Research On Ultraviolet Photo-induced Non-volatile VO₂ Phase Transition And Intelligent Sensing

Human beings receive information from the external environment,where 80%of them obtained through visual perception.The retina can not only detect light signals,but also preform the first stage of image pre-processing.This process of visual perception and cognitive learning has inspired the development of artificial visual systems.CMOS-based artificial intelligence vision systems have physical separation of each functional component,which leads to delays in data access and relatively high-power consumption.Therefore,the development of multifunctional electronic devices integrating sensing,memory and processing functions is an effective way to improve the efficiency of artificial visual systems.Among them,optoelectronic neuromorphic sensors provide a good choice because of their optical sensing and storage properties.Due to the limitation of their working mechanisms,the existing optoelectronic synapses face challenges in large writing non-linearity and difficulty in large-scale integration.To this end,materials that can respond to photo stimulation in conjunction with non-volatile phase transition characteristics could open new avenues for the realization of high-performance neuromorphic sensors.As a strongly correlated oxide,vanadium dioxide（VO₂）has a variety of isomerized phases and VO_xphases with slight differences in oxygen content.Since the formation energy of each phase is similar,transformation between the existing phases can be easily realized by external field excitation.In this study,we found that VO₂had selective recognition and non-volatile memory properties for ultraviolet light（UV）.We utilized a variety of characterization methods to deeply study the physical mechanism,and prepared the UV neuromorphic sensor to demonstrate the functions of image pre-processing,motion recognition and detection.The main details are listed as follows:1.The high quality VO₂ films grown by epitaxy on Al₂O₃substrates are irradiated by different wavelength light,where VO₂only exhibits non-volatile multi-level response to UV light.Different experimental methods are used to character VO₂films before and after UV irradiation.It is found that UV exposure can induce structural and electronic phase transition of VO₂.This is because the photon energy of UV light is higher than the activation energy of oxygen vacancy,which results in the removal of oxygen from VO₂film and leads to photo-induced phase transition.The reversible modulation of this non-volatile response can be achieved by insert the oxygen ions back into the film through the electrolyte gating.2.The potential for mass production of VO₂-based UV neuromorphic sensor was studied.Large scale VO₂thin films were grown on silicon wafers by magnetron sputtering and fabricated into neuromorphic sensor arrays.By testing randomly selected devices,we proved that the film had good uniformity,and also confirmed the silicon-based VO₂film has good UV selective recognition and photo-induced phase transition characteristics.Based on this characteristic,we constructed an artificial neural network and preform pre-processing function and handwritten digit recognition of the image containing red,green,blue and UV light information.3.The application of UV neuromorphic sensor was extended to the field of flexible wearable electronics,which can accommodate various complex environment and realize the detection of UV light under different bending states.Based on van der Waals epitaxial method,we grew large area of VO₂films on atom-level strippable flexible mica substrates.We also verified the homogenous of the film growth and the properties of UV photo-induced phase transition.By testing the UV exposure and electrolyte gating of the films under different bending conditions,and comparing the results with the device performance in the plane state,we proved that the VO₂/mica flexible structure has good bending tolerance and mechanical stability.Finally,we constructed two kinds of neural networks,near-sensor and in-sensor structure,and demonstrated handwritten digit recognition and motion detection based on these two architectures,respectively.Our work extends the application of traditional infrared optical material VO₂to the field of UV intelligent optoelectronic sensing,and provides a new choice for near-sensor/in-sensor calculation design.