Dresden 2009 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Downloads | Hilfe
MM: Fachverband Metall- und Materialphysik
MM 35: Poster Session II
MM 35.36: Poster
Mittwoch, 25. März 2009, 16:30–18:30, P4
Studies on the Phase Diagram of Boron Employing a Neural Network Potential — •Tobias Morawietz1, Jörg Behler1, and Michele Parrinello2 — 1Lehrstuhl für Theoretische Chemie, Ruhr-Universität Bochum, D-44780-Bochum, Germany — 2Department of Chemistry and Applied Biosciences, ETH Zürich, USI-Campus, CH-6900 Lugano, Switzerland
The crystalline phases of elemental boron have a structural complexity unique in the periodic table. The complex connection pattern of the icosahedral building blocks forms a formidable challenge for the construction of accurate but efficient potentials.
We present a high-dimensional Neural Network potential for boron, which is based on first-principles calculations and can be systematically improved. The potential is several orders of magnitude faster to evaluate than the underlying density-functional theory calculations and allows to perform long molecular dynamics and metadynamics simulations of large system. By a stepwise refinement of the potential and an application of the potential in metadynamics simulations we show that starting from random atomic positions the structure of α-boron is predicted in agreement with experiment. Further, pressure-induced phase transitions of α-boron are discussed.