README.md

---
dataset_info:
  features:
  - name: material_id
    dtype: string
  - name: formula
    dtype: string
  - name: space_group_number
    dtype: int64
  - name: crystal_system
    dtype: string
  - name: structure
    dtype: string
  - name: deprecated
    dtype: bool
  - name: uncorrected_energy_per_atom
    dtype: float64
  - name: energy_per_atom
    dtype: float64
  - name: formation_energy_per_atom
    dtype: float64
  - name: energy_above_hull
    dtype: float64
  - name: is_stable
    dtype: bool
  - name: equilibrium_reaction_energy_per_atom
    dtype: float64
  - name: band_gap
    dtype: float64
  - name: cbm
    dtype: float64
  - name: vbm
    dtype: float64
  - name: efermi
    dtype: float64
  - name: is_gap_direct
    dtype: bool
  - name: is_metal
    dtype: bool
  - name: is_magnetic
    dtype: bool
  - name: ordering
    dtype: string
  - name: total_magnetization
    dtype: float64
  - name: total_magnetization_normalized_vol
    dtype: float64
  - name: total_magnetization_normalized_formula_units
    dtype: float64
  - name: num_magnetic_sites
    dtype: int64
  - name: num_unique_magnetic_sites
    dtype: int64
  - name: possible_species
    list: string
  - name: theoretical
    dtype: bool
  - name: e_total
    dtype: float64
  - name: e_ionic
    dtype: float64
  - name: e_electronic
    dtype: float64
  - name: n
    dtype: float64
  - name: universal_anisotropy
    dtype: float64
  - name: homogeneous_poisson
    dtype: float64
  - name: e_ij_max
    dtype: float64
  - name: xas_is_available
    dtype: bool
  - name: bandstructure_is_available
    dtype: bool
  - name: dos_is_available
    dtype: bool
  - name: decomposes_to_is_available
    dtype: bool
  - name: types_of_magnetic_species_is_available
    dtype: bool
  splits:
  - name: train
    num_bytes: 1480702502
    num_examples: 210579
  download_size: 473873714
  dataset_size: 1480702502
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
license: cc-by-4.0
pretty_name: Materials Project
size_categories:
- 100K<n<1M
tags:
- materials-science
- crystal-structures
- pymatgen
---


# Dataset Card for xpanceo-team/materials-project

## Dataset Summary

This dataset is a snapshot of the **Materials Project (MP)** materials database, exported from the MP **Summary** endpoint using `mp-api`.

Snapshot date: **2026-01-18**.

The `structure` column stores **pymatgen `Structure` JSON** serialized as a string. 

Important: For several rich/large MP sub-documents (e.g. XAS, DOS, band structure), this dataset stores only *availability flags* (`*_is_available`) rather than the full objects.

## Source

Upstream data source: the Materials Project database and API documentation.

Materials Project is an open resource for computed materials properties and structures, built using high-throughput DFT workflows and a curated data model.

## Preprocessing

Structures and properties were downloaded using `mp-api` (version `0.45.15`) with `MPRester.materials.summary.search()`.

The export used a fixed field list:
- “regular” scalar fields (energetics, electronic structure, magnetism, elasticity-related, dielectric-related),
- `material_id`, `structure`, and symmetry (used to derive `space_group_number` and `crystal_system`),
- availability checks for selected Emmet (internal document model) sub-documents: `xas`, `bandstructure`, `dos`, `decomposes_to`, `types_of_magnetic_species`.

## Dataset Structure

### Data Instances

Each row corresponds to one Materials Project entry identified by `material_id` (e.g., `mp-149`), with an associated crystal structure and computed properties.

### Data Fields

- `material_id` (string): Materials Project material identifier (e.g., `mp-149`).
- `formula` (string): Reduced chemical formula string.
- `structure` (string): **pymatgen `Structure` JSON** (serialized) for the entry’s structure.
- `space_group_number` (int): International space group number from `mp_doc.symmetry.number`.
- `crystal_system` (string): Crystal system label from `mp_doc.symmetry.crystal_system`.
- `deprecated` (bool): Whether the entry is marked deprecated upstream.

Energetics / thermodynamics:
- `uncorrected_energy_per_atom` (float)
- `energy_per_atom` (float)
- `formation_energy_per_atom` (float)
- `energy_above_hull` (float)
- `is_stable` (bool)
- `equilibrium_reaction_energy_per_atom` (float)

Electronic structure:
- `band_gap` (float)
- `cbm` (float)
- `vbm` (float)
- `efermi` (float)
- `is_gap_direct` (bool)
- `is_metal` (bool)

Magnetism (when available):
- `is_magnetic` (bool)
- `ordering` (string)
- `total_magnetization` (float)
- `total_magnetization_normalized_vol` (float)
- `total_magnetization_normalized_formula_units` (float)
- `num_magnetic_sites` (int)
- `num_unique_magnetic_sites` (int)

Additional properties (when available):
- `universal_anisotropy` (float)
- `homogeneous_poisson` (float)
- `e_total` (float)
- `e_ionic` (float)
- `e_electronic` (float)
- `n` (float)
- `e_ij_max` (float)

Composition helpers:
- `possible_species` (list[string])
- `theoretical` (bool)

Availability flags (full objects are not included in this dataset):
- `xas_is_available` (bool)
- `bandstructure_is_available` (bool)
- `dos_is_available` (bool)
- `decomposes_to_is_available` (bool)
- `types_of_magnetic_species_is_available` (bool)

Note: Numeric fields are preserved as-is from the MP Summary documents. Refer to official Materials Project documentation for definitions and conventions.

### Data Splits

Single split:
- `train`: 210,579 examples

## Usage

Load the dataset:

```python
from datasets import load_dataset

ds = load_dataset("xpanceo-team/materials-project", split="train")
print(ds)
print(ds.column_names)
```

Convert to pandas (may be heavy depending on memory):

```python
df = ds.to_pandas()
```

Parse `structure` with pymatgen:

```python
from pymatgen.core import Structure

row = ds[0]
structure = Structure.from_str(row["structure"], fmt="json")

print(structure.composition, len(structure))
```

## Citation

If you use this dataset, please cite the upstream Materials Project publication and acknowledge this Hugging Face repackaging.

```bibtex
@article{Jain2013MaterialsProject,
  title   = {Commentary: The Materials Project: A materials genome approach to accelerating materials innovation},
  author  = {Jain, Anubhav and Ong, Shyue Ping and Hautier, Geoffroy and Chen, Wei and Richards, William Davidson and Dacek, Stephen and Cholia, Shreyas and Gunter, Dan and Skinner, David and Ceder, Gerbrand and Persson, Kristin A.},
  journal = {APL Materials},
  year    = {2013},
  volume  = {1},
  number  = {1},
  pages   = {011002},
  doi     = {10.1063/1.4812323}
}
```

## License


This dataset is distributed under **Creative Commons Attribution 4.0 (CC BY 4.0)**, consistent with Materials Project’s CC BY 4.0 terms..

If you use or redistribute this dataset (including derivatives), please **cite Materials Project** (see **Citation**) and **indicate if you made changes**, as required by CC BY 4.0.