Artificial Synaptic Transistor Based On Titanium Dioxide Films

Since the concept of artificial intelligence was proposed,lots of breakthroughs have been made in the fields of machine learning,theorem proving,pattern recognition,and artificial intelligence language.However,due to the limitations of von Neumann computer architecture,the performance of traditional computers can no longer meet requirements of artificial intelligence.So new computing methods should be developed.Inspired by human brain,neuromorphic computing is proposed to realize artificial intelligence and reduce energy consumption.As the basic hardware unit of neuromorphic computing,recently artificial synapse has attracted worldwide research interest.Among different kind of device structure,synaptic transistor has attracted more attention due to its unique properties,such as the ability to perform signal transmission and learning at the same time.So they can be used in dynamic spatio-temporal information processing applications.At present,researchers have launched a series of studies on synaptic transistors,with their main work focused on device structure,materials,and working mechanism Electrolyte gated synaptic transistors can effectively use the ions in the electrolyte dielectric layer to adjust the conductivity of the channel.When the gate voltage is low,a double electric layer with high capacitance will form at the interface between the electrolyte and the channel layer.This electrostatic field induced channel conductance change is volatile.When the gate voltage is high,ions enter the channel layer through the interface and change the conductance of the channel layer.This change is generally non-volatile.There are natural advantages in simulating the process of neurotransmitter regulating the change of synaptic weight in the simulation of synaptic behavior.There have been studies that have realized the short-term plasticity,long-term plasticity,and spike time-dependent plasticity of synapses,but they are fundamental to the scalability of devices.According to the published literature,the research on durability,operating speed and electrolyte stability is still insufficient,and there is still room for optimization of the material and structure of the device,and there is still lack of reliable synaptic devices for neuromorphic computing.In this thesis,we first briefly review the development of artificial synapses.Then we select a representative electrolyte gated transistor as the research object and introduce its structure,working principle,and application prospects.At last we report our experimental exploration on this novel deviceAfter thorough literature search,titanium dioxide was selected as the channel layer material in our experiment.The device operating mechanism and feasibility of oxygen ions(oxygen vacancies)regulating its electrical conductivity were discussed.Titanium dioxide and YSZ amorphous thin films with higher flatness were prepared by magnetron sputtering.The thin films were annealed with suitable conditions to obtain a partially crystalline titanium dioxide thin film and a YSZ thin film with higher crystallinity.The transistor device was constructed using hard mask during the thin film depositing process.A home-made probe station was used to measure the transistor's electrical transport properties and simulate its synaptic function characteristics.The experimental results of the transfer characteristics show that the device has basic continuously adjustable characteristics.The single pulse stimulation proved the influence of the gate pulse voltage amplitude and duration on the channel layer conductance adjustment range.The short-term memory and long-term memory characteristics of synapses can be simulated by using gate voltages of different amplitude.Subsequent experiments through temperature change and continuous pulse stimulation showed that the conductance change of the device can be adjusted as needed during the regulation process.Finally,the parameters suitable for artificial synapses were determined,and continuous linear regulation of channel resistance was realized.This regulation remains stable after repeated potentiation/depression.At the same time,we also discussed the shortcoming of this device.The main problem is that the retention characteristics are poor in this transistor.Retention characteristics is a very important factor in artificial synapses.In subsequent experiments,we will try to overcome this disadvantage by optimizing the channel material and the device design In short,this thesis proposes an artificial synaptic transistor device based on titanium dioxide thin film.Our results may provide helpful information to the development and application of artificial synapses in neuromorphic computing and the realization of artificial intelligence.