[molpro-user] Different ROHF-CCSD energies in gamess and molpro

[Luís P. Viegas]

Dear molpro users,

I have encountered a problem for which i cannot find an immediate
explanation. I am performing ROHF-CCSD (antecipating CCSD(T) and CR-CC(2,3)
corrections, with molpro and gamess, respectively) single point energy
calculations on top of DFT geometries. It is an IRC path from a saddle
point to reactants in a doublet system.

I did not expect a perfect superposition of the CCSD curves, since
according to the literature, CCSD energies are not invariant to
canonicalization procedures (in gamess I am using Guest and Saunders and I
read in a recent paper that molpro uses Davidson). The problem lies in the
region near the saddle point (X near zero), where the CCSD energies of
molpro become suspiciously higher than the CCSD energies of gamess. This
can be seen in ccsd.pdf. My first naive thought was to look at the ROHF
reference energies of both methods, and so I plotted the difference between
the ROHF energies of both program packages in hf.pdf. In both plots, the
units of the y axis are in kcal/mol. The ROHF energy differences are very
small, so I guess that the problem is not here.

http://dl.dropbox.com/u/2159175/ccsd.pdf


http://dl.dropbox.com/u/2159175/hf.pdf



Does anyone know how to explain this or at least have a hint on where
should I look at?

Best regards,

Luís Pedro Viegas
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.molpro.net/pipermail/molpro-user/attachments/20111105/5c64f30e/attachment.html>