The physical separation of functional components of traditional AI vision systems results in latency in data access and relatively high power consumption. About 80% of the way humans obtain information from the outside world depends on vision, the retina can detect light stimulation, and can perform preliminary light signal processing, this efficient visual perception and cognitive learning process has inspired the development of artificial visual systems in the future. In this context, neuromorphic intelligent photoelectric sensor devices that integrate perception, storage and computing functions have become cutting-edge research hotspots in recent years.
Jin Kuijuan, researcher of the Key Laboratory of Photophysics of the Institute of Physics of the Chinese Academy of Sciences/Beijing National Research Center for Condensed Matter Physics, and the L03 Group led by Yang Guozhen, academician of the Chinese Academy of Sciences, are committed to the preparation of functional oxide films, state regulation and device applications by laser molecular beam epitaxial method. Ge Chen, an associate researcher in the research group, explored and carried out research on intelligent photoelectric sensor devices for oxide neuromorphics around the basic scientific problems of new oxide photoelectric films and intelligent photoelectric sensing physics (Nano Energy, 89, 106439, 2021). Recently, the team proposed a new neuromorphic photoelectric sensor based on ultraviolet irradiation/electrolyte-regulated VO2 nonvolatile phase transition, which showed good linearity, retention characteristics, silicon-based compatibility, built an artificial neural network and demonstrated functions such as image recognition.
Vanadium dioxide (VO2) is a typical strongly associated oxide, there are a variety of isomeric phases and subtle differences in oxygen content caused by the rich VOx phase, the study shows that through the electric field, light field, pressure field and other field regulation can achieve phase-to-phase conversion. The research team used the laser molecular beam epitaxial method to grow high-quality VO2/Al2O3 thin films, prepared them into phototransistor structures and performed photoelectric tests. The study found that the VO2 film under ultraviolet irradiation under nonvolatile changes, while under visible light illumination only transient photoelectric response. Increasing the dose of UV radiation can even induce a non-volatile phase transition of VO2, from an insulated monoclinic phase to a rutile phase. The series of characterization results show that this is mainly due to the fact that UV irradiation creates oxygen vacancies in the VO2 film, while visible light with photon energy lower than its oxygen vacancy activation energy only produces a transient photoelectric response. In the process of reset, based on the team’s previous research on the direction of ion transport at the electrolyte regulation interface[AdvMaterInaterfaces21500407(2015)；AdvMater301801548(2018)；AdvMater311900379(2019)；AdvFunctMater291902702(2019);NanoEnergy67104268(2020)]proposes a protocol for inserting oxygen ions into oxygen deletion films using electrolyte gating. Therefore, through ultraviolet light irradiation and electrolyte regulation of vo2 oxygen shedding/embedding, can achieve reversible non-volatile regulation of its conductance, and then designed intelligent ultraviolet photoelectric sensor devices.
In addition, the research team grew a large area of VO2 film on a silicon wafer by magnetron sputtering technology and prepared it into an array of neuromorphic sensor devices. The study randomly selected 100 of these devices for testing, which proved that the film exhibited good uniformity. Vo2 films grown on silicon wafers have photogenetic nonvolatile phase transition characteristics and polymorphic reversible regulation characteristics similar to epitaxially grown VO2 films, demonstrating the potential for large-scale integration of this new principle device. Further studies have shown that the nonvolatile change of channel current shows an approximate linear relationship with the dose of ultraviolet radiation, which lays a good foundation for future applications. Recently, studies have demonstrated the possibility of VO2 in the application of flexible smart photoelectric sensor devices (Adanced Functional Materials）。
Figure 1: UV irradiation/electrolyte gating achieves non-volatile reversible regulation of VO2
Figure 2. Properties of VO2 thin film devices grown on silicon wafers
Figure 3:An artificial neural network constructed from a VO2-based neuromorphic UV photoelectric sensor device for image recognition demonstration
Based on the new VO2 neuromorphic photoelectric sensor device, an artificial neural network is constructed and the standard MNIST handwritten digital image is recognized, which can preprocess the images randomly introduced into RGB Gaussian noise and selectively identify the ultraviolet information contained therein. For images containing RGB Gaussian noise, the recognition accuracy is only 24%. In contrast, after the pre-processing of ultraviolet light information using VO2-based neuromorphic photoelectric sensors, the recognition accuracy of the image reaches 93%, which is the same as that of the original MNIST image.
This study expands the application of the traditional infrared optical material VO2 to the field of ultraviolet intelligent photoelectric sensing, providing a new choice for near-sensor computing/in-sensor computing design. Relevant research results toPhoto-induced non-volatile VO2 phase transition for neuromorphic ultraviolet sensorsIt was published in Nature Communications and recommended as Feathered Articles. In addition, the results were endorsed by Nature ElectronicsVanadium dioxide remembers the lightResearch Highlights were made for the title. The research work is supported by the National Key Research and Development Program, the National Natural Science Foundation of China, and the Youth Innovation Promotion Association of the Chinese Academy of Sciences. (Source: Institute of Physics, Chinese Academy of Sciences)
Related paper information:https://doi.org/10.1038/s41467-022-29456-5
Special statement: This article is reproduced only for the purpose of disseminating information, does not mean to represent the views of this website or confirm the authenticity of its content; if other media, websites or individuals reprint from this website, you must retain the “source” indicated on this website, and bear the legal responsibility such as copyright; if the author does not want to be reproduced or contact us for reprint fees, please contact us.