Retrieving the Quantitative Chemical Information at Nanoscale from Scanning Electron Microscope Energy Dispersive...

Area of Science:

• Materials Science
• Nanotechnology
• Machine Learning

Background:

• Accurate chemical composition determination is crucial for nanoscale materials.
• Traditional methods for analyzing nanostructures can be time-consuming and costly.
• Energy dispersive X-ray spectroscopy (EDX) in scanning electron microscopy (SEM) is a common analysis technique.

Purpose of the Study:

• To develop and validate a machine learning-based method for quantitative chemical composition analysis of nanostructures.
• To apply blind source separation (BSS) using non-negative matrix factorization (NMF) to EDX spectrum images.
• To demonstrate a faster and more cost-efficient approach for nanoscale chemical quantification.

Main Methods:

• Utilized blind source separation (BSS) with non-negative matrix factorization (NMF) on EDX spectrum image maps from SEM.
• Decomposed EDX data into distinct source components: nanostructures, semiconductor substrate, and background.
• Validated results with Monte Carlo simulations and cross-sectional transmission electron microscopy (TEM) EDX.

Main Results:

• Successfully determined the quantitative composition of metal alloy nanowires on InSb and gold nanostructures on germanium.
• The BSS-NMF method accurately separated signals from nanostructures, bulk semiconductor, and carbon background.
• Recovered quantitative composition was validated by independent simulation and experimental methods.

Conclusions:

• SEM EDX measurements combined with machine learning BSS-NMF provide a robust method for nanostructure chemical composition determination.
• This approach significantly enhances the speed and cost-efficiency of nanoscale chemical quantification.
• The findings have broad implications for materials characterization across various systems.