Notice: Wiley Online Library will be unavailable on Saturday 27th February from 09:00-14:00 GMT / 04:00-09:00 EST / 17:00-22:00 SGT for essential maintenance. Apologies for the inconvenience.
 Abyssal peridotites, fragments of residual upper oceanic mantle, are believed to have less radiogenic Os compositions and higher Os concentrations than primitive upper mantle (PUM, 187Os/188Os = 0.129 [Meisel et al., 1996]). We have measured 187Os/188Os in 10 whole rock abyssal peridotites representing non-plume-influenced mid-ocean ridge segments. The 187Os/188Os ratios range from 0.1183–0.1582. This large range in Os composition, to both less radiogenic and more radiogenic values than primitive upper mantle, can be attributed to ancient melting and subsequent sequestering of isotopic signatures, melt-rock reaction, or secondary alteration, or a combination of any of the three. Petrographic, electron microprobe, and hand-sample inspection show the peridotites to have experienced varying amounts of serpentinization and weathering, accompanied by heterogeneity in 187Os/188Os values. In addition, a majority of the peridotites studied here are spatially associated with gabbros having equally, if not more highly, heterogeneous Os isotopic compositions and N-MORBs with homogeneous Sr, Nd, and Pb isotopic signatures. This suggests secondary seawater alteration as the dominant influence on present-day Os signatures. We show that careful separation, leaching, and analysis of Cr-spinel from abyssal peridotite largely removes the radiogenic seawater Os isotopic signature, allowing a more accurate assessment of the Os signature of depleted mid-ocean ridge basalt mantle (DMM). Cr-spinel is a highly refractory mantle mineral and commonly host to tiny sulfide inclusions, which are carrier phases for Os. These sulfides are well protected by the Cr-spinel from high-temperature serpentinization and low-temperature seafloor weathering, thereby preserving nonradiogenic DMM-like Os isotopic signatures. Our analyses of treated Cr-spinel fractions finds 187Os/188Os ratios that are dominantly less radiogenic (0.1238–0.1482) than corresponding whole rock compositions. Nonmagnetic Cr-spinels show more highly variable Os compositions and concentrations than magnetic Cr-spinels, overlapping with the more radiogenic whole rock values. Magnetic Cr-spinels are dominantly less radiogenic than PUM, with high but variable Os concentrations, possibly resulting from a sulfide “nugget” effect. The systematically lower 187Os/188Os in Cr-spinels compared to whole rocks shows that whole rock abyssal peridotites are largely compromised by radiogenic seawater interaction, often obscuring the mantle signature. Measurements of treated abyssal peridotite Cr-spinels in some cases can circumvent this seawater alteration problem allowing more straightforward interpretation of isotopic values and thus further constraining the Os isotopic signature for the depleted upper mantle. Our results suggest that the 187Os/188Os for the upper mantle lies in the range 0.120–0.125.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 A cornerstone assumption of chemical geodynamic theory is that mid-ocean ridge basalt is derived from, and is isotopically representative of, the depleted upper MORB mantle (DMM). This presumption has been difficult to test, owing, on the one hand, to the low concentrations of Sr, Nd, Hf, and Pb in abyssal peridotites [Johnson et al., 1990; Snow et al., 1994], and, on the other hand, to the low Os concentrations in MORB [Roy-Barman and Allègre, 1994; Schiano et al., 1997; Roy-Barman et al., 1998]. The most successful test to date showed that Nd isotopes in carefully leached Cpx from abyssal peridotites are similar to those in nearby MORB [Snow et al., 1994].
 When comparing the 187Os/188Os values for abyssal peridotites and zero-age MORB, the test for a cogenetic relationship is obscured by the extreme isotopic heterogeneity seen in abyssal peridotites [Roy-Barman and Allègre, 1994; Snow and Reisberg, 1995] and in MORBs [Schiano et al., 1997; Roy-Barman et al., 1998]. This presents the possibility that abyssal peridotites are not direct petrogenetically related residues of recent mid-ocean ridge basalt (MORB) melting [Esperanca et al., 1999; Brandon et al., 2000], thereby adding uncertainty to the use of MORB for isotopic fingerprinting of the upper mantle. Inherent mantle heterogeneity has been suggested to explain the wide range of osmium compositions on large scales but may be difficult to justify on small scales, especially considering the recent work by Brenan et al.  on closure temperatures for Os diffusion in sulfide. Alternatively, we argue that most abyssal peridotite Os compositions have been compromised by exposure to radiogenic Os during interaction with seawater. This is not surprising, since abyssal peridotites are arguably the most altered mantle rocks routinely studied and are nearly completely serpentinized (50–90%) and frequently weathered [Snow and Dick, 1995]. This study provides evidence that abyssal peridotite whole rock Os compositions are elevated owing to a radiogenic seawater component but that this alteration problem can be circumvented by analyzing Os from carefully treated Cr-spinels separated from abyssal peridotites. Cr-spinels have been utilized successfully in several previous Os isotope studies of peridotitic massifs, layered intrusions, and ophiolites Marcantonio et al., 1993; Meisel et al., 1997; Walker et al., 1996].
 Cr-spinel is a ubiquitous, accessory phase in mantle lithologies [Dick and Bullen, 1984] and is, furthermore, the most resistant mantle phase during low-temperature metamorphism and alteration. Recent work has shown Cr-spinel to be a common host of sulfide inclusions, which have been shown to be one of the principal Os carrier-phases in the mantle [Morgan, 1986; Hart and Ravizza, 1996; Burton et al., 1999]. Utilizing the Os isotopic composition of spinel, we address two questions: (1) what is the mineralogic heterogeneity of Os isotopes in the upper mantle and (2) what is the Os isotope signature of the depleted MORB mantle. The Re/Os system is useful for addressing these questions concerning mantle evolution as, unlike the other long-lived radiogenic isotope systems, Os is compatible in mantle lithologies during melting and will tend to be enriched in residues. Because the parent, Re, is mildly incompatible in mantle lithologies, the residues of melting will have lowered Re/Os ratios, which over time will lead to unradiogenic Os isotope signatures in the residual peridotites. The Re/Os system provides a unique window into, and added constraints on, the mid-ocean ridge melting process; yet a great deal of uncertainty still exists.
