Ferroelectric synaptic devices based on CMOS-compatible HfAlOx for neuromorphic and reservoir computing applications

Abstract

The hafnium oxide-based ferroelectric tunnel junction (FTJ) has been actively researched because of desirable advantages such as low power and CMOS compatibility to operate as a memristor. In the case of HfAlO_x (HAO), the remanent polarization (P_r) value is high and the atomic radius of Al is smaller than that of Hf; therefore, ferroelectricity can be better induced without mechanical force. In this paper, we propose an FTJ using HAO as a ferroelectric layer through electrical analysis and experiments; further, we experimentally demonstrate its capability as a synaptic device. Moreover, we evaluate the maximum 2P_r and TER value of the device according to the difference in conditions of thickness and cell area. The optimized device conditions are analyzed, and a large value of 2P_r (>∼43 μC cm⁻²) is obtained. Furthermore, we show that paired-pulse facilitation, paired-pulse depression, and spike-timing-dependent plasticity can be utilized in HAO-based FTJs. In addition, this study demonstrates the use of an FTJ device as a physical reservoir to implement reservoir computing. Through a series of processes, the synaptic properties of FTJs are verified for the feasibility of their implementation as an artificial synaptic device.