[molpro-user] Strange CCSD(T) behavior

[Glaesemann, Kurt R (Kort)]

On 12/19/08 8:39 AM, "Steven E. Wheeler" <swheele2 at chem.ucla.edu> wrote:
> Additionally, for certain basis sets I get well-behaved CCSD energies
> but again the (T) correction is about three times the size it "should
> be", compared to other basis sets.

There can be "bad" areas of the potential energy surface for which the
energy goes crazy.  Here are two references to papers that I am on that
describe such things (others have looked into this, but it is easier for me
to find my own papers.).  One bad thing is that the larger the basis set,
the more likely you are to have trouble.  Also, when the single reference
wavefunction is a poor representation of the correct result, things are more
likely to go bad, but not always!

This paper discusses examples of how perturbative methods can lead to
divergent behavior:
M.S. Gordon, M.W. Schmidt, G.M. Chaban, K.R. Glaesemann, W.J. Stevens, and
C. Gonzalez, "A Natural Orbital Diagnostic for Multiconfigurational
Character in Correlated Wavefunctions", J. Chem. Phys., 110, 4199-4207
(1999).

This paper just includes some discussion of annoying things that I ran
across with trying to study FeCO+.  A divergence at the MCQDPT2 level and a
kink at the MCSCF level:
K.R. Glaesemann, M.S. Gordon, and H. Nakano, "A Study of FeCO+ with
Correlated Wavefunctions", Phys. Chem. Chem. Phys., 1, 967-975 (1999).


______________________

Kurt R. Glaesemann, Ph.D.
Senior Research Scientist II, Molecular Science Computing
EMSL: Environmental Molecular Sciences Laboratory

Pacific Northwest National Laboratory
902 Battelle Boulevard
P. O. Box 999,  MSIN K8-83
Richland, WA  99352
Tel:  509-371-6165 
Fax:  509-371-6445 
kurt.glaesemann at pnl.gov
www.emsl.pnl.gov