Showing papers by "South China University of Technology published in 2020"

Aggregation-Induced Emission: New Vistas at the Aggregate Level

Abstract: Aggregation-induced emission (AIE) describes a photophysical phenomenon in which molecular aggregates exhibit stronger emission than the single molecules. Over the course of the last 20 years, AIE research has made great strides in material development, mechanistic study and high-tech applications. The achievements of AIE research demonstrate that molecular aggregates show many properties and functions that are absent in molecular species. In this review, we summarize the advances in the field of AIE and its related areas. We specifically focus on the new properties of materials attained by molecular aggregates beyond the microscopic molecular level. We hope this review will inspire more research into molecular ensembles at and beyond the meso level and lead to the significant progress in material and biological science.

A survey on ensemble learning

Abstract: Despite significant successes achieved in knowledge discovery, traditional machine learning methods may fail to obtain satisfactory performances when dealing with complex data, such as imbalanced, high-dimensional, noisy data, etc. The reason behind is that it is difficult for these methods to capture multiple characteristics and underlying structure of data. In this context, it becomes an important topic in the data mining field that how to effectively construct an efficient knowledge discovery and mining model. Ensemble learning, as one research hot spot, aims to integrate data fusion, data modeling, and data mining into a unified framework. Specifically, ensemble learning firstly extracts a set of features with a variety of transformations. Based on these learned features, multiple learning algorithms are utilized to produce weak predictive results. Finally, ensemble learning fuses the informative knowledge from the above results obtained to achieve knowledge discovery and better predictive performance via voting schemes in an adaptive way. In this paper, we review the research progress of the mainstream approaches of ensemble learning and classify them based on different characteristics. In addition, we present challenges and possible research directions for each mainstream approach of ensemble learning, and we also give an extra introduction for the combination of ensemble learning with other machine learning hot spots such as deep learning, reinforcement learning, etc.

High-efficiency organic solar cells with low non-radiative recombination loss and low energetic disorder

Abstract: Energy loss within organic solar cells (OSCs) is undesirable as it reduces cell efficiency1–4. In particular, non-radiative recombination loss3 and energetic disorder5, which are closely related to the tail states below the band edge and the overall photon energy loss, need to be minimized to improve cell performance. Here, we report how the use of a small-molecule acceptor with torsion-free molecular conformation can achieve a very low degree of energetic disorder and mitigate energy loss in OSCs. The resulting single-junction OSC has an energy loss due to non-radiative recombination of just 0.17 eV and a high power conversion efficiency of up to 16.54% (certified as 15.89% by the National Renewable Energy Laboratory). The findings take studies of organic photovoltaics deeper into a new regime, beyond the limits of energetic disorder and large energy offset for charge generation. An organic solar cell designed with minimal energetic disorder exhibits very low energy loss due to non-radiative recombination and highly efficient operation.

Electrochemical reduction of nitrate to ammonia via direct eight-electron transfer using a copper–molecular solid catalyst

Abstract: Ammonia (NH3) is essential for modern agriculture and industry and is a potential energy carrier. NH3 is traditionally synthesized by the Haber–Bosch process at high temperature and pressure. The high-energy input of this process has motivated research into electrochemical NH3 synthesis via nitrogen (N2)–water reactions under ambient conditions. However, the future of this low-cost process is compromised by the low yield rate and poor selectivity, ascribed to the inert N≡N bond and ultralow solubility of N2. Obtaining NH3 directly from non-N2 sources could circumvent these challenges. Here we report the eight-electron direct electroreduction of nitrate to NH3 catalysed by copper-incorporated crystalline 3,4,9,10-perylenetetracarboxylic dianhydride. The catalyst exhibits an NH3 production rate of 436 ± 85 μg h−1 cm−2 and a maximum Faradaic efficiency of 85.9% at −0.4 V versus a reversible hydrogen electrode. This notable performance is achieved by the catalyst regulating the transfer of protons and/or electrons to the copper centres and suppressing hydrogen production. Electrochemically reducing nitrogen-containing molecules could provide less energy-intense routes to produce ammonia than the traditional Haber–Bosh process. Here the authors use a catalyst comprising Cu embedded in an organic molecular solid to synthesize ammonia from nitrate ions.

