Rhenium, Os, 187Os/188Os, Ir, Ru, Pt, and Pd analyses of 43 samples from DSDP/ODP Hole 504B in the eastern equatorial Pacific crust with an age of ∼6.7 Myr are used to calculate the composition of a composite section through the upper oceanic crust. Weighted average Re (1526 pg/g), Os (21 pg/g), 187Os/188Os (0.213), 187Re/188Os (353), initial 187Os/188Os (0.173), Ir (9 pg/g), Pt (304 pg/g), and Pd (265 pg/g) are similar to a composite sample from DSDP Sites 417/418 in the western Atlantic with an age of ∼120 Myr (2153 pg Re/g, 33 pg Os/g, 187Os/188Os = 1.042 ± 0.016, 187Re/188Os = 349; 187Os/188Osi = 0.344, 25 pg Ir/g, 698 pg Ru/g, 804 pg Pt/g, and 2001 pg Pd/g). These sections complement existing data for ODP Site 735B (11–12 Myr) and, combined, provide the best available information on Re-PGE-Os isotope systematics of the upper and middle oceanic crust. Weighted average Os concentrations for samples from the volcanic zone, the transition zone with stockwork mineralization, and the sheeted dike complex decrease from 30 pg/g to 18 pg/g. Rhenium concentrations are highest in the transition zone (3867 pg/g), leading to 187Re/188Os of 885, compared to values of 178 for the volcanic zone and 393 for the sheeted dike complex. The present-day and initial 187Os/188Os are highest in the transition zone (0.329 and 0.230, respectively). This zone also has the highest average Pt (899 pg/g) and Pd (523 pg/g) concentrations. Of all lithologic units, breccias are most enriched in Re (4874 pg/g) and depleted in Os (8 pg/g), Ir (3 pg/g), Pt (131 pg/g), and Pd (173 pg/g) and have the most radiogenic present-day and initial 187Os/188Os (0.710 and 0.324, respectively). We contend that this is indicative of intense alteration, loss of PGE, and early addition of hydrogenous Os. Pillow lavas and massive flows and dikes of the volcanic zone and the sheeted dike complex are also affected by additions of hydrogenous Os, causing superchonditic initial 187Os/188Os. The weighted average Os/Ir of 2.4 for the entire core is significantly higher than values reported for mantle peridotites and estimates of the upper mantle and most likely indicates fractionation during melting and melt extraction or intracrustal differentiation rather than differential mobility of Os and Ir during alteration. Addition of hydrogenous Re to altered oceanic crust, although locally important, does not increase the Re concentration of oceanic crust by more than a few percent. The magnitude of mantle-derived Os loss to seawater cannot be reliably estimated from our data, but a steady state mass balance of hydrogenous Os indicates that the average loss cannot exceed 10% of the crustal inventory. Similarly, average addition of hydrogenous Os does not exceed 10% of the primary crustal inventory. This, however, leads to a significant increase in 187Os/188Os from depleted upper mantle values to ∼0.17. The characteristic fractionated PGE pattern and 187Re/188Os of fresh MORB are not fundamentally changed during hydrothermal alteration, and the magnitude of radiogenic ingrowth during storage of subducted oceanic crust for prolonged periods in the mantle is primarily determined by magmatic processes at the ocean ridges and, possibly, additional processing in subduction zones.
 The oceanic crust is severely depleted in platinum group elements (PGE: Ru, Rh, Pd, Os, Ir, Pt) relative to the Earth's core and mantle, and contains less than 0.005‰ of the terrestrial PGE budget. This depletion reflects the siderophile and chalcophile nature of PGE as well as their compatible behavior during mantle melting [e.g., Barnes et al., 1985; Greenough and Fryer, 1990; Leblanc, 1991; Fleet et al., 1994; Hart and Ravizza, 1996; Burton et al., 2002]. As the 187Os/188Os of unaltered, zero-age MORB should represent the 187Os/188Os of the depleted MORB mantle (DMM) that is being melted, early attempts to analyze the Os isotopic composition of MORB were motivated by characterizing this important mantle reservoir [e.g., Martin, 1991; Roy-Barman and Allègre, 1994]. The contrasting behavior of Re (moderately incompatible) and Os (compatible) during partial melting allows the use of the predicted primitive upper mantle 187Os/188Os (PUM, 187Os/188Os = 0.1296 [Meisel et al., 1996, 2001a]) as an upper limit for the isotopic composition of DMM. Most studies, however, have reported 187Os/188Os values of MORB that are more radiogenic than estimates of the DMM and show a negative correlation between 187Os/188Os and Os concentration [e.g., Martin, 1991; Roy-Barman and Allègre, 1994; Schiano et al., 1997]. Such signatures are generally interpreted as resulting from secondary alteration, either during ascent of magma through altered oceanic crust or after solidification by hydrothermal processes. However, preferential melting of enriched (i.e., high Re/Os) mantle components (pyroxenite) of a “marble cake” mantle [Allègre and Turcotte, 1986] has also been proposed as an explanation [e.g., Roy-Barman and Allègre, 1994; Roy-Barman et al., 1996; Schiano et al., 1997].
 Alteration of oceanic crust not only affects the geochemical evolution of the Earth's mantle but also constitutes an important source/sink for dissolved species in seawater. There is abundant evidence for enrichment of redox-sensitive elements such as Re [e.g., Martin, 1991; Colodner et al., 1993; Hauri and Hart, 1993; Brügmann et al., 1998] and U [e.g., Hart, 1969; Bloch, 1980; Hart and Staudigel, 1982] through hydrothermal circulation in oceanic crust. Hydrothermal leaching of elements with oceanic crustal isotope signatures also affects many marine isotope systems (e.g., Os [Ravizza et al., 1996; Sharma et al., 2000]; Sr [Albarède et al., 1981]; Nd [Piepgras and Wasserburg, 1985]; and Pb [Michard and Albarède, 1985]). Despite some evidence for exchange of Os during hydrothermal circulation [e.g., Ravizza et al., 1996; Brügmann et al., 1998; Sharma et al., 2000], comprehensive studies of the effects of alteration of MORB on the Re-Os system and PGE concentrations are missing. Such studies are needed to place boundary conditions on models of the long-term fate of oceanic crust returned to the mantle in subduction zones. We initiated this study to fill this gap.
2. DSDP/ODP Hole 504B
 DSDP/ODP Hole 504B is one of the very few deep drill holes into oceanic crust. The Site is located approximately 200 km south of the Costa Rica Rift (1°33.6′N, 83°43.8′W) in 6.57–6.94 Ma crust (Chron 3Ar [Alt et al., 1996a, 1996b; Berggren et al., 1995]). A detailed description of the petrology and geochemistry of 504B samples analyzed in this study is given in a companion paper [Bach et al., 2003]. Below we summarize only those parameters that have relevance for the interpretation of PGE data and Re-Os isotope systematics. Chondrite-normalized (subscript: CN) La/Sm of 0.25–0.48 (average: 0.39) in 504B lavas are significantly lower than N-MORB (La/SmCN of 0.65 [Hofmann, 1988]), indicative of the strongly depleted nature of this oceanic crust. Enriched samples with La/SmCN up to 1.6 do exist, but are rare [e.g., Pedersen and Furnes, 2001; Bach et al., 2003]. Despite the depleted nature of 504B basalts three lines of evidence indicate that lavas were sulfur-saturated: 1) The least altered samples (massive flows) have sulfur concentrations identical to those predicted on the basis of sulfur solubility for a given major element composition [Bach et al., 2003]. 2) Sulfides with primary textural features have been reported in massive 504B lavas [Distler et al., 1983]. 3) Median (310,000) and average (500,000) Cu/Pd values of unmineralized 504B samples are typical of N-MORB (S-saturated MORB: Cu/Pd > 100,000 [Momme et al., 2001]).
 Since inception of drilling during Leg 69 in 1981 the hole has been deepened in 6 subsequent drilling campaigns (Legs 70, 83, 111, 137, 140, 148) and provides a cross section through the upper ∼2 km of oceanic crust. The drill hole currently terminates in the lower portion of the sheeted dikes, probably close to the dike to massive gabbro transition. It penetrates a 274.5 m thick sedimentary section, the topmost volcanic section of mainly pillow and sheet lavas, a transition zone containing stockwork sulfide mineralization, and the major portion of the sheeted dike complex. Alteration features have been extensively documented [e.g., Alt et al., 1996a, 1996b; Bach et al., 2003]: A ∼300 m thick low-temperature seawater alteration zone at the top with primarily open circulation is underlain by a ∼250 m thick suboxic-anoxic alteration zone with alteration temperatures <150°C. A ∼209 m thick zone of intercalating flows and dikes that includes a 18 m thick sulfide impregnated stockwork zone marks the transition into the sheeted dikes that are characterized by greenschist facies to lower amphibolite facies mineral assemblages. The temperature of pervasive hydrothermal alteration increases downward in the sheeted dikes (300–500°C). It has been argued that base metals (e.g., Cu, Zn, Pb) and sulfur have been mobilized during high-temperature alteration at small water-rock ratios in the deep “reaction zone”. Copper and sulfur losses are probably related to oxidation of primary sulfides, whereas Zn is thought to be mobilized during replacement of primary Ti-magnetite by sphene and secondary magnetite [Bach et al., 1996, 2003]. Though the thickness of the “reaction zone” is not well known, it appears that Cu-, S-, and Zn-depletions in the recovered lower sheeted dike section, contrary to earlier inferences [Alt et al., 1995; Zuleger et al., 1995], are insufficient to account for enrichment of these elements in the zone of stockwork mineralizations if closed system is assumed [Bach et al., 2003].
