Monte Carlo model of reset stochastics and failure rate estimation of read disturb mechanism in HfOx RRAM

Read disturb is a key failure mechanism in ultra-thin dielectric based HfO_x RRAM devices. While the physical origin of read disturb has been attributed to vacancy perturbations in the conductive filament due to high localized field around the filament constriction or at the dielectric tunnel barrier, there is no statistical model based on physical principles that enables quantification of the disturb failure probability (F_DIST) and the associated variability in the disturb voltage (V_DIST) for a given set of device operating conditions. This study develops a percolation based framework to model the filament configuration in the low resistance state (LRS) and a Kinetic Monte Carlo (KMC) routine to simulate the gradual evolution of the filament during the reset phase and identify the first voltage at which the disturb phenomenon is observed in the post-reset high resistance state (HRS). The model enables us to explain the bimodal nature of filament configuration after reset, wherein the filament may have ruptured (resulting in an oxide tunnel barrier) or it could have undergone shrinkage staying in the Quantum Point Contact (QPC) regime with a constriction that governs its stability. Two key external parameters, current compliance (I_comp) and voltage sweep ramp rate (RR) have been identified and their impact on the disturb failure distribution and reset kinetics has been examined in detail.