Controlling the velocity of the acoustic wave is critical for advancing time-domain signal processing. However, current slow acoustic configurations suffer from narrowband and backscattering loss. Here, we propose broadband topologically protected slow acoustic reciprocal systems via multiwinding modulation in the Brillouin zone (BZ). By coupling the resonance-induced nearly flat bands and the topological edge states (TESs), hybridized TESs with multiple windings across the BZ are formed, which result in broadband slow acoustic waves with controllable velocity. The relative bandwidth of similar to 7% and the velocity of 0.04c0 (c0 is the acoustic wave velocity in air) are achieved, which greatly outperforms conventional methods. Our work paves the way for the development of high-performance acoustic devices.

The demand for miniaturized on-chip spectrometers is growing rapidly in the field of portable optical sensing. However, fabrication-induced defects commonly encountered in current on-chip spectrometers alter the response of the devices, thereby degrading the accuracy of spectral reconstruction. To circumvent the need for complex defect-correction calibrations, we demonstrate a calibration-free on-chip spectrometer using topological photonic waveguides. The spectrometer's inherent topological protection feature withstands manufacturing and defect-induced fluctuations, enabling high accuracy spectral reconstruction despite the presence of manufacturing and defect errors. When defect errors are present, the relative error (epsilon) of spectral reconstruction does not exceed 0.029 across a 176 nm bandwidth. Additionally, with the introduction of random errors up to 5 and 10 nm, the epsilon remains below 0.04 and 0.059, respectively. This design leverages the topological photonic structure's tolerance to defects, ensuring stable performance for robust portable spectrometers.

Fractional vortex beams (FVBs), endowed with unique and complex optical field distributions, exhibit superior potential compared to integer vortex beams in diverse fields such as optical communications. Consequently, the accurate and high-quality sorting of FVBs is of paramount importance. However, traditional methods struggle to meet the stringent requirements for efficiency, crosstalk, and resolution in FVB sorting. Meanwhile, although diffractive optical neural networks (DONNs) have been successfully applied to integer-order orbital angular momentum (OAM) sorting, their application in the non-integer domain remains unexplored. In this paper, we systematically demonstrate the complete workflow and potential efficacy of DONNs in addressing the specific challenges of FVB sorting for the first time. In simulations, we achieve inter-channel crosstalk below-20 dB. Meanwhile, we elaborate on the precise OAM spectrum measurement capability of DONNs and successfully employ DONNs to experimentally measure the spectral distributions of different OAM states for the first time, achieving a measurement fidelity of approximately 99%. This provides a novel all-optical measurement approach for the quantitative evaluation of OAM spectra. Furthermore, we integrate FVB sorting with wavelength-division multiplexing (WDM) technology, enabling our sorting device to maintain stable performance at two wavelengths (532 nm and 660 nm) with an experimental output crosstalk below-10 dB. We believe that this rapid sorting method, which combines the high-speed operation, high efficiency, and high parallelism of optical neural networks, holds tremendous potential for future high-dimensional optical communications, optical metrology, and so on.

A continuously tunable reflective metasurface is proposed for the dynamic generation of vortex beams in the microwave band. Each meta-atom integrates a voltage-controlled varactor diode, enabling near-continuous reflection phase modulation through external bias control. To address phase instability caused by position-dependent biasing vias in large-scale metasurface arrays, a symmetric blind-via design strategy is employed to effectively suppress parasitic-induced phase perturbations. The designed metasurface achieves approximately 340° continuous reflection phase coverage over the frequency range from 5.75 to 7.25 GHz. By programming spatial phase distributions, multiple complex beam types, including conventional vortex beams, focused vortex beams, and non-diffracting vortex beams, can be dynamically generated using a single metasurface configuration. Full-wave simulations validate stable phase control and consistent beamforming performance across the operating band. The proposed metasurface provides a practical and scalable solution for multifunctional wavefront manipulation, with potential applications in microwave antennas, wireless communication, and radar systems.

