PROCEDURE FOR PREPARING COMPARISON WATERS

Six water samples (W09–W14) were prepared using purified water (Milli-Q Gradient A10, Elix water purification system, Merck-Millipore) and guaranteed chemical reagents; more details are provided in Table 1. The Milli-Q water was purified from organic matter using ultraviolet oxidation and ultrafiltration (total organic carbon <3 ppb). The prepared water samples were then stored in gamma-sterilized polyacrylonitrile (PAN) bottles (250 mL; Nikko Hansen & Co., Ltd.). PAN bottles provide an effective barrier against gas exchange (Takahashi et al. Reference Takahashi, Minami, Aramaki, Handa and Matsushita2019).

Table 1

Chemical reagents used in 22 L for water samples W09–W14. The artificial seawater powder was Marine Art SF-1 (Osakayakken Co., Ltd); * indicates an alternative to the NaHCO₃ reagent, because of a supply shortage. The nominal ¹⁴C and DIC values were computed from the mixture ratio of the chemical reagents. The DIC concentrations of the hydroxide solutions were measured by acid titration, using bubbling solutions of NaOH and Ca(OH)₃ during the pre-test experiment.

The nominal ¹⁴C content of the samples was targeted to be 0–95 pMC (Table 1), while chemical contents (Na⁺, K⁺, Mg²⁺, Ca²⁺, Cl⁻, SO₄ ²⁻, HCO₃ ⁻, and CO₃ ²⁻) were similar to those of natural waters (Table 2). The carbon sources included NaHCO₃, gas-bottle CO₂ (Taiyo Nippon Sanso Co.), and natural shell powder (Naitoh Shouten Co., Ltd.; employed as pigment in the Japanese painting industry). The former two sources were considered dead-carbon sources, while the latter was considered modern. Shell powder was reacted with H₃PO₄ or HCl in 1 L plastic bottles to produce CO₂. The CO₂ released from the shell powder and the gas-bottle CO₂ were collected in 20 L Tedlar^® gas-sampling bags; then, the CO₂ gas was dissolved through bubbling in 1 M NaOH or 0.01 M Ca(OH)₂ solutions (360–500 mL). The NaOH and Ca(OH)₂ solutions acted as a control of the cation content and adjust it to natural water chemical compositions (Na and Ca types). The obtained CO₂ gas was pumped through both the hydroxide solutions at ~20 L hr^–1 for 3–5 hr (gas bags were exchanged several times). The bubbling was stopped when the pH of hydroxide solution reached ~7.6. The Ca(OH)₂ solution was filtered under suction (with a 5 µm glass-fiber filter), in order to remove any precipitate generated during bubbling. The carbon contamination to the hydroxide solutions produced during bubbling cannot be eliminated, but it does not really matter because a slight contamination does not influence to perform intercomparison using these waters. Additionally, we made sure that all batches of comparison water had identical ¹⁴C concentrations and chemical compositions.

Table 2

Initial (n = 1) and final (n = 12) chemical features of samples W09–W14 at the elapsed days of ~15 and ~1000, respectively. The errors are given as standard deviations across 12 batches of storage bottles. The saturation index (SI) values, in parentheses, were calculated assuming a Ca²⁺ concentration of 0.1 mg L^–1.

n.d. = not detected.

The reagents (Table 1) were dissolved and mixed in ~10 L of Milli-Q water, inside a 22 L plastic container, and then diluted with Milli-Q water reaching a total volume of 22 L. The solutions were homogenized using a magnetic stirrer for >3 hr, and then left to stand overnight. The precipitates which accumulated in sample W12 during this time were redissolved in the solution by further stirring. Each solution was dispensed by siphoning into a 250 mL PAN bottle (full filling amount = ~310 mL), leaving no headspace. We prepared 66 bottles for each comparison water. All of the bottles were stored in a dark room (~20ºC) until use.

INTER-BATCH HOMOGENEITY

Inter-batch homogeneity was assessed by determining the chemical compositions and δ¹³C values for 12 samples chosen randomly from the PAN bottle batches. The samples were analyzed by ion chromatography (Dionex ICS-2100, Thermo Fisher Scientific) and an isotope-ratio mass spectrometer (IRMS) coupled with a gas chromatography system (GC–IRMS; Delta-V Advantage with GasBench II, Thermo Fisher Scientific), at the Geological Survey of Japan. Measurement errors were lower than ±10% for the chemical composition, ±2% for the DIC concentration, and ±0.1‰ for the δ¹³C.

