Explore the vast range of titles available.

Multimodal deep learning plays a pivotal role in supporting the processing and learning of diverse data types within the realm of artificial intelligence generated content (AIGC). However, most photonic neuromorphic processors for deep learning can only handle a single data modality (either vision or audio) due to the lack of abundant parameter training in optical domain. Here, we propose and demonstrate a trainable diffractive optical neural network (TDONN) chip based on on-chip diffractive optics with massive tunable elements to address these constraints. The TDONN chip includes one input layer, five hidden layers, and one output layer, and only one forward propagation is required to obtain the inference results without frequent optical-electrical conversion. The customized stochastic gradient descent algorithm and the drop-out mechanism are developed for photonic neurons to realize in situ training and fast convergence in the optical domain. The TDONN chip achieves a potential throughput of 217.6 tera-operations per second (TOPS) with high computing density (447.7 TOPS/mm 2 ), high system-level energy efficiency (7.28 TOPS/W), and low optical latency (30.2 ps). The TDONN chip has successfully implemented four-class classification in different modalities (vision, audio, and touch) and achieve 85.7% accuracy on multimodal test sets. Our work opens up a new avenue for multimodal deep learning with integrated photonic processors, providing a potential solution for low-power AI large models using photonic technology. Most photonic processors can only handle a single data modality due to the lack of abundant parameter training in optical domain. Here, authors propose and demonstrate a trainable diffractive optical neural network chip based on on-chip diffractive optics with tunable elements to address these constraints.

Full characterization of light intensity, polarization, and spectrum is essential for applications in sensing, communication and imaging. However, existing schemes rely on discrete, bulky components to capture polarization and spectrum separately, and suffer from detecting only a few values in each dimension. Here, we implement a compact disordered-guiding photonic chip with a neural network for single-shot high-dimensional light field detection. The disordered region introduces complex interference and scattering among polarized components, while the guiding region efficiently collects the outputs to on-chip photodetectors. This design encodes high-dimensional input into multi-channel intensities with high sensitivity, subsequently decoded by the neural network. Experimentally, the accurate detection of broad spectrum with mixed full-Stokes polarization states is realized with a polarization error of 1.2° and spectral resolution as high as 400 pm. Furthermore, the device demonstrates high-dimensional imaging with superior recognition performance over single-dimensional methods. This innovation offers a compact and high-resolution solution for high-dimensional detection. The authors present an on-chip intelligent high-dimensional light field detection system for single-shot accurate measurement of intensity, polarization, and spectrum — achieving superior resolution and integration compared to conventional system

Matrix computation, as a fundamental building block of information processing in science and technology, contributes most of the computational overheads in modern signal processing and artificial intelligence algorithms. Photonic accelerators are designed to accelerate specific categories of computing in the optical domain, especially matrix multiplication, to address the growing demand for computing resources and capacity. Photonic matrix multiplication has much potential to expand the domain of telecommunication, and artificial intelligence benefiting from its superior performance. Recent research in photonic matrix multiplication has flourished and may provide opportunities to develop applications that are unachievable at present by conventional electronic processors. In this review, we first introduce the methods of photonic matrix multiplication, mainly including the plane light conversion method, Mach–Zehnder interferometer method and wavelength division multiplexing method. We also summarize the developmental milestones of photonic matrix multiplication and the related applications. Then, we review their detailed advances in applications to optical signal processing and artificial neural networks in recent years. Finally, we comment on the challenges and perspectives of photonic matrix multiplication and photonic acceleration.This review summarizes the advances of photonic accelerators from the viewpoint of photonic matrix multiplication, providing a guidance for all-optical or optoelectronic-hybrid AI hardware chip system.

