Visualizing and Regulating Dynamic Evolution of Interfacial Electrolyte Configuration during De-solvation Process on Lithium-Metal Anode.

Acting as a passive protective layer, solid-electrolyte interphase (SEI) plays a crucial role in maintaining the stability of the Li-metal anode. Derived from the reductive decomposition of electrolytes (e.g., anion and solvent), the SEI construction presents as an interfacial process accompanied by the dynamic de-solvation process during Li-metal plating. However, typical electrolyte engineering and related SEI modification strategies always ignore the dynamic evolution of electrolyte configuration at the Li/electrolyte interface, which essentially determines the SEI architecture. Herein, by employing advanced electrochemical in situ FT-IR and MRI technologies, we directly visualize the dynamic variations of solvation environments involving Li+-solvent/anion. Remarkably, a weakened Li+-solvent interaction and anion-lean interfacial electrolyte configuration have been synchronously revealed, which is difficult for the fabrication of anion-derived SEI layer. Moreover, as a simple electrochemical regulation strategy, pulse protocol was introduced to effectively restore the interfacial anion concentration, resulting in an enhanced LiF-rich SEI layer and improved Li-metal plating/stripping reversibility.

Compact gas cell for simultaneous detection of atmospheric aerosol optical properties based on photoacoustic spectroscopy and integrating sphere scattering enhancement.

Atmospheric aerosols play a pivotal role in the earth-atmospheric system. Analyzing their optical properties, specifically absorption and scattering coefficients, is essential for comprehending the impact of aerosols on climate. When different optical properties of aerosols are individually measured using multiple devices, cumulative errors in the detection results inevitably occur. To address this challenge, based on photoacoustic spectroscopy (PAS) and integrating sphere (IS) scattering enhancement, a compact gas cell (PASIS-Cell) was developed. The PASIS-Cell comprises a dual-T-type photoacoustic cell (DTPAC) and an IS. IS is coupled with DTPAC through a transparent quartz tube, thereby enhancing the scattering signal without compromising the acoustic characteristics of DTPAC. Concurrently, DTPAC can realize high-performance photoacoustic detection of absorption signal. Experimental results demonstrate that PASIS-Cell can simultaneously invert atmospheric aerosol absorption and scattering coefficients, with a minimum detection limit of less than 1 Mm-1, showcasing its potential in the analysis of aerosol optical properties.

Kinetics Conditioning of (Electro) Chemically Stable Zn Anode with pH Regulation Toward Long-Life Zn-Storage Devices.

The safety, low cost, and high power density of aqueous Zn-based devices (AZDs) appeal to large-scale energy storage. Yet, the presence of hydrogen evolution reaction (HER) and chemical corrosion in the AZDs leads to local OH- concentration increasement and the formation of Znx SOy (OH)z •nH2 O (ZHS) by-products at the Zn/electrolyte interface, causing instability and irreversibility of the Zn-anodes. Here, a strategy is proposed to regulate OH- by introducing a bio-sourced/renewable polypeptide (ɛ-PL) as a pH regulator in electrolyte. The consumption of OH- species is evaluated through in vitro titration and cell in vivo in situ attenuated total reflection surface-enhanced infrared absorption spectroscopy at a macroscopic and molecular level. The introduction of ɛ-PL is found to significantly suppress the formation of ZHS and associated side reactions, and reduce the local coordinated H2 O of the Zn2+ solvation shell, widening electrochemical stable window and suppressing OH- generation during HER. As a result, the inclusion of ɛ-PL improves the cycle time of Zn/Zn symmetrical cells from 15 to 225 h and enhances the cycle time of aqueous Zn- I2 cells to 1650 h compared to those with pristine electrolytes. This work highlights the potential of kinetical OH- regulation for by-product and dendrite-free AZDs.

Protecting Li-metal in O2 atmosphere by a sacrificial polymer additive in Li-O2 batteries.