 Due to the low recovery for the entire drill hole (∼18%) the recovered material does not necessarily represent the lithologic composition of the upper ∼2 km crust at Site 504. The severity of this bias and potential loss of soft lithologies such as breccias and clay-rich alteration zones in Hole 504B cannot be assessed with borehole logging data because the quality of the data is severely compromised by large variations in the borehole caliper. Haggas et al. , for example, recently demonstrated for the neighboring ODP Hole 896A that lithologic abundances estimated from Formation Microscanner logging data and core description data do not agree well for soft lithologies (breccias). The low recovery rates and lack of Formation Microscanner logging data for Hole 504B make it impossible to reliably estimate the statistical error associated with our composite calculation described in section 3.2, but they are likely large (i.e., ±50%, see section 5.1.). Despite this drawback, Hole 504B represents the best available and most intensely studied cross-section through upper oceanic crust. Together with a previous study of gabbroic rocks recovered from the top 500 m of ODP Hole 735B in the southwest Indian Ocean [Blusztajn et al., 2000] this study provides the most complete evaluation of the effects of alteration of oceanic crust on the Re-Os system and PGE concentrations. Nevertheless, we are fully aware that Holes 504B (3.6 cm/yr half-spreading rate) and 735B (0.8 cm/yr half-spreading rate) are not necessarily representative of altered oceanic crust in general. This fundamental issue needs to be resolved through drilling into altered oceanic crust in the next phase of ocean drilling.
3. Samples and Analytical Procedure
3.1. Analytical Procedures
 We selected samples from the working halves of the cores stored at the DSDP West Coast Repository in 1998. Seventy-one samples were cut, sand-blasted, carefully cleaned, air-dried, and ground in a new agate shatter box. We selected a subset of 43 samples for Re-PGE-Os isotope analyses. Depending on the available sample mass between 3–32 grams (typically 10–20 grams) each was spiked with a mixed PGE spike enriched in 99Ru, 105Pd, 190Os, 191Ir, and 198Pt. The PGE were preconcentrated using a NiS fire assay technique developed at WHOI [Ravizza and Pyle, 1997], using equal or less than equal (0.8–1) proportions of flux and sample. 187Os/188Os and Os concentrations were measured by sparging volatile OsO4 directly into a single-collector, magnetic sector ICP-MS (Finnigan ELEMENT) according to a procedure described by Hassler et al. . The external reproducibility of 187Os/188Os for 80 pg Os (in-house Johnson-Mathey standard Os solution) is ∼0.7% (95% confidence interval) for the sparging technique. The average 187Os/188Os of 80 pg in-house standard solution analyses is, within uncertainty, identical to ICP-MS analyses of the same standard at higher concentrations (0.41 and 1.23 ng) as well as the long-term average of the same standard analyzed by N-TIMS (see Table 1).
|Rock Standarda||Gram||Split/Positionb||Rock Type||187Os/188Osc ± 2σm||Os, pg/g||Ir, pg/g||Ru, ng/g||Pt, ng/g||Pd, ng/g||Os/Ir|
|BHVO-1 (USGS)||4.0||58/23||Hawaiian basalt||0.131 ± 0.006||92||97||2.4||2.7||0.95|
|BHVO-1 (USGS)||4.1||58/23||Hawaiian basalt||0.1316 ± 0.0008||91||105||2.7||3.0||0.87|
|BHVO-1 (USGS)||2.9||58/23||Hawaiian basalt||0.1320 ± 0.0011||95||76||1.8||2.7||1.25|
|BHVO-1 (USGS)||3.5||58/23||Hawaiian basalt||0.1309 ± 0.0014||105||103||0.39||2.3||2.8||1.02|
|BHVO-1 (USGS)||3.6||41/5 (URI)||Hawaiian basalt||0.1349 ± 0.0008||90||100||2.9||3.1||0.90|
|95% confidence limitd||0.0018||7||14||0.5||0.2||0.17|
|Relative 95% C.L., %||1.4||7||15||19||7||17|
|Proposed or informational valuese||440||2||3|
|Finnegan et al. ||0.95|
|EN026 10D-3 (URI)||9.1||#4||MORB, Mohns Ridge||0.1545 ± 0.0018||14||23||0.56||1.5||0.61|
|EN026 10D-3 (URI)||9.2||#4||MORB, Mohns Ridge||0.1577 ± 0.0014||20||28||0.72||1.4||0.71|
|EN026 10D-3 (URI)||9.0||#4||MORB, Mohns Ridge||0.1672 ± 0.0019||18||29||0.52||1.5||0.62|
|95% confidence limitd||0.013||6||7||0.2||0.1||0.13|
|Relative 95% C.L., %||8.4||36||25||36||8||21|
|TDB-1 (CCRMP)||5.1||B-502||Diabase, Canada||0.811 ± 0.006||126||72||4.4||1.75|
|TDB-1 (CCRMP)||6.9||B-502||Diabase, Canada||0.888 ± 0.005||125||78||4.3||24.3||1.60|
|TDB-1 (CCRMP)||5.8||B-502||Diabase, Canada||0.881 ± 0.007||120||77||4.3||23.8||1.56|
|TDB-1 (CCRMP)||5.6||B-502||Diabase, Canada||0.8925 ± 0.0026||107||77||4.2||24.6||1.39|
|TDB-1 (CCRMP)||4.8||B-502||Diabase, Canada||0.776 ± 0.006||127||88||4.9||25.1||1.44|
|TDB-1 (CCRMP)||4.6||B-502||Diabase, Canada||0.8397 ± 0.0027||118||70||4.0||1.69|
|TDB-1 (CCRMP)||5.7||B-502||Diabase, Canada||0.786 ± 0.004||124||80||4.4||1.55|
|TDB-1 (CCRMP)||5.0||B-502||Diabase, Canada||0.785 ± 0.004||125||84||4.7||26.2||1.49|
|95% confidence limitd||0.04||5||5||0.2||1.0||0.09|
|Relative 95% C.L., %||4.6||4.3||5.9||5.0||4.1||6.0|
|Certified, provisional, or informational valuesf (95% C.L.)||150||0.3||5.8||22.4|
|Literature Values TDB-1||Analytical Methodsg||Osh||Irh||Ruh||Pth||Pdh|
|Sen Gupta ||GF-AAS||5||25|
|Paukert and Rubeska ||20||Fire-assay, GF-AAS (n = 5)||<3000||<2||6.7 ± 2.4||24.3 ± 0.5|
|Enzweiler et al. ||0.5||ID, fusion, Te-coprec., ICP-MS (n = 2)||0.37||5.7||22.4|
|Farago et al. ||10||ICP-MS (n = 5)||7.15 ± 0.37|
|Farago et al. ||10||ICP-MS (n = 1)||6.68|
|Farago et al. ||10||ICP-MS (n = 1)||4.93|
|Plessen and Erzinger ||50||Fire-assay, ICP-QMS (n = 17 − 25)||120 ± 19||0.34 ± 0.079||3.8 ± 0.59||20 ± 1.7|
|Meisel et al. [2001b]||<2||ID, HPA aqua regia, ICP-MS (n = 12)||112 ± 70||77 ± 11||0.24 ± 0.05||5.02 ± 0.21||23.0 ± 2.9||1.5|
|Rock Standarda||Gram||Split/Positionb||Rock Type||187Os/188Osc ± 2σm||Os, pg/g||Ir, pg/g||Ru, ng/g||Pt, ng/g||Pd, ng/g||Os/Ir|
|WGB-1 (CCRMP)||5.9||B-984||Wellgreen gabbro||0.1889 ± 0.0005||484||170||5.2||14.5||2.85|
|WGB-1 (CCRMP)||4.7||B-984||Wellgreen gabbro||0.183 ± 0.003||1027||674||13.8||22.4||1.52|
|WGB-1 (CCRMP)||5.9||B-984||Wellgreen gabbro||0.1850 ± 0.0012||493||169||4.0||15.1||2.92|
|WGB-1 (CCRMP)||5.5||B-984||Wellgreen gabbro||0.2040 ± 0.0005||407||194||3.9||15.6||2.10|
|WGB-1 (CCRMP)||5.3||B-984||Wellgreen gabbro||0.1909 ± 0.0006||481||236||5.2||14.7||2.04|
|WGB-1 (CCRMP)||5.0||B-984||Wellgreen gabbro||0.2016 ± 0.0006||385||194||3.9||13.5||1.98|
|WGB-1 (CCRMP)||5.6||B-984||Wellgreen gabbro||0.1934 ± 0.0009||547||187||4.5||15.6||2.93|
|WGB-1 (CCRMP)||5.0||B-984||Wellgreen gabbro||0.1524 ± 0.0008||1719||207||5.5||18.2||8.30|
|95% Confidence Limitd||0.012||361||134||2.6||2.2||1.7|
|Relative 95% C.L. (%)||6.7||51||53||45||14||55|
|Certified, provisional, or informational values6 (95% C.L.)||330||0.3||6.1||13.9|
|Literature Values WGB-1||Analytical Methodsg||Os||Ir||Ru||Pt||Pd|
|Sen Gupta ||GF-AAS||0.2||21|
|Paukert and Rubeska ||20||Fire assay, GF-AAS (n = 5)||3300 ± 200||4.6 ± 1.4||5.5 ± 1.7||10.9 ± 2.2|
|Hall and Pelchat ||Fire assay, ICP-MS||11.7 ± 2.4|
|Enzweiler et al. ||0.5||ID, fusion, Te-coprec., ICP-MS (n = 2)||0.28||5.7||14.1|
