Microcrystal\nGrowth Pathways Investigated with Machine\nLearning Segmentation and Classification in Scanning Electron Microscopy

Advances\nin electron microscopy have revolutionized material characterization\non the nano- and microscales, providing important insights into local\nordering, structure, and size and quality distributions. While shape\nand size can be rigorously quantified through microscopy, it is often\nlimited to local structure analysis and fails to describe bulk sample\nquality. Herein, a flexible machine learning (ML) tool is described\nthat can segment and classify faceted crystals in scanning electron\nmicroscopy (SEM) micrographs to determine sample quality through the\ncrystal size and product distribution. As a case study, this tool\nwas applied to investigate crystal growth pathways (classical nucleation\nand growth compared to nonclassical growth) in DNA-mediated nanoparticle\nassembly through size and product (single crystal, fused crystal,\nor noncrystal) distribution of samples containing over 13000 colloidal\ncrystal products. Strong DNA bond strengths (controlled by DNA sequence)\nlead to fast nucleation that exhausts the monomer concentration, resulting\nin smaller colloidal crystals. Alternatively, increased thermal energy\nand crystallization time lead to nonclassical crystallization pathways\n(coalescence) that result in larger colloidal crystals. This tool\nis useful since experimental conditions can now be deliberately identified\nto control colloidal crystal size and size distribution, important\nconsiderations for researchers interested in designing and synthesizing\ncolloidal crystal metamaterials.