Statistics of Resistance Drift Due to Structural Relaxation in Phase-Change Memory Arrays

The phase-change memory (PCM), based on the reversible phase transition in a chalcogenide material, is among the most attractive memory concepts for next-generation nonvolatile memories. Due to the metastable nature of the amorphous state, the memory can exhibit a time variation of resistance after programming as a result of two main mechanisms: 1) structural relaxation (SR), which is an atomic rearrangement to minimize the defect density, and 2) crystallization of the amorphous chalcogenide. SR has been mostly studied at the single-cell level, whereas a statistical analysis and modeling is necessary for device reliability estimation and prediction. This work studies the statistical behavior of SR in PCM devices, through experimental and modeling approaches. Statistical SR data from PCM arrays are shown, and a Monte Carlo model for SR statistics is proposed, based on previous physical modeling of the SR process. This model allows for long-term, physics-based, and array-level reliability extrapolations in large PCM arrays.