The growth of the NaI(Tl) crystal (named NaI-037) was completed on 18 January 2021, using NaI powder purified at IBS [24, 32]. The top and bottom sections of the crystal ingot were cut using a diamond bandsaw, as shown in Figure 1B. After cutting the top and bottom, the NaI-037 crystal is 70 mm in diameter and 51 mm in height. The flat top and bottom surfaces and a barrel-shaped side surface were polished using aluminum oxide films ranging from 400 to 8,000 grit. After the polishing, several layers of soft polytetrafluoroethylene (PTFE) film were wrapped on the barrel surface as a diffusive reflector. 3 mm thick optical pads (silicon rubber EJ-560) were coupled to the top and bottom surfaces as an optical interface. A 3 mm thick copper casing with 5 mm thick quartz windows at each end was encapsulated the crystal hermetically. Hamamatsu 3-inch R12669SEL photomultiplier tubes (PMTs) are coupled via an optical gel (EJ-550) to each end of the crystal. It is the same encapsulation design that was used for previous NaI crystal developments in Reference [28]. It is also similar to encapsulation used in the COSINE-100 [13] and the ANAIS-112 experiments. The entire assembly was performed in a glovebox, where the humidity was maintained to be less than 10 ppm (H₂O) using Ar gas and a molecular sieve trap. Before the assembly, all parts were cleaned using diluted Citranox liquid with sonication and baked in a vacuum oven for more than 12 h. After assembly, the detector was delivered to the Yangyang underground laboratory (Y2L), which has ∼700 m of rock overburden [33]. From the crystal growth to Y2L delivery, it took less than 3 weeks and minimized the cosmogenic activation in the crystal.

The background contamination levels of the NaI-037 crystal were evaluated using the same experimental apparatus as that used for the NaI(Tl) crystal R&D at the Y2L [28, 34]. It includes an array of 12 CsI(Tl) crystals surrounded by 10 cm copper, 5 cm polyethylene, 15 cm lead, and 30 cm liquid scintillator-loaded mineral oil [35, 36] as a radiation shield. The detector was installed inside the CsI(Tl) array, as shown in Figure 2.

The PMTs attached to the NaI-037 crystal had two readouts each, a high-gain signal from the anode and a low-gain signal from the fifth-stage dynode. The anode signal was amplified by a factor of 30, whereas the dynode signal was amplified by a factor of 100 using a custom-made preamplifier. The amplified signals were digitized by 500 MHz, 12-bit flash analog-to-digital converters (FADCs) ¹ . Triggers from the individual channels were generated by the field-programmable gate arrays (FPGAs) embedded in the FADC. The final trigger was generated in the trigger and clock board (TCB) ² when an anode signal corresponding to one or more photoelectrons (PEs) occurred in each PMT within a 200 ns time window. The anode and dynode signals were recorded whenever the anode signal produced a trigger.

Signals from the CsI(Tl) crystals were amplified by a factor of 10 and digitized in a charge-sensitive 62.5 MHz FADC (SADC) ³ . The SADC provided the integrated charge and the time of the signal. An integration time of 2048 ns was used to record the CsI(Tl) signals considering their decay time. The SADC channels did not generate triggers.

If the trigger condition was satisfied, the TCB sent trigger signals to the FADC and SADC to store the signals from the NaI(Tl) and the CsI(Tl) crystals. The FADC stored an 8 μs long waveform starting approximately 2.4 μs before the time of the trigger position. The SADC stored the maximum integrated charge within an 8 μs search window. This system is similar to the one used in the COSINE-100 data acquisition [37].

The energy calibration of the anode signal was done using a 59.54 keV X-ray emitted from ²⁴¹Am. Figure 3 shows the anode energy spectrum. A clear peak at 59.54 keV resulting from the ²⁴¹Am source is shown together with the ¹²⁷I X-ray escape peak around 30 keV. The dynode signal was calibrated using the photopeaks corresponding to ²¹⁴Bi (609 keV) and ⁴⁰K (1460 keV) contaminants in the crystal.

The charge distribution of the single photoelectron (SPE) was obtained by identifying the isolated clusters at the decay tail of the 59.54 keV X-ray signal from the ²⁴¹Am source (3–5 μs after the signal start). The light yield was determined from the ratio of the total deposited charge and the mean of the SPE charge for the 59.54 keV X-ray data. In this crystal, a light yield of 17.8 ± 0.6 number of photoelectron (NPE)/keV was obtained. It is similar to the result for the NaI-036 crystal, which has the highest light yield among the previously developed low-background NaI(Tl) crystals using the Astro-grade powder [28]. This light yield is also larger than those of the detectors used in the COSINE-100 and DAMA/LIBRA experiments, as summarized in Table 1.

⁴⁰K is one of the most problematic background sources in the search for weakly interacting massive particles (WIMP) using NaI(Tl) crystals. The X-rays/Auger electrons from ⁴⁰K decays produce 3.2 keV energy signals, similar to the energy signals expected for a WIMP-nuclei interaction [30, 31, 38]. The ⁴⁰K decays also emit 1460 keV γ rays, which can escape from the NaI(Tl) crystal and hit the surrounding CsI(Tl) crystals, leading to a double coincidence with the 3.2 keV X-rays.

