Feature modeling, which involves feature representation learning and leveraging, plays an essential role in industrial recommendation systems. However, the data distribution in real-world applications usually follows a highly skewed long-tail pattern due to the popularity bias, which easily leads to over-reliance on ID-based features, such as user/item IDs and ID sequences of interactions. Such over-reliance makes it hard for models to learn features comprehensively, especially for those non-ID meta features, e.g., user/item characteristics. Further, it limits the feature leveraging ability in models, getting less generalized and more susceptible to data noise. Previous studies on feature modeling focus on feature extraction and interaction, hardly noticing the problems brought about by the long-tail data distribution. To achieve better feature representation learning and leveraging on real-world data, we propose a model-agnostic framework AdaF ^2 M ^2 , short for Adaptive Feature Modeling with Feature Mask. The feature-mask mechanism helps comprehensive feature learning via multi-forward training with augmented samples, while the adapter applies adaptive weights on features responsive to different user/item states. By arming base models with AdaF ^2 M ^2 , we conduct online A/B tests on multiple recommendation scenarios, obtaining +1.37 ^2 M ^2 has been widely deployed on both retrieval and ranking tasks in multiple applications of Douyin Group, indicating its superior effectiveness and universality.

In this study, we designed highly sensitive photoresist resins, JN05 and BN05, having both a photosensitive DNQ unit as a nonchemical amplification mode and the photo acid-cleavable groups like the ethyl vinyl ether (EVE) and diethyl decarbonate (BOC) protection groups as a chemical amplification mode. Resists with these dual-sensitive chemical reaction modes will undergo photolysis in the DNQ unit after exposure and further the acid-catalyzed deprotection reactions of the EVE and BOC groups after post-exposure bake. This dual-sensitized photolysis of the photoresist occurs at mild reaction conditions, enabling synergistic modifications by multiple functional groups. Upon exposure to doses of 150 mJ/cm2 and 90 mJ/cm2, respectively, JN05 and BN05 produced clear positive tone patterns with a line width of 1.0 mu m (L/S = 1:1). Compared with the Novolac/DNQ system itself, the JN05 and BN05 photoresists combined the benefits of CA photoresists, nonchemically amplified solubility properties, and the multifunctional characteristics achieved by the DNQ system and CA comodification. The present work provides an approach for the improvement of photoresist sensitivity and resolution in the UV resist system.

The negative effects of electromagnetic (EM) pollution on precise electronic devices as well as public health have received serious attention. Although various electromagnetic interference (EMI) shielding materials have emerged, it is still a great challenge to effectively balance the high EMI shielding performance and low secondary emission of EM waves induced by direct reflection. In this study, we designed two type of shielding materials based on 3D carbon nanotube sponge (CNTS) with manufactured surface structure. By combining the 2D MXene layer or magnetic nickel coating with patterned CNTS, we achieve high EMI shielding performance and low reflectivity, as well as adjustable shielding mechanism through the dual adjustment of the structure and electromagnetic properties. The EMI shielding effectiveness of CNTS/MXene up to 90 dB and reflectivity as low as 0.31 (at 18 GHz), a reduction of about 64 % compared to the original CNTS. And the EMI shielding effectiveness of Ni/CNTS up to 67 dB and reflectivity as low as 0.44 (at 12 GHz). Microwave microscope (NFSMM), COMSOL simulation and vector network analyzer consistently confirmed the synergestical surpression effect on the reflectivity from surface structure engineering and electromagnetic regulation. These results not only guide the designing of advanced EMI shielding materials with low reflectivity, but also shed light on the hidden mechanism between interface structures and performances of the composite materials.

Natural protein-based biomaterials with complex hierarchical structures often have incredible and even counterintuitive mechanical properties. Understanding and utilizing the conformational transition mechanisms of natural proteins will further guide the design of natural-inspired biomaterials. In this study, a small static-force-induced spatiotemporal "freezing" phenomenon of silk fibroins confined in porous carbon nanotube sponges has been investigated. The "freezing" silk fibroins not only bring the shape memory effect to elastic carbon nanotube sponges but also enable them to prop up heavy objects with loads exceeding 10,000 times their own weight. Also, the protein/CNTS hybrid achieves an ultrastiffness (over 10 MPa) and superelastic shape recovery (recovery strain >90%). Both experimental and numerical results indicate that the secondary conformational transition of silk fibroin plays a key role, where more α-helices/random coils transform into β-sheets under both confinement and low pressure. Our work reports a conformational transition mechanism of silk fibroin in a confined space, which provides guidance for constructing protein-based biological smart materials with potential applications in textiles, medicine, architecture, and other research fields.

