Performance Optimization And Neuron Application Research Of Novel Memory Device Based On Ge₂Sb₂Te₅

In recent years,the rapid development of artificial intelligence technology has stimulated the demand for"massive data"storage and computing.However,there is always an insurmountable performance barrier between the traditional memory and CPU,which restricts the development of the traditional von Neumann architecture computing system.The"In-memory Computing"technology centered on emerging storage technologies such as phase change memory and resistive switching memory is expected to break through this barrier and build a low-power and low-latency"Computing in Memory"architecture computing system.Ge₂Sb₂Te₅ is a storage medium material with a wide range of applications,relatively mature technology and broad development prospects.However,the high operating current/voltage of the traditional Ge₂Sb₂Te₅-based phase change memory devices and resistive switching memory devices can not comply with the development trend of low power consumption and high density storage.Therefore,for the purpose of optimizing the operating current/voltage,this thesis designs two different device performance optimization schemes for the two different application scenarios of phase change memory and resistive switching memory.The main research contents include:（1）Thesis proposes a design scheme that uses a new blade-type phase change memory to replace the traditional T-type phase change memory to solve the problem of high operating current.Firstly,in the COMSOL Multiphysics software,a three-dimensional thermoelectric model of the blade-type and T-type phase change memory with the same feature size was constructed,and the reset operation process with significant power consumption of the phase-change memory was simulated.The simulation results show that compared with the T-type structure,the reset operation current of the blade-type structure is reduced by more than 10 times.Secondly,the simulation results show that during the reset operation,the high current density and significant temperature change in the phase-change memory cell will cause strong thermoelectric effects,mainly including the Peltier effect at the connection interface between the phase change layer and the heater,and the Thomson effect in the phase change layer.And under positive electrical excitation,the thermoelectric effect can reduce the operating current by enhancing the heating efficiency of the phase change memory cell.Taking advantage of this characteristic,skutterudite materials with better thermoelectric properties were screened out as substitutes for Ti N heater materials to further optimize device performance.The simulation results show that compared with Ti N materials,the reset current of the blade-type phase change memory using skutterudite material as the heater is reduced by 13.8%,and the power consumption is reduced by 17.9%.（2）A novel Ag/MXene/Ge₂Sb₂Te₅/Pt resistive switching memory was prepared to solve the problem of high operating voltage by introducing two-dimensional material Mxene into Ag/Ge₂Sb₂Te₅/Pt three-layer film structure.The electrical test results showed that the introduction of MXene material reduced the operating voltage of the device from 1.6 V to 0.38 V,and the operating power consumption decreased from 2.24 m W to 76 n W.In recent years,the introduction of MXene material in resistive switching memory to improve device performance has been proved to be an effective solution,however,there is no research on the MXene/Ge₂Sb₂Te₅ structure.Therefore,this thesis reveals the benefit of MXene/Ge₂Sb₂Te₅ structure to reduce the operating power consumption of devices through first-principles calculation.In addition,Ag/MXene/Ge₂Sb₂Te₅/Pt devices also exhibit threshold switching characteristics and continuous conductance modulation capability.Taking advantage of this feature,this thesis successfully simulates the integration and firing functions of biological neurons by applying continuous current pulse stimulation to the device,and explores the development prospects of using this device to construct artificial neurons and apply it to"neuromorphic"circuits.