Showing papers by "City University of Hong Kong published in 2017"

Least Squares Generative Adversarial Networks

Abstract: Unsupervised learning with generative adversarial networks (GANs) has proven hugely successful. Regular GANs hypothesize the discriminator as a classifier with the sigmoid cross entropy loss function. However, we found that this loss function may lead to the vanishing gradients problem during the learning process. To overcome such a problem, we propose in this paper the Least Squares Generative Adversarial Networks (LSGANs) which adopt the least squares loss function for the discriminator. We show that minimizing the objective function of LSGAN yields minimizing the Pearson X2 divergence. There are two benefits of LSGANs over regular GANs. First, LSGANs are able to generate higher quality images than regular GANs. Second, LSGANs perform more stable during the learning process. We evaluate LSGANs on LSUN and CIFAR-10 datasets and the experimental results show that the images generated by LSGANs are of better quality than the ones generated by regular GANs. We also conduct two comparison experiments between LSGANs and regular GANs to illustrate the stability of LSGANs.

Lead Halide Perovskite Nanocrystals in the Research Spotlight: Stability and Defect Tolerance

Abstract: This Perspective outlines basic structural and optical properties of lead halide perovskite colloidal nanocrystals, highlighting differences and similarities between them and conventional II–VI and III–V semiconductor quantum dots. A detailed insight into two important issues inherent to lead halide perovskite nanocrystals then follows, namely, the advantages of defect tolerance and the necessity to improve their stability in environmental conditions. The defect tolerance of lead halide perovskites offers an impetus to search for similar attributes in other related heavy metal-free compounds. We discuss the origins of the significantly blue-shifted emission from CsPbBr3 nanocrystals and the synthetic strategies toward fabrication of stable perovskite nanocrystal materials with emission in the red and infrared parts of the optical spectrum, which are related to fabrication of mixed cation compounds guided by Goldschmidt tolerance factor considerations. We conclude with the view on perspectives of use of th...

On-chip generation of high-dimensional entangled quantum states and their coherent control

Abstract: Optical quantum states based on entangled photons are essential for solving questions in fundamental physics and are at the heart of quantum information science. Specifically, the realization of high-dimensional states (D-level quantum systems, that is, qudits, with D > 2) and their control are necessary for fundamental investigations of quantum mechanics, for increasing the sensitivity of quantum imaging schemes, for improving the robustness and key rate of quantum communication protocols, for enabling a richer variety of quantum simulations, and for achieving more efficient and error-tolerant quantum computation. Integrated photonics has recently become a leading platform for the compact, cost-efficient, and stable generation and processing of non-classical optical states. However, so far, integrated entangled quantum sources have been limited to qubits (D = 2). Here we demonstrate on-chip generation of entangled qudit states, where the photons are created in a coherent superposition of multiple high-purity frequency modes. In particular, we confirm the realization of a quantum system with at least one hundred dimensions, formed by two entangled qudits with D = 10. Furthermore, using state-of-the-art, yet off-the-shelf telecommunications components, we introduce a coherent manipulation platform with which to control frequency-entangled states, capable of performing deterministic high-dimensional gate operations. We validate this platform by measuring Bell inequality violations and performing quantum state tomography. Our work enables the generation and processing of high-dimensional quantum states in a single spatial mode.

High dislocation density–induced large ductility in deformed and partitioned steels

Abstract: A wide variety of industrial applications require materials with high strength and ductility. Unfortunately, the strategies for increasing material strength, such as processing to create line defects (dislocations), tend to decrease ductility. We developed a strategy to circumvent this in inexpensive, medium manganese steel. Cold rolling followed by low-temperature tempering developed steel with metastable austenite grains embedded in a highly dislocated martensite matrix. This deformed and partitioned (D and P) process produced dislocation hardening but retained high ductility, both through the glide of intensive mobile dislocations and by allowing us to control martensitic transformation. The D and P strategy should apply to any other alloy with deformation-induced martensitic transformation and provides a pathway for the development of high-strength, high-ductility materials.