Electromagnetic interference (EMI) shielding materials with lightweight, high shielding effectiveness, excellent chemical stability, especially minimized secondary electromagnetic pollution, are urgently desired for integrated electronic systems operating in harsh working environments. Here in this study, by systematically engineering and matching the interfacial properties of carbon-based membrane materials, i.e., graphite paper, whisker carbon nanotube paper (WCNT paper), carbon nanotube film (CNT film), bucky paper (BP), and carbon cloth (CC) with three-dimensional (3D) porous carbon nanotube sponge (CNTS), we successfully constructed a series of multifunctional all-carbon EMI shielding materials, which exhibit excellent average shielding effectiveness of over 90 dB with a thickness of about 1 mm and dramatically minimized secondary electromagnetic reflection. Moreover, benefiting from the all-carbon nature and engineered interfaces, our CMC materials also exhibit excellent photothermal and Joule heating performances. These results not only provide guidance for designing advanced multifunctional all-carbon EMI shielding materials but also shed light on the hidden mechanism between interfaces and performances of composite materials.

User preferences follow a dynamic pattern over a day, e.g., at 8 am, a user might prefer to read news, while at 8 pm, they might prefer to watch movies. Time modeling aims to enable recommendation systems to perceive time changes to capture users' dynamic preferences over time, which is an important and challenging problem in recommendation systems. Especially, streaming recommendation systems in the industry, with only available samples of the current moment, present greater challenges for time modeling. There is still a lack of effective time modeling methods for streaming recommendation systems. In this paper, we propose an effective and universal method Interest Clock to perceive time information in recommendation systems. Interest Clock first encodes users' time-aware preferences into a clock (hour-level personalized features) and then uses Gaussian distribution to smooth and aggregate them into the final interest clock embedding according to the current time for the final prediction. By arming base models with Interest Clock, we conduct online A/B tests, obtaining +0.509% and +0.758% improvements on user active days and app duration respectively. Besides, the extended offline experiments show improvements as well. Interest Clock has been deployed on Douyin Music App.

It is a critical challenge to construct a robust and efficient electromagnetic shielding system with tunable shielding behavior and interfacial properties to adapt to complicated extreme conditions. Here, we fabricate a ternary SiC@high-entropy alloy (HEA)@carbon nanotube sponge (CNTS) composite with designed heterogeneous interfaces and composition on a wide range of scales. On the atomic scale, the electromagnetic character of the composite can be easily regulated through adjusting the elemental composition of HEAs due to their high entropy effect and the cocktail effect. A nano-meter thick silicon carbide (SiC) layer on the HEA further generates a Schottky junction between the metal-SiC and CNTS-SiC to achieve an outstanding average electromagnetic interference (EMI) shielding performance of ca. 90 dB in the X-band on the nano scale (the maximum EMI SE of the 4 mm-thick sample is greater than 100 dB), accompanied with good elasticity and excellent chemical and thermal resistance to endure acid corrosion and high-temperature annealing on the micro scale. This work paves a pathway for designing and developing potential shielding composites with excellent and tunable EMI shielding performance and comprehensive characteristics for complicated application scenarios. A SiC@high-entropy alloy@carbon nanotube sponge composite with heterogeneous interfaces achieves an outstanding EMI shielding performance over 90 dB, good elasticity and excellent chemical and thermal resistance.

