Two-dimensional (2D) photonic crystals (PhCs) have shown unique advantages in integrated optical computing because of their compact footprint, wavelength selectivity, and defect-mode engineering capability. However, for convolution-dominated optical linear operators, it remains challenging to realize weighting units that simultaneously provide compact device size, continuous tunability, and reconfigurable channel control. To address this issue, we propose reconfigurable optical weight banks based on a four-channel defect-engineered 2D PhC structure for compact optical convolution and transposed convolution. The proposed device combines wavelength-selective routing with composition-dependent defect-cavity engineering and TiN-assisted thermooptic tuning of adjacent GaAs coupling rods, enabling continuous transmittance modulation and optical weight loading. We numerically verify the spectral response, reciprocal multiplexing and demultiplexing behavior, and thermo-optic programmability of the proposed PhC device, and further analyze its role in implementing convolution- and transposed-convolution-based optical linear operators. The results show that the designed channels operate in the 1.64-1.69 mu m range with transmission efficiencies not lower than 89 % in the demultiplexing mode and 92 % in the multiplexing mode, while separable low-rank kernel decomposition further reduces the number of required weighting elements and device footprint. Proof-of-concept validations on imagegeneration and image-translation tasks confirm the applicability of the proposed architecture. These results indicate that defect-engineered 2D PhCs provide a feasible device platform for reconfigurable on-chip optical weighting and compact optical linear operators.

Optical resonances with high quality (Q) factor offer advantages for various research fields such as sensing, lasing and filtering. However, designing and optimizing high-Q resonances in all-dielectric metasurfaces is physically complexing, computationally demanding and time-consuming. In this work, we propose a metasurface array with Y shaped a-Si blocks unit cell which could support high-Q Fano resonance. With detailed analysis, it is confirmed that the destructive interference between electric dipole and electric quadrupole reduces the radiative loss of the resonant mode, leading to the formation of a resonant dip in the transmission spectrum. The Q factor of the resonance could be improved when the resonant wavelength approaches the wavelength of the Rayleigh anomaly. A two-stage deep learning framework is further employed to optimize the Q factor of the resonance and inverse design a structure supporting resonance with desired Q factor. At the first stage, a forward neural network (FNN) based on a residual multilayer perceptron (ResMLP) is integrated with the non-dominated sorting genetic algorithm II multi-objective optimizer to explore the trade-off between the Q factor and modulation depth, resulting in an optimized design with a Q factor of 5521-representing a 41% improvement over the highest configuration in the data set. In the second stage, a bidirectional neural network (BNN) with an architecture similar to that of the FNN is proposed, enabling on-demand generation of Fano resonances defined by mathematical parameters in the Fano formula. This model exhibits strong data efficiency and achieves accurate inverse design using a compact dataset of approximately 4000 samples, reaching a mean squared error of 4 & times;10-4. This study offers a general and computationally efficient approach for designing high-Q dielectric metasurfaces, accelerating the discovery of high-performance devices across a wide range of nanophotonic applications.

This study proposes a fiber-optic xenon (Xe) sensor with photothermally tunable gas concentration sensitivity (PTGCS). A polymer microtip doped with Xe-selective material MOF-1 and photothermal material MOF-2 on the single-mode fiber (SMF) end face acts as a Fabry-Pérot (FP) sensor for Xe gas detection. MOF-2 can control the temperature of the polymer microtip by adjusting the power of the 808 nm laser. Based on the adsorption law of MOF-1, we establish the "laser power/temperature-gas concentration-spectral response" model of the sensor. Meanwhile, an ultra-compact plasmonic grating embedded between the optical fiber core and the polymer microtip serves as the temperature indicator. The research results show that when the excitation power of the 808 nm laser is 5 mW (50 °C) and 10 mW (70 °C), the PTGCS values of the sensor are 0.191 nm/ppb and 0.106 nm/ppb, respectively. The ratio of the two PTGCS values is 1.8, which can serve as the identification parameter for Xe gas to distinguish it from other gases, particularly krypton (Kr) gas. This sensor overcomes the limitation of relying solely on the selectivity of sensitive materials to identify Xe and can self-adjust the temperature to resist temperature interference.

