and on-resistance is 2.5% and 7.9%, respectively.

Metaheuristic algorithms play a crucial role in tackling the increasing complexity and challenges in antenna design. The nutcracker optimization algorithm (NOA), a novel metaheuristic inspired by nutcrackers' food-gathering, storing, searching, and retrieving behaviors, has shown excellent performance on 23 standard test functions and CEC—2014/2017/2020 test suites compared to well-established algorithms, yet it remains unapplied in antenna design. This study proposes a multi-strategy improved NOA (MINOA) to resolve NOA's unbalanced exploration and exploitation issues, applying it to ultra-wideband antenna design optimization and linear antenna array sidelobe suppression. MINOA employs Bernoulli chaotic mapping for uniform population initialization, a dynamic boundary strategy for balanced exploration and exploitation, and adaptive t-distribution disturbance to accelerate convergence and enhance local exploitation. Extensive tests on 23 benchmark functions prove MINOA's superiority in optimization accuracy, convergence speed, and stability over advanced algorithms such as NOA, WOA, GWO, SSA, DEA, SCSO, and HBMO. The Wilcoxon signed-rank test validates its significant improvement in accuracy. In broadband antenna optimization via an artificial neural network (ANN)-based surrogate model, MINOA reduces the mean square error (MSE) by 40.9% at the same iteration number and by 28.6% with 10 fewer iterations and 29 fewer fitness function calls compared to NOA during the preliminary training phase, achieving the widest bandwidth (3.62–11 GHz) among the eight algorithms. The Wilcoxon signed-rank test confirms MINOA's superiority. In the 16-element linear antenna array optimization, although MINOA performs slightly worse than DEA and WOA, it still achieves a low-sidelobe level of −41.38 dB, verifying its feasibility.

Higher-order boundary value problems (BVP) of differential equations (DEs) are important in the mathematical description of many real-world processes. Solving such problems for exact or analytical solutions is not always easy to deal. Therefore, to compute their numerical solution, we need some numerical methods. Hence, in this work, a powerful numerical procedure based on Haar Wavelet (HW) method is established to deal with fourteenth-order BVPs linear and nonlinear. A generalized form of the algorithm is developed under general boundary conditions. Then the numerical method is verified on various examples from the literature. Also, maximum and root mean square errors are calculated. Moreover, a comparison between exact and numerical results is shown at different collocation points. Furthermore, convergence rate is approximately 2 at various numbers of nodal points is also calculated.