[molpro-user] active space

[lydia]

Hi Terry,
Thanks so much for your reply. I appreciate it.
I am trying your suggestions.
As you said, it happends because the MCSCF algorithm is too good.
I give you an example. The avtive space I want to choose should
include the 5d empty orbitals of some atom. The output shows that
those corresponding orbitals are not pure 5d of this atom, and the
other half contribution is from the 3d empty orbitals of another atom
which are not I want. I tried to include more states to do the
SA-CASSCF calculations, sometimes the 3d empty orbitals are gone, but
sometimes they are still there. I think those orbitals are high- lying
and do not nearly have electrons occupation, so  it will not influence
the accuracy of the results. How do you think?

lydia



On Fri, Sep 10, 2010 at 2:35 AM, Terry Frankcombe <tjf at rsc.anu.edu.au> wrote:
> I have had a similar problem when trying to look at excited states in
> sigma radicals where there is also a heap of pi bonding in other parts
> of the system.
>
> Essentially, the problem is that Molpro's MCSCF algorithm is too good.
> Bringing in these other orbitals, at the expense of the part of the
> system you're really interested in, increases the calculated correlation
> energy and drops the total energy.
>
> What has mostly worked for me is to tell multi to be a bit more stupid
> and use the older augmented Hessian method.  That's well known for
> getting stuck in local minima... which is exactly what you want here!
> You tell multi to do this through the iterations directive.
>
> Ciao
> Terry
>
>
> On Thu, 2010-09-09 at 19:17 -0400, lydia wrote:
>> Dear all,
>> I have a question about active space in the CASSCF calulations. I
>> found that some empty orbitals which are not the ones I want will mix
>> with the active empty orbitals I choose with almost equal
>> contribution. Probally because those empty orbitals have similar
>> energies with these avtive orbitals I choose. If it has small
>> contribution, it can be ignored. But it is big in this situation. How
>> do you usually solved this problem? If you have similar experiences,
>> please share with me.
>> Thank you.
>>
>> All the best,
>> lydia
>> _______________________________________________
>> Molpro-user mailing list
>> Molpro-user at molpro.net
>> http://www.molpro.net/mailman/listinfo/molpro-user
> --
> Dr. Terry Frankcombe
> Research School of Chemistry, Australian National University
> Ph: (+61) 0417 163 509    Skype: terry.frankcombe
>
>