Surface-enhanced Raman spectroscopy (SERS) holds great promise for trace detection due to its high sensitivity and rapid analytical capability. However, fabricating SERS substrates that simultaneously provide high sensitivity and stability remains a significant challenge. In this study, 2D/3D Ag/TiO2 nanobowl arrays were fabricated using polystyrene (PS) microsphere arrays as templates. By optimizing the nanobowl diameter and silver deposition time, the substrate exhibited optimal SERS performance, achieving a detection limit as low as 10-13 M for Rhodamine 6G (R6G) and a relative standard deviation (RSD) of 7.49%. Adjusting the concentration and volume of the PS microsphere solution allowed control over the number of array layers, demonstrating that the 3D double-layer Ag/TiO2 arrays exhibited a stronger SERS enhancement effect compared to the 2D single-layer arrays. When applied to pesticide detection, the substrate achieved detection limits of 1 mu g/L for acetamiprid, 10 mu g/L for carbendazim, and 10 mu g/L for thiabendazole, with corresponding linear correlation coefficients of 0.992, 0.968 and 0.975, demonstrating high sensitivity and good linearity. The proposed fabrication method for array-based SERS substrates offers a promising strategy for detecting pesticide residues.

Quasi-bound states in the continuum (qBICs) can support resonances with high quality factor. However, the factor of qBICs decreases rapidly as the asymmetry parameter increases. Traditionally, maintaining a high factor requires a small , which brings significant challenges for nanofabrication. In contrast, our study finds an increase in the factor with increasing of in a metasurface array composed of Si blocks with circular notch. For small asymmetry parameter , the factor of the qBIC follows the inverse square law. However, when , the factor increases gradually. Through detailed analysis of the coupling strength and electromagnetic field distributions, we attribute the increase of the factor to the coupling between the qBIC and the lattice surface mode (LSM), which arises from the alignment between the diffraction wavevector and the reciprocal lattice vector. In this case, the factor under large can reach nearly and is maintained over a wide range of . Our work offers a new pathway for the design of high qBIC resonances at large asymmetry and contributes to other research field related to the high- nature of qBIC in photonics.

Within the traditional electronic neural network framework, Generative Adversarial Networks (GANs) have achieved extensive applications across multiple domains, including image synthesis, style transfer and data augmentation. Recently, several studies have explored the use of optical neural networks represented by the diffractive deep neural network (D2NN) for GANs. However, most of these focus on applications of the generative network, and there is currently no well-established D2NN architecture that simultaneously implements generative adversarial functionality. Here, we propose a novel implementation scheme for generative adversarial networks based on all-optical diffraction layers, demonstrating a complete all-optical adversarial architecture that simultaneously realizes both the generative network and the adversarial network (D2NN-GAN). We validated this method on the MNIST handwritten digit dataset, achieving Nash equilibrium convergence with the discriminator accuracy stabilizing around 50%. Concurrently, the average SSIM parameter of generated images reached 0.9573, indicating that the generated samples possess high quality and closely resemble real samples. Furthermore, we extended the framework to the KTH human action dataset, successfully reconstructing the “running” action with a discriminator accuracy of approximately 75%. The D2NN-GAN architecture introduces a fully optical generative adversarial model, providing a practical path for future optical modeling methods, such as image generation and video synthesis.

