All Optical Nonlinear Activation Function Based on Ge/Si Hybrid Micro-Ring Resonator

Bo Wu,Hengkang Li,Jianji Dong,Xinliang Zhang

DOI: https://doi.org/10.1364/acpc.2021.m5i.6

2021-01-01

Abstract:We demonstrate an all optical nonlinear activation function device based on Ge/Si hybrid micro-ring resonator. Three different types of nonlinear function are realized with threshold below 1mW and they can work well beyond 5MHz.

What problem does this paper attempt to address?

Low-threshold All-Optical Nonlinear Activation Function Based on a Ge/Si Hybrid Structure in a Microring Resonator

Bo Wu,Hengkang Li,Weiyu Tong,Jianji Dong,Xinliang Zhang

DOI: https://doi.org/10.1364/ome.447330

2022-01-01

Optical Materials Express

Abstract:Optical nonlinear activation function is an indispensable part of the optical neural network. While linear matrix computation has thrived in an integrated optical neural network, there are many challenges for nonlinear activation function on a chip such as large latency, high power consumption and high threshold. Here, we demonstrate that Ge/Si hybrid structure would be a qualified candidate owing to its property of CMOS-compatibility, low nonlinear threshold and compact footprint. Thanks to the strong thermal-optic effect of germanium in conjunction with micro-ring resonator, we experimentally demonstrate three different types of nonlinear function (Radial basis, Relu and ELU functions) with a lowest threshold of 0.74 mW among our measured nonlinear functions and they can work well with a repetition rate below 100 kHz. Simultaneous size shrinkage of germanium and resonance constraint inside germanium is proposed to speed up response time. Furthermore, we apply our measured nonlinear activation function to the task of classification of MNIST handwritten digit image dataset and improve the test accuracy from 91.8% to 94.8% with feedforward full-connected neural network containing three hidden layers. It proves that our scheme has potential in the future optical neural network.
All-optical nonlinear activation function based on Germanium Silicon composite waveguide

Hengkang Li,Bo Wu,Jianji Dong,Xinliang Zhang

DOI: https://doi.org/10.1109/icocn55511.2022.9901099

2022-01-01

Abstract:Nonlinear activation functions, a part of artificial neuron network, is responsible for nonlinearly mapping the input of the neuron to the output. The optical realization of nonlinear activation functions has become a research hotspot, which is optical activation function. Here, we propose and experimentally demonstrate an all-optical implementation of optical activation function based on germanium silicon composite waveguide, which takes advantage of the intrinsic absorption and the plasma dispersion effect of germanium in the communication band. It realizes two different nonlinear activation functions. This proves that our scheme has a great potential for optical activation functions.
Programmable All-optical Nonlinear Activation Functions Based on a Micro-ring Resonator with Phase-change Materials

Z. L. Fu,R. Jiang,X. Ding,Z. Wang,J. Wang,C. C. Wu

DOI: https://doi.org/10.1109/acp55869.2022.10088664

IF: 3.8

2022-01-01

Optics Express

Abstract:A programmable hardware implementation of all-optical nonlinear activation functions for different scenarios and applications in all-optical neural networks is essential. We demonstrate a programmable, low-loss all-optical activation function device based on a silicon micro-ring resonator loaded with phase change materials. Four different nonlinear activation functions of Relu, ELU, Softplus and radial basis functions are implemented for incident signal light of the same wavelength. The maximum power consumption required to switch between the four different nonlinear activation functions in calculation is only 1.748 nJ. The simulation of classification of hand-written digit images also shows that they can perform well as alternative nonlinear activation functions. The device we design can serve as nonlinear units in photonic neural networks, while its nonlinear transfer function can be flexibly programmed to optimize the performance of different neuromorphic tasks.
All-Optical Nonlinear Activators Using Silicon-Based Waveguides and Microring Resonators for Photonic Neural Networks

