The first goal to improve the spatial resolution of laboratory X-ray microstructural characterization techniques, was initially tackled via classical approaches, which indeed helped, but did not reach our goal. Our next attempt included the use of machine-learning methodologies. Critical issues here are to i) have sufficient data for training, and ii) to perform the training, which typically involves many hours of human-machine interaction. We found a unique way to overcome both these issues: we made virtual samples in the computer, calculated the corresponding diffraction patterns, and let the computer train itself on these patterns, where of course the ground truth is known (Fig1). This means it was straightforward to get enough training patterns, and no human involvement was needed in the training. The results were excellent, and with the new routines, an improvement in spatial resolution of about a factor of two was achieved. Today, machine-learning routines like ours are used in industry for the present purpose.
In the course of this work, we however conceived a completely unexpected and not tried before idea for a new experimental methodology to achieve the higher spatial resolution. It involves the use of X-ray focusing optics, and a sample scanning procedure. The idea is considered so new and with so large potential that we protected it via a patent application filed in July 2020, and a PCT application was filed in July 2021. In 2022 a license agreement was signed with the Danish company Xnovo Technology ApS. Also in 2022 we succeeded securing additional ERC POC funding to further develop and demonstrate the methodology. This work is done in collaboration with Xnovo, and a photo of the set-up is shown in Fig 2. In 2023, we demonstrated that we can map grains smaller than 5 um, with a resolution down to 1um. Our original first overall goal is thus fulfilled beyond expectations.
The second goal to advance metal research by quantifying local microstructural variations in industrially relevant metals, is of course broader. M, and specific achievements less easily measurable. A few outstanding highlights are summarized in the following, but a measurable success criteria, which are fulfilled includeare that i) leading aluminum and steel industries today focus on this aspect and ii) we in 2022 held a very successful Risø International Symposium with this theme (see below).
Further Highlights include:
A common theory for annealing of particle containing metals is that the local microstructural variations near large non-deformable particles will stimulate nucleation (PSN). Using a multimodal X-ray approach, PSN was revisited in 3D in a commercial Al sheet. It was found that PSN was not as dominating as expected (Fig 3), and that sites associated with other microstructural local variations are of similar importance. These results stimulated huge interest from the international aluminum company Novelis, with whom we are now starting a collaboration project.
Boundary migration during recrystallization is of key scientific interest and essential for industrial processing. By 4D X-ray microscopy, we followed the boundary migration near a hardness indent in lightly rolled pure aluminum. Very heterogeneous migration was observed. The experimental results combined with MD simulations revealed that the classic assumption about how velocity is expected to relate to driving force and boundary mobility couldn’t explain the experimental results (Fig 4). The inhomogeneous morphology of the deformation microstructure has to be taken into account.
Gradient and multilayered metals are metals in which heterogeneity is introduced in a controlled manner to obtain tailored properties. We have studied laminated metal Ti-Al by synchrotron X-ray micro-diffraction, and showed that the difference in properties between the layers result in development of significant residual stresses. We therefore also investigated Ti manufactured with a layered structure of fine and coarse grains (Fig 5). This sample exhibits excellent properties combining the strength of the fine grains with the ductility of the coarse grains.
3D printing of metallic components has become popular during recent years. The printing parameters for standard metals are now highly optimized and almost fully dense samples can be made. But few voids are unavoidable. We have analysed effects of voids on boundary migration during recrystallization by in-situ EBSD experiments and phase field simulations. It is clear that voids significantly affect the migration kinetics, and that the position of the voids in relation to the inhomogeneous deformation microstructure is of importance (Fig 6). Our work on 3D printed metals, have inspired a novel research proposal on Microstructural Engineering of Additive Manufactured Metals, which has been approved for funding of approx. 5 mio Euro by a private Danish Research Foundation.
The third goal to link our 4D experimental results to PF, MD and CPFEM modelling, was tackled together with our affiliated partners. As mentioned by a few examples above this collaboration has been very successful, and it turned out that instead of just validating/improving the models by the 4D experimental data, much more scientific benefit was obtained by combining the power of the characterization and modelling to analyze the underlying mechanisms.
Concerning the broader dissemination, highlights are two very successful Risø Symposium organized by us, held in 2019 and 2022 with the themes ‘Metal Microstructures in 2D, 3D and 4D’ and “Microstructural Variability”. Both proceedings are published open access by IOP:
https://iopscience.iop.org/issue/1757-899X/580/1(öffnet in neuem Fenster) and
https://iopscience.iop.org/issue/1757-899X/1249/1(öffnet in neuem Fenster). During 15-20th May 2023 we co-hosted the 8th Conference on Recrystallization and Grain Growth. The conference was held in a two-site mode, at DTU and at Chongqing China, to avoid the uncertainties and travel difficulties related to the pandemic. We had approx. 100 participants at the DTU site, and altogether also this Conference was a great success.
