Accepted
,
Abstract +
In response to the technical issue in Raman distributed optical fiber technology where the traditional meter-level spatial resolution performance is insufficient, leading to a decline in system measurement accuracy within sub-spatial resolution fiber segments along the sensing fiber, a threshold coefficient fitting technique based on a one-dimensional peak-seeking method is proposed in this study. Significant temperature measurement errors of up to tens of degrees Celsius are caused by the overlap of Raman scattering signals from non-detection regions when the detection fiber length is shorter than the system's spatial resolution. This severely limits the technology application in scenarios requiring precise temperature monitoring. To overcome the above bottleneck, a purely algorithmic approach is introduced, which reconstructs the temperature field without requiring hardware modifications. The sensing fiber was globally scanned using the one-dimensional peak-finding algorithm to precisely locate sub-spatial resolution detection fiber regions. Simultaneously, the peak intensity, full width at half maximum (FWHM), and location were extracted from the temperature rise curve within the fiber under test (FUT). Through pre-calibration experiments, a quantitative fitting model was established between peak temperature rise curves and threshold coefficients, revealing a quantitative mapping relationship between FWHM and sensing distance, as well as length of FUT. The results indicated that FWHM exhibited a significant positive linear correlation with sensing distance, independent of temperature variations. This characteristic enabled FWHM to serve as a reliable feature parameter for identifying the actual length of detection fibres. During real-time measurements, the detection fiber length was determined via the mapping model based on extracted FWHM and location. Then the corresponding threshold coefficient fitting model is selected to compensate for distorted temperature rise peaks, thereby reconstructing distributed temperature field. Experimental results demonstrated that over a 10-kilometre sensing distance, the results indicate that the application of this technique significantly enhanced the temperature measurement accuracy within the 30 cm detection fiber, achieving 1.5 °C compared to the baseline accuracy of 34.7 °C before compensation. Conclusions indicate that the proposed threshold coefficient fitting technique, through algorithmic innovation, effectively overcomes the technical limitation of deteriorating temperature measurement accuracy in sub-spatial resolution regions within Raman distributed fibre optics sensing. The constructed FWHM quantitative mapping model provides critical basis for threshold compensation, ultimately achieving precise temperature monitoring of sub-metre regions within long-distance sensing contexts. This solution features a streamlined structure, low cost, and ease of engineering integration. It offers a novel approach for long-term, high-precision temperature monitoring in fields such as power cable fault orienation, oil and gas pipeline micro-leakage early warning, and civil structural health monitoring.
Abstract +
Quantum battery is a new energy storage concept designed based on the principles of quantum mechanics, aimed at overcoming the physical limitations of traditional electrochemical batteries in terms of energy density, charging speed, and efficiency. This review provides a comprehensive synthesis of recent theoretical and experimental progress in the field, emphasizing the underlying theoretical framework and the core physical mechanisms that govern energy storage, transport, and extraction. Central attention is given to the essential roles of quantum coherence and entanglement in enhancing charging performance and enabling collective phenomena. The thermodynamic foundations of quantum batteries are introduced, including stored energy, ergotropy, capacity, power, and energy fluctuations. The review then examines the structural characteristics and charging behaviors of several representative quantum battery models in depth, including light-matter interaction batteries based on the Tavis-Cummings or Dicke framework, spin-chain batteries with various interaction types, high-dimensional (three-level and multi-level) batteries employing adiabatic and shortcut-to-adiabatic control, as well as Rydberg-atom-based batteries featuring switchable strong long-range interactions. For each model, the influence of initial states, coupling strength, system size, and excitation distribution on charging dynamics, capacity, and power scaling is systematically discussed. Furthermore, key challenges faced by quantum many-body battery models in realistic environments are reviewed, particularly in relation to their open-system characteristics. We summarize recent advances in understanding how decoherence, dissipation, and environmental noise degrade battery performance, while highlighting how non-Markovian memory effects can stabilize energy flow or partially restore lost coherence. Measurement-based feedback control, dissipative engineering, and decoherence-free subspace techniques are introduced as promising strategies to suppress decoherence, mitigate self-discharge, and extend battery lifetime. The potential quantum advantages in self-discharge suppression, energy retention, and anti-aging mechanisms are also examined. Finally, the review explores feasible implementation routes toward long-distance or wireless quantum charging, and surveys experimental platforms capable of realizing quantum batteries, including superconducting circuits, trapped ions, cavity-QED systems, optomechanical devices, and Rydberg arrays. Overall, quantum battery research is undergoing rapid expansion, and its progress not only promises transformative innovations in next-generation energy storage technologies, but also provides a powerful experimental platform for advancing quantum thermodynamics, quantum resource theory, and the physics of nonequilibrium quantum systems.
