High-speed 4-dimensional scanning transmission electron microscopy using compressive sensing techniques.

Area of Science:

• Materials Science
• Physics
• Chemistry

Background:

• Four-dimensional scanning transmission electron microscopy (4-D STEM) is crucial for nanoscale characterization, acquiring 2D diffraction patterns at each 2D probe location, resulting in a 4D dataset.
• Conventional 4-D STEM is limited by slow acquisition times (up to 30s for 256^2 locations), leading to issues like drift, beam damage, and sample contamination.
• Fast direct electron detectors exist, but acquisition speed remains a bottleneck, hindering the technique's full potential.

Purpose of the Study:

• To demonstrate the application of compressive sensing for high-speed and reduced electron fluence 4-D STEM data acquisition.
• To investigate the impact of detector downsampling on precision and acquisition speed.
• To validate the method using atomic-resolution yttrium silicide datasets and advanced analysis techniques.

Main Methods:

• Acquired a random subset of probe locations instead of a regular grid for 4-D STEM data collection.
• Simulated compressive sensing acquisition for 4-D STEM, analyzing techniques like ptychography and differential phase contrast.
• Evaluated detector downsampling, confirming inherent oversampling in CBED patterns and its benefit for faster acquisition without precision loss.

Main Results:

• Compressive sensing reduced acquisition times by 100-300 times compared to conventional methods.
• Achieved over 25 dB peak signal-to-noise ratio in recovered phase using only 0.3% of the total data.
• Detector downsampling allowed faster data acquisition without compromising precision.

Conclusions:

• Compressive sensing is a viable method for significantly accelerating 4-D STEM data acquisition.
• The technique effectively reduces electron dose, mitigating beam damage and improving data quality.
• This approach enhances existing frame rates and opens new possibilities for nanoscale imaging and analysis.