Bei Chen,Xiaozhi Liu,Zichao Zhao,Zeyu Yu,Yuehai Wang,Jianyi Yang

DOI: https://doi.org/10.1109/ACP55869.2022.10088743

2022-01-01

Abstract:Silicon-based waveguides and all-pass microring resonators used as the all-optical nonlinear activators for photonic neural networks are demonstrated. For the three-layer, full-connected neural network and the MNIST dataset, the recognition accuracies based on the proposed silicon-based devices all can reach more than 90%.
Optical Nonlinear Activation Functions Based on MZI-Structure for Optical Neural Networks

Qiang Li,Shengping Liu,Yang Zhao,Wei Wang,Ye Tian,Junbo Feng,Jin Guo

DOI: https://doi.org/10.1364/acpc.2020.t4d.4

2020-01-01

Abstract:We experimentally demonstrated an on-chip optical nonlinear activation function circuit for optical neural networks based on a conventional linear transformer, MZI-mesh. The proposed circuit is reconfigurable to perform multiple types of nonlinear activation functions.
Strong Nonlinear Optics In On-Chip Coupled Lithium Niobate Microdisk Photonic Molecules

Min Wang,Ni Yao,Rongbo Wu,Zhiwei Fang,Shilong Lv,Jianhao Zhang,Jintian Lin,Wei Fang,Ya Cheng

DOI: https://doi.org/10.1088/1367-2630/ab97ea

2020-01-01

New Journal of Physics

Abstract:High-quality lithium niobate (LN) thin-film microresonators provide an ideal platform for on-chip nonlinear optical applications. The strict phase-matching condition should be satisfied for an efficient nonlinear optical process, which requires dispersion engineering with an LN microresonator. However, this is challenging in single microresonator, resulting from the fabrication error. Here, we demonstrate strong nonlinear effects in a photonic molecule (PM) structure composed of two strongly coupled lithium niobate microdisks. The size mismatch of the microdisks enables phase matching by employing coupling-induced frequency splitting to compensate for the material and geometric dispersion. With a continuous wave excitation, rich nonlinear optical phenomena including cascaded four-wave mixing and stimulated Raman scattering were observed around the second harmonic signal. Meanwhile, an ultra-high four-wave mixing absolute conversion efficiency of 14% as obtained when the second harmonic signal power is at microwatts level. The LN PM is of great potential for applications in nonlinear integrated photonics.
All-Optical Nonlinear Activation Function Based on Germanium Silicon Hybrid Asymmetric Coupler

Hengkang Li,Bo Wu,Weiyu Tong,Jianji Dong,Xinliang Zhang

DOI: https://doi.org/10.1109/jstqe.2022.3166510

IF: 4.9

2022-01-01

IEEE Journal of Selected Topics in Quantum Electronics

Abstract:Nonlinear activation functions are crucial for optical neural networks (ONNs) to achieve more various functions. However, the current nonlinear functions suffer from some dilemma, including high power consumption, high loss, and limited bandwidth. Here, we propose and demonstrate an all-optical implementation of a nonlinear activation function based on germanium silicon hybrid integration. The principle lies in the intrinsic absorption and the carrier-induced refractive index change of germanium in C -band. It has a large operating bandwidth and a response frequency of 70 MHz, with a loss of 4.28 dB and a threshold power of 5.1 mW. Adopting it to the MNIST handwriting data set classification, it shows an improvement in accuracy from 91.6% to 96.8%. This proves that our scheme has great potential for advanced ONN applications.
Highly-efficient second and third harmonic generation in a monocrystalline lithium niobate microresonator.

Jintian Lin,Yao Ni,Zhenzhong Hao,Jianhao Zhang,Wenbo Mao,Min Wang,Wei Chu,Rongbo Wu,Zhiwei Fang,Lingling Qiao,Wei Fang,Fang Bo,Ya Cheng

2018-01-01

Abstract:Nonlinear optics in whispering-gallery-mode (WGM) microresonators have attracted much attention. Owing to strong confinement of the light in a small volume, a WGM microresonator can dramatically boost the strength of light field, giving rise to enhancement of the nonlinear interactions of light with the resonator material. Here, we demonstrate highly efficient second harmonic generation (SHG) and third harmonic generation (THG) in an on-chip monocrystalline lithium niobate (LN) microresonator. Benefitting from a cyclic phase matching scheme for the transverse-electric WGMs, nonlinear wavelength conversion utilizing the largest second-order nonlinear coefficient d33, which has not been demonstrated until now despite of its obvious advantage, is successfully achieved in an X-cut LN microresonator. We obtain high conversion efficiencies in not only the SHG (3.8% mW^-1) but also the THG (0.3% mW^-2) realized through a cascaded sum-frequency process. Our results represent a major step toward the classical and quantum photonic integrated circuits.
Programmable low-threshold optical nonlinear activation functions for photonic neural networks

