Robustness of diffractive deep neural networks (D 2 NNs) as a key performance for the deployment in real-world remains insufficiently elucidated. We propose a criterion to evaluate the robustness of D 2 NN using the normalized cutoff frequency (NCF), and reveal that the robustness is determined by the spatial frequency propagation characteristics of the light field. We demonstrate that despite differences in D 2 NN architectures, the same robustness of D 2 NNs is attributed to equal NCF values in classification and regression tasks. This principle is physically interpretable and widely applicable to linear and nonlinear D 2 NNs. The relationship between the robustness of D 2 NN and the spatial frequency bandwidth of the optical field is discovered, enabling efficient and accurate prediction of the robustness of large-scale D 2 NN models.

Vectorial metasurface holography is a powerful technique that relies on full-dimensional modulation of optical fields to realize its full potential, laying the foundation for encoding vast amounts of optical information and enhancing optical encryption. However, current methods for achieving such modulation remain challenging, as they typically require tuning multiple structural parameters or employing multilayer metasurfaces to provide sufficient degrees of control. Here, we propose a broadband full-vectorial meta-holography (BFVM) strategy based on a purely dielectric geometric-phase metasurface, which achieves complete control over the amplitude, phase, inhomogeneous polarization, and position of light solely by varying the rotation angles of the meta-units. This approach not only enables the generation of full-vectorial holographic images with complete Stokes polarization distributions, but also supports the creation of up to 50 distinct, ultra-high-capacity vectorial holograms, each encoding a designed polarization structure with spatial variation. These structured polarization states can serve as a previously overlooked degree of freedom for enhancing holographic storage. This advancement expands the frontier of vectorial holography, paving the way for next-generation full-vectorial beam shaping, full-color holographic displays, and immersive AR/VR technologies.

Crosstalk and speckle severely limit the imaging quality of multi-plane holography. Here, we introduce a stochastic gradient descent-assisted k-space translation (KT-SGD) method that simultaneously reduces crosstalk and speckle in multi-plane holographic imaging. By encoding amplitude-only holograms that utilize the translation property of k-space, the method fully recovers object wavefronts and separates depth-dependent patterns in the Fourier domain. Both numerical simulations and optical experiments show that KT-SGD delivers high-quality reconstructions with higher Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index Measure (SSIM) than conventional methods. Additionally, we analyze and experimentally confirm the relationship between inter-plane pattern displacement and the applied tilt phase angle, accounting for aliasing effects arising from limited sampling of the spatial light modulator. This approach offers a versatile solution for high-fidelity multi-plane holographic displays, light-field manipulation, optical encryption, and holographic lithography.

Weak measurements, which enable access to weak values with minimal disturbance to the quantum system, have significantly advanced the study of counterintuitive quantum phenomena, including the quantum Cheshire Cat (QCC). Building on this methodology, we report a single-path realization of the QCC effect by mapping the two interferometric arms onto orthogonal polarization states of a single photon within the orbital angular momentum (OAM) degree of freedom. This approach demonstrates that the pre- and post-selected quantum states yield weak measurement statistics consistent with the polarization and OAM observables behaving as although they are localized differently, effectively decoupling the OAM property from its polarization-encoded carrier. Using a polarization-addressed spatial light modulator, we implement two programmable weak couplings on the same hardware: a weak absorptive perturbation (probing particle-like localization) that yields the particle-detectability weak value from the slope of the normalized post-selection probability and a small OAM rotation (probing wave-like OAM) that yields the OAM-related weak value from the slope of the interference visibility—both operating within the linear weak-coupling regime. By comparing the statistics from these two measurements, we experimentally confirm the effective decoupling of the photon’s particle localization and its OAM. This pointer-free, single-platform architecture preserves full weak-value readout capability while substantially reducing alignment and stabilization complexity, offering a compact and resource-efficient route to QCC-based property separation well suited for high-dimensional quantum applications.

With strong unpredictability, high-dimensional microwave chaotic signals are significant for secure communication. However, conventional electronic techniques face challenges in generating high-dimensional chaotic signal with large bandwidth. In this paper, a dual-loop optoelectronic oscillator with cascaded phase modulators is proposed for high-dimensional microwave chaotic signal generation. In the system, phase modulation is converted into nonlinear intensity modulation by using a Mach-Zehnder interferometer to achieve chaotic oscillation. The dual-loop structure causes numerous characteristic frequencies, enhancing high-dimensional dynamics. Its mathematical model based on the Ikeda equation is established and used for dynamics analysis. An experimental comparison demonstrates that the proposed system generates high-dimensional chaotic signal with a broadband spectrum and low time-delay signature. The nonlinear dynamics are further investigated through numerical simulations, including bifurcation route and dynamics performances, which demonstrate the high-dimensional dynamical characteristics and enhanced unpredictability of the generated chaotic signal.

