Spiking behaviour in laterally-coupled pairs of VCSELs with applications in neuromorphic photonics

Hejda, Matĕj and Vaughan, Martin and Henning, Ian and Al-Seyab, Rihab and Hurtado, Antonio and Adams, Mike (2023) Spiking behaviour in laterally-coupled pairs of VCSELs with applications in neuromorphic photonics. IEEE Journal of Selected Topics in Quantum Electronics, 29 (2). pp. 1-10. ISSN 1077-260X (https://doi.org/10.1109/jstqe.2022.3218950)

[thumbnail of Hejda-etal-IEEEJSTQE-2022-Spiking-behaviour-in-laterally-coupled-pairs-of-VCSELs-with-applications-in-neuromorphic-photonics]
Preview
 Text. Filename: Hejda_etal_IEEEJSTQE_2022_Spiking_behaviour_in_laterally_coupled_pairs_of_VCSELs_with_applications_in_neuromorphic_photonics.pdf
Final Published Version
License: Creative Commons Attribution 4.0 logo
Download (5MB)| Preview

Abstract

We report a theoretical study on laterally-coupled pairs of vertical-cavity surface-emitting lasers (VCSELs) operated under conditions that generate or suppress high-speed optical spiking regimes, and show their potential in exemplar functionalities for use in photonic neuromorphic computing systems. The VCSEL numerical analysis is based on a system of five coupled mode equations, which, for the case of weak coupling, are reduced to a set of three equations that predict the saddle-node stability boundary in terms of device parameters and operating conditions. These results guide numerical simulation to demonstrate multiple neuron-like dynamics, including single- and multiple-spike emission, spiking inhibition, and rebound spiking directly in the optical domain. Importantly, these behaviours are obtained at sub-nanosecond rates, hence multiple orders of magnitude faster than the millisecond timescales of biological neurons. The mechanisms responsible are explained by reference to appropriate phase portraits. The coupled VCSELs model is then used for demonstration of high-speed, all-optical digital-to-spiking encoding and for representation of digital image data using rate-coded spike trains.

CORE (COnnecting REpositories)