Supercapacitors have been highly demanded as advanced energy storage devices. A major obstacle is that, compared to commercial battery systems, their energy density is relatively low. Here, we report a flexible asymmetric supercapacitor with high energy and power densities composed of nanostructured MnO2 nanospheres and Bi2O3 nanoshells decorated on a 3D interconnected carbon nanotube sponge (CNTS) in a neutral hydrogel-electrolyte. Our CNTS@MnO2 electrode and CNTS@Bi2O3 electrode demonstrate an excellent areal capacitance of 5.54 Fcm(-3) and 4.76 Fcm(-3), respectively. In 10,000 charge-discharge cycles, they maintain 93% and 80% of their initial capacitance, respectively. A flexible asymmetric supercapacitor composed of the CNTS@MnO2 anode and the CNTS@Bi2O3 cathode (embedded in a Na2SO4 hydrogel-electrolyte) delivers a high energy density up to 37 Whkg(-1) at a cell voltage of 1.8 V, 7 times higher than that of traditional electrochemical double-layer capacitors (EDLCs). The long-term cyclic stability and robust mechanical stability profit from the flexible conductive skeleton of CNTS and meet the demands of wearable flexible electronics.

Controlling the growth of microbial consortia is of great significance in the biomedical field. Selective bacterial growth is achieved by fabricating silk inverse opal (SIO) scaffolds with varying pore sizes ranging from 0.3 to 4.5 µm. Pore size significantly influences the growth dynamics of bacteria in both single and mixed-strain cultures. Specially, the SIO-4.5 µm scaffold is observed to be more favorable for cultivating S. aureus , whereas the SIO-0.3 µm scaffold is more suitable for cultivating E. coli and P. aeruginosa . By adjusting the secondary conformation of silk fibroin, the stiffness of the SIO substrate will be altered, which results in the increase of bacteria on the SIO by 16 times compared with that on the silk fibroin film. Manipulating the pore size allows for the adjustment of the S. aureus to P. aeruginosa ratio from 0.8 to 9.3, highlighting the potential of this approach in regulating bacterial culture.

Download This Paper Open PDF in Browser Add Paper to My Library Share: Permalink Using these links will ensure access to this page indefinitely Copy URL Copy DOI

Compared to conventional fibers, electrospun porous nanofibers with hierarchical structures often involve additional active sites, interfaces, and internal spaces which boost the performances of functional materials. Here in this study, coaxial composite cellulose acetate@silk fibroin (CA@SF) fibrous membranes are constructed through an electrostatic spinning technique combining solvent-induced phase separation. Hierarchical core-shell structures on the fibers are achieved, which significantly increases the surface area and benefits the mechanical property, flux, as well as the electroless deposition of Ag nanoparticles. The total electromagnetic shielding efficiency of the sandwiched hierarchical CA@SF@Ag composite membrane with a thickness of only 100 μm reaches up to 100 dB, surpassing around 82% beyond nonhierarchical ones. To be noticed, when post-treated by ethanol, the membrane enables an enhanced tensile strength of up to 10 MPa with a thickness of only 50 μm. Our findings pave the way to the application of electrospun fiber membranes in the field of ultrathin electromagnetic shielding films.

AbstractInspired by the Mimosa plant, this study herein develops a unique dynamic shape memory polymer (SMP) network capable of transitioning from hard to pliable with heat, featuring reversible actuation, self‐healing, recyclability, and degradability. This material is adept at simulating the functionalities of artificial muscles for a variety of tasks, with a remarkable specific energy density of 1.8 J g−1—≈46 times higher than that of human skeletal muscle. As an intelligent manipulator, it demonstrates remarkable proficiency in identifying and handling items at high temperatures. Its suitable rate of shape recovery around human body temperature indicates its promising utility as an implant material for addressing acute obstructions. The dynamic covalent bonding within the network structure not only provides excellent resistance to solvents but also bestows remarkable abilities for self‐healing, reprocessing, and degradation. These attributes significantly boost its practicality and environmental sustainability. Anticipated to promote advancements in the sectors of biomedical devices, soft robotics, and smart actuators, this SMP network represents a forward leap in simulating artificial muscles, marking a stride toward the future of adaptive and sustainable technology.

User preferences follow a dynamic pattern over a day, e.g., at 8 am, a user might prefer to read news, while at 8 pm, they might prefer to watch movies. Time modeling aims to enable recommendation systems to perceive time changes to capture users' dynamic preferences over time, which is an important and challenging problem in recommendation systems. Especially, streaming recommendation systems in the industry, with only available samples of the current moment, present greater challenges for time modeling. There is still a lack of effective time modeling methods for streaming recommendation systems. In this paper, we propose an effective and universal method Interest Clock to perceive time information in recommendation systems. Interest Clock first encodes users' timeaware preferences into a clock (hour-level personalized features) and then uses Gaussian distribution to smooth and aggregate them into the final interest clock embedding according to the current time for the final prediction. By arming base models with Interest Clock, we conduct online A/B tests, obtaining +0.509% and +0.758% improvements on user active days and app duration respectively. Besides, the extended offline experiments show improvements as well. Interest Clock has been deployed on Douyin Music App.