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.

We present an ultra-compact 16-channel overlapping arrayed waveguide grating with 1.6 nm channel spacing that achieves low insertion loss (<1 dB), low adjacent channel crosstalk (-22 dB), and a small footprint of 540 × 590 μm2.

We experimentally demonstrate self-guided quantum tomography (SGQT) for states encoded by Laguerre-Gaussian (LG) modes. By iteratively learning the unknown quantum states from adaptively optimized projection measurements, we experimentally achieve a reconstruction fidelity of more than 97.3% in eight-dimensional systems.

Free-space optical (FSO) communication offers high capacity, license-free spectrum, and immunity to electromagnetic interference, making it a promising solution for future wireless links. However, practical deployment of FSO systems remains challenged by issues of link security and reliable transmission under atmospheric turbulence. In this work, we investigate the use of nonseparability encoding based on vectorial structured light as a turbulence-resilient and secure transmission approach. Through simulations and experiments, we compare multi-level nonseparability encoding with conventional amplitude modulation under various turbulence strengths and eavesdropping scenarios. The results demonstrate that vectorial structured light exhibits inherent security advantages, imposing higher optical signal-to-noise ratio penalties on eavesdroppers and making information recovery from partial interception significantly more difficult. These findings confirm the potential of nonseparability-based encoding for enhancing both security and robustness in FSO communication systems.

The significant birefringence in 220 nm silicon-on-insulator (SOI) silicon waveguides makes polarization diversity essential in silicon photonic integrated circuits. To address this, we propose a broadband polarization splitter-rotator (PSR) based on a silicon-nitride-on-silicon (SiN-on-SOI) hybrid integration platform. The device consists of a polarization rotator (PR) unit and an asymmetrical directional coupler (ADC). The PR unit is formed by a double-layer silicon taper coupled with a single-layer silicon nitride taper, while the ADC is implemented using a shallow-etched ridge waveguide. During operation, the TM0 mode is adiabatically converted to the TE1 mode, which is then coupled via the ADC into the narrow waveguide, propagating as the TE0 mode. Meanwhile, the input TE0 mode passes through the original waveguide without conversion. Through systematic optimization of the width parameters in the three-layer taper structure, efficient mode mixing is achieved across the entire 1260-1675 nm band, resulting in excellent broadband performance. Theoretical analysis indicates that the PSR operates effectively over the full optical communication band, with an insertion loss (IL) below 0.34 dB and an extinction ratio (ER) exceeding 17 dB. Experimental results demonstrate a 1dB bandwidth of 190 nm and an ER greater than 14.5 dB, representing the widest operational bandwidth reported for a silicon PSR to date. Furthermore, a comprehensive fabrication tolerance analysis reveals the device's high robustness to silicon nitride dimensional variations, while identifying the silicon slab thickness as the critical parameter for maintaining high extinction ratios. The device is fully compatible with standard silicon photonic foundry processes and holds strong potential for large-scale production.

Lateral scanning interferometry (LSI) is a recently developed non-contact technique for three-dimensional (3D) surface metrology, offering promising potential for rapid, large-area measurement of micro-and nano-scale structures. However, the spatial sampling error caused by straightness error during high-speed lateral scanning limits the accuracy of topography reconstruction, and this issue has not been effectively addressed in existing LSI systems. This paper presents a recovery algorithm based on pre-interpolation and the Hilbert transform (HT) for accurate reconstruction of surface topography. The spatial sampling error is estimated by extracting the amplitude envelope and identifying the envelope peak based on planar regions of each interferogram. Spline interpolation is applied to correct the spatial sampling points, resulting in uniformly sampled interference signals. Subsequently, the HT and a seven-step phase-shifting method are applied to reconstruct the surface topography. Experimental results demonstrate that the proposed method effectively corrects spatial sampling errors during high-speed scanning of periodic nanoscale surface structures.

In non-Hermitian systems, the dynamic encircling of exceptional points (EPs) engenders intriguing chiral phenomena, where the resultant state characteristics are intrinsically dependent upon the encircling handedness. An ingenious approach using simple leaky optical elements has been presented to emulate this chiral behavior without physically encircling an EP. This innovative simplification of EP properties enables a more straightforward implementation of asymmetric switching of polarization and path. Given that photons inherently possess multiple physical degrees of freedom, the research focus has shifted from single-dimensional to multidimensional asymmetric switching. Hence, there is a fundamental challenge of how to achieve multidimensional asymmetric switching through a simple and universally applicable architecture. Here, we propose and experimentally demonstrate a novel topology-optimized architecture, termed EP-encirclement emulation tailoring, enabling multidimensional asymmetric switching. Theoretical analysis reveals that our architecture eliminates the 3-dB inherent loss in conventional architecture by replacing couplers with (de)multiplexers. Building upon this architecture, we harness all-fiber devices to implement a high-performance asymmetric switching of polarization, mode, and orbital angular momentum (OAM). To our knowledge, this is the first experimental demonstration of asymmetric OAM switching to date. Our work provides an efficient topology architecture for emulating dynamic EP encirclement, paving the way for universal and flexible asymmetric switching devices.

