Ferroelectric materials with a negative capacitance (NC) effect, incorporated with a dielectric layer, hold great potential for low-power electronics. However, integrating an NC layer into ultrathin silicon-based electronics still faces critical challenges, including thickness scalability and fatigue reliability concerns. Here, we report two-dimensional (2D) negative capacitance field-effect transistors (NC-FETs) that utilize ferroelectric α-indium selenide (α-In2Se3) and hafnium oxide (HfO2) in the gate dielectric, with a molybdenum disulfide (MoS2) channel. Optimized oxide capacitance matching enables steep-slope switching with a minimum subthreshold swing (SS) of 28.8 mV decade-1 and an on/off ratio of ∼107, via internal ferroelectric voltage amplification. The experimentally measured NC effect and oxide passivation not only enhance device performance but also yield robust long-term stability. Furthermore, a depletion-load inverter built from these NC-FETs demonstrates clear logic functionality and ultralow power consumption, surpassing the Boltzmann limit for switching steepness and paving the way for high-performance and long endurance electronics.

Photodetectors that can achieve both high sensitivity and fast speed remain a challenge due to inherent trade-offs, particularly limiting in photonic integrated circuits. While substantial progress has been made in broadening spectral response, few strategies have directly addressed the simultaneous enhancement of sensitivity and speed. Here, we introduce a two-dimensional photodetector incorporating a functional interlayer, monolayer tungsten oxyselenide (TOS), formed between n-type PdPSe and p-type WSe2. TOS serves multifunctional roles to overcome the trade-offs: suppressing dark current as a hole barrier, enhancing responsivity via photogating, and preserving speed through trap-assisted tunneling and direct tunneling. Under spatially resolved illumination, the device shows a detectivity (D*) of 3.78 × 1015 Jones at 520 nm, with a bandwidth of 0.055 GHz at 785 nm. When integrated onto a silicon photonic platform, it achieves D* of 3.69 × 1011 Jones and a bandwidth of 0.11 GHz at 785 nm, outperforming most reported on-chip photodetectors. These results establish a route to sensitive, high-speed photodetectors for optical interconnect, communications, and sensing applications.

A robust edge termination structure is crucial in realizing very high-voltage IGBT required for the fast charging of next-generation electric vehicles. The conventional Variable Lateral Doping (VLD) termination suffers from severe breakdown voltage (BV) degradation caused by surface charges. To address this issue, not at the expense of increasing the area, we propose a compact field plate-floating ring hybrid termination structure and show that this structure is relatively insensitive to surface-charge induced BV degradation. The underlying mechanisms are elucidated via two-dimensional simulation study and its high-voltage robustness validated experimentally.

Vision systems in vertebrates have evolved energy-efficient and adaptable features in hyperpolarizing photoreceptors that machine vision struggles to mimic. Because semiconducting materials always exhibit a photoconductive effect, attempts to mimic hyperpolarizing photoreceptors have proven to be non-trivial. Sophisticated two-dimensional (2D) material based van der Waals heterostructures and other novel structures/materials pose fabrication and integration challenges. This work aims to address the issue by successfully harnessing the defect dynamics in a ubiquitous transition metal oxide (TMO) hafnia to present, for the first time, a photosensor with characteristics closely resembling those of hyperpolarizing photoreceptors, including on-the-fly adaptation to constant and changing illumination, all in just a single ultrathin (5 nm) layer. This work opens a new prospect for accelerating the development of biomimetic vision systems, given the integral role TMOs have already played in mainstream semiconductor technology.

