COMPAS-3: 39,000 Polybenzenoid Hydrocarbon Isomer Geometries and DFT Calculations

Approximately 39,000 peri-condensed polybenzenoid hydrocarbon isomer geometries, with up to 68 atoms, form the COMPAS-3x data set. Ryan Stocks published this dataset on figshare in 2026, which includes results from double hybrid density functional theory (DHDFT) calculations accelerated on GPUs. The dataset also contains benchmark isomerization energies for a range of LDA, GGA, and MGGA functionals.

Use Cases

• Benchmarking density functional theory (DFT) methods based on isomerization energies for polybenzenoid hydrocarbons.
• Training machine learning models for molecular property prediction based on the calculated DHDFT energies.
• Studying the performance of GPU-accelerated quantum chemistry algorithms on large molecular datasets.
• Analyzing the accuracy of different functional classes (LDA, GGA, MGGA) with and without dispersion corrections.

Strengths

• Contains approximately 39,000 molecular geometries, providing a substantial benchmark set.
• Includes calculations using the revDSD-PBEP86-D4(noFC)/def2-QZVPP method, a specific and high-level double hybrid DFT approach.
• Benchmarks multiple functional types (LDA, GGA, MGGA) against a reference, providing comparative performance data.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Data may reflect methodological bias inherent to the specific computational setup described.

Provenance

Source

Ryan Stocks via figshare.

Collection Method

Computational generation using GPU-accelerated double hybrid density functional theory (DHDFT) calculations on the Perlmutter supercomputer.

Freshness

Last updated 2026-05-05 06:18:30; freshness should be verified.