Metal-43 Dataset

Overview

The Metal-43 Dataset is a comprehensive dataset containing DFT-calculated of 43 metallic elements. Each configuration is stored in extxyz format and includes the atomic positions, energy, forces, and stress at different temperature. This dataset is intended for developing and benchmarking machine learning potentials, especially for metals.

The dataset also provides scripts to compute statistical properties and the sum of Kohn–Sham eigenvalues below the Fermi level ($E_F$) under different smearing conditions.

• *.extxyz files: Each file represents a set of atomic configurations for a particular element at a specific temperature.

  • File name format: {Element}_{Structure}_{Temperature}.extxyz
    Example: Ag_fcc_300.extxyz → Silver in fcc structure at 300 K.

• scripts/: Contains Python scripts to process the dataset:

  • statistics.py: Calculates statistical properties such as average energy and force distributions.
  • eigen.py: Computes the sum of Kohn–Sham eigenvalues below the Fermi level for different smearing values.

Usage:

python statistics.py vasprun.extxyz
python .\eigen.py --mode dL --path EBAND-Ag --window 0 0.05 
python .\eigen.py --mode dEf --path EBAND-Ag\0.00000 --window -0.1 0.1