based memristive devices revealed by density functional theory*

Memristive devices are potential candidates for neuromorphic computing owing to their high speed, low power consumption and high endurance. However, the stochastic nature of conductive filament formation hampers the device accuracy and reliability during operation. Here, we propose using NaNbO₃ (NNO) synthesized under Na-deficient condition as the resistive layer in memristive devices. This method results in the self-assembly of nanopillars that serve as conduction channels. Particularly, our density functional theory (DFT) calculations revealed that there exists a unique O-vacancy site in the planar fault (PF) with much lower formation energy. These low energy O-vacancies introduce defect states that n-dope the system, thereby enhancing conductivity. Importantly, our calculations suggest that there are no energy barriers to hinder the diffusion of O vacancies to these energetically favorable vacancy sites in the PF, thus facilitating the switching process vital for memristive devices.