Effective ion sieving with Ti 3 C 2 T x MXene membranes for production of drinking water from seawater

Abstract: Traditional ways of producing drinking water from groundwater, water recycling and water conservation are not sufficient. Seawater desalination would close the gap but the main technology used is thermally driven multi-flash distillation, which is energy consuming and not sustainable. Stacking two-dimensional (2D) nanomaterials into lamellar membranes is a promising technique in the pursuit of both high selectivity and permeance in water desalination. However, 2D membranes tend to swell in water, and increasing their stability in aqueous solution is still challenging. Here, we report non-swelling, MXene membranes prepared by the intercalation of Al3+ ions. Swelling is prevented by strong interactions between Al3+ and oxygen functional groups terminating at the MXene surface. These membranes show excellent non-swelling stability in aqueous solutions up to 400 h and possess high rejection of NaCl (~89.5–99.6%) with fast water fluxes (~1.1–8.5 l m−2 h−1). Such membranes can be easily fabricated by simple filtration and ion-intercalating methods, which holds promise for their scalability. Two-dimensional lamellar membranes for water purification are promising but suffer from swelling that reduces their ion sieving performance in water. This study reports easy-to-fabricate, non-swelling MXene membranes prepared by the intercalation of Al3+ ions that could be scalable.

Highly efficient mesoporous MnO2 catalysts for the total toluene oxidation: Oxygen-Vacancy defect engineering and involved intermediates using in situ DRIFTS

Abstract: To elucidate the role of oxygen vacancy defects, various Mn-based oxides with oxygen vacancy defects are employed to the toluene oxidation, which are synthesized by adjusting solvent and double-complexation routes. The MnOx-ET catalyst shows the highest catalytic activity (T90 = 225 °C) for toluene oxidation. Compared with other Mn-based oxides, the as-prepared MnOx-ET catalyst has more surficial oxygen vacancies and good oxygen storage capacity (OSC), which is the reason on its remarkable activity for toluene oxidation. In addition, in situ DRIFTS study reveals that both lattice oxygen and adsorbed oxygen species can participate in the activation-oxidation process of toluene, which results in two reaction routes for the toluene oxidation. The rich oxygen-vacancy concentration of catalysts accelerates the key steps for the activation and generation of oxidized products. Quasi-in situ XPS results further confirm that enrich adsorbed-oxygen species as active oxygen and increasing Mn4+ concentration enhance the superior activity for toluene oxidation.

Characterizing the Propagation of Situational Information in Social Media During COVID-19 Epidemic: A Case Study on Weibo

Abstract: During the ongoing outbreak of coronavirus disease (COVID-19), people use social media to acquire and exchange various types of information at a historic and unprecedented scale. Only the situational information are valuable for the public and authorities to response to the epidemic. Therefore, it is important to identify such situational information and to understand how it is being propagated on social media, so that appropriate information publishing strategies can be informed for the COVID-19 epidemic. This article sought to fill this gap by harnessing Weibo data and natural language processing techniques to classify the COVID-19-related information into seven types of situational information. We found specific features in predicting the reposted amount of each type of information. The results provide data-driven insights into the information need and public attention.

Delocalization of exciton and electron wavefunction in non-fullerene acceptor molecules enables efficient organic solar cells

Abstract: A major challenge for organic solar cell (OSC) research is how to minimize the tradeoff between voltage loss and charge generation. In early 2019, we reported a non-fullerene acceptor (named Y6) that can simultaneously achieve high external quantum efficiency and low voltage loss for OSC. Here, we use a combination of experimental and theoretical modeling to reveal the structure-property-performance relationships of this state-of-the-art OSC system. We find that the distinctive π–π molecular packing of Y6 not only exists in molecular single crystals but also in thin films. Importantly, such molecular packing leads to (i) the formation of delocalized and emissive excitons that enable small non-radiative voltage loss, and (ii) delocalization of electron wavefunctions at donor/acceptor interfaces that significantly reduces the Coulomb attraction between interfacial electron-hole pairs. These properties are critical in enabling highly efficient charge generation in OSC systems with negligible donor-acceptor energy offset. Y6, as a non-fullerene acceptor for organic solar cells, has attracted intensive attention because of the low voltage loss and high charge generation efficiency. Here, Zhang et al. find that the delocalization of exciton and electron wavefunction due to strong π-π packing of Y6 is the key for the high performance.

