Search

Cortical spreading depression (CSD) is a slow wave of cortical depolarization closely associated with migraines with an aura. Previously, it was thought that CSD depolarization was mainly driven by neurons, with characteristic changes in neuronal swelling and increased extracellular potassium (K+) and glutamate. However, the role of astrocytes, a member of the neurovascular unit, in migraine with CSD has recently received increasing attention. In the early stages of CSD, astrocytes provide neurons with energy support and clear K+and glutamate from synaptic gaps. However, in the late stages of CSD, astrocytes release large amounts of lactic acid to exacerbate hypoxia when the energy demand exceeds the astrocytes’ compensatory capacity. Astrocyte endfoot swelling is a characteristic of CSD, and neurons are not similarly altered. It is primarily due to K+influx and abnormally active calcium (Ca2+) signaling. Aquaporin 4 (AQP-4) only mediates K+influx and has little role as an aquaporin. Astrocytes endfoot swelling causes perivascular space closure, slowing the glymphatic system flow and exacerbating neuroinflammation, leading to persistent CSD. Astrocytes are double-edged swords in migraine with CSD and may be potential targets for CSD interventions.

Mucosal vaccines offer potential benefits over parenteral vaccines for they can trigger both systemic immune protection and immune responses at the predominant sites of pathogen infection. However, the defense function of mucosal barrier remains a challenge for vaccines to overcome. Here, we show that surface modification of exosomes with the fragment crystallizable (Fc) part from IgG can deliver the receptor-binding domain (RBD) of SARS-CoV-2 to cross mucosal epithelial layer and permeate into peripheral lung through neonatal Fc receptor (FcRn) mediated transcytosis. The exosomes F-L-R-Exo are generated by genetically engineered dendritic cells, in which a fusion protein Fc-Lamp2b-RBD is expressed and anchored on the membrane. After intratracheally administration, F-L-R-Exo is able to induce a high level of RBD-specific IgG and IgA antibodies in the animals’ lungs. Furthermore, potent Th1 immune-biased T cell responses were also observed in both systemic and mucosal immune responses. F-L-R-Exo can protect the mice from SARS-CoV-2 pseudovirus infection after a challenge. These findings hold great promise for the development of a novel respiratory mucosal vaccine approach.

The secretory glutaminyl cyclase (sQC) and Golgi-resident glutaminyl cyclase (gQC) are responsible for N-terminal protein pyroglutamation and associated with various human diseases. Although several sQC/gQC inhibitors have been reported, only one inhibitor, PQ912, is currently undergoing clinic trials for the treatment of Alzheimer’s disease. We report an X-ray crystal structure of sQC complexed with PQ912, revealing that the benzimidazole makes “anchor” interactions with the active site zinc ion and catalytic triad. Structure-guided design and optimization led to a series of new benzimidazole derivatives exhibiting nanomolar inhibition for both sQC and gQC. In a MPTP-induced Parkinson’s disease (PD) mouse model, BI-43manifested efficacy in mitigating locomotor deficits through reversing dopaminergic neuronal loss, reducing microglia, and decreasing levels of the sQC/gQC substrates, α-synuclein, and CCL2. This study not only offers structural basis and new leads for drug discovery targeting sQC/gQC but also provides evidence supporting sQC/gQC as potential targets for PD treatment.

This cross‐sectional study aimed to determine the epidemiology of olfactory and gustatory dysfunctions related to COVID‐19 in China. This study was conducted by 45 tertiary Grade‐A hospitals in China. Online and offline questionnaire data were obtained from patients infected with COVID‐19 between December 28, 2022, and February 21, 2023. The collected information included basic demographics, medical history, smoking and drinking history, vaccination history, changes in olfactory and gustatory functions before and after infection, and other postinfection symptoms, as well as the duration and improvement status of olfactory and gustatory disorders. Complete questionnaires were obtained from 35,566 subjects. The overall incidence of olfactory and taste dysfunction was 67.75%. Being female or being a cigarette smoker increased the likelihood of developing olfactory and taste dysfunction. Having received four doses of the vaccine or having good oral health or being a alcohol drinker decreased the risk of such dysfunction. Before infection, the average olfactory and taste VAS scores were 8.41 and 8.51, respectively; after infection, they decreased to 3.69 and 4.29 and recovered to 5.83 and 6.55 by the time of the survey. The median duration of dysosmia and dysgeusia was 15 and 12 days, respectively, with 0.5% of patients having symptoms lasting for more than 28 days. The overall self‐reported improvement rate was 59.16%. Recovery was higher in males, never smokers, those who received two or three vaccine doses, and those that had never experienced dental health issues, or chronic accompanying symptoms. The incidence of dysosmia and dysgeusia following infection with the SARS‐CoV‐2 virus is high in China. Incidence and prognosis are influenced by several factors, including sex, SARS‐CoV‐2 vaccination, history of head‐facial trauma, nasal and oral health status, smoking and drinking history, and the persistence of accompanying symptoms.

An organophosphorus catalytic method for the synthesis of substituted 2-amidopyridines is reported. The method employs a small-ring organophosphorus-based catalyst and a hydrosilane reductant to drive the conversion of ketoximes and pyridine-N-oxides into 2-amidopyridines through sequential Beckmann rearrangement followed by [2,3]-sigmatropic rearrangement. The readily available ketoximes could be activated to nitrilium ions in PIII/PVredox catalysis and could efficiently participate in the domino reaction of pyridine-N-oxides, thus providing various substituted 2-amidopyridines in moderate to excellent yields. This presented strategy features excellent functional group tolerance and a broad substrate scope.