Instrumental Profile Modelling of a HighResolution Spectrograph based on Gaussian Process Regression
,
Abstract +
Aims : High-resolution spectrographs are central to modern exoplanet research and are particularly effective for detecting Earth-like planets whose radial velocity (RV) signals can be only a few tens of centimeters per second. Achieving this level of precision requires highly accurate wavelength calibration. A key factor in this process is the modeling of the instrumental profile (IP), which describes the response of the spectrograph to incoming light. The true IP of a high-resolution instrument is often complex. It may show asymmetry or extended wings and change across the detector because of optical aberrations, variations in fiber illumination, and environmental effects. These features lead to systematic errors in the measured line centers when traditional parametric models such as Gaussian functions are used, and they limit the achievable RV precision.
Methods: This work introduces a non-parametric IP modeling method based on Gaussian Process Regression (GPR). The IP is treated as a smooth function with a flexible covariance structure instead of being constrained by a predefined analytic form. GPR learns both the global structure and small-scale features of the line shape directly from the data. Since the IP varies slowly across the detector, the method divides each spectral order into several consecutive spatial segments. Each segment is fitted independently, capturing local variations. The model includes measurement uncertainties and provides a probabilistic description of the IP. Adjacent segments are linked with smooth interpolation to ensure a continuous IP across the entire order. Model performance is evaluated using reduced chi-squared and root mean square error (RMSE), allowing quantitative assessment and comparison with traditional approaches.
Results: The method is tested with laser frequency comb (LFC) exposures from the fiber-fed High Resolution Spectrograph (HRS) on the 2.16 m telescope at Xinglong Observatory. The LFC produces a dense and highly stable set of emission lines and is well suited for validating IP reconstruction. Three experiments show clear and consistent improvements. Using odd-numbered lines to predict evennumbered ones within a single exposure reduces the RMSE by 35.6% compared with a Gaussian model, showing better determination of line centers. Applying an IP model trained on one exposure to a later exposure reduces the RMSE by 42.5%, demonstrating improved stability when the model is transferred between exposures. A comparison between two channels in the same exposure shows a 37.1% improvement in calibration consistency, indicating reduced channel-tochannel systematics.
Conclusions: The results show that GPR provides a more accurate description of the instrumental profile and its spatial variation than traditional parametric models. The improved reconstruction of the IP leads to more accurate line center measurements and a more stable and precise wavelength solution. This capability is important for pushing the RV precision of high-resolution spectrographs toward the centimeter-per-second level. GPR offers a promising approach for modeling instrumental profiles and supports the precision required for detecting Earth-like exoplanets.
Methods: This work introduces a non-parametric IP modeling method based on Gaussian Process Regression (GPR). The IP is treated as a smooth function with a flexible covariance structure instead of being constrained by a predefined analytic form. GPR learns both the global structure and small-scale features of the line shape directly from the data. Since the IP varies slowly across the detector, the method divides each spectral order into several consecutive spatial segments. Each segment is fitted independently, capturing local variations. The model includes measurement uncertainties and provides a probabilistic description of the IP. Adjacent segments are linked with smooth interpolation to ensure a continuous IP across the entire order. Model performance is evaluated using reduced chi-squared and root mean square error (RMSE), allowing quantitative assessment and comparison with traditional approaches.
