Research On Memristor Storage Array Based On Metallic Oxide

Among all these non-volatile memories,memristors stand out because of its high storage density,low power consumption,fast switching speed and compatibility with CMOS process.However,traditional digital memristor can only be converted between a high resistive state and a low resistive state,and the resistance switching behaviors of analog memristor are hard to control.To develop stable multi-resistance-level memristor is particularly important and considered to be the key to further improve its storage density.Based on metallic oxide memristor and memristor line array and crossbar array,this thesis explores its resistive switching characteristics as well as retention,SET/RESET,multilevel storage and photoelectric characteristics.Besides,this thesis further simulates and applies the key characteristics of the memristor crossbar array.The main research contents are as follows:1.Using different materials and process parameters,this thesis prepared memristors with Al₂O₃ or Hf O₂ as functional layer and Aluminium,Copper or Tungsten as electrode layer.This thesis explored their basic characteristics of I-V,SET/RESET,retention and photoelectricity as well as the effects of electrode size,materials and thickness of functional layer.The results show that memristors with thinner oxide layer（<10nm）are easy to be broken down and function as a low resistance;memristors with thicker oxide layer（>25nm）have an exorbitant SET voltage and are even hard to be SET;memristors with 15～20nm oxide layer performed best.Moreover,this thesis prepared Al₂O₃/Hf O₂stacked layer memristors and found that they showed the characteristic of Forming-free.Besides,this thesis gained different low resistance states（LRS）and high resistance states（HRS）by adjusting the current limiting and RESET voltage.Thus,they showed the potential for multilevel storage.2.On the basic of single memristor,this thesis prepared memristor line array and memristor crossbar array using photolithographic process and explored their I-V characteristics of the memristor arrays.The results show that the line array has the characteristic like Forming when it is at LRS and has a higher RESET voltage about-1.5V.Then the line array can be SET/RESET by a smaller voltage about±0.2V than a single memristor.The crossbar array has a good transistor output characteristics,but its SET voltage is too high.3.Based on the memristor simulation frame“Memtorch”as well as the memristor model and window functions this frame supports,this thesis models and simulates a single memristor and a memristor crossbar array.And the resistive switching voltage of memristor is very low,at about±0.2V.The size of the memristor array is 64×64×10,Finally,and each input corresponds to the source and drain voltages of the crossbar array.Finally,this thesis uses the memristor crossbar array model and realizes the identification function of MNIST data set.