Inconsistencies in modelling interstitials in FeCr with empirical potentials

We present empirical potential and Density Functional Theory results of interstitials in FeCr and pure Cr. Results show that potentials for the original and revised two-band model, a recently introduced third two-band model, and for the revised concentration-dependent model produce errors of up to multiple eV in formation and binding energies for Fe-containing interstitials in pure Cr. Fe-interstitial binding in Cr is much stronger than Cr-interstitial binding in Fe according to Density Functional Theory, but all four potentials still strongly overestimate the binding strength. At the Fe-rich end errors in empirical potentials are smaller and most of the errors are not a linear extrapolation in concentration of the larger errors in pure Cr. Interstitial formation energies in Fe-rich FeCr are underestimated by all four empirical potentials, but much less so than in pure Cr. In Fe-rich FeCr the revised concentration-dependent model produces Cr-interstitial binding energies quite similar to Density Functional Theory values, while all three two-band models show almost no binding or repulsion.