Ying Huang,Weiping Wang,Lei Qiao,Xiaoyan Hu,Tao Chu

DOI: https://doi.org/10.1364/OL.451287

IF: 3.6

2022-01-01

Optics Letters

Abstract:We experimentally demonstrate two types of programmable, low-threshold, optically controlled nonlinear activation functions, which are challenging to realize in photonic neural networks (PNNs). These devices rely on on-chip integrated Ge-Si photoelectric detectors and silicon electro-optical switches, and they generate rectified linear unit (ReLU) or sigmoid functions with arbitrary slopes without additional electrical processing. Both devices function at an extremely low threshold of 0.2mW. The embedding of these nonlinear activation functions into convolutional neural networks facilitates the attainment of high inference accuracies of up to 95% when applied to Modified National Institute of Standards and Technology (MNIST) handwritten digit-classification tasks. The devices are suitable for low-power PNNs with an arbitrary number of propagation layers in photonic-computing chips. (C) 2022 Optica Publishing Group
A High-Stability All-Optical Nonlinear Activator for Optical Computing

Caihong Teng,Xingyu Tang,Yixuan Huang,Wen Du,Weijie He,Aobo Ren,Jiang Wu,Zhiming Wang

DOI: https://doi.org/10.1109/jlt.2023.3290021

IF: 4.7

2023-01-01

Journal of Lightwave Technology

Abstract:We propose an all-optical nonlinear activator based on the fiber-optic ring resonator (FORR) to implement physically nonlinear activation at low optical signal intensity. The FORR is fabricated from a simple passive optical element, suggesting that optical nonlinearity can be generated by the inherent properties of the device without the need for additional energy consumption. Our device exhibits advantages of excellent nonlinear traits, high stability, versatility, and compatibility in combination with existing optical chips. These features have potential advantages for the scalability of optical neural (ONNs). As a proof of concept, we show the capability of FORR as a nonlinear unit for nonlinear mapping in ONNs. For classification and super-resolution reconstruction tasks, the performance of our model, obtained from experimental measurements, is comparable to that of activation functions commonly used in computers. Our work may provide a feasible avenue for novel layout and wider deployment of all-optical neural networks.
Reconfigurable All-Optical Integrated Nonlinear Activator with Switchable Response Functions for Photonic Neural Networks

Bei Chen,Jian Wang,Zichao Zhao,Xiaowen Xiong,Jianyi Yang,Ming Li,Ninghua Zhu

DOI: https://doi.org/10.1364/ofc.2024.m4c.5

2024-01-01

Abstract:We experimentally demonstrate a reconfigurable all-optical integrated nonlinear activator with switchable response functions, including Gaussian, Radial Basis, Soft-plus, leaky ReLU, Swish and clamped ReLU functions, especially all triggered by low-power inputs.
All-optical Nonlinear Activator for Optical Neural Network with Low Threshold Power and Ultrafast Response Speed

Ziwen Zhou,Ze Zhang,Weiqin Wang,Wenyi Peng,Jiawang Xiong,Siqi Yan,Ming Tang

DOI: https://doi.org/10.1364/cleo_at.2023.jw2a.46

2023-01-01

Abstract:We proposed a novel all-optical nonlinear activator based on the hybrid integration of graphene and silicon/metallic double slot structure, which exhibits low threshold power and ultrafast response speed.
Optical Ring-Resonator Modulator Based on 0.8 Μm CMOS Technology

Wang Fan,Qiu Chen,Xiao Si-Miao,Hao Yin-Lei,Jiang Xiao-Qing,Wang Ming-Hua,Yang Jian-Yi

