Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

O: Fachverband Oberflächenphysik

O 67: Frontiers of Electronic Structure Theory: Focus on Artificial Intelligence Applied to Real Materials 3

O 67.7: Vortrag

Donnerstag, 8. September 2022, 12:00–12:15, S054

Similarity-of-materials analysis for reusability and interoperability of data in materials databases — •Šimon Gabaj, Martin Kuban, Santiago Rigamonti, and Claudia Draxl — Humboldt-Universität zu Berlin, Zum Großen Windkanal 2, 12489 Berlin, Germany

Large computational materials databases, such as NOMAD [1-2], make it possible to reuse already generated materials data. All these data typically come from different sources and have been created for a different purpose, be it geometry optimization, electronic structure, or alike. To support interoperability and thus reusability of such data, we devise a data-analysis workflow making use of similarity fingerprints. First, we encode the electronic density of states (DOS) in a vectorial representation [3] to obtain a descriptor. Then, we employ the Tanimoto coefficient to compute the similarity between all pairs of calculations. We demonstrate our workflow with selected materials, chosen from the NOMAD database. In the ideal case, all calculations of the same material in the same geometry should be identical. This is, however, not observed. Using our approach, we can uncover correlations between the DOS similarity and methodology as well as computational parameters. This way, we can also identify parameters that are relevant for the convergence of results.

[1] Draxl, C., Scheffler, M., MRS Bulletin, 43, 676, (2018)

[2] Draxl, C., Scheffler, M., J. Phys. Mater., 2, 036001, (2018)

[3] Kuban, M., et al., to be published