Machine-Learning-Assisted Development of Gel Polymer Electrolytes for Protecting Zn Metal Anodes from the Corrosion of Water Molecules.

Rechargeable aqueous zinc-ion batteries (RAZIBs) offer low cost, high energy density, and safety but struggle with anode corrosion and dendrite formation. Gel polymer electrolytes (GPEs) with both high mechanical properties and excellent electrochemical properties are a powerful tool to aid the practical application of RAZIBs. In this work, guided by a machine learning (ML) model constructed based on experimental data, polyacrylamide (PAM) with a highly entangled structure was chosen to prepare GPEs for obtaining high-performance RAZIBs. By controlling the swelling degree of the PAM, the obtained GPEs effectively suppressed the growth of Zn dendrites and alleviated the corrosion of Zn metal caused by water molecules, thus improving the cycling lifespan of the Zn anode. These results indicate that using ML models based on experimental data can effectively help screen battery materials, while highly entangled PAMs are excellent GPEs capable of balancing mechanical and electrochemical properties.

TCJA-SNN: Temporal-Channel Joint Attention for Spiking Neural Networks.

Spiking neural networks (SNNs) are attracting widespread interest due to their biological plausibility, energy efficiency, and powerful spatiotemporal information representation ability. Given the critical role of attention mechanisms in enhancing neural network performance, the integration of SNNs and attention mechanisms exhibits tremendous potential to deliver energy-efficient and high-performance computing paradigms. In this article, we present a novel temporal-channel joint attention mechanism for SNNs, referred to as TCJA-SNN. The proposed TCJA-SNN framework can effectively assess the significance of spike sequence from both spatial and temporal dimensions. More specifically, our essential technical contribution lies on: 1) we employ the squeeze operation to compress the spike stream into an average matrix. Then, we leverage two local attention mechanisms based on efficient 1-D convolutions to facilitate comprehensive feature extraction at the temporal and channel levels independently and 2) we introduce the cross-convolutional fusion (CCF) layer as a novel approach to model the interdependencies between the temporal and channel scopes. This layer effectively breaks the independence of these two dimensions and enables the interaction between features. Experimental results demonstrate that the proposed TCJA-SNN outperforms the state-of-the-art (SOTA) on all standard static and neuromorphic datasets, including Fashion-MNIST, CIFAR10, CIFAR100, CIFAR10-DVS, N-Caltech 101, and DVS128 Gesture. Furthermore, we effectively apply the TCJA-SNN framework to image generation tasks by leveraging a variation autoencoder. To the best of our knowledge, this study is the first instance where the SNN-attention mechanism has been employed for high-level classification and low-level generation tasks. Our implementation codes are available at https://github.com/ridgerchu/TCJA.

Optically Tunable Electrical Oscillations in Oxide-Based Memristors for Neuromorphic Computing.

The application of hardware-based neural networks can be enhanced by integrating sensory neurons and synapses that enable direct input from external stimuli. This work reports direct optical control of an oscillatory neuron based on volatile threshold switching in V3O5. The devices exhibit electroforming-free operation with switching parameters that can be tuned by optical illumination. Using temperature-dependent electrical measurements, conductive atomic force microscopy (C-AFM), in situ thermal imaging, and lumped element modelling, it is shown that the changes in switching parameters, including threshold and hold voltages, arise from overall conductivity increase of the oxide film due to the contribution of both photoconductive and bolometric characteristics of V3O5, which eventually affects the oscillation dynamics. Furthermore, V3O5 is identified as a new bolometric material with a temperature coefficient of resistance (TCR) as high as -4.6% K-1 at 423 K. The utility of these devices is illustrated by demonstrating in-sensor reservoir computing with reduced computational effort and an optical encoding layer for spiking neural network (SNN), respectively, using a simulated array of devices.

A generative artificial intelligence framework based on a molecular diffusion model for the design of metal-organic frameworks for carbon capture.

