remoTES: A new design for the cryogenic NaI detectors of the COSINUS experiment

The increasing statistical significance of the DAMA / LIBRA annual modulation signal is a cause for tension in the field of dark matter direct detection. The COSINUS experiment aims at a model-independent cross-check of the DAMA / LIBRA signal claim, using NaI crystals operated as cryogenic scintillating calorimeters at millikelvin temperatures. Such a setup enables measurement of phonon and scintillation light signals via Transition Edge Sensors (TESs) and allows particle discrimination on an event-by-event basis. The non-standard properties of NaI cause an obstacle when attaching a TES directly onto the surface of the crystal. This can be overcome with the “remoTES” design, where the TES is attached to an external wafer crystal. We present the results from a first successful operation of NaI and other crystals as cryogenic calorimeters with the remoTES design.


Introduction
Astronomical and cosmological observations imply that 26 % of the universe's mass content is composed of non-luminous and non-baryonic dark matter (DM) [1]. The nature of DM is still unknown and yields one of the major open questions in modern physics. Various new fundamental particles and extensions to the Standard Model have been proposed as DM candidates [2,3], among those, weakly interacting massive particles (WIMPs) are favored by the DM direct detection community [4].
Despite immense experimental effort, no convincing DM signal has been observed until today and the DM hypothesis has been ruled out for most excess signals seen by experiments such as XENON [5] or CRESST [6]. However, there is one long-standing signal claim remaining, namely by the DAMA/LIBRA experiment [7,8]. DAMA's claim is based on the hypothesis that the DM rate should be subject to an annual modulation due to Earth's motion within the Milky Way. In the last 25 years, DAMA/LIBRA has continuously observed such a modulation with period and phase in-line with the DM hypothesis. By today, the statistical significance of the signal has reached over 13σ in the 1-6 keVee energy range. However, the region of the parameter space corresponding to the DAMA signal under the standard scenario assumptions has been ruled out by the null results of various other experiments [9]. In order to perform a model-independent investigation of the DAMA/LIBRA signal, it is mandatory to utilize the same target material, sodium iodide (NaI) [9,10]. Alongside other planned and active experiments -such as SABRE [11], COSINE [12] and ANAIS [13] -the COSINUS (Cryogenic Observatory for SIgnatures seen in Next-generation Underground Searches) experiment aims at such a model-independent test of the DAMA/LIBRA signal claim [14].
In contrast to its competitors, COSINUS will operate NaI crystals as cryogenic scintillating calorimeters, cooled down to milli-Kelvin temperatures and read out by Transition Edge Sensors (TESs). With this detector setup, each particle interaction in the detector material leads to both a phonon/heat and a light signal. While the particle-independent phonon signal gives a measure of the total deposited energy in the crystal, the scintillation signal is subject to light-quenching and allows for particle discrimination. The COSINUS setup thus enables separation of nuclear recoils (possible DM signal) from the electron/gamma-background (e/γ) on an event-by-event basis.
To absorb and measure the scintillation light, the NaI crystals will be placed in a beakershaped silicon detector equipped with a tungsten TES (W-TES) thermometer as developed and refined by the CRESST collaboration [15]. For measurement of the heat signal, direct fabrication of the TES thermometer onto the absorber crystal would ensure excellent transmission of the non-thermal phonons to the W-film. Unfortunately, this production procedure is not possible for NaI as the crystals are soft, hygroscopic, and have a low melting point. The COSI-NUS collaboration thus developed the so-called remoTES, where the TES is instead placed on a remote wafer crystal that is connected to the absorber. This detector design is described in detail in section 2. Several R&D measurements have been performed with remoTES detectors and different absorber materials, the results of which are summarized in section 3. Finally, in section 4 we conclude and give an outlook on future work planned regarding setup and detector design of COSINUS.