Results: The method is tested with laser frequency comb (LFC) exposures from the fiber-fed High Resolution Spectrograph (HRS) on the 2.16 m telescope at Xinglong Observatory. The LFC produces a dense and highly stable set of emission lines and is well suited for validating IP reconstruction. Three experiments show clear and consistent improvements. Using odd-numbered lines to predict evennumbered ones within a single exposure reduces the RMSE by 35.6% compared with a Gaussian model, showing better determination of line centers. Applying an IP model trained on one exposure to a later exposure reduces the RMSE by 42.5%, demonstrating improved stability when the model is transferred between exposures. A comparison between two channels in the same exposure shows a 37.1% improvement in calibration consistency, indicating reduced channel-tochannel systematics.
Conclusions: The results show that GPR provides a more accurate description of the instrumental profile and its spatial variation than traditional parametric models. The improved reconstruction of the IP leads to more accurate line center measurements and a more stable and precise wavelength solution. This capability is important for pushing the RV precision of high-resolution spectrographs toward the centimeter-per-second level. GPR offers a promising approach for modeling instrumental profiles and supports the precision required for detecting Earth-like exoplanets.
,
Abstract +
Bi2Te3-based materials prepared by traditional zone melting often suffer fro m poor mechanical properties. Although powder metallurgy followed by hot ext rusion can effectively enhance mechanical strength, this approach involves a len gthy, multi-step processes including powdering, sintering, and extrusion. Such a complex procedure has hindered the development of polycrystalline Bi2Te3-bas ed materials and their application in micro-thermoelectric devices. In this work, p-type Bi2Te3-based ribbons were first fabricated via melt spinning. Subsequent ly, a series of highly textured, fine-grained p-type Bi2Te3-based bulk materials were prepared by directly tiling these ribbons and consolidating them through Spark Plasma Sintering (SPS). The as-spun ribbons possess a strong texture, al ong with abundant nanostructures and defects. The subsequent consolidation, ac hieved by directly tiling these ribbons and applying Spark Plasma Sintering (SPS) without any pulverization, effectively preserved their intrinsic preferred orie ntation. This resulted in a strong (1 1 0) texture perpendicular to the pressing direction, which is distinct from that obtained via the conventional ball-milling and SPS route. The sample sintered at 743 K exhibited an orientation factor of 0.37, comparable to that of hot-extruded counterparts. Owing to this strong te xture, the sample exhibited superior electrical transport properties along the dire ction parallel to the pressure. A high power factor of 3.79 mW m-1 K-2 was ac hieved at room temperature. Furthermore, grain refinement led to a significant reduction in thermal conductivity. Consequently, a peak ZT value of 1.30 was obtained at 398 K for the sample sintered at 743 K, representing a 46% enhan cement over traditional zone-melted samples. This study provides a rapid and f acile strategy for fabricating highly textured, fine-grained, high-performance Bi2 Te3-based materials, laying a solid foundation for their engineering applications in Micro-thermoelectric devices.
,
Abstract +
With the rise and wide applications of 3D heterogeneous integration technology, inductive voltage regulators have become increasingly important for mobile terminals and high-computing-power devices, while also offering significant development opportunities for high-frequency soft magnetic films. Based on the requirements of onchip power inductors, we first review the advantages and limitations of three types of magnetic core films: permalloy, ·Co-based amorphous metallic films, and FeCo-based nanogranular composite films, with a focus on the technical requirements and challenges posed by several μm-thick laminated magnetic core films. Secondly, almost all on-chip inductors are hard-axis excited, meaning that the field of inductors should be parallel to the hard axis of the magnetic core. We thus compare the characteristics of two types of large-area film fabrication methods, i.e. applying in-situ magnetic field and oblique sputtering, both of which can effectively induce in-plane uniaxially magnetic anisotropy (IPUMA). Their impacts on the static and high-frequency soft magnetic properties are also compared. The influence of film patterning on the domain structures and highfrequency magnetic losses of magnetic cores, as well as corresponding countermeasures, are also briefly analyzed. Furthermore, the temperature stability of magnetic permeability and anisotropy of magnetic core films is discussed from the perspectives of process compatibility and long-term reliability. Although the Curie temperature and crystallization temperature of the three types of magnetic core films are relatively high, the upper limits of their actual process temperatures are affected by the thermal effects on the alignment of magnetic atomic pairs, microstructural defects, and grain size. Finally, the current bottlenecks in testing high-frequency and large-signal magnetic losses of magnetic core films are addressed, and potential technical approaches for achieving magnetic core films that meet the future demands of on-chip power inductors for higher saturation current and lower magnetic losses are outlined.