Metal-organic frameworks (MOFs) exhibit great promise for CO2 capture. However, finding the best performing materials poses computational and experimental grand challenges in view of the vast chemical space of potential building blocks. Here, we introduce GHP-MOFassemble, a generative artificial intelligence (AI), high performance framework for the rational and accelerated design of MOFs with high CO2 adsorption capacity and synthesizable linkers. GHP-MOFassemble generates novel linkers, assembled with one of three pre-selected metal nodes (Cu paddlewheel, Zn paddlewheel, Zn tetramer) into MOFs in a primitive cubic topology. GHP-MOFassemble screens and validates AI-generated MOFs for uniqueness, synthesizability, structural validity, uses molecular dynamics simulations to study their stability and chemical consistency, and crystal graph neural networks and Grand Canonical Monte Carlo simulations to quantify their CO2 adsorption capacities. We present the top six AI-generated MOFs with CO2 capacities greater than 2m mol g-1, i.e., higher than 96.9% of structures in the hypothetical MOF dataset.

Synergistic Cation Solvation Reorganization and Fluorinated Interphase for High Reversibility and Utilization of Zinc Metal Anode.

Batteries based on zinc (Zn) chemistry offer a great opportunity for large-scale applications owing to their safety, cost-effectiveness, and environmental friendliness. However, the poor Zn reversibility and inhomogeneous electrodeposition have greatly impeded their practical implementation, stemming from water-related passivation/corrosion. Here, we present a multifunctional electrolyte comprising gamma-butyrolactone (GBL) and Zn(BF4)2·xH2O to resolve these intrinsic challenges. The systematic results confirm that water reactivity toward a Zn anode is minimized by forcing GBL solvents into the Zn2+ solvation shell and constructing a fluorinated interphase on the Zn anode surface via anion decomposition. Furthermore, NMR was selected as an auxiliary testing protocol to elevate and understand the role of electrolyte composition in building the interphase. The combined factors in synergy guarantee high Zn reversibility (average Coulombic efficiency is 99.74%), high areal capacity (55 mAh/cm2), and high Zn utilization (∼91%). Ultimately, these merits enable the Zn battery utilizing a VO2 cathode to operate smoothly over 5000 cycles with a low-capacity decay rate of ∼0.0083% per cycle and a 0.23 Ah VO2/Zn pouch cell to operate over 400 cycles with a capacity retention of 77.3%.

Ice crystal sublimation for easily producing MnO2 cathodes with hierarchically porous structure and enhanced cyclic reversibility.

The charge/discharge performance of rechargeable aqueous zinc ion batteries (RAZIBs) at high currents is often unsatisfactory due to the cathode preparation process and the use of hydrophobic binders. By adding freeze-drying treatment to the preparation process of the cathodes, MnO2 cathodes with hierarchically porous structures are obtained, which provide additional channels for ion transfer, thus greatly enhancing the charge/discharge performance in aqueous Zn-MnO2 batteries.

VTSNN: a virtual temporal spiking neural network.

Spiking neural networks (SNNs) have recently demonstrated outstanding performance in a variety of high-level tasks, such as image classification. However, advancements in the field of low-level assignments, such as image reconstruction, are rare. This may be due to the lack of promising image encoding techniques and corresponding neuromorphic devices designed specifically for SNN-based low-level vision problems. This paper begins by proposing a simple yet effective undistorted weighted-encoding-decoding technique, which primarily consists of an Undistorted Weighted-Encoding (UWE) and an Undistorted Weighted-Decoding (UWD). The former aims to convert a gray image into spike sequences for effective SNN learning, while the latter converts spike sequences back into images. Then, we design a new SNN training strategy, known as Independent-Temporal Backpropagation (ITBP) to avoid complex loss propagation in spatial and temporal dimensions, and experiments show that ITBP is superior to Spatio-Temporal Backpropagation (STBP). Finally, a so-called Virtual Temporal SNN (VTSNN) is formulated by incorporating the above-mentioned approaches into U-net network architecture, fully utilizing the potent multiscale representation capability. Experimental results on several commonly used datasets such as MNIST, F-MNIST, and CIFAR10 demonstrate that the proposed method produces competitive noise-removal performance extremely which is superior to the existing work. Compared to ANN with the same architecture, VTSNN has a greater chance of achieving superiority while consuming ~1/274 of the energy. Specifically, using the given encoding-decoding strategy, a simple neuromorphic circuit could be easily constructed to maximize this low-carbon strategy.

