Miniature light-field microscopy is a vital tool for high-speed volumetric imaging in freely moving animals, yet its spatial resolution is constrained by the insufficient sampling inherent in simultaneous spatial-angular acquisition. To address this, we propose a physics inspired light-field characteristic driven 3D reconstruction network integrating three core innovations: spatial-angular feature blocks for aliasing suppression, multi-scale feature blocks for structural fidelity, and a physics-inspired adaptive weighting loss to ensure high-quality reconstruction of sparse biological signals. These modules tackle the sampling deficiency to effectively restore spatial resolution. Crucially, we developed the first miniature dual-path imaging system to provide high-quality paired training data. Simulations demonstrate that our method achieves a lateral resolution of $5.69~\mu\text{m}$ , representing a 22% improvement over state-of-the-art deep learning methods and significantly outperforming traditional physics-based algorithms. Validations on complex biological structures and physical experiments across a large $500~\mu\text{m}$ depth-of-field confirm superior axial stability and reconstruction accuracy, providing a reliable technical pathway for high-resolution in vivo 3D imaging.
Abstract Coherence Scanning Interferometry (CSI) is a non-contacting, optical method that is widely used to measure the topography of engineering surfaces. Several linear models of CSI have been proposed in the literature including the Elementary Fourier Optics (EFO) Model that represents CSI as a two dimensional (2D) filtering operation applied to a phase modulated object field defined in a reference plane, and three dimensional (3D) approaches, including the Foil and Universal Fourier Optics (UFO) Models, that represent the surface explicitly within a volumetric framework and incorporate scattering through a 3D transfer function (TF). Although previous work, has compared the output of virtual instruments based on these models for several test cases, the underlying theory has not been compared directly. By reformulating the 2D EFO Model in an equivalent 3D representation, this paper provides for the first time, a detailed comparison of the TFs that characterise these models. This analysis reveals that all models share a common structure but differ in both their treatment of apodization and more critically the inclusion of a surface scattering term. This term, arising from the imposition of surface boundary conditions, modifies the transfer characteristics and introduces curvature dependent weighting that is absent in the EFO and UFO formulations. Numerical examples using high-curvature hemispherical features demonstrate the fundamental differences between 2D and 3D models, and a more precise estimate of phase at the boundary with the inclusion of the surface scattering term.
Abstract Two-photon microscopy (2PM) constitutes a key-enabling technology for imaging in weakly scattering specimens, delivering high-contrast signals through the quadratic dependence of fluorescence on excitation intensity. Extending 2PM to widefield or scan-less formats is attractive for rapid functional imaging. To maintain axially localized excitation while illuminating large areas, most widefield 2PM implementations rely on temporal focusing (TF) schemes re-compressing ultrashort pulses only on the sample plane. Introducing significant optical and dispersive complexity, typically demanding sub-200-fs sources, careful pre-compensation, and high-numerical-aperture objectives to efficiently deliver pulsed-shaped excitation. In this work, the realization of a widefield-excitation two-photon single-pixel microscope operating without TF while employing standard focusing optics is shown. A digital micromirror device is used to project sequential scanning patterns, and images are reconstructed from a single bucket detector. A scrambled Hadamard-based intensity-correlation pipeline is implemented. Excitation is provided by a 1064 nm industrial-grade laser delivering 300-fs pulses, demonstrating robust image formation with pulse durations above those commonly used in 2PM. High-contrast images are recovered up to 200×200 µm field of view in weakly turbid samples using moderate-NA optics. Partial-sectioning is achieved by taking advantage of the scrambled pattern’s diffractive effect. Overall, we enable TF-free, cost-effective, scalable single-pixel implementation for widefield 2PM.
