Highly siderophile element depletion in the Moon

Abstract

Coupled ¹⁸⁷Os/¹⁸⁸Os and highly siderophile element (HSE: Os, Ir, Ru, Pt, Pd, Re) abundance data are reported for Apollo 12 (12005, 12009, 12019, 12022, 12038, 12039, 12040), Apollo 15 (15555) and Apollo 17 (70135) mare basalts, along with mare basalt meteorites La Paz icefield (LAP) 04841 and Miller Range (MIL) 05035. The most magnesian samples have chondrite-relative HSE abundances and chondritic measured and calculated initial ¹⁸⁷Os/¹⁸⁸Os, with mare basalts having consistently low HSE abundances at ~2 ×10⁻⁵ to 2 ×10⁻⁷ the chondritic abundance. The lower and more fractionated HSE compositions of evolved mare basalts can be reproduced with bulk-partition coefficients of ~2 for Os, Ir, Ru, Pt and Pd and ~1.5 for Re. Lunar mare basalt bulk-partition coefficients are probably higher than for terrestrial melts as a result of more reducing conditions, leading to increased HSE compatibility. The chondritic-relative abundances and chondritic ¹⁸⁷Os/¹⁸⁸Os of the most primitive high-MgO mare basalts cannot be explained through regolith contamination during emplacement at the lunar surface. Instead, mare basalt compositions can be modelled as representing ~5–11% partial melting of metal-free sources with low Os, Ir, Ru, Pd (~0.1 ng g⁻¹), Pt (~0.2 ng g⁻¹) Re (~0.01 ng g⁻¹) and S, with sulphide-melt partitioning between 1000 and 10000.

Apollo 12 olivine-, pigeonite- and ilmenite normative mare basalts define an imprecise ¹⁸⁷Re-¹⁸⁷Os age of 3.0 ±0.6 Ga. This age is within uncertainty of ¹⁴⁷Sm-¹⁴³Nd ages for the same samples and the isochron yields an initial ¹⁸⁷Os/¹⁸⁸Os of 0.109 ±0.008. The Os isotopic composition of the Apollo 12 source indicates that the lunar mantle source of these rocks evolved with Re/Os within ~10% of chondrite meteorites from the time that the mantle source became a system closed to siderophile additions to the time that the basalts erupted. The similarity in absolute HSE abundances between mare basalts from the Apollo 12, 15 and 17 sites, and from unknown regions of the Moon (La Paz mare basalts, MIL 05035) indicates relatively homogeneous and low HSE abundances within the lunar interior. Low absolute HSE abundances and chondritic Re/Os of mare basalts are consistent with ~0.02% late accretion addition that was added prior to the formation of the lunar crust and significantly prior to cessation of lunar mantle differentiation (>4.4 Ga) to enable efficient mixing and homogenization. The HSE abundances are also consistent with the observed, small ¹⁸²W excess (20 ppm) in the bulk silicate Moon relative to the bulk silicate Earth.

1. Introduction

Derivative melts of the Moon’s interior, sampled as mare basalts and pyroclastic glass beads represent the only available materials for establishing lunar mantle composition. Pyroclastic glasses are believed to represent primitive (13–20 wt.% MgO), high-degree partial melts sourced from the deep mantle (>400 km; Shearer & Papike, 1993), whereas mare basalts have witnessed more complex petrogenetic histories (e.g., Neal & Taylor, 1992; Longhi, 2006). In terrestrial igneous systems, high-MgO melts, such as picrites and komatiites have highly siderophile element (HSE: Os, Ir, Ru, Rh, Pt, Pd, Re, Au) abundances most similar to their mantle sources because they result from high degree partial melting, with (nearly) all sulphide host minerals in the parent mantle being incorporated into the melt (e.g., Barnes et al., 1985; Rehkamper et al., 1999). Early studies showed that the Ir contents of picritic high-Ti orange and low-Ti green glasses were only offset from the terrestrial array by a factor of two to four (Ganapathy et al., 1973; Morgan & Wandless, 1979), leading to the conclusion that HSE abundances in the lunar mantle, sampled by picritic glasses, are similar to those in the terrestrial mantle (Ringwood, 1992). However, significant doubt has been cast on such high HSE abundances in the lunar interior. Pristine pyroclastic picritic glasses contain elevated volatile and siderophile abundances that are surface-correlated (e.g., Chou et al., 1975; Krahenbuhl, 1980), and surface-correlated elevated HSE abundances have been ascribed to meteoritic contamination on the outer surfaces of the glass beads after their formation and emplacement (Walker et al., 2004).

In contrast to their exteriors, the interiors of the lunar pyroclastic glass beads have similar HSE abundances to mare basalts from the Apollo 15 and 17 sites (Day et al., 2007). Utilizing the relationship of HSE abundance variations with indices of fractionation (e.g., MgO) in mare basalts (e.g., Warren et al., 1999) it is possible to perform regression analysis to assess the lunar mantle composition. The estimated mantle concentrations obtained (0.01 ng g⁻¹ Re, 0.1 ng g⁻¹ Os, Ir, Ru, Pd and 0.2 ng g⁻¹ Pt) are ~40 times lower than HSE abundances estimated for Earth’s primitive mantle (Becker et al., 2006). An interpretation of low concentrations of the HSE in the lunar mantle is supported by ultra-low HSE abundances measured in pristine lunar crustal rocks (Warren et al., 1980; Day et al., 2010), consistent with a bulk silicate Moon with significantly lower HSE abundances than for bulk silicate Earth. These results have important implications, suggesting far more limited late-accretion to the Moon than for Earth.

