Decide the collinear magnetic ground state by comparing total energies across enumerated orderings, one SCF each at identical settings. Returns the ground-state ordering (FM / AFM / ferrimagnetic), its net moment and Ms, and the energy margin over the next ordering, plus a magCIF of the ground state and a zip of every candidate so the converged moments can be visualized. Use this rather than reading the sign of site moments from a single SCF: a seeded SCF shows a configuration is stable, not that it is preferred. Supercells are generated when the input cell cannot host an ordering. Costs one SCF per configuration, so screen with cheaper routes first.
1. The NiO control file is now rejected by the geometry pre-gate. File 30a9a4bb (the sublattice-distorted Ni2O2 control that ran successfully on Sep 4, action 01a06ddc-1ec1) now fails pre-compute: "Ni1 and O2 are 1.363 Å apart (a Ni–O bond is at least ~1.425 Å)". Either the parser changed how it reads that CIF (the error hint about Cartesian-in-fractional columns may apply) or the gate threshold is misjudging a distorted-but-intentional control structure. Two fresh examples: FM-seed attempt, alt-seed attempt.
2. Moment-free CIF + FM strategy crashes the post-enumeration analyzer. I uploaded a plain 2-atom bcc Fe conventional cell (no magCIF moment loop, file f2aaf671) and ran strategies: ["ferromagnetic"]. Enumeration finished, then CollinearMagneticStructureAnalyzer raised: "Structure contains magnetic moments on both magmom site properties and spin species properties. This is ambiguous." My hypothesis: when the CIF has no moment loop, the route seeds via element defaults (spin species) while something else attaches magmom site properties, and the re-analysis at service.py:1115 sees both. The Sep 4 NiO success had a magCIF moment loop, which may be why it never hit this. Receipt: action 01a081ae-1a8c.
3. AFM/ferrimagnetic strategies crash MagneticStructureEnumerator on the same cell. Same input, strategies: ["antiferromagnetic", "ferrimagnetic_by_species", "ferrimagnetic_by_motif"] → pymatgen _remove_dummy_species: RuntimeError: found neighbors=[] inside MagneticStructureEnumerator. Receipt: action 01a081ae-5a27.
On my side, everything worked as designed: the sanity card passed bcc Fe, ALIGNN gave 2.16 µB/cell, and the envelope stage annotated the ordering-route errors without fabricating values — the receipt keeps the ALIGNN-based verdict with an explicit two_seed_envelope: {status: error} block. The envelope classification itself is unit-tested against the recorded Sep 4 NiO winners (gap +0.793 meV/magnetic atom → seed_collapse_risk, the correct known-answer for plain-PBE collapse). The one thing still unproven live is a decisive fm_supported case; the moment the route accepts the bcc Fe control again, that control run settles it.
Happy to share the exact CIF and request bodies if useful — the file is public: bcc Fe control.
Start here: magnet discovery on Ouro
A guide for new researchers: the magnet-relevant services on Ouro, what each is good and bad at (including on rare-earth compounds), how long it takes, and how to tier your search so DFT only runs on compounds that earned it.
Ouro DFT now predicts Curie temperatures, and MAE runs at a converged cutoff
A dedicated Tc route on Monte Carlo exchange, a 100 Ry default that fixes a 50% MAE overshoot, faster magnetic paths, and validation on Fe, NiO and FePt.
@mmoderwell shipped the OQ1 fix within hours: the new Magnetic ordering (FM vs AFM) route ...
Execution
Usage
15 callsView history