The ABINIT Users Mailing List ( CLOSED )

[Masayoshi Mikami <mmikami@rc.m-kagaku.co.jp>]

Dear Romero,

I cannot guess what you are having with your models
without your full inputs, but I might have a suggestion.

It might not be a bad idea to try other PW codes, such as
http://www.fhi-berlin.mpg.de/th/fhi98md/
and try with the option "Fast initialization by atomic orbitals".
(I have not tried FHI98MD, but I am just curious about
Tight-binding initialization for initial wave function.
Your benchmark may be interesting.)

I have just remembered my previous post:
http://www.abinit.org/wws/arc/forum/2003-10/msg00043.html
So, in big molecule systems like DNA, some tricks may be needed
even with LCAO-base methods.

Best wishes,
Masayoshi

On 2004/04/08, at 3:55, Nichols A. Romero wrote:

> Hi,
>
> I've been having some difficulties getting the SCF cycle to converge 
> for
> DFT-LDA calculation on some C28-derived molecules. I've done
> calculations on molecules before using LCAO basis and PW basis and I
> know the latter is always trickier but I've never had this much 
> trouble.
>
> If I have iscf=5 (default), I will eventually get the infamous
> scfcge: ERROR - Potential-based CG line minimization not converged
>
> I have tried a number of things including using occopt=3 with
> tsmear=0.01 and nline=10.
>
> The only thing that seems to work is iscf=3, but it sometimes takes 
> over
> 200 SCF cycles before it convergence.
>
> Here are some other relevant parameters, maybe someone can let me know
> if I've set one of them incorrectly.
> diemix 0.3
> diemax 2.0
> acell  3*30 (I think that I should have enough vacuum)
>
> I can send a complete input file if necessary.
>
> Bests,
> --
> Nichols A. Romero
>
> 4-133 ESB, MC 704                           508 East Michigan Avenue
> 1110 West Green Street                      Apartment 12
> Urbana, IL 61801-3080                       Urbana, IL 61801-5176
> (217) 333-6108                              (217) 417-5210