Machine Learning Speeds Up Chemical Imaging Using Randomized X-rays
-
Researchers used sandpaper to scramble X-ray beams going through a sample, creating randomized illumination patterns.
-
A machine learning model was trained on the scrambled images and fluorescence signals to map chemical compositions.
-
This technique speeds up traditional raster-scanning methods for resolving chemical imaging.
-
The method was demonstrated on a cathode battery sample, successfully distinguishing different lithium-based particles.
-
The approach could be extended beyond synchrotrons to more accessible lab X-ray sources.