High-Throughput Evaluation of Mechanical Exfoliation Using Optical Classification of Two-Dimensional Materials.

Area of Science:

• Materials Science
• Nanotechnology
• Computational Science

Background:

• Mechanical exfoliation is key for high-quality two-dimensional (2D) materials but suffers from low yield.
• Analyzing large datasets of exfoliated samples is crucial for process optimization.
• Existing machine learning tools for classifying 2D material thickness are slow and require manual input.

Purpose of the Study:

• To develop an open-source, GPU-accelerated software platform for high-throughput analysis of 2D material samples.
• To enable scalable methods for analyzing deposited material and optimizing exfoliation processes.
• To improve the efficiency of identifying usable 2D material flakes.

Main Methods:

• Leveraging parallel computation and optimizing core algorithms for GPU acceleration.
• Automating preprocessing steps for analyzing uncompressed optical images at scale.
• Building upon existing machine learning classification techniques for enhanced performance.

Main Results:

• The software quantifies flake coverage and thickness across large datasets of optical images.
• Processes over 200x more pixel data with a 60x reduction in processing time compared to baseline methods.
• Classifies a dataset of 2916 images in 35 minutes, a significant improvement over the estimated 32 hours for the baseline method.

Conclusions:

• The developed platform enables rapid evaluation of exfoliation results.
• Provides a practical tool for optimizing 2D material deposition techniques.
• Aims to improve the overall yield of high-quality two-dimensional materials.