Optical neural networks, as the next generation of artificial neural networks, have the advantages of high speed and low power consumption. Optical fully connected neural networks and optical convolutional neural networks consisting of interconnected Mach–Zehnder interferometer were investigated with respect to optical neural networks currently under development. The characteristics and implementations of two optical neural networks in terms of algorithms and structures were presented. Using pytorch and INTERCONNECT software, two kinds of optical neural networks were simulated and analyzed, and the performance of the two kinds of optical neural networks were compared in terms of image recognition accuracy, network complexity, and energy consumption, so as to elucidate the advantages and disadvantages of each of the optical fully-connected neural networks and the optical convolutional neural networks, and to provide reference for the development of optical neural networks in the future.

Photocatalysis has attracted great attention as a new method for wastewater treatment. However, in practical applications, particularly in environments with severe water pollution, the penetration of light into deeper layers is challenging, thereby restricting photocatalytic reactions mainly to the water surface. In addition, dispersed catalyst has the problem of residue and difficulty in separation. In this study, TiO2 nanoparticles (NPs) adhered quartz and polydimethylsiloxane (PDMS) optical waveguides are proposed to address the aforementioned issues. By employing this approach, light can effectively reach the deep layers of wastewater, activating the TiO2 NPs on the waveguide surface and facilitating the degradation of pollutants in these deep layers. On the other hand, TiO2 NPs adhered on the waveguide surface reduce the likelihood of secondary pollution. Experiments have demonstrated that TiO2 NPs adhered quartz optical waveguide and PDMS optical waveguide can degrade rhodamine B (Rh B) in its aqueous solution with a concentration of 1.04×10-6 g/mL at degradation rates of 78.9% and 88.8%, respectively, after 10 hours of 365 nm UV irradiation. In the testing of heavily polluted water with Rh B carbon ink mixture, the TiO2 NPs-adhered PDMS waveguide exhibits an averaged degradation rate of 63.1% after 10 hours of irradiation, which is approximately 2.21 times that achieved with a direct TiO2 NPs-dispersed photocatalytic system. The preparation method of TiO2 NPs-adhered waveguides is simple and cost-effective, and has wide application prospects in industrial wastewater and environment purification.

Fluorescent nanodiamonds (FNDs) containing nitrogen vacancy (NV) centers exhibit broad prospects in quantum sensing, quantum computing, and fluorescence imaging due to their unique properties. Functionalizing FNDs with ordered porous metal-organic frameworks (MOFs) provides a controlled environment for quantum applications. This approach facilitates the immobilization of FNDs and enhances their interactions with other substances on their surface. This study explores a composite material of FNDs and MOFs. By encapsulating FNDs within the biocompatible metal organic framework zeolite imidazolate framework-8 (ZIF-8), the fluorescence and quantum sensing performance of FNDs were effectively preserved, and selective encapsulation by ZIF-8 resulted in observable changes in the fluorescence lifetime of FNDs. Furthermore, the study investigates the modulation effects of different synthesis conditions on the size and morphology of FNDs@ZIF-8. Through an in-depth exploration of FNDs@MOFs composite materials, the research unveils the quantum characteristics of FNDs within the composite, offering new avenues for designing more flexible, high-performance fluorescence imaging devices and quantum sensors.

The triggering and propagation mechanisms of particle material avalanche constitute an exceedingly complex issue, with the crucial factor being the variation in the packing structure of the internal static accumulation area. To delve further into the packing structure of the rearrangement region within the avalanche system, this study employed an image-based approach to identify the spatial coordinates of three-dimensional cylindrical particles on the side and utilized Voronoi diagrams to characterize their packing structure. The results indicate that the Voronoi polygon area of the side-stacked particles follows a gamma distribution. Linear structures are more prone to participating in particle rearrangement, and they exhibit greater fragility in the direction perpendicular to the radius of the cylinder. During the avalanche interval, the changes in the scale of the passive layer particle packing structure are positively linearly correlated with the magnitude of the next avalanche and negatively linearly correlated with the time interval until the next avalanche occurs.

A calibration method of Shack-Hartmann wavefront sensor based on single-mode fiber beam splitter is proposed in this paper. The parallel beam and sensor were precisely aligned through single-mode fiber coupling to make the standard centroids of spot array closer to the physical position of the optical axis of the micro-lenses. The calibration setup and calibrating procedure were introduced. And 11 spherical waves with different radius of curvatures were used to verify the calibration accuracy. The experiment results and errors were compared. The results showed that compared with the conventional calibrating method by using plane wave, the average deviation of reconstructed wavefront peak to valley value with proposed method was reduced from 1.024λ to 0.667λ, and the root mean square value was reduced from 0.136λ to 0.033λ, which demonstrated that the proposed method significantly improves the reconstruction accuracy.