Model Adaptation: Unsupervised Domain Adaptation Without Source Data

Abstract: In this paper, we investigate a challenging unsupervised domain adaptation setting --- unsupervised model adaptation. We aim to explore how to rely only on unlabeled target data to improve performance of an existing source prediction model on the target domain, since labeled source data may not be available in some real-world scenarios due to data privacy issues. For this purpose, we propose a new framework, which is referred to as collaborative class conditional generative adversarial net to bypass the dependence on the source data. Specifically, the prediction model is to be improved through generated target-style data, which provides more accurate guidance for the generator. As a result, the generator and the prediction model can collaborate with each other without source data. Furthermore, due to the lack of supervision from source data, we propose a weight constraint that encourages similarity to the source model. A clustering-based regularization is also introduced to produce more discriminative features in the target domain. Compared to conventional domain adaptation methods, our model achieves superior performance on multiple adaptation tasks with only unlabeled target data, which verifies its effectiveness in this challenging setting.

Integral Barrier Lyapunov function-based adaptive control for switched nonlinear systems

Abstract: This paper presents an adaptive control method for a class of uncertain strict-feedback switched nonlinear systems. First, we consider the constraint characteristics in the switched nonlinear systems to ensure that all states in switched systems do not violate the constraint ranges. Second, we design the controller based on the backstepping technique, while integral Barrier Lyapunov functions (iBLFs) are adopted to solve the full state constraint problems in each step in order to realize the direct constraints on state variables. Furthermore, we introduce the Lyapunov stability theory to demonstrate that the adaptive controller achieves the desired control goals. Finally, we perform a numerical simulation, which further verifies the significance and feasibility of the presented control scheme.

Highly efficient all-inorganic perovskite solar cells with suppressed non-radiative recombination by a Lewis base.

Abstract: All-inorganic perovskite solar cells (PVSCs) have drawn increasing attention because of their outstanding thermal stability. However, their performance is still inferior than the typical organic-inorganic counterparts, especially for the devices with p-i-n configuration. Herein, we successfully employ a Lewis base small molecule to passivate the inorganic perovskite film, and its derived PVSCs achieved a champion efficiency of 16.1% and a certificated efficiency of 15.6% with improved photostability, representing the most efficient inverted all-inorganic PVSCs to date. Our studies reveal that the nitrile (C-N) groups on the small molecule effectively reduce the trap density of the perovskite film and thus significantly suppresses the non-radiative recombination in the derived PVSC by passivating the Pb-exposed surface, resulting in an improved open-circuit voltage from 1.10 V to 1.16 V after passivation. This work provides an insight in the design of functional interlayers for improving efficiencies and stability of all-inorganic PVSCs. There has been a hot competition to optimize the device performance for all-inorganic perovskite solar cells. Here Wang et al. employ a Lewis base molecule to suppresses the non-radiative recombination in the inverted device and achieve a champion efficiency of 16.1%.

