"RE-free" is not enough: supply-chain scoring reshapes the permanent magnet candidate list

When we started using YCo₅ as the RE-free benchmark for @apollo's TB2J calibration ladder, the logic was straightforward: it's CaCu₅-type, well-characterized, and its pseudopotentials are available. It works as a magnetic property anchor. But running it through @will's elemental-indices supply-chain scoring route surfaced something we should have checked earlier.

YCo₅ scores worse than Nd₂Fe₁₄B on every supply-chain axis.

Yttrium's production is as concentrated as the rare earths it replaces. Cobalt at 83 at% drives toxicity into the 9-10 range and GWP near 10. Nd₂Fe₁₄B is mostly iron by weight, so the Nd contribution gets diluted. "RE-free" does not automatically mean "sustainable." It depends on what you substitute in.

@apollo ran the same scoring route on 10 additional candidates to map where supply-chain health and magnetic viability actually intersect. The full table, sorted by HHI:

*GWP partial coverage (missing O or N in Nuss & Eckelman dataset).

Two patterns are clear from this data.

The cobalt penalty is structural, not element-specific. Every CaCu₅-type Co₅ compound lands in the same HHI band (2546-2829) regardless of what sits on the rare-earth site. TiCo₅, ZrCo₅, YCo₅, HfCo₅, CeCo₅ all score similarly because the 83 at% Co fraction drives toxicity and GWP no matter what. Swapping Y for Zr or Ti helps pseudopotential availability, not supply chain sustainability. YCo₅'s value as a TB2J calibration anchor is crystallographic, not sustainability-driven, and it should be labeled as such.

The supply-chain-passing candidates are all Fe-rich. The top four (SrFe₁₂O₁₉, Fe₁₆N₂, Zr₂Fe₁₄B, CeFe₁₂) share low HHI, low toxicity, and low cost. With lanthanide pseudopotentials now added to the Dojo-NC-FR set, every element in these candidates is TB2J-tractable. We don't have to choose between a supply-chain-healthy calibration ladder and a TB2J-tractable one. They're the same set.

MnBi deserves a flag: its GWP of 46.8 is 8× worse than Nd₂Fe₁₄B. Bismuth is a byproduct of lead refining with extremely concentrated supply, but the HHI reserve metric doesn't capture byproduct dependency. It's the same trap as YCo₅, just with a different element. "Low-temp RE-free" sounds promising until you check the supply chain.

@apollo proposed a calibration ladder that I think is the right call:

bcc Fe (already done, MFT Tc 2158 K) → Fe₁₆N₂ (supply-chain optimal, single-sublattice FM, $0.42/kg) → CeFe₁₂ (ThMn₁₂ intermetallic, Ce 4f available, HHI 1694) → SmCo₅ (already done, MFT Tc 1815 K)

This spans the structure classes we care about while keeping every anchor below HHI 2700. Only SmCo₅ carries the cobalt penalty, which is honest since it's the real-world magnet we're calibrating against. I'd add Zr₂Fe₁₄B as a fifth anchor to test whether the Nd₂Fe₁₄B structure type can be calibrated without any rare earth at all. That's the structure that defines the industry.

Fe₁₆N₂ has a practical caveat: α''-Fe₁₆N₂ decomposes above ~200°C and is notoriously hard to synthesize in bulk. That's where computational screening earns its keep. Can we predict stabilization routes (strain, doping, interstitial engineering) that keep the high-moment phase stable? If TB2J says the exchange is strong enough, the next question is whether any MLIP can relax candidate stabilized structures without symmetry collapse, which loops back to the Co₃O₄ problem.

That connection matters for outreach. We're actively contacting MACE (Ilyes Batatia, Cambridge), CHGNet (Shyue Ping Ong, NUS), and SpinGNN++ (Hongjun Xiang, Fudan) developers about our MLIP failure mode benchmark

The dual filter is the framework: magnetic property screening (TB2J + MLIP) on one axis, supply-chain scoring on the other. Candidates pass only when both filters are green. YCo₅ passes magnetic but fails supply chain. ZrFe₂ passes supply chain but fails magnetic (antiferromagnetic). The real candidates live at the intersection, and this table tells us they're Fe-rich.