Thermal Properties · foundry-ml
dataset_thermalexp_aflow
A benchmark tabular dataset of 4,886 AFLOW thermal expansion coefficients for inorganic materials modelling.
What dataset_thermalexp_aflow is
dataset_thermalexp_aflow is a curated tabular benchmark dataset created by Conrad L. Clement, Steven K. Kauwe, and Taylor D. Sparks in 2020. Published through Foundry-ML and the Materials Data Facility, it draws from the AFLOW repository of inorganic materials to supply thermal property measurements for computational research.
The dataset addresses the need for standardised physical targets in materials science and solid-state physics. By coupling chemical compositions with thermal expansion coefficients, it provides training and evaluation data for machine learning algorithms applied to thermal-mechanical modelling.
How it was made
The dataset consists of 4,886 tabular records sourced from the AFLOW database. Each entry pairs an input chemical formula describing an inorganic compound with a scalar target representing the thermal expansion coefficient.
Data is distributed in tabular Parquet format and organised into a single training split for tabular regression and classification tasks.
Key facts
- Number of records
- 4,886
- Input feature
- Material composition (formula)
- Target property
- Thermal expansion coefficient
- Target units
- K^-1
- Data format
- Parquet tabular data
- Primary tasks
- Tabular regression and tabular classification
- Source database
- AFLOW
- Release year
- 2020
- DOI
- 10.18126/qmrs-jg02
What you can use it for
- Composition-based thermal expansion prediction
- Training supervised regression models to estimate thermal expansion coefficients directly from chemical formulas.
- Thermal behaviour classification
- Categorising inorganic compounds into low- or high-expansion classes for screening thermal-mechanical materials.
- Benchmark evaluation for materials featurisers
- Benchmarking composition-based machine learning representations against a standard set of inorganic thermal properties.
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_thermalexp_aflow at the moment you asked. Nothing is cached or stored — every row above came from that request.
How the columns are distributed
Over 4,886 rows of default/train
formula
string_text- min
- 2
- median
- 8
- mean
- 7.26
- max
- 15
target
float- min
- 0
- median
- 5.0e-5
- mean
- 6.0e-5
- max
- 4.9e-4
Strengths and limits
Good at
- Provides a focused, clean collection of 4,886 thermal expansion values tied directly to standard chemical formulas.
- Supplies explicit physical units in inverse kelvin (K^-1) for straightforward regression benchmarking.
- Derives from the established AFLOW database for inorganic crystal calculations.
Watch out for
- Contains only composition formulas as inputs, lacking atomic coordinates, space groups, or structural geometries.
- Includes only a single predefined train split, requiring users to manually partition validation and test sets.
Questions
Can I look inside dataset_thermalexp_aflow 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_thermalexp_aflow under?
Its authors declare other. Read the terms on the original page before reusing the data.
What units are used for the target values?
The target property is the thermal expansion coefficient measured in inverse kelvin (K^-1).
What features are provided for model training?
The dataset contains a single input feature column, formula, representing the chemical composition of each inorganic material.
How many data points are available in the dataset?
There are 4,886 individual entries provided in a single training split.
Which domain tasks is this dataset designed for?
It is set up for tabular regression and tabular classification tasks in materials science and physics.
How to cite it
Clement, Conrad L. and Kauwe, Steven K. and Sparks, Taylor D. Benchmark AFLOW Data Sets for Machine Learning (Thermal expansion). Materials Data Facility (2020). https://doi.org/10.18126/qmrs-jg02
Written from dataset_thermalexp_aflow'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.