Ultra-high Tunneling Electroresistance Ratio (2 × 104) & Endurance (108) in Oxide Semiconductor-Hafnia Self-rectifying (1.5 × 103) Ferroelectric Tunnel Junction

In this study, we present a remarkable improvement in the performance of hafnia-based ferroelectric tunnel junctions (FTJs) using oxygen scavenging technology and extremely low-damage (ELD) deposition, leading to a significant increase in the tunneling electroresistance ratio $({\mathrm {TER}}) (\gt 2 \times 10^{4})$, on-current density $(\gt 10^{-2}\mathrm{A} /cm^{2})$, and self-rectifying ratio $({\mathrm {RR}}) (\gt 1.5 \times 10^{3})$. First-principles DFT simulations were also used to evaluate how the asymmetric oxygen vacancy (VO) distribution affected FTJs. As an array-level demonstration of the proposed approach, we experimentally built an FTJ-based XNOR synapse array and verified its operation for binary neural networks (BNN).

Introduction:

A number of significant advantages of the FTJ over conventional technologies have been demonstrated, including their capacity for non-destructive read operations, low power consumption, and fast switching operation. Despite having a lot of merits, FTJs still have some device issues that need to be resolved in order to reach their full potentials, such as low TER values and low on-current levels [1] –[6]. In this study, we have made an impressive improvement in TER and significantly increased on-current density (Fig. 1). ELD deposition process allows us to reduce the disorder in the HZO thin film, producing high ferroelectricity even at relatively thin thickness. By deliberately increasing V_O at one interface, oxygen scavenging technology enhances the asymmetry of FTJ. (Fig. 2) Furthermore, we have examined the effect using first-principles computations. In addition, we built an FTJ-based XNOR synapse array for BNN to provide its array-level demonstration of our suggested approach.