Siderophile element constraints on the origin of the Moon

1. Introduction

Siderophile (iron-loving) elements provide important leverage for geochemically tracking planetary formation and terrestrial global geodynamic processes through time. The siderophile elements comprise the highly siderophile elements (HSE; including Re, Os, Ir, Ru, Pt, Rh, Au and Pd), a suite of elements characterized by low-pressure metal–silicate distribution coefficients of more than 10⁴, and the moderately siderophile elements (MSE; including W, Co, Ag, Ni, Ge and Mo), with low-pressure metal–silicate partition coefficients that are typically more than 10 but less than 10⁴. Assuming generally chondritic abundances of HSE and MSE in the bulk Earth, mass balance calculations suggest that approximately 98% of the Earth's HSE and approximately 90% of its MSE reside in its metallic core [1,2] (table 1 and figure 1). The absolute and relative abundances of the siderophile elements in the silicate Earth have been affected by both mantle and crustal processes including melting, metasomatism, crystal–liquid fractionation and crustal recycling [7–9]. Many of the siderophile elements are also typically chalcophile, so they are affected by processes that involve the growth and breakdown of sulfides [10,11].

In addition to the utility of interpreting the absolute and relative abundances of siderophile elements in planetary mantle materials and derivative melts, the MSE and HSE include several radiogenic isotope systems that can be used to examine global and planetary processes. For example, the short-lived ¹⁸²Hf–¹⁸²W isotope system (¹⁸²Hf→¹⁸²W+β⁻, where t_1/2=8.9 Myr) can be used to trace the fractionation of the lithophile and incompatible trace element Hf from the moderately siderophile and highly incompatible trace element W [12,13]. This short-lived system records the effects of processes that generally occurred within the first approximately 50 Ma of Solar System history. The observation that the ¹⁸²W/¹⁸⁴W of terrestrial rocks is approximately 200 ppm more radiogenic (enriched in ¹⁸²W) than chondrites led to the interpretation of early formation of the Earth's core [14–18]. The isotopic difference between the mantle and chondrites, together with mass balance constraints, also implies that the Earth's core is a W-rich reservoir with ¹⁸²W/¹⁸⁴W that is approximately 220 ppm lower than terrestrial silicates. The great differences in W isotopic compositions and abundances between large W-bearing reservoirs, such as the mantle, core and impactors with bulk chondritic compositions, mean the system can also be used to assess mixing between these reservoirs at any time in Solar System history.

The HSE include the long-lived ¹⁸⁷Re–¹⁸⁷Os and ¹⁹⁰Pt–¹⁸⁶Os isotopic systems (¹⁸⁷Re→¹⁸⁷Os+β⁻, where t_1/2=42 Ga; ¹⁹⁰Pt→¹⁸⁶Os+α, where t_1/2≈460 Ga). The Re–Os system has been used, among other applications, as a highly sensitive tracer of the long-term Re/Os of the bulk silicate Earth (BSE) [19]. Projections of mantle data to primitive, undepleted compositions provide strong evidence that the Re/Os of the BSE is within the very limited range of ratios defined by chondritic meteorites [3,20]. As discussed below, this provides an important constraint on the origin of HSE in the mantle. The Pt–Os system displays little variation among most terrestrial samples [21]. This reflects the fact that ¹⁹⁰Pt is a minor isotope of Pt and decays very slowly. Only large fractionations in Pt/Os, such as resulting from liquid metal–solid metal fractionation, together with long periods of time can result in measurable variations in ¹⁸⁶Os/¹⁸⁸Os [22].

