Application of a novel local and automatic PCA algorithm for diffraction pattern denoising in TEM-ASTAR analysis in...

Area of Science:

• Materials Science
• Crystallography
• Data Analysis

Background:

• Transmission Electron Microscopy - Atom Probe Tomography (TEM-APT) generates 4D diffraction pattern (DP) datasets.
• Noise in DP datasets complicates accurate crystallographic analysis and quantitative grain characterization.
• Existing denoising methods may lack efficiency or adaptability to localized crystallographic signals.

Purpose of the Study:

• Introduce a novel denoising method, LAT-PCA (Local Automatic Thresholding - Principal Component Analysis), for TEM-APT DP datasets.
• Enhance the efficiency and accuracy of crystallographic analysis by reducing noise in DP data.
• Reduce data acquisition and post-processing times for TEM-APT analysis.

Main Methods:

• Implemented Principal Component Analysis (PCA) on localized windows of the 4D DP dataset.
• Utilized a Marchenko-Pastur Distribution to automatically threshold noise in principal components.
• Focused on higher-order principal components containing the physical crystallographic signal.

Main Results:

• LAT-PCA effectively reduces noise while preserving essential DP features.
• The method demonstrates significant reductions in acquisition and post-processing times.
• Denoised data facilitates more straightforward and accurate phase mapping and grain orientation determination.
• Experiments on a silicon-germanium-carbon sample validated the method's reliability and improvements over lower signal-to-noise data.

Conclusions:

• LAT-PCA is an effective and automatic solution for denoising TEM-APT DP datasets.
• The localized processing and automatic thresholding enhance computational efficiency and adaptability to varying noise levels.
• This method improves dataset quality, reduces analysis time, and minimizes artifacts, leading to more accurate material characterization.