Polaritons are quasiparticles formed via light-matter interaction (e.g., with electrons or phonons). They confine light to subwavelength scales, enhance electromagnetic fields in Fabry-P & eacute;rot polaritonic resonators, and tune cavity resonance via adjusting external parameters (light frequency or dielectric environment). Thus, we propose tuning Fabry-P & eacute;rot polaritonic resonators by modifying the dielectric environment through substrate phase transitions. We perform mid-infrared nanoimaging of alpha-MoO3 nanocavities on phase-change VO2 at varying temperatures. At 25 degrees C (VO2 insulating phase), adjusting incident light frequency tunes the highly sensitive Fabry-P & eacute;rot phononic polaritonic resonator order from 10 to 2; at 90 degrees C (VO2 metallic phase), the order tunes from 7 to 2. With incident light fixed at 992 cm-1, the alpha-MoO3 nanocavity/VO2 heterojunction undergoes a low-high-low temperature cycle, showing a maximum Fabry-P & eacute;rot resonance order change of 3, along with reversible tuning ability and delayed recovery. Simulations indicate higher-order Fabry-P & eacute;rot resonant modes can be achieved in thinner and wider alpha-MoO3 nanocavities. This method provides new means and experimental support for designing tunable Fabry-P & eacute;rot resonant devices.

With the growing computational demands of artificial intelligence, optical neural networks (ONNs) have gained widespread attention due to low energy usage and high parallelism. Nonlinear activation functions (NAFs) play a crucial role in enhancing network performance. Here, we propose a nonlinear activation induced by the signal light in a two-dimensional photonic crystal (PhC) cavity, using embedded graphene oxide (GO) for a large response range modulation of both intensity and phase in complex-valued ONNs. The proposed device demonstrates substantial enhancements in classification accuracy, outperforming the best-performing electronic NAF by 6.07 % on Fashion-MNIST and 9.08 % on CIFAR-10 datasets. This device is capable of implementing nonlinear activation over a wide intensity range, making it well-suited for use as an activation function in future large-scale optical computing networks, and demonstrating the significant potential of PhC devices in this field.

Manipulation of plasmonic nanostructures with direction control has significant impact in science and engineering applications 1-5. We demonstrated the creation of the tailoring of wavevector dispersion through self-organization of plasmonic gold nanoplates via 3D printing.

Optical power splitters (OPSs) are essential components in photonic integrated circuits. The OPS with continuously tunable power splitting ratio (PSR) in multiple spectral bands can increase the data capacity and improve the flexibility of photonic integrated circuits. In this paper, we propose inverse-designed 1 x 2 OPSs on a silicon platform capable of operating in multi-wavelength bands by using topology optimization method. The PSR can be continuously tunable from 6:1 to 1:6 within a maximum of four spectral bands ranging from 1450 to 1650 nm. Furthermore, the proposed OPSs have ultracompact footprints, varying from 1.5 x 1.5 mu m2 to 2.5 x 2.5 mu m2, and good fabrication tolerances. Our design has the advantages of compactness, high flexibility, and fabrication robustness, which can be used in complex photonic integrated circuits.

With the rapid development of information technology, artificial intelligence and large-scale models have exhibited exceptional performance and widespread applications. Photonic hardware offers a promising solution to meet the growing demands for computational power and energy efficiency. Researchers have aimed to develop an efficient integrated photonic computing chip capable of supporting a wide range of application scenarios in both static and dynamic temporal domains. However, with several mainstream photonic components already well-developed, achieving fundamental breakthroughs at the level of basic computing units remains highly challenging. Here, we report a novel algorithm-hardware co-design strategy that enables in situ reconfigurability across diverse neural network models, all within a unified photonic configuration. We unlock the intrinsic capabilities of a compact cross-waveguide coupled microring component to natively support both static and dynamic temporal tasks. As a proof of concept, we experimentally integrated a turnkey soliton microcomb as the light source on the photonic computing platform, demonstrating the realization of fully connected, convolutional, and recurrent neural network models within a unified structure. The chip achieves area computing efficiency of up to 2.45 TOPS/mm2 for 208 tunable components. We evaluate the performance of the proposed chip by implementing image classification tasks on the MNIST and CIFAR-10 datasets, achieving measured test accuracies of 92.93% and 56.57%, respectively. Sentiment analysis on the IMDB dataset achieves a measured test accuracy of 80.81%. Furthermore, speech recognition is implemented by combining three neural networks within a scaled-up architecture. This work addresses the challenges of performing versatile computations on integrated photonic platforms, offering a promising solution for chip-integrated multifunctional photonic information processing.