2. The origin of siderophile elements in the terrestrial mantle

It has been known for more than 40 years that the abundances of the HSE and some MSE present in the mantle are substantially higher than would be expected for metal–silicate segregation at relatively shallow depths [7,23,24]. To explain this, a variety of processes including high-pressure and high-temperature metal–silicate equilibration, incomplete core segregation and late accretion have been called upon by different authors to account for the observed abundances [25]. Although still heavily debated, different processes are commonly favoured today to account for the mantle abundances of the MSE versus the HSE. Some early studies predicted that metal–silicate partition coefficients for siderophile elements at high pressures and temperatures might be greatly diminished, compared with partition coefficients measured at relatively low pressures and temperatures [23,26]. These authors speculated that the higher than expected abundances of many of the MSE (and HSE) in the mantle resulted from metal–silicate equilibration under high-pressure and high-temperature conditions. Such conditions might be expected to obtain at the base of a global magma ocean. Subsequent experimental studies, focused on MSE, have confirmed that the abundances of most MSE can be accounted for by metal–silicate partitioning at pressures of 40–60 GPa [27–29]. Although a number of issues remain, such as the mechanisms of metal–silicate equilibration [30], and the effects of multiple generations of magma oceans or seas where this took place [31,32], much of the geochemical community currently appears to accept the general outline of this hypothesis.

In contrast to the MSE, high-pressure and high-temperature metal–silicate segregation does not appear to well explain the high absolute and generally chondritic relative abundances of all HSE in the mantle [33,34]. Consequently, it has been argued that the present inventory of the HSE, as well as a small portion of the MSE in the BSE, were added as a ‘late veneer’ of late accreted materials, subsequent to the termination of core segregation [35,36]. Late stage accretion may have added materials with bulk chondritic compositions comprising as much as approximately 0.8% of the total mass of the Earth [37]. This hypothesis predicts the imposition of chondritic relative abundances of HSE onto the mantle and is consistent with the chondritic ¹⁸⁷Os/¹⁸⁸Os and ¹⁸⁶Os/¹⁸⁸Os projected for the BSE [19,38,39]. These ratios require precisely chondritic Re/Os and, to a lesser extent, Pt/Os over the 4.5 Gyr history of the Earth. As with the MSE and magma ocean hypothesis, some issues remain with the late accretion hypothesis. For example, current estimates of HSE abundances in the BSE [40,41] suggest modest enrichments in Ru/Ir and Pd/Ir above chondritic values, yet the delivery of HSE with chondritic relative abundances to the mantle has been a major assumption of the late accretion hypothesis. Also, the mismatch in estimated mantle abundances between the Earth and Moon (see below) requires some planetary dynamical gymnastics to explain [42].

Although the origins of the MSE and HSE were collectively discussed by many of the earlier studies that addressed siderophile element excesses in the mantle [25,35,43,44], the origins of the MSE and HSE have only rarely been collectively discussed since. This mainly reflects the greater emphasis recent studies have given to the study of specific MSE or HSE, not because of perceived major flaws in combining hypotheses. Mass balance permits late accretionary delivery of essentially 100% of the observed HSE to the mantle, while adding no more than approximately 10% of most MSE. Thus, the mantle abundances of the MSE could primarily have been set by high pressure and temperature metal–silicate partitioning, as may have occurred as a consequence of global melting resulting from a giant impact [45]. The HSE abundances in the mantle could have been set by late accretion subsequent to that event.

Appealing to a mixed model for the origin of the MSE and HSE, however, is accompanied by some potential pitfalls. Although metal–silicate distribution coefficients for some HSE, such as Ir, do not decrease to values approaching those necessary to account for mantle abundances, even at very high pressures, the distribution coefficients of some other HSE, such as Pt and Pd, may decrease to values that would result in excess abundances of these elements in the mantle and lead to some non-chondritic HSE ratios [34,46]. This may be consistent with the apparent, slightly suprachondritic Pd/Ir of the BSE. However, current partitioning data at highest pressures appear to be inconsistent with the slightly suprachondritic Ru/Ir, and chondritic Pt/Ir estimates for the BSE.

