Structure with f-electron element symmetry change after relaxation

[yujia_teng]

Dear admin,
I came across with the problem that the symmetry (space group/Bravais lattice) of my structure is changed after relaxation, either isif=3 fully relax or isif=2 atomic position only, even if I increased the precision by setting SYMPREC=1E-8. The problem is with the supercell (primitive cell is fine), but it's also the magnetic primitive cell. Is it due to f-electron? I didn't met with this kind of problem before. The input files are attatched.

Best

[henrique_miranda]

I am not sure I understand your question exactly.
Do you mean that the spacegroup symmetry is lost during the ionic relaxation?
Or do you mean that the symmetry changes because of a particular setting of MAGMOM?
Looking at your "primitive cell" files I see:

% grep symmetry OUTCAR
Analysis of symmetry for initial positions (statically):
Routine SETGRP: Setting up the symmetry group for a
The static configuration has the point symmetry C_3v.
Analysis of symmetry for dynamics (positions and initial velocities):
Routine SETGRP: Setting up the symmetry group for a
The dynamic configuration has the point symmetry C_3v.
Analysis of structural, dynamic, and magnetic symmetry:
Routine SETGRP: Setting up the symmetry group for a
The magnetic configuration has the point symmetry C_3v.

in the case "relax atomic position only" I see:

% grep symmetry OUTCAR
Analysis of symmetry for initial positions (statically):
Routine SETGRP: Setting up the symmetry group for a
The static configuration has the point symmetry C_3v.
Analysis of symmetry for dynamics (positions and initial velocities):
Routine SETGRP: Setting up the symmetry group for a
The dynamic configuration has the point symmetry C_3v.
Analysis of structural, dynamic, and magnetic symmetry:
Routine SETGRP: Setting up the symmetry group for a
The magnetic configuration has the point symmetry C_1h.

which means that the system has a lower symmetry because of the MAGMOM setting.
Now in the case of "fully relax" I see:

% grep symmetry OUTCAR
Analysis of symmetry for initial positions (statically):
Routine SETGRP: Setting up the symmetry group for a
The static configuration has the point symmetry C_1h.
Analysis of symmetry for dynamics (positions and initial velocities):
Routine SETGRP: Setting up the symmetry group for a
The dynamic configuration has the point symmetry C_1h.
Analysis of structural, dynamic, and magnetic symmetry:
Routine SETGRP: Setting up the symmetry group for a
The magnetic configuration has the point symmetry C_1 .

In this last case you are starting from a structure that already does not have the same symmetry as in the previous cases (C_1h instead of C_3v).
The MAGMOM tag only breaks the symmetry further to C_1.
In these cases it can happend that the symmetry is recovered (if that is the lowest energy structure) but it is not guarateed because lattices vectors and positions have to become symmetric enough for VASP to find the correct symmetry again. Can you try running "fully relax" but starting from the same structure as "relax atomic position only"?

[yujia_teng]

I mean space group symmetry is lost during the ionic relaxation.

The starting structure is same. Different symmetry recognized by vasp is caued by SYMPREC = 1.0E-08 higher precision. I re uploaded one with default SYMPREC value.

Check the CONTCAR file, the relaxed structure. What I mean is after relaxation, the space group is changed from P6_3mc (No.186) to Cmc2_1 (No. 36). (CONTCAR space group different from POSCAR) This happens both when I fully relax it (ISIF=3) or atomic position only (ISIF=2). I'm not sure why, and doubt it may due to f-electron. Because the structure is reported in experiment paper, and is stable.