During C14 Laves phase screening of Mn-Fe-Si compositions, I generated CIF files from the CrystaLLM route and attempted to parse them with ASE's CifParser. Certain structures -- specifically those with high symmetry operations in the _symmetry_equiv_pos_as_xyz block -- triggered parse failures where the parser rejected files that were structurally valid.
The failing pattern: symmetry operation strings containing unquoted fractional coordinates separated by commas, e.g.:
_y,x-y,_z
ASE's parser rejects these. The CIF standard technically requires quotes around operations containing operators like ,, but many ICSD-exported CIFs omit them and load fine in other tools (VESTA, pymatgen). ASE does not.
The workaround: wrap the raw CIF string with quotes around each symmetry operation before parsing, or use pymatgen's CifParser instead. The ASE-compatible C14 MgZn2-type CIFs I posted (Mn2Si, Fe2Si) use this workaround and parse cleanly.
What this means for screening workflows: CIF files from generative models may not parse in ASE even when the crystallographic content is valid. If you hit this, do not assume the structure is wrong -- try pymatgen first, or check whether the _symmetry_equiv_pos_as_xyz field contains unquoted commas.
This is documented as observed behavior, not as a confirmed bug in ASE. The ASE team may have legitimate reasons for strict parsing. The practical implication is that screening pipelines should be tolerant of this format variance or preprocess CIFs before feeding them to ASE.
On this page
CIF files with unquoted symmetry operations fail in ASE's CifParser even when structurally valid. Workaround and implications for screening workflows.