Coupled mode theory (CMT) is a universal method for studying resonant systems in various disciplines in science. Combined with traditional fitting methods, implicit physical parameters of the resonant systems can be revealed. However, this methodology fails in tackling the scenario of multi-solution for a given resonant system, resembling a fundamental challenge that has not been addressed yet. In this work, we propose and experimentally demonstrate a CMT physics and data co-driven deep neural network (CMT-NN) that can predict the implicit physical parameters of complex resonant systems in a rapid and precise way. More importantly, the challenge of multi-solution is mitigated by incorporating physical eigenvalues and response of the system in evaluating the physics consistency of the neural network. The applicability and generality of CMT-NN are demonstrated by simulations and experiments, where the CMT-NN can capture subtle spectral features and learn the coupling physical properties effectively. Compared with the traditional fitting method, the average computation time has been reduced by three orders of magnitude and the prediction performance is improved by more than two orders of magnitude. Displacement sensing experiments further validate the robustness of CMT-NN. It is anticipated that the CMT-NN can provide a paradigm shift in using the CMT for studying resonant systems and shed new light on the understanding, design and optimization of various coupled resonant systems.
Distributed acoustic sensing (DAS) has attracted considerable attention across various fields, and artificial intelligence (AI) technology plays a vital role in DAS applications for event recognition and denoising. Existing AI models require real-world data (RWD), whether labeled or not, for training, which is contradictory to the reality of limited available event data in practical scenarios. Here, a physics-informed DAS neural network paradigm is proposed, which eliminates the need for real-world event data during training. By physically modeling the target events along with real-world and DAS system constraints, physical functions are derived to train a generative network for the synthesis of DAS event data. A DAS noise-removal network is then trained using the generated data to effectively eliminate background noise in DAS measurements. The effectiveness of the proposed paradigm is demonstrated in two applications: event identification based on a public DAS spatiotemporal dataset, and belt conveyor fault monitoring based on DAS time-frequency data. In both cases, the paradigm achieves comparable or superior performance to data-driven networks trained with RWD. Owing to the incorporation of physical information and the ability to remove background noise, the proposed approach shows strong generalization capability across different sites within the same application. Notably, a fault diagnosis accuracy of 91.8% is achieved in a real belt conveyor field using networks transferred from a simulation test site, without using any fault event data from the target field during training. The proposed paradigm offers a potential solution to the critical challenges of limited data availability and intense noise in practical DAS applications.
The short communications.
Metasurfaces represent a promising platform for improving coverage in future communication systems. Passive designs are especially attractive because they need no power supply and can be manufactured at low cost. However, most passive metasurfaces work well only for one polarization, frequency band, or incidence angle, which limits their practical use. Here we propose passive intelligent panels, termed metacrystals, that overcome these limitations by enabling highly complex multiplexed responses to multiple incident waves simultaneously and independently. This capability is enabled by a compact volumetric architecture that goes beyond conventional metasurfaces by exploiting a finite, yet still modest, thickness to unlock substantially more degrees of freedom. Through simulations and experiments, we demonstrate all-dielectric metacrystals capable of simultaneously controlling anomalous reflection and absorption, both in transmission and reflection regimes. Designed using inverse topology optimization, these metacrystals combine structural integrity, straightforward scalability, and compatibility with low-cost 3D printing for operation up to 100 GHz.
Optical fibres, owing to their ultra-low transmission loss, underpin global telecommunications. However, this remarkable low-loss performance has not been extended to integrated photonic devices, which are increasingly critical for data-intensive communications in the era of artificial intelligence (AI). Here, we translate the widely adopted mass-production process for fibre manufacturing-flame hydrolysis-to wafer-scale integrated photonics, and demonstrate ultrahigh-Q integrated microresonators. By leveraging high GeO2 doping, the deposited germano-silicate (Ge:silica) films achieve full densification at moderate thermal budgets, while also allowing for a post-processing furnace-reflow technique that has the capability to both repair any etch-induced defects and enhance optical Q, leading to a high degree of process tolerance. When combined with deep-UV lithography, these films form microresonators exhibiting ultrahigh Q factors of up to 566 million at 1064 nm, corresponding to a waveguide propagation loss as low as 0.07 dB m-1. Like their fibre counterparts, these devices exhibit a broad transmission window with Q factors surpassing 100 million demonstrated from the telecommunications band to the violet spectrum. Moreover, the width dependence of Q factor and a two-order-of-magnitude Q recovery enabled by the furnace reflow process are also demonstrated. This work extends high-quality, high-rate flame hydrolysis deposition (FHD) from optical fibre manufacturing to integrated photonics, establishing a scalable route towards fibre-level loss in photonic integrated circuits.