Inferior absolute strength and dissolution properties are the main bottlenecks for the widespread application of dissolvable magnesium alloys in complex working environments for unconventional oil and gas resources. Here, a novel functional peak-aged Mg-9.5Gd-2.7Y-0.9Zn-0.8Cu-0.4Ni (wt.%) alloy for fracturing tools is reported, and it possesses an ultimate tensile strength of 457.6 MPa, ultimate compressive strength of 620.7 MPa and dissolution rate of ∼43.7 mg·cm−2·h−1 in 3 wt.% KCl solutions at 93 °C. The excellent strength of the aged-alloy is primarily attributed to the combination of grain refinement, long-period stacking ordered (LPSO) strengthening, and precipitation strengthening induced by stacking fault and β’ phase, among which the precipitation strengthening is dominant. Further investigations confirm that the corrosion is triggered from the micro-galvanic coupling between the Mg matrix and the cathodic lamellar and block LPSO phases. Strip-shaped corrosion pits along with LPSO phases are subsequently formed, significantly accelerating corrosion. The β’ precipitates can effectively improve the strength without compromising the dissolution rate because of their nanoscale size. This study provides an excellent material selection for dissolvable fracturing tools and presents a strategy by which a synergistic combination of strength and dissolution rate is achieved via peak-aging treatment.

Spider silks, mainly constructed of spidroin, have received extensive attention for their excellent mechanical properties, slow bio-degradability, and high biocompatibility. However, due to uncontrollable protein-folding processes, the structural transition of spidroin, especially when composited with other functional materials under confinement, is insufficiently understood. Herein, we report a pressure-induced conforma-tional transition process of the spidroin which is confined within carbon nanotube (CNT) sponge matrixes. The structural transition of spidroin from alpha-helix to fi-sheet can be induced by a very small hydrostatic pressure (several megapascals) and recover easily through a subsequent solvent vapor annealing process in an ambient atmosphere. Therefore, the spidroin/CNT sponge exhibits reversible vapor-/pressure-sensitive "shape-memory" behav-ior with the recovery efficiency close to 100%. Our observation reveals a crucial mechanism for the conformational transition of spidroin under confinement, which paves the way toward the fabrication of spider silk-based products with superior performances.

Natural spider silks with striking performances achieve extensive investigations. Nonetheless, a lack of consensus over the mechanism of the natural spinning hinders the development of artificial spinning methods where the regenerated spider silks generally show poor performances compared with the natural fibers. As is known, the Plateau-Rayleigh instability tends to break solution column into droplets and is considered a main challenge during fiber-spinning. Here in this study, by harnessing the viscoelastic properties of the regenerated spidroin dope solution via organic salt-zinc acetate (ZA), this outcome can be avoided, and dry-spinning of long and mechanically robust regenerated spider silk ribbons can be successfully realized. The as-obtained dry-spun spider silk ribbons show an enhanced modulus up to 14 ± 4 GPa and a toughness of ≈51 ± 9 MJ m-3 after the post-stretching treatment, which is even better than that of the pristine spider silk fibers. This facile and flexible strategy enriches the spinning methodologies which bypass the bottleneck of precisely mimicking the complex natural environment of the glands in spiders, shining a light to the spider-silk-based textile industrial applications.

Electromagnetic interference (EMI) shielding materials with high thermal resistivity, lightweight and flexibility have been intensively demanded to protect precision electronic instruments in various cutting-edge fields. Herein, using a low-temperature growth strategy, we successfully constructed a carbon nanotube (CNT) @ silicon carbide (SiC) coaxial 3D porous composite sponge with comprehensive outstanding performances, including low density of 15-32 mg center dot cm(-3); super elasticity with high shape-recovery of 95 % at compression strain of 96 %; reversible compression-release stability with shape retention of 86 % and stress retention of 85 % after 1000 loading-unloading cycles; excellent thermal resistivity of over 1000 degrees C in Argon or 700 degrees C in Air; eye-catching EMI shielding performance with EMI shielding effectiveness of up to 75.7 +/- 6 dB in X-band which, to the best of our knowledge, is the highest value for silicon carbide (SiC) based porous EMI shielding materials. Our results not only provide guidance for the further development of the next-generation EMI shielding materials under extreme conditions, but also cast a light to explore fundamental mechanisms behind porous EMI shielding materials.

