AI-NERD: Elucidation of relaxation dynamics beyond equilibrium through AI-informed X-ray photon correlation...

Area of Science:

• Physics and Materials Science
• Data Science and Machine Learning

Background:

• Probing in situ dynamics of functional materials across length and time scales presents a significant challenge.
• X-ray photon correlation spectroscopy (XPCS) is effective for characterizing materials dynamics but interpretation is complicated by spatial and temporal heterogeneity.

Purpose of the Study:

• To develop an unsupervised deep learning (DL) framework for automated classification of relaxation dynamics from experimental data.
• To demonstrate the framework's ability to accelerate exploration of large datasets and identify samples of interest.
• To correlate microscopic dynamics with macroscopic properties in a model system.

Main Methods:

• Development of an unsupervised deep learning (DL) framework.
• Automated classification of relaxation dynamics from X-ray photon correlation spectroscopy (XPCS) data.
• Application to a model system for correlating microscopic dynamics with macroscopic properties.

Main Results:

• Successful automated classification of relaxation dynamics from XPCS data without prior physical knowledge.
• Demonstrated acceleration of large dataset exploration for identifying samples of interest.
• Established direct correlation between microscopic dynamics and macroscopic properties.

Conclusions:

• The developed DL framework is material and process agnostic.
• This work represents a significant step towards autonomous materials discovery.
• The framework offers a powerful tool for interpreting complex materials dynamics.