The spatiotemporal extension/expansion of mine areas is affected by multiple factors. So far, very little has been done to examine the interaction between mine areas and political or economic realities. The (ultra‐)mafic magmatic mines in China played a specific role in supporting national development and providing an ideal research subject for monitoring their interrelationship. In this study, remote sensing and mining‐related GIS data were used to identify and analyze 1233 (ultra‐)mafic magmatic mine area polygons in China, which covered approximately 322.96 km2of land and included a V–Ti–Fe mine, a copper–nickel mine, a chromite mine, an asbestos mine, and a diamond mine. It was found that (1) the areal expansion of mines is significantly related to the mine types, perimeter, topography, and population density. (2) The mine area variation also reflects market and policy realities. The temporal expansion of the mine area from 2010 to 2020 followed an S‐shaped pattern (with the turning point occurring in 2014), closely related to iron overcapacity and tightened mining policies. (3) The complexity (D) of the mine area may reflect mine design and excavation practices. To be specific, lower Dindicates early‐stage or artisanal/small‐scale mining, whereas higher Drepresents large‐scale mining. This study demonstrates that the detailed mapping of mine land can serve as an indicator to implement mining‐related market and policy changes. The (ultra‐)mafic mines area data set can be accessed at https://zenodo.org/record/7636616#.Y-p0uXaZOa0. A data set of China's (ultra‐)mafic magmatic mine was presented, including 1233 polygons. The spatiotemporal extension of mines is conceptualized to mirror the mining‐related economic and political realities in China. A data set of China's (ultra‐)mafic magmatic mine was presented, including 1233 polygons.The spatiotemporal extension of mines was conceptualized to mirror the mining‐related economic and political realities in China. A data set of China's (ultra‐)mafic magmatic mine was presented, including 1233 polygons. The spatiotemporal extension of mines was conceptualized to mirror the mining‐related economic and political realities in China.

Deep learning accelerators (DLAs) based on compute-in-memory (CIM) technologies have been considered promising candidates to drastically improve the throughput and energy efficiency for running deep neural network models. In this review, we analyze DLA designs reported in the past decade, including both fully digital DLAs and analog CIM based DLAs, to provide insights regarding the current status of CIM technologies and prospective of this emerging field. We observed that the reported CIM designs, even in their early research stage, do provide energy efficiency advantages from measured silicon data over digital DLA. Additionally, it is revealed that the main advantage comes from completely eliminating the run-time DRAM access for weights. For performance benchmarks, we performed a top-down analysis using a generic DLA design and illustrated how fully-weight-stationary CIM DLAs, being no longer bounded by the memory bottleneck, offer large throughput advantages compared to traditional digital DLAs. The benchmark was performed by computing popular models deployed in Google’s TPU accelerator.

Iron plaque, as a natural barrier between rice and soil, can reduce the accumulation of pollutants in rice by adsorption, contributing to the safe production of rice in contaminated soil. In this study, we unveiled a new role of iron plaque, i.e., producing hydroxyl radicals (·OH) by activating root-secreted oxygen to degrade pollutants. The ·OH was produced on the iron plaque surface and then diffused to the interfacial layer between the surface and the rhizosphere environment. The iron plaque activated oxygen via a successive three-electron transfer to produce ·OH, involving superoxide and hydrogen peroxide as the intermediates. The structural Fe(II) in iron plaque played a dominant role in activating oxygen rather than the adsorbed Fe(II), since the structural Fe(II) was thermodynamically more favorable for oxygen activation. The oxygen vacancies accompanied by the structural Fe(II) played an important role in oxygen activation to produce ·OH. The interfacial ·OH selectively degraded rhizosphere pollutants that could be adsorbed onto the iron plaque and was less affected by the rhizosphere environments than the free ·OH. This study uncovered the oxidative role of iron plaque mediated by its produced ·OH, reshaping our understanding of the role of iron plaque as a barrier for rice.

Herein, the de novosynthesis of atropisomeric benzofurans through the Cu/spirocyclic-pyrrolidine-oxazoline (SPDO) complex catalyze asymmetric formal [3 + 2] cycloaddition between quinones and ynamides is explored, producing a range of structurally diverse axially chiral benzofurans in satisfactory yields with good to excellent enantioselectivities. Furthermore, density functional theory calculations were applied to explain the possible reaction mechanism.

Autonomous manipulation using robot hands can benefit from tactile sensing, as it can collect information on variations in applied force and surface properties. This article presents a capacitive tactile sensor placed on the robot’s hand fingers. Due to its unique structure and high sensitivity to material permittivity, this sensing system can obtain capacitive data both when a robot finger is approaching an object and when it has touched the object. With three-dimensional reduction methods, that is, principal component analysis (PCA), independent component analysis (ICA), and multidimensional scaling (MDS), a dataset is transformed to be 2-D and then fed into two supervised classifications algorithms, that is, k-nearest neighbors (KNNs) and support vector machines (SVMs). In comparison to previous studies, the MDS-based SVM achieves high material recognition accuracy, up to 98% for recognition of three different material classes, that is, plastic, paper, and glass using capacitance data only. Furthermore, it performs well in recognition of five different materials, that is, dry plastic, plastic with water drops, paper, dry glass, and glass with water drops. The recognition accuracy is as high as 93%. Computational time can be reduced by about 60% by combining the dimension reduction methods with classification algorithms. The results indicate that different material properties can be identified efficiently using the proposed method.

