Compared to traditional bulk materials, two-dimensional transition metal dichalcogenides (2D TMDCs) hold the potential in low power logic, photoelectric, and nonvolatile memory devices due to a tunable band structure, pure heterojunction interface, and photodetection for a wide spectral range. ReS 2 is chosen as the channel material in our work because it possesses excellent photoresponsivity. In addition, a BN dielectric is inserted between SiO 2 and ReS 2 as a gate dielectric. The pure heterojunction interface at BN/ReS 2 enhances the electric characteristics, including negligible hysteresis window and higher mobility. The photoelectric measurement results show that ReS 2 devices with or without the BN dielectric have an outstanding photoresponsivity of up to ∼10 6 A/W and a specific detectivity of up to ∼10 13 Jones as well as a fast photoresponse time of less than 100 ms under an extremely low optical power density of 0.47 μW/cm 2 , which fully proves that ReS 2 has the ability to detect very weak light. Also, the contact resistance and Schottky barrier height are also extracted with variable temperature measurements for the ReS 2 device. The higher contact resistance and unsymmetric output currents illustrate that the mirror force is the main factor leading to the reduction of the Schottky barrier height. Finally, the influences of the ReS 2 channel thickness on mobility and photoresponsivity are also studied by the device-to-device variability. The work in this paper further demonstrates that nanosheet-based ReS 2 can be regarded as an excellent candidate for application into the field of light-sensitive sensors under the condition of the CMOS-compatible process.

With the development of semiconductor technology, the size of traditional metal oxide semiconductor field effect transistor devices continues to decrease, but it cannot meet the requirements of high performance and low power consumption. Low power tunneling field effect transistor (TFET) has gradually become the focus of researchers. This paper proposes a novel T-shaped gate TFET based on the silicon with the negative capacitance (NC-TGTFET). On the basis of TGTFET, ferroelectric material (HZO) is used as gate dielectric. The simulation results show that, compared with the traditional TGTFET, the opening order and sensitivity of the two tunneling junctions are different. The influences of thickness and the doping concentration of pocket and ferroelectric material properties on the characteristics of NC-TGTFET is also discussed by Sentaurus simulation tool. Furthermore, the negative capacitance of ferroelectric material makes NC-TGTFET have a very steep subthreshold swing (18.32 mV/dec) at the range of drain current from 1×10 −15 to 1×10 −7 A μm −1 . And the on-state current (V g = 0.5 V, V d = 0.5 V) is 1.52×10 −6 A μm −1 .

No abstract

No abstract

While mislabeled or ambiguously-labeled samples in the training set could negatively affect the performance of deep models, diagnosing the dataset and identifying mislabeled samples helps to improve the generalization power. Training dynamics, i.e., the traces left by iterations of optimization algorithms, have recently been proved to be effective to localize mislabeled samples with hand-crafted features. In this paper, beyond manually designed features, we introduce a novel learning-based solution, leveraging a noise detector, instanced by an LSTM network, which learns to predict whether a sample was mislabeled using the raw training dynamics as input. Specifically, the proposed method trains the noise detector in a supervised manner using the dataset with synthesized label noises and can adapt to various datasets (either naturally or synthesized label-noised) without retraining. We conduct extensive experiments to evaluate the proposed method. We train the noise detector based on the synthesized label-noised CIFAR dataset and test such noise detector on Tiny ImageNet, CUB-200, Caltech-256, WebVision and Clothing1M. Results show that the proposed method precisely detects mislabeled samples on various datasets without further adaptation, and outperforms state-of-the-art methods. Besides, more experiments demonstrate that the mislabel identification can guide a label correction, namely data debugging, providing orthogonal improvements of algorithm-centric state-of-the-art techniques from the data aspect.