AI-assisted on-chip nanophotonic convolver based on silicon metasurface

Liao, Kun; Gan, Tianyi; Hu, Xiaoyong; Gong, Qihuang

doi:10.1515/nanoph-2020-0069

Cited by 40 publications

(21 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In the future, the free-space optical components can be replaced by the on-chip lenses which have been achieved in the nanophotonic convolver [39], on-chip deep learning systems [40][41][42][43] and integrated tunable varifocal lenses performing like error backpropagation in neural networks [44]. Once combined with on-chip lenses, our nonlinear JTC will have high energy efficiency, compact volume and high speed, and therefore promotes the potential convolution-related applications in many aspects, including image classification with large and deep convolutional neural networks, speech recognition and translation with the combination of convolutional neural networks and deep recurrent neural networks, autonomous driving by mapping raw pixels with convolutional neural networks, inverse design (like nanophotonic / plasmonic structures, self-adaptive microwave cloak) problem solving based on convolutional neural networks, and robotic manipulation with deep reinforcement learning.…”

Section: Discussionmentioning

confidence: 99%

Nonlinear Optical Joint Transform Correlator for Low Latency Convolution Operations

George¹,

Solyanik‐Gorgone²,

Yang³

et al. 2022

Preprint

View full text Add to dashboard Cite

Convolutions are one of the most relevant operations in artificial intelligence (AI) systems. High computational complexity scaling poses significant challenges, especially in fast-responding networkedge AI applications. Fortunately, the Convolution Theorem can be executed 'on-the-fly' in the optical domain via a joint transform correlator (JTC) offering to fundamentally reduce the computational complexity. Nonetheless, the iterative two-step process of a classical JTC renders them unpractical. Here we introduce a novel implementation of an optical convolution-processor capable of near-zero latency by utilizing all-optical nonlinearity inside a JTC, thus minimizing electronic signal or conversion delay. Fundamentally we show how this nonlinear auto-correlator enables reducing the high O(n 4 ) scaling complexity of processing two-dimensional data to only O(n 2 ). Moreover, this optical JTC processes millions of channels in time-parallel, ideal for large-matrix machine learning tasks. Exemplary utilizing the nonlinear process of four-wave mixing, we show light processing performing a full convolution that is temporally limited only by geometric features of the lens and the nonlinear material's response time. We further discuss that the all-optical nonlinearity exhibits gain in excess of > 10 3 when enhanced by slow-light effects such as epsilon-near-zero. Such novel implementation for a machine learning accelerator featuring low-latency and non-iterative massive data parallelism enabled by fundamental reduced complexity scaling bears significant promise for network-edge, and cloud AI systems.

show abstract

Section: Discussionmentioning

confidence: 99%

Nonlinear Optical Joint Transform Correlator for Low Latency Convolution Operations

George¹,

Solyanik‐Gorgone²,

Yang³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…The limited resolution of data sampling is constrained by minimum waveguide spacing between the output ports to avoid coupling between the waveguides. To further improve the resolution, the output waveguides can fan out, 32 as opposed to being parallel to each other (as shown in Figure 1d). However, this method will lead to a very large device area.…”

Section: ■ Discussionmentioning

confidence: 99%

1D Self-Healing Beams in Integrated Silicon Photonics

et al. 2021

View full text Add to dashboard Cite

Since the first experimental observation of optical Airy beams, various applications ranging from particle and cell micromanipulation to laser micromachining have exploited their nondiffracting and accelerating properties. The later discovery that Airy beams can self-heal after being blocked by an obstacle further proved their robustness to propagate under scattering and disordered environment. Here, we report the generation of Airy beams on an integrated silicon photonic chip and demonstrate that the on-chip 1D Airy beams preserve the same properties as the 2D beams. The 1D meta-optics used to create the Airy beam has the size of only 3µ𝑚 × 16µ𝑚, at least three orders of magnitude smaller than the conventional optic. The onchip self-healing beams demonstrated here could potentially enable diffraction-free light routing for on-chip optical networks and high-precision micromanipulation of bio-molecules on an integrated photonic chip.

show abstract

“…At the same time, a desired phase profile can be obtained by etching air slots of different lengths in a selected substrate . The desired profile for a broad-band lens can be calculated by an analysis or by using the inverse design to determine the length of each slot. , The main difference is that in our design we set broad boundaries for parameter space using the objective-first algorithm and then transform the output to a hole-based structure. Basically, the main purpose of both methods is to approach the desired phase with the help of the selected method.…”

Section: Sampling Inhomogeneous Medium With Bruggeman Effective Mediu...mentioning

confidence: 99%

A Broad-Band Achromatic Polarization-Insensitive In-Plane Lens with High Focusing Efficiency

et al. 2021

View full text Add to dashboard Cite

Until now, many pioneering studies have been conducted on multiwavelength achromatic lens designs that can eliminate chromatic aberrations at various wavelengths. In these studies, the working bandwidth is still not wide enough for practical applications, and an effective achromatic design is achieved with a long design cycle that determines critical elements of the unit cell. To solve the limitations of the unit cell method, we use inverse design to simultaneously explore structures with broad bandwidth and high efficiency. In this study, we show that it is possible to obtain an achromatic cylindrical lens that can focus transverse electric and transverse magnetic polarization between 1300 and 1750 nm using the objective-first algorithm. We implement the generalized Bruggeman effective medium theory to binarize the lens with randomly varying index profiles while maintaining the optical performance of the cylindrical inverse-designed lens. The binarized lens is produced via three-dimensional printing and tested in a microwave regime, exhibiting high bandwidth operation and high focusing efficiency (average 62%).

show abstract

AI-assisted on-chip nanophotonic convolver based on silicon metasurface

Cited by 40 publications

References 29 publications

Nonlinear Optical Joint Transform Correlator for Low Latency Convolution Operations

Nonlinear Optical Joint Transform Correlator for Low Latency Convolution Operations

1D Self-Healing Beams in Integrated Silicon Photonics

A Broad-Band Achromatic Polarization-Insensitive In-Plane Lens with High Focusing Efficiency

Contact Info

Product

Resources

About