remoTES detector design
The idea behind the remoTES detector was first proposed by M. Pyle et al. [16] as a simple and reproducible alternative to the direct fabrication of TESs onto absorber crystals. The COSINUS implementation of the remoTES detector is schematically depicted in Figure 1, showing both 033.2 the absorber and wafer crystal [17]. The fabrication-intensive W-TES is attached to the waferlike separate crystal and connected to the heat bath via a gold thermal link. The absorber crystal itself is connected to the TES via a system of gold pads and gold bonding wire(s). Any signal created by thermal and athermal phonons in the absorber crystal is then transmitted to the TES via electron-phonon coupling in the gold pad. Although the heat capacity of the gold pad and wire bond is diminishing the overall created signal, the factor ten higher electronphonon coupling in gold compared to tungsten [16,18] is expected to compensate for this loss. Moreover, the remoTES design avoids signal loss due to acoustic mismatch between different materials, as would occur in the so-called composite design which has been previously suggested and probed by the COSINUS collaboration [19].
While the application of the W-TES to the wafer crystal (Al 2 O 3 in all prototypes) is sa standardized procedure, the specifics of the remaining remoTES design vary between different prototypes. In this work we present the construction details and results of three different prototypes, featuring silicon (Si, 2.33g), tellurium dioxide (α-TeO 2 , 2.27g), and sodium iodide (NaI, 3.67g) absorber crystals. The process of attaching the gold pad to the absorber crystal can be adapted to the physical and chemical properties of the crystal. The same is true for the method of attaching the gold wire to the pad on the absorber, as standard wedge-or ballbonding may not be possible for soft and fragile absorber crystals. The details of the remoTES design for each of the three prototype absorbers are summarised in Table 1.

Results from prototype measurements
The two measurements with Si and TeO 2 absorber crystals were performed at an above-ground wet dilution refrigerator of the CRESST group at the Max-Planck Institute for Physics in Munich. The most recent measurement of a NaI remoTES, reported here, was instead performed at the CRESST test cryostat located at the INFN Gran Sasso Underground Laboratory (LNGS). Specifics of the setup and the analysis chain for the above-ground runs can be found in [17]. For the NaI underground run, where the remoTES was operated in a module together with a beaker-shaped Si light detector, the analysis chain is comparable. In all runs at least two types of particle pulse event classes could be observed, one for hits in the absorber crystal and one for hits in the wafer crystal. More detail on these event classes and the discrimination process can be found again in Ref. [17].  In all runs 57 Co and 55 Fe sources were used for the energy calibration. The X-rays of 5.89 keV (K α ) and 6.49 keV (K β )) produced by the 55 Fe-source can be seen in the spectra for the Si and TeO 2 measurements shown in Figure 2. For a measurement with the NaI absorber, which included a neutron source (AmBe), we show the combined result of light and remoTES phonon signal in a so-called light yield plot. The 2D histogram in Figure 3 shows the total recoil energy as measured by the remoTES on the x-axis and the light yield LY = light signal phonon signal on the y-axis. In this presentation of the data, light quenching takes full effect and one can discriminate the e/γ events, at a light yield of one, from the quenched nuclear recoil events. The events visible between the two bands at energies higher than 50 keV can be attributed to inelastic recoils off iodine nuclei, whereas the dense population of events around 6 keV consists of X-ray events produced by the 55 Fe source. For all measurements, the baseline resolution is determined following the well-established procedure described in [17] and can be found in Table 1. The threshold of the detectors can be estimated as five times the baseline resolution.

Conclusion and outlook
Through various prototype measurements, we have shown, that the remoTES detector design gives an easy-to-fabricate alternative to the direct deposition of TESs onto absorber materials. The successful operation of a remoTES with a NaI absorber crystal, enabling particle discrimination, was an important milestone for the COSINUS experiment. With an achieved baseline resolution of under 400 eV, corresponding to a true recoil energy threshold of <2 keV, the detector prototypes are closing in on the final design goal of a 1 keV threshold. While work is ongoing within the collaboration to refine the design of the detector modules (remoTES plus Si beaker light detector), the construction of the experimental setup is currently taking place in Hall B of LNGS. The first phase of data taking with the COSINUS experiment is planned to start in autumn 2023, and the first DM results are expected by the end of 2024.