amorphouspy

A Python package for atomistic simulations of glasses.

amorphouspy provides capabilities in the form of Python functions from generating initial structural models through running molecular dynamics simulations with LAMMPS, all the way to computing material properties and performing detailed structural analysis.

amorphouspy_api strings these Python functions together to end-to-end workflows using executorlib. It then exposes these workflows through web endpoints, which can e.g. be used by an agent powered by a large language model.

Find the documentation at glasagent.github.io/amorphouspy.

Key Features

amorphouspy

• Structure Generation: Create random oxide glass structures from composition dicts (e.g. {"SiO2": 75, "Na2O": 15, "CaO": 10}) with automatic density estimation using Fluegel's empirical model.
• Interatomic Potentials: Built-in support for PMMCS (Pedone), BJP (Bouhadja), SHIK (Sundararaman), Du/Teter, BMP (Bertani, Menziani, Pedone) classical force fields with automatic LAMMPS input generation. Support for machine-learning interatomic potentials will come later.
• Melt-Quench Simulations: Multi-stage heating/cooling protocols with potential-specific temperature programs and ensemble control.
• Structural Analysis: RDFs, coordination numbers, Qn distributions, bond angle distributions, ring statistics, cavity analysis.
• Property Calculations: Elastic moduli (stress-strain finite differences), viscosity (Green-Kubo formalism), coefficient of thermal expansion (temperature scan or from NPT fluctuations).

amorphouspy_api

• End-to-end Workflows: Submit a composition, get back the glass structure and all of its predicted properties
• Job management API: Fully fledged job management API, including results database
• Visualization: Interactive Plotly-based visualizations of glass structure and properties

Installation

Install the bleeding edge version via

# Install pixi (if not already installed)
curl -fsSL https://pixi.sh/install.sh | bash

# Clone repo and install environment from PyPI/conda-forge
git clone https://github.com/glasagent/amorphouspy.git
cd amorphouspy
pixi install

You can also install the amorphouspy package into existing environments from PyPI or conda-forge, see the Installation guide for details.

Quick Start

See the Tutorial for a step-by-step introduction.

Authors

Developed at the Bundesanstant für Materialforschung und -prüfung (BAM) in collaboration with SCHOTT AG and the Max Planck Institute for Sustainable Materials.

• Achraf Atila — BAM — Core framework, analysis tools, potentials
• Marcel Sadowski — Schott AG — CTE simulation module
• Jan Janssen — MPI-SusMat — pyiron integration, lammpsparser
• Leopold Talirz — Schott AG — API layer, project coordination

Acknowledgement

Supported by the Federal Ministry of Research, Technology and Space via the GlasAgent MaterialDigital 3 project.

How to Cite

If you use amorphouspy in your research, please cite it using the citation information provided on our Zenodo record (see the "Citation" box on the right-hand side), or use the reference below.

@misc{amorphouspy,
  doi = {10.5281/ZENODO.19564118},
  url = {https://zenodo.org/doi/10.5281/zenodo.19564118},
  author = {Leopold Talirz, and Achraf Atila, and Jan Janssen, and Marcel Sadowski,  },
  title = {glasagent/amorphouspy: v0.8.1},
  publisher = {Zenodo},
  year = {2026},
  copyright = {Apache License 2.0}
}

License

Apache License 2.0. See LICENSE.