Several artificial photo-synaptic devices have recently emerged to replicate photonic synaptic plasticity for neuromorphic computing. The integration of an artificial biological neuromorphic vision sensor and a photodetector into a single device poses a notable challenge. Here, for the first time, a novel photo-synaptic memristor based on monochalcogenide GeS was developed through defect engineering. The fabricated device exhibited rapid photo-response and persistent photoconductivity behavior with both pristine and defect-engineered GeS, owing to its distinct trapping state relaxations. First Principles based DFT calculations reveal that additional energy states (acting as photon traps) are present in GeSOx. GeS and GeSOx possess an indirect energy bandgap of 1.62 and 1.30 eV, respectively. Consequently, these devices could generate photocurrent upon light exposure, mimicking neuronal behavior. Moreover, they exhibit essential synaptic functionalities, such as STM, LTM, EPSC, PPF, and transition from short-term memory to long-term memory. Particularly, the device exhibits outstanding image memory and letter recognition optical wavelength sensitive responses, mimicking the biological retina. The simulated machine vision system with the GeS retina device as the processing core presents excellent accuracies of 96.75 % for MNIST and 85.43 % for fashion-MNIST datasets. Thanks to the photosensitivity of GeS, these devices can operate at low bias voltage of 0.1 V and consume only similar to 85 pJ of energy per usage. Furthermore, the logic function "AND" was incorporated into the optoelectronic simulation. The findings of this study present the way for the integration of sophisticated robotic vision systems and advancing neuromorphic computing capabilities.

Memristors are pivotal for energy-efficient artificial intelligence (AI) hardware, potentially eliminating the von Neumann bottleneck by in-memory realizations of synaptic operations. However, the dynamic requirements of neuromorphic computing on specific electronic devices pose reliability and universality challenges, limiting progress toward more widely applicable computing platforms. Here, a 2D high-kappa dielectric-based memristor with the desired reconfigurable resistive switching behavior is successfully demonstrated. Utilizing a few layered Bi2SeO5 possessing excellent electrical insulation properties as the switching medium, the device features a low operating voltage (approximate to 0.5 V), low operation current (10 pA), long memory retention (>103 s), large switching window (approximate to 108), steep slope (<1 mV dec-1), fast switching speed (40 ns), and low energy dissipation (approximate to 1 pJ). The switching characteristics between volatile and non-volatile memory can be achieved on demand by regulating compliance currents, offering the possibility of implementing multiple neural computational primitives. A simulated convolutional neural network (CNN) based on long-term potentiation/depression (LTP/D) achieves 85% accuracy in complex image recognition. Furthermore, MNIST and fashion-MNIST recognition with built reservoir computing (RC) utilizing volatile behaviors reach 97% and 85% accuracy, respectively. This work opens new opportunities for 2D high-kappa dielectrics in next-generation AI hardware with enhanced energy efficiency and computational versatility.

Emerging bio-inspired computing systems simulate the cognitive functions of the brain for the realization of future computing systems. For the development of such efficient neuromorphic electronics, the emulation of short-term and long-term synaptic plasticity behaviors of the biological synapses is an essential step. However, the electronic synaptic devices suffer from higher variability issues which hinder the application of such devices to build neuromorphic systems. For practical applications, it is essential to minimize the cycle-to-cycle and device-to-device variations in the synaptic functions of artificial electronic synapses. This study involves the fabrication of diffusive memristor devices using WTe2 chalcogenide as the main switching material. The choice of the switching material provides a facile solution to the variability problem. The greater uniformity in the switching characteristics of the WTe2 based memristor offers higher uniformity for the synaptic emulation. These devices exhibit both volatile and nonvolatile switching properties, allowing them to emulate both short-term and long-term synaptic functions. The WTe2 -based electronic synaptic devices present a high degree of uniformity for the emulation of various essential biological synaptic functions including short-term potentiation (STP), long-term potentiation (LTP), long-term depression (LTD), spike-rate-dependent plasticity (SRDP), and spike-timingdependent plasticity (STDP). A higher recognition accuracy of similar to 92 % is attained for pattern recognition using the modified National Institute of Standards and Technology (MNIST) handwritten digits, which is attributed to the enhanced linearity and higher uniformity of LTP/LTD characteristics. (c) 2024 Published by Elsevier Ltd on behalf of The editorial office of Journal of Materials Science & Technology.

Active control of the surface-enhanced Raman scattering enhancement performance was achieved by a WOx/MoOx hybrid-based SERS FET.

