Development of highly radiopure NaI(Tl) scintillator for PICOLON dark matter search project

The re-crystallization (RC) method helps remove the radioactive impurities that are well soluble in water. The solubility of NaI in water is 75.14 at 100 ^∘C, and it decreases 64.76 at 25 ^∘C. The solubility is defined as the mass of solute in 100 g of water solution. The difference in the solubility between 100 ^∘C and 25 ^∘C enables us to get the pure NaI sediment.

The COSINE group showed the significant ⁴⁰K suppression in the RC methodShin2018 . Potassium in the aqueous solution of sodium iodide forms a well soluble potassium iodide (KI) or potassium hydroxide (KOH). The solubility of KI and KOH in water are as high as 59.7 and 54.2 at 25 ^∘C, respectively, as shown in Table 1. Since the concentration of potassium is much lower than its solubility, potassium ions remain in the water during the NaI RC process. The COSINE group reported the effectiveness of the RC method Shin2018 ; Park2020 .

We verified effectiveness of the RC method by measuring the ⁴⁰K activity in NaI before and after this procedure. The activity of ⁴⁰K in the NaI was determined using an ultra-low background HPGe detector which is installed at the KamLAND underground facility of the Kamioka neutrino observatory. The HPGe detector was constructed by joint efforts of Kavli IPMU (Tokyo Univ.) and RCNS (Tohoku Univ.) research groups. More information about the HPGe detector and underground clean room facility can be found in Kozlov2020 ; Kozlov2019 . We found an apparent reduction of gamma-ray intensity by the RC method. The initial radioactivity of ⁴⁰K in the original (non-purified) NaI powder was in the range of $18\sim 32$ mBq/kg at 90% CL. After a single RC cycle, the activity of ⁴⁰K was reduced to $0.2\sim 10$ mBq/kg at 90% CL.

We prepared 100 ^∘C NaI saturated solution in a bottle filled with pure nitrogen. The bottle was slowly cooled down to room temperature to avoid adhering of NaI crystals to the bottle’s surface. The NaI crystals were separated from the water solution by suction filtration. The filtration was done in a glove box flushed with pure nitrogen to prevent contact with air containing radon.

The COSINE group reported the re-crystallization worked well to reduce ²¹⁰Pb Shin2018 ; Park2020 . Our results also showed a significant reduction by double and triple RC methods, as listed in Table 6. Nevertheless, we need a higher reduction to get a pure NaI(Tl) crystal below a few tens of $\mu$Bq/kg. We tried to combine another strategy to remove Ra and Pb ions.

We extensively searched for the source of ²¹⁰Pb and found it was not only ²²²Rn in pure water but also ²¹⁰Pb in original NaI powder. We combined several methods to remove both ²²²Rn and ²¹⁰Pb from the water solution of NaI. The gaseous ²²²Rn in water was removed by bubbling the pure water with pure nitrogen gas. The longer bubbling, typically one hour, gives a significant reduction. We tried to catch Pb in the water solution of NaI by both cation exchange resin and crown ether.

We searched for the best combination of the resins by processing the NaI aqueous solution, which was initially added 4.9 ppm of lead ion. We tested four resins delivered by two companies. The resin A is suitable to remove strontium, uranium, and lead ions. The resin B is designed to remove the lead ion. The resins C and D, which another company delivers, are applied to remove uranium and iron ions.

We measured the concentration of the Pb ion in the sample by ICP-MS, Agilent 7900 at Osaka University and Agilent 7700 at Osaka Sangyo University. The measured values of the reduction factor of lead ion are listed in Table 2. According to the result, we selected the resin B to reduce lead ion. In addition to resin B, we apply the resins A to reduce not only ²¹⁰Pb but also U, Ra, and Th.

We took necessary precautions to avoid contacts between the room’s air and NaI powder during drying process. We dried the purified NaI solution by a rotary evaporator. We selected a flask made of synthetic silica glass to avoid the possible pollution from it. We filled pure nitrogen gas into the flask to break the vacuum when the drying finished. Additional drying was done by using a vacuum oven before growing NaI(Tl) crystal.