3. New insights gained by combining ¹⁸²W and highly siderophile element data

The possible juxtaposition of two different mechanisms for establishing the abundances of MSE and HSE in the terrestrial mantle creates some opportunities to combine MSE and HSE data to provide new insights to Earth accretionary processes and the origin of the Moon. One important consequence of the proportionally different contributions of the MSE and HSE by late accretion is the fact that the ¹⁸²W/¹⁸⁴W of the mantle would have been lowered, on average by approximately 10–30 ppm, relative to pre-late accretionary mantle. The level of uncertainty in the effect of late accretion on W isotopic composition stems from uncertainty in the mass of materials with bulk chondritic composition required to generate the HSE abundances estimated for the BSE. This, in turn, stems from the fact that HSE abundances in bulk chondrites vary by about a factor of three [4,47]. Volatile-rich materials, such as some carbonaceous chondrites, tend to be characterized by lower than average concentrations of HSE. If late accreted materials were dominated by such materials, the addition of approximately 0.8 wt% of the Earth's total mass by late accretion would be required to account for present HSE abundances. Metal-rich chondrites, such as some enstatite chondrites, tend to have higher than average concentrations of HSE, so if late accretion was dominated by the addition of such materials, as little as 0.3 wt% would be required.

Based on the mass balance arguments presented above, if pre-late accretionary mantle or crustal materials survive in the terrestrial rock record, they should be characterized by ¹⁸²W enrichments of approximately 10–30 ppm, coupled with substantial HSE depletions, relative to the modern mantle. Recently, analytical techniques have advanced sufficiently that such small W isotopic variations can be resolved [48,49]. Willbold et al. [48] reported approximately 13±4 ppm positive ¹⁸²W anomalies in 3.8 Ga supracrustal rocks from Isua, Greenland (figure 2). They concluded that the Isua rocks were derived from mantle domains that formed prior to and remained isolated from late accretion until production of this crust. They speculated that most HSE and a modest proportion of the MSE (including W) were added to the mantle after the mantle precursors to the Isua lithologies formed. Tungsten isotopic anomalies resulting from this type of process are best termed exogenous, as they would have been generated by accretionary additions to the Earth long after ¹⁸²Hf was no longer extant. Their model predicts low HSE abundances in mantle precursors, and presumably the rocks with the ¹⁸²W anomalies as well, but corresponding HSE abundances were not measured by that study. Nevertheless, consistent with the hypothesis of [48], several other studies have provided elemental evidence for the poorly mixed state of late accreted materials in early Earth history. For example, based on gradually increasing concentrations of Pt, estimated for the parental melts of 3.6–2.9 Ga komatiites, Maier et al. [51] argued for slow, progressive downward mixing of a HSE-rich late veneer into deep mantle source regions of the komatiites.

Touboul et al. [50] reported similar, uniform approximately 15±5 ppm positive ¹⁸²W anomalies in the 2.8 Ga Kostomuksha komatiites (Russian, Karelia) (figure 2). These rocks had previously been described as typical, late Archaean, Al-undepleted komatiites [52]. They were targeted for W isotopic study because a prior study determined that they were generated by melting of a mantle source that was enriched in ¹⁸⁶Os/¹⁸⁸Os and ¹⁸⁷Os/¹⁸⁸Os, compared to chondritic Os evolution models for the mantle [52]. That study interpreted the Os isotopic enrichments as evidence for derivation of the Os from the outer core, given that some fractionation models for inner core crystallization suggest the outer core may have evolved with suprachondritic Re/Os and Pt/Os [22]. The observation that the W elemental systematics of these komatiites are broadly consistent with a magmatic origin, coupled with a lack of evidence for crustal interaction in the komatiitic melt, suggests their W isotopic composition must reflect that of their mantle source. Given that the ¹⁸²W/¹⁸⁴W of the outer core is likely strongly depleted relative to the BSE (−220 ppm), the W isotopic results indicate that the W, and probably also the Os, isotopic anomalies were not the result of core–mantle interaction. Further, in comparison to the HSE-depleted mantle source assumed for Isua, Puchtel & Humayun [53] reported that the projected Kostomuksha mantle source had essentially ‘normal’ abundances of HSE. Consequently, unlike the Isua supracrustal rocks, these komatiites could not have been derived from preserved, pre-late accretionary mantle. Hence, an exogenous model to explain this anomaly fails.

