We've taken a couple stabs at it, but none of them are perfect just yet.
Make a quest for this. Here are the requirements:
Wide dataset, references to the CIF files
Choose the relaxed version of a CIF if it exists. Use separate file ref columns to store the unrelaxed vs relaxed, if relevant.
Only take properties from relaxed, if relevant
Report predicted magnetic properties from Estimate magnetic moments and Ms from a CIF, MAE from Calculate magnetic anisotropy energy (MAE)
Report stability from Calculate energy above the convex hull, dynamical stability from Calculate phonon dispersion and band structure
Do NOT use any ALIGNN predictions
The properties will be sparse! It's okay. Don't do any new calculations yet. But do collect everything we have.
Now go!
Done. The dataset is live: RE-free permanent magnet candidates: properties from route calculations
Quest: Comprehensive RE-free permanent magnet candidate dataset (v2)
4,324 rows, one per unique RE-free magnetic compound from all CIF files on the platform. 29 columns including:
3 file reference columns (native FK to assets): unrelaxed_cif_id, relaxed_cif_id, primary_cif_id — relaxed preferred when both exist (3,586 compounds have a relaxed version)
5 action reference columns (native FK to actions): moments_action_id, mae_action_id, hull_action_id, phonon_action_id, curie_action_id — each links to the specific route execution that produced the value
Property columns: magnetic_moment_uB, Ms_A_per_m, MAE_eV, MAE_easy_axis, energy_above_hull_eV_per_atom
Properties pulled exclusively from existing route executions of:
Calculate magnetic anisotropy energy (DFT MAE — provides MAE, easy axis, total magnetic moment, and saturation magnetization)
ALIGNN route (de2d96c5) was fully excluded — 1,620 ALIGNN rows filtered out from 88,050 total.
Property | Filled | % |
|---|---|---|
Magnetic moment (µB) | 2,070 | 47.9% |
Saturation magnetization (A/m) | 2,069 | 47.8% |
MAE (eV) + easy axis |
Phonon calculations exist on the platform (89 results across 14 files) but were run on CIFs outside this compilation (thermoelectrics team screening files), so none matched. The phonon_action_id column is ready to populate if we run phonon routes on PM candidates.
The thermodynamically_stable column shows 1/0 (boolean from the hull route's predicted_stable field) — I can convert to true/false text if preferred.
The Estimate magnetic moments and Ms from a CIF route (CHGNet-based, d1fdf6d1) has ~10 actions on the platform but was not captured in the calculated properties
thermodynamically_stablephonon_stablecurie_temp_Kband_gap_eVscf_total_energy_eVRelax a crystal structure (used for relaxed/unrelaxed matching)
New DFT routes: Magnetic moments, Ground-state SCF, Band gap, Band structure
2,069 |
47.8% |
Energy above hull (eV/atom) | 2,547 | 58.9% |
Curie temperature (K) | 2,556 | 59.1% |
Phonon stability | 0 | 0% |
Band gap (eV) | 5 | 0.1% |
No new calculations were run. Everything here is harvested from existing route executions.