Diffractive neural networks, as a representative approach to free-space optical diffractive information processing, exploit the intrinsic advantages of light, including low power consumption and parallelism, to efficiently perform various visual tasks. For a specific visual task, such as optical classification, a physical decoder composed of cascaded diffractive surfaces must be carefully trained and subsequently fabricated with high precision. However, the precision manufacturing of diffractive processors typically involves substantial cost and produces devices that are not reprogrammable, thereby limiting the achievable parallelism for handling multiple targets. In this work, linear optical decoders in diffractive computing are virtualized as meta-decoders without a physical embodiment. This approach enables a hybrid optical-electronic classification framework that exploits correlations between optically inferred fields and computer-generated virtual reference fields. The proposed scheme integrates computational ghost diffraction with diffractive computing, referred to as ghost classification. It provides several advantages, including single-point detection, a lens-free configuration, pattern-independent flexibility, reprogrammability, and the ability to classify multi-class targets in parallel. This work leverages the complementary strengths of hybrid optical-electronic inference while incorporating lightweight electrical computations through multiplication-only correlation operations. The resulting framework serves as a transitional architecture in which each processing unit remains physically interpretable rather than a black box.

A Swin-ReconGAN network model is proposed to address the instability in MMF imaging caused by environmental disturbances. Speckle feature transfer is performed to adapt the pretrained reconstruction network, mitigating the impacts of fiber perturbations, speckle drift, and varying scattering media on image reconstruction. The proposed network achieves effective feature transfer imaging for individual fiber bending states using only 200 image-speckle pairs, significantly reducing the required training dataset size and data acquisition costs compared to traditional neural networks. After 50 independent runs in the cross-bending states feature transfer imaging, the Swin-ReconGAN achieved an average structural similarity index measure (SSIM) of 0.705, outperforming both the Transfer Learning U-Net (0.565) and the Scratch U-Net (0.620). Similarly, in the cross-medium feature transfer imaging, the Swin-ReconGAN achieved an average SSIM of 0.680 with a coefficient of variation (CV) of only 1.27%, significantly outperforming the Transfer Learning U-Net (average SSIM 0.497, CV 1.46%) and the Scratch U-Net (average SSIM 0.573, CV 9.84%). Beyond individual states, the Swin-ReconGAN also demonstrates robust capability in multiple discrete bending states feature transfer imaging. By directly performing feature transfer on speckle patterns, this model provides a practical approach for robust speckle reconstruction in small-sample scenarios.

In conventional photoconductive detectors, the shared collection path for dark current and photocurrent imposes a fundamental compromise between noise suppression and signal responsivity. To break this trade-off, we introduce a device architecture featuring a strategically integrated third electrode within a MAPbI3/Nb2CTx heterostructure. This design establishes concurrent yet distinct transport pathways: a dedicated shunt redirects dark current, and a high-efficiency channel selectively collects photogenerated carriers. The resultant device achieves an order-of-magnitude reduction in dark current compared to conventional two-terminal detectors, while concurrently enhancing the photocurrent by tenfold (from 200 to 2000 nA), and yielding a high on/off ratio of 5.2 & times; 104. Consequently, it demonstrates clear photoresponse (signal-to-noise ratio, SNR >= 3) under weak 650 nm illumination (28 nW cm-2), confirming its exceptional sensitivity for low-light imaging. Even at an irradiance of 160 nW cm-2 (approximately two-fifths of typical moonlight), the device maintains excellent resolution and produces high-contrast images with well-defined edges and rich grayscale detail. This work presents a generalizable strategy for high-performance perovskite photodetectors by synergizing heterojunction engineering with multi-terminal active modulation, significantly advancing detection capabilities under extreme low-light conditions. (sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic),(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic),(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic).(sic)(sic)(sic)(sic)(sic)(sic)(sic),(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)MAPbI3/Nb2CTx(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic).(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic):(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic),(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic).(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic),(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic),(sic)(sic)(sic)(sic)(sic)(sic)(sic)10(sic)((sic)200 nA(sic)(sic)2000 nA),(sic)(sic)(sic)(sic)(sic)5.2 & times; 104.(sic)(sic),(sic)(sic)(sic)(sic)650 nm(sic)(sic)(sic)(sic)(28 nW cm-2)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)((sic)(sic)(sic)SNR >= 3),(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic).(sic)(sic)(sic)160 nW cm-2(sic)(sic)(sic)((sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic))(sic),(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic),(sic)(sic)(sic)(sic)(sic)(sic)(sic),(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic).(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic),(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic),(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic)(sic).

