Large-Scale Atom Probe Tomography Data Mining: Methods and Application to Inform Hydrogen Behavior.

Area of Science:

• Materials Science
• Analytical Chemistry
• Data Science

Background:

• Atom probe tomography (APT) generates large, complex datasets.
• Statistical analysis of APT data is crucial for material characterization.
• Handling and interpreting vast amounts of spectral data presents a significant challenge.

Purpose of the Study:

• To develop an efficient method for analyzing large collections of APT spectra.
• To enable automated searching and clustering of APT spectra based on similarity.
• To investigate the behavior of hydrogen as a contaminant in APT experiments.

Main Methods:

• Collected and organized numerous APT datasets into a database.
• Represented APT spectra as a point cloud.
• Utilized a city block distance metric for spectral dissimilarity measurement.
• Performed statistical analysis on spectral clusters to infer material properties.

Main Results:

• Developed an automated, similarity-based spectral searching technique.
• Identified clusters in the spectral point cloud corresponding to similar materials.
• Correlated hydrogen behavior (H2+/H+ ratio) with electric field, voltage, and base material.
• Observed a decrease in H2+/H+ ratio with increasing electric field, suggesting postionization.
• Found that absolute amounts of H2+ and H+ often increase during APT experiments.

Conclusions:

• The point cloud approach offers an effective strategy for managing and analyzing large-scale APT data.
• This method facilitates rapid material identification and characterization from spectral data.
• The study provides valuable insights into the behavior and detection of hydrogen contamination in APT.
• Understanding hydrogen behavior is critical for accurate material analysis using APT.