Microcrystal Growth Pathways Investigated with Machine Learning Segmentation and Classification in Scanning Electron...

Area of Science:

• Materials Science
• Nanotechnology
• Data Science

Background:

• Electron microscopy offers detailed local structural analysis but struggles with bulk sample quality assessment.
• Quantifying crystal size and distribution is crucial for material properties, especially in nanomaterials.

Purpose of the Study:

• To develop a machine learning (ML) tool for automated segmentation and classification of faceted crystals in scanning electron microscopy (SEM) images.
• To determine colloidal crystal sample quality by analyzing crystal size and product distribution.
• To investigate crystal growth pathways in DNA-mediated nanoparticle assembly.

Main Methods:

• A flexible machine learning tool was developed to process SEM micrographs.
• The tool segments and classifies faceted crystals, analyzing size and product distributions (single crystal, fused crystal, noncrystal).
• Applied to over 13,000 colloidal crystal products from DNA-mediated nanoparticle assembly.

Main Results:

• Machine learning analysis revealed distinct crystal size and product distributions correlating with growth pathways.
• Strong DNA bonds resulted in faster nucleation and smaller colloidal crystals.
• Increased thermal energy and crystallization time promoted nonclassical growth (coalescence) yielding larger crystals.

Conclusions:

• The developed ML tool accurately assesses bulk sample quality from SEM images.
• Experimental conditions can be precisely controlled to tailor colloidal crystal size and distribution.
• This facilitates the design and synthesis of colloidal crystal metamaterials.