Accelerating water evaporation from salty droplets on polar substrate: a molecular dynamics study.

Evaporation of seawater containing various ions is the largest source of rainfall, affecting the global climate. In industrial areas, water evaporation finds applications in the desalination of seawater to get fresh water for arid coastal regions. Understanding how ions and substrates influence the evaporation of sessile salty droplets on a substrate is essential to modulate the evaporation rate. In the present study, we investigate the effect of ions (Mg2+, Na+, Cl-) on the evaporation of water molecules from sessile droplets on the solid surface using molecular dynamics simulations. The electrostatic interactions between water molecules and ions suppress water evaporation. However, the interactions between molecules and atoms in the substrates accelerate the evaporation. We increase the evaporation of salty droplets by 21.6% by placing the droplet on the polar substrate.

Durably Self-Sustained Droplet on a Fully Miscible Liquid Film.

Droplets impacting onto a solid or liquid surface inducing wetting, floatation, splash, coalescence, etc. is ubiquitous in nature and industrial processes. Here, we report that liquid droplets exhibit spherical caps upon contact with a fully miscible liquid film of lower surface tension, despite the spontaneous mixing of the two liquids. Such a spherical cap on a continuous liquid surface sustains a long lifespan up to minutes before ultimately merging into the film. Benefiting from large viscous forces in a thin film as a result of spatial confinement, the surface flow is substantially suppressed. Therefore, the surface tension gradient responsible for this phenomenon is maintained because the normal diffusion of film liquid into the droplet can timely dilute film liquid supplied by uphill Marangoni flow at the droplet surface. The present finding removes the conventional cognition that droplet coalescence is prompt on fully miscible continuous liquid surfaces, thus benefiting design of new types of microfluidic devices.

Hydrophilic, Clean Graphene for Cell Culture and Cryo-EM Imaging.

The wettability of graphene is critical for numerous applications but is very sensitive to its surface cleanness. Herein, by clarifying the impact of intrinsic contamination, i.e., amorphous carbon, which is formed on the graphene surface during the high-temperature chemical vapor deposition (CVD) process, the hydrophilic nature of clean graphene grown on single-crystal Cu(111) substrate was confirmed by both experimental and theoretical studies, with an average water contact angle of ∼23°. Furthermore, the wettability of as-transferred graphene was proven to be highly dependent on its intrinsic cleanness, because of which the hydrophilic, clean graphene exhibited improved performance when utilized for cell culture and cryoelectron microscopy imaging. This work not only validates the intrinsic hydrophilic nature of graphene but also provides a new insight in developing advanced bioapplications using CVD-grown clean graphene films.

Layer-Tunable Nonlinear Optical Characteristics and Photocarrier Dynamics of 2D PdSe2 in Broadband Spectra.

Layered 2D transition metal dichalcogenides (TMDCs) exhibited fascinating nonlinear optical (NLO) properties for constructing varied promising optoelectronics. However, exploring the desired 2D materials with both superior nonlinear absorption and ultrafast response in broadband spectra remain the key challenges to harvest their greatest potential. Here, based on synthesizing 2D PdSe2 films with the controlled layer number, the authors systematically demonstrated the broadband giant NLO performance and ultrafast excited carrier dynamics of this emerging material under femtosecond visible-to-near-infrared laser-pulse excitation (400-1550 nm). Layer-dependent and wavelength-dependent evolution of optical bandgap, nonlinear absorption, and photocarrier dynamics in the obtained 2D PdSe2 are clearly revealed. Specially, the transition from semiconducting to semimetallic PdSe2 induced dramatic changes of their interband absorption-relaxation process. This work makes 2D PdSe2 more competitive for future ultrafast photonics and also opens up a new avenue for the optical performance optimization of various 2D materials by rational design of these materials.

Constructing a Reinforced and Gradient Solid Electrolyte Interphase on Si Nanoparticles by In-Situ Thiol-Ene Click Reaction for Long Cycling Lithium-Ion Batteries.

