TiNbFe4 (requested SG: I4/mmm #139, calculated: F-43m #216, optimized: 402 steps, cell relaxed, symmetry refined)
Ran a structure sanity card on this CIF. Short version: this structure is clean and the I4/mmm → F-43m transition is a legitimate symmetry refinement, not an artifact.
All checks pass. Minimum pair distance is 2.393 Å (Fe-Fe), all 24 sites are ordered, density is 7.75 g/cm³, and the reference-structure matcher shows zero displacement from ideal F-43m positions. F-43m is robust across every tolerance from 0.01 to 1.00 Å.
The bond lengths are remarkably uniform. Fe-Nb at 2.813 Å with zero spread, Fe-Ti at 2.812 Å with zero spread, Nb-Ti at 2.937 Å with zero spread. This is a textbook well-ordered intermetallic — no distortion, no disorder, no corruption.
The symmetry change is the right kind. Unlike the MgMn₂Co₂S₆ case (where Fm-3m collapsed to P1 during MLIP relaxation), here the symmetry went up: I4/mmm (tetragonal) → F-43m (cubic). The GGen action logs show that post-relaxation symmetry refinement detected the higher symmetry, re-expanded the cell from 12 to 24 atoms, and re-relaxed in 2 steps. The optimizer found that the tetragonal starting point was a suboptimal description of a structure that actually has cubic symmetry. That's a refinement, not a collapse.
Site analysis confirms a Heusler-type arrangement: Ti at 4a (0,0,0), Nb at 4c (¼,¼,¼), Fe at 16e (~⅛,⅛,⅝). Six atoms per primitive cell. This is a well-ordered F-43m intermetallic, and the structure is trustworthy as-is.
The one caveat: the relaxation used torch-sim (MLIP FIRE), same as the sulfide. But here the MLIP found higher symmetry than the starting point, which is the opposite of the symmetry-erasure pattern. The structure also converged cleanly in the post-refinement re-relaxation (2 steps). No red flags.
This CIF is good to go for downstream property prediction. The F-43m result is physically reasonable for a TiNbFe₄ intermetallic.