Publication date: 15th April 2025
Memristors have attracted significant research interest due to their potential for non-volatile random-access memory and artificial synapses in neuromorphic computing applications. However, achieving all the desired memristive and bio-synaptic characteristics in a single device is challenging due to some unavoidable intrinsic shortcomings in diverse classes of materials under consideration. To improve these properties various strategies such as chemical doping, nanostructure growth, heterostructure formation of active materials have reported to be highly effective [1,2]. In this report, we demonstrate an enhanced memory and synaptic properties of WO3-x-based memristors by inserting a Cu-doped ZnO (Cu:ZnO) layer between the substrate and WO3-x layer. The results reveal that the bilayer devices exhibit more excellent memristive performances in terms of memory window (on/off ratio), retention time, endurance, and variations in operating voltages compared to the single-layered devices. In addition, the bilayer devices show distinguishable non-volatile multi-level resistance states, making it suitable for high-density multi-bit memory storage. Most importantly, the devices exhibit improved bio-synaptic functionalities such as long-term potentiation/depression (LTP/LTD) and spike-timing-dependent plasticity, compared to their single-layer counterparts, as shown in Fig. (a). The conduction mechanisms of high resistance state (HRS) and low resistance states (LRS) can be explained by trap-assisted Poole-Frenkel emission and trap-free space charge limited current conduction mechanisms. The origin of the improved resistive switching (RS) is investigated through electrochemical impedance spectroscopic (EIS) measurements in both the HRS and LRS, as shown in Fig. (b). The findings disclose that the Cu:ZnO layer acts as a virtual electrode during the RS process, which facilitates a more controlled formation and rupture of conducting filaments through the migration of oxygen vacancies and Ag ions. The results indicate that heterostructure-based memristors potentially pave the way for next-generation in-memory neuromorphic computing applications.
【Reference】[1] Li et al., J. Mater. Chem. C, 2020, 8, 16295--16317. [2] Ismail et al., ACS Materials Lett. 2023, 5, 11, 3080–3092.