DScribe

DScribe is a Python package for transforming atomic structures into fixed-size numerical fingerprints. These fingerprints are often called “descriptors” and they can be used in various tasks, including machine learning, visualization, similarity analysis, etc. To get started you can check the basic tutorial.

Note

We are incrementally adding support for calculating the derivatives of descriptors with respect to the atom positions. In version 1.0.0 you can find an implementation for getting derivatives of non-periodic systems for the SOAP descriptor. Please check the new documentation on derivatives and the new new tutorial on predicting energies and forces using these derivatives.

Note

Version 1.0.x introduces some improvements that are not directly backwards compatible with previous versions. Check the changelog for details.

Capabilities at a Glance

DScribe currently includes the following descriptors:

• Coulomb matrix

• Sine matrix

• Ewald sum matrix

• Atom-centered Symmetry Functions (ACSF)

• Smooth Overlap of Atomic Positions (SOAP)

• Many-body Tensor Representation (MBTR)

• Local Many-body Tensor Representation (LMBTR)

Check the tutorials for more information.

Go Deeper

You can find more details in our open-access article: DScribe: Library of descriptors for machine learning in materials science

Documentation for the source code can be found here. The full source code with examples and regression tests can be explored at github.