Generalized stacking-faults and screw-dislocation core-structure in bcc iron: A comparison between ab initio calculations and empirical potentials

Abstract

Generalized stacking fault energies and screw dislocation core structures are reported for two sets of models for iron: density functional theory (DFT) calculations and empirical potentials. A thorough comparison between various DFT approaches has been performed on {110} and {211} γ-lines, which give a first indication on dislocation properties: (i) the effect of the exchange-correlation functional, LDA versus GGA, is significant in the pseudopotential approximation but not in the PAW approximation or in paramagnetic calculations; and (ii) the discrepancy due to the basis set between SIESTA and plane-wave results is rather small. Three empirical potentials for iron have been benchmarked on these DFT results. They all yield similar energies, but different shapes for the γ-lines. Using the criterion suggested by Duesbery and Vitek, the γ-line results point to non-degenerate core structures for the DFT calculations and for the Ackland and Ackland–Mendelev potentials but not for the Dudarev–Derlet potential. The direct calculations of the dislocation core structures show that the Ackland potential is an exception to the Duesbery–Vitek rule. More insight into the stability of the core structure can be gained by looking at the response to the polarization of the core. The Dudarev–Derlet and Ackland potentials have similar polarizations, but the energy difference between degenerate and non-degenerate cores is much larger with the Dudarev–Derlet potential, as expected from the γ-lines. The polarizability of the non-degenerate core is smaller with the Ackland–Mendelev potential than in DFT, indicating that the energy landscape is flatter in this direction.

Keywords:

• generalized stacking fault
• γ-surface
• bcc iron
• dislocation
• density functional theory

View correction statement:

Corrigendum

Acknowledgements

We thank C.C. Fu (CEA/Saclay, France) for her help with the SIESTA code and L. Paulatto (SISSA and DEMOCRITOS, Trieste, Italy) for providing PAW datasets for Fe. This work was funded by the EFDA MAT-REMEV programme and by the SIMDIM project under contract No. ANR-06-BLAN-250. It was performed using HPC resources from GENCI-CINES (Grant 2009-096020).