Berlin 2018 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
MM: Fachverband Metall- und Materialphysik
MM 62: Methods in Computational Materials Modelling (methodological aspects, numerics)
MM 62.4: Vortrag
Donnerstag, 15. März 2018, 16:30–16:45, TC 006
DFTB for lazy theorists: automatic transferable parametrization with particle swarm optimization — •Chiara Panosetti1, Reinhard J. Maurer2, and Karsten Reuter1 — 1Technische Universität München — 2University of Warwick
In terms of numerical efficiency and accuracy, tight-binding is intermediate between first-principles and empirical methods. Among the many flavors used in materials modelling, Self-Consistent Charge Density Functional Tight Binding (SCC-DFTB)[1] is particularly geared to cover a broad range of bonding situations while retaining access to electronic properties. This renders it appealing for use, e.g., as a lower-level pre-screening technique in hierarchical approaches. However, its pairwise parametrization requires a procedure which constitutes an N2 effort within the periodic table and is not only tedious, but also rather arbitrary and difficult to document and reproduce.
Building on the work of Chou et al.,[2] we here propose, extend and critically discuss a new parametrization strategy where all the quantities determining the SCC-DFTB interaction are optimized with a particle swarm algorithm to minimize a generalized observable-based cost function. Careful use of such a powerful tool should facilitate reproducibility and provide a straightforward recipe to fine tune the balance between transferability and accuracy.
[1] M. Elstner, Theor. Chem. Acc. 116, 316 (2006).
[2] C. P. Chou et al., J. Chem. Theory Comput. 12, 53 (2016).