Effects of Cathode GDL Gradient Porosity Distribution along the Flow Channel Direction on Gas-Liquid Transport and Performance of PEMFC.

The gas diffusion layer (GDL) is an important component of proton exchange membrane fuel cells (PEMFCs), and its porosity distribution has considerable effects on the transport properties and durability of PEMFCs. A 3-D two-phase flow computation fluid dynamics model was developed in this study, to numerically investigate the effects of three different porosity distributions in a cathode GDL: gradient-increasing (Case 1), gradient-decreasing (Case 3), and uniform constant (Case 2), on the gas-liquid transport and performance of PEMFCs; the novelty lies in the porosity gradient being along the channel direction, and the physical properties of the GDL related to porosity were modified accordingly. The results showed that at a high current density (2400 mA·cm-2), the GDL of Case 1 had a gas velocity of up to 0.5 cm·s-1 along the channel direction. The liquid water in the membrane electrode assembly could be easily removed because of the larger gas velocity and capillary pressure, resulting in a higher oxygen concentration in the GDL and the catalyst layer. Therefore, the cell performance increased. The voltage in Case 1 increased by 8% and 71% compared to Cases 2 and 3, respectively. In addition, this could ameliorate the distribution uniformity of the dissolved water and the current density in the membrane along the channel direction, which was beneficial for the durability of the PEMFC. The distribution of the GDL porosity at lower current densities had a less significant effect on the cell performance. The findings of this study may provide significant guidance for the design and optimization of the GDL in PEMFCs.

Facile synthesis approach of bifunctional Co-Ni-Fe oxyhydroxide and spinel oxide composite electrocatalysts from hydroxide and layered double hydroxide composite precursors.

Zinc-air batteries (ZABs) are promising candidates for the next-generation energy storage systems, however, their further development is severely hindered by kinetically sluggish oxygen evolution reaction (OER) and oxygen reduction reaction (ORR). Facile synthesis approaches of highly active bifunctional electrocatalysts for OER and ORR are required for their practical applications. Herein, we develop a facile synthesis procedure for composite electrocatalysts composed of OER-active metal oxyhydroxide and ORR-active spinel oxide containing Co, Ni and Fe from composite precursors consisting of metal hydroxide and layered double hydroxide (LDH). Both hydroxide and LDH are simultaneously produced by a precipitation method with a controlled molar ratio of Co2+, Ni2+ and Fe3+ in the reaction solution, and calcination of the precursor at a moderate temperature provides composite catalysts of metal oxyhydroxides and spinel oxides. The composite catalyst shows superb bifunctional performances with a small potential difference of 0.64 V between a potential of 1.51 V vs. RHE at 10 mA cm-2 for OER and a half-wave potential of 0.87 V vs. RHE for ORR. The rechargeable ZAB assembled with the composite catalyst as an air-electrode exhibits a power density of 195 mA cm-2 and excellent durability of 430 hours (1270 cycles) of a charge-discharge cycle test.

Role of the caudal peduncle in a fish-inspired robotic model: how changing stiffness and angle of attack affects swimming performance.

