All-optical Networks Research Articles

AWGR-based all-optical switching network for distributed machine learning

With the increase sizes of training datasets and models, the bottleneck in distributed machine learning (DML) training has shifted from computation to communication. To address this bottleneck, we propose an all-optical switching network architecture for accelerating the communication phase of DML training. Experimental results validate packets with error-free and 385 ns server-to-server low-latency communication at traffic load of 0.9. Small-scale DML training experiment deployed in the proposed architecture shows that Resnet50, Resnet101, and Vgg19 can be accelerated by 1.16x to 1.48x compared to electrical switching network. The proposed architecture demonstrates a 58.9% enhancement in cost efficiency and a 60.9% improvement in power efficiency compared to the 3-tier fat-tree architecture.

Tilted-Mode All-Optical Diffractive Deep Neural Networks.

Diffractive deep neural networks (D2NNs) typically adopt a densely cascaded arrangement of diffractive masks, leading to multiple reflections of diffracted light between adjacent masks, thereby affecting the network's inference capability. It is challenging to fully simulate this multiple-reflection phenomenon. To eliminate this phenomenon, we designed tilted-mode all-optical diffractive deep neural networks (T-D2NNs) and proposed a theoretical model for diffraction propagation in the tilted mode. Simulation results indicate that T-D2NNs address the performance degradation caused by interlayer reflections in D2NNs constructed with high-index diffractive masks. In classification tasks, T-D2NNs achieve better classification results compared to D2NNs that consider interlayer reflections.

All-Optical Metro-Access Integration Network (MAIN) Bidirectional Transmission Enabled by Coherent Digital Subcarrier Multiplexing

All-Optical Metro-Access Integration Network (MAIN) Bidirectional Transmission Enabled by Coherent Digital Subcarrier Multiplexing

Tunable Photoinduced Liquid Crystal Retarders for All-Optical Diffractive Deep Neural Networks

Tunable Photoinduced Liquid Crystal Retarders for All-Optical Diffractive Deep Neural Networks

Real-time target recognition with all-optical neural networks for ghost imaging.

The generation and structural characteristics of random speckle patterns impact the implementation and imaging quality of computational ghost imaging. Their modulation is limited by traditional electronic hardware. We aim to address this limitation using the features of an all-optical neural network. This work proposes a real-time target recognition system based on an all-optical diffraction deep neural network for ghost imaging. We use a trained neural network to perform pure phase modulation on visible light, and directly complete the target recognition task by detecting the maximum value of light intensity signals at different positions. We optimized the system by simulating the effects of parameters, such as the number of layers of the network, photosensitive pixel, unit area etc., on the final recognition performance, and the accuracy of target recognition reached 91.73%. The trained neural network is materialised by 3D printing technology and experiments confirmed that the system successfully performs real-time target recognition at a low sampling rate of 1.25%. It also verified the feasibility and noise resistance of the system in practical application scenarios.

1 × 2 broadband polarization-independent beam splitter based on air annular photonic crystal designed by a particle swarm optimization algorithm

We propose a 1×2 broadband polarization-independent beam splitter based on air annular photonic crystals. The inner and outer radii of the air annular photonic crystals at the V-shaped waveguide side and corners can influence the transmittance and broadband performance. To improve the design efficiency and transmittance of the broadband splitter, particle swarm optimization was used to reverse-design the splitter. The performance of the designed splitter was studied, and the results show that the designed splitter has a wide bandwidth range of 1530–1565 nm. It exhibits a minimum transmittance of 84% and an average transmittance of 87.8% for TE polarization and a minimum transmittance of 81% and an average transmittance of 83% for TM polarization. The response times for both polarizations are below 0.7 ps. This device has promising applications in all-optical communication networks and photonic high-density integration.

All-optical reconfigurable optical neural network chip based on wavelength division multiplexing.

