Ultra-low background germanium assay at the Boulby Underground Laboratory

As we move to an era where next generation ultra-low background particle physics experiments begin to be designed and constructed, the ability to assay materials with high sensitivity and at speed with a variety of techniques will be key. This paper describes the Mirion Technologies (Canberra) specialty ultra-low background detectors installed and commissioned at the Boulby Underground Laboratory between 2017 and 2021. The low background levels of the detectors combine with low background shielding and a radon-reduced dry nitrogen purge system to give sensitivity approaching the best in the world without the need for intricate shielding solutions. For an optimised sample geometry, run for 100 d, it would be possible to reach close to 10 μBq kg-1 (10-12 g/g) for background radionuclides of interest in neutrinoless double-beta decay.


Introduction
The Boulby Underground Laboratory is located in the north-east of England at Boulby Mine.The laboratory is at a depth of 1100 m (2840 m water equivalent) in a salt strata deposited some 250 million years ago when the Zechstein sea evaporated.The Boulby UnderGround Screening (BUGS) Facility has been operational since 2015 and has been important in the material characterisation efforts of several leading low-background particle physics experiments [1,2] and a number of environmental studies [3][4][5].The original detectors which still operate in the facility have previously been discussed in depth [6].In 2017, the facility began a process of upgrade to meet the material characterisation requirements of next generation dark matter and neutrinoless double-beta decay experiments [7].This process included the purchase of three specialty ultra-low background (S-ULB) detectors developed by Mirion Technologies (Canberra).Following the success of the original detectors, it was decided that the same philosophy should be employed to provide sensitivity from a few keV to 3000 keV to allowing characterisation of gamma-rays from 210 Pb (46.5 keV) through to 208 Tl (2614.5 keV).No single detector is suitable for maximising sensitivity across the whole energy range so a combination of Broad Energy Germanium (BEGe) and standard p-type coaxial germanium detectors were acquired.In addition, the Lumpsey SAGe-well detector was previously shown to have a background level which, while in specification, substantially reduced its potential sensitivity.This detector was returned to Mirion Technologies (Canberra) for refurbishment to the S-ULB standard.BUGS also operates two XIA UltraLo-1800 surface alpha counters, a dual detector low background radon emanation facility and an Agilent-8900 ICP mass spectrometer [8].The addition of these assay techniques prompted the change in the acronym for BUGS (formerly the Boulby Underground Germanium Suite).

The Detectors
For the manufacture of S-ULB detectors, careful selection is made of all detector materials that will be located inside the lead shield.All materials close to the High Purity Germanium (HPGe) crystal are selected based on their activity to minimise the impact on the detector background.Screening of electronic components, and material such as aluminum and copper together with a complete control of all stages of the detector manufacturing process is key to the consistent and reproducible level of background.Furthermore, any material that is susceptible to cosmogenic activation is stored underground (800 meters water equivalent) to minimise the presence of potentially long half-life contaminants.This procedure was applied to Rosberry, Belmont, Merrybent and Lumpsey.There is no compromise between low radioactivity material and spectroscopic performance when compared to the standard version of each detector.
Roseberry -Roseberry is a Mirion Technologies (Canberra) BEGe BE6530 detector.This planar detector has a 65 cm 2 area on its front face and a 30 mm thickness.The crystal is held in an ultra-low background cryostat and is shielded from any gamma-rays coming from the crystal holder and high voltage feedthrough by a thin disk of low-background lead.In addition to selecting low-radioactivity material, the mass of material in the detection head is optimised such that it only contains essential components.In addition, a thin outer contact is used on the HPGe to have a very thin (∼µm) residual dead-layer.This minimises low-energy gamma-ray absorption.Figure 1 shows the internal configuration of the Roseberry detector with the low-background lead disk visible.Belmont -Belmont is a 600 cubic centimetre p-type coaxial high-purity germanium detector.A number of the materials used to construct the cryostat for this detector were assayed using the original BUGS germanium detectors to ensure as low a background as possible.With such a high relative efficiency, Belmont has the potential to assay samples with a precision and sensitivity far beyond that of any other detector in the BUGS germanium suite.Merrybent -Merrybent is a 375 cubic centimetre p-type coaxial high-purity germanium detector.Although smaller than Belmont, this detector is still able to assay samples with high efficiency and has a background far lower and a performance far better than the only other equivalent detector, Lunehead [6].
Both Belmont and Merrybent have large masses of HPGe (3.2 kg and 2.0 kg, respectively) but still provide excellent energy resolution meaning the size of the crystal does not compromise performance.In addition, freshly pulled high purity germanium crystals were used to limit the cosmogenic activation of the sensitive element of these two detectors.Lumpsey -Lumpsey is a SAGe well detector.Previously configured in a standard cryostat, it was not able to meet our sensitivity requirements for a number of materials.To maximise our sensitivity to low energy gamma-rays for small samples, this detector was refurbished to the S-ULB standard by Mirion Technologies (Canberra).The refurbishment of the detector followed the same standard as any S-ULB detector with a careful selection of material and electronics.As with any SAGe-Well detector, a particular focus was applied to select proper materials for the high voltage capacitor.In addition, the original configuration of this detector included a lithium drifted contact in the well.This left the detector susceptible to damage in the event of a warm up which could happen if access to the underground laboratory were to be restricted.The lithium contact was removed and replaced with a stable thin layer contact which minimises the absorption of low energy gamma-rays.This process involved increasing the diameter of the sample well in the germanium crystal from 28 mm to 31.2 mm meaning that samples ∼ 20% larger can be assayed when compared to the previous configuration.Finally, the refurbishment marks the first BUGS germanium detector to be electrically cooled.An internal schematic, modelled in MCNP, is shown Figure 2 which shows the characteristic shape of the detector crystal.

