| In recent years,the rapid development of artificial intelligence based on neural networks has led to a significant increase in computing demand,prompting the urgent need to improve computing hardware.However,the conventional von Neumann architecture is limited by the bottleneck caused by the storage-computing separation structure,and cannot keep up with the demand for improved computing power in the future.To address this,the memristor has emerged as a promising candidate due to its ability to integrate memory and computing,as well as its high computing parallelism,low power consumption,and high integration density.In the field of neural network computing,memristors need to have multiple resistance states during training,and after training,their resistance states should remain stable.This study uses single crystal Li Nb O3 film memristors to investigate resistance state control methods,laying the foundation for the development of memristors with multiple stable resistance states.The study also examines the impact of external environmental factors,including irradiation,atmosphere,and temperature,on device characteristics.The main work are as follows:（1）Firstly,this thesis aims to investigate the resistance state control method,specifically by analyzing the impact of the electroforming process and the number of scanning voltage cycles on the number,distribution,and stability of the memristor’s resistance state.The results show that increasing the compliance current of the electroforming process from 5×10-5 A to 1×10-4 A enables more resistance states.However,when the compliance current is increased to 1×10-3 A,the number and distribution of memristors remain the same,but the stability decreases.Additionally,increasing the number of scanning voltage cycles leads to more regular resistance state distribution and better stability.Based on these findings,this study regulates the resistance state of memristor by pulse voltages,enabling the memristor to have eight stable and distinguishable resistance states.（2）The study also investigates the impact of external environmental factors on device characteristics,focusing on the effects ofγradiation,atmosphere,and temperature,and analyzing the influence mechanism.The characteristics of the memristor remain stable as the irradiation dose increases from 0 Gy to 3733 Gy,but its resistance gradually decreases.This is because an increase in irradiation dose can lead to an increase in the number of oxygen vacancies in the film.Besides,in a vacuum and atmospheric environment,the characteristics of memristors show no significant change.This is mainly because the existing oxygen vacancies in the film are sufficient to form conductive filaments,and there is no need to use the generated oxygen vacancies,whose generating process is influenced by the atmosphere,to form conductive filaments.As the temperature decreases from 300 K to 200 K,the switching ratio of the memristor also decreases,as a decrease in temperature makes it more difficult for oxygen vacancies to move under the electric field. |
PDF Full Text Request