The rapid evolution of eavesdropping technologies has encouraged regular updates and improvement of encryption systems. Developing a detector-dependent optical encryption scheme to tightly connect the decryption and imaging processes offers great potential to prevent eavesdropping. By designing an optically programmable dual-band photodetector, a color image encryption scheme where the photodetector functions as both a detector and a critical decryption key is demonstrated here. The distinctive optically programmable property of the photodetector enables the manipulation of its long-wavelength sensitivity via short-wavelength photonic stimulation, leading to different imaging outputs between single-pixel imaging and point-scan imaging, which therefore demonstrates a capability to decrypt information hidden in color images. This detector-dependent decryption method can effectively prevent potential information leaks when other detectors are used as eavesdropping devices. Our encryption paradigm opens new avenues for color image encryption using photodetectors, enhancing encryption security by introducing a device-based dimension.

We demonstrate the dynamic visualization of the photorefractive effect in an iron-doped lithium niobate crystal using single-pixel complex-amplitude imaging (SPCI). A 532 nm Gaussian beam locally pumps the crystal, while a uniform 632.8 nm illumination probes the photoinduced refractive index changes. Benefiting from its common-path self-referencing architecture, SPCI exhibits strong environmental robustness compared to traditional interferometry, enabling stable, continuous observation of refractive index evolution over extended periods. Our experimental results clearly resolve the spatial anisotropy and temporal progression of the photorefractive effect, further revealing a distinct polarization-dependent saturation behavior. This approach provides direct experimental access to underlying photorefractive dynamics, facilitating both fundamental understanding and device optimization.

Light-induced scattering is a key process for energy transfer in photorefractive crystals with a diffusion-based nonlinearity mechanism, which is the foundation of various devices like light amplifiers, phase conjugators, and oscillators. However, the intricate nonlinear encoding of light-induced scattering in photorefractive crystals poses a challenge for optical information retrieval and processing. In this study, we propose a deep learning approach for accurately achieving the decoding of the slowly forming fanning speckles formed by using a structured light to impinge onto the BaTiO3 crystal. We employ a U-Net to learn the mapping function from multiple fanning speckle-object pairs, enabling digital decoding of fanning beams and real-time intermediate processes. Through comparison with linear speckles generated by ground glass, we find that under the same image entropy, nonlinear speckles exhibit higher encoding capability than linear speckles. Moreover, using the same deep neural network, nonlinear speckles achieve superior decoding performance. This work not only deepens the understanding of light-induced scattering formation dynamics but also demonstrates that photorefractive crystals can serve as robust nonlinear encoders, suggesting promising applications in nonlinear optical encryption and optical computing.

Infrared metasurfaces featuring artificially designed structures provide a versatile platform for tailoring sensor properties, holding great promise for next-generation broadband surface-enhanced mid-infrared absorption spectroscopy. In particular, the over-coupled metasurfaces provide broader sensing bandwidths with a simpler fabrication compared to under-coupled pixelated metasurfaces. However, over-coupled metasurfaces has encountered several technical bottlenecks, particularly in the numerical simulation of electromagnetically induced absorption mechanisms and in the extraction of broadband signal. Herein, we propose a metasurface design based on extinction property analysis that modularly controls quasi-bound states in the continuum and dual over-coupled resonances to enable trace detection and spectral fingerprinting identification, respectively. The quasi-bound states in the continuum with surface sensitivity of 0.79 nm/nm serves as an intrinsic calibration reference, delivering a sharp spectral marker for high-fidelity signal retrieval. The calibrated framework allows accurate retrieval of broadband vibrational signatures, while the over-coupled resonances collectively amplify molecular absorption from 1800 to 1000 cm-1. Our results demonstrate that extinction-based analysis offers superior frequency resolution for visualizing the coupling between resonators and molecules. It underscores the potential of dual over-coupled metasurfaces for identifying complex analytes such as microplastics and biomarkers, paving the way for advanced mid-infrared sensing platforms.

