papers-autoplex-rss

This repository contains workflow configurations, raw data, and plotting scripts to accompany the article:

An automated framework for exploring and learning potential-energy surfaces
Yuanbin Liu, Joe D. Morrow, Christina Ertural, Natascia L. Fragapane, John L. A. Gardner,
Aakash A. Naik, Yuxing Zhou, Janine George, Volker L. Deringer

This work showcases the use of autoplex (v0.0.7), a software framework for the automated generation of high-quality machine-learned interatomic potentials across a variety of material systems, including Si, Ti–O, silica, water/ice, Ge-Sb-Te, and In-Sb-Te.

Contents

1. Workflows/: YAML configuration files and Python scripts for running the automated potential-generation workflows.
2. Figures/: Raw data and Jupyter notebooks used to generate the figures presented in the paper.
3. Tables/: Raw data used to reproduce the tables reported in the paper.
4. Structures/: MD trajectories and DFT-optimized structures.

Note: the force-field parameter files and reference data are available in Zenodo.

Installation

To install the required packages to run the full workflow from scratch and reproduce all figures, first create a clean Python environment (e.g., using conda or venv), then run:

pip install -r requirements.txt

Key references

The following tools were used for performing some data analysis in this study:

1. graph-pes: a user-friendly interface for training graph-based machine-learned interatomic potentials.

2. RingsStatisticsMatter.jl: a Julia package for shortest-path ring statistics analysis.

3. OVITO Python API (DOI): a library that provides access to OVITO's post-processing and visualization capabilities.

License

The contents of the workflows folder are licensed under the MIT License (LICENSE_CODE), while the rest of this repository is released under the CC BY license (LICENSE_DATA).