Two-dimensional (2D) semimetal heterojunctions are emerging as a promising platform for next-generation photodetectors, with notable strengths in broadband response, gate tunability, and polarization sensitivity. This review looks at four key areas: heterojunction architecture, dark current suppression, gate-voltage-mediated performance control, and polarization detection. Type-I, type-II, and type-III band alignments are examined in terms of how they affect carrier separation and noise, with a focus on Schottky barrier engineering as a primary route to reduce thermally generated dark current. Electrostatic gating is also discussed—how it shifts the Fermi level and modifies Schottky barrier heights—and how that enables dynamic tuning of responsivity, detectivity, and speed. For polarization-sensitive detection, both linear and circular polarization regimes are evaluated, drawing on intrinsic crystal anisotropy and asymmetric device designs. A side-by-side comparison of graphene, black phosphorus, and topological semimetal systems shows clear trade-offs in performance. Remaining challenges like contact resistance, environmental stability, and scalability are critically assessed, and some future directions are pointed out, including machine-learning-assisted design and silicon photonic integration. Overall, this review offers a unified framework to help move 2D semimetal heterojunction photodetectors from lab prototypes toward practical use.

Gallium oxide-based memristors have attracted increasing attention for non-volatile memory, artificial synapses, and neuromorphic computing because of the ultra-wide bandgap, high resistance, chemical stability, and defect-tunable properties of Ga₂O₃/GaOₓ. This paper reviews the material properties, typical device structures, resistive switching mechanisms, optimization strategies, and neuromorphic applications of gallium oxide-based memristors. The switching behavior is mainly related to oxygen-vacancy migration, conductive filament formation and rupture, metal-ion-assisted switching, and interface barrier modulation. Device performance is strongly influenced by oxide stoichiometry, electrode activity, film thickness, and interface quality. Approaches like defect adjustment, changes to the electrode and interface work, designing very thin or two-dimensional Ga₂O₃ structures can enhance switching stability, reduce the operating voltage, and improve analog conductance modulation. In neuromorphic applications, GaOₓ-based memristors can emulate synaptic functions, such as emulating potentiation, depression, paired-pulse facilitation, and adjusting multilevel weights. Finally, the remaining challenges in endurance, uniformity, large-scale integration, and CMOS compatibility are discussed.

Personalisation of product rankings in e-commerce is needed because different users have different interests, demands and browsing conditions. A user-item network model can be employed to represent clicks, favourites, additions to shopping carts, ratings and purchases for personalised Top-k ranking in this paper. This review connects PageRank and Personalised PageRank with graph collaborative filtering, multi-behaviour graph learning, temporal self-supervision and industrial pre-ranking systems. According to the above research, different types of behaviour, the order of interaction and time information, graph construction methods, negative sampling, etc., can affect the ranking results. However, there are still many problems such as scarce and noisy implicit feedback, cold start, popularity bias, preference drift, scalability, privacy, fairness and lack of interpretability. In the future, research will continue to be carried out in the field of combining dynamic heterogeneous graphs, multi-behavioural objectives, causal and self-supervised learning, privacy-preserving computation and graph foundation models. In short, the user-item network model applies linear algebra, probability theory, graph propagation and representation learning to solve the problem of personalised e-commerce ranking. It is convenient to conduct a comparison of the models and decide which one to use.

Quarkonium, which is a bound state of a heavy quark and its corresponding antiquark, is similar to the hydrogen atom in particle physics, and its spectrum can provide information about the interquark potential of quantum chromodynamics. Numerical solution of the radial Schrödinger equation for charmonium uses the Cornell potential model, which combines the short-range Coulomb term and the linearly rising confining term. Fourth-order Numerov integration is used in combination with inward and outward shooting; they are connected at the classical turning point, and a node labelled energy scan is added to link each root to an excited state. First, this way is validated against the analytical hydrogen spectrum; it has been reproduced to better than one millielectronvolt. Then, based on the experimental measurement of the three lowest S-wave vector states of charmonium, the mass of the charm quark, the strong coupling constant and the confinement coefficient are determined by fitting, and a coarse random search followed by Nelder-Mead simplex refinement is employed for the fit. The converged fit is within 0.01 MeV of the experimental masses, and its root mean square deviation is 0.005 MeV; however, the fitted string tension is lower than its independent value and the fitted coupling is higher than its independent value, which may be due to the omission of spin-spin interaction, relativistic correction and open charm coupled-channel effects. The calculated wave functions have a node structure and their spatial extent grows slowly; it is characteristic of confinement.

This paper, based on Lyapunov stability analysis, designs a more flexible triggering mechanism for a class of nonlinear systems with time-delay effects. The proposed event-triggered pulse control scheme can ensure global input-to-state stability (ISS) and strictly prevent Zeno behavior. The advantages and optimized performance of the findings are verified through typical numerical examples. Simulation results show that, under the same external disturbance conditions, traditional schemes generate a large number of pulse events, whereas the adaptive mechanism proposed here only requires a few pulses to maintain system stability, significantly reducing communication resource usage. At the same time, under strong disturbance conditions, the system's peak state deviation is noticeably suppressed, showing better robustness and lower triggering frequency. Considering that in practical engineering, pulse actions can introduce additional delays, future research will further analyze scenarios where pulse execution has delays and explore delay-related pulse issues within the proposed triggering framework.

