This loop asked whether ZrFe₈Co₄, the cell that read 4 MJ/m³ in loop 36, can be changed so that its cobalt stays on the iron site that gives that reading.
New candidate: ZrFe₉Co₂Ti. ZrFe₈Co₄ is ruled out.
ZrFe₈Co₄ cannot be fixed. Its reading needs every cobalt on one of the three iron sites, cobalt moves off that site for almost nothing, and none of 19 changes we tried (15 substituting elements, hafnium for zirconium, and hydrogen, boron or carbon in the empty site) raises the cost to a quarter of what an anneal would need. Two results from the same loop promote a different cell from loop 36's table. First, the too-large calculated moment that loop 36 found turns out to be a problem of iron-only cells; in the measured yttrium analogues it is gone by three cobalt, so a cell with cobalt can be read at its own moment. Second, ZrFe₉Co₂Ti does not care where its cobalt sits: its three arrangements are equal in energy to within 1 meV/atom, and all three are uniaxial at 1.25–1.78 MJ/m³ and 1.58 T, with no unstable phonons.
Against the campaign's standing best, Mn₁₆Al₁₅Ti (1.65 MJ/m³ at 1.02 T, from an earlier private loop), ZrFe₉Co₂Ti is level on anisotropy and has 55 % more magnetization. It is behind on hardness (κ 0.8–0.9 against 1.4), on distance from the hull, and on cost, and nobody has made its parent phase without silicon. So the loop leaves the campaign with a second candidate of a different kind, a high-magnetization iron-based one, and with one fewer open direction.
DFT runs are PBE in a double-zeta basis, k-spacing 0.2 Å⁻¹, on 26-atom cells with c along z, relaxed with DFT structure relaxation and read with Magnetic anisotropy energy
ZrFe₉Co₂Ti with Ti on 8i and the two Co on each pair of iron sites they can share:
Co sites | Cell | Energy after relaxation, against 8j,8j (meV per 26 atoms) | K, own moment (MJ/m³) | K, 0.94 of the moment (MJ/m³) | Js (T) | Hull, Orb (meV/atom) | Phonons |
|---|---|---|---|---|---|---|---|
8j, 8j |
Targets: K ≥ 1 MJ/m³ with one easy axis, Js ≥ 1 T, hull ≤ 75 meV/atom, no imaginary phonons, Tc ≥ 450 K, cost at or under Nd₂Fe₁₄B's 15.69 $/kg. Mean K over the three is 1.48 at the cells' own moments and 1.64 at 0.94, and c is the easy axis against both in-plane directions in every cell. The 8j,8j row and its remaining numbers are from loop 36: hull 42.7 meV/atom against published-PBE references, Curie route 1205 K (about 630 K after the factor of 1.9 by which the route overreads YFe₁₁Ti), and 10.46 $/kg
Cell + ionic relaxation with ABACUS DFT; 0.04 eV/Å threshold; final energy = -84003.1282 eV; energy change = -0.2796 eV
ZrFe₉Co₂Ti, lowest arrangement: Zr on the large 2a site, Ti on one 8i site, both Co on 8j.
The cell the loop set out to fix. Each row is ZrFe₈Co₄ with one change, relaxed with the Orb potential through Energy above the convex hull
Change to ZrFe₈Co₄ | Site it takes | Cost to move one Co from 8f to 8j (meV) | Hull of the lowest cell (meV/atom) |
|---|---|---|---|
Nothing (bare ZrFe₈Co₄) | — |
Cell + ionic relaxation with ABACUS DFT; 0.04 eV/Å threshold; final energy = -90606.8691 eV; energy change = -0.2547 eV
ZrFe₈Co₄ with all Co on 8f (from loop 36), the ordering that reads 4 MJ/m³ and that nothing here holds in place.
Loop 36 found that our calculation gives YFe₁₁Ti a spin moment about 15 % above the measured one, and that this alone turns its anisotropy from +2 MJ/m³ to easy-plane. The question for a cobalt cell is whether the same correction applies. In Y(Fe,Co)₁₁Ti, which has been measured, it does not:
Our moment falls with cobalt while the measured one rises to a maximum near three Co, so the overshoot goes from +15 % to about −6 %. The measured values are literature, not our measurements: 19.97 μB is a single crystal at 5 K (Herper et al., Acta Mater. 242, 118473, 2023); the maximum at x = 0.3 is Wang et al., J. Phys.: Condens. Matter 13, 1617 (2001), as cited and plotted by Ke and Johnson (Phys. Rev. B 94, 024423, 2016). The 23.1 figure is from an abstract we could not open in full, so treat the −6 % as approximate and the direction as the result. No Zr compound with cobalt has a measured moment; this rests on the yttrium analogue.
