Research On High-Speed Reservoir Computing Based On Short-Cavity Semiconductor Lasers

In today’s era of rapid information development,the Photonics Reservoir Computing(PRC)system is playing an increasingly important role in dynamic information processing.It has been widely used in tasks such as time series prediction,pattern recognition and classification.Higher requirements are placed on the information processing speed,which means that the information processing rate of the PRC system needs to be further improved.Therefore,it is particularly important to find a high-speed PRC system.Since the short photon lifetime in the laser leads to a faster dynamic response,it greatly affects the information processing rate.Meanwhile,the laser photon lifetime is closely related to the internal cavity length,and it can be obtained from the laser rate equation(Lang-Kobayashi)that changing the internal cavity length will cause changes in the feedback strength,injection strength,threshold current and other parameters required for the laser to enter each dynamic state,which in turn affects the parameter space required for the PRC system to exhibit high performance.Existing studies have confirmed that in the PRC system,when the virtual node interval is 0.2 times the relaxation oscillation frequency,the system can obtain better performance.This indicates that the high relaxation oscillation frequency contributes to obtaining a smaller node interval and thus achieving a higher information processing rate.The high relaxation oscillation frequency of short-cavity semiconductor lasers can meet the requirements of high-speed information processing.Therefore,the influence of the internal cavity length of the semiconductor laser on the performance of the PRC system and the corresponding parameter space is studied in this paper.In addition,we found that the circular-side hexagonal resonator(CSHR)microlaser has the characteristics of ultra-short photon lifetime,small size,and high relaxation oscillation frequency(about 11 GHz)under high bias current,which can be well used in PRC high-speed information processing of the system.The main work of this paper are as follows:(1)Numerical simulation of the influence of the internal cavity length on the performance of a distributed feedback laser-based PRC system.The effects of laser feedback strength,injection strength,frequency detuning,and the number of virtual nodes on the performance of the RC system are further analyzed.The results show that the short cavity length can significantly improve the information processing rate.(2)The influence of the internal cavity length on the parameter range is studied,and it is found that when the system performs the Santa-Fe chaotic time series prediction task,under the internal cavity length of 128 μm,the parameter space that makes the system show good prediction performance is nearly 20%higher than that of the cavity length of 600 μm,and the 10th order nonlinear autoregressive moving average task is improved by about 40%.(3)A reservoir computing system based on the optical feedback CSHR microlaser is constructed.Numerical simulations are performed to find the optimal operating point of the system by adjusting key parameters(such as feedback strength,injection strength,frequency detuning,number of virtual nodes,etc.).The results show that the system can achieve an information processing rate of 10 Gbps in large parameter space,and is expected to achieve an information processing rate of 25 Gbps with the development of technology.