Zhifei Jian , Wenhua Li , Lin Zhang , Xingui Tang , Renkai Zhao , Yongxi Liang , Yanping Jiang , Xiaobin Guo , Guowu Tang , Kai Yan
{"title":"Ferroelectric polarization modulated optoelectronic synapses based on BaTiO3/TiO2 heterojunction for non-volatile visual memory","authors":"Zhifei Jian , Wenhua Li , Lin Zhang , Xingui Tang , Renkai Zhao , Yongxi Liang , Yanping Jiang , Xiaobin Guo , Guowu Tang , Kai Yan","doi":"10.1016/j.jcis.2025.138398","DOIUrl":null,"url":null,"abstract":"<div><div>The growing demand for neuromorphic computing architectures that mimic biological information processing has driven extensive research on optoelectronic synapses with multimodal neuromodulation capabilities. In this study, BaTiO<sub>3</sub>/TiO<sub>2</sub> optoelectronic synaptic devices with high non-volatile memory characteristics were constructed by interfacial energy band engineering. This heterojunction synaptic device achieves a 1350 % enhancement in relaxation time (<em>τ</em><sub><em>2</em></sub> = 84.14 s) compared to conventional BaTiO<sub>3</sub> device (<em>τ</em><sub><em>2</em></sub> = 6.21 s). Moreover, short-term to long-term memory conversion and the cognitive process of “learning experience” are achieved by adjusting light pulse parameters. Through further investigation, a synergistic ferroelectric polarization strategy is proposed, demonstrating that downward polarization extends <em>τ</em><sub><em>2</em></sub> to 202.93 s, with image retention time exceeding 4800 s. The synaptic device demonstrates biological-level energy efficiency (10.45 fJ) while achieving 97.5 % and 89.05 % recognition accuracy on MNIST and Fashion-MNIST datasets, respectively, through convolutional neural networks. This work not only exhibits the application prospect of ferroelectric semiconductor-based heterojunction in artificial optoelectronic synapse but also provides new ideas for the design and application of multimodal neuromorphic devices.</div></div>","PeriodicalId":351,"journal":{"name":"Journal of Colloid and Interface Science","volume":"700 ","pages":"Article 138398"},"PeriodicalIF":9.4000,"publicationDate":"2025-07-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Journal of Colloid and Interface Science","FirstCategoryId":"92","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S0021979725017898","RegionNum":1,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"CHEMISTRY, PHYSICAL","Score":null,"Total":0}
引用次数: 0
Abstract
The growing demand for neuromorphic computing architectures that mimic biological information processing has driven extensive research on optoelectronic synapses with multimodal neuromodulation capabilities. In this study, BaTiO3/TiO2 optoelectronic synaptic devices with high non-volatile memory characteristics were constructed by interfacial energy band engineering. This heterojunction synaptic device achieves a 1350 % enhancement in relaxation time (τ2 = 84.14 s) compared to conventional BaTiO3 device (τ2 = 6.21 s). Moreover, short-term to long-term memory conversion and the cognitive process of “learning experience” are achieved by adjusting light pulse parameters. Through further investigation, a synergistic ferroelectric polarization strategy is proposed, demonstrating that downward polarization extends τ2 to 202.93 s, with image retention time exceeding 4800 s. The synaptic device demonstrates biological-level energy efficiency (10.45 fJ) while achieving 97.5 % and 89.05 % recognition accuracy on MNIST and Fashion-MNIST datasets, respectively, through convolutional neural networks. This work not only exhibits the application prospect of ferroelectric semiconductor-based heterojunction in artificial optoelectronic synapse but also provides new ideas for the design and application of multimodal neuromorphic devices.
期刊介绍:
The Journal of Colloid and Interface Science publishes original research findings on the fundamental principles of colloid and interface science, as well as innovative applications in various fields. The criteria for publication include impact, quality, novelty, and originality.
Emphasis:
The journal emphasizes fundamental scientific innovation within the following categories:
A.Colloidal Materials and Nanomaterials
B.Soft Colloidal and Self-Assembly Systems
C.Adsorption, Catalysis, and Electrochemistry
D.Interfacial Processes, Capillarity, and Wetting
E.Biomaterials and Nanomedicine
F.Energy Conversion and Storage, and Environmental Technologies