Technical Comments

Questions Regarding Precambrian Sulfur Isotope Fractionation

Science  15 Jun 2001:
Vol. 292, Issue 5524, pp. 1959
DOI: 10.1126/science.292.5524.1959a

The discovery by Farquhar et al. (1) of mass-independent isotope fractionation of33S and 36S in rocks formed more than ∼2 billion years ago (Ga), but not in younger rocks, has boosted the theory postulating a dramatic change from an anoxic to an oxygen-rich atmosphere about 2 Ga. That is because the only known natural process that may cause mass-independent fractionation of both 33S and 36S involves atmospheric photochemical reactions by ultraviolet light in the absence of an ozone shield (1). Here, we suggest the strong possibility that the fractionation observed by Farquhar et al. (1) may have been created by analytical rather than natural processes.

The magnitude of mass-independent fractionation of 33S or36S, expressed as Δ33S or Δ36S, is defined as the deviation of a measured δ33S or δ36S value from the δ33S or δ36S value expected from the mass-dependent fractionation relationships. That is,Embedded Image(1)andEmbedded Image(2)when |Δ33S| or |Δ36S| ≥ 0.1‰ mass-independent fractionation is indicated.

Farquhar et al. (1) determined the values of δ33Smeas, δ34Smeas, and δ36Smeas on SF6 gas generated from a sample. When the SF6 is not completely purified by a gas chromatographic column, the presence of impurity gases (C-F-S-O-H compounds) may cause the δ33Smeas and especially the δ34Smeas and δ36Smeas values to differ considerably from the true values (2), and thereby to create an apparent mass-independent fractionation. Four observations suggest that such problems exist in the data reported by Farquhar et al. (1, 3).

1) Most of the δ34S values determined by Farquharet al. (1) differed significantly from those determined on the same samples by other researchers using the conventional SO2 method, a more reliable approach for determining δ34S values but not for determining δ33S and δ36S (2,4). The difference between the δ34Smeas values obtained by Farquhar et al. (1) and those obtained by the other investigators was typically between 1 and 10 per mil (‰), rather than the acceptable difference of less than ± 0.5‰. As equations 1 and 2 suggest, if the δ34Smeas value differs from the true value by +3‰ but the δ33Smeas and δ36Smeas values are accurate, the apparent Δ33S and Δ36S values will become –1.5‰ and –5.7‰, respectively—and will therefore pass the threshold of significant mass-independent fractionation—even if there is no natural mass-independent fractionation in the original rock sample. If the δ33Smeas or δ36Smeas is also inaccurate, as may have been the case in the Farquhar et al. study (as discussed below), the apparent Δ33S and Δ36S values can differ greatly from the values above.

2) On some samples, Farquhar et al. (1) generated two to three sets of SF6 gas through successive acid treatments. Essentially, all of the sulfur in these rocks (except for one sample) was originally in the form of disseminated pyrite, but some may have been converted to sulfate by recent oxidation (5). The SF6 gases from such a rock sample may have different sets of δ33S, δ34S, and δ36S values because of kinetic isotope effects during the recent oxidation or the acid treatments. They should, however, have identical sets of Δ33S and Δ36S values to within ±0.05‰, if there were no analytical problem. This was not the case among the samples analyzed by Farquhar et al. (Fig. 1).

Figure 1

Observed mass-independent fractionation of33S and 36S on carbon-rich samples, adapted from (1). Tielines connect the Δ33S and Δ36S values of different SF6 gases from the same rock samples.

3) Some natural processes produce mass-independent fractionation of33S with or without mass-independent fractionation of36S, but no natural process is known to produce mass-independent fractionation of only 36S (1). The samples dating from less than 2.0 Ga, however, showed mass-independent fractionation of only 36S (Fig. 1), an indication that serious analytical problems exist for Δ36S, not only for these younger samples but also for older samples. True mass-independent fractionation of 36S thus may not have existed in any of the rock samples.

4) Mass-independent fractionation of 33S or 36S is commonly found in carbon-rich rocks, such as shales bearing disseminated pyrite. Our own experience suggests that such rocks are more likely to produce impurity gases (C-F-S-O-H compounds) during the analytical process. None of the samples younger than 500 Ma analyzed by Farquhar et al. showed mass-independent fractionation of either 33S or 36S (5); these samples were all carbon-poor, simple mineral separates (5) that were less likely to produce impurity gases. Therefore, the conclusion by Farquhar et al. (1) that the mass-independent fractionation was found only in rocks dating from more than 2.0 Ga is analogous to comparing apples (carbon-rich rocks) and oranges (carbon-poor rocks).

Farquhar et. al. (1) are to be commended for recognizing the possible presence of mass-independent fractionation of sulfur isotopes in geological samples. The observations reported above, however, lead us to seriously question the validity of the major conclusion of that study that photochemical reactions were responsible for mass-independent fractionation in rocks dating from earlier than 2.0 Ga. To establish the true changes in mass-independent fractionation of sulfur isotopes over geologic time, more systematic investigations must be carried out on a variety of samples of all ages—especially the disseminated-pyrite–bearing shales from <2.0 Ga, which have not yet been investigated.


Response: Ohmoto et al. contend that the mass-independent fractionations we reported are an artifact of our analytical technique. Here, we report repeat analyses using our technique and reanalyses using an independent technique (laser fluorination) that attest to the robustness and accuracy of isotopic data previously reported in (1). These analyses, and further independent measurement of the effect by secondary ion mass spectrometry (2), confirm our original measurements and support our conclusions.