A sensitive and practical surface-enhanced Raman scattering (SERS) substrate, consisting of one-dimensional TiO2/Ag nanowire arrays was prepared by the high-voltage electrostatic spinning method. The fabricated substrate showed high sensitivity to the SERS signals of rhodamine 6G biomolecules, with good stability and a detection limit as low as 10_ 12 M. The one-dimensional composite nanostructures exhibit excellent SERS properties, with enhancement factors of 107.

Planar metalenses have distinct advantages over their traditional bulky refractive lens in terms of being lightweight and integrable. Their remarkable ability to modulate the phase and amplitudes of incident light without restrictions offers a revolutionary approach for a multitude of frontier applications. In recent years, tunability has become a prominent direction to pursue for the investigation of planar metalens. However, existing studies on tunable metalenses predominantly concentrate on adjusting the focal length to achieve zooming effects in optical imaging, while less attention has been dedicated to the tunability of the focal field itself. This aspect, if explored, could significantly broaden the scope and flexibility of their applications, particularly in multi-mode optical imaging. In this work, a flexible and stretchable metalens is proposed and theoretically demonstrated for the dynamic tuning of the focal field morphology. To fulfill the phase requirement during dynamic modulation, the diatomic coupled resonator is applied as the basic element, which possesses higher-order freedom of phase modulation capability. Through the symmetry reforming process by transverse stretching along the horizontal direction, the focal field of the metalens can be converted from a diffraction-limited airy spot into a uniform transverse optical needle. The length of the transverse optical needle can be precisely tailored according to the degree of deformation of the metalens. This research presents a method for light field modulation and holds extensive potential for applications in dual-mode laser-scanning confocal microscopy, laser processing, optical manipulation, etc.

Silicon-based refractive index sensors are of significance in the detection of gases, biological substances and chemical compounds. Among these, optical microcavities can confine the optical field to the micrometre-scale region, and possess the advantages of high Q factor, small size and easy integration. In this paper, a trapezoidal subwavelength grating (SWG) is introduced into a slot micro-ring resonator, and the mode splitting is employed to enrich the supported standing wave modes and optimize the spatial profiles of the resonant modes. The modes' Q factor is improved and the high sensitivity and low detection limit is achieved. The optimal trapezoidal subwavelength grating double slot micro-ring resonator (T-SWGDSMRR) structure is obtained by designing the structural parameters and analyzing their effects on the sensing performance parameters and spectral characteristics. The T-SWGDSMRR, designed for detecting the glucose solution, demonstrated a low detection limit of 3.3 x 10-5 RIU and an ultra-high Q factor of up to 100825, accompanied by a refractive index sensitivity of 424 nm/RIU. Finally, a cascaded double micro-ring sensor is proposed using the vernier effect, through cascading the T-SWGDSMRR with a referential ring, the sensitivity is enhanced to 12828 nm/RIU, and the limit of detection is 3.12 x 10-6 RIU.

Optical neural networks (ONNs) have demonstrated unique advantages in overcoming the limitations of traditional electronic computing through their inherent physical properties, including high parallelism, ultra-wide bandwidth, and low power consumption. As a crucial implementation of ONNs, on-chip diffractive optical neural network (DONN) offers an effective solution for achieving highly integrated and energy-efficient machine learning tasks. Notably, wavelength, as a fundamental degree of freedom in optical field manipulation, exhibits multidimensional multiplexing capabilities that can significantly enhance computational parallelism. However, existing DONNs predominantly operate under single-wavelength mechanisms, limiting the computational throughput. Here, we propose a multi-wavelength visual classification architecture termed PhC-DONN, which integrates two-dimensional photonic crystal (PhC) components with diffractive computing units. The architecture comprises three functional modules: (1) a PhC convolutional layer that enables multi-wavelength feature extraction; (2) a three-stage diffraction layer performing parallel modulation of optical fields; and (3) a PhC nonlinear activation layer implementing wavelength nonlinear computation. The results demonstrate that the PhC-DONN achieves classification accuracies of 99.09 % on the MNIST dataset, 66.41 % on the CIFAR-10 dataset, and 92.25 % on KTH human action recognition. By introducing a wavelength-parallel classification mechanism, the architecture accomplishes multi-channel inference during a single light propagation pass, resulting in a 32-fold enhancement in computational throughput compared to conventional DONNs while improving classification accuracy. This work not only establishes a novel optical classification paradigm for multi-wavelength optical neural network, but also provides a viable pathway towards constructing large-scale photonic intelligence parallel processors.