Photoluminescent Ti3C2 MXene Quantum Dots for Multicolor Cellular Imaging

Abstract: The fabrication of photoluminescent Ti3 C2 MXene quantum dots (MQDs) by a facile hydrothermal method is reported, which may greatly extend the applications of MXene-based materials. Interestingly, the as-prepared MQDs show excitation-dependent photoluminescence spectra with quantum yields of up to ≈10% due to strong quantum confinement. The applications of MQDs as biocompatible multicolor cellular imaging probes and zinc ion sensors are demonstrated.

Virtual-to-real deep reinforcement learning: Continuous control of mobile robots for mapless navigation

Abstract: We present a learning-based mapless motion planner by taking the sparse 10-dimensional range findings and the target position with respect to the mobile robot coordinate frame as input and the continuous steering commands as output. Traditional motion planners for mobile ground robots with a laser range sensor mostly depend on the obstacle map of the navigation environment where both the highly precise laser sensor and the obstacle map building work of the environment are indispensable. We show that, through an asynchronous deep reinforcement learning method, a mapless motion planner can be trained end-to-end without any manually designed features and prior demonstrations. The trained planner can be directly applied in unseen virtual and real environments. The experiments show that the proposed mapless motion planner can navigate the nonholonomic mobile robot to the desired targets without colliding with any obstacles.

Photo-Inspired Antibacterial Activity and Wound Healing Acceleration by Hydrogel Embedded with Ag/Ag@AgCl/ZnO Nanostructures.

Abstract: Ag/Ag@AgCl/ZnO hybrid nanostructures are embedded in a hydrogel by a simple two-step technique. The Ag/Ag@AgCl nanostructures are assembled in the hydrogel via ultraviolet light chemical reduction followed by incorporation of ZnO nanostructures by NaOH precipitation. The hydrogel accelerates wound healing and exhibits high antibacterial efficiency against both Escherichia coli and Staphylococcus aureus under visible light irradiation. The Ag/Ag@AgCl nanostructures enhance the photocatalytic and antibacterial activity of ZnO due to the enhancement of reactive oxygen species by visible light. This hydrogel system kills 95.95% of E. coli and 98.49% of S. aureus within 20 min upon exposure to simulated visible light, and rapid sterilization plays a crucial role in wound healing. In addition, this system provides controllable, sustained release of silver and zinc ions over a period of 21 days arising from the reversible swelling–shrinking transition of the hydrogel triggered by the changing pH value in the bio...

Skin-Inspired Multifunctional Autonomic-Intrinsic Conductive Self-Healing Hydrogels with Pressure Sensitivity, Stretchability, and 3D Printability.

Abstract: The advent of conductive self-healing (CSH) hydrogels, a class of novel materials mimicking human skin, may change the trajectory of the industrial process because of their potential applications in soft robots, biomimetic prostheses, and health-monitoring systems. Here, the development of a mechanically and electrically self-healing hydrogel based on physically and chemically cross-linked networks is reported. The autonomous intrinsic self-healing of the hydrogel is attained through dynamic ionic interactions between carboxylic groups of poly(acrylic acid) and ferric ions. A covalent cross-linking is used to support the mechanical structure of the hydrogel. Establishing a fair balance between the chemical and physical cross-linking networks together with the conductive nanostructure of polypyrrole networks leads to a double network hydrogel with bulk conductivity, mechanical and electrical self-healing properties (100% mechanical recovery in 2 min), ultrastretchability (1500%), and pressure sensitivity. The practical potential of CSH hydrogels is further revealed by their application in human motion detection and their 3D-printing performance.

Graphitic Nitrogen Triggers Red Fluorescence in Carbon Dots.