Social media platforms have become important spaces where learners participate in online educational discussion, express learning experiences, and form peer interaction. Comments and replies contain rich emotional information related to confusion, frustration, encouragement, and recognition. Focusing on learner emotion transfer in educational social media discourse, this study applies natural language processing to public educational video comments, Reddit learning community posts, and open course discussion texts. It builds an experimental process that includes data collection, manual annotation, RoBERTa-based emotion recognition, and reply-chain transfer modeling. The results show that RoBERTa-base achieves strong performance in emotion classification, with an Accuracy of 0.846 and a Macro-F1 of 0.804. It performs better than VADER and TF-IDF plus SVM baselines. The emotion transfer results further indicate that frustration tends to persist in reply chains, but it can also shift into encouragement. Confusion is more likely to trigger supportive responses, while joy can promote positive interaction diffusion. These findings show that emotion in educational social media discourse is not isolated expression. It is continuously transmitted and regulated through interaction structures. The results provide computational evidence for affective support, comment moderation, and learner interaction guidance on online education platforms.

Diabetes is a growing chronic disease burden, and early screening is important for identifying individuals who may require further clinical testing. However, laboratory biomarkers such as glycated hemoglobin (HbA1c) may not always be available in low-cost population screening. Using the NHANES 2017-March 2020 pre-pandemic public-use data, we examined whether non-laboratory variables can support diabetes risk screening in adults. Adults aged 20 years or older were included. Diabetes status was defined using self-reported doctor diagnosis and HbA1c-based outcome labelling, while laboratory biomarkers were excluded from the main predictor set. The main non-laboratory predictors included demographic, socioeconomic, anthropometric, blood pressure, and lifestyle variables. Two tree-based machine learning models, Random Forest and XGBoost, were trained and evaluated on a testing set. Model performance was assessed using AUC, accuracy, sensitivity, specificity, F1-score, and Brier score. The final main sample contained 5,728 adults, including 1,071 diabetes cases. In the non-laboratory setting, XGBoost achieved higher AUC and much higher sensitivity than Random Forest. Adding HbA1c in a supplementary model substantially improved classification performance. These findings suggest that non-laboratory variables may support preliminary risk screening, but laboratory testing remains important for accurate diabetes classification.

Gate-oxide thickness is a first-order design variable that controls electrostatic coupling, threshold behavior, and leakage in scaled field-effect transistors. This paper presents a compact Silvaco TCAD study of a silicon MOSFET in which the oxide thickness tox is swept from 1 to 5 nm while the drain bias is fixed at VD = 0.05 V. The simulation flow builds the device structure, defines the mesh and regions, assigns material/doping parameters, solves the bias sweep, and extracts transfer characteristics, threshold voltage, and subthreshold swing. The simulated ID-VG curves show a systematic positive shift with increasing tox, indicating weakened gate control. Using the maximum-slope definition of SS and a constant-current threshold criterion of ID = 1 × 10−7 A, the extracted SS increases from 70.57 to 91.31 mV/dec, while Vth increases from 0.029 to 0.517 V as tox increases from 1 to 5 nm. These results provide a clear TCAD-based visualization of the trade-off between gate dielectric scaling, switching efficiency, and threshold-voltage engineering.

This work presents a compact-model-level comparison between the MIT virtual-source (MVS) model and a 65-nm foundry model for short-channel NMOS devices and simple load-dependent test circuits. A short-channel device schematic based on the MVS concept is first used to establish the physical picture of virtual-source-controlled transport. Two single-transistor test structures are then evaluated: a reference branch without a source resistor and a branch with a source load. Using the provided DC voltage, DC current, and AC frequency-response characteristics, the two models are compared in terms of output-voltage transition, current build-up, and bandwidth roll-off behavior. The results show that both models capture the qualitative impact of loading, but they differ noticeably in the transition location ofVDC, in the low-to-intermediate-bias rise ofIDC, and in the onset of AC roll-off. Under the present test conditions, the MVS model exhibits earlier DC transition, faster current establishment, and a later AC roll-off than the 65-nm foundry model. These observations indicate that a virtual-source-based compact description is not merely an alternative fitting form, but a physically meaningful modeling layer that can influence circuit-level predictions even in very simple transistor test structures.

Ultrafast Electron Diffraction (UED), built on the pump‑probe framework, has long been an indispensable tool at the cutting edge of interdisciplinary research spanning physics, chemistry, and biology. Who would have thought that, with its dual advantages of femtosecond temporal resolution and sub‑angstrom spatial resolution, it could directly "visualize" the ultrafast coherent coupling evolution of lattices, electrons, and spins in momentum space? This not only breaks the limitation that traditional spectroscopy can only indirectly infer molecular structural dynamics, but also truly pushes condensed‑matter physics, photochemical reactions, and transient quantum materials into a new era of real‑time atomic‑scale visualization. This paper systematically reviews the landmark breakthroughs of UED in uncovering the microscopic mechanisms of extreme nonequilibrium states of matter, covering phase transition dynamics, electron‑phonon coupling, molecular dynamics, and quantum ultrafast manipulation, fully demonstrating the irreplaceable scientific value of this technology.