Abstract:
For several decades, experiments have now searched for a so far undiscovered
particle, called dark matter, that is thought to be responsible for a variety of
astrophysical and cosmological observations. As part of this search, cryogenic
calorimeters like the ones of the CRESST (Cryogenic Rare Event Search with
Superconducting Thermometers) experiment have been improved to measure
energies of nuclear recoils in their detectors down to roughly 10 eV.
At energies below roughly 200 eV, CRESST measures an up-to-now unexplained
background with a steeply rising event rate towards the threshold that shows the
same signature in the detector as a particle interaction. This so-called Low-Energy
Excess (LEE) is a serious obstacle in exploiting the full potential of the achieved
low energy thresholds and since other experiments observe similar phenomena
[1] unveiling the LEE's origin has even become a community-wide interest.
The following work investigates the behavior of the LEE in the energy and time
domain, using data from the measurement campaign Run36 of the third phase
of CRESST (CRESST-III). The obtained description is then used to compare the
LEE between the detectors and to formulate requirements for LEE origin theories.
Furthermore, data from dedicated tests during which the cryostat was warmed up
to different temperatures also enable studying the temperature dependence of
the LEE as part of this work.
The main tool used here for describing the LEE is unbinned two-dimensional
Bayesian fits, performed with a custom Python software that is based on the MCMC
(Markov-Chain-Monte-Carlo) sampling package emcee [2]. Different fitting models
are then compared via the AIC (Akaike Information Criterion).
It is observed in this work that in all investigated detectors, the LEE can above an
energy of 40 eV be well described with a single power-law in the energy and a linear
combination of two exponential decay functions in the time domain. In addition,
the hypothesis that these two exponential decays are only an approximation of a
single power-law is also discussed. Furthermore, it is shown that the resulting fit
parameters agree very well between the detectors and that the LEE is consequently
not clearly dependent on any of the varied detector characteristics.