TCTP-TCSP p1/3 k4 abc1d5 scaffold — Mo6B4Mn4 converged boride variant - relaxed

just kidding, not too crazy

good lordy @mmoderwell this is a banger

also @hermes add this to your post about the other Mn-Mo magnet, put it through the same gauntlet

Done. Put Mo₆B₄Mn₄ through the same gauntlet and added the full analysis to the post.

Short version: it's a tradeoff. Mo₆B₄Mn₄ wins on magnetization — 0.306 T vs 0.232 T (Js), 28.5 vs 21.1 emu/g (σs) — and has clean phonons (zero imaginary modes, no artifact this time). But Tc drops to 83 K vs 181 K for Mn₃(BMo₂)₂.

The killer is the 2.43 Å Mn-Mn NN distance. That's 0.5 Å shorter than the 2.93 Å pair in Mn₃(BMo₂)₂, and the AFM exchange at that distance is −30.2 meV — nearly 6× stronger. Two of the four Mn sites have J₀ = −55.7 meV, so deeply AFM that the mean J₀ across sites is actually net negative. The system may not even have a simple FM ground state.

Same two enhancement strategies apply, and the upside is bigger here because the AFM penalty is so extreme:

• Fe substitution at the 2.43 Å pair could flip −30.2 meV to FM (+60 meV to the pair sum)

• Lattice expansion to push the pair past ~2.7 Å would weaken the direct exchange substantially

If you build either CIF I'll run CHGNet + TB2J on it.

lets give the Fe substitution a try @hermes

Done. Built two Fe-substituted variants of Mo₆B₄Mn₄, relaxed both, and ran the full gauntlet (phonons, CHGNet moments, TB2J exchange). Also the three Cmmm TB2J runs from the original post have landed. Full results are in the updated post.

Short version: Fe doesn't help. It makes the short pair worse.

• Mo₆B₄Mn₂Fe₂ (Fe at the 2.43 Å pair): Fe–Fe exchange at 2.42 Å is −69.6 meV — more than double the original Mn–Mn (−30.2). Tc drops to 0 K (net AFM).

• Mo₆B₄Fe₄ (full Fe): Fe–Fe at 2.42 Å is −58.6 meV. Tc drops to 58 K (from 83).

Both relaxed clean (P2/m preserved, zero imaginary phonons). CHGNet moments also drop: 0.306 → 0.230 → 0.174 T as you go from Mn₄ to Mn₂Fe₂ to Fe₄.

The problem is the distance. At 2.4 Å, the d-electron overlap is so large that it doesn't matter whether it's Mn–Mn or Fe–Fe — both are strongly AFM. The "Fe–Mn is often FM" rule works at 3+ Å (Heuslers, etc.), not at direct-contact distances.

The strategy that actually works is lattice expansion. On the Cmmm structure (Mn₃(BMo₂)₂), 5% ab-expansion pushes the NN from 2.93 to 3.07 Å and the exchange flips from AFM (−5.45) to FM (+3.0). Tc jumps from 181 to 361 K. That's the Bethe–Slater crossover happening in real time.

For Mo₆B₄Mn₄, the 2.43 Å pair would need ~25% expansion to reach the crossover point — too much for strain. This structure's short pair is a dead end for Tc enhancement. The Cmmm structure with expansion is the path forward.