Three aspects of lunar mare basalt genesis currently limit our ability to more accurately estimate lunar mantle HSE composition in order to elucidate early differentiation, and the timing and extent of late-accretion to the Moon. First, contamination of mare basalt lavas by even minor amounts (<0.01% by mass) of HSE-rich impactor material residing on the lunar surface would lead to increased HSE contents, flatter chondrite-normalized HSE patterns, and chondritic ¹⁸⁷Os/¹⁸⁸Os, potentially leading to an overestimate of lunar mantle HSE abundances. Second, residual metal in the source of mare basalts after melt-extraction could be problematic to understanding lunar mantle composition. At low fO₂ conditions in the lunar mantle, residual metal could lead to strongly fractionated HSE inter-element ratios in mantle-derived melts, and significant ¹⁸⁷Os/¹⁸⁸Os variation would be established from the time of mantle source crystallization (>4.4 Ga) to the eruption of mare basalts (~3.8–3.0 Ga). Third, while there is clear evidence for mineralogical heterogeneity and fractionation of stable isotopes and long-lived radiogenic lithophile isotope systems in mare basalt source regions (e.g., Warren & Taylor, 2014), the degree to which the lunar mantle is heterogeneous with respect to the HSE is not well constrained. Understanding the distribution of the HSE is important since it is possible that magma ocean differentiation, possibly accompanied by late accretion, led to strong HSE fractionations in the lunar mantle. In turn, these observations can affect interpretation of the isotopic composition of other siderophile elements, such as W, measured in mantle-derived lunar materials (Walker, 2014). In this contribution, new data are reported for Apollo 12, 15 and 17 mare basalts, as well as two lunar meteorites, to improve understanding of HSE abundances in the lunar interior.

2. Samples

Mare basalts from the Apollo 12 mission (12005, 12009, 12019, 12022, 12038, 12039, 12040) that have crystallization ages of 3.2 ±0.1 Ga, span a range in MgO contents from 6.7 to 20 wt.% and include olivine-, pigeonite-, ilmenite-, and feldspar-normative basalts, were analysed for Re-Os isotopes and HSE abundances. A summary of petrology, mineralogy, ages and compositions for these samples is provided in Table S1. Analyses of Apollo 12 samples are complemented by new analyses of two evolved low-Ti mare basalt meteorites from Antarctica: La Paz (LAP) 04481 (~7 wt.% MgO) and Miller Range (MIL) 05035 (7.4 wt.% MgO). La Paz mare basalt meteorites are paired, crystallised at ~3 Ga and have possible affinities to Apollo 12 mare basalts (e.g., Righter et al., 2005; Day et al., 2006), whereas MIL 05035 is a 3.8 Ga coarsely-crystalline meteorite and lacks a pronounced Eu-anomaly (Liu et al., 2009). Powder splits of 15555 and 70135 that were measured in the study of Day et al. (2007) were also measured for inter-laboratory comparison and all of these data are compared with a compiled dataset of HSE abundance data on mare basalts (Table S2).

3. Analytical Methods

Samples were requested from the Curation and Analysis Planning Team for Extraterrestrial Materials (CAPTEM), and from the Meteorite Working Group (MWG), as freshly-split interior portions of main sample masses. None of the received samples had adhering materials or metal markings when examined under a high-powered binocular microscope. Samples were disaggregated using an alumina mortar and pestle dedicated to processing lunar mare basalts, in a clean laboratory environment. To obtain the most precise Os isotope and HSE abundance data using isotope dilution, we used data previously obtained for mare basalts by Day et al. (2007) for guidance. Up to three separate digestions of each sample were made, with masses ranging from 0.26 to 1.33 g.

Osmium isotope and HSE abundance analyses were performed at the University of Maryland. Samples were sealed in 10- to 20-cm long borosilicate Carius tubes, or in quartz 50 mL high-pressure asher tubes (HPA), with isotopically enriched multi-element spike (⁹⁹Ru, ¹⁰⁶Pd, ¹⁸⁵Re, ¹⁹⁰Os, ¹⁹¹Ir, ¹⁹⁴Pt), and 4 to 11 mL of acid mixture (1 part 12 M HCl and 2 parts 15.7 M HNO₃). All reagents were multiply Teflon distilled. Samples were digested in Carius tubes at a maximum temperature of 270°C in an oven for >72 hours, and for six hours in an Anton Paar High Pressure Asher (HPA) device at a maximum temperature of 320 °C and >150 bar. Osmium was triply extracted from the acid phase into CCl₄ (Cohen & Waters, 1996) and then back-extracted from the solvent into concentrated HBr, followed by purification by micro-distillation (Birck et al., 1997). The remaining HSE were recovered and purified from residual solutions using an anion exchange separation technique. We deliberately refrained from using an HF digestion-step on residual solutions, as has been proposed by recent studies (Ishikawa et al., 2014). An HF digestion-step has been reported to access up to 9 to 15% more Re from within silicate phases, compared with a classic Carius tube/HPA digestion (Li et al., 2014), at least in some instances. Therefore, measured Re/Os obtained by HF silicate digestion after Os extraction will not reflect those in equilibrium with measured ¹⁸⁷Os/¹⁸⁸Os. This is a particular disadvantage for obtaining chronological information, where Re/Os may be increased significantly, leading to younger apparent ages.

Osmium isotopic compositions were measured as OsO₃⁻ ions in negative ion mode using a ThermoFisher Triton thermal ionisation mass-spectrometer and Re, Pd, Pt, Ru and Ir were measured using Aridus or Aridus II desolvating nebulisers coupled to a ThermoFisher Element 2 inductively coupled plasma-mass spectrometer in low-resolution mode, using identical methodologies for measurement and data correction described previously (Day et al., 2010). External precision of ¹⁸⁷Os/¹⁸⁸Os ratios, determined from individual measurements of 3.5 to 35 pg Os standards (UMCP Johnson and Matthey) during the analytical campaign, was better than ± 0.2% (¹⁸⁷Os/¹⁸⁸Os = 0.11373 ±9; n = 19; 2σ). External reproducibility (2σ) was better than 0.2% for 0.01 to 1 ppb solutions of Re and Pt, 0.3% for Ir and better than 0.5% for Ru and Pd.

