Learned Image Compression With Gaussian-Laplacian-Logistic Mixture Model and Concatenated Residual Modules.

Recently deep learning-based image compression methods have achieved significant achievements and gradually outperformed traditional approaches including the latest standard Versatile Video Coding (VVC) in both PSNR and MS-SSIM metrics. Two key components of learned image compression are the entropy model of the latent representations and the encoding/decoding network architectures. Various models have been proposed, such as autoregressive, softmax, logistic mixture, Gaussian mixture, and Laplacian. Existing schemes only use one of these models. However, due to the vast diversity of images, it is not optimal to use one model for all images, even different regions within one image. In this paper, we propose a more flexible discretized Gaussian-Laplacian-Logistic mixture model (GLLMM) for the latent representations, which can adapt to different contents in different images and different regions of one image more accurately and efficiently, given the same complexity. Besides, in the encoding/decoding network design part, we propose a concatenated residual blocks (CRB), where multiple residual blocks are serially connected with additional shortcut connections. The CRB can improve the learning ability of the network, which can further improve the compression performance. Experimental results using the Kodak, Tecnick-100 and Tecnick-40 datasets show that the proposed scheme outperforms all the leading learning-based methods and existing compression standards including VVC intra coding (4:4:4 and 4:2:0) in terms of the PSNR and MS-SSIM. The source code is available at https://github.com/fengyurenpingsheng.

Improving the linearity of synaptic plasticity of single-walled carbon nanotube field-effect transistors via CdSe quantum dots decoration.

The linearity of synaptic plasticity of single-walled carbon nanotube field-effect transistor (SWCNT FET) was improved by CdSe quantum dots decoration. The linearity of synaptic plasticity in SWCNT FET with decorating QDs was further improved by reducing the P-type doping level from the atmosphere. The synaptic behavior of SWCNT FET is found to be dominated by the charging and discharging processes of interface traps and surface traps, which are predominantly composed of H2O/O2redox couples. The improved synaptic behavior is mainly due to the reduction of the interface trap charging process after QDs decoration. The inherent correlation between the device synaptic behavior and the electron capture process of the traps are investigated through charging-based trap characterization. This study provides an effective scheme for improving linearity and designing new-type SWCNT synaptic devices.

A highly tunable photoelectric response of graphene field-effect transistor with lateral P-N junction in channel.

This paper reports a highly tunable photoelectric response of graphene field-effect transistor (GFET) with lateral P-N junction in channel. The poly(sulfobetaine methacrylate) (PSBMA) provides strong N-type doping on graphene due to the dipole moment of pendent groups after ultraviolet annealing in high vacuum. A lateral P-N junction is introduced into the channel of the GFET by partially covering the graphene channel with PSBMA. With such P-N junction in the channel, the GFET exhibits a highly tunable photoelectric response over a wide range of exciting photon wavelength. With a lateral P-N junction in the channel, the polarity of photocurrent (Iph) of the GFET switches three times as the back-gate voltage (VBG) scan over two Dirac-point voltages. The underlying physical mechanism of photoelectric response is attributed to photovoltaic and photo-induced bolometric effect, which compete to dominatingIphat variousVBG. This provides a possible strategy for designing new phototransistors or optoelectronic device in the future.

Analysis of Low-Frequency 1/f Noise Characteristics for MoTe2 Ambipolar Field-Effect Transistors.

Low-frequency electronic noise is an important parameter used for the electronic and sensing applications of transistors. Here, we performed a systematic study on the low-frequency noise mechanism for both p-channel and n-channel MoTe2 field-effect transistors (FET) at different temperatures, finding that low-frequency noise for both p-type and n-type conduction in MoTe2 devices come from the variable range hopping (VRH) transport process where carrier number fluctuations (CNF) occur. This process results in the broad distribution of the waiting time of the carriers between successive hops, causing the noise to increase as the temperature decreases. Moreover, we found the noise magnitude for p-type MoTe2 FET hardly changed after exposure to the ambient conditions, whereas for n-FET, the magnitude increased by nearly one order. These noise characteristics may provide useful guidelines for developing high-performance electronics based on the emerging transition metal dichalcogenides.

Study of the Structure, Electronic and Optical Properties of BiOI/Rutile-TiO2 Heterojunction by the First-Principle Calculation.

