Direct Measurement of Femtomoles of Osmium and the 187Os/186Os Ratio in Seawater

See allHide authors and affiliations

Science  09 Oct 1998:
Vol. 282, Issue 5387, pp. 272-274
DOI: 10.1126/science.282.5387.272


Two depth profiles of the osmium concentration and the187Os/186Os isotopic ratio in the Indian Ocean showed that the osmium concentration seems to be unaltered by chemical or biological processes occuring in seawater; accordingly, osmium is conservative. These data were obtained from an experimental method that eliminated the problems related to osmium preconcentration. This method led to a new evaluation of the concentration of osmium in seawater; the mean concentration of osmium and the187Os/186Os ratio are equal to 10.86 ± 0.07 picograms per kilogram and 8.80 ± 0.07, respectively. The results suggest the existence of an organocomplex that dominates the speciation of osmium in seawater.

The187Os/186Os ratios in marine sediments are considered to indicate the relative variations of continental, mantle, and meteoritic input to seawater and to provide a record of continental weathering. However, to link the Os isotopic composition of seawater to geologic events, one must know the residence time of Os in the oceans. The residence time determines the response of the ocean to variations in the input of Os and, consequently, the shortest period and smallest amplitude of change that are resolvable by the present analytical precision of Os measurements (1). Until recently (2, 3), no direct measurement of the present-day composition of seawater Os could be made because of analytical difficulties. Here we present two depth profiles of the concentration and isotopic composition of Os that were measured in the Indian Ocean with an analytical technique (4) that had been adapted to water analysis.

Seawater samples were obtained in August 1997 at two different sites along the southwest Indian Ridge, during the EDUL cruise of the N/OMarion Dufresne. Seven samples were taken from the CTD4 site (27°52′S, 63°51′W), and nine samples were taken from the CTD12 site (34°11′S, 55°37′W) (5) (CTD, conductivity-temperature-depth). Our measurements, made on unfiltered samples, can be compared with measurements that were obtained earlier on unfiltered samples from the Atlantic and the Pacific Oceans (2).

The chemical separation method that we used was derived from the method described by Birck et al. (4, 6). Osmium isotopic ratios (7) and concentration measurements (Table 1) were constant within analytical uncertainty along each profile and from one site to another (the sites are separated by several hundred kilometers). Duplicate measurements were made on some samples to confirm the reproducibility of the results. The total procedural blank amounted to 4% of the sample total and remained constant within 15% throughout the duration of this study (8). The best statistical estimate of the mean of the187Os/186Os ratios (Fig. 1A) over the two profiles is 8.80 ± 0.07 (2σ error). This ratio is in agreement with the ratios measured by Sharma et al. (2) (8.7 ± 0.2 for the North Atlantic Deep Water and 8.7 ± 0.3 for the central Pacific Ocean) but is higher than most of the ratios obtained from bulk or leached marine sediments (9), with the exception of the ratios obtained from some organic-rich sediments (10). This discrepancy is understandable because most of these sediments contain nonradiogenic Os in addition to their hydrogenous source. Although our profiles were on a ridge, we did not observe any decrease in the187Os/186Os ratio with depth, contrary to the measurements from the Juan de Fuca ridge (2).

Figure 1

Two depth profiles of (A) the187Os/186Os ratio and (B) the normalized Os concentration (to 35 per mil salinity) in the Indian Ocean. Solid circles represent CTD4 sites, and open circles represent CTD12 sites. The vertical lines represent the average values. Error bars are 2 SD.

Table 1

Concentration and isotopic composition of Os in the Indian Ocean water. Concentrations and isotopic compositions are blank corrected. Precisions are expressed in 2σ. N, normalized to 35 per mil; σ, standard deviation; (1), first sample; (2), second sample.

View this table:

The best statistical estimate of the mean concentration of the samples is 10.86 ± 0.07 pg/kg (normalized to 35 per mil salinity) or 10.77 ± 0.07 pg/kg (unnormalized; 2σ error). This concentration is three times that previously measured in the Atlantic and the Pacific Oceans (2, 3). Because the range of the individual uncertainties (2 to 7%, except sample CTD12-7) of the Os concentration is similar to the whole variation over the two profiles, we cannot assess whether there is any correlation with salinity, but the constancy of the concentration profiles within 6% (Fig. 1B) strongly indicates that Os concentration is unaltered by chemical or biological processes that occur in seawater; hence, Os concentrations exhibit a conservative behavior.

