A part per trillion isotope ratio analysis of 90Sr/88Sr using energy-filtered thermal ionization mass spectrometry

Strontium-90 is a major radioactive nuclide released by nuclear accidents and discharge waste. Input of such radioactive nuclide into earth surface environment causes potential threat of long-term internal exposure when taken up by organism. Rapid and precise measurement of 90Sr in variety of environmental sample is important to understand the distribution and dynamics of 90Sr in the local environment after the accident and to assess the effect of radioactive nuclide inputs on bodies. However, previous 90Sr measurement techniques have drawbacks such as long measurement times for radiometry and high detection limits for mass spectrometry. Here we present a technique to accurately measure a significantly small amount of 90Sr in natural environmental samples using an energy-filtered thermal ionization mass spectrometry. Our technique achieved a 90Sr detection limit of 0.23 ag, which corresponds to a 90Sr activity of 1.2 µBq. The detection limit was lowered by two orders of magnitude compared with the previous mass spectrometric 90Sr analyses. The ability of our technique will expand the applicability of mass spectrometric 90Sr survey not only to the rapid 90Sr survey upon nuclear accidents but also to study a long-term environmental diffusion of radioactive materials using size-limited environmental and biological samples.

www.nature.com/scientificreports/ Conventionally, radiometric methods using solid/liquid scintillators or gas ionization detectors were used to detect 90 Sr in various environmental samples 5,[12][13][14][15] . The radiometric method is fundamentally a counting experiment of the decaying atoms. Because 90 Sr has a half-life of 28.79 years 24 , the fraction of the decaying atom is significantly small: approximately 1.3 × 10 9 atoms of 90 Sr corresponds to 1 Bq (i.e. decay per second) of activity. Therefore, for precise radiometric 90 Sr determination, a relatively large sample size or long analysis time is required 12,17 . For example, 100 g of soil samples are measured by nitrate precipitation-low background gas-flow counting method (nitrate precipitation-LBC) with 2 weeks 9 , and 1 L of seawaters are measured by gross beta radiometric counting with iron-barium co-precipitation method 14 . Because 90 Sr is a β-particle emitter, radiometric determination of 90 Sr includes complex radiochemical procedures and typically requires 2 weeks or more measurement time 17, 18 . This slow sample processing speed is insufficient for an urgent environmental survey in response to a nuclear emergency 27 .
An alternative and less time-consuming method to analyze 90 Sr is mass spectrometry. Inductively coupled plasma mass spectrometry (ICP-MS) is a new mass spectrometric technique for 90 Sr detection [27][28][29][30][31] . Unlike radiometry, the target of mass spectrometric measurement is all the existing 90 Sr atoms in the sample. The large difference in target numbers suggests that mass spectrometric methods are more sensitive for detecting 90 Sr. However, measuring a minor isotope is a challenging task in mass spectrometry. Mass spectrometry's sensitivity to detect trace 90 Sr is hindered by high noise signals due to isobaric interferences of 90 Zr and molecular ions, as well as peak tailing of the highly abundant 88 Sr ions. ICP-MS's major source of noise signals is 90 Zr 2,24,25 . In environmental samples, the ratio between minor 90 Sr and more abundant stable isotopes of Sr or isobaric isotope 90 Zr is extremely high (i.e., at least nine orders of magnitude 15 ). ICP ion source efficiently ionizes 90 Zr, which remains in the sample solution in trace amounts even after Sr extraction chemistry. The reaction of the ions with O 2 gas is used in a dynamic reaction cell (DRC) technique to reduce the 90 Zr ion transmission 2,24,25 . Typically, this technique achieves abundance sensitivities, defined as the intensity ratio between 88 Sr peak tail on m/z = 89.908 and 88 Sr, on the order of 10 −9 2,19 . Furthermore, the introduction of a new technique, triple quadrupole ICP-MS (ICP-MS/MS), combined with the O 2 gas reaction, effectively reduced the 90 Zr ion transmission in the mass spectrometer and lowered the noise signal to 0.1 cps, which correspond to a 90 Sr detection limit of 0.11 fg (0.6 mBq) 28 . With a sample size of 4 µg of Sr, the abundance sensitivity for the 90 Sr/ 88 Sr ratio achieved by this technique was 5 × 10 −12 28 . However, reducing the 90 Sr detection limit with the ICP ion source is difficult. The highly efficient ICP ion source mainly produces abundant polyatomic or polyvalent ions from the sample solution's solvents, Ar gas, and trace impurities in the sample solution, and such ions exist across the entire m/z range. It also emits Ar gas-related ions such as Ar + and ArO + at significantly high intensities, which may cause non-spectrum interference of the peak tails. All such interferences are increasing the noise level and limiting the sensitivity of ICP-MS analysis. Even with the ICP-MS/MS technique, the background signal was reported as 0.1-0.2 cps when aspirating a blank solution 28 , showing that noise signals of sub-cps level are inevitable while using the ICP ion source.
