[Wien] Slab calculations

[Peter Blaha]

> I am trying to do a 6 layer slab calculation of Mg(0001) and I get energy
> oscillations in the scf cycle. I already used a Rkmax of 10 and gmax of 14.
> Does anybody have experience for these kind of calculations? What is the
> crucial parameter to reach convergence? I have not optimized the positions,
> but this should not play a role for convergence, right?
>
> Thanks a lot
>
> Mg MT sphere: 3
> unit cell: 25 angstroem
> vacuum thickness 12 angstroem

What kind of E-oszillations ? Ry, mRy or even less ?

NEVER (almost never) use RKMAX=10 !
Mg as a very soft sp element can probably even be done with RKMAX=5 !,
but of course 6 or 7 is better. This speed up the calculation by a
factor of 100 !?! and you avoid ghostbands.

MT sphere  3   ??  bohr ??  seems very large ??

Mg is a metal, so what is your k-mesh ? You need a good one!

There's one (not easily solvable) problem, and this is that the k-mesh
along z (you should have just one in your mesh!) is determined by the number
of layers through "backfolding". Improvement is only possible by using
more layers.
This will cause some funny oszillations in the DOS (but not that I know
in the energy).

So use a good k-mesh and eventually a broadening scheme (TEMP 0.010 ).

For surfaces, one should "correct" the E-parameters in case.in1.
Replace all "0.3" by appropriate values (below your EF).


                                      P.Blaha
--------------------------------------------------------------------------
Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna, A-1060 Vienna
Phone: +43-1-58801-15671             FAX: +43-1-58801-15698
Email: blaha at theochem.tuwien.ac.at    WWW: http://info.tuwien.ac.at/theochem/
--------------------------------------------------------------------------