Chalcogenide-Based Brain-Inspired Photo-Synapses for Neuromorphic Vision Sensor: An Experimental and Theoretical Study.

2D chalcogenide-based memristors have the potential to be used in artificial biological visual systems since their synaptic behavior can be optically and electrically modulated. Furthermore, 2D van der Waals materials such as SnS can be used to integrate multifunctional optoelectronic devices by employing a rational design. Here, the simulation of a human biological visual system is reported by using multifunctional optoelectronic synaptic devices. First-principles-based DFT calculations show that SnS is a semiconductor with a bandgap of 2.47 eV. The electrical and optical inputs can be controlled to perform memory and logic functions consistent with those in the brain's visual cortex. In particular, an SnS memristor shows outstanding letter recognition and image memory as a function of wavelength-sensitive responses, mimicking the biological retina. When the SnS retina device is employed as the processing core, machine vision simulations indicate an excellent accuracy of 98.51% for Modified National Institute of Standards and Technology (MNIST) datasets. Owing to the excellent photosensitivity of SnS, these devices can operate at an ultralow voltage of 0.1 V, with an energy consumption of 0.345 nJ per event. Notably, the SnS-based photo-synaptic device can perform the OR and AND logical operations by varying the optical input wavelengths. These findings are expected to pave the way for the development of advanced robotic vision systems with innovative neuromorphic computing capabilities. The integration of an exfoliated 2D SnS-based optoelectronic memristor with a hybrid AI framework is the key innovative motivation of this study, wherein empirically determined synaptic characteristics are employed to guide the training of a neural network model for image recognition tasks.