Li-O2 batteries (LOBs) with Li-metal as the anode are characterized by their high theoretical energy density of 3500 W h kg-1 and are thus considered next-generation batteries with an unlimited potential. However, upon cycling in a harsh O2 atmosphere, the poor-quality solid electrolyte interphase (SEI) film formed on the surface of the Li-metal anode cannot effectively suppress the shuttle effect from O2, superoxide species, protons, and soluble side products. These issues lead to aggravated Li-metal corrosion and hinder the practical development of LOBs. In this work, a polyacrylamide-co-polymethyl acrylate (PAMMA) copolymer was innovatively introduced in an ether-based electrolyte as a sacrificial additive. PAMMA was found to preferentially decompose and promote the formation of a dense and Li3N-rich SEI film on the Li-metal surface, which could effectively prohibit the shuttle effect from a series of detrimental species in the Li-O2 cell during the discharge/charge process. Using PAMMA, well-protected Li-metal in a harsh O2 atmosphere and significantly enhanced cycling performance of the Li-O2 cell could be achieved. Thus, the use of a sacrificial polymer additive provides a promising strategy for the effective protection of Li-metal in Li-O2 cells in a severe O2 atmosphere during practical applications.

Full-Dimensional Analysis of Electrolyte Decomposition on Cathode-Electrolyte Interface: Establishing Characterization Paradigm on LiNi0.6Co0.2Mn0.2O2 Cathode with Potential Dependence.

Cathode electrolyte interphase (CEI) layers derived from electrolyte oxidative decomposition can passivate the cathode surface and prevent its direct contact with electrolyte. The inorganics-dominated inner solid electrolyte layer (SEL) and organics-rich outer quasi-solid-electrolyte layer (qSEL) constitute the CEI layer, and both merge at the junction without a clear boundary, which assures the CEI layer with both ionic-conducting and electron-blocking properties. However, the typical "wash-then-test" pattern of characterizations aiming at the microstructure of CEI layers would dissolve the qSEL and even destroy the SEL, leading to an overanalysis of electrolyte decomposition pathway and misassignment of CEI architecture (e.g., component and morphology). In this study, we established a full-dimensional characterization paradigm (combining Fourier transform infrared, solution NMR, X-ray photoelectron spectroscopy, and mass spectrometry technologies) and reconstructed the original CEI layer model. Besides, the feasibility of this characterization paradigm has been verified in a wide operating voltage range on a typical LiNixMnyCozO2 cathode.

Unveiling Atom Migration Abilities Affected Anode Performance of Sodium-Ion Batteries.

In sodium-ion batteries (SIBs), the low initial coulombic efficiency (ICE) is commonly induced by irreversible phase conversion and difficult desodiation, especially on transition metal compounds (TMCs). Yet the underlying physicochemical mechanism of poor reaction reversibility is still a controversial issue. Herein, by using in situ transmission electron microscopy and in situ X-ray diffraction, we demonstrate the irreversible conversion of NiCoP@C is caused by the rapid migration of P in carbon layer and preferential formation of isolated Na3 P during discharge. By modifying the carbon coating layer, the migration of Ni/Co/P atoms is inhibited, thus the improvement of ICE and cycle stability is realized. The inhibiting of fast atom migration which induces component separation and rapid performance degradation might be applied to a wide range of electrode materials, and guides the development of advanced SIBs.

Anomalous Inferior Zn Anode in High-Concentration Electrolyte: Leveraging Solid-Electrolyte-Interface for Stabilized Cycling of Aqueous Zn-Metal Batteries.

