Highly sensitive photovoltage detection in gigahertz/terahertz (THz) graphene sensors enables direct probing of electron spin resonances (ESRs) and provides access to the intrinsic spin–orbit coupling (SOC) and sublattice-potential gaps in graphene-based systems. In this work, we employ a sub-THz photoconductivity-based technique to investigate spin-flip transitions in bilayer graphene and graphene/MoS 2 heterostructures at low temperatures. In both systems, three spin resonances are observed corresponding to Zeeman, intrinsic SOC, and sublattice-potential splittings. The resonance frequencies and extracted [Formula: see text]-factors and spin relaxation times are nearly identical in the two structures, indicating that the proximity of MoS 2 does not significantly enhance graphene’s SOC under the present conditions. The energy gaps show a pronounced temperature dependence, increasing steadily as the temperature decreases and suggesting a contribution from thermally induced strain or electron–phonon coupling. Notably, all resonances are accompanied by a characteristic drop in photovoltage, consistent with spin-dependent recombination of photoexcited carriers. These findings establish THz photovoltage spectroscopy as a powerful, contact-free, and on-chip method for probing spin-dependent electronic structure and relaxation phenomena in the material forming the detector.
We review recent progress in the development of graphene-based topological structures for terahertz (THz) plasmonic applications. Our focus is on perforated graphene layers (PGLs), composed of coplanar graphene microribbon (GMR) arrays interconnected by graphene nanoribbon (GNR) bridges. These structures exhibit resonant plasmonic behavior and nonlinear inter-ribbon transport, enabling efficient THz detection through rectification and bolometric mechanisms. The plasmonic frequency is tunable via bias-induced carrier injection. Depending on the structural and transport parameters, the PGL devices can exhibit S-type current–voltage characteristics due to strong carrier heating and double injection feedback. Numerical analysis of the responsivity and spectral characteristics demonstrates strong resonance effects and indicates the feasibility of compact, high-performance THz sensors and photomixers based on these structures.
We experimentally studied the optical properties of a new type of AlGaN/GaN grating-gate terahertz plasmonic crystals with and without graphene on the top of the metal grating. It was found that CVD graphene removes the localized plasmon mode observed in the plasmonic crystals without graphene at the gate voltages below the threshold and only weakly affects the delocalized modes seen above the threshold. These effects were explained by the high resistance of the CVD graphene, which does not screen the plasmon modes but provides good control of the electron concentration at the AlGaN/GaN interface due to the Schottky barrier formed on AlGaN.
The output power of photovoltaic (PV) plants is primarily influenced by solar radiation, which is closely related to their geographical location. As a result, PV output exhibits significant spatiotemporal correlation characteristics. To address the inherent volatility and complexity in PV power forecasting, an ultra-short-term PV power output prediction method based on multi-dimensional IoT data collection and Attention-BiLSTM is proposed. This paper first extracts the correlation between geographical locations and weather conditions across multiple photovoltaic stations. Then, through an attention mechanism, we compute a weight matrix. This attention matrix is subsequently fed into a BiLSTM for training iterations, further uncovering and leveraging the interrelationships among multiple stations in both geospatial and photovoltaic power generation data. By combining the historical data of 10 adjacent PV power stations collected by IoT with weather forecast data, a new attention matrix is generated, enabling the model to focus on spatial environmental variations at the same time point through deep learning. Compared with existing methods, the proposed model shows better accuracy in forecasting random environmental disturbances. It improves the forecasting robustness and accuracy under random disturbances such as cloud and wind speed changes, and opens up new ways for efficient utilization of the PV energy.
Data collected from distributed photovoltaic systems via the Internet of Things (IoT) exhibit randomness and continuous data gaps, rendering traditional data imputation methods prone to low accuracy. This paper proposes a “data-physical” driven photovoltaic output data completion model based on xLSTM-Hankel tensor CP decomposition. The model is designed with a dual branch structure. Branch 1 employs xLSTM to fully extract the global temporal features of PV data, while Branch 2 uses the Hankel tensor CP decomposition method, which accounts for the intermittent nature of PV output, to enhance the model’s generalization capability. Finally, data repair experiments were conducted using actual PV data collected from a northern province in China. The experimental results show that the filling algorithm in this paper has an RSE that is 2.08% higher than that of the GAN algorithm and an MAE that is 4.06% lower when faced with long-term data missing.
