Red

The ability to study events where very little energy is released in a LAr-TPC (Liquid Argon Time Projection Chamber) has gained immense scientific interest, as it allows researchers to search for the existence of WIMPs (Weakly Interacting Massive Particles) in a low-mass range, around or below 1 GeV/c², which remains largely unexplored. These types of events are characterized by the presence of only the ionization signal, as the primary scintillation signal is too weak to detect.

However, the recombination effects of ionization charges produced by nuclear recoils could depend on the orientation of the track along which they are generated, relative to the electric field in the TPC. Currently, there is no definitive theory explaining this recombination phenomenon. The possibility of experimentally verifying this, using Machine Learning (ML) applications, represents a highly innovative approach and could potentially open extremely interesting pathways for the direct search for Dark Matter.

The ReD TPC is a miniaturized version of the TPC to be used in DarkSide-20k. Its light detection system employs modern Silicon Photomultipliers (SiPMs), operating at cryogenic temperatures. Neutrons emitted by a 252Cf radioactive source are used to produce nuclear recoils, simulating WIMP interactions. The energy of these recoils is determined from the time of flight, measured using a spectrometer composed of plastic scintillators.

Machine Learning techniques are well-suited for extracting information and finding correlations within a dataset, learning directly from the data to deduce an effective model for explaining the phenomenon. ML algorithms have already been successfully applied to ReD data. This research area is advancing both in the development of additional ML algorithms for comparison and in incorporating new experimental parameters, such as the temporal distributions generated individually by each SiPM, which are not currently utilized.