Speaker
Description
All FEL attosecond photon-pulse generation schemes currently employed at EuXFEL rely on extremely narrow, high-current spikes. However, no longitudinal diagnostic presently offers sufficient resolution to resolve these critical features. We present a machine-learning reconstruction that fuses complementary coherent-radiation measurements to infer the current profile from multi-band spectra. In addition to the facility’s THz spectrometer (CRISP), an infrared spectrometer has been integrated into the diagnostics path, broadening spectral coverage and improving the conditioning of the inverse problem. The model is trained on simulated data with physics-motivated constraints and transferred to experiment using domain-adaptation strategies. We detail the reconstruction pipeline and validation methodology, emphasizing uncertainty characterization, sensitivity to calibration and bandwidth, and the incremental benefit of combining THz and IR channels. Using EuXFEL single-shot data, we show representative current-profile reconstructions and comparisons to the existing CRISP-based workflow, including ablation studies isolating the contributions of each spectrometer. We also discuss practical aspects of online use during attosecond-mode tuning and outline the path toward routine, facility-integrated deployment.
| In which format do you inted to submit your paper? | LaTeX |
|---|---|
| Preprint marking on your proceeding paper | I do not intend to submit a paper associated with this contribution. |