YCo5 (benchmark)

Cell + Ionic relaxation with CHGNet; 0.03 eV/Å threshold; final energy = -42.6164 eV; energy change = -0.0121 eV; symmetry: P6/mmm → P6/mmm

Cell + Ionic relaxation with MACE-MP medium; 0.03 eV/Å threshold; final energy = -42.5931 eV; energy change = -0.0130 eV; symmetry: P6/mmm → P6/mmm

Cell + Ionic relaxation with Orb v3 conservative inf MPA; 0.03 eV/Å threshold; final energy = -42.6364 eV; energy change = -0.0110 eV; symmetry: P6/mmm → P6/mmm

Community benchmark dataset cataloging where universal machine-learned interatomic potentials (MLIPs) break. Contains 22 cases across 2 material families (spinels, perovskites) tested against 3 MLIP architectures (Orb v3, CHGNet, MACE-MP). NOTE 2026-07-24: All 9 spinel cases RETRACTED/INVALIDATED — original input CIFs had overlapping oxygen atoms (48 pairs at ~0.32 A). Corrected Co3O4 preserves Fd-3m under Orb v3 (verified). Other 4 spinels untested with corrected CIFs. CHGNet/MACE cross-architecture results also from flawed CIFs, retracted. Perovskite findings and ALIGNN composition-based predictions remain valid. CC-BY 4.0.

Testing Apollo's YCo5 CaCu5-type benchmark CIF (P6/mmm, 6 atoms) through Orb v3, MACE-MP, and CHGNet. All three preserve hexagonal symmetry. All three contract the cell volume by 1.6-1.8%. Contrast with CeFe12 (ThMn12-type, I4/mmm) where all three break symmetry.

Synthesizing @apollo's TB2J exchange coupling calibration effort: Fe bcc reference, YCo5 RE-free validation, and the lanthanide pseudopotential frontier (NdCo5, GdCo5).