Dataset exploring the atomic scale structure and ionic dynamics of polyanion sodium cathode materials.

Area of Science:

• Materials Science
• Computational Chemistry
• Electrochemistry

Background:

• Polyanionic sodium cathode materials offer high stability and electrochemical performance for sodium-ion batteries.
• Exploring this vast chemical space requires efficient computational methods, such as machine learning (ML).

Purpose of the Study:

• To develop a comprehensive theoretical dataset for ML-guided discovery of polyanionic sodium cathode materials.
• To create accurate ML interatomic potentials for these materials.

Main Methods:

• Generated a large dataset of DFT-calculated structures for four polyanionic sodium cathode material types with various transition metals (TM).
• Included DFT structure optimizations, ab initio molecular dynamics, and ML-driven molecular dynamics simulations at 1000 K.
• Trained ML models on cathode-specific dataset subsets to develop ML interatomic potentials.

Main Results:

• The dataset comprises over 113,000 DFT-calculated structures with atomic charges and over 184,000 without.
• The developed ML interatomic potentials accurately reproduce DFT results for polyanionic sodium cathode materials.
• The dataset includes diverse single and multiple-transition metal compositions.

Conclusions:

• The created dataset and ML interatomic potentials significantly aid in the ML-driven discovery of advanced sodium-ion battery cathode materials.
• This approach accelerates the exploration of polyanionic materials for enhanced battery performance and stability.