Supporting Dataset for Benchmarking Universal Machine-Learned Interatomic Potentials for H

Connor Edwards provides computational data supporting a research article on machine-learned interatomic potentials. The dataset includes ab initio molecular dynamics trajectories for nine metal-organic frameworks, CP2K input files, and analysis scripts. It was last updated on 2026-04-28.

Use Cases

• Benchmarking model performance based on Mean Absolute Error (MAE) calculations described in the evaluation utilities.
• Analyzing structural evolution of MOFs based on trajectory files containing energies, forces, and stress tensors.
• Training set generation for interatomic potentials using the provided sampling utilities and scripts.
• Computing structural properties like carbon cluster size, molecule counts, and radial distribution functions using the analysis functions.

Strengths

• Includes trajectories for nine distinct metal-organic frameworks: CALF-20, MIP-206, MOF-5, MOF-10, NU-1000, UiO-66, UiO-66-NH2, UiO-67, and ZIF-8.
• Provides a comprehensive suite of Python analysis scripts for evaluating model performance and computing structural properties.
• Contains CP2K input files for AIMD simulations at 2000 K, matching the trajectory data.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

Connor Edwards via figshare

Collection Method

Generated from ab initio molecular dynamics (AIMD) simulations using CP2K.

Freshness

Last updated 2026-04-28 05:38:26; freshness should be verified.