Department High Brightness Beams
Beam Dynamics and Instrumentation
Simulated Commissioning Studies for BESSY III: Charting the Path to Brightness
BESSY III, the next-generation multi-bend achromat (MBA) electron storage ring, will set new synchrotron light source performance standards. As part of the accelerator design team, our group focuses on simulated commissioning. Through computational beam dynamics studies, we analyze the effects of inevitable errors in magnets, RF systems, injection, and diagnostics in the electron beam. To counteract these effects, we design a commissioning process that incorporates corrector magnets and beam diagnostics, and utilizes available control parameters in the accelerator. Our goal is to bring the accelerator as close as possible to its ideal operating conditions, optimizing key factors such as dynamic aperture and beam brilliance. Beyond simulations, we are currently testing and refining commissioning routines at the existing BESSY II light source to ensure a smooth and safe implementation at BESSY III in the future. Through this work, we lay the groundwork for a high-performance, next-generation light source that will enable groundbreaking research and innovation.
Dynamic Optimization of the BESSY III lattice.
Beam Dynamics of SRF Photoinjectors – From Analytical Models to Machine Learning Optimization
SRF photoinjectors are powerful electron sources, delivering MeV beam energies, ultrashort pulses, and high brightness at high repetition rates—essential for applications like ultrafast electron diffraction, Terahertz experiments, energy recovery linacs, and free-electron laser injectors. However, optimizing beam properties is complex due to the vast number of machine parameter combinations.
Analytical model of emittance contributions at the SRF photoinjector of SEALab.
Our approach leverages the full range of modeling tools—from analytical approximations to machine learning-assisted optimization—to enable fast, ideally real-time beam tuning based on simulations and measurement-driven models. We significantly reduce computational time by replacing costly simulations with neural network approximations. Additionally, our reinforcement learning strategy uses fast derivative computation to optimize beam properties more efficiently than conventional methods while maintaining a defined accuracy.
Scheme for a learning loop of a surrogate model to optimize the SRF photoinjector parameters.
Publications and Talks
- E. Brookes, Mechanisms for commissioning and implementation of the SEALab beam modes: from modelling to optimization strategies, talk at ERL Workshop 2024.
- B. Alberdi Esuain, Beam Dynamics and Instrumentation for MeV Electron Scattering with an SRF Photoinjector, HU thesis 2024.
- D. Meier et al, Optimizing a superconducting radio-frequency gun using deep reinforcement learning, PRAB article 2022.