Perovskite Properties · foundry-ml
dataset_perovskite_formatione
Formation energy data for 9,646 perovskite compounds to support machine learning screening for material stability.
What dataset_perovskite_formatione is
This tabular materials science dataset contains 9,646 perovskite formation energy records compiled to aid the computational screening of perovskite metal oxides for solar light capture. Originally generated by Ivano E. Castelli, Thomas Olsen, Soumendu Datta, David D. Landis, Søren Dahl, Kristian S. Thygesena, and Karsten W. Jacobsen, the data focuses on assessing compound stability and synthesisability.
Formation energy indicates whether a crystal structure is thermodynamically favoured to form relative to its constituent elements. By pairing chemical compositions directly with calculated formation energies, the dataset enables computational chemists and materials scientists to train predictive models for candidate material evaluation without repeating costly first-principles calculations.
How it was made
The data is structured as a two-column tabular dataset. The input column provides the material composition as a chemical formula, and the target column provides the corresponding formation energy expressed in electronvolts per atom (eV/atom).
All 9,646 records are provided within a single training split, formatted for tabular regression and classification workflows.
Key facts
- Data points
- 9,646
- Input features
- formula (material composition)
- Target variable
- target (formation energy in eV/atom)
- Data format
- Parquet
- Task categories
- Tabular regression, tabular classification
- Publication year
- 2011
- DOI
- 10.18126/xmh8-d711
What you can use it for
- Perovskite stability regression
- Train machine learning regressors to predict formation energy directly from chemical formulae.
- Synthesisability screening
- Screen candidate perovskite compositions to classify whether materials meet energetic thresholds for synthesisability.
- Solar material discovery
- Prioritise thermodynamically stable perovskite metal oxides for further evaluation in solar light capture workflows.
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/dataset_perovskite_formatione at the moment you asked. Nothing is cached or stored — every row above came from that request.
How the columns are distributed
Over 9,646 rows of default/train
formula
string_text- min
- 4
- median
- 10
- mean
- 9.41
- max
- 12
target
float- min
- -0.64
- median
- 1.06
- mean
- 1.10
- max
- 4.24
Strengths and limits
Good at
- Provides 9,646 evaluated perovskite compositions with standardised formation energies.
- Presents a clean, two-feature tabular structure directly suitable for baseline composition-based machine learning models.
- Supplies target values in uniform energy units of electronvolts per atom (eV/atom).
Watch out for
- Includes only a single train split, requiring users to manually partition validation and test sets.
- Contains only composition formulas, lacking 3D atomic coordinates, lattice parameters, or space group details.
- Reports only formation energy without additional electronic or optical property targets.
Questions
Can I look inside dataset_perovskite_formatione 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 dataset_perovskite_formatione under?
Its authors declare other. Read the terms on the original page before reusing the data.
What target property does this dataset measure?
It measures the formation energy of perovskite compounds in electronvolts per atom (eV/atom).
What input features are available for training models?
The only input feature provided is the chemical formula representing the material composition.
How many data splits are defined?
The dataset contains a single split labelled train.
What machine learning tasks is the dataset suitable for?
It is configured for tabular regression to predict continuous formation energy values and tabular classification to predict stability categories.
How to cite it
Castelli, Ivano E., Olsen, Thomas, Datta, Soumendu, Landis, David D., Dahl, Søren, Thygesena, Kristian S., Jacobsen, Karsten W. Computational screening of perovskite metal oxides for optimal solar light capture. Materials Data Facility, 2011. https://doi.org/10.18126/xmh8-d711
Written from dataset_perovskite_formatione'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.