Bidirectional Invisible Photoresponse Implemented in a Traps Matrix-Combination toward Fully Optical Artificial Synapses

Abstract

Fully optical artificial synapses are crucial hardware for neuromorphic computing, which is very promising to address the future large-scale computing capacity problem. The key characteristic required in a semiconductor device to emulate synaptic potentiation and depression in a fully optical artificial synapse is the bidirectional photoresponse. This work integrates wide-band-gap TiO₂ polycrystals and narrow-band-gap PbS quantum dots into a graphene transistor simultaneously, providing the device with both near-ultraviolet and near-infrared photoresponses through the photogating effect. Moreover, the TiO₂ serves as a hole-trapping matrix and the PbS as an electron-trapping matrix, which impose opposite effects to the device after photoexcitation, resulting in a photoresponse in the opposite polarity. As a result, the device demonstrates a wavelength-dependent bidirectional photoresponse, which enables it to be utilized as a fully optical artificial synapse. By using near-ultraviolet or near-infrared lights as stimuli, the device successfully mimics synaptic plasticity, including synaptic potentiation/depression, paired-pulse facilitation, and spike-rating-dependent plasticity, as well as the human brain-like transition of short-term memory and long-term memory and learning-experience behavior. This work validates the methodology of combining different trap matrices to achieve the bidirectional photoresponse, which can significantly inspire future research in fully optical artificial synapses.