Using the first-principle calculation that is based on the density functional theory (DFT), our group gains some insights of the structural, electronic and optical properties of two brand new types of BiOI/TiO2 heterojunctions: 1I-terminated BiOI {001} surface/TiO2 (1I-BiOI/TiO2) and BiO-terminated BiOI {001} surface/TiO2 (BiO-BiOI/TiO2). The calculation illustrates that BiOI/TiO2 heterojunction has excellent mechanical stability, and it shows that there is a great possibility for the BiOI/TiO2 heterojunction to be used in visible-light range, hence the photocatalytic ability can be enhanced dramatically. Especially, from the calculation, we discovered that there are two specific properties: the band-gap of 1I-BiOI/TiO2 heterojunction reduces to 0.28 eV, and the BiO-BiOI/TiO2 semiconductor material changes to n-type. The calculated band offset (BOs) for 1I-BiOI/TiO2 heterojunction indicates that the interfacial structure contributes a lot to a suitable band alignment which can disperse the photo-generated carriers into the opposite sides of the interface, so this could effectively weaken the electron-hole recombination. Meanwhile, the built-in potential around the interface accelerates the movement of the photo-generated electron-hole pairs. We believe this is the reason that the BiOI/TiO2 material shows perfect photocatalytic performance. This paper can provide theoretical support for the related research, especially the further research of the BiOI-based material.

Study on the Thermal Conductivity Characteristics for Ultra-Thin Body FD SOI MOSFETs Based on Phonon Scattering Mechanisms.

The silicon-on-insulator (SOI) metal-oxide-semiconductor field-effect transistors (MOSFETs) suffer intensive self-heating effects due to the reduced thermal conductivity of the silicon layer while the feature sizes of devices scale down to the nanometer regime. In this work, analytical models of thermal conductivity considering the self-heating effect (SHE) in ultra-thin body fully depleted (UTB-FD) SOI MOSFETs are presented to investigate the influences of impurity, free and bound electrons, and boundary reflection effects on heat diffusion mechanisms. The thermal conductivities of thin silicon films with different parameters, including temperature, depth, thickness and doping concentration, are discussed in detail. The results show that the thermal dissipation associated with the impurity, the free and bound electrons, and especially the boundary reflection effects varying with position due to phonon scattering, greatly suppressed the heat loss ability of the nanoscale ultra-thin silicon film. The predictive power of the thermal conductivity model is enhanced for devices with sub-10-nm thickness and a heavily doped silicon layer while considering the boundary scattering contribution. The absence of the impurity, the electron or the boundary scattering leads to the unreliability in the model prediction with a small coefficient of determination.

Photoinduced Hysteresis of Graphene Field-Effect Transistors Due to Hydrogen-Complexed Defects in Silicon Dioxide.

Photoinduced hysteresis (PIH) of graphene field-effect transistors (G-FETs) has attracted attention because of its potential in developing photoelectronic or nonvolatile memory devices. In this work, we focused on the role of SiO2 dielectric layer on PIH, where G-FETs have only a SiO2 dielectric layer. Adsorbates are effectively removed before the PIH test. The effects of laser wavelength, laser power density, and temperature on the PIH are systematically investigated. The PIH is significantly enhanced by increasing the hydrogen flow in a hydrogen-atmosphere device thermal annealing. This strongly suggests proton-related defects that play a key role. The pure electronic process for PIH is further ruled out by the significant dependence of the doping rate on the temperature. A mechanism of PIH based on proton generation after hole trapping at [O3≡Si-H] is proposed. The proposed mechanism is well-supported by our experimental data: (1) the observed threshold photon energy for PIH is between 2.76 and 2.34 eV, which is close to the energy barrier for [O3≡Si-H], releasing a proton. (2) No obvious carrier mobility degradation after the PIH process suggests that the bulk defects in SiO2 are the major contributors rather than graphene/SiO2 interface defects. (3) The dependence of the doping rate on the temperature and the laser power density matches a theoretical model based on the random hopping of H+. The results in this work are also valuable for the study of degradation of other oxide dielectric materials in various field-effect transistors.

A Gas Sensing Channel Composited with Pristine and Oxygen Plasma-Treated Graphene.

Oxygen plasma treatment has been reported as an effective way of improving the response of graphene gas sensors. In this work, a gas sensor based on a composite graphene channel with a layer of pristine graphene (G) at the bottom and an oxygen plasma-treated graphene (OP-G) as a covering layer was reported. The OP-G on top provided oxygen functional groups and serves as the gas molecule grippers, while the as-grown graphene beneath serves as a fast carrier transport path. Thus, the composite channel (OP-G/G) demonstrated significantly improved response in NH3 gas sensing tests compared with the pristine G channel. Moreover, the OP-G/G channel showed faster response and recovering process than the OP-G channel. Since this kind of composite channel is fabricated from chemical vapor deposited graphene and patterned with standard photolithography, the device dimension was much smaller than a gas sensor fabricated from reduced graphene oxide and it is favorable for the integration of a large number of sensing units.