Dear CCLers,
Many thanks for all of the replies for our orginal question:
In the "standard" BSSE equation (for a dimer AB):
BSSE(AB) = E(AB)^AB - E(A)^A - E(B)^B + [E(A')^A - E(A')^AB + E(B)^B - E(B)^AB]
I've got most of the parameters, but how in G98W do I calculate E(A)^AB and E(B)^AB, i.e. the energy of the monomer units in terms of the dimer basis set.
Sam Abrash wrote:
"You replace the other molecule's atoms with ghost atoms that are still assigned the basis functions of the other molecule and then you do the calculation.  It's the counterpoise method."
I actually did this originally, but I obtained "strange" results, which I thought meant that I had done something wrong.
It turns out that the keyword to use is MASSAGE:
Wolfgang Roth wrote:
"to calculate E(A)^AB and E(B)^AB usually the atomic charge of the non
intesting atoms , e.g. B resp. B, is set to zero. Thus, the basic functions
of these atoms are still present. In Gaussian, the massage keyword is used"
But many thanks go to Tanja van Mourik who gave me an excellent reference list:
"With Gaussian, the keyword Massage has to be used for counterpoise
corrections (assuming it's the same for G98 and G98W).  Have a look at the
Gaussian manual:
There is an example there as well. However, don't believe the sentence
about counterpoise being only a crude estimate, have a look at the
following papers to learn more about BSSE and Counterpoise:

-  F.B. van Duijneveldt, J.G.C.M. van Duijneveldt-van de Rijdt, and
   J.H. van Lenthe, Chem. Rev. 94, 1873, 1994 (on counterpoise theory)

-  S. Simon, M. Duran, J.J. Dannenberg, J. Chem. Phys. 105, 1996
   (on geometry optimizations on BSSE-corrected potential energy surfaces).

-  F.B. van Duijneveldt, "Basis Superposition Error"
   in "Molecular Interactions" S. Scheiner (ed.), Wiley (1997)

-  T. van Mourik, A.K. Wilson, K.A. Peterson, D.E. Woon, T.H. Dunning, Jr.,
   Adv. Quant. Chem. 31, 105, 1999 (showing BSSE effects on interaction
   energies, distances, and frequencies)."
Ta very much Tanja!