Publicação
Efficient computational methods to index crystallographic (S)TEM images and ED patterns
| Resumo: | Electron Microscopy (EM) of nanomaterials relies on grey-scale images to display the material’s atomic arrangement, and a high resolution EM can simultaneously capture multiple atomic structures into a single image. However, the extraction of useful information from these images is still limited to the determination of the material’s orientations, an underutilisation of the powerful features of an EM equipment and less productive EM sessions. This is due to the compute-intense tasks that have not been automated yet. This dissertation aims to significantly reduce the time required to extract useful data from EM images and to remove the user bias when analysing high resolution (S)TEM images, by automating most user routine tasks and integrating them into a software tool, Im2Cr. The deployed Im2Cr tool aimed to aid an EM user to find the most probable atomic structure orientation of a nanomaterial in a single 2D image from a set of pre-defined materials, with a minimal user interaction. Im2Cr was designed and built with a simple and intuitive Graphical User Interface (GUI) that runs on a common modern laptop. It takes as input a high resolution (S)TEM image and multiple CIF files with candidate atomic structures to describe the material under observation. After performing the Fourier Transform (FT) on selected Regions Of Interest (ROI) in the image, the tool automatically detects periodic information related to the atom’s positions by the brighter spots on the image FT. With a set of geometric computations it tries to match the theoretical values computed with the measured ones by assigning a custom made merit index. This quantitative evaluation avoids possible user bias and/or errors on image characterisation. Im2Cr outputs at the end a report with the best matching crystallographic structure, its orientation and the indexation table. This tool was successfully tested for robustness and execution efficiency in a wide range of high resolution (S)TEM images from crystalline nanomaterials, with domain size ranging from 4 to 100 nm. The autonomous indexation with preset parameters has a very high success rate and runs in a small fraction of typical (S)TEM images acquisition time by taking advantage of the inherent hardware parallelism. Alternatively, the user can change some relevant parameters related to the ROI selection on the (S)TEM image and on the FT peaks detection. Im2Cr promising results point to the possibility of real-time image analysis with reduced user interaction, allowing for an increased (S)TEM characterisation yield and also enabling the interpretation of complex images, such as those from nanocrystalline materials imaged in high-order zone axis orientations. |
|---|---|
| Autores principais: | Silva, André Sá |
| Assunto: | Engenharia e Tecnologia::Engenharia Eletrotécnica, Eletrónica e Informática |
| Ano: | 2018 |
| País: | Portugal |
| Tipo de documento: | dissertação de mestrado |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Universidade do Minho |
| Idioma: | inglês |
| Origem: | RepositóriUM - Universidade do Minho |
| Resumo: | Electron Microscopy (EM) of nanomaterials relies on grey-scale images to display the material’s atomic arrangement, and a high resolution EM can simultaneously capture multiple atomic structures into a single image. However, the extraction of useful information from these images is still limited to the determination of the material’s orientations, an underutilisation of the powerful features of an EM equipment and less productive EM sessions. This is due to the compute-intense tasks that have not been automated yet. This dissertation aims to significantly reduce the time required to extract useful data from EM images and to remove the user bias when analysing high resolution (S)TEM images, by automating most user routine tasks and integrating them into a software tool, Im2Cr. The deployed Im2Cr tool aimed to aid an EM user to find the most probable atomic structure orientation of a nanomaterial in a single 2D image from a set of pre-defined materials, with a minimal user interaction. Im2Cr was designed and built with a simple and intuitive Graphical User Interface (GUI) that runs on a common modern laptop. It takes as input a high resolution (S)TEM image and multiple CIF files with candidate atomic structures to describe the material under observation. After performing the Fourier Transform (FT) on selected Regions Of Interest (ROI) in the image, the tool automatically detects periodic information related to the atom’s positions by the brighter spots on the image FT. With a set of geometric computations it tries to match the theoretical values computed with the measured ones by assigning a custom made merit index. This quantitative evaluation avoids possible user bias and/or errors on image characterisation. Im2Cr outputs at the end a report with the best matching crystallographic structure, its orientation and the indexation table. This tool was successfully tested for robustness and execution efficiency in a wide range of high resolution (S)TEM images from crystalline nanomaterials, with domain size ranging from 4 to 100 nm. The autonomous indexation with preset parameters has a very high success rate and runs in a small fraction of typical (S)TEM images acquisition time by taking advantage of the inherent hardware parallelism. Alternatively, the user can change some relevant parameters related to the ROI selection on the (S)TEM image and on the FT peaks detection. Im2Cr promising results point to the possibility of real-time image analysis with reduced user interaction, allowing for an increased (S)TEM characterisation yield and also enabling the interpretation of complex images, such as those from nanocrystalline materials imaged in high-order zone axis orientations. |
|---|