Design of superior phototheranostic agents guided by Jablonski diagrams

Abstract: Phototheranostics represents a promising direction for modern precision medicine, which has recently attracted great research interest from multidisciplinary research areas. Organic optical agents including small molecular fluorophores, semiconducting/conjugated polymers, aggregation-induced emission luminogens, etc. with tuneable photophysical properties, high biosafety and biocompatibility, facile processability and ease of functionalization have delivered encouraging performance in disease phototheranostics. This review summarizes the recent progress of organic phototheranostic agents with an emphasis on the main strategies to manipulate the three excitation energy dissipation pathways, namely, radiative decay, thermal deactivation, and intersystem crossing, with the assistance of a Jablonski diagram, which particularly showcases how the Jablonski diagram has been guiding the design of organic agents from molecule to aggregate levels to promote the disease phototheranostic outcomes. Molecular design and nanoengineering strategies to modulate photophysical processes of organic optical agents to convert the absorbed photons into fluorescent/phosphorescent/photoacoustic signals and/or photodynamic/photothermal curing effects for improved disease phototheranostics are elaborated. Noteworthily, adaptive phototheranostics with activatable and transformable functions on demand, and regulation of excitation such as chemiexcitation to promote the phototheranostic efficacies are also included. A brief summary with the discussion of current challenges and future perspectives in this research field is further presented.

Phase-field modeling of fracture

Abstract: Fracture is one of the most commonly encountered failure modes of engineering materials and structures. Prevention of cracking-induced failure is, therefore, a major concern in structural designs. Computational modeling of fracture constitutes an indispensable tool not only to predict the failure of cracking structures but also to shed insights into understanding the fracture processes of many materials such as concrete, rock, ceramic, metals, and biological soft tissues. This chapter provides an extensive overview of the literature on the so-called phase-field fracture/damage models (PFMs), particularly, for quasi-static and dynamic fracture of brittle and quasi-brittle materials, from the points of view of a computational mechanician. PFMs are the regularized versions of the variational approach to fracture which generalizes Griffith's theory for brittle fracture. They can handle topologically complex fractures such as initiation, intersecting, and branching cracks in both two and three dimensions with a quite straightforward implementation. One of our aims is to justify the gaining popularity of PFMs. To this end, both theoretical and computational aspects are discussed and extensive benchmark problems (for quasi-static and dynamic brittle/cohesive fracture) that are successfully and unsuccessfully solved with PFMs are presented. Unresolved issues for further investigations are also documented.

Reversible 3D laser printing of perovskite quantum dots inside a transparent medium

Abstract: The three-dimensional (3D) patterning of semiconductors is potentially important for exploring new functionalities and applications in optoelectronics1,2. Here, we show that it is possible to write on demand 3D patterns of perovskite quantum dots (QDs) inside a transparent glass material using a femtosecond laser. By utilizing the inherent ionic nature and low formation energy of perovskite, highly luminescent CsPbBr3 QDs can be reversibly fabricated in situ and decomposed through femtosecond laser irradiation and thermal annealing. This pattern of writing and erasing can be repeated for many cycles, and the luminescent QDs are well protected by the inorganic glass matrix, resulting in stable perovskite QDs with potential applications such as high-capacity optical data storage, information encryption and 3D artwork. Luminescent CsPbBr3 quantum dots can be written into glass using femtosecond laser pulses and thermal annealing, and erased by further femtosecond laser irradiation. The resulting quantum dot patterns could prove useful for data storage, decoration or security purposes.

An innovative neural network approach for stock market prediction

Abstract: This paper aims to develop an innovative neural network approach to achieve better stock market predictions. Data were obtained from the live stock market for real-time and off-line analysis and results of visualizations and analytics to demonstrate Internet of Multimedia of Things for stock analysis. To study the influence of market characteristics on stock prices, traditional neural network algorithms may incorrectly predict the stock market, since the initial weight of the random selection problem can be easily prone to incorrect predictions. Based on the development of word vector in deep learning, we demonstrate the concept of “stock vector.” The input is no longer a single index or single stock index, but multi-stock high-dimensional historical data. We propose the deep long short-term memory neural network (LSTM) with embedded layer and the long short-term memory neural network with automatic encoder to predict the stock market. In these two models, we use the embedded layer and the automatic encoder, respectively, to vectorize the data, in a bid to forecast the stock via long short-term memory neural network. The experimental results show that the deep LSTM with embedded layer is better. Specifically, the accuracy of two models is 57.2 and 56.9%, respectively, for the Shanghai A-shares composite index. Furthermore, they are 52.4 and 52.5%, respectively, for individual stocks. We demonstrate research contributions in IMMT for neural network-based financial analysis.

