We study the Fermi surface topology of a two-dimensional electron gas (2DEG) proximitized by d-wave superconductors in a linear superconductor–normal–superconductor (SNS) Josephson junction with a π phase difference. Owing to nodal quasiparticles and the anisotropic gap, the d-wave case differs qualitatively from the isotropic s-wave case: the nonlocal conductance is no longer directly tied to the number of critical points on the Fermi surface. Instead, we show that the rectified conductance remains quantized at low bias and faithfully encodes the Fermi surface topology via the Euler number χ _F . This quantized response persists even for complex or multi-pocket Fermi surfaces, establishing rectified conductance as a robust and experimentally accessible probe of Fermi surface topology in gapless superconductors.

Electrical controllable quantum spin pumping, independent of any ferromagnetic materials, has always been a hot research topic. This study proposes a theoretical framework to realize quantum spin pumping in a zigzag graphene-like ribbon by applying only two asymmetric side gates with opposite magnitudes. Two current peaks were observed while split between the two spin species under side gates, with one peak being above zero and the other below zero, resulting in spinpolarized currents. The spin polarization stems from the opposite effects of the side gates on the Fermi velocities and transmission coefficients in opposite directions, which are increased in one direction and decreased in the other, and the two spin species are affected in opposite ways. Moreover, pure spin currents occur at zero Fermi energy, and the spin polarization can be converted between 1 and -1 by adjusting the driving frequencies. This study proposes a novel approach for achieving spin-polarized and pure spin currents in zigzag graphene-like ribbons, offering potential applications in the development of spintronic devices.

We investigate Andreev reflection and valley-dependent transport in a graphene/line-defect/superconductor junction. Using a tight-binding model combined with the Bogoliubov–de Gennes formalism and a scattering matrix approach, we show that the line defect acts as a universal valley filter for both electrons and holes, independent of their band index. This universal behavior originates from the invariance of the pseudospin structure governing transmission across conduction and valence bands. Based on this property, we identify two distinct transport mechanisms associated with different Andreev reflection regimes. In the retro-Andreev reflection regime, where electrons and holes reside in the same band, the defect induces a double-filtering mechanism that enhances valley selectivity and produces strongly asymmetric angular distributions. In contrast, in the specular regime, the interband nature of Andreev reflection leads to a complementary filtering mechanism, which suppresses transport at large incident angles and restores angular symmetry. These features are quantitatively captured by an analytical equation for coherent multiple scattering, which accurately reproduces the zero-bias differential conductance. Our results establish a direct connection between microscopic valley filtering and macroscopic transport observables, demonstrating that differential conductance measurements provide a clear signature of both the Andreev reflection regime and the efficiency of valley filtering.

Based on the tight-binding model and Keldysh nonequilibrium Green's function method, we investigated single-parameter non-adiabatic quantum charge and spin pumping in a-T3 lattice ribbons featuring a flat band that cuts through the two Dirac cones. The key finding is that when a= 0, the pumped current originating from the zigzag edge band remains finite, whereas that contributed by the flat band exhibits an exponential increase, characterized by a series of current peaks when the driving frequency matches the energy gap between the flat and propagating bands. In contrast, for a nonzero a, new pumped currents arise due to the interactions between the zigzag edge band and either the flat or propagating bands. Spin-polarized currents can be realized by applying a local ferromagnetic insulator to the boundary atoms to modulate the energy bands of the flat and propagating bands, and the spin polarization can be transformed from 1 to-1 by manipulating the driving frequency of the pumping potential. These findings pave the way for spintronic devices based on a-T3 lattice model materials.

Ferromagnetic bearded zigzag graphene nanoribbons (GNRs) are proposed as promising candidates for achieving robust spin polarization and giant magnetoresistance (GMR). Numerical calculations using non-equilibrium Green’s function, based on a GNR device with width in the order of 10 nm, reveal significant spin-dependent transport properties under varying magnetizations and Fermi energy. In the parallel magnetization configuration, the conductance exhibits quantized plateaus of e2/h, corresponding to perfect spin polarization. In the antiparallel configuration, the conductance vanishes, ensuring a high GMR ratio. The spin-dependent band structure analysis demonstrates the half-metallic nature of the bearded GNRs, which act as conductors for one spin orientation while insulating the opposite. The spin-polarized current remains robust even in the presence of Anderson disorder, ensuring stability against localized scattering effects. The proposed GNR device operates as an efficient spin filter and valve. These findings highlight the potential of ferromagnetic GNRs for advanced spintronic applications, enabling electrically controlled spin-polarized transport and giant magnetoresistance in nanoscale devices.