The large volume NaI(Tl) crystal was produced by the Bridgeman-Stockbarger technique. The purified and dried NaI powder and thallium iodide were filled into a high-purity graphite crucible. Certain details describing purification of the crucible material can be found in Kozlov2020 ; Kozlov2019 . The crucible was installed in a quartz vessel filled with inert gas at the center of the furnace. The crucible was heated to 700 ^∘C to melt the NaI powder, and it was slowly moved to crystallize. The crystallization takes at least two weeks of cooling and annealing to make a transparent, colorless, and large single crystal without any inclusion. The melted NaI does not interact and infiltrate the pure graphite crucible because of the special coating on the surface of the crucible. In the end of crystal formation the NaI(Tl) ingot detaches from the crucible. We see no evidence of interaction between the crucible and melted NaI. We think that there is no significant pollution of the NaI(Tl) ingot occurs by the crucible material.

The original NaI(Tl) ingot was used to make a NaI(Tl) crystal shaped as a cylinder (a 76.2 mm in diameter and 76.2 mm long). An optical window made of synthetic quartz with 10 mm in thickness was attached on one side of the crystal edge. Other surfaces of the crystal were covered with an enhanced specular sheet (ESR) made of polyethylene. The whole crystal was encapsulated into a 4 mm-thick black acrylic container. We used the acrylic housing to minimize absorption of low energy gamma and X-rays emitted by external calibration sources.

An acrylic housing enables a low energy calibration down to 6.4 keV_ee. The K_β X-rays emitted after the electron capture of ¹³³Ba are between 34.9 keV and 35.8 keV with total intensity of 11.5 % TOI . After the photoelectronic effect on iodine, the ionized iodine emits its K_α X-ray, 28.6 keV and 28.3 keV TOI . If these K_α X-rays escape from the crystal a residual energy of approximately 6.4 keV is deposited in a NaI(Tl) scintillator. This low energy peak is useful to test the stability of low energy calibration.

Acrylic housing has a risk of moisture permeability. We tested deliquesce when the NaI(Tl) detector was installed in the shield with pure nitrogen. The NaI(Tl) detector has not suffered any deliquescence for at least 1.8 years.

We made several trials of purification. The details of the NaI(Tl) detectors are listed in Table 3. Hereafter, we identify the NaI(Tl) detector by crystal ID shown in Table 3.

We measured activity of radionuclides remained in the crystals #68, #71 and #73 using an ultra low-background setup Kozlov2019 at the KamLAND underground facility. The radiopurity of two other crystals (#24 and #85) was measured in a surface laboratory at Tokushima University.

In Kamioka, the DAQ system consists of VME based MoGURA electronics Terashima2008 produced by Tokyo Electron Device Ltd. Several supplementary NIM modules were used to reject majority of noise pulses from a photomultiplier tube (PMT) Kozlov2020 . The MoGURA on-board flash analog-to-digital converters (FADC) can be used to record up to 10 $\mu$sec long waveforms. Signal selection as well as rejection of the remaining noise pulses from PMT were done by offline pulse shape analysis (PSD).

In the surface laboratory, the current pulse from a PMT was divided into three routes and digitized by CAMAC ADC Ichihara2003 . The one was used for a trigger. We introduced one current signal of the PMT to a charge-sensitive ADC (CSADC: REPIC RPC-022) channel through a 200 ns cable delay. The CSADC integrates the total charge of the current pulse for 1 $\mu$s. The other current signal was introduced into the other ADC channel directly to integrate it partially.

The PSD was performed to extract the alpha-ray event. The decay time of the NaI(Tl) for an alpha-ray, $\tau_{\alpha}$ is 190 ns, on the other hand, the one for a beta or gamma-ray, $\tau_{\beta}$ is 230 nsBernabei2008 . This difference enables clear discrimination of an alpha-ray and others. The parameter $R$ was calculated as

where $I(t)$ was the input current. The upper limit and the lower limits of the integration are defined as $t_{2}$, $t_{0}$, and $t_{1}$, respectively. The present values of $t_{0}$, $t_{1}$, and $t_{2}$ were listed in Table 4.