Optical computing has become an important way to achieve low power consumption and high computation speed. Optical neural network (ONN) is one of the key branches of optical computing due to its wide range of applications. However, the integrated ONN schemes proposed in previous works have some disadvantages, such as fixed network structure, complex matrix-vector multiplication (MVM) unit, and few all-optical nonlinear activation function (NAF) methods. Moreover, for the most compact MVM schemes based on wavelength division multiplexing (WDM), it is infeasible to employ intrinsic nonlinear effects to implement NAF, which brings frequent O-E-O conversion in ONN chips. Besides, it is also hard to realize a reconfigurable ONN with coherent MVMs, while it is much easier to implement in WDM schemes. We propose for the first time an all-optical silicon-based ONN chip based on WDM by adopting a new adjustment mechanism: optical gradient force (OGF). The proposed scheme is reconfigurable with tunable layers, variable neurons per layer, and adjustable NAF curves. In the task of classification of the MNIST dataset, our chip can realize an accuracy of 85.13% with 4 full-connected layers and only 50 neurons in total. In addition, we analyze the influence of the OGF-based NAF under fabrication errors and propose a calibration method. Compared to the previous works, our scheme has the two-fold advantages of compactness and reconfiguration, and it paves the way for the all-optical ONN based on WDM and opens the path to unblocking the bottleneck of integrated large-dimension ONNs.

Optical activation function using a metamaterial waveguide for an all-optical neural network.

In this study, we experimentally demonstrated that the nonlinear optical coefficient of the original Si can be enhanced by incorporating a metamaterial structure into an existing silicon waveguide. The two-photon absorption coefficient enhanced by the metamaterial structure was 424 cm/GW, which is 1.2 × 103 times higher than that of Si. Using this metamaterial waveguide-based nonlinear optical activation function, we achieved a high inference accuracy of 98.36% in the handwritten character recognition task, comparable to that obtained with the ReLU function as the activation function. Therefore, our approach can contribute to the realization of more power-efficient and compact all-optical neural networks.

Research on CdSe/ZnS Quantum Dots-Doped Polymer Fibers and Their Gain Characteristics.

Polymer fibers are considered ideal transmission media for all-optical networks, but their high intrinsic loss significantly limits their practical use. Quantum dot-doped polymer fiber amplifiers are emerging as a promising solution to this issue and are becoming a significant focus of research in both academia and industry. Based on the properties of CdSe/ZnS quantum dots and PMMA material, this study experimentally explores three fabrication methods for CdSe/ZnS quantum dots-doped PMMA fibers: hollow fiber filling, melt-drawing, and melt extrusion. The advantages and disadvantages of each method and key issues in fiber fabrication are analyzed. Utilizing the CdSe/ZnS quantum dots-doped PMMA fibers that were fabricated, we theoretically analyzed the key factors affecting gain performance, including fiber length, quantum dots doping concentration, and signal light intensity. Under the conditions of 1.5 W power and 445 nm laser pumping, a maximum on-off gain of 16.2 dB was experimentally achieved at 635 nm. Additionally, using a white light LED as the signal source, a broadband on-off gain with a bandwidth exceeding 70 nm and a maximum gain of 12.4 dB was observed in the 580-650 nm range. This research will contribute to the development of quantum dots-doped fiber devices and broadband optical communication technology, providing more efficient solutions for future optical communication networks.

Terahertz spoof plasmonic neural network for diffractive information recognition and processing

All-optical diffractive neural networks, as analog artificial intelligence accelerators, leverage parallelism and analog computation for complex data processing. However, their low space transmission efficiency or large spatial dimensions hinder miniaturization and broader application. Here, we propose a terahertz spoof plasmonic neural network on a planar diffractive platform for direct multi-target recognition. Our approach employs a spoof surface plasmon polariton coupler array to construct a diffractive network layer, resulting in a compact, efficient, and easily integrable architecture. We designed three schemes: basis vector classification, multi-user recognition, and MNIST handwritten digit classification. Experimental results reveal that the terahertz spoof plasmonic neural network successfully classifies basis vectors, recognizes multi-user orientation information, and directly processes handwritten digits using a designed input framework comprising a metal grating array, transmitters, and receivers. This work broadens the application of terahertz plasmonic metamaterials, paving the way for terahertz on-chip integration, intelligent communication, and advanced computing systems.

Design and simulation of a compact polarization beam splitter based on dual-core photonic crystal fiber with elliptical gold layer

For the polarization multiplexing requirements in all-optical networks, this work presents a compact all-fiber polarization beam splitter (PBS) based on dual-core photonic crystal fiber (PCF) and an elliptical gold layer. Numerical analysis using the finite element method (FEM) demonstrates that the mode modulation effect of the central gold layer effectively reduces the dimensions of the proposed PBS. By determining reasonable structural parameters of the proposed PCF, the coupling length ratio (CLR) between X- and Y-polarized super-modes can approach 2, achieving a minimal device length of 0.122 mm. The PBS exhibits a maximum extinction ratio (ER) of − 65 dB at 1.55 μm, with an operating bandwidth spanning 100 nm (1.5–1.6 μm) and a stable insertion loss (IL) of ~ 1.5 dB at 1.55 μm. Furthermore, the manufacture feasibility and performance verification scheme are also investigated. It is widely anticipated that the designed PBS will play a crucial role in the ongoing development process of miniaturization and integration of photonic devices.

