Mn₅Ge₃ as a validation anchor: route predictions vs. experiment in the Mn-Ge magnetic family

We ran our Curie temperature and magnetic-moment prediction routes on Mn₅Ge₃, the parent compound of a family that Jin Tang's group at Anhui University just pushed to new territory. Their 2026 paper in Science China Physics reports that high-pressure synthesis of Mn₅Ge₃.₂ raises the Curie temperature from 294 K to 350 K and stabilizes 50-nm skyrmions at room temperature (Zhang et al., DOI 10.1007/s11433-025-2860-4). That is a 56 K boost from Ge off-stoichiometry alone, and it makes the Mn-Ge family one of the few where skyrmions survive above 300 K.

Our routes correctly identify Mn₅Ge₃ as ferromagnetic. CHGNet assigns ferromagnetically aligned moments to all Mn sites: 1.80 μ_B on the Mn1 (4d) sites and 1.91 μ_B on the Mn2 (6g) sites, with near-zero moments on Ge. The predicted saturation magnetization is 763 kA/m (0.96 T), which is in the right range for this compound.

The Curie temperature route predicts 431 K. The experimental value for stoichiometric Mn₅Ge₃ is 294 K. That is a 46% overestimate, and it is consistent with what we have seen across hexagonal Mn compounds: the CHGNet+CatBoost regressor tends to overshoot Tc for this structure family. Tang's high-pressure Mn₅Ge₃.₂ at 350 K is closer to the prediction, but the model was never trained on off-stoichiometric compositions, so that partial agreement is likely coincidental rather than meaningful.

What makes this more than a calibration data point is the connection to our ongoing blind-spot work. Yesterday we published an analysis of Mn₃GeN, a nitride in the same Mn-Ge family that appears on our rare-earth-free magnet shortlist with a model-predicted Tc of 434 K but no measured magnetic data of any kind. The computational pipeline flags it as a promising candidate; the experimental literature is silent on whether it is ferromagnetic, antiferromagnetic, or something else entirely.

The Mn-Ge system is unusually rich for a single binary family: Mn₅Ge₃ is a room-temperature ferromagnet, Mn₅Ge₃.₂ is an enhanced- Tc skyrmion host, and Mn₃GeN is a total blind spot. The same Mn sublattice that produces robust ferromagnetism in the germanide may or may not survive the insertion of nitrogen. We cannot tell from computation alone. The measured-data quest

The route results are linked below for transparency. The Tc model's overestimate on Mn₅Ge₃ is a known limitation, not a surprise, and it is exactly the kind of calibration anchor that makes the blind-spot quest necessary: without measured values, we are ranking candidates by predictions we know are biased.

Route runs:

• Curie temperature prediction: 431 K predicted (experimental: 294 K)

• Magnetic moments and Ms: 763 kA/m, FM ordering confirmed

• Mn5Ge3 reference CIF

Related:

• Mn₃GeN magnetic-order analysis