Optimal statistical experiment design for detecting and locating light atoms using quantitative high resolution (scanning) transmission electron microscopy
13 February 2017
Campus Groenenborger, U0.24 - Groenenborgerlaan 171 - 2020 Antwerpen (route: UAntwerpen, Campus Groenenborger
Organization / co-organization:
Department of Physics
Sandra Van Aert & Jan Sijbers
Public defence of the PhD thesis of Mrs. Julie Gonnissen - Faculty of Science - Department of Physics
In the last decade, visualising light atoms like lithium and hydrogen has gained serious interest, as they play a key-role in many industrial applications, such as lithium batteries or energy storage materials. Therefore, the optimisation of different techniques in Transmission Electron Microscopy (TEM) for such applications has become an important issue. However, it is extremely difficult to visualise materials containing light elements and quantify their structure and chemical composition at the atomic-scale, since the interaction with the electron beam weakens for lighter atoms.
The main goal of this PhD research was to optimise the experiment design of the electron microscope in order to characterise nanostructures containing light atoms, using advanced and new techniques in quantitative TEM. The purpose is then to retrieve the atomic structure of light atom nanocrystals from experimental images. A direct qualitative interpretation of experimental images gives unreliable results, if the difference in atomic number Z of neighbouring atom columns is small or when the signal-to-noise ratio (SNR) of the images is low. Thus, quantitative methods are necessary in order to characterise the chemical composition of crystals containing light elements. Statistical parameter estimation theory in combination with detection theory is therefore used, where in this thesis the parameters that have to be estimated are the atom types present (or absent) in the structure, the position coordinates of the projected atom columns, and the number of atoms in the projected atom columns (i.e. the column thickness).
The goal is then to find the optimal experiment design that minimises the probability of error or maximises the attainable precision for the estimated parameters. Both conventional TEM and STEM are investigated and compared for detecting and locating light atoms, and for counting the number of atoms in a projected atom column. As a practical application in the research to quantitatively characterise light atom structures, the local oxygen-octahedral coupling at perovskite heterostructural interfaces in different epitaxial thin films is determined. Furthermore, the domain wall in a LiNbO3 crystal is quantified, and the atomic shift of the domains next to the domain wall is determined as well as the width of the transition region between both domains, using statistical parameter estimation theory.