Radium interference during radon measurements in water: comparison of one- and two-phase liquid scintillation counting

Sample preparation and counter calibration followed the method described by the US Environmental Protection Agency (US EPA) (13), which is appropriate for Rn determination in drinking water from groundwater and surface water sources in both one- and two-phase samples (20). The calibration factor CF (cpm/Bq) (detection efficiency of ²²²Rn or ²²⁶Ra) was determined based on the calibration sample (Ra standard) count:

where S (cpm) is the calibration standard count, B (cpm) background sample count, C (Bq/L) is the concentration of ²²⁶Ra standard solution, and V (L) is the volume of the calibration standard per analysed sample (10 mL in our experiments).

The activity concentration of ²²²Rn [A (Bq/L)] was calculated using the following formula:

where G (cpm) is the sample count and D the decay correction factor for ²²²Rn. This factor should be calculated for the time between sampling and midpoint of the counting (t), as follows:

For Rn half-life [ T 1/ 2 (²²²Rn)] we assumed 3.824 days.

Calibration involved preparation of calibration standards (10 mL of distilled water spiked with the known ²²⁶Ra activity, mixed with 10 mL of scintillation cocktail, shaken, and set aside for 30 days to attain secular ²²⁶Ra/²²²Rn equilibrium) and background samples (10 mL of distilled water mixed with 10 mL of scintillation cocktail) and was carried out as described earlier (21).

Two-phase samples must be shaken vigorously for a few minutes at least to ensure efficient Rn transfer from water to the organic phase (15). For recounts we repeated shaking and then waited for 2 h for the ²²⁶Ra/²²²Rn equilibrium to restore, as a number of studies have shown that repeated shaking of two-phase ²²⁶Ra standard calibration samples before re-counting returns ²²²Rn and ²²²Rn progenies into the aqueous phase, which slightly reduces Rn detection efficiency (5–10 %) until the equilibrium is restored (22–23).

The reliability of Rn measurement in water depends greatly on the sampling technique, as inadequate procedure can lead to error. The drinking water we took from the fountain, for example, must not get in contact with air at any point during sampling or storage (24). We minimised Rn desorption by collecting water from a non-aerated spigot and filling a glass beaker (5 L) until it was overflowing, after which we submerged the 1.5 L glass bottle into the beaker upside-down and turned it up slowly to fill it with water and to eliminate any air bubbles (19). We then capped it, still submerged, with a teflon-lined cap.

The 1220 Quantulus^TM LSC we used in our experiments is convenient for ultra-low-level measurements because of its own background reduction system that involves a passive shield and active guard detector based on anticoincidence counting (25). Samples with each cocktail were prepared in three probes and measured on LSC for 100 minutes in six cycles.

Rn measurements also depend on the correctly adjusted pulse shape analysis (PSA) parameter (26), whose values range between 1 and 256. Proper PSA settings make it possible to discriminate alpha from beta signals, that is, to measure both alpha and beta activities at the same time by directing alpha and beta signals in two separate spectra (25). The lowest limit of Rn detection on Quantulus can be achieved when optimal PSA value has been experimentally determined, PSA is activated (which greatly reduces alpha backgrounds), and Rn content is calculated from the alpha spectrum (14).

Spectral data were acquired and evaluated with the WinQ and EASYView software (PerkinElmer Life Sciences, Turku, Finland).

The experiments involved the use of a radioactive source, aqueous ²²⁶Ra standard (Czech Metrology Institute, Brno, Czech Republic, ref. date 1/10/2013) with certified activity A(²²⁶Ra) of 39.67 Bq/mL and combined standard uncertainty of 0.5 %. All experiments were completed in the early 2018. ²²⁶Ra progenies had been purified from the standard solution five years earlier to minimise interference from ²¹⁰Po, ²¹⁰Pb and ²¹⁰Bi to max. 4 % (27).

All samples were prepared in high-performance 20 mL glass vials (Perkin Elmer). For the emulsifying scintillation cocktail we used the Ultima Gold AB and for the three water-immiscible cocktails we used High Efficiency Mineral Oil Scintillator, Opti-Fluor O, and Ultima Gold F, all by Perkin Elmer. The Ultima Gold AB cocktail was reported to generate the alpha background spectrum, which suggests that this cocktail contains ²²⁶Ra as an impurity (14). The Ultima Gold F cocktail uses di-isopropylnaphtalene (DIN) as solvent and needs more time for a clear phase separation than older (mineral oil or pseudocumene-based) kinds of cocktails but is more suitable for alpha-beta discrimination (6). The High Efficiency Mineral Oil Scintillator cocktail is a mixture of mineral oil (70–75 %) and pseudocumene (25–30 %) (15).

