DSD-double hybrid functionals

The DSD-PBEP86, DSD-PBEPBE and DSD-BLYP used to be in older versions of PSI but have been removed. In the functionals.py there is the comment:

removed 26 Feb 2014 pending better handling of SS/OS DH coeff

Was this an efficiency issue or was the implementation giving erroneous results?

I guess the gist of my question is whether it would be OK to use these functionals or would I receive unreliable results?



There was a commit a couple years ago that cleaned up how total functionals were built from their component parts (e.g., B3LYP = B3 + 0.81 * LYP + 0.19 * VWN3 …). It cleared up a parameter-definition ambiguity but made it impossible to define SCS-DH functionals purely within the functionals spec. Being unable to specify those functionals you listed, they were removed from the functionals list. No one among the developers has needed them since, so they’re still off the list (though, naturally, we have a ticket, https://github.com/psi4/psi4/issues/39).

You’re the first person who’s wanted them, so thanks for the interest, though I can’t promise we’ll get to them soon. If you uncommented the relevant pieces from functionals.py, what I would expect you would get is the proper underlying functional (the hard part) with the SCS applied incorrectly or not at all (that is, a plain DH). An ad hoc means of computing these wouldn’t be hard.


Thanks for getting back to me so quickly.

Will wait, I guess, until this gets fixed.

All the best,