| Re: Problem with Copper hard pseudopotential ( No.1 )|
- Date: 2012/10/18 15:08
- Name: T.Ozaki
>Hello, I am having some strange results when using Cu6.0H.pao/Cu8.0H.pao and Cu_PBE11H.vps:
>When running a calculation for just the atom, the DOS seems to have a strong dependence on cell size (I used ~1nm^3).
I did similar calculations using Cu6.0H-s3p3d3 and Cu_PBE11H, and changed the lattice
constant of the cubic cell, 10, 13, 17 Ang. My results for the three cases are almost
equivalent to each other.
> When running bulk calculations (FCC structure), the DOS seems to be wrong when using
> Cu8.0H.pao. It's not what it should be, and it seems to be vary a lot depending on the
> basis set used (s4p3d3, s3p2d2f1,etc).
Probably, the computational condition must be very close to the overcompleteness rather
than a problem arising from pseudopotential. Could you try to use the following keyword?
scf.ProExpn.VNA off # default=on
Switching off the projector expansion tends to avoid the erratic situation.
| Re: Problem with Copper hard pseudopotential ( No.2 )|
- Date: 2012/11/05 19:58
- Name: Alfonso <firstname.lastname@example.org>
- Dear T.Ozaki,
Thank you for your reply. Turning off scf.ProExpn.VNA did solve the problem. I searched the openmx manual and couldn't find any information regarding this keyword. I read your PRB titled "Efficient projector expansion for the ab initio LCAO method". I would like to understand this overcompleteness problem, as I don't understand why using the projector expansion can cause this problem. If you could briefly explain or recommend some literature about that I'd appreciate it.
What's the downside of turning off this projector expansion? Are its advantages only noticeable in the total energy calculation? in which situations may be problematic to turn it off? (besides geometry optimization)
Thanks again for your help