This loop asked whether an iron-based magnet of the kind found in loop 37 can be moved as close to stability as ZrFe₁₀Si₂, the only rare-earth-free compound of this structure that anyone has made.
No new candidate.
Of 47 combinations of stabilising elements, only the one already made, two silicon per formula unit, lies near the stable phases; mixing silicon with titanium, the element loop 37's candidate uses, is the worst combination of all. Cobalt and nickel do dissolve in the made compound at little cost, but they take away the property we need: with one cobalt the anisotropy is 0.94 MJ/m³, just under the target of 1, and with nickel or two cobalt it is 0.27.
So loop 37's ZrFe₉Co₂Ti stays the iron-based candidate, with its distance from stability as it was found: this loop shows that distance cannot be reduced by borrowing silicon from the made compound. What we know now that we did not: the two hosts respond to cobalt in opposite ways (it adds anisotropy next to titanium and removes it next to silicon), and which elements can share this structure is decided by the compounds they would form with each other outside it. ZrFe₉CoSi₂ is the nearest miss; its phonons, Curie temperature and sensitivity to the moment were not measured, since it is under the target as it stands.
Hull distances are from Energy above the convex hull
Cobalt and nickel in the made host, each in its lowest arrangement (both Si on 8f, Co or Ni on 8j):
Cell | Geometry | Novelty | K (MJ/m³) | c against the nearer in-plane direction (MJ/m³) | Js (T) | Hull, Orb (meV/atom) | Hull, PBE references (meV/atom) | Cost ($/kg) |
|---|---|---|---|---|---|---|---|---|
Targets: K ≥ 1 MJ/m³ with one easy axis, Js ≥ 1 T, hull ≤ 25 meV/atom preferred (75 at most), no imaginary phonons, Tc ≥ 450 K, cost at or under Nd₂Fe₁₄B's 15.69 $/kg. None of the three meets the K target. "New composition" means absent from Materials Project, Alexandria, GNoME and COD, and five literature searches found nothing on cobalt or nickel in Zr-only Fe₁₀Si₂; structurally each is one composition of a solid solution on the made compound, not a new structure type. The cells are our own builds. The cobalt and silicon in these ordered cells make the two in-plane directions unequal, which a real, disordered sample would average. For comparison, the host without cobalt, with silicon on the same site, read +2.13 MJ/m³ at 1.49 T in an earlier, private loop (a larger basis and a different silicon arrangement), and loop 37's ZrFe₉Co₂Ti reads 1.25–1.78 at 1.58 T with hull 27 and 42.7.
Cell + ionic relaxation with ABACUS DFT; 0.04 eV/Å threshold; final energy = -72911.3434 eV; energy change = -0.1982 eV
ZrFe₉CoSi₂ after relaxation, the nearest miss: Zr on the large 2a site, both Si on 8f, Co on 8j.
Which stabilisers the structure accepts. Each row is the lowest of all symmetry-distinct arrangements of that combination in a 13-atom cell; ZrFe₁₂ with no stabiliser is at 30.1 (loop 36). The full screen is 266 cells over 47 combinations of Si, Ti, V, Mo, Cr, Al, Ga and Ge; these are the lowest and the ones that matter for the argument.
Host | Sites in the lowest cell | Hull, Orb (meV/atom) | Hull, PBE references (meV/atom) |
|---|---|---|---|
ZrFe₁₀Si₂ | Si on 8f, Si on 8f |
With titanium and silicon together the cell decomposes into Fe₂TiSi, ZrFe₃ and iron; with silicon alone the competitors are Fe₅Si and the cubic Zr₆Fe₁₆Si₇. Gallium and germanium hosts would also miss the cost target.
What holds silicon on the 8f site, where the cobalt-free host reads its highest anisotropy. In 26-atom cells, moving one of the four Si from 8f to an 8j site costs about half an electron-volt whether or not cobalt or nickel is present: +539 meV in ZrFe₁₀Si₂, +512 with one Co, +490
A host mixing silicon with Ti, V, Mo or Cr comes within 10 meV/atom of the made compound: refuted. Predicted ZrFe₁₀SiTi at 8–20 meV/atom; it is at 52. The nearest host with any of the four is ZrFe₉MoSi₂ at 18.1, against 4.2.
An atom that prefers 8j (Al or Ga) pins both Si on 8f, a guess with no precedent: refuted. The gap had to reach 200 meV per formula unit. With Al it is 55, the same as without it, and Ga and Ge reverse it.
Cobalt enters the made host at under 25 meV/atom with silicon staying on 8f: supported. 12.9 with one Co and 19.7 with two, Co on 8j, both Si on 8f; predicted 9–16 and 14–28.
One nickel does what two cobalt do, at a lower hull cost, a guess with no precedent: supported, to no use. The nickel cell is lower on the hull than either cobalt cell and reads the same anisotropy as the two-cobalt cell, 0.27 MJ/m³.
The cobalt cells keep K ≥ 1 MJ/m³: refuted. Predicted 1.2–2.0 for one Co; measured 0.94. All three anisotropy predictions were too high.
A follow-up guess, that manganese (one electron fewer than iron) would push the silicon host the other way, failed its first check: ZrFe₉MnSi₂ is 26.9 meV/atom
ZrFe₁₀Si₂ is alone near the hull among Zr hosts of this structure. Hull distances of two stabilisers cannot be estimated by adding the effects of each.
The silicon host and the titanium host sit on opposite sides of the anisotropy maximum of this structure. Cobalt, nickel and, by the same argument, a true moment below the calculated one all lower the silicon host; that last step is an inference from loop 36's curves, not a measurement here.
The cell that is close to stable has no anisotropy to spare, and the cell with the anisotropy is 14 meV/atom further from stability than ZrFe₉CoSi₂ and about 22 further than the made compound, on either surface. ZrFe₉Co₂Ti remains the iron-based candidate and Mn₁₆Al₁₅Ti the campaign's standing result.
Earlier in this campaign: loop 36
new composition |
+0.61 |
1.47 |
new composition | +0.02 | 1.42 |
DFT relaxation, stopped at 0.041 eV/Å | new composition | −0.18: the easy axis is in the plane | 1.44 |
ZrFe₁₀GeSi | Ge on 8j, Si on 8j |
ZrFe₁₀Ga₂ | Ga on 8j, Ga on 8j |
ZrFe₉MoSi₂ | Mo on 8i, Si on 8f, Si on 8f |
ZrFe₁₀GaMo | Ga on 8j, Mo on 8i |
ZrFe₁₀GaTi | Ga on 8j, Ti on 8i |
ZrFe₁₀MoSi | Mo on 8i, Si on 8j |
ZrFe₁₀Mo₂ | Mo on 8i, Mo on 8i | not run |
ZrFe₁₀Ti₂ | Ti on 8i, Ti on 8i | not run |
ZrFe₁₀Al₂ | Al on 8j, Al on 8j | not run |
ZrFe₁₀AlSi | Al on 8j, Si on 8j | not run |
ZrFe₉Si₃ | Si on 8j, Si on 8j, Si on 8j | not run |
ZrFe₁₀TiSi | Si on 8f, Ti on 8i | not run |
ZrFe₉TiSi₂ | Si on 8f, Si on 8f, Ti on 8i | not run |
ZrFe₉Ti₂Si | Si on 8f, Ti on 8i, Ti on 8i | not run |
47 stabiliser combinations on the Zr 1:12 frame: only the made ZrFe₁₀Si₂ is near the hull. Co and Ni dissolve in it cheaply but lower the anisotropy (ZrFe₉CoSi₂ +0.94 MJ/m³ at 1.47 T).