Abstract: Carbon dots (CDs) are a stable and highly biocompatible fluorescent material offering great application potential in cell labeling, optical imaging, LED diodes, and optoelectronic technologies. Because their emission wavelengths provide the best tissue penetration, red-emitting CDs are of particular interest for applications in biomedical technologies. Current synthetic strategies enabling red-shifted emission include increasing the CD particle size (sp2 domain) by a proper synthetic strategy and tuning the surface chemistry of CDs with suitable functional groups (e.g., carboxyl). Here we present an elegant route for preparing full-color CDs with well-controllable fluorescence at blue, green, yellow, or red wavelengths. The two-step procedure involves the synthesis of a full-color-emitting mixture of CDs from citric acid and urea in formamide followed by separation of the individual fluorescent fractions by column chromatography based on differences in CD charge. Red-emitting CDs, which had the most negat...

Molecular Fluorescence in Citric Acid-Based Carbon Dots

Abstract: Nitrogen-doped carbon dots synthesized from citric acid as a carbon precursor have recently been considered to contain fluorescent derivatives of citrazinic acid, which contribute to their emission in the blue spectral range. To study the impact of such molecular fluorescent species on the optical properties of carbon dots, we synthesized three samples employing citric acid and three different nitrogen sources: ethylenediamine, hexamethylenetetramine, and triethanolamine. On the basis of the analysis of the nitrogen content and its coordination by X-ray photoelectron spectroscopy, FTIR spectra, and systematically comparing absorption, steady-state emission, and photoluminescence decays of each kind of carbon dot, we derive the influence of the molecular precursors and gain further understanding of the complex structure of carbon dots highlighting the strong impact of molecular fluorescence in the samples produced with ethylenediamine and hexamethylenetetramine.

CREST: Convolutional Residual Learning for Visual Tracking

Abstract: Discriminative correlation filters (DCFs) have been shown to perform superiorly in visual tracking. They only need a small set of training samples from the initial frame to generate an appearance model. However, existing DCFs learn the filters separately from feature extraction, and update these filters using a moving average operation with an empirical weight. These DCF trackers hardly benefit from the end-to-end training. In this paper, we propose the CREST algorithm to reformulate DCFs as a one-layer convolutional neural network. Our method integrates feature extraction, response map generation as well as model update into the neural networks for an end-to-end training. To reduce model degradation during online update, we apply residual learning to take appearance changes into account. Extensive experiments on the benchmark datasets demonstrate that our CREST tracker performs favorably against state-of-the-art trackers.

Texturing in situ: N,S-enriched hierarchically porous carbon as a highly active reversible oxygen electrocatalyst

Abstract: Facile yet rational design of an efficient reversible oxygen electrocatalyst is critical for many renewable energy conversion and storage technologies. Here we report a simple and general synthetic protocol for fabricating a hierarchically porous and heteroatom doped carbon catalyst, which exhibited outstanding oxygen reduction/evolution activities (with a metric potential difference of 0.72 V in 1 M KOH, the best value for metal-free catalysts reported to date) with good stability in different electrolytes. The excellent performances of the catalyst were primarily endowed by our synthetic protocol, which integrates good conductivity, abundant accessible dopant species and suitable porous architectures within an in situ pyrolysis reaction. As a result, the performances of rechargeable Zn–air batteries based on the optimized catalyst substantially outperform those afforded by a benchmark Pt/C catalyzer. Our work is expected to open up new avenues for developing other efficient catalysts in a facile and viable way.

Metal-Ion-Modified Black Phosphorus with Enhanced Stability and Transistor Performance.

Abstract: Black phosphorus (BP), a burgeoning elemental 2D semiconductor, has aroused increasing scientific and technological interest, especially as a channel material in field-effect transistors (FETs). However, the intrinsic instability of BP causes practical concern and the transistor performance must also be improved. Here, the use of metal-ion modification to enhance both the stability and transistor performance of BP sheets is described. Ag+ spontaneously adsorbed on the BP surface via cation-π interactions passivates the lone-pair electrons of P thereby rendering BP more stable in air. Consequently, the Ag+ -modified BP FET shows greatly enhanced hole mobility from 796 to 1666 cm2 V-1 s-1 and ON/OFF ratio from 5.9 × 104 to 2.6 × 106 . The mechanisms pertaining to the enhanced stability and transistor performance are discussed and the strategy can be extended to other metal ions such as Fe3+ , Mg2+ , and Hg2+ . Such stable and high-performance BP transistors are crucial to electronic and optoelectronic devices. The stability and semiconducting properties of BP sheets can be enhanced tremendously by this novel strategy.

