Logo image
Statistics of disturb events in OxRAM devices — A phenomenological model

Statistics of disturb events in OxRAM devices — A phenomenological model

The Institute of Electrical and Electronics Engineers, Inc. (IEEE) Conference Proceedings, p.PM-3.1
01/01/2017
Clustering Derivation Devices Electric potential Fittings High resistance Oxygen Percolation Reasoning Reliability Rupture Shrinkage Statistics Switching Vacancies Weibull distribution
Conference Title: 2017 IEEE International Reliability Physics Symposium (IRPS) Conference Start Date: 2017, April 2 Conference End Date: 2017, April 6 Conference Location: Monterey, CA, USA Read disturb (RD) is a key reliability metric for OxRAM devices, more so in the high resistance state (HRS), when small voltages might be sufficient to perturb the oxygen vacancy configuration in the conducting filament that may have undergone “shrinkage” or “rupture” during the RESET sweep. There are very few studies dealing with read disturb phenomena, more so, on its statistical nature. The Weibull distribution seems to be the default choice for time to disturb (tnisr) or voltage to disturb (Vdist) data with no physical reasoning provided. We present here a fundamental derivation of the disturb voltage / time distribution using the percolation cell framework, accounting for the localized electric field distributions and inherent variability in the HRS state vacancy configuration for different switching cycles that play a major role in determining the statistical nature of this phenomenon. The clustering model is shown to be the best choice for fitting the disturb data and this has been verified using several test data published in the literature.
1
Logo image