We calculated the interaction energy of a system in two methods, B3LYP/aug-cc-pVTZ (Gaussian) and SAPT2+3/aug-cc-pVTZ (PSI4) and got an energy difference of about 8 kcal/mol. How reliable this is?
@Travis: SAPT0 IS NOT comparable to the HF interaction energy! The latter energy lacks the dispersion contribution, while SAPT0 includes it on the uncoupled level. Similarly SAPT(DFT) IS NOT comparable to the supermolecular DFT method for the same reason. See WIREs Comput Mol Sci 2012, 2: 254–272 doi: 10.1002/wcms.86
Often, you cannot compare single point potential energies between different basis sets or methods. Only energy differences between states measured using the same representation will have any meaning.
In the mentioned complex I have an anionic species and a neutral molcule interacting to each other where I see the most deviation. Otherwise, in case of two neutral molecules I see a little deviation, say 0.5-1 kcal/mol, comparing against the MP2/aug-cc-pVTZ energy values.
Thanks for the well-explained answer and the paper. :)
@Travis: SAPT0 IS NOT comparable to the HF interaction energy! The latter energy lacks the dispersion contribution, while SAPT0 includes it on the uncoupled level. Similarly SAPT(DFT) IS NOT comparable to the supermolecular DFT method for the same reason. See WIREs Comput Mol Sci 2012, 2: 254–272 doi: 10.1002/wcms.86