New Technologies for Ultra-low Dose-rate Imaging in the STEM

Abstract

Electron microscopy arose from the need to image materials beyond the resolution optical microscopes could achieve. Though taking many years and technological advancements, the modern scanning transmission electron microscope (STEM) can readily image a diverse range of materials at atomic resolution. However, using such intense electron beams causes sample-damage to such an extent that it often becomes the limiting factor, instead of the microscope?s performance. Therefore, new low-dose imaging techniques are required. Imaging with a low beam current and a short pixel dwell time is identified as a universally accessible approach to low-dose imaging. However, images captured under these conditions are often excessively noisy due to signal streaking, caused by the ?streaking? of signal from one pixel into subsequent pixels due to finite detector response times. When imaging at short dwell times this becomes unavoidable, and a solution is needed. New hardware which digitises the signal from the detector is developed, recording all electrons with equal intensity and localising them to a single time value, eliminating signal streaking. As only electrons are detected as signal, Gaussian noise and detector afterglow are also eliminated. Image comparisons of a biological tissue are shown, demonstrating how the technique produces low-dose, high signal-to-noise ratio images of fragile specimens. A lamella is imaged to show the absence of detector afterglow and a cluster of five silicon atoms on graphene is imaged at 31 f.p.s., demonstrating both high temporal and spatial resolution in the same dataset. As all electrons are now recorded with equal intensity, detector inhomogeneity is reduced. Eight detector maps are analysed, with their flatness, roundness, and smoothness values improving by 6.78 %, 9.97 %, and 32.06 % respectively after digitisation. The response time of these same detectors range from 200 ns to over 1.5 ?s, but have an instant response in the digital signal. Signal streaking is added to image simulations to isolate its effects, and is used to show the loss of information in Fourier transform of images with signal streaking, allowing detector performance to be evaluated virtually. The line flyback time occupies an increasingly larger portion of the imaging time with reducing dwell time, lowering scanning efficiency. While the line flyback time can be reduced, this results in a compression artefact due to hysteresis in the scanning coils. A semi-empirical model of this compression is created, allowing it to be corrected at any imaging settings. This increases scanning efficiency by 20 %, with the route to increase this a further 25 % identified. The findings of this thesis are accessible, retrofittable, and sustainable ways to increase microscope performance, whether to allow low-dose imaging, or extend the lifespan of existing microscopes.