The ABINIT Users Mailing List ( CLOSED )

["Anglade Pierre-Matthieu" <anglade@gmail.com>]

PS: after watching you output it seems pretty clear that your systems
problems comes from the Kerker's preconditioner. Reduce diemix a
little bit and determine the best value for iprcel (every simple
mixing step until the X in "iprcel 4X shall converge and 4X should be
the biggest possible). For instance if you use a mixing of 0.4 +
iprcel 4X you should reach perfect convergency no later than after
60-70 step.  If this is not enough it is the testimony of problems
during the wave function calculations.  Using nnsclo=2 and nline=6 as
Matthieu suggested will almost double the computation time while
ensuring convergency at this level.

On 10/23/06, Anglade Pierre-Matthieu <anglade@gmail.com> wrote:
> Hi,
>
> Since you are working with a relatively small system I would suggest you tu 
> use
> iprcel 4X instead of diemac. This will speed up your true SCF
> convergence.  (For inhomogeneous systems diemac will lead energy
> change to reduce without ever reaching convergency.)
>
> If ths system keeps refusing to converge it means that there is too
> many states near the Fermi energy. The solution then is to get
> convergency with a small electronic temperature, treating the system
> as if it was metallic (tsmaer and occopt).
>
> regards
> PMA
>
>
>
> On 10/23/06, 张�s <zhangting1980323@gmail.com> wrote:
> > Dear all:
> >
> >        I'm calculating phonon band of 5,5 CNT with GGA-PBE XC. I've
> > optimized the CNT's sturcture to the max force 1.6e-6 Ha/Bohr.
> > But when I start to perform phonon calculation, I found that the SCF dataset
> > can not converge to the suggested tolvrs=1.0d-18, but to
> > oscillate at about 1.0d-10. I'm testing a similar SCF calculation with LDA,
> > Teter Pade XC, it seems converge well. Why does this happen? Thanks!
> >
>
>
> --
> Pierre-Matthieu Anglade


--
Pierre-Matthieu Anglade