Research on electronic synaptic simulation of HfO2-based memristor by embedding Al2O3

Abstract

Abstract The potential of neuromorphic computing in synaptic simulation has led to a renewed interest in memristor. However, the demand for multilevel resistive switching with high reliability and low power consumption is still a great resistance in this application. In this work, the electronic synaptic plasticity and simulated bipolar switching behavior of Pt/Al2O3(2 nm)/HfO2 (10 nm)/Al2O3 (2 nm)/Ti tri-layer memristor is investigated. The effect of Al2O3 layer embedded at the top electrode and the bottom electrode on the resistive performance of the memristor was studied. It is found that both of them can effectively improve the reliability of the device (104 cycles), the resistive window (>103), the tunable synaptic linearity and reduce of the operating voltage. RRAM with Al2O3 embedded at the top electrode have higher uniformity and LTP linearity, while those with Al2O3 embedded at the bottom electrode significantly reduce the operating current (∼10 μA) and improve LTD linearity. Electron transport mechanisms were compared between single-layer HfO2 and tri-layer Al2O3/HfO2/Al2O3 samples under DC scanning. The results showed that the thin Al2O3 layer at the top electrode led to Fowler Northeim tunneling in the low-resistance state, while the thin Al2O3 layer at the bottom electrode led to Schottky emission in the high-resistance state. The Al2O3/HfO2/Al2O3 memristors were successfully used to achieve synaptic properties, including enhancement, inhibition, and spike time-dependent plasticity, demonstrating an important role in high-performance neuromorphic computing applications.