Automatic Reaction KInetic MEchanism DEsign
arkimede is a Python framework for studying catalytic reactions on solid surfaces. It combines transition-state (TS) searches, machine-learning potentials (MLPs), and density-functional theory (DFT) calculations within unified workflows based on the Atomic Simulation Environment (ASE). The goal of arkimede is to facilitate the systematic exploration of catalytic mechanisms of surface reactions with efficient TS-search algorithms and active-learning strategies.
Clone and install the repository:
git clone https://github.com/raffaelecheula/arkimede.git
cd arkimede
pip install -e .Core dependencies:
- NumPy
- SciPy
- Matplotlib
- Atomic Simulation Environment (ASE)
Other dependencies:
- Sella (TS-searches methods)
- MLP codes (CHGNet, MACE, OCP, FAIRChem)
- DFT codes (Quantum Espresso, VASP)
- PyYAML (YAML files management)
- Shephex (SLURM jobs management)
The examples directory contains tutorial-style workflows.
Run TS searches using universal MLPs (CHGNet, MACE, OCP, FAIRChem). First, the initial and final states of elementary reactions are relaxed. TS guess structures are then generated using the IDPP interpolation method. Optionally, NEB calculations can be performed to improve the TS guesses. Finally, single-structure TS-search methods (Dimer, ARPESS, Sella, and BA-Sella) are used to calculate TS structures and energies.
Run single-structure TS searches with DFT (Quantum Espresso or VASP), starting from TS guess structures obtained with the IDPP interpolation method.
Run single-structure TS searches with DFT (Quantum Espresso or VASP), starting from TS guess structures obtained with TS searches using a MLP.
Run active learning TS searches, in which MLPs are iteratively fine-tuned using forces evaluated with DFT single-point calculations. A different MLP is fine-tuned for each TS structure.
Run active learning TS searches, in which MLPs are iteratively fine-tuned using forces evaluated with DFT single-point calculations. The same MLP is fine-tuned using data from all TS structures.
Contributions to arkimede are welcome. If you have suggestions, bug reports, or would like to contribute code, please open an issue or submit a pull request on the GitHub repository.
This project is licensed under the GNU General Public License v3.0. See the LICENSE file for details.
Raffaele Cheula (email: cheula.raffaele@gmail.com)