replace triu by triu_indices

[samwaseda]

Closes #246

[coveralls]

Pull Request Test Coverage Report for Build 1884463820

• 4 of 4 (100.0%) changed or added relevant lines in 1 file are covered.
• 203 unchanged lines in 4 files lost coverage.
• Overall coverage increased (+0.03%) to 70.227%

---

Files with Coverage Reduction	New Missed Lines	%
pyiron_atomistics/atomistics/master/murnaghan.py	14	72.98%
pyiron_atomistics/lammps/control.py	28	89.05%
pyiron_atomistics/vasp/base.py	50	65.29%
pyiron_atomistics/sphinx/base.py	111	81.82%

Totals
Change from base Build 1841593007:	0.03%
Covered Lines:	11685
Relevant Lines:	16639

---

💛 - Coveralls

[pmrv]

As discussed in #246 I've tried all combinations of running the job for an HCP cell. Results are below first as matrices then as bar plots.

Generally orthorhombic structures seem to work best, but notice in the bar plot the inconsistency between c31 and c32 w/ pressure and c44 and c55 w/o pressure, which are supposed to be equal in HCP.

[samwaseda]

As discussed in #246 I've tried all combinations of running the job for an HCP cell. Results are below first as matrices then as bar plots.

Generally orthorhombic structures seem to work best, but notice in the bar plot the inconsistency between c31 and c32 w/ pressure and c44 and c55 w/o pressure, which are supposed to be equal in HCP.

For primitive I have frankly no idea why it's not working. Otherwise is it possible that the lattice parameter is not relaxed beforehand?

[pmrv]

As discussed in #246 I've tried all combinations of running the job for an HCP cell. Results are below first as matrices then as bar plots.
Generally orthorhombic structures seem to work best, but notice in the bar plot the inconsistency between c31 and c32 w/ pressure and c44 and c55 w/o pressure, which are supposed to be equal in HCP.

For primitive I have frankly no idea why it's not working. Otherwise is it possible that the lattice parameter is not relaxed beforehand?

I relaxed all the structures with calc_minimize(pressure=[0]*6, ionic_force_tolerance=1e-6) beforehand.

[samwaseda]

I relaxed all the structures with calc_minimize(pressure=[0]*6, ionic_force_tolerance=1e-6) beforehand.

hm ok then I'm going to include a symmetrisation step

[samwaseda]

I realised that it's related to the fact that the point group symmetries that spglib do not help us symmetrize those values, so that for now it can be only avoided by increasing the number of points. In theory I guess it's possible to solve this problem by reducing the structure to the smallest unit and get all the symmetries there, but I'm not really sure how to do it properly. One way or other, this is more a fundamental problem, which is not directly related to this PR, so I'd suggest to merge this one and open another one as soon as we find a solution.

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[pmrv]

I realised that it's related to the fact that the point group symmetries that spglib do not help us symmetrize those values, so that for now it can be only avoided by increasing the number of points. In theory I guess it's possible to solve this problem by reducing the structure to the smallest unit and get all the symmetries there, but I'm not really sure how to do it properly. One way or other, this is more a fundamental problem, which is not directly related to this PR, so I'd suggest to merge this one and open another one as soon as we find a solution.

I think both spglib and pymatgen have methods to get reduced unit cells from super cells, but I haven't looked into how they work. If increasing the number of points works, should we just bump the default value for that?