Can I estimate the bond energy by running a single Gaussian calculation of the fragments at very long separation (say, 40 angstroms)? Or do I have to calculate each fragment without the other?