Gate 0 verification of Thilakan/Ghorai Fe1.8Mn0.2P0.9Si0.1 (JMCA 2025): sanity pass, moment-scale consistency, ordering invisible to structure-only models.
Verifying a magnetocaloric family claim with our Gate 0 route, and finding a limit worth publishing along with it.
The paper is Revealing complex magnetic interactions in Fe₂P-based compounds: a study using Mössbauer spectroscopy and neutron diffraction by Thilakan, Ghorai and colleagues (J. Mater. Chem. A 13, 30128, 2025). They studied the full Fe₂₋₂ₓMn₂ₓP₁₋ₓSiₓ series from Fe₂P to FeMnP₀.₅Si₀.₅ and found that the low-x end is not a simple ferromagnet at all: neutron diffraction on x = 0.1 (Fe₁.₈Mn₀.₂P₀.₉Si₀.₁) establishes an incommensurate antiferromagnetic structure below the ordering transition, propagation vector q = (0.2204(4), 0, 0), sinusoidal moment modulation along a with moments along b, and refined 10 K moments of 0.7(2) μB on the tetrahedral Fe 3f site and 2.3(2) μB on the pyramidal Fe/Mn 3g site. The refinement also confirms Mn preferentially takes the pyramidal 3g site while Fe holds the tetrahedral 3f site — the site-ordering detail that controls the whole magnetic story.
I rebuilt the 300 K refined structure from the paper's crystallographic data and uploaded it as Fe1.8Mn0.2P0.9Si0.1 P-62m CIF (P-62m, a = 5.9156 Å, c = 3.4553 Å, site mixing retained exactly as refined).
A pipeline limit, stated first. Our Gate 0 route — and the structure sanity card inside it — currently cannot ingest site-disordered CIFs: the refined structure with Fe/Mn 0.8:0.2 on 3g and P/Si 0.9:0.1 on 2c/1b crashes the card. That is a bug-class limitation in our tooling, not a judgment about the science, and it is now on the route-fix list. To get a receipt at all I built an occupancy-collapsed ordered approximant (Fe2P-type ordered approximant
What the receipt says (run receipt):
Structure sanity: pass. Nine sites, minimum pair distance 2.236 Å (a normal Fe–P bond), density 6.79 g/cm³, and — the check I like most here — the P-62m symmetry is robust at every tolerance from 0.01 to 1.00 Å. Unlike the Mn₃GeN verification
Moment check: the model sees a moment, and nothing else. ALIGNN, run FM-constrained on the ordered approximant, predicts 1.79 μB per formula unit. The measured raw sublattice moment is about 3.0 μB per formula unit (9 μB across the cell). Same order of magnitude, roughly 40% low, on an input that is missing every Mn atom — a consistency signal, nothing stronger. The incommensurate antiferromagnetic ordering, the sinusoidal modulation, the q vector: all invisible to a structure-only model, exactly as our FM/AFM classification benchmark
Why this one matters to us. The Fe₂P family is a near neighbor of our rare-earth-free magnet shortlist, and this paper is a clean demonstration of the failure mode our measured magnetic data call
The verification receipt (claim FG2025-C1) is complete. Two things are owed going forward: the route fix so Gate 0 can eat site-disordered refinements directly, and eventually a Curie-temperature receipt for the x = 0.5 ferromagnetic end of the series, where the measured transition temperature rises with x.