We analyzed dependence of the concentration of ⁴⁰K in the NaI(Tl) detectors #68, #71, and #73 on various purification methods listed in Table 3.

We measured the concentration of ⁴⁰K in the NaI(Tl) crystal at Kamioka Underground laboratory, Tohoku University. The ⁴⁰K decays both beta decay (89.26%) and electron capture (10.72%). We need a low background environment to measure the tiny amount of radioactivity because of the enormous external background. The Kamioka Underground laboratory is suitable to reduce the background from cosmic rays. To suppress environmental gamma-rays, the detectors were placed inside of an ultra-low background passive shielding made of a $15\sim 20$ cm-thick lead and a 5 cm-thick 99.99% pure copper layers. Structure and composition of the passive shielding were also reported in the earlier publications Kozlov2020 ; Kozlov2019 . Figure 1 shows crystal #73 surrounded by the copper shielding layer. An ultra-low background PMT, Hamamatsu Photonics 4-inch R13444X Kozlov2020 , was attached to the optical window of the NaI(Tl) crystal by an optical grease. A pure nitrogen gas was supplied into the shield to purge out air containing radon.

The MoGURA data acquisition system recorded the pulse shape of the PMT output. The pulse shape analysis was done, as shown in the left panel of Figure 2. The Energy-Ratio distribution makes three loci. The PSD distribution shows the significant distortion above 4000 keV_ee due to the saturation of PMT output. The beta/gamma-ray energy spectrum was obtained without distortion to check the existence of the beta-ray ($E_{\beta\mathrm{max}}=1311.09$ keV) and gamma-ray ($E_{\gamma}=1461$ keV) TOI . We selected the beta- and the gamma-ray events by PSD analysis. The energy spectra observed by the crystals #68 and #71 are shown in the right panel of Figure 2. There was a clear difference in the energy spectra between crystals #68 and #71. There was a prominent continuum below 1.3 MeV_ee of crystal #68 due to the beta-rays of ⁴⁰K. The prominent beta-ray spectrum insisted that the ⁴⁰K was contained in the NaI(Tl) crystal of #68. The concentration of ^natK in the crystal #68 was calculated as 130 ppb by Geant 4.10 Monte Carlo simulation, assuming the natural abundance of ⁴⁰K in natural potassium is $0.0117$%. On the other hand, there were no significant components of both beta-ray and gamma-ray from ⁴⁰K in the energy spectrum of the crystal #71. The beta-ray event rate in crystal #71 was at least 1/6 smaller than in crystal #68. It corresponds to the concentration of ^natK was less than 20 ppb at 90% C. L.

The reproducibility and the effectiveness of the triple RC method were tested by ingot #73. The upper limit of ^natK was 30 ppb at 90% C. L. We found that the double RC method was enough to reduce potassium contamination. We confirmed the double RC method’s effectiveness by later ingots #76 and #83.

We measured the alpha concentration in purified NaI(Tl) crystals. After establishing the potassium reduction, we optimized the reduction method of alpha-ray emitters. The present crystal #85 was made by the combined method of double RC and resins. We also measured the alpha-rays in #24 to verify the reduction.

We accumulated the data for the live-time of 36.33 day$\times 1.279$ kg for #85 and 11.2 day$\times 1.279$ kg for #24. The result of the PSD analysis with the crystal #85 is shown in Figure 3. It shows the faint concentration in the region $0.4<R<0.45$. A large locus around $R=0.53$ is the events of beta-rays, gamma-rays, and cosmic-rays. We compared the energy spectrum of alpha-rays with one of our previous best NaI(Tl) crystal, #24 fushimi2014kamlandpico . The right panel of Figure 3 is the energy spectra taken by crystals #24 (red) and #85 (black). The horizontal axis is the electron equivalent energy. There was a considerable improvement in the event rate in the present work. The crystal #24 contains $163\pm 33$ $\mu$Bq/kg of ²²⁶Ra, $149\pm 11$ $\mu$Bq/kg of ²²⁸Th, and $86\pm 25$ $\mu$Bq/kg of ²¹⁰Pb. We analyzed the constituent of the alpha-ray spectrum of crystal #85 as described below.