Thermal ionization mass spectrometry (TIMS) is a standard technique for measuring isotope ratios of Sr, such as 87 Sr/ 86 Sr and 88 Sr/ 86 Sr 34 . Compared with the ICP ion source, the thermal ionization ion source is energylimited, and thus ionization is mostly limited to target elements with very few polyatomic ions. In TIMS, Argas-related species, solvent-related species, and polyvalent ions are mostly absent. With less spectrum and nonspectrum interferences, TIMS can reduce noise signal levels on 90 Sr and thus increase sensitivity upon 90 Sr detection. Recently, several attempts have been made to detect 90 Sr using the TIMS technique. Previous studies have failed to detect traces of 90 Sr in environmental samples 27,28 . Kavasi et al. measured 90 Sr/ 88 Sr ratios of 90 Srcontaining reference materials of wild berry and lake sediment using a sample size of 1000 ng of Sr and reports an abundance sensitivity for the 90 Sr/ 88 Sr ratio as 2.1 × 10 −10 , corresponding to a 90 Sr noise signal of 0.77 cps 37 . However, when the 90 Sr/ 88 Sr ratio is lower, their measured 90 Sr/ 88 Sr ratio shows a systematic bias toward higher values. To account for such a significant bias, additional empirical "relative bias" errors must be introduced into the analytical uncertainty. This inaccuracy is obvious when the 90 Sr/ 88 Sr ratio is lower than 1.2 × 10 −9 and is likely to be caused by an inaccurate noise correction scheme for the 90 Sr signal. Ito et al. focused on analyzing smallsized samples and constructed an isotope-dilution total-evaporation (ID-TE-) TIMS technique to analyze 90 Sr with a sample size of 5-20 ng of Sr 38 . Their study did not report abundance sensitivity, but it can be estimated as 2 × 10 −8 based on the reported analytical conditions of 90 Sr noise level of approximately 5 cps and 88 Sr target intensity of 4 V 38 . Among the mass spectrometric 90 Sr analysis methods, TIMS is the only method that has been confirmed by an independent IAEA proficiency test 39 . During the TIMS analysis, the ionization of 90 Zr + is suppressed, whereas Sr is still present and ionizing from the filament 38 because of the difference in the evaporation and ionization potentials between the two elements. Therefore, the peak tailing of the highly abundant 88 Sr ions is the main source of background signal for 90 Sr in TIMS measurements. Peak tail ions can be efficiently reduced using an energy filtering device placed immediately before the detector. However, none of the previous 90 Sr analyses used an effective energy filtering device to account for the 88 Sr tailing [35][36][37][38] . Kavasi et al. used a WARP energy filter. However, they did note that the WARP energy filter eliminates low energy ions but has no effect on high-mass side peak tailing, implying that the WARP energy filter does not work for 88 Sr peak tail on 90 Sr. The ability of TIMS to detect a significantly small amount of 90 Sr remains unknown.
