> I am using the PBE-GGA functional. According to the > http://prb.aps.org/abstract/PRB/v79/i8/e085104 paper, it is indeed the > best choice for Fe. However, LDA seems to be the best for Ge. Well, it > looks like that the Fe-Ge system is not very convenient for this kind of > optimizations.
That's perhaps too negative as conclusion. Even though you might not be able to predict the absolute volume with 1% accuracy, relative trends should be much better, more or less independent of the functional (e.g. volume change upon the introduction of a given amount of vacancies). STefaan
