GALJ-MOF

GALJ-MOF (Gaussian-Approximated Lennard-Jones for Metal-Organic Frameworks)

Usage

1. Run python gaussian_approximation.py.

2. Paramaters of 8 Gaussians for each element will be saved in the gaussian_params directory.

3. To generate GI descriptor grids, execute the command python get_gi.py. Include the following arguments for customization:

• --mof-dir: path to the directory containing CIF files
• --sigma: a list of sigma parameters to generate GI descriptors
• --grid-spacing: grid spacing of GI grids (default: 0.2)

The generated GI descriptor grids will be stored in the grids directory under the filename mofname.npy. To execute the get_gi.py script, use the command like:

python get_gi.py --mof-dir MOFs --sigma 0.1 0.2 0.3 --grid-spacing 0.5

4. The first 3 columns in mofname.npy represent the x, y, and z coordinates of grid points. Subsequent columns contain GI descriptors with their corresponding --sigma values.

---

If you use GALJ-MOF in a scientific publication, please cite the following paper:

S. Choi, D. S. Sholl, and A. J. Medford, Gaussian Approximation of Dispersion Potentials for Efficient Featurization and Machine-Learning Predictions of Metal-Organic Frameworks, J. Chem. Phys. 2022, 156, 214108. DOI: https://doi.org/10.1063/5.0091405

---

Dependencies

• NumPy
• SciPy
• Atomic Simulation Environment (ASE)
• PyTorch
• Skorch
• AMPTorch

Acknowledgements

• This works was supported by the Department of Energy, Office of Science, Basic Energy Sciences, under Award #DE-SC0020306.
• The codes in get_gi.py is adapted from AMPtorch/CEMT. For installation and instruction of the AMPtorch package, please refer to its official GitHub repo.