In fish, the tail is a key element of propulsive anatomy that contributes to thrust during swimming. Fish possess the ability to alter tail stiffness, surface area and conformation. Specifically, the region at the base of the tail, the caudal peduncle, is proposed to be a key location of fish stiffness modulation during locomotion. Most previous analyses have focused on the overall body or tail stiffness, and not on the effects of changing stiffness specifically at the base of the tail in fish and robotic models. We used both computational fluid dynamics analysis and experimental measurements of propulsive forces in physical models with different peduncle stiffnesses to analyze the effect of altering stiffness on the tail angle of attack and propulsive force and efficiency. By changing the motion program input to the tail, we were able to alter the phase relationship between the front and back tail sections between 0° and 330°. Computational simulations showed that power consumption was nearly minimized and thrust production was nearly maximized at the kinematic pattern whereφ= 270°, the approximate phase lag observed in the experimental foils and in free swimming tuna. We observed reduced thrust and efficiency at high angles of attack, suggesting that the tail driven during these motion programs experiences stalling and loss of lift. However, there is no single peduncle stiffness that consistently maximizes performance, particularly in physical models. This result highlights the fact that the optimal caudal peduncle stiffness is highly context dependent. Therefore, incorporating the ability to control peduncle stiffness in future robotic models of fish propulsion promises to increase the ability of robots to approach the performance of fish.

Machine-Learning-Accelerated Development of Efficient Mixed Protonic-Electronic Conducting Oxides as the Air Electrodes for Protonic Ceramic Cells.

Currently, the development of high-performance protonic ceramic cells (PCCs) is limited by the scarcity of efficient mixed protonic-electronic conducting oxides that can act as air electrodes to satisfy the high protonic conductivity of electrolytes. Despite the extensive research efforts in the past decades, the development of mixed protonic-electronic conducting oxides still remains in a trial-and-error process, which is extremely time consuming and high cost. Herein, based on the data acquired from the published literature, the machine-learning (ML) method is introduced to accelerate the discovery of efficient mixed protonic-electronic conducting oxides. Accordingly, the hydrated proton concentration (HPC) of 3200 oxides is predicted to evaluate the proton conduction that is essential for enhancing the electrochemical performances of PCCs. Subsequently, feature importance for HPC is evaluated to establish a guideline for rapid and accurate design and development of high-efficiency mixed protonic-electronic conducting oxides. Thereafter, screened (La0.7 Ca0.3 )(Co0.8 Ni0.2 )O3 (LCCN7382) is prepared, and the experimental HPC adequately corresponds with the predicted results. Moreover, the PCC with LCCN7382 exhibits satisfactory electrochemical performances in electrolysis and fuel cell modes. In addition to the development of a promising air electrode for PCC, this study establishes a new avenue for ML-based development of mixed protonic-electronic conducting oxides.

Highly Deformable, Conductive Double-Network Hydrogel Electrolytes for Durable and Flexible Supercapacitors.

Developing flexible energy storage devices with the ability to retain capacitance under extreme deformation is promising but remains challenging. Here, we report the development of a durable supercapacitor with remarkable capacitance retention under mechanical deformation by utilizing a physical double-network (DN) hydrogel as an electrolyte. The first network is hydrophobically associating polyacrylamide cross-linked by nanoparticles, and the second network is Zn2+ cross-linked alginate. Through soaking such a DN hydrogel into a high concentration of ZnSO4 solution, a highly deformable electrolyte with good conductivity is fabricated, which also shows adhesion to diverse surfaces. Directly attaching the hydrogel electrolyte to two pieces of an active carbon cloth facilely produces a flexible supercapacitor with a high specific capacitance and theoretical energy density. Remarkable capacitance retention under tension, compression, and bending is observed for the supercapacitor, which can also maintain above 87% of the initial capacitance after 4000 charge-discharge cycles. This study provides a simple way to fabricate hydrogel electrolytes for deformable yet durable supercapacitors, which is expected to inspire the development of next-generation flexible energy storage devices.

How to Improve Entrepreneurship Education in "Double High-Level Plan" Higher Vocational Colleges in China.

