Imaging 3D Chemistry and Structure at Nanometer Resolution

Abstract

Microscopists relentlessly strive to observe structure at the highest possible resolution. Over the past century, there has been a transition away from photonic wave sources, including light and X-rays, towards harnessing the potentials of electron optics. Whereas optical microscopes encountered their inherent far-field diffraction constraints long ago, thereby limiting characterization to the micron or nanoscale, electron microscopes can now routinely image materials by using accelerated electrons.

The pursuit of achieving high-resolution micrographs has significantly advanced through the development of aberration correctors. These devices utilize multi-pole magnetic lenses to rectify imperfections in the electron probe. As a result, electron probes can now reach sub-angstrom length scales, improving spatial resolution. This relentless pursuit of improving resolution played a vital role in expanding our understanding of materials and biological structures at the atomic and nanoscale levels.

Despite the remarkable progress in corrector hardware, there remains a increasingly need for advanced algorithms capable of efficiently extract valuable information from sparse low-dose acquisitions. The challenges are particularly pronounced in both atomic-scale measurements and 3D imaging, where substantial radiation dosages may exceed the dose tolerance of materials. This necessity assumes an even greater significance for chemically sensitive measurements, wherein the likelihood of capturing inelastic scattering events markedly diminishes.

To mitigate beam damage, reducing electron exposure is a common strategy but can lead to a decrease in signal-to-noise ratio (SNR) -- hindering the analysis process. The extent radiative damage is particularly problematic for tomography experiments, where designing experimental configurations suitable for a material's tolerance is crucial for successfully obtaining accurate 3D measurements.

Over the past few decades, efforts to enhance image SNR have occurred simultaneously with developments in detector sensitivity and algorithms that faithfully reveals hidden structure lost in noisy signals. By using the concept of `compressive sensing,' it is possible to reduce dose through a computational procedure that assumes signals should be sparse in some domain. More importantly, compressive sensing provides theoretical guarantees for recovering signals. This thesis extends this framework into the field of electron microscopy and uncovers several applications that benefit.

The dissertation introduces image processing algorithms that enables such sparse chemical and 3D structural imaging of materials. These algorithms are designed to maintain maximal quality with minimal radiative dosages. The development of several computational tools designed for the visualization of high-resolution of material's 3D structure and chemistry are presented in this thesis; these tools are applied to several physical systems including catalytic, semi-organic and oxide systems from projections micrographs collected by scanning transmission electron microscopes. Most of the methods presented in this thesis have already been published in several journals include textit{Nature Communications, npj Computational Materials} and textit{Ultramicroscopy}. Overall, this research aims to push the boundaries of electron microscopy and unlock new insights into the 3D structure and chemistry for a wider class of materials.