# Third-party notices SODA builds on scientific software developed by other projects. Its original code is distributed under the {repo}`MIT License `; incorporated code, external libraries, and model files retain their respective licenses. This page summarizes the attribution recorded in the repository's {repo}`Third-Party Notices ` and reproduces the four upstream license texts stored with those notices. That document also identifies the individual SODA files derived from, inspired by, or adapted to each project. For references to include in a publication, see [Citing SODA](citations.md). ## At a glance | Project | Contribution to SODA | Recorded license | | --- | --- | --- | | [EON](#eon) | Components for atomistic optimization, saddle searches, and adaptive kinetic Monte Carlo. | BSD 3-Clause | | [EZGA](#ezga) | Evolutionary structure exploration through SODA's EZGA adapters. | MIT | | [NEBScape](#nebscape) | Geometry-pair preparation and transition-state search through SODA's NEBScape adapters. | MIT | | [SAGE](#sage) | Native structure and individual data types used by the EZGA integration. | Personal, academic, or non-commercial research use; commercial use requires written permission. | ## EON SODA contains code derived in whole or in part from [EON, SVN revision r2545](https://theory.cm.utexas.edu/svn/eon/?p=2545). This archived revision identifies the upstream source used by SODA. The incorporated components include optimizers, saddle-search methods, atomic structures, potential interfaces, and AKMC state handling. Many have been substantially modified for SODA. The {repo}`EON provenance table ` maps the relevant SODA files to their upstream counterparts. It also distinguishes SODA's local workflow-controller refactoring from identified EON-derived components. The {ref}`citation guide ` lists the AKMC method paper and the 2014 EON software paper separately. Copyright © 2010, EON Development Team. The retained license is reproduced below from {repo}`third_party/eon/LICENSE `. ```{literalinclude} ../../third_party/eon/LICENSE :language: text ``` ## EZGA SODA integrates [EZGA](https://gitlab.mpcdf.mpg.de/fhi-theory/EZGA) through adapters for its engine, agents, populations, workers, and configuration. These adapters preserve reviewed EZGA behavior while SODA supplies MPI communication, execution, and HDF5 storage. The reviewed distribution is **`ezga-lib==1.2.31`**. The recorded source reference is the PyPI distribution metadata; no reviewed EZGA Git commit is recorded in the notices. See the {repo}`EZGA provenance table ` for the adapted interfaces and behavior. Copyright © 2026 Juan Manuel Lombardi, Felix Riccius, Charles W.P. Pare, Hendrik H. Heenen, Karsten Reuter, and Christoph Scheurer. The retained license is reproduced below from {repo}`third_party/ezga/LICENSE `. ```{literalinclude} ../../third_party/ezga/LICENSE :language: text ``` ## NEBScape SODA integrates [NEBScape](https://gitlab.mpcdf.mpg.de/hjung/nebscape) through adapters for geometry-pair preparation, interpolation, and nudged elastic band calculations. These preserve the reviewed serial scientific behavior while SODA supplies in-memory data exchange, MPI execution, and HDF5 storage. The reviewed source revision is **`124b41add7a929a18b6fc58b367a51eed0d1c825`**. The installers select the Python distribution **`nebscape==0.3.1`**; the source revision is retained separately for provenance and the license reference. See the {repo}`NEBScape provenance table ` for the adapters and settings derived from that review. Copyright © 2025 Jung Hyunwook. The retained license is reproduced below from {repo}`third_party/nebscape/LICENSE `. ```{literalinclude} ../../third_party/nebscape/LICENSE :language: text ``` ## SAGE [SAGE (`sage-lib`)](https://pypi.org/project/sage-lib/0.1.9.29/) supplies the native `Partition` and `AtomPositionManager` data structures used to construct EZGA individuals from SODA's atomic data. SODA's EZGA installers pin **`sage-lib==0.1.9.29`**, the version whose license is retained here. The {repo}`SAGE provenance entry ` records the bridge and the upstream source-distribution link. **SAGE has a separate license for personal, academic, or non-commercial research use.** Commercial use requires prior written permission from its copyright holder. Redistribution is permitted only for non-commercial purposes and must retain the license and copyright notice. These terms apply to SAGE even though EZGA's own license is MIT. Copyright © 2026 Juan Manuel Lombardi. The complete retained terms, including the licensing contact, are reproduced below from {repo}`third_party/sage/LICENSE `. ```{literalinclude} ../../third_party/sage/LICENSE :language: text ``` ## Other libraries and model files An installation also uses libraries selected by its build options and workflow. The four notices above cover the recorded source provenance and integrations; they are not a complete license inventory of every installed dependency. | Component | Role | License information | | --- | --- | --- | | LAMMPS | Native simulation engine and potential evaluation. | [GNU GPL v2 license text](https://github.com/lammps/lammps/blob/c10219da9e366addec5aa2be8883fe9c419d91a0/LICENSE). | | Eigen | C++ linear algebra. | [Licensing overview](https://gitlab.com/libeigen/eigen/-/blob/3147391d946bb4b6c68edd901f2add6ac1f31f8c/COPYING.README); primarily MPL 2.0, with other licenses identified per file. | | pybind11 | C++/Python bindings. | {repo}`BSD 3-Clause license `. | | cpptrace | Stack traces and diagnostics. | [MIT license](https://github.com/jeremy-rifkin/cpptrace/blob/ce639ebfcec47a7c74233b4bab50017cb34e615b/LICENSE). | | Lyra | Command-line argument parsing. | [Boost Software License 1.0](https://github.com/bfgroup/Lyra/blob/56998c3ce72ea3240145de062503c278b75939dc/LICENSE.txt). | | Catch2 | C++ tests. | [Boost Software License 1.0](https://github.com/catchorg/Catch2/blob/2b60af89e23d28eefc081bc930831ee9d45ea58b/LICENSE.txt). | | pincheck | HPC process/thread placement diagnostics. | [MIT license](https://gitlab.mpcdf.mpg.de/khr/pincheck/-/blob/98eb3bcef222c8d7a75ef91eaefa82f21b89fc7f/LICENSE). | MPI and HDF5 implementations, Python packages, GPU toolkits, and optional potential backends also carry their own notices. Consult the license files supplied with the versions installed on your machine. For machine-learning potentials, check both the software license and the license accompanying the exact model checkpoint and any redistributed training data. For example, the [MACE model listings](https://github.com/ACEsuit/mace#pretrained-foundation-models) identify licenses for individual model families; a package's license alone does not describe every model it can load. The [citation guide](citations.md#potentials-models-and-simulation-engines) explains which scientific references to record with your model.