Strange SCF energy for large ANO basis sets

[marand]

Dear all!

I am trying to obtain excitation energies of bithiophene using a series of the ANO basis sets of increasing sizes in order to track the evolution of the excitation energies with basis set size. I have used the DZ, TZ, QZ, augDZ, augTZ and aug QZ versions of the ANO basis, as well as the full RCC-ANO basis set recommended for accurate calculations.

I have come across a strange behavior of the SCF module, which does converge smoothly in all basis sets I used, but starting from the QZ basis set, the final energies are by about 0.5 Hartree higher than those obtained in smaller basis sets. Morover, the virial theorem is not so well satisfied in the large bases (coeff=2.0125) as it is in the smaller ones (coeff=1.9998).

Sth is wrong, but I have no clue as to how to get the correct energies. I would really welcome some advice. As attachments I include the control and output files for the TZ and QZ basis sets.
Sorry, if it be anything of the obvious mistakes, but right now I cannot see it.

Best regards
Marcin Andrzejak

[uwe]

Hi,

which ANO basis sets did you use? The old ones or the newer ones? See here:

http://www.molcas.org/ANO/

Regards,

Uwe

[marand]

Hello!

Indeed! The ANO-RCC basis set from the EMSL portal is the new, corrected one. The augmented double zeta ANO basis set, as well as the augmented treiple zeta ANO basis set are take from the works of Roos and co-workers from 1990 and 1991.

This is still a little surprising (form the variational point of view) that the improved basis set gives the SCF energy that is considerably larger (by over 0.6 Hartree) from the one obtained with the 'old' version of the basis set. I know that for approximate methods energy is not everything and even if energy is not the best, the properties still may be better reproduced, but still...

Anyway, thanks a lot for this tip!
Marcin