Stability and Aging of Phase Change Materials : An Ab Initio Perspective (Invited)

Data recording with Phase Change Materials is a much studied topic as the writing/erasing characteristics, cyclability and downscaling properties of these materials allow for efficient data storage in future generations of devices. Nevertheless, some aspects of phase change materials are limiting their performances and delaying their wider technological application.
First, aging phenomena are common to all amorphous structures, but of special importance PCMs since it impedes the realization of multi-level memories. Different interpretations have been proposed, but we focus here on the structural relaxation of amorphous GeTe, chosen because it is the simplest system that is representative of the wider class of GST alloys, lying along the GeTe-Sb2Te3 composition line of the GeSbTe phase diagram. One difficulty encountered in the simulation of these amorphous systems is that the direct generation of an amorphous structure by quenching a liquid using Density Functional Theory (DFT) based Molecular Dynamics leads to one sample with a small number of atoms, and, hence of small number of atomic environments. Here we sample a large number of local atomic environments, corresponding to different bonding schemes, by chemically substituting different alloys, selected to favor different local atomic structures. This enables spanning a larger fraction of the configuration space relevant to aging. Our results support a model of the amorphous phase and its time evolution that involves an evolution of the local (chemical) order towards that of the crystal. On the other hand its electronic properties drift away from those of the crystal, driven by an increase of the Peierls-like distortion of the local environments in the amorphous, as compared to the crystal [1].
A second problem faced by PCMs is the fact that data recording is limited at high temperature due to the increased propensity to recrystallize. One approach to counter this is to stabilize the PCM using impurity atoms such as C or N. Using DFT and the analysis of the mechanical properties (constraints theory), we demonstrate how these impurity atoms modify the rigidity of the network, which is experimentally correlated with the activation energy for crystallization [2].
Finally, the crystal phase itself has been shown to have variable conductivities depending on the thermal history and annealing conditions. If this could be used profitably for multi-level recording, it also indicates that the crystal is undergoes some temporal evolution. Using DFT, we clarify the stability behavior of GST crystal and show that the metal-insulator transition is driven by the migration of intrinsic vacancies and an Anderson localization transition [3].
[1] J.Y Raty, W. Zhang, J. Luckas, C. Chen, R. Mazzarello, C. Bichara and M. Wuttig, Nat. Comm. (2015)
[2] G. Ghezzi, J.Y. Raty, S. Maitrejean, A. Roule, E. Elkaim and F. Hippert, Applied Physics Letters, 99 (2011) 151906
[3] W. Zhang, A. Thiess, P. Zalden, R. Zeller, P. H. Dederichs, J-Y. Raty, M.Wuttig, S. Blügel et R. Mazzarello, Nature Materials 11 (2012) 952