ase-gpatom 1.0.0

Atomistic simulation tools based on Gaussian processes

gpatom is a Python package which provides several tools for geometry optimisation and related tasks in atomistic systems using machine learning surrogate models. gpatom is an extension to the Atomic Simulation Environment.

gpatom consists of two independently maintained sub-repositories, gpatom/aidts-gpmin and gpatom/ase-gpatom.

gpatom/aidts-gpmin includes:

• GPMin: An atomistic optimization algorithm based on Gaussian processes.

• AIDMin: Energy structure minimization through the Artificial-Intelligence framework.

• AIDNEB: Nudged Elastic Band calculations through the Artificial-Intelligence framework.

• AIDTS: Transition State Search through the Artificial-Intelligence framework.

• AIDMEP: Minimum Energy Pathway through the Artificial-Intelligence framework.

gpatom/ase-gpatom includes:

• BEACON: Bayesian Exploration of Atomic CONfigurations.

  BEACON does global optimization by Bayesian optimization by training the model with the DFT forces on atoms. Represents the atoms with a global structural fingerprint. Works generally well for different kinds of atomic systems: clusters, surfaces, bulk systems. For usage, see Gitlab Wiki: https://gitlab.com/gpatom/ase-gpatom/-/wikis/How-to-use-BEACON

List of related publications for gpatom/ase-gpatom:

• BEACON:

  Global optimization of atomic structures with gradient-enhanced Gaussian process regression

  S. Kaappa, E. G. del Río, K. W. Jacobsen

  Physical Review B, vol. 103, 174114 (2021). https://doi.org/10.1103/PhysRevB.103.174114

• ICE-BEACON:

  Atomic Structure Optimization with Machine-Learning Enabled Interpolation between Chemical Elements

  S. Kaappa, C. Larsen, K. W. Jacobsen

  Physical Review Letters, vol. 127, 166001 (2021). https://doi.org/10.1103/PhysRevLett.127.166001

• Ghost-BEACON:

  Machine-learning-enabled optimization of atomic structures using atoms with fractional existence

  C. Larsen, S. Kaappa, A. L. Vishart, T. Bligaard, K. W. Jacobsen

  Physical Review B, vol. 107, 214101 (2023). https://doi.org/10.1103/PhysRevB.107.214101

• BEACON with MACE prior:

  Bayesian optimization of atomic structures with prior probabilities from universal interatomic potentials

  P. Lyngby, C. Larsen, K. W. Jacobsen

  Physical Review Materials, vol. 8, 123802 (2024). https://doi.org/10.1103/PhysRevMaterials.8.123802

• ICE, Ghost and Hyperspatial optimization generalized to arbitrary many elements and variable unit cell:

  Global atomic structure optimization through machine- learning-enabled barrier circumvention in extra dimensions

  C. Larsen, S. Kaappa, A. L. Vishart, T. Bligaard, K. W. Jacobsen

  npj computational materials, 11, 222 (2025). https://doi.org/10.1038/s41524-025-01656-9

Contact

Please join our #gpatom channel on Matrix.

Installation cheat sheet

To install latest release from pypi, use:

$ pip install ase-gpatom

To install a developer version (allows in-place edits of the code), clone the sourcecode and go to the toplevel gpatom directory, then run:

$ git clone https://gitlab.com/gpatom/ase-gpatom.git
$ pip install --editable ase-gpatom

Testing cheat sheet

To run the tests, go to the test directory and run:

$ pytest

Run the tests in parallel on n cores (requires pytest-xdist):

$ pytest -n 4

Show tests without running them:

$ pytest --collectonly

Run tests in particular module:

$ pytest test_module.py

Run tests matching pattern:

$ pytest -k <pattern>

Show output from tests:

$ pytest -s

Note that since many tests write files, temporary directories are created for each test. The temporary directories are located in /tmp/pytest-of-<username>/. pytest takes care of cleaning up these test directories.

Use pytest.ini and test/conftest.py to customize how the tests run.