Firefly and PC GAMESS-related discussion club

Learn how to ask questions correctly

Alex Granovsky
gran@classic.chem.msu.su

In addition, I'd suggest you to set $contrl wide=.t. $end
in all calculations - both CASSCF and XMCQDPT2. This will
punch/read MOs with the full precision so that there will
not be any errors in calculations introduced by the conversion
of MOs to the text format truncating several significant digits,
as it is done by default. And the option inorb=2 is somewhat
better that inorb=1 as it guarantees exact orthogonality of MOs.
Note in this case the MOs of initial guess are used; so you
must specify $guess guess=moread with norb=all_mos

Regards,
Alex Granovsky



On Fri May 28 '10 8:54pm, Andrei V Scherbinin wrote
---------------------------------------------------
>Hello Andrey,

>There is a useful option INORB=1 in the $XMCQDPT group
>that allows to read SOME converged orbitals from a $VEC
>group, and pass immediately to the XMCQDPT computation.
>(A complete $VEC including all virtuals must be given).

>In particular, the user may read in the same MOs (say,
>those resulting from the state-average ALDET CASSCF
>calculation) in order to perform separate XMCQDPT runs
>for different spin multiplicities (of course, no guarantee
>that such a perturbation treatment always makes sense).

>Best wishes,

>Andrei
>
>
>
>On Fri May 28 '10 4:54pm, Andrey wrote
>--------------------------------------
>>I have performed state-averaged calculations by using CISTEP=ALDET. The density was averaged over two states, namely, singlet and triplet, with the same Sz=0. The problem is I cannot do XMCQDPT2 calculations for both states at once based on averaged density from above ALDET-calculations.
>>I understand that these states should not be mixed in Heff. The idea was (and is) to estimate both states with the same quality. Because, if I do all calculations (CASSCF followed by MRPT) for each state separately as state-specific ones, I will get, strongly speaking, different quality of description for different states.

>>Can someone help me to overcome this problem?

>>thank you very much in advance!
>>Andrey

[ This message was edited on Sat May 29 '10 at 2:19pm by the author ]