Diffractive optical neural networks (DONNs) provide a promising route towards high-speed and energy-efficient computing. However, the absence of a general framework for nonlinear activation remains a fundamental limitation. Here we propose a high-order optical neural network (HONN) framework under partially coherent illumination, which enables effective nonlinearity through a unified system-level description. In this framework, the nonlinear response arises from the interplay of coherence-dependent detection and input-dependent pseudo-nonlinearity to a controllable order. We experimentally demonstrate this model in visible spectrum on Digit MNIST and Fashion MNIST, and results show that HONN can successfully capture the performance evolution across varying coherence conditions and nonlinear orders. Further validations on RAF and MedMNIST verify that our HONN surpasses conventional DONNs with competing performance to traditional electrical neural networks with <10% of their computational complexity only. Furthermore, our model possesses superior coherent robustness, laying a solid foundation for practical real-world deployment. These results establish a general framework for high-order optical inference beyond traditional diffractive models and provide a practical route towards optical computing in machine vision, computational imaging and sensing.

Acoustic pulling provides an additional degree of freedom for precise acoustic manipulation in acoustofluidics, with applications in separation, assembly, and cell characterization. However, most current approaches rely on elaborately designed scattering to redistribute momentum and generate pulling forces, which inevitably produces strong backward-propagating waves and consequently limits the realization of sustained long-distance pulling. Herein, we propose a novel strategy for long-distance acoustic pulling through a self-induced intensity gradient field generated by the manipulated object itself. Using a phononic crystal with a precisely engineered bandgap structure, we establish a unique transmission mode that propagates as a guided mode in the absence of the object but becomes prohibited upon object insertion, resulting in the formation of a bandgap. This inhibition leads to persistent negative intensity gradient fields in the object, yielding continuous acoustic pulling forces. By leveraging phononic crystals with precisely engineered bandgap structures, we establish a distinct transmission regime that gives rise to a continuous negative intensity gradient within the object, thereby generating a sustained acoustic pulling force. Furthermore, this approach accommodates a broad range of object sizes and offers a versatile platform for acoustic manipulation in biomedicine and related fields.

Massive multiple-input multiple-output (MIMO) technologies have been greatly developed for the sixth-generation (6G) communication systems. Based on full-digital antenna arrays, asymmetric massive MIMO communication systems can provide distinct beamforming patterns for uplink and downlink chains. In this paper, a terahertz (THz) channel model for asymmetric massive MIMO communication systems considering the influence of different beam patterns is proposed. The proposed model incorporates channel characteristics of massive MIMO based on spherical-wavefront modeling and distance-dependent steering vectors. The asymmetry between uplink and downlink is considered through antenna beam patterns and array configurations. The statistical properties of the channel model are derived. These channel characteristics and system performance, such as channel capacity, are simulated and compared under different beam patterns. The results show that the space-time-frequency (STF) correlations increase while the delay spread and angular spread decrease when the beam patterns are more concentrated. It is also found that the channel capacity can be increased by utilizing high-gain and wide beam patterns.

Conventional metasurfaces are frequently constrained by single-functionality and limited integration capabilities, hindering their application in emerging 5 G/6 G communications and intelligent sensing scenarios that demand multifunctionality, lightweight design, and high compatibility. To address these challenges, we propose a transmissive-reflective dual-functional metasurface based on a hexagonal unit cell architecture. Within the operating frequency band, this metasurface achieves full phase coverage in the reflection domain for x-polarized waves and in the transmission domain for y-polarized waves, respectively. The dual-functional capability is further verified by generating mode l = 1 vortex beams in the reflection space and mode l = 2 vortex beams in the transmission space. Furthermore, we introduce a feedforward neural network (FNN) framework for the inverse design of the metasurface, enabling the direct mapping of target electromagnetic responses to matched structural parameters. The deep neural network is trained and validated using full-wave simulation datasets. Validation results demonstrate excellent agreement between the network-predicted electromagnetic responses and the target responses, with the average phase error maintained within 5 degrees.