 Our peridotite suite includes samples from the Atlantis II Fracture Zone (31°S, 57°E) and the Du Toit Fracture Zone (FZ) (53°S, 25.5°E), both part of the Southwest Indian Ridge, and samples from the 15°20' FZ (15.3°S, 46°W), on the Mid-Atlantic Ridge (Figure 1). Detailed studies of these abyssal peridotites and their respective locations are presented elsewhere [Dick et al., 1991; Fisher et al., 1986; Roest and Collette, 1986; Snow et al., 1987; le Roex et al., 1989; Grindlay et al., 1998]. Samples from these fracture zones were chosen for study because they are situated between “normal” ridge segments far from any mantle plumes and thus escape the chemical influence of hot spot material. Although these three locations are all non-plume-influenced fracture zones, the spreading ridges upon which they are situated are variable (i.e., spreading rate). For example, the SWIR geochemistry is highly variable along its length from the Bouvet Triple Junction to the Indian Ocean Triple Junction. This may be a function of the changing spreading rates or reflective of heterogeneous mantle source regions. In this study the samples are grouped together as normal MORB peridotites, but it should be remembered that the isotopic and geochemical differences between fracture zone samples could be highly dependant on the spreading ridge characteristics and the underlying mantle lithologies.
 Our sample suite consists of harzburgites, lherzolites, and dunites, all of which have experienced varying degrees of serpentinization and weathering. The peridotites show a variety of textures and alteration stages. Rock color often indicates the degree of serpentinization and/or weathering: orangish-brown to tan coloring indicating moderate to high weathering of olivine to clay, while green and black colors indicate fresher, less weathered but usually highly serpentinized rocks. Petrographically, the peridotites show typical alteration features. Replacement of olivine, opx, and cpx is fairly widespread; a point count of sample RC 27-9-6-2 (lherzolite) reported by Snow and Dick  puts the pre-alteration modal proportions at 48.2% olivine, 45.1% orthopyroxene, 6.1% clinopyroxene, 1.2% spinel, and 0.41% plagioclase. Primary olivine modal proportions in the rest of the harzburgites and lherzolites are generally >70% with orthopyroxene between 15 and 30% and clinopyroxene ranging from 0 to 11%. Protogranular textures dominate the less altered peridotites, but little can be said concerning the primary textural relationships in the peridotites with high proportions of secondary minerals. Replacement of primary silicates is often accompanied by reddish-brown, Fe-oxidation minerals, and indicated by high proportions of serpentine (lizardite) and magnetite.
 Spinel is a common accessory mineral in abyssal peridotites, generally amounting to 0.25–1.0% in harzburgites and lherzolites. In dunites, spinel may be absent or, in rare cases, as much as 16% of the modal rock composition. The resistant nature of spinel results in the presence of fresh to moderately altered grains in peridotites that have otherwise undergone significant serpentinization. Sample RC27-9 6-2 has been shown to contain a significant amount of sulfide, present as interstitial grains, inclusions in pyroxene, and separate grains within alteration veins [Lee, 1997]. Petrographic investigation of the other peridotites found scattered sulfides (>1 μm) occasionally occurring within silicate hosts and along pyroxene exsolution lamellae. In a few thin sections there exist sulfide grains up to ∼60 μm, often times associated with multiple smaller sulfide globules. The lack of visible sulfides in some samples containing high concentrations of osmium indicates that the bulk of the osmium must be contained in “ultra-trace phase” sulfide micro-inclusions [Shirey and Walker, 1998] within spinel.
3. Sample Preparation and Leaching
3.1. Whole Rock
 From each peridotite, we crushed 150–200 g of rock to centimeter-sized chips using a steel jaw crusher. Careful inspection of the sample for shards of steel was conducted prior to further crushing. In a later group of whole rock samples, we tested a new method of rock pulverization called “electric pulse disintegration.” In this technique the sample chips sit in a water bath, which is charged with rapid electric pulses from a high-voltage power source. The sample is virtually exploded, which occurs preferentially along grain boundaries. As a result, individual undamaged mineral grains can be recovered. This method proved satisfactory as a prepowdering step and will be used in future sample preparation, thereby eliminating the threat of metal contamination. (Note that the whole rock sample ABP 16-71-22 was measured using both crushing techniques, and the Os compositions are nearly identical, indicating no detectable metal contamination from the jaw crusher.) After thorough inspection, the sample was powdered using an agate puck grinder. The powdered whole rock sample (1.5 g) was spiked with an in-house mixed spike enriched in 99Ru, 105Pd, 190Os, 191Ir, and 198Pt. The sample and spike mixture was mechanically mixed and left overnight in a covered ceramic crucible at 40°C. The mixture was then combined with 3.6 g of flux, consisting of 0.20 grams of sulfur, 0.40 grams of nickel, and 3.0 grams of Na-borate; 1:2:15 flux proportions. Whole rock crucibles were fused at 1020°C for 1.5 hours. Complete tracer/sample equilibration is obtained during fusion and can be monitored during sparging. A detailed description of NiS fire assay is given in Ravizza and Pyle . Sample chemistry for ICP-MS sparging is briefly described in sections 3.3 and 4.2 and in detail by Hassler et al. .
 We selected whole rock sample RC27-9-34-84 to test for potential low-temperature Os degassing prior to fusion. We placed 1.5 g of sample powder in a ceramic crucible in the furnace at 400°C for 5 hours. The sample was cooled and then spiked before proceeding with the presparging chemistry as described in section 4.2.
 The whole rock peridotite (150–200 g) was crushed to chip size and then placed in a disc mill and ground to <1 mm. From microprobe analysis we know that alteration rinds of “ferrit-chromite” and/or magnetite exist on and within the spinel and are due to serpentinization and low-temperature weathering. This secondary component is formed in the presence of radiogenic seawater Os, thus we used a two-step technique to ensure removal of as much of this seawater-compromised component as possible. Through continued crushing, using a steel diamond mortar, to <500 μm, we were able to expose most of the altered spinel surfaces. Samples were inspected under the microscope for steel shards, which were subsequently removed by hand. Thorough decanting removed the finest size fraction in preparation for heavy liquid separation. The sample was poured into a glass separation column with methylene iodide (specific gravity = 3.32). The heavy mineral separate was dominated by spinel (∼70–80% on average), with some olivine and clinopyroxene. Both mineral fractions were thoroughly rinsed with acetone and dried. The spinel-rich fraction was then separated further into a magnetic and nonmagnetic fraction using a hand magnet. Both magnetic and nonmagnetic fractions were purified by extensive handpicking of silicate phases under a microscope. Prior to leaching, the samples were 85–99% pure spinel. Much of the remaining silicate was not easily separated mechanically because it existed interstitially with spinel.