The alpha-ray energy spectrum consists of two components (see Figure 4). One is the lower energy region (below 4000 keV_ee), and the other is the higher energy region (above 4000 keV_ee). Both components have no prominent structure. We considered that the events above 4000 keV_ee are an accidental spread of $R$ of high energy cosmic rays. We assumed there is a constant background of the cosmic-ray events below 4000 keV_ee. The energy spectrum below 4000 keV_ee consists of constant cosmic-ray, alpha-ray emitted in the NaI(Tl) crystal, and alpha-ray emitted on the ESR reflector sheet.

We measured the surface concentration of alpha-ray emitters on the ESR sheet to determine the alpha-ray intensity. We applied the low-background and position-sensitive alpha-ray tracker, $\mu-$PICHashimoto2020 ; Ito2020 . No significant excess beyond the background was observed, as shown in Figure 5. The upper limit of the surface contamination was calculated as $1.77\times 10^{-3}$ alpha/hr/cm² (90% C. L.).

We simulated the energy spectrum from the ESR sheet by Geant4.10 Monte Carlo simulation Allison2016 . We assumed the origin of the alpha-ray was ²¹⁰Po because the ESR sheet attracts the progeny of ²²²Rn. ²¹⁰Pb terminates the decay chain from ²²²Rn, and accumulated ²¹⁰Pb generates the ²¹⁰Po. A red line in Figure 4 shows the calculated energy spectrum of ²¹⁰Po from the ESR sheet. There is significant excess in the event rate in the low energy region above the red line. We concluded the excess is due to the alpha-rays emitted in the NaI(Tl) crystal.

The energy spectrum of alpha-ray from the NaI(Tl) crystal showed no prominent structure. The event number of alpha ray was too small to perform further analysis such as time-correlation analysis. We calculated the yield by integrating the number of events instead of the peak fitting. We set the integration interval according to the quenching factor of alpha-ray and energy resolution. The alpha-rays which have near energies make a cluster in the energy spectrum because of low energy resolution. We analyzed the alpha-ray intensity according to the clusterized energy ranges was calculated according to the quenching factor of alpha-ray in NaI(Tl) scintillator fushimi2014kamlandpico . The energy listed in Table 5 is the FWHM (Full Width Half-Maximum) interval of the clusterized peak. The chemical symbol in the parenthesis indicates the decay chains of uranium and thorium. The electron equivalent energy of alpha-ray, $E_{\mathrm{ee}}$, is derived by

Where $f_{\alpha}$ is the quenching factor of alpha-ray. We determined $f_{\alpha}=0.58$ from the shape of the energy spectrum.

The alpha-ray of ²²⁴Ra lies between (C) and (D) clusters. The alpha-rays of ²²⁴Ra contributes both event rates of (C) and (D) because of low energy resolution. We divided two clusters at the peak energy of the alpha-ray of ²²⁴Ra. The event rates of the clusters (C) and (D) contain half of the ²²⁴Ra events. We confirmed the result of this analysis by the data from crystal #24. The present analysis agrees with the result by peak fitting.

The concentration of radioactivity was calculated from the event numbers in Table 5 and the total exposure 36.33 day$\times$1.279 kg. We assumed that all the radioactivities of isotopes in the decay chain of thorium were in secular equilibrium. The isotopes of uranium-series, ²²⁶Ra, ²²²Rn, ²¹⁸Po, and ²¹⁴Po were in secular equilibrium. We did not include ²¹⁴Po in the present analysis since we cannot take all the events of alpha-rays from ²¹⁴Po. The half-life of ²¹⁴Po is as short as 164 $\mu$sec, and the dead-time of the present data acquisition system was as long as several hundreds of microsecond. The concentration of ²¹⁸Po was easily derived from the result of (D) as $14\pm 4$ $\mu$Bq/kg. The one of ²²⁶Ra was derived from the result of (B) as $13\pm 4$ $\mu$Bq/kg; both values are consistent with each other.

No significant excess was found in the concentration of ²¹⁰Po. We set the upper limit on the concentration as 5.7 $\mu$Bq/kg at 90%C. L.