Perovskite Properties · foundry-ml
dataset_perovskite_stability_updated
A curated DFT dataset for benchmarking machine learning models predicting perovskite stability and durability.
What dataset_perovskite_stability_updated is
dataset_perovskite_stability_updated is a tabular materials science dataset developed by Ryan Jacobs, Jian Liu, Harry Abernathy, and Dane Morgan. Published via Foundry-ML and the Materials Data Facility, it compiles electronic structure features and thermodynamic stability indicators derived from density functional theory (DFT) calculations.
The dataset addresses the challenge of predicting the durability and catalytic viability of perovskite compositions. By supplying precomputed DFT properties alongside multiple energy above the convex hull metrics under varying environmental conditions, it serves as a benchmark for training regression and classification algorithms in energy and photovoltaic material design.
How it was made
The collection consists of 2,844 perovskite entries derived from DFT calculations. For each entry, structural and electronic descriptors are recorded, including composition with site notations, total cell energy, DFT-calculated band gap, oxygen p-band centre, and the nominal transition metal d-electron count.
Thermodynamic stability is quantified via three specific target metrics: energy above hull at ultra-high vacuum and 1,200 Kelvin (E_hull), energy above hull for a closed system (E_above_hull_closed), and energy above hull for an open system exposed to room air at 500 °C (E_above_hull_open).
Key facts
- Total samples
- 2,844
- Data format
- Tabular (Parquet)
- Tasks
- Tabular regression, tabular classification
- Input features
- Composition, composition (no brackets), O pband (eV), energy, Nominal d #, Band gap (eV)
- Target properties
- E_hull (meV/atom), E_above_hull_closed (meV/atom), E_above_hull_open (meV/atom)
- Release year
- 2023
- Authors
- Ryan Jacobs, Jian Liu, Harry Abernathy, Dane Morgan
- DOI
- 10.18126/xcye-zy28
What you can use it for
- Thermodynamic Stability Regression
- Training regression models to predict convex hull energy metrics (E_hull) across closed or open atmospheric conditions directly from chemical formula and DFT electronic features.
- Perovskite Durability Classification
- Classifying whether a candidate perovskite remains stable under elevated temperatures (1,200 K) or atmospheric open-system conditions (500 °C in room air).
- Screening Catalytic and Photovoltaic Materials
- Filtering potential perovskite compositions by combining stability targets with electronic descriptors such as the oxygen p-band centre and band gap.
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_stability_updated at the moment you asked. Nothing is cached or stored — every row above came from that request.
How the columns are distributed
Over 2,844 rows of default/train
Band gap (eV)
float- min
- 0
- median
- 0
- mean
- 0.39
- max
- 4.21
E_above_hull_closed (meV/atom)
float- min
- 0
- median
- 79.48
- mean
- 100.08
- max
- 946.19
E_above_hull_open (meV/atom)
float- min
- 0
- median
- 178.17
- mean
- 219.60
- max
- 1556.38
E_hull (meV/atom)
float- min
- -0
- median
- 192.05
- mean
- 232.18
- max
- 1143.19
Nominal d #
float- min
- -1
- median
- 3.25
- mean
- 3.38
- max
- 11
O pband (eV)
float- min
- -6.33
- median
- -2.27
- mean
- -2.61
- max
- -0.58
composition
string_text- min
- 13
- median
- 18
- mean
- 18.74
- max
- 39
composition (no brackets)
string_text- min
- 7
- median
- 12
- mean
- 12.74
- max
- 33
energy
float- min
- -440.54
- median
- -294.98
- mean
- -293.57
- max
- -195.61
Strengths and limits
Good at
- Provides multi-condition stability targets, distinguishing between closed systems, high-temperature ultra-high vacuum, and open air at 500 °C.
- Combines composition identifiers with primary DFT-calculated electronic properties including band gap and oxygen p-band centre.
- Formats all data into a clean, unified tabular schema ready for machine learning pipelines.
Watch out for
- Contains only 2,844 data points, which represents a relatively small sample for training high-capacity deep learning models.
- Does not include explicit 3D atomic coordinates or crystal structure files, restricting representation to compositional strings and bulk electronic descriptors.
- Lacks predefined validation or test splits, providing only a single training split.
Questions
Can I look inside dataset_perovskite_stability_updated 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_stability_updated under?
Its authors declare other. Read the terms on the original page before reusing the data.
What primary target properties are included in the dataset?
The dataset contains three stability targets measured in meV/atom: E_hull at ultra-high vacuum and 1,200 K, E_above_hull_closed for a closed system, and E_above_hull_open for an open system at 500 °C in ambient air.
Which input features are available for model training?
Inputs comprise material compositions (both with site notation and plain text), total energy (eV/cell), band gap (eV), oxygen p-band centre (eV), and nominal transition metal d-electron count.
What machine learning tasks is this dataset structured for?
It is structured for tabular regression tasks predicting continuous energy values and tabular classification tasks categorising materials by stability.
How to cite it
Jacobs, Ryan and Liu, Jian and Abernathy, Harry and Morgan, Dane. Machine Learning Design of Perovskite Catalytic Properties. Materials Data Facility, 2023. https://doi.org/10.18126/xcye-zy28
Written from dataset_perovskite_stability_updated'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.