Characterization of a 220Rn source for low-energy electronic recoil calibration of the XENONnT detector

Low-background liquid xenon detectors are utilized in the investigation of rare events, including dark matter and neutrinoless double beta decay. For their calibration, gaseous 220Rn can be used. After being introduced into the xenon, its progeny isotope 212Pb induces homogeneously distributed, low-energy (<30 keV) electronic recoil interactions. We report on the characterization of such a source for use in the XENONnT experiment. It consists of four commercially available 228Th sources with an activity of 55 kBq. These sources provide a high 220Rn emanation rate of about 8 kBq. We find no indication for the release of the long-lived 228Th above 1.7 mBq. Though an unexpected 222Rn emanation rate of about 3.6 mBq is observed, this source is still in line with the requirements for the XENONnT experiment.


Introduction
Detectors utilizing liquid xenon to search for neutrinos and rare events have undergone rapid growth in both mass and radiopurity during the past decade [1][2][3][4].The time projection chambers (TPCs) used in these experiments utilized the high-Z property of the xenon target to shield the fiducial volume from surrounding radioactivity.However, this means any external gamma calibration source will not reach the center of the TPC, as the mean free path for example of MeV gammas is of O(cm) and thus much smaller than the diameter of the meter-scale TPCs.
In this paper, we study an internal source to calibrate the low-energy (keV-scale) electronic recoil response of the XENONnT TPC [5,6]. 220Rn emanated from 228 Th is carried by the xenon gas flow and mixed with the liquid target.The 220 Rn decay chain produces a variety of radiation: In the 228 Th decay chain, 212 Pb is relatively long-lived with a half-life of 10.6 h, providing a betaspectrum with a uniform component below 200 keV electronic recoil energy, and a high event rate above this energy.This calibrates the major background from the beta emitter 214 Pb originating from 222 Rn, which constitutes about 50% of the electronic recoil (ER) background below 10 keV in the first science data set of the XENONnT [1] experiment.Such a low-energy calibration source is essential to estimate the electronic recoil background contributing to the region of interest to the dark matter search [7].Following tests in a small detector [8], the first calibration using a 220 Rn source was implemented in XENON100 [9].As the active mass of dark matter detectors has since increased by 2 orders of magnitude, a higher radon emanation rate is desired to produce the required statistics for calibration; meanwhile, the average ER background rate (events per keV×tonne×year) has decreased by 2 orders of magnitude, which places a stringent limit on the radiopurity of the calibration source with a more precise radioactive background estimation.These requirements for the source are crucial for applications for low-background measurements.
The design of the 228 Th source is described in section 2, while measurements of its 220 Rn and 222 Rn emanation rate are reported on in section 3 and section 4. To prevent contamination of the experiment with the long-lived 228 Th, its release has been excluded by a measurement which is presented in section 5. Taken together, these measurements confirm that this source meets or exceeds the requirements of a low-background experiment such as XENONnT.

Source Preparation
Four 228 Th source discs were procured from Eckert & Ziegler [10], with a nominal activity of ∼13.8 kBq each, as of April 2022.The 228 Th oxide was electroplated to a 5 mm-diameter active area at the center of each 25.5 mm-diameter platinum disc.The usual gold covering on the active area was not added to enhance radon emanation.The four discs were placed in an emanation vessel with Conflat ® CF-50 stainless steel (SS) flanges.The source was held by oversized washers and nuts that were mounted on three M6 SS threaded rods that were attached to a CF-50 blind flange.The discs were mounted with the active sides facing each other, as shown in Figure 1.To minimize contamination, all parts of the support structure were thoroughly cleaned prior to assembly according to the procedure outlined in [11].

