[Wien] unphysical growth of mag. moment with supercell size

[Maciej Zwierzycki]

Dear WIEN2k Users,

I am trying to calculate the electronic structure (DOS, LDOS) of
magnetic impurity/add on atom (Co) on the surface of silicene
(i.e. Si graphene). While checking the convergence vs. the supercell
size I've noticed the following artifact: the size of the magnetic
moment on Co stays roughly constant (about 0.8 \mu_B) up until certain size of the
supercell (say 9x9) and then starts growing, in some cases up to the
atomic values (~3 \mu_B). I don't see how this could be physical, as
the Co-silicene distance and the local geometry around impurity stays
the same, however I was not able to find any obvious errors in my
calculations.

The calculations for bulk silicene work well.

I'm running 14.2 version, the vertical separation between silicene
sheets is 14 Angst, the lattice constant is a=3.86 Angst. I've used
rKmax between 3 and 7. The k-mesh was generated using (3,3,1)
divisions but I've also tested with denser 2D meshes and the
automatically generated meshes. The symmetry, determined during the
initialization phase was the same in all the cases, e.g.:

H   LATTICE,NONEQUIV.ATOMS: 61 156 P3m1                                        

I've started these calculations by extracting the geometry of local Co
neighborhood from the relaxed calculations with 4x4 supercell and
embedding it into bulk silicene when running larger supercells. The
calculations were then performed with thus fixed geometry. However the
same trends can be observed with unrelaxed geometry. In the latter
case I fashioned the supercell  out of bulk unit cell and then
attached Co atom at arbitrary height. 

I'd be happy to provide further informations (structure files) and grateful
for any hints you may give me.

Kind Regards
Maciej Zwierzycki

PS. At the moment I'm trying to establish to what extent the results
are sensitive to separation between silicene sheets.