Spencer

Electro-Optic Sampling is a non-destructive, non-invasive longitudinal current diagnostic for relativistic electron bunches. Advanced spectral encoding methods allow for single-shot measurement capabilities, and reconstruction methods such as Phase-Diversity Electro-Optic Sampling (DEOS) allow for high bandwidth pulse measurements. Shaped electron bunches are of special interest to beam-driven wakefield acceleration, and DEOS is well suited to measure these shaped current profiles.

DEOS measurement is modeled from first principles. Phase-retardance signal simulated from THz pulse spectrum and ZnTe crystal properties, and reconstruction modeled in addition to detector resolution.

Machine Learning (ML) methods are being explored and developed to enhance the reconstruction accuracy of DEOS even further. By accounting for the real-world properties of the measurement, such as the detector resolution and temporal walk-off, an algorithm can be trained to accept the DEOS reconstruction and predict the underlying THz pulse free of resolution limitations and errors.

Experimentally, shaped beams are generated using Emittance-Exchange (EEX) beamline, and current profile is measured using DEOS. Measurement is benchmarked using Transverse Deflecting Cavity (TDC), and measured profiles are used to train ML algorithm.