Semimetal Sb Based ReRAM Devices And Its Applications In Artificial Synapse

With the advent of the era of big data,the amount of information increases exponentially,requiring more advanced and efficient information processing systems and storage units.The currently adopted von Neumann computing architecture causes a von Neumann bottleneck due to the impact of the data transmission bandwidth between the CPU and the memory.This gap is becoming more and more obvious in the era of big data,which seriously limits the further improvement of information processing speed and efficiency.At the same time,the silicon-based memory in the storage system is faced with the severe limitation of Moore's Law reaching the limit,and can no longer meet the higher storage requirements in the era of big data.Therefore,research on new types of memory and more efficient computing architectures becomes the key to this field.Resistive memory,as a representative of new memory technology,is becoming a strong candidate to solve this problem.Conductive bridge random access memory(CBRAM),also known as electrochemical metallized memory(ECM),has been widely used in resistive memory and the realization of artificial synapses.Although the traditional CBRAM using metal as anode material has high scalability and large dynamic range,as an artificial synapse,the sudden change of conductance during the formation of metal conductive filaments makes CBRAM based on metal ion transport only limited weight states.quantity and low linearity.In addition,the high conductance of metal conductive filaments greatly increases the power consumption of the device,and these problems severely limit the application of CBRAM devices as artificial synapses for neural computing networks.Metalloids have low single-atom conductance due to their low density of states or semiconducting properties,making them potential to realize low-power artificial synapses.Based on this,the main research contents of this paper are as follows:(1)Research on the resistance-switching performance of metalloid anode CBRAM.The implementation of memristive synaptic devices in neuromorphic computing is hindered by a limited number of weighted states and power consumption.Conventional metal-based memristors are particularly susceptible to temporal and spatial variations in set voltage and resistance states due to the random formation of metal filaments.Here,we report a metalloid Sb-based CBRAM device that exhibits the advantages of low operating current,consistent settling voltage,long retention time,and multi-level resistive states.Furthermore,first-principles calculations and characterization results reveal the migration of Sb⁺(2)Metalloid anode CBRAM artificial synaptic plasticity research.Synaptic plasticity is considered to be the basis of human brain learning and memory,and the simulation of synaptic plasticity is conducive to the construction of neural networks with memristive devices as electronic synapses for neuromorphic computing.In this chapter,the W/Al Sb/Si O₂/W artificial synaptic device was simulated for synaptic plasticity,and it was found that it successfully realized the basic characteristics of synaptic plasticity such as PPF and STP,and successfully realized that the conductance(synaptic weight)adjustment linearity was good.The LTP and LTD behaviors of CBRAM-based artificial synapses effectively solve the problems of small number of weights and poor linearity caused by metal filaments.