Neuromorphic Computing And Degradable Applications Based On Memristors

In recent years,memristors have stood out among many electronic synaptic devices and degradable electronic devices due to their high scalability(<10 nm),multi-level function and low power consumption(sub-pJ).Memristors based on different materials can have different applications in different fields.Through choosing appropriate materials to prepare memristors,further studying their switching mechanisms,optimizing the devices through various physical parameters,could finally achieve stable performance memristors which could have specific applications towards specific field.Based on those prepared memristors,simulation experiments can provide technical support for large-scale applications of memristors in field of neuromorphic computing,degradable electronic devices and so on.This paper explores the application of nitride semiconductor material memristors for neuromorphic calculations,building the foundation for memristors to be used in hardware neural network calculations.At the same time,it explores memristors made of lactose(biological material)and uses them to explore degradable applications.To be summarized,this paper has done works including the following:(1)Successfully designed and manufactured a series of AlN-based semiconductor memristors,and selected TiN/AlN/Pt memristors considering of its best resistance switching performance,which proved to have reliable and stable resistance switching performance.(2)This work particularly examined whether the TiN/AlN/Pt have important features of the brain synapse which is long-term potentiation and long-term depression and improved the linearity of the memristor's conductance by optimizing the pulse conditions and finally achieved symmetric and linear synaptic properties.(3)The handwriting recognition data set in the database of the National Institute of Standards and Technology and the conductance of the memristors were used as synaptic weights in the neural network to perform handwritten digit recognition simulation.From the results,we could know that the AlN-based memristors achieved 95% accuracy of handwritten digit recognition.(4)In order to explore the performance and application of memristors made of organic materials,we used ?-lactose extracted from milk to fabricate solid electrolyte layers in memristors and explored the capabilities and suitability of ?-lactose memristors and applications suit for their properties.The devices can achieve repeatable switching with uniform switching voltages,and has multi-level storage capability.In addition,the Ag/?-lactose/ITO memristor device was immersed in deionized water.After immersion for a very short time(3 seconds),the ?-lactose film completely disappeared in deionized water and the water-insoluble Ag electrode left the surface of ITO and floated in water.Obviously,with the disappearance of the ?-lactose film,the resistance conversion behavior will also disappear.The experimental and simulation results in this paper show that TiN/AlN/Pt memristors have reliable and stable resistance switching performance,and by adjusting long-term potentiation and long-term depression features,TiN/AlN/Pt memristors can be used for neuromorphic computing.From our results,we can see that the AlN-based memristor can achieve 95% accuracy of handwritten digit recognition.This means our work shows that AlN-based memristors have the potential to be used as electronic synapses in future hardware neuromorphic systems.Also,this thesis shows that through experiments,memristors made by ?-lactose,a biomaterial,also have stable resistance switching performance and multi-level storage capacity.In particular,immersion of a lactose memristor in water can cause its switching ability to disappear which indicates that the ?-lactose memristor is very suitable for making transient electronic devices and using them as digestible bioelectronic devices or medical devices.