[Wien] Re : Supercell problem of Al2O3

[Michael Fischer]

Hello,

I suspect that I misunderstood the definition of "hexagonal cell" in 
WIEN2k. I presumed that the translations
  (2/3,1/3,1/3) +
(1/3,2/3,2/3) +
are automatically applied when this lattice type is chosen, like a 
translation (1/2, 1/2, 1/2) that is automatically applied for 
body-centered lattices.
Apparently this is not the case, and I now understand that this is 
logic, because this translation is only applicable for rhombohedral 
structures in hexagonal setting (where the hexagonal cell is a kind of 
supercell, three times larger than the primitive one), but not for 
"genuine" hexagonal structures.
This explains why initialization of "a.struct" comes up with another, 
"genuinely hexagonal" space group.

I suspect that  "A_latest.struct" does not work because you did enter 
the fractions exactly, i.e. you used 0.33333000 instead of 0.33333333. 
If you use W2Web, I suggest you enter the fractions as "1/3" or even 
something like "0.25+1/3", this should also work and guarantee highest 
possible precision.
Before you tackle this, it might be sensible to check the 
outputnn-file. If you find that all bond lengths are very close to the 
expected values, the structure should be "nearly" correct, and the fact 
that the correct symmetry is not found is probably caused by some error 
like the one suggested above. If some bond lengths deviate strongly 
from the expected values, the error is probably more severe.
(By the way, check the following: Shouldn't it be (0.30625 0 0.25), 
(0.69375 0 0.75) etc. for oxygen, rather than (0.30625 0 0.75) etc.?)

Hope this helps, Greetings
Michael