 Mineral separates were then leached (section 3.3) and powdered under acetone, using an agate mortar. The sample powders were weighed in a ceramic crucible and then spiked with one of two tracers, both enriched in varying proportions of 99Ru, 105Pd, 190Os, 191Ir, and 198Pt. Sample/spike mixtures were left on the hotplate overnight at 40°C to dry. Added to each sample in the following proportions 8:2:1 were 2.0 grams Na-borate, 0.5 grams nickel, and 0.25 grams sulfur. Each crucible was then placed in the oven for 1.5 hours at 1020°C. The nickel sulfide beads were weighed to evaluate fusion sample recovery (60–70% on average), followed by sparging chemistry as explained in section 4.2.
 Each Cr-spinel fraction (magnetic and nonmagnetic) was leached prior to powdering under acetone. The following three solutions comprised the leaching sequence: (1) 6.2 N HCl, (2) 70% concentrated HF, 15% 6.2 N HCl, 15% concentrated HNO3, and (3) concentrated HCl. Up to 250 mg of sample was placed in 1–2 mL of 6.2 N HCl at room temperature in a sealed, 4 mL Teflon vial. After 1 hour, the HCl was pipetted away, and the sample was rinsed with ultrapure water and dried. To the beaker 1–2 mL of HF/HCl/HNO3 solution was added and sealed, then placed on a hotplate at 100°C for 5 hours. (Note that the samples with high Ca content, likely present as interstitial clinopyroxene, formed CaF during this step. This precipitate appeared as a thin white layer on the spinel. To deal with CaF precipitation we completely dried down the HF/HCl/HNO3 solution, allowing fluorine to fume off, thereby breaking down the CaF). Samples that did not precipitate CaF had the HF/HCl/HNO3 solution pipetted away and were rinsed with ultrapure water and dried. The final step used 1–2 mL of concentrated HCl. The beaker was sealed and sonicated for 45 min at room temperature. The HCl was pipetted away, and the sample was thoroughly rinsed with ultrapure water and dried.
 Sample RC 27-9-34-84m (previously measured 187Os/188Os = 0.1289) was subjected to a particularly rigorous leaching sequence to evaluate the affects of extensive leaching. The sample was leached with the same solutions, but the length of each leaching step was increased. Step 1, which involved the 6.2 N solution, was stretched from 1 hour to 3 days. Step 2, with the HF/HCl/HNO3 solution, was lengthened from 5 hours to 20 hours at 100°C. Step 3 was increased from 45 min to 1 hour of sonication, and an additional hour in the beaker was added as well.
 We carefully separated spinel leachate solutions for direct ICP-MS sparging. From each spinel leach we pipetted/decanted the leachate into a 4 mL Teflon beaker. To dilute the platinum group element (PGE) concentrations of the solution, 1 mL of ultra-pure water was added. The leachates were split in half, with the second split set aside for future PGE spiking and ICP-MS sparging to determine Os concentration. To each of the HCl leachates (both 6.2 N and concentrate) a drop of ethanol was added to maintain a reducing environment. The HCl leachates were dried down and then put back into solution with concentrate HNO3 and transferred into a 24 mL screw cap beaker. The diluted (1 mL ultrapure water) HF/HCl/HNO3 leachates were transferred to a 24 mL Teflon beaker and sealed tightly with Teflon tape to limit loss of oxidized Os. Just prior to sparging, the leachates were chilled on ice, further diluted with ultrapure water, and sparged directly into the ICP-MS.
4. Analytical Methods
4.1. Mineral Data
 Spinel and sulfide chemistry was determined using the JOEL JXA-733 Superprobe at the MIT Electron Microprobe Facility. Multiple spinel and sulfide grains from selected samples were analyzed at 15 kV accelerating voltage and a probe current of 10 nA. The spot size was held constant at 10 microns, and the counting time per element was 10–40 s. Count times during sulfide analysis were generally 40 s, except for S at 120 s. Analyses were calibrated using synthetic standards. The data were reduced with the CITZAF program [Armstrong, 1995] using the atomic number correction of Duncumb and Reed, Heinrich's tabulation of mass absorption coefficients, and the fluorescence correction of Reed.
 The NiS beads from the whole rock and Cr-spinel fusions were dissolved in 6.2 N HCl at ∼150°C and filtered through a 0.45 μm cellulose filter. The filter paper was then dissolved in 1–2 mL concentrated HNO3, in a 24 mL Teflon screw-cap beaker at room temperature. Oxidation of Os to OsO4 was achieved by placing the tightly sealed beaker on a hotplate at 100°C the night prior to sparging. Before preparing the solution for sparging, it was chilled on ice for at least 1 hour. The acid solution was then diluted ∼10-fold with ultrapure water and a screw cap with inflow and outflow tubes was placed on the beaker. This cap allowed argon to bubble through the solution, via perforated tubing, thereby carrying the volatile OsO4 into the torch for analysis. See Hassler et al.  for a detailed description of ICP-MS sparging method.
 Osmium composition and concentration was obtained by sparging [Hassler et al., 2000] of volatile OsO4 into our Finnigan Element Magnetic Sector ICP-MS. Argon sample gas flow rates were variable to optimize ion beam intensity but averaged around 1.35 L/min. Each analysis consisted of 30 runs with 100 passes (analysis time is 7 min). We report the following Os sparging standards on ICP-MS: 80 pg standard 187Os/188Os = 0.17439 ± 0.56% (1 SD, n = 4), 400 pg standard 187Os/188Os = 0.17432 ± 0.55% (1 SD, n = 24), 1.25 ng standard 187Os/188Os = 0.17399 ± 0.29% (1 SD, n = 42). The whole rock fusion blank measured during sparging session had 187Os/188Os ratio of 0.2364 ± 0.0104 and a concentration of 0.93 ± 0.01 pg/g and the Cr-spinel blank had 187Os/188Os = 0.1392 ± 0.0061 with a concentration of 7.36 ± 0.22 pg/g. These blanks were used to calculate the “% blank correction” listed in Table 3.
