In this work, a silicon-on-insulator based heterojunction tunnel field-effect transistor (SOI-HJ-TFET) biosensor with nanocavity is proposed first time for the detection of key lung cancer biomarkers, namely carcinoembryonic antigen (CEA), cytokeratin-19 fragment 21-1 (CYFRA 21-1), and neuron-specific enolase (NSE). The proposed device utilizes a low-bandgap magnesium silicide ($Mg_{2}Si$) source to enhance band-to-band tunneling, improving sensitivity. The nanocavity is functionalized with an APTES-glutaraldehyde matrix for the immobilization of biomarker molecules. The detection methodology employs a charge deduction approach, where biomarker concentrations are converted into an equivalent interface charge, modulating the device's electrical characteristics. Using calibrated SILVACO ATLAS TCAD simulations, the device's performance is evaluated for both non-tumorigenic (NTG) and tumorigenic (TG) cell concentrations. The results demonstrate that the presence of cancerous cells leads to a significant reduction in drain current, electron concentration, and band-to-band tunneling rate due to a higher negative equivalent interface charge. The biosensor exhibits high inverse sensitivity, with values up to 12×$10^{3}$ for TG CYFRA 21-1 cells. Key performance metrics, including changes in subthreshold swing ($\Delta SS$ up to 203 mV/decade) and threshold voltage ($\Delta V_{T}$ up to 637 mV), provide clear differentiation between healthy and cancerous CEA and CYFRA 21-1 cells, respectively. Additionally, sensing time analysis shows rapid response ranging from nanoseconds to microseconds, making the device suitable for real-time biosensing. The proposed SOI-HJ-TFET biosensor exhibits high inverse sensitivity, fast response, and strong inverse selectivity, particularly for CYFRA 21-1, indicating its potential for efficient early-stage lung cancer detection in point-of-care diagnostic platforms.
Predicting the performance of biological tunneling field-effect transistor (Bio-TFET) sensors typically requires computationally intensive and time-consuming Technology Computer-Aided Design (TCAD) simulations. This study proposes an alternative approach—a hybrid Technology Computer-Aided Design–machine learning (TCAD–ML) framework that accelerates and enhances performance prediction for a novel dual source vertical charge plasma-based BioTFET (DSVCP-TFET). This device uses a Si/SiO₂ material system in a vertical setup and a charge plasma technique. For the machine learning side, we designed a fully connected artificial neural network (ANN) with a 32-32-16-8-1 structure. They trained it using TCAD-generated current-voltage data for four different biomolecules: air (k = 1), Urease (k = 1.64), Ferro-cytochrome c (k = 4.7), and Gelatin (k = 12). The trained model demonstrates excellent predictive accuracy, with R² values consistently exceeding 0.997 across all datasets. When it comes to drain current sensitivity for Gelatin, the ML model predicted 1.27 × 10⁵, which matches up well with the TCAD result of 1.79 × 10⁵. This strong agreement further validates the effectiveness of the hybrid framework. The ML model also successfully mirrors the selectivity metrics for different biomolecule pairs. Furthermore, this TCAD–ML method shows that data-driven strategies aren't just a shortcut—they offer a fast, powerful option compared to sticking with pure simulation.
An optical mode switch enables reconfigurable mode-division-multiplexing (MDM) techniques. Here, an integrated 3 × 3 non-blocking dual-mode thermo-optic switch is proposed utilizing a Mach-Zehnder interferometer integrated with multimode interference (MMI) couplers. The design of the dual-mode 3-dB power splitter leverages the self-imaging phenomenon facilitated by a taper-assisted MMI coupler. Meanwhile, a mode-insensitive phase shifter is utilized to alter the two modes phases simultaneously. The insertion loss for the TE0 and TE1 modes can reach 0.4 dB and 0.45 dB at 1550 nm; for the C-band, it is less than 0.57 and 0.58 dB, respectively. The crosstalk for the TE0 and TE1 modes can achieve -39.59 dB and -39.54 dB, respectively, at 1550 nm, and for the C-band, it is below -35.5 dB for both modes. The device has a compact size, measuring 2045 μm × 12.5 μm. Our innovative dual-mode switch offers low-power consumption benefits. It is engineered for effortless integration with MDM systems, catering to the needs of upcoming high-capacity communication networks.
Conventional processing-in-memory (PIM) architectures suffer from limited efficiency due to transistor-intensive adder trees and analog-to-digital converter (ADC) overhead. This work presents BRAIN-AD, a bit-serial, ReRAM-based digital PIM macro for energy-efficient autonomous driving assistance systems (ADAS). The proposed design integrates a compact 1-bit multiply-accumulate (MAC) unit combining a 3T1R Re-XNOR non-volatile bit-cell for in-memory multiplication with an area-efficient 10T pass-transistor full adder for sequential accumulation. A 16 Kb (128×128) macro employs sparsity-aware power gating and supports scalable fixed-point computation from 1 to 16 bits via bit-serial execution. Post-layout simulations in 65 nm CMOS achieve peak throughput of 0.72 TOPS and 112 TOPS/W energy efficiency, providing approximately 1.8× higher throughput and 1.95× higher energy efficiency than state-of-the-art digital PIM designs. System-level evaluation using a quantised INT4 NVIDIA PilotNet model shows less than 2.5% accuracy degradation relative to the FP32 baseline. These results establish BRAIN-AD as a robust, scalable, and practical digital PIM solution for resource-constrained ADAS workloads. This work highlights digital ReRAM-based bit-serial PIM as a scalable and robust alternative to analog CIM for safety-critical edge-AI applications.