Low-loss silica waveguide 1×8 thermo-optic switch based on large-scale multimode interference couplers

Optical switches play key role in signal crossing and interconnection. Low-loss and compact optical switches are highly demanded in on-chip optical routing. A silica waveguide 1 × 8 thermo-optic switch based on cascaded 1 × 8 multimode interference (MMI) and 8 × 8 MMI couplers is experimentally demonstrated. Beam propagation method is adopted in the theoretical design and optimization. Standard CMOS technology fabrication is used in switch preparation. Air trenches are introduced to enhance the thermal modulation. At wavelength 1550 nm, the measured insertion loss and crosstalk of this 1 × 8 switch is less than 3.69 dB and −16.29 dB, respectively. And the extinction ratio is larger than 16.68 dB for all routing states. The rise time and fall time are 1.0 ms and 1.24 ms respectively. Compared with the 8-channel switch based on cascaded stages structure, the size reduces by 30 %. With future bandwidth improvement, this demonstrated switch has good potentials in the application of all-optical network routing.

Research on the Inhibition and Transmission Properties of Photonic Spiking Dynamics in Semiconductor Ring Lasers

Significant progress has been made in the research of all-optical neural networks in recent years. In this paper, we theoretically explore the properties of a neural system composed of semiconductor ring lasers (SRLs). Our study demonstrates that external optical signals generated by a tunable laser (TL) are injected into the first semiconductor ring laser photonic neuron (SRL1). Subsequently, the responses of SRL1 in the clockwise (CW) and counterclockwise (CCW) directions are unidirectionally injected into the CW and CCW directions of the second semiconductor ring laser photonic neuron (SRL2), respectively, which then exhibits similar spiking inhibition behaviors. Numerical simulations reveal that the spiking inhibition behavior of the SRL response can be precisely controlled by adjusting the perturbation time and intensity of the external injection signal, and this behavior is highly repeatable. Most importantly, we successfully achieve the stable transmission of these responses between the two SRL photonic neurons. These inhibition behaviors are analogous to those of biological neurons, but with a response speed reaching the sub-nanosecond level. Additionally, we indicate that SRL photonic neurons undergo a refractory-period-like phenomenon when subjected to two consecutive perturbations. These findings highlight the immense potential for the design and implementation of future all-optical neural networks, providing critical theoretical foundations and support for them.

Numerical simulation of an in-fiber broadband plasmonic polarization filter using photonic crystal fiber with dual elliptical gold layers

The demand for advanced and compact photonic filters is increasing in today's optical communication systems. This work introduces a new all-fiber broadband polarization filter using photonic crystal fiber with dual elliptical gold layers. The filtering performance is evaluated and analyzed by the finite element method. The simulation results demonstrate that the two elliptical gold layers can effectively induce a significant change in the energy of x-polarized light, leading to a great difference between the two polarized lights. By selecting reasonable structural parameters of the proposed PCF, the 1 mm-long in-fiber filter has a maximum extinction ratio of -108.90 dB, a low insertion loss of 0.64 dB around 1.31 µm and an ultra-broad bandwidth of 1020 nm, covering two common communication windows of 1.31 µm and 1.55 µm. Additionally, the filtering experimental scheme is discussed. This filter is expected to drive the advancement of optical integrated chips and all-optical communication networks.

Terahertz optical pattern recognition with rotation and scaling enhanced by a 3D-printed diffractive deep neural network.

Optical pattern recognition (OPR) has the potential to be a valuable tool in the field of terahertz (THz) imaging, with the advantage of being capable of image recognition with single-point detection, which reduces the overall system costs. However, this application is limited in the traditional OPR that rotation and scaling of the input image will bring about an offset of the recognition spot. Here we demonstrate a full-diffractive method to maintain the recognition spot at a fixed position, even when the input image is rotated or scaled, by using an all-optical diffractive deep neural network. The network is composed of two layers of diffractive optical elements (DOEs) without a 4f-system, and 3D-printed all-in-one. Experimental results show that our device can achieve a stable recognition of the input image regardless of its rotation (from 0° to 360°) or scaling (with a ratio from 1 to 1/1.9). This work is expected to provide enhanced functionality for compact THz systems in imaging and security applications.

