Pablo Villanueva Perez

Research

Our research focuses on the development of X-ray techniques to explore nature in 3D and 4D (3D+time), from the microscale down to the nanoscale, and from seconds down to femtoseconds. The unique penetration power of X-rays allows the study of the inner structure of matter in a non-destructive manner. Furthermore, their short wavelength allows the study from the macroscopic down to the atomic scale. These properties, together with the advent of high-brilliance sources, such as diffraction-limited storage rings (MAX IV) and X-ray free-electron lasers (European XFEL), offers the possibility to develop novel X-ray microscopes to explore nature in ways not possible before.

We are frequent users of the MAX IV synchrotron facility and the European XFEL.

Our research topics include:

• Simulation of X-ray instruments at high-brilliance sources
• Conceptual design and implementation of X-ray microscopes
• Coherent X-ray imaging
• Phase retrieval
• Fast reconstruction algorithms based on machine learning and other state-of-the-art approaches.

Some ongoing research projects are described below:

Projects

3D and 4D AI+Physics X-ray imaging reconstruction from sparse projections

X-rays have been widely used in scientific research and industry to study the structure and composition of materials. Among X-ray techniques, X-ray tomography is a commonly used 3D imaging technique in which a sample is scanned from different angles by rotating it. However, this process is typically slow and unsuitable for capturing fast, sensitive dynamics. To overcome these limitations, state-of-the-art single-shot 3D imaging techniques have been developed to capture multiple projections simultaneously from different directions [1, 2], but reconstructing time-resolved 3D movies from sparse experimental data remains a significant challenge. To address this problem, we developed two unsupervised deep learning reconstruction algorithms: ONIX (Optimized Neural Implicit X-ray Imaging) [3] and 4D-ONIX [4]. These algorithms combine advanced computer vision and AI methods with the physics of X-ray imaging and offer the potential to reconstruct high-quality 3D and 4D (dynamic 3D) structure and dynamics of the sample from ultra-sparse projections,  as few as two projections spaced 24° apart. These approaches allow researchers to visualize dynamic phenomena, such as droplet collisions and metal melting, up to 1,000 times faster than conventional methods. They provide a powerful tool for exploring the structure and dynamics of various samples and open new spatiotemporal frontiers in materials science and fluid dynamics.

References

[1] Villanueva-Perez, Pablo, et al. "Hard x-ray multi-projection imaging for single-shot approaches." Optica 5.12 (2018): 1521-1524.

[2] Liang, Xiaoyu, et al. "Sub-millisecond 4D X-ray tomography achieved with a multibeam X-ray imaging system." Applied Physics Express 16.7 (2023): 072001.

[3] Zhang, Yuhe, et al. "ONIX: An X‐ray deep‐learning tool for 3D reconstructions from sparse views." Applied Research 2.4 (2023): e202300016.

[4] Zhang, Yuhe, et al. "4D-ONIX: A deep learning approach for reconstructing 3D movies from sparse X-ray projections." arXiv preprint arXiv:2401.09508 (2024).

[5] Makowska, Malgorzata G., et al. "Operando tomographic microscopy during laser-based powder bed fusion of alumina." Communications Materials 4.1 (2023):73.

[6]Malgorzata Makowska. High-resolution time-resolved tomography of Selective Laser Melting of alumina, 2023 (on doi.psi.ch).

X-ray Multibeam ptychography

Multibeam ptychography is an innovative X-ray imaging technique that extends the capabilities of single-beam ptychography to overcome limitations in speed and field of view. While traditional ptychography offers high-resolution and quantitative phase contrast imaging, it is constrained by slow data acquisition over large areas. By dividing the X-ray beam into multiple coherent probes, multibeam ptychography enables simultaneous imaging of multiple regions, significantly improving efficiency while maintaining the resolution. This approach makes it ideal for studying a wide range of samples, from advanced materials to functional devices, enabling faster and more comprehensive measurements in cutting-edge research.