Abstract Optical tweezers using metal nanostructures have leveraged extreme subwavelength focusing to circumvent the diffraction limit, enabling the isolation and label-free sensing of single nanoparticles. Over the past two decades, nanoaperture optical tweezers (NOTs) have matured into a useful tool for biophysical analysis, monitoring the conformational dynamics, binding affinities, and structural mutations of single proteins without perturbing labels or tethers. Driven by a post-machine-learning shift in biophysics toward understanding the sequence-structure-dynamics-function paradigm, NOTs have recently achieved real-time mapping of single-protein energy landscapes. Future improvements will aim to resolve the sub-microsecond protein folding dynamics by improving the signal-to-noise ratio while navigating the impacts of surface interactions and thermophoresis. This work forecasts key technical innovations over the next five years to achieve nanosecond-scale temporal resolution. By transitioning to smaller metal nanostructures (which includes moving away from nanoapertures) and operating at longer near-infrared wavelengths, near-field sensitivity can be maximized while mitigating laser-induced heating. Augmented functionalities—including integrated Raman spectroscopy (for applications like peptide identification and single-cell proteomics), and enantioselective chiral trapping will expand the utility of metal nanostructure optical tweezers. Combined with machine learning models to maximize data extraction from low signal-to-noise environments and train future predictive models on protein dynamics, these advances aim to deliver a robust platform to understand the dynamics of biomolecules.
Abstract Programmable mode selection is essential for integrated photonic networks, yet selectively isolating individual modes without delicate balancing of gain and loss remains challenging. Here we introduce a gauge-engineering method in non-Hermitian directed-graph networks that support geometry-protected pure decay modes—eigenstates exhibiting smooth exponential amplitude decay along directed paths. In fully connected configurations, a single dominant mode emerges naturally, separated from the remaining modes by a large, tunable energy gap. By introducing synthetic gauge fields through phase-compensated non-reciprocal hopping, any desired pure decay mode can be promoted to the dominant position while its amplitude profile is preserved, which can be naturally interpreted by spectral graph theory. The approach further extends to simultaneous selection of paired modes in half-connected graphs and customizable multi-mode distributions in higher dimensions via orthogonal folding. Our method enables robust, loss/gain-free control over mode profiles, advancing applications in single-mode lasers, sensors, and quantum processing.
Single spatial light modulator (SLM), full-frame-rate RGB displays require a complex off-axis architecture for the reference beams. A recent approach (Pi et al 2022 Opt. Lett. 47 4379) provides a simpler on-axis solution, but its use of conventional zero-order imaging limits SLM bandwidth utilization to 17%. We overcome this inefficiency by introducing wide-angle multi order holographic near eye display (MoHNED), exploiting higher diffraction orders and multi-order imaging. The developed wavelength-dependent spatial-frequency allocation rule, based on a sine-diffraction constraint, co-aligns the angular components of the RGB channels while keeping them spectrally separated. This raises bandwidth utilization to 86% for single-high-order and 88% for multi-order operation. As a result, the field of view (FoV) expands from 2° × 33° to 12° × 33°, without increasing SLM resolution. To ensure symmetric reconstruction in the center of the FoV, we implement an analytically derived tilted-SLM configuration. Experimental validation of MoHNED confirms high-quality, full-color 3D imaging with a large depth of field, successfully driving 4K holographic video at a native 58 Hz refresh rate.
Laser beam cutting is a widely used manufacturing process that requires precise monitoring to ensure consistent, high-quality results. To enable robust quality control, a series of cutting experiments were conducted using multiple sensors connected to a smart control platform based on FPGA-MPSoC. This platform provides time-synchronized acquisition, time-encoded labeling, and storage of visual data from a camera and acoustic signals from a microphone. A particular focus of this work is the investigation of acoustic process monitoring for laser cutting, a sensing approach that has received only limited attention in existing research, as well as the comparison and fusion of acoustic and optical sensor information. The collected process data are preprocessed and analyzed through Python-based workflows, incorporating comprehensive feature extraction from the sensor signals. Furthermore, a lightweight, interpretable machine-learning approach based on support vector machines (SVM) is employed to predict quality indicators such as dross formation. In addition, a novel class-based categorization of cut-edge quality is introduced, enabling a more differentiated representation of process outcomes and forming the basis for the proposed quality prediction framework. The proposed methodology demonstrates the feasibility of acoustic process monitoring, optical-acoustic data fusion, and interpretable quality prediction for laser beam cutting and is designed for future real-time process monitoring and control, paving the way for enhanced quality assurance in industrial applications.
