Speaker
Description
Electro-optic sampling (EOS) is an attractive technique for non-destructive measurement of electron bunch profiles. Its importance continues to grow for next generation high-energy accelerators, where intercepting diagnostics are no longer viable. In EOS, the measured signal corresponds to the phase retardance, rather than a direct measurement of the electric field. The electro-optic (EO) response acts like a finite-bandwidth transfer function, which attenuates high-frequency components and makes the reconstruction ill-conditioned. In this work, we investigate machine learning (ML) as a data-driven approach to improve reconstruction of the underlying THz pulse from EOS measurements. While the fundamental information content remains constrained by the EO response, the ML framework leverages prior knowledge of physically plausible pulse structures to stabilize the inversion.
| I have read and accept the Privacy Policy Statement | Yes |
|---|---|
| Supervisor's name | Gwanghui Ha |
| Supervisor's email | gha@niu.edu |
| Paper submission software | LATeX |