|Pattou et al. ||15||ID, fire assay, Te-coprec., ICP-MS||220||0.26||5.3||12.6|
|Plessen and Erzinger ||50||Fire assay, ICP-QMS (n = 28 − 30)||200 ± 34||0.20 ± 0.04||3.8 ± 1.0||13 ± 1.1|
|Amossé ||10||Fusion, Te-coprec., ICP-QMS (n = 6)||240 ± 10||0.78 ± 0.12||5.4 ± 0.6||15.1 ± 0.49|
|Morcelli and Figueiredo ||0.5||RNAA||200 ± 59|
|Müller and Heumann ||0.2||ID, HPA aqua regia, ICP-QMS (n = 4)||650 ± 170||1.7 ± 0.5||6.0 ± 1.7||10.6 ±2.7|
|Müller and Heumann ||0.2||ID, HPA AR + HF, ICP-QMS (n = 4)||640 ± 280||1.6 ± 0.9||6.3 ± 1.2||8.0 ± 0.7|
|Schmidt et al. ||10||Fire assay, INAA||430 ± 20||220 ± 10||<0.3||11.2 ± 0.1|
|Schmidt et al. ||10||Fire assay, INAA||440 ± 20||220 ± 10||<0.3||11.0 ± 0.1|
|Meisel et al. [2001b]||<2||ID, HPA aqua regia, ICP-MS (n = 7)||550 ± 120||270 ± 70||0.16 ± 0.02||4.71 ± 0.4||11.7 ± 1.3|
|Literature Values WGB-1||Gram||Analytical Methodsg||187Os/188Osc ± 2σm|
|Schmidt and Snow ||10||Fire assay, Br2 + microdist., N-TIMS||0.15824 ± 0.00019|
|Schmidt and Snow ||10||Fire assay, Br2 + microdist., N-TIMS||0.15815 ± 0.00017|
|Procedural Blanks||Gram||Fusion Reagents||187Os/188Osc ± 2σm||Os, pg/g||Ir,i pg/g||Ru, pg/g||Pt, pg/g||Pd, pg/g|
|Average values ± 95% confidence interval||5–20||Powdered borax, Ni, S (30/2/1.2 weight ratio)||0.61 ± 0.06||0.39 ± 0.05||0.70 ± 0.16 (1.7 ± 0.9)||64 ± 26||26 ± 6||14 ± 6|
|Number of Analyses||20||20||16 (21)||4||24||15|
 Subsequent to 187Os/188Os analyses the liquid residues after sparging were analyzed for Ru, Pd, Ir and Pt concentrations using conventional liquid uptake through a desolvating nebulizer into the ICP-MS. To check the internal consistency of the calculated concentrations, Ru, Pd, and Pt concentrations were calculated using two isotope ratios. In most cases, the calculated concentrations agree to within a few percent of each other. However, Ru concentrations calculated using 99Ru/101Ru and 99Ru/102Ru, corrected for interferences from Pd, often did not agree well. We do not know the cause for this discrepancy but we suspect the presence of an unidentified molecular interference on mass 102. Ruthenium concentrations are therefore reported only if both Ru concentrations agree to better than 30%. Concentrations (with 95% confidence intervals) of analytical blanks per gram flux are 0.4 ± 0.05 pg Os with a 187Os/188Os of 0.613 ± 0.062 (n = 20), 1.7 ± 0.9 pg Ir (n = 21), 64 ± 26 pg Ru (n = 4), 26 ± 6 pg Pt (n = 24), and 14 ± 6 pg Pd (n = 15). The higher and less precise average Ir blank compared to the average Os blank [cf. Meisel et al., 2001b] is caused by a bimodal distribution of Ir blank data. The majority of Ir blanks (n = 16) yield average values of 0.7 ± 0.2 pg, but five Ir blanks yielded concentrations between 2 and 8 pg/g. We decided to correct bulk Ir concentrations >2 pg/g with the average of all Ir blank determination, but disregard Ir blanks >2 pg/g for samples with bulk Ir concentrations <2 pg/g. As most samples were fused with a flux to sample ratio less than unity (0.8–1) blank corrections normalized to sample mass can be up to 20% less than the values listed above for flux-normalized blanks. Blank-corrected concentrations are listed in Table 2 and reported uncertainties include uncertainties of the blank composition.
|Hole-Core Section, cm, Pieceb||dbsf,c m||Lithologyd||W.F.e||187Os/188Osf||±2σ||Os, pg/g||Re, pg/g||187Re/188Osg||187Os/188Osih||Ir, pg/g||Ru, pg/g||Pt, pg/g||Pd, pg/g|
|504B-7R3, 32–35, 478||310.8||VZ/PR||0.02163||0.181||0.007||46||1179||124||0.167||15||74||247||419|
|504B-13R4, 47–51, 806||366.5||VZ/PG||0.02163||0.187||0.006||48||941||96||0.176||30||232||783||1161|
|504B-16R4, 0–12, 1003||388.6||VZ/PR||0.02163||0.1616||0.0019||30||358||59||0.155||27||85||536||612|
|504B-19R1, 20–25, 1115||403.2||VZ/PR||0.02163||0.182||0.015||63||667||51||0.176||11||50||224||222|
|504B-23R1, 33–40, 1315||439.4||VZ/PG||0.02163||0.326||0.013||7||547||398||0.281||8||185||403|
|504B-28R1, 90–99, 1496||475.9||VZ/PG||0.02163||0.210||0.012||15||2704||871||0.112||1||120|
|504B-28R2, 103–108, 1514||477.6||VZ/PR||0.02163||0.507||0.017||4||643||776||0.420||4||140||86||205|
|504B-28R3, 58–64, 1529||478.6||VZ/PG||0.02163||0.42||0.03||3||1555||3054||0.072||5||59||46||60|
|504B-36R4, 26–34, 329||548.3||VZ/MR||0.01127||0.1358||0.0007||259||1781||33||0.132||51||169||89|
|504B-37R3, 5–12, 390||555.6||VZ/PR||0.02163||0.573||0.006||3||297||442||0.524||1||71||46|
|504B-39R2, 75–84, 490||572.8||VZ/PG||0.02163||0.2600||0.0026||11||1340||600||0.192||6||127||121|
|504B-40R3, 130–138, 558r||583.8||VZ/MR||0.01127||0.234||0.003||24||677||136||0.219||14||155||304||366|
|504B-40R3, 130–138, 558g||583.8||VZ/MG||0.01127||0.1663||0.0026||47||1582||163||0.148||25||107||418||577|
|504B-49R2, 25–30, 938||657.8||VZ/B||0.00626||0.485||0.020||5||816||808||0.394||1||95||44||146|
|504B-51R1, 81–87, 981||670.3||VZ/MG||0.01127||0.1609||0.0022||55||1250||110||0.149||20||981||1444|
|504B-60R1, 76–85, 1264||742.3||VZ/MG||0.01127||0.2899||0.0029||9||1423||821||0.198||5||89||79|
|504B-64R1, 14–20, 1404||773.2||VZ/B||0.00626||0.517||0.018||4||1132||216||0.336||1||195||41||55|
|504B-66R2, 130–136, 1522||793.8||VZ/PG||0.02163||1.97||0.05||1||1445||10672||0.771|
|504B-70R1, 23–30, 1550||827.3||VZ/B||0.00626||1.34||0.11||1||1353||6916||0.619||38|
|504B-73R1, 74–80, 5E||853.3||TZ/MG||0.01045||0.2585||0.0022||9||1075||586||0.192||5||43||136|
|504B-79R2, 51–59, 3||906.6||TZ/MG||0.01045||0.323||0.006||11||1977||914||0.220||6||263||116||118|
|504B-80R1, 62–70, 5||910.7||TZ-SW/PG||0.01212||0.343||0.008||69||8551||614||0.274||30||123||785||1911|
|504B-80R2, 10–16, 1||911.6||TZ-SW/PG||0.01212||0.244||0.023||28||2923||518||0.185||3||368||77|
|504B-80R2, 124–132, 11||912.8||TZ-SW/B||0.00173||0.2515||0.0028||28||2126||376||0.209||12||380||476||815|
|504B-80R3, 47–53, 4A||913.5||TZ-SW/B||0.00173||0.210||0.004||31||368||59||0.203||14||69||501||875|
|504B-81R1, 10–23, 2||919.7||TZ-SW/B||0.00173||0.290||0.012||5||1598||1564||0.114||4||63||118|
|504B-82R2, 26–34, 4||930.3||TZ/B||0.00173||0.349||0.018||4||2697||3299||−0.022||1||78||77||190|
|504B-82R3, 9–13, 2A||931.6||TZ/B||0.00173||0.483||0.017||4||660||837||0.388||3||308||117|
|504B-83R1, 82–90, 5A||938.4||TZ/PG||0.01212||0.2433||0.0026||30||1579||255||0.215||13||267||515||1162|
|504B-84R1, 141–146, 11||947.9||TZ/PG||0.01212||0.2803||0.0026||8||786||463||0.228||8||76||157|
|504B-85R2, 89–99, 4B||957.9||TZ/MG||0.01045||0.210||0.013||33||1311||194||0.188||24||678||12918||957|
|504B-87R1, 17–23, 2B||967.2||TZ/B||0.00173||4.07||0.09||3||10376||29930||0.707||1||1351||43|
|504B-89R1, 23–28, 2C||985.8||TZ/PG||0.01212||1.77||0.08||7||14994||12503||0.360||4||64||192|
|504B-91R3, 45–51, 6B||1007.0||TZ/B||0.00173||0.193||0.005||33||2149||321||0.157||18||62||548||781|
|504B-94R2, 20–29, 2B||1032.2||TZ/MG||0.01045||0.287||0.009||6||1260||1061||0.168||3||118||313||97|
|504B-99R1, 44–52, 6C||1071.9||SDC/D||0.09159||0.21||0.04||21||1148||536||0.151||55|
|504B-101R1, 105–111, 10B||1090.6||SDC/D||0.09159||1.27||0.05||0.5||825||9147||0.185||2||64||144|
|504B-107R1, 104–111, 14||1144.6||SDC/D||0.09159||0.195||0.016||20||1221||293||0.162|