Magnetization reversal by multiple optical pulses for a photonic spiking neuron with the leaky integrate and fire model

Photonic accelerators are anticipated to be the next generation of hardware processors, replacing traditional electronic accelerators. In current photonic accelerators based on artificial neural networks, photonic integrated circuits are incorporated with electronic integrated circuits to leverage their strengths: photonic circuits are used to perform linear calculations, while electronic circuits are used to perform nonlinear calculations. However, this architecture requires optoelectric conversion at each layer and is unable to leverage the superiority of light. We propose a novel photonic spiking neuron with a magneto-optical synapse and an all-optical spiking neural network. This study experimentally demonstrates that the magnetization reversal of CoFeB, which occurs during thermal accumulation owing to multiple optical pulses, is similar to the behavior of the leaky integrated and fire model of spiking neurons.

Role of depth in optical diffractive neural networks.

Free-space all-optical diffractive neural networks have emerged as promising systems for neuromorphic scene classification. Understanding the fundamental properties of these systems is important to establish their ultimate performance. Here we consider the case of diffraction by subwavelength apertures and study the behavior of the system as a function of the number of diffractive layers by employing a co-design modeling approach. We show that adding depth allows the system to achieve high classification accuracies with a reduced number of diffractive features compared to a single layer, but that it does not allow the system to surpass the performance of an optimized single layer. The improvement from depth is found to be limited to the first few layers. These properties originate from the constraints imposed by the physics of light, in particular the weakening electric field with distance from the aperture.

Low-Polarization, Broad-Spectrum Semiconductor Optical Amplifiers.

Polarization-insensitive semiconductor optical amplifiers (SOAs) in all-optical networks can improve the signal-light quality and transmission rate. Herein, to reduce the gain sensitivity to polarization, a multi-quantum-well SOA in the 1550 nm band is designed, simulated, and developed. The active region mainly comprises the quaternary compound InGaAlAs, as differences in the potential barriers and wells of the components cause lattice mismatch. Consequently, a strained quantum well is generated, providing the SOA with gain insensitivity to the polarization state of light. In simulations, the SOA with ridge widths of 4 µm, 5 µm, and 6 µm is investigated. A 3 dB gain bandwidth of >140 nm is achieved with a 4 µm ridge width, whereas a 6 µm ridge width provides more output power and gain. The saturated output power is 150 mW (21.76 dB gain) at an input power of 0 dBm but increases to 233 mW (13.67 dB gain) at an input power of 10 dBm. The polarization sensitivity is <3 dBm at -20 dBm. This design, which achieves low polarization sensitivity, a wide gain bandwidth, and high gain, will be applicable in a wide range of fields following further optimization.

Exploration On the Construction of Smart Campus All-Optical Network

With the rapid development of information technology, the construction of education informatization is urgent, and it is particularly important to promote the gradual development of colleges and universities towards digital transformation. Among them, the construction of smart campus has become an important means to enhance the level of education informatization and improve the quality of education and teaching. As an important part of the new generation of information technology, all-optical network has been widely used in the construction of smart campus. This paper first summarizes the existing network bottlenecks of traditional networks under the trend of digital campus transformation, including the problems of operation and maintenance management, network bandwidth, service expansion and weak current security, then introduces the necessity of all-optical network construction from the aspects of network bandwidth, stability and maintenance cost, and finally demonstrates the construction value of all-optical network through application scenarios. It provides a useful reference for the construction of smart campus all-optical network.

Inverse design of convolutional neural networks via nanophotonic kernels

We proposed an ultra-compact neural network based on inversely designed photonic kernels, which has the advantages of a small footprint, high computing speed, and low power consumption. The core layers of the neural network, composed of automatically optimized photonic multiplication kernels and addition kernels, can perform arbitrary photonic computations at the speed of the light wave, offering more advantages than their electronic counterparts. Furthermore, the convolution, pooling, and fully-connected layers can be realized by the appropriate permutation of the proposed optical kernels. We constructed an all-optical convolutional neural network with a high accuracy of 97.7% for recognizing handwritten digits. Our results could significantly promote the inverse design of photonic neural networks.