[CP2K:3975] CP2K PM6 result does not consistent with MOPAC PM6

Michael Banck mba... at debian.org
Sat Aug 18 19:15:33 UTC 2012


On Fri, Aug 17, 2012 at 01:42:58PM -0400, Bharat Sharma wrote:
> Au parameters are correct as reported in paper. I started the calculation
> with the Ag-O distance of 2.03Angstrom both in CP2K and MOPAC (simply same
> initial geometry). The uptimized bond length of Ag-O from CP2K is 1.90 A
> where as from MOPAC is 2.03A which is consistent with PM6 results in "AM1*
> parameters for palladium and silver, J Mol Model (2011) 17:2585–2600)".
> I did not one shot energy calculation with Ag-O distance of 2.03A in CP2K
> but the heat of formation is 135.7645 kcal/mole but from MOPAC is
> 98.21kcal/mole.
> I did the test calculation of AgH+, Ag(NH3)4+, AgF from both in CP2K and
> MOPAC but results are fairly similar( error ~ or less 1 kcal/mole).
> In all test calculations, the initial geometries for CP2K and MOPAC are
> same.

I repeated the calculations using Gaussian09 and get basically the same
results as with CP2K: a bond length of 1.905 A and a heat of formation
of 135.04 kcal/mol.

This is the G09 input file I used:

#P pm6 opt


0  2
O                 -2.48815585    0.31201747    0.00696372
Ag                -4.60495278    0.41447825    0.06239400

That said, I never got consistent PM6 results between MOPAC on the one
side and G09/CP2K on the other, especially for heat of formations and
even with simple systems like an isolated water molecule (whereas AM1
energies are pretty close).

I'd be very much interested if somebody could explain what is going on.


More information about the CP2K-user mailing list