Speaker
Description
A Generative Neural Network for the Prediction of Radio Pulses from Extensive Air Showers
Using radio emission from Extensive Air Showers (EAS) to measure cosmic rays has been gaining traction in recent years. Several large arrays of antennas have been planned or deployed in order to measure extensive showers and can give us insights into shower evolution with more and more precise measurements.
Simulations of radio emission from EAS are essential for reconstructing various shower parameters from the measured radio signals. Traditional microscopic simulations superpose the emission of all the electrons and positrons in the shower for every antenna. As modern experiments use more and more antennas, the computational cost of these simulations can get prohibitively large. Furthermore, modern reconstruction approaches such as Information Field Theory require fast, accurate and differentiable forward models for the prediction of air-shower radio pulses.
In this work, we present a novel neural network which predicts radio pulses for the environmental parameters of the Pierre Auger Observatory when provided with shower parameters including arrival direction, electromagnetic energy and depth of shower maximum for antenna positions on a star-shape grid. We compare the pulses generated by the network to CoREAS simulations and assess the network's ability to predict the fluence pattern and the total radiation energy. Finally, we demonstrate the viability of using air shower pulses predicted by the network for Xmax reconstruction.