Aqueous Zn-metal batteries (AZMBs) are promising large-scale energy storage devices for their high safety and theoretical capacity. However, unstable Zn-electrolyte interface and severe side reactions have excluded AZMBs from long cycling required by practically reversible energy storage. Traditional high-concentration electrolyte is an effective strategy to suppress dendrites growth and resolve the poor electrochemical stability and reversibility of Zn-metal anodes, yet how scientifically universal such strategy is for hybrid electrolyte of different concentrations remains unclear. Herein, we studied the electrochemical behaviors of AZMBs comprising a ZnCl2 -based DMSO/H2 O electrolyte of two distinct concentrations (1 m vs. 7 m). The electrochemical stability/reversibility of Zn anodes in both symmetric and asymmetric cells with high-concentration electrolytes are unusually inferior to the ones with low-concentration electrolyte. It was found that more DMSO components in the solvation sheath of low-concentration electrolyte exist at the Zn-electrolyte interface than in high-concentration counterpart, enabling higher organic compositions in solid-electrolyte-interface (SEI). The rigid inorganic and flexible organic compositions of SEI decomposed from the low-concentration electrolyte is accounted for improved cycling and reversibility of Zn metal anodes and the respective batteries. This work reveals the critical role of SEI than the high concentration itself in delivering stable electrochemical cycling in AZMBs.

Active noise reduction for a differential Helmholtz photoacoustic sensor excited by an intensity-modulated light source.

A highly sensitive differential Helmholtz photoacoustic sensor with active noise reduction was reported. Coupled to one cavity of the photoacoustic cell, an intensity-modulated excitation light would reflect multiple times to produce photoacoustic signal, and meanwhile cause the solid-state photoacoustic effect forming differential mode noise with the frequency same as the photoacoustic signal, which could not be suppressed by conventional differential technology. Wavelength modulation technology is a splendid method to restrain this effect, which is not suitable for light sources with not adjustable wavelength. To suppress this kind of noise, an intensity-modulated compensation light was coupled to another cavity, whose central wavelength was at the non-absorption line of the measured gas. The compensation light was of the same frequency, phase, and power as the excitation light, by which the solid-state photoacoustic effects were produced to form destructive interference called active noise reduction. The experiment results showed that the active noise reduction significantly improved the signal-to-noise ratio and signal-to-background ratio. Compared with the differential, the differential with active noise reduction improved signal-to- noise ratio by about 1.2 times and signal-to-background ratio by about 9.4 times. When low-power near-infrared lasers were employed as the two light sources, the minimum detection limits for acetylene and methane reached 21 and 200 ppb, respectively.

Highly Dispersed Ru-Co Nanoparticles Interfaced With Nitrogen-Doped Carbon Polyhedron for High Efficiency Reversible Li-O2 Battery.

The lithium-oxygen (Li-O2 ) battery with high energy density of 3860 Wh kg-1 represents one of the most promising new secondary batteries for future electric vehicles and mobile electronic devices. However, slow oxygen reduction/oxygen evolution (ORR/OER) reaction efficiency and unstable cycling performance restrain the practical applications of the Li-O2 battery. Herein, Ru-modified nitrogen-doped porous carbon-encapsulated Co nanoparticles (Ru/Co@CoNx -C) are synthesized through reduction of Ru on metal-organic framework (MOFs) pyrolyzed derivatives strategies. Porous carbon polyhedra provide channels for reactive species and stable structure ensures the cyclic stability of the catalyst; abundant Co-Nx sites and high specific surface area (353 m2 g-1 ) provide more catalytically active sites and deposition sites for reaction products. Theoretical calculations further verify that Ru/Co@CoNx -C can regulate the growth of Li2 O2 to improve reversibility of Li-O2 batteries. Li-O2 batteries with Ru/Co@CoNx -C as cathode catalyst achieve small voltage gaps of 1.08 V, exhibit excellent cycle stability (205 cycles), and deliver high discharge specific capacity (17050 mAh g-1 ). Furthermore, pouch-type Li-O2 batteries that maintain stable electrochemical performance output even under conditions of bending deformation and corner cutting are successfully assembled. This study demonstrates Ru/Co@CoNx -C catalyst's great application potential in Li-O2 batteries.

Design of a high-sensitivity differential Helmholtz photoacoustic cell and its application in methane detection.