Without imaging, single-pixel sensing aims to utilize a single-element detector to directly extract features of interest from a target. We demonstrate an optoelectronic framework for image-free tracking of a few moving targets. It integrates a Fourier filter for background suppression and a nonlinear power-law model that uniquely encodes spatial positions into optical responses. The all-optical encoder is designed based on a specific target template, which allows it to filter out the background and other types of irrelevant targets. The responses are decoded into coordinates via least-squares inversion with cross-target optimization. Using only 12 patterns, the system tracks two and three points with average errors of 1.00 and 1.77 macro-pixels experimentally at 0.67 ms temporal resolution. This work advances single-pixel tracking beyond single-target scenarios toward simultaneous localization of multiple targets with background suppression.

Intensity correlation in random light beams serves as the foundation for numerous applications, including intensity interferometry and ghost imaging. With the advances in light field manipulation techniques, an increasing number of random light sources exhibiting anomalous correlation properties have been successfully generated. In this study, we experimentally generate and characterize two novel classes of correlated light sources—holographic super-thermal light with g(2)>2 and its variant with anti-bunching-like cross-correlation with g(2)<1—filling two previously unexplored regions in a proposed two-dimensional classification framework for random light beams. When applied to ghost imaging, holographic super-thermal light can enhance the imaging quality, and four distinct types of ghost images (upright-positive, upright-negative, inverted-positive, and inverted-negative) can all be found in the second-order interference between holographic thermal light and laser light. Theoretically, we derive the modified Siegert relation for non-Gaussian holographic super-thermal light and establish the analytical model for its interference with laser light, which is corroborated by numerical simulations. Our work fills the vacant regions in the random light classification framework, enriches classical correlation modes, and provides profound insights into non-Gaussian correlations in classical optics.

Static metasurfaces enable versatile wavefront manipulation but inherently lack post-fabrication tunability, limiting their use in dynamic optical systems. Integrating liquid crystals (LCs) with metasurfaces offers a promising route toward reconfigurability; however, achieving efficient, multi-channel dynamic control over vectorial light fields remains challenging. Here, we propose and experimentally demonstrate an electrically reconfigurable LC-metasurface hybrid device for dynamic vectorial holography. Our design employs an iterative optimization algorithm to generate complex holographic images with spatially variant polarization states. The metasurface is integrated with a nematic LC cell, enabling dynamic modulation of the incident polarization state via an applied voltage. As a result, distinct vectorial holograms can be electrically switched without mechanical moving parts. We experimentally demonstrate a 16-channel Greek alphabet display. With high polarization conversion efficiency, compact footprint, and broad operational bandwidth, this platform holds promise for next-generation displays, optical encryption, and LiDAR.

Optical intensity correlation is a fundamental property of light and an essential resource for numerous optical applications. In this work, we introduce the concept of classical correlation swapping, a classical analogue of quantum entanglement swapping, to generate all-purpose spatial correlations between two independent thermal light sources. Using a spatially unresolved Mach-Zehnder interferometer and a medium variable, we theoretically and experimentally demonstrate the feasibility of the correlator with two distinct schemes for independent pseudo-thermal light beams. Notably, the resulting photon correlations can be readily tailored to exhibit either bunching (peak) or anti-correlated (dip) characteristics, with a significantly reduced number of post-selection measurements. Leveraging this classical correlator, we further demonstrate the first classical ghost imaging experiment using uncorrelated or unknown light. Numerical simulations also confirm that the correlator can not only operate well in other spatial degrees of freedom (e.g., orbital angular momentum) but also be used to establish specific spatial correlations between pseudo-thermal light sources possessing distinct correlation properties. This work opens a new avenue for harnessing uncorrelated classical light in correlation-based optical applications.