Significant advancements in artificial neural networks (ANNs) have driven the rapid progress of artificial intelligence and machine learning. While current feedforward neural networks primarily handle static data, recurrent neural networks (RNNs) are designed for dynamical systems. However, RNNs demand extensive training on specific tasks, limiting their scalability and affordability for edge computing. Physical reservoir computing (RC) offers an alternative approach by mapping inputs into high-dimensional states, allowing for pattern analysis within a fixed reservoir. Unlike RNNs, RC is well-suited for temporal and sequential data processing with rapid speed and low training costs. This makes RC suitable for hardware implementation across various research domains. Nonetheless, existing demonstrations of RC remain constrained to small-scale device arrays. As electronic synapse arrays aim to approach very large-scale and highly complex hardware as in the human brain, managing heat dissipation becomes a formidable challenge. In this work, we successfully developed the neuristors based on textured h-BN films, prepared using a CMOS-compatible technique, and constructed a physical RC system based on as-fabricated devices. Our approach leverages vertically aligned BN to provide aligned diffusion paths for the reproducible migration process of metal ions from the electrodes and offers a potential solution for thermal management in electronic devices. This achievement highlights the promising potential of our neuristors for future high-density and energy-efficient neuromorphic computing.

To fulfill the demands of three-dimensional (3D) cross-point array architectures in embedded memories, selectors with low voltage and low off current with high selectivity are imperative. Selector devices with reliable pulsed switching performance at a logic-compatible voltage are crucial for achieving the ultimate goal of embedded 3D universal cross-point memory. While resistive ReRAM systems offer high density and ultralow power consumption, the 3D cross-point array introduces challenges such as sneak path currents during read and write operations. To overcome these problems, the selectors must be integrated into each cross-point cell. Although selector devices have been extensively studied, no singular material system has currently fulfilled all the vital needs for integration with nonvolatile memory. Here, we demonstrate an Ag-gated GeS-based bidirectional threshold switching selector device (Ag/Ti/GeS/Ag) with a highly uniform cycle-to-cycle and device-to-device switching performance. The device demonstrates highly uniform bidirectional threshold switching with promising attributes, including low threshold voltages (similar to +/- 0.22 V), low off current (similar to pA), large selectivity window (similar to 3 x 108), an endurance of more than 109 cycles, and an on/off switching latency of 50 ns at 1 V pulsed operations. The simple material configuration of the selector device (Ag/Ti/GeS/Ag) facilitates easy fabrication, and its uniform switching performance makes it suitable for 3D cross-point integration with nonvolatile memory cells in a one-selector-one-resistor array configuration.

The emerging concept of a physical reservoir (PR) presents a viable solution for simplifying neural network architecture and training, and for addressing the limitations inherent in the von Neumann architecture. In this work, we propose an approach for accelerating the processing of event stream video (DVS128 dataset) by employing a silicon FET array as the PR. The inherent volatile memory of the FET array captures spatiotemporal correlations and compresses event sequences into analogue current values, thereby enabling a convolutional neural network to directly classify the different gestures. In a 10-class classification task, the hybrid system— combining the PR with an eight-layer ResNet—achieved a mean accuracy of 96.5% for 2-s event streams, following a 410-s training process. This research introduces a hardware-software cooperative system that facilitates near real-time computing, while also addressing the challenges of increasing architectural complexity and energy consumption.

In this work, forming-free and self-limited resistive switching characteristics are demonstrated in the transition-metal-chalcogenide-based conductive bridge RAM devices. Owing to the choice of a suitable solid electrolyte, the proposed WTe2-based devices present excellent switching characteristics offering highly desirable attributes such as high pulse endurance (> $2\times 10$ 7 cycles) and stable data retention (10 years at 72°C). The devices also present excellent device-to-device and cycle-to-cycle uniformity which is highly desirable for the practical implementation of resistive random access memory (RRAM) devices in large crossbar arrays. The demonstration of robust and highly repeatable multilevel switching further provides multibit data storage capability for high-density memory.