Entrepreneurship education complements vocational education in helping students develop their career prospects. This empirical study comprehensively analyzed sample data of 13,885 students from 40 "Double High-level Plan" higher vocational colleges in China using robust standard error regression analysis and other methods. The results showed that Entrepreneurship Practice (EP), Entrepreneurship Curriculum (EC), and Integration of Entrepreneurship Education and Professional Education (IEEPE) have a significant positive effect on Entrepreneurship Education Performance (EEP), with EP being the most important factor. Furthermore, ascribed factors (gender, household registration, only child or not, whether parents have entrepreneurial experience) and self-achieved factors (double high-level type, school area, subject major, whether to accept social entrepreneurship education) were found to affect students' perception of investment in entrepreneurship education. The study summarizes the existing problems of entrepreneurship education in "Double High-level Plan" higher vocational colleges and proposes four suggestions: pursue the integrated development of entrepreneurship education and "Double High-level" construction, advance both theoretical education and practical education, promote digital reform of the "three teaches" (teachers, teaching materials, and teaching methods), and develop entrepreneurship education in a comprehensive and balanced manner. This has certain theoretical and practical significance for the improvement of entrepreneurship education in other developing countries.

Reducing Water Activity by Zeolite Molecular Sieve Membrane for Long-Life Rechargeable Zinc Battery.

Aqueous electrolytes offer major advantages in safe battery operation, green economy, and low production cost for advanced battery technology. However, strong water activity in aqueous electrolytes provokes a hydrogen evolution reaction and parasitic passivation on electrodes, leaving poor ion-transport in the electrolyte/electrode interface. Herein, a zeolite molecular sieve-modified (zeolite-modified) aqueous electrolyte is proposed to reduce water activity and its side-reaction. First, Raman spectroscopy reveals a highly aggressive solvation configuration and significantly suppressed water activity toward single water molecule. Then less hydrogen evolution and anti-corrosion ability of zeolite-modified electrolyte by simulation and electrochemical characterizations are identified. Consequently, a zinc (Zn) anode involves less side-reaction, and develops into a compact deposition morphology, as proved by space-resolution characterizations. Moreover, zeolite-modified electrolyte favors cyclic life of symmetric Zn||Zn cells to 4765 h at 0.8 mA cm-2 , zinc-VO2 coin cell to 3000 cycles, and pouch cell to 100 cycles. Finally, the mature production technique and low-cost of zeolite molecular sieve would tremendously favor the future scale-up application in engineering aspect.

Learning Multi-Level Density Maps for Crowd Counting.

People in crowd scenes often exhibit the characteristic of imbalanced distribution. On the one hand, people size varies largely due to the camera perspective. People far away from the camera look smaller and are likely to occlude each other, whereas people near to the camera look larger and are relatively sparse. On the other hand, the number of people also varies greatly in the same or different scenes. This article aims to develop a novel model that can accurately estimate the crowd count from a given scene with imbalanced people distribution. To this end, we have proposed an effective multi-level convolutional neural network (MLCNN) architecture that first adaptively learns multi-level density maps and then fuses them to predict the final output. Density map of each level focuses on dealing with people of certain sizes. As a result, the fusion of multi-level density maps is able to tackle the large variation in people size. In addition, we introduce a new loss function named balanced loss (BL) to impose relatively BL feedback during training, which helps further improve the performance of the proposed network. Furthermore, we introduce a new data set including 1111 images with a total of 49 061 head annotations. MLCNN is easy to train with only one end-to-end training stage. Experimental results demonstrate that our MLCNN achieves state-of-the-art performance. In particular, our MLCNN reaches a mean absolute error (MAE) of 242.4 on the UCF_CC_50 data set, which is 37.2 lower than the second-best result.

Porous ß-Mo2C nanoparticle clusters supported on walnut shell powders derived carbon matrix for hydrogen evolution reaction.