We study the crossed Andreev reflection and the nonlocal transport in the staggered graphene/superconductor/periodic line defect superlattice (LDGSL) junctions. The staggered pseudospin potential in the left graphene electrode suppress the local Andreev reflection, while the elastic cotunneling of K' valley electrons is inhibited due to the exclusive rightward motion of K valley electrons in the right LDGSL electrode, thereby enabling the realization of dominant intravalley crossed Andreev reflection for incident electrons from the K' valley. Meanwhile, the intravalley elastic cotunneling occurs while both local Andreev reflection and crossed Andreev reflection are completely eliminated for incident electrons in the K valley. Furthermore, the probability of intervalley crossed Andreev reflection scattering is significantly lower than that of intravalley CAR scattering across a broad range of incident angles and electron energies. Our results are helpful for designing the flexible and high-efficiency Cooper pair splitter based on the valley degree of freedom.

Previous trackers based on Siamese network and transformer do not interact with the feature extraction stage during the feature fusion, excessive weight of the target features in the template area when the target deformation is large during feature fusion, causing target loss. This paper proposes a target tracking framework with target perception based on Siamese network and transformer. First, feature extraction was performed on the template area and search area and the extracted features were enhanced. A concatenation operation is used to combine them. Second, we used the feature perception obtained during the final stage of attention enhancement by searching for images to rank them and extracted the features with higher scores to enhance the feature fusion effect. Experimental results showed that the proposed tracker achieves good results on four common and challenging datasets while running at real-time speed with a speed of approximately 50 fps on a GPU.

We theoretically investigate the generation and manipulation of pure spin current in a zigzag graphene nanoribbon with the quantum pumping effect, where a magnetized bearded graphene nanoribbon is inserted into the zigzag graphene nanoribbon. It is found that the spin polarized current can be generated for nonzero Fermi energy while the pure spin current will occur for zero Fermi energy in singleparameter quantum pumping. However, the pure spin current is independent of the driving frequency of the AC field and can be easily modulated by tuning the driving frequency or the phase difference in two-parameter quantum pumping. Moreover, the pumped spin current has a cosine function relationship with respect to the phase difference due to spin splitting. This indicates a useful method for manipulating the pure spin current in graphene nanoribbons and is important for spintronics applications.

We theoretically investigate nonadiabatic quantum spin pumping in zigzag/bearded graphene nanoribbons, in which two bearded graphene nanoribbon regions are deposited by local ferromagnetic insulators to induce spin splitting. We show that spin-polarized currents, spin separation, and even pure spin pumping can be achieved by tuning the driving frequency and the magnetization orientations of the two ferromagnetic insulators. Meanwhile, for the two ferromagnetic insulators with antiparallel/parallel magnetization orientation, the left and right electrodes can simultaneously generate an equal amount of pumped currents with opposite/same spin polarization. Moreover, in the two-parameter spin pumping regime, the flowing directions of the pumped currents can be tuned by the phase difference. This suggests a useful method for manipulating and separating spin in graphene nanoribbons, which is important for spintronic applications.

In this paper, we propose a novel Transformer-based target tracking framework. In previous Transformer-based trackers, the decoder predicted the spatial location of the target object by learning the query embeddings. However, learned embeddings do not have corresponding physical representations, which makes it impossible to focus on a specific region. In order to make the target query have clear physical meaning, we design the target query in the tracker as the target query based on anchor point. In other words, the anchor is encoded as the target query. In addition, we applied an attention variant RCDA (row-column decoupled attention) that decouples key 2D features into 1D row features and 1D column features and then performs row attention and column attention sequentially. The application of RCDA can achieve better tracking results than basic attention. Our approach is end-to-end and does not require postprocessing steps. Under the same ResNet-50 backbone network, Anchor STARK’s AO score reaches 0.694 https://github.com/Renyiam/Anchor-STARK .