In terahertz (THz) applications, Field-effect transistors (FETs) have emerged as prime candidates for the next generation of THz and sub-THz electronics. One of the main advantages of TeraFETs over state-of-the-art commercial THz electronics based on Schottky diodes is their ability to tune the plasma frequency over a wide range via gate voltage, which adjusts the sheet carrier density (electrons or holes) in the device channel. In this paper, we demonstrate that employing a Quantum Channel (QC) design, where the back barrier is positioned in close proximity to the front barrier, resulting in a channel thickness between 1 and 10 Bohr radii, offers significant advantages for TeraFET applications. This configuration enables increasing the maximum achievable density. In electro-optics applications, the QC design provides a unique opportunity for implementing an opto-FET, where optical absorption is modulated by the gate voltage. The key advantage of the QC-HEMT is the ability to shift the absorption edge by a large energy on the order of the Fermi level by changing the sheet carrier concentration in the device channel by modulating the gate voltage and, thereby, modulating the Moss–Burstein shift, effectively altering the optical band gap sensed by incoming radiation optical radiation by an electric field. A large sheet carrier concentration achieved in QC-HEMTs allows for a much broader modulation range of plasma frequencies and facilitates frequency multiplication via nonlinear plasmonic resonance (approximately 3 times higher). A large sheet carrier concentration also facilitates frequency multiplication via nonlinear plasmonic resonance. Such nonlinear frequency rectification dramatically increasing the frequency range (up to 1 THz or even higher) at much higher achievable powers, since the power is introduced via uniform gate voltage pumping by large-area grating gate structures. For an optical QC-HEMT, a large Moss-Burstein shift makes the optical QC-HEMT design superior as an optical modulator and optical spectrometer.
In this work, we report on terahertz rectification, up to 3.9 THz, using an asymmetric dual-grating gate graphene-based field effect transistor. The device, at 8K, was excited by terahertz radiation at three tones (2.5, 2.9 and 3.9 THz). A maximum of photocurrent, around 100 pA, was measured for the excitation at 3.9 THz. This intensity was increased by a factor of 3 when the bias of both top gates were of opposite sign. This behavior was explained as due to the ratchet effect induced by the asymmetric structure and terahertz radiation. The photocurrent remains unchanged when the modulation frequency was increased up to 2 kHz demonstrating the high speed response feature of the device.
The plasma oscillations in high-mobility field-effect transistors (HEMTs) have emerged as a key physical mechanism for manipulating electromagnetic radiation in the sub-terahertz (sub-THz) and THz frequency ranges. These collective electron excitations can be excited and tuned electrically offering a compelling route to compact, integrable components for a wide range of next-generation technologies, including sixth-generation (6G) wireless networks, high-resolution biomedical and chemical spectroscopy, industrial process monitoring, and advanced security and defense systems. For these applications, plasmonic crystals – periodic arrays of many strongly coupled FET channels– are particularly promising. In this work, we report on a new class of collective excitations in plasmonic crystals termed rotonic plasmons, which arise at plasmonic mode crossings and exhibit a parabolic dispersion law reminiscent of soft-mode and roton-like spectra. We show that uniform gate modulation across plasmonic crystal unit cells induces periodic variations in the sheet carrier concentration and, consequently, in the plasma frequency. This time-periodic modulation drives nonlinear plasmonic parametric resonances enabling RF-to-THz conversion. By solving the generalized Mathieu equation with damping, we demonstrate that high-amplitude gate pumping enables frequency multiplication and, at cryogenic temperatures (77K) leads to parametric instabilities due to enhanced electron mobility. In plasmonic crystals with lower mobility, RF-to-THz conversion can instead be realized via periodic short-pulse excitation, a regime we introduce as Time-Domain Frequency Multiplication (TDFM). Investigation of AlGaN/GaN low-high plasmonic crystals confirm their potential as tunable, compact THz sources.
Elliptical concrete-filled steel tubes (CFST) are structural elements where concrete is embedded in an elliptical steel tube, offering significant advantages such as effective axial compression resistance. These structures are widely applied in civil engineering and critical infrastructure like airport terminals and bridges. While most CFST research focuses on circular and rectangular cross-sections, there is limited study on elliptical CFST. The emergence of neural networks offers a novel approach to calculate the compressive capacity of elliptical CFST. Neural networks can learn from sample data without predefined assumptions, uncovering nonlinear relationships between the data and results. Furthermore, they can predict new variables, overcoming the constraints of traditional experimental and finite element methods. This paper proposes an advanced calculation method to quickly and accurately assess the axial compressive bearing capacity of elliptical CFST short columns. The model integrates neural network capabilities with unified strength theory, simulating the bearing capacity of elliptical CFST stub columns. Through simulation training, the method is validated for calculating the compressive performance of elliptical CFST. The findings demonstrate that this approach not only enhances computational efficiency but also provides a reliable and optimized tool for structural evaluation. The calculation efficiency of the compressive capacity of elliptical CFST based on a neural network is increased by about 23.7%.