A high-sensitivity differential Helmholtz photoacoustic cell based on multiple reflection was reported, and its performance parameters and gas replacement time were optimized by finite element simulation. To realize the long absorption path of the measured gas, the collimated excitation light was reflected multiple times on the gold-plated wall of the absorption cavity, and the wavelength modulation technology was used to reduce the multiple reflection noise. Additionally, the differential could suppress external co-phase noise and double the photoacoustic signal. When a laser with a central wavelength of 1653 nm was employed as the excitation light source, the minimum detection limit of 177 ppb (signal-to-noise ratio, SNR = 1) for methane was achieved within a detection time of 1 s, and the corresponding normalized noise equivalent absorption coefficient was 4.1×10-10 cm-1WHZ-1/2.

Chromosome-level Genomes Reveal the Genetic Basis of Descending Dysploidy and Sex Determination in Morus Plants.

Multiple plant lineages have independently evolved sex chromosomes and variable karyotypes to maintain their sessile lifestyles through constant biological innovation. Morus notabilis, a dioecious mulberry species, has the fewest chromosomes among Morus spp., but the genetic basis of sex determination and karyotype evolution in this species has not been identified. In this study, three high-quality genome assemblies were generated for Morus spp. [including dioecious M. notabilis (male and female) and Morus yunnanensis (female)] with genome sizes of 301-329 Mb and were grouped into six pseudochromosomes. Using a combination of genomic approaches, we found that the putative ancestral karyotype of Morus species was close to 14 protochromosomes, and that several chromosome fusion events resulted in descending dysploidy (2n = 2x = 12). We also characterized a ∼ 6.2-Mb sex-determining region on chromosome 3. Four potential male-specific genes, a partially duplicatedDNA helicase gene (named MSDH) and three Ty3_Gypsy long terminal repeat retrotransposons (named MSTG1/2/3), were identified in the Y-linked area and considered to be strong candidate genes for sex determination or differentiation. Population genomic analysis showed that Guangdong accessions in China were genetically similar to Japanese accessions of mulberry. In addition, genomic areas containing selective sweeps that distinguish domesticated mulberry from wild populations in terms of flowering and disease resistance were identified. Our study provides an important genetic resource for sex identification research and molecular breeding in mulberry.

Comparison of fungal communities and nonvolatile flavor components in black Huangjiu formed using different inoculation fermentation methods.

Black Huangjiu (BH) is a traditional alcoholic beverage in China, which is very popular among people. The different methods (simultaneous inoculation, sequential inoculation), were applied to ferment BH in this study, which were investigated the changes in the composition of fungal communities and non-volatile flavor compounds (nVFCs) by high-throughput sequencing (HTS) and ultraperformance liquid chromatography-tandem mass spectrometer (UPLC MS/MS). The results showed that Rhizopus and Saccharomyces were the predominant fungal genera throughout fermentation, and 471 nVFCs were detected in BH after fermentation. Compared to that observed simultaneous inoculation, Rhizopus increased at the end of sequential fermentation, and the contents of the organic acids and their derivatives increased significantly [variable importance in the projection (VIP) > 1.0, p < 0.05, fold change (FC) > 2], while that of lipids and lipid-like molecules decreased significantly (VIP > 1.0, p < 0.05, FC < 0.5). Through the correlation analysis of 32 nVFCs with significant differences (VIP > 1.0, p < 0.05, FC >32 or < 0.03) and the community, it was found that lipids and lipid-like molecules (12) and organic acids and their derivatives (10) were significantly (p < 0.05) negatively correlated with Saccharomyces, but they were significantly (p < 0.05) positively correlated with Rhizopus. Compared with simultaneous inoculation, BH fermented by sequential inoculation, the taste was stronger, sweeter, mellow, and softer. Our findings provide information on nVFC dynamics and will aid in the selection of beneficial strains to improve BH quality.

