Role of metal nanocrystals on the breakdown statistics of flash memory high-κ stacks

Abstract

We present a statistical perspective of the dielectric breakdown characteristics of flash memory devices constituting Au nanocrystallites (NCs) embedded in HfO2 successfully by using a defect generation and clustering model. The significant clustering and apparent high breakdown voltage in the NC sample are well-explained with the present of non-uniform spatial electric field patterns in the vicinity of the NCs. Simulation results for NC-embedded high-κ dielectrics reveal some regions with suppressed and some with enhanced electric field in the dielectric film, with the field magnitude depending on the accumulated charge (Qi) in the NC and the size of NC. From the breakdown tests, a positive correlation between the Weibull slope (β) and the ramp rate (RR) is observed, which is attributed to oxide thinning effect. In addition, temperature dependent current-voltage (I-V) characteristics clearly reveal that the change in E-field in the NC embedded stacks influences the voltage transition point for different conduction mechanisms to occur.