Object Motion Detection\nEnabled by Reconfigurable\nNeuromorphic Vision Sensor under Ferroelectric Modulation

Increasing the\ndemand for object motion detection (OMD) requires\nshifts of reducing redundancy, heightened power efficiency, and precise\nprogramming capabilities to ensure consistency and accuracy. Drawing\ninspiration from object motion-sensitive ganglion cells, we propose\nan OMD vision sensor with a simple device structure of a WSe<sub>2</sub> homojunction modulated by a ferroelectric copolymer. Under optical\nmode and intermediate ferroelectric modulation, the vision sensor\ncan generate progressive and bidirectional photocurrents with discrete\nmultistates under zero power consumption. This design enables reconfigurable\ndevices to emulate long-term potentiation and depression for synaptic\nweights updating, which exhibit 82 states (more than 6 bits) with\na uniform step of 6 pA. Such OMD devices also demonstrate nonvolatility,\nreversibility, symmetry, and ultrahigh linearity, achieving a fitted <i>R</i><sup>2</sup> of 0.999 and nonlinearity values of 0.01/–0.01.\nThus, a vision sensor could implement motion detection by sensing\nonly dynamic information based on the brightness difference between\nframes, while eliminating redundant data from static scenes. Additionally,\nthe neural network utilizing a linear result can recognize the essential\nmoving information with a high recognition accuracy of 96.8%. We also\npresent the scalable potential via a uniform 3 × 3 neuromorphic\nvision sensor array. Our work offers a platform to achieve motion\ndetection based on controllable and energy-efficient ferroelectric\nprogrammability.