Journal of Computational Electronics最新文献

Machine learning has recently been introduced into metasurface design to improve the optimization efficiency. Nevertheless, existing methods are primarily limited to single-objective tasks and often yield suboptimal efficiency. Therefore, this paper proposes a general multi-objective optimization framework based on NSGA-II for the design of VO₂ (vanadium dioxide) metasurface absorbers. Results demonstrate that the optimization framework successfully optimizes two different fitness functions (absorptance and operational bandwidth) and exhibits exceptional iterative efficiency, converging rapidly within just 7 iterations, faster than the conventional methods (around 50 iterations). Optimized absorber achieves near-perfect absorption of 99.41% and ultra-wide operational bandwidth of 18.5 THz. The research mechanism shows that the high absorptance stems from the strong mode coupling of FP and LSPP (Localized Surface Plasmon Polariton) modes. The absorber reveals a good impedance matching with the free space, where the real and imaginary parts are close to 1 and 0, which also explains the perfect absorption from the impedance matching perspective. Ultimately, the absorber displays an excellent angle tolerance and polarization-insensitive properties. The proposed efficient multi-objective optimization framework has the potential for the design of achromatic metalenses, sensors, and detectors.

Solar absorbers are essential components in solar thermal energy systems, as they directly capture and convert sunlight into heat while aiming to achieve high efficiency with minimal thermal losses. The engineering of multilayer absorber structures offers a promising approach to maximize spectral absorption performance across a wide range of incident radiation. In this study, a graphene-integrated tri-layer absorber based on Cr-Ti-TiC materials is designed to enhance optical absorption and thermal conversion efficiency. The proposed structure demonstrates remarkable broadband absorption, achieving 97.49% absorption at 1000 nm and 94.13% absorption at 2800 nm, covering the ultraviolet, visible, and infrared wavelength regions. To further improve performance and ensure optimal design parameters, machine learning regression techniques are applied, enabling accurate prediction of absorber behavior and supporting intelligent geometric optimization. The comprehensive computational results indicate that the developed absorber is well suited for advanced solar thermal applications, offering significant potential for both household and industrial energy systems. Additionally, this work provides valuable insights into scalable fabrication approaches for multilayer graphene-based absorbers, contributing to future research and development in renewable energy harvesting technologies.

Antenna array synthesis is one of the most significant optimization problems in electromagnetics. In addressing the challenge of balancing convergence accuracy and computational efficiency in existing optimization algorithms, this paper introduces a chaos-enhanced adaptive multi-mutation arctic puffin optimization (CEAM-APO) algorithm. The proposed algorithm establishes a three-stage enhancement framework that significantly improves optimization performance in electromagnetic applications, including sidelobe suppression and null control. First, the presented approach integrates the tent chaotic mapping strategy for uniform initialization during the initialization stage, effectively reducing the risk of premature convergence to local optima. Next, a fusion mutation strategy based on t-distribution perturbation with adaptive scaling is used to adjust the mutation probability adaptively during the middle stage, improving both convergence speed and accuracy. Subsequently, a local refinement operator incorporating the golden sine for directed refinement is introduced to update the tail individual in the last stage, enhancing overall solution performance. To verify the performance of this framework, evaluations on four classical benchmark functions demonstrate its performance over the original APO and other mainstream algorithms (PSO, JS, GWO, SA) in both convergence speed and solution accuracy, achieving a 100% success rate. Furthermore, when applied to linear and planar array synthesis, the proposed CEAM-APO method also demonstrates stable and superior results in key performance metrics such as sidelobe suppression and null control.

In this study, the impact of transition metal doping (Cu and Fe) on the structural, electronic, and interconnect properties of scandium nitride (ScN) nanoribbons of width 6 investigated for application in VLSI interconnect technology. Atomic substitutions of one and two Cu or Fe atoms were modeled, and the resulting configurations analyzed based on binding energy, electronic structure (Fermi energy and bandgap), and parasitic interconnect parameters. The calculated quantum resistance, kinetic inductance, quantum capacitance, and RC-delay are presented. DFT-based calculations reveal that Fe doping, particularly ScN-2Fe, yields the highest structural stability (-6.44 eV), while Cu doping increases the bandgap, improving semiconducting behavior. In interconnect performance, ScN-1Cu possesses the optimal combination of high Fermi velocity, lowest parasitic inductance and capacitance, and lowest RC-delay (12.6 µs) and is therefore a perfect choice for high-speed, low-loss VLSI interconnects. These findings underscore the key significance of dopant selection and concentration in determining the stability vs. conductivity vs. signal delay compromise in nanoscale interconnect materials.

We present the design and simulation of a wavelength-tunable high-slope-efficiency (SE) and on-chip-output-power single-longitudinal-mode (SLM) erbium–ytterbium (Er–Yb)-co-doped waveguide laser (EYCDWL) operating in the C-band using OptiSystem software. The laser is implemented using a 1-cm-long EYCDW built on a lithium niobate-on-insulator (LNOI) platform and a single 980-nm backward pump. Wavelength tuning of 35 nm in the range of 1530–1565 nm is simulated with a side-mode suppression ratio (SMSR) over 59 dB for coupling ratios of 40%, 50%, and 60%. Simulation results indicate that the highest SE and on-chip output power of 67.2% and 318 mW, respectively, are obtained at a lasing wavelength of 1545 nm for a coupling ratio of 50%. Similarly, a minimum linewidth (LW) of 125 MHz is observed for a lasing wavelength of 1550 nm considering a 60% coupling ratio. Power variation of around 0.13 mW is observed for a lasing wavelength of 1545 nm considering a 50% coupling ratio based on 12 iterations carried out at 5-min intervals. Finally, the impact of energy transfer upconversion (UC) on the output power of the EYCDWL is also analyzed. A power penalty of around 100 mW is observed in on-chip lasing power. This tunable laser is a candidate for integration with other photonic components including wavelength lockers, modulators, and photodetectors, realizing complete photonic integrated chips (PICs) for various applications, such as short-range free-space optical communication, sensing, and next-generation computing applications.

