yig

Laser-plasma accelerators make large, conventional particle accelerators more compact, less costly, as well as increase broad availability and access in science, industry and medicine. However, comprehension of the involved physics requires sophisticated and computationally demanding algorithms for simulation and reconstruction. The Helmholtz AI young investigator group aims to close the loop between theory and experiment by researching data-driven digital twinning techniques that stimulate theoretical comprehension as well as experimental validation of the very complex dynamics involved in laser-particle acceleration.

Research focus

Our declared goal is to research digital twins of future Laser-driven particle accelerators which comprises of surrogate modeling, uncertainty quantification / outlier detection as well as solution of inverse imaging problems for fast understanding of experimental data.

We therefore research recent surrogate modelling techniques such as Physics-informed Neural Networks for acceleration of state-of-the-art Particle-In-Cell simulations (PIConGPU) and identification of unmodelled dynamics from data (PDE identification & learning). Those dynamics are experimentally observed, e.g. X-ray diffraction experiments, by reconstruction of diffraction data by reliable neural networks. The latter revolutionize the way scattering experiments are carried out by fast and reliable data analysis leveraging large amounts of training data and injection of prior knowledge into the inference procedure of neural networks.

Team

Nico Hoffmann is heading our Helmholtz AI young investigator’s group AI for Future Photon Sciences. His research interest are physics-guided techniques for solving inverse imaging problems by generative models as well as the application of Reinforcement Learning to planning and control problems.

Ritz-Ann Aguilar is researching generative models for inversion of Proton-Beam diagnostics at HZDR as well as X-ray diffraction patterns (e.g. Holography, SAXS) acquired at EuropeanXFEL together with experimentalists, computational physicists and data scientists.

Erik Thiessenhusen is working on data-driven methods for fast and reliable reconstruction of Xray diffraction data acquired at e.g. our HIBEF beamline at EuropeanXFEL within the scope of his PhD project at HZDR.

Maksim Zhdanov is working at the intersection of Bayesian Computation, probabilistic modeling and ML-based surrogate modeling for uncertainty quantification & reliable analysis of GI-SAXS data of e.g. EuropeanXFEL.

Anna Willmann is a PhD student in the YIG. She is researching generative models for developing a data-driven digital twin of the Free-Electron Laser beamline COXINELL at ELBE center.

Jeyhun Rustamov maintains our Neural Solvers library. He is currently developing code for in-memory training of neural operators via continual learning.