Computational crystal structure prediction with high-through-put Ab initio and data mining methods

Crystal structure prediction is an essential step in rational materials design. Unfortunately, there is no general tool for reliably predicting crystal structures of new alloys. Total energy ab initio approaches can be used to accurately compare energies of different candidate structures, but developing a manageable list of candidate structures for comparison is still very challenging. A powerful new tool to tackle this problem is "high-throughput" ab initio computation, which makes use of robust automated techniques to perform many thousands of calculations. High-throughput ab initio can be enhanced with data mining techniques,which can be used to accelerate structure prediction in new alloys. We have used high-throughput methods to calculate over 14,000 full ab initio structural optimizations on 80 intermetallic binary alloys, and implemented a novel data mining scheme that shows potential to dramatically reduce the time necessary for identify stable cry structures in new alloys.

Duke Authors

Cited Authors

  • Morgan, D; Ceder, G; Curtarolo, S

Published Date

  • 2004

Published In

  • JOM

Volume / Issue

  • 56 / 11

Start / End Page

  • 70 -