Non-negative matrix factorization-aided phase unmixing and trace element quantification of STEM-EDXS data.

Area of Science:

• Materials Science
• Geochemistry
• Electron Microscopy

Background:

• Scanning transmission electron microscope-energy-dispersive X-ray spectroscopy (STEM-EDXS) is vital for material chemical characterization.
• Quantification in STEM-EDXS is difficult for samples with overlapping phases and common elements.
• Electron beam-sensitive materials, like those from Earth's deep mantle, pose additional analytical challenges.

Purpose of the Study:

• To develop a semi-automated methodology for identifying, segmenting, and unmixing phases with significant spectral and spatial overlap in STEM-EDXS data.
• To retrieve accurate energy-dispersive X-ray spectra and phase abundance maps for complex materials.
• To enable reliable trace element quantification in challenging sample matrices.

Main Methods:

• Combining non-negative matrix factorization (NMF) with a priori knowledge of the sample.
• Applying the methodology to an electron beam-sensitive mineral assemblage from Earth's deep mantle.
• Semi-automated phase identification, segmentation, and spectral unmixing.

Main Results:

• Successfully retrieved true EDX spectra for constituent phases in a complex mineral assemblage.
• Generated accurate phase abundance maps, resolving spatial overlap.
• Achieved reliable quantification of trace elements down to approximately 100 ppm concentration levels.
• Demonstrated the method's effectiveness on an electron beam-sensitive sample.

Conclusions:

• The developed methodology effectively overcomes challenges in STEM-EDXS quantification caused by phase overlap and limited signal-to-noise ratio.
• This approach enhances the chemical characterization capabilities for diverse materials systems.
• The technique is adaptable for analyzing various materials with complex spectral and spatial characteristics.