Multiple detector analysis in the CRESST Dark Matter experiment

During the past century, from the first observation of the velocity anomalies in the Coma Cluster made by Zwicki, several observational evidences were collected at both galactic and cosmological scale that the a large amount of matter in our universe is dark. The most recent results of the Planck mission measured the ratio between ordinary matter and dark matter to be around 5. Still the nature of dark matter, despite the significant effort of the scientific community, remains unknown. The CRESST (Cryogenic Rare Event Search with Superconducting Thermo- meters) experiment based at the underground facility of the Laboratori Nazion- ali del Gran Sasso (LNGS) is one of the most sensitive experiments aiming at the direct detection of dark matter particles via elastic scattering off nuclei in CaWO4 scintillating crystals. The CRESST crystals are operated as cryogenic calorimeters, each equipped with a cryogenic light detector for the detection of the scintillation light coming from the crystals. When a particle interacts inside the detector produces a phonon and a light signal; the latter is used to discriminate nuclear recoils, possibly induced by dark matter scattering, from electron recoils induced by the dominant electron/gamma background. Defining the Light Yield as the ratio between the energy measured in the light channel and the one measured in phonon channel, it is possible to establish regions (bands) where the different interactions are expected. The results of CRESST Run 32 showed an excess of events in the acceptance region that could not be explained in therms of known backgrounds. Interpreting these events as a dark matter signal, under the most standard assumptions for the galactic dark matter halo, identifies regions in the cross-section vs mass parameters space compatible with the observation. In the following CRESST physics run, thanks to a substantial global background reduction, it was possible to partially exclude the dark matter interpretation of the excess of events. This work aims to test with a high statistic the CRESST run 32 excess, to prove (and possibly exclude) the dark matter interpretation. The analysis is per- formed on the full dataset collected during Run 33 using a Likelihood approach to combine the data from multiple detectors. For this purpose data collected during CRESST Run 33 were used. These data, acquired between July 2013 and August 2015, provide a very large exposure, of the order of 130 kg·day for each detector, allowing to achieve an improvement of the the CRESST exclusion limit for dark matter masses above 1 GeV/c2. The description of the CRESST experimental framework in which the ana- 4 lysis has been portrayed, together with all the steps of the analysis workflow needed to go from the collected raw data to the fit of the bands in the Light Yield - Energy plane are described. The steps needed for extending the range of the detector, the operations necessary for the energy calibration, alongside the different cuts to remove pile-up and spurious events for the different detectors are thoroughly discussed. The relevant parameters for the electron/gamma and nuclear recoil bands that are used to produce the exclusion limit are computed for each individual detector. The exclusion limit is computed in this work with an Extended Maximum Likelihood approach, which allows to combine the data of multiple detectors and benefit of the cumulative exposure. The final part of the thesis focuses on comparisons with previous results and possible future improvements. The exclusion of the dark matter interpretation of the excess observed in Run 32 is discussed.