Chip-to-chip optical multimode communication with universal mode processors DOI Creative Commons
Bo Wu, Wenkai Zhang, Hailong Zhou

и другие.

PhotoniX, Год журнала: 2023, Номер 4(1)

Опубликована: Окт. 27, 2023

Abstract The increasing amount of data exchange requires higher-capacity optical communication links. Mode division multiplexing (MDM) is considered as a promising technology to support the higher throughput. In an MDM system, mode generator and sorter are backbone. However, most current schemes lack programmability universality, which makes link susceptible crosstalk environmental disturbances. this paper, we propose intelligent multimode using universal processing (generation sorting) chips. processor consists programmable 4 × Mach Zehnder interferometer (MZI) network can be intelligently configured generate or sort both quasi linearly polarized (LP) modes orbital angular momentum (OAM) in any desired routing state. We experimentally establish chip-to-chip system. basis freely switched between four LP OAM modes. also demonstrate capability at rate 25 Gbit/s. proposed scheme shows significant advantages terms intelligence, resistance crosstalk, disturbances, fabrication errors, demonstrating that MZI-based reconfigurable chip has great potential long-distance systems.

Язык: Английский

Monolithic back-end-of-line integration of phase change materials into foundry-manufactured silicon photonics DOI Creative Commons

Maoliang Wei,

Kai Xu, Bo Tang

и другие.

Nature Communications, Год журнала: 2024, Номер 15(1)

Опубликована: Март 30, 2024

Язык: Английский

Процитировано

16
Fully nonlinear neuromorphic computing with linear wave scattering DOI Creative Commons
Clara C. Wanjura, Florian Marquardt

Nature Physics, Год журнала: 2024, Номер 20(9), С. 1434 - 1440

Опубликована: Июль 9, 2024

Abstract The increasing size of neural networks for deep learning applications and their energy consumption create a need alternative neuromorphic approaches, example, using optics. Current proposals implementations rely on physical nonlinearities or optoelectronic conversion to realize the required nonlinear activation function. However, there are considerable challenges with these approaches related power levels, control, efficiency delays. Here we present scheme system that relies linear wave scattering yet achieves processing high expressivity. key idea is encode input in parameters affect processes. Moreover, show gradients needed training can be directly measured experiments. We propose an implementation integrated photonics based racetrack resonators, which connectivity minimal number waveguide crossings. Our work introduces easily implementable approach computing widely applied existing state-of-the-art scalable platforms, such as optics, microwave electrical circuits.

Язык: Английский

Процитировано

16
Fully forward mode training for optical neural networks DOI Creative Commons
Zhiwei Xue, Tiankuang Zhou,

Zhihao Xu

и другие.

Nature, Год журнала: 2024, Номер 632(8024), С. 280 - 286

Опубликована: Авг. 7, 2024

Optical computing promises to improve the speed and energy efficiency of machine learning applications

Язык: Английский

Процитировано

15
Photonics for Neuromorphic Computing: Fundamentals, Devices, and Opportunities DOI
Renjie Li, Yuanhao Gong,

Hai Huang

и другие.

Advanced Materials, Год журнала: 2024, Номер unknown

Опубликована: Июнь 21, 2024

In the dynamic landscape of Artificial Intelligence (AI), two notable phenomena are becoming predominant: exponential growth large AI model sizes and explosion massive amount data. Meanwhile, scientific research such as quantum computing protein synthesis increasingly demand higher capacities. As Moore's Law approaches its terminus, there is an urgent need for alternative paradigms that satisfy this growing break through barrier von Neumann model. Neuromorphic computing, inspired by mechanism functionality human brains, uses physical artificial neurons to do computations drawing widespread attention. This review studies expansion optoelectronic devices on photonic integration platforms has led significant in where integrated circuits (PICs) have enabled ultrafast neural networks (ANN) with sub-nanosecond latencies, low heat dissipation, high parallelism. particular, various technologies employed neuromorphic accelerators, spanning from traditional optics PCSEL lasers examined. Lastly, it recognized existing encounter obstacles meeting peta-level speed energy efficiency threshold, potential new devices, fabrication, materials, drive innovation also explored. current challenges barriers cost, scalability, footprint, capacity resolved one-by-one, systems bound co-exist with, if not replace, conventional electronic computers transform foreseeable future.

Язык: Английский

Процитировано

13
Seven Bit Nonvolatile Electrically Programmable Photonics Based on Phase-Change Materials for Image Recognition DOI
Jian Xia, Tianci Wang, Zixuan Wang

и другие.