Touboul et al. [50] concluded that the elevated ¹⁸²W/¹⁸⁴W present in these komatiites, as well as the corresponding enrichments in ¹⁸⁶Os/¹⁸⁸Os and ¹⁸⁷Os/¹⁸⁸Os, are best explained as resulting from the endogenous ingrowth of ¹⁸²W in a high Hf/W, Pt/Os and Re/Os mantle reservoir that formed very early in Earth history. This reservoir could have resulted from metal–silicate fractionation in a basal magma ocean, or via fractionation in an early formed, upper mantle magma ocean [50]. Given the short-lived nature of ¹⁸²Hf, the event that led to fractionated Hf/W in this reservoir likely occurred within 30 Myr after Solar System formation. Thus, it may have formed at a time when the Earth was considerably less massive than at present. Given the current estimate of the age of the Moon to be more than or equal to 52 Myr after Solar System formation [54], this reservoir would have formed well before the putative giant impact that led to formation of the Moon.

4. The Earth–Moon highly siderophile element conundrum

The present HSE abundances in the lunar mantle must reflect either the combined outcome of lunar coalescence and differentiation, and/or late accretionary additions to the Moon prior to complete formation of the lunar crust (which would have blocked further additions of HSE to the mantle). Unfortunately, unlike for Earth, we do not have bona fide samples of the lunar mantle, and must currently be content to work with derivative materials, primarily basaltic and picritic materials generated by partial melting of the lunar mantle. The HSE abundances present in planetary mantles have commonly been estimated from the HSE contents of derivative melts via plots of a HSE concentration versus a major element whose abundance may reflect degree of melting, such as Mg [44,55,56]. A particularly good HSE for this purpose is Pt, because at least for Earth, it is little fractionated between mantle and crust.

Relative to terrestrial rocks with comparable MgO contents, the Pt contents of lunar basalts and volcanic glasses are generally lower by more than a factor of approximately 20 [57,58] (figure 3). Consequently, several prior studies have concluded that the lunar mantle sources of basalts, as well as volcanic orange and green glasses, were depleted in the HSE by at least a factor of 20 relative to the terrestrial mantle [57–59]. This is somewhat surprising, as calculations that consider planetary cross sections and gravitational focusing suggest that the lunar and terrestrial mantles should have similar HSE concentrations if both bodies were exposed to equivalent late accretionary fluxes [36,57,60,61].

The large offset in HSE concentrations between terrestrial and lunar mantles could reflect several different processes. For example, it may suggest that the lunar core formed subsequent to late accretion and that the ‘missing’ HSE were extracted into the small lunar core. This possibility is not currently supported by the limited Os isotopic data for lunar basalts. The efficient removal of HSE from the mantle during lunar core segregation would likely have led to the generation of high Re/Os in the silicate mantle [28,62], yet the limited Os isotopic data for lunar basaltic rocks indicate derivation from sources with chondritic ¹⁸⁷Os/¹⁸⁸Os at their times of crystallization [57,58]. This means that the lunar mantle sources of these rocks evolved for 0.5–1 Gyr with chondritic Re/Os from the time of core formation until the time of mantle melting to generate the basalts. During this interval of time, the ¹⁸⁷Os/¹⁸⁸Os of the mantle would likely have evolved to ratios considerably higher than the chondritic initial ratios measured.

It is also possible the ‘missing’ HSE reside in the lunar crust, which is both ancient and thick, and may have served as an effective barrier against HSE addition to the mantle from late stage impactors [36,63]. However, unless the late accretionary impactors were largely metal, the bulk composition of the lunar crust appears to be insufficiently similar to chondritic to account for the necessary late accretionary flux [57].

Finally, the presumed low HSE abundances in the lunar mantle could also mean that either the Moon received proportionally much less late accreted material compared with the Earth, or that the late accretionary clock for the Earth began prior to that for the Moon. These latter two possibilities are further considered in §5.

