A collinear screen with Prophet-Spin ranks Fe, Co, Ni, NiO, MnO, and Cr correctly. Moments are within about 12% of experiment, except chromium.
Prophet is now on Ouro. I ran the collinear magnetism screen on six materials whose magnetic order is known from experiment.
The screen predicts local moments for a short menu of collinear seeds: all moments up, a two-sublattice antiferromagnet, and a flip of the least-abundant magnetic element. It ranks those states by energy. Saturation magnetization is the aligned ferromagnetic moment divided by the experimental cell volume. This is a few collinear patterns, not a search over magnetic space groups.
The ground state matched experiment on all six. Iron, cobalt, and nickel are ferromagnetic. Chromium, NiO, and MnO come out compensated, 36 to 471 meV/atom below the ferromagnetic state, with net magnetization of zero.
Material | Experiment | Prophet | Local moment | μ0Ms | ΔE vs ferromagnetic |
|---|---|---|---|---|---|
bcc Fe | Ferromagnetic | Ferromagnetic | 2.76 vs 2.22 μB | 2.73 vs 2.15 T | AFM +183 meV/atom |
hcp Co | Ferromagnetic | Ferromagnetic | 1.63 vs 1.72 μB | 1.72 vs 1.82 T | AFM +281 meV/atom |
fcc Ni | Ferromagnetic | Ferromagnetic | 0.73 vs 0.61 μB | 0.78 vs 0.61 T | AFM +78 meV/atom |
NiO | AFM-II | Cations antiparallel, net 0 | 1.71 vs 1.90 μB | 0.01 vs 0 T | Compensated −36 meV/atom |
MnO | AFM-II | Cations antiparallel, net 0 | 4.67 vs 4.58 μB | 0 vs 0 T | Compensated −43 meV/atom |
bcc Cr | Incommensurate AFM | Commensurate AFM | 1.70 vs 0.62 μB | 0 vs 0 T | AFM −471 meV/atom |
On Fe, Co, Ni, NiO, and MnO the mean absolute moment error is about 12%. Iron and nickel run high, by about 24% and 20%. Cobalt is slightly low. MnO is within 2%. Ferromagnetic μ0Ms follows the same error, because it is that moment spread over the experimental cell volume.
NiO and MnO are labeled ferrimagnetic. The nickel and manganese moments are antiparallel and the net moment is about zero, which is the experimental antiferromagnet. The ferrimagnetic name is there because the moment head also seeds oxygen, so the oxygen moments stay small and positive while the cations oppose.
Chromium is an incommensurate spin-density wave. The route only scored the two-atom commensurate cell. The order is antiferromagnetic, and the moment is 1.70 μB against the experimental amplitude of 0.62 μB.
The model is Prophet, from Kairos Materials. Code is MIT. Weights are CC-BY-4.0. Paper: Prophet.
Predict energy, forces, and stress is the Matbench Discovery checkpoint.
Predict local moments for a few collinear seeds and rank them by energy: ferromagnetic, a two-sublattice antiferromagnet when the cell has one magnetic element, and ferrimagnetic patterns that oppose one element when several are present. Saturation magnetization is the aligned ferromagnetic moment divided by the cell volume. Ground-state magnetization uses the net moment of the lowest ordering, so an antiferromagnetic winner is near zero. This is a short collinear menu, not a magnetic space-group search.
Low-temperature metal moments and μ0Ms are the standard handbook values (Fe 2.22 μB and 2.15 T, Co 1.72 μB and 1.82 T, Ni 0.61 μB and 0.61 T). Oxide moments are neutron values, about 1.90 μB for NiO and 4.58 μB for MnO. Chromium’s 0.62 μB is the spin-density-wave amplitude.
Screen collinear magnetic order and saturation magnetization is the screen above.
On rare-earth magnets, Prophet models the iron and cobalt, not the rare earth
Relax and Curie routes on nine rare-earth magnets. Nd, Sm, and Y carry no moment. Gd does, but couples the wrong way. Cobalt Curie temperatures run about 25% low, and Sm2Fe17 is twice too high.
Prophet relaxes magnets to about 1%, and its ferromagnetic Curie temperatures are off by 20% on average
Relax and Curie routes on ten materials with known lattices and ordering temperatures. Lattices land within about 1%, except MnBi's c axis. MnBi's order and the oxide Néel temperatures miss.