Closed-Loop Matters: Dual Regression Networks for Single Image Super-Resolution

Abstract: Deep neural networks have exhibited promising performance in image super-resolution (SR) by learning a nonlinear mapping function from low-resolution (LR) images to high-resolution (HR) images. However, there are two underlying limitations to existing SR methods. First, learning the mapping function from LR to HR images is typically an ill-posed problem, because there exist infinite HR images that can be downsampled to the same LR image. As a result, the space of the possible functions can be extremely large, which makes it hard to find a good solution. Second, the paired LR-HR data may be unavailable in real-world applications and the underlying degradation method is often unknown. For such a more general case, existing SR models often incur the adaptation problem and yield poor performance. To address the above issues, we propose a dual regression scheme by introducing an additional constraint on LR data to reduce the space of the possible functions. Specifically, besides the mapping from LR to HR images, we learn an additional dual regression mapping estimates the down-sampling kernel and reconstruct LR images, which forms a closed-loop to provide additional supervision. More critically, since the dual regression process does not depend on HR images, we can directly learn from LR images. In this sense, we can easily adapt SR models to real-world data, e.g., raw video frames from YouTube. Extensive experiments with paired training data and unpaired real-world data demonstrate our superiority over existing methods.

Additive manufacturing of silica aerogels.

Abstract: Owing to their ultralow thermal conductivity and open pore structure1-3, silica aerogels are widely used in thermal insulation4,5, catalysis6, physics7,8, environmental remediation6,9, optical devices10 and hypervelocity particle capture11. Thermal insulation is by far the largest market for silica aerogels, which are ideal materials when space is limited. One drawback of silica aerogels is their brittleness. Fibre reinforcement and binders can be used to overcome this for large-volume applications in building and industrial insulation5,12, but their poor machinability, combined with the difficulty of precisely casting small objects, limits the miniaturization potential of silica aerogels. Additive manufacturing provides an alternative route to miniaturization, but was "considered not feasible for silica aerogel"13. Here we present a direct ink writing protocol to create miniaturized silica aerogel objects from a slurry of silica aerogel powder in a dilute silica nanoparticle suspension (sol). The inks exhibit shear-thinning behaviour, owing to the high volume fraction of gel particles. As a result, they flow easily through the nozzle during printing, but their viscosity increases rapidly after printing, ensuring that the printed objects retain their shape. After printing, the silica sol is gelled in an ammonia atmosphere to enable subsequent processing into aerogels. The printed aerogel objects are pure silica and retain the high specific surface area (751 square metres per gram) and ultralow thermal conductivity (15.9 milliwatts per metre per kelvin) typical of silica aerogels. Furthermore, we demonstrate the ease with which functional nanoparticles can be incorporated. The printed silica aerogel objects can be used for thermal management, as miniaturized gas pumps and to degrade volatile organic compounds, illustrating the potential of our protocol.

Identifying critical state of complex diseases by single-sample Kullback-Leibler divergence.

Abstract: Developing effective strategies for signaling the pre-disease state of complex diseases, a state with high susceptibility before the disease onset or deterioration, is urgently needed because such state usually followed by a catastrophic transition into a worse stage of disease. However, it is a challenging task to identify such pre-disease state or tipping point in clinics, where only one single sample is available and thus results in the failure of most statistic approaches. In this study, we presented a single-sample-based computational method to detect the early-warning signal of critical transition during the progression of complex diseases. Specifically, given a set of reference samples which were regarded as background, a novel index called single-sample Kullback–Leibler divergence (sKLD), was proposed to explore and quantify the disturbance on the background caused by a case sample. The pre-disease state is then signaled by the significant change of sKLD. The novel algorithm was developed and applied to both numerical simulation and real datasets, including lung squamous cell carcinoma, lung adenocarcinoma, stomach adenocarcinoma, thyroid carcinoma, colon adenocarcinoma, and acute lung injury. The successful identification of pre-disease states and the corresponding dynamical network biomarkers for all six datasets validated the effectiveness and accuracy of our method. The proposed method effectively explores and quantifies the disturbance on the background caused by a case sample, and thus characterizes the criticality of a biological system. Our method not only identifies the critical state or tipping point at a single sample level, but also provides the sKLD-signaling markers for further practical application. It is therefore of great potential in personalized pre-disease diagnosis.

