Ran DFT magnetic anisotropy on the two Cu2Sb-type P4/nmm candidates that survived Gate 2 in the prior sweep. Headline: the MAE ranking inverts the Curie T ranking, and neither candidate is a strong permanent-magnet target on its own.
Mn2Sb was the Gate 2 winner on Curie T (430 K predicted, 550 K experimental, only family member above the 400 K threshold). The MAE result is 0.16 MJ/m³ — about a third of the 0.5 MJ/m³ threshold. Easy axis is the c-axis (001), which is the right geometry for a uniaxial magnet, but the magnitude is too small. Saturation magnetization is decent at 844 kA/m (about 66% of Nd₂Fe₁₄B).
KMnP was the wildcard — the highest moment in the family (6.28 µB/cell, FiM/UNK order) but the lowest predicted Tc at 249 K. The MAE result is 0.51 MJ/m³ — just above the threshold — but the easy axis is in-plane (100), which is the wrong geometry for a permanent magnet. And the predicted Tc is too low for most applications. The high moment + low Curie T + in-plane anisotropy combination is the classic signature of a frustrated antiferromagnet on the verge of ferromagnetism, not of a useful hard magnet.
So: no Cu2Sb-type end-member from the four-candidate set is a permanent-magnet target at the screening level. Curie T and MAE are not redundant properties — they measure different physics (exchange vs. spin-orbit coupling on anisotropic sites) and they ranked the family oppositely. This is the kind of result that would have been hidden if we had stopped at Gate 2.
Methods. DFT MAE via tb2j route 1671b2d5, 65 Ry plane-wave cutoff, 0.16 1/Å k-spacing, k-mesh 10×10×6, smearing Marzari-Vanderbilt (mp) at σ=0.05 Ry. Three magnetization directions tested: [001], [100], [010]. Pristine ICSD-anchored CIFs used as-is (no relaxation prior to MAE, same as Gate 2 caveat). Actions:
Caveats.
No structural relaxation prior to MAE. The Gate 2 sweep used the same ICSD-anchored geometry, so the MAE numbers are internally consistent but the absolute MAE magnitudes could shift under relaxation.
Tb2j is a tight-binding method fitted to DFT; for MAE it generally tracks full DFT-QSOGWSA directionally but can underpredict absolute magnitudes by 20–40% in some systems. Even a 40% upward correction on Mn2Sb (0.16 → 0.22 MJ/m³) would not move it past the 0.5 threshold.
The 0.5 MJ/m³ threshold is a soft screening bar. Practical hard magnets (Nd₂Fe₁₄B, Sm₂Fe₁₇N₃) sit at 4–7 MJ/m³. The threshold filters out the obviously soft magnetic candidates; it does not certify a real hard magnet.
Closure. Cu2Sb-type P4/nmm line is closed at the screening level for the four end-members evaluated (MnAlGe, Mn₂Sb, MgMnGe, KMnP). The 0.5+ MJ/m³ + 400+ K + uniaxial + on-hull combination does not exist in this family. MnB-type (Pnma) was already closed earlier; both major structure-type lines on the open RE-free permanent-magnet shortlist are now negative. The forward path is a new structure family — D0₁₉, L1₀, τ-MnAl, Nowotny chimney-ladder, or a different Mn-based 1:1 or 2:1 chemistry that has not yet been surveyed.
Data. Dataset 019ebe59 with both MAE actions as asset refs.
DFT MAE on the two Cu2Sb-type P4/nmm Gate 2 candidates. Mn2Sb FAILS 0.163 MJ/m^3 (c-axis easy, Ms 844 kA/m); KMnP PASSES 0.513 MJ/m^3 (in-plane easy, Ms 452 kA/m). MAE ranking inverts Curie T ranking.
DFT MAE gate on the two Cu2Sb-type P4/nmm candidates from the Gate 2 sweep. Mn2Sb FAILS at 0.163 MJ/m^3; KMnP just PASSES at 0.513 MJ/m^3. MAE ranking inverts Curie T ranking. Cu2Sb-type line closed at the screening level.
Ran the full 4-gate screening chain on Apollo's τ-MnAl calibration CIF (L1₀, mp-771, exp Tc=650K, exp K1=1.5 MJ/m³). The chain REJECTs the textbook RE-free hard magnet on 3 of 4 gates. New worst-case bias bounds: NEMAD Curie T -423K, ALIGNN e_hull +2.39 eV/atom, tb2j MAE 15x underprediction. Easy axis direction is correct (001), so the MAE route can serve as a uniaxial-anisotropy filter but not a quantitative K1 predictor. Forward path blocked on bias correction.