Total procedural blanks (TPB), using both internally cleaned Carius tubes and the 50 mL HPA quartz tubes, gave ¹⁸⁷Os/¹⁸⁸Os of 0.195 ±0.072, and 0.24 ±0.07 pg [Os], 2.1 ±0.9 pg [Ir], 1.6 ±0.7 pg [Ru], 12.2 ±8.6 pg [Pt], 13.7 ±10.8 pg [Pd], and 0.90 ±0.19 pg [Re] (1σ; n = 8). In order to address the origin of blank contributions, we performed complete HF-HNO₃ digestions in Teflon^©, as well as 1:2 concentrated HCl-HNO₃ Carius tube digestions of both quartz and pyrex glass powders used to make Carius tubes (Table 1 and Supplementary Information). Complete HF-HNO₃ dissolutions reveal high concentrations of some HSE in quartz and Pyrex glass (~500 pg g⁻¹ Ir, >300 pg g⁻¹ Pt, >100 pg g⁻¹ Pd). Carius tube digestions of the glass powders indicate that significant and variable quantities of some HSE can be etched from the Pyrex (20 pg g⁻¹ Os, >60 pg g⁻¹ Ir, >300 pg g⁻¹ Pt) and quartz (315 pg g⁻¹ Pd) during digestions and that the ¹⁸⁷Os/¹⁸⁸Os values are relatively radiogenic (0.146 to 0.231). Therefore, in order to obtain low and consistent TPB from Carius tubes, we used a cleaning method of initial boiling of Carius tubes in Aqua Regia, followed by sealing of a 1:2 concentrated HCl - HNO₃ acid mixture in the Carius tube and heating in an oven to 200°C overnight (the same cleaning methods were employed for the later phase of analysis in Day et al., 2010). This cleaning method significantly reduces TPB; for example Pt TPB improved from 26 to 2 pg and Pd TPB improved from 27 to 3 pg.

Table 1:

HSE abundances (in pg g ⁻¹) and ¹⁸⁷Os/¹⁸⁸Os composition of quartz and pyrex glass

Sample	Glass Type	Method	Mass (g)	Os	Ir	Ru	Pt	Pd	Re	187Os/188Os	±2σ
Pyrex Glass	Neck	CT	1.035	18.8	105	4.7	5.7	0.3	3.3	0.142	0.004
	Neck	HF-HNO3	0.201	NA	413	234	435	106	55	NA
	Base	CT	1.059	21.5	29.0	2.5	640.9	9.2	2.8	0.231	0.025
	Base	HF-HNO3	0.228	NA	647	101	978	25	26	NA
Quartz Glass	Neck	CT	1.016	6.2	13.3	1.7	2.7	36.8	2.8	0.168	0.005
	Neck	HF-HNO3	0.198	NA	402	31	124	134	25	NA
	Base	CT	0.991	6.5	17.1	1.7	3.3	594	2.2	0.166	0.003
	Base	HF-HNO3	0.198	NA	613	21	115	185	22	NA
	Middle	HF-HNO3	0.210	NA	430	16	68	104	31	NA

Approximately 50 g of pyrex or quartz used to make Carius tubes were cleaned in boiling Aqua Regia and then crushed and powdered using alumina grinding devices. The powders were spiked and digested using Carius tubes (3 days, 270°C) or teflon vials (3 days, 150°C). Because of the volatility of osmium tetraoxide during high-temperature oxidation, we do not report Os concentrations or Os isotopic compositions for HF-HNO₃ experiments. Ru concentrations may also be lower in these experiments due to partial oxidation of Ru^+(VIII) to ruthenium tetraoxide (boiling point = 40°C). Samples are blank corrected to the total procedural blanks run with the experiments (For HF-HNO₃ experiments= 19 pg Ir, 21 pg Ru, 1 pg Pt, 60 pg Pd, 8 pg Re; for Carius Tubes experiments= 0.2 ± 0.2 pg Os, 1.8 ± 1.1 pg Ir, 1.1 ± 1.2 pg Ru, 9.4 ± 5.2 pg Pt, 5.9 ± 0.63 pg Pd, 0.91 ± 0.19 pg Re, ¹⁸⁷Os/¹⁸⁸Os = 0.165 ± 0.011, n = 4 - note, these TPB differ from the TPB cohort run with lunar mare basalt samples). NA = not analysed.

All concentrations and isotopic compositions are blank-corrected according to blank contributions measured for individually prepared experimental batches, resulting in a range of calculated uncertainties for samples. For small aliquants, the associated uncertainties relating to blank additions are large, and in most cases, these uncertainties exceed analytical uncertainties (Table S3). Because of the proportionally large relative uncertainties in Re blank contributions (e.g., ~11–76% variation in Re for TPB), the value Re* is also reported; this is the concentration of Re calculated assuming chondritic ¹⁸⁷Os/¹⁸⁸Os at the assumed time of sample crystallisation. External reproducibility of measurements were also monitored by regular analysis of diluted and concentrated reference materials (Allende, UB-N, HARZ-01, GP-13) in the Maryland laboratory, results of which are given in Walker et al. (2004), Puchtel et al. (2008), and Day et al. (2012).

4. Results

New HSE abundance results for three separate aliquants of 15555 (950, 955, 958) agree well with previous measurements of these samples by Day et al. (2007). Similarly, the HSE abundance measurements for LAP 04841, which is paired with the La Paz mare basalt meteorites (Hill et al., 2008), overlap values for LAP 02205, LAP 02224, LAP 02226, LAP 02436 and LAP 03632 measured by Day et al. (2007). Because of the close correspondence in the two datasets, generated from two different laboratories, we consider our new results in context with these published data.

4.1. Highly siderophile element abundance data

Apollo 12, 15 and 17 mare basalts and mare basalt meteorites span a similar range of absolute HSE abundances (0.5 to 65 pg g⁻¹ Os, 0.1 to 50 pg g⁻¹ Ir, 6.6 to 153 pg g⁻¹ Pt), ranging from three to seven orders of magnitude lower than abundances of these elements in Carbonaceous Ivuna (CI)-chondrites (Figure 1). Previous measurements of Ir concentrations by neutron activation analysis in 12009, 12022, 12038, 12040 and 15555 agree within a factor of one to ten with our own measurements (Table S2). Concentrations of Os, Ru, Pt, Pd and Re correlate broadly positively with Ir content and also exhibit a positive correlation with MgO content, indicating the strong compatibility of the HSE during fractional crystallization of mare basalt melts, as is also observed for terrestrial systems (Figure 2 and Figure S2). Samples with the highest MgO contents from the Apollo 12 (12005 [20 wt.%], 12040 [17.1 wt.%]), Apollo 15 (15016 [12.3 wt.%], 15555 [~11.8 wt.%]) and Apollo 17 (74255 [10.5 wt.%]) sample suites generally have flat, chondrite-relative HSE patterns. Samples with low MgO abundances within these suites have more fractionated HSE patterns, with low measured Ir, Ru and Pd concentrations in 12038 and 12039. Mare basalt meteorites, which have the lowest MgO contents of the samples analysed (<8 wt.% MgO), have more fractionated HSE patterns, with higher (Pt+Pd+Re)/(Os+Ir+Ru) ratios, when compared with Apollo 12, 15 and 17 mare basalts.