Polaritons in anisotropic materials result in exotic optical features, which can provide opportunities to control light at the nanoscale 1 – 10 . So far these polaritons have been limited to two classes: bulk polaritons, which propagate inside a material, and surface polaritons, which decay exponentially away from an interface. Here we report a near-field observation of ghost phonon polaritons, which propagate with in-plane hyperbolic dispersion on the surface of a polar uniaxial crystal and, at the same time, exhibit oblique wavefronts in the bulk. Ghost polaritons are an atypical non-uniform surface wave solution of Maxwell’s equations, arising at the surface of uniaxial materials in which the optic axis is slanted with respect to the interface. They exhibit an unusual bi-state nature, being both propagating (phase-progressing) and evanescent (decaying) within the crystal bulk, in contrast to conventional surface waves that are purely evanescent away from the interface. Our real-space near-field imaging experiments reveal long-distance (over 20 micrometres), ray-like propagation of deeply subwavelength ghost polaritons across the surface, verifying long-range, directional and diffraction-less polariton propagation. At the same time, we show that control of the out-of-plane angle of the optic axis enables hyperbolic-to-elliptic topological transitions at fixed frequency, providing a route to tailor the band diagram topology of surface polariton waves. Our results demonstrate a polaritonic wave phenomenon with unique opportunities to tailor nanoscale light in natural anisotropic crystals. Hyperbolic phonon polaritons that exhibit long-distance, ray-like propagation and oblique wavefronts are described at the surface of an anisotropic bulk crystal.

Silicon photonics is promising for artificial neural networks computing owing to its superior interconnect bandwidth, low energy consumption and scalable fabrication. However, the lack of silicon-integrated and monitorable optical neurons limits its revolution in large-scale artificial neural networks. Here, we highlight nonlinear germanium-silicon photodiodes to construct on-chip optical neurons and a self-monitored all-optical neural network. With specifically engineered optical-to-optical and optical-to-electrical responses, the proposed neuron merges the all-optical activation and non-intrusive monitoring functions in a compact footprint of 4.3 × 8 μm 2 . Experimentally, a scalable three-layer photonic neural network enables in situ training and learning in object classification and semantic segmentation tasks. The performance of this neuron implemented in a deep-scale neural network is further confirmed via handwriting recognition, achieving a high accuracy of 97.3%. We believe this work will enable future large-scale photonic intelligent processors with more functionalities but simplified architecture. Large-scale silicon-based integrated artificial neural networks lack of silicon-integrated optical neurons. Here, Yu et al, report a self-monitored all-optical neural network enabled by nonlinear germanium-silicon photodiodes, making the photonic neural network more versatile and compact.

Various optical crystals possess permittivity components of opposite signs along different principal directions in the mid-infrared regime, exhibiting exotic anisotropic phonon resonances. Such materials with hyperbolic polaritons—hybrid light–matter quasiparticles with open isofrequency contours—feature large-momenta optical modes and wave confinement that make them promising for nanophotonic on-chip technologies. So far, hyperbolic polaritons have been observed and characterized in crystals with high symmetry including hexagonal (boron nitride), trigonal (calcite) and orthorhombic (α-MoO 3 or α-V 2 O 5 ) crystals, where they obey certain propagation patterns. However, lower-symmetry materials such as monoclinic crystals were recently demonstrated to offer richer opportunities for polaritonic phenomena. Here, using scanning near-field optical microscopy, we report the direct real-space nanoscale imaging of symmetry-broken hyperbolic phonon polaritons in monoclinic CdWO 4 crystals, and showcase inherently asymmetric polariton excitation and propagation associated with the nanoscale shear phenomena. We also introduce a quantitative theoretical model to describe these polaritons that leads to schemes to enhance crystal asymmetry via the damping loss of phonon modes. Ultimately, our findings show that polaritonic nanophotonics is attainable using natural materials with low symmetry, favouring a versatile and general way to manipulate light at the nanoscale. Scanning near-field optical microscopy measurements show that polaritonic nanophotonics is attainable in natural low-symmetry materials, leading to a general way to manipulate light at the nanoscale.

The growing demand for enhanced computational power and energy efficiency has driven the development of optical Ising machines for solving combinatorial optimization problems. However, existing implementations face challenges in integration density and energy efficiency. Here, we propose a monolithically integrated four-spin Ising machine based on optoelectronic coupled oscillators. This system integrates a custom-designed Mach-Zehnder interferometer (MZI) symmetric matrix with a high-efficiency optical-electrical coupled (OEC) nonlinear unit. The OEC unit has an ultra-compact 0.01 mm² footprint and achieves a power efficiency of 4 mW per unit, ensuring scalability. The reconfigurable real-valued coupling matrix achieves a mean fidelity of 0.986. The spin evolution time is measured as 150 ns, with a 1.71 ns round-trip time confirmed through bandwidth measurements. The system successfully finds ground states for various four-spin Ising problems, demonstrating its effectiveness. This work represents a significant step toward monolithic integration of all-optical physical annealing systems, minimizing footprint, power consumption, and convergence time. Researchers demonstrate a monolithically integrated four-spin optical Ising machine using optoelectronic coupled oscillators. The system achieves high efficiency (4 mW per unit), fast spin evolution (150 ns), and reconfigurable coupling, advancing scalable optical computing.

