pypolymlp 0.15.5

This is the pypolymlp module.

A generator of polynomial machine learning potentials

pypolymlp is a Python code designed for the development of polynomial machine learning potentials (MLPs) based on datasets generated from density functional theory (DFT) calculations. The code provides functionalities for fitting polynomial models to energy, force, and stress data, enabling the construction of accurate and computationally efficient interatomic potentials. In addition to potential development, pypolymlp allows users to compute various physical properties and perform atomistic simulations using the trained MLPs.

Polynomial machine learning potentials

A polynomial MLP represents the potential energy as a polynomial function of linearly independent polynomial invariants of the O(3) group. Developed polynomial MLPs are available in Polynomial Machine Learning Potential Repository.

Citation of pypolymlp

“Tutorial: Systematic development of polynomial machine learning potentials for elemental and alloy systems”, A. Seko, J. Appl. Phys. 133, 011101 (2023)

@article{pypolymlp,
    author = {Seko, Atsuto},
    title = "{"Tutorial: Systematic development of polynomial machine learning potentials for elemental and alloy systems"}",
    journal = {J. Appl. Phys.},
    volume = {133},
    number = {1},
    pages = {011101},
    year = {2023},
    month = {01},
}

Required libraries and python modules

• python >= 3.9
• numpy != 2.0.*
• scipy
• pyyaml
• setuptools
• eigen3
• pybind11
• openmp (recommended)

[Optional]

• phonopy
• phono3py
• symfc
• sparse_dot_mkl
• spglib
• pymatgen
• ase
• joblib

How to install pypolymlp

• Install from conda-forge

Version	Last Update	Downloads	Platform	License

conda create -n pypolymlp-env
conda activate pypolymlp-env
conda install -c conda-forge pypolymlp

• Install from PyPI

conda create -n pypolymlp-env
conda activate pypolymlp-env
conda install -c conda-forge numpy scipy pybind11 eigen cmake cxx-compiler
pip install pypolymlp

Building C++ codes in pypolymlp may require a significant amount of time.

• Install from GitHub

git clone https://github.com/sekocha/pypolymlp.git
cd pypolymlp
conda create -n pypolymlp-env
conda activate pypolymlp-env
conda install -c conda-forge numpy scipy pybind11 eigen cmake cxx-compiler
pip install . -vvv

Building C++ codes in pypolymlp may require a significant amount of time.

How to use pypolymlp

Polynomial MLP development

• Command line

• Python API

• Utilities

  • DFT structure generator
    • Random atomic displacements with constant magnitude
    • Random atomic displacements with sequential magnitudes and volume changes
    • Random atomic displacements, cell expansion, and distortion
  • Compression of vasprun.xml files
  • Automatic division of DFT dataset
  • Atomic energies
  • Enumeration of optimal MLPs
  • Estimation of computational costs

• Experimental features

  • SSCHA free energy model
  • Electronic free energy model
  • Substitutional disordered model

Calculations using polynomial MLP

In version 0.8.0 or earlier, polymlp files are generated in a plain text format as polymlp.lammps. Starting from version 0.9.0, the files are generated in YAML format as polymlp.yaml. Both formats are supported by the following command-line interface and the Python API.

• Energy, forces, stress tensor

• Equation of states

• Local geometry optimization

• Elastic constants

• Phonon properties, Quasi-harmonic approximation

• Force constants

• Polynomial invariants

• Experimental features

  • Self-consistent phonon calculations
  • Molecular dynamics
  • Thermodynamic integration using molecular dynamics
  • Thermodynamic property calculation
  • Evaluation of atomic-configuration-dependent electronic free energy
  • Global structure optimization
  • Structure optimization at finite temperatures

• How to use polymlp in other calculator tools

  • LAMMPS
  • phonopy and phonon3py
  • ASE

Tutorials

1. Development of on-the-fly MLP
2. Development of general-purpose MLP