Dual-phase nanostructuring as a route to high-strength magnesium alloys

Abstract: Combining the benefits of nanocrystals with those of amorphous metallic glasses leads to a dual-phase material—comprising sub-10-nanometre-sized nanocrystalline grains embedded in amorphous glassy shells—that exhibits a strength approaching the ideal theoretical limit. Nanostructuring of crystalline metal alloys can yield high-strength materials, but these tend to soften as the strain is increased. Ge Wu et al. describe a strategy that combines the benefits of nanocrystallinity with those of single-phase amorphous metallic glasses to yield a dual-phase material—nanocrystalline grains each enclosed in an amorphous glassy shell—that exhibits strength approaching the ideal theoretical limit. They demonstrate this approach with a magnesium alloy and prepare the strongest thin films yet achieved for any magnesium alloy. The authors suggest that this material could be a promising coating for wear-resistant surfaces. It is not easy to fabricate materials that exhibit their theoretical ‘ideal’ strength. Most methods of producing stronger materials are based on controlling defects to impede the motion of dislocations, but such methods have their limitations. For example, industrial single-phase nanocrystalline alloys1,2 and single-phase metallic glasses3 can be very strong, but they typically soften at relatively low strains (less than two per cent) because of, respectively, the reverse Hall–Petch effect4 and shear-band formation. Here we describe an approach that combines the strengthening benefits of nanocrystallinity with those of amorphization to produce a dual-phase material that exhibits near-ideal strength at room temperature and without sample size effects. Our magnesium-alloy system consists of nanocrystalline cores embedded in amorphous glassy shells, and the strength of the resulting dual-phase material is a near-ideal 3.3 gigapascals—making this the strongest magnesium-alloy thin film yet achieved. We propose a mechanism, supported by constitutive modelling, in which the crystalline phase (consisting of almost-dislocation-free grains of around six nanometres in diameter) blocks the propagation of localized shear bands when under strain; moreover, within any shear bands that do appear, embedded crystalline grains divide and rotate, contributing to hardening and countering the softening effect of the shear band.

An Overview of the Development of Flexible Sensors.

Abstract: Flexible sensors that efficiently detect various stimuli relevant to specific environmental or biological species have been extensively studied due to their great potential for the Internet of Things and wearable electronics applications. The application of flexible and stretchable electronics to device-engineering technologies has enabled the fabrication of slender, lightweight, stretchable, and foldable sensors. Here, recent studies on flexible sensors for biological analytes, ions, light, and pH are outlined. In addition, contemporary studies on device structure, materials, and fabrication methods for flexible sensors are discussed, and a market overview is provided. The conclusion presents challenges and perspectives in this field.

An Intrinsically Stretchable and Compressible Supercapacitor Containing a Polyacrylamide Hydrogel Electrolyte

Abstract: Stretchability and compressibility of supercapacitors is an essential element of modern electronics, such as flexible, wearable devices. Widely used polyvinyl alcohol-based electrolytes are neither very stretchable nor compressible, which fundamentally limits the realization of supercapacitors with high stretchability and compressibility. A new electrolyte that is intrinsically super-stretchable and compressible is presented. Vinyl hybrid silica nanoparticle cross-linkers were introduced into polyacrylamide hydrogel backbones to promote dynamic cross-linking of the polymer networks. These cross-linkers serve as stress buffers to dissipate energy when strain is applied, providing a solution to the intrinsically low stretchability and compressibility shortcomings of conventional supercapacitors. The newly developed supercapacitor and electrolyte can be stretched up to an unprecedented 1000 % strain with enhanced performance, and compressed to 50 % strain with good retention of the initial performance.

