Abstract

Photonic neuromorphic engineering is an emerging field that aims to combine the merits of non-von-Neumann architectures with the high bandwidth and parallelism of the optical domain [1]. Especially, laser-neurons can enhance these features, by exhibiting a wide pallet of dynamics. Recently, we demonstrated, an optically triggered inhibitory multi-section laser neuron, where nanosecond responses were linked to quantum-dot (QD) waveband transitions [2]. In our approach, the inhibitory neuron is activated through ground-state (GS) input (considered excitation signal), but contrary to conventional excitatory neurons, generates excited state (ES) output, which we link to an inhibition signal. In this work, we utilize a numerical model that targets a simpler structure (single-section laser). Results confirm the role of intra-band transport time to inhibitory operation, whereas dynamic analysis confirmed integrate and fire functionality alongside picosecond spikes, thus providing increased temporal resolution and simpler implementation compared to electro-optic or multi-section devices [2].

© 2019 IEEE