DOI: https://doi.org/10.1088/0256-307x/26/10/104206

2009-01-01

Chinese Physics Letters

Abstract:Optical ring-resonator-based modulators are fabricated on the silicon-on-insulator material through the mature commercial 0.8 mu m complementary metal oxide semiconductor foundry. The device configuration is based on a single ring resonator coupled to one bus waveguide. The waveguide widths are about 1 mu m. The p-i-n junctions are employed to inject currents. The experimental result shows that the ring resonators with the quality factor of above 40000 are obtained. The maximum extinction ratio of the modulators is larger than 10dB. The speed is tens of nanoseconds, and the corresponding injected current is smaller than 10 mA.
All-optical neural network nonlinear activation function based on the optical bistability within a micro-ring resonator

Hui Zhang,Jin Wen,Zhengwei Wu,Qian Wang,Huimin Yu,Ying Zhang,Yu Pan,Lan Yin,Chenglong Wang,Shuangchao Qu

DOI: https://doi.org/10.1016/j.optcom.2024.130374

IF: 2.4

2024-05-01

Optics Communications

Abstract:Training all-optical neural networks in itself remains an unresolved problem, and the challenges compound when the problem is turned into the hardware implementations. In this paper, we propose a nonlinear activation function based on optical bistability within a micro-ring resonator (MRR), achieving threshold control without external modulation. Furthermore, a convolutional neural network similar to the Le-Net-5 architecture is designed, in which all nonlinear activation functions are composed of optical bistable hysteresis loop. The numerical simulation results demonstrate that the recognition rate on the Fashion-MNIST dataset can achieve 91.3%, which means that the optical neuromorphic computation can be implemented by utilizing the nonlinear optical effects themselves in the all-optical hardware. Such a scheme promises access to the all-optical neural network training in the optical hardware environment compared to numerical activation functions.

optics
Cascadable Optical Nonlinear Activation Function Based on Ge-Si.

Baiheng Zhao,Bo Wu,Hailong Zhou,Jianji Dong,Xinliang Zhang

DOI: https://doi.org/10.1364/ol.539722

IF: 3.6

2024-01-01

Optics Letters

Abstract:To augment the capabilities of optical computing, specialized nonlinear devices as optical activation functions are crucial for enhancing the complexity of optical neural networks. However, existing optical nonlinear activation function devices often encounter challenges in preparation, compatibility, and multi-layer cascading. Here, we propose a cascadable optical nonlinear activation function architecture based on Ge-Si structured devices. Leveraging dual-source modulation, this architecture achieves cascading and wavelength switching by compensating for loss. Experimental comparisons with traditional Ge-Si devices validate the cascading capability of the new architecture. We first verified the versatility of this activation function in a MNIST task, and then in a multi-layer optical dense neural network designed for complex gesture recognition classification, the proposed architecture improves accuracy by an average of 23% compared to a linear network and 15% compared to a network with a traditional activation function architecture. With its advantages of cascadability and high compatibility, this work underscores the potential of all-optical activation functions for large-scale optical neural network scaling and complex task handling.
Ultrafast Silicon/Graphene Optical Nonlinear Activator for Neuromorphic Computing

Ziwen Zhou,Chen Liu,Weiwei Zhao,Jingze Liu,Ting Jiang,Wenyi Peng,Jiawang Xiong,Hao Wu,Chi Zhang,Yunhong Ding,Francesco Da Ros,Xingyuan Xu,Kun Xu,Siqi Yan,Ming Tang

