Firefly and PC GAMESS-related discussion club



Learn how to ask questions correctly


Re^6: CASSCF optimized orbitals are not orthogonal for XMCQDPT

Vladimir A. Mironov
vladimir.a.mironov@gmail.com


FORS-MCSCF and CASSCF are synonyms.
I mean that your MCSCF calculation with restricted excitations is not a CASSCF. You would easily get an arbitrary result from the XMCQDPT2 run.

On Sat Mar 26 '11 10:45pm, Gena wrote
-------------------------------------
>But won't it lead to a dramatical rise of CSFs? I understood from previous CASSCF calculations that CSFs with more than 4 excitations almost don't contribute to the excited states i'm interested in at all.
>Also, the other system i mentioned doesn't cause such problems, and i also use (12,12) active space with only 4 excitations there.

>On Sat Mar 26 '11 9:06pm, Vladimir A. Mironov wrote
>---------------------------------------------------
>>I guess, the problem arises from the IEXCIT=4 keyword in $DRT group. As far as I know, XMCQDPT calculation is only compatible with CASSCF reference wavefunction. Replace IEXCIT=4 with fors=.t.

>>On Sat Mar 26 '11 4:56pm, Gena wrote
>>------------------------------------
>>>On Sat Mar 26 '11 4:37pm, sanya wrote
>>>-------------------------------------
>>>>Probably, you have included CASSCF natural orbitals instead of optimized. Both sets of orbitals are punched; OPTIMIZED are in the last $VEC group.

>>>>An alternative is to set INORB=2 to ignore orthogonality check. But I'm not sure about the results with nonorthogonal orbitals.

>>>>On Fri Mar 25 '11 7:31pm, Gena wrote
>>>>------------------------------------
>>>>>On Fri Mar 25 '11 5:58pm, Gena wrote
>>>>>------------------------------------
>>>>>>Hello!
>>>>>>I try to calculate electronic absorption spectra using XMCQDPT. But after CASSCF calculation is successfully finished, XMCQDPT calculation with optimized MOs from CASSCF PUNCH-file fails in the very beginning listing unorthogonal to each other orbitals, and as i see it - they're all unorthogonal. Firefly also advises me to set INORB=0 to reorthogonalize them. But as XMCQDPT isn't parallel, it isn't convenient. Moreover, i don't understand the difference, i thought that would be the same CASSCF calculation, only in serial mode. Sorry for stupid question, i'm just studying multiconfigurational methods. What am i doing wrong?
>>>Thank you, Sanya! No, of course i understand the difference between optimized and natural orbitals. Moreover, i tried to use natural orbitals to see what error would it cause. And i got immediately I/O error.
>>>The interesting for me thing is that i tried another compound with the same algorithm of actions, and the xmcqdpt calculation started successfully, passing orthogonality check without problems! (Though it isn't over yet).
>>>I too dislike to ignore unorthogonality as i really wanna get reliable results, because i need them in my experimental work.


[ Previous ] [ Next ] [ Index ]           Sun Mar 27 '11 4:24pm
[ Reply ] [ Edit ] [ Delete ]           This message read 634 times