CarNet

Cartesian Natural Tensor Networks - atomistic machine learning for interatomic potentials, scalar properties, tensorial properties and beyond.

Installation

It is recommended to create a virtual environment first (e.g. using conda) before installing.

1. Install PyTorch

Follow the official PyTorch installation guide to install PyTorch for your platform and CUDA version.

2. Install natt

pip install git+https://github.com/wengroup/natt.git

3. Clone and install CarNet

git clone https://github.com/wengroup/carnet.git
cd carnet
pip install -e .

(Optional) To include dependencies for testing:

pip install -e ".[test]"

Features

• Interatomic potentials: train machine learning interatomic potentials with interfaces for:
  • ASE via a calculator
  • LAMMPS via the MLIAP package
• Tensorial properties: predict tensorial properties (e.g. polarizability, dielectric tensors, elastic tensors) of molecules and materials

Examples

See the examples directory for more details on how to use CarNet in ASE and LAMMPS.

Citation

Chen, Q., Pattamatta, A.S.L., Wang, B., Srolovitz, D.J. and Wen, M., 2026. Atomistic Machine Learning with Irreducible Cartesian Natural Tensors. arXiv preprint arXiv:2510.04015.

@article{chen2026atomistic,
  title   = {Atomistic Machine Learning with Irreducible Cartesian Natural Tensors},
  author  = {Chen, Qun and Pattamatta, ASL and Wang, Boyu and Srolovitz, David J and Wen,
  Mingjian},
  journal = {arXiv preprint arXiv:2510.04015},
  year    = {2025},
  doi     = {10.48550/arXiv.2510.04015},
}