Figure 2. A typical QMS spectrum of a highly enriched18O water from m/z 1-40. The insert plot shows the partial pressure (Pa, plotted on a logarithmic scale) with respect to m/z 14 to 24.
In the m/z range 16-24, several signals could not be used for our fitting analysis of the 18O concentration, either because of the interference of other species, or because of the very low signal. Oxygen ([16O]+) from air interferes with oxygen ([16O]+) from H216O (m/z 16). Injecting water without air is virtually impossible, and small leakages are always present as well. The origin of this interference is air is clear from its correlation with m/z 14 ([14N]+ from air). Therefore m/z 16 was disregarded from the fit. This interference is small, and hence the consequential interferences on m/z 17 and m/z 18 from air due to [17O]+ and [18O]+ are orders of magnitude smaller and therefore negligible. Additionally, the very minor signals arising from the various clumped isotopocule ions on m/z 22-24 (see Table 5 in supplement) are too small to be of use.
In the spectrum of an 18O enriched water (Figure 2) a very small signal from the recombined ion [18O2]+ is visible at m/z 36. Approximately 1.5% of the spectrum is in the form of [O]+ (all three different oxygen isotopes together), (see Table 5 in the supplement). Including the signal at m/z 36 in the spectral fit showed that about 7% of the [18O]+ ions, recombines to [18O2]+. This signal at m/z 36 doesn’t significantly influence the fitted18O value and was therefore neglected in our fitting program.
At m/z 1 and 2 signals from [1H+] and [2H+]/[H2+] are visible in the spectrum (Figure 2). As these hydrogen fragments do not contain oxygen, they were not included in the fitting program.
Five m/z values in the range 17-21 could be used for a successful fit, yielding the 18O concentration. The fitting program was written in R. The output of this R program, the fit parameters, were besides the abundance of 18O, the size of the fractions [H2O]+, [OH]+, [O]+ and thus the size of the complementary fraction [H3O]+ as well. Next to the signals m/z 17-21, the abundances of 17O and2H were also input parameters for the fitting program. The abundances of 17O and 2H of the highly enriched 18O water were separately determined to reduce the number of fitting parameters, which is necessary as17O and 18O in the fit are in fact quite correlated. Determination by dilution and comparison with reference waters is adequate in these two cases, as neither2H nor 17O abundances are very critical in the fitting process. This is due to the fact that both abundances are low anyway: 2H because it is in the natural range, and 17O because we deal with highly enriched 18O waters. Because of that, there is only room for ≤ 1% 17O, and determination of the17O abundance with a relative precision of 5% is already more than adequate. Such precision is well-achievable using dilution.
For determining 17O, the enriched waters were diluted and measured alongside references IAEA 607, 608 and 609 (Faghihi15 and CIO laboratory standards, using the LGR-LWIA. For determining 2H concentration, the diluted enriched waters were measured alongside CIO laboratory standards using the LGR-LWIA as well. In both cases we calculated the abundances from our isotope delta-measurements using the literature values for the abundances in VSMOW (Hageman4 for 2H and Li14 for 17O). The results of the 18O, 2H and17O abundances corresponding to all the highly enriched 18O water portions are presented in Table 3 (results section), along with their uncertainties.
SLAP-rep-O was mixed with highly 18O enriched water to mimic VSMOW, and referred to as VSMOW-rep-O (analogous to SLAP-rep-O, VSMOW-rep-O refers to water with an isotopic δ18O value close to VSMOW). We added a known quantity of highly enriched 18O water needed to shift the δ18O to 0‰ when measured against VSMOW. SLAP-rep-O was weighed on a precision balance (readability 0.01 g) in a 1 L Duran brown glass flask. H218O was weighed on an analytical balance (readability 0.01 mg) in a small glass vial. This vial was submerged in the 1 L flask with SLAP-rep-O. To ensure complete mixing, the resulting mixture, VSMOW-rep-O, was stirred for at least 48 hrs. Accurate determination of the weights of the mixing water portions is extremely critical in the whole calculation chain, therefore weights are also corrected for buoyancy effects, as the density of H218O water is significantly larger than that of H216O (1.11 g/mL instead of 1 g/mL). The weighing was performed as fast as possible to keep evaporation of water to a minimum.
Stable isotope measurements were performed using the LGR-LWIA. The replicates were measured alongside the real VSMOW and SLAP, such that scale contraction issues played no role (see below).
