News Article

A joint research team from the Technical University of Denmark, Lund University in Sweden and the European Synchrotron Radiation Facility (ESRF) has developed a non-destructive method for mapping the three-dimensional stress distribution inside plastically deformed metallic materials using x-rays. The sample studied is a plastically deformed polycrystalline aluminum alloy AA1050. The experiments were performed at the ESRF synchrotron beamline ID11 using 48.5 keV microbeam x-rays (3 μm × 3 μm) and the sample was mounted on a device integrating a tensile tester and a diffraction goniometer. Using the Debye–Scherrer camera principle, x-ray diffraction spots from the sample are recorded by a two-dimensional x-ray detector. The stress–strain curve of the AA1050 alloy is measured in situ and the relationship between deformation and fracture is studied under plastic deformation conditions, where a load of approximately 85 MPa is applied and the alloy is elongated by 32%, which is close to the ultimate elongation. Microbeam x-rays are scanned over the sample in the XY direction to obtain x-ray diffraction data for many different points on the sample. In addition, a rotational scan is performed to collect data at many different projection angles, even at the same point of x-ray exposure. Thus, using the same image reconstruction method as x-ray computed tomography, three-dimensional data on crystal structure and stress distribution can be obtained. Although similar experimental studies have been performed in the past, most of them have been limited to the study of the macroscopic stress distribution. As a result, in the case of large plastic deformation, x-ray diffraction patterns from different parts of the sample overlap, making it difficult to interpret the data correctly. One of the key points of this study was the use of microbeams, which allowed us to separate the stress distributions of different grains. For more information, see the article, A. Henningsson, M. Kutsal, J. P. Wright, W. Ludwig, H. O. Sørensen, S. A. Hall, G. Winther, and H. Poulsen, “Microstructure and Stress Mapping in 3D at Industrially Relevant Degrees of Plastic Deformation,” Scientific Reports 14 (2024): 20213, https://www.nature.com/articles/s41598-024-71006-0. XAFS imaging is one of the most popular methods in synchrotron radiation experiments. In particular, by focusing on chemical shifts at the absorption edge, it can visualize differences in chemical states, such as the valence of elements in a sample, and is often used to obtain useful information in chemistry and materials science. Currently, many synchrotron radiation facilities around the world are equipped with transmission x-ray microscopes and beamlines for nanobeam XAFS imaging. The transmission x-ray microscopes used in modern XAFS imaging are projection-type microscopes that acquire information over the entire field of view at once, rather than scanning-type microscopes that scan each point of the sample. While this method is very good, it does not always produce satisfactory images when analyzing real samples. A research group at the NSLS-II synchrotron radiation facility at Brookhaven National Laboratory in the United States has developed a method to overcome such problems and greatly improve the image quality of chemical state imaging by XAFS. At their beamline, a single exposure on the order of 0.01 s is used to take a snapshot of the absorption contrast image of a sample in the field of view, and the imaging is repeated while sweeping the incident x-ray energy near the absorption edge of an element. These data can be acquired rapidly with a spatial resolution of 20–30 nm over a field of view of up to several hundred microns. Tools for data analysis are publicly available. (See the article, M. Ge and W.-K. Lee, “PyXAS—An Open-Source Package for 2D x-ray Near-Edge Spectroscopy Analysis,” Journal of Synchrotron Radiation 27 (2020): 567–575, https://doi.org/10.1107/S1600577520001071). One of the practical problems with XAFS imaging is that the goal of high spatial resolution with a large number of pixels to obtain excellent quality images results in a significant reduction in signal-to-background per pixel. These problems were well-known before the advent of nanobeam and other advanced optical devices. They are particularly severe for relatively low concentrations of elements in the sample. In some cases, the samples that could be measured were limited, or compromises had to be made in the images and quantitative information that could be obtained. In this study, we focused on the statistical variation due to inhomogeneity of the beam profile and low counts as the main causes of image quality degradation. Since normalizing the data as it