We present a turbulence-resilient all-optical classifier that learns turbulence-resilient features using randomized phase screens. Our system achieves all-optical classification of MNIST digits through turbulent media, offering a new paradigm for robust free-space optical communication. © 2026 The Author(s)

We design, fabricate and experimentally characterize a weakly-guiding triple-ring ringcore fiber that supports 24 orbital angular momentum modes including radial high-order ones with low loss and crosstalk, showing its potential for advancing high-capacity fiber-optic communications.

Seeing clearly through fog, murky water, or other scattering materials is difficult with conventional imaging systems. One powerful tool is polarimetric imaging, which captures how light is polarized to reveal hidden details and material properties. But traditional systems must choose between a wide aperture (for gathering light) and a large depth of field (for keeping the whole scene in focus), limiting their effectiveness in challenging environments. Here we report a metasurface-based polarimetric light field camera capable of single-shot acquisition of five-dimensional (5D) light field data, comprising spatial, angular, and polarization information. A polarization-multiplexed metasurface placed between the main lens and the image sensor serves as an integrated optical encoder, enabling complete 5D capture without moving parts or stacked polarization optics. This configuration supports near-infinite DOF imaging through computational refocusing while maintaining a large aperture. As a proof of concept, we demonstrate all-in-focus polarimetric imaging over a 35 cm depth range in highly turbid water, achieving up to 6.63- and 7.30-fold contrast enhancement relative to the scattered images obtained with a pair of orthogonal linear polarizations. These results establish a compact and scalable pathway to high-dimensional imaging with potential applications in scientific instrumentation, industrial inspection, and environmental sensing.

We propose an environment-coupled dynamic model to analyze the mode splitting of orbital angular momentum (OAM) beams propagating in complex environments and suppress the mode splitting through pre-compensated OAM. The feasibility of the model in analyzing mode splitting and splitting suppression was demonstrated through spatial light modulators, generated turbulence, and water flow experiments. Furthermore, we designed the mode content and relative argument for multiple-emitted-beam multiplexing and realized binary encoding based on mode content control. This study provides a physically analyzable approach for the stable propagation of OAM beams passing through an inhomogeneous medium, and also provides a research paradigm for the propagation analysis of other structured beams in complex environments.