Shielding
An important requirement of running the BUGS germanium detectors is to have a high level of confidence in the quality and stability of detector backgrounds.The first factor which determines this is the quality of the shielding material which surrounds the detectors.The shields for these detectors were produced by Lead Shield Engineering.All the detectors described above use multilayer shielding with the inner layer comprising 10 cm of high-purity copper and the outer layer comprising 10 cm of lead.In addition, the shield surrounding the Belmont detector includes a 10 mm ultra-pure copper liner.As with the original BUGS germanium detectors, this lead and copper was sourced from stock which had previously been used to shield former low background experiments hosted at the Boulby Underground Laboratory.

Nitrogen Purge
The second requirement for ensuring the quality and stability of the background in a germanium detector is to control the atmosphere which surrounds it.The Boulby Mine ventilation system consists of a single intake down one of the access shafts.The air is circulated around the mine and then is exhausted up the second access shaft.Fig 3 shows the monthly averages of radon detected both on the surface, in the main Boulby Underground Laboratory and in the BUGS facility during 2022.It can be seen that the underground radon levels mirror those seen on the surface.Over this measurement period, the average radon levels are (2.10 ± 0.03) Bq/m 3 , (2.23 ± 0.03) Bq/m 3 and (2.09 ± 0.03) Bq/m 3 for the surface, main lab and BUGS, respectively.This agreement shows that, with adequate ventilation, the radon levels measured underground are identical to those on the surface.This means that the radon emanation rate from material surrounding the laboratory is sufficiently low such that it is effectively mitigated by the intake ventilation air.The rock surrounding the facility has relatively low levels of uranium when compared to similar underground facilities [9].Daily averages show a variance of approximately 1.1 Bq/m 3 .Figure 4 shows the radon increase measured in the laboratory when the main ventilation fans for the mine are turned off for routine maintenance.It can be seen that the radon levels in the laboratory increase to approximately 60 Bq/m 3 .This figure also shows the related increase in detector background for a sample running concurrently which peaks at approximately seven times higher than before and after the spike.Germanium data are taken in hour long snapshots so any periods with substantially higher radon levels can be identified and removed.Any radon which enters the cavity of a germanium detector shield will cause unwanted background, particularly in spectral gamma-ray peaks associated with 214 Pb and 214 Bi which are products of the decay of 222 Rn.To prevent the ingress of radon, a positive pressure of pure N 2 gas is used to purge the air from the cavity.This purge is often performed using boil-off from liquid nitrogen dewars.As the boiling point of radon (−61.7 °C) is substantially higher than that of liquid nitrogen (−195.8°C), it stays liquid meaning that pure nitrogen is produced.In principle, this would be the ideal source of purge gas but in practice it can be difficult to maintain constant pressure and, therefore, flow rate.
Given the overall low levels of radon observed at the Boulby Underground Laboratory, we were confident that it would be possible to purge detectors with nitrogen gas generated on site.For the original BUGS germanium detectors, a simple system with a combination of Noblegen NG5 and NG6 nitrogen gas generators was used.This achieved acceptable results for the assay program for LZ [1] but, as part of the upgrade, we wanted to ensure that the purest possible purge gas was available.Although the radon levels at Boulby are low, some impurities still remain in the initial nitrogen purge.To reduce this further, the gas is dried and passed through a charcoal trap held at −80 °C.To ensure that a constant flow of nitrogen is seen by each detector, the purge is controlled using Bronkhorst El-Flow Select mass-flow controllers which can flow at 0.2-10.0litres per minute (±0.6% precision).This allows the identification of any periods with sub-optimal purge which, in This increase is mirrored in data from a sample running concurrently in the Roseberry detector.Acquisition is split into hourly slices so data with higher rates can be easily identified and removed.
turn, allows the removal of affected data.

