Time-delayed reservoir computing based on a dual-waveband quantum-dot spin polarized vertical cavity surface-emitting laser

Abstract

In this work, we present numerical results concerning a time-delayed reservoir computing scheme, where its single nonlinear node, is a quantum-dot spin polarized vertical cavity surface-emitting laser (QD s-VCSEL). The proposed photonic neuromorphic scheme exploits the complex temporal dynamics of multiple energy states present in quantum dot materials and uses emission from two discrete wavebands and two polarization states, so as to enhance computational efficiency. The benchmark task used for this architecture, is the equalization of a distorted 25Gbaud PAM-4 signal after 50Km of transmission at 1550 nm. Results confirm that although typical ground-state emitting quantum-dot nodes offer limited performance, due to bandwidth limitations; by exploiting dual emission, we achieved a one-hundred-fold improvement in bit-error rate. This performance boost can pave the way for the infiltration of quantum-dot based devices in high-speed demanding neuromorphic driven applications.