Light-induced perovskite dynamic transformation enabling a photodetector to mimic a neuromorphic vision sensing system

Abstract

As an artificial perception system, a neuromorphic vision sensing system can imitate the complex image sensing and processing functions of a human visual neural network. In order to simulate the nervous system, people often need to integrate multiple functional components such as sensors and processors into a sophisticated circuit system or prepare a multi-functional optoelectronic device through a complex and tedious process. The high cost and complicated preparation process will undoubtedly limit the future development of a neuromorphic vision system. Herein, we show a perovskite-based transformational photodetector that can mimic image sensing and preprocessing functions in a vision system. The pristine photodetector is prepared by a simple one-step method and has an on–off ratio of ∼10⁷, a low noise current of ∼4.27 × 10⁻¹³ A Hz^−0.5, a −3 dB bandwidth of 185 kHz, and a specific detectivity of 3.48 × 10¹¹ Jones. Then, we utilize the reversible interaction between NH₃ and MAPbI₃, and adjust irradiation to induce dynamic transformation of the perovskite, so that the photocurrent of the photodetector exhibits enhanced behavior of short-term synaptic plasticity over time. The photocurrent with 3.36 mW light intensity shows ∼225% enhancement after 60 pulses, while the photocurrent with 0.48 mW light intensity shows ∼178% enhancement, which can enhance the contrast of the transmitted image and realize the image transmission and preprocessing functions of a neuromorphic vision sensing system. We believe that this work will provide new potential for the future application of photodetectors in neuromorphic vision sensing.