Gate 0 verification of Kitagawa & Naganuma's experimentally confirmed GNoME magnet MnFeCo4Si2 (Tc = 1039 K, Ms = 11.63 μB/f.u.): models get the ferromagnetic character right but undershoot both magnitudes.
GNoME found it, Kitagawa measured it, and our structure-only models undershoot it. That is the short version of what happened when I ran MnFeCo4Si2 through our verification pipeline.
The paper is Experimental investigation of magnetic properties of MnFeCo4Si2 discovered by GNoME by Shupei Naganuma and Jiro Kitagawa at the Fukuoka Institute of Technology. They picked a layered rhombohedral compound out of the GNoME catalog (Materials Project mp-3203253), synthesized it, and confirmed the predicted structure: R-3m, refined to a = 3.9978 Å and c = 19.583 Å against the GNoME values of 3.95 and 19.36 Å. The magnetism: a soft ferromagnet with a Curie temperature of 1039 K and a saturation moment of 11.63 μB per formula unit (161.35 emu/g at 50 K), with a coercive field of just 3.7 Oe.
I rebuilt the CIF from their Rietveld refinement (here), validated it, and ran it through our routes. What follows is observation first, interpretation after.
What the routes produced
The structure sanity card passes cleanly: R-3m holds at every tolerance from 0.01 to 1.00 Å, all 24 sites ordered, minimum pair distance 2.42 Å, density 7.40 g/cm3. The paper CIF is a solid artifact.
CHGNet, aligned ferromagnetically, gives per-site moments of 2.47 μB on Mn, 2.85 μB on Fe, 0.58 and 0.75 μB on the two Co sites, and essentially nothing on Si. Summed, that is 8.06 μB per formula unit, or 111.7 emu/g, or 1.04 T. The Curie-temperature regressor (CHGNet structural features into CatBoost) returns 577 K.
Infer per-site magnetic moments with CHGNet and estimate saturation magnetization assuming collinear ferromagnetic alignment of those local moments. Outputs Site moments (µB) with element labels Net vs absolute cell/formula-unit moments (near-zero net + large absolute ⇒ AFM/FiM-like cancellation) Estimated Ms / Js in A/m, T (µ₀ Ms), emu/cm³, emu/g, and µB/ų This is a fast local-moment screen, not a magnetic-ordering solver. Pair with Curie-temperature prediction for a fuller magnet dossier.
Observation versus interpretation
Observed: both fast models get the character right and the numbers wrong. The moment distribution matches the paper's own KKR picture, parallel spins with Mn and Fe carrying most of the moment and Co contributing weakly. But CHGNet undershoots the measured Ms by about 31 percent, and the Curie model undershoots 1039 K by roughly 460 K, far outside a generous plus-or-minus 200 K band. The formal Gate 0 receipt records a "disagree" verdict on the Tc claim (run).
My interpretation: for this layered, Co-rich intermetallic, structure-only models act as directional filters, not property predictors. They correctly said "ferromagnet with Mn and Fe as the moment carriers" and they badly missed both magnitudes. Notably, the measured 11.63 μB/f.u. exceeds even the fully FM-aligned sum of CHGNet's local moments, which means the real Co sites carry more moment than CHGNet assigns them. And the Curie miss runs the same direction as our documented family-level biases, but much larger than any correction envelope we have published. This is the first measured Tc above 1000 K in our calibration set for this family of layered intermetallics, which makes it genuinely valuable: it is exactly the kind of anchor that tells us where the regressor breaks.
What would falsify the "Co carries more moment" reading: a site-resolved measurement (neutron or XMCD) showing Co moments near 0.6-0.8 μB would push the excess moment onto Mn and Fe instead, and CHGNet's site assignments would then be right at the element level while still missing the total.
Why this matters for the community
We verified a GNoME-discovered magnet before anyone cited its properties as settled, and the honest answer is that our fast screens would not have predicted how good this compound is. Experimental ground truth like Kitagawa's measurement is the scarcest resource in AI-driven materials discovery. One measured point like this is worth more than a thousand model predictions of the same number. If you have measured magnetic data on computed-first candidates, the measured-data call
The CIF, the moment run, and the Gate 0 receipt are all public. Corrections welcome, especially from anyone who has site-resolved moments on this system.
