Study On Memristor Characteristics And Synaptic Plasticity Based Complex Metal Oxide Thin Films

Memristors can mimic the function of human brain neuromorphic calculation,breaking through the von Neumann framework.However,the formation and breaking of the conductive filament inside a conventional memristor is random,which results in discrete distribution of the threshold voltage of the device,which makes it difficult to mimic the function of biological synapse stably.This problem has become the biggest factor that hinders memristor applications.Therefore,the research of memristors based on new materials and structures has been put on the agenda.In response to these problems,this article has conducted an in-depth analysis of the testing results of memristors.In this paper,three kinds of metal oxide films,BaTiO₃?Hf_0.5Zr_0.5O₂ and NdNiO₃ are taken as examples to study the characteristics and neuromorphic plasticity of high-performance memristors.The details are as follows:Firstly,the ferroelectric tunnel memristors with different interfaces terminations based on the Pt/BaTiO₃(BTO)/La_0.7Sr_0.3MnO₃(LSMO)/SrTiO₃(STO)structure were prepared.We have made an in-depth discussion on the changes in the performance of ferroelectric tunnel memristors by nanoscale interface engineering.In this work,the ultra-thin single crystal BTO ferroelectric thin films was prepared.According to the test results,it indicated that the performance of different kinds of ferroelectric tunnel memristors can be controlled by interfaces terminations.Different interfaces terminations can make devices have different resistance switching behaviors.The most important thing is that different interfaces can adjust the intrinsic energy band of the device's heterojunction,which make the devices have different resistance gradient ranges,and it has different synaptic learning characteristics in the field of neuromorphic computing.This result in two kinds of ferroelectric tunnel memristors can play different roles in the Spiking Neuron Network(SNN)and Artificial Neuron Network(ANN).Secondly,a ferroelectric memristor based on Au/Hf_0.5Zr_0.5O₂(HZO)/p⁺-Si structure was prepared and its multi-level storage characteristics and synaptic plasticity characteristics were discussed.In this work,we prepared a ferroelectric HZO film.According to the testing,it illustrated that the proposed device has stable switching characteristics and obvious resistance state switching behavior at room temperature,and the synaptic plasticitycan can be faithfully mimicked,such as spike-timing dependent plasticity(STDP)and paired-pulse facilitation(PPF).This work shows that the proposed ferroelectric memristor has stable multi-level storage characteristics and the function of simulating biological synapses,which shows its potential to be applied to next-generation storage devices.Thirdly,a phase change memristor based on the Pd/NdNiO₃(NNO)/n^--Si structure was prepared.In this study,the test results show that the memristor based on NNO film has the excellent current-voltage(I-V)curves and stable retention characteristics.And it illustrated that the synaptic behavior of the human brain from a short-term plasticity(STP)to a long-term plasticity(LTP)state can be mimicked by the proposed devices,which lays the foundation for its successful application in neuromorphic networks.More importantly,the test datas shows that the synaptic plasticity characteristics can be faithfully mimicked by the NNO film-based memristor.These results paved the way for the research and development of memristors based on rare earth nickelate films.At the same time,these results have stimulated further research on the characteristics of high-performance NNO films.