Efforts to advance the use of analog SRAM compute-in-memory (SRAM-CIM) macros for high-precision multiply-and-accumulate (MAC) operations must deal with issues pertaining to energy efficiency, computing latency (TAC), and area overhead. This brief presents a novel SRAM-CIM structure that utilizes (1) a high input precision computing cell (HIPCC) to perform 8b-MAC operations with high multiplication throughput, and (2) a global bitline-combining (GBL-comb) scheme to improve energy efficiency by reducing the number of analog-to-digital converters (ADCs). A 28nm 384-kb SRAM-CIM macro with 20-bit output precision (near-full precision) was fabricated using a foundry-provided 28nm logic process for MAC operations with 8b-input, 8b-weight, and 16 accumulations. The resulting macro achieved a $T_{\mathrm{ AC}}$ of 3.6 ns with energy efficiency of 14.97 TOPS/W when applied to 8-bit MAC operations.

The production of hydrogen from electrolytic water using renewable energy is a highly effective way to reduce carbon emissions and has garnered significant attention from scholars around the world. As the crucial link between the electrolyzer and the DC bus, the performance of the DC-DC buck converter is paramount. Therefore, a multiphase stacked interleaved buck converter (MSIBC) is proposed in this paper, which offers superior performance in terms of low voltage, high current, and minimal current ripple. This paper delves into the circuit design methods and current ripple elimination mechanisms of this topology. Furthermore, a detailed analysis is conducted to compare the performance of MSIBC with that of interleaved buck converter (IBC) and stacked interleaved buck converter (SIBC), covering factors such as efficiency, inductor current stress, and switching current stress. The results indicate that MSIBC offers superior efficiency, low inductor current stress, and switching current stress. Additionally, MSIBC enables redundant control of the system. Finally, experimental and MATLAB/SIMULINK simulation results confirm the accuracy of ripple cancellation.

Computing-in-memory (CIM) chips have demonstrated the potential high energy efficiency for low-power neural network (NN) processors. Even with energy-efficient CIM macros, the existing system-level CIM chips still lack deep exploration on sparsity and large models, which prevents a higher system energy efficiency. This work presents a CIM NN processor with more sufficient support of sparsity and higher utilization rate. Three key innovations are proposed. First, a set-associate blockwise sparsity strategy is designed, which simultaneously saves execution time, power, and storage space. Second, a ping-pong weight update mechanism is proposed for a higher utilization rate, enabling simultaneous execution of CIM and write operations. Third, an efficient CIM macro is implemented with adaptive analog-digital converter (ADC) precision for better sparsity utilization and performance-accuracy trade-off. The 65-nm fabricated chip shows 9.5-TOPS/W system energy efficiency at 4-bit precision, with 6.25 $\times $ actual improvement compared with a state-of-the-art CIM chip. Besides, this work supports high CIM execution accuracy on the ImageNet dataset.

Nonvolatile near-memory-compute (nvNMC) macros are promising candidates for edge artificial intelligence (AI) devices requiring high energy efficiency, short wakeup-to-compute latency, and robust inference accuracy with high precision of inputs (IN), weights ( $W$ ), and outputs (OUT). Nonetheless, the practical application of nvNMC macros is hindered by inherent design challenges: 1) high energy consumption in reading repetitious weight data, 2) low energy efficiency due to high bitstream toggling rate in digital multiply-and-accumulate (MAC) circuits, 3) narrow signal margin and high memory readout latency, and 4) redundant MAC computation in digital MAC circuits under input-stationary flow. In addressing these challenges, we developed a number of schemes based on the concept of system-circuit co-design, including 1) a weight-feature aware read (WFAR) scheme; 2) a toggling-aware weight-tuning (TAWT) scheme; 3) a differential charge-accumulating margin-enhanced voltage-sensing amplifier (DCME-VSA); and 4) an input-sparsity aware pre-computing unit (ISAPU). An 8-Mb spin-transfer torque magnetic random access memory (STT-MRAM) nvNMC macro fabricated using foundry-provided 22 nm STT-MRAM achieved read bandwidth of 436 GB/s, MAC computing latency of 20 ns, and energy efficiency of 53.6–190.2 TOPS/W when performing eight-bit input and eight-bit weight MAC operations with 26-bit output and 576 accumulations.

On-chip non-volatile compute-in-memory (nvCIM) enables artificial intelligence (AI)-edge processors to perform multiply-and-accumulate (MAC) operations while enabling the non-volatile storage of weight data in power-off mode to enhance energy efficiency. However, the design challenges of nvCIM-based AI-edge processors include: 1) lack of a nvCIM-friendly computing flow; 2) a tradeoff between usage of memory devices versus process variations, computing yield and area overhead; 3) long computing latency and low energy efficiency; and 4) small-signal margin and large bitline current. This article presents an nvCIM-friendly AI-edge processor that uses a hybrid-mode resistive random access memory nvCIM (hmRe-nvCIM) macro to overcome the abovementioned challenges by three processor-level schemes: 1) a multimode nvCIM engine controller (mmCIM-EC); 2) a bitwise-input-sparsity and place-value-aware dynamic accumulation (BIS-PVA-DA); and 3) a bitwise weight column inversion (BWCI) and two macro-level schemes: 1) a dynamic-accumulation-aware current quantization (DACQ) and 2) a current–voltage-hybrid analog-to-digital converter (CVH-ADC). The proposed AI-edge processor fabricated using 22-nm technology achieved 51.4 TOPS/W and 472.7- $\mu \text{s}$ wake-up to response time, while the hmRe-nvCIM macro achieved 67.2 TOPS/W under 8-bit input, 8-bit weight, and 22- or 24-bit output precision.