Bidirectional Interface Protection of a Concentrated Electrolyte, Enabling High-Voltage and Long-Life Aqueous Zn Hybrid-Ion Batteries.

Prussian blue analogues (PBAs) as a promising high-voltage cathode material for aqueous zinc-ion batteries (ZIBs) are usually subjected to an ephemeral lifespan and low Coulombic efficiency due to the irreversible phase change and high Zn2+ insertion potential. Besides, Zn dendrites, H2 evolution reaction, and corrosion derived from a Zn anode interface remain huge challenges. Given this, a highly stable zinc hexacyanoferrate (KZnHCF) cathode together with a mixed concentrated electrolyte is prepared to realize a high-voltage and long-life aqueous ZIB, in which the mixed concentrated electrolyte consisting of 30 m KFSI + 1 m Zn(CF3SO3)2 possesses a unique Zn2+ solvation sheath (Zn(CF3SO3)0.3(FSI)3.1(H2O)2.6) that can not only stabilize the cathode interface and improve the Coulombic efficiency but also fundamentally solve the Zn anode interface issues. As a result, the aqueous KZnHCF/Zn battery achieves an ultralong life over 3000 cycles without any capacity decay even under a high discharge voltage of 1.78 V (vs Zn2+/Zn). Such extraordinary performance represents significant progress in aqueous PBA-based ZIBs. This work shares guidance to improve the performance of aqueous ZIBs through optimizing the electrolyte in tuning the stable operation of the cathode and the zinc anode.

Stabilizing Li-O2 Batteries with Multifunctional Fluorinated Graphene.

As a full cell system with attractive theoretical energy density, challenges faced by Li-O2 batteries (LOBs) are not only the deficient actual capacity and superoxide-derived parasitic reactions on the cathode side but also the stability of Li-metal anode. To solve simultaneously intrinsic issues, multifunctional fluorinated graphene (CFx, x = 1, F-Gr) was introduced into the ether-based electrolyte of LOBs. F-Gr can accelerate O2- transformation and O2--participated oxygen reduction reaction (ORR) process, resulting in enhanced discharge capacity and restrained O2--derived side reactions of LOBs, respectively. Moreover, F-Gr induced the F-rich and O-depleted solid electrolyte interphase (SEI) film formation, which have improved Li-metal stability. Therefore, energy storage capacity, efficiency, and cyclability of LOBs have been markedly enhanced. More importantly, the method developed in this work to disperse F-Gr into an ether-based electrolyte for improving LOBs' performances is convenient and significant from both scientific and engineering aspects.

Red yeast rice dietary intervention reduces oxidative stress-related inflammation and improves intestinal microbiota.

Inflammation and oxidative stress play key roles in the aging process, while red yeast rice (RYR), a traditional Chinese fermented food, has anti-oxidant and anti-inflammatory effects. To understand the anti-aging function of RYR in vivo, this study established a D-galactose-induced aging mouse model to verify the positive effects of RYR dietary intervention on aging and explore the related underlying mechanism. Eight weeks of RYR dietary intervention was shown to have a significant inhibitory effect on cognitive decline and hippocampal damage. The molecular mechanistic studies showed that the anti-aging effects of RYR were achieved by (i) improving the oxidative stress-related damage (increasing SOD, CAT, and GSH, and reducing MDA), (ii) regulating the NF-κB inflammation pathway induced by oxidative stress (decreasing the pro-inflammatory cytokines IL-6, TNF-α, IFN-γ, iNOs, and IL-1ß, increasing the anti-inflammatory cytokine IL-10, and decreasing the expression of the NF-κB protein), (iii) slowing down apoptosis caused by oxidative stress (reducing the expression of P21 and P53), (iv) restoring the abundance of Lactobacillus, Lachnospiraceae and Rikenellaceae downregulated by D-galactose, and (v) reducing the abundance of Akkermansia and Helicobacter enriched by D-galactose. Mass spectrometry revealed orange pigments (rubropunctatin and monascorubrin) as the main antioxidant components in RYR, which might play key roles in aging inhibition. This study provides theoretical support for the wide application of orange pigments as an antioxidant dietary supplement.

