Is def2-TZVP O.K

[asaf]

I calculated the energy of a flour atom with def2-TZVP and with def-TZVP and the results are different. is it O.K.
I tried:
def-SV(P) .vs. def2SV(P),
def-QZVP .vs. def2-QZVP,
and the results was exactly the same, so it  seems a little bit strange that in the first case there is a difference.
The functional was   b-lyp, grid size was 7, and a dft calculation was performed with dscf key word.
Thanks,

[amavr78]

It makes absolute sense , since def2-TZVP and def-TZVP are different. One additional f function has been added to the def2- , if you take a look into the basis file.

def2-QZVP is exactly the same as def-QZVP . So are def-SVP and def2-SVP .

Cheers.

[asaf]

So now, the question is, if def2-QZVP and def-QZVP, as well as def-SVP and def2-SVP are exactly the same, why do we need two definition (def- and def2-) to the same basis set.

[antti_karttunen]

Hi,

def-SVP and def2-SVP are not the same for all elements. There are differences for alkali metals and transition metals, for example. The details can be found from the original def2-article (F. Weigend, R. Ahlrichs; Phys. Chem. Chem. Phys. 7, 3297(2005)). Anyway, usually it is simpler to just use the def2-bases everywhere, they are currently the best choice. In some cases using def-bases for some elements might be reasonable if the def2-variant is a bit too heavily polarized.

[asaf]

Thanks,
So, by the way, currently, can def2 be considered as the "basis set limit" ?
 

[antti_karttunen]

Hi,

I would say that considering the price/performance ratio the def2-bases are the best currently available segmented GTO-type basis sets. But for getting closer to the basis set limit you would need basis sets that are still much larger than the largest available def2-QZVPP. The literature has lots of discussion on this, just search with "basis set limit".

[asaf]

Good,
So, If I want to search for an appropriate basis set for my system (which is big), an accuracy of which order is considered good ? i.e., If I compare to a reference highly accurate basis set, an error of 5.0*10^-3 atomic units is considered good.
By the way, this is approximately the difference which I get for a very simple system - the flour atom (dscf  calculation, pbe functional, gridsize 7) with the following basis sets:
def2-TZVP --> -99.66912313248
def2-QZVP --> -99.67510981418
Thanks

[antti_karttunen]

Hi,

You probably shouldn't worry too much about the errors in the absolute energies, as it's the relative energies that are of chemical relevance. Make some tests for your system and see what basis set yields an appropriate accuracy in terms of relative energies (or molecular properties). The original def2-publication (PCCP 2005) does have helpful discussion on what level of accuracy is expected for each type of basis set.