Search

We propose a novel synaptic design of more efficient neuromorphic edge-computing with substantially improved linearity and extremely low variability. Specifically, a parallel arrangement of ferroelectric tunnel junctions (FTJ) with an incremental pulsing scheme provides a great improvement in linearity for synaptic weight updating by averaging weight update rates of multiple devices. To enable such design with FTJ building blocks, we have demonstrated the lowest reported variability: σ / μ = 0.036 for cycle to cycle and σ / μ = 0.032 for device among six dies across an 8 inch wafer. With such devices, we further show improved synaptic performance and pattern recognition accuracy through experiments combined with simulations.

Owing to their switchable spontaneous polarization, ferroelectric materials have been applied in various fields, such as information technologies, actuators, and sensors. In the last decade, as the characteristic sizes of both devices and materials have decreased significantly below the nanoscale, the development of appropriate characterization tools became essential. Recently, a technique based on conductive atomic force microscopy (AFM), called AFMpositive- up-negative-down (PUND), is employed for the direct measurement of ferroelectric polarization under the AFM tip. However, the main limitation of AFM-PUND is the low frequency (i.e., on the order of a few hertz) that is used to initiate ferroelectric hysteresis. A significantly higher frequency is required to increase the signal-to-noise ratio and the measurement efficiency. In this study, a novel method based on high-frequency AFM-PUND using continuous waveform and simultaneous signal acquisition of the switching current is presented, in which polarization–voltage hysteresis loops are obtained on a highpolarization BiFeO3 nanocapacitor at frequencies up to 100 kHz. The proposed method is comprehensively evaluated by measuring nanoscale polarization values of the emerging ferroelectric Hf0.5Zr0.5O2 under the AFM tip. [ABSTRACT FROM AUTHOR]

The band structure of many semiconducting monolayer transition metal dichalcogenides (TMDs) possesses two degenerate valleys with equal and opposite Berry curvature. It has been predicted that, when illuminated with circularly polarized light, interband transitions generate an unbalanced nonequilibrium population of electrons and holes in these valleys, resulting in a finite Hall voltage at zero magnetic field when a current flows through the system. This is the so-called valley Hall effect that has recently been observed experimentally. Here, we show that this effect is mediated by photogenerated neutral excitons and charged trions and not by interband transitions generating independent electrons and holes. We further demonstrate an experimental strategy, based on wavelength dependent spatial mapping of the Hall voltage, which allows the exciton and trion contributions to the valley Hall effect to be discriminated in the measurement. These results represent a significant step forward in our understanding of the microscopic origin of photoinduced valley Hall effect in semiconducting transition metal dichalcogenides and demonstrate experimentally that composite quasi-particles, such as trions, can also possess a finite Berry curvature. [ABSTRACT FROM AUTHOR]

We present a novel method for fabricating large-area field-effect transistors (FETs) based on densely packed multichannel graphene nanoribbon (GNR) arrays using advanced direct self-assembly (DSA) nanolithography. The design of our strategy focused on the efficient integration of the FET channel and using fab-compatible processes such as thermal annealing and chemical vapor deposition. We achieved linearly stacked DSA nanopattern arrays with sub-10 nm half-pitch critical dimensions (CD) by controlling the thickness of topographic Au confinement patterns. Excellent roughness values (∼10% of CD) were obtained, demonstrating the feasibility of integrating sub-10 nm GNRs into commercial semiconductor processes. Based on this facile process, FETs with such densely packed multichannel GNR arrays were successfully fabricated on 6 in. silicon wafers. With these high-quality GNR arrays, we achieved FETs showing the highest performance reported to date (an on-to-off ratio larger than 10²) for similar devices produced using conventional photolithography and block-copolymer lithography. [ABSTRACT FROM AUTHOR]

The development of deep learning technology has resulted in great contributions in many artificial intelligence services, but adversarial attack techniques on deep learning models are also becoming more diverse and sophisticated. IoT edge devices take cloud-independent on-device DNN (deep neural network) processing technology to exhibit a fast response time. However, if the computational complexity of the denoizer for adversarial noises is high, or if a single embedded GPU is shared by multiple DNN models, adversarial defense at the on-device level is bound to represent a long latency. To solve this problem, eDenoizer is proposed in this paper. First, it applies Tucker decomposition to reduce the computational amount required for convolutional kernel tensors in the denoizer. Second, eDenoizer effectively orchestrates both the denoizer and the model defended by the denoizer simultaneously. In addition, the priority of the CPU side can be projected onto the GPU which is completely priority-agnostic, so that the delay can be minimized when the denoizer and the defense target model are assigned a high priority. As a result of confirming through extensive experiments, the reduction of classification accuracy was very marginal, up to 1.78%, and the inference speed accompanied by adversarial defense was improved up to 51.72%.

Obesity is associated with various human disorders, such as type 2 diabetes, cardiovascular diseases, hypertension, and cancers. In this study, we observed that knockout (KO) of CCN5, which encodes a matricellular protein, caused mild obesity in mice. The CCN5 KO mice also exhibited mild diabetes characterized by high fasting glucose levels and impaired insulin and glucose tolerances. Cardiac hypertrophy, ectopic lipid accumulation, and impaired lipid metabolism in hearts were observed in the CCN5 KO mice, as determined using histology, quantitative RT-PCR, and western blotting. Fibrosis was significantly greater in hearts from the CCN5 KO mice both in interstitial and perivascular regions, which was accompanied by higher expression of pro-fibrotic and pro-inflammatory genes. Both systolic and diastolic functions were significantly impaired in hearts from the CCN5 KO mice, as assessed using echocardiography. Taken together, these results indicate that CCN5 KO leads to lipotoxic cardiomyopathy with mild obesity and diabetes in mice.

Superconductor-graphene-superconductor (SGS) junction provides a unique platform to study relativistic electrodynamics of Dirac fermions in graphene combined with proximity-induced superconductivity. We report the observation of the Josephson effect in proximity-coupled superconducting junctions of graphene in contact with Pb1-xInx (x=0.07) electrodes for temperatures as high as T = 4.8 K, with a large value of IcRN (~255 μ V). This demonstrates that Pb1-xInx SGS junction would facilitate the development of the superconducting quantum information devices and superconductor-enhanced phase-coherent transport of graphene. [ABSTRACT FROM AUTHOR]