[molpro-user] df-ccsd(t)

Gerald Knizia knizia at theochem.uni-stuttgart.de
Mon May 26 12:27:53 BST 2008

On Friday 23 May 2008 18:36, Gregersen, Brent wrote:
> It looks like df-ccsd(t) is not a supported calculation type (although
> it probably isn't too hard to include it in future versions).

The problem with full df-ccsd is that it is not a good method. There are 
certain terms in the CCSD equations (3- and 4-external contributions) which 
require large DF basis sets to be approximated and where DFing does not 
improve the scaling or runtime. It might actually be slower than a 
conventional calculation.
(In the local case the situation is completely different, however).

> In the interim, are there options that could be supplied to df-lccsd(t)
> that would approximate a df-ccsd(t) calculation? 

You can specify {df-lccsd(t0),domsel=1.0}. I think that should do the trick.
Gerald Knizia

More information about the Molpro-user mailing list