Computational Electron Microscopy for the Characterization of Soft, Hard, and Hybrid Materials


Connecting structure and function in nanoscale engineered materials and devices relies on the analysis of the fundamental arrangement of matter, frequently under dynamic conditions. The demand to image structures at fundamental length scales has touched inorganic materials, biology, and frequently hybrid hard/soft materials with unique phenomena driven by heterogeneous components. Electron microscopy has been at the core of imaging and interrogating structures at this scale for many decades. Improvements in hardware and digital control have pushed the boundaries on the information measurable from modern systems. Operating as an ‘electron scattering’ experiment the ability to push copious amounts of current into small volumes has exacerbated the sensitivity of many materials to radiative damage. Thus, developing new autonomous methods of imaging which rely on advanced computation or machine learning is playing a larger role in the reliable and high-speed imaging of next-generation materials. Such methods potentially allow an investigator to image materials dynamically in ways not previously possible, partly due to the reduction in area doses and partly due to drastic time-savings. In Chapter 1, the major advances in electron microscopy over the last few years are highlighted, as well as outstanding challenges in the reliable characterization of dose sensitive materials. Novel methods based on a hybrid hardware-software approach are highlighted, as well as advances modeling the process of image formation for effective signal restoration. A high-level background to Compressive Sensing is offered, as well as practical discussions on the likelihood of successful implementation in electron microscopy. Expanding on this, Chapter 2 discusses the practical implementation of a coupled hardware/software approach to general imaging of radiation sensitive materials, particularly where multiple measurements will be made. Chapter 3 further advances this topic by placing control of the microscope directly in the hands of an autonomous machine learning platform, which efficiently plans out the application of dose within the column for maximum information retrieval. The work shifts gears slightly in Chapter 4 for a discussion of a novel crystalline orientation mapping tool enabled by advances in image processing, as well as its potential as a non-destructive method for understanding complex defect structures. Finally, Chapter 5 will discuss future directions and potential expansions on the preliminary work discussed in this thesis. The consistent theme of this work is the practical application of statistical learning and computation to enable new ways of interrogating structures, particularly with an eye towards reducing the time or integrated electron dose for key information. In many cases, it will be shown that considerably superior methods are available with simple innovations for almost no cost, and the future of electron microscopy will likely be at the interface of hardware and software, as data management and acquisition increasingly become core to the technique.

Date created
Resource type
Rights statement