Optoelectronic Synapse Based on Te/SnS2 Heterostructure with Integrated Sensing-Memory-Computing for Neuromorphic Visual System

Abstract

The rapid development of artificial intelligence greatly stimulates the development of advanced multifunctional optoelectronic devices that integrate sensing, memory, and computing functions into one device. Herein, a 1D-2D Te/SnS₂ mixed-dimensional heterostructure device is constructed to demonstrate the multifunctional optoelectronic synapse for all-in-one neuromorphic visual systems. Owing to the formation of type-II p-n junction, the device achieves large rectification ratio of 10³ and current on/off ratio of 10⁵. As a photodetector, the device exhibits prominent gate tunable photoresponse with high detectivity (1.14 × 10¹¹ Jones), responsivity (19.0 A W⁻¹), external quantum efficiency (4.44 × 10⁷%). Combing light and gate voltage as inputs, the device realizes an optoelectronic logic gate “AND”, indicating its information processing ability. Furthermore, the device also exhibits superior nonvolatility and multi-bit optoelectronic programmable characteristics. As results, the device can stimulate the synaptic plasticity, including short-term/long-term plasticity and paired-pulse facilitation. By coupling light and electrical signals, the device realizes Pavlovian associative learning and retina-like light adaptation. Further applying the device to the artificial neural network system for handwritten digit recognition achieves a high accuracy of 94.4%. This work demonstrates the great potential of our device for neuromorphic visual perception and provides new insight for developing next-generation integrated optoelectronic systems.