CCS_fit - Fitting using Curvature Constrained Splines
The CCS_fit package is a tool to construct two-body potentials using the idea of curvature constrained splines.
Getting Started
Package Layout
ccs_fit-x.y.z
├── CHANGELOG.md
├── LICENSE
├── MANIFEST.in
├── README.md
├── bin
│ ├── ccs_build_db
│ ├── ccs_export_sktable
| ├── ccs_export_FF
│ ├── ccs_fetch
│ ├── ccs_fit
│ └── ccs_validate
├── docs
├── examples
│ └── Basic_Tutorial
│ └── tutorial.ipynb
│ └── Advanced_Tutorials
│ ├── CCS
│ ├── CCS_with_LAMMPS
│ ├── DFTB_repulsive_fitting
│ ├── ppmd_interfacing
│ ├── Preparing_ASE_db_trainingsets
│ ├── Search_mode
│ └── Simple_regressor
├── logo.png
├── poetry.lock
├── pyproject.toml
├── src
│ └── ccs
│ ├── ase_calculator
│ ├── common
│ ├── data
│ ├── debugging_tools
│ ├── fitting
│ ├── ppmd_interface
│ ├── regression_tool
│ └── scripts
│ ├── ccs_build_db.py
│ ├── ccs_export_FF.py
│ ├── ccs_export_sktable.py
│ ├── ccs_fetch.py
│ ├── ccs_fit.py
│ └── ccs_validate.py
└── tests
ccs_build_db- Routine that builds an ASE-database.ccs_fetch- Executable to construct the traning-set (structures.json) from a pre-existing ASE-database.ccs_fit- The primary executable file for the ccs_fit package.ccs_export_sktable- Export the spline in a dftbplus-compatible layout.ccs_export_FF- Fit the spline to commonly employed force fields; Buckingham, Morse and Lennard Jones.ccs_validate- Validation of the energies and forces of the fit compared to the training set.main.py- A module to parse input files.objective.py- A module which contains the objective function and solver.spline_functions.py- A module for spline construction/evaluation/output.
(Recommended) installing from pip
pip install ccs_fit
Installing from source using poetry
git clone https://github.com/Teoroo-CMC/CCS_fit.git ccs_fit
cd ccs_fit
# Install python package manager poetry (see https://python-poetry.org/docs/ for more explicit installation instructions)
curl -sSL https://install.python-poetry.org | python3 -
# You might have to add poetry to your PATH
poetry --version # to see if poetry installed correctly
poetry install # to install ccs_fit
Tutorials
We provide tutorials in the examples folder. To run the example, go to one of the folders. Each contain the neccesery input files required for the task at hand. A sample CCS_input.json for O$_2$ is shown below:
{
"General": {
"interface": "CCS"
},
"Train-set": "structures.json",
"Twobody": {
"O-O": {
"Rcut": 2.5,
"Resolution": 0.02,
"Swtype": "sw"
}
},
"Onebody": [
"O"
]
}
The CCS_input.json file should provide at a minimum the block “General” specifying an interface. The default is to look for input structures in the file structure.json file. The format for structure.json is shown below :
{
"energies":{
"S1": {
"Energy": -4.22425752,
"Atoms": {
"O": 2
},
"O-O": [
0.96
]
},
"S2": {
"Energy": -5.29665634,
"Atoms": {
"O": 2
},
"O-O": [
0.98
]
},
"S3": {
"Energy": -6.20910363,
"Atoms": {
"O": 2
},
"O-O": [
1.0
]
},
"S4": {
"Energy": -6.98075271,
"Atoms": {
"O": 2
},
"O-O": [
1.02
]
}
}
}
The structure.json file contains different configurations labeled (“S1”, “S2”…) and corresponding energy, pairwise distances (contained in an array labelled as “O-O” for oxygen). The stoichiometry of each configuration is given under the atoms label (“Atoms”) as a key-value pair (“O” : 2 ).
To perform the fit :
ccs_fit
The following output files are obtained:
CCS_params.json CCS_error.out ccs.log
CCS_params.json - Contains the spline coefficients, and one-body terms for two body potentials.
error.out - Contains target energies, predicted energies and absolute error for each configuration.
ccs.log - Contains debug information
Funding
This project has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 957189, and the Swedish National Strategic e-Science programme eSSENCE. The project is part of BATTERY 2030+, the large-scale European research initiative for inventing the sustainable batteries of the future.
License
This project is licensed under the GPLv3 License - see the LICENSE file for details.
Acknowledgement
We want to thank Pavlin Mitev, Christof Köhler, Matthew Wolf, Kersti Hermansson, Bálint Aradi and Tammo van der Heide, and all the members of the TEOROO-group at Uppsala University, Sweden for fruitful discussions and general support.