Accurate identification of the target and tracking it at high speeds using drone-mounted cameras and compute hardware finds military and commercial applications. Conventional frame-based cameras and convolutional neural networks (CNNs) extract detailed spatial information to show high accuracy but suffer from lower throughput caused by large models. Alternatively, event cameras capture the motion information as an asynchronous event stream with high temporal resolution. Spiking neural networks (SNNs) can be used to process these data at high speed, but the sparse sensing and difficulty in training SNN limit the accuracy. Fusing the complementary spatial and temporal advantages of the frame and event-based pipelines allows high-speed identification and tracking while preserving accuracy. The SNN processes the event stream continuously to provide high-speed target estimates with lower accuracy, while periodic anchors provided by the reliable CNN restore the accuracy. In this work, we present a heterogeneous programmable ARM Cortex-based system-on-a-chip (SoC) in 40-nm Taiwan Semiconductor Manufacturing Company (TSMC) ultra low power (ULP) technology with power-efficient RRAM compute-in-memory (CIM) for CNN and high-speed SRAM compute-near-memory (CNM) for SNN for the modality-matched acceleration of the hybrid vision. Our SoC incorporates: 1) two levels of power gating to save 91.8% of total chip power with non-volatile RRAM-CIM; 2) embedded triple error correction (TEC) within RRAM CIM macro to suppress the raw bit errors in reading by >5 orders of magnitude; and 3) parallelly operating CNN and SNN modules to provide >100 outputs/s. Such cross-layer hybrid approaches can mitigate fundamental tradeoffs in sensing and processing.

Advanced artificial intelligence edge devices are expected to support floating-point (FP) multiply and accumulation operations while ensuring high energy efficiency and high inference accuracy. This work presents an FP compute-in-memory (CIM) macro that exploits the advantages of computing in the time, digital, and analog-voltage domain for high energy efficiency and accuracy. This work employs: 1) a hybrid-domain macrostructure to enable the computation of both the exponent and mantissa within the same CIM macro; 2) a time-domain computing scheme for energy-efficient exponent computation; 3) a product-exponent-based input-mantissa alignment scheme to enable the accumulation of the product mantissa in the same column; and 4) a place-value-dependent digital–analog-hybrid computing scheme to enable energy-efficient mantissa computations of sufficient accuracy. A 22-nm 832-kB FP-CIM macro fabricated using foundry-provided compact 6T-static random access memory (SRAM) cells achieved a high energy efficiency of 72.14 tera-floating-point operations per second (TFLOPS)/W while performing FP-multiply-and-accumulate (MAC) operations involving BF16-input, BF16-weight, FP32-output, and 128 accumulations.

Compute-in-Memory (CIM) implemented with Resistive-Random-Access-Memory (RRAM) crossbars is a promising approach for accelerating Convolutional Neural Network (CNN) computations. The growing size in the number of parameters in state-of-the-art CNN models, however, creates challenge for on-chip weight storage for CIM implementations, and CNN compression becomes a crucial topic of exploration. Tensor Train (TT) decomposition can be used to decompose a tensor into smaller ones with fewer parameters, at the cost of increased number of computations. In this work we propose a technique to minimize intermediate operations across the full convolution operation and improve hardware utilization to implement TT-CNNs in CIM systems. We first use an iterative decompose-and-fine-tune method to prepare TT-CNNs. We then propose an inter-convolutional-step reuse scheme to reduce the required operation count and post-mapping RRAM count for TT-CNN implementation in tiled-CIM architecture. We demonstrate that through proper mapping, pipelining, and reuse, effective compression ratio of 12 and 20 with 0.8% and 1.4% accuracy drop, respectively for WRN; and effective compression ratio of 6 and 11 with 0.9% and 1.2% accuracy drop for VGG8. We also show that around 30% higher hardware utilization than the original CNN format can be achieved using the proposed TT-CIM approaches.

Due to the characteristic differences of decision makers (DMs), such as growth experience, educational background, social status, and personal beliefs, individual decision information often varies. It becomes more prominent in large-scale group decision making (LSGDM). Therefore, consensus discussion is necessary to make the final decision result represent all DMs’ opinions as much as possible. In this regard, this article analyzes the conflicts between individual and subgroup adjustments and between subgroup and group adjustments by the built models. Concerning intrasubgroup adjustment surplus and intersubgroup adjustment surplus, we regard them as cost allocation problems in cooperative games under the consensus constraint. Then, the two-stage Nash-bargaining consensus adjustment game (NBCAG) is introduced to make the allocation result as fair as possible. Meanwhile, the Nash-bargaining consensus adjustment scheme for LSGDM is proposed. Considering the heterogeneity of DMs and subgroups, two-stage asymmetrical NBCAGs are further constructed. In light of these results, a new LSGDM method is proposed. Numerical and comparative analysis is also performed. This article offers the first LSGDM method that emphasizes the fairness and Pareto optimality of consensus adjustment by two-stage Nash-bargaining games.