Constructing a stable solid electrolyte interphase (SEI) on high-specific-capacity silicon (Si) anode is one of the most effective methods to reduce the crack of SEI and improve the cycling performance of Si anode. Herein, the authors construct a reinforced and gradient SEI on Si nanoparticles by an in-situ thiol-ene click reaction. Mercaptopropyl trimethoxysilane (MPTMS) with thiol functional groups (SH) is first grafted on the Si nanoparticles through condensation reaction, which then in-situ covalently bonds with vinylene carbonate (VC) to form a reinforced and uniform SEI on Si nanoparticles. The modified SEI with sufficient elastic Lix SiOy can homogenize the stress and strain during the lithiation of Si nanoparticles to reduce their expansion and prevent the SEI from cracking. The Si nanoparticles-graphite blending anode with the reinforced SEI exhibits excellent performance with an initial coulombic efficiency of ≈90%, a capacity of 1053.3 mA h g-1 after 500 cycles and a high capacity of 852.8 mA h g-1 even at a high current density of 3 A g-1 . Moreover, the obtained anode shows superior cycling stability under both high loadings and lean electrolyte. The in-situ thiol-ene click reaction is a practical method to construct reinforced SEI on Si nanoparticles for next-generation high-energy-density lithium-ion batteries.

KeMRE: Knowledge-enhanced medical relation extraction for Chinese medicine instructions.

Medicine instructions usually contain rich medical relations, and extracting them is very helpful for many downstream tasks such as medicine knowledge graph construction and medicine side-effect prediction. Existing relation extraction (RE) methods usually predict relations between entities from their contexts and do not consider medical knowledge. However, understanding a part of medical relations may need some expert knowledge in the medical field, making it challenging for existing methods to achieve satisfying performances of medical RE. In this paper, we propose a knowledge-enhanced framework for medical RE, which can exploit medical knowledge of medicines to better conduct medical RE on Chinese medicine instructions. We first propose a BERT-CNN-LSTM based framework for text modeling and learn representations of characters from their contexts. Then we learn representations of each entity by aggregating representations of their characters. Besides, we propose a CNN-LSTM based framework for entity modeling and learn entity representations from their relatedness. In addition, there are usually many different instructions for the same medicine, which usually share general knowledge on this medicine. Thus, to obtain medical knowledge of medicines, we annotate relations on a randomly-sampled instruction of each medicine. Then we build knowledge embeddings to represent potential relations between entities from knowledge of medicines. Finally, we use an MLP network to predict relations between entities from their representations and knowledge embeddings. Extensive experiments on a real-world dataset show that our method can significantly outperform existing methods.

Macroscopic superhydrophobicity achieved by atomic decoration with silicones.

Modulating wetting states of water droplets on a solid surface has great potential in numerous applications such as self-cleaning, heat transfer, and water purification. Present strategies to achieve controlled wettability usually employ surface roughness and/or chemical coating at the nanometer-to-micrometer scale. Here we propose that atomic-scale decoration and patterning with silicones can be used to fabricate superhydrophobic surfaces. Atomic decoration, differing from chemical coatings in the characteristic size and partial coverage, controls wetting states with the atomic precision (several angstroms) at low cost and maintains a long lifetime. We attribute the mechanism of atomic decoration to the decrease in effective potential-well depth, inducing weak interactions between the substrate and water droplet. Our work pushes water wetting to the limit of atomic scales, hinting for a paradigm shift from macroscopic to atomically-precise engineering of wetting states and providing new insights for fabricating superhydrophobic surfaces.

Optimization of CNN through Novel Training Strategy for Visual Classification Problems.

The convolution neural network (CNN) has achieved state-of-the-art performance in many computer vision applications e.g., classification, recognition, detection, etc. However, the global optimization of CNN training is still a problem. Fast classification and training play a key role in the development of the CNN. We hypothesize that the smoother and optimized the training of a CNN goes, the more efficient the end result becomes. Therefore, in this paper, we implement a modified resilient backpropagation (MRPROP) algorithm to improve the convergence and efficiency of CNN training. Particularly, a tolerant band is introduced to avoid network overtraining, which is incorporated with the global best concept for weight updating criteria to allow the training algorithm of the CNN to optimize its weights more swiftly and precisely. For comparison, we present and analyze four different training algorithms for CNN along with MRPROP, i.e., resilient backpropagation (RPROP), Levenberg-Marquardt (LM), conjugate gradient (CG), and gradient descent with momentum (GDM). Experimental results showcase the merit of the proposed approach on a public face and skin dataset.

