Workshop: Hard X-rays imaging opportunities at ALBA
Maxwell Auditorium
ALBA Synchrotron
Invited speakers
Marcus Carlsson, Associate Professor in Mathematics at Lund University
Title: Joint Holotomography Solver for Large-Scale Experimental Data Processing
Abstract: Multi-distance holotomography enables quantitative three-dimensional phase-contrast imaging with coherent hard X-rays and routinely achieves sub-100 nm resolution. At such resolutions, reconstructions become sensitive to probe structure and scanning positions. Conventional workflows address these effects through sequential processing steps, which can limit accuracy.
We present a scalable joint reconstruction framework that simultaneously estimates the object, illumination probe, and scan-position corrections from the measured data. Using a bilinear-Hessian-based optimization strategy and a multi-GPU implementation, the method enables practical joint reconstruction of large experimental datasets and reduces artifacts compared with conventional pipelines.
Pablo Villanueva-Perez, Associate Professor in Synchrotron Radiation Research at Lund University's Department of Physics
Title: 3D and 4D X-ray imaging opportunities at diffraction-limited storage rings
Abstract: The advent of diffraction-limited storage rings, such as MAX IV, SIRIUS, ESRF-EBS, APS-U, and the upcoming ALBA-II, opens the door to new spatiotemporal frontiers in X-ray imaging. Unlocking this potential, however, requires not only advances in reconstruction algorithms but also new acquisition strategies capable of efficiently capturing high-dimensional information from dynamic systems.
In this talk, I will present our recent progress in both acquisition and reconstruction methods for 3D and 4D (3D+time) X-ray imaging. First, I will introduce X-ray multi-projection imaging (XMPI), a technique that overcomes the rotational constraints of conventional time-resolved tomography by generating multiple angularly resolved beams that probe the sample simultaneously. This approach enables the recovery of 4D information without the need for sample rotation or any scanning.
Second, I will present end-to-end, self-supervised AI reconstruction frameworks for 3D and 4D imaging. These methods incorporate physical models and sample priors to produce high-fidelity reconstructions from sparse data, enabling up to an order-of-magnitude faster acquisitions without sacrificing spatial resolution in both full-field and coherent imaging modalities.
Finally, I will highlight the spatiotemporal capabilities that emerge from combining XMPI with these AI-based approaches. This includes, for example, the ability to observe 4D dynamics at frequencies up to 25 kHz with sub-10 µm spatial resolution during advanced manufacturing processes.
We anticipate that the combination of XMPI and AI-driven reconstruction, together with the capabilities of ALBA-II and similar facilities, will enable access to previously unobservable dynamic phenomena across a wide range of scientific and industrial applications.
