FEL2026 - 42nd International Free Electron Laser Conference

Machine-learning-based four-dimensional phase space reconstruction for future S3FEL electron beam diagnosis

MOP33

24 Aug 2026, 16:00

2h

Poster Presentation Session 3: FEL theory and Machine Learning Monday Poster Session

Speakers

Cheng-xin Wu (Institute of Advanced Light Source Facilities Shenzhen) Zhenbiao Sun (Institute of Advanced Light Source Facilities Shenzhen)

Description

Four-dimensional (4D) transverse phase space (TPS) reconstruction of electron beam is an important issue in accelerator beam diagnosis, since the 4D-TPS is vital for beam manipulation and cannot be directly determined from limited two-dimensional screen measurements. A machine-learning-based differentiable reconstruction framework is developed to infer the hidden phase-space distribution, and it has been verified with ELEGNAT simulation code. The reconstructed 4D-TPS can be used to resolve the 4D emittance and analyze the x-y transverse coupling. What’s more, it can support creating the surrogate model for practical accelerator beam tracking. This framework is being examined on a superconducting continuous-wave Electron Source Test Facility. In the future, it will also be applied to the under-construction Shenzhen Superconducting Soft-X-ray Free Electron Laser (S3FEL) facility for beam diagnosis and manipulation.