ABINIT Discussion Forums

Hi,

in trying to look at rhombohedral Selenium, I run into the following problem. In this structure, Z=3 and Se sits on Wyckoff site 18 f. But this is only the conventional, not primitive, cell. I use the following input:

acell 11.362 11.362 4.429 angstrom
angdeg 90.0 90.0 120.0
spgroup 148
brvltt -1

natom 6
natrd 1
xred 0.16021 0.20227 0.1205

and get as error:
symlatt : BUG -
One of the coordinates of rprimd in axes is non-integer,
or non-half-integer (if centering).
coord= 6.6666666667E-01 6.6666666667E-01 -3.3333333333E-01
-3.3333333333E-01 6.6666666667E-01 6.6666666667E-01
6.6666666667E-01 -3.3333333333E-01 6.6666666667E-01
fact= 1
Action : contact ABINIT group.

This *seems* like a genuine bug but maybe my input is wrong. Any comments or suggestions?

thanks,
Joe

Hello Joe,

I'm puzzled by you crystal setup in the input. Starting from http://www.crystallography.net/cif/9/9013138.cif, why don't you let the symmetriser do its job with a plain ...

acell 11.362 11.362 4.429 angstrom
angdeg 90.0 90.0 120.0
spgroup 148
natom 18
natrd 1
xred 0.16021 0.20227 0.12045

... I'm maybe in a very bad crystallographic karma today.

Kind regards,

Alain

Hi Alain,

your input suggestion doesn't work because the conventional cell as given by the cif file is not primitive, the primitive cell has only 6 Se atoms in it (I'm pretty sure) but the abinit symmtetry finder is failing to construct it properly (I think). Anyway, I tried your suggestion with all permutations of brvltt and natom, if brvltt /= -1 it stops by saying the cell is not primitive, if brvltt = -1 then natom = 6 is required; and if brvltt = -1 and natom = 6, it stops with the bug message I reported first.

Joe

Right the unit cell in the cif isn't primitive.
Did you try with chkprim 0 ? Default is 1 and Abinit will stop if the cell isn't primitive.

Alain

Yes I know, that's what the brvltt -1 is for, to force it to find the primitive cell. This usually works (for example, with body-centered tetragonal cells and so forth).

Joe

This case seems really buggy in the symmetry finder. If I input the conventional cell, and use chkprim 0, then use prtkpt 1 to generate a lattice of k points, abinit stops saying that the kpt grid breaks symmetry. There really seems to be a problem here. Who is the expert on the symmetry finder system?

Joe

Definitely the Boss but I'm afraid that he is swamped with paperwork for the moment.
I'll try to help by having a look again this weekend.

Alain

Hi Boys, did you miss me?

in abinit spgroup 148 is incompatible with the hexagonal cell you are imposing: rhombohedral primitive is the trigonal cell with
angdeg 3*<stuff>
acell 3*<astuff>

Of course this also means you have to convert the xred for the single irred atom... Possibly go by cartesian coordinate instead?

Matthieu

Hello Matthieu,

Mea culpa, I forgot you were introduced in the symmetrizer black magic. It's a nice suggestion - en bon belge, mon franc vient de tomber But I would like to reproduce the prtkpt behavior reported by Joe; seems odd to me.

Alain

Hi,

I'll try your suggestion Matthieu but I don't understand why brvltt -1 doesn't obtain that result for you. In the .cif file the cell is always given as hexagonal in the rhombohedral case; is the conversion not coded? I'm comparing for example to body centered tetragonal, where you input a tetragonal cell and abinit converts internally to the correct primitive cell.

Joe

Indeed... I think it's just not coded. Getting the Innen or FCC primitive axes from conventional is slightly easier than trigonal from hexagonal. Matteo Giantomassi, Xavier, or Razvan Caracas should know exactly where the brvlatt -1 treatment happens.

cheers

Matthieu

Matthieu, I tried your suggestion: generated the rhombohedral cell two ways, one by cartesian vectors input as rprim with acell = 3*1.0, and also by acell 3*12.7063922007622
and angdeg 3*115.32088917441888. I also found the irred atom in cartesian coords, inputted back with xcart or with xred with the new cell, I always get the same error message:


symlatt : the Bravais lattice is hR (rhombohedral)

symlatt : BUG -
One of the coordinates of rprimd in axes is non-integer,
or non-half-integer (if centering).
coord= 6.6666666667E-01 6.6666666667E-01 -3.3333333333E-01
-3.3333333333E-01 6.6666666667E-01 6.6666666667E-01
6.6666666667E-01 -3.3333333333E-01 6.6666666667E-01
fact= 1
Action : contact ABINIT group.

Any ideas? Is this a real bug or am I still inputting the xtal data wrong? I've attached the cif file I'm using in case that might be helpful.