Real-time tracking of structural evolution in 2D MXenes using theory-enhanced machine learning.

Area of Science:

• Materials Science
• Data Science
• Spectroscopy

Background:

• In situ Electron Energy Loss Spectroscopy (EELS) and Transmission Electron Microscopy (TEM) are crucial for analyzing material structure and composition.
• Current EELS and TEM capabilities allow monitoring ultrafast transient changes, requiring advanced analytical methods.
• Understanding atomic-scale transformations in 2D MXenes is vital for their electronic and optical properties.

Purpose of the Study:

• To develop a machine learning (ML) framework for real-time assessment and characterization of in operando EELS Spectrum Images (EELS-SI).
• To enable precise control over atomic-scale structural transformations in 2D MXenes.
• To facilitate automated, on-the-fly synthesis and characterization for materials discovery.

Main Methods:

• Developed a novel ML framework for real-time EELS-SI analysis.
• Utilized Variational Autoencoders (VAE) to integrate computational and experimental MXene datasets into a unified latent space.
• Employed a unique training method requiring fewer labeled data points than traditional deep learning.

Main Results:

• The ML framework accurately predicts structural evolutions of 2D MXenes.
• Achieved prediction latencies suitable for closed-loop processing within TEM.
• Demonstrated a method that requires significantly less labeled training data.

Conclusions:

• The developed ML framework advances real-time characterization of in situ EELS data.
• This approach enhances capabilities for materials discovery and precision engineering of functional materials.
• Enables automated, on-the-fly synthesis and characterization at the atomic scale.