It is also possible that current estimates for HSE abundances in the lunar mantle are significantly in error, and actual HSE abundances are considerably higher. For example, presumptions regarding the limited fractionation of Pt between the lunar mantle and derivative volcanic rocks may not be valid. It has been shown that mantle fO₂ and fS₂ on the Earth have moderately strong effects on the partitioning of Os and Ir between mantle and basaltic melts [64]. Differences in mantle fO₂ and fS₂ between the terrestrial and lunar mantles may, therefore, have resulted in widely different partitioning characteristics for Pt. Partitioning data for Pt relevant to conditions of lunar mantle melting do not yet exist.

These issues highlight the need for continued experimental study of HSE partitioning between possible residues and melts for conditions relevant to the Moon. They also highlight the need to pursue the sampling of lunar mantle by future sample return missions. For the purposes here, it will be presumed that HSE abundances in the lunar mantle are 20 or more times lower than in the terrestrial mantle.

5. Ramifications of existing observations

One important ramification of the recent combined ¹⁸²W–HSE data is that the long-term preservation of a very early formed, ¹⁸²W-enriched endogenous reservoir in the terrestrial mantle implies that the putative Moon-forming giant impact did not result in complete global melting and/or mixing of mantle. Noble gas evidence for the survival of pre-giant impact terrestrial reservoirs [65] is also consistent with this conclusion. Some recent models for Moon formation [66] do not require complete global melting, and may not even permit it, so this is not necessarily a problem for giant impact models.

A second major ramification of these observations is that, if the interpretations of Willbold et al. [48] and Maier et al. [51] for slow mixing of late accreted materials into the mantle are correct, they lend considerable support to the late accretion hypothesis. They also provide an indication that the upper mantle remained poorly mixed with respect to siderophile elements for tens to hundreds of millions of years after the Moon-forming event. These studies collectively predict that early Archaean komatiites that are characterized by low Pt should be accompanied by enrichments in ¹⁸²W. The 3.5 Ga Komati komatiites of southern Africa are the only low Pt, early Archaean komatiites yet examined for W isotopic composition. Touboul et al. [54] reported an ¹⁸²W/¹⁸⁴W for these rocks that overlaps within uncertainties of modern rocks (figure 2), but with a hint of minor enrichment. If early Archaean komatiites with low presumed mantle source abundances of HSE are found not to be enriched in ¹⁸²W, then alternative explanations for the HSE depletion will have to be developed.

A highly debated tenet of the late accretion hypothesis for the Earth is that metal–silicate equilibration resulting from large impacts led to periodic drawdowns in siderophile element abundances, especially the HSE [31,37]. This leads to the question of whether or not the putative Moon-forming giant impact was a final global clearinghouse event that stripped previously accumulated HSE from the terrestrial mantle. The efficiency of HSE drawdown would depend on how metal from the impactor chemically interacted with the mantle, and ultimately merged with the terrestrial core. Rapid merging of cores would allow little metal–silicate equilibration, and presumably little drawdown of HSE. Conversely, partial to complete break-up of the impactor core might lead to addition of sufficiently small metal droplets such that the sinking of the droplets through a magma ocean, ponding of metal at the base of a magma ocean or even downward percolation of metal through solid silicate mantle, would enable metal–silicate equilibration and potentially highly efficient drawdown of HSE abundances from the mantle [30,67].

If HSE drawdown occurred as a result of the putative Moon-forming giant impact, then the late accretionary clocks for the Earth and Moon were likely started at the same time, and the apparent HSE difference between the two bodies would most likely reflect proportionally different extents of late accretion. One dynamical process that can potentially explain the differences in HSE abundances is stochastic late accretion [42]. Stochastic late accretion is the process whereby most of the late accretionary mass was added to Earth (and Mars) by a small number of relatively large, Pluto-mass bodies (approx. 0.2 Earth mass). For this hypothesis, it is presumed that the smaller cross section of the Moon allowed it to statistically elude the few large impactors that struck Earth. If this interpretation is correct, late accretion was not dominated by a gentle rain of small impactors, as is normally implied by the concept of a late veneer, but rather a harsh peppering of the Earth by collisions with a limited number of moderate-sized impactors. Late accretion of this type would likely have resulted in the formation of periodic magma lakes and seas. Such a scenario is not well described by the term ‘late veneer’.