5.1. Mineralogical and Petrographic Analysis
 Electron microprobe analysis of selected spinel grains showed a large compositional range in mineral chemistry, covering nearly the entire solid-solution face of the spinel prism between Cr-spinel-hercynite-magnetite. Specific phases included Al-Cr-spinel and Mg-Al-Cr-spinel to more Fe-rich picotite and pleonaste, and even altered “Ferrit-Chromite,” originally recognized by Spangenberg  as an alteration product of Cr-spinel. Because of the variability in spinel composition, we will henceforth use the term Cr-spinel to represent the range in spinel solid-solution compositions shown in Table 1, unless specifically stated. Bulk crystal average Mg # and Cr # (Mg # = molecular Mg × 100/(Mg + Fe2+); Cr # = molecular Cr × 100/(Cr + Al); Fe203 was obtained by stoichiometric recalculation from FeO) ranged from 47.6 to 74.7 and 13.8 to 84.3, respectively (Figure 2). The extremely high Cr # range is a result of ferrit-chromite dominated grains and/or magnetite rims in one or two samples.
Table 1. Representative Chromium Spinel Analyses
Notes: Representative microprobe analyses have been recalculated by stoichiometry based on the method of Bence and Albee . RFeO = recalculated FeO. Mg# = Mg*100/(Mg + Fe2+), Cr# = Cr*100/(Cr + Al), Fe3+# = Fe3+*100/(Cr + Al + Fe3+).
Lower case letter indicates specific spinel crystal. Upper case letters indicate type of analysis: C = averaged core analyses (usually 5 points), R = rim analyses (10–30 um from crystal edge), IP = inclusion within plagioclase.
 Petrographic analysis under plain polarized light (Figure 3a) depicts a typical, nearly opaque to translucent brown Cr-spinel “RC27-9 19-1-b-C,” within various secondary silicate phases. A variety of crystal habits exist in the peridotites, but generally the Cr-spinels are anhedral to subhedral with scalloped edges, often accompanied by holly-leaf and/or symplectite texture. Figure 3b is a backscattered electron image of Figure 3a (RC27-9 19-1-b-C from Table 1), a subhedral to angular, interstitial Al-Fe-Cr-spinel. Replacement of primary silicate phases was pervasive, yet because of the resistant nature of Cr-spinel, its primary interstitial igneous shape was well preserved. The interior of the grain appears to be unaltered, with only minor fractures and little magnetite. Microprobe analysis of the rim (RC27-9 19-1-b-R), the highly reflective portion of the grain, found the composition to be magnetite. Anomalous fine-grained zones (top and bottom of grain) may represent relict pyroxene, which has been permeated by highly reflective magnetite, or possibly exsolution of a Cr-Fe3+-rich composition from the Al-rich core. In the same area, a rather interesting transition zone or alteration front exists between the fine-grained silicate-magnetite mixture and the massive Cr-spinel. Magnetite is fairly thin on most edges, although a few sections seem to have well-developed rinds resulting from metamorphic overprinting of Cr-spinel. Other areas (right and left edges) have a distinctive needle-like growth texture, signifying new magnetite crystal-growth, rather than replacement. Overall, this Cr-spinel has very little ferrit-chromite or magnetite and thus would appear to have experienced limited serpentinization or alteration.
 The second backscattered electron photomicrograph (Figure 3c) shows a slightly more iron-rich Al-Fe-Cr-spinel “RC27-9 34-88-a-C”. This Cr-spinel is angular and interstitial in nature and has much more veining. Analysis “RC27-9 34-88-a-R” was taken along the grain's highly reflective rim and is compositionally different than the rim analysis in Figure 3b. The high amount of Cr2O3 even in the highly reflective rim material indicates a ferrit-chromite composition rather than magnetite, which formed as Cr-spinel underwent moderate to low-temperature serpentinization (<500°C). Reflective ferrit-chromite laths delineate multiple extensional veins, which also contain secondary silicate minerals. The ferrit-chromite growth pattern within the veins and also in the upper right corner of the grain is illustrative of symplectite texture. Despite the presence of grain boundaries with little to no ferrit-chromite, this Cr-spinel, based on the proportion of metamorphic overprinting and hence the proportion of magnetite plus ferrit-chromite, is likely to have a more radiogenic Os signature.
 Preliminary petrographic and electron microprobe analysis (Table 2) of sulfide within these peridotites gave compositions similar to those measured by Lee  in sample RC27-9 6-2. The petrographic occurrence of sulfide was three-fold: (1) sulfide hosted by silicates, either in the interior of the crystal or along exsolution lamellae, (2) sulfide hosted by spinel, dominantly as micro-sulfides (< 1 μm), and 3) sulfide interstitially between silicates. Microsulfides within spinel are not represented in Table 2 owing to the difficulty associated with locating and analyzing a grain at the micron scale. The dominant sulfide phase is pentlandite, from the Fe-Ni-S system. Interestingly, the amount of Cu in the sulfides is highly variable, indicating the possibility of exsolution of chalcopyrite. There is also petrographic evidence of pyrrhotite exsolution. Two of the analyses in Table 2 are average compositions from multiple spot measurements within a grain. Unfortunately, no sulfide grains were located and successfully analyzed in samples also measured for Os.
Table 2. Sulfide Microprobe Data
RC 27-9 34-40a
RC 27-9 34-40b
RC 27-9 34-40c
RC 27-9 34-58a
Samples with n>1 (where n is number of spot analyses per sulfide grain) are averaged compositions.
 ICP-MS analysis of 10 abyssal peridotites gave whole rock 187Os/188Os between 0.1268 and 0.1582. Whole rock Os concentrations vary from 0.25–3.0 ppb. Osmium isotopic compositions, concentrations, and corresponding precisions for each sample are listed in Table 3; 187Os/188Os is plotted versus Os concentration in Figure 4. Osmium isotopic compositions for magnetic Cr-spinel fractions ranged between 0.1238 and 0.1371, with concentrations ranging from 1.50 ppb up to 166 ppb. Nonmagnetic Cr-spinel fractions showed a similar compositional range but extended to more radiogenic values (0.1539). Concentrations varied from 0.57 to 8.50 ppb. Some early measurements of osmium in Cr-spinel were done by negative thermal ionization mass-spectrometry (NTIMS), as indicated in Table 3, and used different sample preparation and sample chemistry [Hart and Ravizza, 1996]. Most of the samples analyzed using both techniques show similar Os compositions and concentrations, with no apparent systematic differences between NTIMS and ICP-MS techniques.
