In the framework of the FLASH2020+ project, the FLASH1 beamline will be upgraded to deliver seeded FEL pulses for users. This upgrade will be achieved by combining high gain harmonic generation and echo-enabled harmonic generation with a wide-range wavelength-tunable seed laser, to efficiently cover the 60-4 nm wavelength range. The undulator chain will also be refurbished entirely using new radiators based on the APPLE-III design, allowing for polarization control of the generated light beams. With the superconducting linac of FLASH delivering electron beams at MHz repetition rate in burst mode, laser systems are being developed to seed at full repetition rates. In the contribution, we will report about the progress of the project.

The operation of light source accelerators is a complex process that involves a combination of empirical and theoretical physics, simulations, and data-intensive methodologies. For example, the FLASH1 beamline at DESY is upgrading to an external seeding FEL light source*. We utilize special diagnostics, machine learning algorithms, and comprehensive simulations to achieve this. To optimize resources, we constantly look to improve our approach, allowing us to robustly control the accelerator and meet the desired stability of our users. Machine learning and GPU-based algorithms have become crucial, enabling us to employ advanced optimization techniques and possibly AI. However, in many cases, it is imperative to establish a robust mechanism for simulations involving large particle numbers to ensure that future upgrades and experiments can effectively and sustainably leverage these computational strategies.