Variability in Al2O3-based resistive memories, a kinetic Monte Carlo approach

We propose a kinetic Monte Carlo (kMC) model to analyse resistive switching and variability in filamentary Al 2 O 3 -based resistive memories (RRAM) built on TiN/Ti/Al 2 O 3 /W stacks. The simulator describes the formation and rupture of conductive filaments driven by oxygen vacancy generation and recombination, incorporating electric field and thermal calculations, including Joule heating and a steady-state heat conduction scheme to capture local temperature profiles. The model is tuned against extensive experimental data obtained under ramped voltage stress, and it reproduces typical current–voltage characteristics in forming, set, and reset processes using Poole–Frenkel conduction in the high-resistance state and an ohmic approach in the low-resistance state. We then use the calibrated kMC tool to study cycle-to-cycle variability in set and reset events, combining advanced numerical extraction of switching voltages and currents with one- and two-dimensional statistical descriptors, such as coefficients of variation and cumulative distribution functions. Finally, we investigate the impact of oxygen vacancy density inside conductive filaments in the reset process, showing that the reset voltage exhibits a non-monotonic dependence on defect concentration due to changes in filament connectivity and Joule heating, which has direct implications for reliability and variability-aware design of filamentary memristors.