U.S. Department of Energy

Pacific Northwest National Laboratory

Stevens presents at 2015 Materials Research Society Meeting

Abstract

Transmission electron microscopy (TEM) and/or Scanning TEM (STEM) is a widely used experimental method to study biological structures and materials science structures, interfaces and defects under both static and operando conditions. Images in TEM are acquired in a projection mode where the scattering from the sample is collected on a pixelated detector. In the STEM mode of acquisition, the images are acquired by integrating the total signal in the detector that originates from the beam as it is rastered over the specimen. In many cases the resolution in the image is finally determined by the ability of the sample to withstand the electron dose. Another concern, of equal importance, is acquisition speed, which is related to dose, but is limited by data transfer speeds. One approach that has been developed to help with these experimental issues is compressive sensing (CS). In CS the idea is that a signal can be acquired without the constraints of the Nyquist-Shannon sampling rule—we can achieve high spatial resolution with a much lower sampling rate. The use of compressive sensing for (S)TEM data has already shown in simulation that without any prior information, the amount of data needed to achieve atomic spatial resolution images reduced by as much as 93%. In addition to allowing experiments to be performed much more quickly, the specimen will enjoy a much lower electron dose. This will allow experiments to be performed on samples that cannot currently be studied in the TEM. We apply Bayesian models of sparsity-based CS and manifold-based CS to image and video acquisition in (S)TEM. For STEM we are able to reduce the electron dose by a factor of 20 (using compressively acquired data), by using a technique called inpainting. Inpainting allows missing pixels to be inferred from a small subset of acquired pixels. In TEM video we have simulated an increase in framerate by a factor of 20. The technique for TEM uses a random (binary) coded aperture to encode several subframes that are subsequently integrated into a single frame on the camera. The CS inversion process recovers the subframes and performs inpainting on the missing pixels.

Stevens AJ, L Kovarik, QM Ramasse, L Carin, and ND Browning.  2015.  "Bayesian Machine Learning for Compressive (S)TEM Image/Video Acquisition."  Abstract submitted to 2015 Materials Research Society Fall Meeting, Boston, MA.  PNNL-SA-110947. 

| Pacific Northwest National Laboratory