Loop 51 found that iron, cobalt, nickel and manganese do not hold together on their own. This loop asked whether a binder element gives them an ordered, uniaxial, iron-based cell, starting from a published prediction for layered Fe₂NiAl.
New candidate: Fe₄Ni₂AlSi (layered Fe₂NiAl₀.₅Si₀.₅). It is a lead, not a result: computed easy-axis anisotropy of 0.98–1.01 MJ/m³ at 1.26 T, a Curie estimate near 960 K, 7.5 meV/atom above its competitors on DFT, and 4.60 $/kg in raw materials. Against the campaign's best it has less magnetization than Fe₈CoNiSi₂ (1.69 T) and about the same anisotropy, but it is far closer to stable (7.5 against 64–68 meV/atom) and its end compounds, Fe₂NiAl and Fe₂NiSi, are both real, arc-melted materials.
Two things are not settled. The anisotropy is on the 1.0 MJ/m³ target line in a basis set that reads the one published benchmark of this family about 30 % low, so the true value is more likely higher than lower, but that has not been shown. And the anisotropy exists only if iron and nickel sort into alternating planes; the compounds as made so far are cubic, and nobody has reported the layered order in a sample. The loop also mapped how the anisotropy depends on what is in the cell: cobalt turns the easy axis into an easy plane, a half-and-half mix of aluminium and silicon is the best of what was tried, and manganese does not enter the structure at all. Cost of the loop: about $4.
Screens: every symmetry-distinct arrangement of 18 compositions on one 8-atom bcc-derived cell (619 cells), relaxed with Orb v3 and scored on the Orb hull with the hull route. The lowest arrangement at each composition then went through phonons, a DFT relaxation and the anisotropy route: PBE, DZP basis, k-spacing 0.2, five directions, each on its own relaxed cell. All eight cells are phonon-stable and isotropic in the plane.
Cell | 3d planes along c | Geometry | Orb hull (meV/atom) | K₁ (MJ/m³), + is easy c | Js (T) |
|---|---|---|---|---|---|
Targets: one easy axis with K₁ at least 1.0 MJ/m³, Js at least 1 T, hull within 75 meV/atom for a measured cell (25 to pass a screen), Curie point at least 450 K, at most 20 $/kg. K₁ is at the default 0.1 eV band-sum width; at 0.05 eV Fe₄Ni₂AlSi reads 1.01. The Fe₂CuAl cell grew 11.5 % in the DFT relaxation, so its Orb hull distance is for a cell far from the one measured.
For Fe₄Ni₂AlSi the other criteria: Curie route 959 K, where the same route reads made cubic Fe₂NiAl at 847 K against 964 K measured; raw materials 4.60 $/kg
Fe₄Ni₂AlSi: planes of iron and planes of nickel alternate along c, with layers of iron and aluminium, then iron and silicon, between them.
Cell + ionic relaxation with ABACUS DFT; 0.04 eV/Å threshold; final energy = -22946.0621 eV; energy change = -0.0304 eV
The parent, layered Fe₂NiAl, and the cell where cobalt planes replace the nickel planes and the axis is lost:
Cell + ionic relaxation with ABACUS DFT; 0.04 eV/Å threshold; final energy = -22893.9728 eV; energy change = -0.0152 eV
Cell + ionic relaxation with ABACUS DFT; 0.04 eV/Å threshold; final energy = -22090.4868 eV; energy change = -0.0260 eV
Fourteen DFT single points in one setup put Fe₄Ni₂AlSi +7.5 meV/atom above Fe₅Si + Al₂SiNi₅ + Al₃Ni₅. Within its own composition the layered arrangement is 12.8 meV/atom below the next one and 27.3 below the cubic inverse-Heusler arrangement, and it is 5.8 below the line between layered Fe₂NiAl and layered Fe₂NiSi. The one published reaction available as a check, layered Fe₂NiAl against iron plus NiAl, came out 22 meV/atom more favourable here than in the paper (−10.3 against +12.1), in a balance that contains elemental iron, which this setup handles poorly. So the margin is best read as somewhere between +7 and +30.
The published layered Fe₂NiAl is the lowest arrangement and near the hull: supported. Lowest of 24 at 18.9 meV/atom on Orb, 26.8 below the inverse Heusler (the paper: about 25; our DFT: 30.6).
Cobalt or manganese in the frame gives a near-hull layered cell with more moment: half right. Cobalt cells are layered and near the hull (15–21) with 0.12–0.15 T more, but manganese is rejected: every manganese arrangement is 54–76 meV/atom up.
The layered cells are easy-axis ferromagnets: true for nickel, false for cobalt. Fe₂NiAl reads +0.74 (the published plane-wave value is 1.05); cobalt planes give −0.66.
Anisotropy rises with electron count and passes 1 MJ/m³ at 30 electrons: partly. It rises from 28 to 29.5 electrons per formula unit and then falls back with silicon alone; the top is the aluminium–silicon mix.
Cobalt in the iron plane reaches the top with more moment: refuted. Same electron count as the aluminium–silicon cell, two thirds of its anisotropy, and less magnetization, not more.
The lead is within 40 meV/atom of its competitors and layered is its lowest arrangement: supported, with the calibration caveat above.
There is now a bcc-derived, aluminium-bound frame in which a near-hull cell is uniaxial without cobalt, a rare earth or a heavy element, built from compounds that have been made.
Manganese, the starting point of loop 51, has no place in it. Cobalt costs the easy axis wherever it sits.
What faces the iron plane decides the sign of the anisotropy; the total electron count is a guide, not the rule.
Not a measured candidate: K₁ is on the target line in the DZP basis, and has not been run in a larger one.
Not calibrated against experiment: no layered compound of this type has a measured anisotropy. The only benchmark is a published calculation, which our route reads at 0.70.
Not shown to be makeable in the layered form. Fe₂NiAl and Fe₂NiSi are made as cubic compounds (Zhang et al., arXiv:1301.7489); whether their iron and nickel will layer on a low-temperature anneal is an experimental question.
The magnetization is likely a little high: the same DFT reads made cubic Fe₂NiAl at 4.99 μB per formula unit against 4.46 measured.
Only 8-atom arrangements were compared, and only the ferromagnetic state.
Earlier: loop 51
Fe / Co |
1.36 |
Fe / Co–Ni | 1.33 |
Fe / Ni | 1.21 |
Fe / Ni | 1.26 |
Fe / Ni | 1.27 |
Fe / Cu | 1.07 |
Fe–Co / Ni | 1.18 |
Fe / Fe–Ni | 1.39 |
Layered Fe₄Ni₂AlSi reads 0.98–1.01 MJ/m³ easy-axis at 1.26 T and sits 7.5 meV/atom above its competitors on DFT; cobalt removes the axis and manganese does not enter the frame.