Dielectric Properties · foundry-ml
dielectric_constant_v1-1
DFT-calculated dielectric and optical properties of 1,056 inorganic compounds for machine learning material screening.
What dielectric_constant_v1-1 is
dielectric_constant_v1-1 is a tabular dataset containing density functional theory (DFT) calculated dielectric and optical properties across 1,056 inorganic materials. Compiled by Ioannis Petousis, Fritz B. Prinz, Kristin A. Persson, and collaborators, the dataset was published through the Materials Data Facility and Foundry-ML to support high-throughput computational screening.
Identifying dielectric materials with targeted permittivity is critical for designing modern electronic devices, including high-k dielectric capacitors and semiconductor field-effect gates. By combining crystallographic geometry with tensor-resolved dielectric metrics, this dataset enables researchers to train regression and classification models without performing resource-intensive quantum chemical simulations for every candidate compound.
How it was made
The dataset pairs structural and electronic descriptions of 1,056 relaxed crystal structures with their DFT-derived optical and dielectric response properties. Input features include unit cell volumes, site counts, space group numbers, pymatgen structure representations, and structural files in CIF and POSCAR formats, alongside calculation metadata and bandgaps obtained from the Materials Project.
Target features capture both direction-dependent tensors and direction-averaged estimates. These include the full electronic and total dielectric constant tensors, polycrystal estimates for electronic and total permittivity, refractive indices, logarithmic transformations of total polycrystal values, and a binary flag identifying potential ferroelectricity.
Key facts
- Dataset size
- 1,056 materials
- Data type
- Tabular
- Storage format
- Parquet
- Supported tasks
- Tabular regression, tabular classification
- Input formats
- CIF, POSCAR, Pymatgen structure representation
- Underlying method
- Density functional theory (DFT)
- Database cross-reference
- Materials Project
- Publication year
- 2022
What you can use it for
- High-k dielectric screening
- Train machine learning regression models on structural representations to predict polycrystal dielectric constants for capacitive device screening.
- Optical property prediction
- Predict the refractive index of inorganic crystals directly from composition, space group, and volume data without running DFT simulations.
- Ferroelectric candidate identification
- Construct classification models using crystal symmetry, bandgap, and composition to flag candidate ferroelectric materials.
- Dielectric tensor forecasting
- Model anisotropic permittivity by training algorithms to output the full 3x3 electronic and total dielectric tensors from atomic configurations.
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/dielectric_constant_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 1,056 rows of default/train
band_gap
float- min
- 0.11
- median
- 1.73
- mean
- 2.12
- max
- 8.32
cif
string_text- min
- 932
- median
- 1,885
- mean
- 1875.90
- max
- 2,545
e_electronic
string_text- min
- 66
- median
- 119
- mean
- 113.06
- max
- 191
e_total
string_text- min
- 70
- median
- 137
- mean
- 132.82
- max
- 196
formula
string_text- min
- 2
- median
- 5
- mean
- 5.50
- max
- 10
log(poly_total)
float- min
- 0.32
- median
- 1.02
- mean
- 1.05
- max
- 2.44
material_id
string_text- min
- 6
- median
- 8
- mean
- 7.71
- max
- 9
meta
string_text- min
- 1,733
- median
- 2,762
- mean
- 2754.36
- max
- 3,805
n
float- min
- 1.28
- median
- 2.19
- mean
- 2.43
- max
- 16.03
nsites
int- min
- 2
- median
- 8
- mean
- 7.53
- max
- 20
poly_electronic
float- min
- 1.63
- median
- 4.79
- mean
- 7.25
- max
- 256.84
poly_total
float- min
- 2.08
- median
- 10.54
- mean
- 14.78
- max
- 277.78
poscar
string_text- min
- 164
- median
- 344.50
- mean
- 337.74
- max
- 714
pot_ferroelectric
bool- True67%
- False33%
space_group
int- min
- 1
- median
- 163
- mean
- 142.97
- max
- 229
structure
string_text- min
- 230
- median
- 493
- mean
- 490.62
- max
- 1,036
volume
float- min
- 13.98
- median
- 145.94
- mean
- 166.42
- max
- 597.34
Strengths and limits
Good at
- Provides full dielectric constant tensors alongside polycrystal scalar averages for detailed electro-optic analysis.
- Supplies complete structural representations across several standard formats, including CIF, POSCAR, and pymatgen objects.
- Supports both continuous property regression and categorical classification of potential ferroelectricity.
Watch out for
- The collection contains 1,056 materials, which is a comparatively small sample size for broad inorganic chemical space.
- The data includes only a single default training partition, lacking pre-established validation or test splits.
- All dielectric and refractive properties are computed via theoretical DFT simulations rather than laboratory experiments.
Questions
Can I look inside dielectric_constant_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 dielectric_constant_v1-1 under?
Its authors declare cc-by-4.0. Read the terms on the original page before reusing the data.
What target properties does the dataset contain?
The dataset provides the electronic dielectric constant tensor, total dielectric constant tensor, refractive index, polycrystal electronic and total dielectric estimates, log-transformed polycrystal totals, and a potential ferroelectric indicator.
In what formats are the crystal structures provided?
Structures are provided as pymatgen structure representations, standard Crystallographic Information Files (CIF), and VASP POSCAR files, alongside scalar parameters like volume and space group.
What machine learning task categories are supported?
It supports tabular regression for continuous permittivity and optical properties, as well as tabular classification for potential ferroelectricity.
Where do the material identifiers and bandgaps originate?
Material identifiers and bandgap values in electronvolts are drawn directly from the Materials Project database.
How to cite it
Petousis, Ioannis, Mrdjenovich, David, Ballouz, Eric, Liu, Miao, Winston, Donald, Chen, Wei, Graf, Tanja, Schladt, Thomas D., Persson, Kristin A., Prinz, Fritz B. High-throughput screening of inorganic compounds for the discovery of novel dielectric and optical materials. Materials Data Facility, 2022. https://doi.org/10.18126/racd-go9m
Written from dielectric_constant_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.