[molpro-user] CASSCF energies using FROZEN

[Lorenzo Lodi]

Dear Kirk,
thanks for your answer. I understand what you are saying; unfortunately
I have deleted the input file which produced the unexpected results
which spurred my first question and I have been unable to reproduce
them. It might just have been a matter of numerical convergence (I'm
dealing with molecules containing transition metals) and convergence of
casscf to different solutions depending on starting guess or other
settings is common.

Lorenzo



On 13/04/13 17:59, Peterson, Kirk wrote:
> Dear Lorenzo,
>
> no, I don't believe this is correct.  If you check your outputs carefully, you'll see that when you use frozen those orbitals are also always
> doubly occupied in all CSFs, just like closed.  The orbitals are just not optimized.  So if you use frozen instead of closed, the
> energy should always be higher in the frozen case (if both have converged to the global minimum).  One can see this with the very simple input:
>
> geom={
> C
> O  1  2.1322}
> hf
> {multi;frozen,2}
> {multi;closed,2}
>
> Using  closed,2;frozen,2  will give the exact same energy as   closed,0;frozen,2 or just frozen,2
>
> regards,
>
> -Kirk
>
> On Apr 13, 2013, at 8:18 AM, Lorenzo Lodi <l.lodi at ucl.ac.uk> wrote:
>
>> I think I understand what was going on in my calculations, I was misinterpreting my inputs.
>>
>> I'll explain my mistake, in case it might be useful for other people.
>> A line such as
>> CLOSED, 2; FROZEN,0;
>> will create CSF where the two lowest orbitals are restricted to be doubly occupied, but the orbitals themselves are variationally optimized.
>> On the other hand a line like
>> CLOSED, 0; FROZEN,2;
>> will impose no restriction on the CSF occupancy (hence producing a large active space than in the previous case) but will not optimize the two lowest orbitals.
>> As the active space is larger it is quite possible that the total energy will be lower than in the previous case.
>> What I really wanted to do was
>> CLOSED, 2; FROZEN,0;
>> where the active space is the same as in case 1 but the two lowest orbitals are not optimized. The total energy will then be higher than in case 1, as expected.
>>
>> Lorenzo
>>
>>
>> Il 12/04/2013 18:30, Lorenzo Lodi ha scritto:
>>> Hello,
>>> I have noticed that in some cases I get lower CASSCF energies when using
>>> FROZEN orbitals than when specifying the same orbitals by CLOSED. I
>>> would like to know if this is caused by convergence to a local minimum
>>> of the CLOSED calculation or if this situation is somehow 'correct'.
>>> Naively, I would expect that keeping orbitals FROZEN instead of CLOSED
>>> would always increase the total energy. Am I wrong?
>>>
>>> Lorenzo
>> _______________________________________________
>> Molpro-user mailing list
>> Molpro-user at molpro.net
>> http://www.molpro.net/mailman/listinfo/molpro-user
>