In the mixture of 10 mL of Opti-Fluor O and 10 mL of water, ²²²Rn partition coefficients for water:cocktail:air are 1:48:2 (28). These coefficients reflect ²²²Rn detection efficiency, which corresponds to the one third of the CF value (as the alpha spectrum contains ²²²Rn, ²¹⁸Po, and ²¹⁴Po) (29). In our previous research (21) we investigated the dependence of CF on PSA settings (in the range from 30–90) and established that Opti-Fluor O and Mineral Oil had very similar CF in the 30–70 PSA range. Ultima Gold F had a slightly lower CF, while the emulsifying Ultima Gold AB cocktail had about 25 % higher CF than its water-immiscible counterparts.

The experiments were divided in two parts. In the first part, vials with 10 mL of spa water and 10 mL of respective cocktail (48 samples in total) were kept in dark to avoid photoluminescence (this is a common LSC practice, but photoluminescence reactions occur in low-energy regions and would probably not have interfered with the Rn spectra). After 5 h, all 48 samples were measured for mean baseline ²²²Rn activity concentrations (A₀). Moreover, the activities for days ~4 (about one ²²²Rn half-life), ~9 (two half-lives), and ~31 were also measured in 12 samples that were not spiked with Ra solution.

In the second part, the remaining 36 samples were spiked with 20 μL, 100 μL, or 200 μL of the ²²⁶Ra standard, which corresponds to ²²⁶Ra activity concentrations of 79.34, 396.7, or 793.4 Bq/L, respectively. Each sample type was prepared in triplicate. These were counted in six cycles of 100 min at different time points (~5, ~11, ~31, and ~64 days after sample preparation), and the obtained counts were used to calculate Rn activity concentrations according to the US EPA method (13).

Based on ²²²Rn decay, we also calculated its ²²²Rn progeny activity concentrations (what we refer to as theoretical value) for the mean moment of counting (~5, ~11, ~31, and ~64 days after sample preparation). Since Rn/Ra equilibrium occurs after ~30 days, the measurements on days 31 and 64 in fact enabled determination of ²²⁶Ra activity concentration.

The following subsection describes the optimisation of the alpha/beta discrimination circuit that can be applied for Perkin Elmer instruments. Conventional recommendation for optimum PSA setting involves counting of pure alpha and beta calibration samples (often ²¹⁴Am and ⁹⁰Sr) and adjusting PSA where the spillover of alpha pulses into the beta spectrum and vice versa is equal and minimal (19). A PSA discriminator set in this manner is to some extent inadequate for ²²²Rn determination by an LS counter, since PSA can be more precisely regulated with a ²²²Rn standard, provided that the lowest beta spillover and alpha-to-beta count ratio match the theoretical value (15). The third possibility is the simplest but not as precise as the previous two. It requires recording the dependence of CF on PSA and selection of a working PSA discriminator within the range in which CF factor does not vary significantly (7, 29). This technique provides an optimal range for PSA selection, even though its value may lower the counting precision for up to a few percent, but the advantage is that this range does not vary significantly between different scintillation cocktails. We therefore made one experiment to obtain the optimal PSA range setting and used it for all scintillation cocktails as described below.

Figure 1 represents CF dependence on the chosen PSA value, which shifted in the experiments across its range from 1 to 256. The calibration factor CF (detection efficiency for ²²²Rn, ²¹⁸Po, and ²¹⁴Po radionuclides) was consistent for all ²²⁶Ra concentration activities. Similar CF dependencies on PSA values have been reported by other authors (6, 7, 29). It is clear that CF plateaus when PSA values are set between 40 and 90. Consequently, all measurements presented in this paper were performed with PSA set at 70.

Figure 1

The first line of experiments established the accuracy of one- and two-phase LSC methods. Our findings seem to support earlier observations that Mineral Oil yields an overestimate and Ultima Gold AB an underestimate of ²²²Rn activity concentrations (21), although the variability of our results excludes definitive conclusion on this point. Figure 2 shows measured ²²²Rn activity concentrations at baseline (A₀) and on days 4, 9, and 31 after sample preparation as well as theoretical ²²²Rn concentrations calculated with radioactive decay formula. The consistency between these calculated theoretical predictions and ²²²Rn activity concentrations measured on day 4, 9, and 31 for all scintillation cocktails confirms the reliability of the LSC methods to measure ²²²Rn in water samples.

