A Three-Dimensional Neuromorphic Photosensor Array for Nonvolatile In-Sensor Computing

In-sensorcomputing hardware based on emerging reconfigurable photosensorscan effectively reduce redundant data and decrease power consumption,which can greatly promote the evolution of machine vision. However,because of the complex device structures and low integration abilities,the common architectures mainly lie in two dimensions, resulting inlow time and area efficiencies. Here we propose a three-dimensional(3D) neuromorphic photosensor array for parallel in-sensor image processing.It is constructed on a vertical Graphite/CuInP2S6/Graphite photosensor unit, where the directional Cu+ ionmigrations after voltage pulse programming enable a reconfigurablephotovoltaic effect and an in-sensor computing capability. With amemristor-like device structure, van der Waals interfaces, and a highuniformity with a low crosstalk problem, a 10 x 10 array is fabricatedfor intelligent image recognition. Furthermore, using a verticallystacked 3D 3 x 3 x 3 array, we demonstrate an in-sensorconvolution strategy with high time and area efficiencies.