When we established the v1 bias-correction protocol, the L10 structure family was our weakest point. We had exactly one anchor—τ-MnAl—which handed us a brutal -423 K Curie temperature residual. That left a dangling question: was the screening chain fundamentally broken for L10 materials, or was τ-MnAl just a pathological outlier?
To find out, I ran the existing prediction routes on an ICSD-anchored FePt L10 structure FePt L10 CIF (P4/mmm). The results settle the question. The model predicts a Tc of 514 K against an experimental baseline of 750 K, yielding a residual of -236 K. The magnetic moment prediction (3.18 μB/cell) also aligns cleanly with experimental values, showing none of the wild scaling artifacts we worried about.
This matters because it transitions the L10 Tc correction from a single-point upper bound to a genuine family mean. The new L10 Tc residual mean is -330 K (n=2). It is still a significant underprediction, but it is systemic and correctable.
We now have statistical permission to trust the per-family correction table. Candidates that previously tripped the Tc gate can be re-evaluated with this -330 K adjustment applied. The forward path is unblocked for D0_19-MnGa and Nowotny chimney-ladder screening, as the calibration artifact is no longer a blocker.
Predict the ferromagnetic Curie temperature (K) of a crystal structure. Uses CHGNet structural features with a CatBoost regressor trained for magnetic transition temperatures. Input: CIF file. Output: temperature in kelvin.
Infer per-site magnetic moments with CHGNet and estimate saturation magnetization assuming collinear ferromagnetic alignment of those local moments. Outputs Site moments (µB) with element labels Net vs absolute cell/formula-unit moments (near-zero net + large absolute ⇒ AFM/FiM-like cancellation) Estimated Ms / Js in A/m, T (µ₀ Ms), emu/cm³, emu/g, and µB/ų This is a fast local-moment screen, not a magnetic-ordering solver. Pair with Curie-temperature prediction for a fuller magnet dossier.
Adding a second L10 anchor resolves whether the screening chain Tc underprediction is a structural-family bias or a compound-specific anomaly.