Figure 1 -.

Chondrite-normalised (chondrite Ivuna [CI]-type Orgueil; Horan et al., 2003) highly siderophile element (HSE) patterns for (a) Apollo 12 olivine-normative (12009, 12040), pigeonite-normative (12012, 12039), ilmenite-normative (12005, 12022) and feldspathic-normative (12038) mare basalts, (b) Apollo 15 low-Ti and Apollo 17 high-Ti mare basalts, including new measurements for 15555 and 70135, and (c) low-Ti mare basalt meteorites. Also shown are Re* estimates for the samples (dashed lines – see text for details). Previously published data for Apollo 15, Apollo 17 and mare basalt meteorites are from Day et al. (2007).

Figure 2 –

Bulk-rock MgO content versus (a) Os concentration (pg g⁻¹) and (b) ¹⁸⁷Os/¹⁸⁸Os for mare basalts. Symbols and data sources are the same as for Figure 1.

4.2. Rhenium-osmium isotope data

There is a relationship of increasing dispersion of measured ¹⁸⁷Os/¹⁸⁸Os ratios with decreasing MgO in mare basalts (Figure 2). Apollo 12 mare basalts span a range in ¹⁸⁷Os/¹⁸⁸Os from 0.1268 to 0.208. This range is similar to ¹⁸⁷Os/¹⁸⁸Os ratios for Apollo 15 (0.1220 to 0.160) and Apollo 17 (0.1255 to 0.173) mare basalts, but does not extend to the high ¹⁸⁷Os/¹⁸⁸Os ratios from some La Paz mare basalts (0.1362–0.463). MIL 05035 has the lowest measured ¹⁸⁷Os/¹⁸⁸Os of any mare basalt meteorite measured to date (0.1244). Ratios of ¹⁸⁷Re/¹⁸⁸Os for Apollo 12 mare basalts vary from 0.38 to 1.88 and this range of values is more restricted than for Apollo 15 (0.15 to 6.7), Apollo 17 (0.4 to 7.9) or mare basalt meteorites (La Paz = 2.5 to 656; MIL 05035 = 25.2). None of the mare basalt suites define isochronous relationships, although a number of aliquants of Apollo 12 mare basalts lie at or on a 3.2 Ga reference isochron (Figure 3). A large number of mare basalt samples lie to the right of the 3.2 Ga isochron, corresponding to young apparent ages.

Figure 3 -.

¹⁸⁷Re/¹⁸⁸Os versus ¹⁸⁷Os/¹⁸⁸Os for mare basalts with (a) ¹⁸⁷Re/¹⁸⁸Os = 0 to 2.5 and (b) ¹⁸⁷Re/¹⁸⁸Os = 0 to 35. Also shown are isochrons, tied to Solar System initial ¹⁸⁷Os/¹⁸⁸Os, which span Solar System formation (4.568 Ga), the mean Apollo 12 crystallization age (3.2 ±0.1 Ga) and comparative isochron lines at 1 Ga and 10 Ma. Despite crystallization ages of ≥3 Ga (Table S1), data for some mare basalts lie to the right of the 3.2 Ga isochron, in some cases to young apparent ages (<10 Ma), implying open-system behaviour of Re and/or Os. Data are from this study and Day et al. (2007).

Given the 3 to 3.8 Ga crystallization ages for mare basalts determined by independent means (Table S2), the dispersion on the ¹⁸⁷Re/¹⁸⁸Os-¹⁸⁷Os/¹⁸⁸Os plot reflects disturbance of the ¹⁸⁷Re-¹⁸⁷Os system in many samples, yielding a range of calculated initial γ_Os values for Apollo 12 (γ_Osi = +29 to −41), Apollo 15 (γ_Osi = +8 to −323) and Apollo 17 (γ_Osi = −1.3 to −424) mare basalts and negative γ_Osi values for all of the mare basalt meteorites (γ_Osi −17 to >−10,000). Consequently, the calculated Re value, assuming a chondritic initial ¹⁸⁷Os/¹⁸⁸Os at the time of sample crystallization [Re*], differs little from measured Re in some Apollo 12 mare basalts, to being significantly lower than measured in MIL 05035 and the La Paz mare basalt meteorites (Supplementary Information). There is no correlation between measured ¹⁸⁷Os/¹⁸⁸Os and γ_Osi values for the dataset. Regression of reciprocal Os concentrations to measured ¹⁸⁷Os/¹⁸⁸Os yield intercept values of: Apollo 12 = 0.1269 [r² = 0.38] (without 12038, 258 HPA Aliquant = 0.1188 [r² = 0.83]); Apollo 15 = 0.1231 [r² = 0.66]; Apollo 17 = 0.1206 [r² = 0.97]; La Paz = 0.1382 [r² = 0.87] (Supplementary Information).

5. Discussion

5.1. Post-crystallisation disturbance

Significant disturbance of the ¹⁸⁷Re-¹⁸⁷Os system has been observed, even in lunar rocks and soils with relatively high Re and Os abundances (Laul et al., 1972; Chen et al., 2002; Walker et al., 2004; Day et al., 2010). Mare basalts are crystalline, with Apollo mare basalts analysed in this study showing no petrographic evidence for impact-induced disturbance. Nonetheless, anomalously low initial Os isotope compositions (γ_Osi) calculated for some Apollo mare basalts, expressed on ¹⁸⁷Re/¹⁸⁸Os-¹⁸⁷Os/¹⁸⁸Os diagrams as deviations to the right of the 3.2 Ga isochron (Figure 3), indicate open-system behaviour of Re by addition, or loss of Os. Extreme examples are the Apollo 17 suite, which plot at or around a 10 Ma reference isochron, despite having ¹⁴⁷Sm-¹⁴³Nd and ⁴⁰Ar-³⁹Ar crystallization ages in excess of 3.7 Ga (Table S2). It is unlikely that this non-isochronous behaviour results from analytical issues. Despite low abundances of Re and Os in measured samples, sample-to-blank ratios were sufficiently high and reproducible to obtain quality data. Instead, two factors suggest that the most probable cause of open-system behaviour in mare basalts was addition of Re. First, relatively unfractionated Os/Ir (chondrite-normalised Os/Ir, excluding sample 12038 = 1.3 ± 0.9) for mare basalts are consistent with observations in terrestrial igneous systems that Os and Ir behave in a geochemically similar manner, and are normally not strongly fractionated, even in relatively evolved mafic rocks (Day, 2013). Second, the very low calculated Re* concentrations for mare basalts (range = 0.14 to 7 pg g⁻¹; mean = 1.5 ±1.4 pg g⁻¹) renders Re highly susceptible to contamination.

