CRESST-II Phase 2

As the backgrounds and the potential signal were in the same order in CRESST-II phase 1 and the uncertainties in the background models were large, the major goal of CRESST-II phase 2 was to clarify the nature of the signal excess above background.

6A possible explanation for the excess events was published in [78]. Simulations showed that the energy spectrum of radon-induced events can have a steep rise at low energies due to sputtering in the surrounding materials, when the surface roughness of these materials is not neglected.

7The presence of two maxima can be ascribed to the presence of different nuclei in the target material CaWO4. M1 describes a possible WIMP with a mass ofmχ=25.3 GeV/c² which is heavy enough to produce detectable recoils of tungsten, whereas the lighter WIMP (mχ=11.6 GeV/c²) described by M2 only produces detectable scatters off oxygen and calcium nuclei. Both possibilities lead to a similar spectral distribution in terms of recoil energy and could in principle be discriminated by the light yield distribution of the signal events. However, in this measurement the overlap of the three nuclear recoil bands was to large for this, due to the finite resolution of both, phonon and light detector.

Figure 2.15: Parameter space for elastic spin-independent dark matter nucleon scattering.

The favored parameter space reported by CRESST-II phase 1 [73] is shown as brown region.

The exclusion limits of CRESST-II phase 2 of the modules TUM40 [67] and Lise [74] are shown in red and magenta, respectively. Additionally the expected sensitivities from a data-driven background model (1σconfidence limit (C.L.), light red and light magenta band) are displayed.

For comparison, exclusion limits (90 % C.L.) of liquid noble gas experiments are shown in blue [37, 38], and of germanium-based experiments in green [44, 45]. The favored parameter spaces reported by CDMS-Si [43] and CoGeNT [40] are drawn as green shaded regions. Marked in grey is the parameter space where coherent neutrino-nucleus scattering, dominantly from solar neutrinos, will be an irreducible background for a CaWO4-based dark matter search experiment [49].

The aim was to reduce neutron,α and in particular²⁰⁶Pb recoil backgrounds by more than one order of magnitude. CRESST-II phase 2 has been acquiring data from July 2013 until August 2015. A total of 18 detector modules corresponding to an overall total target mass of 5 kg has been installed.

The data show that the neutron background has been efficiently reduced by the in-stallation of an additional 3.5 cm thick PE shield inside the outer vacuum can. The background of degradedα’s, originating from a bulk contamination of the bronze hold-ing clamps, has been completely removed by introduchold-ing a newly produced ultra-pure bronze material [63] for the clamps. However, the background from²⁰⁶Pb nuclear recoil background could not be completely suppressed in the conventional modules despite ef-forts in improving material selection and radon-prevention methods [74]. In order to reduce the exposure to radon, the assembling and mounting of the detectors was carried out in radon-depleted air supplied by the CUORE experiment [79]. This background could only be suppressed efficiently in three new actively discriminating detector module designs, that were introduced in phase 2.

Actively Discriminating Module Designs

Actively discriminating modules were designed to efficiently reject surface α-events by only having scintillating and/or active surfaces inside the module. Three different designs, with two modules each, were installed in phase 2. Two of these designs feature a fully-scintillating housing of the detector module. Thereby, events induced by α-decays from all inner surfaces are shifted out of the region of interest by producing an additional light signal (see section 2.3). One design still uses non-scintillating clamps, but instead accomplishes to surround the crystal with only active surfaces.

Module with Big Carrier One design has a large TES carrier with the same diameter as the crystal and a thickness of 7 mm. Thereby, the carrier crystal, glued to the main absorber crystal, can be held by three pairs of clamps, which are covered with the scintillating polymer Parylene C. Events in the absorber and the carrier crystal experience a pulse shape difference, which is exploited in a pulse-shape analysis to discriminate possible relaxation events introduced by the parylene [76]. This module is mounted together with a standard light detector. Details about the performance of one of these modules can be found e.g. in [76].

