Neural Network interatomic potentials for phase change materials

Abstract

Phase change materials are attracting an increasing interest worldwide for applications in Phase Change non volatile Memories (PCM). A PCM is essentially a resistor of a thin film of a chalcogenide material (typically Ge₂Sb₂Te₅, GST) with a low field resistance which changes by several orders of magnitude depending on the state of GST, metallic in the crystalline form and insulating in the amorphous phase. Programming the memory requires a relatively large current to heat up the GST and induce reversibly the phase change, either the melting of the crystal and subsequent amorphization or the recrystallization of the amorphous. A very attractive option to reduce the programming current, which is still an issue for future scaled PCM technologies, involves the change of the device geometry with the use of chalcogenide nanowires (GST or GeTe). In this project we plan to perform atomistic simulations of the phase change dynamics in GeTe to assess the dependence of the melting/crystallization temperature on the size/shape of the nanoparticles. To this aim we will develope empirical interatomic potentials with ab-initio accuracy by fitting large ab-initio databases within a novel neural network (NN) scheme we have recently validated with the simulations of the phase diagram of silicon (Behler et al, Phys. Rev. Lett. 100, 185501 (2008)).