Lane changing is a high-risk maneuver in driving that plays a crucial role in road safety. Before executing a lane change, accurately assessing potential risks in the driving environment is essential to ensuring a successful maneuver. To provide readers with a comprehensive understanding of the characteristics of vehicle lane-changing models, this paper examines the key factors influencing lane-changing behavior in different road environments. These factors include driver behavior characteristics, the impact of the lane-changing environment, the vehicle’s driving state, and the dynamic status of surrounding vehicles — all of which significantly affect lane-changing decisions. Existing lane-changing models can be broadly categorized into three types: game theory-based models, reinforcement learning-based models, and deep learning-driven models. Based on specific lane-changing scenarios and optimization objectives, these three types of models are further subdivided, and their respective advantages and disadvantages are analyzed in detail. Furthermore, the core challenges of various lane-changing models are explored from two key perspectives: input variables and inference algorithms. Finally, through a comprehensive analysis of these three model types, future research directions are proposed.
The spread of infectious diseases is affected by a variety of factors such as climate, environment, population movement, and social behavior, and the interactions among these factors are complex and difficult to accurately quantify. At the same time, the mutation and adaptation of pathogens make the transmission pattern of infectious diseases constantly change, which increases the response time for early warning. To comprehend the dynamics of infectious disease transmission patterns amidst global climate change, this study employs neural networks to forecast such patterns under these altered climatic conditions. Initially, we standardize the impact data on infectious disease transmission, which is influenced by global climate change. Subsequently, we employ principal component analysis (PCA) to extract the key components from these data. Based on this theoretical foundation, we construct a prediction model for infectious disease transmission patterns. To enhance the model’s accuracy, we utilize a genetic algorithm (GA) for optimization, ultimately achieving precise predictions of transmission patterns. The experimental results demonstrate that the method put forward can enhance the prediction effect, and the prediction response time is lower than 14.7 s, which has a greater application value.
The rapid evolution of intelligent logistics has driven the need for efficient and adaptive warehouse scheduling solutions, particularly in dynamic and uncertain environments. Traditional scheduling methods, including rule-based heuristics and mathematical optimization models, struggle to handle real-time variability and computational complexity, limiting their practical deployment. These conventional approaches often fail to integrate real-time data and adapt to changing supply chain conditions, leading to inefficiencies in routing, resource allocation, and delivery times. To address these challenges, we propose an adaptive logistics optimization framework that integrates reinforcement learning with real-time decision-making. Our model formulates warehouse scheduling as a multi-objective optimization problem, balancing cost, efficiency, and service quality while incorporating real-time traffic conditions, shipment demands, and uncertainty modeling. We leverage a dynamic reinforcement-based strategy that continuously refines decision-making using historical data and real-time feedback. The approach is designed to optimize vehicle routing, order fulfillment, and fleet coordination under stochastic conditions. Experimental evaluations demonstrate that our proposed method outperforms traditional optimization techniques, achieving significant improvements in cost reduction, delivery efficiency, and adaptability to dynamic disruptions. This research contributes to the advancement of intelligent logistics by offering a robust, data-driven scheduling optimization framework suitable for modern supply chain environments.
With the rapid development of smart grids, distribution networks are becoming increasingly intelligent and complex. However, their structural diversity and high operational uncertainty present significant challenges to ensuring secure and stable performance. To address these issues, this paper presents an AI-powered edge impedance analysis approach for risk situation awareness in smart distribution networks. The proposed method leverages edge computing to conduct real-time impedance-based monitoring and integrates artificial intelligence to enhance the real-time identification of abnormal conditions using AI-assisted analysis, classification, and fault prediction capabilities. First, the operational characteristics and risk types of distribution networks are systematically analyzed. Then, an AI-driven framework combining edge impedance sensing with intelligent analysis models is designed. Finally, a case study is conducted to validate the effectiveness and practicality of the proposed approach. The results demonstrate that this method significantly improves the timeliness and accuracy of risk awareness, offering valuable support for smart grid operation and management.
