[molpro-user] Open shell CC woes

[Kirk Peterson]

Dear Mikael,

one trick that often works is to use the mcscf program (multi) to do your scf since it has a much more robust
convergence algorithm.  Just make sure you specify occ and closed sufficiently so that you recover just the
HF configuration. Write the orbitals using the canonical keyword.

regards,

Kirk

On Mar 24, 2010, at 12:08 AM, Mikael Johansson wrote:

> 
> Dear Kirk,
> 
> Thanks for the hint on how to use previous orbitals, that one works! The reusing of orbitals, that is; the larger basis set SCF's still converge to the wrong states, up to 1 hartree above energies encountered during the SCF, with no help from playing around with level shifts and stuff :-/ So I'll try some other tricks, like slightly breaking the degenerate symmetry.
> 
> Have a nice day,
>    Mikael J.
>    http://www.iki.fi/~mpjohans/
> 
> 
> On Wed, 17 Mar 2010, Kirk Peterson wrote:
> 
>> Dear Mikael,
>> 
>> since it sounds like a smaller basis set works ok, just follow your current input with
>> another set that is identical (starting from the basis set block) except for changing the basis set to VTZ.
>> The second ROHF will pick up the last set of orbitals even if a different basis was used.
>> 
>> regards,
>> 
>> Kirk
>> 
>> On Mar 16, 2010, at 4:43 PM, Mikael Johansson wrote:
>> 
>>> 
>>> Hello All!
>>> 
>>> I'm trying to compute the energy difference between the doublet and sextet
>>> states of D3h symmetric FeF3 at CCSD(T) level (in C2v). I'm having great
>>> difficulty in converging the sextet state ROHF, which surprised me, as it
>>> should be the ground state. The doublet I got to converge.
>>> 
>>> Now, I haven't done any open-shell calcs with Molpro before, so perhaps
>>> I'm just doing something wrong. I would appreciate it if someone would
>>> care to comment if even my input structure is correct. Also any general
>>> comment on the usefulness of computing spin state splittings with RCCSD(T)
>>> are most welcome.
>>> 
>>> For the sextet I've used the following input:
>>> 
>>> {rhf,maxit=500;
>>> occ,14,8,5,2;
>>> closed,12,7,4,1;
>>> wf,53,1,5}
>>> rccsd(t),maxit=500
>>> 
>>> And for the doublet, the following:
>>> 
>>> {rhf,maxit=500;
>>> occ,13,7,5,2;
>>> closed,12,7,5,2;
>>> wf,53,1,1}
>>> rccsd(t),maxit=500
>>> 
>>> The complete input is at the end of the mail. But as mentioned, the sextet
>>> ROHF just refuses to converge with a basis bigger than Ahlrichs-VDZ (or
>>> converges to a local minimum). Is it possible somehow within Molpro to
>>> expand a converged lower basis set solution onto a larger basis set? I
>>> tried to google for that, unsuccessfully.
>>> 
>>> I'm using 2006.1 patch level 151 in case it matters. Any help with this
>>> would be most appreciated!
>>> 
>>> Have a nice day,
>>>    Mikael J.
>>>    http://www.iki.fi/~mpjohans/
>>> 
>>> 
>>> ***,FeF3-sextet
>>> memory,80,m
>>> GPRINT,orbital
>>> GTHRESH,twoint=1.d-13
>>> set,charge=0
>>> geomtyp=xyz
>>> file,1,FeF3-6-pVTZ-s.int
>>> file,2,FeF3-6-pVTZ-s.wfu
>>> punch,FeF3-6-pVTZ-s.punch
>>> geometry={
>>>   4
>>>       FeF3 experimental D3h
>>> Fe         0.00000        0.00000        0.00000
>>> F          0.88150        1.52680        0.00000
>>> F          0.88150       -1.52680        0.00000
>>> F         -1.76300        0.00000        0.00000
>>> }
>>> basis={
>>> spdfg,Fe,cc-pVTZ;C
>>> spdf,F,cc-pVTZ;C
>>> }
>>> {rhf,maxit=500;
>>> occ,14,8,5,2;
>>> closed,12,7,4,1;
>>> wf,53,1,5}
>>> rccsd(t),maxit=500
>>> 
>>> _______________________________________________
>>> Molpro-user mailing list
>>> Molpro-user at molpro.net
>>> http://www.molpro.net/mailman/listinfo/molpro-user
>> 
>>