Ty-Perez/AMP20

---
license: mit
task_categories:
- other
tags:
- materials-science
- crystal-structures
- alexandria
- mp20
pretty_name: Alex-MP-20
size_categories:
- 500K<n<1M
---

# Alex-MP-20

A dataset of crystal structures derived from the Alexandria (Alex) and
Materials Project (MP-20) databases. It contains DFT-computed properties
for ~675 k unique structures split into train / val / test partitions.

## Schema

| Column | Type | Description |
|--------|------|-------------|
| `pos` | List[List[float]] | Fractional atomic coordinates |
| `cell` | List[List[float]] | Unit-cell lattice vectors (3×3) |
| `atomic_numbers` | List[int] | Atomic number for each site |
| `pbc` | List[int] | Periodic boundary flags (3,) |
| `ids` | string | Structure identifier |
| `space_group` | string | International space-group symbol |
| `chemical_system` | string | Element set (e.g. "Fe-O") |
| `energy_above_hull` | float64 | Thermodynamic stability (eV/atom) |
| `dft_band_gap` | float64 | DFT band gap (eV) |
| `dft_bulk_modulus` | float64 | DFT bulk modulus (GPa) |
| `dft_mag_density` | float64 | DFT magnetic moment density (muB/Ang^3) |
| `hhi_score` | float64 | Herfindahl–Hirschman elemental scarcity index |
| `ml_bulk_modulus` | float64 | ML-predicted bulk modulus (GPa) |

## Splits

| Split | Examples |
|-------|----------|
| train | 540 162 |
| val | 67 521 |
| test | 67 521 |