 Abundant sulfide from sample RC 27-9-6-2 allowed direct osmium measurement on a sulfide separate. Owing to the high variation of Os content in sulfides, the sulfide powder was not spiked because we did not want to compromise the Os isotopic measurement by under or overspiking. The 187Os/188Os for the sulfide was 0.1278 (the corresponding whole rock from this study gave 187Os/188Os = 0.1275).
 A duplicate whole rock powder from “RC 27-9-34-84” was heated to 400°C prior to spiking to assess the potential for low-temperature Os degassing. The subsample is reported in Table 3 as “Whole Rock 2” indicating that the concentration was significantly lowered by this preheating (from 0.45 to 0.12 ppb). The 187Os/188Os ratio is much more radiogenic (0.1725) compared to the initial whole rock measurement (0.1458), suggesting loss of Os from dominantly unradiogenic sites.
 We also separated and measured leachates from selected samples. Osmium isotopic compositions from the sequential leachates were highly heterogeneous (Table 3). For example, Cr-spinel separate RC 27-9-34-84m has an 187Os/188Os = 0.1289, a 6.2 N leachate 187Os/188Os = 0.3816, a HF/HCl/HNO3 leachate 187Os/188Os = 0.1296, and a Conc. leachate 187Os/188Os = 0.1840.
6.1. Whole Rock Osmium
 The whole rock osmium isotopic data (Table 3) overlaps the published whole rock abyssal peridotite database and enlarges it by nearly 40% (Figure 4). About half of the data points plot at values more radiogenic than the estimated primitive upper mantle (PUM) composition of 0.129 ± 0.0009 [Meisel et al., 1996]; about half plot below PUM, within the proposed range of the present-day depleted MORB mantle (DMM) composition (0.125 ± 0.0014) [Snow and Reisberg, 1995]. Hattori and Hart  derived a similar value (0.1248) from analysis of Os-rich minerals in ophiolites and peridotite massifs of various ages. A single anomalously low whole rock value (ABP 16-71-22) of 0.1183 is well below primitive upper mantle and is thought to be the least radiogenic abyssal peridotite whole rock measurement published.
 It should be noted that this estimated value for primitive upper mantle differs slightly from the chondritic reference composition. Meisel et al.  stated that PUM represents the modern Os composition of the bulk upper mantle that would have evolved from the late accretionary period to present assuming no modification of Re/Os. The modern “chondritic” Os composition (0.127) is derived from ordinary chondrites (187Os/188Os = 0.1286 ± 0.0010 and carbonaceous chondrites (187Os/188Os = 0.1258 ± 0.0005) [Meisel et al., 1996]. The incompatible behavior of Re versus the compatible behavior of Os during silicate mantle melting should produce residual peridotites that are less radiogenic and have higher Os concentrations than bulk earth. This is not the case for a large portion of the global whole rock data set. The scatter seen in Figure 4 includes our whole rock measurements and extends to very radiogenic compositions with decreasing concentration. Many of the abyssal peridotites have more radiogenic compositions and lower concentrations than primitive upper mantle.
Snow and Reisberg  proposed that low-temperature seafloor alteration of abyssal peridotites produces elevated 187Os/188Os values in some whole rocks. As indicated in Figure 4, the present-day seawater Os composition (187Os/188Os ∼ 1.06) is significantly more radiogenic than PUM, suggesting that, even at water/rock ratios of 104, alteration of peridotite involving seawater could compromise the existing mantle Os signature [Martin, 1991; Roy-Barman and Allègre, 1994; Snow, 1993]. This does not mean that measurement of whole rock abyssal peridotites may not occasionally yield 187Os/188Os similar to or less radiogenic than PUM. However, knowing that abyssal peridotites are pervasively altered over a range of temperatures by serpentinization and/or weathering and considering the compositional heterogeneity displayed in Figure 4, we interpret the scatter of data to indicate that 187Os/188Os whole rock values are largely compromised by seawater interaction. The tendency for Os concentrations to be lower than PUM may suggest oxidation and removal of Os-rich sulfide during serpentinization and weathering.
6.2. Small-Scale Os Isotope Heterogeneity
 A longstanding issue in mantle geochemistry is the determination of the scale of chemical heterogeneity within the mantle. This is not easily addressed here because of the lack of understanding concerning the behavior of the Re-Os system during mantle melting. Hart and Ravizza  addressed the issue of the distribution of Os in mantle phases by measuring Os compositions and concentrations in carefully separated phases from a Kilbourne Hole lherzolite xenolith. They found the Os content in sulfide to be 4.11 ppm. Their results were similar to those of Morgan and Baedecker , who reported Os contents in Kilbourne Hole lherzolite sulfide particles ranging from 3.5 to 11 ppm Os. The other phases measured by Hart and Ravizza  (olivine, clinopyroxene, orthopyroxene, and spinel) were found to have Os contents of 36, 200, 60, 36 ppt, respectively. This confirms the notion that the vast majority of the Os budget lies within the various sulfide phases present in mantle lithologies. It is important to note that the low Os concentration reported by Hart and Ravizza  for spinel relative to silicates is surely due to a greater proportion of sulfide inclusions in the pyroxenes. In the present study, however, the Os budget is dominated by spinel, which is similar to other work done on Kilbourne Hole xenoliths [Burton et al., 1999] and other mantle lithologies [Walker et al., 1996].
 In an effort to understand the nature of the phases hosting Os in abyssal peridotites and evaluate the potential for seawater contamination, we have incorporated data [Lee, 1997] from a detailed study of the least serpentinized sample (RC 27-9-6-2) from the study of Snow and Reisberg  (this sample was dredged from the east wall of the Atlantis II fracture zone). The boulder was divided into a number of sub-sections, as illustrated in Figure 5 [Lee, 1997], and each section was analyzed for osmium composition and concentration. In addition, a magnetic size fraction was analyzed from sub-sample E1. Also, subsample powder H1 was leached by a variety of reagents, which were analyzed to assess the impact of seawater contamination.