Backgrounds
In early 2021, the detectors were running with the nitrogen purge operating in a stable and monitored condition with no additional improvements planned.By means of comparison, Figure 5 shows the background levels for standard sub-222 Rn lines in the Belmont detector with three configurations.Firstly shielded but with no purge, secondly with the previous purge configuration (no radon reduction and minimal flow control) and finally using the radon reduced nitrogen flowing at 3 litres per minute through an MFC. Figure 6 shows a comparison between the backgrounds measured in Chaloner (a standard BE530 detector) and Roseberry.As expected, the background in Roseberry is substantially below that measured in Chaloner even with identical radon-reduced nitrogen purges.Finally, a comparison between Lumpsey pre-and post-refurbishment is presented in Figure 7.In its previous configuration, the background was high enough that the nitrogen purge made no appreciable difference to the measured background.We see that the Lumpsey background in the new configuration has a substantially lower background, with an overall 70 times reduction.Table 1 summarises the count rates for several gamma-ray peaks in the detectors presented in this study and in the original BUGS germanium detectors [6] now running with the optimised radon-reduced nitrogen purge.The background counts in Chaloner and Lunehead are dominated by impurities in the detector and shield construction rather than from airborne radon, so are not affected by improvements in the N 2 purge.A measurement for Lumpsey in its previous configuration is also presented, showing the substantial reduction in background rate.
In its current configuration, Merrybent is not achieving the lowest background possible for late chain 238 U due to what is believed to be contamination in the shield.Several assays with large, dense, and radiopure materials give counts in the 214 Pb and 214 Bi peaks at rates below the background rate [10].This infers that the elevated background does not come from the detector and the performance of the N 2 purge in Roseberry and Belmont (noting again that all detectors share a common source of N 2 for the purge) infers that there should not be an issue with the purity of the gas.The lowest rates seen in this detector occurred during assay of ultra-pure gadolinium sulphate where count rates of 1.1 kg/day and 0.6 kg/day were observed for the 351 keV and 609 keV lines, respectively.

Table 1:
Count rates for the Boulby HPGe detectors in their current configuration.These values may not represent the ultimate backgrounds that could be achieved with these detectors but this is for the very simple shield and purge configuration described in this paper.With the exception of the pre-refurbishment Lumpsey background, these runs were all performed in early 2021.The values for Lunehead, Chaloner and Lumpsey (pre-refurbishment) are broadly in agreement with the measurements discussed in [6].

Calibration
The detectors are periodically calibrated using a check-source which includes 1 µCi of 155 Eu and 1 µCi of 22 Na.This serves to monitor the stability of peak resolution and can be used to monitor dead-layer thickness in the Roseberry BEGe detector.For the former, peak widths are extracted from gamma-ray peaks at 86.5 keV and 105.3 keV from 155 Eu and at 1274.5 keV from 22 Na.The 511 keV e + e − annihilation peak from 22 Na may also be used although Doppler broadening means that this is not useful to monitor absolute resolution.To monitor the stability of the dead layer, the ratio of the count rate in the 86. to thicken, the area of the lower energy peak would reduce relative to the higher energy peak.A similar technique using a collimated 241 Am source is described in [12] although we only monitor the overall dead-layer thickness, not the position dependent thickness.Figure 8a shows the stability of FWHM resolution for the Roseberry detector for the calibration energies.Figure 8b shows the count rates over time, demonstrating the agreement between the decrease in counts per second and the half lives of 155 Eu and 22 Na which are 4.76 years and 2.60 years, respectively.Figure 8c shows the stability of the 105.3/86.5 peak area ratio over a period of 1.6 years from March 2021 to November 2022.This gives a constant ratio of 0.674 ± 0.001 with a reduced chi-squared of 0.87.

