Classical-nucleation-theory analysis of priming in chalcogenide phase-change memory

Abstract

The chalcogenide Ge2Sb2Te5 (GST) is of interest for use in phase-change memory. Crystallization is the rate-limiting step for memory operation, and can be accelerated by the prior application of a “priming” heating pulse. There is characteristic fading of the priming effect if there is a time interval between the priming pulse and the main heating pulse to achieve crystallization. We apply classical nucleation theory to interpret these effects, based on a fitting of nucleation kinetics (steady-state and transient) over the full temperature range of the supercooled liquid. The input data come from both physical experiments and atomistic simulations. Prior studies of conventional glass-formers such as lithium disilicate preclude any possibility of fading; the present study shows, however, that fading can be expected with the particular thermodynamic parameters relevant for GST and, possibly, other phase-change chalcogenides. We also use the nucleation analysis to highlight the distinction between GST and the other archetypical chalcogenide system (Ag,In)-doped Sb2Te. Classical nucleation theory appears to be applicable to phase-change chalcogenides, and to predict performance consistent with that of actual memory cells. Nucleation modeling may therefore be useful in optimizing materials selection and performance in device applications.