In this study, we focused on bringing out the performance of modern TIMS instruments to perform an accurate 90 Sr/ 88 Sr measurement of environmental and biological samples with low 90 Sr activity using an effective energy filtering technique for 90 Sr detection. The Retarding Potential quadrupole (RPQ) lens act as high selectivity filter for ions with disturbed energy or angle 40 . The use of RPQ lenses for energy filtering coupled with reduction of the multiple noise signal sources resulted in a significantly low and highly stable noise signal for 90 Sr compared with the previous studies. With significantly low and stable noise signals, an appropriate noise correction scheme was used in this study to allow accurate measurement of 90

Results and discussion
Noise reduction schemes upon 90 Sr detection by energy-filtered TIMS. Peak tailing of 88 Sr, which has the largest abundance among Sr isotopes, is the main limiting factor in TIMS detection of 90 Sr. The RPQ lenses used in this study eliminate peak tailing from the high-mass and low-mass sides of the 88 Sr peak (Fig. 1). The RPQ lens parameters must be fine-tuned to effectively eliminate the 88 Sr peak tailing on 90 Sr while sacrificing only a small amount of ion transmission rate and peak shape ( Figure S1 in supporting information). Finally, after the introduction and fine-tuning of the RPQ lenses, a 1000-fold reduction of the 88 Sr peak tail signal was achieved ( Figure S2 in supporting information). Because the 88 Sr peak tail signal has been reduced to approximately 0.01 cps on 90 Sr, signals from other noise sources must also be controlled at this noise level. The filament material, rather than the sample-derived Zr, is the major source of the isobaric 90 Zr signal in TIMS measurement. Typically, filaments were baked out at 4500 mA (approximately 2000 °C) before the use. With this level of bake out, about 1 cps and about 0.1 cps of the 90 Zr + signals were observed from Re and Re-ZR filaments, respectively, during Sr isotope ratio measurement even with the presence of Sr sample on the filament. These were not acceptable levels of noise in this study. Therefore, filaments were baked at higher temperatures at a 5.5 A filament current (approximately 2200 °C) to reduce the filament-derived Zr + signal. After this high temperature bakeout, the Re and the Re-ZR filaments tend to show 90 Zr + signals on the 0.01 cps order when heated up to 1550 °C without the sample material. Even after this high temperature bakeout, a few Re and Re-ZR filaments still demonstrate higher 90 Zr + emission. Therefore, after the bakeout, all the filaments were inspected for 90 Zr intensity to ensure that filaments with higher 90 Zr + emission were not used. The 90 Zr intensity threshold at 1550 °C was empirically set to 0.04 cps. Note that the actual 90 Zr intensity during Sr isotope measurement is negligible (far lower than 0.04 cps) because Zr + emission is suppressed during the presence of Sr on the filament.
Ito et al. observed that organic material with m/z 90.0 is a significant isobaric interfering molecule on 90 Sr, and this could be perfectly mass separated by slightly shifting the axial m/z from 89.908 to 89.777 38 . We detected another isobaric interference molecule spectrum at m/z 89.908 with intensities on the order of 0.1 cps or less ( Figure S3 in supporting information). The intensity of this noise signal is relatively high at the beginning of the measurement and decreases with time ( Figure S4 in supporting information), indicating that this noise signal is related to organics or volatiles. The most plausible molecule for this mass spectrum is 88 SrH 2 + , with hydrogen derived from organics or residual H 2 O. This noise signal was eliminated by preheating of the sample filament under vacuum before the measurement and by avoiding the use of parafilm/catheter during sample loading.
Abundance sensitivity and detection limit of the 90 Sr/ 88 Sr measurement. The 90 Sr-free NIST SRM-987 was measured using all these noise reduction schemes to validate the noise level and noise stability on 90 Sr. The results of the three analytical sessions with slightly different analytical conditions are summarized in Table 1, Fig. 2, and Figure S5 in supporting information.
