Multiplier-less approach in the neural network trigger algorithm for a detection of cosmic rays

Abstract

Nowadays astrophysics is focused on understand the origin of the ultrahigh-energy cosmic rays (UHECR). Finding sources of UHECR is difficult, due to deflection of charged particles in intergalactic magnetic fields. This problem can be, however, avoided by detecting electrically neutral particles, such as neutrinos, which are created by the UHECR particles in interactions during propagation. Due to the very low cross section of the neutrinos, the detection technique requires a very sophisticated algorithm.Our trigger algorithm is based on an analysis of signal shapes by an artificial neural network (ANN). This approach can efficiently separate air showers which started at the top of the atmosphere ("old" showers) from air showers initiated very close to detection level, which can be potentially initiated by neutrinos ("young" showers). The main disadvantage of our algorithm is high FPGA resource usage. Optimizing the size of ANN and a multiplier-less approach can decrease used resources.