A third element, or Hf, pins cobalt to 8f in ZrFe₈Co₄: refuted. Predicted that no element would reach the bar and that the best would land between 40 and 150 meV; the best is Mn at 57, eight above the bare cell. Every element that stabilizes the structure from 8i (Ti, V, Cr, Nb, Mo, Ta) makes cobalt on 8j the ground state.
An interstitial on the empty 2b site pins cobalt: refuted. That site touches only 8j atoms, so an atom there should make 8j less welcoming to cobalt. Carbon does the most (61 meV for one Co, 90 per Co if all four move) in a cell 90 meV/atom above the hull.
The moment overshoot shrinks with cobalt: supported, and by more than predicted (0–8 % at three Co; found about −6 %).
ZrFe₉Co₂Ti keeps its anisotropy on every cobalt arrangement: supported. Predicted 1.2–1.9 and 1.3–2.1 MJ/m³ for the two new cells; found 1.25 and 1.42. Predicted energies within 100 meV per cell of the 8j,8j arrangement; found 15 and 16.
Cobalt has no usable site preference in this structure. A reading that depends on which iron site holds the cobalt is a reading of one arrangement, and a cobalt cell should be judged by its spread over arrangements. At four Co in the bare cell that spread is 4 MJ/m³ and changes sign; at two Co in the Ti cell it is 0.5 and does not.
Loop 36 put ZrFe₁₀CoTi (+1.12 MJ/m³) ahead of ZrFe₉Co₂Ti because its reading did not move with the moment. With the moment error traced to iron-only cells, that caution is no longer needed for a two-cobalt cell, and ZrFe₉Co₂Ti is the stronger of the two on every arrangement.
Relaxing with DFT lowers these readings by about 0.5 MJ/m³ against the Orb-relaxed cells (2.24 to 1.78, 1.81 to 1.25, 1.98 to 1.42). Rankings made on Orb cells in this family should allow for that.
The three arrangements respond to a lower moment in different directions (one falls, one is flat, one rises), so a sample holding all three is less sensitive to the moment than any one cell suggests.
Nothing here has been made. The only rare-earth-free compound reported in this structure is Zr(Fe,Si)₁₂, and it is metastable; ZrFe₉Co₂Ti sits 27 meV/atom above the Orb hull and 43 above published-PBE references. Three ordered 26-atom cells stand in for an alloy with cobalt spread over two sites; eight other arrangements exist and were screened only with Orb. The calculations use one functional and a small basis, and the Curie temperature, PBE-reference hull and cost are measured on one arrangement. A hardness κ below 1 means a magnet made from it would need help from grain shape or microstructure to resist demagnetization. "New composition" means absent from Materials Project, Alexandria, GNoME and COD and not found in our literature searches; all structures here were built by us and carry no third-party license.
Earlier in this campaign: loop 36
1.58 |
8f, 8j | 1.58 |
8f, 8f | 1.59 |
Hf in place of Zr | large site |
Mn for one Fe | 8j |
Al for one Fe | 8j |
Zn for one Fe | 8j |
Ga for one Fe | 8j |
Ge for one Fe | 8j |
Sn for one Fe | 8i |
Cu for one Fe | 8j |
Ni for one Fe | 8j |
Ti for one Fe | 8i |
Cr for one Fe | 8i |
Mo for one Fe | 8i |
V for one Fe | 8i |
Ta for one Fe | 8i |
Nb for one Fe | 8i |
Si for one Fe | 8f, pushing a Co off it |
H in the empty 2b site | 2b interstitial |
B in the empty 2b site | 2b interstitial |
C in the empty 2b site | 2b interstitial |
Measured total moment at low temperature | 19.97 | — | about 23.1 (at 3.3 Co) |
Nothing holds the cobalt of ZrFe₈Co₄ on the site its 4 MJ/m³ needs. ZrFe₉Co₂Ti does not depend on where its cobalt sits and becomes the campaign's iron-based candidate.