This archive contains various files and scripts linked with the articles listed in the How to cite section below. It includes sample LAMMPS input files, potential files for different machine learning interatomic potentials (MLIPs) for borophene on silver, and various scripts for structure generation, post-treatment of data, and adaptive training of neural network potentials.

The MLIPs files included are for the following methods:

• DeePMD
• n2p2
• NNMP

Please cite these articles if you use any of the files in this archive.

• Contents of this archive
• Description of the scripts
  • Structure generators
  • Post-treatment scripts
  • Adaptive training scripts
• How to cite
• Author
• License
• Acknowledgments

• potential: The potential files for the n2p2, DeePMD and NNMP MLIPs
• lammps: Sample LAMMPS input files for launching an MD simulation with the n2p2, DeePMD and NNMP MLIPs
• scripts: Various scripts used for post-treating the data or generating structures

Install all necessary libraries with:

pip install -r requirements.txt

• BoroPy.py:
  • Streamlit GUI tool to build and visualize borophene structures
  • Usage (local): streamlit run BoroPy.py
  • Or use the online app
• xgenerate-structure:
  • Generate a borophene structure to stdout
  • Usage: python xgenerate-structure -h to get the help
• generatorfunctions.py:
  • Set of functions to generate borophene structures, write LAMMPS input files, etc. Called in other scripts.

• gofr.c:
  • C code to compute the radial distribution function from a LAMMPS dump file.
  • Compile with gcc gofr.c -o GofR -lm
  • Usage: gofr -h to get the help
• xconvert:
  • Conversion from and to VASP, N2P2 and LAMMPS
  • Usage: python xconvert -h to get the help
• xGDOS:
  • python code to read a LAMMPS dump file containing atomic velocities and compute the GDOS
  • Usage: python xGDOS -h to get the help
• xLAMMPStoNNP:
  • Convert and concatenate many dump files into a single file to use with N2P2. Also look for structure generating extrapolation warnings and store them apart.
  • Usage: python xLAMMPStoNNP -h to get the help
• xOUTCARtoLAMMPS:
  • Convert an OUTCAR trajectory file into a LAMMPS dump file
  • Usage: python xOUTCARtoLAMMPS -h to get the help
• xplotLAMMPSlog:
  • Plot a LAMMPS log file
  • Usage: python xplotLAMMPSlog -h to get the help
• xprepareDPdata:
  • Prepare the data for a DeepMD potential training from a n2p2 data file
  • Usage: python xprepareDPdata -h to get the help
• xreadLAMMPSlog:
  • Read a LAMMPS log file and extract the thermodynamic properties, prints to stdout
  • Usage: python xreadLAMMPSlog -h to get the help
• xSTM:
  • Compute an STM image from a CHGCAR or PARCHGCAR file
  • Usage: python xSTM -h to get the help

You will need to adapt these scripts to your own cluster and problem... Especially the xjobadaptive and adaptive_learning/SlurmJob.py scripts where some paths and cluster configuration are hardcoded.

• adaptive_learning/adaptive_training.py:
  • Script using the following classes to perform an adaptive training of a NNP and distribute jobs on the fly on a SLURM cluster
• adaptive_learning/AdaptiveTraining.py:
  • Class AdaptiveTraining to perform an adaptive training of a NNP
• adaptive_learning/Cluster.py:
  • Class Cluster to help distributing jobs on a cluster
• adaptive_learning/SlurmJob.py:
  • Class SlurmJob to help launch and follow jobs on a SLURM cluster
• adaptive_learning/functions.py:
  • Some user-defined functions
• adaptive_learning/xjobadaptive:
  • Launch an adaptive training of a NNP on a SLURM cluster. You need to edit the script to set the correct paths and cluster definition for you.

Please cite the following articles if you use any of the files in this archive (click to see the bibtex entry):

@article{mignon_neural_2023,
  title = {Neural Network Approach for a Rapid Prediction of Metal-Supported Borophene Properties},
  author = {Pierre Mignon and Abdul-Rahman Allouche and Neil Richard Innis and Colin Bousige},
  journal = {Journal of the American Chemical Society},
  year = {2023},
  doi = {10.1021/jacs.3c11549},
  volume = {145},
  number = {50},
  pages = {27857-27866}
}

@article{bousige_portable_2025,
  title = {A portable dataset for borophene growth modeling with reactive neural network potentials},
  author = {Colin Bousige and Anouar-Akacha Delenda and Abdul-Rahman Allouche and Pierre Mignon},
  journal = {(submitted)},
  year = {2025},
  doi = {},
  volume = {},
  number = {},
  pages = {}
}

Colin BOUSIGE, CNRS, Laboratoire des Multimatériaux et Interfaces, Lyon, France

This project is licensed under the MIT License - see the LICENSE file for details

This work was supported by the French National Research Agency grant ANR-21-CE09-0001-01.