Recently, Transformer networks have been used for feature extraction and calculation of similarity in object tracking. This new structure is called one stream structure, and has achieved good results. However, the one stream structure of the Transformer tracker has too many network parameters, which limits the tracking speed of the network. For this reason, this work has designed a one stream Transformer structure that uses soft split operations to significantly reduce model parameters and computational complexity. To further improve the accuracy of tracking, this work proposes a multi-level residual perception structure to enhance the feature information of the target and reduce the background feature information, thereby enhancing the fore-ground and background discrimination ability of the model. To prove the speed and accuracy of this method, this work not only compared it with algorithms using deep neural network models, but also compared it with UAV tracking algorithms using shallow networks. Experimentally, the UAV123 dataset reached the level of SOTA; the inference speed can reach 130 FPS.

Graphene, a two-dimensional material with remarkable electronic properties, offers significant potential for valley-based electronic devices. In this study, we explore a novel mechanism to achieve valley-dependent, near-perfect crossed Andreev reflection (CAR) in graphene-based junctions by utilizing the valley degree of freedom in a graphene/superconductor/line defect superlattice (LDGSL) structure. The LDGSL introduces unique valley-filtering effects. By incorporating staggered pseudospin potentials and intrinsic spin-orbit coupling in the left graphene electrode, the system selectively enhances CAR for electrons in the K ' valley, while simultaneously suppressing local Andreev reflection and elastic cotunneling (ECT). Numerical simulations reveal that CAR is nearly perfect for K ' valley electrons with spin-up, while for K valley electrons with spin-down, only ECT is observed. Our results demonstrate the viability of this approach for valley-polarized CAR in graphene/ superconductor junctions, providing a pathway for the development of valley-based quantum information devices.

We study the crossed Andreev reflection and the nonlocal transport in the staggered graphene/superconductor/periodic line defect superlattice (LDGSL) junctions. The staggered pseudospin potential in the left graphene electrode suppress the local Andreev reflection, while the elastic cotunneling of K′ valley electrons is inhibited due to the exclusive rightward motion of K valley electrons in the right LDGSL electrode, thereby enabling the realization of perfect intravalley crossed Andreev reflection for incident electrons from the K′ valley. Meanwhile, the intravalley elastic cotunneling occurs while both local Andreev reflection and crossed Andreev reflection are completely eliminated for incident electrons in the K valley. Furthermore, the probability of intervalley crossed Andreev reflection scattering is significantly lower than that of intravalley CAR scattering across a broad range of incident angles and electron energies. Our results are helpful for designing the flexible and high-efficiency Cooper pair splitter based on the valley degree of freedom.

Molecular dynamics simulations show that two poly(para-phenylene) (PPP) chains can self-assemble helically to form a regular double-helix structure under the inducement of a fullerene molecule. The cross section of the PPP double helix shows a dumbbell-like shape consisting of two highly strained bulbs on two edges. The contribution of system energy and each energy component to the helical self-assembly is discussed, and the conditions and mechanism are explained. The fullerene diameter, PPP length, temperature, and relative position all have great influence on the helical self-assembly process. The thermal stability of the formed double helix is further tested. Multiple fullerenes, arranged in a string, can easily cause the helical self-assembly of two PPP chains. Three to six PPP chains have a certain probability of forming regular multiple helices under the inducement of fullerenes with an appreciate diameter. This work provides a new idea and theoretical basis for the controllable fabrication of regular helical polymers and related functional nanodevices.

Representation learning of users and items is the core of recommendation, and benefited from the development of graph neural network (GNN), graph collaborative filtering (GCF) for capturing higher order connectivity has been successful in the recommendation domain. Nevertheless, the matrix sparsity problem in collaborative filtering and the tendency of higher order embeddings to smooth in GNN limit further performance improvements. Contrastive learning (CL) was introduced into GCF and alleviated these problems to some extent. However, existing methods usually require graph perturbation to construct augmented views or design complex CL tasks, which limits the further development of CL-based methods in the recommendation. We propose a simple CL framework that does not require graph augmentation, but is based on dropout techniques to generate contrastive views to address the aforementioned problem. Specifically, we first added dropout operation to the GNN computation, and then fed the same batch of samples twice into the network for computation. Using the randomness of dropout, a pair of views with random noise was obtained, and maximizing the similarity of the view pairs is set as an auxiliary task to complement the recommendation. In addition, we made a simple modification to the computation of the GNN to alleviate the information loss due to embedding smoothing by means of cross-layer connected graph convolution computation. We named our proposed method as Simple Contrastive Learning Graph Neural Network based on dropout (SimDCL). Extensive experiments on five public datasets demonstrate the effectiveness of the proposed SimDCL, especially on the Amazon Books and Ta-Feng datasets, where our approach achieves 44% and 43% performance gains compared to baseline.

