A model for osmium isotopic evolution of metallic solids at the core mantle boundary

Transcription

1 Article Volume 12, Number 3 23 March 2011 Q03007, doi: /2010gc ISSN: A model for osmium isotopic evolution of metallic solids at the core mantle boundary Munir Humayun Department of Earth, Ocean, and Atmospheric Science and National High Magnetic Field Laboratory, Florida State University, 1800 East Paul Dirac Drive, Tallahassee, Florida 32310, USA (humayun@magnet.fsu.edu) [1] Some plumes are thought to originate at the core mantle boundary, but geochemical evidence of coremantle interaction is limited to isotopes in samples from Hawaii, Gorgona (89 Ma), and Kostomuksha (2.7 Ga). The isotopes have been explained by physical entrainment of Earth s liquid outer core into mantle plumes. This model has come into conflict with geophysical estimates of the timing of core formation, high pressure experimental determinations of the solid metal liquid metal partition coefficients (D), and the absence of expected 182 W anomalies. A new model is proposed where metallic liquid from the outer core is partially trapped in a compacting cumulate pile of Fe rich nonmetallic precipitates (FeO, FeS, Fe 3 Si, etc.) at the top of the core and undergoes fractional crystallization precipitating solid metal grains, followed by expulsion of the residual metallic liquid back to the outer core. The isotopic composition of the solids and liquids in the cumulate pile is modeled as a function of the residual liquid remaining and the emplacement age using 1 bar D values, with variable amounts of oxygen (0 10 wt %) as the light element. The precipitated solids evolve isotope compositions that match the trends for Hawaii (at an emplacement age of Ga; 5% 10% oxygen) and Gorgona (emplacement age < 1.5 Ga; 0% 5% oxygen). The Fe rich matrix of the cumulate pile dilutes the precipitated solid metal decoupling the Fe/Mn ratio from and W isotopes. The advantages to using precipitated solid metal as the host include a lower platinum group element and Ni content to the mantle source region relative to excess iron, miniscule anomalies in 182 W (<0.1 "), and no effects for Pb isotopes, etc. A gradual thermomechanical erosion of the cumulate pile results in incorporation of this material into the base of the mantle, where mantle plumes subsequently entrain it. Fractional crystallization of metallic liquids within the CMB provides a consistent explanation of both isotope correlations, W isotope systematics, and Fe/Mn evidence for core mantle interaction over the entire Hawaiian source. Components: 15,100 words, 9 figures, 4 tables. Keywords: core mantle boundary; osmium isotopes; iron; Hawaii; mantle plumes; ocean island basalt. Index Terms: 1015 Geochemistry: Composition of the core; 1025 Geochemistry: Composition of the mantle; 1033 Geochemistry: Intra-plate processes (3615, 8415); 1038 Geochemistry: Mantle processes (3621); 1040 Geochemistry: Radiogenic isotope geochemistry. Received 1 July 2010; Revised 31 January 2011; Accepted 8 February 2011; Published 23 March Humayun, M. (2011), A model for osmium isotopic evolution of metallic solids at the core mantle boundary, Geochem. Geophys. Geosyst., 12, Q03007, doi: /2010gc Copyright 2011 by the American Geophysical Union 1 of 23

2 1. Introduction [2] Some mantle plumes are thought to originate at the core mantle boundary (CMB) [e.g., Montelli et al., 2006], raising the possibility of exploring chemical signals of core mantle interaction [Walker et al., 1995; White, 2010]. The discovery of radiogenic 186 / 188 (from the decay of 190 Pt) correlated with 187 / 188 in Hawaiian lavas provided the first substantive evidence of chemical interaction between the liquid outer core and the mantle [Brandon et al., 1998, 1999]. The mechanism proposed to explain this data involved differentiation of the bulk core into a solid inner core and a liquid outer core with fractionation of Pt/ and Re/ ratios analogous to that of magmatic iron meteorites, followed by physical entrainment of liquid outer core into the mantle at the coremantle boundary [Walker et al., 1995; Brandon et al., 1999]. Subsequent critiques of the coremantle interaction hypothesis have centered on this mechanism [Lassiter, 2006; Van Orman et al., 2008; Hayashi et al., 2009], or on the lack of correlated tracers, particularly the absence of evidence for deficits in 182 W that may be expected from simple core mantle interaction involving physical entrainment of the outer core [Scherstén et al., 2004]. Other criticisms have focused on the lack of elemental excesses of Ni, W, and siderophile elements [McDonough, 2003; Sobolev et al., 2005; Ireland et al., 2009a]. The proposed bulk core differentiation can only explain the isotopic data if all the essential parameters: partition coefficients, light element content (by virtue of its impact on partition coefficients), and timing of inner core growth, are given their maximum possible values. For example, given the highest plausible Pt/ ratios that can be expected to be generated by solid metal/liquid metal partitioning, a very early age of the inner core (>3.5 Ga [Walker et al., 1995; Brandon et al., 2003; Puchteletal., 2005]) is required to generate the observed 186 / 188 variations. This age is inconsistent with heat loss models that indicate a younger age [Labrosse et al., 2001; Labrosse, 2003; Lassiter, 2006]. The inner core represents only 5.5% by mass of the bulk core, so if it is to play an important role in increasing the Re/ and Pt/ ratios of the outer core, very large D values are required. Recent experimental studies [Van Orman et al., 2008; Hayashi et al., 2009] have found that D values may be large at low pressures, but decrease with increasing pressure and temperature, so that at the inner core outer core boundary these D values are ineffective at increasing the outer core s Pt/ (and Re/) ratios required as a prelude to radiogenic ingrowth for isotopes. Thus, an explanation in terms of physical entrainment of liquid outer core has had serious difficulties in explaining the Hawaiian data. An alternative physical model involving core mantle interaction to account for the isotopic correlation between 186 / 188 and 187 / 188 in Hawaiian lavas is required. Models for generating the 186 / 188 signal without involving the core have been proposed [Ravizza et al., 2001; Luguet et al., 2008], but a compelling alternative has yet to be developed. [3] The discovery of excess iron in the sources of Hawaiian lavas [Humayun et al., 2004] relative to mid ocean ridge basalts (MORB) and some other ocean island basalts (OIBs) revealed that all Hawaiian lavas analyzed to date are characterized by Fe/Mn ratios that are 10% 20% higher than lavas from Iceland or MORB [e.g., Qin and Humayun, 2008], a finding implicating the role of the core as the largest reservoir of iron on Earth. The 187 / 188 isotopic ratios of Hawaiian lavas exhibit a range of variation from Primitive Upper Mantle (PUM) like (0.129) [Meisel et al., 2001; Becker et al., 2006] to higher ratios (0.145), but the range in Fe/Mn within Hawaiian lavas is limited [Huang et al., 2007], so that a correlation between Fe/Mn and isotope ratios is not observed. This would be very disturbing for the notion of coremantle interaction if the outer core entrainment mechanism were correct. However, critics of the core mantle interaction hypothesis [Scherstén et al., 2004; Sobolev et al., 2005; Lassiter, 2006] have not sufficiently distinguished between criticisms of the proposed physical entrainment mechanism and the fundamental concept of chemical exchange at the core mantle boundary. The coremantle boundary is envisaged as an extremely complex structure [Garnero, 2000], where a multitude of physical and chemical processes may occur [e.g., Walker, 2000; Brandon and Walker, 2005]. [4] Physical entrainment of liquid outer core into the mantle would increase the platinum group element (PGE) content of the mantle by as much as an order of magnitude, which was tested by analyses of komatiite lavas [Puchtel and Humayun, 2000, 2005], a test more difficult to apply for lower degree partial melts like Hawaiian picrites where the PGE contents are buffered by the presence of sulfide in the source [Bennett et al., 2000; Ireland et al., 2009b]. The earliest inklings that the physical entrainment mechanism for core mantle 2of23

3 interaction might be insufficient to explain all isotopic and PGE abundances in komatiites was shown in a series of combined isotopic PGE analyses aimed at this very issue [Puchtel and Humayun, 2000, 2005; Puchteletal., 2004a, 2004b, 2005] which concluded that for these komatiites with radiogenic 186 / 188 there was no resolvable increase in the PGE content of the mantle sources, and that the physical entrainment of outer core liquid was implausible. Puchtel and Humayun [2000] and Humayun et al. [2004] argued for alternative mechanisms involving chemical exchange, including equilibration at the CMB, to transfer isotopic signatures without entraining liquid outer core to the mantle, and Humayun et al. [2004] further proposed mechanisms that decoupled the, W, and Fe tracers. Hayden and Watson [2007] argued that diffusive transport of siderophile elements across the CMB was plausible. However, the issue remains unresolved. [5] While the iron meteorite analogy originally proposed by Walker et al. [1995] remains a potent explanation for the correlated isotope systematics observed in Hawaii [Brandon et al., 1999], Gorgona [Brandon et al., 2003], and Kostomuksha [Puchtel et al., 2005], a more robust model for explaining the role of the Earth s core is required. The concept of a cumulate pile floating on the outer core has been advanced [Buffett et al., 2000] to explain seismic observations of a core rigidity zone [Rost and Revenaugh, 2001]. In this study, I examine the possibility that radiogenic 186 / 188 is generated in such a cumulate pile [Humayun, 2009]. This model has the advantage that it allows the intercumulus liquid to evolve beyond the degree of fractional crystallization (F L ) that characterizes the bulk core (F L 0.95), allowing the D values to be compositionally determined regardless of the pressure temperature conditions. The subsequent transport of the resulting