Machine Learning Enhanced Optical Microscopy for the Rapid Morphology Characterization of Silver Nanoparticles.

Area of Science:

• Materials Science
• Nanotechnology
• Computational Imaging

Background:

• Accurate nanoparticle characterization (size, shape) is crucial for determining material properties and applications.
• Conventional microscopy methods can be time-consuming or lack precision for nanoscale analysis.

Purpose of the Study:

• To develop a computational imaging platform for rapid nanoparticle characterization using conventional optical microscopy.
• To establish a machine learning model for precise size and morphology prediction of nanoparticles.

Main Methods:

• Utilized through-focus scanning optical microscopy (TSOM) on a standard optical microscope.
• Developed and trained a machine learning model using TSOM image data.
• Applied the model to characterize silver nanocubes and silver nanowires.

Main Results:

• Achieved <5% estimation error for individual silver nanocube size.
• Reported 1.6% error for averaged size and 0.4 nm error for standard deviation at the ensemble level.
• Successfully identified sharp-tip vs. blunt-tip silver nanowires with 82% accuracy.

Conclusions:

• The computational imaging platform offers a rapid and accurate method for nanoparticle characterization.
• The developed machine learning model demonstrates potential for online monitoring of nanoparticle synthesis.
• The approach is extendable to more complex nanomaterials, including anisotropic and dielectric nanoparticles.