Berlin 2018 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
MM: Fachverband Metall- und Materialphysik
MM 61: Topical Session (Symposium MM): Big Data in Materials Science - Managing and exploiting the raw material of the 21st century
MM 61.2: Talk
Thursday, March 15, 2018, 16:15–16:30, H 0107
Cluster expansions with CELL: a novel python package with a focus on complex alloys — •Santiago Rigamonti1, Maria Troppenz1, Martin Kuban1, Axel Huebner1, Christopher Sutton2, Luca Ghiringhelli2, Matthias Scheffler2, and Claudia Draxl1,2 — 1Humboldt-Universität zu Berlin — 2Fritz-Haber Institut
The cluster expansion (CE) technique allows for obtaining compact models of configuration-dependent properties in alloys with ab initio accuracy, by expanding these properties in terms of clusters (sets of crystal sites). In this work, we present the code CELL [1] which is an object-oriented, modular and user-friendly python package for building accurate CE models. Its focus is on complex surfaces and bulk alloys possessing large parent cells (>30 atoms), for which a full structure enumeration is impossible. We note, though, that CELL is not limited to complex alloys. The creation of training data-sets is incorporated into a user-friendly setup of parent- and supercell objects. For selecting the optimal set of clusters, compressed-sensing schemes, such as LASSO and the split Bregman method, and various cross-validation strategies are available. In addition, finite-temperature properties and the characterization of phase transitions are achieved by applying the Wang-Landau method and diffusive nested sampling. We will demonstrate CELL’s capabilities by selected examples.
[1] S. Rigamonti, M. Troppenz, M. Kuban, A. Huebner, C. Sutton, L. Ghiringhelli, M. Stournara, M. Scheffler, and C. Draxl. CELL: python package for cluster expansions with large parent cells. In preparation.