Investigation On Electrical Performance And Synaptic Bionic Of Titanium Oxide Memristor

Recently,with the rapid development of information technology,it is also facing the severe challenge of high energy consumption and low efficiency in the massive data processing of the classic"von Neumann"computer.As novel nonlinear electronic component,memristor has the characteristics of resistance with current dynamic adjustment and storage integration,which is the most promising one to surpass Moore's Law.Among them,the memristor has attracted much attention to its non-uniformity and non-ideal simulation performance in the field of storage and neuromorphic computing.Aimed at further optimizing its uniformity and simulated cognitive characteristics,the TiO_x-based memristor is chosen to develop excellent endurance and ideal simulation characteristics in this thesis,the main work is as follows:First,a novel ITO/TiO_x/Ti N memristor of abrupt and gradual types coexisting is fabricated based on the optimization of different oxygen partial pressures.This device can be used as digital and analog memristors under different conditions,and the digital memristor can cycle 10~4times with the window up to 33 under 10%oxygen partial pressure;the thickness of the function layer is compared based on the device yield,and the device with the 10 nm function layer thickness has the optimized yield over 75%.Next,the Ta O_xlayer is inserted in ITO/TiO_x/TiN structure,the operating current of the device can be reduced from 1 m A to the n A level by optimizing the insertion position and thickness;when the thickness of TaO_xlayer is 5 nm,the uniformity of HRS/LRS of the device can be significantly improved and the endurance is improved to 10~5cycles without attenuation under certain conditions of operating current.Through the analysis of the conductive filament mechanism model,it can be seen that the TaO_xlayer acts as a virtual electrode to provide the basis for the re-formation and rupture of the conductive filament,so that the randomness of the conductive filament distribution is reduced and the device uniformity is improved.Finally,in order to optimize simulation characteristics,a multilayer titanium oxide-based analog memristor is fabricated based on function layers with different stoichiometric ratios;The multi-value storage characteristics of the device are verified under DC scanning by exploring the appropriate insert position of the TiO_ylayer and adjusting the TiO_y/TiO_xthickness ratio;when a 5 nm TiO_ylayer is inserted to the upper interface can realize the synaptic bionic function of ITO/TiO_y(5 nm)/TiO_x(10nm)/Ti N analog memristor,and the synaptic plasticity linearity is significantly improved(0.69),which overcomes the non-ideal simulation obstacles.It is revealed that an oxygen concentration gradient in the device effectively limits the number of oxygen vacancies involved in the formation process of conductive filaments at the beginning,which improves the conductance linearity of the dominant modulated oxygen vacancies.