Tungsten (W) is one of the most promising plasma-facing materials for future fusion devices. Although its melting point is the highest among all metals, it still has great risk of melting under extremely high plasma heat fluxes, which is a big concern for ITER and future reactors. Actively cooled W plasma-facing components (PFCs) with both monoblocks and flat-type structure have been successfully installed in the lower divertor of the EAST tokamak since 2021, and provide a good opportunity for direct comparison of the damage mechanism for the two types of PFCs. Various in situ melting phenomena on the lower divertor have been observed by CCD cameras, which have been further verified by post-mortem inspections. Severe melting and even exfoliation of the edge-beveled W plates were observed on some W/Cu flat-type components at horizontal outer targets. Many droplets were ejected during long-pulse operations, which induced a significant increase of W impurities and total irradiation in the core plasma, and thus greatly deteriorated the plasma performance and even caused disruptions. Two different shaping structures of flat-type PFCs show different positions of melting and the corresponding mechanisms. Slight melting was found on the sharp leading edges of W/Cu monoblocks between cassette modules (inter-CM) for horizontal targets with small droplet ejection, which was much improved compared to that observed on the upper W divertor, illustrating that the application of a large-sized bevel chamfer inter-CM was generally effective. In addition, an unexpected melting phenomenon on the dome plate was attributed to the extreme transient heat flux during disruption with runaway electrons. The application of both types of W/Cu PFCs for the divertor provides important experiences and lessons for the engineering design and optimization of divertor PFCs in future fusion devices.

The perpendicular magnetic anisotropy of Nd-Fe-B films decreases when the film thickness increases above a critical value, owing to the formation of the in-plane texture. To avoid this dilemma, the influences of Ta spacer layers on the microstructure and magnetic properties of Nd-Dy-Fe-Co-B thick films are systematically investigated by varying the number and thickness of Ta spacer layers. With increasing the number and thickness of Ta spacer layers, a great enhancement of the perpendicular magnetic anisotropy is found, while the coercivity maintains around 2 T. According to the analysis of microstructural results, some interfaces are formed, with the Nd-Fe-B columnar crystal and the Nd element aggregation area being the two sides of the interface. It is found that the Ta spacer layer facilitates the formation of the out-of-plane texture. In addition, the Ta spacer layer has the effect of enhancing the pinning effect leading to a high coercivity. Based on first-order reversal curve analysis, the Ta spacer layer can enhance the dipole interaction and reduce the exchange interaction, which reveals that Nd-Fe-B grains are isolated by the Nd-rich phase. This work can help us to prepare thick Nd-Fe-B-based films with high perpendicular magnetic anisotropy and high coercivity for applications in micromechanical systems.

It is important to understand the relationship between coercivity and microstructure for the development of high performance rare-earth permanent magnets. In this paper, the effects of film thickness on perpendicular magnetic anisotropy, coercivity mechanism, magnetization reversal process and microstructure of Nd-Dy-Fe-Co-B thick films are systematically investigated by varying film thickness from 1 to 12 mu m. It is found that the coercivity mechanism transforms from a mixed type dominated by pinning to another mixed type dominated by nucleation. In addition, the evolution of grain shape from irregular spherical crystals to columnar ones occur at film thickness of 3 mu m. The columnar crystals form on the capping layer and grow toward the buffer layer. When the film is thicker than 6 mu m, Nd elements accumulate on the Ta buffer layer. With increasing the film thickness, the columnar crystal structure can grow throughout the whole thick film, and the Nd-rich phase is extruded onto the buffer layer, while the pinning effect weakens and the nucleation mechanism is dominated. This work can help us better understand the relationship between coercivity mechanism and microstructure of thick rare-earth permanent magnetic films. (C) 2022 Elsevier B.V. All rights reserved.