An extreme case of Re-Os isotope disturbance is evident in mare basalt meteorites. The potential causes of disturbance for these samples are numerous. Unlike Apollo mare basalts, which were collected at the lunar surface, mare basalt meteorites experienced impact removal from the Moon, followed by fusion melting during fall through Earth’s atmosphere and, finally, residence within the Antarctic ice. Both the LaPaz and MIL 05035 mare basalt meteorites show evidence for shock, with partially maskelyntised plagioclase, heavily fractured pyroxene and melt veins (Day et al., 2006; Liu et al., 2009). While these meteorites do not exhibit obvious geochemical or petrological evidence for terrestrial alteration, even minor weathering of sulphide or metal phases in meteorites can lead to ¹⁸⁷Re-¹⁸⁷Os disturbance, and this type of weathering is often evident in Antarctic differentiated achondrite meteorites (e.g., Brandon et al., 2012; Day et al., 2012). As with Apollo mare basalts, the apparent open-system behaviour is better attributed to Re addition, rather than Os loss, because of relatively consistent Os/Ir, and because of the anomalously high Re in MIL 05035 and the LaPaz mare basalts, compared to Apollo basalts (e.g., Figure 1).

A notable aspect of the LaPaz basalts is that they are characterized by a range of Re/Os (2.2 to ~>500) and ¹⁸⁷Os/¹⁸⁸Os (0.136 to 0.463), and all plot to the right of the 3 Ga isochron in Figure 3. The maximum residence time on the Antarctic ice for the La Paz mare basalts is defined by a cosmic ray exposure age (interpreted as an ejection age) of 55 ±5 ka (Nishiizumi et al., 2006). If open-system disturbance is responsible for Re-Os isotopic variations in the mare basalt meteorites, one interpretation of the data is that, for ~3 Ga, closed-system ingrowth of Re led to ¹⁸⁷Os/¹⁸⁸Os of ~0.14 in the LaPaz mare basalts (¹⁸⁷Re/¹⁸⁸Os = <0.6), but in the last 55 ka, some La Paz samples have seen open-system addition of Re and an increased ingrowth to ¹⁸⁷Os/¹⁸⁸Os >0.4 (¹⁸⁷Re/¹⁸⁸Os = >2). Low HSE abundance mare basalt meteorites offer an extreme example of the complexities of ¹⁸⁷Re-¹⁸⁷Os disturbance that can occur through meteorite ejection and terrestrial residence processes.

Given the evidence for Re-Os isotope disturbance, it is important to evaluate whether other HSE may have been affected by post-crystallization processes. In addition to having elevated abundances of Re, some mare basalts also have elevated Pt abundances (LAP 02224, 17 = 1095 pg g⁻¹ Pt). The maximum Pt abundances measured in some aliquants of LaPaz mare basalt meteorites are significantly higher than in Apollo 12, 15 or 17 mare basalts (<155 pg g⁻¹). Furthermore, the enrichments in Pt also correspond to high Re and Pd in the samples. Although Pt concentrations ~80 pg g⁻¹ have been measured in layers of Antarctic ice cores, these are exceptional samples and typical concentrations of both Pt and Ir are less than 1 pg g⁻¹ (Petaev et al., 2013). Therefore, Pt addition to mare basalt meteorites from residence in the Antarctic ice is unlikely.

An alternative explanation for disturbance of both Re and Pt is neutron capture effects by ¹⁸⁶W and ¹⁹⁵Pt (e.g., Herr et al., 1971; Michel et al., 1972; Wittig et al., 2013; Kruijer et al., 2013). For Re, such effects would hypothetically lead to overestimation of the quantities of Re, due to increase of the non-spike isotope, ¹⁸⁷Re (Day et al., 2010). In contrast, Pt fluence effects would lead to an excess of ¹⁹⁶Pt, due to ¹⁹⁵Pt capturing thermal, epithermal and fast neutrons more effectively than ¹⁹⁶Pt or ¹⁹⁸Pt. This would lead to an overestimate of the spike isotope - ¹⁹⁶Pt - and would reduce the resultant Pt concentration. However, for this process to occur an already high Pt content would need to be present in the sample and so it does not offer an explanation for the high Pt in some La Paz mare basalt samples.

In summary, isotope-dilution analyses of Os, Ir, Ru, Pt and Pd in mare basalts show no clear indications of disturbance from post-crystallization processes. Rhenium abundances, however, are clearly disturbed in some samples, but no compelling non-lunar explanation for high Pt abundances is recognized, so high Pt and Pd instead are interpreted to result from magmatic fractionation processes in lunar melts. As emphasised previously (Day et al., 2007; 2010), if any of the HSE were added during or after sample crystallization, such contamination would lead to the conclusion that absolute HSE abundances in mare basalts are even lower than reported values.

5.2. Magmatic fractionation processes

A remarkable aspect of the mare basalt HSE dataset is that high MgO variants from all of the landing sites studied to date have the highest HSE abundances, chondrite-relative HSE patterns and near-chondritic Os isotopic compositions. In contrast, lower MgO mare basalts tend to have lower HSE abundances, more fractionated inter-element HSE variations, and can have measured ¹⁸⁷Os/¹⁸⁸Os higher than present-day chondritic values. The most obvious explanation for these differences is that fractional crystallization led to progressive fractionation of more compatible HSE from less compatible HSE in the order of Ir ≥ Os > Ru > Pt > Pd ≥ Re (Supplementary Information). An illustration of olivine fractionation on Apollo 12 olivine- and pigeonite-normative mare basalts is shown in Figure 4. For modelling it was assumed fractionation was driven by the crystallization of olivine in equilibrium with an olivine-normative primitive melt composition (Fo₇₃; Rhodes et al., 1977). Two sets of partition coefficients were applied to the models; those from the terrestrial literature and ‘best-fit’ values. The partition coefficients for terrestrial compositions suggest higher incompatibility of some HSE in terrestrial mafic melts, especially for Pt, Pd and Re, and cannot reproduce the observed fractionations in the Apollo 12 mare basalts. The S-undersaturated nature of mare basalts (Bombarderi et al., 2005) indicates that sulphide likely did not precipitate from lunar melts. While multiple phases (i.e., more than just olivine) and phase-compositions (i.e., <Fo₇₃) were likely variable during evolution of the mare basalt parental composition(s), the iterative calculations of partitioning of the HSE between olivine and melt indicate that Os, Ir, Ru, Pt, Pd and Re all behaved compatibly in the relative order observed in fractionation trends for Apollo 12 mare basalts (Supplementary Information). For some Apollo 15 mare basalts and mare basalt meteorites, the high Pt abundances could also reflect the increased incompatibility of this element in their parental melts (kD = ~ <1), relative to Apollo 12 mare basalts.