solids and liquids into the mantle is then explored. The links between the various chemical tracers:, W, PGE abundances, and Fe, expected from such a model are also developed. [6] Prior to this discussion, the constraints rigorously imposed on any model for isotopic evolution by the Re Pt isotope systematics of OIBs are reviewed, here. The original outer core entrainment model [Walker et al., 1995; Brandon et al., 1998, 1999; Brandon and Walker, 2005] invoked two component mixing of isotopes to explain the correlation observed in Hawaiian lavas. In the light of subsequent isotopic analyses of chondrites [Brandon et al., 2005], a reinterpretation of this mixing trend is required. The concept of a Pt depletion age analogous to that of the Re depletion age of peridotites [Walker et al., 1989] is introduced here, and it is shown that the normal endmember of the Hawaiian isotopic correlation is depleted in Pt/, and that the entire Hawaiian trend can be explained as a product of core mantle interaction. The implications of fractional crystallization of metallic liquids within the CMB for isotope correlations, W isotope systematics, and Fe/Mn evidence for core mantle interaction, are explored here. 2. The Geochemical Nature of the OIB Reservoir [7] Radiogenic 186 / 188 ratios correlated with 187 / 188 have now been observed in both modern [Brandon et al., 1998, 1999, 2003] and ancient [Puchtel et al., 2005] lavas. However, radiogenic 186 / 188 does not always accompany radiogenic 187 / 188 in OIB lavas, the example being Icelandic picrites that exhibit a similar range of 187 / 188 to Hawaiian lavas, but do not exhibit correlated radiogenic 186 / 188 ratios [Brandon et al., 2007]. The isotopic data for picrites from Hawaii and Iceland, and Cretaceous komatiites from Gorgona, are shown in Figure 1. Also shown in Figure 1 are the 186 / 188 data available for abyssal peridotites [Brandon et al., 2000], an estimate of the bulk core composition based on chondrites, and the present estimate of Primitive Upper Mantle (PUM) that is anchored on 187 / 188 ratios of fertile peridotites [Meisel et al., 2001]. Figure 1 also shows that the Hawaiian picrites extend a range in 186 / 188 that includes both the least radiogenic 186 / 188 ratios and the most radiogenic 186 / 188 ratios observed in the mantle. These data were previously interpreted in terms of a normal mantle component isotopically mixing with an enriched component, inferred to be entrained outer core [Brandon et al., 1999]. [8] A meaningful discussion of these data requires knowledge of the chondritic parameters for the 190 Pt 186 system. An initial measurement of the Allende meteorite obtained 186 / 188 = [Walker et al., 1997], and subsequent measurements of 7 abyssal peridotites found modern mantle 186 / 188 = ± 1, consistent with the Allende data [Brandon et al., 2000]. The subsequent discovery of nucleosynthetic anomalies in extracted from primitive chondrites [Brandon et al., 2005; Yokoyama et al., 2007; Reisberg et al., 2009] 3of23

4 Figure 1. Summary of the currently available data for modern mantle rocks (Hawaii [Brandon et al., 1999], Iceland [Brandon et al., 2007], Gorgona [Brandon et al., 2003], and Kane Fracture Zone abyssal peridotites [Brandon et al., 2000]) compared with chondritic mantle (shaded box) and Primitive Upper Mantle (PUM). The 187 / 188 values for PUM [Meisel et al., 2001] and chondrites [Walker et al., 2002] are from the literature, with 186 / 188 assigned as ± 2 (see text for discussion). Note that Hawaiian picrites with 187 / 188 ratios comparable to PUM exhibit unradiogenic 186 / 188 ratios that require 190 Pt/ 188 ratios a factor of 2 lower than PUM in their source regions. requires revision of the chondritic 186 / 188. A representative 186 / 188 ratio for chondrites, that avoids the isotopic artifacts introduced by the dissolution process, is then obtained by averaging the isotopic compositions of ordinary (St. Marguerite, Forest Vale, Weston, Allegan) and enstatite (Yilmia, Daniel s Kuil) chondrites of metamorphic grade 4 6 [Brandon et al., 2005], which yields 186 / 188 = ± 2. [9] About half of the Hawaiian data in Figure 1 plot below the revised chondritic values, exhibiting a long term chemical depletion of the Pt/ ratio, with PUM like Re/ ratios, an effect that is difficult to conceive of in terms of melt depletion and enrichment of a PUM like mantle source. To better understand the implications of the extreme endmembers of the Hawaiian trend, two concepts are introduced here: (1) the Pt depletion age, analogous to a Re depletion age [Walker et al., 1989], and (2) the Pt/ ratio of the source assuming a singlestage evolution from T = 4.57 Ga. The isotopic evolution of a mantle reservoir that differentiated at time (t) from a mantle of chondritic composition (CHUR) is given by, x ¼ CHUR 190 Pt 188 x 190 Pt 188 CHUR e t 1 ð1þ where x denotes the source region of a sample having nonchondritic 186 / 188 ratios today, i.e., the differentiated reservoir. This equation has two unknowns, the Pt/ ratio of the reservoir or source region, and the time of differentiation, and cannot be rigorously solved for a single isotopic composition. However, two model solutions can be obtained as follows: [10] 1. Assuming that during differentiation of the mantle source all Pt was removed, a Pt depletion age (T 0 ) is calculated by setting ( 190 Pt/ 188 ) x =0 in equation (1) to obtain, 20 1 T 0 ¼ ð 190 PtÞ ln 6B 4@ CHUR 190 Pt x C A ð2þ CHUR Deriving the Pt/ ratio of the mantle source from the Pt/ ratio of erupted lavas is a difficult exercise [Bennett et al., 2000; Puchtel and Humayun, 2000; Puchtel et al., 2004a; Puchtel and Humayun, 2005; Ireland et al., 2009b]. Nonetheless, a Pt/ = 0 is too extreme, since Pt is not as incompatible as Re, implying that Pt depletion ages are minimum ages, and the real time of differentiation was significantly earlier. 4of23

