Materials Science · LeMaterial
LeMat-Bulk-DFT-Hull-All
A bulk materials DFT convex hull dataset for computational materials science and stability discovery.
What LeMat-Bulk-DFT-Hull-All is
LeMat-Bulk-DFT-Hull-All is a large-scale computational materials science dataset produced by LeMaterial. It compiles density functional theory (DFT) convex hull data specifically tailored for bulk materials discovery across physics, chemistry, and engineering.
Evaluating thermodynamic stability against the convex hull—the lowest-energy envelope of competing phases at absolute zero—is a standard step when assessing whether hypothetical or candidate bulk materials can be synthesised. This collection provides comprehensive DFT stability reference data across millions of bulk crystal compositions.
How it was made
It is formatted in Apache Parquet to facilitate parallel and out-of-core data access using tools such as Polars, Dask, and ML-ready dataset loaders.
Key facts
- Category
- Materials science
- Size
- 1M to 10M records
- Data format
- Parquet
- Modalities
- Tabular, text
- Compatible tools
- Polars, Dask, Datasets, MLCroissant
- Creation date
- 5 October 2025
What you can use it for
- Bulk materials screening
- Identify synthesisable or metastable bulk crystal structures by querying precomputed DFT convex hull metrics.
- Machine learning potential training
- Use large-scale tabular bulk DFT data to train surrogate models and predict thermodynamic stability.
- Thermodynamic phase mapping
- Assess phase competition and stability across diverse chemical spaces in bulk solid-state compounds.
Explore the data
No licence declared
The authors have not published a licence, which means no permission to redistribute the data has been granted to anyone. The rows are not served here; read them on the original dataset page.How the columns are distributed
Over 5,335,286 rows of default/train
dft_hull
float- min
- -0
- median
- 0.21
- mean
- 0.43
- max
- 650.98
immutable_id
string_text- min
- 4
- median
- 12
- mean
- 11.96
- max
- 13
mace_mp_energy
float- min
- -2844.13
- median
- -32.54
- mean
- -48.72
- max
- 13072.00
0.00% empty
mace_mp_hull
float- min
- -0.69
- median
- 0.19
- mean
- 0.40
- max
- 823.88
0.00% empty
mace_omat_energy
float- min
- -2847.84
- median
- -32.02
- mean
- -48.12
- max
- 16048.71
mace_omat_hull
float- min
- -19.85
- median
- 0.27
- mean
- 0.47
- max
- 1009.92
nsites
int- min
- 1
- median
- 6
- mean
- 9.26
- max
- 444
orb_conserv_inf_energy
float- min
- -2845.75
- median
- -32.02
- mean
- -48.14
- max
- 542.78
orb_conserv_inf_hull
float- min
- -2.65
- median
- 0.27
- mean
- 0.47
- max
- 64.11
orb_direct_20_energy
float- min
- -2845.54
- median
- -32.01
- mean
- -48.13
- max
- 1998.74
orb_direct_20_hull
float- min
- -2.60
- median
- 0.27
- mean
- 0.48
- max
- 205.22
species_at_sites
list- min
- 1
- median
- 6
- mean
- 9.26
- max
- 444
true_energy
float- min
- -2846.53
- median
- -32.34
- mean
- -48.49
- max
- 5186.10
uma_energy
float- min
- -2845.65
- median
- -32.01
- mean
- -48.13
- max
- 1519.59
uma_hull
float- min
- -2.64
- median
- 0.27
- mean
- 0.48
- max
- 98.41
Strengths and limits
Good at
- Provides extensive coverage with between 1 and 10 million bulk structure entries.
- Stored in columnar Parquet format for fast querying and low-memory operations via Dask and Polars.
Watch out for
- Focuses exclusively on bulk systems, excluding low-dimensional materials, interfaces, or surfaces.
- Specific DFT exchange-correlation functionals and computational convergence parameters are not detailed in the dataset summary.
Questions
Can I look inside LeMat-Bulk-DFT-Hull-All without downloading it?
Not here: it declares no licence, so no permission to redistribute its rows has been granted. Its description and statistics are still shown.
What licence is LeMat-Bulk-DFT-Hull-All under?
Its authors have not declared a licence, which means no permission to reuse or redistribute it has been granted.
What primary property does this dataset describe?
It provides density functional theory (DFT) convex hull stability data for bulk crystal systems.
How is the data distributed?
The data is structured as tabular and text data stored in Apache Parquet format.
Written from LeMat-Bulk-DFT-Hull-All's own dataset card. Statements citing a figure that does not appear there are removed before publication. Licence and running details come from our registry.