,
Abstract +
In this study, single-crystal Si3N4:Eu2+ nanowires were successfully synthesized via a direct current arc plasma nitridation method. The as-synthesized product, characterized by X-ray diffraction, X-ray photoelectron spectroscopy, energy-dispersive X-ray spectroscopy, scanning electron microscopy, and transmission electron microscopy, consists of tightly packed bundles of nanowires. These nanowires have diameters ranging from tens to hundreds of nanometers and lengths up to several tens of micrometers. Under ultraviolet excitation, the nanowires display a bright yellow emission band centered at approximately 589 nm, which is assigned to the 4f65d1→4f7 transition of Eu2+ ions. The photoluminescence properties were investigated under hydrostatic pressure up to 30 GPa. As the pressure increases, the Eu2+ emission band shows a significant and monotonic red shift at a rate of approximately 1.45 nm·GPa-1. This shift is primarily attributed to pressure-induced modifications in the energy level structure, resulting from reduced interionic distances and enhanced ionic interactions. Concurrently, the full width at half maximum (FWHM) of the emission band broadens with a pressure coefficient of about 0.8% GPa-1, which can be explained by the combined effects of an enhanced crystal field, intensified electron-phonon coupling, lattice strain, and distortion. A pressure-sensing model based on chromaticity coordinate analysis was established, demonstrating high performance with a maximum sensitivity of 0.78% GPa-1. The stable correlation between these optical parameters and applied pressure enables high-precision sensing. The developed optical sensor exhibits a suite of advantageous characteristics, including high sensitivity, a broad pressure detection range (up to 30 GPa), and excellent signal stability (maintaining 38% of the initial intensity at 23 GPa). These results indicate significant application potential for Si3N4:Eu2+ nanowires in high-pressure sensing under extreme conditions, such as deep-sea exploration, studies of planetary interiors, and monitoring of ultra-heavy construction.
,
Abstract +
This paper investigates the dynamic control of non-reciprocal propagation for vortex beams in a Rydberg atomic ensemble mediated by flying spin atomic clusters. The system comprises a target Rydberg atomic ensemble with a five-level N-type structure and two flying spin atomic clusters moving at velocity v, coupled via position-dependent non-resonant dipole-exchange interactions to form a hybrid quantum system exhibiting dipole-exchange-induced transparency. The macroscopic relative motion between the flying spin clusters and the stationary target ensemble induces optical non-reciprocity. Utilizing the split-step Fourier propagation method combined with the superatom model, we perform numerical simulations to analyze the spatial evolution of a probe Laguerre-Gaussian (LG) vortex beam. To quantify nonreciprocity, we introduce the LG nonreciprocity index CLG, defined via the normalized mean absolute intensity difference between output spots for left and right incidence. Our findings show that the spin cluster velocity v and the probe detuning (∆p) are key parameters governing the non-reciprocal response. By tuning v and ∆p, we can flexibly manipulate both the intensity and phase profile of the transmitted two-dimensional vortex wavefront. In the presence of dipole-exchange interaction, the output spot undergoes marked stretching deformation, departing from an ideal annular shape, and its stretching direction (e.g., along x or y) can be precisely switched via parameter adjustment. Moreover, the input direction of the probe beam influences the output phase pattern, producing counterclockwise phase rotation for left incidence and clockwise rotation for right incidence. This work reveals a dynamic control mechanism for non-reciprocal propagation of structured light via macroscopic motion of spin clusters and underscores the potential of dipole-exchange-induced transparent systems for designing nonreciprocal optical devices. The results provide a theoretical foundation for optical information processing and quantum communication, and suggest a viable technique for two-dimensional vortex beam shaping with broad application prospects.