Topological physics in classical systems, such as photonic and acoustic systems, is fast becoming an exciting field in fundamental and applied research. However, almost all existing topological acoustic materials are restricted to systems with conventional lattices and specific time-space symmetries. Here, we present a unique design of topological phononic crystals with moiré superlattice that are constructed by simply sliding a layer of moiré phononic crystals. Assisted by the sliding degree of freedom, sliding moiré phononic crystals exhibit nontrivial topology characterized by a dynamical Chern number, leading to topological pumping of special topological moiré edge states that are completely insensitive to the edge geometry. More interestingly, moiré phononic crystals possess one-dimensional moiré flat bulk bands, giving rise to one-dimensional localized states in the bulk that result in an array of compact and nearly independent one-dimensional signal channels. Various applications such as acoustic communications and isolations in integrated devices can be anticipated from the intriguing acoustic one-dimensional localized bulk states and topological moiré edge states enriched by the sliding modulation.
Simulation is a valuable tool for traffic planning that requires reliable modeling of traffic dynamics. Free flow speeds in bicycle traffic depend on the characteristics and preferences of the bicyclists, infrastructure design, and environmental conditions. However, existing models are limited in capturing changes in the speed during free riding, thereby reducing their applicability in bicycle traffic analysis. This study advances microscopic bicycle traffic simulation by developing and evaluating simulation models for free riding dynamics, aiming to capture the heterogeneous and context-dependent effects of infrastructure design (slopes, curves, and presence of intersections) and wind on bicyclist behavior. We implement three models within SUMO -(1) context-based speed distributions, (2) a speed regression model, and (3) a physics-based speed model derived from power output- and benchmark them against built-in models. All models are evaluated using empirical trajectory data from 57 bicycle commuters in semi-controlled experiments in Sweden and Germany. Results demonstrate that the proposed models outperform the existing baselines in replicating speed patterns. In this regard, the physics-based model provides the closest alignment to observed speeds. Omitting context dependency in free riding can result in substantially larger errors in speeds on uphills and downhills, and in deceleration on curves or when crossing intersections. Context-sensitive models enhance the accuracy of bicycle traffic simulation, thereby increasing their usefulness in planning and evaluating bicycling facilities that accommodate the diverse preferences of bicyclists.
Neuromorphic computing based on artificial synapses requires devices capable of gradual, repeatable, and energy-efficient conductance modulation. Ionically gated transistors are promising candidates because their ion dynamics naturally produce synaptic behavior under low-voltage operation. However, how key device characteristics-conductance range, number of accessible conductance states N, and weight-update nonlinearity β-jointly influence neural network performance remains insufficiently understood. Here, we investigate MoS2-based ionically gated synaptic transistors using a combined experimental and modeling framework that links device physics to hardware-aware artificial neural network (ANN) simulations across image-classification tasks of varying complexity. We show that under fixed-amplitude pulsing, increasing N introduces a fundamental trade-off: finer weight resolution is accompanied by stronger update nonlinearity. ANN simulations further reveal that, within the nonlinearity range studied here, classification accuracy is governed by a task-dependent optimal weight resolution rather than a simply maximized number of states. To overcome the nonlinear weight updates, we employ a physics-informed transient model to develop a predictive pulse-engineering algorithm and experimentally demonstrate that it can linearize synaptic weight evolution in the same device. These linearized updates improve ANN accuracy by 1.5%-5.2% for MNIST, 4.0%-5.2% for FMNIST, and 1.2%-12% for KMNIST across the tested state numbers, establishing a quantitative link between device-level dynamics and neural network performance in ionically gated synaptic transistors.
Spatiotemporal optical vortices (STOVs) carrying transverse orbital angular momentum (OAM) fundamentally expand light-structuring capabilities. However, current rigid-body generation paradigms constrain transverse OAM to a single scalar property, leaving rich internal wavepacket dynamics inaccessible. This rigidity contrasts with ubiquitous natural vortices where symmetry breaking is the norm. Here, we break rotational symmetry via the nonlinear mapping of the azimuthal phase gradient, realizing programmable spatiotemporal flux breathing. We theoretically and experimentally demonstrate that local phase gradient variations induce instantaneous group velocity anisotropy. This compels local OAM density to spontaneously reorganize into stable, multilobed lattice structures while strictly preserving global topological charge. Furthermore, we harness these structures' modulation frequency for free-space information transfer, achieving high-fidelity encoding and decoding of spatiotemporal topological states. This work transitions STOVs from passive scalar objects to structured functional carriers, opening avenues for high-dimensional optical communications, ultrafast spatiotemporal manipulation, strong-field physics, and high-dimensional quantum entanglement.
