Towards an automated data cleaning with deep learning in CRESST.

The CRESST experiment employs cryogenic calorimeters for the sensitive measurement of nuclear recoils induced by dark matter particles. The recorded signals need to undergo a careful cleaning process to avoid wrongly reconstructed recoil energies caused by pile-up and read-out artefacts. We frame this process as a time series classification task and propose to automate it with neural networks. With a data set of over one million labeled records from 68 detectors, recorded between 2013 and 2019 by CRESST, we test the capability of four commonly used neural network architectures to learn the data cleaning task. Our best performing model achieves a balanced accuracy of 0.932 on our test set. We show on an exemplary detector that about half of the wrongly predicted events are in fact wrongly labeled events, and a large share of the remaining ones have a context-dependent ground truth. We furthermore evaluate the recall and selectivity of our classifiers with simulated data. The results confirm that the trained classifiers are well suited for the data cleaning task.

Secular equilibrium assessment in a CaWO4 target crystal from the dark matter experiment CRESST using Bayesian likelihood normalisation.

CRESST is a leading direct detection sub-GeVc-2 dark matter experiment. During its second phase, cryogenic bolometers were used to detect nuclear recoils off the CaWO4 target crystal nuclei. The previously established electromagnetic background model relies on Secular Equilibrium (SE) assumptions. In this work, a validation of SE is attempted by comparing two likelihood-based normalisation results using a recently developed spectral template normalisation method based on Bayesian likelihood. Albeit we find deviations from SE in some cases we conclude that these deviations are artefacts of the fit and that the assumptions of SE is physically meaningful.

Geant4-based electromagnetic background model for the CRESST dark matter experiment.

The CRESST (Cryogenic Rare Event Search with Superconducting Thermometers) dark matter search experiment aims for the detection of dark matter particles via elastic scattering off nuclei in CaWO 4 crystals. To understand the CRESST electromagnetic background due to the bulk contamination in the employed materials, a model based on Monte Carlo simulations was developed using the Geant4 simulation toolkit. The results of the simulation are applied to the TUM40 detector module of CRESST-II phase 2. We are able to explain up to ( 68 ± 16 ) % of the electromagnetic background in the energy range between 1 and 40 keV .

Limits on Momentum-Dependent Asymmetric Dark Matter with CRESST-II.

The usual assumption in direct dark matter searches is to consider only the spin-dependent or spin-independent scattering of dark matter particles. However, especially in models with light dark matter particles O(GeV/c^{2}), operators which carry additional powers of the momentum transfer q^{2} can become dominant. One such model based on asymmetric dark matter has been invoked to overcome discrepancies in helioseismology and an indication was found for a particle with a preferred mass of 3 GeV/c^{2} and a cross section of 10^{-37} cm^{2}. Recent data from the CRESST-II experiment, which uses cryogenic detectors based on CaWO_{4} to search for nuclear recoils induced by dark matter particles, are used to constrain these momentum-dependent models. The low energy threshold of 307 eV for nuclear recoils of the detector used, allows us to rule out the proposed best fit value above.

Low-level γ-ray spectrometry at the underground laboratory Garching.

Two low-background setups for material screening based on HPGe detectors were built in the Garching Underground Laboratory with an overburden of ~10 m.w.e. They include several layers of passive shielding as well as an active muon veto. The first setup (GEM) comprises a 150% efficiency HPGe detector which can optionally be surrounded by a NaI(Tl) scintillation detector that serves as anti-Compton veto. The second setup (LoAx) consists of two smaller HPGe detectors which are arranged face-to-face to cover a larger solid angle around the sample and to allow coincidence measurements. For a 5.6 kg piece of copper after 11 days of measurement we have reached a sensitivity for (226)Ra and (228)Ra/(228)Th of ~5 mBq kg(-1) with the GEM setup. In the LoAx setup we have achieved limits of less than 100 mBq kg(-1) for (234)Th and (210)Pb with a 156 g sample of PPO wavelength shifter after 18 days of measurement.