Intrinsic retention statistics in phase change memory (PCM) arrays

Introduction: Recently, phase change memory (PCM) has entered the commercial stage in a 45 nm technology [1]. To better assess the potential scaling and application as embedded memory [2], data retention and its statistics must be carefully understood and optimized. This work studies crystallization statistics in 1 Gb arrays of PCM devices. We evidence (i) retention stabilization by tuning of the programming conditions, and (ii) erratic retention due to crystallization variability. A new retention model is developed, which is capable of predicting cell-to-cell and cycle-to-cycle variability as a function of programming conditions.