Insights behind multi-level conductance transitions in HfOx memristors
Published in Device Research Conference (DRC), 2024
Abstract: Binary-oxide memristors are being explored as analog computing primitives, as they can be reversibly transitioned between hundreds of nonvolatile conductance states. These states originate from the growth/dissolution of conductive oxygen-deficient nanofilaments. The number of distinguishable states has been increased by controlling the morphology of these filaments through compositional or thermal engineering. While the effect of these design parameters can be directly measured, their influence on filament structure has only been hypothesized, as the mechanisms leading to intermediate conductances are not well understood. Further enhancements of the analog operating regime requires, however, a deep understanding of the atomic-level processes involved. Here, through ab initio materials/device simulations, we resolve the kinetics and electronic current flow behind analog conductance transitions in an ultrascaled TiN-amorphous HfOx-Ti/TiN device. We confirm the valence change effect by correlating spatial current flow with oxygen coordination, and then connect this static picture to the kinetics of ion movement to identify the mechanisms leading to multi-level conductance states.