Dataset for "Evaluating Mechanical Property Prediction across Material Classes using Molecular Dynamics Simulations with Universal Machine-Learned Interatomic Potentials"

Supporting dataset for: “Evaluating Mechanical Property Prediction across Material Classes using Molecular Dynamics Simulations with Universal Machine-Learned Interatomic Potentials”, Konstantin Stracke, Connor W. Edwards and Jack D. Evans, 2025

Contains (in a zipped directory):

0 - Barostat parameter fine-tuning: Scripts for barostat optimization including job submission (submit.py, batch_compressibility.submit), thermal expansion calculations (variable_thermal_expansion.py), volume analysis (volume.py), and plotting (plot.py).

1 - Run scripts: Simulation setup and execution scripts for both NPT ensemble (structure optimization and thermal expansion runs) and NVT ensemble (batch optimization and variable volume runs).

2 - Calculation scripts: Python scripts for computing mechanical properties including negative thermal expansion (calculate_NTE.py), bulk modulus via NPT (calculate_bulk_NPT.py) and NVT (calculate_bulk_NVT.py) methods, anisotropic properties (calculate_anisotropic.py), densities (calculate_densities.py), decomposition analysis (calculate_decomp.py), and volume/pressure fluctuation methods with correlation analysis.

3 - Analysis scripts: Scripts for generating figures including metric comparisons (Figure 1), violin plots (Figure 2), method scoring (Figure 3), isotropy vs anisotropy comparisons (Figure 4a), structural comparisons (Figure 4b), and MOF vs non-MOF comparisons (Figure SI).

4 - Data: Cached simulation data (combined_methods.npz), pressure and volume correlation results for NVT and NPT ensembles, bulk moduli results from multiple methods, coordination data (bonds.npz), stability results, negative thermal expansion reference data, computed property deltas from reference values, and job runtime information.