Fabrication And Performance Study Of Ion-gate Metal Oxide Synaptic Transistor

In recent years,data-intensive computing applications such as pattern recognition,video processing,database engine and network router have increased dramatically with the rapid development of big data and artificial intelligence（AI）,and the computing power demand of the current computing system has experienced explosive growth.The traditional digital computing system based on von Neumann system architecture,its miniaturization process is subject to the technological process increasingly approaching the physical limit,and it is difficult to realize the increase of the ratio of computing power to energy consumption by expanding the core scale and optimizing the advanced process.In terms of processing explosive data,it is a little difficult,and it is urgent to find a new computing paradigm.Human brain is a particularly ideal model to deal with the above problems.Human brain has a total of tens of billions of neurons and synapses,with low power consumption,high fault tolerance and efficient parallel computing capabilities.The central nervous system of organisms is a super-large multi-core parallel distributed information storage and computing network.This kind of complex network based on plasticity connection has more fault-tolerant learning and memory ability than any applied digital computing system.Therefore,inspired by the structure and function of the human brain,scientists have developed various artificial synaptic devices,among which the ion gate transistor is a promising candidate for constructing hardhardy-level neural networks due to its similar structure and working mechanism to that of biological synapses.In this study,metal oxide synaptic transistors were constructed by sol-gel and electrospinning techniques respectively,and a variety of biological synaptic plasticity was simulated.Specific work is as follows:1.The indium oxide（In₂O₃）synaptic transistor was prepared by simple and non-toxic water induction method at low temperature,and lactose was dissolved into citric acid solution as gate electrolyte.At low and high bias voltages,both short and long-term synaptic plasticity were simulated by ion relaxation in the gate electrolyte.Furthermore,based on the potentiation/depression characteristics,an artificial neural network for image recognition is developed.The recognition rate of the network is as high as 94.6%.This research can provide a green biomaterial-based platform for the development of environmentally friendly,economical and high-performance neuromorphic electronic devices.2.Hollow SnGaO nanofibers were prepared by electrospinning technology combined with Kirkendal effect,hollow nanofibers can be prepared by reducing the heating rate during annealing,and synaptic transistors based on SnGaO nanofibers were constructed.The lactose dissolved in citric acid solution is used as the gate electrolyte,which successfully simulates various biological synaptic functions,and achieves excellent long-term plasticity and ideal linear conductance modulation.The results show that the hollow nanofibers increase the interface area of the ion coupling,which significantly enhances the long-term plasticity of the synaptic transistor,and provides a new strategy for regulating the long-term plasticity of the synaptic transistor.