Figure 2

Considering the limits of detection for measuring ²²²Rn and ²²⁶Ra laid down by the Council Directive 2013/51/ EURATOM (30), the methods used should be able to detect activity concentrations as low as 10 Bq/L and 40 mBq/L, respectively (30). In our earlier report (21) minimal detectable activities (MDA) for ²²²Rn over 300 minutes of counting were 38 mBq/L for Ultima Gold AB, 104 mBq/L for Ultima Gold F, 65 mBq/L for Mineral Oil, and 104 mBq/L for OptiFluor O, which shows that LSC methods are suitable for both direct ²²²Rn and indirect ²²⁶Ra detection, since the required MDA of 40 mBq/L for the latter is easily achievable if longer counting times are applied.

The average ²²²Rn activity concentration in the baseline (A₀) spa water samples was 473(±65) Bq/L (average of four A₀ values in all four cocktails, Figure 2). Relying on the law of radioactive decay, we used four functions from Figure 2 to predict Rn activity over time. As mentioned before, the theoretical ²²²Rn activity of the spa water was calculated based on decay at the moment of counting. Tables 1–4 show mean baseline, measured, and theoretical ²²²Rn activity concentrations 5, 11, 31, and 64 days after sampling.

The first two measurements were carried out while ²²⁶Ra and ²²²Rn were still not in the equilibrium, but on the third and the fourth measurement the equilibrium was there (after 30 days or eight ²²²Rn half-lives). The latter measurements made it possible to evaluate the accuracy of LSC in one- and two-phase samples, since the original, naturally occurring ²²²Rn from the spa water had completely decayed by that time and any leftover ²²²Rn activity measured in these samples is therefore the result of the decay of spiked ²²⁶Ra standard. The baseline theoretical ²²²Rn activity represents measured activity corrected for sampling date and time, while all other values were derived from the law of radioactive decay. The activity concentration in the first measurement, made about five days after sampling, was corrected with respect to ²²²Rn half-life in order to evaluate to which extent ²²⁶Ra presence influenced ²²²Rn results. These corrections were not carried out for other measurements, simply because it is not common laboratory practice to wait for more than two half-lives of a radionuclide of interest to count the samples, and it is clear that such corrected values would be enormous because of high spiked ²²⁶Ra activities.

Tables 1–3 showing measurements in two-phase samples reveal no substantial differences between the cocktails. The results obtained for samples with the lowest spiked ²²⁶Ra concentration (79.34 Bq/L) were satisfactory for all three cocktails and for all four dates of counting, which indicates that Ra interferes with the Rn spectra because of the migration of its Rn progeny from the aqueous to the organic phase. With higher spiked ²²⁶Ra concentrations (396.7 Bq/L and 793.4 Bq/L), however, it is clear that only a smaller fraction of ²²⁶Ra presence (one fourth to one fifth) can be detected in samples with radioactive equilibrium, regardless on the measurement day. It is possible that some saturation effect occurs that limits Rn transfer to the organic phase. This is an interesting hypothesis that should be verified in a larger number of samples with different higher ²²⁶Ra concentrations for better statistics.

The lowest activity concentrations were obtained with the Opti-Fluor O cocktail (Table 2), which points to the lowest interference from Ra decay. Namely, even with ²²⁶Ra and ²²²Rn in equilibrium, only about one fifth of ²²²Rn from ²²⁶Ra decay entered the organic phase.

This brings us to the main conclusion about two-phase samples, namely that Ra progeny ²²²Rn entering the organic phase interferes with measuring the original Rn activity in a sample.

On the other hand, some of the obtained activities slightly exceed spiked ²²⁶Ra content, even in the fourth measurement when all radon from the sample had decayed. The explanation lies in inadequately adjusted PSA (PSA=70 was probably not optimal value for the samples prepared with selected cocktails). If the activity concentration exceeds spiked ²²⁶Ra concentration, there has been a spill of beta particles into the alpha spectrum during counting.

Ra interference is even more evident in one-phase LSC samples, especially before Rn/Ra equilibrium was achieved. The obtained activities with the Ultima Gold AB cocktail were close to ²²⁶Ra + ²²²Rn cumulative activities, while two-phase samples (regardless of the cocktail) gave much lower activity concentrations. Obviously, the two-phase method provides more accurate ²²²Rn measurements in the presence of high ²²⁶Ra content in samples if counting is performed in the first few days after the sampling, that is, before radioactive equilibrium is achieved.

The one-phase method offers a more reliable indirect assessment of ²²⁶Ra concentration later, when radioactive equilibrium is achieved, while the two-phase LSC had satisfying accuracy only with the lowest spiked ²²⁶Ra concentration. At higher spiked activity concentrations all three water-immiscible cocktails showed some kind of “saturation” limit that allows only so much of Ra in the form of its ²²²Rn progeny to enter the organic phase from water when the ²²²Rn/²²⁶Ra equilibrium is reached.

