Self-Rectifying Short-Term Memory Phenomena Through Integration of TiOx Oxygen Reservoir and Al2O3 Barrier Layers for Neuromorphic System

Abstract

In this study, a tri-layer Pt/Al/TiOx/HfOx/Al2O3/Pt memristor device is fabricated and analyze its electrical characteristics for reservoir computing and neuromorphic systems applications. This device incorporates an oxygen reservoir of a TiOx and a barrier layer of an Al2O3, enabling stable bipolar switching characteristics without the need for an electroforming process over 10(3) cycles. It also exhibits self-rectifying properties under a negative bias. Based on these characteristics, it is investigated essential synaptic functions such as long-term potentiation (LTP), long-term depression (LTD), paired-pulse facilitation (PPF), spike-rate-dependent plasticity (SRDP), spike-duration-dependent plasticity (SDDP), spike-number-dependent plasticity (SNDP), and spike-amplitude-dependent plasticity (SADP), to assess their suitability for neuromorphic applications that mimic biological synapses. Furthermore, utilizing the short-term memory characteristics of the device, reservoir computing (RC) measurement from [0000] to [1111] in 4-bit representation is conducted. This capability enables us to achieve a high accuracy of 95.5% in MNIST pattern recognition tasks. Lastly, the natural decay characteristics caused by oxygen ion migration in the device, examining the transition from short-term to long-term memory in image memorization tasks is explored. The potential for deployment in high-density crossbar arrays by calculating the read margin based on the device I-V curve and programming scheme is also evaluated.