Abstract +
Topological materials, characterized by symmetry-protected nontrivial band structures such as Dirac cones and Weyl nodes, exhibit a rich variety of quantum states and novel physical phenomena. These materials hold great promise for applications in quantum transport, spintronics, and nonlinear optics. In recent years, ultrafast pump-probe spectroscopy has become a powerful tool for studying nonequilibrium dynamics in quantum materials. With femtosecond temporal resolution, this technique enables direct observation of charge, spin, orbital, and lattice interactions on their intrinsic timescales, offering new insights into the coupling mechanisms in topological systems. This review summarizes the latest progress in applying ultrafast spectroscopy to topological insulators, topological semimetals, and magnetic topological materials. We first discuss the distinct relaxation pathways of surface and bulk electronic states after photoexcitation, focusing on electron-phonon scattering, surface-bulk charge transfer, and ultrafast spin conversion. We then describe population inversion phenomena in Dirac and Weyl semimetals, spin polarization dynamics induced by tilted Weyl bands, and the influence of magnetic order on topological states, including coherent phonon and magnon excitations, magnetically driven topological transitions, and terahertz pulse generation. Furthermore, we review photoinduced topological phase transitions driven by electronic correlations, lattice distortions, and magnetic order under strong optical fields, highlighting the potential for nonthermal optical control of quantum phases. Finally, we discuss future research directions, emphasizing the integration of multidimensional ultrafast spectroscopic techniques—spanning temporal, energy, momentum, and spin resolution—with advanced theoretical simulations to construct a comprehensive picture of nonequilibrium topological states. This work aims to serve as a reference for studies on the ultrafast dynamics of topological quantum materials and to promote their practical applications in high-speed, low-power information processing, spintronics, and quantum computation.
,
Abstract +
Dielectric strength (Er) is a critical factor in screening and evaluating for SF6 replacement gas. The conventional experimental methods for measuring Er are not only exceptionally time-consuming but also prohibitively expensive. This work constructed an Er prediction model for SF6 replacement gases based on machine learning methods. First, an exhaustive literature survey is performed to collect 88 high-quality experimental Er values. Second, a total of 32 insightful microscopic descriptors are accurately calculated for each compound based on density functional theory, including both global parameters and molecular electrostatic potential parameters. Furthermore, five state-of-the-art machine learning algorithms, which have been carefully modified based on five-fold cross-validation and hyperparameter optimization, are utilized to train and test the 88 experimental Er data and their relevant microscopic descriptors. Finally, the result reveals that Ada Boost regression model demonstrates superior predictive performance with a coefficient of determination of 0.90, a mean absolute error of 0.17, and a root mean square error of 0.18. Moreover, Shapley Additive exPlanations analysis is used to reveal the correlation between the microscopic descriptors and Er. The results indicate that polarizability emerges as the predominant factor significantly affecting Er, which accounts for as high as 17.3%, followed by the molecular weight (14.1%). Specifically, molecules with high α are more prone to deformation under the action of an electric field, and their electron clouds are more likely to be polarized, which has a positive correlation with Er. There is an approximately positive correlation between the molecular weight and the Er of gases. To confirm the reliability of Ada Boost regression model for Er prediction, the Er of SF6 and six known environmentally friendly replacement gases were tested within an absolute error of 0.02-0.33. This study provides a feasible pathway to accelerate the search for SF6 replacement gases.
