Deep learning for improving non-destructive grain mapping in 3D.

Laboratory X-ray diffraction contrast tomography (LabDCT) is a novel imaging technique for non-destructive 3D characterization of grain structures. An accurate grain reconstruction critically relies on precise segmentation of diffraction spots in the LabDCT images. The conventional method utilizing various filters generally satisfies segmentation of sharp spots in the images, thereby serving as a standard routine, but it also very often leads to over or under segmentation of spots, especially those with low signal-to-noise ratios and/or small sizes. The standard routine also requires a fine tuning of the filtering parameters. To overcome these challenges, a deep learning neural network is presented to efficiently and accurately clean the background noise, thereby easing the spot segmentation. The deep learning network is first trained with input images, synthesized using a forward simulation model for LabDCT in combination with a generic approach to extract features of experimental backgrounds. Then, the network is applied to remove the background noise from experimental images measured under different geometrical conditions for different samples. Comparisons of both processed images and grain reconstructions show that the deep learning method outperforms the standard routine, demonstrating significantly better grain mapping.

Improved grain mapping by laboratory X-ray diffraction contrast tomography.

Laboratory diffraction contrast tomography (LabDCT) is a novel technique for non-destructive imaging of the grain structure within polycrystalline samples. To further broaden the use of this technique to a wider range of materials, both the spatial resolution and detection limit achieved in the commonly used Laue focusing geometry have to be improved. In this work, the possibility of improving both grain indexing and shape reconstruction was investigated by increasing the sample-to-detector distance to facilitate geometrical magnification of diffraction spots in the LabDCT projections. LabDCT grain reconstructions of a fully recrystallized iron sample, obtained in the conventional Laue focusing geometry and in a magnified geometry, are compared to one characterized by synchrotron X-ray diffraction contrast tomography, with the latter serving as the ground truth. It is shown that grain indexing can be significantly improved in the magnified geometry. It is also found that the magnified geometry improves the spatial resolution and the accuracy of the reconstructed grain shapes. The improvement is shown to be more evident for grains smaller than 40 µm than for larger grains. The underlying reasons are clarified by comparing spot features for different LabDCT datasets using a forward simulation tool.

A flexible and standalone forward simulation model for laboratory X-ray diffraction contrast tomography.

Laboratory X-ray diffraction contrast tomography (LabDCT) has recently been developed as a powerful technique for non-destructive mapping of grain microstructures in bulk materials. As the grain reconstruction relies on segmentation of diffraction spots, it is essential to understand the physics of the diffraction process and resolve all the spot features in detail. To this aim, a flexible and standalone forward simulation model has been developed to compute the diffraction projections from polycrystalline samples with any crystal structure. The accuracy of the forward simulation model is demonstrated by good agreements in grain orientations, boundary positions and shapes between a virtual input structure and that reconstructed based on the forward simulated diffraction projections of the input structure. Further experimental verification is made by comparisons of diffraction spots between simulations and experiments for a partially recrystallized Al sample, where a satisfactory agreement is found for the spot positions, sizes and intensities. Finally, applications of this model to analyze specific spot features are presented.

Boundary migration in a 3D deformed microstructure inside an opaque sample.

How boundaries surrounding recrystallization grains migrate through the 3D network of dislocation boundaries in deformed crystalline materials is unknown and critical for the resulting recrystallized crystalline materials. Using X-ray Laue diffraction microscopy, we show for the first time the migration pattern of a typical recrystallization boundary through a well-characterized deformation matrix. The data provide a unique possibility to investigate effects of both boundary misorientation and plane normal on the migration, information which cannot be accessed with any other techniques. The results show that neither of these two parameters can explain the observed migration behavior. Instead we suggest that the subdivision of the deformed microstructure ahead of the boundary plays the dominant role. The present experimental observations challenge the assumptions of existing recrystallization theories, and set the stage for determination of mobilities of recrystallization boundaries.

A method to characterize the roughness of 2-D line features: recrystallization boundaries.

A method is presented, which allows quantification of the roughness of nonplanar boundaries of objects for which the neutral plane is not known. The method provides quantitative descriptions of both the local and global characteristics. How the method can be used to estimate the sizes of rough features and local curvatures is also presented. The potential of the method is illustrated by quantification of the roughness of two recrystallization boundaries in a pure Al specimen characterized by scanning electron microscopy.

Applications of high-energy synchrotron radiation for structural studies of polycrystalline materials.

The large penetration power of high-energy X-rays (>60 keV) raises interesting prospects for new types of structural characterizations of polycrystalline materials. It becomes possible in a non-destructive manner to perform local studies, within the bulk of the material, of the fundamental materials physics properties: grain orientations, strain, dislocation densities etc. In favourable cases these properties may be mapped in three dimensions with a spatial resolution that matches the dimensions of the individual grains. Imbedded volumes and interfaces become accessible. Moreover, the high energies allow better in-situ studies of samples in complicated environments (industrial process optimization). General techniques for research in this energy range have been developed using broad-band angle-dispersive methods, on-line two-dimensional detectors and conical slits. Characterizations have been made at the level of the individual grains and grain boundaries as well as on ensembles of grains. The spatial resolution is presently of the order of 10-100 micom. Four examples of applications are presented along with an outlook.