[Molpro-user] availability of integral direct open-shell MP2

[Gerald Knizia]

On Thu, 25 Sep 2008, Ben Shepler wrote:
> I am a little confused about the availability of integral direct open- 
> shell MP2 methods in Molpro 2008.1.  The manual states that all methods 
> can be performed integral direct except (T) and FCI. 
>
> [doesn't work when called directly but from UCCSD]

This appears to be a bug in either the documentation or in the code 
calling rmp2. For a work-around, you could try the following options:

  o call {rmp2,maxit=20}. This runs an iterative rmp2 procedure (instead 
of the normal one using semi-canonical formulas) which might get slow when 
very large molecules are involved. This uses the same routine as UCCSD 
calls.
  
  o use df-rhf and df-rmp2 instead of normal RMP2. These will also be
much faster when the molecules are large enought that integral direct 
calculations offer an advantage. If you need symmetry to fix certain 
states for open-shell calculations, you can precede df-rhf with a 
symmetry-enabled non-df rhf in a smaller basis, or use df-rmp2 combined 
with normal rhf. Canonical DF-RMP2 has been implemented in Molpro 2008.1 
as a subprogram of DF-RMP2-F12, and simply calling it as "df-rmp2" should 
work.
-- 
Gerald Knizia