DOI: https://doi.org/10.1002/adom.202401686

IF: 9

2024-10-05

Advanced Optical Materials

Abstract:In this article, an ultrafast and energy‐efficient all‐optical nonlinear activator is proposed that leverages the saturable absorption and ultrafast carrier relaxation time of monolayer graphene, combined with the strong light confinement of a double‐slot structure. This breakthrough in speed and energy efficiency of all‐optical nonlinear activators opens the door to significant improvements in the performance and applicability of ONNs. Optical neural networks (ONNs) have shown great promise in overcoming the speed and efficiency bottlenecks of artificial neural networks. However, the absence of high‐speed, energy‐efficient nonlinear activators significantly impedes the advancement of ONNs and their extension to ultrafast application scenarios like real‐time intelligent signal processing. In this work, a novel silicon/graphene ultrafast all‐optical nonlinear activator, leveraging the hybrid integration of silicon slot waveguides, plasmonic slot waveguides, and monolayer graphene is demonstrated. Exploiting the exceptional picosecond‐scale photogenerated carrier relaxation time of graphene, the response time of the activator is markedly reduced to ≈93.6 ps, establishing all‐optical activator as the fastest known in silicon photonics to knowledge. Moreover, the all‐optical nonlinear activator holds a low threshold power of 5.49 mW and a corresponding power consumption per activation of 0.51 pJ. Its feasibility and capability for use in ONNs, manifesting performance comparable with commonly used activation functions are experimentally confirmed. This breakthrough in speed and energy efficiency of all‐optical nonlinear activators opens the door to significant improvements in the performance and applicability of ONNs.

materials science, multidisciplinary,optics
Graphene/Silicon Heterojunction Enabling Reconfigurable Phase-Relevant Optical Activation Function

Hongtao Lin,Kun Liao,Xiaoyong Hu,Chuyu Zhong,Hui Ma,Maoliang Wei

DOI: https://doi.org/10.1364/cleo_si.2024.sm4g.1

2024-05-05

Abstract:Dual-functional graphene/silicon heterojunctions in microring resonators enabled highly reconfigurable optical activation function devices with phase activation. Special nonlinear functions are fed into a complex-valued ONN to challenge image recognition tasks, showing improved results.

Physics,Engineering
MXene‐Based Broadband Ultrafast Nonlinear Activator for Optical Computing

Zhan Yang,Weimin Tan,Tianju Zhang,Chenduan Chen,Zixin Wang,Yu Mao,Chenxi Ma,Qing Lin,Wanjun Bi,Fei Yu,Bo Yan,Jun Wang

DOI: https://doi.org/10.1002/adom.202200714

IF: 9

2022-06-30

Advanced Optical Materials

Abstract:This work presents an all‐optical nonlinear activator using a combination of microfiber and 2D materials. Theoretically, the nonlinear activator can compute 100 times faster than commonly used electronic computers. A set of criteria for judging nonlinear activators is proposed for the requirements of high speed and versatility of optical computing. Relying on the criteria, related research will greatly accelerate the development of nonlinear activators in the field of optical computing. Optical neural networks (ONNs) are particularly advantageous owing to their inherent parallelism and low energy consumption. However, one of the obstacles to the implementation of ONNs is the lack of optical nonlinearity. In this study, optical nonlinear activators for ONNs are prepared by combining Ti3C2Tx MXene with microfibers and their principles are verified. Activation functions obtained from experimental measurements are used to simulate multiclassification and super‐resolution reconstruction tasks with performance comparable to that of activation functions commonly used in computers. Four necessary criteria are proposed and validated for evaluating the performance of the nonlinear activator: recovery time, deviation from linearity, the activation function close to identity mapping, and reconfigurability of the configuration. Theoretically, the nonlinear activator can compute 100 times faster than commonly used electronic computers and can be used as a nonlinear activation unit for ONNs to help the integration of ONNs with artificial intelligence.

materials science, multidisciplinary,optics
All‐Optical Nonlinear Neuron Based on Metallic Quantum Wells

Junru Niu,Hua Shao,Yiming Feng,Bingtao Gao,Yiyun Zhang,Ying Li,Hongsheng Chen,Haoliang Qian

