A light-stimulated synaptic device based on graphene hybrid phototransistor

TLDR

Neuromorphic chips emulate biological brains and are increasingly used for sensory data processing, yet they remain largely electronic, limiting access to the superior computing features of photons for vision tasks. The study introduces a light‑stimulated synaptic device built from a graphene‑carbon nanotube hybrid phototransistor. It uses this hybrid phototransistor to convert optical stimuli into synaptic responses. The device exhibits neuron‑like optical responses with tunable short‑ and long‑term plasticity, enabling spatiotemporal processing of optical spikes and offering a silicon‑compatible, multifunctional photosensitive synapse that expands neuromorphic systems’ complexity and functionality.

Abstract

Neuromorphic chips refer to an unconventional computing architecture that is modelled on biological brains. They are increasingly employed for processing sensory data for machine vision, context cognition, and decision making. Despite rapid advances, neuromorphic computing has remained largely an electronic technology, making it a challenge to access the superior computing features provided by photons, or to directly process vision data that has increasing importance to artificial intelligence. Here we report a novel light-stimulated synaptic device based on a graphene-carbon nanotube hybrid phototransistor. Significantly, the device can respond to optical stimuli in a highly neuron-like fashion and exhibits flexible tuning of both short- and long-term plasticity. These features combined with the spatiotemporal processability make our device a capable counterpart to today's electrically-driven artificial synapses, with superior reconfigurable capabilities. In addition, our device allows for generic optical spike processing, which provides a foundation for more sophisticated computing. The silicon-compatible, multifunctional photosensitive synapse opens up a new opportunity for neural networks enabled by photonics and extends current neuromorphic systems in terms of system complexities and functionalities.