It seems unlikely that the discrepancy between Os concentration data from this study and data from previous ones (2,3) represents a real difference between the major oceans because the homogeneity of the 187Os/186Os ratio in the oceans [deduced from our data and the data in (2)] indicates that Os is reasonably well mixed. Thus, the observed difference probably results from differences in the analytical procedures.

It has been thought that some Os could be adsorbed to the walls of the storage vessel. With radioactive 185Os as a hexachloroosmate (OsCl6 2–), Koide et al. (3) estimated a loss of Os tracer in Pyrex bottles that was equal to 0.5% per day. Two of our samples, CTD4-9 and CTD12-11, were duplicated after 41 and 36 days, respectively. They should have lost 17 and 19% of their Os, respectively (3); instead, we observed no Os loss. Moreover, the same Os concentration was obtained for all our samples, even though we completed our analyses over a period of 3 months (which should imply a loss of 36% of Os between the first and last analyses). These results show that natural Os does not behave in the same way as the OsCl6 2– tracer and that the recovery yield of Os that Koide et al. estimated is not indicative of the true yield.

To design a procedure that would enable the separation of Os from seawater, Sharma et al. (2) conducted experiments using seawater solutions that were doped with an Os standard, to which an Os tracer was added (both as OsCl6 2–). They obtained nearly quantitative recovery yields and tracer sample equilibration. The lower Os concentrations that they measured suggest that their recovery of natural Os was not quantitative (even if reproducible) and, again, that Os is not in the OsCl6 2– state in seawater.

Homogenizing tracer and sample Os before extraction is difficult. The critical period for the Os concentration measurement with our method occurs during the oxidation step, which permits the isotopic equilibration and the separation of Os from seawater into Br2. Duration tests were made on sample CTD4-6 to ensure that the oxidation step was long enough to oxidize all the Os contained in the sample (Fig. 2). It appears that 48 hours are required for a sample heated to 90°C in an oven to reach an Os concentration plateau (11). The long duration of the oxidation step in rather harsh conditions suggests that Os speciation is not, or is not dominantly, H2OsO5 or H3OsO6 , in which species Os is already in the most oxidized state.

Figure 2

(A) Plot of the187Os/186Os ratio versus the duration of the oxidation step of the chemical separation procedure. The isotopic ratio is constant, and its mean is 8.72 ± 0.10, in agreement with the mean of all samples. (B) Plot of Os concentration versus the duration of the oxidation step of the chemical separation procedure. On a heating plate, the duration is much longer because the heating conditions are much weaker; the vessel is only heated from below. Solid circles represent aliquots that were heated in an oven at 90°C, and open circles represent aliquots that were heated on a plate at 160°C. Error bars are 2 SD.

Hence, none of the species that were previously proposed on the basis of thermodynamic calculations seem to satisfy our observations (12). In these calculations, neither the presence of particles nor the presence of organic matter was considered, owing to a lack of thermodynamic data.

The small mass of particulate matter in seawater (13) and the low Os concentration of continental material (14) show the contribution of terrigenous particles to be negligible. Meteoritic dust is also inadequate because of the discrepancy between the187Os/186Os ratio of seawater (8.8) and that of meteorites (∼1) (15).

However, our analytical procedure is able to oxidize scavenged metals and destroy oceanic organic matter (16), which did not occur in previous studies (2, 3). The difference in concentration between our data and the previous data could be interpreted as a result of organic trace metal speciation. A notable affinity of the platinum-group elements (to which Os belongs) for organic compounds (17) strengthens this hypothesis. Particulate organic matter (POM) concentration declines rapidly through the thermocline and reaches a low level in deep seawater (13). This enrichment of the surface water is due to the biological activity that photosynthesizes and permanently recycles organic matter through the diverse trophic levels. Only a part of the POM escapes being recycled in the surface water and is oxidized below; it is unlikely that this phenomenon would induce the same concentration of Os in the deep ocean as in the surface water. If Os comes from organic matter, the matter is certainly dissolved organic matter.

The residence time of Os (τOs) has been estimated indirectly with several different methods (1, 2,10, 18, 19), all of which give estimates that are between 104 and 105 years. Knowing the Os concentration of seawater is a first step toward a direct calculation; to complete this calculation, we also need an evaluation of the input or output fluxes. Analyses of the Os concentration of some of the world's largest rivers (19) show that their Os concentration is similar to that in seawater. The precision is not yet sufficient to reduce the uncertainty by much, because the variation of the mean Os concentration of rivers could be estimated as being between one-half and twice the Os concentration in seawater. This results in the same residence time estimate [between 6.5 × 104 and 1.6 × 104 years (20)] as previous estimates.


Stay Connected to Science

Navigate This Article