5 [11] 2. The time integrated average ( 190 Pt/ 188 ) ratio for the source can be obtained from equation (1) by setting T = 4.57 Ga to yield, 190 Pt 190 Pt ¼ T CHUR CHUR e T For a sample that has unradiogenic 186 / 188, the ( 190 Pt/ 188 ) T ratio represents the maximum value of the source ( 190 Pt/ 188 ) x ratio after differentiation while, for a sample with radiogenic 186 / 188 relative to CHUR, the ( 190 Pt/ 188 ) T ratio represents the minimum value of the source ( 190 Pt/ 188 ) x ratio after differentiation. Thus, the ( 190 Pt/ 188 ) T ratio captures the minimum range of real source 190 Pt/ 188 variation exhibited by a set of samples, e.g., the Hawaiian picrites, that span the range of 186 / 188 from unradiogenic to radiogenic. In Table 1, values of both T 0 and ( 190 Pt/ 188 ) T are shown for most of the available modern sample 186 / 188 data. To facilitate comparisons between CHUR and ( 190 Pt/ 188 ) T the final column of Table 1 shows the model 190 Pt/ 188 ratios normalized to chondritic values. [12] Examination of the Pt depletion ages for the available 186 / 188 data on Hawaiian lavas in Table 1 reveals that source depletions occurred at least 2.5 Ga ago for the most depleted sample, and the chondrite normalized Pt/ ratio exhibits a range from 0.5 to 1.7, a range of a factor of three (Table 1). Age corrected 186 / 188 ratios for Gorgona komatiites (89 Ma) analyzed for 186 / 188 correspond only to the enriched portion of the Hawaiian trend. The Icelandic picrites generally derive from sources with Pt/ of x PUM. Abyssal peridotites from the Kane Fracture Zone have 186 / 188 ratios compatible with a time integrated Pt/ ratio of x chondritic and a T 0 range of Ga. Measured Pt/ ratios are significantly higher than this in abyssal peridotites indicating recent Pt enrichment [Brandon et al., 2000]. A recent study of PGE abundances in mantle peridotites found the majority of samples to exhibit a Pt/ ratio similar to chondritic values [Becker et al., 2006], supporting the use of chondritic parameters as representative of the mantle s isotopic evolution. [13] Surprisingly, neither Hawaiian picrites nor Gorgona komatiites define trends that originate from or pass through the proposed CHUR value (Figure 1). In contrast, Icelandic picrites could X ð3þ arguably be derived from a source material with CHUR characteristics by variable Re addition, e.g., by mixing with a crustal component [Brandon et al., 2007; Sobolev et al., 2008]. The fact that the Hawaiian trend originates from a Pt depleted source that has high Re/ ratios is difficult to reconcile with a magmatic differentiation process. [14] Given that the core contains the vast majority (>99%) of the Earth s inventories of Re,, and Pt, and that the outer core is a convecting, possibly reactive, fluid interacting with the base of the mantle, it is well worth considering alternative models of core mantle interaction that avoid the pitfalls of an early model [Lassiter, 2006; Van Orman et al., 2008; Hayashi et al., 2009]. Below, I describe a model by which fluid outer core is partially trapped and fractionally crystallized within the base of the mantle or at the top of the core, where it evolves radiogenic that provides a suitable match for from some OIBs. The concept of the model is depicted in Figure Model Parameters [15] Figure 3 presents a simple box model showing the cumulate pile (Figure 2) as a series of small boxes through which the residual metallic liquid passes precipitating solid metal. Residual metallic liquid is returned to the core, while precipitated solid metal is entrained with the FeO rich cumulates into the lower mantle sources of some OIBs. Chondritic (CHUR) parameters used in the model are summarized in Table 2, and were obtained as follows. Modern chondritic 186 / 188 was obtained as described above. The 190 Pt/ 188 ratio ( ± 11) for the model bulk core was taken from the average of 31 analyses of 16 carbonaceous chondrites [Horan et al., 2003; Fischer Gödde et al., 2010] (excluding an Ornans analysis from Horan et al. [2003] that was unusually low in Pt/), and this number is identical within error to average ordinary chondrites. Walker et al. [2002] documented that carbonaceous chondrites have lower 187 / 188 ( ± 6) and 187 Re/ 188 (0.393) ratios than either ordinary ( ± 17, 0.422) or enstatite chondrites, the cause of which is not known. Walker et al. [1995] used Re/ ratios of ordinary chondrite metal for the Earth s core. Humayun and Campbell [2002] observed that metal in ordinary chondrites exhibited higher Re/ ratios than bulk ordinary chondrites due to chemical transport of Re from the matrix during metamorphism. The processes involved in creating the observed Re/ variations in bulk chondrites 5of23