The design and optimization of epitaxial doping profiles for heterojuction bipolar transistorsrequire expert know-how and extensive trial-and-error using technology computer-aided design (TCAD) simulations. The vastness of the design exploration space along with time-intensive device simulation quickly renders conventional approaches infeasible. In this work, we propose a data-driven inverse design framework based on a surrogate Bayesian neural network model that not only predicts device performance with high accuracy, but also provides uncertainty estimates for each prediction. Leveraging these uncertainties, we implement an active learning scheme to iteratively refine the surrogate model by selecting the most informative new simulation points and thus ensuring maximal performance gain with minimal simulation times. Our results demonstrate that active learning leads to (>45%) improvement in network performance over random optimization approaches—paving a promising path to automated and efficient optimization. Finally, the efficacy of the approach is showcased by obtaining candidate epitaxial profiles through the framework that surpass expert-tuned state-of-the-art epitaxial designs for linearity and gain figure-of-merits.

This study presents an SPR-based biosensor designed to effectively detect Mycobacterium tuberculosis, leveraging 2D materials and a Kretschmann configuration, with all analyses and performance evaluations conducted through comprehensive simulation studies. The biosensor utilizes a CaF₂ prism, niobium pentoxide (Nb₂O₅), silver (Ag), ferric oxide (Fe₂O₃), and black phosphorus (BP). The work concentrates on the design and optimization of the SPR structure to obtain an exceptionally high detection sensitivity for mycobacterium tuberculosis (MTB). Strong plasmonic characteristics are provided by the combination of Ag and Fe₂O₃, and sensitivity is greatly increased by the BP-tunable bandgap and improved light–matter interactions. By facilitating surface functionalization and improving chemical stability, Nb₂O₅ makes the sensor reliable for practical uses. The primary modeling approach was the transfer matrix method (TMM) and finite element method (FEM) was employed to verify the simulation results in order to further validate the design. A maximal sensitivity of 622.33 (deg./RIU) was achieved by fine adjusting the thickness of the Nb₂O₅, Ag, Fe₂O₃, and BP layers in order to optimize the sensing structure. The sensor’s impressive QF of 142.02 (RIU⁻¹), high FOM of 140.09, and narrow FWHM of 4.382 (deg.) suggest strong resolution and detection precision. Moreover, the biosensor exhibited a consistent ability to detect refractive index (RI) fluctuations within the biologically pertinent range of 1.343–1.3551. Superior angular sensitivity and improved optical performance are demonstrated by the proposed structure in comparison with existing SPR biosensor configurations. These findings suggest that the proposed SPR sensor could be an effective and rapid tool for point-of-care testing for TB.

Graphical Abstract

This work presents the numerical design and computational analysis of a biosensing platform that utilizes a two-dimensional photonic crystal structure with a silicon rods-in-air configuration, intended for the precise detection and measurement of urinary biomarkers such as glucose, urea, and albumin. The design incorporates a linear waveguide coupled to a central hexagonal ring resonator, in which changes in the refractive index of the analyte induce measurable resonant wavelength shifts that form the basis of the sensing mechanism. The photonic band structure of the sensor is computed using the plane-wave expansion method, whiles its transmission spectrum and resonant characteristics are evaluated through a finite-difference time-domain approach. Through parametric optimization of the sensing rod radius, the design achieves a high sensitivity of 900 nm/RIU and a high quality factor of 1.4267(times)10(^textrm{6}), along with a low detection limit of 1.276(times)10(^{-7}) RIU and an impressive figure of merit of 7.8372(times)10(^textrm{5}) RIU(^{-1}). The sensor demonstrates excellent linearity, with correlation coefficients of 0.99995 for glucose and 0.99957 for urea. Furthermore, the impact of fabrication-induced disorders on performance is analyzed to assess robustness. A three-dimensional simulation validates the design’s feasibility, and a feasible fabrication process is outlined for future experimental realization. With a compact footprint of 130.235 (upmu text {m}^2), the proposed design is well suited for photonic integrated circuit applications.

The small-signal intrinsic noise behavior of a GaN high electron mobility transistors (HEMTs) was modeled using the whale optimization algorithm-hybrid kernel extreme learning machine (WOA-HKELM) algorithm. This algorithm not only addresses the issues associated with extreme learning machine (ELM), such as the challenge of determining the number of hidden-layer nodes and the occurrence of overfitting, but also identifies the optimal regularization coefficient C and kernel parameters S that support the HKELM algorithm through the utilization of the WOA. To verify the superiority of the proposed WOA-HKELM algorithm, small-signal noise modeling experiments were conducted on GaN HEMT devices with different gate dimensions under various bias conditions. The modeling performance of WOA-HKELM was compared with that of the improved Sparrow Search Optimized Hybrid Kernel Extreme Learning Machine (MShOA-HKELM), HKELM, and other algorithms. Experimental demonstrate that the WOA-HKELM achieves higher accuracy and stability in modeling the intrinsic small-signal noise characteristics of GaN HEMTs.