Surface acoustic wave (SAW) devices have become key building blocks in communications, sensing, and emerging neuromorphic hardware, yet their growing complexity in materials, architectures, and operating environments exposes limitations in conventional design and signal processing approaches. Concurrently, machine learning (ML) provides powerful tools for Modeling high-dimensional systems and extracting information from noisy multiparameter data. This review surveys the rapidly developing interface between SAW technology and ML, spanning device-level Modeling, multiparameter sensing, and SAW-driven neuromorphic hardware. We first discuss how ML is used as a surrogate for traditional simulation and compact Modeling. Examples include neural network-based inverse design of SAW resonators from target performance indicators, and regression models used to extract coupling of modes (COM) parameters from simulated or measured responses, which improves the speed and reliability of resonator design. We then examine ML-enhanced SAW sensing, with emphasis on thin-film and flexible Aluminium Scandium Nitride (AlScN) -based devices operating under complex environmental and mechanical conditions. Case studies include flexible UV sensors that use ML regression to decouple bending-induced strain from the target signal, and stacking ensemble frameworks that exploit scattering parameter features to suppress cross-interference between temperature, humidity, and UV intensity. Finally, we highlight how SAWs are being explored as low-power actuators in neuromorphic and AI hardware, enabling acoustic control of phase transitions in Iron-Rhodium alloy (FeRh) and the creation of magnetic skyrmions for artificial synapses and neurons. Across these domains, we identify common patterns in data representation, model selection, and evaluation, and we outline open challenges in dataset curation, interpretability, and deployment on edge and flexible platforms. By bridging the RF design, sensing, and neuromorphic communities, this review charts a path toward ML-native SAW systems that draw on both the rich physics of acoustic waves and the adaptability of modern data-driven methods….
Pediatric chest computed tomography (CT) requires optimization of image quality while minimizing the radiation dose, especially for low-contrast anatomical structures. This study aimed to develop and utilize novel, Various Age-size Pediatric Chest Phantoms (VAPC) models for 1-, 4-, and 7-year-old pediatric patients to systemically evaluate the effect of varying adaptive statistical iterative reconstruction-Veo (ASiR-V) levels under dose conditions consistent with American Association of Physicists in Medicine Report 246 recommendations for pediatric chest CT (as detailed in the Methods section). Phantom models were generated from datasets in the Digital Imaging and Communications in Medicine format, obtained from The Cancer Imaging Archive, and scanned using five protocols with Adaptive Statistical Iterative Reconstruction-Veo (ASiR-V) levels of 20%, 40%, 60%, and 80%. The absorbed doses in the lungs, soft tissue, spine, and heart were measured using Gafchromic LD-V1 films. The image quality was evaluated using the signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), and noise power spectrum. Increasing the ASiR-V percentage consistently improved image quality, with 80% ASiR-V producing significant SNR gains (e.g. 5.63-7.93, in 1 year-old), increased CNR, and 42%-61% noise reduction compared with filtered back projection. The 1 year-old phantom showed an 80.7% reduction in lung dose (2.99-0.58 mGy), and the 4 year-old phantom showed a 42.01% reduction in lung dose (3.07-1.78 mGy) at the same parameters (80% ASiR-V). In contrast, the 7 year-old phantom showed the largest gain in image fidelity. Overall, ASiR-V showed optimal performance in younger phantoms (1 year) while maintaining anatomical details and achieving improved dose efficiency in the 7 year-old phantom, supporting its effectiveness for low-dose pediatric chest-CT imaging.
Ultra-wide bandgap gallium oxides offer tremendous possibilities to develop short-wave optoelectronic devices. However, it is formidably challenging to produce single-crystal gallium oxide wafer and develop high-performance high-dimensional optoelectronics. Here we show a liquid-metal-assisted strategy to directly synthesize and transfer single-crystal, large-area and ultrathin β-Ga2O3. Benefiting from the UV-exposure oxidation of liquid gallium and strong interaction with gallium, our β-Ga2O3 film shows a 4-inch wafer-scale size, a 7.5-nm thickness and a flexible transfer operation. The solar-blind β-Ga2O3 detector achieves high responsivity (16.3 A W-1), fast response (<150 μs) and wide linear dynamic range (120 dB). By employing metasurface design, the anisotropy ratio reaches a record high value of 28.8 for Ga2O3-based detectors. Moreover, we develop a sundial-inspired metasystem to simultaneously detect the incident direction, polarization, and intensity of solar-blind irradiation. These findings illustrate the potential of high-quality Ga2O3 wafer for high-dimensional photodetection, paving the way for next-generation solar-blind communications.