Geometric Efficiency and Minimum Detectable Activity
To provide a comparison between detectors, we employ a similar technique to determine geometric efficiency as described in [6].For the larger p-type detectors, sensitivity is maximised when the combination of sample mass and efficiency is maximised.For Belmont and Merrybent, the assay of a powder filled 4 litre Marinelli beaker (∼4.5 kg) is simulated.This is about the largest sample these detectors would encounter.In Figure 9, it can be seen that the best results are for the 212   (61 min) and 208 Tl (3.1 min), we can assume that these isotopes are in secular equilibrium and, as such, assume that a measure of the 238 keV line from 212 Pb will give a good prediction for the 2614.5 keV line from 208 Tl.
For Roseberry and Lumpsey, a smaller sample of 9 g is used.For both Roseberry and Lumpsey, a pot suitable for insertion in the well is also simulated.However, this is a sub-optimal geometry for the planar BEGe detector.To compare optimised assay configuration for an identical sample, the Roseberry simulation is repeated but with the 9 g sample in a petri dish.Figure 10 shows the two different configurations on the Roseberry detector.
In Figure 11, it can be seen that the highest efficiency overall for the 9 g sample is on Lumpsey.This figure also highlights that the petri dish sample on Roseberry gives a much better efficiency than does an identical mass sample in a well suitable pot.With both geometric efficiency and background count rates determined, it is possible to predict the minimum detectable activity (MDA) for U/Th/K in each of the detectors.To calculate MDA for each isotope, we calculate the critical limit of number of counts required above background to give a 90 % confidence level measurement.From this, we calculate the corresponding sample-specific activity [13].
Figure 12 shows the evolution of the MDAs with acquisition time for the 210 Pb line at 46.5 keV for the 9 g sample in the three configurations on Roseberry and Lumpsey.It can be seen that both detectors are able to assay to similar sensitivities as long as the geometry of the sample used is optimised.It is also seen that there has been approximately a 5 times reduction in the MDA for 210 Pb in Lumpsey when compared to pre-refurbishment.
Table 2 shows the MDAs that could be achieved for the new detectors and highlights the   1).The lines visible at 511 keV, 811 keV, 834 keV, 1173 keV and 1332 keV are due to cosmogenic activation of the copper crystal holder as discussed in [11].and in a pot suitable for the well on Lumpsey (blue) and Roseberry (red).Lumpsey gives moderately lower MDAs for this 9 g sample however, Roseberry can improve on these MDAs for larger samples.The black dashed line is the MDA for a large petri dish sample on Roseberry.For comparison, the equivalent MDA for Lumpsey pre-refurbishment is shown by the blue dashed line.There is roughly a 5 times reduction in the MDA.
0 events.This is also true for isotopes with a   end-point energy greater than the   energy of the double-beta decay isotope of interest.Of particular concern is 214 Bi as it has a   end-point energy of 3.27 MeV which is above most   energies.Table 3 shows the MDAs for isotopes which can cause unwanted background for 0 isotopes of interest.The measurement at 351.2 keV is for 214 Pb but this can be used to infer the level of 214 Bi assuming secular equilibrium.The calculated MDAs are comparable to those discussed in [14] showing that the BUGS HPGe detectors are approaching the current world leading sensitivities.The level of sub-222 Rn isotopes in detector construction materials is also of particular importance to an experiment such as Super Kamiokande Gd where ultra-pure water is loaded with gadolinium sulfate to allow diffuse supernova neutrino background (DSNB) searches.Even at low levels, 238 U, 235 U, 232 Th, and their progeny can adversely affect measurements of conventional physics targets such as solar neutrinos because the high energy gamma-ray emissions from radioactive impurities in the Gd sulfate may mimic detection signatures from neutron captures on Gd.The enhanced sensitivity of the BUGS detectors has aided the identification of gadolinium sulfate which meets the experimental requirements [15].

Conclusions
Germanium detectors of ever increasing sensitivity to samples of all shapes and sizes are going to be of key importance to material radioassay for next generation low-background particle physics experiments.The S-ULB detectors operating in the BUGS facility aim to meet the needs of these experiments by providing high sensitivity and high throughput measurements that can span the entire energy spectrum of interest to include measurements of the 210 Pb decay line at 46.5 keV all the way up to 208 Tl at 2615 keV.