Long-Life Aqueous Zn-I2 Battery Enabled by a Low-Cost Multifunctional Zeolite Membrane Separator.

Aqueous zinc iodide (Zn-I2) batteries are promising large-scale energy-storage devices. However, the uncontrollable diffuse away/shuttle of soluble I3- leads to energy loss (low Coulombic efficiency, CE), and poor reversibility (self-discharge). Herein, we employ an ordered framework window within a zeolite molecular sieve to restrain I3- crossover and prepare zeolite molecular sieve particles into compact, large-scale, and flexible membranes at the engineering level. The as-prepared membrane can confine I3- within the catholyte region and restrain its irreversible escape, which is proved via space-resolution and electrochemical in situ time-resolution Raman technologies. As a result, overcharge/self-discharge and Zn corrosion are effectively controlled by zeolite separator. After replacing the typically used glass fiber separator to a zeolite membrane, the CE of Zn-I2 battery improves from 78.9 to 98.6% at 0.2 A/g. Besides, after aging at the fully charged state for 5.0 h, self-discharge is restrained and CE is enhanced from 44.0 to 85.65%. Moreover, the Zn-I2 cell maintains 91.0% capacity over 30,000 cycles at 4.0 A/g.

Nonvolatile and Nonflammable Sulfolane-Based Electrolyte Achieving Effective and Safe Operation of the Li-O2 Battery in Open O2 Environment.

The Li-O2 battery should operate effectively/safely in an open O2 environment for practical applications, but not trapped in sealed/closed atmosphere. However, the typical use of volatile and flammable electrolyte restricts Li-O2 battery to be able to be running in open O2 environment. We report herein, for the first time, a highly electrochemical reversible Li-O2 battery operated in an open O2 environment, i.e., under the condition of keeping O2 flowing continuously based on a nonvolatile and nonflammable sulfolane (TMS) solvent. The electrochemical irreversibility of Li2O2/O2 conversion and incompatibility between Li metal anodes and electrolyte have been addressed via dissolving LiNO3 in concentrated TMS electrolyte. The tuned electrolyte not only enables a stable solid electrolyte interphase (SEI) with conformal inorganic components (including LiF, LiNxOy, and Li2O) that promotes a uniform Li electro-plating/stripping process but also results in a low charge overpotential, a stable discharge terminal plateau, and reversible O2 generation of the Li-O2 battery conducted in an open O2 environment.

Non-Structured DNN Weight Pruning-Is It Beneficial in Any Platform?

Large deep neural network (DNN) models pose the key challenge to energy efficiency due to the significantly higher energy consumption of off-chip DRAM accesses than arithmetic or SRAM operations. It motivates the intensive research on model compression with two main approaches. Weight pruning leverages the redundancy in the number of weights and can be performed in a non-structured, which has higher flexibility and pruning rate but incurs index accesses due to irregular weights, or structured manner, which preserves the full matrix structure with a lower pruning rate. Weight quantization leverages the redundancy in the number of bits in weights. Compared to pruning, quantization is much more hardware-friendly and has become a "must-do" step for FPGA and ASIC implementations. Thus, any evaluation of the effectiveness of pruning should be on top of quantization. The key open question is, with quantization, what kind of pruning (non-structured versus structured) is most beneficial? This question is fundamental because the answer will determine the design aspects that we should really focus on to avoid the diminishing return of certain optimizations. This article provides a definitive answer to the question for the first time. First, we build ADMM-NN-S by extending and enhancing ADMM-NN, a recently proposed joint weight pruning and quantization framework, with the algorithmic supports for structured pruning, dynamic ADMM regulation, and masked mapping and retraining. Second, we develop a methodology for fair and fundamental comparison of non-structured and structured pruning in terms of both storage and computation efficiency. Our results show that ADMM-NN-S consistently outperforms the prior art: 1) it achieves 348× , 36× , and 8× overall weight pruning on LeNet-5, AlexNet, and ResNet-50, respectively, with (almost) zero accuracy loss and 2) we demonstrate the first fully binarized (for all layers) DNNs can be lossless in accuracy in many cases. These results provide a strong baseline and credibility of our study. Based on the proposed comparison framework, with the same accuracy and quantization, the results show that non-structured pruning is not competitive in terms of both storage and computation efficiency. Thus, we conclude that structured pruning has a greater potential compared to non-structured pruning. We encourage the community to focus on studying the DNN inference acceleration with structured sparsity.