The average intensity of the 90 Sr noise signal of the three analytical sessions were 0.0118 ± 0.0029 (2SD, n = 4), 0.0168 ± 0.0034 (2SD, n = 13) and 0.0136 ± 0.0029 (2SD, n = 17) for Bremen, Fukushima-1st and Fukushima-2nd sessions, respectively. Note that these noise signals include dark noise counts because they were not corrected. Slightly higher noise intensity of Fukushima-1st compared with the Bremen and Fukushima-2nd reflects the fact www.nature.com/scientificreports/ that higher SEM operating voltage resulted in higher dark noise level in Fukushima-1st session ( Table 1). The average dark noise count during the Fukushima-2nd session was 0.0096 cps, accounting for 70% of the total noise signal. This indicates that our noise reduction scheme has successfully reduced noise signals other than the detector's intrinsic dark noise to the level of 0.004 cps (Fig. 2). The small variability of the noise signal indicates that the noise signals are well controlled at this level. In this study, the level and stability of the noise signal achieved are superior compared with the noise signals reported by non-energy filtered TIMS technique 37 (0.77 ± 0.82 cps, 2SD) and by ICP-MS/MS technique 28 (0.1 cps). Noted that measurement with such a low signal requires a long measurement time (1 h in this study) to count the substantial number of ions. The stable nature of the TIMS ion source allows long measurement with stable analytical conditions. The abundance sensitivity and detection limit of the 90 Sr/ 88 Sr ratio can be determined from these data. Finally, the abundance sensitivity ( 90 Sr/ 88 Sr ratio) achieved in the Fukushima-2nd session was 8.3 ± 1.8 × 10 −12 (2SD, n = 17; Table 1). A detection limit of a signal is defined by 3σ of the variability of the zero-point (or blank) analysis. Therefore, the detection limit of 90 Sr/ 88 Sr ratio finally achieved in the Fukushima-2nd session was estimated to be 2.7 × 10 −12 (Table 1 and Figure S5c in supporting information). The fact that the detection limit is lower than the abundance sensitivity demonstrates TIMS's excellent noise stability and its good control on 90 Sr. 90 Sr/ 88 Sr measurement of the reference materials and natural samples. Table 2 summarizes the details of the analyzed environmental and biological samples, as well as the 90 Sr activity parameters found in the literature (more details are shown in Table S1 in supporting information). As a 90 Sr-free sample, a seawater reference material NASS-6 (Atlantic surface water) issued by the National Research Council Canada and a geochemical reference material JCp-1 (modern coral) issued by the Geological Survey of Japan were analyzed. The 90 Sr-containing samples analyzed in this study include the International Atomic Energy Agency's certified refer-  www.nature.com/scientificreports/ ence materials IAEA 156 (clover) and IAEA 330 (spinach). We also analyzed the ash from crayfish and smallmouth bass samples collected in Fukushima prefecture, Japan, after the Fukushima Daiichi Nuclear Power Plant accident. The 90 Sr activity of these samples was analyzed using a radiometric method and published elsewhere 20 . Table 2 also summarized the measured Sr abundance of the 90 Sr containing samples. The decay corrected 90 Sr/ 88 Sr reference value of these samples was calculated for each analytical session using this measured Sr abundances together with either the certified 90 Sr activity, for IAEA 156 and IAEA 330, or the radiometrically measured 90 Sr activity, for Fukushima samples, respectively.

Performance and applicability of 90 Sr analysis by energy filtered TIMS.
Among the environmental and biological samples analyzed, Sr ion beams run short before 220 cycles in some of the measurements with a sample amount of 100 ng of Sr. Nevertheless, 100 ng measurements mostly show precision indistinguishable with the measurement of larger sample sizes. From these observations, we defined that the minimum amount of sample required for our 90 Sr/ 88 Sr measurement as 100 ng of Sr. This is one of the essential parameters to evaluate the performance of TIMS measurement. Given the minimum sample size of 100 ng of Sr, and the 90 Sr/ 88 Sr ratio detection limit of 2.7 × 10 -12 , the detection limit of the absolute 90 Sr amount can be estimated as 0.23 ag or 0.0012 mBq. The previous TIMS 90 Sr measurements have reported a 90 Sr detection limit of 0.17 fg (or 0.88 mBq) 37 and 0.029 fg (or 0.15 mBq) 38 . Our technique has succeeded to lower the 90 Sr detection limit of TIMS by three orders of magnitude.
