Fe22B6W (mp-1224794) is in the wrong magnetic state in Materials Project

While checking a B-Fe-W energy-above-hull run for a Fe12BW4 candidate, one reference entry stood out. Our phase diagram puts Fe22B6W (mp-1224794) on the convex hull. Materials Project puts it 0.148 eV/atom above the hull.

The diagram comes from this run of the energy-above-hull route. The route doesn't reuse MP's DFT energies. It re-relaxes every MP reference with Orb v3 (conservative, inf, MPA) and builds the hull from Orb energies, so the disagreement is Orb versus MP's GGA calculation.

The MP calculation is almost non-magnetic

Fe22B6W is the τ-phase Fe23B6 (Cr23C6-type, 29-atom primitive cell) with one Fe replaced by W. MP's own Fe23B6 entry is the natural comparison:

Fe23B6 has the expected moment for an iron-rich boride. The W-substituted entry has almost none. Replacing one Fe in 23 with W can't plausibly wipe out magnetism across the whole cell, which suggested MP's run converged to a low-spin solution.

Checking it with DFT

To test that directly, I ran the Magnetic moments route twice on MP's own Fe22B6W structure (CIF

Ferromagnetic (nspin: collinear):

Non-magnetic (nspin: non_spin):

The ferromagnetic state is 6.12 eV per cell lower, or 211 meV/atom (278 meV per Fe). B carries a small antiparallel moment (−0.33 μB) and W carries −0.82 μB, as expected for an early 5d atom in an Fe host. The saturation magnetization at this geometry is about 1.8 T. The converged moments are in the ferromagnetic magCIF.

That 211 meV/atom is larger than the 148 meV/atom separating MP's entry from the hull. Apply it to MP's number and Fe22B6W moves from 0.148 above the hull to roughly 0.06 eV/atom below it (formation energy about −0.26 eV/atom). That's consistent with Orb placing it on the hull.

How firm is this

• The direction is solid. MP's Fe22B6W energy is for a nearly non-magnetic state that is far from the ground state. The conclusion that it belongs on or near the hull survives unless the magnetic energy is overestimated by more than 30%.

• The exact magnitude isn't. The spin-state energy difference comes from a different code, basis, and pseudopotentials than MP's VASP PAW calculation. Magnetic energies are broadly transferable between codes, but not to the meV. The honest statement is "on or somewhat below the hull," not "60 meV below."

• Geometry favors the magnet further. MP's geometry was relaxed in the low-spin state. A ferromagnetic relaxation would expand the cell (magnetovolume effect) and lower the energy further, so using MP's geometry is conservative.

• The full fix would be to recompute Fe22B6W, Fe23B6, and the relevant B-Fe-W references ferromagnetically at one consistent setting and rebuild the hull. That's worth doing if Fe22B6W becomes a candidate in its own right.

Takeaway for hull screening

MP energy_above_hull values for magnetic compounds are only as good as the magnetic state each calculation converged to. A quick sanity check is to compare total_magnetization against a close structural analog. An Fe-rich entry with a near-zero moment is a red flag. Here it hid a phase that is probably thermodynamically stable behind a 148 meV/atom penalty, and an MLIP hull caught it.