,
Abstract +
This study employs a multi-unit thermoradiative device (TRD) for automotive exhaust waste heat recovery. A coupled model integrating radiative heat transfer, current-voltage characteristics, and fluid heat exchange is established. Based on Fourier’s law of heat conduction and thermal radiative transfer theory, the energy constraint equations, total power output, and conversion efficiency of the system are derived. The variations of exhaust temperature, TRD operating temperature, and ambient temperature with unit number are obtained through numerical simulations, thereby revealing the regulation mechanisms of voltage and semiconductor bandgap on energy conversion performance. Results show that the temperatures of the exhaust gas and the hot side of the TRD decrease with increasing unit number and also decline with increasing current at the same unit position. In contrast, the cold side of the TRD and the ambient temperature rise due to heat accumulation and cascading heating effects, and further increase with higher current, reflecting the coupling between electrical output and thermal processes. Increased voltage suppresses radiative recombination, leading to reduced current, while the electrical power reaches a maximum at a specific operating point. The total heat flux is reduced as voltage increases. Due to the nonlinear relationship between electrical power and heat flux, the efficiency attains an optimum value at a certain voltage, achieving a balance between electrical output and heat dissipation. This study demonstrates that the locally optimal power reaches a global maximum of 170.45 W at a bandgap of 0.06 eV, while the locally optimal efficiency increases monotonically with bandgap before saturating gradually. To address the inherent trade-off between power and efficiency, a target function Z defined as the product of locally optimal power and efficiency is introduced. Numerical analysis reveals that Z attains its maximum value of 49.74 W at a bandgap of 0.105 eV, effectively balancing the competing objectives of power output and energy conversion efficiency. This approach offers a new pathway for performance optimization in thermoelectric systems.
,
Abstract +
Metal-organic chemical vapor deposition (MOCVD) remains the dominant technique for the growth of III-nitride semiconductors; however, the complex growth kinetics and defect formation mechanisms continue to limit the achievable material quality and device performance. In recent years, the rapid advancement of in situ X-ray characterization techniques—particularly those enabled by high-brightness synchrotron radiation—has provided unprecedented opportunities for probing real-time structural evolution during nitride epitaxy. This review summarizes the latest international progress in in situ X-ray studies of III-nitride MOCVD growth, with emphasis on the development of in situ MOCVD growth platforms, emerging X-ray methodologies, and their applications in monitoring surface and interfacial dynamics.
We present the principles and implementation of in situ X-ray reflectivity (XRR), crystal truncation rods (CTR), grazing-incidence diffraction, and microbeam/coherent scattering techniques（XPCS） in nitride epitaxy. Using representative case studies from GaN and InGaN, we discuss how these tools reveal key dynamical processes—including early-stage nucleation, strain relaxation, step-flow behavior, alloy segregation, and interface roughening—under realistic growth conditions. Special attention is given to transient non-equilibrium phenomena such as compositional fluctuations and interface reconstruction in high-In content alloys, which remain inaccessible to conventional in situ probes.
Furthermore, we highlight emerging trends enabled by next-generation synchrotron sources, including millisecond- to microsecond-resolved measurements, nanoscale spatial mapping, and in situ coherent X-ray diffraction imaging (CXDI/XPCS). These capabilities are expected to provide direct atomic-to-mesoscale insights into island nucleation, step dynamics, defect evolution, and strain-composition coupling in complex heterostructures. Finally, we outline future research directions, such as integrating data-driven structure inversion, multi-scale modeling, and closed-loop “growth-measurement-feedback” control to accelerate the understanding and optimization of nitride epitaxy.
This review demonstrates that in situ X-ray techniques have become a powerful and indispensable bridge between microscopic structural evolution and macroscopic device performance, and will play a key role in enabling precise, controllable epitaxy of next-generation wide-bandgap semiconductor materials.
We present the principles and implementation of in situ X-ray reflectivity (XRR), crystal truncation rods (CTR), grazing-incidence diffraction, and microbeam/coherent scattering techniques（XPCS） in nitride epitaxy. Using representative case studies from GaN and InGaN, we discuss how these tools reveal key dynamical processes—including early-stage nucleation, strain relaxation, step-flow behavior, alloy segregation, and interface roughening—under realistic growth conditions. Special attention is given to transient non-equilibrium phenomena such as compositional fluctuations and interface reconstruction in high-In content alloys, which remain inaccessible to conventional in situ probes.
Furthermore, we highlight emerging trends enabled by next-generation synchrotron sources, including millisecond- to microsecond-resolved measurements, nanoscale spatial mapping, and in situ coherent X-ray diffraction imaging (CXDI/XPCS). These capabilities are expected to provide direct atomic-to-mesoscale insights into island nucleation, step dynamics, defect evolution, and strain-composition coupling in complex heterostructures. Finally, we outline future research directions, such as integrating data-driven structure inversion, multi-scale modeling, and closed-loop “growth-measurement-feedback” control to accelerate the understanding and optimization of nitride epitaxy.
