Stochastic failure model for endurance degradation in vacancy modulated HfOx RRAM using the percolation cell framework

Endurance is a key performance metric for non-volatile memory devices. For resistive switching random access memory (RRAM) technology, which operates based on the reversible drift of oxygen ions in a bipolar switching scheme, the degradation in endurance can be stochastically modeled by using the percolation cell framework, as the SET and RESET events are akin to the dielectric breakdown and recovery phenomena in logic gate stacks. Based on our understanding of the physical mechanisms postulated for endurance degradation and the electrical test results on the memory window closure, we propose a generic statistical model in this study that can be used to simulate the endurance cycle distribution as a function of the pulse voltage, pulse duration, compliance level, activation energy, dielectric thickness, filament temperature and filament size / shape. Fitting the endurance model to real test data helps in estimating the value of the activation energy for oxygen ion transport. Our focus here is only on the “failure to RESET” mechanism, whereby the filament does not rupture to reach the high resistance state due to a large imbalance in the oxygen ion and oxygen vacancy count.