This loop asked whether a new family, with no precedent in the literature, could give a rare-earth-free magnet. The family is iron-rich compounds whose iron-group atoms sit on the same kind of site that gives YCo₅ its anisotropy. The concrete target was a made Fe–Zr–Y phase, (Y,Zr)Fe₃, which nobody had read as a magnet.
No new candidate. Every cell keeps about 1.17–1.19 T, but each prefers its magnetization to lie in a plane (easy-plane anisotropy) instead of along one axis.
A survey of the crystal database found that, among rare-earth-free compounds near the hull with 0.9 T or more, the YCo₅-type site occurs only in intergrowths of the CaCu₅ and Laves structures built on Y, Zr or Hf. With iron on that site these are easy-plane. The made phase and its niobium and titanium cousins only shuffle the second metal; that moves the anisotropy by a few tenths of an MJ/m³ and never across zero. A recently published way to turn YFe₃ uniaxial, by putting Nb, V or Ta on an iron site, turns out to sit 45 meV/atom or more off the hull, because niobium would rather take the second metal's site. That may be why the authors' arc-melted samples did not take Ta into YFe₃. The loop closes this direction for about $2.3 and leaves the campaign's standing candidates as they were.
PBE in the DZP basis, kspacing 0.2 Å⁻¹. Each cell is relaxed in DFT, then measured on its own relaxed cell as a 12-atom rhombohedral primitive that keeps the threefold axis. K is the energy with the moment in the plane minus the energy along c, so a negative value means easy-plane; the in-plane directions agreed within 0.01 MJ/m³. Hull distances are on the Orb surface.
Cell | Geometry | Novelty | K along c (MJ/m³) | Js (T) | Orb hull (meV/atom) |
|---|---|---|---|---|---|
YFe₃ (calibration) | |||||
The targets are K ≥ 1 MJ/m³ along one easy axis and Js ≥ 1 T. The YNb₂Fe₉ value is the one at the narrowest band-sum width; it reads −0.45 at the default width and −0.31 at half of it. YZr₂Fe₉ costs 12.42 $/kg. Our route reads YFe₃ at 0.74–0.78 of a published full-potential PBE value (−0.90 MJ/m³, arXiv:2610.18303), with the sign seen in samples, so the sizes above are likely somewhat small and the signs are right.
YZr₂Fe₉ as relaxed: Y sits in the CaCu₅-type layer and Zr in the Laves-type layers. The iron atoms in the Y plane are the YCo₅-type sites, and they carry the easy plane:
A free screen of 174 arrangements of Y and Zr on the two metal sites gave three results. Zr strongly prefers the Laves layers: with Zr on the CaCu₅-type site the cell sits 58 meV/atom above the hull, against 5 with Zr in the Laves layers at the same composition. Mixed Y/Zr arrangements lie only 1.7–5 meV/atom above the line between YFe₃ and YZr₂Fe₉, so on this machine-learned surface the made phase is a near-ideal solid solution. None of the arrangements reproduces the made phase's small orthorhombic distortion. Putting Nb on an iron site of YFe₃ costs 45 meV/atom
Zr beside yttrium turns YFe₃ uniaxial (no precedent): refuted. Predicted −0.3 to +1.2 MJ/m³ with an easy axis above +0.3; measured −0.94, a stronger easy plane than YFe₃'s −0.70.
The made orthorhombic phase is YFe₃ with Zr ordered on the Laves sites: partly. The site preference and composition fit, but no Y/Zr order makes the cell orthorhombic (b/√3a 0.9998–1.0015 against 0.984 measured).
Nb and Ti beside yttrium also stay easy-plane: held. Ti −0.96; Nb −0.29, the only one that moves toward zero.
Nb on the Laves site plus one Nb on an iron site gives an easy axis at ≥ 1 T: refuted at the stability gate. Every iron site is 57–68 meV/atom off the hull, so no DFT was spent.
The YCo₅-type site is not a general lever for rare-earth-free magnets. Near the hull it exists only in CaCu₅–Laves intergrowths, and iron on it gives an easy plane. Cobalt on it gives an easy axis, but in every iron-rich host we have tried, cobalt prefers another site.
The second metal in the Laves layers of YFe₃ (Zr, Ti, Nb) moves the anisotropy by −0.26 to +0.41 MJ/m³ without touching the moment, and never crosses zero.
The published iron-site route to easy-axis YFe₃ is a computed state about 45 meV/atom off the hull, not a phase a furnace will give.
Our anisotropy route now has a PuNi₃-type calibration (YFe₃): sign right, size 0.74–0.78 of full-potential PBE.
The standing candidates are unchanged; see loop 58
made |
1.173 |
computed-only (Alexandria); end member of the made (Y,Zr)Fe₃ |
1.177 |
computed-only (Alexandria) |
1.192 |
computed-only (Alexandria) |
1.169 |
Cell + ionic relaxation with ABACUS DFT; 0.04 eV/Å threshold; final energy = -34547.6877 eV; energy change = -0.0276 eV; not fully force/stress converged
A new family with no precedent: the made Fe–Zr–Y phase (Y,Zr)Fe₃ and its Nb and Ti cousins keep 1.17–1.19 T but stay easy-plane; the published Fe-site flip is 45 meV/atom off the hull.