Research On Design And Neural Synapse Bionics Of Memristor Based On Oxide Heterojunction

With the rapid development of the Internet and artificial intelligence,the demand for computing power is increasing day by day,and the computing power under the existing computer system is facing increasingly severe challenges.The traditional von Neumann computer system’s structure of separating memory from computing and the speed limit of a single bus expose existing computers to problems such as memory walls,power consumption walls,and loss of multi-core synchronization performance in the face of huge computing power demands.The neuromorphic computing system that simulates the characteristics of the human brain has the advantages of integrated storage and computing,low power consumption,etc.,and has become a new means to break through the bottleneck of traditional computing systems.Memristors have excellent electrical properties and synaptic characteristics similar to biological synapses,which can be used to simulate neuromorphic computing systems and become excellent candidates for building the next generation of high-performance computing systems.In addition,building a brain-inspired intelligent system is an important research direction for the application of neuromorphic computing,and among them,the artificial visual system is the most important link.In order to endow devices with the ability to perceive and process light signals,exploring memristors with photoelectric synaptic properties has become an important way to simulate artificial visual systems.Based on the above discussion,this dissertation conducts research around the following aspects:（1）Research on memristors with electrical synaptic properties:To overcome the computing power bottleneck in the von Neumann system,this study aims to develop memristors with electrical synaptic properties and to build a neuromorphic computing system that integrates memory and computing.The Ag/WO_x/Zn O/ITO memristor structure is fabricated by depositing Zn O and WO_x thin films and an Ag top electrode on an ITO coated glass substrate using magnetron sputtering technology.Synaptic plasticity of the devices is verified and tested using a semiconductor analyzer.The memristor reveals a gradual change in conductance upon application of a sweeping voltage,indicating the potential of the device to mimic biological synaptic plasticity.Furthermore,by designing pulse signals with different parameters and forms,the device exhibits short-term and long-term synaptic plasticity.The physical mechanism of the resistive switching effect of the device was explained in detail by fitting the IV curves and studying the migration of oxygen vacancies and oxygen ions.Based on the device’s long-term synaptic plasticity simulation,a cross array of memristors is constructed for neuromorphic computing,which successfully achieves image classification and recognition tasks.（2）Research on memristors with photoelectric synaptic properties:This study aims to explore the feasibility of using memristors to simulate the visual system by developing memristors with photoelectric synaptic properties.Mo S₂ thin films with excellent optoelectronic properties are grown on ITO coated glass substrates using the hydrothermal method.Then,the IGZO thin film and Ag top electrode are deposited on the substrate by magnetron sputtering to prepare the Ag/IGZO/Mo S₂/ITO photoelectric synaptic memristor.Using the same method as above,the electrical synaptic plasticity of the device was tested and verified.To investigate the potential application of the device in artificial visual systems,UV light pulses with varying parameters were used to probe the optoelectronic synaptic of the device.The device exhibited light-induced short-term and long-term synaptic plasticity.Meanwhile,under the effects of light illumination with different parameters,the device showed a transition from short-term memory to long-term memory.Based on the device’s electrical and photoelectric synaptic properties,a retinal array and a visual cortical array are constructed to simulate light perception,denoising,and pattern recognition capabilities of the visual system.This device provides a new solution for constructing artificial visual systems.