Figure 4 –

Olivine fractionation model for Apollo 12 olivine- and pigeonite normative mare basalts. Sample data (gray dashed lines) are compared with fractionation of olivine (Fo₇₃) in equilibrium with a primitive magma composition in one percent increments using crystal-liquid partition coefficients for MgO-rich volcanic rocks from the terrestrial literature (red dashed lines; kD_Os = ~1.3; kD_Ir = ~0.7; kD_Ru = ~1.7; kD_Pt = ~0.08; kD_Pd = ~0.03; kD_Re = ~0.1; Day, 2013) and iteratively calculated best-fit crystal-liquid partition coefficients (black lines; kD_Os = 2.2; kD_Ir = 2.4; kD_Ru = 2; kD_Pt = 2.1; kD_Pd = 2.1; kD_Re = 1.5). In contrast to terrestrial magmas, the HSE appear to be strongly compatible in lunar igneous environments. CI-chondrite normalization data from Horan et al. (2003).

The results of the modelling of fractional crystallization on the HSE are consistent with trends that are evident in bulk geochemical data for Apollo 12, 15 and 17 mare basalts, and mare basalt meteorites, indicating crystallization and removal of Cr-spinel, pyroxene and olivine (Chappell & Green, 1973; Rhodes & Hubbard, 1973; Rhodes et al., 1976; 1977; Day & Taylor, 2007; Schnare et al., 2008; Liu et al., 2009). Increased compatibility required to explain mare basalt HSE abundance systematics, however, relative to terrestrial values, indicates a fundamental difference in HSE behaviour during fractional crystallization on Earth versus the Moon. The most obvious cause of these differences arises from the more reducing conditions of lunar compared with terrestrial melts, such that phases including FeNi metals (e.g., Reid et al., 1970), and spinel, into which some HSE are likely to be more compatible under low fO₂ conditions (e.g., Righter et al., 2004; Finnigan et al., 2008), occur. Birck & Allegre (1994) argued that Re is about three orders of magnitude more incompatible during terrestrial basalt petrogenesis as a function of oxygen fugacity inherited from the source regions of basalts. Results of modelling support this conclusion, but also indicate that all of the HSE, with the possible exception of Pt, are more compatible during lunar – compared with terrestrial – petrogenesis.

5.3. Crustal (regolith) contamination

The HSE are potentially extremely sensitive indicators of crustal contamination of mare basalts close to the lunar surface because of the high HSE contents of ‘chondritic’ impactor materials (Horan et al., 2003; Fischer-Gӧdde et al., 2011). Prior studies have shown that the upper portions of the lunar crust are variably enriched in the HSE through impactor contamination (Gros et al., 1976; Hertogen et al., 1977; Korotev, 1987; Warren et al., 1989; Norman et al., 2002; Puchtel et al., 2008; Fischer-Godde & Becker, 2012), so the possibility of contamination of a melt extruded onto the lunar surface would be high, if mare basalts extruded into impactor- and HSE-rich regolith at 3.8 to 3.0 Ga. Only a limited number of mare basalts, however, have been considered to have experienced significant crustal contamination (Neal et al., 1994b; Jerde et al., 1994), and severe constraints have been placed on the amount of crustal assimilation possible (typically <<3%: Finila et al., 1994). A model of impactor contamination for lunar crustal rocks shows that the most Os- and MgO-rich mare basalts could potentially have experienced <0.01% assimilation of regolith material (Figure 5). Such limited amounts of assimilation would have virtually no detectible impact on other ‘indices of contamination’ (e.g., Al₂O₃, Sr-Nd-Pb isotopes), indicating that the HSE are valuable highly sensitive tracers of contamination, for mare basalts emplaced into regolith.

Figure 5 –

Os concentration (pg g⁻¹) versus ¹⁸⁷Os/¹⁸⁸Os for lunar mare basalts, lunar pristine crustal rocks (Day et al., 2010), lunar impact-melt breccias (LIMB; Puchtel et al., 2008) and chondrites (Horan et al., 2003). The curve shows mixing between a hypothetical pristine crustal composition (¹⁸⁷Os/¹⁸⁸Os - ~0.22; Os = 1 pg g⁻¹) and an average chondrite composition (¹⁸⁷Os/¹⁸⁸Os = 0.1275; Os = 840,000 pg g⁻¹). Percentages (by mass) of the chondritic component along the mixing line are labelled. While some previously defined pristine crustal rocks show evidence for impactor contamination (Day et al., 2010), mare basalts do not show strong mixing characteristics with HSE-rich impactor-contaminated regolith and instead show progressive Re/Os fractionation with suites, during fractional crystallization of mafic parental melts.

The possibility of regolith contamination for mare basalts studied to date can be excluded, however, for two primary reasons: (1) Apollo 12 pigeonite-normative basalts are considered to derive from partial (~3%) assimilation of lunar crustal materials (Neal et al., 1994b), yet the studied pigeonite-normative basalts (12019; 12039) have more radiogenic ¹⁸⁷Os/¹⁸⁸Os and lower Os contents than inferred olivine-normative ‘parental’ compositions, inconsistent with contamination by HSE-rich impactor regolith material; (2) the siderophile elements Ni and Co are found in high concentrations in chondrites (typically >8,000 μg g⁻¹ and 500 μg g⁻¹, respectively) and have high Ni/Co (>10), however, high- and low-MgO mare basalts have much lower Ni/Co (<1.7; Supplementary Information). These lines of evidence indicate that the chondritic measured ¹⁸⁷Os/¹⁸⁸Os and chondrite-relative HSE compositions of the most magnesian mare basalts reflect derivation of the HSE from their mantle sources, not from assimilation of HSE-rich regolith materials. The general consensus is that the majority of lunar regolith formation occurred before 3.5 Ga, when the impact flux was assumed to be higher (McKay et al., 1991). Mare basalts from the Apollo 12, 15 and 17 sites and from regions of the Moon sampled by mare basalt meteorites do not appear to have sampled this regolith material during their emplacement.

