[CPMD-list] slow convergence with ODIIS

[Atte Sillanpää]

On Wed, 14 Apr 2004 tangxin at student.dlut.edu.cn wrote:

> &DFT
>   NEWCODE
>   FUNCTIONAL PW91
>   GC-CUTOFF
>   0.1e-05
> &END

Hello TangXin,

I had similar problems with BLYP earlier. I think the convergence
problems were related to the numerically nasty form of the exchange
functional. Apparently the Perdew functional behaves similarly.
(please correct if I'm wrong) However, if you choose PBE instead, you
get rid of all those problems.  I also think that PBE results should
be pretty close to PW91 results, so there should be no real obstacles
to switch to that functional.  You can find several Vanderbilt pseudos
for PBE created by Bernd Meyer (come with the uspp generator from
David's homepage), and it should be trivial to modify your own pseudos
to use that functional. Note that with PBE you don't need to increase
the density cutoff, but you can use the default value 1.d-8.

Hope this helps,

Atte

  atte.sillanpaa at oulu.fi	+358 (0)8 553 1681 (work), KE 368
  Dept. of Physical Chemistry	+358 (0)40 592 7369 (gsm)
  Po-BOX 3000
  FIN-90014 University of Oulu
  FINLAND