Demonstration of Efficient Radon Removal by Silver-Zeolite in a Dark Matter Detector

TL;DR

This work tackles radon background in low-background detectors by demonstrating a room-temperature radon trap based on silver-zeolite Ag-ETS-10 and benchmarking it against coconut-shell activated charcoal. Using a closed-loop gas system around a spherical proportional counter, the authors implement in-situ trap regeneration at $160^{\circ}$C and perform phase-resolved radon tests with rigorous data cuts and MC-based rate modeling. They quantify trap performance via the radon-reduction factor $R(t)$, estimated from phase III decay trends and phase IV observations, utilizing MCMC to propagate uncertainties and account for pileup and dead-time. The results show silver-zeolite achieving near-complete radon removal, with $R$-values three orders of magnitude higher than those for activated charcoal, indicating substantial practical benefits for underground lab operations and future ton-scale experiments in dark matter and neutrino physics.

Abstract

We present the performance of an efficient radon trap using silver-zeolite Ag-ETS-10, measured with a spherical proportional counter filled with an argon/methane mixture. Our study compares the radon reduction capabilities of silver-zeolite and the widely used activated charcoal, both at room temperature. We demonstrate that silver-zeolite significantly outperforms activated charcoal by three orders of magnitude in radon capture. Given that radon is a major background contaminant in rare event searches, our findings highlight silver-zeolite as a highly promising adsorbent, offering compelling operational advantages for both current and future dark matter and neutrino physics experiments. Furthermore, this not only offers great promise for developing future radon reduction systems in underground laboratories, but also paves the way for innovative, multidisciplinary advancements with far-reaching implications in science, engineering and environmental health.

Figures (7)

• Figure 1: Closed-loop circulation system for the radon removal campaigns, made of stainless steel fittings, valves and tubing. The setup includes a radon source to diffuse radon into the SPC detector. The gas from the detector (red) circulates through the trap, where radon is adsorbed. The purified gas (blue) is pumped back to the detector, passing through a particle filter, a binary gas analyzer (BGA), and a custom-made laser absorption spectroscopy system (LAS) Obrien:2021Garrah:2023.
• Figure 2: Measured rate with either quality or $^{214}$Po cut applied during the three radon campaigns, with the trap filled with either silver-zeolite or activated charcoal, at room temperature. The blue distribution corresponds to the background run (phase I), orange is the radon diffusion in the SPC (phase II), green is the radon decay (phase III), and red is the events recorded when the radon trap is open (phase IV). The light and dark green curves (with either quality cut or $^{214}$Po cut applied, respectively) are the fitted/extrapolated decay rates for each campaign, depicting many random Markov Chain Monte Carlo (MCMC) samples. The vertical dashed line defines the reference time where the expected rate with quality cut is equal to 76.5 Hz. Note that there are some visible gaps in the data (campaigns 2 and 3) due to brief pauses in data collection.
• Figure 3: Ratio between the rate in phase IV (trap open) at the reference time and the rate in phase I (background run) for each campaign with the trap filled with silver-zeolite (campaigns 1 and 2) and activated charcoal (campaign 3). The blue circle and red square represent the rate ratio obtained with either quality cut or $^{214}$Po cut, respectively. Complete radon removal correspond to a rate ratio of 1.
• Figure 4: R-values at 90% lower confidence limit (LCL), obtained at the reference time, with the trap filled with silver-zeolite (campaigns 1 and 2) and activated charcoal (campaign 3). The blue and red bars represent the R-values obtained with either quality cut or $^{214}$Po cut (with 50% reduced signal efficiency), respectively.
• Figure S1: Amplitude distribution of events for approximately 6 hours of data in phase III of campaign 1, recorded by the SPC (black markers). The distribution is modeled with an adaptive-bandwidth Gaussian kernel density estimation silvermanwang2011bandwidthkde_articlekde_code with bootstrapped efron statistical uncertainty (red curve and $1\sigma$ shaded band). The inset shows the amplitude peak positions corresponding to the maximum energy deposition of three alpha decays versus the corresponding alpha energies (red markers with error bars), with a linear fit (blue curve and shaded band).