Regensburg 2016 – wissenschaftliches Programm

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

MM: Fachverband Metall- und Materialphysik

MM 55: Methods in Computational Materials Modelling III: Machine learning and statistics

MM 55.3: Vortrag

Donnerstag, 10. März 2016, 12:15–12:30, H53

Automated Convergence Checks with the Python Based Library PyIron — •Jan Janßen, Tilmann Hickel, and Joerg Neugebauer — Max-Planck-Institut für Eisenforschung GmbH, Düsseldorf, Germany

Over the last years methodological and computational progress in atomistic simulations has substantially improved the predictive power in materials design. To guarantee the desired numerical precision, simulation tools that run the necessary complex simulation protocols and convergence checks automatically are critical but scarce. Based on our recently developed PyIron library which provides an efficient framework to implement such protocols we have developed an automatized tool for fitting the Birch-Murnaghan equation of states with prescribed accuracy targets. The tool includes an approach to handle the intrinsic energy fluctuation in DFT simulations. By making a Monte-Carlo based sensitivity analysis that includes the intrinsic model errors as well as the energy fluctuations we determine the numerically most efficient volume range for the approximation. The achievable accuracy as well as the performance gains of such an approach will be discussed.