6 Table 1. mium Isotopic Compositions, Pt Depletion Ages (T 0 ), Model 190 Pt/ 188 Ratios for the Mantle Sources Assuming Single StageEvolutionFromT=4.57Ga,andthePUM Normalized Model 190 Pt/ 188 Ratios (PUM norm)ofpicriticor Komatiitic Lavas From Hawaii, Iceland, and Gorgona; of Abyssal Peridotites From the Kane Fracture Zone; and of PUM Sample 187 / / 188 T 0 ( 190 Pt/ 188 ) T PUM norm Hawaii a MKR MKR Kil Kil Kil ML ML LO LO KOH H Iceland b ICE ICE ICE 4a ICE 4b DMF ICE ICE 8a ICE 8b ICE 9a ICE Gorgona c GOR GOR GOR GOR KFZ peridotites d AP AP AP AP AP AP AP PUM e f a Brandon et al. [1999]. b Brandon et al. [2007]. c For Gorgona komatiites, 187 / 188 ratios were age corrected using 187 Re/ 188 and 187 / 188 ratios from the same analyses [Brandon et al., 2003] as follows (abbreviations and their superscripts adopted from original source): GOR (average of CT a and Na peroxide fusion a ), GOR 94 7 (NiS Fusion a ), GOR (CT b ), GOR 521 (CT a ); 186 / 188 ratios were age corrected using single 190 Pt/ 188 ratios [Brandon et al., 2003]. d Brandon et al. [2000]. e Meisel et al. [2001]. f Brandon et al. [2005]. are not understood sufficiently well to ascribe either bulk carbonaceous chondrite or bulk ordinary chondrite Re/ ratios to the bulk Earth. We have chosen bulk ordinary chondrite 187 / 188 ratios (0.1283) and 187 Re/ 188 (0.422) for the bulk core (Table 2). The CI abundances of Re,, and Pt, were updated from Anders and Grevesse [1989] using new CI chondrite data [Walker et al., 2002; Horan et al., 2003; Fischer Gödde et al., 2010]. Abundances of and Pt were calculated for the bulk core using initial CI chondritic abundances given in Table 2, with abundances enhanced in the core by a factor of 6.3 relative to CI chondrites to account for the difference in volatiles between bulk 6of23

7 Figure 2. An illustration depicting a cumulate pile of core sediments (marbled gray) undergoing compaction resulting in expulsion of differentiated metallic liquid back to the core. The rigid and compacted material at the top of the pile is sheared off into the mantle (green) and becomes entrained by rising plumes (yellow). Recycled crust and other mantle heterogeneities are depicted as brown blobs. Modified after Buffettetal. [2000]. partitioning of highly siderophile elements is rather limited [Chabot et al., 2007] and can be ignored for present purposes. The nonmetallic alloying components dominate the partitioning by removing Fe from a metallic state (e.g., Fe) to an ionic state (e.g., FeO), with the highly siderophile elements preferring the metallic solid crystals or the iron domains in the metallic liquid. This effect has been parameterized for S, C, and P [Chabot and Jones, 2003; Chabot et al., 2009]. While the light element abundances of the core are not well known, oxygen is regarded as one of the important candidate elements [e.g., Poirier, 1994; Asahara et al., 2007; Ozawa et al., 2009]. In the treatment below, oxygen has been used as the candidate light element, but the effect of other light elements is generally similar and selection of a different light Earth and CI chondrites ( 1.7), the enhanced refractory element content of the bulk Earth ( 1.2) [McDonough and Sun, 1995], and the enhancement in concentration of the highly siderophile elements into the Earth s core (3.1) [McDonough, 2003]. The Re abundance of the bulk core was calculated from the Re/ ratio used for the core. [16] Since the inner core represents only 5.5% of the mass of the outer core today, the effect of inner core removal on outer core composition has been neglected in the present study. This is immaterial for core models with T > 2.5 Ga, but must be considered in models of the outer core that are younger than 2.5 Ga [Labrosse et al., 2001]. Including this effect would move model compositions toward higher degrees of fractionation by 5.5%. [17] It has long been known that the presence of sulfur, and other alloying agents, plays a more significant role in determining the partition coefficients between solid metal and liquid iron alloy (D SM/LM ) then other intensive variables like temperature and pressure [Jones and Malvin, 1990; Chabot et al., 2003; Chabot and Jones, 2003; Chabot et al., 2008]. For example, the variation of sulfur content from pure Fe to the Fe S eutectic results in an increase in D() from 1.5 to 10 3 [Chabot et al., 2003, 2009]. The effect of Ni, and by extension Co, and other minor metals, on the Figure 3. Box model for isotopic fluxes associated with the cumulate pile model described in the text. For each box, the abundance (ng/g) and the 190 Pt/ 188 ratio normalized to the chondritic value are also shown. The cumulate pile is considered as a multibox system where outer core liquid enters in the first box and precipitates