Preparation And Performance Research On Tantalum Oxide Based Self-Rectifying Memristors

In the era of rapid development of information technology,the huge amount of data attached to the"von Neumann"computing system is facing a major challenge of high consumption and low efficiency when processing and computing,and the emergence of memristors will likely solve this critical problem,with its unique resistance adjustable.It is expected to become a powerful tool to break Moore’s Law.Memristor is a two-terminal device,which not only can achieve fast transition between different resistive states during operation,but also can accomplish good compatibility with COMS process during preparation,so it is often used in the field of information storage and neurosynaptic simulation.In order to meet the huge amount of data storage and operation,memristors are often prepared into cross-array style structure to improve device integration,so there may be crosstalk current,the solution can be connected in series with a transistor or diode and other selective units in the circuit to achieve the effect of crosstalk current suppression,however,using this approach will affect the integration of the device,so the rectification characteristics of the memristor itself is used to Therefore,it is more reasonable to use the rectification characteristics of the memristor itself to suppress the crosstalk current in the circuit.In this paper,tantalum oxide is used as the dielectric layer of the memristor to study the self-rectifying memristor.The main research contents and results are as follows.First of all,the W/TaO_x/TiN memristor with a single dielectric layer is constructed,and the oxygen partial pressure,a key parameter during the deposition of tantalum oxide dielectric layer,is optimized,and the performance of three structural memristors,Ti/TaO_x/TiN,ITO/TaO_x/TiN,and W/TaO_x/TiN,is discussed from the perspective of electrode engineering,and it is concluded that the optimal oxygen partial pressure is 10%,the memristor with W as the top electrode has the best retardation characteristics and the largest window,and the analog memristor at this time has a cycle endurance of 5000 cycles and a retention time of more than 2000s.Second,based on the W/TaO_x/TiN memristor structure,the devices with different electrode sizes are prepared by the photolithography process,and the resistance transition mechanism is elaborated from the potential barrier perspective by combining the electrical properties of the devices with different electrode sizes,and the synaptic performance simulations in artificial neural networks are performed by applying pulse tests based on the optimized devices.Finally,a self-rectifying memristor is constructed for the stacked structure of tantalum oxide and alumina,and the rectification ratio of the device is optimized mainly by changing the position and thickness of the alumina layer,and the rectification ratio of the device is increased to 403.79.Further,for different alumina thicknesses,the causes of the phenomenon are discussed in detail by the film element composition,device microscopic characterization and double logarithmic fitting,which may be related to the effective Schottky barrier thickness change and tunneling effect.The adaptable size of the array calculated by Kirchhoff’s equation is increased from N=4 to N=1425 without rectification characteristics,and the amnestic resistors with different thicknesses of alumina are also increased from N=70 to N=1425,showing the great potential of suppressing leakage currents and simulating neural synapses after the devices are integrated into the array,which provides a reference value for the general application of amnestic resistors in the field of neural computing in the future.