ACEfriction.jl

Documentation	Build Status	License

About ACEfriction.jl

ACEfriction.jl facilitates simulation and machine learning of configuration-dependent friction tensor models from data. In more general terms, ACEfriction.jl provides methods for efficient representation, learning, and evaluation of $E(3)$-equivariant symmetric positive semi-definite matrix-valued functions on 3D-point clouds, i.e., $E(3)$-equivariant functions of the form

$${\bf \Gamma}({\bf r}_1, \dots, {\bf r}_N, {z_1},\dots,{z_N}) \in \mathcal{SPSD}_{3N},$$

where ${\bf r}_i \in \mathbb{R}^3,\; (i=1,\dots,N)$, are the positions of points/particles in the point cloud, the $z_i$s are some discrete features (e.g., chemical element types) and $\mathcal{SPSD}_{3N} \subset \mathbb{R}^{3N \times 3N}$ is the set of $3N\times 3N$-dimensional positive semi-definite matrices.

The underlying representation is based on an equivariant Atomic Cluster Expansion and, as such, size-transferrable, i.e., models can be trained and evaluated on 3D-point clouds comprised of an arbitrary number, $N$, of points/particles.

Documentation

For details, please refer to the Documentation, which includes a function manual and Workflow Examples of fitting an electronic friction tensor as well as a momentum-conserving friction tensor model as commonly employed in Dissipative Particle Dynamics.

Installation

To install ACEfriction.jl run the following code in a Julia-REPL:

] registry add https://github.com/ACEsuit/ACEregistry
] add ACEfriction

More detailed instructions can be found in the Installation Guide of the Documentation.

Reference

If you use this code, please cite our paper:

@article{sachs2024equivariant,
  title={Equivariant Representation of Configuration-Dependent Friction Tensors in Langevin Heatbaths},
  author={Sachs, Matthias and Stark, Wojciech G and Maurer, Reinhard J and Ortner, Christoph},
  journal={arXiv preprint arXiv:2407.13935},
  year={2024}
}

License

ACEfriction.jl is published and distributed under the MIT License.