Designing compact and energy-efficient resistive RAM (RRAM) macros is challenging due to: 1) large read/write circuits that decrease storage density; 2) low-conductance cells that increase read latency; and 3) the pronounced effects of routing parasitics on high-conductance cell read energy. Multiple-bits-per-cell RRAM can boost storage density but has further challenges resulting from reliability problems due to conductance relaxation and slow write due to narrow conductance levels. This work presents a multiple-bits-per-cell RRAM macro called Efficient Multiple-Bits-per-Cell Embedded RRAM (EMBER), which: 1) demonstrates read/write circuit compaction through constrained optimization of driver and pass gate transistor sizes; 2) introduces a common-mode bleed conductance at the sense amplifier inputs, reducing read settling time by $11.35\times $ for low-conductance cells, and 3) cuts read path capacitance to further reduce read access time and energy. To address reliability and write speed, EMBER contains a configurable on -chip read/write controller. We present a level allocation scheme that uses array-level characterization data to find sufficiently reliable allocations, while simultaneously maximizing write bandwidth. EMBER is the first embedded RRAM storage macro to achieve fully integrated multiple-bits-per-cell readout and write-verification without any off -chip reference generation or sensing. The macro operates at 100 MHz with 64k $\times $ 48 = 3 M cells in TSMC 40-nm CMOS, achieving 1 b/cell read operation with 1.0 pJ/bit energy at 2.4 Gbps, and 2 b/cell read with 1.1 pJ/bit at 1.6 Gbps. 1 b/cell write-verify operates with 0.40 nJ/bit energy at 12.4 Mbps (BER < 6e-4), and 2 b/cell write-verify operates with 1.2 nJ/bit at 3.8 Mbps (BER < 3e-3). The array-level endurance is found to be 10 K for 1–2 b/cell. Normalizing for process scaling, the macro demonstrates the highest effective RRAM cell density to date of 5.6e-3 b/ $\text{F}^{2}$ for 1 b/cell and 1.3e-2 b/ $\text{F}^{2}$ for 2 b/cell, an improvement of 31% and 204%, respectively, over the best prior work.

This article presents a novel static random access memory computing-in-memory (SRAM-CIM) structure designed for high-precision multiply-and-accumulate (MAC) operations with high energy efficiency (EF), high readout accuracy, and short compute latency. The proposed device employs 1) a time-domain incremental-accumulation (TDIA) scheme to enable high-accumulation MAC operations while maintaining a large signal margin across MAC values (MACVs), 2) a dynamic differential-reference (D2REF) scheme based on software-hardware co-design to reduce read energy consumption, and 3) a low-dMACV-aware recursive time-to-digital converter (LMAR-TDC) for implementation with the D2REF scheme to further suppress readout energy consumption. A 28 nm 1 Mb SRAM-CIM macro fabricated using foundry-provided compact 6T-SRAM cells achieved EF of 39.31 TOPS/W and compute latency of 6.6 ns for 8b-MAC operations with 64 accumulations per cycle and near-full output precision (22b).

Convolutional neural networks (CNNs) play a key role in many deep learning applications; however, these networks are resource intensive. The parallel computing ability of computing-in-memory (CIM) enables high energy efficiency in artificial intelligence accelerators. When implementing a CNN in CIM, quantization and pruning are indispensable for reducing the calculation complexity and improving the efficiency of hardware calculations. Mixed-precision quantization with flexible bit widths provides a better efficiency-accuracy tradeoff than fixed-precision quantization. However, CIM calculations for mixed-precision models are inefficient because the fixed capacity of CIM macros is redundant for hybrid precision distributions. To address this, we propose a software and hardware co-design static random-access memory (SRAM)-based CIM architecture called SUN, including a CIM-adaptive mixed precision joint pruning quantization algorithm and dynamic hybrid precision CNN accelerator. Three techniques are implemented in this architecture: 1) a mixed precision joint pruning algorithm for reducing the memory access and removing the redundant computing; 2) a CIM-adaptive filter-wise and paired mixed-precision quantization for improving CIM macro utilization; and 3) an SRAM-based CIM CNN accelerator in which the SRAM CIM macro is used as the processing element to support sparse and mixed-precision CNN computation with high CIM macro utilization. This architecture achieves a system area efficiency of 428.2 TOPS/mm 2 and throughput of 792.2 GOPS on the CIFAR-10 dataset.

Reperfusion injury, which is distinct from ischaemic injury, occurs when blood flow is restored in previously ischaemic brain tissue, further compromising neurons and other cells and worsening the injury. There is currently a lack of pharmaceutical agents and therapeutic interventions that specifically mitigate cerebral ischaemia/reperfusion (I/R) injury. Ginsenoside Rg1 (Rg1), a protopanaxatriol-type saponin isolated from Panax ginsengC. A. Meyer, has been found to protect against cerebral I/R injury, but its intricate protective mechanisms remain to be elucidated. Numerous studies have shown that autophagy plays a crucial role in protecting brain tissue during the I/R process and is emerging as a promising therapeutic strategy for effective treatment. In this study, we investigated whether Rg1 protected against I/R damage in vitro and in vivo by regulating autophagy. Both MCAO and OGD/R models were established. SK-N-AS and SH-SY5Y cells were subjected to OGD followed by reperfusion with Rg1 (4–32 μM). MCAO mice were injected with Rg1 (30 mg·kg−1·d−1. i.p.) for 3 days before and on the day of surgery. Rg1 treatment significantly mitigated ischaemia/reperfusion injury both in vitro and in vivo. Furthermore, we demonstrated that the induction of autophagy contributed to I/R injury, which was effectively inhibited by Rg1 in both in vitro and in vivo models of cerebral I/R injury. Rg1 inhibited autophagy through multiple steps, including impeding autophagy initiation, inducing lysosomal dysfunction and inhibiting cathepsin enzyme activities. We revealed that mTOR activation was pivotal in mediating the inhibitory effect of Rg1 on autophagy. Treatment with Torin-1, an autophagy inducer and mTOR-specific inhibitor, significantly reversed the impact of Rg1 on autophagy, decreasing its protective efficacy against I/R injury both in vitro and in vivo. In conclusion, our results suggest that Rg1 may serve as a promising drug candidate against cerebral I/R injury by inhibiting autophagy through activation of mTOR signalling.

