Initially I executed calculation in the "spirit of force theorem" with just one iteration and obtained well converged K value calculated from EIGEN. However reviewer mentioned that force theorem is not obeyed in case of +U calculation - corrections are not in second order. He is right, the K value obtained from full scf calculations and total energy difference is ~40% higher. For just GGA both approaches gives the same result.

This depends on how you start the EIGN calculation. From a scalar relativistic GGA+U calculation or just from GGA ??

And in addition it is probably again a problem of symmetry. Reliable values can only be obtained with identical symmetry in the scalarel and SO calculation. So you have to perform the scalarel. calc with the same struct file (symmetry operations) and case.in2c (LM combinations) as the 2 SO calculations.

It is NOT only a matter of k-points and cannot be "fixed by x kgen -fbz
What surprises me is the non-monotonic dependence of K value as a number of k-points. At the same time K calculated from the total energy converged. While, the band structure energy is part of the total energy.
Not a big surprise. During the scf cycle the potential changes. Remember, by adding a constant potential, the EIGEN values would change by shift*NE, but the potential energy would also shift in opposite direction by -shift*NE, so Etot remains constant, but EIGEN not. (NE .. number of electrons)

Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna, A-1060 Vienna
Phone: +43-1-58801-165300
Email:    WIEN2k:
Wien mailing list

Reply via email to