The Mechanisms for Preferential Attachment of Nanoparticles in Liquid Determined Using Liquid Cell Electron Microscopy, Machine Learning, and Molecular Dynamics

Taylor Woehl, David Welch, Chiwoo Park, Roland Faller, James Evans, Nigel Browning
2016 Microscopy and Microanalysis  
Optimization of colloidal nanoparticle synthesis techniques requires an understanding of underlying particle growth mechanisms. Non-classical particle-mediated growth mechanisms are particularly important as they affect nanoparticle size and shape distributions and can lead to formation of superstructures [1] . Nanomaterial growth by non-classical particle-mediated growth mechanisms is complex and often proceeds via multiple assembly pathways simultaneously [2] . Ensemble techniques for
more » ... rizing the dynamics of these growth pathways, such as x-ray or visible light spectroscopy, average the dynamics associated with these multiple pathways, making them impossible to distinguish. Liquid cell electron microscopy has emerged as an in situ technique providing an ideal combination of high spatial resolution visualization of nanomaterial structure and in situ dynamics. Measurements of growth and assembly dynamics and nanoparticle interactions from in situ movies are typically performed manually using simple image analysis algorithms; however, these simple algorithms suffer from human error and bias and lack in statistics [3] . The use of advanced machine learning-based image analysis algorithms [3] to interpret dynamic liquid cell electron microscopy in situ movies promises to provide insights into complex nanoparticle growth pathways. Here we employ liquid cell scanning transmission electron microscopy (STEM), machine learning-based image analysis [4], and steered molecular dynamics (SMD) simulations to demonstrate that the experimentally observed preference for end-to-end attachment of silver nanorods is a result of weaker solvation forces occurring at rod ends [5] . Silver nanoparticle growth and attachment processes were directly observed and initiated via liquid cell STEM [6] . Silver nanocrystals grew on the silicon nitride window surface, with their size increasing monotonically with time through monomer addition (of reduced silver ions) as well as aggregation of neighboring nanocrystals [7] . For the experimental parameters used in this study, nanoparticles often grew into asymmetric shapes, such as nanorods (Figure 1a ). These mobile particles were observed to eventually attach to neighboring particles, and showed a preference for end-toend attachment. Machine learning-based image analysis algorithms were used to identify interparticle attachment events and extract the orientation angles of the nanorods at the time of attachment (Figures 1b and 1c) [4]. SMD revealed that when the side of a nanorod approached another rod, perturbation of the surface-bound water at the nanorod surface created significant energy barriers to attachment (Figure 1d) . Additionally, rod morphology (i.e., facet shape) effects can explain the majority of the side attachment events that are observed experimentally [8] . 812
doi:10.1017/s1431927616004918 fatcat:a4ae2b4mk5hpjlngcaqku5qsz4