Abstract

A machine-learned, DFT -accurate, linear-scaling potential (GAP) has been developed for the ternary material system, Ge-Sb-Te, in which the composition Ge2Sb2Te5, along the GeTe-Sb2Te3 pseudo-binary tie-line, is the canonical material for next-generation phase-change random-access memory (PCRAM) applications. The potential has been used to generate glassy models as large as 24,300 atoms, and ensembles of smaller glassy models, prepared under identical conditions, for a statistical study of the electronic properties, calculated using DFT methods, in particular the prevalence and nature of ‘defect’ states with energies lying in the bandgap of this semiconducting material.

[1] F Mocanu et al, J. Phys. Chem. 122, 8998 (2018) [2] K Konstantinou et al, Nat. Comm. 10, 3065 (2019)