5.4. Osmium isotope and HSE homogeneity in lunar mantle sources

In order to explain lithophile major- and trace-element abundance and lithophile isotope systematics, mineralogically distinct sources have been proposed for mare basalts (e.g., Rhodes & Hubbard, 1973; Rhodes et al., 1977; Neal et al., 1994a; Snyder et al., 2000; Spicuzza et al., 2007; Liu et al., 2010). This mantle heterogeneity is generally considered (e.g., Warren & Taylor, 2014) to result from magma-ocean crystallisation, the generation of cumulate mantle, followed by overturn and mixing of the cumulate layers due to density instabilities. Thus, while it is potentially possible to generate quartz- and olivine-normative Apollo 15 compositions through fractional crystallization of a homogeneous parental magma (Schnare et al., 2008), this is not possible for Apollo 12 mare basalts due to the most extreme differences in Sr and Nd isotopic composition, suggesting a mineralogically heterogeneous source (Nyquist et al., 1979). There are, however, no systematic trends between mare basalts at the Apollo 12, 15 or 17 sites that can obviously be attributed to source heterogeneities. In turn, there are no clear differences in HSE abundances of the highest MgO basalts from these Apollo sites, indicating relatively consistent distribution of the HSE within their source regions. While these mare basalts may not represent the entire lunar mantle, they indicate that Apollo 12, 15 and 17 source regions have chondritic-relative proportions of the HSE. The mare basalts for the different Apollo landing sites are therefore considered as having similar HSE abundances, although it is recognised that potential source heterogeneities may exist for other elements.

The HSE abundances of the most MgO-rich mare basalts are consistent with their derivation from a lunar mantle with chondritic-relative proportions of the HSE. In terrestrial settings, partial melting promotes HSE fractionation, primarily because the HSE are strongly controlled by sulphide and HSE-rich alloys during partial melting of the mantle, allowing approximation of melting conditions from partial melting models (Barnes et al., 1985; Rehkamper et al., 1999). For the Moon, an additional complication is the potential effect of remnant metal in the source, due to low oxygen activity. The evidence for residual metal in the mantle source after partial melting is limited, however, with correlated W/Th and W/U in mare basalts that are indicative of a metal-free source, or complete exhaustion of metal during melting (Supplementary Information). Not least, the effect of metal in the source of mare basalts would be to strongly fractionate the HSE due to the variable but high metal-silicate partition coefficients (e.g., Mann et al., 2012), which is not observed in the mare basalt data. Indeed, the relatively unfractionated patterns of high MgO mare basalts indicate similar compatibility of the HSE during partial melting in the lunar mantle. Either metal did not exist in the source of mare basalts, or it was completely exhausted during mare basalt parental melt extraction.

To approximate melting conditions in the lunar mantle, a non-modal batch-melting columnar (cylindrical) melting regime was assumed with uniform melting of a mantle source, an initial lunar mantle S content of 75 μg g⁻¹ and the melt composition of 1050 μg g⁻¹ estimated for the Apollo 12 mare basalt source (Bombarderi et al., 2005), as well as lunar mantle HSE abundances based on previous estimates (Day et al., 2007). Partitioning was assumed to be similar to that derived to calculate observations made in Figure 4. The results of these models are shown in Figure 6. The low estimated S content of the lunar mantle results in rapid exhaustion at ~11% partial melting, rather than >20% for terrestrial mantle conditions (250 μg g⁻¹). Consequently, between 5 and ≤11% partial melting is required to explain the HSE abundances measured in high-MgO mare basalts. While the models are only an approximation due to a lack of appropriate bulk distribution coefficients for lunar melting processes, they illustrate that sulphide-silicate melt partitioning of between 1000 and 10000 best describes the lunar HSE abundance data for high MgO mare basalts. The greater compatibility of the HSE, but the rapid exhaustion of S during partial melting of the lunar interior led to chondritic-relative HSE abundances in erupted, primitive mare basalt melts.

Figure 6 –

Melting models for (a) Ir concentrations as a function of partial melting for different sulphide-melt partitioning (1000, 10000, 100000) in a columnar melting regime, assuming 75 μg g⁻¹ S in the source, 1050 μg g⁻¹ S in the melt composition (Bombarderi et al., 2005) and a lunar mantle HSE composition similar and estimates of partial melting from prior studies (e.g., Day et al., 2007). Palladium is assumed to be perfectly incompatible in silicates. Solid lines in (b) show 1–15% partial melting for sulphide-melt partitioning of 1000 and dashed lines show 1–15% partial melting for sulphide-melt partitioning of 10000. The grey shaded region shows the field for the highest HSE abundance Apollo 12 mare basalts. Partition coefficients for the model are taken from lunar calculations of Figure 4.

Day et al. (2007) calculated terrestrial and lunar mantle compositions by regressing basalt data and obtained chondritic-relative HSE abundances for both bodies, as well as suggesting long-term chondritic Re/Os from measured ¹⁸⁷Os/¹⁸⁸Os of mare basalts. The regression of terrestrial basalt HSE data to a mantle composition gave similar results to HSE abundances measured in terrestrial mantle peridotites. The new results for Apollo 12 mare basalts reinforce the previous estimate of HSE abundances in the lunar interior, implying that the source regions of Apollo 12, 15 and 17 mare basalts have remarkably homogeneous HSE abundances at ~100 pg g⁻¹ Os, Ir, Ru and Pd, 200 pg g⁻¹ Pt, and 10 pg g⁻¹ Re. Previously, the disturbance of Re in Apollo 15 and 17 prevented calculation of a precise initial ¹⁸⁷Os/¹⁸⁸Os for samples from this mission. Conversely, the Apollo 12 suite can be used to calculate an errorchron age of 3.0 ±0.6 Ga and an initial ¹⁸⁷Os/¹⁸⁸Os = 0.1085 ±0.0079, if only pigeonite-, ilmenite- and olivine-normative mare basalts are considered in the calculation. While highly imprecise, the errorchron age is within uncertainty of the crystallization ages of Apollo 12 mare basalts measured using lithophile isotope chronometers (Table S2). At the time of crystallization, Apollo 12 mare basalts originated from a lunar source with chondritic ¹⁸⁷Os/¹⁸⁸Os (γOs = +1 ±4; 1σ) and long-term Re/Os within ~10% of chondritic values. This value represents the most precise initial ¹⁸⁷Os/¹⁸⁸Os composition measured to date for any lunar locality (Figure 7).