 Sample RC 27-9-6-2 is somewhat unique in that it contains a 1 cm vein of diopside with a thin zone of dunite around it. In the matrix, there is both an abundance of sulfide and accessory biotite, andesine feldspar, apatite, and relict Fe-Ti oxides. These later minerals are all interpreted as owing to impregnation of the host peridotite by late melts associated with emplacement and crystallization of the vein. On the basis of the Nd isotope data, both the whole rock and the vein are in the range of typical Indian Ocean N-MORB mantle. The bulk sample 143Nd/144Nd is 0.512965, Cpx separated from the bulk sample is 0.512994 [Snow et al., 1994], and Cpx separated from the vein is 0.513091 [Lee, 1997].
 The Os isotopic heterogeneity of this single sample is enormous and spans almost the total range of the global abyssal peridotite data set shown in Figure 4. Interestingly, we measured a separate whole rock sample for Os isotopic composition and our value of 0.1275 is almost exactly the mean of the two previous whole rock measurements; Snow and Reisberg's  at 0.1255 and Lee's  average at 0.1310. The leached residues increase in Os concentration when silicates are dissolved (HF and HCl) and decrease when sulfides are dissolved (HNO3 and CrO3), suggesting that Os is dominantly hosted by the sulfide. Some of the leachates had Os ratios up to 0.1573, strongly implicating seawater as the source of the observed small-scale heterogeneity in this dredge boulder. However, the Os isotopic ratios do not bear any simple relationship to the exposed surfaces of the dredge boulder (Figure 5); the highest (E) and lowest (D) segments are both interior samples, suggesting the seawater introduction of Os took place under moderate to high-T conditions at depth, before the peridotite was exposed to weathering.
Lee  also studied and electron-probed virtually every sulfide occurrence in two selected areas of this rock, covering an area of almost 60 cm2. Sulfides occurred in three types: (1) tiny individual inclusions and curvi-planar arrays of inclusions, dominantly hosted in Cpx; (2) larger interstitial sulfide grains on grain boundaries between silicates, and (3) larger irregular grains in mesh-textured serpentine surrounding relict olivine “islands.” The sulfides are predominantly pyrrhotite, pentlandite, and chalcopyrite, with minor bornite, marcasite, heazlewoodite, millerite, polydymite, and rare pyrite. Most represent exsolution from the sulfur-poor end of the monosulfide solid solution. There is no correlation between sulfide minerals and textural type (i.e., pyrrhotite-pentlandite are common in all three textures). It is likely that all textural types were recrystallized and altered at moderately high temperature during serpentinization, with only the bornite and millerite bearing witness to lower temperature alteration (<200°C). There is no obvious correlation between sulfide texture or abundance and the Os isotopic composition (though the sulfide survey did not cover the entire slab surface). We suggest that the Ni-Fe-rich sulfides “scavenge” Os from circulating seawater, during serpentinization. This would be consistent with the low closure temperature recently reported for Os diffusion in pyrrhotite [Brenan et al., 2000]. On the basis of their data, pyrrhotite grains of 100-micron diameter will equilibrate in 10,000 years at 380°C or in only 100 years at 470°C. Interstitial sulfides will thus readily equilibrate with circulating seawater at modest levels in the oceanic crust. Only sulfides encapsulated in silicates or chromites can be expected to retain primary Os signatures. The distribution of sulfide, the local water/rock ratio, and the effective “scavenging” of Os would almost certainly lead to isotopic “patchiness” and small-scale variability.
 It may also be possible that some of the osmium isotopic heterogeneity is primary, due in part to magmatic impregnation during formation of the diopside vein, as minor isotopic heterogeneity is evident in the Nd isotopes. Brandon et al.  reported significant Os isotopic heterogeneity in drilled samples from a single section of abyssal peridotite from the Kane Fracture Zone. Samples within 80 cm of each other were as much as 8% different in 187Os/188Os; Brandon et al.  argued that this heterogeneity was primary and mantle-derived. It is difficult to rationalize this claim with the low closure temperatures reported by Brenan et al. . For a mantle upwelling rate under a ridge of 30 km/million years, the adiabatic cooling rate is ∼10°C/million years. Even large 1 mm sulfide grains in such a mantle domain will have closure temperatures for Os of only 400°C. Long-term preservation of isotopic heterogeneity in the convecting upper mantle seems highly unlikely on the length scales involved in the Brandon et al.  sample suite. Furthermore, the correlation of 187Os/188Os with Pt/Os ratio, which they put forward as prime evidence for the primary nature of the heterogeneity, could easily be produced by interaction of sulfides with seawater, as seawater has both high 187Os/188Os and a high Pt/Os ratio. Open system behavior of sulfides with respect to Os has also been documented in a suite of peridotitic xenoliths [Handler et al., 1999]. Further understanding of these complex processes will require microbeam Os isotope analysis of sulfides [Alard et al., 2000].
6.3. Regional Os Heterogeneity
 One of the main assumptions made here is that serpentinization/alteration of abyssal peridotites results in elevated osmium isotopic compositions due to interaction with seawater. It is evident from Figure 4 that less-altered abyssal peridotites can be carefully selected and analyzed which yield whole rock 187Os/188Os values less than or equal to primitive upper mantle. The explanation for the heterogeneity in 187Os/188Os is then attributed to a variably enriched and depleted mantle source. However, the question then becomes, is it possible to recover the mantle osmium signature from a common serpentinized abyssal peridotite. We present rather compelling evidence illustrating that Os isotopic heterogeneity in abyssal peridotite whole rock measurements is the result of a radiogenic seawater Os component and not solely due to inherent mantle heterogeneity.
