Reconstructing Time-of-Flight Detector Values of Angular Streaking Using Machine Learning

TL;DR

This paper introduces a machine learning approach to denoise and reconstruct faulty Time-of-flight detector data in angular streaking experiments, enabling real-time analysis and improved data quality for attosecond science.

Contribution

A novel machine learning method trained on simulation data that effectively denoises and reconstructs multiple failed detectors in real-time during experiments.

Findings

Successful denoising of noisy detector data

Reconstruction of up to three failed detectors

Applicable online during experiments

Abstract

Angular streaking experiments enable for experimentation in the attosecond regions. However, the deployed Time-of-flight detectors are susceptible to noise and failure. These shortcomings make the outputs of the Time-of-flight detectors hard to understand for humans and further processing, such as for example the extraction of beam properties. In this article, we present an approach to remove high noise levels and reconstruct up to three failed Time-of-flight detectors from an arrangement of 16 Time-of-flight detectors. Due to its fast evaluation time, the presented method is applicable online during a running experiment. It is trained with simulation data, and we show the results of denoising and reconstruction of our method on real-world experiment data.