Speaker
Description
The COSINUS experiment aims at the direct detection of dark matter, operating sodium iodide crystals as cryogenic calorimeters using the remoTES design. In this design, the TES is deposited on a seperate wafer and connected to the phonon collector on the absorber via a gold bonding wire. The resulting pulse shape upon a particle interaction is well described by a three-node thermal model.
To gain a more in-depth understanding of the detector physics, a Bayesian pulse-shape inference framework using BAT.jl was developed and validated on several prototypes. The configuration of the framework allows for the accommodation of posterior distributions for parameters such as thermal response and collection efficiency. These parameters are instrumental in quantifying phonon propagation and detector response. Consequently, this framework facilitates applications including detector diagnostics and event-type classification.
To address the increasing volume of data from large detector arrays, automated analysis modules have been developed. The purpose of these modules is twofold: firstly, to streamline data processing, and secondly, to reduce manual intervention. The development of all these modules was implemented using a low background data set from a COSINUS measurement conducted at LNGS. Subsequently, the performance was validated using above-ground detectors that exhibited a considerably elevated background and noise level.
It has been demonstrated that the modules possess the capacity to generate a noise power spectrum(NPS) entirely autonomously.A comparison of this NPS with a handmade NPS shows no significant difference. Furthermore, they employ a neural network to facilitate the cleaning of data. This is followed by a process of event type classification, which results in the generation of a standard event for each event type, thereby obviating the need for human intervention.