Significance The integration of femtosecond laser processing with topological principles has revolutionized the precision fabrication and functional design of three-dimensional (3D) micro/nanostructures. Its significance lies in advancing topological photonics by enabling accurate simulations of condensed matter phenomena through laser-direct-written waveguide arrays in transparent media, thereby establishing foundations for designing topologically protected optical devices and quantum simulation platforms. Simultaneously, it empowers innovations in biomimetic functional structures, where laser additive/subtractive techniques precisely construct geometrically complex microstructures to replicate superior bio-inspired physicochemical properties, driving technological breakthroughs in biomedical engineering and soft robotics. Crucially, this synergy overcomes limitations of traditional manufacturing, achieving nanoscale 3D fabrication that underpins next-generation functional materials and smart devices. Progress Femtosecond laser fabrication exhibits significant potential in micro/nanoscale topological structures through dual aspects: photonic topological structures and geometric topological structures. The former concentrates on light propagation and manipulation, while the latter addresses material/fluid transport and responses. Both dimensions fundamentally leverage the distinctive robustness and functional advantages endowed by topological configurations. Recent advances leveraging femtosecond-laser fabrication have yielded breakthroughs in topological photonics and geometric micro/nano-topologies. Significant advances in topological photonics include Rechtsman et al.'s demonstration of a helical-waveguide hexagonal lattice based on Floquet topological insulator principles (Fig. 1(a)), where laser-induced effective gauge fields break time-reversal symmetry, enabling unidirectional robust propagation immune to defects and lattice dislocations. Expanding into non-Hermitian regimes, Yu et al. fabricated multi-unit cascaded waveguides in glass (Fig. 2(c)), experimentally verifying universal rules for exceptional point (EEP) manipulation in multi-level systems, thereby establishing a theoretical framework for high-order non-Hermitian topological control. For geometric micro/nanostructures, Reddy et al. pioneered hybrid photonic structures integrated on single-mode fiber tips (Fig. 3(d)), using femtosecond laser writing to achieve on-chip and fiber-end generation/dynamic modulation of structured light fields, offering compact solutions for high-dimensional optical communications. Thiel et al. leveraged laser writing for 3D chiral architectures, fabricating uniaxial helical photonic crystals (Fig. 5(c)) that exhibit strong orbital-angular-momentum-based chirality with 5% transmittance for specific circularly polarized light versus 95% for other polarizations, substantially broadening spin-based chiral optics. In biomimetic applications, Arakawa et al. generated vascular networks spanning multiple scales via photo-degradable peptide hydrogels (Fig. 6(c)), establishing near-physiological microchannel systems with spatiotemporal remodeling capacity, crucial for advanced tissue engineering and programmable vascular modification. Collectively, these advancements highlight femtosecond laser fabrication as a versatile platform for engineering multifunctional topological architectures. By enabling precise manipulation of both photonic degrees of freedom and geometric symmetries, this approach establishes a cross-disciplinary framework for realizing unprecedented robustness and reconfigurability in optical, quantum, and biomedical systems. Such capabilities position it as an indispensable tool for advancing next-generation photonic integration, quantum control paradigms, and biologically inspired engineering platforms. Conclusions and Prospects This study confirms femtosecond laser fabrication as an irreplaceable platform for 3D topological structures, having validated innovative mechanisms like topologically protected transport and fluidic optimization through photonic simulations and biomimetic functional structures. Nevertheless, challenges persist in limited large-scale fabrication efficiency, complex 3D band structure design, insufficient dynamic control, and weak compatibility with emerging functional materials. Future research must prioritize overcoming manufacturing bottlenecks via multi-beam parallel processing and machine-learning-assisted inverse topological design to enhance efficiency, alongside developing stimuli-responsive materials for light/thermal/electrically driven dynamic reconfiguration of topological architectures. Moreover, deeper multidisciplinary integration should explore emerging applications at quantum-bio interfaces and multi-scale integrated systems. Manufacturing bottlenecks are being resolved, and innovative designs are maturing. Enhanced synergistic responses between materials and structures are advancing, particularly through frontier interdisciplinary convergence with quantum technologies, flexible electronics, and multiphysics devices. Femtosecond laser processing of 3D topological structures plays a pivotal role in exploring novel fundamental physical effects. This capability will also drive transformative breakthroughs in multiple cutting-edge domains. These applications include novel high-density photonic integration, high-performance optical sensing, and advanced quantum technologies. Further applications encompass wearable smart systems, biomimetic engineering, biomedical diagnostic/therapeutic devices, and efficient energy conversion systems. Consequently, this technology demonstrates exceptional scientific potential and strong strategic importance.

The Doppler effect is a fundamental and widely observed physical phenomenon. In recent years, the conventional Doppler effect has evolved into the rotational and vectorial forms with the study of structured light. Particularly, for beams with angularly-distributed phase, the skew Poynting vector would give rise to Doppler shifts when interacting with transverse motions, namely the rotational Doppler effect (RDE). Here, we highlight recent breakthroughs in both principles and applications of Doppler effects associated with structured fields. One such breakthrough is the demonstration of the structure-shearing Doppler effect (SDE). Recent studies suggest that the spatial structures of light beams would provide a modification of the group and phase velocities of optical fields, and then an open question remains how the spatial confinement on optical fields affects the Doppler shifts of structured beams. We found that the transverse structure of optical fields naturally causes an extra red shift on the original Doppler shift. Another breakthrough concerns the miniaturization of optical metrology and sensing instrument. The first all-fiber rotational Doppler velocimetry (AF-RDV) is designed and fabricated with mode-sculpted fiber-optic elements, enabling compact, cost-effective, and reference-free angular velocity sensing. By leveraging mode-selective coupling, the system reveals the mode-change-dependent nature of the RDE and shows great promise for biomedical and industrial applications.

The emergence of topological photonics has revolutionized the paradigm of photonic device design, with its core principle being the utilization of topological invariants to achieve robust control over light propagation and localization. In recent years, this concept has been successfully introduced into fiber optics, giving rise to topological photonic crystal fibers (TPCFs). The Dirac-vortex TPCFs, based on the Jackiw-Rossi zero mode, are realized by introducing a generalized Kekulé modulation in their cross section. This approach exhibits remarkable properties, including a controllable number of modes, a large bandwidth for single-polarization single-mode operation, and robustness against structural disorder. In this paper, we give a comprehensive overview of recent advances in Dirac-vortex TPCFs, including the physical mechanisms with its origin of topological photonic crystals and photonic crystal fibers, theoretical design, and experimental realization. We also discuss opportunities and challenges of Dirac-vortex TPCFs for future applications.