The microstructure evolution and mechanical properties of a Fe–0.12C–0.2Si–1.6Mn–0.3Cr–0.0025B (wt.%) steel with different initial microstructures, i.e., hot rolled (HR) and cold rolled–annealed (CRA), were studied through optical microscopy, scanning electron microscopy, electron channeling contrast imaging, microhardness and room temperature uniaxial tensile tests. After water quenching from 930 °C to room temperature, a fully martensitic microstructure was obtained in both as-quenched HR and CRA specimens, which shows a microhardness of 480 ± 5 HV, and no significant difference in microstructure and microhardness was observed. Tensile test results show that the product of tensile strength and total elongation (UTS × TE) of the as-quenched HR specimen, i.e., 24.1 GPa%, is higher than that of the as-quenched CRA specimen, i.e., 18.9 GPa%. While, after being tempered at 300 °C, the martensitic microstructures and mechanical properties of the two as-quenched specimens change significantly due to the synergy role of the matrix phase softening and the precipitation strengthening. Concerning the maximum UTS × TE, it is 18.9 GPa% obtained in the as-quenched CRA one, while that is 24.4 GPa% obtained in the HR specimen after tempered at 300 °C for 5 min.

Compute-in-memory (CIM) has recently emerged as a promising design paradigm to accelerate deep neural network (DNN) processing. Continuously better energy and area efficiency at the macrolevel had been reported through many testchips over the last few years. However, in those macro design-oriented studies, accelerator-level considerations, such as memory accesses and processing of entire DNN workloads have not been investigated in-depth. In this article, we aim to fill this gap starting with the characteristics of our latest CIM macro fabricated with cutting-edge FinFET CMOS technology at 4-nm node. We then study, through an accelerator simulator developed in-house, three key items that would determine the efficiency of our CIM macro in the accelerator context while running MLPerf Mobile suite: 1) dataflow optimization; 2) optimal selection of CIM macro dimensions to further improve macro utilization; and 3) optimal combination of multiple CIM macros. Although there is typically a stark contrast between macro-level peak and accelerator-level average throughput and energy efficiency, the aforementioned optimizations are shown to improve the macro utilization by $3.04\times $ and reduce the energy-delay product (EDP) to $0.34\times $ compared to the original macro on MLPerf Mobile inference workloads. While we exploit a digital CIM macro in this study, the findings and proposed methods remain valid for other types of CIM (such as analog CIM and analog–digital–hybrid CIM) as well.

Double-walled carbon nanotubes (DWCNTs) make up a unique class of carbon nanotubes (CNTs) that are particularly intriguing for scientific research and are promising candidates for technological applications. A more precise level of control and greater yields can be achieved via catalytic chemical vapor deposition (CCVD), which involves the breakdown of a carbonaceous gas over nanoparticles. The addition of molybdenum to the system can increase the selectivity with regard to the number of walls that exist in the obtained CNTs. As reported herein, we have designed and synthesized a novel Co–Mo-MOF, [Co(3-bpta)1.5(MoO4)]·H2O (where 3-bpta = N,N′-bis(3-pyridyl)terephthalamide), and employed the Co–Mo-MOFas a bimetallic catalyst precursor for the CCVD approach to prepare high-quality DWCNTs. The Co–Mo-MOFwas employed after being calcined in N2and H2at 1100 °C and decomposing into CoO, CoMoO4, and MoO3. Existing CoMoO4is unaltered after reduction in H2at 1100 °C, while CoO and MoO3are converted into Co0and MoO2, and more CoMoO4is created at the expense of Co0and MoO2without clearly defining agglomeration. Finally, the interaction between metallic Co particles and C2H4is what initiates the formation of DWCNTs. In-depth discussion is provided in this paper regarding the mechanism underlying the high selectivity and activity of Co–Mo catalysts in regulating the development and structure of DWCNTs. The DWCNTs also offer excellence performance when they are used as water purification agents and as selective sorbents. This work opens a feasible way to use MOFs as a way to produce MWCNTs, thus blazing a new trail in the field of MOF-derived carbon-based materials.

