Time and place
26 April 2024 13:00-16:00, building 303A, aud. 43, DTU Lyngby Campus
Principal supervisor
Professor Thomas Willum Hansen
Co-supervisor
Professor Jakob Schiøtz, DTU Fysik
Examiners
Shima Kadkhodazadeh, Senior Researcher, DTU Nanolab
Marc-Georg Willinger, Professor, TU Munich
Nestor Zaluzec, Senior Scientist, Argonne National Laboratory
Abstract
High resolution transmission electron microscopy (HRTEM) is an important tool for atomic scale structural characterization of materials such as nanoparticles for catalytic applications.
A major drawback is the necessity of an intense high-energy electron beam, which can be disruptive to the specimen. Hence, novel analysis methods with high accuracy and efficiency for analyzing images with low signal-to-noise ratio are in high demand.
HRTEM image series recorded at different electron intensities (dose rate) are investigated. Novel quantities of signal-to-noise ratio (SNR) models are suggested and tested on the image series. Another aspect of image recording which is investigated is the influence of the electron detector and how this influences the SNR models. In order to track dynamic events over time and dose rate, a novel approach of structural similarity index measurement (SSIM) is tested. Each frame is compared to its adjacent frame and the difference is quantified. This is tested on the HRTEM image series.
Results show the novel SNR depends on dose rate and electron camera type. At some regimes, the novel SNR models grow rapidly with increasing dose rate. Data has determined a dose rate where the direct electron camera provides higher SNR than a conventional scintillator-based camera and vice versa. The novel SSIM approach can potentially track dynamic events at high dose rates while it becomes challenging at lower electron dose rates.
These approaches provide novel methods for quantifying the interpretability of HRTEM images and detect subtle atomic-scale events. Furthermore, they can facilitate threshold determinations for efficient automated analysis via artificial intelligence.
