The Study Of Ion Transport-based Oxide Memristive Neurons And Their Neural Pulse Firing Characteristics

In order to meet the needs of massive data processing under the development of social informatization and the development of artificial intelligence,building a brain-like neuromorphic system with low power consumption and the ability to process complex information is an important method to accelerate the development of computing systems.Artificial neurons are the key components in the development of artificial intelligence,which are very important in brain-like electronic devices and become the basic hardware and hot research object of constructing neural morphology system.Memristors are not only simple in structure but also have unique resistance switching characteristics,which can use their resistance changes to simulate the function of neurons,and are considered to be the most potential candidates for building biomimetic artificial neural components.Exploring compact and low-power artificial neurons is critical to the development of neuromorphic computing.However,the artificial neurons based on memristor constructed by researchers still need other circuit components with high power consumption.According to the above problems,this paper studies the neuron characteristics of different ion memristors,and demonstrates the application.The main work is as follows:1.A Pt/HfO₂/Ag memristive device was prepared,and its nerve impulse firing characteristics were studied.The driving voltage of silver ions is low,and the silver conductive filaments are spontaneously formed and broken under the action of electric field and self-diffusion,so that the resistance switching process of the device occurs.Without the assistance of an external circuit,a constant bias current was directly applied to the device,and the device exhibited the neuronal voltage spiking characteristics.The frequency of the voltage pulse spike signal of the device is in a certain input current range,and the frequency increases with the increase of the input current intensity,which is similar to the excitatory neuron characteristics of biological neurons.The power consumption of a single pulse firing of the device is about 72 p J.In addition,the memristive neuron also exhibits phase transition,burst and pulse spike adaptive discharge behavior,which has application potential in the construction of low-power artificial neural networks.Using Pt/HfO₂/Ag memristive spiking neurons to simulate the recognition of handwritten letters,the recognition rate is over 96%.2.A Pt/Co₃O_4-x/ITO memristive device was prepared,and its nerve impulse firing characteristics were studied.The resistive switching mechanism of the device is the formation and fracture process of oxygen vacancy conducting filaments under the action of electric field and electric field-induced Joule heating.The device can simulate various firing behaviors of neurons under the input of constant bias current.It is applied to adaptive neurons according to their adaptive properties of voltage pulse spike frequency that is time-dependent under current input.The Pt/Co₃O_4-x/ITO memristor and sensor structure are used to construct an adaptive artificial tactile sensing neuron,which can effectively sense the subtle changes of the pressure signal over time;the neural network simulation results highlight the system’s ability to identify multiple Superiority in accuracy and sensitivity for a dynamic touch object.The Pt/HfO₂/Ag and Pt/Co₃O_4-x/ITO memristive devices are based on the migration and diffusion process of silver ions and oxygen vacancies,respectively,to realize the resistance switching process of the memristor.Both ionic memristors can provide rich neural spiking function under the action of bias current.This simple device structure and ease of operation will pave the way for high-density and low-power neuromorphic systems.