|504B-122R1, 76–85, 6||1214.3||SDC/B||0.01145||1.89||0.05||2||3480||11927||0.545||69||52|
|504B-130R2, 69–82, 4A||1281.3||SDC/D||0.09159||0.1859||0.0018||15||1077||351||0.146||8||170||272||175|
|504B-134R1, 136–143, 19||1314.4||SDC/D||0.09159||0.229||0.013||27||1211||216||0.205||209||351|
|504B-141R1, 50–57, 7||1346.0||SDC/B||0.01145||2.75||0.06||6||14593||17103||0.826||3||160||177|
 Rhenium concentrations were determined by isotope dilution ICP-MS on separate 0.1–0.2 g splits of the basalt powder used for Os isotope and PGE analyses. It should be noted that some of the scatter in isochron plots may be caused by not analyzing Re and Os on the same sample split (nugget effect). The samples were dissolved in HF-HNO3-HCl and equilibrated with a 185Re spike in all-Teflon microwave vessels. Subsequently, Re was separated on a 1 ml column of AG1X8 (200–400 mesh) in 0.5N HNO3, eluted with 8N HNO3 and analyzed by ICP-MS. Blank corrections do not exceed 1% of the total Re. Additional details on the analytical procedures are given in Table 1 and, for complementary major and trace elements, by Bach et al. .
3.2. Accuracy and Precision of PGE Data
 Low PGE concentrations, inhomogeneous distribution of PGE-rich trace phases (nugget effect), and problems with complete dissolution of rock powders have made accurate and precise analysis of PGE in basalts notoriously difficult. Analytical problems are compounded by the lack of international rock standard reference materials (SRM) with certified PGE concentrations at sub-ng/g levels. While Rehkämper et al.  have shown that replicate analyses of 5–10 gram Icelandic basalt BTHO yield reproducible (i.e., <10%, 1σ) Ir, Ru, Pt and Pd concentrations, Meisel et al. [2001c] demonstrate that replicate measurements of up to 3 gram well-homogenized pyroxenite powder (sample 97So20) can yield nearly six-fold variations in Os concentrations. Similarly, variations in Ir concentrations of more than an order of magnitude were observed by Crocket  for 0.1–0.25 g splits of altered Kilauea basalts. We emphasize two points: Firstly, excellent reproducibility of PGE data does not guarantee accuracy of the data. For instance, identical splits of a basaltic SRM analyzed for Ir gave three times higher concentrations when NiS fire assay was used compared to very reproducible results obtained by dissolving the rock powder using mineral acids in Teflon beakers (J.-G. Schilling, personal communication, 2003). These results corroborate the notion that acid digestion in Teflon is an unreliable method for accurate PGE analysis (T. L. Meisel et al., Re-Os systematics of UB-N, a serpentinized peridotite reference material, submitted to Chemical Geology, 2003) (hereinafter referred to as Meisel et al., submitted manuscript, 2003). Secondly, at the concentration level of MORB nugget effects are nearly impossible to avoid. For instance, addition of a 15 μm PGE-rich nugget with 30 wt% Os and a density of 10 g/cm3 to 100 gram sample powder will lead to an increase Os concentration by ∼50 pg/g. If PGE-rich trace phases are present in samples with low PGE contents, poor reproducibility of PGE concentrations is all but unavoidable. In order to evaluate the reproducibility and accuracy of our analytical method we have therefore analyzed four SRM: 1) the Hawaiian basalt standard BHVO-1, prepared by the USGS but not certified for PGE; 2) the in-house MORB from the Mohns Ridge, Greenland Sea, prepared by Jean-Guy Schilling at the University of Rhode Island (EN026 10D-3). We also analyzed two rock SRM prepared by the Canadian Certified Reference Materials Project (CCRMP) with certified Pt and Pd concentrations; 3) the Mesoproterozoic (∼1265 Ma) diabase TDB-1 from Tremblay Lake, Saskatchewan, Canada, and 4) the gabbro WGB-1 from the Triassic Wellgreen Complex, Yukon Territory, Canada. Provisional (Ir) or informational (Ru) values are also available for these Canadian SRM. It should be noted that Hall and Pelchat  and Meisel et al. [2001b] (for sample sizes ≤2 gram) provide evidence that WGB-1 is less homogenous than TDB-1.
 Given the susceptibility of geologic materials to inhomogeneous distribution of PGE we argue that it is justified to use the best reproducibility of any SRM analyzed as a measure of analytical performance. Our results for these rock SRM (Table 1) give a complex picture of analytical accuracy, precision and the quality of SRM: Data for BHVO-1 (n = 5) indicate that our analytical method can reproduce, at the 95% confidence level, 187Os/188Os to 1.4%. Platinum group element concentrations in TDB-1, apparently the most homogenous SRM, are reproducible to better than 6%. The somewhat poorer reproducibilities of Ir and Pt concentrations in BHVO-1 are caused by a single analysis. Results from TDB-1 alone (n = 8) indicate that we can reproduce 187Os/188Os, as well as Os, Ir, Pt and Pd concentrations to better than 6% (95% confidence interval). Measured Pt and Pd concentrations agree well with certified and proposed values. Three analyses of the low-level SRM EN026 10D-3 (e.g., 17 pg/g Os) indicate that 187Os/188Os and Os concentrations are reproducible to 9% and 36%, respectively (95% confidence interval). It is unclear at this point whether the poorer reproducibility for this low-level reference material reflects heterogeneity, analytical difficulties at this low concentration level, or a combination of both. Surprisingly, results for the SRM with the highest PGE concentrations (WGB-1) do not reproduce well at sample sizes of 4–6 gram. Uncertainties at the 95% confidence level range from 7% for 187Os/188Os up to 53% for Ir concentrations. 187Os/188Os are negatively correlated with Os concentrations, indicative of the presence of at least two phases, one with unradiogenic 187Os/188Os and high PGE concentrations, the other with more radiogenic 187Os/188Os and lower PGE concentrations. The poor reproducibility is caused by one analysis with high concentrations for all PGE, and one analysis with elevated Os and Pd concentrations. Our average Os concentration data for WGB-1 are within uncertainty identical to those reported by Schmidt et al.  and Meisel et al. [2001b], but our 187Os/188Os is significantly more radiogenic (0.1870 vs. 0.15820) than that of Schmidt et al. . This difference may indicate heterogeneity between the powder splits used in both labs, or indicate differences in analytical techniques used in both studies (see also Meisel et al. [2001b, 2001c, submitted manuscript, 2003] for the effect of different analytical techniques on reported PGE concentrations). As the grand average of eight WGB-1 analyses agrees within uncertainties with the certified Pd and Pt concentrations and data from many other laboratories, we contend that the poor reproducibility reflects inhomogeneity of this SRM at sample sizes of 4-6 gram. Homogeneity tests were performed by CCRMP on 25 gram samples and analytical work as part of the certification process was done using ≥10 gram (mostly 15–30 gram) samples.
