TURBOMOLE Users Forum

Forum General => Miscellaneous => Topic started by: bademoor on April 23, 2008, 05:02:54 PM

Title: comparison of absolute energies of gaussian and turbomole
Post by: bademoor on April 23, 2008, 05:02:54 PM

Hi,

In our research group we are frequently using the QM-codes gaussian and turbomole. Some people currently using gaussian, are planning to switch to turbomole (much faster  :)), however they would like to retain the work they've done with gaussian (i.e. they do not want to redo all the calculations in tm, if possible) meaning that they would like to be able to compare an absolute energy obtained with gaussian with an absolute energy obtained with tm. From the post on this forum (http://www.turbo-forum.com/index.php?topic=54.0) I understand that, in principle this should be possible unless using the same functional, basis set, etc.

So, as tm-user I started to do some quick test-calculations for them. I did a single point energy calculation on methane with B3LYP/6-31g*; here are the results:
1) in Gaussian I find an energy of -40.5184205770 hartree
2) in tm I find an energy of -40.51765327052 hartree

this gives a difference of about 2 kJ/mol between the absolute energies, although I expected to find a much lower value.
Is it possible to comment on this? Is this difference expected or not?

Best regards,

Bart

PS: I've (double) checked that I'm really using the same basis set in both programs, so this can't be the problem. Next to this, I used the b3-lyp_Gaussian functional in tm and a m3 gridsize.

Title: Re: comparison of absolute energies of gaussian and turbomole
Post by: antti_karttunen on April 24, 2008, 07:24:24 AM

Hi,

I did some test calculations and it looks like this could be a 5d vs. 6d problem. When using 6-31G*, Gaussian uses by default a set of six d-functions. However, for Turbomole the default is to use five d-functions. By giving keyword 5D to Gaussian, the energies should be practically equivalent.

By the way, while the b3-lyp_Gaussian functional is quite handy if you want to quickly utilize your previous Gaussian results, it has some limitations (some Turbomole features have not been implemented with it). In the long run, it is probably best to switch to normal b3-lyp (VWN5). The B3LYP (VWN5) functional can also be used in Gaussian, if you still want to use Gaussian and Turbomole in the same project.

Title: Re: comparison of absolute energies of gaussian and turbomole
Post by: bademoor on April 24, 2008, 04:57:05 PM

Hi Antti,

thx for the quick answer, it is indeed a 5d vs. 6d problem ... just checked it (I didn't know this).
With regard to you thoughts about using b3-lyp_Gaussian, you're completely right (I share your opinion)! Now I just have to convince my colleagues  ;)

Greetings,

Bart