Nickel–Cobalt Diselenide 3D Mesoporous Nanosheet Networks Supported on Ni Foam: An All-pH Highly Efficient Integrated Electrocatalyst for Hydrogen Evolution

Abstract: Novel 3D Ni1−xCoxSe2 mesoporous nanosheet networks with tunable stoichiometry are successfully synthesized on Ni foam (Ni1−xCoxSe2 MNSN/NF with x ranging from 0 to 0.35). The collective effects of special morphological design and electronic structure engineering enable the integrated electrocatalyst to have very high activity for hydrogen evolution reaction (HER) and excellent stability in a wide pH range. Ni0.89Co0.11Se2 MNSN/NF is revealed to exhibit an overpotential (η10) of 85 mV at −10 mA cm−2 in alkaline medium (pH 14) and η10 of 52 mV in acidic solution (pH 0), which are the best among all selenide-based electrocatalysts reported thus far. In particular, it is shown for the first time that the catalyst can work efficiently in neutral solution (pH 7) with a record η10 of 82 mV for all noble metal-free electrocatalysts ever reported. Based on theoretical calculations, it is further verified that the advanced all-pH HER activity of Ni0.89Co0.11Se2 is originated from the enhanced adsorption of both H+ and H2O induced by the substitutional doping of cobalt at an optimal level. It is believed that the present work provides a valuable route for the design and synthesis of inexpensive and efficient all-pH HER electrocatalysts.

Ultralight Graphene Foam/Conductive Polymer Composites for Exceptional Electromagnetic Interference Shielding

Abstract: Ultralight, high-performance electromagnetic interference (EMI) shielding graphene foam (GF)/poly(3,4-ethylenedioxythiophene):poly(styrenesulfonate) (PEDOT:PSS) composites are developed by drop coating of PEDOT:PSS on cellular-structured, freestanding GFs. To enhance the wettability and the interfacial bonds with PEDOT:PSS, GFs are functionalized with 4-dodecylbenzenesulfonic acid. The GF/PEDOT:PSS composites possess an ultralow density of 18.2 × 10–3 g/cm3 and a high porosity of 98.8%, as well as an enhanced electrical conductivity by almost 4 folds from 11.8 to 43.2 S/cm after the incorporation of the conductive PEDOT:PSS. Benefiting from the excellent electrical conductivity, ultralight porous structure, and effective charge delocalization, the composites deliver remarkable EMI shielding performance with a shielding effectiveness (SE) of 91.9 dB and a specific SE (SSE) of 3124 dB·cm3/g, both of which are the highest among those reported in the literature for carbon-based polymer composites. The excelle...

Enhanced selectivity in mixed matrix membranes for CO2 capture through efficient dispersion of amine-functionalized MOF nanoparticles

Abstract: Mixed matrix membranes (MMMs) for gas separation applications have enhanced selectivity when compared with the pure polymer matrix, but are commonly reported with low intrinsic permeability, which has major cost implications for implementation of membrane technologies in large-scale carbon capture projects. High-permeability polymers rarely generate sufficient selectivity for energy-efficient CO2 capture. Here we report substantial selectivity enhancements within high-permeability polymers as a result of the efficient dispersion of amine-functionalized, nanosized metal–organic framework (MOF) additives. The enhancement effects under optimal mixing conditions occur with minimal loss in overall permeability. Nanosizing of the MOF enhances its dispersion within the polymer matrix to minimize non-selective microvoid formation around the particles. Amination of such MOFs increases their interaction with thepolymer matrix, resulting in a measured rigidification and enhanced selectivity of the overall composite. The optimal MOF MMM performance was verified in three different polymer systems, and also over pressure and temperature ranges suitable for carbon capture. Mixed matrix membranes can separate CO2 from flue gas mixtures but increasing selectivity without sacrificing permeability remains challenging. Selectivity can be increased with little loss in permeability by using nanoparticulate, amine-functionalized metal–organic framework fillers.

Occurrence and Characteristics of Microplastic Pollution in Xiangxi Bay of Three Gorges Reservoir, China

Abstract: Microplastic pollution in inland waters is receiving growing attentions. Reservoirs are suspected to be particularly vulnerable to microplastic pollution. However, very limited information is currently available on pollution characteristics of microplastics in reservoir ecosystems. This work studied the distribution and characteristics of microplastics in the backwater area of Xiangxi River, a typical tributary of the Three Gorges Reservoir. Microplastics were detected in both surface water and sediment with concentrations ranging from 0.55 × 105 to 342 × 105 items km–2 and 80 to 864 items m–2, respectively. Polyethylene, polypropylene, and polystyrene were identified in surface water, whereas polyethylene, polypropylene, and polyethylene terephthalate, and pigments were observed in sediment. In addition, microplastics were also detected in the digestion tracts of 25.7% of fish samples, and polyethylene and nylon were identified. Redundancy analysis indicates a weak correlation between microplastics and w...

Observer-Based Event-Triggering Consensus Control for Multiagent Systems With Lossy Sensors and Cyber-Attacks

Abstract: In this paper, the observer-based event-triggering consensus control problem is investigated for a class of discrete-time multiagent systems with lossy sensors and cyber-attacks. A novel distributed observer is proposed to estimate the relative full states and the estimated states are then used in the feedback protocol in order to achieve the overall consensus. An event-triggered mechanism with state-independent threshold is adopted to update the control input signals so as to reduce unnecessary data communications. The success ratio of the launched attacks is taken into account to reflect the probabilistic failures of the attacks passing through the protection devices subject to limited resources and network fluctuations. The purpose of the address problem is to design an observer-based distributed controller such that the closed-loop multiagent system achieves the prescribed consensus in spite of the lossy sensors and cyber-attacks. By making use of eigenvalues and eigenvectors of the Laplacian matrix, the closed-loop system is transformed into an easy-to-analyze setting and then a sufficient condition is derived to guarantee the desired consensus. Furthermore, the controller gain is obtained in terms of the solution to certain matrix inequality which is independent of the number of agents. An algorithm is provided to optimize the consensus bound. Finally, a simulation example is utilized to illustrate the usefulness of the proposed controller design scheme.

CREST: Convolutional Residual Learning for Visual Tracking

Abstract: Discriminative correlation filters (DCFs) have been shown to perform superiorly in visual tracking. They only need a small set of training samples from the initial frame to generate an appearance model. However, existing DCFs learn the filters separately from feature extraction, and update these filters using a moving average operation with an empirical weight. These DCF trackers hardly benefit from the end-to-end training. In this paper, we propose the CREST algorithm to reformulate DCFs as a one-layer convolutional neural network. Our method integrates feature extraction, response map generation as well as model update into the neural networks for an end-to-end training. To reduce model degradation during online update, we apply residual learning to take appearance changes into account. Extensive experiments on the benchmark datasets demonstrate that our CREST tracker performs favorably against state-of-the-art trackers.

Recent Progress on Flexible and Wearable Supercapacitors

Abstract: Recently, wearable electronic devices including electrical sensors, flexible displays, and health monitors have received considerable attention and experienced rapid progress. Wearable supercapacitors attract tremendous attention mainly due to their high stability, low cost, fast charging/discharging, and high efficiency; properties that render them value for developing fully flexible devices. In this Concept, the recent achievements and advances made in flexible and wearable supercapacitors are presented, especially highlighting the promising performances of yarn/fiber-shaped and planar supercapacitors. On the basis of their working mechanism, electrode materials including carbon-based materials, metal oxide-based materials, and conductive polymers with an emphasis on the performance-optimization method are introduced. The latest representative techniques and active materials of recently developed supercapacitors with superior performance are summarized. Furthermore, the designs of 1D and 2D electrodes are discussed according to their electrically conductive supporting materials. Finally, conclusions, challenges, and perspective in optimizing and developing the electrochemical performance and function of wearable supercapacitors for their practical utility are addressed.