ACEpotentials.jl: A Julia implementation of the atomic cluster expansion

Area of Science:

• Computational materials science
• Quantum mechanics
• Atomistic modeling

Background:

• Developing accurate interatomic potentials is crucial for large-scale atomistic simulations.
• Existing methods can be computationally expensive and require large datasets.
• The Atomic Cluster Expansion offers a robust framework for descriptor development.

Purpose of the Study:

• Introduce ACEpotentials.jl, a Julia package for constructing interatomic potentials.
• Leverage the Atomic Cluster Expansion for data-efficient and systematically improvable potentials.
• Demonstrate the package's capabilities in standard atomistic modeling workflows.

Main Methods:

• Utilizing the Julia programming language for high-performance computing.
• Implementing the Atomic Cluster Expansion for describing atomic environments.
• Employing linear models and Bayesian active learning for potential development.

Main Results:

• ACEpotentials.jl provides a simple, interpretable, and robust tool for potential construction.
• The package demonstrates high performance in prototypical atomistic modeling tasks.
• Systematically improvable and data-efficient potentials are generated.

Conclusions:

• ACEpotentials.jl facilitates the creation of accurate and efficient interatomic potentials.
• The software package enhances the application of the Atomic Cluster Expansion in materials science.
• This tool supports advanced techniques like Bayesian active learning for model refinement.