Switching in Atomic Memristors—The Role of Defects and Interfaces

Memristors utilize a phenomenon known as resistive switching where the electrical resistance of a material can be modulated between a high-resistance state and a low-resistance state with each able to be maintained with zero-power supply provoking strong interest for memory and neuromorphic computing applications. Memristors based on 2D materials are especially interesting as they promise the enabling of low power, high density, and scalable computinng technology. The vertical metal-insulator-metal (MIM) device architecture is thought to be operated on the migration of metal atoms into intrinsic vacancies of the atomic layer. We previously reported the first observation of this phenomenon with 2D materials in a vertical MIM device configuration and uncovered the one-to-one correlation between gold atom substitution into sulfur vacancies within nonvolatile changes in resistance within a vertical Au/MoS₂/Au memristor [1][2]. Subsequently we developed a dissociation-diffusion-adsorption model to describe the kinetics of resistive switching, and in this work, we expand the model to further describe the role of defects in the 2D material and include the role of imperfections in the electrode surface [2]. We study the Ag/hBN/Ag vertical atomic memristor with scanning tunnelling microscopy (STM) for imaging, spectroscopy, and transport measurements to characterize the defect sites of hBN in addition to the underlying Ag surface and emulate an MIM structured device at an atomic scale. We see that changes in the defect states cause critical differences in the switching performance of monolayer hBN based memory devices as the switching voltage and power consumption are tied to the local defect structures in the hBN layer. Furthermore, the defect sites on the Ag electrode including grain boundaries, atomic steps and individual vacancies act as low-energy dislocation sites for Ag atom dissociation. The insights gained from this work can be used to extend the functional behavior of atomic memristive devices in future memory or neuromorphic computing applications. [1] R. Ge et al., <i>Nano Lett.</i>, 18, 434-441, <b>2018</b> [2] S. Hus et al., <i>Nat. Nanotechnol</i>, 16, 58-62, <b>2021</b> [3] R. Ge et al., <i>Adv. Mater</i>. , 33, 7, <b>2020</b>