Research On Ultrathin Memristors Based On Silver Ion Migration

Nowadays,the traditional von Neumann architecture has confronted serious challenges to deal with explosively increased data because of the gigantic difference in read/write speed between the fast processor and the slow memory.Neuromorphic computing,which has been proposed conceptually by Mead to simulate the biological brain,is a promising candidate to overcome the bottleneck.Memristors are similar to biological synapses because their resistance can be precisely modulated by electric charges flowing through them,which provides an ideal platform for overcome the von Neumann bottleneck and investigating neuromorphic computing.Then enormous interest has been triggered to exploit the memristive properties.However,most of the work on memristors focuses on the selection of device materials and the characterization of memristive properties,and little attention is paid to the effect of device scale on the resistive behavior,especially in vertical scale.The key to integration has changed from the reduction of the two-dimensional size to the development of 3D stacking technology and reducing the vertical scale has become a research direction to the integration.In this thesis,the main work and conclusions are as follows:1.The high-quality Ag/SrTiO₃/La_0.67Sr_0.33MnO₃/SrTiO₃ devices with smooth surface and good phase formation were fabricated by optimizing the growth parameters of La0.67Sr0.33MnO3.2.The effect of the SrTiO₃ thickness on the resistive behavior of Ag/SrTiO₃/La_0.67Sr_0.33MnO₃/SrTiO₃ devices was investigated.With the SrTiO₃ active layer down to 3 unit cells in thickness,efficient control of Ag⁺-ion migration gave rise to enhanced memristive properties with the conductance continuously modulated within a large memory window of～26000%between an ohmic low resistance state and an electron-tunneling high resistance state.The memristor exhibited good performance of maintaining and reversing.In addition,the resistance behavior of the memristor was not related to oxygen vacancy and ferroelectricity but with the Ag electrode and SrTiO₃active layer.3.The memristive properties of Ag/SrTiO₃/La_0.67Sr_0.33MnO₃/SrTiO₃ memristor were investigated.The memristor exhibited excellent linearity and long-term plasticity and the spike-timing-dependent plasticity was dependent upon the waveform,amplitude,pulse width of the spike and the initial state of the memristor.In the high resistance state,excellent spike-timing-dependent plasticity characteristics with a large modulation of synaptic weight of～3500%and sensitive response to applied electrical stimuli of as low as～1.0 V and as fast as～0.01 ms were observed,which was faster by more than 3 orders of magnitude than the biological synapses.Adopting the spike-timing-dependent plasticity results as database,the neuromorphic computation,namely the recognition of handwritten digits,showed a high accuracy rate of 95.5%.The Ag/SrTiO₃/La_0.67Sr_0.33MnO₃/SrTiO₃ memristor exhibited nearly the thinnest thickness of active layer and the largest modulation of synaptic weight of spike-timing-dependent plasticity compared with similar devices,which facilitated the design and fabrication of high-density artificial neural networks with reduced device size in vertical.4.The effect of the ratio of the space volume occupied by ions on the resistive behavior was investigated.The resistive behavior of LaAlO₃ and La FeO₃ devices with the same thickness of active layer was compared and investigated,and the results showed that the resistive behavior did not have a distinct dependence on the ratio of the space volume occupied by ions.