As the two-dimensional counterpart of metamaterials, metasurfaces have demonstrated exceptional potential in optics due to their remarkable ability to flexibly manipulate electromagnetic waves, offering a platform for multidimensional wavefront manipulation. However, existing multiplexing techniques predominantly focus on a single physical dimension, and multi-parameter joint multiplexing still faces challenges such as limited design degrees of freedom and channel crosstalk. This paper proposes a wavelength-polarization multiplexed terahertz metadevice based on anisotropic meta-atoms, which enables independent wavefront manipulation of three-band responses under orthogonal linear polarizations within a single-layer structure, thereby creating six fully decoupled transmission channels. We designed and numerically validated a dual-polarization triple-frequency decoupled metalens, a dual-polarization achromatic vortex beam generator, and a six-channel holographic display device. This research provides a compact and efficient solution for multidimensional optical field manipulation, with promising applications in terahertz communications, high-capacity information encryption, and holographic displays.
Room-temperature midwave infrared (MWIR) detection is crucial for applications such as gas sensing, thermal imaging, and optical communications. However, conventional narrow-band gap semiconductors (such as HgCdTe, InAsSb, and PbSe) suffer from high dark current, leading to significant noise at room temperature. To address this challenge, we propose a hybrid-dimensional PbSe/WTe2 van der Waals heterostructure. The formed Schottky barrier effectively modulates carrier transport through the built-in electric field, resulting in significantly suppressed dark current density (as low as 6.9 × 10-5 A·cm-2) and noise current (8.3 × 10-13 A Hz-1/2). The dark current is reduced by approximately 1 order of magnitude compared to that of pure PbSe devices. Further, the device demonstrates a broad spectral response range from 808 to 4000 nm, with a peak responsivity of 0.75 A/W and specific detectivity of 4.81 × 109 cm Hz1/2 W-1 at 2600 nm, which are comparable with commercial MWIR detectors. Finally, we demonstrate a room-temperature clear blackbody and thermal target imaging by PbSe/WTe2 heterojunction, providing an effective pathway for low-noise room-temperature MWIR detectors.
The topological properties of optical spin skyrmions provide an additional degree of freedom for data encoding in photonic networks. Although optical spin skyrmions can be realized at subwavelength scales by surface plasmonics, they fail to radiate into free space as information carriers due to spatial confinement to metal-dielectric interfaces. To date, free-space radiative spin skyrmions have relied on cascaded, bulky optical elements. Here, we report a direct approach to generating free-space optical spin skyrmions using a single surface plasmonic device named a plasmonic geometric phase aperture. Distinct from surface-bound modes, the skyrmion textures of the radiated fields are engineered by spin-orbit interaction in metallic nanoslits via Pancharatnam-Berry geometric phases. It transforms a portion of the circularly polarized waves into a vortex beam that, nested with the residual light, forms spin skyrmion fields, which can be experimentally visualized through spin-selective, phase-resolved scanning near-field optical microscopy. Importantly, the spin skyrmions are generated in the intermediate field region several micrometers above the device surface, bridging the crucial spatial gap between the optical near-field and far-field demanded for on-chip interconnection. The findings provide an ideal solution for high-capacity and robust chip-to-chip optical communications using optical skyrmions.
Supercoupling in near-zero-index (NZI) media enables geometry-insensitive electromagnetic (EM) transport through narrow channels with near-zero phase delay. However, most studies have focused on single-channel, point-to-point configurations, leaving EM power-flow distribution in complex structures largely unexplored. Here we extend NZI supercoupling to complex structures and show that EM power flow follows a passive, deterministic, and quasi-static distribution governed by boundary conditions and impedance contrasts, with PEC-terminated branches carrying no propagating power flow. We interpret this behavior using a pressure-driven flow analogy and directly visualize it in a waveguide-emulated plasmonic platform with photonic doping. This quasi-static power-flow distribution follows an "Ohm's law of ideal EM power flow", where the potential is set by boundary conditions and the effective impedance by each branch's length-to-width ratio. Beyond the physical interpretation, our results suggest an impedance-designed approach to passive multi-port EM interconnects, offering insights for NZI physics and on-chip networks at millimeter-wave and terahertz frequencies.
