Data for "Modeling the Phase-Change Memory Material, Ge2Sb2Te5, with a Machine-Learned Interatomic Potential"

Change log
Mocanu, FC 
Konstantinou, Konstantinos  ORCID logo

This data set contains:

  1. Trajectories of glass and liquid Ge2Sb2Te5 models generated with the GAP. There are 5 models labeled 001-005 of 315 atoms and one large model of 7,200 atoms labeled 7K.

  2. Structural analysis data including: radial distribution functions, rings, voids, tetrahedral angular order parameters.

  3. The inter-atomic potential used to carry out the simulations.

  4. The crystallization trajectory of one of the GAP models is provided separately due to the size.

Software / Usage instructions
Extended xyz files of the structures were generated with the quippy python library ( from the original LAMMPS ( trajectories. CSV files were generated with the Pandas python library ( and can be read by any library or software that works with this format. The inter-atomic potential xml files were generated with the QUIP teach_sparse program. ( The files are compressed in the common zip format.
GAP, machine-learning, atomic structures, glasses, RDF, rings, voids