Abstract
Retention is one of the key reliability metrics for non-volatile memory devices. In oxygen ion transport based resistive switching memory (OxRAM), the retention phenomenon has been well studied from an electrical perspective and physical mechanisms explaining the origin of retention loss have also been speculated to support the observed data. However, the stochastic aspects of retention loss and its variability deserve to be investigated so that the time-dependent shift in the resistance distribution and the retention failure time statistics can be better quantified and estimated for a given set of operating conditions. We propose here a phenomenological Markovian multi-state model combined with the percolation framework and ion diffusion theory to analyze the distributions of retention failure in the low resistance state for OxRAM devices. (C) 2015 Elsevier Ltd. All rights reserved.