Based on the applicability of this technique to environmental and biological sample measurements, detectable 90 Sr activity concentrations in this study are different among samples with different stable Sr abundances (Fig. 4). This is because the TIMS technique measures isotope ratios and thus the detection limit of the 90 Sr analysis is primarily determined as 90 Sr/ 88 Sr ratio. For example, seawater is one of the typical environmental samples and has a stable Sr abundance of approximately 8 mg/kg. With this energy-filtered TIMS technique, the detection limit of seawater 90 Sr activity concentration is estimated as 0.09 Bq/kg. The minimum quantity of the seawater sample required for the measurement is as small as 12.3 µL. River water and groundwater have stable Sr abundances range from several tens to several hundreds of ppb. In this case, the detection limit of 90 Sr activity concentration will be as low as 0.1 mBq/kg. A sample size of 10 g will be required to analyze such a dilute sample.
Comparing the capability of 90 Sr analysis between TIMS and ICP-MS is not straightforward. The capability of ICP-MS is limited by the detection limit of the absolute amount of 90 Sr and the abundance sensitivity defined as the 90 Sr/ 88 Sr ratio (Fig. 5). It is not limited by the amount of stable Sr unless the 90 Sr/ 88 Sr ratio was interfered with the abundance sensitivity limit. However, the capability of TIMS is limited by the detection limit of 90 Sr/ 88 Sr ratio and the minimum amount of sample required for the analysis (Fig. 5). The latter is required to keep the fixed ion beam intensity of 88 Sr during the measurement because the performance of the 90 Sr/ 88 Sr ratio measurement by TIMS is mainly determined by the ion beam intensity of the most abundant 88 Sr. The detection limit of the absolute amount of 90 Sr is a derivative of these two parameters for TIMS. Compared with the previous TIMS and www.nature.com/scientificreports/ ICP-MS/MS techniques, a significant reduction in the 90 Sr/ 88 Sr detection limit achieved in this study had widely expanded the applicability of this technique to analyze samples with low 90 Sr contents (Fig. 5). Additionally, an order of magnitude reduction of the minimum sample amount compared with the previous TIMS technique allows the analysis of 10 times smaller sample sizes. These features are advantageous in analyzing trace amounts of 90 Sr on size-limited environmental samples such as soil exchangeable fraction and eolian dust as well as biological samples with low Sr abundances.

Methods
Reagents and standards. For sample preparation, ultrahigh purity grade acids, HNO 3 , HCl, HClO 4 , and HF (TAMAURE AA-100; Tama Chemicals Co., Ltd, Japan.), as well as ultra-pure water produced by a Milli-Q Element system (Millipore, USA) were used. The Sr isotopic reference material SRM 987 issued by NIST was used as a 90 Sr-free reference standard.
Sample preparation. The IAEA reference material's certified 90 Sr activity values were calculated using their dry weight. Therefore, these materials were weighed and digested as they were. Concentrated HNO 3 and HClO 4 acids were used to digest IAEA 156 and IAEA 330. The Fukushima samples were analyzed on an ashed sample, which were used in the previous 90 Sr activity analysis 20 . The ashed crayfish and smallmouth bass samples were weighed and digested successively using 2.4 M HCl and conc HNO 3 . JCp-1 was digested with 5% CH 3 COOH. Finally, all the samples were dissolved in 3 M HNO 3 . NASS-6 was mixed with 6 M HNO 3 to make a 3 M HNO 3 solution.
