Design and modeling of niobium oxide-tantalum oxide based self-selective memristor for large-scale crossbar memory

Abstract

Memristor-based crossbar architecture has emerged as a promising candidate for 3-D memory, logic, and neuromorphic computing system as it offers remarkably high integration density, low power consumption, fast operation, and easy integration with CMOS technology. However, the fundamental obstacle for their development is the sneak current, which causes misreading and write-crosstalk. In this regard, we present the TiN/NbO2/TiN/TaOx/TiN based self-selective memristor by combining the threshold switching properties of niobium oxide (NbO2) and memory switching properties of tantalum oxide (TaOx) in a single device. The performance investigation is carried out using the finite element simulation method, based on self-consistent solutions of joule heating equation, drift-diffusion continuity equation, and current continuity for accurately capturing the temperature and field-dependent transport of vacancies. The results reveal that NbO2-TaOx based self-selective memristor can allow significantly lower OFF current (1.22 μA), higher read window (32.6), and higher non-linearity (141) than that of TiN/TaOx/TiN based memristor. We demonstrate that the self-selective memristor exhibits good speed with the operation time constant of 70 ns. Furthermore, the crossbar array using a self-selective memristor has shown excellent performance with an improved readout margin up to 27 word lines. Our material-to-circuit performance analysis promises a reliable and energy-efficient crossbar array using NbO2-TaOx cell that can be further utilized for implementing 3-D cross-bar array.