Channel models with a good balance of pervasiveness, accuracy, and efficiency are important for the design and optimization of the sixth generation (6G) wireless communication systems. In this paper, a pervasive beam domain channel model (BDCM) capable of modeling all frequency bands and scenarios in 6G is proposed. Unlike traditional geometry-based stochastic models (GBSMs) that describe channels between antenna pairs in the space domain, the pervasive BDCM reformulates the channel in terms of beam pairs to describe special channel characteristics in the beam domain, such as sparsity and Doppler insensibility. The proposed BDCM incorporates essential spatial wideband and spherical wavefront effects for ultra-massive multiple-input multiple-output (MIMO) by considering the nonlinear phase variations across antenna arrays. The pervasive transform matrices for different antenna configurations are derived to enable flexible conversions between the pervasive GBSM and pervasive BDCM. In addition, key statistical properties of the BDCM are derived and analyzed. The proposed pervasive BDCM in different frequency bands and scenarios are validated by measurement data and compared with the GBSM results. The complexity analysis reveals that the proposed pervasive BDCM significantly reduces the computational complexity compared with the pervasive GBSM under different scatterer densities.

Vortex beams carry orbital angular momentum (OAM) and exhibit a ring-shaped intensity distribution, adding a new dimension compared to Gaussian beams. In cloudy and foggy environments, using vortex beams for detection and imaging can partially improve the signal-to-noise ratio ( SNR ) affected by backscattering compared to Gaussian beams. However, the improvement is limited at high concentrations. We introduce a novel approach to improve the SNR of vortex beam detection under these conditions. First, we utilized a ring filter for preliminary noise reduction, then applied polarization information to divide the data into different polarization directions. We then performed weighted summation on the one-dimensional photon counting echo data from these directions to further reduce noise. Simulation results demonstrated that this method improved SNR across various parameters. Specifically, at low reflectivity, the peak signal-to-noise ratio ( PSNR ) increased from 0.333 to 2.14, improving ranging accuracy. In imaging, the SNR of the processed range profile rose from 9.86 dB to 15.6 dB, and the structural similarity index ( SSIM ) improved from 0.590 to 0.894, indicating enhanced image quality. Therefore, our method effectively enhances both ranging accuracy and imaging quality of vortex beams under cloud and fog conditions, with potential applications in fields such as remote sensing.

