1303.2802 (Zhong-Li Liu)

MUSE: Multi-algorithm collaborative crystal structure prediction    [PDF]

Zhong-Li Liu

The multi-algorithm collaborative crystal structure prediction was realized and constructed as a Multi-algorithm-collaborative Universal Structure-prediction Environment (MUSE). The evolutionary algorithm was coupled with the simulated annealing and the basin hopping algorithms to efficiently find the stable and metastable structures of materials under certain conditions. After introduced two new operators, slip and twist, the diversity of structures are exceedingly enhanced. In particular, in order to achieve the self-adaptive evolution of structures, MUSE uses the competition techniques of ten variation operators to increase the diversity of crystal structures. The multi-algorithm coupling, the ten variation operators, the symmetry constraints in the first generation and the self-adaptive algorithms are all key techniques to improve the search efficiency of MUSE. To show the search ability of MUSE and how it works, I presented some predicted systems including metallic, covalent and ionic systems. All the test results show MUSE has high efficiency and almost 100% success rate.

View original: http://arxiv.org/abs/1303.2802