Measurement of the 220 Rn emanation rate
During the measurement, the assembly shown in the right of Figure 1 was placed directly inside the detection volume of the electrostatic radon monitor [12], whose working principle is shown in Figure 2. The four-liter hemispheric detector volume is filled with nitrogen at a slight over-pressure of 1050 mbar and is instrumented with a Si-PIN diode mounted to the top flange.By applying a -1 kV bias voltage, the positively charged radon progeny ions are collected at the surface of the diode.Alpha particles emitted during their subsequent decays can then be detected, in case they are directed towards the diode.Figure 3 shows the alpha-particle energy spectrum of the collected radon daughters, with clear emission lines from 216  Rn progeny from the source, measured in the electrostatic radon monitor (black) over the four-day-long measurement period.Note that the emission line of 220 Rn itself is not visible, as the detector relies on the collection of charged radon progeny. 212Bi branching ratio, the rate of 216 Po (blue hatched region) is found to be suppressed.This is a consequence of the isotope's short half-life (145 ms), which is of the same order of magnitude as the median duration of the collection process.Therefore, a certain fraction of the 216 Po ions disintegrates along their way, before reaching the Si-PIN diode.Note that this does not affect the even shorter lived 212 Po, as its progenitor isotopes are already collected on the diode.The subdominant peak at 10.8 MeV is compatible with the rare "long-range alpha" transitions between To prevent the uncertainty of the 216 Po rate due to its in-drift-decay, the 220 Rn emanation rate of the source is determined solely from the equilibrium rate of 212 Po decays.They are selected within the green hatched area shown in Figure 3.The fraction of events falling outside this selection is estimated from a fit to spectrum and amounts to 2%.Due to the 11 hour-long half-life of the preceding 212 Pb isotope, the activity evolution of 212 Po features a time-delayed increase towards its equilibrium value.Therefore, its initial activity  init 212 Po that is determined from a time interval at the beginning of the measurement, needs to be up-scaled by the fraction  eq 212 Po to which its equilibrium activity has been reached (see Equation 3.1).We determine this fraction from an analytical calculation [17] taking into account the complete decay chain dynamics.To reduce the impact of local charge accumulation caused by the high activity of the source, a time interval of approximately 6 hours ( 1 = 20 min,  2 = 240 min) was chosen from the four-day-long measurement.
Due to the steep increase of the 212 Po rate at the beginning of the measurement, a delay between the start of the radon progeny collection and the beginning of data acquisition would lead to a variation of up to 8% of the final result.This was estimated by shifting the selected time-window [ 1 ,  2 ] by ±5 minutes in either direction.
Finally, the total detection efficiency for 220 Rn has to be taken into account.Since a 220 Rn reference source was unavailable, it could not be measured directly.Therefore, the approach that has been applied in [8] was followed, and the efficiency was estimated using the one for the radon isotope 222 Rn.The latter has been measured for our detector under similar conditions to be  222 Rn 214 Po = (35 ± 2)% [18,19], when using the detected rate of the 222 Rn progenitor isotope 214 Po exclusively.Since 212 Po and 214 Po appear on the same location along their respective decay chain, their collection probability should be very similar.After accounting for the 64% branching fraction of 212 Bi (see decay chain in Equation 1.1), a detection efficiency of  220 Rn 212 Po = (22.4± 1.3)% for 220 Rn can be assumed.We attribute an additional uncertainty of 1.4% accounting for field-free regions generated by the presence of the 220 Rn source (e.g.below the source assembly), from which no ions can be collected.
The 220 Rn emanation rate  220 Rn of the source is then given by and a value of  220 Rn = (8.2± 0.8) kBq is obtained.The uncertainty corresponds to the squared sum of the individual contributions discussed above, with the statistical uncertainty being found to be negligible.The measured radon emanation efficiency is 15%, which denotes the fraction of the 220 Rn emanated from the source over the total amount produced by the decay of 228 Th.This result is compatible to the one of the radon source in Ref. [20], 3 times more than the source produced by PTB (Germany) for XENON1T [8], and 3 orders of magnitude higher than natural thorium compound sources [21].This improvement is attributed to the increased surface area, achieved by distributing the total activity among four individual discs, by which a more efficient radon release is obtained.

222 Rn emanation of the source
Due to the long half-life of 222 Rn (T 1/2 = 3.8 days), an excessive emanation of this isotope to the XENONnT detector must be prevented.To measure the 222 Rn emanation rate, the source vessel has been filled with helium and was left for accumulation of the radon for several days.Then the gas was transferred into an evacuated 20 liter expansion vessel, in which the short-lived 220 Rn was left to decay to a negligible level over a time of 4.5 hours.The remaining 222 Rn was then extracted, purified and collected by an activated carbon trap that is cooled to liquid nitrogen temperature [22].This radon is subsequently transferred along with a counting gas mixture consisting of 90% argon and 10% methane (P10) into a miniaturized proportional counter [22,23].Although these highly radio-pure counters do not allow for a full spectroscopic reconstruction of the radon alpha spectrum, alpha decays can be clearly separated from background events induced by muons, betaor gamma-rays.Thus, we use these detectors as simple counters with a fixed energy threshold of 50 keV.We calibrate the total 222 Rn transfer and counting efficiency with an acidic 226 Ra standard solution producing a well known amount of 222 Rn.From all alpha decays of the 222 Rn decay chain, (1.48 ± 0.06) alpha events are recorded for each radon atom, on average [22].
Two such measurements were performed, consistently revealing a 222 Rn emanation rate of (3.62 ± 0.14) mBq.Given that the previous 228 Th source produced by the PTB (Germany) did not show any significant 222 Rn emanation above < 50 µBq [8], this finding was unexpected.Given that Ref. [20] reported a comparable 222 Rn emanation from a similar source manufactured by Eckert & Ziegler, it is likely that the 222 Rn is released by trace impurities introduced during the production process.Under the assumption that both radon isotopes have the same emanation fraction, the level of 226 Ra impurity in the 228 Th deposit can be estimated to be 350 ppm.
While the 222 Rn emanation of this source corresponds to about 10% of the total 222 Rn emanation of the XENONnT experiment [11], its effect is expected to become subdominant approximately one week after the end of the 220 Rn calibration.This waiting time is further reduced by the radon removal system of the XENONnT experiment [24] and will thus not appreciably impact the data-taking efficiency.For future large-scale experiments, such as DARWIN/XLZD [25,26] and nEXO [27], however, this might not necessarily be true due to their more demanding background requirements.Therefore, it will be worth investigating possibilities to identify the origin of this impurity in order to mitigate it in future source productions.