Significance Artificial neural network (ANN) is a mathematical model that emulates the structure and function of the biological nervous system in data processing. As a fundamental architecture of artificial intelligence (AI), artificial neural networks are extensively utilized across various domains including image reconstruction, face recognition, speech recognition, and text generation. To address increasingly complex problems, the number of parameters in AI models has grown exponentially, necessitating greater computational resources. As a result, energy consumption for model training and device operation has increased substantially. With Moore's Law approaching its physical limits, there is a critical need to explore alternative computing paradigms. The emergence of optical (or photonic) computing, which uses optical fields as information carriers and optical devices for computations, represents an innovative and promising approach with the potential to transform multiple aspects of computing and information processing. Optical neural networks (ONNs), classified as analog optical computing, offer a promising solution to overcome computational limitations of traditional electronic hardware. By harnessing the inherent parallelism, high speed, and low latency of light, ONNs demonstrate potential for accelerating AI tasks, enabling ultra-fast image processing, low-power computing, and real-time data handling. Their potential integration with quantum and neuromorphic systems may establish a new frontier in computational science. Progress This paper presents a comprehensive review of progress, applications and future challenges associated with optical neural networks. Based on physical realization methods, ONNs can be categorized into three typical frameworks: diffractive optical neural networks (DONNs), on-chip waveguides optical neural networks (OCONNs) and optoelectronic neural networks. Initially, this review examines multiplexing methods and linear optical matrix-vector multiplication (MVM) for ONNs. In optical systems, encoding input information through distinct orthogonal optical states represents an effective approach to enhance data processing efficiency. Large models typically require substantial data throughput at the input stage. Optical systems inherently possess multiple degrees of freedom (DOFs)-including wavelength, spatial mode, and polarization state-enabling parallel processing of high-dimensional datasets. Optical multiplexing encompasses wavelength division multiplexing (WDM), space division multiplexing (SDM), time multiplexing and other methods (Fig. 2). The linear transformations in neural networks are fundamentally reducible to MVM. Optical MVM has reached maturity, capable of achieving both linear weighting and linear convolution. Optical MVM (Fig. 3) can be implemented through diffractive units, such as spatial lights modulators (SLMs) and metasurface, or through waveguides, such as Mach-Zehnder interferometers (MZIs) and micro-ring resonators (MRRs). The review then addresses methods for optical nonlinearity (Fig. 4). While nonlinear effects are widely utilized, activating them requires high light intensity, creating a fundamental conflict with ultra-low-power objectives in photonic computing implementations. Research priorities include developing lower threshold nonlinear effects, where quantum interference mechanisms can amplify nonlinear optical responses under low optical power regimes. Additionally, investigating nonlinear encoding paradigms for linear systems presents a crucial developmental pathway for energy-efficient optical computing systems. Nonlinearity can be introduced into a linear physical system through specific encoding strategies for input or transfer matrix, offering a novel approach for nonlinearity realization. The paper analyzes two typical training methods: in silica training and in situ training. In silica training involves deploying trained parameters directly to optical devices (Fig. 5). In situ training integrates software and hardware rather than maintaining their separation. In-situ training is typically implemented through optical backpropagation (Fig. 6). Additionally, the fully optical forward method and optical spiking process are also employed as alternative training approaches (Fig. 7). The paper examines ONN applications in image processing (Fig. 8), computing acceleration (Fig. 9), telecommunication and quantum simulation (Fig. 10). Image processing represents the most natural application for ONNs. Diffractive neural networks offer a novel paradigm for high-speed image processing. Within optoelectronic neural network architectures, the integration of all-optical layers, mostly based on-chip waveguides, plays a critical role in computing acceleration. The inherent parallelism of optical analog computing enhances system efficiency, bypassing the bottlenecks of conventional digital electronic systems. Conclusions and Prospects ONNs demonstrate significant advancement in both theoretical foundations and practical implementation. While optical neural networks have achieved superior performance in certain applications compared to traditional electronic devices, significant challenges persist-particularly in function diversity, mechanisms, efficient nonlinear activation, and device tunability. Researchers continue to address these challenges, proposing potential solutions. Future research directions will emphasize developing optical neural networks with enhanced computational efficiency, reduced power consumption, improved integration scalability, and superior reconfigurability and generalization capabilities to demonstrate ONN's potential across broader applications. The concurrent development of dedicated ONNs, general-purpose ONNs, hybrid optical-electronic neural networks, and all-optical neural networks remains essential. Furthermore, in-situ training will be crucial in advancing scalable training processes for ONNs. Through the integration of optics, material science, computer science and related disciplines, combined with comprehensive utilization of AI tools, we anticipate the emergence of an era characterized by high-performance, general-purpose optical computing.

Defect cavities have been extensively studied for their ability to efficiently manipulate light transmission. However, integrating defect cavities into conventional photonic crystal waveguides typically incurs significant reflection loss, resulting in high energy dissipation in on-chip devices. To address this challenge, we incorporate topological concepts into the defect cavity design and propose a topological defect cavity (TDC) based on valley photonic crystals. Benefiting from the defect-immunity of topological waveguides, the TDC-topological waveguide directly coupled system enables low reflection loss transmission under non-resonant conditions, while allowing flexible tuning of the cavity's resonant frequency. These characteristics make this TDC an excellent platform for on-chip functional device design. As a demonstration, we design a multifunctional logic device combining this TDC and a power splitter, capable of delivering diverse output states as needed. This work holds potential for applications in integrated micro-nano photonics.

The optical force exerted on a dipole particle can be divided into gradient force, scattering force, and spin–curl force, all of which can be derived from Maxwell’s stress tensor with the dipole approximation. Here, we identify an additional spin–curl force for arbitrary objects beyond the dipole approximation, which is named the generalized spin–curl force in this paper. The generalized spin–curl force originates from the Minkowski force density and depends on the imaginary parts of the permittivity, permeability, and chirality of the object. However, it remains imperceptible in conventional optical force calculations due to its exact cancellation by a compensatory surface force during MST surface integration. The study of the generalized spin–curl force provides critical insights into elucidating the mechanisms underlying optical momentum transfer and internal force distribution within complex media. Furthermore, the generalized spin–curl force offers a novel mechanism for enhancing optical sensors, enabling highly sensitive detection of absorptive or chiral perturbations in systems such as microcavities and metasurfaces. Its ability to manipulate internal force distributions also provides new pathways for advancing optical force probes and chirality-selective sensing at the nanoscale.

