Enhancing Element Identification by Expectation-Maximization Method in Atom Probe Tomography.

Area of Science:

• Materials Science
• Analytical Chemistry
• Data Science

Background:

• Atom Probe Tomography (APT) is a powerful technique for nanoscale chemical analysis.
• Accurate elemental identification is crucial for interpreting APT data.
• Existing methods can struggle with low signal-to-noise ratios or overlapping mass spectra.

Purpose of the Study:

• To present a novel Bayesian approach for elemental assignment in APT data.
• To enhance data visualization and composition mapping capabilities in APT.
• To improve the analysis of challenging datasets, such as those with low counts or overlapping peaks.

Main Methods:

• Utilizes Bayesian assignation of labels via the expectation-maximization (EM) algorithm.
• Assumes characteristic shapes for mass-over-charge peaks in mass spectra.
• Determines the probability of assigning each atom to a specific element.

Main Results:

• The method effectively assigns elemental identities to atoms in APT datasets.
• Demonstrates enhanced data visualization and composition mapping.
• Particularly efficient for small count experiments with low signal-to-noise ratios.
• Successfully resolves overlapping Si and N peaks in a steel sample.

Conclusions:

• The proposed Bayesian EM method offers an effective alternative for elemental identification in APT.
• It enhances the ability to extract meaningful information from raw APT data.
• Complements existing single-ion decomposition methods and is valuable for complex analyses.