GRIM: A General, Real-Time Deep Learning Inference Framework for Mobile Devices Based on Fine-Grained Structured Weight Sparsity.

It is appealing but challenging to achieve real-time deep neural network (DNN) inference on mobile devices, because even the powerful modern mobile devices are considered as "resource-constrained" when executing large-scale DNNs. It necessitates the sparse model inference via weight pruning, i.e., DNN weight sparsity, and it is desirable to design a new DNN weight sparsity scheme that can facilitate real-time inference on mobile devices while preserving a high sparse model accuracy. This paper designs a novel mobile inference acceleration framework GRIM that is General to both convolutional neural networks (CNNs) and recurrent neural networks (RNNs) and that achieves Real-time execution and high accuracy, leveraging fine-grained structured sparse model Inference and compiler optimizations for Mobiles. We start by proposing a new fine-grained structured sparsity scheme through the Block-based Column-Row (BCR) pruning. Based on this new fine-grained structured sparsity, our GRIM framework consists of two parts: (a) the compiler optimization and code generation for real-time mobile inference; and (b) the BCR pruning optimizations for determining pruning hyperparameters and performing weight pruning. We compare GRIM with Alibaba MNN, TVM, TensorFlow-Lite, a sparse implementation based on CSR, PatDNN, and ESE (a representative FPGA inference acceleration framework for RNNs), and achieve up to 14.08× speedup.

StructADMM: Achieving Ultrahigh Efficiency in Structured Pruning for DNNs.

Weight pruning methods of deep neural networks (DNNs) have been demonstrated to achieve a good model pruning rate without loss of accuracy, thereby alleviating the significant computation/storage requirements of large-scale DNNs. Structured weight pruning methods have been proposed to overcome the limitation of irregular network structure and demonstrated actual GPU acceleration. However, in prior work, the pruning rate (degree of sparsity) and GPU acceleration are limited (to less than 50%) when accuracy needs to be maintained. In this work, we overcome these limitations by proposing a unified, systematic framework of structured weight pruning for DNNs. It is a framework that can be used to induce different types of structured sparsity, such as filterwise, channelwise, and shapewise sparsity, as well as nonstructured sparsity. The proposed framework incorporates stochastic gradient descent (SGD; or ADAM) with alternating direction method of multipliers (ADMM) and can be understood as a dynamic regularization method in which the regularization target is analytically updated in each iteration. Leveraging special characteristics of ADMM, we further propose a progressive, multistep weight pruning framework and a network purification and unused path removal procedure, in order to achieve higher pruning rate without accuracy loss. Without loss of accuracy on the AlexNet model, we achieve 2.58× and 3.65× average measured speedup on two GPUs, clearly outperforming the prior work. The average speedups reach 3.15× and 8.52× when allowing a moderate accuracy loss of 2%. In this case, the model compression for convolutional layers is 15.0× , corresponding to 11.93× measured CPU speedup. As another example, for the ResNet-18 model on the CIFAR-10 data set, we achieve an unprecedented 54.2× structured pruning rate on CONV layers. This is 32× higher pruning rate compared with recent work and can further translate into 7.6× inference time speedup on the Adreno 640 mobile GPU compared with the original, unpruned DNN model. We share our codes and models at the link http://bit.ly/2M0V7DO.