Application of data science tools to quantify and distinguish between structures and models in molecular dynamics...

Area of Science:

• Materials Science
• Computational Materials Science
• Data Science

Background:

• Extracting knowledge from complex material structures requires effective quantification methods.
• High-dimensional data from multiscale simulations and experiments pose a challenge for analysis.
• Bridging the gap between rich hierarchical structures and low-dimensional representations is crucial.

Purpose of the Study:

• To develop and demonstrate a data-driven approach for structure quantification at the atomic scale.
• To address the challenge of transforming high-dimensional material structure data into low-dimensional representations.
• To enable efficient knowledge extraction from materials simulations and experiments.

Main Methods:

• Digital representation of material structures.
• Extraction of structure measures using n-point spatial correlations.
• Identification of low-dimensional measures via principal component analyses (PCA).

Main Results:

• Successfully applied novel protocols to molecular dynamics (MD) simulation datasets.
• Demonstrated the classification of datasets based on input parameters like interatomic potential and temperature.
• Validated the efficacy of the data-driven approach for atomic-scale structure quantification.

Conclusions:

• The developed data-driven approach effectively quantifies material structure at the atomic scale.
• This method facilitates the classification and understanding of simulation data.
• It offers a valuable tool for materials knowledge extraction from multiscale data.