Kelling team
AI FOR FUTURE PHOTON SCIENCE
Jeffrey Kelling - Group leader
CLOSING THE LOOP BETWEEN SIMULATION AND EXPERIMENT
Intense laser fields give rise to extreme states of matter where large forces are exerted on groups of charged particles. In such systems matter can be put under extreme pressure, such as in inertial-confinement fusion, or accelerated at small distances, such as in laser-plasma accelerators, which provides a path towards more compact, less costly particle accelerators, broadening availability in science, industry and medicine. Understanding these systems requires computationally demanding simulations and reconstruction algorithms to interpret experimental data.
We aim to close the loop between theory and experiment in wider photon science by researching data-driven digital twinning techniques in conjunction with large simulations to help guide and control experiments and extract more information from large amounts of data.
We research and apply data-driven smart optimization techniques to guide experiments and improve control of complex research installations around high-power lasers in conjunction with recent surrogate modelling techniques to supplement with data-driven digital twins. In building physics-informed neural networks and foundation models as surrogates for laser-plasma systems, we closely interface with state-of-the-art particle-in-pell simulations (PIConGPU). We study recent physics-based models and join simulation and experimental data to solve ill-posed inverse problems in X-ray scattering and particle detection in order to help reveal the detailed mechanism in laser-plasma interactions.
Research lines
- Surrogate Modeling and Digital Twins based on both simulation and experimental data;
- Parameter Optimization to guide experimental campaigns and devices;
- Modeling Inverse Problems for indirect measurements and phase retrieval.
