17
$\begingroup$

I have heard that TURBOMOLE works extremely fast (even compared to MOLPRO) with def2 basis sets, but is slow if Dunning basis sets are used.

What is it about def2 basis sets that allows calculations to be so fast in TURBOMOLE, and why is this unique to def2 basis sets? Why can't we get similar speed with Dunning basis sets?

Improve this question
$\endgroup$

2 Answers 2

Reset to default
7
$\begingroup$

What systems are you running? Turbomole is designed for basis sets employing segmented contractions (like the Karlsruhe sets are), whereas the Dunning sets are generally contracted.

While any code using segmented contractions works also with generally contracted basis sets, it is horrendously slow for heavy atoms since primitive integrals are re-computed hundreds of times.

Improve this answer
$\endgroup$
7
$\begingroup$

The definition of fast and slow is a bit complicated.

I just tried an organic molecule with about 100 atoms and ran two jobs: One with def2-TZVP (1647 AO basis sets, 1872 CAOs) and one with cc-pVTZ (1938 AOs, 2205 CAOs). Pure Hartree-Fock single-point energy calculation, both jobs needed 13 SCF iterations.

Timings: def2-TZVP 31 minutes, cc-pVTZ 44 minutes

Same job but with RI-DFT using PBE functional:

Timings: def2-TZVP 81 seconds (15 iterations), cc-pVTZ 94 seconds (13 iterations)

So, yes, Dunning basis sets are slower, but for real-life applications the difference is not really grave. Although Susi is of course right that for heavy elements and really large basis sets things are worse...

Improve this answer
$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .