Thinning algorithm for the simulation of large signals in surface detector stations at the Pierre Auger Observatory

Extensive air showers induced by ultra-high-energy cosmic rays result in a massive number of particles reaching Earth's surface. Simulations of the surface detector of the Pierre Auger Observatory consist of injecting these particles into virtual volumes containing its water-Cherenkov detector stations. In such simulations, the shower core sometimes lands very close to such a station resulting in the injection of a colossal number of particles and correspondingly large run-times. As the electronics in such stations saturate, disproportionately large amounts of CPU and memory resources are therefore used to simulate signals that will eventually be clipped and discarded. Here, we present a station-level algorithm that greatly reduces run-times in such cases by randomly decimating injection candidates and scaling resulting signals accordingly. The method eliminates outliers in the distribution of simulation run-times and reduces the average run-time by dozens of hours without a significant cost to the quality of simulated signals. Meanwhile, it still allows for the full simulation of station signals up to the point of saturation, which differs from other speed-oriented algorithms relying on parameterizations that often inadequately describe the rare interactions and odd geometries relevant in the case of low statistics.