Standard deviations for chemical compositions and δ¹³C values after storage (~1000 days) were <1% and <0.1‰, respectively (Table 2). We concluded that the inter-batch homogeneity was high enough to allow an intercomparison.

MONITORING OF SAMPLE STABILITY

The stability of the intercomparison water samples was assessed by monitoring DIC, ¹⁴C, and δ¹³C values during storage. The δ¹³C and DIC values were determined by GC–IRMS as described above. CO₂ was extracted from water samples by a headspace gas-purge procedure (Torres et al. Reference Torres, Mix and Rugh2005; Molnár et al. Reference Molnár, Hajdas, Janovics, Rinyu, Synal, Veres and Wacker2013) in order to measure ¹⁴C. The sample water was injected into a pre-evacuated glass bottle containing H₃PO₄ sealed by a butyl rubber septum using syringe. The CO₂ released in the head-space of the glass bottle was purged and the carrier gas transferred the CO₂ to the vacuum line through a needle. Afterwards, the CO₂ gas was purified cryogenically and reduced to graphite with hydrogen gas (Kitagawa et al. Reference Kitagawa, Masuzawa, Makamura and Matsumoto1993) to make ¹⁴C measurement targets. We measured the ¹⁴C contents at two accelerator mass spectrometry (AMS) facilities, Nagoya University and National Institute for Environmental Studies (NIES), relative to NIST oxalic acid (HOx-II) and corrected it for isotopic fractionation using the ¹³C/¹²C measurements in each of the AMS systems. Corresponding results are shown in Figure 1, while the chemical features are given in Table 2.

Figure 1

Results of the stability monitoring for ¹⁴C (circles), δ¹³C (triangles), and DIC concentration (diamonds).

Takahashi et al. (Reference Takahashi, Minami, Aramaki, Handa and Matsushita2019) reported that, when DIC concentration was relatively high (~200 mM), the ¹⁴C content of a NaHCO₃ solution stored in a PAN bottle was stable over periods of up to 900 days, even if the ¹⁴C content was very low (approximating the ¹⁴C-free level). However, when DIC concentration was low (~2 mM), the ¹⁴C concentration increased from 0.6 to 1.8 pMC. Because of the low storage PAN bottle performance, it is difficult to prepare comparison water having very low DIC and ¹⁴C concentrations.

During storage, samples W09 and W12 showed significant changes in their ¹⁴C and DIC concentrations. The reasons for the decrease in ¹⁴C concentration of W09 is uncertain; possibly, it was triggered by air-exchange between the storage room (low ¹⁴C; atmospheric ¹⁴CO₂ was lower than global levels; Takahashi et al. Reference Takahashi, Konohira, Hiyama, Minami, Nakamura and Yoshida2002) and the sample. The change in ¹⁴C concentration observed for W12 was almost identical to the increase seen during the performance test in a PAN bottle (Takahashi et al. Reference Takahashi, Minami, Aramaki, Handa and Matsushita2019). The results suggested that W12 should not be used in the intercomparison.

The concentrations of most chemical species remained constant within the margin of error (error = ±10%) during storage (Table 2). Nevertheless, Ca²⁺ and DIC concentrations decreased in samples W10, W12, and W13, while Ca²⁺ was not detectable in samples W11 and W14 after storage. Moreover, a precipitate was found in sample W12 after storage. We measured ¹⁴C and δ¹³C values of the precipitate by AMS and IRMS with dual inlet system (Delta-V advantage) to be 0.8 pMC and −8.5‰, respectively. The ¹⁴C content of the precipitate was identical with that of DIC in W12 (Figures 1 and 2), indicating that the carbonate mineral was directly precipitated from DIC of W12. The x-ray diffraction analysis (CuKα; Rigaku MiniFlex) of the precipitate indicated the presence of CaCO₃ (aragonite) and MgO; perhaps, the relatively high concentration of Mg²⁺ in the sample induced the precipitation of aragonite (Burton and Walter Reference Burton and Walter1987). No precipitates were found in the other samples, although W10 and W12 displayed significant chemical and δ¹³C changes during storage (Table 2; Figure 1). Likely, some calcite precipitated at the invisible scale in low-Mg²⁺ solutions (e.g., W10, which showed an increase in the saturation index of carbonate minerals).

Figure 2

Results of the intercomparison exercise involving laboratories A–F: ¹⁴C (solid symbols), δ¹³C (open symbols). CO₂ extraction procedures: headspace-gas purging (circles), carbonate precipitation (triangles), gas bubbling (diamonds). The treatment periods were different between laboratory A (27–976 elapsed days; Figure 1) and the other laboratories (378–431 elapsed days). The ¹⁴C concentrations of some samples from laboratory A indicated secular changes (Figure 1), and therefore, a notice is necessary for the comparison of results from laboratory A with those from the other laboratories.

