Classification of apatite structures via topological data analysis: a framework for a 'Materials Barcode'...

Area of Science:

• Materials Science
• Computational Chemistry
• Data Science

Background:

• Classifying complex inorganic crystal structures is challenging.
• Traditional methods like structure maps have limitations in uncovering nuanced relationships.
• Unsupervised machine learning offers potential for automated discovery in crystal chemistry.

Purpose of the Study:

• To introduce topological data analysis (TDA) as an unsupervised machine learning tool for crystal chemistry.
• To apply TDA using persistent homology to apatite chemistry to identify classification criteria.
• To develop a novel visualization scheme, the
• Materials Barcode
• , for understanding structural similarity.

Main Methods:

• Utilized topological data analysis (TDA) and persistent homology.
• Employed crystal chemistry descriptors as input for TDA.
• Analyzed the topological connectivity of apatite stoichiometries.

Main Results:

• TDA automatically identified a hierarchical classification scheme within apatites.
• Classification was based on the commonality of discrete coordination polyhedra.
• A visualization scheme, the
• Materials Barcode
• , was developed to track similarity persistence.

Conclusions:

• TDA provides an automated, exploratory machine learning tool for crystal chemistry.
• The
• Materials Barcode
• offers a nuanced insight into structural associations and compound similarity.
• This approach advances the analysis of complex crystal chemistry databases.