linear dependent basis

[Peter Knowles]

The error message that you see is certainly driven by the lowest
eigenvalue of the basis overlap being less than 1.0d-8, and you will
find the print of the eigenvalues so tested immediately above the
error message. If this isn't the case, I guess further progress can
come only if you give your input file so that somebody can try to
reproduce it. In principle the 10^{-8} threshold could be made
adjustable, but I wonder if there is any case with a more singular
metric than this for which one could really solve the SCF equations in
64-bit arithmetic.

At Tue, 24 Sep 2002 15:46:12 -0600,
Elena Jakubikova wrote:
> 
> Hello,
> 
> I am doing calculations with very large basis set that is nearly linearly
> dependent and I am getting the  error message
> "ERROR: BASIS LINEARLY DEPENDENT OR WRONG S" too, which is only
> understandible.
> However I am getting the message when the  smallest eigenvalue of S matrix is
> smaller than 1d-4 rather than 1d-8. Is
> there any way to change this value from 1d-4 to 1d-8 so I can proceed with
> calculation? I understand that very small
> eigenvalues of S matrix may cause numerical problems in the course of
> calculation, but from the calculations I
> have run with the same basis set using Gaussian I believe it is safe to
> proceed.
> I would appreciate any comments.
> 
> Elena Jakubikova
> Department of Chemistry
> Colorado State University
> 
> 
> >   ------------------------------------------------------------------------
> > [Date Prev] [Date Next] [Index][Response]
> >
> > linear dependent basis
> >
> >   ------------------------------------------------------------------------
> >
> >    * To: molpro-user <molpro-user at stchem.bham.ac.uk>
> >    * Subject: linear dependent basis
> >    * From: "H. -J. Werner" <werner at theochem.uni-stuttgart.de>
> >    * Date: Wed, 28 Aug 2002 08:55:44 +0200
> >    * Sender: owner-molpro-user at molpro.bham.ac.uk
> >    * User-agent: Mutt/1.3.16i
> >
> >   ------------------------------------------------------------------------
> >
> > Dear Molpro users,
> > there was recently some discussion about the error
> > message " ERROR: BASIS LINEARLY DEPENDENT OR WRONG S"
> > I looked at one of the problems and found that this
> > was simply an input error (bond distances almost
> > zero). If you get this message, you should first
> > check your geometry, in particular the printed
> > bond distances!! The error message comes when the
> > smallest eigenvalue of the AO overlap matrix is
> > smaller than 1.d-8, and this indicates that
> > something is really wrong.
> > Joachim Werner
> >
> > --
> > Prof. Hans-Joachim Werner
> > Institute for Theoretical Chemistry
> > University of Stuttgart
> > Pfaffenwaldring 55
> > D-70569 Stuttgart, Germany
> > Tel.: (0049) 711 / 685 4400
> > Fax.: (0049) 711 / 685 4442
> > e-mail: werner at theochem.uni-stuttgart.de
> >
> >   ------------------------------------------------------------------------
> >   ------------------------------------------------------------------------
> > To add an item, send mail to molpro-user at molpro.net
> >
> >   ------------------------------------------------------------------------

--
Prof. Peter J. Knowles              
Email P.J.Knowles at bham.ac.uk  Phone +44-121-414-7472  Fax +44-121-414-7471
School of Chemical Sciences, University of Birmingham, Birmingham, B15 2TT, UK
WWW http://www.tc.bham.ac.uk/~peterk/