 Os/Ir in the four reference materials vary from low values of 0.64 (±21%, 95% confidence interval) in EN026 10D-3 to high values of 3.1 (±55%, 95% confidence interval) in WGB-1. The fact that uncertainties are largest in the SRM with the highest Os and Ir concentrations indicates that sample heterogeneity, and not analytical problems, cause most of the scatter in the data. This notion is supported by Os/Ir data for TDB-1 that are reproducible to within 6% at the 95% confidence level. BHVO-1 has a Os/Ir of ∼1 (±17%, 95% confidence interval) that agrees with data by Finnegan et al. .
 In summary, we contend that large differences in PGE concentrations and 187Os/188Os, observed in several 504B basalts (e.g., 504B-28R1, 90–99, 1496, or 504B-85R2, 89–99, 4B), are most likely caused by nugget effects [e.g., Meisel et al., 2001c] and not analytical problems. The fact that replicates of these samples define trends that are sub-parallel to a 6.7 Myr reference line (shown for sample 504B-85R2, 89–99, 4B in the middle panel of Figure 2) in a Re-Os isochron diagram corroborate this notion. The most homogenous SRM for low-level PGE analyses and 187Os/188Os, BHVO-1, unfortunately is not longer available in large enough quantities to serve as a SRM for the PGE community in the future. Until a similarly homogenous SRM with low PGE concentrations becomes widely available we recommend the use of SRM TDB-1 as a measure of analytical performance at concentration levels typical of crustal rocks.
4.1. Analytical Data
 Analytical results for 187Os/188Os as well as Re and PGE concentrations are listed in Table 2. Complementary data for major and trace elements, S- and Sr-isotopes, and petrographic data are published in a companion paper [Bach et al., 2003, Tables 1 and 2]. Rhenium and Os concentrations range from 0.3–15 ng/g and <1–276 pg/g, respectively. 187Re/188Os range from 32–30,000 (Figures 1 and 2). Such high 187Re/188Os lead to rapid ingrowth of radiogenic Os, resulting in present-day 187Os/188Os of up to 4.07 in Os-poor samples (Figure 1, filled symbols). With few exceptions the samples plot above a 6.75 Myr reference isochron (Figure 2). Initial 187Os/188Os, calculated assuming closed system and using Re and Os data on different sample splits, range from −0.02–0.83 and are on average superchondritic (Figure 1, open symbols). Replicate analyses of several samples show that Os (and other PGE) concentrations can vary by up to an order of magnitude (504B-85R2, 89-99, 4B; 3 and 33 pg/g) and are generally negatively correlated with 187Os/188Os. The fact that Os concentrations and 187Os/188Os reproduce well in most rock SRM and some 504B rocks despite low Os content (504B-101R1, 105–111, 10B; 0.4, 0.5, and 0.6 pg/g; 187Os/188Os = 1.018, 1.271, and 1.679) indicates that sample heterogeneity and not analytical problems determine the reproducibility of replicate analyses. Repeat Re determinations agree to within 3–6% (504B-36R4, 26–34, 329; 504B-130R2, 69–82, 4A; 418 SC). In contrast, Ir, Ru, Pt, and Pd concentrations in 504B basalts vary by several orders of magnitude from <1–51 pg/g, <50–678 pg/g, <10–∼13,000 pg/g, and <34–1911 pg/g, respectively. In about half of the repeat PGE measurements variations in absolute concentration and concentration ratios exceed a factor of two, independent of absolute concentration. Variations in primitive mantle-normalized PGE ratios are large even for element pairs such as Os/Ir (0.5 to ∼15) that usually show minimal magmatic fractionation despite orders of magnitude variability in absolute concentration. Large deviations of Os/Ir from unity are not restricted to samples with low concentrations and do not correlate with Os or Ir concentrations. Despite significant variability between individual samples the primitive mantle-normalized pattern show the distinct fractionation between Os-Ir and Pt-Pd that is typical of MORB (Figure 3) [see also Hertogen et al., 1980; Rehkämper et al., 1999], but note that Hartmann  found largely unfractionated primitive mantle-normalized PGE pattern in MORB from the East Pacific Rise and the Northern Lau Basin. We consider these results questionable because isotope dilution was not used and Ir concentrations are near the detection limit.
4.2. Weighted Averages
 The most striking difference between the proportions of recovered [Alt et al., 1996b] to analyzed lithologies is the overabundance of breccias among the analyzed samples. In addition, we analyzed disproportionately many samples from the transition and volcanic zones, and disproportionately few samples from the sheeted dike complex. This bias is partly justified by the greater lithologic complexity of the volcanic and, particularly, the transition zones compared to the sheeted dike complex. In an attempt to correct for our sampling and analytical bias we use petrographic data for recovered material from 504B [Alt et al., 1996b] together with similar petrographic data for our samples [Bach et al., 2003] to calculate weighting factors (W.F., see Table 2). These factors are used together with PGE concentrations and 187Os/188Os to calculate the average weighted isotopic composition and concentrations for each lithologic unit and magmatic zone. We emphasize that this procedure assumes that the recovered material faithfully represents the crustal section drilled. Due to the overall low recovery of only ∼18% this assumption is afflicted with large, but unquantifiable uncertainties. However, this procedure makes best use of the existing petrographic and geochemical data. The basic data used for calculating weighted averages are listed in Table 3. The weighting factor (W.F.) for a lithology i (i.e., breccias, lava, massive flow/dike) from zone j (i.e., volcanic zone, transition zone, sheeted dike complex) is a function of the thickness (Δm) of each zone relative to the total thickness (m) of the magmatic section in hole 504B, the number of samples analyzed from each lithology (Ni) and the fraction of this lithology within a given zone (fij), with the sum of all weighting factors being equal to one.
The weighting factors are listed in Table 2, and the results of the weighting procedure for the various lithologies and magmatic zones are summarized in Tables 4 and 5. We emphasize that average weighted 187Os/188Os and 187Re/188Os values were calculated by taking differences in concentrations and isotopic compositions into account. We also use the weighted average values to calculate a supercomposite for the entire drill hole and compare it to a physical mixture representing a supercomposite of DSDP holes 417/418 in ∼120 Ma Atlantic Ocean crust [Hart and Staudigel, 1982, 1989; Staudigel et al., 1995]. As this procedure assumes that recovered and analyzed samples faithfully represent the crustal section drilled, it is very difficult to assign uncertainties to estimated composite compositions. Due to the low recovery and large scatter in absolute concentrations the uncertainties are likely on the order of ±50%.
|Volcanic Zone (VZ)||Transition Zone (TZ)||Sheeted Dikes (SDC)|
|Recovered core, m|
|Red halo pillows||99.9||0||0|
|Red halo massive flows||23.66||0||0|
|Gray massive flow/dike||79.21||76.4||0|
|Recovered core, %|
|Analyzed samples: Number, %|
|Pillows||11 (58)||5 (31)||0 (0)|
|Flows/dikes||5 (26)||4 (25)||6 (75)|
|Breccias||3 (16)||7 (44)||2 (25)|
|504B SC||Volcanic Zone||Transition Zone||Sheeted Dike Complex||Pillows||Flows/Dikes||Breccia||417/8 SC|
|Os (pg/g)||21 (20)||30||22||18||22||23||8||33 (28)|
|417/8 SC||504B SC||504B VZ||504B TZ||504B SDC||504B P||504B F/D||504B B|
 Element abundance patterns normalized to N-MORB [Hofmann, 1988] are shown in Figure 4 for the different zones and lithologies (see figure caption for details). Reference pattern for 504B (red line) and 417/418 (gray line) supercomposites are also shown. The depleted nature of 504B lavas is most clearly seen in the least altered samples (i.e., massive flows/dikes and samples from the sheeted dike complex). However, the apparent N-MORB nature of average pillow lavas (Figure 4, lower panel) is in part caused by a single E-MORB sample (504B-19R1, 20–25, piece 1115). Without this sample average pillow lavas had slightly lower normalized Rb, U, and K, concentrations, but normalized Ba and Th concentrations were similar to average breccia and flows/dikes. Relative Re enrichments are most pronounced in breccias and rocks from the transition zone. Absolute Re enrichments critically depend on the choice of Re concentration in N-MORB (1000 pg/g [Hauri and Hart, 1997]; ∼960 pg/g [Schiano et al., 1997]). Major and trace element systematics are discussed in more detail in a companion paper [Bach et al., 2003].
