Reversing amorphous aging associated with resistance drift in phase-change memory via atomistic rejuvenation

Resistance drift in the amorphous RESET state limits phase-change memory (PCM) for multilevel storage and neuromorphic computing. Most mitigation strategies slow the drift rather than reverse the structural aging that develops after programming. Here, we use large-scale molecular dynamics simulations with a Y–Sb–Te neuroevolution potential to examine atomistic rejuvenation in Sb–Te-based glasses. Aging and rejuvenation are quantified using potential energy, radial distribution functions, global Steinhardt order parameters, ring statistics, and homopolar bonding, which track short-range relaxation, topological ordering, and bond rearrangement. Binary Sb–Te models first establish competing thermal pathways: low-temperature annealing promotes energetic relaxation and ordering, whereas high-temperature reheating followed by requenching restores higher-energy, less-aged amorphous states while avoiding prolonged residence in the crystallization-prone window. In Y-doped Sb2Te3, Y reduces energetic relaxation, delays four-membered-ring ordering, and reduces bonding fluctuations while preserving reversible rejuvenation. Device-scale simulations further show that rejuvenation pulses reset key aging-sensitive descriptors in a 6.5 × 105-atom mushroom-cell PCM model. These results demonstrate that atomistic rejuvenation can reverse aging-sensitive structural features in Sb–Te-based PCM glasses, providing a structural route toward mitigating resistance drift.