Statistical Modeling and Understanding of HRS Retention in 2.5 Mb HfO2 based ReRAM

Despite observed fluctuations in the read current of individual ReRAM devices, the current distribution over a large set of devices is typically stable at room temperature, and follows log-normal statistics (LNS). Hence, to capture the retention process in ReRAM, we should not follow the resistance changes of individual devices, but investigate the evolution of these current distribution statistics in time. In this work, for the first time, we propose such a statistical retention model that explains the observed changes of HRS read current distributions of a 2.5 Mb HfO
2
based ReRAM as function of baking temperature and time. The model is based on fluctuations of oxygen vacancies in the conductive filament, accounting for both the shift and tilt of the experimental distributions. Furthermore, a limiting effect at higher read current is observed and explained by the limited number of possible oxygen vacancy configurations.