 Variations in Re and Os concentrations, 187Re/188Os and present-day as well as initial 187Os/188Os with depth are shown in Figure 1 together with weighted average values for each zone (see next paragraph). Rhenium concentrations are lowest in pillow lavas from the topmost (oxic) part of the volcanic zone. Samples with exceedingly high Re concentrations are found exclusively in the deeper (predominantly reducing) sections of the hole. Osmium concentrations generally appear to decrease with depth (see weighted averages, Figure 1). Samples with exceedingly high 187Re/188Os and present-day 187Os/188Os are restricted to rocks with <10 pg/g Os from the reducing portion of the hole. Initial (6.75 Myr) 187Os/188Os are on average superchondritic (0.173), particularly in breccias (0.324) and in rocks from the transition zone (0.23; see Figures 1 and 2).
5.1. Estimating the Chemical Composition of Magmatic Zones, Major Lithologic Units, and a Supercomposite 504B
 The combined petrographic and geochemical data have been used to estimate the weighted chemical and isotopic composition of major lithologic units, magmatic zones, and the entire core (see Figures 1–4). We emphasize that our weighted averages deviate slightly from those of Bach et al. [2003, Table 4]. Bach et al.  used the entire data set (n ≤ 69) and not just those samples analyzed for Re-Os-PGE (n ≤ 43). For the supercomposite, for instance, relative differences between both data sets never exceed 38%, indicating that our error estimate of ±50% is conservative. Ruthenium was not included in our weighting exercise because Ru concentrations could not be determined in 16 out of 43 samples. In contrast, concentrations in only five (Ir), three (Pt), and six (Pd) samples were below detection limit. The results indicate that all magmatic zones and lithologies are affected to various degrees by alteration. All except four samples have superchondritic initial 187Os/188Os, with average values as high as 0.324 in breccias, the most permeable lithology in the core [Bach et al., 2003, and references therein]. Extremely variable 187Re/188Os indicate redistribution of Os and/or Re throughout the hole, most significantly in breccias and rocks from the transition zone. Large variations in Re and, to a minor degree, Os concentrations result in very variable 187Re/188Os values. Rhenium concentrations are positively correlated with S and H2O contents, indicative of redistribution during hydrothermal alteration (Figure 5).
 The following discussion of Re and PGE mobility is fundamentally hampered by the lack of reliable estimates for initial PGE distributions throughout Hole 504B. While Re-Os isotope systematics complement Re and Os concentrations and thus place additional constraints on the mobility of Re and Os, no such isotopic indicators exists for Ir, Pt and Pd. Due to the complex alteration history, the lack of unaltered samples and the fact that MORB glass, for reasons outlined below, may not be a good proxy for the PGE inventory and Re/Os of the upper oceanic crust, we cannot reliably determine primary concentrations for Pt, Pd, and Ir.
5.2. Mobility of Re
 187Re/188Os of MORB are 2–3 orders of magnitude higher than those of DMM and reflect the moderately incompatible behavior of Re and the compatible behavior of Os [e.g., Morgan and Lovering, 1967; Martin, 1991; Hauri and Hart, 1993, 1997; Schiano et al., 1997]. The weighted average 187Re/188Os of ∼350 in 504B and 417/418 basalts is about a factor of two lower than 187Re/188Os of pristine MORB glass from the Atlantic, Indian and Pacific Oceans [∼570; Schiano et al., 1997]. The fact that both depleted MORB (504B [Pedersen and Furnes, 2001]) normal MORB (417/418 [Staudigel et al., 1995]) and an enriched MORB sample from Hole 504B exhibit low 187Re/188Os compared to pristine MORB glasses indicate that mechanisms other than differences in the degree of melting and/or source depletion cause this discrepancy. Unpublished data from a study comparing Os concentrations in MORB glasses and crystalline matrix show that the matrix is enriched in Os relative to pristine glass, sometimes by orders of magnitude, resulting in high 187Re/188Os in glass compared to crystalline matrix [Gannoun et al., 2002; P. Schiano and K. Burton, personal communication, 2003]. Crystalline MORB matrix appears to dominates the Os budget and has to be taken into account when PGE budgets for oceanic crust are considered.
 Several samples with low Os and high Re concentrations are characterized by 187Re/188Os >10,000. This most likely indicates significant addition of Re, rather than loss of Os, because Or/Ir in samples with high 187Re/188Os are indistinguishable from the remainder of the samples. In oxic seawater Re occurs as very soluble perrhenate (ReO4−). Under reducing conditions, however, Re is much less soluble and is efficiently lost to the solid phase [Colodner et al., 1995]. The high Re content of breccias in the deeper part of the crustal section where reducing conditions prevail is thus consistent with addition of hydrogenous Re. This interpretation is supported by the fact that samples with the highest Re concentrations also have high water contents (Figure 5a). Enrichment of Re and loss of Pd and Pt in breccias clearly demonstrate differential mobility of PGE and Re during hydrothermal alteration.
 The positive correlation between Re and S contents, typical of pristine MORB glass [e.g., Schiano et al., 1997] has been significantly modified in 504B basalts by hydrothermal alteration (Figure 5c). Rocks from the oxic part of the volcanic zone are characterized by low Re and S contents, whereas rocks from the deeper non-oxidative portion of 504B either have roughly primary Re and S contents or have experienced Re and S addition. Similar redox-sensitive behavior has been observed by Brügmann et al.  in samples recovered from the TAG hydrothermal system. As the most intense Re enrichment is restricted to rocks with high permeability (breccias) it is difficult to estimate if Re loss caused by oxidation of sulfides in the upper volcanic zone balances Re addition in the deeper, reducing part of hydrothermal convection system in 504B. Most likely oxidative loss is not sufficient to balance reductive addition, and net addition of Re from seawater, indicated by the Re-depleted nature of hydrothermal vent fluids [Colodner et al., 1993], has to be invoked.
 Despite the fact that Re and U are redox-sensitive and less mobile under reducing conditions, enrichment of U is most pronounced in the upper oxic alteration zone (high Fe3+/Fetot; Figure 6a), and weighted average values for magmatic zones (colored stars in Figure 6a) and lithologic units (large open symbols in Figure 6a) show a positive correlation between oxidation state and U concentration. In contrast, the most pronounced Re enrichments are restricted to samples from the more reducing transition zone (orange star in Figure 6b) and breccias in the sheeted dyke complex (green triangles in Figure 6b). Uranium enrichment is probably primarily caused by incorporation into carbonates and Fe-oxyhydroxides [Bach et al., 2003; Kelley et al., 2003], whereas Re is neither significantly associated with carbonates nor with Fe-oxyhydroxides. Though addition of hydrogenous Re can, locally, lead to very high 187Re/188Os [see also Brügmann et al., 1998], data for high-temperature (HT) vent fluids place limits on the overall hydrothermal Re-uptake by altered oceanic crust. Colodner et al.  show that end-member HT vent fluids are depleted in Re, analogous to other redox-sensitive elements in seawater such as U and Mo. Maximum estimates of the HT fluid flux through the oceanic crust (e.g., 1 × 1014 kg yr−1 [Edmond et al., 1979]), seawater Re concentrations of ∼40 pM [Anbar et al., 1992; Colodner et al., 1993], and an annual crust production rate of ∼25 km3 [Reymer and Schubert, 1984] limit average hydrothermal uptake to ∼10 pg Re/g crust, equivalent to ∼1% of the magmatic Re inventory. Even if HT hydrothermal enrichment is restricted to the upper third of the oceanic crust, average enrichment will not exceed a few percent of the primary magmatic Re budget. Though Re concentrations for low-temperature hydrothermal fluids are not available, the lack of significant Re enrichment in the upper volcanic section of 504B indicates that low-temperature ridge-flank alteration does not add significantly to crustal Re inventory. Hydrothermal alteration of oceanic crust is thus a small sink for hydrogenous Re compared to suboxic and anoxic marine sediments [Colodner et al., 1993; Crusius et al., 1996; Morford and Emerson, 1999; Nameroff et al., 2002] and altered oceanic crust will not differ significantly from unaltered oceanic crust with respect to Re.