solid metal, and then the residual liquid is advected to the next box, where the process repeats, until the final box where the residual metallic liquid returns to the outer core. Since the boxes are composed of Fe rich cumulate phases (FeO, FeSi, etc.) that may be regarded as free, the concentration of in the box was weighted by the fraction of precipitated metal, taken to be uniform at 2% by mass. The values used for the abundances and 190 Pt/ 188 ratios are based on a model with 5% initial oxygen in the metallic liquid, with step size of F L labeled in each box. The arrows are scaled to indicate the relative magnitudes of the fluxes. Material from any of the boxes can be transported into the overlaying mantle ( 1% 2% by mass fraction) and contribute Fe,, and W to the sources of OIBs. 7of23

8 Table 2. Model Parameters a D 0 b CI Abundance (ppm) Bulk Core (ppm) Re Pt W Re/ / Pt/ / " 182 W 2.0 a Sources are as follows: D 0 and b values are from Chabot and Jones [2003]. CI abundances are modified from Anders and Grevesse [1989] (except Re and W) using recent chondrite data [Walker et al., 2002; Horan et al., 2003; Fischer Gödde et al., 2010] (see text for discussion). Bulk core abundances are modified from McDonough [2003] to reflect new chondrite data and were calculated as 6.3 CI abundances for Pt and. Compositions of average ordinary chondrites used for 187 Re/ 188 and 187 / 188 are from Walker et al. [2002], average of carbonaceous chondrites for 190 Pt/ 188 is from Horan et al. [2003] and Fischer Gödde et al. [2010], average of metamorphosed ordinary and enstatite chondrites for 186 / 188 is from Brandon et al. [2005], and bulk core " 182 W is from Jacobsen [2005]. element would not materially affect the conclusions reached. [18] A generalized expression for the effect of any nonmetal alloying component on siderophile element solid metal liquid metal partitioning is available [Chabot and Jones, 2003]. The application of this expression to oxygen has been made here since oxygen forms FeO analogous to FeS, and is written for oxygen as, 1 X O D ¼ D 0 ð4þ 1 2X O where, D 0 and b are constants for a given siderophile element independent of the alloying light element [Chabot and Jones, 2003], and X O is the mole fraction of oxygen in the metallic liquid. Any element that does not bind effectively with oxygen relative to FeO (i.e., all highly siderophile elements, including Re,, and Pt) will partition into the metallic domains of the liquid, and strongly prefer the metallic solid. The situation is less clear for W, which is only slightly more siderophile than Fe, and less siderophile than Ni, at high temperatures and 1 bar pressures. The effect of oxygen on W may be analogous to the effect of carbon on W, where W carbides become stable and result in W becoming incompatible in Fe C rich liquids [Chabot et al., 2006, 2008]. If W oxides are soluble in the outer core (analogous to FeO solubility), then W may become more incompatible than predicted by using equation (4) with 1 bar experiments. The Ni NiO redox buffer becomes less siderophile at high pressures relative to Fe FeO [Campbell et al., 2009], but the corresponding shift in the W WO 2 buffer is not known. [19] The model for fractional crystallization of outer core liquid metal explicitly calculated the partition coefficients of Re,, and Pt, at 1% intervals of relative change in the fraction of liquid remaining, F L, assuming that oxygen is perfectly incompatible in solid metal. The model also calculated the Re/ and Pt/ ratios after each 1% relative change in F L using the partition coefficients calculated at each step. The isotopic composition of the quenched liquid, the instantaneous solid metal in equilibrium with that liquid, and the accumulated solid metal, were calculated using equation (1) above. In the model, these calculations were performed for a range of light element contents here assumed to be oxygen, the abundance of which was varied from 0% to 10%, and for time of isolation that was varied from 4.5 to 0.5 Ga. Table 3 provides a selected set of results for elemental abundances and isotopic compositions of residual metallic liquids and precipitated metallic grains for three different initial oxygen contents (0, 5, 10 wt %), and three different times of isolation from the outer core (4.5 Ga, 2.5 Ga, 0.5 Ga) as a function of F L. In section 4, the results of the model output are presented. 4. Isotopic Composition of Coexisting Metallic Liquids and Solids [20] Figure 4 shows the present day isotopic composition of a set of quenched liquids, and coexisting solids, isolated at T = 4.5 Ga with oxygen content varied from 0% to 10%. The isotopic compositions of the quenched liquids is not particularly sensitive to the assumed oxygen content at the time of emplacement, but depends importantly on the degree of fractionation (F L ). It should be noted from Figure 4 that the quenched liquids formed by fractionation of bulk core have the right slope but are systematically offset from the Hawaiian picrite data of Brandon et al. [1999]. Advocates of a liquid metal emplacement model could remedy this by shifting the 187 / 188 from to for the modern bulk core implying a Re/ ratio arbitrarily higher by 10% than even the highest chondritic values [Walker et al., 2002; Fischer Gödde et al., 2010]. Exclusive of nucleosynthetic anomalies, shifting the 186 / 188 from 8of23

