A soft touch with electron beams: Digging out structural information of nanomaterials with advanced scanning low energy electron microscopy coupled with deep learning

• Published in: Ultramicroscopy 2024-08, Vol.262, p.113965-113965, Article 113965
• Main Authors: Materna Mikmeková, Eliška, Materna, Jiří, Konvalina, Ivo, Mikmeková, Šárka, Müllerová, Ilona, Asefa, Tewodros
• Format: Article
• Language: English
• Subjects: Deep learning; Mesoporous silica; Scanning low energy electron microscopy
• ISSN: 0304-3991; 1879-2723
• DOI: 10.1016/j.ultramic.2024.113965

•Scanning low energy electron microscopy (SLEEM) eliminates local charging effects.•Denoised image via deep learning acquires clearer structural features.•Deep learning as a great tool to advance our understanding about the properties and their structure-property relationships of various nanomaterials.•Approach can be applied for the observation of nanoporous structures near the surface. Nanostructured materials continue to find applications in various electronic and sensing devices, chromatography, separations, drug delivery, renewable energy, and catalysis. While major advancements on the synthesis and characterization of these materials have already been made, getting information about their structures at sub-nanometer resolution remains challenging. It is also unfortunate to find that many emerging or already available powerful analytical methods take time to be fully adopted for characterization of various nanomaterials. The scanning low energy electron microscopy (SLEEM) is a good example to this. In this report, we show how clearer structural and surface information at nanoscale can be obtained by SLEEM, coupled with deep learning. The method is demonstrated using Au nanoparticles-loaded mesoporous silica as a model system. Moreover, unlike conventional scanning electron microscopy (SEM), SLEEM does not require the samples to be coated with conductive films for analysis; thus, not only it is convenient to use but it also does not give artifacts. The results further reveal that SLEEM and deep learning can serve as great tools to analyze materials at nanoscale well. The biggest advantage of the presented method is its availability, as most modern SEMs are able to operate at low energies and deep learning methods are already being widely used in many fields. [Display omitted]