5.3. Mobility of Os and Ir
 The clear indication of early addition of hydrogenous Os to 504B basalts, inferred from Re-Os isotope systematics, is the strongest argument for hydrothermal mobility of Os in oceanic crust. In this section we attempt to quantify net gain/loss of Os and explore why the large majority of samples have primitive-mantle normalized Os/Ir > 1 (up to ∼15). The three exceptions with 0.5 < Os/Ir < 1 have low Ir concentrations of ≤8 pg/g. The weighted average Os/Ir of ∼2.4 for the entire core is significantly higher than the few existing data for other MORB that are suggestive of chondritic Os/Ir or enrichment of Ir over Os [Hertogen et al., 1980], the opposite of 504B pattern. Three potential explanations for the deviation from chondritic Os/Ir need to be considered. Firstly, the observed ratio could reflect the Os/Ir of the local mantle source. Secondly, melting, melt extraction or differentiation may have modified primary chondritic mantle values. In this case the observed Os/Ir can be considered magmatic in origin. Thirdly, hydrothermal processes may have added Os or extracted Ir from the section of crust studied. We briefly discuss these alternatives below and explain why we believe that the observed inventories of Os and Ir are primarily magmatic in origin.
 Average Os/Ir ∼2.4 significantly exceed values reported for MORB [Hertogen et al., 1980; see also MORB EN026 10D-3 in Table 1 with an average Os/Ir of 0.6], the upper mantle as well as oceanic mantle peridotites variously depleted by melt-extraction [e.g., Snow and Schmidt, 1998; Schmidt et al., 2000; Snow et al., 2000]. Even highly melt-depleted abyssal peridotites from the Garrett Fracture Zone, for instance, apparently have chondritic Os/Ir [Schmidt et al., 1998]. Minor deviations from chondritic LPGE (i.e., Ru, Rh, Pd)/HPGE (i.e., Os, Ir, Pt) have been reported for oceanic mantle-derived rocks [e.g., Pattou et al., 1996]. Significant deviations from chondritic Os/Ir so far have been reported only in arcs (see discussion in section 6). We therefore consider it unlikely that superchondritic Os/Ir in 504B lavas reflect the composition of the local mantle.
 Though it is unlikely that superchondritic Os/Ir are characteristic of the mantle source, several lines of evidence support a magmatic origin for the superchondritic Os/Ir. For instance, the data define an overall positive trend between Ni and magmatic Os concentrations (Figure 7). This trend is similar in shape, but shifted to slightly higher Os concentrations compared to the trend for pristine MORB glasses from the three major ocean basins [Schiano et al., 1997] (gray open symbols in Figure 7). This trend can be explained by fractional crystallization of olivine (± Os-rich trace phases included in olivine [Roy-Barman et al., 1998]) from a primary model melt containing 480 μg Ni/g and 400 pg Os/g, assuming DOs/DNi = 2 [Hart and Ravizza, 1996]. It should be noted that the apparent compatibility of Os in olivine [Hart and Ravizza, 1996] assumed in this calculation may reflect a mixture of clean olivine with low partition coefficient for Os à la Burton et al.  with an Os-rich phase included in the olivine separate analyzed by Hart and Ravizza . It is equally possible that olivine fractionation controls Os only indirectly by inducing sulfide saturation and subsequent sulfide precipitation, as suggested by Burton et al. . Fractionation of sulfide and PGE alloys follows essentially horizontal trajectories in Figure 7 (% sulfide fractionation shown to the left of the “effective” olivine fractionation trend, calculated using a DOs of 48,000 [Roy-Barman et al., 1998]), whereas fractional crystallization of olivine with a very low partition coefficient for Os follows a nearly vertical trend [Burton et al., 2002]. Thus the fractionation line attributed to “effective” olivine in Figure 7 may in reality represent coupled crystallization of olivine and very small amounts (<0.01%) of sulfides or other PGE-rich trace phases. The lack of a trend between S content and primary magmatic Os concentrations indicates that Os is apparently not affected by oxidation of primary sulfides and formation of secondary sulfides in the transition zone (Figure 5d), or that the majority of Os is not associated with sulfides.
 The fact that Os/Ir does not correlate with any other parameter determined in our study (see also companion paper by Bach et al. ) indicates that Os and Ir inventories are controlled by a (or multiple) trace phase(s) with non-chondritic Os/Ir. Though Luguet et al.  found sulfides with Os/Ir as low as 0.4 in an in situ study of sulfides in abyssal peridotites, the majority of the sulfides analyzed had nearly chondritic values. Our preferred interpretation of the superchondritic Os/Ir involves fractionation or assimilation of a trace phase (possibly laurite) with non-chondritic Os/Ir during melting, melt extraction, storage, or emplacement. As we do not know whether neighboring (lateral transport) or deeper portion (vertical transport) of the oceanic crust near Site 504 constitutes a complementary reservoir with subchondritic Os/Ir, we cannot exclude intracrustal fractionation of Os/Ir. The only data for deeper oceanic crust, ODP Hole 735B [Blusztajn et al., 2000], indicate that the top 500 m of the gabbro section has Os/Ir of ∼1.05, within uncertainty identical to chondritic ratios.
 The lack of correlation of Os/Ir with indicators of hydrothermal alteration in Hole 504B (e.g., H2O, U, Rb, Fe3+/Fetot, Sr- and S-isotopes; petrographically determined extent of alteration [Bach et al., 2003, Table 1] correlates positively and linearly with H2O content; r2 = 0.86) indicates that hydrothermal processes are not the primary cause for non-chondritic Os/Ir. This is consistent with the average initial 187Os/188Os of 0.173, far less radiogenic than expected for a ∼1.3:1 mixture of seawater Os (187Os/188Os of 0.85–1.05 for <6.75 Myr old seawater) and primary magmatic Os that is required to create average Os/Ir of ∼2.4 by addition of hydrogenous Os. An Os isotope mass balance restricts the fraction of hydrogenous to less than 6% of the total Os inventory. This is confirmed by the absence of a significant correlation between hydrogenous Os concentration, calculated by isotope mass balance after accounting for ingrowth of 187Os, and water content (Figure 5b). Hydrogenous Os accounts for only 1–2 pg of the total Os. Significant loss of Os (and Ir) from breccias, however, cannot be excluded as a viable explanation for the low average Os (8 pg/g) and Ir (3 pg/g) concentrations in this lithologic unit. As breccias constitute a minor proportion of the section the overall Os and Ir inventories will not be significantly affected by such loss.
 If, as we argue, the observed Os inventory is primarily magmatic, pervasive loss of Ir from upper oceanic crust has to be explored as an explanation for Os/Ir > 1. The lack of correlation between Os/Ir and other indicators of alteration, however, does not support pervasive loss of Ir by hydrothermal processes. Furthermore, the following mass balance illustrates that pervasive Ir loss of a magnitude required for generating average Os/Ir of ∼2.4 significantly affects the marine Ir cycle. The weighted average Os/Ir can be used to predict average Ir concentrations in HT fluids assuming an average loss of 12 pg Ir per gram altered crust (i.e., 21 pg Os minus 9 pg Ir), and ocean crust production rates as well as HT fluid flux identical to those assumed above. The predicted Ir concentration of ∼9 ng Ir per kg HT fluid is 2–3 orders of magnitude higher than Os concentrations in vent fluids [Sharma et al., 2000] and is easily detectable, unless Ir is efficiently scavenged during or shortly after venting. Data on near-vent sedimentary burial fluxes of Ir are scarce, but Ir burial rates in sediments beneath the Rainbow hydrothermal plume [Cave et al., 2003] are at least one order of magnitude too small to account for necessary near-field scavenging. The hypothetical hydrothermal Ir flux is an order magnitude larger than the inferred riverine flux [Anbar et al., 1996], mandating an order of magnitude shorter marine residence time than predicted by Anbar et al.  (2 × 103–2 × 104 yr). In summary, we argue that neither addition of hydrogenous Os nor pervasive extraction of crustal Ir is a reasonable explanation for the observed Os/Ir.