9 Table 3. Selected Model Output for Models With 0%, 5%, and 10% Oxygen Initially in Metallic Liquid, With Isolation Times of 4.5 Ga, 2.5 Ga, and 0.5 Ga Compositions of Residual Metallic Liquids Compositions of Precipitated Metallic Solids 187 F L O (wt %) (mg/g) Re (mg/g) Pt (mg/g) W (mg/g) / / 188 (mg/g) Re (mg/g) Pt (mg/g) W (mg/g) 187 / / 188 D D Pt Isolation Time = 4.5 Ga Isolation Time = 2.5 Ga of23

10 Table 3. (continued) Compositions of Residual Metallic Liquids Compositions of Precipitated Metallic Solids FL O (wt %) (mg/g) Re (mg/g) Pt (mg/g) W (mg/g) 187 / / 188 (mg/g) Re (mg/g) Pt (mg/g) W (mg/g) 187 / / D DPt Isolation Time = 0.5 Ga of 23

11 Table 3. (continued) Compositions of Residual Metallic Liquids Compositions of Precipitated Metallic Solids 187 F L O (wt %) (mg/g) Re (mg/g) Pt (mg/g) W (mg/g) / / 188 (mg/g) Re (mg/g) Pt (mg/g) W (mg/g) 187 / / 188 D D Pt to is not permitted by the chondritic data. This is because Pt/ ratios for all chondrite groups are nearly identical (C chondrites 1.87 ± 11 (excluding Ornans); O chondrites 1.88 ± 11; E chondrites 1.91 ± 17), unlike for Re/ ratios, and a decrease in Pt/ ratio of 2x is needed to effect the required change in 186 / 188. Since there is no cosmochemical process known to operate in chondrites that can fractionate Pt/ ratios by this much, the bulk Earth, and by extension the bulk core, must have a chondritic Pt/ ratio. Extraction of an inner core is expected to increase the Pt/ ratio, not decrease it. [21] The solid metal that is in equilibrium with the fractionating liquid metal at each step of the calculation (instantaneous solid metal) is offset from the quenched liquid isotopic composition (Figure 4) by about the right amount to explain the Hawaiian trend [Brandon et al., 1999]. Because of the increase in the partition coefficients with increasing oxygen content, the instantaneous solid metal shows more of a dependence on the oxygen content. A model assuming zero oxygen in the emplaced liquid is equivalent to a model with constant partition coefficients. Such a model provides a successful match to the more radiogenic Hawaiian picrites, but cannot match the unradiogenic end member. At higher oxygen contents ( 10%), the instantaneous solid metal curve is within analytical error of all the Hawaiian picrite isotope compositions, except for the most radiogenic compositions in 186 / 188. All isotope compositions can be successfully accounted for if a mixture of solids with trapped liquid is assumed. Such models do not account for the Iceland picrite isotope compositions [Brandon et al., 2007]. Conceivably, the Gorgona Island komatiites exhibit isotope compositions that can be accounted for by trapped outer core liquids, although this is not a necessary interpretation (see below). [22] Figure 5 shows the effect of changing the timing of emplacement of the metallic liquids and solids from 4.5 Ga to 0.5 Ga, each curve calculated for 5% oxygen in the bulk core neglecting the effect of inner core growth. The effect of emplacement age is most notable in the slope of the correlation between 186 / 188 and 187 / 188, which becomes steeper as the time of differentiation is moved closer to the present day (Figure 5). The relative effect of moving the timing of emplacement from 4.5 Ga to 0.5 Ga on the isotope compositions of the quenched liquids is noticeable only for the youngest ages of emplace- 11 of 23

12 Figure 4. Model output for the isotopic composition of in quenched liquid metal and corresponding precipitated solid metal formed by differentiation of bulk outer core (gray box) at T = 4.5 Ga in a cumulate pile at the top of the core or within the CMB. The three types of curves are for initial oxygen contents of 0, 5, and 10 wt %. The effect of increasing the oxygen content on the evolution of the quenched liquids is insignificant, but the impact on the precipitated solid metal is important. The model curves can fit most of the Hawaiian picrite samples [Brandon et al., 1999] but do not fit the more radiogenic end of the Icelandic samples [Brandon et al., 2007]. Because the quenched liquid metal curves are parallel to the precipitated solid metal curves, a fit could be achieved with quenched liquid metal if the bulk core 187 / (today), which is hard to justify and could not fit the least radiogenic samples. Attempting to shift the bulk core 186 / (today) would imply a Pt/ ratio of the bulk core of , about half the chondritic value. This is impossible to argue since chondritic Pt/ ratios are very tightly constrained at ± 12, and the core is the main reservoir of both Pt and. mium isotope data sources are the same as in Figure 1. ment, Ga (Figure 5, top). What is more important for the quenched liquids compositions is that F L is not limited to 0.945, so that even recent emplacement ages give rise to substantial 186 / 188 variations. [23] The isotopic