ABSTRACT In conventional photoconductive detectors, the shared collection path for dark current and photocurrent imposes a fundamental compromise between noise suppression and signal responsivity. To break this trade‐off, we introduce a device architecture featuring a strategically integrated third electrode within a MAPbI 3 /Nb 2 CT x heterostructure. This design establishes concurrent yet distinct transport pathways: a dedicated shunt redirects dark current, and a high‐efficiency channel selectively collects photogenerated carriers. The resultant device achieves an order‐of‐magnitude reduction in dark current compared to conventional two‐terminal detectors, while concurrently enhancing the photocurrent by tenfold (from 200 to 2000 nA), and yielding a high on/off ratio of 5.2 × 10 4 . Consequently, it demonstrates clear photoresponse (signal‐to‐noise ratio, SNR ≥ 3) under weak 650 nm illumination (28 nW cm −2 ), confirming its exceptional sensitivity for low‐light imaging. Even at an irradiance of 160 nW cm −2 (approximately two‐fifths of typical moonlight), the device maintains excellent resolution and produces high‐contrast images with well‐defined edges and rich grayscale detail. This work presents a generalizable strategy for high‐performance perovskite photodetectors by synergizing heterojunction engineering with multi‐terminal active modulation, significantly advancing detection capabilities under extreme low‐light conditions.

Low-saturation structural colors, such as muted or pastel hues, offer visual comfort, reduced eye fatigue, and aesthetic versatility for human-centric optical applications yet remain largely unexplored compared to their vivid counterparts. Here, we numerically investigate a thermally tunable metasurface that generates such colors by leveraging symmetry-protected bound states in the continuum (SP-BICs). Introducing controlled asymmetry into a dual-period silicon grating on a metallic substrate activates high-Q quasi-BIC resonances, selectively absorbing narrow spectral bands while maintaining >90% reflectance elsewhere. This results in structural colors with saturation below 10% in the CIE 1931 space. Integrated joule heating enables dynamic tuning of the absorption peak (~0.1 nm/K), allowing continuous control of pastel hues from 460 to 520 nm. As a simulation-based proof of concept, this design provides a promising route toward eye-friendly displays, low-glare optical labeling, and thermal sensing where low-saturation coloration is desired.

Hole injection layers (HILs) are pivotal for the performance of quantum dot light-emitting diodes (QLEDs), with solution-processed inorganic HIL materials being a primary means for the commercial application of QLEDs. Transition metal oxides (TMOs), due to their excellent stability, have been widely employed as inorganic HILs by thermal evaporation in QLEDs; however, the hole injection ability of solution-processed TMO film necessitates further enhancement owing to inferior film quality. In this study, a solution-processed molybdenum oxide (MoOx) film was used as a HIL, and its hole injection ability in the QLED was improved by tuning the oxygen states. Oxidation treatments on the MoOx layer can effectively mitigate oxygen vacancies, and consequently, the conduction band minimum (CBM) of MoOx is elevated, which enhances the hole injection through easier electron extraction from the highest occupied molecular orbitals (HOMO) of the hole transport layer. The MoOx-based red QLED exhibits significantly longer working lifetimes (T50@100 cd·m-2 of ∼66,892 h) and comparable current efficiencies (13.7 cd·A-1) compared to the Poly(3,4-ethylenedioxythiophene)-poly(styrenesulfonate) (PEDOT:PSS)-based QLEDs. Our research not only proposes a promising approach to high-performance MoOx-based QLEDs but also paves the way for further applications of TMOs in QLEDs.