Figure 4 shows the tagged low-energy spectra from the NaI(Tl) crystal by requiring the detection of the 1460 keV γ ray in the CsI(Tl) crystals. The ⁴⁰K background level in the NaI(Tl) crystal was determined by comparing the measured coincidence rate from a GEANT4-based simulated data, as described in Reference [11]. By accumulating more than 6 months of data, the K level was measured to be 8.3 ± 4.6 ppb, which was compared with the other NaI(Tl) crystals listed in Table 1. It is well below our goal of 20 ppb, consistent with the results from the DAMA/LIBRA crystals [34, 39] and previously developed NaI-035 and NaI-036 crystals with the Astro-grade powder.

Alpha-induced events inside the NaI(Tl) crystals can be identified using the fast decay times of their corresponding signals. The charge-weighted duration time, called the meantime, is defined as ⟨ t ⟩ = Σ i A i t i Σ i A i , (1) where A and t are the charge and time of the i-th digitized bin of a signal waveform, respectively. The meantime is estimated within 1500 ns from the pulse starting timing. Figure 5 shows a scatter plot of the energy versus the meantime for the NaI-037 crystal. In the figure, the populations of γ/β and α events can be separated clearly due to the faster decay times of the α-induced events.

In the NaI(Tl) crystal experiments, the dominant background source in the low-energy signal region is from ²¹⁰Pb [31, 40, 41]. The ²¹⁰Pb activity can be estimated from the alpha-decay studies, because the α decays of ²¹⁰Po originate from the β decays of ²¹⁰Pb. Due to the decay time of 200 days of ²¹⁰Po, the amount of ²¹⁰Pb produced can be estimated using a time-dependent fit of the alpha rate as follows: N = N Pb 210 1 − e − t − t 0 / τ Po 210 + C , (2) where N is the total alpha rate, N _Pb210 is the amount of ²¹⁰Pb at the equilibrium state, t ₀ is the time difference between ²¹⁰Pb contamination and the start time of data taking, τ _Po210 is the mean lifetime of ²¹⁰Po, approximately 200 days, and C represents the long-lived components from ²³⁸U and ²³²Th chains. Figure 6 shows the measured total alpha rates in the NaI-037 crystal over 4.5 months of detector running time. The ²¹⁰Pb activity in the crystal was estimated to be 0.38 ± 0.10 mBq/kg, which is lower than the COSINE-100 crystals and is consistent with the activity in the NaI-036 crystal produced using the Astro-grade powder. However, this activity is slightly higher than the DAMA crystals and another crystal NaI-035 grown with the same Astro-grade powder. The 0.4 mBq/kg level contamination of ²¹⁰Pb is enough to reach 1 count/kg/keV/day background level in the 1–6 keV energy region, similar to the activity in the DAMA/LIBRA detectors, as described in Reference [28].

Contaminants from the ²²⁸Th subchain in the ²³²Th family can be estimated by exploiting the time-delayed α–α coincident events of ²²⁰Rn and ²¹⁶Po. The alpha decay of ²¹⁶Po has a half-life of 0.145 s following its production via alpha decay of ²²⁰Rn. Owing to the short half-life of ²¹⁶Po, it is straightforward to select two successive α particles with almost no random coincident events.

The presence of the coincident events with 4.5 months of exposure is shown in Figure 7A as the distribution of the time gap between those two α events. The exponential component indicates the contamination from ²³²Th, corresponding to below 0.78 ppt (90% confidence level). The ²³²Th concentration in the NaI-037 crystal is the lowest among the other NaI(Tl) crystals, as summarized in Table 1.

²³⁸U is one of the common radioisotopes because of its long half-life. The ²³⁸U content in the background can be studied using the time-delayed β–α coincident events of this radioactive chain, similar to the calculation of the ²³²Th background. This method exploits the α decay of ²¹⁴Po with a half-life of 164.3 μs, while ²¹⁴Bi, the parent particle of ²¹⁴Po, undergoes β decay. Due to the 50 μs dead time of the trigger system, the coincident events with delay times greater than 50 μs can be tagged. The results with 4.5 months of exposure are shown in Figure 7. The ²³⁸U activity of NaI-037 was below 1.98 ppt (90% confidence level, similar to that observed for the other NaI(Tl) crystals, as given in Table 1.

Because of the small size of the NaI-037 crystal and no liquid scintillator active veto, a significantly higher background contribution is expected from the external background compared to those found in the COSINE-100 crystals [30, 40]. The PMTs attached to the NaI(Tl) and the CsI(Tl) crystals are the primary sources of external background. In this study, the external background contributions were simulated using the GEANT4-based simulation toolkit used for the COSINE-100 background modeling [30, 40].

The cosmogenic production of radioactive isotopes in the NaI(Tl) crystal is mainly due to long-lived nuclides such as ³H and ²²Na [30, 43]. The NaI-037 crystal was grown in Daejeon, Korea (70 m in altitude) and delivered underground within a month. Based on the previous study, 1-month exposure time near sea level can produce 0.004 mBq/kg of ³H and 0.05 mBq/kg of ²²Na [43], respectively.