Nonreciprocal all-optical diodes have attracted great interest for the function of light forward transmission and backward cut-off. In this paper, we present an efficient design for multi-channel photonic crystal (PC) diodes based on nonlinear Fano resonances. By optimizing the structural parameters of the asymmetric PC structure, a dual-channel PC diode is achieved at the working wavelengths of 1519.67 nm and 1533.59 nm, with the wavelength spacing of about 14 nm. The nonreciprocal transmission contrasts are 0.993 and 0.988, respectively. Moreover, by introducing an elliptical nonlinear Fano cavity, a tri-channel PC diode with ultra-small wavelength interval is realized at the working wavelengths of 1533.64 nm, 1533.97 nm and 1534.21 nm. Therefore, dynamic manipulation of multi-channel nonreciprocal unidirectional transmission becomes feasible. Our multi-channel PC diodes are compact and can be applied in high-density photonic integrated circuits.

Wavelength routers, capable of manipulating light with different wavelengths into distinct channels, are the key components in photonic integrated circuits (PICs). However, traditional wavelength routers suffer from significant losses due to backscattering caused by defects, which diminishes their transmission efficiency. Recently, topological devices based on all-dielectric valley photonic crystals (VPCs) have attracted much attention due to their robustness, small size, and suitability for fabrication. In this paper, we present multi-channel topological wavelength routers based on VPCs, which are constructed via adjusting the dielectric cylinders at the interface or the translation parameters. The maximum extinction ratios of different channels can reach 38.0 dB and 36.1 dB, respectively. Furthermore, beyond wavelength routing, this device can also simultaneously achieve multiple functions, including beam splitting and mode routing. The designed wavelength routers exhibit robustness against defects, as demonstrated by simulation results. This work offers an effective approach for the realization of topological devices and establishes a foundation for the development of high-density PICs.

Due to the weak far-field radiation and strong local field enhancement properties of Mie lattice resonances, transmission resonance with significant high Q factor is achieved which could be used in applications such as refractive index sensing, filtering and nano lasing. In this paper, we design an semi-cylinder with cuboid defect metasurface array working in the near-infrared, where a resonance with extremely narrow linewidth in transmission spectrum has been demonstrated at the wavelength of 870.9408 nm. The full width at half-maximum of the resonance is as narrow as 0.0032 nm with a corresponding Q factor of 272169. According to distributions of the electric field and displacement current, it is confirmed that the narrow linewidth resonance is generated from the Mie lattice resonance which is originated from the coupling between the out-of-plane magnetic dipole mode with Rayleigh Anomalous diffraction. It is further verified that the out-of-plane magnetic dipole mode is excited due to the induced defect in the semi-cylinder. When being applied in refractive sensing, a sensitivity and a figure of merit of 333 nm/RIU and 104063 RIU -1 are numerically achieved simultaneously. Our study here provide a new way for achieving resonances with high Q-factors.

Optical logic gates play a crucial role in all-optical signal processing systems. Traditional methods of designing logic gates require manual adjustment of structural parameters. In this paper, we utilize a genetic algorithm for inverse design, and the optical AND, OR, and NOT logic gates are achieved on a silicon platform at the working wavelength of 1.55 μm. The total area of the logic gates is fixed at 2.2 μm × 2.2 μm, convenient to be integrated with other functional devices, the optimized structural parameters are acquired for different logic gates and the contrast ratios of the OR, AND, and NOT gates are 8.55, 5.32, and 4.14 dB, respectively. The design is characterized by a compact structure, high contrast, and a high degree of freedom, offering a valuable reference for photonic integrated circuits.