High-speed energy-efficient memristor confined in sub-5 nm space with elemental oxygen reservoir layer.

Li, Chenfei; Niu, Wencheng; Wan, Da; Tang, Lin; Xie, Zhengdao; Zhang, Kai; Liu, Yuan; Liu, Qi et al. · Nat Commun · 2026

basic_science · Level V

Where this comes from

Abstract

Random migration of oxygen vacancies (V<sub>O</sub>) leads to unpredictable formation and rupture of conductive filaments (CFs) in oxide-based memristors. In this work, an atomically flat 4.5 nm hafnium oxide (HfO<sub>x</sub>) switching layer and a 3.5 nm elemental oxygen reservoir (EOR) layer are confined between two-dimensional HfS<sub>2</sub> and MoS<sub>2</sub> layers, ensuring a homogeneous electric field distribution. The migration and redistribution of V<sub>O</sub> within the ultrathin HfO<sub>x</sub> switching layer enable the memristive behavior of the device. The EOR-based memristors achieve high set/reset transition speeds of 8 ns and 15 ns, respectively. The electroneutral EOR layer interacts with V<sub>O</sub> in the HfO<sub>x</sub> switching layer and, together with the HfO<sub>x</sub> tunnel layer above the HfS<sub>2</sub>, forms a barrier to suppress the high-resistance state current. Reliable endurance up to 10<sup>5</sup> cycles, and long retention up to 10<sup>5 </sup>s are simultaneously obtained. Finally, a high recognition accuracy of 97.0% is achieved, demonstrating potential for low-power neuromorphic computing applications.