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

MM: Fachverband Metall- und Materialphysik

MM 4: Methods in Computational Materials Modelling (methodological aspects, numerics)

MM 4.3: Vortrag

Montag, 1. April 2019, 10:45–11:00, H45

Automated error analysis and control for ab initio calculations — •Jan Janssen, Tilmann Hickel, and Joerg Neugebauer — Max-Planck-Institut für Eisenforschung, Düsseldorf, Germany

Over the last years methodological and computational progress in atomistic simulations have substantially improved the predictive power in materials design. A critical prerequisite to ensure a reliable comparison between the ab initio computed data with experimental data is to quantify the various sources of uncertainty present in the ab initio calculations. These include systematical errors due to insufficient convergence, statistical or numerical errors due to incomplete sampling and model errors for derived quantities. A well-known example is the determination of the equilibrium lattice constant and bulk modulus, which requires a careful analysis of the fit of the ab initio data on an approximate analytic form such as the Murnaghan equation of state.

To automatize the complex analysis we have developed an algorithm which takes the precision in the derived quantity as a convergence goal and automatically determines the convergence parameter to achieve it. This algorithm is implemented using pyiron (http://pyiron.org) - an integrated development environment (IDE) for computational material science. This tool provides an efficient and user friendly environment to implement complex simulation protocols and allows to run them as high-throughput simulations over the periodic table. Our investigations revealed that many of the commonly used rules of the thumb for fitting ground state materials properties become invalid for high precision calculations.

100% | Bildschirmansicht | English Version | Kontakt/Impressum/Datenschutz
DPG-Physik > DPG-Verhandlungen > 2019 > Regensburg