ABINIT Discussion Forums

Hello,

I'm trying to compute positron properties of metals. As a start I considered a delocalized positron in B2-NiAl. I use the PAW-formalism. As explained in the description of the input variable 'positron', I first ran a self-consistent ground-state calculation with positron=0, and then a calculation with positron=1, using the output electron density of the previous run as input electron density. Two cuts through the resulting positron output density are shown in the attached figure: The blue line is a cut from 0,0,0 to 0,0,1 in the cubic notation, where 0,0,0 and 1,0,0 are occupied by Al-atoms. The green line is from 1/2,1/2,0 to 1/2,1/2,1, where the Ni-atom sits at 1/2,1/2,1/2. As it should be, the positron density is highest far from the atoms. What is puzzling, however, is the negative density at the position of the Ni-atom. How can this be?

To be specific, I performed both calculations with practically the same input file, just adding positron 1 ixcpositron 3 posocc 0.01 for the positron calculation. This means that I did the positron calculation with a rather high number of k-points. I would have thought that this should be no problem, it should just populate the lowest state. Looking into the output file, however, at all k-points the lowest band is filled with 0.01 (given by posocc), even though this band does show some dispersion. tsmear has no influence on this.

Granted, it is stated both in X. Gonze et al., Computer Phys. Comm. (2009) and in the description of the input variable 'positron' that for delocalised positrons one should use only the Gamma point. Doing that, the negativity of the positron density at the Ni-site vanishes, it even gets slightly, but significantly positive there (now why is that?).

So the question is: What is the reason for the negative density? Shouldn't that be forbidden somewhere deep inside (I mean in the code, in physics it is clearly forbidden)? Why is the occupation of the lowest band constant everywhere (to be sure, the negative density doesn't come from negative occupations due to smearing)? In the light of this, can I trust the results, which look plausible when I only consider the Gamma point?