Exclusion of 228 Th release
Contamination due to 228 Th and its decay products must be avoided in all components in the XENONnT experiment.Therefore, it must be guaranteed that no thorium is released from the source.To test this, the source was flushed for 9 days with argon at a flow rate of 700 SCCM using the setup sketched in Figure 4. Non-gaseous components in the argon stream are collected by two 0.2 µm PTFE membrane filters.To check for potential contamination with 228 Th, both downstream filters (F1 & F2) were then measured using high purity germanium (HPGe) spectrometers located underground at a shallow depth of 15 meters of water equivalent [28].The activity on both filters was evaluated via the 228   is initially dominated by 212 Pb accumulated in the plate-out of 220 Rn progeny.This contribution quickly decays with a half-life of about 11 hours (dashed light blue line in Figure 5).Afterward, the evolution follows the decay of 224 Ra, which is released via recoil from the source and has a half-life of 3.6 days (dashed medium blue line).Finally, a contribution from 228 Th would cause a persisting, constant activity.As there is no detectable activity after 42 days, an upper limit of ≤ 1.7 mBq at 90% C.L. can be placed as indicated by the hatched dark blue area.This is sufficient given the requirements of the XENONnT experiment.The activity evolution of the backing filter (F2) is found to follow the decay of 212 Pb without any indication for 224 Ra.This indicates that all non-noble gas components have been collected already by the main filter (F1).Only the 220 Rn progeny produced in between F1 and F2 are then collected on the backing filter.

Summary
Internal calibration sources are an essential tool to calibrate large-scale liquid xenon detectors used for rare event searches.In this work, we have characterized a new 228 Th source, which has been thereafter installed and successfully applied in the XENONnT experiment.This source produces gaseous 220 Rn, that can be introduced and mixed into the liquid xenon and allows for a homogeneous calibration via the low energy electronic recoil signals from the decay of 212 Pb.The new source consists of four commercially available 228 Th-plated discs (see Figure 1).This design showed an enhancement of the 220 Rn emanation rate as compared to similar sources applied in other experiments [8,21].Though an unexpectedly large 222 Rn emanation rate has been found, the source is still compatible with an application in the XENONnT experiment.Furthermore, the release of the long-lived 228 Th from the source was excluded experimentally.We summarize the results of our measurements in Table 1.Gaseous sources providing 220 Rn emanation will be an important tool also for the next generation of liquid xenon detectors such as nEXO [27] and DARWIN/XLZD [29].

Figure 1 .
Figure 1.Left: Individual 25.5 mm-diameter platinum disc with 228 Th oxide deposition visible in the center.Right: Final assembly of the four source discs on a CF-50 blind flange using M6 threaded rods and nuts.The active sides of the sources are facing each other with spaces to maximize the radon emanation.

Figure 2 .
Figure 2. A schematic drawing of the electrostatic radon monitor.Positively charged radon progeny particles are attracted by an electric field toward the Si PIN diode, which acts as an alpha-detector.Figure adapted from [13].

Figure 3 .
Figure 3. Energy spectrum of electrostatically collected220 Rn progeny from the source, measured in the electrostatic radon monitor (black) over the four-day-long measurement period.Note that the emission line of 220 Rn itself is not visible, as the detector relies on the collection of charged radon progeny.

Figure 4 .Figure 5 .
Figure 4. Sketch of the setup used to certify that no228 Th is removed from the source.Any potential228 Th removed from the source is carried with the argon purge and collected on either of the two downstream filters (F1&F2).After 9 days of continuous flushing, the activity on both filters is measured using high-purity germanium (HPGe) spectrometers.
Th progeny 212 Pb.For this a weighted mean of its own gamma emission

Table 1 .
Total 228 Th activity of the new XENONnT source as provided by the supplier with results from the characterization measurements carried out in this work.The elevated 222 Rn emanation is likely caused by trace impurities introduced during source manufacturing.