Computational spectrometer based on spectral encoding directly captures the "fingerprint" of an object. However, the small effective detection area of conventional encoders leads to reduced sensitivity, limiting their practical applications. In this work, we propose a high-sensitivity near-infrared miniaturized spectrometer based on bilayer-metasurface, with a size of only 21 x 21 mu m, and has potential to be integrated with CMOS chips to achieve spectral imaging. The sensitivity is about 60 % higher than that of conventional spectrometer, and the spectral reconstruction fidelity reaches 93.5 %. Our work provides an effective strategy for application in next generation of machine vision tasks for artificial intelligence.

The photonic spin Hall effect (PSHE) manifests as a spin-dependent lateral shift at an interface due to a spin-orbit interaction. When circularly polarized light is incident on a particle at the surface, it typically generates an optical lateral force (OLF) of the order of similar to 0.05 pN/(mWmu m-2) governed by PSHE [Nature Photonics 9, 809 (2015)]. Intuitively, the net OLF vanishes when two beams of equal intensity with opposite circular polarizations (e.g., left- and right-handed) are incident simultaneously. In this work, we exploited the phase-engineered PSHE by superposing two chiral beams with opposite circular polarizations and a controlled phase difference. Both theoretical analysis and experimental results demonstrate that this approach, combining engineered phase difference with circular polarization control, significantly enhances the OLF up to the order of similar to 1.0 pN/(mWmu m-2). This large optical lateral force (LOLF) enables new applications in PSHE-based systems and optical micromanipulation.

Optical manipulation technology has rapidly developed into a significant research area in recent years,which utilizes light-matter interactions to capture and control objects.Due to its non-contact,non-destructive,and high-precision manipulation,this technology exhibits substantial potential in fields such as life science,quantum information,and precision measurement.Notably,optical pulling and lateral forces,as critical mechanical effects that transcend traditional limitations of optical manipulation,have garnered considerable research in light manipulation.This article reviews the progress in optical pulling and lateral forces by analyzing their generation mechanisms,implementation approaches,and application potential in areas like optical sorting,micro-nano robotics driven,and life science.Additionally,it systematically reviews and discusses the development histories of optical lateral and pulling forces,and explores future prospects and challenges within this dynamic field.

In remote rotational velocity measurements, atmospheric turbulence-induced phase distortion of the vortex beam increases velocity measurement error (VME). Previous studies overlooked the reference to new dimensional information for measurement error analysis and accuracy enhancement. Our work proposes the Optimal Joint Reference VME (OJR-σ) method as a, to our knowledge, novel error optimization method; it references the measurement error information from the left- and right-handed polarized components (LP and RP) of the polarized vortex beam and optimizes the velocity measurements values weights of LP and RP in the result to minimize the VME. Combined with the GS phase recovery algorithm, this method effectively reduces system VME, enabling distortion compensation and optimal VME mode distribution evaluation. The results indicate that the OJR-σ method achieves a lower VME advantage across all modes compared to both the General Joint Reference VME (GJR-σ) and General VME (G-σ) methods, with maximum VME decreases of 29% and 71% for the High VME decline rate modes, respectively. Additionally, the OJR-σ method exhibits fewer High VME modes, resulting in an average VME of 83.6% and 71.0% compared to the other two methods. After GS compensation, the VME of High VME modes decreases by 6.12%, 4.7%, and 6.78% for the three error methods, respectively. Furthermore, the OJR-σ method proves more effective than GJR-σ in reducing the VME for high topological charge modes, achieving a decline reaching 69.9%. Our work combines the phase recovery algorithm with the reference of measurement error information from both polarization dimensions, significantly reducing VME and demonstrating the potential of polarized vortex beams in high-precision applications. This innovatively provides, to our knowledge, a novel method and theoretical support for further enhancing the accuracy of free-space rotational velocity measurements.