DOI: https://doi.org/10.1002/adom.202300223

IF: 9

2023-04-25

Advanced Optical Materials

Abstract:An effectual all‐optical nonlinear neuron is proposed for implementing the sigmoid activation function in optical neural networks. The device uses a double‐ring resonator and metallic quantum wells exhibiting a large and ultrafast Kerr nonlinearity. It achieves a sigmoid activation function with a high modulation depth of 18.78 dB and a rapid transition rate of 13 GW cm−2 per 3 dB. Optical nonlinear neuron, as an essential implementation method for optical interconnection, could greatly promote the development of optical neuron networks, which appears to be a promising alternative to electronic neural networks that are limited by computing speed and the end of Moore's law. However, restricted by weak nonlinearities, the modulation performance of optical nonlinear neuron networks is much lower compared to their electronic counterparts. Therefore, an effectual all‐optical nonlinear neuron is proposed by combining a double‐ring resonator with field‐enhancement and metallic quantum wells (MQWs) with a large Kerr nonlinear response. Based on the large and ultrafast Kerr nonlinearity provided by MQWs, the device can achieve a modulation extinction ratio of 18.78 dB and a transition rate of 13 GW cm−2 per 3 dB. The work may provide a new route for integrated, ultrafast, and high‐efficient optical nonlinear neurons for on‐chip neuron networks.

materials science, multidisciplinary,optics
High-quality activation function for applications in neuromorphic photonic chips realized using low-quality nonlinear optical resonators

Ivan A. Pshenichnyuk,Kamil R. Taziev,Sergey S. Kosolobov,Vladimir P. Drachev

2024-06-25

Abstract:Integrated optical devices that can realize a threshold filtration of signals are in demand in photonics. In particular, they play a key role in neuromorphic chips, acting as optical neurons. A list of requirements exist for thresholders to be practically applicable in this context. A value of the threshold, in general, should be independently tunable for each neuron. A sharpness of the corresponding step function should be also dynamically variable, to allow switching between deterministic and stochastic algorithms, for example, in optical Ising machines. Nonlinear ring resonators draw the attention of researchers in this field since they potentially can provide the required type of threshold behavior. Here we suggest the switching mechanism that implements the property of resonators to provide extremely sharp (with respect to the wavelength) $\pi$ phase shifts near the critical coupling regime. Adding a variable source of losses into the resonator, a well controlled broadening of the threshold can be achieved. Sharpness of the step in this case is independent on the quality factor of resonators and corresponding width of resonances. The nonlinear switching mechanism allows to use this feature to construct efficient optical neurons, that operate at small intensities. It also allows to use conventional materials like silicon with a relatively weak Kerr nonlinearity. The obtained results potentially lead the way to fast nonlinear Kerr effect based activation functions that can operate in continuous wave regime.

Optics,Emerging Technologies

All Optical Nonlinear Activation Function Based on Ge/Si Hybrid Micro-Ring Resonator

Low-threshold All-Optical Nonlinear Activation Function Based on a Ge/Si Hybrid Structure in a Microring Resonator

All-optical nonlinear activation function based on Germanium Silicon composite waveguide

Programmable All-optical Nonlinear Activation Functions Based on a Micro-ring Resonator with Phase-change Materials

All-Optical Nonlinear Activators Using Silicon-Based Waveguides and Microring Resonators for Photonic Neural Networks

Optical Nonlinear Activation Functions Based on MZI-Structure for Optical Neural Networks

Strong Nonlinear Optics In On-Chip Coupled Lithium Niobate Microdisk Photonic Molecules

All-Optical Nonlinear Activation Function Based on Germanium Silicon Hybrid Asymmetric Coupler

Highly-efficient second and third harmonic generation in a monocrystalline lithium niobate microresonator.

Programmable low-threshold optical nonlinear activation functions for photonic neural networks

A High-Stability All-Optical Nonlinear Activator for Optical Computing

Reconfigurable All-Optical Integrated Nonlinear Activator with Switchable Response Functions for Photonic Neural Networks

All-optical Nonlinear Activator for Optical Neural Network with Low Threshold Power and Ultrafast Response Speed

Optical Ring-Resonator Modulator Based on 0.8 Μm CMOS Technology

All-optical neural network nonlinear activation function based on the optical bistability within a micro-ring resonator

Cascadable Optical Nonlinear Activation Function Based on Ge-Si.

Ultrafast Silicon/Graphene Optical Nonlinear Activator for Neuromorphic Computing

Graphene/Silicon Heterojunction Enabling Reconfigurable Phase-Relevant Optical Activation Function

MXene‐Based Broadband Ultrafast Nonlinear Activator for Optical Computing

All‐Optical Nonlinear Neuron Based on Metallic Quantum Wells

High-quality activation function for applications in neuromorphic photonic chips realized using low-quality nonlinear optical resonators