is under these conditions is likely to result in large errors, a new normalization method using deep learning was introduced. Since XAFS is a spectral change when the incident energy is varied, each pixel in an XAFS image also spans a series of images and should have XAFS information. Although each acquired image is separate data and can be degraded by statistical variations due to inhomogeneities in the beam profile and small counts, since pixels at a given coordinate across a series of images have fixed XAFS spectral information, each image data are actually not independent, but rather in a certain mutually constrained relationship. Therefore, the research group performed processing using deep neural networks, which are trained to take into account information about physical properties, such as the shape of the spectra for each element and chemical state. In this study, the LiNi x Mn y Co1−x−y O2 system, a model sample of battery materials, is used as a material and it is shown that the valence of Ni and Co can be evaluated with high confidence. For more information, see the article, Z. Li, T. Flynn, T. Liu, S. Liu, W.-K. Lee, M. Tang, and M. Ge, “Highly Sensitive 2D X-Ray Absorption Spectroscopy via Physics Informed Machine Learning,” npj Computational Materials 10 (2024): 128, https://www.nature.com/articles/s41524-024-01313-7. The source code for this data processing is included in the PyXAS public package. https://github.com/gmysage/pyxas. X-ray tomography is a technique that uses mathematical operations to create a three-dimensional image of the electron density distribution inside an object based on a large number of images taken by irradiating the object with x-rays. Fifty-three years ago, in 1971, Godfrey Hounsfield, an electrical engineer at EMI in England, performed the world's first 3D x-ray imaging of the brain at Atkinson Morley Hospital in London (for more details, see the article, J. Boone and C. McCollough, “Computed Tomography Turns 50,” Physics Today 74, no. 9 (2021): 34, https://physicstoday.scitation.org/doi/10.1063/PT.3.4834). The spatial resolution of early x-ray tomography was very limited by today's standards. The optical elements used to focus and image the x-rays, and the detectors used to image the x-rays were insufficient to achieve high spatial resolution. Today, in the 21st century, there have been very significant advances in all of these areas. The most promising modern x-ray imaging technique is now considered to be the use of coherent x-ray sources, such as ultra-brilliant synchrotron radiation or x-ray free-electron lasers, in the form of ptychographic techniques. This method is designed to achieve the highest spatial resolution, and images are obtained by phase recovery from diffracted images of the object without the use of optical elements. x-ray ptychographic tomography is a well-established 3D imaging technique capable of identifying volumes as small as a few 10 nm cubic. Recently, a team of researchers at the Paul Scherrer Institute in Switzerland achieved an even better spatial resolution, setting a new record of 4 nm. Their work introduces two new methods: burst ptychography and tomographic back-propagation reconstruction. Burst ptychography is an ultra-fast image acquisition method with speeds of up to 14,000 images per second. This method overcomes the instability of the absolute coordinates (X, Y, Z) caused by various factors. The new spatial resolution record of 4 nm can be attributed to the success of this method. Tomographic back-propagation reconstruction allows imaging of objects approximately 10 times larger than the conventional depth of field. In this study, the researchers successfully imaged integrated circuits manufactured using Intel's 7-nm process. The advantage of x-ray tomography is that it can acquire 3D images of the interior in a non-destructive manner, and each tomographic image can be evaluated without sectioning. However, its spatial resolution lagged behind that of other techniques, particularly electron microscopy. At some point, however, the spatial resolution broke through the 1-μm barrier and reached the tens of nm range, leading to many real-world applications, and since then further progress has been made to reach around 10 nm and now 4 nm. The level of difficulty may continue to increase, but the competition to reach the 1 nm barrier is likely to heat up. For more information, see the article by T. Aidukas, N. W. Phillips, A. Diaz, E. Poghosyan, E. Müller, A. F. J. Levi, G. Aeppli, M. Guizar-Sicairos, and M. Holler, “High-Performance 4-nm-Resolution X-Ray Tomography Using Burst Ptychography,” Nature 632 (2024): 81–88, https://www.nature.com/articles/s41586-024-07615-6. The author declares no conflict of interest.