Avalanche photodiodes (APDs) demand multiplication materials with low ionization coefficient ratio ( k ) for high-speed and high-sensitivity photodetection. Germanium/Silicon (Ge/Si) APDs have been preferred for a decade, leveraging the exceptional multiplication property of Si and inherent complementary metal-oxide-semiconductor (CMOS) compatibility. However, the bandwidth remains tens of gigahertz, fundamentally limited by unexpected dual-carrier multiplication in high- k Ge. Here, we transcend this material limitation by introducing a uni-multiplication-carrier concept. Through a separated absorption-charge-cliff-multiplication structure, we elaborately tailor the electric field to gradient distribution within a thin Ge region, establishing electron-dominated multiplication with a significantly reduced k . Experimentally, the device achieves a record-high bandwidth of 105 GHz at a gain of 7. This enables 8×260 Gb/s signal reception, previously only achieved by gainless photodetectors, while providing 9 dB sensitivity improvement. This work paves the way for amplifier-free optical communications, ultra-precise optical sensing, and large-scale optical computing. Researchers shatter the bandwidth record of avalanche photodiodes to 105 GHz using a uni-multiplication-carrier concept. This breakthrough paves the way for next generation optical interconnects and computing.

Vessel trajectory data are currently the most important data source for vessel trajectory data mining research. However, vessel AIS data have a short sampling time interval and a large amount of data redundancy, which hampers the efficient utilization of AIS data. In order to effectively remove redundant information from AIS data and improve its usage efficiency, a compression algorithm for vessel trajectory data compression algorithm considering critical region identification (VATDC_CCRI) is proposed. The VATDC_CCRI algorithm identifies the critical regions of a vessel’s trajectory by analyzing the distribution of node variation rates. It employs the Douglas–Peucker (DP) algorithm to compress the data in these critical regions, reducing the distortion of the trajectory after compression. Additionally, the algorithm utilizes a sliding window approach to process the initial trajectory to improve the quality of the compressed vessel trajectories and retain as many spatiotemporal characteristics of the original trajectories as possible. It combines the feature nodes from the crucial regions in the vessel’s trajectory with the results obtained from the sliding window algorithm, effectively compressing the vessel’s trajectory. Experiments conducted on individual and multiple trajectories demonstrate that the VATDC_CCRI algorithm achieves higher compression rates and exhibits faster processing speeds compared to other classical vessel trajectory compression algorithms while preserving the shape of the vessel’s trajectory significantly.

Graphene-based photodetectors have attracted significant attention for high-speed optical communication due to their large bandwidth, compact footprint, and compatibility with silicon-based photonics platform. Large-bandwidth silicon-based optical coherent receivers are crucial elements for large-capacity optical communication networks with advanced modulation formats. Here, we propose and experimentally demonstrate an integrated optical coherent receiver based on a 90-degree optical hybrid and graphene-on-plasmonic slot waveguide photodetectors, featuring a compact footprint and a large bandwidth far exceeding 67 GHz. Combined with the balanced detection, 90 Gbit/s binary phase-shift keying signal is received with a promoted signal-to-noise ratio. Moreover, receptions of 200 Gbit/s quadrature phase-shift keying and 240 Gbit/s 16 quadrature amplitude modulation signals on a single-polarization carrier are realized with a low additional power consumption below 14 fJ/bit. This graphene-based optical coherent receiver will promise potential applications in 400-Gigabit Ethernet and 800-Gigabit Ethernet technology, paving another route for future high-speed coherent optical communication networks. Graphene-based photodetectors have many advantages for applications. Here, the authors demonstrate a high-speed optical coherent receiver for optical communications based on graphene-on-plasmonic slot waveguide photodetectors.