Module with Si Beaker Light Detector Similar to the first design, the second design also uses a large TES carrier glued to the target crystal. Additionally the crys-tal is held inside a beaker-shaped silicon light detector. Thereby, the cryscrys-tal is surrounded by the light detector on all sides except the side glued to the carrier crystal. The geometry is designed so that the absorber crystal has no line-of-sight to the bronze clamps holding the carrier or any other surface than the light de-tector and the carrier. Again events from the non-scintillating clamps hitting the carrier can be discriminated by pulse-shape analysis. Details about the perfor-mance of these modules can be found e.g. in [76]

Crystal Held by CaWO₄ Sticks In the third design the target crystal is held by scin-tillating CaWO4 sticks which are held in place by clamps mounted outside the detector housing. In this way, no non-scintillating parts are present inside the housing. Additionally, in this design the crystal is a cuboid instead of a cylinder, for enhanced light output [80]. The light is detected with a standard light detector as in the conventional module. A schematic drawing of this design is shown in figure 2.16. A detailed description of this design can be found in [81].

All three designs performed well and effectively vetoed any ²⁰⁶Pb nuclear recoil back-ground. Therefore, a module housing with fully-scintillating or fully-active surfaces is the solution to suppress all surface-related backgrounds completely.

Crystal Quality

Since 2011 crystals have been grown at the crystal growing facility of the Technische Universit¨at M¨unchen (TUM) [82]. The aim is to improve the background and in-crease the light output compared to commercially available crystals. In phase 2 in total four TUM-grown crystals have been installed in actively discriminating modules in the CRESST setup. These crystals show a significant improvement in terms of radiopurity

light detector (with TES) block-shaped target crystal

reflective and scintillating housing CaWO4 sticks

(with holding clamps)

Figure 2.16: Schematic drawing of an actively discriminating module with crystal held with CaWO4 sticks. The sticks are held with clamps from outside the housing. A standard light detector as used in the conventional modules is mounted above the crystal.

with respect to the commercially available crystals. A lowering of thee⁻/γ-background in the ROI by a factor 2 to 10 with respect to commercial crystals (with an average rate of 3.44 /[kg keV day]) is accompanied with a significant reduction in the alpha contam-ination corresponding to a total intrinsic alpha activity of about 3 mBq/kg [77, 83].

Due to the improvement in background in addition to a low threshold of 603 eV the module TUM40 has the best overall performance in phase 2. An analysis of this module, with an exposure of 29.35 kg live days collected in 2013, was published in 2014 [67]. The superior performance of this module results in a gain in sensitivity for low mass dark matter and an at that time world leading limit on dark matter particles down to 1 GeV/c² was set (see figure 2.15).

The limit derived from data was compared with a Monte Carlo simulation, based on a background model assuming the presence of e⁻/γ-background only [83], which gives the light-red band (1σ C.L.) in figure 2.15 [67]. Limit and simulation agree throughout the whole mass range indicating that the events in the acceptance region may solely be explained by leakage from the e⁻/γ-band.

This result excludes the lower mass maximum (M2) from phase 1, but more statistics is required to improve the limit at higher masses and thereby to clarify the origin of the higher mass maximum (M1).

Low-Threshold Detectors

The importance of a low threshold was demonstrated with the result of the conventional detector module named Lise, which is the module with the lowest energy threshold operated in phase 2 [74]. With a threshold of 307 eV the sensitivity for light dark matter was significantly enhanced. In a data set with an exposure of 52.2 kg live days dark matter masses down to 0.5 GeV/c² are explored (see figure 2.15), which is a novelty in the direct dark matter search field. The improvement compared to the TUM40 result is a consequence of an almost constant background level down to the low threshold.

The lower the mass of the dark matter particle, the more relevant these improvements become. In contrast, the reduced sensitivity for higher masses is a result of a worse background level in the crystal and the poor performance of the light detector. With the

result of Lise CRESST demonstrated that detector performance is the key requirement to achieve sensitivity to dark matter particles ofO(1 GeV/c²) and below.

In figure 2.15 it can also be observed that the CRESST exclusion limits rise more moderately towards lower dark matter masses than the limits from other experiments.

This distinctive feature is a result of the ability to probe a possible dark matter signal on light nuclei (O and Ca) as well as on heavy W nuclei in addition to measure nuclear recoil energies with little systematic uncertainty down to a low energy threshold.