As the production of additive manufacturing (AM) metal parts for service-ready and performance-critical components increases, verification of part quality is becoming an increasingly essential stage of the manufacturing pipeline. Process monitoring is a primary method for inspecting the quality of parts during fabrication and is expected to provide an avenue for feedback control in the future. Machine learning (ML) is increasingly utilised for rapid processing and interpretation of process monitoring data, allowing rapid condition monitoring both offline and in-line. Many ML classifier and predictor studies rely on instantaneous data, underutilising the inherent context of the continuous nature of laser-based AM technologies. This work leverages the contextual information encoded in the time-sequenced process monitoring data as input into a ML algorithm known as the long short-term memory (LSTM) network. The design and optimisation of an LSTM architecture utilising multi-axis infrared monitoring is presented, and the efficacy of single-source vs multi-source data investigated. This study demonstrates that multi-source LSTMs can achieve a fast, consistent, and accurate model for the prediction of bulging, depression, porosity, and over- or under-building build conditions in Laser-beam Directed Energy Deposition. F1-scores exceeding 0.95 and inference times of 2 ms per 0.5 s observation length are obtained.
Photoconductive gain in extrinsic (doped) semiconductor devices has traditionally been described by the ratio of the constant carrier lifetime to the carrier transit time. In this work, the constant-lifetime approximation is replaced by an effective lifetime, obtained through spatial averaging of photocurrent density over the device length, resulting in photocurrent saturation with applied voltage bias. The concept of effective mobility, which incorporates contributions from both majority and minority carriers, is introduced. Analytical expressions for effective carrier distributions based on effective mobility and for photoconductive gain governed by effective carrier distributions and majority-carrier photocurrent are derived. It is further shown that the maximum gain depends only on the ratio of majority to minority carrier mobilities, eliminating the unphysical prediction of unbounded gain at high bias voltages, extreme mobilities, or short device lengths. The analysis reconciles classical theory with simulation and experimental results, recovering correct gain behaviour across low- and high-drift regimes.
We demonstrate a high-performance 1G1M hybrid thermal imaging system combining a silicon refractive lens with a broadband achromatic metalens. By employing a concave surface design to achieve spatial separation of field angles, the system achieves a 30° field of view and a small f -number of 1.05. Compared to single-layer telecentric metalenses, our hybrid approach increases optical throughput by fourfold and doubles the spatial resolution. Experimental results across the 8–12 μ m range show clear thermal imaging at distances up to 20 m, with a measured modulation transfer function of 0.3 at 22 lp mm ^−1 . This work provides a compact, high-efficiency solution for next-generation long-wave infrared applications by overcoming the fundamental limitations of single-layer meta-optics.
Silicon photonic biosensors are emerging as transformative platforms for next-generation diagnostics, combining label-free, real-time detection with high sensitivity, mass scalability, and complementary metal-oxide-semiconductor compatibility. This review examines the multidisciplinary ecosystem driving their continuous advancement, from interferometric and resonant architectures achieving attomolar detection limits, to optimized device performance and multiplexing strategies that enable simultaneous multi-analyte detection on a single chip. We analyze critical enabling technologies, including surface biofunctionalization protocols, microfluidic integration, optoelectronic packaging, and system-level architectures ranging from fully integrated monolithic solutions to hybrid cartridge-based platforms. In this way, this work maps the trajectory toward miniaturized biosensor systems capable of addressing complex challenges in point-of-care diagnostics, personalized medicine, and global health equity.
Spatiotemporal couplings (STCs) in ultrashort optical pulses have been extensively studied for beams with simple spatial structure; however, their characterization in vortex pulses remains fundamentally more complex due to phase singularities and nontrivial modal structure. In this work, we develop a systematic methodology for the quantitative analysis of STCs in ultrashort vortex pulses, enabling consistent characterization of spatial chirp, angular dispersion, and pulse-front tilt in beams with nontrivial phase topology. The approach is formulated to provide results directly comparable to experiments and is applicable to structured beams beyond the Gaussian paradigm. Applied to vortex pulses generated by spiral phase plates and spiral axicons, the results reveal that STCs originate from the interplay between chromatic dispersion and wavelength-dependent modal conversion, which redistributes spectral components across spatial modes. In particular, the spectral dependence of conversion efficiency governs the magnitude of these distortions. As a representative case, broadband achromatic spiral phase plates exhibit a nearly threefold larger efficiency bandwidth ( $\approx$ 250 nm at 550 nm) compared to conventional designs ( $\approx$ 90 nm), resulting in significantly reduced STCs. These findings provide physical insight and design guidelines for ultrafast structured light.
