Integrated silicon-photonic reservoir computing with an on-chip Si3N4 microcavity

We propose an integrated silicon photonic reservoir computing (RC) scheme using a silicon nitride (Si3N4) microcavity hybrid coupled with a semiconductor laser chip. The device is more than 10 times smaller than existing analogs, owing to the highly integrated Si3N4 microcavity. The device is chaoti...

Full description

Saved in:
Bibliographic Details
Published inOptics communications Vol. 591; p. 132133
Main Authors Zhao, Lehan, Bai, Qingsong, Wu, Mingjie, Ran, Chongchong, Zeng, Yongcan, Deng, Changmao, Wu, Zhengmao, Yang, Wenyan, Wu, Jiagui
Format Journal Article
LanguageEnglish
Published Elsevier B.V 01.10.2025
Online AccessGet full text
ISSN0030-4018
DOI10.1016/j.optcom.2025.132133

Cover

More Information
Summary:We propose an integrated silicon photonic reservoir computing (RC) scheme using a silicon nitride (Si3N4) microcavity hybrid coupled with a semiconductor laser chip. The device is more than 10 times smaller than existing analogs, owing to the highly integrated Si3N4 microcavity. The device is chaotic in a radio-frequency range exceeding 20 GHz, with subnanosecond-scale nonlinear temporal fluctuations. An RC system is constructed and tested on time-series prediction and handwritten-digit recognition tasks. The normalized mean square error is 0.014 on the Santa Fe test with 109 samples/s, while the error rate on the MNIST handwritten-digit dataset is approximately 5.3 %. These results are promising for highly compact RC schemes with complementary metal-oxide semiconductor compatibility.
ISSN:0030-4018
DOI:10.1016/j.optcom.2025.132133