AI-Assisted Real-Time Tracking of Subnanometer Strain Relaxation during Heteroepitaxy

Area of Science:

• Materials Science
• Surface Science
• Artificial Intelligence in Science

Background:

• Controlling interfacial structure and film quality in heteroepitaxy requires understanding strain relaxation.
• Direct real-time quantification of strain relaxation during initial growth stages is challenging.

Purpose of the Study:

• To develop and demonstrate an AI-assisted in situ framework, AutoRHEED, for tracking subnanometer strain relaxation.
• To enable direct, quantitative measurement of interfacial strain evolution during epitaxial growth.

Main Methods:

• AutoRHEED integrates image quality filtering, autonomous streak identification, and physics-guided calibration.
• Quantitative in-plane interplanar spacing (d spacing) evolution is extracted from reflection high-energy electron diffraction (RHEED) videos.
• The framework was applied to highly lattice-mismatched AlSb/Si(001) growth.

Main Results:

• AutoRHEED successfully captured the transient strain relaxation from initial states to full relaxation.
• The method demonstrated monolayer-level sensitivity in tracking strain evolution.
• In situ-derived relaxation trajectories showed excellent agreement with ex situ HAADF-STEM analysis.

Conclusions:

• AutoRHEED provides a reliable and temporally accurate method for in situ strain relaxation quantification.
• The framework enhances control over interfacial structure and film quality in epitaxial growth.
• This AI-assisted approach offers structural fidelity and temporal reliability for advanced materials characterization.