The line defect of graphene has significant applications in valleytronics, which has received extensive attention in recent years. It is found experimentally that there exists local deformation around the line defect. Current studies generally believe that the influence of local deformation on the valley transport properties can be negligible, because the modifications to the nearest neighbour hopping energy is less than 5% under the small deformation. Based on the first-principles calculations and the non-equilibrium Green’s function method, we investigated the effect of local deformation on the valley transport properties of two different kinds of line defects, the 58 ring line defect and the 57 ring line defect. It is found that for the 58 ring line defect, the effect of local deformation on the valley transmission coefficient is not evident at lower energies. However, at higher energies, the impact of local deformation is obvious, and the maximum valley transmission coefficient does not decrease with increasing energy, but can be maintained 1 within a large energy range. In contrast, the influence of local deformation on the valley transmission coefficient of the 57 ring line defect indeed can be negligible, regardless of the level of energy. Further investigation indicates that the change of the C—C bond length connected to the two defect atoms in the 58 ring plays a key role in the transmission of the valley states across the line defect. If this part of the influence is not taken into account, the valley transmission coefficient is nearly unaffected by the local deformation. The valley state enters the right side of the line defect directly through the bond connected to the line defect, so the change in bond length connected to the line defect has a significant impact on the valley transmission. This special structure does not exist in the 57 ring, where the valley states will have to pass through a narrow region containing 57 ring to enter the right side of the line defect, resulting in different valley scattering phenomena. By constructing two parallel line defects, the 100% valley polarization can be achieved in a large angular range with the 58 ring line defect. The finding has important implications for the design of graphene line defect based valley filters.

Abstract Learning graph structure-based representations of user-item interaction data has become the core of modern recommender systems, and graph neural networks (GNNs) show great potential for mining high-quality user-item representations. Therefore, GNN-based collaborative filtering (CF) models have been very successful. However, we believe that further improvement in CF model performance is limited owing to data sparsity and interaction noise in CF. Consequently, the GNN-based CF model is highly prone to problems such as overfitting and poor generalization after multiple graph convolution operations with limited training data. To resolve these problems, we propose a new solution,That is, contrast learning based on stochastic masking and feature-level enhancement(MFCL); specifically, we perform a random masking operation on the resulting embeddings after each layer of graph convolution to prevent the model from learning relations containing excessive sampling noise. In recommender systems, comparative learning better resolves the data sparsity problem than CF because it can extract self-supervised signals from raw data. To further enhance the robustness of the model, we adopt a robust strategy based on feature enhancement. This strategy, called feature blending, blends the feature information of the original graph to obtain an enhanced graph, thereby strengthening the representation learning ability of the nodes. We experimentally demonstrate the superiority and effectiveness of the proposed method using two datasets (MovieLens-1M and Gowalla).

We explore the influence of strain on the valley-polarized transmission of graphene by employing the wave-function matching and the non-equilibrium Green's function technique. When the transmission is along the armchair direction, we show that the valley polarization and transmission can be improved by increasing the width of the strained region and increasing (decreasing) the extensional strain in the armchair (zigzag) direction. It is noted that the shear strain does not affect transmission and valley polarization. Furthermore, when we consider the smooth strain barrier, the valley-polarized transmission can be enhanced by increasing the smoothness of the strain barrier. We hope that our finding can shed new light on constructing graphene-based valleytronic and quantum computing devices by solely employing strain.

To meet the demand for accurate recommendation and personalized learning in online education, an online course recommendation algorithm with multilevel fusion of user features and item features is proposed for content‐based recommendation systems applied to online education courses that have weak generalization ability and cannot cope well with data sparsity. The algorithm is improved on the deep learning recommendation algorithm named Wide & Deep Learning (WDL), which is a CTR (Click‐through rate prediction) prediction algorithm. MMF (Multi‐level Fusion Feature) uses collaborative filtering to replace linear methods in the wide part of WDL, introduces feature interactions to model user representations and item representations separately, and uses ResNet (Residual Network) ideas to improve deep neural network (DNN) in the deep part to reduce the performance degradation caused by overfitting. The experimental validation was conducted on the online education data set and the public data set MovieLens‐1M, and the AUC was improved by 1.21% and 1.46%, respectively. Meanwhile, the effect brought by this improved algorithm is interpretable.