You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Symmetry unfolding (ibz2fbz) follow-ups: ignored rotations, magnetic time reversal, phases and orbitals under rotation #283
Found by independent verifiers during #280 and #281 (round 3 of the #277-#279 work). All items exist on dev at 4cde607; none was introduced by those PRs.
ibz2fbz overwrites its rotations argument with [] on its first line, so the parameter is ignored.
ibz2fbz(kgrid_info=None) raises NameError on new_in_original_grid_indices (the only in-tree caller always passes kgrid_info).
docs/source/user-guide/ebs.rst:68 calls an ebs.ibz2fbz method that does not exist.
ebs.py ~2125 -np.fmod(k+6.5,1)+0.5 negates every k-point, so each value lands at the image under -g. Harmless for time-reversal-symmetric data; wrong otherwise.
ebs.is_grid returns a false positive when the IBZ's distinct coordinates happen to form a full sub-grid; ibz2fbz is then skipped and the constructor raises.
ibz2fbz adds time reversal for every system, including magnetic groups without it. QE's time-reversal-flagged and lattice-only symmetries are treated as plain rotations.
projected_phase is not transformed on unfolded images (not conjugated under time reversal, not rotated).
Atoms and orbitals are not remapped under a rotation (px/py, dxy/dx2-y2 are not mixed; atoms are not permuted), so orbital-resolved characters and orbital-resolved spin at rotated images are approximate (measured on reduced BiSb: per-orbital spin error ~60% of typical values; summed spin ~2%).
Fixed in dev by #299 (5884b80). All eight items are done. The verdicts are on #299. The open limits are listed in the PR body: orbital weights without phases rotate by |M|^2; QE orbitals stay unrotated until #298 adds names; a magnetic non-collinear VASP mesh gets the time-reversal fill-in silently.
Activity
test(ebs): reproduce the #283 ibz2fbz defects
test(ebs): reproduce the #283 ibz2fbz defects
fix(ebs): make ibz2fbz unfold magnetic groups, phases, atoms and orbi…
lllangWV commentedon Oct 4, 2026
Fixed in dev by #299 (5884b80). All eight items are done. The verdicts are on #299. The open limits are listed in the PR body: orbital weights without phases rotate by |M|^2; QE orbitals stay unrotated until #298 adds names; a magnetic non-collinear VASP mesh gets the time-reversal fill-in silently.