We theoretically demonstrate strong enhancement of non-reciprocal magneto-optical (MO) cross-polarization coupling in metasurfaces mediated by quasi-bound states in the continuum (q-BICs). We do so by utilizing n-doped InSb micropillar metasurfaces under an external magnetic field in the polar Kerr configuration. q-BIC-driven MO responses follow the resonant absorption features that are further tuned by structural tilting. Tilts parallel to the incidence, yield an enhanced MO Brewster-like effect, while non-parallel tilts introduce reciprocal geometric coupling that can be isolated to reveal a strongly enhanced purely MO, non-reciprocal response. These results highlight q-BICs as an efficient route to amplify non-reciprocal effects in MO metasurfaces.
Non-destructive, on-site screening methods capable of molecular analysis directly through commercial packaging remain a significant challenge for safety, security, and quality assurance. Here, we introduce an original non-invasive photonics approach based on Raman spectroscopy that combines wavefront shaping with wavelength modulation. The wavefront shaping acts to limit the signal contribution from the packaging, while wavelength modulation further suppresses fluorescence to enhance the method's sensitivity. As a result of this judicious combination, we are able to enhance the signal-to-noise ratio of the Raman scattering obtained through the packaging up to 12-fold. We demonstrate the capability of the system by quantifying methanol in bottled spirits. Our method achieves quantification of methanol through coloured spirit bottles with a limit of detection of 0.2% (v/v) methanol in 40% ethanol, well below the reported maximum tolerable methanol concentration of 2% (v/v). In contrast to previous approaches, this method remains robust across a diverse range of coloured glass bottles, validated by measurements on real spirit bottles and samples. More broadly, this geometry establishes a versatile Raman sensing platform for assessing authenticity, composition, and contaminants directly through packaging.
Abstract Ratiometric optical thermometry based on thermally coupled green emission levels of Er 3+ ions in upconversion (UC) phosphors provides a sensitive approach for non-contact temperature measurement. In this work, a gadolinium oxysulfide (Gd 2 O 2 S) phosphor co-doped with 9% Yb 3+ and 1% Er 3+ is proposed for UC-based temperature sensing. Under 980 nm excitation, strong UC emissions corresponding to multiple Er 3+ transitions are observed, with intensities that are highly temperature dependent. The fluorescence intensity ratio (FIR), derived from integrated emission intensities of selected transition pairs, was analyzed as a function of temperature. By systematically analyzing FIR relationships across all observable emission pairs, only the thermally coupled levels, 2 H 11 / 2 → 4 I 15 / 2 and 4 S 3 / 2 → 4 I 15 / 2 , were found to exhibit linear Boltzmann behavior, thereby confirming the applicability of the FIR model for ratiometric thermometry. The phosphor exhibits a maximum relative thermal sensitivity of 1.71% K −1 at 273 K, a temperature uncertainty of 0.35 K at 300 K, and an FIR model error of 2.069% based on the fitted and experimentally measured energy gaps over the investigated temperature range of 273–873 K. These results indicate that the proposed phosphor is a promising candidate for high-precision, non-contact temperature sensing applications.
Abstract The evolution toward 6G systems is driven by the need for ultra-high data-rated, low-latency, and reliable terahertz (THz) communications. Topological valley photonics has recently emerged as a robust platform for waveguided THz integrated circuits (TICs), offering advantages in terms of robustness against disorder, fabrication imperfections and negligible bending losses. In parallel, phase change materials enable non-volatile and reversible refractive index modulation, although their integration within THz topological platforms remains largely unexplored. This work presents the theoretical design and simulated performance of novel topological resonant electrically addressed 1 × 2 and 1 × M switches implemented on a silicon photonic–electronic platform and operating at the THz band, specifically around 1 THz and 10 THz. The switching mechanism relies on Sb 2 Se 3 thin film enabling low loss and CMOS compatible programmability, activated through a voltage-controlled graphene Joule heater. Numerical results validate the feasibility of the proposed 1 × 2 switch, representing the first investigation of Sb 2 Se 3 on silicon devices at 1 THz and highlighting their potential for reconfigurable and energy-efficient TICs for future 6G systems.