Sr was separated from the other elements, including Zr, using extraction chromatography with Sr Resin (Eichrom Technologies Inc., USA) using a handmade PTFE column of 0.2 mL volume 34 . After sample solution loading, the column was rinsed with 2 mL of 6 M HNO 3 and 0.5 mL of 3 M HNO 3 successively, and Sr was eluted with 2 mL of 0.05 M HNO 3 . Finally, the separated Sr fraction was reacted with one drop of concentrated HNO 3 to decompose resin-derived organics. The total yield of the Sr separation chemistry was approximately 94% 34 . After the Sr separation chemistry, the Zr/Sr ratio was reduced to 1.2 × 10 -5 times the original ratio of the sample. Instrumentation. In this study, two thermal ionization mass spectrometers were used, the Triton™ XT (Thermo Scientific™) at Thermo Fisher Scientific, Bremen, Germany, and the Triton™ Plus (Thermo Scientific™) at Fukushima University, Japan. Both instruments are equipped with eight movable Faraday cups, an axial Faraday cup, and an axial secondary electron multiplier (SEM). The Faraday cups are connected with the standard 10 11 Ω amplifiers. A set of RPQ lenses, which is an energy filtering device, is equipped in front of the SEM detector. The configuration of the detectors was summarized in Table S4 in supporting information. The peak position of 88 Sr and 90 Sr with this collector setting was shown in Figure S7 in supporting information.

Quantitative analysis of
This study uses both Re and Re-ZR (rhenium zone-refined) single filaments. High temperature bakeout of the filaments at 5.5 A filament current were performed under high vacuum for 30 min (Bremen) or 60 min (Fukushima). In Fukushima, the baked filaments are further inspected for 90 Zr emission. 90 Zr intensity of the  Sr ratio was determined as the 2σ internal error, which is two times the standard error of the 90 Sr/ 88 Sr ratios of 220 or 860 cycles. In all the measurements, the analytical error of the 90 Sr/ 88 Sr ratio was comparable with the √n counting error of the accumulated 90 Sr count before noise correction. Instrumental isotope fractionation is a possible error source of the 90 Sr/ 88 Sr ratio measured using mass spectrometry. However, the variability of the average 88 Sr/ 86 Sr values caused by the instrumental mass fractionation during the measurement was ± 0.5% ( 88 Sr/ 86 Sr values from 8.31 to 8.41) and was negligible compared with the analytical error of 90 Sr/ 88 Sr ratio (> 10%, 2RSD) in this study. Therefore, no correction for instrumental isotope fractionation was made.
Concentration of Zr in NIST 987 and Ta-activator solutions were measured ICP-MS with a dilution ratio of 10,000. For both reagents, Zr concentrations were under detection limit, which limits the Zr concentration of the original solutions as < 1.6 × 10 2 ppb. Noise correction scheme for 90 Sr signal. The stable nature of the noise signal allows precise noise correction on 90 Sr measurement. Our noise correction scheme is as follows. First, the average 90 Sr noise signal value was determined using multiple measurements of 90 Sr-free NIST SRM-987. Then, the value is subtracted from 90 Sr intensities of every measurement cycle in the run. All the data in the measurement were re-calculated offline using the noise corrected 90 Sr intensities, and finally, the noise corrected 90 Sr/ 88 Sr ratios are determined. The correction value should be determined every analytical session because the noise intensity level is sensitive to the SEM parameters and analytical conditions. A single correction value is used for all measurements within the analytical session.

Conclusions
This study demonstrates the ability of a TIMS instrument to perform an accurate 90 Sr/ 88 Sr ratio measurement of samples with 90 Sr/ 88 Sr ratios as low as 2.7 × 10 −12 . Not only the introduction of effective hardware such as the RPQ lenses but a thorough investigation and elimination of noise signals have resulted in a constantly low noise signal of 90 Sr on the 0.001 cps order. The sample throughput of this technique was 40 samples per day for Sr separation chemistry and 21 samples per 2 days for isotope ratio measurement. This is not as fast as the ICP-MS techniques but is significantly faster than radiometric methods. To detect 0.0012 mBq of 90 Sr, this energy-filtered TIMS technique requires a sample size of 100 ng of Sr. This sample size corresponds to 12.3 µL of seawater or less than 10 mg of biological samples with Sr abundances larger than 10 mg/kg, which seems easy to handle. The lowered 90 Sr detection limit and smaller sample size of this technique are suitable for studying environmental diffusion of radioactive materials, as well as environmental elemental cycling studies using nuclear test origin 90 Sr in the earth surface materials.