 This is especially true at the Atlantis II Fracture Zone, where is has been shown by Snow  that the MORB mantle is typical N-MORB. Sr, Nd, and Pb isotopic compositions from multiple abyssal basalts, taken along the Atlantis II FZ, show very homogeneous N-MORB-like signatures. This is quite important considering that a large portion of the SWIR, dominantly between Bouvet and Marion, displays significant variation in basalt Sr, Nd, and Pb isotope systematics. The Du Toit Fracture Zone, to the west of Marion, has been characterized by both N-MORB and E-MORB type basalts [Le Roex, 1989], and thus our case for alteration-induced Os isotopic heterogeneity for samples from this fracture zone is not as strong. Similarly, the isotopic systematics at 15′20° FZ are complex and suggest that some primary mantle heterogeneity may underlie the dominant seawater alteration signal. On the basis of the anomalously low 187Os/188Os values we report for the two 15′20° FZ samples, it is apparent that there is significant complexity associated with the melting dynamics, and it has previously been suggested that sub-continental lithospheric lithologies within the upper mantle may play a role in the anomalous isotopic signatures [Esperanca, 1999].
 Recent PGE and Os isotopic work by Blusztajn et al.  on gabbros drilled at the Atlantis II Fracture Zone lend further support for the presence of an alteration— induced radiogenic seawater Os component. Os isotopic compositions in the gabbros range from 0.140 to 0.467 and correlate with Rb/Cs, suggesting that seawater alteration affected the Os isotopic system. The effect of serpentinization on abyssal peridotite Os compositions can also be seen when comparing harzburgites to dunites from the Atlantis II FZ. One potential explanation for the rough correlation seen in Figure 2b could be that dunite is serpentinized and weathered more readily than harzburgite and thus interacts with a larger volume of radiogenic seawater Os. Because of the lesser proportion of olivine in harzburgite and hence greater proportion of more resistant pyroxene, the Os-hosting sulfides contained within Cr-spinel and pyroxene may be less affected by seawater interaction than sulfide found in dunite. Whether this difference in peridotite lithology would lead to a difference in Os isotopic composition is clearly speculative and needs further attention. Another potential explanation is related to the variable melting dynamics associated with harzburgites versus dunites and the differing spinel mineralogy between these two lithologies.
6.4. Cr-Spinel and Serpentinization
 As mantle melting progresses, the chrome content in residual spinel increases, while the magnesium content decreases, owing to the effect of Cr on the partitioning of Fe-Mg between olivine and spinel [Irvine, 1965]. This produces the apparently contradictory result of an inverse correlation between abyssal peridotite spinel Mg # and Cr # [Dick et al., 1984]. It follows that dunites, products of melt focusing [Kelemen et al., 2000], should generally have higher Cr # than the enclosing peridotitic residues of melting (Figure 2). Hybrid peridotites, melt impregnated and late melt-reacted rocks, are frequently characterized by the presence of plagioclase and spinel compositions with TiO2 > 0.15% [Allan and Dick, 1996] (Figure 2b). Spinel from harzburgite sample RC27-9-30-33 has low Cr # and high Mg #, suggesting a lower degree of melting and thus a less depleted residual peridotite. As long as the dunites are products of harzburgite melting, the two lithologies should have similar Cr-spinel 187Os/188Os ratios. Although we have limited data to base a comparison on, we do see a difference in 187Os/188Os values of harzburgite versus dunite.
 Serpentinization of abyssal peridotite commonly occurs at crustal temperatures as high as 500°C, continuing until near-extrusion on the seafloor. As the most resistant mantle phase, Cr-spinel preserves original isotopic and trace element signatures as the abyssal peridotite undergoes serpentinization and low-temperature seafloor weathering. In an extensive study, which looked at alteration of Cr-spinel in serpentinites, Burkhard  found the composition of the spinel to control the degree of spinel alteration. Cr-spinel which is rich in Cr and Fe2+ is much more susceptible to alteration than Cr-spinel rich in Al and Mg. This suggests that dunitic spinel with high Cr #'s may tend have more radiogenic Os compositions than harzburgitic spinel. On the basis of this alteration argument, the spinel compositions from the dunites, which have higher Cr #, Fe3+ #, and TiO2, typical of rocks that have reacted with MORB-like melts, should have more radiogenic 187Os/188Os compositions than the harzburgitic spinels. This is evident in Figure 2b, as Cr-spinel samples RC 27-9-30-33 and RC 27-9-34-63 have Cr #'s comparable to moderately depleted harzburgite and are two of the least radiogenic spinels. This may also be explained by melt transport through the dunites. More radiogenic Os compositions in dunites resulting from melt-interaction infer that more radiogenic source lithologies are melting and these melts are reacting with the dunites. In order to verify this possible petrogenetic relationship, Os from a much larger number of harzburgite and dunite spinels from a single ridge segment or fracture zone need to be studied.
6.5. Cr-Spinel Osmium
 The separated and leached Cr-spinels from each of the abyssal peridotites show a similar heterogeneity in 187Os/188Os to the whole rock compositions (Figures 4 and 6). Upon first glance, this would tend to indicate that we have not succeeded in removing any of the heterogeneity attributed to seawater alteration. If we consider all of the Cr-spinels together, this may be the case, but it is interesting to note the difference in Os composition between the various Cr-spinel fractions. The magnetic Cr-spinels display a much smaller range in 187Os/188Os than nonmagnetic Cr-spinels and, except for a single non-magnetic Cr-spinel, have much higher Os concentrations. In fact, all but one of the magnetic Cr-spinels have a greater Os concentration than PUM, while the majority of the nonmagnetic Cr-spinels have concentrations below PUM. This division between magnetic and nonmagnetic fractions in Os isotope space stands to illustrate a number of potential differences. The variation in Os concentration, especially with nonmagnetic Cr-spinel dominantly below PUM, suggests a distinct difference in sulfide abundance between the spinel fractions, possibly coupled with variable leaching effectiveness. It would appear as though the nonmagnetic Cr-spinels have less Os due to lower abundances of sulfide. However, this has not been petrographically verified.
 Alternatively, it may reflect a difference in sulfide occurrence. Magnetic Cr-spinels have more sulfide inclusions that are not affected by leaching, whereas nonmagnetic Cr-spinels have higher amounts of interstitial sulfide that lose Os during leaching. Furthermore, this variation in sulfide occurrence could also explain the range in Os composition seen in the nonmagnetic Cr-spinels versus the magnetic Cr-spinels. The nonmagnetic fraction in many cases includes a small (<20%), but potentially influential, proportion of interstitial Cr-spinel/silicate grains. These multiphase grains failed to separate along true grain boundaries during mechanical separation. Highly altered (secondary) silicate phases could add additional radiogenic Os, thereby further compromising some of these Cr-spinels. Despite rigorous leaching (HF/ HCl /HNO3), it seems that the removal of silicates may not be as effective as removal of Fe-oxides. Regardless of the differences between magnetic and nonmagnetic spinels, on the whole, both fractions have less radiogenic 187Os/188Os values than their associated whole rock osmium compositions.
