[molpro-user] strange CCSD(T) results -- possible answer

[hinde at ion.chem.utk.edu]

On June 28 I wrote:

Dear MOLPRO community:

I am a novice MOLPRO user trying to compute three-body terms
in weakly bound van der Waals trimers at the CCSD(T) level
of theory. I have come across some strange results that
suggest that I'm doing something wrong or that there is a
bug in the MOLPRO version that is installed on the machine
I'm using (2002.6).

[the rest of the message is deleted but can be found in the
archives]

-----

Thanks to Sebastian Riedel (Univ. of Helsinki), who visited
me in person (!) to offer the suggestion that I use wf and
occ cards to explicitly specify the Hartree-Fock
wavefunction, instead of relying on MOLPRO's guess at the
initial orbital occupation.

With this addition to my input files, the calculations give
the same CCSD(T) energy regardless of the order in which
they are performed, which is as it should be.

It's still a bit interesting and puzzling to me that even
though the HF energies for both calculations were correct
(and are not changed by adding the wf and occ cards) the
CCSD(T) energies change when the wf and occ cards are added.
The reason this puzzles me is that the HF phase of the
calculation converges to the correct orbital occupation
whether or not the wf and occ cards are employed, but
somehow the CCSD(T) phase is sensitive to the presence of
the wf and occ cards.

Dr. Riedel's advice:  "Always use wf and occ!"  I agree!

Thanks also to Ben Shepler (Emory Univ.), Andreas Krapp
(Univ. of Marburg), and David Moore (Lehigh Univ.) for
offering suggestions and attempting to reproduce my
calculations to help determine where the problem was.

Robert Hinde
Univ. of Tennessee