Surface relief gratings (SRGs) are fundamental diffractive elements for optical wavefront control, but in azobenzene-containing polymers their inscription at large modulation depths is limited by growth saturation and profile distortions, which hinder efficient operation at infrared (IR) wavelengths. Here, we develop a quantitative optical modeling framework that combines Fresnel propagation, vectorial focusing, and finite-element electromagnetic simulations to describe how the writing field interacts with the evolving azopolymer surface during holographic inscription. The analysis shows that increasing surface modulation progressively reduces the effective optical driving force for further growth through the combined action of periodicity-dependent pattern reconstruction and relief-induced perturbation of the writing field. Experiments performed on SRGs with different periodicities reveal the same periodicity-dependent saturation trend predicted by the simulations. Within the investigated parameter range, a periodicity of 7.5 μ m provides the best compromise between achievable depth, profile fidelity, and diffraction angle. Under these conditions, we inscribe near-sinusoidal SRGs with modulation depths approaching 3 μ m in a single all-optical step, and, using a stitching strategy, extend the patterned area to 1 mm ^2 while preserving modulation-depth uniformity. This enables an azopolymer-based grating operating at the telecommunication wavelength of 1.55 μ m. These results establish a quantitative framework for understanding deep-SRG inscription limits in holographically written azopolymer gratings and provide practical design rules for reconfigurable IR diffractive photonics.
We demonstrate a novel method to purge the core of long lengths of hollow-core optical fibres where the original gas content is pushed out by helium permeating through the fibre jacket along the full fibre length. Once helium has replaced the gas in the fibre, the fibre ends can be sealed, whereupon the helium is left to permeate back out through the glass, leaving an evacuated fibre. The viability and the limits of this technique are investigated by a numerical analysis whose accuracy is validated by proof-of-principle experiments. The model predicts that the evacuation times of hollow-core fibres of hundreds of metres to kilometres length can be reduced by orders of magnitude compared to traditional methods, from hundreds of days to about one day.
Off-axis digital holography microscopy (DHM) systems have evolved from laboratory prototypes to mature metrology platforms with a wide range of applications that spans fields such as life sciences, material research, and semi-conductor industry. While fascinating developments of the core technology are still emerging, the main areas of development nowadays focus on specific applications to deliver dedicated technological solutions and resulting measurements metrics. DHM and its one-shot full-field capability have been particularly useful in the study of micro-electromechanical systems (MEMS) and of their dynamic behavior. However, the technology provides raw data that is different from incumbent technologies, and there is currently no benchmark method for comparing performances. In this article, we use a rigorous methodology based on fundamental noise definitions to define the detection limit of DHM for measuring MEMS amplitude displacement maps called vibration maps. We carefully evaluate the effect of all significant experimental parameters on the minimum detection level and discuss their optimization. We validate this study by measuring the vibration mode of a phononic crystal, requiring a detection level of one picometer. An intuitive reference metric related to the theoretical minimum noise for stroboscopic vibration measurements is also proposed. It can be used as a universal comparison between systems based on DHM or incumbent technologies while also allowing users to define their experimental parameters to achieve the desired resolution.
Several techniques have been developed for real-time tissue characterization; however, none are likely to meet the sensitivity and specificity requirements for medical diagnosis on their own. Multimodal sensing is thereby an important path towards improving the accuracy of real-time tissue characterization. In endoscopy, it is often nontrivial to accomplish such sensing in the form factor required for some applications considered. We present a compact side-viewing endoscopic platform that integrates a chip-on-tip micro-camera with time-domain diffuse optical spectroscopy, enabling surface imaging and depth-resolved tissue characterization within a 4.0 mm diameter catheter. System performance was validated using standard MEDPHOT phantoms, while a custom hybrid multilayer phantom was employed to assess accuracy, depth sensitivity, and imaging resolution. Representative in vivo measurements were made at the forearm location to demonstrate the depth sensitivity of the probe. The results illustrate that the platform reliably retrieves tissue optical properties and produces clear structural images.