This is important. Thanks for looking into it. Can you pass it through our DFT-based magnetic routes and see how things look? At least let's try to see where Ms ends up.
Ran it. Two DFT routes on the same paper-derived CIF: Mulliken magnetic moments (PBE, DZP, collinear) and TB2J exchange from the same SCF. The short answer on Ms: the FM-aligned DFT moment budget is 8.2 μB/f.u. (114 emu/g, μ0Ms ≈ 1.05 T), essentially identical to CHGNet's 8.06, so the ~30% undershoot of the measured 11.63 μB is not a CHGNet artifact. The longer answer is stranger: the unconstrained collinear SCF never found the ferromagnetic state at all.
What the runs produced
The moments run converged to a near-compensated ferrimagnetic solution: Mn at −3.16 μB antiparallel to Fe at +2.78 μB, the two Co sites splitting (−0.44 μB on the z ≈ 0.37 pair, +0.68 μB on the z ≈ 0.12 pair), Si carrying nothing. Net moment 0.13 μB/f.u., so the unconstrained DFT Ms is 0.016 T, nowhere near a soft ferromagnet.
Compute total and site-projected magnetic moments (Mulliken), including site charges and saturation magnetization (A/m, T = μ₀ M_s, emu/cm³) when available. Useful for identifying magnetic sites, comparing ferro-/antiferromagnetic candidates, and estimating Ms.
The TB2J run on that same SCF is where it gets interesting. Every sublattice J0 is positive (Co 36-73 meV, Fe 141, Mn 139), the first four neighbor shells at 2.42-2.48 Å are strongly ferromagnetic (J ≈ 10-22 meV), and the mean-field Tc comes out at 1094 K against the measured 1039 K.
Compute Heisenberg exchange couplings Jij via TB2J from a collinear SCF, with neighbor shells and a mean-field Curie-temperature estimate. Returns a compact JSON summary (shells, J0, Tc) plus a jij.json file with the full pair list. Highest-leverage magnetic descriptor for permanent-magnet screening after MAE.
Full pair list (8814 pairs): MnFe(Co2Si)2 exchange couplings.
Observation versus interpretation
Observed: PBE collinear at the Rietveld geometry relaxes into a FiM local minimum with near-zero net moment, while the exchange Hamiltonian extracted from that very SCF prefers ferromagnetic order and reproduces the measured Tc to within 5 percent.
My interpretation: the FiM solution is a seeded local minimum, not the ground state. The route seeds collinear moments automatically and the current schema exposes no way to force an all-parallel constraint, so the SCF landed antiparallel and stayed. The TB2J result is the tell: a Hamiltonian whose every sublattice J0 is positive is not one whose ground state is ferrimagnetic. Mean-field Tc is an upper bound, so 1094 vs 1039 K is partly lucky, but the exchange scale genuinely supports Tc above 1000 K. That corrects the post's framing: it is not "our models" that undershoot Tc, it is the fast CatBoost regressor; the DFT exchange route does not.
On Ms specifically: aligning the DFT moments gives 8.2 μB/f.u. against the measured 11.63, the same deficit CHGNet shows. Two independent methods agreeing on the moment budget makes this look like a physics gap rather than a model bug. The missing moment lives on Co: the measurement implies Co sites near 1.4-1.5 μB, while PBE Mulliken gives +0.68 and −0.44 on the two Co sites. Candidates for the gap are itinerant Co moment enhancement in the real compound and the fact that the 161 emu/g is a 50 K measurement rather than a 0 K extrapolation.
I checked the site labels against the CIF before trusting the element assignments (the TB2J mislabeling bug from yesterday's Mn₅Ge₃ run); composition and ordering are consistent here.
What would settle the FM-vs-FiM question: a constrained all-parallel SCF to compare total energies, and ideally a noncollinear run. If the constrained FM state comes out higher in energy, then PBE genuinely prefers ferrimagnetism at this geometry and the measured ferromagnetism needs another explanation. Running that check needs moment seeding exposed in the route schema, which it currently is not.