Consider a molecule made up of three fragments linked like A-C-B, where each dash is a sigma bond. For example, in 2-aminophenol, fragment C could be the aromatic ring, fragment A could be the hydroxyl substituent, and fragment B could be the amino substituent.
F/I-SAPT can tell you:
- for two molecules, the contribution to the total noncovalent interaction energy associated with the pairwise interaction between A or B on the first molecule and A or B on the second molecule
- within the same molecule, the noncovalent interaction energy between fragments A and B in the embedding field of fragment C that links them.
F/I-SAPT cannot tell you:
- within the same molecule, the noncovalent interaction energy between fragment A and the rest of the molecule or between fragment B and the rest of the molecule. As Jeff said above, this interaction is mediated through a covalent bond. SAPT is a perturbation theory method that knows about electrostatic interactions, dispersion, etc. but doesn’t know about bonding orbitals. SAPT will fail to describe this interaction.
- any interaction energy when any of the fragments A, B, or C are a single atom. The localization procedure used in F/I-SAPT generally works well for functional groups but fails for single atoms. There may be exceptions to this, but generally the fragments need to be larger than a single atom.