compositions of solids formed in chemical equilibrium with the quenched liquids are shown in Figure 5 (bottom). Unlike the tightly bunched curves for the quenched liquids, this set of curves covers the range from the Hawaiian picrite data [Brandon et al., 1999] to the Gorgona komatiite data [Brandon et al., 2003], with the curve for the most recent emplacement age (T = 0.5 Ga) being within the field of quenched liquid isotopic compositions. The reason for the large offset from the quenched liquids is that the partition coefficients create large differences in Pt/ and Re/ from the bulk core initial composition, which then undergo differential isotopic evolution. For the more recent emplacement ages, the chemical fractionation is identical to that of the T = 4.5 Ga curve, but the differential isotopic evolution is smaller. For example, for the T = 0.5 Ga curve the bulk core has already experienced 4.0 Ga of isotopic evolution before the solid metals are fractionated from it so that these solids cannot evolve unradiogenic 186 / 188 ratios with a Pt depletion age older than 0.5 Ga. It should be noted in Figure 5 (bottom) that the Gorgona komatiites plot very close to the isotopic compositions of solids that evolved from the bulk core at T = 0.5 Ga. It should be appreciated from Figure 4 that any of the curves in Figure 5 (bottom) could be displaced to the left by reducing the oxygen content of the bulk core composition below 5%, or to the right by increasing the bulk core oxygen content above 5%, to obtain better matches to the analytical data. [24] Further, it should be noted that the Icelandic picrites [Brandon et al., 2007] display a wider than expected range of 186 / 188 ratios and plot to the more radiogenic 187 / 188 side of plausible corederived solid metal compositions (Figure 5, bottom). This requires some process of Re addition to the source of the Iceland picrites. Such Re addition 12 of 23

13 Figure 5. Model output for the isotopic composition of in (top) quenched liquid metal and (bottom) corresponding precipitated solid metal formed by differentiation of bulk outer core (gray box) at variable time intervals of 1 Ga (T = Ga) and at an oxygen content of 5%. The curves are color coded by time interval as black (T = 4.5 Ga), blue (T = 3.5 Ga), green (T = 2.5 Ga), orange (T = 1.5 Ga), and red (T = 0.5 Ga). Figure 5 (top) represents the isotopic composition evolved from differentiated liquid compositions, which originate from the chondritic values used for the bulk core, and are tightly bunched. Figure 5 (bottom) represents the isotopic composition evolved from solid metal precipitated from each of the liquids, showing the impact of differentiating the bulk core at different times. Other data are as in Figure 4. observed in Figures 4 and 5. In Figure 6, the fraction of liquid remaining required to evolve Pt/ ratios for an emplacement age of 4.5 Ga sufficiently high to create 186 / 188 ratios of is shown as a function of the initial oxygen content of the metallic liquid (bulk core). The more oxygen assumed in the initial metallic liquid, the less the degree of fractionation required to evolve the liquid to the desired Pt/ ratio. For comparison, a number of estimates of the oxygen content of the core are shown, including those by Anderson and Isaak [2002], Badro et al. [2007], Asahara et al. [2007], and Ozawa et al. [2008]. Anderson and Isaak [2002] constrained the amount of light element required to explain the core density deficit at 3% 7% from seismological constraints and mineral physics experiments, which they regarded to be composed of Si and S. Badro et al. [2007] estimated 5.3% oxygen with 2.8% Si, of which only the oxygen content is shown in Figure 6. Asahara et al. [2007] and Ozawa et al. [2008] estimated the oxygen content of the Earth s outer core from partitioning of FeO between ferropericlase (a.k.a. magnesiowüstite) and liquid iron, and obtained similar estimates of 7% 8% shown as a single broad arrow. It should be appreciated from Figure 6, that the degree of fractionation required to generate the high Pt/ and Re/ ratios is relatively modest and can accommodate a wide range of light element contents. Further, although all the does not necessarily require core related processes since mixing of derived from a chondritic mantle, a depleted mantle, and subducted oceanic crust, provide an alternative [Brandon et al., 2007; Sobolev et al., 2008] The Effect of the Nature and Abundance of the Light Element [25] Since the light element content of the core is not well known, we next consider the effect of variation in oxygen content on the fractionation processes that can generate the Hawaiian trend Figure 6. The effect of increasing initial oxygen in the metallic liquid is that the degree of fractionation required to produce a Pt/ ratio that can evolve to 186 / 188 = over 4.5 Ga is significantly reduced. Oxygen contents inferred to account for the core density deficit are depicted as follows: Anderson and Isaak [2002], rectangle (3% 7%); Badro et al. [2007], thin arrow (5.3% O); Asahara et al. [2007] and Ozawa et al. [2008], thick arrow (7% 8% O). 13 of 23