Statistical Characterization of the Morphologies of Nanoparticles through Machine Learning Based Electron Microscopy...

Area of Science:

• Materials Science
• Nanotechnology
• Computational Science

Background:

• Transmission electron microscopy (TEM) is vital for nanoparticle morphology analysis.
• Quantitative and statistical analysis of nanoparticle morphology via TEM is challenging.
• Current methods lack the throughput for large-scale statistical analysis.

Purpose of the Study:

• To develop a high-throughput, statistically robust method for nanoparticle morphology analysis using TEM images.
• To enable precise quantitative analysis of large nanoparticle populations.
• To establish guidelines for representative sampling in nanoparticle characterization.

Main Methods:

• Application of a genetic algorithm to image analysis for nanoparticle morphology.
• Development of a mass-throughput analysis pipeline.
• Clustering of nanoparticles based on morphological similarity for statistical analysis.

Main Results:

• Analysis of over 150,000 nanoparticles with 99.75% precision and 0.25% false discovery rate.
• Identification of at least 1,500 nanoparticles needed for 95% credible interval representation.
• Demonstration of the importance of statistical distribution for optical property estimation.

Conclusions:

• The developed method significantly enhances the statistical analysis of nanoparticle morphology.
• This approach facilitates big data analysis in nanoparticle research.
• The findings provide critical insights for accurate representation and optical property prediction of nanoparticles.