Polymer Dataset for Accelerated Property Prediction and Design

1,073 polymers and related materials with uniformly prepared first-principles calculations. The dataset includes optimized structures, atomization energies, band gaps, and dielectric constants to assist in designing high dielectric constant polymers. It was developed by Huan D. Tran and is progressively expanded with new materials and properties.

Use Cases

• Predict dielectric constants from optimized structures and atomization energies using machine learning models.
• Analyze relationships between band gaps and dielectric constants across 1,073 polymer entries.
• Use the dataset of optimized structures for high-throughput screening of polymers with targeted electronic properties.
• Train property prediction models on the uniformly calculated atomization energies and band gaps.

Strengths

• Contains 1,073 uniformly prepared polymer and material entries.
• Includes four key calculated properties: optimized structures, atomization energies, band gaps, and dielectric constants.
• Data is prepared using first-principles calculations, ensuring a consistent methodological foundation.

Limitations

• Dataset size of 1,073 entries may be limited for complex deep learning models requiring larger training sets.
• Initial focus is on dielectric constant polymers, which may limit generalizability to other material property predictions.
• Properties are calculated, not experimentally measured, which may introduce computational artifacts.

Provenance

Source

http://khazana.uconn.edu/

Collection Method

Uniformly prepared using first-principles calculations with structures from other sources or structure search methods.

Freshness

Last updated in June 2020.