CCL: AIM theory with weak interactions

["Gonzalo Jim nez-Os s" <gjimenez. ~ .unizar.es>]

 Sent to CCL by: "Gonzalo  Jim  nez-Os  s" [gjimenez.]|[.unizar.es]
 Dear CCL'ers,
 I have some theoretical doubts about the aplicability of the Atoms In Molecules
 theory in systems in which the interactions between two molecular fragments are
 middle- or long-range and mainly related to (non-local) electronic correlation
 effects (dispersion) like those ocurring in van der Waals complexes. For
 example, I've been performing some calculations with the well-known benchmarck
 benzene dimer. Whereas the p-p stacked minimum is located only with correlated
 ab initio methods (common DFT fails and separate the two monomers), when the
 wavefunction (the SCF density) calculated through DFT (i.e. B3LYP) on the
 optimized MP2 geometry is analyzed within the AIM context (EXT94b program),
 exactly the same number of critical points (CP) are found with respect to those
 found in the MP2 wavefunction, including seven CPs between the two fragments in
 the interaction region.
 So, my question is: if common DFT (or in the extreme case, HF) cannot account
 for dynamic correlation effects, how these critical points related to "weak
 interactions" between the two fragments can be located in the density
 topology? Is there any physical/theoretical inconsistency on these results?
 Moreover, the value of density (rho) of these zero-gradient points is very
 similar in fully uncorrelated methods (HF), DFT (BLYP, B3LYP) and MP2. Is there
 any plausible explanation for these results?
 Thank you very much in advance,
 Gonzalo Jimnez-Oss
 University of Zaragoza