CCL: Spin Delocalization/Localization using two different functionals

 Sent to CCL by: Mariusz Radon []
 On 06/13/2014 01:09 AM, Kerry Jeffry Wrighton wrote:
 > Sent to CCL by: "Kerry Jeffry Wrighton" [kerry.wrighton*]
 > Dear CCL team
 > I have a question related to equilibrium geometries using B3LYP (hybrid)
 > BP86 (GGA) functionals using the same basis set for both methods (TZV). I
 > optimized the structure of a hexavanadium cluster, which contains 4 V(IV)
 > 2 V(V) atoms. When I used the B3LYP functional, spin density is perfectly
 > localized over 4 V(IV) atoms. In the case of the BP86 functional the spin
 > density is full delocalized over the 6 V atoms. In both cases, the
 > are validated through the frequency analysis and comparison with bond
 > distances of crystallographic data. Why did it happen?
 > Best Regards
 > Kerry Wrighton-Araneda
 Dear Kerry:
 It is a known tendency of pure functionals (like BP86) to overestimate
 an electron delocalization due to the self interaction error. This error
 can be compensated, in part, by mixing an approximate exchange
 functional with the exact exchange (like in B3LYP).  See for instance
 [DOI: 10.1021/ct800531s].
 It is thus perfectly understandable that introducing the exact exchange
 in B3LYP,  as compared with BP86, leads to the more localized unpaired
 electrons. Recently I reported the similar behavior of these two
 functionals for a tetravanadium cluster
 [DOI: 10.1016/j.inoche.2013.12.034 + Supporting Information].
 Best regards,
 Dr Mariusz Radon, Ph.D.
 Coordination Chemistry Group
 Faculty of Chemistry
 Jagiellonian University
 ul. Ingardena 3, 30-060 Krakow, Poland