Advancing optical spiking neural networks with exciton-polaritons

Abstract In recent years, exciton–polariton microcavities have gained recognition as a highly promising platform for optical neural networks. Furthermore, recent breakthroughs have established exciton-polaritons as a compelling alternative for enabling optical spiking neural networks (SNNs). The rich repertoire of physical effects, such as polariton condensation, strong energy blueshift, lateral confinement, or particle propagation within optical microcavities, can be exploited to significantly expand the range of neuronal functionalities that can be mimicked in optical spiking neurons while maintaining relatively low energy consumption. This versatility presents the potential for breakthroughs in optical neural computing, overcoming limitations observed in earlier technologies, like the lack of cascadability. In this perspective, we summarize the current state-of-the-art in optical SNN and review the key nonlinear phenomena inherent to exciton–polariton microcavities to demonstrate how they can be leveraged to enhance optical spiking neuron architectures. We further provide a forward-looking assessment of how polariton-based platforms may surpass the current capabilities of optical systems, unlocking new perspectives for developing advanced spiking neurons. We emphasize the capability of polaritons to process picosecond pulses and sustain the nodes interconnectivity in lateral and longitudinal directions, towards enhanced network scaling. This paper aims to highlight the untapped potential of polariton technology in pushing the boundaries of optical neuromorphic computing.