ReRAM Switching Performance based on Single-Walled Carbon Nanotubes Wire Electrode

Abstract

From the past, the occurrence of bottlenecks in memory/computation functions has been a problem for Von Neumann architecture. As the demand for computing and data storage increases, it is necessary to develop high-performance memory to overcome the limitations of operating speed and low power consumption. Recently, neuromorphic systems have been in the spotlight with computing technology that imitates the human brain. Neuromorphic architecture can be an important factor in implementing spike neural network (SNN) – based system hardware with human synaptic learning. Among the candidates for neuromorphic memory systems, resistive random-access memory (ReRAM) is attracting attention for its simple structure and easy design with silicon-based CMOS technology. By spike timing dependent plasticity (STDP), known as the learning rule of synapses, the change in the weight of synapses is equal to the change in the resistance of the synaptic ReRAM device.Therefore, this study proposed a nano-wire electrode based non-volatile synaptic devices with metal-CNTs-oxide-Si (MCOS) structure by synthesizing single-walled carbon nanotubes (SWCNTs). SWCNTs have excellent electrical properties as they include higher conductivity than copper and have carrier mobility than Si. As the top electrode wire was applied as SWCNTs, the leakage current was reduced, and the high switching performance was shown through self-integration of each cell. Ultimately, it will be an important parameter of device implementation that improves the reliability of memory and the stability of conductance filament (CF).