ACS Photonics, Год журнала: 2024, Номер 11(2), С. 723 - 730

Опубликована: Янв. 10, 2024

With the rapid development of Internet Things, how to efficiently store, transmit, and process massive amounts data has become a major challenge now. Optical neural networks based on nonvolatile phase change materials (PCMs) have breakthrough point due their zero static power consumption, low thermal crosstalk, large-scale, high efficiency. However, current photonic devices cannot meet multilevel requirements in neuromorphic computing limited storage capacity. Here, new way for increasing capacity is paved from perspective modulation crystallization kinetics PCMs. A more progressive transition amorphous crystalline states achieved through grain-refinement phenomenon induced by nitrogen (N) element doping Ge2Sb2Te5 (GST), giving precise access multibit states. By integrating N-doped (N-GST) with waveguide, high-capacity device enabling over 7 bits (∼222 levels) first time. The increased nearly times compared state-of-the-art (∼32 levels). An optical convolutional network successfully established MINIST handwritten digit recognition task mapping synapse weight multiple levels, accuracy up 96.5% achieved. Our work provides strategy integrated demonstrates enormous application potential field large-scale networks.

Язык: Английский

Процитировано

12
Inverse design of compact nonvolatile reconfigurable silicon photonic devices with phase-change materials DOI Creative Commons

Maoliang Wei,

Xiaobin Lin,

Kai Xu

и другие.

Nanophotonics, Год журнала: 2024, Номер 13(12), С. 2183 - 2192

Опубликована: Янв. 12, 2024

In the development of silicon photonics, continued downsizing photonic integrated circuits will further increase integration density, which augments functionality chips. Compared with traditional design method, inverse presents a novel approach for achieving compact devices. However, compact, reconfigurable devices that employs modulation method exemplified by thermo-optic effect poses significant challenge due to weak capability. Low-loss phase change materials (PCMs) Sb

Язык: Английский

Процитировано

12
A Reconfigurable Bipolar Image Sensor for High-Efficiency Dynamic Vision Recognition DOI
Jia Yang, Y. Cai, Feng Wang

и другие.

Nano Letters, Год журнала: 2024, Номер 24(19), С. 5862 - 5869

Опубликована: Май 6, 2024

Dynamic vision perception and processing (DVPP) is in high demand by booming edge artificial intelligence. However, existing imaging systems suffer from low efficiency or compatibility with advanced machine techniques. Here, we propose a reconfigurable bipolar image sensor (RBIS) for in-sensor DVPP based on two-dimensional WSe2/GeSe heterostructure device. Owing to the gate-tunable reversible built-in electric field, its photoresponse shows bipolarity as being positive negative. High-efficiency incorporating front-end RBIS back-end CNN then demonstrated. It recognition accuracy of over 94.9% derived DVS128 data set requires much fewer neural network parameters than that without RBIS. Moreover, demonstrate an optimized device vertically stacked structure stable nonvolatile bipolarity, which enables more efficient hardware. Our work demonstrates potential fabricating devices simple structure, efficiency, outputs compatible algorithms.

Язык: Английский

Процитировано

12
Integrated photonic neuromorphic computing: opportunities and challenges DOI
Nikolaos Farmakidis, Bowei Dong, Harish Bhaskaran

и другие.

Nature Reviews Electrical Engineering, Год журнала: 2024, Номер 1(6), С. 358 - 373

Опубликована: Июнь 6, 2024

Язык: Английский

Процитировано

12
Direct electromagnetic information processing with planar diffractive neural network DOI Creative Commons

Ze Gu,

Qian Ma, Xinxin Gao

и другие.

Science Advances, Год журнала: 2024, Номер 10(29)

Опубликована: Июль 19, 2024

Diffractive neural network in electromagnetic wave-driven system has attracted great attention due to its ultrahigh parallel computing capability and energy efficiency. However, recent networks based on the diffractive framework still face bottlenecks of misalignment relatively large size limiting their further applications. Here, we propose a planar (pla-NN) with highly integrated conformal architecture achieve direct signal processing microwave frequency. On basis printed circuit fabrication process, could be effectively circumvented while enabling flexible extension for multiple stacking designs. We first conduct validation fashion-MNIST dataset experimentally build up using proposed recognition different geometry structures space. envision that presented architecture, once combined advanced dynamic maneuvering techniques topology, would exhibit unlimited potentials areas high-performance computing, wireless sensing, wearable electronics.

Язык: Английский

Процитировано

12
Single-chip photonic deep neural network with forward-only training DOI
Saumil Bandyopadhyay, Alexander Sludds, Stefan Krastanov

и другие.

Nature Photonics, Год журнала: 2024, Номер unknown

Опубликована: Дек. 2, 2024

Язык: Английский

Процитировано

12