Regional Cortical Surface Area in Adolescents: A Preliminary MRI Twin Study of Genetic and Environmental Contributions.

Cortical surface area (CSA) has particular relevance for understanding development, behavior, and the connection between brain structure and function. Little is known about genetic and environmental determinants of CSA during development. We utilized bivariate twin methods to identify global and regionally specific genetic factors which influence CSA in a preliminary sample of typically-developing adolescents, with hypotheses based on findings in middle-aged adults. Similar to previous findings, we observed high heritability for total CSA. There was also significant evidence for genetic influences on regional CSA, particularly when these were not adjusted for total CSA, with highest heritability in frontal cortex and relatively fewer genetic contributions to medial temporal cortical structures. Adjustment for total CSA reduced regional CSA heritability dramatically, but a moderate influence of genetic factors remained in some regions. Both global and regionally-specific genetic factors influence regional CSA during adolescence.

Microscopic insight into surface wetting: relations between interfacial water structure and the underlying lattice constant.

We report simulation evidence that the structure of the first water layer next to the surface can strongly affect the contact angle of water droplets. Molecular dynamics simulations show that a small uniform strain (± 3%) applied to the lattice constant of a multilayer hydrophilic surface can introduce a marked change in the wetting tendency. In particular, when the lattice constant of a hydrophilic surface matches the projected oxygen-oxygen distance of bulk water to the surface, a contact-angle minimum is resulted. In stark contrast, such a lattice strain has little effect on the wetting properties of hydrophobic surfaces. The structure of the first water layer next to the hydrophilic surface gradually loses characteristics of liquid water when moving away from the contact-angle minimum. Our results demonstrate a close correlation among the length of lattice constant, contact angle of the water droplet, and the structure and dynamics of vicinal water.

Semi-Supervised Hierarchical Graph Classification.

Node classification and graph classification are two graph learning problems that predict the class label of a node and the class label of a graph respectively. A node of a graph usually represents a real-world entity, e.g., a user in a social network, or a document in a document citation network. In this work, we consider a more challenging but practically useful setting, in which a node itself is a graph instance. This leads to a hierarchical graph perspective which arises in many domains such as social network, biological network and document collection. We study the node classification problem in the hierarchical graph where a "node" is a graph instance. As labels are usually limited, we design a novel semi-supervised solution named SEAL-CI. SEAL-CI adopts an iterative framework that takes turns to update two modules, one working at the graph instance level and the other at the hierarchical graph level. To enforce a consistency among different levels of hierarchical graph, we propose the Hierarchical Graph Mutual Information (HGMI) and further present a way to compute HGMI with theoretical guarantee. We demonstrate the effectiveness of this hierarchical graph modeling and the proposed SEAL-CI method on text and social network data.

Differentially private knowledge transfer for federated learning.

Extracting useful knowledge from big data is important for machine learning. When data is privacy-sensitive and cannot be directly collected, federated learning is a promising option that extracts knowledge from decentralized data by learning and exchanging model parameters, rather than raw data. However, model parameters may encode not only non-private knowledge but also private information of local data, thereby transferring knowledge via model parameters is not privacy-secure. Here, we present a knowledge transfer method named PrivateKT, which uses actively selected small public data to transfer high-quality knowledge in federated learning with privacy guarantees. We verify PrivateKT on three different datasets, and results show that PrivateKT can maximally reduce 84% of the performance gap between centralized learning and existing federated learning methods under strict differential privacy restrictions. PrivateKT provides a potential direction to effective and privacy-preserving knowledge transfer in machine intelligent systems.

FET-LM: Flow-Enhanced Variational Autoencoder for Topic-Guided Language Modeling.