Samples W10 and W12 showed decreases in DIC concentration and increases in δ¹³C values. The isotopic mass-balance estimations suggested that the δ¹³C values, corresponding to the carbon removed from the W10 and W12 solutions, were −23‰ and −16‰, respectively. These values were much lower than that of precipitate (aragonite) found in sample W12. Therefore, the change in δ¹³C for W12 is caused by not only precipitation but also by other processes. The pH values of samples W10 and W12 were lower than those of other samples with higher carbonate saturation index (SI) values (Table 2). The relatively high SI values and low pH condition likely caused DIC changes during storage; waters with higher pH were apparently more stable. In conclusion, in order to maintain maximum stability, the intercomparison samples should not have very low DIC and ¹⁴C values, while pH should be >8.

TRIAL INTERCOMPARISON

The water samples prepared in this study had to be tested to determine whether they can be used in the intercomparison program. A trial intercomparison campaign was undertaken by six laboratories (laboratories A–F), in order to reveal possible problems. A set of six water samples was delivered to each laboratory. The CO₂ extraction, purification and graphitization processes were performed using usual procedures of respective laboratories. CO₂ extraction was done in three different procedures: by carbonate precipitation (laboratories B, D, and E), gas bubbling (laboratories C, D, and F), and headspace-gas purging (laboratory A). The precipitation method involved the addition of BaCl₂ or SrCl₂ to the water sample to trigger the precipitation of carbonates, and the addition of H₃PO₄ to release CO₂ (Haynes and Haas Reference Haynes and Haas1980; Drimmie et al. Reference Drimmie, Aravena, Wassenaar, Fritz, Hendry and Hut1991; Aucour et al. Reference Aucour, Sheppard, Guyomar and Wattelet1999). The bubbling procedure followed the procedure in the WOCE program (Dickson et al. Reference Dickson, Sabine and Christian2007): the CO₂ gas was extracted from acidified water by bubbling it with a carrier gas. The extracted CO₂ gas was then cryogenically purified in a vacuum line. The graphitization procedure of all participating laboratories followed that described in Kitagawa et al. (Reference Kitagawa, Masuzawa, Makamura and Matsumoto1993). The ¹⁴C content was measured by five AMS facilities in Japan: Nagoya University, JAEA-Tono, NIES, Paleo Labo Co. Ltd., and Institute of Accelerator Analysis Ltd.; isotopic fractionation was corrected using the ¹³C/¹²C measurements in each of the AMS systems. The δ¹³C values were determined by IRMS with a dual inlet system using purified CO₂.

Across all laboratories, most of the ¹⁴C results were consistent, although there were large variations in ¹⁴C and δ¹³C at low DIC (sample W09) and ¹⁴C (sample W12) concentrations. The results suggested that waters with low DIC content were probably susceptible to the carbon contamination during the CO₂ extraction process. Sample W14 presented a wide range of δ¹³C values, whereas its ¹⁴C contents were relatively similar across the laboratories. High DIC concentrations of W14 could cause carbon isotopic fractionation during the CO₂ extraction and purification processes, because of very small amount (less than 1 mL) of water used. The target range of DIC concentrations varied among laboratories: some of them could not treat such a small amount of water.

Samples W10 and W13 had the highest SO₄ ^2– concentrations. The effect of SO₄ ^2– on the δ¹³C measurements became appreciable when the mole fraction of CO₃ ²⁻ in the precipitate BaCO₃/(BaCO₃+BaSO₄) was <0.3, due to low carbon recovery (Kusakabe Reference Kusakabe2001). Conversely, at high SO₄ ^2– concentrations, there was a positive shift in δ¹³C values; this was caused by the occurrence of isotopic fractionation during precipitation. The ¹⁴C content was not expected to be affected in this case. However, the ¹⁴C data results for W10 and W13 from laboratories C and E, obtained through precipitation, differed significantly from other samples (Figure 2). Possibly, the low carbon recovery accentuated the contamination of modern carbon. Across all laboratories, the ¹⁴C results for W13 varied quite a lot, regardless of the technique used. Perhaps, high SO₄ ^2– concentrations impeded the CO₂ extraction and graphitization processes. The δ¹³C values obtained through precipitation procedure tended to deviate from other results, regardless of SO₄ ^2– concentrations (Figure 2).