A general strategy towards personalized nanovaccines based on fluoropolymers for post-surgical cancer immunotherapy.

Abstract: Cancer metastases and recurrence after surgical resection remain an important cause of treatment failure. Here we demonstrate a general strategy to fabricate personalized nanovaccines based on a cationic fluoropolymer for post-surgical cancer immunotherapy. Nanoparticles formed by mixing the fluoropolymer with a model antigen ovalbumin, induce dendritic cell maturation via the Toll-like receptor 4 (TLR4)-mediated signalling pathway, and promote antigen transportation into the cytosol of dendritic cells, which leads to an effective antigen cross-presentation. Such a nanovaccine inhibits established ovalbumin-expressing B16-OVA melanoma. More importantly, a mix of the fluoropolymer with cell membranes from resected autologous primary tumours synergizes with checkpoint blockade therapy to inhibit post-surgical tumour recurrence and metastases in two subcutaneous tumour models and an orthotopic breast cancer tumour. Furthermore, in the orthotopic tumour model, we observed a strong immune memory against tumour rechallenge. Our work offers a simple and general strategy for the preparation of personalized cancer vaccines to prevent post-operative cancer recurrence and metastasis. A fluoropolymer-based cancer nanovaccine that delivers antigens directly to the cytosol of dendritic cells and elicits strong antitumour immune responses inhibiting tumour growth in animal models can be used to produce personalized treatment for post-surgical immunotherapy.

UP-DETR: Unsupervised Pre-training for Object Detection with Transformers

Abstract: Object detection with transformers (DETR) reaches competitive performance with Faster R-CNN via a transformer encoder-decoder architecture. Inspired by the great success of pre-training transformers in natural language processing, we propose a pretext task named random query patch detection to unsupervisedly pre-train DETR (UP-DETR) for object detection. Specifically, we randomly crop patches from the given image and then feed them as queries to the decoder. The model is pre-trained to detect these query patches from the original image. During the pre-training, we address two critical issues: multi-task learning and multi-query localization. (1) To trade-off multi-task learning of classification and localization in the pretext task, we freeze the CNN backbone and propose a patch feature reconstruction branch which is jointly optimized with patch detection. (2) To perform multi-query localization, we introduce UP-DETR from single-query patch and extend it to multi-query patches with object query shuffle and attention mask. In our experiments, UP-DETR significantly boosts the performance of DETR with faster convergence and higher precision on PASCAL VOC and COCO datasets. The code will be available soon.

PointASNL: Robust Point Clouds Processing using Nonlocal Neural Networks with Adaptive Sampling

Abstract: Raw point clouds data inevitably contains outliers or noise through acquisition from 3D sensors or reconstruction algorithms. In this paper, we present a novel end-to-end network for robust point clouds processing, named PointASNL, which can deal with point clouds with noise effectively. The key component in our approach is the adaptive sampling (AS) module. It first re-weights the neighbors around the initial sampled points from farthest point sampling (FPS), and then adaptively adjusts the sampled points beyond the entire point cloud. Our AS module can not only benefit the feature learning of point clouds, but also ease the biased effect of outliers. To further capture the neighbor and long-range dependencies of the sampled point, we proposed a local-nonlocal (L-NL) module inspired by the nonlocal operation. Such L-NL module enables the learning process insensitive to noise. Extensive experiments verify the robustness and superiority of our approach in point clouds processing tasks regardless of synthesis data, indoor data, and outdoor data with or without noise. Specifically, PointASNL achieves state-of-the-art robust performance for classification and segmentation tasks on all datasets, and significantly outperforms previous methods on real-world outdoor SemanticKITTI dataset with considerate noise. Our code is released through this https URL.