Firefly and PC GAMESS-related discussion club

Learn how to ask questions correctly

Re^9: CIS optimization state energy moves unreasonably

Alex Granovsky


for MCSCF, you always select the overall number of lowest energy
states to track, not just the single state. The rule of thumb is
to solve CI problem some additional states more than you are going
to track - because in the case of flipping of states #5 and #6,
tracking will definitely fail if state #6 was not computed as well.

Alex Granovsky

On Sun Aug 1 '10 9:10pm, Jonas Baltrusaitis wrote
>An update: I selected state 5 for tracking but I am getting lots of warnings in output:

>    Warning - state tracking failed!
>    Warning - state tracking failed!

>   1   -1750.913520872   -1750.913520872  0.014771 4.950E-03  1   0.0000
>    Warning - state tracking failed!
>    Warning - state tracking failed!
>   2   -1750.920811196      -0.007290324  0.006324 3.845E-03  1   0.0000
>    Warning - state tracking failed!
>    Warning - remapping target state #   4 to converged CI state #   5
>    Warning - state tracking failed!
>    Warning - remapping target state #   5 to converged CI state #   4
>   3   -1750.922493625      -0.001682429  0.004601 4.839E-03  1   0.0000
>    Warning - remapping target state #   4 to converged CI state #   5
>    Warning - state tracking failed!
>On Sun Aug 1 '10 5:50pm, Jonas Baltrusaitis wrote
>>Well, tried optimization off converged MCSCF guess, turns out I can't average?.. That's weird


>>Should I just ask for a brightest state gradient? Also, activate tracking of all states? It's 4 excited states I am working with, so would this work?

>>$MCSCF ISTATE=5 ntrack=5

>>Would this optimize to 4th excited state and track all 5 states (ground+4 excited)? Is there any value of state tracking in state averaged calculations?


>>On Sat Jul 31 '10 1:56am, Jim Kress wrote
>>>NBOs and NOs are not the same.  In addition, the NBO analysis will specifically decomopose the NBO orbitals according to nuclear center, bonding type, and AO contribution.

>>>Look here for more info:



>>>On Thu Jul 29 '10 4:49am, Jonas Baltrusaitis wrote
>>>>I got similar suggestion from Sanya only he suggested natural orbitals of interest.

>>>>Maybe I am not interpreting right: all I need is emission spectrum. So I do not have particular bonds or orbitals in mind, I only need to optimize to certain excited state or averaged states to get decent representation of fluorescence spectra. Does that change anything or I still have to select bond or orbital wise?

>>>>Also, would that make these calculations state specific or state average?

>>>>On Wed Jul 28 '10 6:35pm, Jim Kress wrote
>>>>>Try doing a NBO analysis.  Then select your active space from the set of NBOs (or even NLMOs) that correspond to the atoms and/or bonds of primary concern to you.

>>>>>This is how I usually set-up my MCSCF/ CASSCF calculations.


>>>>>On Wed Jul 28 '10 3:15am, sanya wrote
>>>>>>On Tue Jul 27 '10 4:16pm, Jonas Baltrusaitis wrote
>>>>>>>From what I've read in the literature CASSCF should be a good thing for my molecules since it can deal with multireference character and bond breaking. I am a bit hesitant since I am not sure how to select the active space. Would TDDFT do in selecting which orbitals and electrons to correlate? Is there a rule of thumb?

>>>>>>DFT (as well as HF) orbitals are not very good as starting MOs. Natural orbitals are better. Unfortunately, there is no rule of thumb in selecting the active space, just a common sense and a chemical insight :) In fact, it is individual for each molecular system.

[ Previous ] [ Next ] [ Index ]           Wed Aug 4 '10 6:12pm
[ Reply ] [ Edit ] [ Delete ]           This message read 801 times