CC2 vs MP2

[evgeniy]

Dear Users and Developers of TM,

I recently faced a disagreement of CC2 and MP2 in predicting
the most thermodynamically favorable conformation of an organic
molecule. More exactly there are two isomers which are very close
in energy (the energy difference is less than 1 kcal/mol). Nonetheless
I expected that CC2 and MP2 should agree in predicting the
lowest energy isomer. But they do not!

I checked these isomers with DFT (B3LYP and B97-D) and the results
are in agreement with CC2. So I am somewhat baffled why pure ab
initio methods give different results. Any opinion would be greatly appreciated!

Best,
Evgeniy

[resofidentity]

Dear Evgeniy,
for single reference cases, the CC2 gives worse energies in the basis set limit than MP2 compared to full-CI. Also geometries should be slightly worse. But in a closed shell organic molecule the differences should'nt be large.
When you want to compare with the experiment and you expect energetic differences of +-1 kcal/Mol you are in trouble anyway. From experiments you usually get the Gibbs free energy difference at the experimental temperature and in a certain medium (at most). So you need vibrational corrections, at least. The errors in these easily can be larger than 1 or more kcal/Mol in cases. Especially, if you have anharmonic vibrations or large amplitude motions with frequencies smaller than 100 cm^{-1}.

My personal experience is, that in cases were anharmonicity is not an issue the total energy differences from MP2/TZVPP for single reference cases, especially closed shell organic molecules, is usually not a bad guess (though there might be even better density functionals).

regards
 R

P.S.: Do not forget to account for symmetry when you calculate deltaG values!

[evgeniy]

Dear Resofidentity,

Thanks for your detailed reply. I am, at the moment, not going to compare these
results with experiment as there is no experiment. I am trying to understand
why CC2 and MP2 disagree in this case and who gives the correct result. Actually,
I am very curious If full CI were possible for this system would its result agree with CC2
or MP2? As I understood you would bet on MP2, wouldn't you?

Regards,
Evgeniy