Journal Article

All‐in‐one perovskite memristor with tunable photoresponsivity

Zhang, Hai‐Nan, Wang, Yu‐Hao, Chai, Yang, Lu, Tian, Peng, Jia‐Li, Zhu, Zheng‐Qiang, Zhou, Peng, Ren, Tian‐Ling, Tan, Xi‐Chao, Qin, Ken, Dun, Guan‐Hua, Li, Yan, Wu, Xiaoyu, He, Yi‐Chu, Li, Yuan‐Yuan, Wu, Fan, Wang, Ze‐Shu, Xie, Dan, Zhang, Jia‐He, Deng, Ning‐Qin, Lanza, Mario, Tian, He, Zhao, Yu‐Han, Zhao, Xiao‐Tong, Geng, Xiang‐Shun, Liu, Xian‐Zhu, Hu, Weida, Tian, Shi‐Wei, Wu, Xing, Yang, Yi

2025

Overview

Photoelectric memristors have shown great potential for future machine visions, via integrating sensing, memory, and computing (namely “all‐in‐one”) functions in a single device. However, their hard‐to‐tune photoresponse behavior necessitates extra function modules for signal encoding and modality conversion, impeding such integration. Here, we report an all‐in‐one memristor with Cs2AgBiBr6 perovskite, where the Br vacancy doping‐endowed tunable energy band enables tunable photoresponsivity (TPR) behavior. As a result, the memristor showed a large tunable ratio of 35.9 dB, while its photoresponsivity presented a maximum of 2.7 × 103 mA W−1 and a long‐term memory behavior with over 104 s, making it suitable for realizing all‐in‐one processing tasks. By mapping the algorithm parameters onto the photoresponsivity, we successfully performed both recognition and processing tasks based on the TPR memristor array. Remarkably, compared with conventional complementary metal–oxide–semiconductor counterparts, our demonstrations provided comparable performance but had ~133‐fold and ~299‐fold reductions in energy consumption, respectively. Our work could facilitate the development of all‐in‐one smart devices for next‐generation machine visions. Background: Photoelectric memristors are integral for next‐gen machine vision systems, aiming for an “all‐in‐one” device with sensing, memory, and computing capabilities. However, their photoresponse behavior has been challenging to control. Research: We developed a Cs2AgBiBr6 perovskite‐based memristor with tunable photoresponsivity through Br vacancy doping, addressing the integration challenge. Findings: The memristor achieved a 35.9 dB tunable ratio, a maximum photoresponsivity of 2.7 × 103 mA W−1, and a memory retention over 104 s, enabling all‐in‐one processing. Impact: By mapping algorithm parameters to the tunable photoresponsivity, recognition and processing tasks were successfully executed. The device demonstrated comparable performance to CMOS technology but with ~133‐fold to ~299‐fold lower energy consumption, propelling the development of energy‐efficient, all‐in‐one smart devices for advanced machine vision.

• Publisher: John Wiley & Sons, Inc,Wiley
• Subject: Algorithms / CMOS / Computer memory / Electrodes / Energy bands / Energy consumption / Memristors / Perovskites / Photoelectricity / Sensors / Signal encoding / Spectrum analysis / Vision systems