Short bit errors constitute a primary limitation to the endurance of Spin-Transfer Torque Magnetoresistive Random-Access Memory (STT-MRAM). To address this challenge, we propose a novel short-aware error correction method employing $Z$-error correction and ($Z+1$)-error detection error correction coding (ECC). This approach achieves enhanced error correction capability for ($Z+1$) errors. We validate the method through demonstration of triple-error correction using dual-parallel double-error correction and triple-error detection (DEC-TED) code blocks. Comparative analysis with conventional ECC reveals that the proposed method improves short-bit error rate correction capabilities by at least an order of magnitude. Furthermore, it achieves substantial hardware efficiency improvements, demonstrating typical 4–12% area reduction and 46% latency reduction compared to conventional ECC implementations with equivalent error correction capabilities.
A photonic crystal waveguide based on a perforated silicon slab has been optimized for the efficient TE resistive TM-notch filtering. The TE polarization suppression mechanism has been introduced in the proposed design to deal with the inevitable birefringence of silicon photonics. At least 20 dB suppression has been achieved in a 116 nm bandwidth including the TM notch wavelength. Here, with the proposed design mechanism both the features have been implemented simultaneously in the same device footprint. A detailed analysis has been presented of both the TE and TM polarization of light for the PhC slab waveguide using plan wave expansion method and all findings have been re-verified with a 3D-FDTD method. The maximum extinction ratio of the TE polarization is $\approx$ 30dB along with $\approx$ 19dB suppression of the TM notch wavelength. Moreover, the thermal sensitivity of 0.033nm$\textbf {/}^{o}K$ and a refractive index sensitivity of 310 nm/RIU of the design enables precise detection of temperature and refractive index changes.
In this study, an AlGaN/GaN High Electron Mobility Transistor (HEMT) based sensor with an Al$_{2}$O$_{3}$ functionalization layer is designed and optimized for the detection of Cd$^{2+}$ ions in water. To enhance ion sensing performance, step graded barrier and back barrier (epi-layer) HEMT design strategies are proposed. SILVACO ATLAS TCAD is used to optimize and evaluate the effectiveness of the proposed techniques. The device performance was evaluated based on its sensitivity to drain current and threshold voltage. Simulation results indicate that the proposed HEMT device achieves an improved drain current sensitivity of 47.33 mA/(fg/L) and a threshold voltage sensitivity of 5.03 V/(fg/L). Therefore, the application of the proposed step graded barrier and back barrier techniques in AlGaN/GaN HEMT is considered promising for Heavy Metal Ions (HMIs) sensors.
Pulsed DC magnetron-deposited α-In2Se3 thin films (∼100 nm) on TiO2-coated PVC substrates were investigated for electromagnetic interference (EMI) shielding in the 1.5-6.0 GHz range. The α-In2Se3 films, a layered van der Waals 2D material reported to exhibit ferroelectric-related polarization behavior in the literature, showed shielding effectiveness (SE) decreasing from ∼9.2 dB at 1.5 GHz to 7.6 dB at 6.0 GHz. Although lower than metals such as gold (∼33–25 dB), this SE is significant for ultrathin, lightweight films. α-In2Se3 absorbed ∼99 % of incident power, with reflection and transmission each <1 %, whereas gold reflected ∼76 %, absorbed 24 %, and transmitted <1 %. Electrical conductivity (∼10-3-10-6 S/cm) measured, while AC conductivity decreased with frequency, consistent with absorption-dominant shielding. TEM analysis indicated that the observed diffraction features are consistent with the previously reported phase of α In2Se3 structure, EDS showed an In:Se ratio of ∼2:3, and XPS binding energies agreed with literature. Thickness-dependent simulations based on experimentally derived parameters were used to analyze shielding effectiveness. Compared with other 2D semiconductors, α-In2Se3 is reported to exhibit polarization related properties; however, ferroelectric behavior was not measured. These results demonstrate suitability for lightweight EMI shielding in electronics and microwave devices.
This research presents a novel design of a CMOS active block termed as: Multiple Output-Differential Voltage Current Conveyor Buffered Transconductance Amplifier (MO DVCCBTA). The designed active block offers enhanced signal processing capabilities by involving multiple voltage and current output terminals, which makes the design suitable for analog circuit design. A detailed port characteristic of MO-DVCCBTA is verified by plotting the behaviour in PSPICE. In addition, the functionality of MO-DVCCBTA is validated by designing a fractional order inductor (FOI), by incorporating a fractional order capacitor (FOC), which is suitable for both grounded and floating operation. The floating FOI (FFOI) is extended for the modelling of the Human Eardrum Type-II model, which demonstrates its effectiveness in biomedical signal processing. The proposed MO-DVCCBTA design feasibility and performance, along with all application is simulated using TSMC 0.18 μm CMOS technology in PSPICE, OrCAD, and Cadence tools Suite using gpdk 180 nm technology.
Two-dimensional materials offer exceptional properties for precision devices, yet mechanical exfoliation remains limited by poor scalability and repeatability. This study systematically quantifies how peel angle and speed govern graphene yield and contamination using an automated exfoliation stage and high-throughput image analysis. Graphene flakes were produced across six substrate angles (0 to 120°) and four peel speeds (10 to 5000 μm/s), with three trials per condition, generating over 200,000 optical microscopy images. A segmentation model classified flake thickness and adhesive residue to construct flake-yield and residue maps over angle-speed space. Slower peels generally increased single-layer yield but also increased residue coverage, revealing an inherent tradeoff between quality and cleanliness. Notably, some high-angle/high-speed conditions (e.g., 120° at 5000 μm/s) achieved $\geq$80% of the best single-layer yield while keeping residue $\leq$30%. These results demonstrate that peel parameters can be tuned to balance yield and residue, providing practical guidance for optimizing mechanical exfoliation. More broadly, this work shows how automation and image-based classification enable scalable, statistically robust studies of 2D material synthesis.