Variational autoencoder (VAE) is widely used in tasks of unsupervised text generation due to its potential of deriving meaningful latent spaces, which, however, often assumes that the distribution of texts follows a common yet poor-expressed isotropic Gaussian. In real-life scenarios, sentences with different semantics may not follow simple isotropic Gaussian. Instead, they are very likely to follow a more intricate and diverse distribution due to the inconformity of different topics in texts. Considering this, we propose a flow-enhanced VAE for topic-guided language modeling (FET-LM). The proposed FET-LM models topic and sequence latent separately, and it adopts a normalized flow composed of householder transformations for sequence posterior modeling, which can better approximate complex text distributions. FET-LM further leverages a neural latent topic component by considering learned sequence knowledge, which not only eases the burden of learning topic without supervision but also guides the sequence component to coalesce topic information during training. To make the generated texts more correlative to topics, we additionally assign the topic encoder to play the role of a discriminator. Encouraging results on abundant automatic metrics and three generation tasks demonstrate that the FET-LM not only learns interpretable sequence and topic representations but also is fully capable of generating high-quality paragraphs that are semantically consistent.

Oxidization-Temperature-Triggered Rapid Preparation of Large-Area Single-Crystal Cu(111) Foil.

The controlled preparation of single-crystal Cu(111) is intensively investigated owing to the superior properties of Cu(111) and its advantages in synthesizing high-quality 2D materials, especially graphene. However, the accessibility of large-area single-crystal Cu(111) is still hindered by time-consuming, complicated, and high-cost preparation methods. Here, the oxidization-temperature-triggered rapid preparation of large-area single-crystal Cu(111) in which an area up to 320 cm2 is prepared within 60 min, and where low-temperature oxidization of polycrystalline Cu foil surface plays a vital role, is reported. A mechanism is proposed, by which the thin Cux O layer transforms to a Cu(111) seed layer on the surface of Cu to induce the formation of a large-area Cu(111) foil, which is supported by both experimental data and molecular dynamics simulation results. In addition, a large-size high-quality graphene film is synthesized on the single-crystal Cu(111) foil surface and the graphene/Cu(111) composites exhibit enhanced thermal conductivity and ductility compared to their polycrystalline counterpart. This work, therefore, not only provides a new avenue toward the monocrystallinity of Cu with specific planes but also contributes to improving the mass production of high-quality 2D materials.

[Research on the method of copper converting process determination based on emission spectrum analysis].

A method of copper converting process determination based on PbO/PbS emission spectrum analysis was described. According to the known emission spectrum of gas molecules, the existence of PbO and PbS was confirmed in the measured spectrum. Through the field experiment it was determined that the main emission spectrum of the slag stage was from PbS, and the main emission spectrum of the copper stage was from PbO. The relative changes in PbO/PbS emission spectrum provide the method of copper converting process determination. Through using the relative intensity in PbO/PbS emission spectrum the copper smelting process can be divided into two different stages, i.e., the slag stage (S phase) and the copper stage (B phase). In a complete copper smelting cycle, a receiving telescope of appropriate view angle aiming at the converter flame, after noise filtering on the PbO/PbS emission spectrum, the process determination agrees with the actual production. Both the theory and experiment prove that the method of copper converting process determination based on emission spectrum analysis is feasible.

Molecular origin of fast evaporation at the solid-water-vapor line in a sessile droplet.

By analyzing the behaviors of water molecules at the solid-water-vapor contact line, we explore the molecular origin of large evaporation rates at the contact line and find new ways to increase the evaporation of the droplet. In contrast to previous models considering macroscopic surroundings and the geometry of the droplet, here we study the behaviors of water molecules by introducing cohesive energy which includes interactions of water molecules with both other water molecules in the droplet and atoms in the substrate. Molecules at the contact line bear the smallest evaporating energy barrier and therefore, possess the largest possibility to evaporate. Further analyses show that the evaporation rate of the droplet is enhanced through the large length of the contact line. These analyses are corroborated by experiments, where the evaporation rate of the droplet is enhanced up to 30% by incorporating hollow glass spheres in the droplet. Our theoretical and experimental efforts illustrate the underlying molecular mechanisms of large evaporation rates of a droplet, providing new avenues to accelerate droplet evaporation.

