Sprecher
Beschreibung
Coherent elastic neutrino–nucleus scattering (CE$\nu$NS) is a Standard Model process in which a neutrino scatters coherently off an entire nucleus, provided that the momentum transfer is sufficiently small compared to the inverse nuclear size. This process serves as a powerful probe of weak interactions, neutron distributions within nuclei, and potential new physics scenarios, including non-standard neutrino interactions, light mediators, and neutrino electromagnetic properties.
The Mitchell Institute Neutrino Experiment at Reactor (MINER) employs 72 g cryogenic sapphire ($\mathrm{Al_2O_3}$) detectors instrumented with phonon sensors and operated at millikelvin temperatures. In a recent search for CE$\nu$NS, MINER observed no statistically significant excess above background, with the measurement sensitivity being primarily limited by reactor-correlated backgrounds. In addition, an excess of low-energy events was observed that could not be fully explained by the simulated background model. This unexplained component constitutes one of the dominant limitations to the sensitivity of the present measurement.
Similar low-energy excesses have been reported in several cryogenic detector experiments. Although their physical origin remains uncertain, common phenomenological features have emerged. The event rate typically rises steeply toward lower energies and often reappears following detector warm-up cycles, subsequently decreasing with detector operation time. Understanding and mitigating this background is therefore crucial for future CE$\nu$NS and rare-event searches employing cryogenic detector technologies.
Within MINER, we investigated this background using deep-learning-based waveform analysis techniques. We identified a characteristic long rise-time feature in the pre-pulse region and developed a classification framework capable of rejecting approximately 50% of the low-energy excess events. The resulting background reduction improves the projected CE$\nu$NS detection sensitivity by about 10%, demonstrating the potential of machine-learning approaches for enhancing the performance of future cryogenic neutrino experiments. The developed methodology is broadly applicable to experiments employing similar cryogenic detector technologies and can be used to mitigate low-energy backgrounds, thereby extending their physics reach. These studies provide new insight into the nature of low-energy backgrounds in cryogenic detectors and represent an important step toward improving the sensitivity of future CE$\nu$NS and other rare-event searches.