This review demonstrates that in situ X-ray techniques have become a powerful and indispensable bridge between microscopic structural evolution and macroscopic device performance, and will play a key role in enabling precise, controllable epitaxy of next-generation wide-bandgap semiconductor materials.
,
Abstract +
Mode division multiplexing technology can augment the single-wavelength capacity via orthogonal eigenmodes within a multi-mode optical waveguide. In order to achieve the low-loss, small-size, multi-channel mode multiplexers, an 8-channel mode-polarization multiplexer consisting of digital structures and waveguide structures is designed, fabricated, and measured to improves the integration and transmission capacity of optical interconnects. The digital structures optimized by a novel adaptive direct binary search (ADBS) algorithm works for the TE0, TE1, TE2, and TE3 modes. The ADBS algorithm can adaptively escape local convergences and greatly reduce time costs because of the introductions of the pooling operator and mutation operator. The waveguide structures designed by the cascaded asymmetrical directional couplers (ADCs) works for the TM1, TM2, TM3, and TM4 modes. The ADC structures further expands the number of channels based on the digital structure through a cascade strategy. To the best of our knowledge, the digital-waveguide structure with an effective length of about 100 μm is the smallest 8-channel mode-polarization multiplexer. The device is fabricated on a SOI wafer with a 220 nm-thick top silicon layer and a 1 μm-thick silica cladding. The measured insertion loss (IL) and crosstalk (CT) are less than 1.2 dB and lower than -12.5 dB at 1550 nm, respectively. At the same time, the measured ILs and CTs are less than 1.7 dB and lower than -9.3 dB from 1540 nm to 1560 nm, respectively. In addition, clear and open eye diagrams are obtained during the transmission of 256 Gbps signals, verifying the stable and high-speed data transmission capability.
, , Received Date: 2025-08-25
Abstract +
,
Abstract +
Density functional theory (DFT) stands as the predominant workhorse for electronic structure calculation across physics, chemistry, and materials science. However, its practical application is fundamentally constrained by a computational cost that scales cubically with system size, rendering high-precision studies of complex or large-scale materials prohibitively expensive. This review addresses the pivotal challenge by surveying the rapidly evolving paradigm of integrating machine learning (ML) with first-principles calculations to dramatically accelerate and scale electronic structure prediction. Our primary objective is to provide a comprehensive and critical overview of the methodological advances, physical outcomes, and transformative potential of this interdisciplinary field.
The core methodological progression involves a shift from black-box property predictors to symmetry-preserving, transferable models that learn the fundamental Hamiltonian—the central quantity from which diverse electronic properties derive. We detail this evolution, beginning with pioneering applications in molecular systems using graph neural networks (e.g., SchNOrb, DimeNet) to predict energies, wavefunctions, and Hamiltonian matrices with meV-level accuracy. The review then focuses on the critical extension to periodic solids, where preserving symmetries like E(3)-equivariance and handling vast configurational spaces are paramount. We systematically analyze three leading model families that define the state-of-the-art: the DeepH series, which employs local coordinate message passing and E(3)-equivariant networks to achieve sub-meV accuracy and linear scaling; the HamGNN framework, built on rigorous equivariant tensor decomposition, excelling in modeling systems with spin-orbit coupling and charged defects; and the DeePTB approach, which leverages deep learning for tight-binding Hamiltonian parameterization, enabling quantum-accurate simulations of millions of atoms.
These methods yield significant physical results and computational breakthroughs. Key outcomes include: 1) Unprecedented accuracy and speed. Models consistently achieve Hamiltonian prediction mean absolute errors (MAE) below 1 meV (e.g., DeepH-E3: ~0.4 meV in graphene; HamGNN: ~1.5 meV in QM9 molecules), coupled with computational speedups of 3 to 5 orders of magnitude compared to conventional DFT. 2) Scale bridging. Successful applications now span from small molecules to defect-containing supercells with over 10,000 atoms (e.g., HamGNN-Q on a 13,824-atom GaAs defect) and even to millions of atoms for optoelectronic property simulations (DeePTB). 3) Expanded application scope. The review highlights how these ML-accelerated tools are revolutionizing research in previously intractable areas: predicting spectroscopic properties of molecules (e.g., DetaNet for NMR/UV-Vis spectra), elucidating electronic structures of topological materials and magnetic moiré systems, computing electron-phonon coupling and carrier mobility with DFT-level accuracy but far greater efficiency (HamEPC framework), and enabling high-throughput screening for materials design.
