4. DISCUSSION
As explained in the introduction, the consensus values for SLAP with respect to VSMOW were established in 1976. The established δ2H value was based on the absolute abundance measurements, however, the same for δ 18O was lacking and thus the mean δ 18O value, -55.5‰, of an interlaboratory calibration exercise performed at that time was chosen by consensus. Among the representatives of the several participating laboratories, there was already a discussion that possible memory effects would contract the scale, so probably a more negativeδ 18O value would have been more appropriate.
In later years, thanks to improvements to both equipment and analysis procedures such as correction for cross-contamination (Meijer19, laboratories indeed determined more negative values for SLAP. In our laboratory, we typically find values around δ 18O = -55.85‰ by IRMS (CO2-H2O equilibration) and, more recently to our surprise, δ 18O = -55.7‰ using the LGR-LWIA. We expected that by having a well-maintained IRMS and using the appropriate corrections, our results for SLAP would be close to the real values.
However, Kaiser (2009) already suggested a re-analysis of the data of an intercomparison exercise of 7 expert laboratories described in Verkouteren and Klinedinst (2004), resulting in a much more negativeδ 18O value for SLAP, i.e., -56.1 ± 0.2‰. On the other hand, the δ 18O value of -55.11‰ for SLAP measured by Barkan and Luz8, is puzzling. The method Barkan and Luz used was also based on the isotopic exchange equilibration between H2O and CO2 in sealed ampoules, but followed by a fluorination of water using CoF3 to produce O2. Although this approach is different from the standard equilibration method, results should be identical as long as the fluorination is complete. However, their approach consistently points towards less negative values of -55.11‰ (Hillaire-Marcel22).
For a robust locking of the second anchor of the VSMOW scale we performed the work described in this paper. The reliability of our method of quantitative 18O abundance determination of18O water using Quadrupole Mass Spectrometry is crucial for our results. Taking various effects such as fragmentation difference of H216O and H218O into account, and by validating the method with a dilution series, and considering the excellent agreement of the fitted QMS signals and the measured ones, we are confident that the method is reliable.
A systematic deviation of our 18O abundance result of 0.1% higher/lower values would lead to a more/less negative result for SLAP of 0.05‰. However, such a deviation is highly unlikely: It is good to realize that, as we use very highly-enriched 18O water (batches of 98% and 99% 18O), in fact we do not measure this high 18O abundance, but rather quantify the remaining part of 16O exactly by QMS. Since there is only room for 1 to 2% 16O, it is in fact this amount that has to be measured with an accuracy of ≤ 0.1%, which is not a high relative accuracy. Furthermore, if we would still suffer from some systematic deviation, one can expect this deviation to be larger for the water portions with 2% 16O remaining (the Rotem waters) than those with 1% (Cortec). We see no such effect in our results (Figure 6 and in the supplementary material Table 4). The portion of 17O only plays a minor role in the 18O/16O ratio, and this abundance can be determined using a dilution method.
The uncertainty in 18RVSMOW leads to a systematic uncertainty in our final answer of ± 0.013‰, small compared to our final uncertainty.
So, the result of this study is δ 18O -56.33 ± 0.03‰, thus a very negative δ 18O value for SLAP, which was an unanticipated finding.
The implication of this is that apparently complete understanding of all IRMS effects (not to mention those in optical spectroscopy) is still lacking. Measuring cross-contamination-effects (Meijer19) obviously is not enough for correcting the isotope measurement such that the measured delta values are very close to the real delta values.
One of the issues emerging from this lack of complete understanding of all IRMS effects, relates specifically to second order measurements such as 17O excess (Δ 17O) in water. For these measurements, in which the small deviation of the measured δ17O from the natural relation between δ18O and δ17O is determined (Meijer and Li20, Aron21), the question raises, how well these very small deviations (around 0.02‰ or less) can be defined, if there are such large discrepancies between measured δ18O and real δ18O values. The assumption that 17O and 18O will fully obey mass dependent fractionation in the ion source of the IRMS may not be completely true. To put it another way: if the measured scale for δ18O is already so much contracted, who can guarantee that the δ17O scale contracts exactly according to the equilibrium relation between δ17O and δ18O.
Also clumped isotope measurements, which determine the minute deviations from stochastic distribution of the delta values for multiply substituted isotopologues, can probably not rely on the fully mass-dependent scale contraction of their machines. Also, here, full understanding of IRMS effects is key.
The oxygen isotope compositions are typically reported on the VSMOW scale, not only for water samples, but also for other types of samples, such as oxides and silicates. The VPDB scale is mostly used for reporting the stable isotope (carbon and oxygen) results of carbonate minerals and also for oxygen isotope measurements in atmospheric CO2. These two coexisting stable isotope scales for reporting 18O/16O ratios or δ18O values, can be converted into each other (Hillaire-Marcel22). For both scales an extra conversion step to CO2 is necessary, because the measurand in the IRMS is CO2. This extra reaction step is for the VSMOW-CO2 scale, water equilibration of VSMOW with CO2 under standard conditions (first described by Epstein and Mayeda23, Meijer17 and for the VPDB-CO2 scale, acidification of IAEA-603 (formerly NBS-19) with phosphoric acid (McCrea24, Meijer17, Hillaire-Marcel22. The difference between VSMOW-CO2 and VPDB-CO2 on the two δ 18O scales is 0.28-0.29‰ (Hillaire-Marcel22). In our laboratory, we have the habit to realize the two scales (water and carbonate) independently and use this scale difference as a quality check. When using two-point calibration scales, the result of a more negative δ18O value for SLAP, (the second anchor of the VSMOW scale), could give potential discrepancies in the transfer ofδ 18O from and to the VPDB scale. Considering the fact that the water equilibration reaction is more robust and easier to control (and therefore more reliable and accurate) than the carbonate-acid reaction, we propose the VSMOW-CO2δ18O scale be defined as the primary δ18O scale. The definition of the VPDB-CO2 scale could then simply be expressed in terms of the VSMOW-CO2 scale. Final decisions about these isotopic scales are under the auspices of the commission on isotopic abundances and atomic weights (CIAAW).
Identical treatment of sample and references, the frequent use of international reference materials and clear guidelines about how to express the results on the international scale(s) is key to provide normalized interlaboratory-comparable stable isotope measurements. This study does not affect those measurements; the VMSOW-SLAP scale can be taken as is. However, knowing the absolute ratios and/or abundances of all scale determining references would give us clear insight how large the scale contraction processes really are. In fields where VSMOW-SLAP-scaled δ18O values are converted into absolute abundances and vice versa, our new δ18O value for SLAP does matter. An example of such a field is energy expenditure measurements using doubly-labelled water, in which the used enriched reference waters will change their delta value (Faghihi15).