Figure 1 :
Figure 1: A GEANT4 rendering of the Roseberry BEGe detector (left).The HPGe crystal is shown in grey and the ultra-low background aluminium crystal holder is shown in orange.Behind the crystal (shown in red) is the ultra-low background lead disk which reduces any potential background from the back of the detector.The lead disk used is also shown in the picture on the right.

Figure 2 :
Figure 2: Internal drawing of the Lumpsey detector.Key dimensions are included (in mm) and major components are labelled.In this drawing, ULB means ultra-low background and the symbol corresponds to the diameter.

Figure 3 :
Figure 3: Monthly averages for radon activities in Bq/m 3 for the surface, main Boulby Underground Laboratory and BUGS facility for 2022.It can be seen that the values are in reasonable agreement showing that there is no systematic increase in the radon levels underground as compared to the surface.The average values for the surface, main laboratory and BUGS are (2.10 ± 0.03) Bq/m 3 , (2.23 ± 0.03) Bq/m 3 and (2.09 ± 0.03) Bq/m 3 , respectively.

Figure 4 :
Figure 4: During routine mine ventilation fan maintenance, it can be seen that there is a substantial increase in radon levels in the underground laboratory (lower panel).The drop in pressure associated with the fans turning off (shown in the upper panel) allows air from deeper in the mine to travel back to the laboratory.This increase is mirrored in data from a sample running concurrently in the Roseberry detector.Acquisition is split into hourly slices so data with higher rates can be easily identified and removed.

Figure 5 :
Figure 5: Comparison between the 295 keV, 352 keV and 609 keV sub 222 Rn lines in the Belmont detector running with (black) no purge, (red) standard N 2 purge with no radon reduction and no MFC and (blue) the final purge setup with precisely 3 lpm of radon reduced N 2 delivered using an MFC.
Pb isotope which forms part of the 232 Th chain.Due to the short half lives of 212 Pb (10.6 hr),212 Bi keV/kg/day]

Figure 6 :
Figure 6: Comparison between the low background Chaloner (black) detector and the specialty ultra-low background detector Roseberry (red). keV/kg/day]

Figure 7 :
Figure 7:Comparison between Lumpsey running prior to refurbishment (black), and following refurbishment (red) with the final purge setup with precisely 3 lpm of N 2 delivered using an MFC.The overall background rate has reduced by a factor of 70 times but some individual peaks have reduced to far lower levels (see Table1).The lines visible at 511 keV, 811 keV, 834 keV, 1173 keV and 1332 keV are due to cosmogenic activation of the copper crystal holder as discussed in[11].

Figure 8 :
Figure 8: Calibration plots for a multi-isotope ( 155 Eu and 22 Na) calibration source on the Roseberry detector.

Figure 9 :
Figure 9: Comparison between the MDAs for the isotopes listed in Table3on the Merrybent and Belmont detectors.It can be seen that the sensitivity is generally better in Belmont with the exception of 40  where a lower count rate in Merrybent (despite a lower efficiency) for this line, leads to a better predicted result.

Figure 10 :
Figure 10: A 9 g powder sample in a petri dish (left) and a pot which is suitable for insertion in the well of Lumpsey (right).In this figure, both samples are shown sitting on the front face of the Roseberry detector.

Figure 12 :
Figure 12: Comparison between the MDAs for 210 Pb for a 9 g sample on Roseberry in a petri dish (black),

Table 2 :
Minimum detectable activities for each of the detectors in the BUGS facility.For Belmont, Merrybent and Lunehead the MDAs are for a 4 L Marinelli beaker filled with ∼4.5 kg powder.For Roseberry and Chaloner, a petri dish containing 100 g of powder is used.For Lumpsey, a pot containing 9 g of powder is used.For 238 U, 232 Th and 40 K, Lumpsey provides similar performance to Chaloner with a 90 % smaller sample.

Detector Mass (g) MDA at 21 days 609 keV 238 keV 1461 keV 46.5 keV 238 U (ppt) 232 Th (ppt) 40 K (ppb) 210 Pb (mBq/kg)
FWHM resolution measured for calibration lines on the Roseberry detector between March 2021 and November 2022.Statistical errors are included but are not visible on this scale.Counts per second for calibration lines on the Roseberry detector over the same period as seen in the figure above.The calculated half-lives are consistent with those for 155 Eu (4.76 years) and 22 Na (2.60 years).Statistical errors are included but are not visible on this scale.Ratio of the 105.3 keV and 86.5 keV lines from decay of the 155 Eu isotope in the multi-gamma calibration source.