Herein, we choose the waste walnut shell as the carbon source, and ammonium heptamolybdate as the molybdenum source to prepare the ß-Mo2C catalyst supported on carbon matrix (Mo2C@C) by the calcination method for hydrogen evolution reaction (HER). The open pores in the porous Mo2C nanoparticle clusters can facilitate electrolyte permeation and hydrogen molecules release as well as the carbon matrix can enhance the conductivity. As a result, the optimal Mo2C exhibits an efficient HER performance, with an overpotential of 140 mV at 10 mA cm-2 and a Tafel slope of 63 mV dec-1 as well as excellent electrochemical stability. The strategy changing waste walnut shell into the effective catalysts sets an example for the searching and designing rational energy materials.

Crosstalk-aware spectrum defragmentation by re-provisioning advance reservation requests in space division multiplexing enabled elastic optical networks with multi-core fiber.

Space division multiplexing enabled elastic optical networks (SDM-EONs) with multi-core fiber (MCF) have become a promising candidate for future optical transport networks, due to their high capacity and flexibility. Meanwhile, driven by the development of cloud computing and data centers, more types of requests are allowed in the networks, i.e., the usual immediate reservation (IR) requests, which need to be served immediately, and advance reservation (AR) requests, which support initial-delay tolerance services. However, with the introduction of AR requests, spectrum fragments occur frequently in both spatial and time dimension as lightpaths are set up and torn down, and the issue of spectrum fragmentation could be much more serious in SDM-EONs than in simple EONs. To measure fragments status in both spatial and time dimension in SDM-EONs, we first design a metric, i.e., time-dimensional spectrum compactness (TSC). Then, based on TSC, we propose a crosstalk-aware AR requests re-provisioning algorithm with two strategies to optimize the fragments in SDM-EONs. The performance of the proposed algorithm is evaluated via software simulation in terms of spectrum compactness, blocking probability, spectrum utilization, average moving times, average re-provisioning latency and average start time delay. The results show that the proposed re-provisioning algorithm can effectively improve spectrum compactness and spectrum efficiency of the networks. We also evaluate the proposed re-provisioning algorithm in different TSC thresholds, and it turns out that the proposed re-provisioning algorithm in higher threshold performs better in terms of spectrum compactness and spectrum utilization.

CD4+ T cell-mediated cytotoxicity eliminates primary tumor cells in metastatic melanoma through high MHC class II expression and can be enhanced by inhibitory receptor blockade.

Metastatic melanoma is a rapidly progressing disease with high mortality rate and limited treatment options. Immunotherapy based on tumor-targeting cytotoxic T cell responses represents a promising strategy. To assist in its development, we examined the possibility and efficacy of using CD4+ cytotoxic T cells. The regulatory mechanisms controlling CD4+ T cell-mediated cytotoxicity were also investigated. We found that naturally occurring granzyme B and perforin-expressing CD4+ cytotoxic T cells can be recovered from metastatic melanoma patients at significantly elevated frequencies compared to those from healthy controls. These CD4+ cytotoxic T cells were also capable of killing autologous tumor cells harvested from metastatic melanoma, independent of CD8+ T cells or any other cell types. However, several restricting factors were observed. First, the cytolytic activity by CD4+ T cells required high MHC class II expression on melanoma cells, which was not satisfied in a subset of melanomas. Second, the granzyme B and perforin release by activated CD4+ cytotoxic T cells was reduced after coculturing with autologous melanoma cells, characterized by low LAMP-1 expression and low granzyme B and perforin secretion in the supernatant. This suggested that inhibitory mechanisms were present to suppress CD4+ cytotoxic T cells. Indeed, blockade of PD-1 and CTLA-4 had increased the cytolytic activity of CD4+ T cells but was only effective in MHC class II high but not MHC class II low melanomas. Together, our study showed that CD4+ T cell-mediated cytotoxicity could eliminate primary melanoma cells but the efficacy depended on MHC class II expression.