The Atomic Energy Network

Atomic interaction potentials based on artificial neural networks.


The Atomic Energy Network (ænet) is a software package [1,2] for the construction and usage of atomic interaction potentials based on artificial neural networks (ANNs). In essence, ænet offers tools to train ANNs to the potential energy of atomic reference structures. In addition, the package provides C and Fortran libraries that can be integrated in existing simulation software to actually use ANN potentials in atomistic simulations.


ænet is free and open-source software: you can redistribute it and/or modify it under the terms of the Mozilla Public License as published by the Mozilla Foundation.

This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the Mozilla Public License for more details.

Method and implementation

The ænet code is a modern Fortran implementation of the ANN potential method (with C interoperability) [1]. Publications that made use of ænet for various applications can be found on the publications page.

Obtain ænet

We welcome other scientists to use ænet for their research and to contribute to the development of the package and the methodology.

Release notes and a link to the source code are on the download page.


Many questions regarding the usage of ænet are addressed in the documentation or have already been answered in the forum and mailing list. If you run into problems with ænet or if you have a general question, please subscribe to the aenet forum and mailing list so that the entire ænet community benefits from the conversation.


This work used the Extreme Science and Engineering Discovery Environment (XSEDE), which is supported by National Science Foundation grant number ACI-1053575. Since 2019 ænet development has used resources of the Center for Functional Nanomaterials, which is a U.S. DOE Office of Science Facility, at Brookhaven National Laboratory under Contract No. DE-SC0012704. Development has also been supported by the Department of Chemical Engineering at Columbia University.


[1] N. Artrith and A. Urban, Comput. Mater. Sci. 114 (2016) 135-150. (Editor’s Choice)
[2] N. Artrith, A. Urban, and G. Ceder, Phys. Rev. B 96 (2017) 014112.