Preparation And Performance Modification Of HfO_x-based RRAMs On Flexible Subtrates

Resistive random access memory(RRAM)has become a probable candidate for the next-generation storage technology due to its excellent non-volatile storage performance.Transparent oxide based RRAMs on flexible substrates show important application potential in visible electronics.In addition,the computer formed by the separated computing and storage space has von Neumann bottleneck and storage wall problems.The emerging neural form calculation is considered as an effective way to break through the above-mentioned bottleneck.By using the multi-resistance gradient of the resistive device to mimic the connection weight of the synapses,we can realize the huge circuit with the function of the neural network.Then the data can be processed efficiently in parallel way,and the deficiency of the traditional computing architecture can be overcome.Therefore,this paper mainly studies the preparation and optimization of the HfOx-based flexible and transparent RRAM,and analyzes its resistance access memory behavior,and explores the possibility of the device as an electronic synaptic device.We also qualitatively simulate the resistance change process of the device,and analyze its electric-heat coupling process to provide a basis for better understanding the working mechanism of the device.The results show that the ITO/HfOx/ZnO/ITO RRAM device based on the PET substrate has good flexibility,high optical transmittance and stable resistivity storage characteristics.Compared with the single dielectric layer RRAM(ITO/HfOx/ITO),the interface barrier of HfOx/ZnO RRAMs limited the charge transport,which makes the device's maximum operating current down from mA to ?A magnitude(device power also decreased to ?W).In addition,the high and low resistance state conduction mechanism of the single dielectric layer RRAM device is the space charge limiting current mechanism and the filament-assisted ohmic conduction mechanism,respectively.And the high and low resistance state of the HfOx/ZnO RRAM device conforms to the Schottky emission mechanism and the Poole-Frank transmission mechanism,respectively.By controlling the HfOx/ZnO RRAM device,we realized the continuous adjustment of the device resistance value,and the RRAM device was modified into an electronic synaptic device successfully.In the experiment,the learning function of the synaptic device was trained by pulse stimulation.The results showed that the synaptic device had similar memory and forgetting characteristics as the biological synapse,and the forgetting process may be derived from the detachment of the trapped charge.By examining the spike timeing dependent plasticity(STDP)of the synaptic device,the typical asymmetric Hebbian learning rule was obtained,in which the weight change(?w>100%)of the synaptic device is comparable or even more than the ?w of biological synapses.In this paper,the electric-heat coupling model of the HfOx/ZnO RRAM device was established by using the COMSOL Multiphysics software.After simulation,the potential(?),current density(J)and temperature(T)of the RRAM device was obtained.The working current and temperature obtained by simulation is consistent with the experimental results,which shows that the model is basically correct,and the oxygen vacancy distribution under the control of the electric and temperature field is the core of the continuous resistance adjustment of RRAM devices.