Data challenges of in situ X-ray tomography for materials discovery and characterization

Since its development in the 1970s (Hounsfield, Br J Radiol 46(552):1016–1022, 1973) [1], X-ray tomography has been used to study the three dimensional (3D) structure of nearly every type of material of interest to science, both in the laboratory (Elliott and Dover, J Microsc 126(2):211–213, 1982) [2] and at synchrotron facilities (Thompson et al., Nucl Instrum Methods Phys Res 222(1):319–323, 1984) [3]. The ability to nondestructively image internal structures is useful in the medical community for patient diagnosis. For this same reason, it is critical for understanding material structural morphology. X-ray tomography of static materials can generate a true 3D structure to map out content and distribution within materials including voids, cracks, inclusions, microstructure, and interfacial quality. This technology is even more useful when applying a time component and studying the changes in materials as they are subjected to non-equilibrium stimulations. For example, testing mechanical properties (e.g., compressive or tensile loading), thermal properties (e.g., melting or solidification), corrosion, or electrostatic responses, while simultaneously imaging the material in situ, can replicate real world conditions leading to an increase in the fundamental understanding of how materials react to these stimuli. Mechanical buckling in foams, migration of cracks in composite materials, progression of a solidification front during metal solidification, and the formation of sub-surface corrosion pits are just a few of the many applications of this technology. This chapter will outline the challenges of taking a series of radiographs while simultaneously stressing a material, and processing it to answer questions about material properties. The path is complex, highly user interactive, and the resulting quality of the processing at each step can greatly affect the accuracy and usefulness of the derived information. Understanding the current state-of-the-art is critical to informing the audience of what capabilities are available for materials studies, what the challenges are in processing these large data sets, and which developments can guide future experiments. For example, one particular challenge in this type of measurement is the need for a carefully designed experiment so that the requirements of 3D imaging are also met. Additionally, the rapid collection of many terabytes of data in just a few days leads to the required development of automated reconstruction, filtering, segmentation, visualization, and animation techniques. Finally, taking these qualitative images and acquiring quantitative metrics (e.g., morphological statistics), converting the high quality 3D images to meshes suitable for modeling, and coordinating the images to secondary measures (e.g., temperature, force response) has proven to be a significant challenge when a materials scientist ‘simply’ needs an understanding of how material processing affects its response to stimuli. This chapter will outline the types of in situ experiments and the large data challenges in extracting materials properties information.