This article presents an intelligent vision sensor (IVS) with embedded tiny convolutional neural network (CNN) model and programmable processing-in-sensor (PIS) circuit for real-time inference applications of low-power edge devices. The proposed imager realizes the full computing functions of a customized three-layers tiny network, which includes a $3 \times 3$ convolution layer (stride $=$ 3) with activation function of rectified linear unit (ReLU), a $2 \times 2$ maximum pooling (MP) layer (stride $=$ 2), and a $1 \times 1$ fully connected (FC) layer for inference. A 0.8 V $128 \times 128$ IVS prototype was fabricated and verified in TSMC 0.18 $\mu \text{m}$ standard CMOS technology. In normal image mode, it consumed 76.4 $\mu \text{W}$ with full-resolution ( $126 \times 126$ active resolution) image output at 125 f/s. In CNN mode, it consumed 134.5 $\mu \text{W}$ at 250 f/s and an achieved iFoMs of 33.8 pJ/pixel $\cdot $ frame. Using the proposed mixed-mode PIS circuits, the prototype is configured to demonstrate a “human face or not detection” task with an achieved accuracy of 93.6%.

A near-αTi−7Al−2V alloy was designed using cluster formula approach and prepared by laser additive manufacturing, whose specific strength is better than that of Ti−6Al−4V alloy. Its composition formula α-{[Al−Ti12](AlTi2)}15+β-{[Al−Ti14](V3)}2features an enhancedα-Ti by increasing αunit proportion of 15/17 (with respect to 12/17 of Ti−6Al−4V alloy) and stabilized β-Ti via V alloying. This alloy possesses a good laseradditivemanufacturing processibility. At the as-deposited state, the microstructures are composed of fine basket-weave regions, coarse basket-weave regions and ultrafine αWidmanstättenlath regions. The surface roughness of coarse basket-weave regions is much smaller than that of fine basket-weave regions and ultrafine αWidmanstättenlath regions. The α-phase distribution in fine basket-weave regions is more uniform than that in the coarse basket-weave regions. Its ultimate tensile strength of 971−1005 MPa, yield strengthof 891−921 MPa andelongation of 4.5%−6.6% are close to those of Ti−6Al−4V alloy, and particularly, its specific strength of 224−232 kN·m/kg is better than that of Ti−6Al−4V alloy.

Shield tunneling and post-tunneling steady seepage are accompanied by stress and displacement variations, which could induce and influence the soil arching effect. Although there are many studies on the tunneling-induced soil arching effect, the research about the effect of seepage on soil arching effect is extremely lacking. In this study, a numerical model is firstly established and verified by field data. Then, a series of numerical models, whose simulation method of steady seepage is verified by adopting the conformal mapping technique, are established to study the soil arching evolution of deep-buried tunneling and post-construction steady seepage. The results indicate that seepage leads to an increase in effective vertical stress, which is consistent with the existing theory. The seepage weakens the soil arching effect resulting in the height of the arch zone reducing from 2.38D(Dis the tunnel diameter) to 1.25D. The seepage leads to the further development of ground consolidation settlement, but the differential displacement in the soil mass decreases. The ground reaction curve in the steady seepage condition shows a bigger value than that after excavation. It is reasonable to control the ground loss ratio in the range of 0.5–1.0%, which can minimize overburden pressure with moderate ground deformation.

Chalkiness is one of the most important agronomic traits in rice breeding, which directly affects the quality of rice seed. In this study, we identified a chalkiness endosperm mutant, chalk-h, from N-methyl-N-nitrosourea (MNU)-induced japonicarice cultivar Hwacheong (HC). Compared with wild type (WT)-HC, chalk-hshowed severe chalkiness in the endosperm, yellowish green leaves, as well as reduced plant height. Scanning electron microscopy (SEM) analysis showed that starch grains in the chalk-hmutant were irregular in size and loosely arranged, with large gaps between granules, forming ovoid or orbicular shapes. MutMap analysis revealed that the phenotype of chalk-his controlled by a single recessive gene LOC_Os11g39670encoding seryl-tRNA synthetase, which is renamed as CHALK-H. A point mutation occurs in chalk-hon the sixth exon (at nucleotide 791) of CHALK-H, in which adenine (A) is replaced by thymidine (T), resulting in an amino acid codon change from glutamine (Glu) to valine (Val). The chalk-hmutant exhibited a heat-sensitive phenotype from the 3-leaf stage, including yellow-green leaves and reduced pigment content. The transcriptional expression of starch synthesis-related genes was down-regulated in the chalk-hmutants compared to WT-HC at different grain-filling stages. With an increase in temperature, the expression of photosynthesis-related genes was down-regulated in the chalk-hmutant compared to WT-HC. Overexpression of CHALK-Hrescued the phenotype of chalk-h, with endosperm and leaf color similar to those of WT-HC. Our findings reveal that CHALK-His a causative gene controlling chalkiness and leaf color of the chalk-hmutant. CHALK-His the same gene locus as TSCD11, which was reported to be involved in chloroplast development under high temperature. We suggest that CHALK-H/TSCD11plays important roles not only in chloroplast development, but also in photosynthesis and starch synthesis during rice growth and development, so it has great application potential in rice breeding for high quality and yield.