 To further verify that seawater alteration is responsible for most of the radiogenic Os isotopic heterogeneity, we conducted a crude mass balance experiment by collecting sequential leachates from each of the three acid-cleaning steps and measuring the Os composition and concentration stripped from three Cr-spinel samples. Results show that the Os compositions of the leachates for each sample are highly variable (Table 3). Sample “PROT 5 10-186nm” and “RC 27-9-34-84m” have Cr-spinel compositions that are less radiogenic than any of their three associated leachates, indicating that leaching is removing a more radiogenic seawater component. For example, leached spinel “10–186nm” has an 187Os/188Os of 0.1327, with corresponding 6.2 N HCl, HF/ HCl /HNO3, and conc. HCl leachate 187Os/188Os values of 0.1390, 0.1379, and 0.1504, respectively. This clearly shows that a radiogenic seawater Os component is present in both the silicate phases (leached by the HF/ HCl /HNO3 solution) and the Fe-oxide alteration phases (leached by HCl). This relationship between leachates and Cr-spinel did not hold for all samples, possibly due to loss of volatile OsO4 during separation of leachates.
6.6. Cr-Spinel Versus Whole Rock Osmium
 The pervasive serpentinization and low-temperature seawater alteration make whole rock Os measurements of abyssal peridotites difficult to interpret when attempting to characterize the present-day Os composition of DMM. Snow and Reisberg  showed that some of the heterogeneity in the existing abyssal peridotite Os data is the result of a seawater alteration component. They propose a present-day oceanic mantle composition of 0.125 (green band in Figures 4, 6, and 7), based on a combination of carefully selected data from the literature and their measured values. We take a different approach to the same problem and propose that measurements made on separated and carefully treated Cr-spinels from abyssal peridotites will result in Os compositions reflecting the upper mantle Os signature.
 This relationship between whole rock and Cr-spinel is evident in Figure 7. The dominant trend, delineated by arrows going from whole rock compositions to Cr-spinel compositions, is from radiogenic Os compositions at low Os concentrations to less radiogenic compositions at higher concentrations. This is precisely the trend that would be expected if a radiogenic seawater Os component were removed from a Cr-spinel with an unradiogenic Os signature. Several arrows show different trends, with Cr-spinel compositions more radiogenic than the whole rocks from which they were separated. Cr-spinel is not going to universally protect all sulfide-hosted Os signatures, as some highly fractured or veined grains will incur pervasive seawater alteration that is not easily removed by our leaching technique.
 Furthermore, while less than half of all Cr-spinels studied here are within the proposed DMM field (Figure 6), the majority is less radiogenic than PUM. On the basis of Re-Os systematics during melting, the depleted mantle composition should be less radiogenic than primitive upper mantle; even those samples that do plot at 187Os/188Os values higher than PUM are less radiogenic than their corresponding whole rock compositions. So, although the Cr-spinel Os compositions do not distinctly define the present-day composition of DMM, they do provide a more tightly constrained range for the depleted MORB source. This range may continue to shrink as further refinements of the Cr-spinel technique emerge, and subsequently the actual DMM Os signature can be better defined. The uncertainty in the Os composition of the “normal” depleted MORB source is not only a function of the small quantity of data available but also the complicated Os isotopic interpretation required of whole rock abyssal peridotite analyses.
 Although isotopic evidence has shown specific fracture zones along the SWIR and MAR (i.e., Du Toit F.Z.) to have heterogeneous mantle source regions, this study documents that much of the Os isotopic heterogeneity seen in N-MORB residual mantle is due to an alteration-induced radiogenic seawater Os component. Therefore, in many cases, interpretation of the upper mantle isotopic signature can be obscured. In order to circumvent this problem, we separated resistant Cr-spinel from a number of abyssal peridotites and carefully treated them to remove the seawater alteration component. ICP-MS sparging of the Cr-spinel solutions has produced Os compositions which still show heterogeneity, but which define two different populations. Nonmagnetic Cr-spinels have Os concentrations generally equal to or less than PUM and compositions spanning a range similar to the whole rock values. Magnetic Cr-spinels have higher concentrations than PUM and compositions generally equal to or less radiogenic than PUM, thereby illustrating that Cr-spinel can preserve a DMM-like Os signature. Despite the heterogeneity in the nonmagnetic Cr-spinel fraction, both fractions, when compared to their corresponding whole rock 187Os/188Os values, dominantly have less radiogenic 187Os/188Os. Not only does this trend, from radiogenic whole rocks to less radiogenic Cr-spinels, illustrate that the primary depleted upper mantle Os signature is obtainable, but it further constrains the Os isotopic signature for DMM.
 It is also interesting to note that Cr-spinels from the two harzburgites included in this study have two of the lowest 187Os/188Os ratios and TiO2 contents, consistent with the idea that the melt-reacted rocks (dunites) would have more radiogenic 187Os/188Os and high TiO2. This leaves open the possibility that the melts from which the dunites formed were not in equilibrium with the ambient mantle. However, the data set is neither large enough nor sufficiently refined to demonstrate this as fact, and further work is needed to address this issue in more detail. We have, however, shown that the radiogenic seawater Os component, which compromises the isotopic signature of most abyssal peridotites, can be removed, allowing more straightforward interpretation of the Os signature. This then provides the means to define, albeit still quite vaguely, the Os isotopic signature of N-MORB depleted upper mantle.
 This work was supported by NSF-EAR 9804891, NSF-OCE9416620, and NSF-OCE0096634. We thank Parker Hackett for conducting Cr-spinel separations during early phases of this work. Larry Ball at the WHOI ICP-MS facility has been extremely helpful, and we thank him for his patience and flexibility during Os sparging. Thanks is also extended to Neel Chatterjee in the MIT Electron Microprobe Facility for his help. Insightful discussions with Greg Ravizza and continual help in the lab from Tracy Abbruzzese were much appreciated. We would also like to thank Doug Pyle and Alan Brandon for their thoughtful and extremely helpful reviews and comments.