The electromagnetic response of materials serves as the foundation for a broad range of vital applications, from imaging, to sensing, to classical and quantum communications. Here we demonstrate, theoretically and experimentally, a fundamentally new regime of electromagnetic material response originating from inherent material nonlocality. We show that by structuring materials on the intrinsic scale of this nonlocal response, it becomes possible to alter the electromagnetics of the composite, revealing the inherent nonlocal behavior of the constituent components. These intrinsically nonlocal metamaterials exhibit strong intrinsic (as opposed to effective) nonlocality, easily detectable at room temperatures, in realistic (lossy), macroscopic materials. Intrinsically nonlocal metamaterials open a new design space for electromagnetic composites, beyond photonic crystals, metasurfaces, and effective medium composites. This allows the control of electromagnetic fields at a deep subwavelength scale, revealing a new dimension for control of light-matter interactions.
Integrated radio-frequency (RF) photonics plays a pivotal role in wireless communications, sensing, and radar applications. However, the bulky feature of essential RF components still constrains practical deployments in covert, conformal, and space-limited applications. As a promising solution through synergistic integration of both RF and photonics, we demonstrate a photonic RF receiver chip integrating a bow-tie antenna and a microring modulator on thin-film lithium niobate platform. The chip leverages a dual-resonance enhancement mechanism, combining RF and optical resonances, to significantly boost the RF-to-optical conversion efficiency. A record-high figure of merit of 3.88 W-1/2 was achieved within a compact footprint of 2 × 1.7 mm2. Following full packaging, the receiver enabled centimeter-level radar ranging accuracy, 3.2 Gbps wireless communication capacity, and real-time high-definition video transmission in moving scenarios. This work paves a viable way toward covert, conformal, and miniature photonic RF frontends for unmanned aerial vehicles, high-speed trains, and electronic warfare systems.
All-optical information processing, featuring ultrafast response and immunity to electromagnetic interference, plays a pivotal role in future communications and computing technologies. Graphene, with its exceptional nonlinear optical properties and mechanical flexibility, emerges as an ideal candidate for flexible photonic devices. However, graphene-based photonic components are typically functionally fixed and lack dynamic reconfigurability. Here, we integrate graphene with polydimethylsiloxane (PDMS) to fabricate a flexible graphene/PDMS composite and investigate its spatial self-phase modulation (SSPM) effect under mechanical strain. Our results demonstrate that the nonlinear optical response of the composite can be dynamically tuned by strain. Under 532 nm laser excitation (intensity 35 W/cm2), increasing the tensile strain from 0% to 40% continuously suppresses the number of SSPM diffraction rings from 8 to 0, while accompanied by a reduction in the third-order nonlinear susceptibility χ monolayer ( 3 ) $\chi _{{\mathrm{monolayer}}}^{(3)}$ from 1.357 × 10-7 to 6.125 × 10-8 e.s.u. This tunable SSPM effect originates from strain-induced modifications in both the effective number of optically interacting layers and the electronic band structure of graphene. Leveraging this mechanism, we further designed a strain-gated optical switch and reconfigurable optical logic gates, enabling flexible switching between "OR" and "AND" gates. This work opens new avenues for graphene-based tunable nonlinear photonic devices.
Ring laser gyroscopes (RLGs) measure rotation via the Sagnac effect: a slight difference in the frequency of the two counter-propagating beams within the resonator. However, at low rotation rates, an intrinsic limitation in RLGs, known as the lock-in phenomenon, counteracts this effect, precluding the widespread adoption of RLGs as motion sensors. Past efforts to avoid this phenomenon include mechanical dithering1 and magneto-optic non-reciprocity techniques2. Such techniques require external components that limit the miniaturization of RLGs. Here we present a self-biased method that overcomes this limitation through chiral spontaneous symmetry breaking and nonlinear frequency pulling in a He-20Ne RLG without inserted elements. Supported by a theoretical model that reveals phase transition conditions with spontaneous symmetry breaking and the dynamics of bistable chiral states, our experiments demonstrate deterministic chirality switching synchronized with rotation direction. Remarkably, the chiral RLG has a linear frequency response at near-zero rotation rates, achieving an open-loop bias instability of 2.2 × 10-2 degrees per hour at a 10 s integration time. Our work presents a strategy for the development of all-solid-state, high-precision and miniaturized laser gyroscopes, which could be used for the exploration of the interplay of nonlinear dynamics and spontaneous symmetry breaking in photonic systems.