If, on the other hand, the putative giant impact that created the Moon did not drawdown HSE abundances in the mantle, such as may have occurred by rapid merging of cores without metal–silicate equilibration, then the difference in HSE abundances between terrestrial and lunar mantles could simply reflect a longer accumulation period for the Earth. If this scenario can be proved, then it would suggest that late accretion had waned considerably by the time of the formation of the Moon. This important issue can potentially be settled by comparing the ¹⁸²W/¹⁸⁴W of lunar and terrestrial mantles.

6. The case for reinvestigating the W isotopic composition of the lunar mantle

As noted above, it has been proposed that late accretion of materials with chondritic bulk compositions added approximately 0.4–0.8 wt% of mass to the Earth, and approximately 0.05 wt% mass to the Moon. If this is true, and if late accretion to the mantles of the Earth and Moon largely post-dated the formation of the Moon, then mass balance calculations predict that the W isotopic composition of the lunar mantle should be more radiogenic than the Earth by approximately 10–30 ppm, as the higher proportion of late accreted materials added to the Earth would have lowered the ¹⁸²W/¹⁸⁴W of the terrestrial mantle towards the isotopic composition of chondrites (approx. −200 ppm relative to present Earth) more than for the lunar mantle. The mass balance arguments to arrive at this conclusion are very similar to those used by Willbold et al. [48] to account for the W isotopic compositions of Isua rocks.

Accurate and precise measurement of the ¹⁸²W/¹⁸⁴W of the lunar mantle is critical to test this prediction. Because of the effects of cosmic rays on ¹⁸²W, particularly production resulting from cosmic ray interactions with ¹⁸¹Ta [68], most studies of the W isotopic composition of the lunar mantle have focused on Ta-free metal, separated from impact melt rocks and basalts [69,70]. Most recently, Touboul et al. [54] analysed separated metal from a suite of diverse lunar basalts and potassium, rare earth element and phosphorus (KREEP)-rich impact melt rocks. That study reported isotopic homogeneity among the different metals measured and reported a group average ¹⁸²W/¹⁸⁴W that is 9±10 ppm higher than terrestrial standards (based on the 2 standard errors of 15 measurements; figure 4). Touboul et al. [54] interpreted the W data to mean that the Moon formed more than or equal to 52 Myr after formation of the Solar System, and also that the lunar magma ocean crystallized after ¹⁸²Hf was no longer extant, more than or equal to 60 Myr after Solar System formation.

The results from Touboul et al. [54] provided a hint that the lunar mantle may be enriched in ¹⁸²W, relative to the modern terrestrial mantle, and within the predicted level of enrichment (figure 4). However, with uncertainties for individual measurements exceeding 10 ppm in that study, precision was insufficient to definitively test the prediction. Now that precision of individual measurements can be less than 5 ppm, it is necessary to re-measure the isotopic compositions of W in metals extracted from a similar set of lunar volcanic rocks as examined by Touboul et al. [54]. A uniform positive offset of 10–30 ppm, relative to the present terrestrial mantle, could be interpreted as evidence for the hypothesized proportional disparity in late accretion to the Earth and Moon. It will also be important to re-assess whether there is resolvable W isotopic heterogeneity among the mantle reservoirs sampled by lunar basalt. Small, but analytically resolved heterogeneity could be interpreted as evidence for uneven mixing of late accreted materials into different portions of the lunar mantle, or for very minor, variable ingrowth of ¹⁸²W while ¹⁸²Hf was still live. These possibilities may be discriminated based on the nature of the hypothetical variations and how they may correlate with estimates of HSE abundances, or projections for Hf/W fractionations in the sources of the diverse lunar volcanic rock suite.

It is also possible that future, high-precision measurements of lunar metals will reveal that there is no W isotopic difference between the Earth and Moon above the 5 ppm level. If this is the case, it may be cause for a reconsideration of the late accretion hypothesis altogether.

Funding statement

This study was supported by NSF grant no. 1160728, and NASA grant nos. NNX09AJ20A and NNA09DB33A. These sources of support are also gratefully acknowledged.