The significant boost in surface-enhanced Raman scattering (SERS) by the chemical enhancement of semiconducting oxides is a pivotal finding. It offers a prospective path toward high uniformity and low-cost SERS substrates. However, a detailed understanding of factors that influence the charge transfer process is still insufficient. Herein, we reveal the important role of defect-induced band offset and electron lifetime change in SERS evolution observed in a MoO3 oxide semiconductor. By modulating the density of oxygen vacancy defects using ultraviolet (UV) light irradiation, SERS is found to be improved with irradiation time in the first place, but such improvement later deteriorates for prolonged irradiation even if more defects are generated. Insights into the observed SERS evolution are provided by ultraviolet photoelectron spectroscopy and femtosecond time-resolved transient absorption spectroscopy measurements. Results reveal that (1) a suitable offset between the energy band of the substrate and the orbitals of molecules is facilitated by a certain defect density and (2) defect states with relatively long electron lifetime are essential to achieve optimal SERS performance.

An intermittent electrocatalytic mode for the oxygen evolution reaction (OER) inspired by the finding of room-temperature thermal reduction, and it reduces the OER overpotential significantly due to a reversible valence-change mechanism.

In this work, we consider eight silicon wet etching conditions using 20% or 25% Tetramethyl Ammonium Hydroxide (TMAH), each at 60°C,70°C,80°C or 90°C in order to study the TMAH kinetics and planar dependence on formation of upright pyramids by anisotropic etching on a (100) Si substrate. With the aid of negative square mask patterning, we are able to fabricate upright micro-pyramids having well defined facets with a smooth morphology. A new methodology termed as the 'Variable Time Index Method' is employed to understand the etch rate variation as a function of time and identify the etch rates along the (111) family of planes and consequently the planar dependence due to TMAH etch conditions. For each of the etching conditions, analysis using Scanning Electron Microscopy (SEM) is conducted to investigate the etched pyramid morphology and suitably identify the conditions under which deviation from the desired plane direction and surface are obtained. Ideal micro-pyramid morphology, vertex alignment and orientation are obtained for etch conditions of 20% and 25% TMAH, at 80°C.

Growth of data eases the way to access the world but requires increasing amounts of energy to store and process. Neuromorphic electronics has emerged in the last decade, inspired by biological neurons and synapses, with in-memory computing ability, extenuating the 'von Neumann bottleneck' between the memory and processor and offering a promising solution to reduce the efforts both in data storage and processing, thanks to their multi-bit non-volatility, biology-emulated characteristics, and silicon compatibility. This work reviews the recent advances in emerging memristive devices for artificial neuron and synapse applications, including memory and data-processing ability: the physics and characteristics are discussed first, i.e., valence changing, electrochemical metallization, phase changing, interfaced-controlling, charge-trapping, ferroelectric tunnelling, and spin-transfer torquing. Next, we propose a universal benchmark for the artificial synapse and neuron devices on spiking energy consumption, standby power consumption, and spike timing. Based on the benchmark, we address the challenges, suggest the guidelines for intra-device and inter-device design, and provide an outlook for the neuromorphic applications of resistive switching-based artificial neuron and synapse devices.

A double stacked monochalcogenide GeS-based conducting-bridge random access memory (CBRAM) device with a IGZO buffer layer is investigated for highly improved resistive memory characteristics. The IGZO/GeS double layer is found to provide the CBRAM with a markedly improved sub-1V DC set/reset-voltage distributions (<±0.1 V variation). High endurance (>107 cycles) and retention (>105 s at 85 °C) performance are also achieved. The metal ion diffusion and migration rates in the solid electrolytes along with the redox reaction rates at the electrodes determine the respective resistive switching (RS) mechanism in the CBRAM device. Considering this fact, it is proposed that Ag diffusion into IGZO creates a virtual electrode, when coupled with strong ionic transport in GeS, consistently mediate the formation/dissolution of Ag filament there, thus reducing switching variation. Understanding the RS mechanism based on the materials' physical and chemical properties and tailoring the device structure allow an optimal control over cycle to cycle and device to device variability. The findings show that this material combination or similar oxide/chalcogenide stacks may offer a facile means for mitigating CBRAM variability.