The primary basis set superposition error (BSSE) results from the artificial lowering of the energy of each subunit of a pair by the presence of ‘‘ghost orbitals’’ of its partner. In addition, these ghost orbitals perturb the one-electron properties of the molecule, causing a change in the interaction energy, an effect known as secondary BSSE which is not corrected by the counterpoise procedure. The primary and secondary BSSE are calculated for the interactions of NH3 and H2O with Li+, using a variety of different basis sets. It is found that the 2° BSSE can be quite large, comparable in magnitude to the 1° component at both the SCF and MP2 levels. There is no basis found for the supposition that 2° BSSE improves the calculated interaction energy, nor do the 1° and 2° effects cancel one another in general. While the MP2 BSSE tends to be smaller than the SCF analog, the former can be similar in magnitude to the ‘‘true’’ MP2 contribution to the interaction; failure to remove the BSSE can hence lead to a qualitatively incorrect interpretation of the effects of electron correlation. Comparison with a system in which basis set superposition is rigorously excluded suggests that subtraction of both the full 1° and 2° BSSE is appropriate and does not overcorrect the potential. Addition of a diffuse sp shell, especially if coupled with orbital exponent reoptimization, leads to a lowering of the 1° and 2° BSSE, which moreover take on opposite sign and cancel one another to some extent.
Read full abstract