Electronic Structure · foundry-ml
foundry_mp_band_gaps_v1-1
A curated dataset of Materials Project band gaps for machine-learning benchmarking in inorganic electronic structure.
What foundry_mp_band_gaps_v1-1 is
foundry_mp_band_gaps_v1-1 is a curated tabular dataset designed for machine-learning applications in electronic structure prediction. Published through the Materials Data Facility and Foundry-ML by Xiang-Guo Li, Ben Blaiszik, Marcus Schwarting, Ryan Jacobs, Aristana Scourtas, KJ Schmidt, Paul Voyles, and Dane Morgan, it aggregates inorganic materials data originating from the Materials Project.
The dataset addresses the challenge of benchmarking graph neural networks and other predictive models against electronic structure properties. It provides crystal structures, chemical compositions, and associated crystallographic identifiers alongside calculated band-gap values in electronvolts.
How it was made
The dataset compiles tabular records of inorganic compounds with structural and electronic properties. Each entry links material identity—including Inorganic Crystal Structure Database (ICSD) identifiers, reduced chemical composition, crystal space group, and full structural coordinates—with computational methodology and experimental context.
The primary prediction target is the band-gap value in electronvolts (eV), accompanied by descriptor information such as the band-gap type (direct or indirect) and the computational functional used (such as PBE).
Key facts
- Data format
- Parquet
- Data type
- Tabular
- Domain
- Materials science, physics, chemistry
- Category
- Electronic structure
- Publisher
- Materials Data Facility / foundry-ml
- Release year
- 2021
- DOI
- 10.18126/vjwr-5bs9
- Target property
- Band-gap value (eV)
What you can use it for
- Band-gap regression
- Train and benchmark regression models, including graph neural networks, to predict numerical band-gap values in electronvolts directly from crystal structures and compositions.
- Electronic property classification
- Develop classification algorithms to categorise electronic structure features, such as distinguishing direct versus indirect band gaps or metallic versus semiconducting states.
- Materials informatics benchmarking
- Evaluate new crystal representation techniques and machine-learning architectures against a standardized benchmark of Materials Project electronic structures.
Explore the data
The schema, real rows, full-text search and a SQL filter — read in your browser, nothing downloaded.
Explore
Read the real rows without downloading anything
Reading rows…
LiveRead from foundry-ml/foundry_mp_band_gaps_v1-1 at the moment you asked. Nothing is cached or stored — every row above came from that request.
How the columns are distributed
Over 74,992 rows of default/train
bandgap type
string_label- I100%
bandgap value (eV)
float- min
- 0
- median
- 0.06
- mean
- 1.16
- max
- 17.90
comments
string_label- MP metalic material100%
comp method
string_label- PBE100%
composition
string_text- min
- 1
- median
- 7
- mean
- 7.50
- max
- 22
icsd_id
list- min
- 0
- median
- 1
- mean
- 1.04
- max
- 131
reference
string_label- Materials Project100%
space group
string_label- Fm-3m8%
- P2_1/c7%
- Pnma5%
Strengths and limits
Good at
- Provides standardized structural data together with symmetry space groups and reduced chemical formulas.
- Supplies both the continuous band-gap magnitude and discrete metadata such as band-gap type.
- Cross-references source literature and ICSD compound identifiers for provenance tracking.
Watch out for
- Contains only a single defined train partition without pre-split validation or test sets.
- Entries depend on specific computational functionals, requiring care when comparing against experimental measurements.
Questions
Can I look inside foundry_mp_band_gaps_v1-1 without downloading it?
Yes. This page reads its schema, real rows and per-column statistics in the browser, with full-text search and a SQL filter. Nothing is downloaded to your machine.
What licence is foundry_mp_band_gaps_v1-1 under?
Its authors declare cc-by-4.0. Read the terms on the original page before reusing the data.
What is the primary target variable in this dataset?
The target is 'bandgap value (eV)', representing the compound's band gap measured in electronvolts.
What structural and chemical descriptors are included?
It includes the crystal structure, reduced composition, crystallographic space group, and ICSD ID.
What types of machine learning tasks are supported?
The dataset is annotated for both tabular regression (predicting continuous band gaps) and tabular classification (such as predicting band-gap types).
How to cite it
Li, Xiang-Guo and Blaiszik, Ben and Schwarting, Marcus and Jacobs, Ryan and Scourtas, Aristana and Schmidt, KJ and Voyles, Paul and Morgan, Dane. Graph Network Based Deep Learning of Band Gaps - Materials Project PBE Band Gaps. Materials Data Facility, 2021. https://doi.org/10.18126/vjwr-5bs9
Written from foundry_mp_band_gaps_v1-1'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.