As cyber-attacks and network vulnerabilities get more sophisticated, optimizing protection systems utilizing advanced computational methodologies is critical. Big data approaches are a viable way to simulate network attacks and improve protection measures. But conventional approaches frequently can’t handle massive volumes of real-time data or quickly adjust to changing threats. The goal of the research is to create a large data-driven computer-aided network attack simulation and defense system that uses advanced Machine Learning (ML) techniques to optimize defensive techniques, improve threat detection, and increase system adaptability. A hybrid system that combines big data analytics with Hyperbolic Tangent Particle Swarm Optimized Decision Tree (HTPSO-DT) methods was presented to simulate and predict possible cyber-attacks. Big data refers to the methodologies for collecting, processing by min–max scaling, and extracting insights from diverse, high-volume, and high-velocity data sets using Discrete Wavelet Transform (DWT). The system simulates attack scenarios and optimizes defense responses using real-time network traffic data; behavior analysis and predictive modeling. The defense system adjusts by continuously learning from simulations and continually improving its techniques. The proposed method of HTPSO-DT has performed and achieved precision at 99.12%, recall at 99.15%, [Formula: see text]1 score at 99.17%, [Formula: see text]2 score at 99.09%, [Formula: see text]beta score at 98.95%, and ROC-AUU at 0.88. The method significantly improves attack detection accuracy and reduces defense response time. The suggested solution improves the effectiveness of network defense strategies by integrating big data and ML, allowing for real-time, adaptive protection.
In the context of the AI era and the digital transformation of society, the employment and entrepreneurial intentions of university graduates are increasingly influenced by both technological evolution and place-based emotional attachment. As critical agents of regional innovation, graduates possess the potential to drive economic revitalization, especially in rural areas. Drawing upon the Theory of Planned Behavior (TPB), this study incorporates two extended constructs — Place Attachment (PA) and Entrepreneurial Self-efficacy (ESE) — to explore their impact on graduates’ intentions to return to their hometowns for entrepreneurship. A sample of 1,151 university graduates from diverse Chinese institutions was analyzed using structural equation modeling to identify the relational pathways among the variables. The findings reveal that PA and ESE significantly enhance Hometown-based Entrepreneurial Intention (HEI). Additionally, graduates’ perceptions of artificial intelligence play a contextual role by shaping opportunity recognition and influencing their readiness to adopt AI-driven tools in local entrepreneurial ecosystems. The study offers practical insights for regional development policy by highlighting the need to foster both emotional ties to place and digital competencies among youth. It suggests that future policies should support AI-enabled entrepreneurship, particularly in urban and rural transition zones where a strong sense of belonging coexists with emerging digital infrastructure. This dual emphasis is essential for bridging urban–rural divides and promoting sustainable, innovation-led local development.
Objective: This study proposed an integrated and interoperable data collection framework for ophthalmic clinical research based on future networks and edge computing by incorporating electronic data capture (EDC), medical imaging storage, and deep learning-based disease classification. Methods: The framework integrates OpenClinica, OpenEMR, and Bluelight, and uses future networks and edge computing to achieve efficient management of imaging clinical trial data. DICOM images are uploaded to a picture archiving and communication systems (PACS) for storage and retrieval. A retrained Inception V3 model is deployed via a Flask RESTful API to assist in disease classification. To enhance interoperability, structured reports compliant with FHIR standards are generated through a reporting module. Additionally, data mapping is performed to align the CCD output from OpenEMR with the CDISC output in OpenClinica, enabling standardized data exchange. The Elasticsearch platform supports clinical cohort retrieval and analysis. The feasibility of the framework is evaluated through a pilot study involving retinal imaging clinical trials. Results: The framework demonstrated the capability to integrate and manage eCRFs in an EDC system with images stored in PACS and OpenEMR, enhancing the management of clinical trials involving medical images. By combining EDC with AI analysis of medical images, the framework streamlines data flow in clinical trials. In the pilot retinal imaging trial, retinal eCRFs, images, and SRs were successfully collected, and SRs were retrievable via a search engine. Conclusion: This framework offers a feasible solution for clinical researchers to collect data in an integrative and interoperable manner, warranting further validation in other medical imaging clinical trials. The system is architected over 5G to exploit its ultra-low latency and high throughput for faster EDC-PACS data transfers versus traditional architectures.