Communication-efficient federated learning via knowledge distillation.

Federated learning is a privacy-preserving machine learning technique to train intelligent models from decentralized data, which enables exploiting private data by communicating local model updates in each iteration of model learning rather than the raw data. However, model updates can be extremely large if they contain numerous parameters, and many rounds of communication are needed for model training. The huge communication cost in federated learning leads to heavy overheads on clients and high environmental burdens. Here, we present a federated learning method named FedKD that is both communication-efficient and effective, based on adaptive mutual knowledge distillation and dynamic gradient compression techniques. FedKD is validated on three different scenarios that need privacy protection, showing that it maximally can reduce 94.89% of communication cost and achieve competitive results with centralized model learning. FedKD provides a potential to efficiently deploy privacy-preserving intelligent systems in many scenarios, such as intelligent healthcare and personalization.

Oxygen Plasma Modified Carbon Cloth with C=O Zincophilic Sites as a Stable Host for Zinc Metal Anodes.

Aqueous zinc-ion batteries (ZIBs) are currently receiving widespread attention due to their merits of environmental-friendly properties, high safety, and low cost. However, the absence of stable zinc metal anodes severely restricts their potential applications. In this work, we demonstrate a simple oxygen plasma treatment method to modify the surface state of carbon cloth to construct an ideal substrate for zinc deposition to solve the dendrite growth problem of zinc anodes. The plasma treated carbon cloth (PTCC) electrode has lower nucleation overpotential and uniformly distributed C=O zincophilic nucleation sites, facilitating the uniform nucleation and subsequent homogeneous deposition of zinc. Benefiting from the superior properties of PTCC substrate, the enhanced zinc anodes demonstrate low voltage hysteresis (about 25 mV) and stable zinc plating/stripping behaviors (over 530 h lifespan) at 0.5 mA cm-2 with 15% depth of discharge (DOD). Besides, an extended cycling lifespan of 480 h can also be achieved at very high DOD of 60%. The potential application of the enhanced zinc anode is also confirmed in Zn|V10O24·12H2O full cell. The cells with Zn@PTCC electrode demonstrate remarkable rate capability and excellent cycling stability (95.0% capacity retention after 500 cycles).

A federated graph neural network framework for privacy-preserving personalization.

Graph neural network (GNN) is effective in modeling high-order interactions and has been widely used in various personalized applications such as recommendation. However, mainstream personalization methods rely on centralized GNN learning on global graphs, which have considerable privacy risks due to the privacy-sensitive nature of user data. Here, we present a federated GNN framework named FedPerGNN for both effective and privacy-preserving personalization. Through a privacy-preserving model update method, we can collaboratively train GNN models based on decentralized graphs inferred from local data. To further exploit graph information beyond local interactions, we introduce a privacy-preserving graph expansion protocol to incorporate high-order information under privacy protection. Experimental results on six datasets for personalization in different scenarios show that FedPerGNN achieves 4.0% ~ 9.6% lower errors than the state-of-the-art federated personalization methods under good privacy protection. FedPerGNN provides a promising direction to mining decentralized graph data in a privacy-preserving manner for responsible and intelligent personalization.

Research on covert communication channel based on modulation of common compressed speech codec.

As is well known, multimedia has been widely used in VoIP and mobile communications. Research on how to establish covert communication channel over the above popular public applications has been flourishing in recent years. This paper tries to present a novel and effective method to construct a covert channel over common compressed speech stream by embedding sense information into it. In our method, after analysing the characteristic features of the excitation pulse positions of the ITU-T G.723.1 and G.729A speech codec, we design a novel and effective covert communication channel by finely modulating the codes of excitation pulse positions of the above two codecs in line with the secret information to be hidden. To improve the embedding capacity of the proposed method, we also use all the odd/even characteristics of pulse code positions to conduct information hiding. To test and verify the proposed approach, experiments are conducted on several different scenarios. Experimental results show that our methods and algorithms perform a higher degree of secrecy and sound information embedding efficacy compared with exiting similar methods.