Datasets collected and annotated manually are inevitably contaminated with label noise, which negatively affects the generalization ability of image classification models. Therefore, designing robust classification algorithms for datasets with label noise has become a hot research topic. The main issue with existing methods is that self-supervised learning pre-training is time-consuming and still includes a large number of noisy samples after sample selection. This paper introduces the AllMix model, which reduces the time required for pre-training. Based on the DivideMix model, the AllMatch training strategy replaces the original MixMatch training strategy. The AllMatch training strategy uses focal loss and generalized cross-entropy loss to optimize the loss calculation for labeled samples. Additionally, it introduces a high-confidence sample semi-supervised learning module and a contrastive learning module to fully learn from unlabeled samples. Experimental results show that on the CIFAR10 dataset, the existing pre-trained label noise classification algorithms are 0.7%, 0.7%, and 5.0% higher in performance than those without pre-training for 50%, 80%, and 90% symmetric noise ratios, respectively. On the CIFAR100 dataset with 80% and 90% symmetric noise ratios, the model performance is 2.8% and 10.1% higher, respectively.

In order to quickly identify the upconversion luminescence ability of rare earth doped crystal materials, Yb3+/Er3+ co-doped MLuxFy (M = Li, Na, K, Rb, or Cs; x = 1, y = 4, or x = 2, y = 7) crystal materials were designed and then synthesized using a liquid-solid transfer method. The materials were characterized by using X-ray diffraction, transmission electron microscope, Raman spectroscopy, and fluorescence spectrophotometer. The results indicated that green and red upconversion luminescence of Yb3+/Er3+ co-doped MLuxFy submicron/nanocrystals were obtained under a 980 nm laser excitation, whereas these submicron/nanocrystals differed from one another in their ability to give out upconversion luminescence. In the experiment, a pump power threshold (PPT) was used to quickly distinguish the difference in upconversion luminescence ability of Yb3+/Er3+ co-doped MLuxFy submicron/nanocrystals. According to analysis, it is found that there is a correlation between the PPT and the maximum phonon energies of Yb3+/Er3+ co-doped MLuxFy submicron/nanocrystals. Therefore, in the screening of rare earth doped upconversion luminescent crystal materials, the PPT can be used to quickly identify their upconversion luminescent ability.

This study introduces a method for multiparticle force measurement and rheological analysis using a high-precision optical tweezers system, aiming to address the challenge of simultaneous characterization of interaction forces and medium viscoelasticity at the micro/nanoscale. Experiments utilizing the Aresis Tweez 300 scanning optical tweezers system achieved trapping and manipulation of 2.5-μm-diameter SiO2 particles, enabling high-resolution measurement of solvent-mediated interparticle forces (approximately 20 fN) and investigation of cellulose solution rheology (1% mass concentration), including storage modulus and viscosity. The results demonstrated that the scanning optical tweezers system could accurately measure interparticle interaction forces and characterize the dynamic response of complex fluids. By combining multicycle averaging techniques and real-time tracking methods, the signal-to-noise ratio of the system was significantly enhanced, rendering it suitable for dynamic analysis at the micro/nanoscale. This research highlights the broad application potential of scanning optical tweezers technology in colloid science, biophysics, and materials science, providing a novel methodology for micro/nanoscale force and rheology studies.

The development of topological photonics has introduced a wealth of methods for controlling light fields, with topologically protected edge states, offering exceptional properties such as unidirectional transport and immunity to defect and impurity scattering. These characteristics hold significant research and application potential in fields like optical micro-manipulation, optical communication, and optical computation. Topological metasurfaces with Dirac-like dispersion are ideal systems for generating topological edge states, where the Dirac mass term determines the width of the photonic bandgap. However, the mechanisms for controlling this mass term remain underexplored. In this work, we propose a method to extensively tune the photonic mass term by modulating the unit cell structural parameters, which alters both intra- and inter-cell coupling strengths, thereby enabling control over the bandgap and localization properties of edge states. The mass term exhibits a broad range and high sensitivity in manipulating the localization ability of topological edge states across several topological metasurfaces, demonstrating significant potential for applications in planar photonic device design.

With the continuous breakthroughs in laser technology and optical theory, as well as the rapid development of micro/nano technology, all-optical manipulation, a non-contact method, is important in scientific research and has its application scenarios ranging from fluid environments to dry solid interfaces. Recently, remarkable progress has been made in utilizing all-optical methods to manipulate micro/nano objects at dry solid interfaces. This review begins by introducing the research background of all-optical manipulation, which highlights the numerous challenges associated with manipulating micro/nano objects at dry solid interfaces. It focuses on two key theoretical frameworks: opto-thermo-elastic wave theory and photothermal-shock theory. The review analyzes related studies in detail to demonstrate the advantages of all-optical methods in this field. Additionally, other approaches based on different theories for manipulation at dry solid interfaces are discussed. Finally, the review summarizes effective solutions to current challenges and provides an outlook on the future applications of all-optical methods for manipulating micro/nano objects at dry solid interfaces, aiming at offering systematic theoretical guidance and practical references for this field.