Regensburg 2013 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
MM: Fachverband Metall- und Materialphysik
MM 59: Computational Materials Modelling - Phase Stability III
MM 59.4: Talk
Thursday, March 14, 2013, 16:30–16:45, H24
Structure optimization via “local heat pulse”-quench cycles — •Arnulf Möbius1 and Christian Schön2 — 1Institute for Theoretical Solid State Physics, IFW, Dresden — 2Max Planck Institute for Solid State Research, Stuttgart
Structure prediction for crystals with a large number of atoms in the primitive cell, as well as for clusters of many atoms, are often impeded by the corresponding energy landscape exhibiting a huge number of local minima. For such tasks, we present an optimization procedure which is based on "local heat pulse"-quench cycles (LHPQC). It was originally developed for combinatorial optimization tasks [1].
This approach is applied to a lattice structure prediction problem. In that, we use the general utility lattice program (GULP) by J.D. Gale and co-workers [2] as local search code. As a test case, the energy landscape of the Mg10Al4Ge2Si8O36 lattice is investigated, where the interactions are modelled by Coulomb, Buckingham, and three-body potentials [3], and where the cell parameters are free to vary.
The results of our computer experiments testify that the LHPQC procedure is robust and far more efficient than the previous approaches to the same test problem in Ref. 3. Finally, we show how our procedure can be easily parallelized: Its efficiency is considerably improved by treating an ensemble of local minima instead of a single one.
[1] A. Möbius, A. Neklioudov, et al., Phys. Rev. Lett. 79 (1997) 4297.
[2] J.D. Gale and A.L Rohl, Mol. Simul. 29 (2003) 291.
[3] A.R. Oganov, J.C. Schön, et al., in "Modern methods of Crystal Structure Prediction", ed. A.R. Organov, (Wiley, 2011), p. 223.