Firefly and PC GAMESS-related discussion club

 
Learn how to ask questions correctly  
 
 

 

Alex Granovsky
gran@classic.chem.msu.su

just to stress out what has already been suggested by Forum members
and to provide a brief summary:

1. One can use a symmetry in Firefly while explicitly forcing CI
programs, either ALDET or GUGA, to work in C1 group. To do this,
one needs to input group=c1 in $det or $drt. This will results  
in symmetric MOs and pure symmetry states, although one may  
need to tighten convergence criterion of Davidson's CI.

2. With Firefly v. 8.0.0, one can force MCQDPT2 and XMCQDPT2
programs to work in C1 group using istsym=-1 option.

3. The pures option pertains only to spin symmetry but not
to irreps of spatial symmetry group.

I have attached archive with sample inputs and outputs.
I first generated state averaged CIS natural orbitals using
15 lowest CIS excited states (while symmetry was not disabled,
the default settings of CIS program is to compute states without
any symmetry restrictions), then ran FOCAS SA8-CASSCF(18e,13o)
for 3 iterations to improve starting guess a bit, then converged
SA8-CASSCF(18e,13o) with SOSCF, and finally run XMCQDPT2 for
these 8 states  

Personally, I'd strongly suggest you to use state-averaged CIS
natural orbitals to create initial guess orbitals for MCSCF.
There should be a sample attached to one of my posts on this
Forum, you can search for it using "search" button on the left.

Kind regards,
Alex Granovsky


On Tue Mar 5 '13 10:43am, Thomas wrote
--------------------------------------
>Dear all,

>I would like to make a small addition to Pavel's suggestions. With Firefly 8.0.0, it is no longer necessary to specify $CONTRL NOSYM=1 $END if one would like to perform a XMCQDPT2 calculation with states beloning to different irreps. Instead, one can now specify $XMCQDPT ISTSYM=-1 $END to disable symmetry during the XMCQDPT2 step. This naturally also applies to MCQDPT2 calculations.
>
>
>Kind regards,
>Thom
>
>
>
>On Tue Mar 5 '13 1:01am, Solntsev Pasha wrote
>---------------------------------------------
>>Dear Antonio.

>>I had the same problem with some macrocycles. But in my case the symmetry was much higher. I "bugged" Alex a lot about this issue to get answer on the same question as you have. As far as i know, with FF8 you can't do SA-MCSCF over states with different symmetries. You can use different symmetries (using ALDET code), but not symmetries.  The strategy i used is follow:
>>Do geometry optimization using full point group symmetry. Then ,if states of interest have different symmetry (such as ground state and excited states) set group=c1 and use wstate(1) to select desire states (for example, ground state and only symmetry allowed states). If you have transition from single degenerated state to a double degenerated one, don't forget to include both configurations (transition double degenerated) into SA.
>>After SA-MCSCF is done you the results were checked and you love them, you can setup XMCQDPT with $CONTRL NOSYM=1 $END. This option switch off the symmetry completely during XMCQDPT job. Again, you may use KSTATE(1), WSTATE(1), AVECOE(1) to select the states of interest.

>>My observation about SA-MCSCF/XMCQDPT code. It is really very fast. For your molecule it is not a problem at all even without symmetry. Do some search withing this forum and you will find more professional comments from experts.

>>Good luck.

>>Pavel.
>>
>>
>>On Mon Mar 4 '13 11:26pm, sanya wrote
>>-------------------------------------
>>>You may try PURES=.f., it permits state-averaging over states with different multiplicity. It may also work for the states of different symmetry, but I am not sure, you'll better try and see.

>>>On Sat Mar 2 '13 3:23pm, Antonio Carlos Borin wrote
>>>---------------------------------------------------
>>>>Dear Prof. Alex,

>>>>In reply to your message, I´m posting the input file and I will try to better explain what I want to do.

>>>>Please, note that it is the input file you sent me in reply to a previous question about excited states I posted on Mon Feb 18 '13 4:01pm.

>>>>Dr. Solntsev Pasha was of great help, but maybe you can find other solution.

>>>>I want to compute the excited states of adenine, employing Cs symmetry; therefore, I need to compute A' and A'' states.

>>>>In MOLPRO and ORCA, for instance, we can do state-average casscf (sa-casscf) averaging over states belonging to different irreducible representations. So, the A' and A'' states can be computed in the same run. In MOLCAS this is not possible and we have to compute the A' states in a step and the A'' states in other step.

>>>>My question is: is it possible to do the sa-casscf and XMCQDPT calculations averaging over the A' and A'' states in the same job? Or we need to do it as with MOLCAS.

>>>>As Dr. Solntsev Pasha said, it is possible to do it without symmetry, the "default" solution. But I would like to know if it is possible to do it employing symmetry (Cs in this case).

>>>>Which keywords do I need to add to the attached input file to do it?

>>>>Thank you for your attention
>>>>Best regards,
>>>>Antonio

This message contains the 677 kb attachment [ adenine2.rar ] Some adenine samples