Research On The Characteristics Of Memristors And Neurosynapses Based On Simple Oxides

Memristors can not only meet the needs of high-density information storage and high-performance computing for next-generation memory,but also have the functions of non-volatile logic operations and brain-like neuromorphic computing.Therefore,since its introduction,the memristor has made great research progress.However,memristor devices based on traditional single-layer oxide materials or metal electrodes have major bottleneck in switching voltage distribution,low power consumption,and high speed.Finding some new device structures or new oxide materials to further optimize the memristor performance and improve the stability of the device is a top priority.Thus,this article mainly studies the improvement of memristor performance based on simple oxides and the realization of synaptic plasticity.The problem of uneven voltage distribution of the memristor is improved by the preparation of a double-layer device structure,and graphene electrodes or two-dimensional titania nanosheets are used to reduce the power consumption of the device,and the synaptic bionic function of high-performance memristors is studied.The main contents are as follows:1.A Ta/TaO_x/AlN/Pt double-layer memristor device was fabricated by inserting a wide band gap AlN film.Compared with the device without AlN film,the double-layer device exhibits abrupt resistive switching behavior,and has better stability and lower switching voltage.The Ta/TaO_x/AlN/Pt devices provide an ideal choice for the next generation of non-volatile electronic chip systems,but in order to better mimic biological synaptic behavior,it is necessary to prepare a memristor with bidirectional progressive adjustment.Therefore,a double-ended memristor with Ta/Ta₂O₅/AlN/Graphene structure was fabricated,which is based on the multi-layer graphene bottom electrode instead of the traditional metal electrode.The device has stable electrical characteristics under direct current voltage scanning.More importantly,the energy value of a write event can be as low as 37 f J,which further proves the low power consumption.In addition,the memristor can completely mimic the function of biological synapses and synaptic plasticity,including spiking time dependent plasticity,paired pulse facilitation,excitatory postsynaptic current,long-term memory,and short-term memory.According to the fitting result of the current-voltage curve,the conduction mechanism is attributed to trap-assisted tunneling.Therefore,this work has excellent potential in artificial synaptic neuromorphic computing with high stability and low power consumption.2.The titania nanosheets film was prepared by spin-coating method to construct a memristor with Ag/2D-Ti O_x/Pt structure.After the voltage-current testing,the device has typical bipolar resistive switching behavior,and has a low switching voltage(set voltage of0.37 V,reset voltage of-0.2 V),high switch resistance ratio(over 10⁵)and low switching power(set power is 10^-9 W,reset power is 10^-5 W).It is worth noting that a single device exhibits uniform and stable different resistance states under different compliance currents,thereby realizing multi-level storage.The four memristor devices are connected to the actual storage circuit,and the memristor device is selectively stimulated to obtain different output current values,and finally realizes storable and programmable characteristics.In addition,the device conductance can be adjusted by positive and negative alternating current voltage pulses,which is conductive to simulating biological synaptic functions.Therefore,this work provides an excellent solution for high information storage and neuromorphic computing applications.3.The exploration field effect transistor and the memristor are connected in series to form a 1T1R structure,which effectively suppresses the leakage current and realizes a high-density cross array.This chapter studies transistor characteristics in detail,and uses Cadence software to simulate transistor gate characteristics and further draw the layout.This structure is compatible with CMOS technology and provides new insights for the application of system-level chips.