Near-sensor analog computing system based on low-power and self-assembly nanoscaffolded BaTiO3:Nd2O3 memristor

• Published in: Nano today Vol. 55; p. 102144
• Main Authors: Zhang, Yinxing, Jia, Xiaotong, Xu, Jikang, Guo, Zhenqiang, Zhang, Weifeng, Wang, Yongrui, Li, Pengfei, Sun, Jiameng, Zhao, Zhen, Yang, Biao, Yan, Xiaobing
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01-04-2024
• Subjects: Edge detection; Ferroelectric memristor; Near-sensor analog computing system; Object shape recognition; Sensor; Near-sensor analog computing system; Edge detection; Ferroelectric memristor; Sensor; Object shape recognition
• ISSN: 1748-0132; 1878-044X
• DOI: 10.1016/j.nantod.2023.102144

Near-sensor analog computing systems have received a lot of attention as they can effectively reduce the large amount of redundant data transferred between sensor terminals and computing units, thereby shortening the data processing time and reducing power consumption. However, ensuring the reliability and stability of memristor devices used in the hardware circuits of near-sensor analog computing systems remains a considerable challenge. In this paper, we describe a robust ferroelectric memristor based on Pd/BaTiO3:Nd2O3/La0.67Sr0.33MnO3 grown on a silicon structure with SrTiO3 as the buffer layer. Through optimized growth temperature, the device exhibits a low coercive field voltage (−1–2 V), robust endurance characteristics (>1010 cycles), and a power consumption as low as 0.45 fJ per synaptic event. Also in this study, a near-sensor analog computing system based on an array of pressure sensors and ferroelectric memristors was constructed. It is shown that this system can accurately calculate multiple raw analog pressure signals in real time without the need for peripheral circuitry and that the system can classify object shapes and perform edge detection with a maximum deviation of only about 58.6 nA. This study highlights the great potential of ferroelectric memristors for use as fundamental components of near-sensor analog computing systems.
[Display omitted]
•Through optimized growth temperature, the device exhibits a low coercive field voltage (−1–2 V), which is lower than that previously reported for BaTiO3-based memristors [Advanced Materials 29, 1602795 2017; Nature communications 11, 1439 2020; Adv Mater 34, e2110343 2022].•The device has robust endurance characteristics (>1010 cycles), which is the best reported so far that we know of. More importantly, the power consumption per synaptic event is as low as 0.45 fJ, which is much lower than previously reported [Nat Commun 11, 1439, 2020; Nat Nanotechnol 7, 101-104, 2011; Nano Energy 89, 2021].•The near-sensor analog computing system we built accurately calculates multiple raw analog pressure signals in real time without peripheral circuitry and is capable of classifying object shapes and detecting edges with a maximum deviation of only about 58.6 nA..