February_EDFA_Digital

edfas.org 1 1 ELECTRONIC DEV ICE FA I LURE ANALYSIS | VOLUME 24 NO . 1 OPTIMAL SAMPLING AND RECONSTRUCTION STRATEGIES FOR SCANNING MICROSCOPES Nigel D. Browning 1,2,3 , Daniel Nicholls 1 , Jack Wells 4 , Alex W. Robinson 1 1 Liverpool University, U.K. 2 Pacific Northwest National Laboratory, Richland, Wash. 3 Sivananthan Laboratories, Bolingbrook, Ill. 4 University of Liverpool, U.K. nigel.browning@liverpool.ac.uk EDFAAO (2022) 1:11-16 1537-0755/$19.00 ©ASM International ® INTRODUCTION The advent of aberration-corrected scanning transmis- sion electron microscopy (STEM) [1,2] has led to an unprec- edented increase in the achievable spatial resolution from all forms of imaging and spectroscopy (Z-contrast, annular bright field, electron energy-loss spectroscopy, energy-dispersive spectroscopy, etc.), but this has also been accompanied by a simultaneous increase in the operational probe current under typical imaging condi- tions. While the increased current is advantageous for observations of atomic scale dopants in some samples, typical electron doses are now several orders of mag- nitude higher than many materials can withstand. [3] Dose considerations have now become the most critical experimental parameters when imaging beam sensitive materials, which usually leads to a practical reduction in the electron dose and dose rate being implemented at the cost of decreased signal-to-noise ratios and a poorer spatial resolution than themicroscope is capable of deliv- ering at the higher dose/rate levels. This article examines the main issues associated with minimizing beam dose in the STEM (and as the methodology is essentially the same in all scanned methods, this applies to any other scanned imaging system) and proposes the use of a sub- sampling and inpainting methodology (generally falling within the mathematical field of compressive sensing) as a method to overcome the effects of the beam, leading to an improvement in the resolution and reproducibility of high resolution analyses. [4-8] PRACTICAL SCANNING SYSTEMS In a standard STEM, the way the scan system usually works is that it moves the beam from left to right across a single row with a dwell time (typically ~5 µs) for each pixel in that row (Fig. 1). At the end of the row, the beam flies back to the left-hand-side, moves down one pixel and then completes a row again (this is like the way a Fig. 1 Examples of various scanning patterns in a 9 x 9 grid. Number and color indicate scanning order. (a) Raster scanning is the traditional method of scanning in STEM. (b) Space filling random scanning has been shown to reduce beam damage in beam sensitive samples. [8] (c,d) Two scanning patterns possible using probe sub-sampling; sub-sampled random scanning and sub-sampled line-hop scanning at 33.3% sampling ratio. (a) (b) (c) (d)