Research On Memritive Device And Its Synaptic Plasticity Based On Two-dimension Material

When Information and Communication Technology processes and produces huge data,computer based on von-Neumann structure runs into the bottleneck of“memory wall”,which is difficult to satisfy the user in processing speed.Thus,the Neuromorphic computing stucture,which is focus on simulate the human brain,is proposed,and attracts extensive attention.The structure is usually composed of Neuron and Synapse,and stores the state and synaptic weight,respectively.However,the established Neuromorphic chip,whatever“Truenorth”or“Cambrian”,was built by traditional transistor.The ratio of the number of“neuron”and“synapse”is inappropriate.Using the Micro-nano device to build the simple synapse model,might be the key of reducing the scale of Neuromorphic chip.And now the most promising device is memrisor.The analog-type HfO₂ memristor has already realized the synapse function in one device,and the diffusion-type SiO₂:Ag memristor has achieved neuron function in several devices.The novelty device which is based on two-dimension materials like graphene and the transition metal dichalcogenide also show an advantage in electrical properties.In this thesis,all the studies are around the MoS₂ device.The memristor based on MoS₂ stacked film has been fabricated,and its performance was improved by modifying the surface state.The MoS₂ synaptic transistor with chemical-electrical mix synaptic function has also been demonstrated.Spercifically,the main results are as followed:?1?Investigation on the influence of nitrogen plasma surface treatment in MoS₂stacked film.It was indicated that the deposit of carbonitrides compound is the key point to cause the resistive switching?RS?behavior.The multistate RS ratio is relatively stable,resistance range is 100?¹00 k?,cycle for more than 100 times.The volatile RS behavior carry with 0.25 V set voltage,and the gradual analog RS could adjust the conductance from 12?G²5?G.?2?Analyse the mechanism by fitting the I-V curve and component characterization.Build the behavior model of Ag conductive filaments and try to explain the multistate phenomena with a universal theory.?3?We further investigated the synaptic plasticity of MoS₂ memristor.A simulation software NeuroSim was used to evaluate the nonlinearity of weight change.The accuracy of MNIST task is⁷5%.?4?Fabricate the ion gated MoS₂ transistor by mechanical exfoliation and dry transfer method.The width of the excitatory post-synaptic current?EPSC?spikes are studied.Owing to stable lithium ionic intercalcation,the short-term plasticity and the long-term plasticity?LTP?behavior have been mimicked,respectively.?5?When the high frequency signal input through the channel of transistor,a spike-rate-dependent plasticity?SRDP?behavior was observed.Thus,we proposed a“mixed-artifacial”concept,attempting to efficiently control the facilitation function by STP.Under the 500 Hz input signal,the facilitation of channel can be increased 5 times the initial state.