TURBOMOLE Users Forum
TURBOMOLE Modules => Mpshift => Topic started by: janap on March 14, 2008, 08:42:36 pm

Is it possible to run calculations of NMR chemical shifts in the presence of COSMO solvent? That is  are GIAO + COSMO coupling terms implemented in Turbomole?
I know it's available in g03 but g03 is rather slow for my purpose ...

Hello,
COSMO is currently not implemented in mpshift. You will simply get an error message when you try it.
However, you could try to run a geometry optimization with COSMO, remove the $cosmo keyword from the control file and then run mpshift (remember to set the size of the twoint file in $scfintunit to a nonzero number when you do HartreeFock or use a hybrid functional).
I am far away from being an expert, but I guess that since the geometry and the molecular orbital have been computed with COSMO, the missing terms in the GIAO algorithm will be much smaller than the contribution you have included. Just give it a try if you can compare to experimental results.
Regards,
Uwe

Thanks Uwe for your comment. I did what you advised me to do but I obviously get wrong results. I ran the same claculation with Gaussian many times before and I also have a hint from experiment, so I know approximately what numbers I should get.
The results of the 31P chemical shielding tensor in dimethylphosphate are:

iso 11 22 33
from g03 289.5 176.1 231.0 461.4
from TM 70.1 81.9 9.7 301.9
Strangely, I actually get the wrong results with gaussian as well but only if CPCM is used. With PCM and IEFPCM keywords, my results are OK.
The same problem happens if I calculate the wavefunction in Turbomole and use the orbitals obtained to compute the shielding tenso in MAG with GIAO. On the contrary, wavefunction in Turbomole + shielding calculation in MAG with IGLO for gauge origin gives correct results.
So it doesn't seem to be just a Turbomolebound problem, I don't know what I should think of that ::) Any idea?
Thanks a lot :)
Jana

Hi Jana,
dimethylphosphate? I have tried it, used B3LYP, defTZVP basis set, C1 symmetry.
Gas phase:
ATOM p 3 ISOTROPIC: 302.3316402 ANISOTROPIC: 206.5928077
total magnetic shielding:
Trace = 302.33164022
Tensor :
354.88930968 105.84710904 0.06198524
110.42662899 297.53037250 0.39775309
0.23001188 0.58536927 254.57523849
Gas phase geometry, singlepoint COSMO calculation:
ATOM p 3 ISOTROPIC: 298.8953214 ANISOTROPIC: 177.4833376
total magnetic shielding:
Trace = 298.89532140
Tensor :
350.76987979 90.12337352 0.03473749
91.82104276 287.30184845 0.32949959
0.25463815 0.50951864 258.61423595
Finally, optimized geometry with COSMO:
ATOM p 3 ISOTROPIC: 299.4545554 ANISOTROPIC: 175.5218084
total magnetic shielding:
Trace = 299.45455542
Tensor :
348.83998562 92.12898188 0.06244784
88.85976003 290.57082554 0.71849355
0.42249537 1.17594918 258.95285509
The difference is just about 2 ppm between gas phase and COSMO, and I do not get completely wrong results as you did. If you want, you can send your input files to the Turbomole Support.
Regards,
Uwe

Hi Uwe,
thanks for your response and results of your calculations. After reading your post, I ran numerous test calculations and it appears that the problem is in my basis set.
I normally use IGLOIII downloaded from EMSL. If I use defTZVP instead (as you did), everything is OK.
Strangely, all those problems with my results only appear when COSMO is switched on. Calculations without COSMO employing IGLOIII provide results I expect to get (i.e. isotropic chemical shielding somewhere around 300 ppm). Thus, the problem is not a bad choice of a basis set , just something goes wrong when IGLOIII and COSMO are used tohether, but no idea what >:(
J.

Hello,
COSMO is currently not implemented in mpshift. You will simply get an error message when you try it.
However, you could try to run a geometry optimization with COSMO, remove the $cosmo keyword from the control file and then run mpshift (remember to set the size of the twoint file in $scfintunit to a nonzero number when you do HartreeFock or use a hybrid functional).
I am far away from being an expert, but I guess that since the geometry and the molecular orbital have been computed with COSMO, the missing terms in the GIAO algorithm will be much smaller than the contribution you have included. Just give it a try if you can compare to experimental results.
Regards,
Uwe
Hello,
I was just wondering whether something changed since 2008 when I posted here last time. Has COSMO been implemented in MPSHIFT in the meantime or not?
Thanks,
Jana

Hello,
COSMO development currently is being done in other directions than NMR shieldings. I am afraid that the answer to your question is no.
Regards,
Uwe