> I've run an orbital localization procedure during a geometry
> optimization and after 1000 iterations it didn't converge. It never even
> went past the first orbital localization cycle.

It is possible that you have a case where there are many (perhaps
infinitely many) sets of localized orbitals with a similar or the same
degree of localization. See, e.g. P. Scheurer, W.H.E. Schwarz, Int. J.
Quantum Chem. 76 (1999) 428-433, and 420-427. According to my experience
with the Boys & Foster localization as it it implemented in ADF, for
qualitative studies it hardly makes sense to try harder to get convergence
if it doesn't converge during the first 1000 steps. Usually by that time
you have a sufficiently well converged set of localized orbitals that can
be used for interpretation purposes. However, it depends on what you
actually want to use them for.

There is also the possibility to freeze certain problematic MOs if they
cause convergence problems in the localization. This might yield better
convergence in consecutive localization runs.

>
> Is there any way I can alter the orbital localization parameters - no.
> of cycles, convergence criteria etc. ?

I don't think so, except in the source code. It is in
$ADFHOME/adf/local/locmo.d
You might want to increase the parameter "maxit".

Best regards,
Jochen

-----------------------------------------------------
 Dr. Jochen Autschbach

 Lehrstuhl fuer Theoretische Chemie
 Universitaet Erlangen-Nuernberg
 Egerlandstr. 3
 D-91058 Erlangen
 Germany

 EMail : jochen.autschbach_at_chemie.uni-erlangen.de
 Phone : (+49)-9131-85-25021
 FAX : (+49)-9131-85-27736
 WWW : http://www.chemie.uni-erlangen.de/autschbach
-----------------------------------------------------
Received on 2003-07-29 15:26:35