5.4. Mobility of Pt and Pd
 The very restricted range in element concentrations indicative of magmatic fractionation (average MgO = 8.32–9.00 wt.%, average Ni = 83–130 μg/g), large scatter in PGE concentrations in individual rock samples and extensive hydrothermal alteration severely limits our ability to estimate primary Pt and Pd concentrations. In addition, Pd/Pt values in mantle-derived rocks and sulfides are significantly more variable than Os/Ir, as exemplified by in situ analyses of sulfides from abyssal peridotites, that show three orders of magnitude variability in Pd/Pt [Luguet et al., 2001]. A comparison of weighted average PGE concentrations in major lithologic units and magmatic zones is therefore the most suitable indicator of the mobility of elements that do not exhibit isotope variations indicative of PGE sources. Average Pt and Pd concentrations are highest in rocks from the transition zone (Figures 3 and 8), but the high average Pt concentrations are mainly caused by a single sample with high Pt concentrations (504B-85R2, 89–99, 4B; see Figures 8a and 8c). Of all lithologies, massive flows and dikes are characterized by highest average Pt concentrations (325 pg/g), whereas average Pd concentrations are highest in pillow lavas (446 pg/g). Average breccias are characterized by low Pt (131 pg/g) and Pd (173 pg/g), but very high Re concentrations (4874 pg/g). The lack of correlation between Pt and Pd concentrations with H2O and S contents is obvious when the entire data set is considered (Figure 8). However, upon closer inspection it is clear that most breccias do have systematically low Pt concentrations (triangles in Figures 8a and 8c). The few exceptions are mineralized breccias from the transition zone (yellow triangles in Figure 8). If those mineralized breccias were excluded from the population used to calculate weighted averages, average Pt concentrations in breccias were even lower. The observed depletions are significantly larger than those caused by dilution with H2O-rich, PGE-depleted alteration products (see dilution trends in Figures 8a and 8b). Despite the lack of a correlation between Pt concentrations with H2O and S contents we tentatively interpret the low Pt concentration in unmineralized breccias as a result of Pt-loss during hydrothermal alteration, possibly related to alteration of primary PGE containing phases.
 Whether Pd is enriched in pillow lavas (446 pg/g; large open circle in Figures 8b and 8d) or depleted in flows/dikes and breccias (179 and 173 pg/g; large open square and large open triangle in Figures 8b and 8d, respectively) relative to pillow lavas is less clear. Disregarding rocks form the mineralized transition zone, weighted average Pd concentration are lowest in magmatic zones and lithologic units with highest H2O contents (Figure 8d). This observation supports the notion that hydrothermal alteration caused net loss of Pd from the most altered lithology, breccias. In summary, data for Pt and Pd tentatively support the notion that both elements are mobile during hydrothermal alteration. This is consistent with experimental data indicating significantly greater mobility of PPGE (i.e., Rh, Pd, Pt) relative to IPGE (i.e., Ru, Os, Ir) under hydrothermal conditions [e.g., Wood, 1987; Xiong and Wood, 2000]. Given the lack of information on primary magmatic concentrations in 504B basalts the extent of mobility cannot be quantified. The question whether Hole 504B has experienced a net loss or gain, rather than redistribution, of Pt and Pd by hydrothermal processes therefore cannot be answered with our data.
 Most of the above discussion has focused on changes in the Re-Os/PGE systematics caused by hydrothermal alteration. It is therefore important to emphasize that none of the primary characteristics of average MORB with respect to PGE inventory and Re-Os systematics, except the initial 187Os/188Os, have been fundamentally changed by hydrothermal alteration. This conclusion confirms results from a PGE study on altered ocean island basalts from the Réunion hot spot, recovered during ODP Leg 115 [Greenough and Fryer, 1990]. The majority of the 504B basalts are offset by ∼0.05 toward higher 187Os/188Os relative to a 6.75 Myr reference isochron with depleted upper mantle initial ratios, indicative of early addition of minor amounts (<10% of the total crustal inventory) of hydrogenous Os (Figure 2). This seems to be a general feature of altered MORB, as data for the supercomposite DSDP Sites 417/418 also indicate superchondritic initial 187Os/188Os. Such pervasive addition of hydrogenous Os appears to prevent the use of 187Os/188Os in bulk MORB samples as a tracer of mantle heterogeneity. Addition of hydrogenous Re, on average, did not significantly increase the 187Re/188Os. Average 187Re/188Os values are thus primarily determined by magmatic processes, resulting in a strong correlation between 187Os/188Os and the age of upper oceanic crust. Despite clear evidence for mobility of Os, and–to a minor degree - Pt and Pd during hydrothermal alteration, PGE abundance pattern still reflect the primary magmatic fractionation trends [e.g., Hertogen et al., 1980; Barnes et al., 1985; Ravizza and Pyle, 1997; Rehkämper et al., 1999; Blusztajn et al., 2000]. Unless highly altered portions of subducting oceanic crust are separated from less altered portions during subduction and storage in the mantle and evolve as distinct reservoirs, the effects of hydrothermal alteration on the Re-Os/PGE systematics will hardly be noticeable in the geochemistry of melts extracted from mantle reservoirs. This study provides data for the altered upper oceanic crust supporting previous models that indicate steep ingrowth trajectories of 187Os/188Os in recycled oceanic crust [e.g., Martin, 1991, Figure 6; Pegram and Allègre, 1992, Figure 4; Hauri and Hart, 1993, Figures 4 and 5; Roy-Barman and Allègre, 1995, Figures 3, 6; Blusztajn et al., 2000, Figure 5; Becker, 2000, Figure 6]. The supercomposite samples 504B and 417/418 (values in parenthesis) will reach 187Os/188Os values of 3.17 (3.96), 6.14 (6.91), 12.17 (12.87) after 0.5 Gyr, 1 Gyr, and 2 Gyr of closed system evolution, respectively. Osmium concentrations after such prolonged storage times will be 29 (42) pg/g, 37 (56) pg/g, and 53 (78) pg/g, respectively, with 187Os accounting for 30–64 atom% of the total Os inventory.
 Three issues severely limit our present understanding of the petrologic and geochemical alterations of oceanic crust and the effect of subducted altered oceanic crust on the geochemical evolution of the mantle. Firstly, the deeper part of the lower oceanic crust has not been drilled in situ and our understanding of the Re-Os-PGE systematics of deeper oceanic crust is shaped by investigations of sections exposed on land. Data similar to those presented in this paper are needed for typical lower oceanic crust that dominates the crustal PGE budgets. Secondly, the small number of deep drill holes into oceanic crust and the limited recovery may not give us a representative view of altered oceanic crust. This situation can only be improved by drilling complete sections of oceanic crust in the next phase of the Ocean Drilling Program. Thirdly, the mobility of PGE during subduction and transport to the overlying arc mantle and crust is still poorly understood. Existing data for arc rocks indicate that PPGE are significantly more mobile than IPGE [Wood, 1987; Fleet and Wu, 1993; McInnes et al., 1999; Kepezhinskas et al., 2002], and that Re and Os are mobile in Cl-rich oxidizing fluids [Xiong and Wood, 1999, 2000; Widom et al., 2003]. Several studies provide evidence that Os [e.g., Brandon et al., 1996; McInnes et al., 1999] and Re [e.g., Becker, 2000] are extracted from the subducting slab. However, other studies indicate neither significant loss of Os from the subducting slab nor significant addition of slab-derived Os to arc rocks [Becker, 2000; Borg et al., 2000; Woodland et al., 2002; Chesley et al., 2002]. Fractionation of Os from Ir in subduction zones may be controlled by trace phases such as laurite [Kepezhinskas et al., 2002], whereas high Pt/Pd observed in some arc lavas and subarc mantle (xenoliths) may result from melting of previously depleted mantle in the presence of Pt-Fe alloys [e.g., Kepezhinskas and Defant, 2001]. While these studies provide evidence for the relative mobility of PGE in subduction zones, the magnitude of these fluxes are very difficult to constrain. Consequently, we cannot estimate how significantly PGE patterns in the subducting slab are modified by extraction of fluids. For instance, it is currently not known whether Re/Os and Pt/Os fractionations in subduction zones significantly affect the long-term evolution of 187Os/188Os and 186Os/188Os in subducting slabs.
 We thank Lary Ball and Dave Schneider in the WHOI ICP Facility for the excellent support in using the Finnigan MAT ELEMENT for Os isotope, Re and PGE measurements. We also acknowledge the longstanding collaboration with Hubert Staudigel on preparing and investigating composite samples from DSDP Sites 417/418. All samples used in this study were supplied through the assistance of the Deep Sea Drilling Project and Ocean Drilling Program, particularly Dr. Jerry Bode at the WCR. Brian Schroeder was instrumental in preparing sample powders. Stimulating discussion during the weekly Geochemistry Seminar at WHOI shaped many ideas articulated in this manuscript. We also thank Jean-Guy Schilling for splits of his in-house MORB standard EN026 10D-3 and his BHVO-1 split, as well as for sharing analytical results. Thomas Meisel generously shared analytical data for all four rock standards analyzed in this study. Pierre Schiano and Kevin Burton kindly shared information on Os concentrations in pristine MORB glass and crystalline matrix. We very much appreciate the critical comments by associate editor Terry Plank, John Lassiter, and an anonymous reviewer that helped improve the manuscript significantly. We also thank Bill White for his generous editorial handling of the manuscript. Financial support was provided by the U.S. NSF OCE 9811209. This is WHOI contribution 10,894.