| Re: Magnetic moment in cubic SrRuO3 ( No.1 )|
- Date: 2006/11/04 01:41
- Name: T.Ozaki
Have you tried to evaluate the magnetic moment by the Voronoi scheme ?
Also, did you check the band dispersion ?
Since the cutoff radii of basis function you used are relatively longer
for bulk systems, I am wondering that you encountered the overcompleteness.
Above all the possibilities are not your case, the discrepancy may be
attributed to the pseudopotentials.
| Re: Magnetic moment in cubic SrRuO3 ( No.2 )|
- Date: 2006/11/04 06:32
- Name: JessK
thank you for the reply. There was an error in my previous message, indeed I used O5.0-s2p1 for oxygen. When I used O5.0-s2p2d1 I got magnetic moment of 1.01mB on Ru atom, and fitted lattice constant is equal to experimental one.
>Since the cutoff radii of basis function you used are relatively longer
>for bulk systems, I am wondering that you encountered the overcompleteness.
How to define is there overcompleteness? I see that program 'polB' can check it. When I run it, the code passed this step, so I think I avoided overcompleteness. Am I wrong?
| Re: Magnetic moment in cubic SrRuO3 ( No.3 )|
- Date: 2006/11/05 00:05
- Name: T.Ozaki
When some of eigenvalues, E, of the overlap matrix become minums,
the basis set is overcomplete. Since the orthogonal trasformation requires
the calculation of 1/sqrt(E), such a trasformation is ill-defined.
Even if E is not minus, for a small E the calculation of 1/sqrt(E)
produces numerical error. In this case, the eigenvalues of the Kohn-Sham
eq. tend to be erratic, and such erratic eigenvalues can be found
in the band dispersion. Or if you monitor the eigenvalues during the
SCF calc, then you will see the erractic eigenvalues.
For cluster or molecular systems, the state with a small eigenvalue can
be eliminated. However, the eigenvalue spectrum of the overlap matrix
is continuous for the band calculation, and it is not easy to remove
ill-conditioned states unfortunately.