---
pretty_name: MP-20
tags:
- chemistry
- crystal
- material
size_categories:
- 10K<n<100K
---

# MP-20

MP-20 contains 45,229 structures from the Materials Project with 1 to 20 atoms per unit cell, as released with CDVAE. The CDVAE paper reports 45,231 structures; the released files hold 45,229.

This repository holds the complete, unfiltered dataset (45,229 structures) preprocessed to
HDF5 with pycrystalgen, one file per split of the original release.

## Files

| File | Structures | Atoms per structure |
|---|---|---|
| `train.hdf5` | 27,136 | 1–20 |
| `val.hdf5` | 9,047 | 1–20 |
| `test.hdf5` | 9,046 | 1–20 |

## Labels

| Field | Source column | Meaning |
|---|---|---|
| `formation_energy_per_atom` | `formation_energy_per_atom` | formation energy in eV/atom |
| `energy_above_hull` | `e_above_hull` | energy above the convex hull in eV/atom |
| `band_gap` | `band_gap` | band gap in eV |

Fields with no source in this dataset are filled with NaN: `total_energy`.

Source columns that are not stored: `spacegroup.number`, `pretty_formula`, `elements`, `material_id`. Row *i* of a split file is row *i* of the
corresponding source CSV.

## Format

Each file stores flat concatenated arrays:

| Dataset | Shape | Type | Content |
|---|---|---|---|
| `lattice` | (n_struct, 3, 3) | float32 | lattice vectors as rows, in angstrom |
| `num_atoms` | (n_struct,) | int64 | number of atoms of each structure |
| `ptr` | (n_struct,) | int64 | index of the first atom of each structure in `x` and `z` |
| `x` | (n_atoms, 3) | float32 | fractional coordinates in [0, 1) |
| `z` | (n_atoms,) | int64 | atomic numbers |
| `total_energy`, `formation_energy_per_atom`, `energy_above_hull`, `band_gap` | (n_struct,) | float32 | labels, NaN when unknown |

The lattice is rebuilt from the cell lengths and angles of the source CIF, so it is in the
standard orientation (first vector along x). The `batch` array is an artefact of the writer
and is not used when reading.

## Usage

```python
from pycrystalgen.data import HDF5Dataset

dataset = HDF5Dataset.from_hub("materials-toolkits/mp-20", "train.hdf5")
structures = dataset[:32]  # batched Structures
```

## Source and processing

- Source files: `https://github.com/txie-93/cdvae/tree/main/data/mp_20` (`train.csv`, `val.csv`, `test.csv`).
- Built with pycrystalgen commit `8e39dbb`, `scripts/build_cdvae.py`: CSV to CIF, then CIF to
  HDF5, with no filter.
- Checked against the source: same number of rows, every stored label equal to its source
  column row by row, and a sample of 300 structures per split compared with pymatgen's
  parsing of the same CIF (elements, fractional coordinates, lattice).

The data is redistributed from the sources below; refer to them for licence and terms of use.

## References

- T. Xie et al., *Crystal Diffusion Variational Autoencoder for Periodic Material Generation*, ICLR 2022.
- A. Jain et al., *Commentary: The Materials Project: A materials genome approach to accelerating materials innovation*, APL Materials 1, 011002 (2013).