In conclusion, ML-accelerated electronic structure calculation has matured into a powerful paradigm, transitioning from a proof-of-concept to a tool capable of delivering DFT-fidelity results at dramatically reduced cost for systems of realistic scale and complexity. However, challenges remain, including model interpretability ("black-box" nature), transferability to unseen elements, and seamless integration with existing plane-wave DFT databases. Future directions point towards physics-constrained unsupervised learning (e.g., DeepH-zero), development of more universal and element-agnostic architectures, and the creation of closed-loop, artificial intelligence (AI)-driven discovery pipelines. By overcoming current limitations, these methods hold the potential to fundamentally reshape the materials research landscape, accelerating the journey from atomistic simulation to rational material design and discovery.
The core methodological progression involves a shift from black-box property predictors to symmetry-preserving, transferable models that learn the fundamental Hamiltonian—the central quantity from which diverse electronic properties derive. We detail this evolution, beginning with pioneering applications in molecular systems using graph neural networks (e.g., SchNOrb, DimeNet) to predict energies, wavefunctions, and Hamiltonian matrices with meV-level accuracy. The review then focuses on the critical extension to periodic solids, where preserving symmetries like E(3)-equivariance and handling vast configurational spaces are paramount. We systematically analyze three leading model families that define the state-of-the-art: the DeepH series, which employs local coordinate message passing and E(3)-equivariant networks to achieve sub-meV accuracy and linear scaling; the HamGNN framework, built on rigorous equivariant tensor decomposition, excelling in modeling systems with spin-orbit coupling and charged defects; and the DeePTB approach, which leverages deep learning for tight-binding Hamiltonian parameterization, enabling quantum-accurate simulations of millions of atoms.
These methods yield significant physical results and computational breakthroughs. Key outcomes include: 1) Unprecedented accuracy and speed. Models consistently achieve Hamiltonian prediction mean absolute errors (MAE) below 1 meV (e.g., DeepH-E3: ~0.4 meV in graphene; HamGNN: ~1.5 meV in QM9 molecules), coupled with computational speedups of 3 to 5 orders of magnitude compared to conventional DFT. 2) Scale bridging. Successful applications now span from small molecules to defect-containing supercells with over 10,000 atoms (e.g., HamGNN-Q on a 13,824-atom GaAs defect) and even to millions of atoms for optoelectronic property simulations (DeePTB). 3) Expanded application scope. The review highlights how these ML-accelerated tools are revolutionizing research in previously intractable areas: predicting spectroscopic properties of molecules (e.g., DetaNet for NMR/UV-Vis spectra), elucidating electronic structures of topological materials and magnetic moiré systems, computing electron-phonon coupling and carrier mobility with DFT-level accuracy but far greater efficiency (HamEPC framework), and enabling high-throughput screening for materials design.
In conclusion, ML-accelerated electronic structure calculation has matured into a powerful paradigm, transitioning from a proof-of-concept to a tool capable of delivering DFT-fidelity results at dramatically reduced cost for systems of realistic scale and complexity. However, challenges remain, including model interpretability ("black-box" nature), transferability to unseen elements, and seamless integration with existing plane-wave DFT databases. Future directions point towards physics-constrained unsupervised learning (e.g., DeepH-zero), development of more universal and element-agnostic architectures, and the creation of closed-loop, artificial intelligence (AI)-driven discovery pipelines. By overcoming current limitations, these methods hold the potential to fundamentally reshape the materials research landscape, accelerating the journey from atomistic simulation to rational material design and discovery.
- 1
- 2
- 3
- 4
- 5
- ...
- 16
- 17
