Artist’s representation of the autonomous dark-field scanning microscopy experiment at the Advanced Photon Source (APS). Credit: Argonne National Laboratory
Researchers at the U.S. Department of Energy's (DOE) Argonne National Laboratory have recently developed a novel autonomous microscopy technique utilizing the power of AI. Unlike traditional raster scans, the novel method identifies and hones in on clusters of intriguing features.
Traditional point-by-point raster scan techniques are time-consuming as they methodically scan every inch of a sample. The method developed by the researchers at the DOE Argonne Laboratory instead uses artificial intelligence to identify and hone in on interesting features and disregards uniform sections, allowing future researchers to speed up the experimental process.
"Many regions of a sample can be safely disregarded or at least not sampled heavily, but regions where there are discontinuities and boundaries can instead contribute the vast majority of information about the sample," said Charudatta (C.D.) Phatak, a materials scientist at Argonne.
The AI algorithm used, which is published in Nature Communications, begins by selecting a random set of points on the sample. It then gathers data from these points at the same time it predicts subsequent points of interest. This on-the-fly prediction functionality drastically reduces data acquisition time and eliminates the need for human intervention. "Taking the human component out of the prediction process saves a great deal of time and really speeds up the experiment," said Saugat Kandel, a postdoctoral researcher at Argonne. "There's also only a small number of scientists who can perform these experiments effectively as they are done now."
The technique presented has broad applications in a variety of microscopic analysis fields from x-ray microscopy to electron microscopy. Its implementation will offer unparalleled speeds and create new avenues for exploration. "The ability to automate experiments with AI will significantly accelerate scientific progress in the coming years," said Argonne group leader and computational scientist Mathew Cherukara. "This is a demonstration of our ability to do autonomous research with a very complex instrument."