Ohmoto et al. argue that C-F-O-H-S contaminants derived from organic material generate mass interferences that jeopardize our analyses. Rumble et al. (3) have demonstrated, however, that a single pass through a gas chromatograph is adequate to purify SF6 for Δ33S analysis. We have used gas chromatography to purify SF6 before mass-spectrometry for more than 10 years (1,4–11) and have repurifed samples that we have previously analyzed as secondary checks for δ36S. Although we have not found it necessary to repurify our samples for δ33S, we did so for the very first mass-independent sample we analyzed, pprg 199 (1). We found our analyses reproduced within 0.04‰ for both δ34S and Δ33S. We also note that the mass interference invoked by Ohmoto and colleagues produces an effect that is in the wrong direction to explain the Δ33S that we found in carbon-rich shales.

Although Ohmoto et al. hold that the SF6technique is less reliable than the SO2 technique, the SF6 technique is actually well established (3, 4, 6, 7,9–15); indeed, it has been validated by the decision of the International Atomic Energy Agency (IAEA) to define the isotopic composition for the international Canyon Diablo Troilite (CDT) standard as identical to the value measured by the SF6 technique, and not the companion measurements using the SO2 technique (16,17). The isotopic composition for seawater sulfate is also defined using the SF6 technique rather than the SO2 technique (12, 13). To demonstrate the reproducibility of our fluorination technique for the Archean samples, we refluorinated previously extracted sulfur (in the form of pure Ag2S) from samples highlighted by Ohmotoet al. (Table 1A and Fig. 1). The δ34S of the samples reproduce within their stated uncertainties of ± 0.3‰ and are confirmed to differ from the previously reported values cited by Ohmoto et al. Triplicate analysis of Ag2S from sample pprg 480, for example, yielded 12.8 ± 0.1‰, which differs from the previously reported value of 6.0‰. Independent analyses of the Ag2S by two of the authors of this response (Hu and Rumble) fall within the error of our previous δ34S measurements (Fig. 1A).

Table 11

Sulfur isotope analyses.

View this table:
Figure 1

(A) Plot of δ34S data remeasured in 2001 against δ34S data reported in (1). (B) Plot of Δ33S data remeasured in 2001 against Δ33S data reported in (1). Data were determined from the same Ag2S samples and illustrate that the discrepancy between previously reported δ34S values (19) and those reported in (1) is not due to mass interference associated with the SF6 technique but instead likely traces to sample heterogeneity, such as that reported in (20). Data also illustrate that the reproducibility of Δ33S is comparable to the uncertainties reported in (1). UCSD, repeat analyses undertaken at The University of California, San Diego, using techniques of (1); Geophysical Lab, reanalyses of samples from (1) undertaken at Geophysical Laboratory, Carnegie Institution; Strauss and Moore, values reported in (19).

Ohmoto et al. maintain that our different acid extractions “should . . . have identical sets of Δ33S and Δ36S values to within ± 0.05‰.” Our reanalysis of sulfur extracted from sample pprg 2777 (Table 1A and Fig. 1) reproduce the mass-independent measurement well within the stated ± 0.3‰ analytical uncertainties. Likewise, our previous measurements of barite (Table 1B) illustrate the reproducibility of our technique for determination of Δ33S for sulfur from samples that are not carbon-rich shales. Independent analyses of our Ag2S by laser fluorination fall within error of our previous Δ33S measurements (Fig. 1B).

According to Ohmoto et al., our Archean Δ36S data are problematic because “samples dating from less than 2.0 Ga . . . showed mass-independent fractionation of only36S.” In (1), we defined Δ36S as equal to δ36S – 1000[(1+δ34S/1000)1.90–1], with the exponent of 1.90 chosen on the basis of statistical-thermodynamic theory (18). When we regress all of our measurements of present-day δ36S against δ34S, we obtain an empirical exponent of 1.84. As far as we are aware, such a value is not an allowed theoretical mass-dependent slope, and we are now working on understanding its origin. The source of the negative Δ36S that Ohmoto et al. note in samples younger than 2.0 Ga is an artifact of the calculation using the exponent 1.90 instead of 1.84. When the data plotted by Ohmoto et al. are recalculated with an exponent of 1.84 instead of 1.90 (Fig. 2), the supposedly anomalous younger samples plot near the origin (within experimental error), and the relationship between Δ33S and Δ36S for samples older than 2.4 Ga persists. Also, we point out that in many cases the σ error bars overlap for data from the same sample.

Figure 2

Plot of Δ36S against Δ33S for selected data from (1) and (21), using exponent of 1.84 instead of 1.90. Error bars are from (1) and (21).

Ohmoto and colleagues imply that we observed anomalous Δ33S only for sulfur from carbon-rich shales. The barite data and selected other data (Table 1, B and C), however, came from pure mineral separates and are not from analyses of organic-rich whole-rock shales. We observed mass-independent fractionations in carbon-poor rocks and in minerals, as well as in carbon-rich rocks that are older than 2.0 Ga. Over the past 10 years we have analyzed hundreds of samples for δ34S, and we have yet to observe mass-independent fractionations in any rock type—carbon rich or carbon poor—younger than 2.0 Ga. We will continue to test this hypothesis with additional analyses, but at present have no indication that the assertion of Ohmoto et al. is valid.

We have considered the criticisms made by Ohmoto and colleagues with the seriousness they deserve; however, our analytical data and reanalysis using independent methods argue against those criticisms, and support our original interpretations. We invite other laboratories to undertake further tests of our hypotheses and methods and will continue to do so ourselves.


Related Content

Navigate This Article