Learn how to ask questions correctly

Pedro Silva

pedros@ufp.edu.pt

I have been using PCM methods, as provided in the G03, Gamess(US) and Firefly packages for a long time, though my lack of mathematical sophistication unfortunately prevents me from following the algorithmical/theoretical details of the methods (abundantly described in many publication by Mennucci, Tomasi and others). I have now come across some strange behaviour which I cannot understand and I would appreciate some advice if any member of this list can spare some time for that:

- In a Li+:F- system, the solvation energy computed by PCM varies (at moderate-to-large separations, i.e. >4 angstrom) with the inverse of the Li+:F- distance. However, the electrostatic interaction energy is more negative at large separations (-165 kcal/mol at 1000 angstrom, vs. -132 kcal/mol at 9 angstrom and -113 kcal/mol at 6 angstrom) , in contrast to what I would have expected if those numbers came from direct application of Coulomb's law. The absolute value of the energies energies, however, precisely track the ones predicted from Coulomb's law (i.e., plotting them vs. Coulomb's law yields a perfect straighline with slope almost equal to 1 ).

- I have found that the electrostatic solvation energies (computed by PCM) for a system containing two components (a neutral organic radical:superoxide) are different (less negative by 7 kcal/mol) from the values I obtain by summing the separately-computed components (even taking into account possible BSSE). Interestingly, the values obtained for the analogous non-radical system (anion: triplet O2) do not depend on the choice of computational methodology (that is, the value obtained with the bimolecular system is equal to the sum of the values of independently-computed O2 and organic anion). In gas phase, no differences between the two strategies were found for any of the systems.

At first I thought that the difference in PCM behavior might be due (like the previous Li+:F- example) from "Coulomb-like" distance dependence between the negative charges on the superoxide and the partial charges in the neutral radical, but direct computation using both Lowdin and Mulliken charges showed me that such an effect would be almost precisely zero.

I guess that my troubles come from some misunderstanding (on my part) regarding the way the partial charges in one moiety interact (or not) with the solvent-interacting surface of the cavity of the other moiety, or with the choice of reference. I would be very grateful if anyone could shed some light on this.

Wed Sep 14 '16 10:02pm

This message read