Mechanistic Analysis of Oxygen Vacancy-Driven Conductive Filament Formation in Resistive Random Access Memory...

Electrically switchable resistive random access memories have drawn much interest as nonvolatile memory device candidates based on metal-insulator-metal (MIM) structure concepts. However, atomic level mechanisms that lead to conductive filament (CF) formation in MIM structures are often lacking, such as for the system with NiO as the oxide layer, which was found promising for resistive random access memory (RRAM) device applications. In this work, using density functional theory with a Hubbard-type on-site Coulomb correction, which we carefully benchmarked, we analyzed the intrinsic propensity toward CF formation in NiO upon introduction of oxygen vacancies, including interfacial effects of Ag or Pt electrodes. First, for stoichiometric MIM structural models, contributions from metal-induced gap states to the electronic density of states (DOS) were identified, accommodating oxygen vacancy states and showing that the interface region is reduced more easily than the bulklike region, for example, for the Ag/NiO/Ag structure. Moreover, a tendency toward oxygen vacancy clustering was demonstrated, important for CF formation. Indeed, by introducing ordered oxygen vacancies into the oxide layer for both MIM models, several extended defect states within the forbidden gap have resulted, which lead to defect-assisted transport. These were shown to be influenced by the spatial distribution and number of oxygen vacancies in the filament, where the degree of reduction of Ni atoms changes based on the immediate surroundings. Projected electronic DOS for individual Ni atoms in regions near and away from oxygen vacancies indicated that those Ni close to oxygen vacancies contribute most to the conductivity. Interestingly, based on charge analyses, these atoms are revealed to undergo significant reduction, generating a locally conductive region in the oxide layer that consists of metallic/near-metallic Ni (Ni⁰), formed through local reduction.