The Potential for Bayesian Compressive Sensing to Significantly Reduce Electron Dose in High Resolution STEM Images
The use of high resolution imaging methods in the scanning transmission electron microscope (STEM) is limited in many cases by the sensitivity of the sample to the beam and the onset of electron beam damage (for example in the study of organic systems, in tomography and during in-situ experiments). To demonstrate that alternative strategies for image acquisition can help alleviate this beam damage issue, here we apply compressive sensing via Bayesian dictionary learning to high resolution STEM images. These experiments successively reduce the number of pixels in the image (thereby reducing the overall dose while maintaining the high resolution information) and show promising results for reconstructing images from this reduced set of randomly collected measurements. We show that this approach is valid for both atomic resolution images and nanometer resolution studies, such as those that might be used in tomography datasets, by applying the method to images of strontium titanate and zeolites. As STEM images are acquired pixel by pixel while the beam is scanned over the surface of the sample, these post acquisition manipulations of the images can, in principle, be directly implemented as a low-dose acquisition method with no change in the electron optics or alignment of the microscope itself.
- Research Organization:
- Pacific Northwest National Lab. (PNNL), Richland, WA (United States)
- Sponsoring Organization:
- USDOE
- DOE Contract Number:
- AC05-76RL01830
- OSTI ID:
- 1129350
- Report Number(s):
- PNNL-SA-95635
- Journal Information:
- Journal of Electron Microscopy, 63(1):41 - 51, Journal Name: Journal of Electron Microscopy, 63(1):41 - 51
- Country of Publication:
- United States
- Language:
- English
Similar Records
$\mathrm{SIM}$-$\mathrm{STEM}$ $\mathrm{Lab}$: Incorporating Compressed Sensing Theory for Fast STEM Simulation
Compressive Classification for TEM-EELS