We propose a modular designed over-coupled(OC) metasurface for the broadband surface-enhanced infrared absorption spectroscopy(SEIRAS) by analyzing the combined properties in the far field and near field. The customized sensors can independently modify the coupling mode, the resonance frequency, and the coupling efficiency by adjusting the vertical and horizontal structures and hybrid dielectric layers of the metasurface, respectively. Based on the independent regulation of the sensor properties, the influence of the detuning properties, the level of OC coupling, and the coupling efficiency of the signal amplification can be clearly presented through the single variable-controlling approach. These design principles are universal for customized sensors and herald possibilities for machine-learning-aided surface-enhanced infrared absorption(SEIRA) biosensing.

We propose single-pixel dual-mode microscopy (SPDM), which utilizes a digital micromirror device (DMD) to modulate object light and then performs bucket detection and zero-frequency detection in the two reflective arms of the DMD to acquire magnitude and wrapped phase images simultaneously. Benefiting from dual-mode imaging, SPDM is suitable for observing various samples, including transparent ones. Our experimental results fully demonstrate its powerful information acquisition capability. SPDM can provide a spatial resolution of up to 1.95 mu m (a spatial frequency of 512 lp/mm) and produce a pair of real-time video streams of 128 x 128 pixels at 0.51 frames per second. In addition, SPDM can be achieved as an add-on module to ordinary microscopes, which facilitates its adoption in optical microscopy. This work opens up new opportunities for biomedical imaging and industrial inspection applications.

In this work, we present a quantum-inspired, guide-star-free wavefront correction technique termed correlation adaptive optics (CAO) that leverages intensity correlations in even-symmetrical thermal light (ETL) to achieve label-free adaptive optical imaging. The sum-projection of the intensity correlation function across a set of centrosymmetric points acts as a feedback metric for iterative aberration correction, maintaining effectiveness even under object occlusion conditions. Compared to entangled photon pairs produced in spontaneous parametric downconversion, the ETL used in the CAO scheme is much easier to generate with high brightness using a commercial spatial light modulator, which thus paves the way for the adaptive aberration correction in various advanced computational imaging systems using structured illumination.

Microfluidic cavity sensors, owing to their arrayed and high-throughput design, can introduce trace fluid samples into surface-localized electromagnetic fields, enabling high-sensitivity refractive-index sensing and solvent sample monitoring. Compared to existing antenna-type microcavities, the development and design of metal-insulator-metal metamaterial microcavities with stronger local-field effects are expected to further enhance the sensitivity of microcavity refractive-index sensing and solute identification. However, such microcavity designs lack guidance from numerical solutions that integrally account for material and sample dielectric constants. Herein, independent control of resonance frequency and absorptance is demonstrated in a critically coupled microcavity sensor based on extinction properties in a metal-insulator-metal structure. Additionally, the extinction properties of the microcavity provide a dispersive method for analysing resonance modes near the centre frequency, replacing recent phenomenological theories. The results show that the metal-insulator-metal microcavity can achieve a near-perfect absorption resonance for refractive-index sensing and effectively suppress absorption interference in aqueous solutions for solute identification, yielding a sensitivity of 3667 nm RIU-1 and enhanced infrared absorptance at the fingerprints of serum proteins. This work provides a numerical framework for the inverse design of metamaterial-based microcavities in future studies.

Thermal light by nature exhibits the well-known photon bunching effect with the zero-delay second-order correlation coefficient g((2)) ideally equal to 2. The Photon extra-bunching effect, as a special class of the superbunching effect (g((2))>2), describes the enhanced photon correlation (g((2))=3) through the interference of bright twin beams produced in the spontaneous parametric down-conversion. In this work, we report that this effect can exist in classical beams and that g((2)) can approach 4 in some extended extreme models. We experimentally demonstrate the nontrivial extra-bunching correlation (similar to 1.4x gain) in pairwise-generated holographic thermal light beams with a Dove prism-embedded balanced Mach-Zehnder interferometer. The interaction of such a pair of conjugated thermal light beams can generate a pair of uncorrelated superthermal light beams, enabling ghost imaging and Young's intensity interference with enhanced visibility. The quantum-inspired interference-induced spatial extra-bunching effect provides novel insights into inducing superbunching correlation in classical light and holds the potential for application in the coherent two-photon Lidar system to enhance sensitivity.