Semiconductor Fano lasers boost neuromorphic computing


Reservoir computing (RC) has proved effective in solving complex practical time-dependent problems. Typically, RC systems are high-speed and lightweight machine learning models capable of processing different types of information after training, achieved through the nonlinear mapping between the input and output of the system layers. The recent development of a novel RC scheme based on a time-delayed feedback mechanism paved the way for further advancement of exiting RC systems. One of the key advantages of this new configuration was its ability to significantly reduce the reservoir hardware operation difficulty by replacing the real network of the original RC system with a virtual network. Additionally, it preserved the vital characteristics of the reservoirs, such as fading memory ability and the high state space dimension. To this end, there has been growing research interest in time-delayed RC systems, focusing on expanding their applications.

Generally, the nonlinear response of the reservoir is usually realized through different passive nonlinearity techniques, including spintronic oscillators, semiconductor lasers (SLs) and optical amplifiers. Among them, RC configurations consisting of semiconductor lasers and photonic integrated circuits are gaining interest in the realization of miniaturized and energy-efficient RCs. However, when conventional semiconductor lasers are applied in time-delayed RC systems, their inherent sensitivity towards external feedback may lead to dynamic instability of the system, thus degrading its performance. To address the dynamic instability problem, the SL output has been stabilized by injecting external signals. Although it is a promising solution for enhancing the performance of RCs, it introduces additional parameters that further complicate the system design and integration.

Herein, Dr. Yu Huang, Professor Pei Zhou, Mr Yigong Yang and Professor Nianqiang Li from Soochow University proposed a high-speed photonic RC system based on compact Fano laser (FL) coupled with optical feedback under electrical modulation. The most fundamental feature of the Fano laser is the realization of one or all of the laser mirrors via Fano interference between the waveguide mode and the localized resonance of the nanocavity. The nonlinear dynamic characteristics of the FL were explored in detail. The performance and applicability of the proposed RC system were numerically validated and compared to that of the conventional RC systems. Their work is currently published in the research journal, Optics Letters.

Results showed that FL exhibited a wider dynamic steady-state region than most conventional laser types because it is relatively insensitive to external feedback. Its advantages over conventional lasers include its extremely small footprint, ultrashort photon lifetime a faster dynamic, wider RC implementation range and improved prediction and classification performance with an information processing rate of about 10 Gbps. Two separate regions of the excellent RC performances corresponding to two separate scenarios of the FL’s dynamic steady-state were observed. Consequently, improved RC performance with respect to the feedback phase was reported in one of the steady-state regions, especially where the laser was not destabilized for low external reflectivity. Furthermore, the results suggested that Fano laser did not require extra stabilization effects like that imposed by injected signals.

In summary, a high-speed RC system based on FL with optical feedback was numerically demonstrated under electrical modulation. The proposed RC system exhibited superior characteristics that were beneficial in improving its prediction and classification performances. The highest RC performance was obtained in the region characterized by the suppression of the coherent collapse, and it was dependent on the feedback phase. Considering its advantages, Professor Nianqiang Li, a statement to Advances in Engineering, stated that FL is a promising candidate for integrated time-delayed RC, and the study results would contribute to the understanding of its fundamental mechanism.

Semiconductor Fano lasers boost neuromorphic computing - Advances in Engineering

About the author

Nianqiang Li is now working as a full professor at School of Optoelectronic Science and Engineering, Soochow University, Suzhou, China. He received the B.S. degree in communication engineering and the Ph.D. degree in optoelectronics from Southwest Jiaotong University, China, in 2008 and 2016, respectively. From 2013 to 2014, he was a Visiting Scholar with the Georgia Institute of Technology (Georgia Tech), USA. From 2016 to 2018, he was with the School of Computer Science and Electronic Engineering, University of Essex, as a Postdoctoral Researcher focusing on a collaborative EPSRC funded project in U.K. From May, 2018 to December 2018, he was with the School of Electrical Engineering and Computer Science, University of Ottawa, Canada, working on microwave photonics. He has authored or co-authored more than 80 peer reviewed journal papers. His current research mainly focuses on the area of the laser dynamics, chaos-based communication and random number generation, and microwave photonics.


Huang, Y., Zhou, P., Yang, Y., & Li, N. (2021). High-speed photonic reservoir computer based on a delayed Fano laser under electrical modulationOptics Letters, 46(24), 6035-6038.

Go To Optics Letters

Check Also

28.3% Efficient III–V Tandem Solar Cells Fabricated Using a Triple-Chamber Hydride Vapor Phase Epitaxy System - Advances in Engineering

28.3% Efficient III–V Tandem Solar Cells Fabricated Using a Triple-Chamber Hydride Vapor Phase Epitaxy System