What this research suggests, however, is that LS counts in one-phase samples can be precise if the PSA level is set correctly. An even more precise, yet indirect, measurement of native ²²²Rn (i.e. not the Ra progeny) would involve purging it from a water sample with nitrogen or argon or by boiling it to evaporate. The obtained “background” sample would then contain only the remaining radionuclides and serve to measure ²²⁶Ra activity concentration (if the presence of other radionuclides can be excluded by some other chemical pre-treatments), which can be deducted from the baseline measurement in the one-phase sample. However, this would take more time and resources.

Speaking of ²²⁶Ra activities, we would also like to address the issue of extremely high findings compared to the ones in the non-spiked samples. The purpose of our experiments was to investigate the effect Ra on the accuracy of native ²²²Rn measurements, which, we assumed, would be the most obvious if spiking involved high ²²⁶Ra activities, and we were right about that, but interference rendered these measurements completely useless (which is why we did not use them in analysis). Further research would, therefore, do better by exploring the demonstrated phenomena in the presence of much lower ²²⁶Ra activities than the ones we used for spiking.

Throughout the measurements we noticed that the spectra of the spiked samples broadened over time (expanding partly into the lower-channel region). We therefore had to recalibrate and recalculate all results in Tables 1–4 for a wider ROI (given in each table), for which we repeated calibration procedure as described elsewhere (21). Changes in the spectral shape for all four cocktails are presented in Figure 3. The initial ROI expanded for approximately 100 channels towards the lower energy region. Figure 3b shows the peaks of certain radionuclides relevant for all other spectra.

Figure 3

One of the reasons why the spectral shapes evolved diversely (still following a similar pattern) between the water-immiscible cocktails was unequal transfer of ²¹⁰Po from the ²²⁶Ra standard solution to different organic cocktails, which appears as the ²²²Rn + ²¹⁸Po peak broadening to the left (15, 27). The greatest affinity of ²¹⁰Po for the organic phase was observed in the Mineral Oil Scintillator (15). This broadening to the left also indicates that alpha energies were absorbed in water at the interface (15). Smaller discrepancies between the three water-immiscible cocktails can also be the result of different ²¹⁰Pb and ²¹⁰Bi transfer to the organic phase during shaking (15). It is also possible that shaking does not achieve 100 % water-to-cocktail transfer of ²²²Rn progeny of ²²⁶Ra from the standard solution in all cocktails over time, and that short-lived progenies of ²²²Rn, ²¹⁴Pb and ²¹⁴Bi, migrate differently from various cocktails to the water phase (these effects could lower detection efficiency by up to several percent) (15).

In similar experiments, spectral evolution of ²²⁶Ra over five months was a result of the presence of several radionuclides in disequilibrium between the two phases, whose spectra overlapped. However, the counts varied within the 5 % margin (31). The instability of the two-phase samples over longer periods of time is owed to radiological and chemical equilibration between the liquid (water), cocktail (organic), and air (gaseous) phases, which are affected by vial shaking, temperature, type of cocktail and vial, and by the presence of a carrier that reduces accumulation of long-lived progeny on the glass walls of the vial (22). The ²²⁶Ra standard solution used in this study contains Ba as a carrier (the solution is composed of 1 g/L of BaCl₂ + 10 g/L of HCl), and it was determined earlier that ²¹⁰Pb can migrate to the water phase if the glass and the cocktail phase are saturated with the carrier (15).

Glass vial fluorescence generates low-energy peaks in the alpha spectrum in channels 1–300 (32), but it is also possible that some of the beta particles spilled over to the alpha spectrum (Cherenkov effect of beta progenies, dominantly ²¹⁰Bi) if PSA was not precisely adjusted.

During all LSC measurements, we monitored the quench parameter, since quenching can greatly impact detection efficiency and spectral shape. The vials were tightly sealed all the time to exclude oxygen as potential quencher. We observed no spectral shift owed to quenching in the ²²⁶Ra standard solution investigated in an earlier research (26) (Figure 3), nor did the spectral quench parameter of the external standard [SPQ(E)] alter significantly, which is a reliable indicator that quenching did not occur in our experiments.

However, sample stability and the overall spectral evolution presented in Figure 3 were certainly affected by the chemical composition of the spa water sample, which depends on local geology.

The fact that spectral evolution was unique for each cocktail points out that calibration can be challenging if measurements span across longer periods of time and calls for careful consideration.