The ABINIT Users Mailing List ( CLOSED )

[Steven Homolya <steven.homolya@spme.monash.edu.au>]

On Mon, 10 May 2004 03:33 pm, a.soon@auckland.ac.nz wrote:
> Dear all,
> I've the following error message:
>
>  scfcge : ERROR -
>   Potential-based CG line minimization not converged after 13 restarts.
>   Action : read the eventual warnings about lack of convergence.
>   Some might be relevant. Otherwise, raise nband.
>
>  leave_new : decision taken to exit ...
>
> Could anybody please enlighten me whatz going on? Cheers!

This is a common problem. Here is one quick fix to try: use iscf=3, instead 
of 
iscf=5

My latest version (4.1.4) of Infos/varbas.htm incorrectly says the following 
about iscf:

"The preferred option is 5, which is quite robust. The value 3 can be faster, 
but sometimes the SCF iterations will not converge with iscf=3 !"

The opposite is true from my experience. iscf=3 can be agonisingly slow but 
converges more reliably than iscf=5. I believe there have been posts on this 
list that also confirm this.

Might pay to run 5-10 SCF cycles with iscf=3 and then switch back to iscf=5 
and continue, say using multi dataset mode. If you're also doing geometry 
optimisation it's going to be more painful and might be worth going iscf=3 
all the way.

Cheers,

Steve