Firefly and PC GAMESS-related discussion club

 
Learn how to ask questions correctly  
 
 

 

Alex Granovsky
gran@classic.chem.msu.su

A couple of comments concerning your findings:

First of all, the basis sets you are using are of spherical type.
This means you need to add d5=.t. option to your input files.
This will greatly reduce the degree of linear dependence of
these basis sets.

Next, you need some extra precision, esp. for diffuse basis sets.

The proper input line could be then as follows:

 $CONTRL SCFTYP=MCSCF RUNTYP=ENERGY d5=.t. inttyp=hondo $END

You may also try to switch to direct transformation with fastints code as described here

This way you'll restore correct degeneracy of CI/MCSCF roots.  

Finally, cc-type basis sets are designed for explicitly correlated
methods. This means that they are not necessary better for MCSCF
unless you perform XMCQDPT2 after MCSCF step. In addition, there
can be some symmetry constrains preventing high-L basis set functions
from being populated in MCSCF (the latter depends on the structure of
the active space you are using). Anyway, they will be populated in
the subsequent XMCQDPT2 treatment.

As to DK-type basis sets, they are designed for relativistic
calculations. In particular this means that their part describing
wavefunction in the vicinity of nuclei is different. As you are
performing non-relativistic calculations, it is logical to expect
better results with non-relativistic basis sets as the contributions
of inner shells into total energy are the dominant ones.

As to fully uncontracted aug-cc-type basis sets, they are only seldom
of use. First, the problems caused by quasi-linear dependence are
much more significant with these basis sets. Second, these basis sets,
even being fully uncontracted, lack functions describing core-valence
correlations. This is not too significant problem for MCSCF but if
you perform PT2 after MCSCF it would be logical to include these
additional functions into correlation treatment. For instance, you may
need to look at aug-cc-cwCVTZ basis set etc...

Kind regards,
Alex Granovsky




On Sat Mar 10 '12 5:34pm, Andrey wrote
--------------------------------------
>Sorry, dear collegues, in my calculations of the O2 excitation spectra I see some strange (to my up to now not very experienced point of view) dependence of the results on the basis size.

>! ================================================================    
>! Input file for FireFly    
>! ================================================================    
> $CONTRL SCFTYP=MCSCF RUNTYP=ENERGY EXETYP=RUN $END
> $SYSTEM MWORDS=12 $END  
> $CONTRL COORD=UNIQUE UNITS=ANGS $END 
> $MCSCF CISTEP=ALDET SOSCF=1 $END
> $DET GROUP=C1 NSTATE=20 NCORE=2 NELS=12 NACT=8 $END
>  
> $DATA
>Oxygen molecule
>Dnh  4
>  
>O  8.000000 0.000000 0.000000 -0.6035
>! here was the basis set data !
>            
> $END

> $END
> $GUESS GUESS=MOREAD NORB=40 $END
> $VEC
>! here were the guess vectors !
> $END
>
>
>I made a set of calculations with different basises,
>the basis from the internal GAMESS databese 6-311G** (the smallest one among the tested ones), then with the basis sets from the EMSL web page :
>PVTZ-Gamess, aug-cc-pVDZ, aug-cc-pVDZ-DK, aug-cc-pVTZ and finally the totally uncontracted aug-cc-pVTZ. I expected that the ground state (triplet X³\Sigma⁻_g ) energy and the energies of the lowest excited states ( singlets a¹\Delta_g and b¹\Sigma⁺_g ) will lower with increase of the basis size, but instead I got the minimal values for the PVTZ-Gamess case and, on the contrary, increase of the total energies for aug-cc-pVDZ and aug-cc-pVDZ-DK.

>Moreover, for the larger basis sets, aug-cc-pVTZ, in the converged situation I see, that the double degeneracy of the \Delta state (and also for the other higher states with the angular momentum projection > 0, that must be doubly degenerated in the case of O2 molecule) has disappeared, i.e. these levels arise to be splitted ! In contrast, for the smaller basis sets the degeneracy was practically perfect.

>What can be the reason of such strange unphysical dependence ? What I am doing in a wrong way ?

>Thanks in advance for any help.
>Sincerely
>Andrey
>