As a global pollutant, mercury emission is increasingly restricted in recent years. It is urgent to explore a new and efficient mercury removal technology for coal-fired power plants. A new acid-assisted electrochemical oxidation (AEO) technique for mercury removal was proposed using platinum plate as cathode and fluorine-doped tin dioxide (FTO) glass as anode. The effects of acid type, acid concentration, applied direct current (DC) voltage, electrolyte type, SO2, NO and O2on the Hg0removal efficiency were carried out. The results indicated that the mercury removal efficiency increased with the increase of DC voltage and nitric acid concentration. When the concentration of nitric acid increased to 0.15 mol/L, the mercury removal efficiency remained unchanged. SO2and NO inhibited the removal of Hg0in AEO system, but the inhibition was reversible. Compared with the mercury removal efficiency under single experimental conditions, the mercury removal efficiency of electrochemical oxidation can reach 96% under the experimental conditions of 0.1 mol/L nitric acid and 4V DC voltage, suggesting that the synergistic effect of nitric acid and DC voltage plays a key role. According to the experimental results, the mechanism of Hg0removal in AEO system was analyzed. At the anode, Hg0was oxidized by hydroxyl radical (.OH) generated by the oxidation reaction on the anode surface. At the cathode, dissolved oxygen or O2adsorbed on the surface of Pt is reduced to form anionic superoxide radicals (.O−2). Moreover, parts of .O−2would produce .OH with the aid of electron at acidic condition. Free radicals capture experiments showed that .O−2and .OH were the main active substances for the removal of Hg0by acid-assisted electrochemical method. The research is helpful for the development of effective electrochemical techniques for industrial mercury removal and recycling of industrial acid waste.

This article presents a 0.56/0.8 V multimode vision sensor for temporal and spatial information extraction with processing in sensor (PIS) technique. The proposed six transistors and one capacitor (6T1C) temporal contrast pixel (TCP) structure using the exposure compensation scheme (ECS) can realize in-pixel temporal contrast calculation and motion event reporting (ER) with global shutter exposure and seamless frame difference (FD) based on the pulsewidth modulation (PWM) operation. In addition, the column-parallel local binary pattern (LBP) extraction provides further spatial feature information without the power-consuming analog-to-digital conversion to meet the power-efficient requirement for always-on-edge devices. According to the temporal and spatial information extracted from FD and LBP modes, this work realizes the PIS operations for dynamic depth calculation of a stereo vision system, which filters out the static scene and provides the depth information of the dynamic objects only. To further alleviate the data transfer bandwidth, the region of interest (ROI) capability is also embedded in the sensor for data windowing and object locating. The prototyped vision sensor was fabricated in TSMC’s standard 0.18 $\mu \text{m}$ CMOS process and verified. The vision sensor provides five operating modes that can adapt to different tasks, which include the raw image mode (IM), FD mode, ER mode, LBP mode, and ROI mode. The measurement results show a maximum frame rate of 540/819/540 frames/s at a power consumption of 390/162.6/151.9 $\mu \text{W}$ with an achieved iFoM of 45.5/12.5/17.7 pJ/pixel $\cdot $ frame in IM/ER/LBP mode, respectively.

$1.5~\mu $ m optical amplification in a ridge-type Er-doped LiNbO3-on-insulator (Er:LNOI) photonic wire (PW) pumped at 1480 nm wavelength is studied theoretically. The study considers a three-level Er3+ model with inclusion of excited-state absorption and cooperative upconversion. Based on the model, steady-state rate equations and pump/signal power propagation equations have been established, and closed expression was derived for population density of each level involved. Quantitative relations of population inversion, signal gain or gain coefficient to propagation distance, pump power, scattering loss of pump/signal wave and Er3+ concentration are established. In comparison with Ti-diffused Er:LiNbO3 (Ti:Er:LN) waveguide, the Er:LNOI PW exhibits much more efficient gain performance because of ultra-compact mode field and hence ultra-high light intensity (photon density) in it. These include two orders of magnitude lower threshold pump power, tens of times lower saturation pump power and three times higher saturation gain. Due to the same reason, the $^{4}\text{I}_{11/2}$ population of Er3+ in the Er:LNOI PW is four times higher than that in the Ti:Er:LN waveguide. Consequently, the 1480 nm upconversion process of Er3+ in the Er:LNOI PW is much stronger than that in the Ti:Er:LN. It strengthens at higher pump power and definitely affects the $1.5~\mu \text{m}$ gain at higher pump level. Combined effects of ultra-high pump intensity, upconversion process and large non-radiative multi-phonon relaxation rate from $^{4}\text{I}_{11/2}$ to $^{4}\text{I}_{13/2}$ level result in some interesting features, including non-monotonic variation of gain coefficient with propagation distance, slight drop of gain in saturation region and gain increase with pump loss at higher pump power. These combined effects make it possible to realize $2.7~\mu \text{m}$ lasing and an output power of sub-mW is theoretically expected for a 1480 nm pump power of 100 mW. The study also shows that amplified spontaneous emission is significant in the Er:LNOI PW but affects little the signal gain performance.

To investigate the influence of the lateral inflow on the vortical structures and hydraulic losses inside the mixed-flow pump, entropy production, and Q-criterion were applied to numerical simulation analysis of transient flow. Three inflow runner schemes were designed to compare the effects of the middle-isolated pier and rear wall of the “ω” structure on restraining the occurrence and transmission of a vortex and stabilizing the internal flow state of the pump device. The results showed that when the inflow condition transformed from parallel inflow to lateral inflow, the design efficiency and head deviated toward the over-load flow rate, and the maximum deviations were obtained under the over-load flow rate, reaching 4.79% and 50.35%, respectively. The hydraulic losses and vortical structures increased to a large extent, influenced by the lateral inflow, particularly the flow region near the impeller hub. The setting of the vortex suppression facilities effectively reduced the deviation ratios in the design efficiency and head, which decreased to 1.31% and 7.59%, respectively. The vorticity intensity and entropy dissipation rate were also significantly decreased. By contrast, the rear wall of the “ω” structure had a greater stabilization effect on the internal flow patterns than that of the middle-isolated pier.