Amid the rapid integration of Internet of Things (IoT) technologies in agriculture, digital traceability, sensor-driven quality monitoring, and smart packaging are reshaping consumer expectations. It is therefore critical to understand how to translate these capabilities into strong brand perceptions. This study frames three core brand value dimensions, such as functional, emotional, and symbolic, within an IoT-enabled context and applies the Stimulus-Organism-Response (SOR) model alongside Brand Core Value Theory to explore how these values drive brand associations through consumer attitudes. Drawing on 368 valid questionnaires and Partial Least Squares Structural Equation Modeling (PLS-SEM), we demonstrate that IoT-reinforced functional value enhances cognitive trust and thereby strengthens brand associations; that IoT-enabled transparency and interactive engagement amplify emotional value, which influences associations via cognitive trust, emotional trust, and behavioral intention; and that data-driven storytelling and personalization underpin symbolic value, which shapes associations primarily through cognitive and emotional trust. These findings underscore IoT’s pivotal role in enabling and communicating brand value and suggest future research pathways incorporating direct IoT metrics.
As global populations age, the prevalence of Parkinson’s disease (PD), a chronic degenerative condition affecting the central nervous system, continues to rise. Traditional assessment methods for PD often face limitations due to cost and complexity, hindering early detection and timely treatment adjustments. This study introduces an IoT-enabled dynamic assessment system for evaluating leg agility in PD patients, integrating computer vision, deep learning, and distributed edge-cloud computing. The proposed method addresses variability in body shape and appearance by normalizing skeletal data and centering the coordinate system. Positions of ankle and knee joints are determined through a human posture algorithm, normalized according to individual height, and analyzed using the Gramian angular field (GAF). The study explores the effectiveness of recurrent neural networks for time-series signal prediction and convolutional neural networks for GAF image analysis. A transformer encoder model enhanced with positional encoding and a multi-head self-attention mechanism is employed for sequence analysis. The model incorporates a feed-forward layer, layer normalization, and skip connections to ensure training stability. A residual CNN, based on channel fusion, is utilized to learn GAF image features effectively. Lastly, a multilayer perceptron (MLP) is used for feature fusion, integrating feature vectors from the self-attention encoder and GAF image, offering a robust strategy for diverse feature integration. This IoT-AI integration advances personalized tele-rehabilitation, enabling scalable, real-time quantification of Parkinsonian motor symptoms for precision medicine applications.
This study investigates the impact of integrating artificial intelligence (AI) and edge computing on enhancing English learners’ cross-cultural critical thinking within intelligent learning environments. Grounded in socio-cultural theory’s concepts of “internalization” and “scaffolding,” along with the action framework for cross-cultural critical thinking competence, the study proposes four IT-assisted learning strategies: situational construction, topic resource interaction, scaffolding regulation, and human–computer interpersonal feedback evaluation. These strategies were implemented through a 16-week oral English practice program, employing a blended learning mode to foster cross-cultural critical thinking competence. Learning data, collected via edge computing and analyzed using AI algorithms, revealed significant improvements in learners’ cross-cultural awareness, critical thinking competence, and English language proficiency. This research underscores the transformative potential of education informatization, advocating for dynamic, learner-centered, and intelligent learning environments that enhance digital literacy and cross-cultural critical thinking competence. By providing a practical framework for cultivating cross-cultural critical thinking competence in the digital era, this study contributes to the advancement of IT-assisted language education and offers valuable insights for educators, policymakers, and technology developers.
The purpose of this paper is to explore the digital reconstruction of traditional patterns and their application in modern product design. The introduction first elaborates on the cultural value and artistic charm of traditional patterns, as well as the importance of incorporating traditional patterns into modern product design. On this basis, the potential application directions of digital methods in traditional pattern reshaping are explored, and the purpose and profound significance of this study are elucidated. In terms of research methods, the paper summarizes the changes and progress of traditional patterns in the historical process, deeply analyzes the application and trends of traditional patterns in modern product design, and discusses the difficulties and limitations encountered in modern design practice. Subsequently, the application foundation of digital technology in pattern design is introduced, and the algorithm model for the digital reconstruction of traditional patterns is discussed. A digital reconstruction method that combines the characteristics of traditional patterns is proposed. The experimental results show that digital reconstruction patterns perform well in terms of quality, innovation, and practicality. The case analysis further proves the application effect of digital reconstruction of traditional patterns in modern product design.
