Reset-induced variability of retention characteristics in phase change memory (PCM)

Data retention in phase change memory (PCM) is limited by the metastable nature of the amorphous phase, which transforms into the stable crystalline phase at elevated temperatures. A deeper understanding of PCM retention statistics, carried out through an extensive statistical characterization at array level and a physical-based model, is necessary for proper failure testing and prediction of PCM reliability. This work addresses retention statistics both at array level (cell-to-cell statistics) and at single-cell level (cycle-to-cycle statistics), for which we provide evidences for analog and digital random fluctuations in the retention characteristics. We present a physical-based Monte-Carlo model, based on a gaussian spread in the activation energy, which is capable of explaining both the cell-to-cell and the cycle-to-cycle statistics.