The mixing process started with the characterization of the individual 1 liter batches of SLAP-rep-O water, by direct comparison with SLAP. We then measured the produced VSMOW-rep-O by direct comparison with original VSMOW water. The difference between this measurement and the calculated value translates directly into a best δ18O value for SLAP with respect to VSMOW. As we took care that both the δ18O differences SLAP-rep-O vs SLAP and VSMOW-rep-O vs VSMOW are small, their differences could be determined precisely. As these differences between the replicates and the genuine VSMOW and SLAP are small, the δ18O difference between the officially δ18O values (VMSOW-SLAP scale) and the ’true’ isotopic difference did not play a role. The calculation of the resulting δ18O value for SLAP is straightforward, and has been performed with the help of a validated spreadsheet (Faghihi16).
2.3.2 Approach 2
Highly enriched 18O water is not enriched in deuterium (on the contrary, compared to water with natural abundances it is depleted in deuterium). Therefore, after adding H218O to SLAP-rep-O, δ18OVSMOW-rep-O is close to δ18OVSMOW but δ2HVSMOW-rep-O is still close toδ 2HSLAP. In principle, this does not matter for our experiment, as we only are interested in the18O side. However, to exclude the possibility that this large difference in deuterium content between our VSMOW-rep-O and the real VSMOW would influence the absorption of the18O line in the LGR-LWIA (and thus its determination of the δ 18O difference between VSMOW-rep-O and VSMOW), in addition an extra step in the process was introduced. Before SLAP-rep-O was mixed with H218O, pure2H2O was added to mimic VSMOW in deuterium (called VSMOW-rep-D, δ 2H ≈ 0‰). Subsequently VSMOW-rep-D and highly enriched 18O water were mixed to get VSMOW-rep-OD (δ 2H ≈ 0‰ andδ 18O ≈ 0‰). This second approach is shown on the righthand side of Figure 1. It rules out spectroscopic biases in the measurements, but otherwise is not different from the process described in 2.3.1.
We started with the same Antarctic water as described before and added Groningen tap water to produce SLAP-rep-O. Then we added2H2O to mimic VSMOW in deuterium and therefore very precise quantification of the2H2O content was key. Determination of2H abundance of the enriched2H2O water by QMS, however, was not as straightforward as determination of 18O abundance of the enriched H218O. This may be caused by the more complex spectrum for 2H2O. The 2H2O spectrum, Figure 3, illustrates that m/z peaks 17,19 and 21 are about two orders of magnitude smaller than the adjacent m/z peaks 18 and 20. In the supplementary material, Table 6 shows a highly enriched2H water with the various isotopologues and fragments for this range of m/z values.
Peak tailing and leading of the larger peaks makes it difficult to integrate the smaller peaks. These alternating small and large peaks, are not present in the 18O spectrum (Figure 2, see logarithmic insert plot). Another possible explanation is the common knowledge that in high vacuum stainless steel tubes, there is always outgassing of hydrogen. If this is the case in the QMS, H-exchange will affect deuterium abundance measurements with the QMS, especially for these nearly pure 2H2O waters. This outgassing of hydrogen will obviously not affect the determination of oxygen isotope abundances.
Furthermore, the m/z 1 the signal was much larger than we expected from a nearly pure 2H2O water (m/z 1 is approximately 1% of m/z 2, see Figure 3, top insert plot), an observation that worried us initially. But after personal communication with the manufacturer of the QMS, Extorr, we learned that this was probably a source pressure related artifact. Working at higher pressures can cause scattering. If the QMS is not tuned for these low m/z values, a fraction of the scattered ions passes through the mass filter below 0.5. The actual m/z 1 is therefore not resolved well. This fact made signals m/z 1 and 2 useless for obtaining the deuterium concentration of the almost pure 2H2O water. Therefore, we used a similar m/z signal range as was used for the18O determination.
In conclusion, this discrepancy of measured (and fitted)2H abundance (of about 99.7%) and real (specified)2H abundance (99.98%) must be attributed to reasons mentioned above: the more complex spectrum and the continuous outgassing of hydrogen in vacuum stainless tubes. To verify the specification of the supplier we performed NMR analysis for accurate 2H concentration analysis of the highly enriched deuterated water, which corroborated the specified value, and also excluded the possibility that sample handling of these highly enriched waters would lead to dilution due to admixture of water (vapour) from the surroundings.
<figure 3>