[molpro-user] Linearly dependence of basis when expanding the basis

[Gerald Knizia]

On Thursday 21 April 2011 18:48, Łukasz Miądowicz wrote:
> I got a question. I'm calculating electronic structure of diatomic
> molecule. I want to calculate high excited states so my expanding the basis
> set.
>
> I'm expanding the basis set that is given in molpro MANUALLY by
> putting new coefficients.

Do you mean exponents?

> Here's what I'm doing:
>
> 1) I'm putting
>       one coefficient to s orbital of one atom
>       and one coefficient to s orbital of the second atom
> 2) I run molpro
>
> 3) I get an error "BASIS LINEARLY DEPENDENT OR WRONG S"
>
>
> Is there a way to determine which coefficient causes linearity of the
> basis set ?

You can put a gprint,basis in the input in order to see what you actually set 
up. When Molpro complains about linearly dependent basis sets then in most 
cases something is wrong with either the basis or the geometry (e.g., 
angstrom/bohr conversion etc). Molpro is really very good at dealing with 
badly conditioned sets.

> Can molpro expand accurately and automatically the existing basis set
> e.g. ECP2SDF ?

typically, what you'd do if you want to extend sets manually is something like 
this:

memory,50,m;
geometry={C}
basis={
    spdf,C,cc-pVTZ;c;
    spdf,C,even,1;
}
{rhf; wf,spin=2}

This adds one shell of even tempered diffuse functions to the angular momenta 
s to f, for the basis set cc-pVTZ in this case. In your case you would supply 
your concrete set at hand instead of cc-pVTZ. If you need more diffuseness, 
you can add additional shells (e.g., even,2 adds two even tempered diffuse 
shells for each angular momentum). Diffuse basis sets created in this way are 
usually sensible and work rather well.
-- 
Gerald Knizia