Dear Pablo, I have no idea on graphene, so my suggestion is based on "common sense". Check what kind of impurity level/band nitrogen is forming, and if your half-integer occupation per spin comes from a too large dispersion of this band, it may be an artefact of not big enough supercell. Otherwise if the Fermi level crosses genuine band of graphene, maybe this is OK? - like in a non-magnetic metal with odd electron number; a legitimate case of not-too-small doping.
Essentially, this is the question about localization of N states. The difficulty may be both technical (supercell size) and basic (DFT underestimating localization). Try to make sense of whatever is going on around your Fermi level... Eliminate possible technical sources of doubt (k-points etc.)... Best regards Andrei Postnikov > Dear Siesta Users, > > I did some calculations for 4x4 graphene > doped with nitrogen. when I was doing the optimization (cell parameters > and > atomic position) I found two solutions. One of the correspond has a net > magnetic moment of 1 uB ( 65 spin up and 64 spin down). The other minima > has > zero magnetic moment (Qup 64.5 and Qdown 64.5). The latter is about 1 eV > lower in energy and the structures are not that different. > > I have seen this problem in DFT anf HF calculations for open shell > molecules > with strong spin contamination, but what is confusing me is that the > wavefunciton with zero magnetic moment is lower in energy. I would expect > the opposite situation based on the results for molecules. > > I would like to know if somebody faced the same situation or knows have > understand it. > > Many thanks... > > Best regards, > > Pablo > >