Figure 7 -.

Calculated initial ¹⁸⁷Os/¹⁸⁸Os values for lunar mare basalts, pristine lunar crustal rocks, martian meteorites, and terrestrial komatiites, peridotites and field of modern-day ocean island basalts (OIB) plotted relative to the averaged evolution ‘chondritic evolution’ curve of carbonaceous chondrites (from Walker et al., 2002). The Apollo 12 value is computed from regression of undisturbed aliquants that yield an age of 3.0 ±0.6 Ga and an initial ¹⁸⁷Os/¹⁸⁸Os = 0.109 ±0.008. Figure modified from Day (2013) with references for published data therein.

5.5. The importance of late-accretion additions to Earth and the Moon

The elevated absolute and chondritic relative HSE abundances estimated for Earth’s mantle (~0.01 × CI-chondrite) have been interpreted to reflect minor (~0.5%) continued accretion following the cessation of core segregation (e.g., Kimura et al., 1974; Chou, 1978) and has been termed ‘late accretion’. Similar constraints can be placed on Mars, from the analysis of martian shergottites (Brandon et al., 2012), indicating that the martian mantle has chondritic-relative HSE abundances ~0.007 × CI-chondrite (Day, 2013). Without late accretion, variable pressure- and temperature experiments predict low abundances and significant inter-element fractionation of the HSE through metal-silicate equilibration (e.g., Mann et al., 2012). Assuming late accretion occurred subsequent to Earth-Moon system formation, the silicate portion of the Moon should have HSE abundances similar to those of Earth’s mantle, even when accounting for the effects of size, impactor retention and gravitational focussing (Morgan et al., 2001; Walker et al., 2004). The Moon’s mantle, however, does not appear to have such high HSE abundances. Based on available data (Table 2, Table S3), the lunar mantle has a relatively homogeneous distribution of the HSE, in chondritic-relative proportions, but that are ~0.0002 × CI-chondrite.

Table 2:

Highly siderophile element data (in pg g ⁻¹) and ¹⁸⁷Os/¹⁸⁸Os for mare basalts

Sample	Method	Mass (g)	Type	MgO (wt.%)	Os	±	Ir	±	Ru	±	Pt	±	Pd	±	Re	±	Re*	187Re/188Os	±2σ	187Os/188Os	±2σ	γOsi
12009, 136	10cm CT	0.332	ONB	11.6	10.54	0.41	10.19	5.05	20.81	3.61	72.00	20.08	60.58	35.27	1.25	0.89	1.45	0.573	0.009	0.14260	0.00046	4.8
12009, 136	10cm CT	0.778	ONB	11.6	11.81	0.28	9.20	1.31	28.14	2.56	107.89	8.07	63.41	4.39	0.91	0.48	1.58	0.371	0.006	0.14167	0.00072	14.8
12009, 136	20cm CT	0.960	ONB	11.6	10.06	0.24	10.52	2.59	20.36	2.00	74.42	5.54	63.28	9.13	1.54	0.46	1.39	0.738	0.011	0.14274	0.00031	−3.9
12040, 200	10cm CT	0.351	ONB	16.7	26.90	0.40	41.09	7.70	48.80	3.81	99.88	20.84	61.91	34.88	1.42	0.95	2.21	0.254	0.004	0.12752	0.00077	7.7
12040, 200	10cm CT	0.973	ONB	16.7	45.42	0.23	41.61	1.19	67.27	2.18	113.74	6.57	53.19	3.52	3.96	0.68	3.80	0.420	0.006	0.12789	0.00002	−0.9
12040, 200	20cm CT	1.157	ONB	16.7	42.19	0.21	33.72	2.63	77.57	1.80	94.34	4.72	66.41	7.81	4.19	0.48	4.06	0.479	0.007	0.13133	0.00003	−0.8
12039, 38	10cm CT	0.595	PNB	5.8	10.04	0.36	5.28	1.45	21.85	3.15	19.48	7.20	8.76	3.62	1.50	0.52	1.73	0.724	0.011	0.15217	0.00015	5.8
12039, 38	HPA	0.932	PNB	5.8	4.11	0.22	1.39	0.85	18.64	2.51	18.92	1.83	9.88	2.58	1.59	0.59	1.53	1.882	0.028	0.20775	0.00017	−3.6
12019, 15	10cm CT	0.610	PNB	9.2	4.50	0.34	1.92	0.97	16.09	2.93	27.93	7.96	6.97	3.23	1.09	0.59	1.10	1.174	0.018	0.17210	0.00047	0.6
12019, 15	HPA	1.134	PNB	9.2	10.58	0.19	5.50	1.34	19.52	2.13	32.47	1.58	15.50	2.42	1.49	0.59	0.84	0.681	0.010	0.12683	0.00008	−15.9
12022, 298	10cm CT	0.538	INB	11.3	21.88	0.41	12.61	1.88	41.97	3.71	84.16	10.97	26.94	5.45	1.98	0.84	1.94	0.436	0.007	0.12931	0.00012	−0.4
12022, 298	HPA	1.337	INB	11.3	12.21	0.16	8.59	1.27	28.09	1.89	50.01	1.37	19.25	2.16	1.69	0.55	1.33	0.668	0.010	0.13473	0.00006	−7.8
12005, 63	10cm CT	0.652	INB	20.0	65.11	0.34	50.34	1.76	113.78	3.26	150.52	9.73	91.69	5.30	7.27	1.04	7.03	0.539	0.008	0.13454	0.00044	−1.0
12005, 63	20cm CT	0.750	INB	20.0	31.76	0.32	25.33	3.75	52.57	2.69	75.10	6.96	38.80	10.10	2.50	0.62	2.62	0.379	0.006	0.12757	0.00005	1.0
12005, 63	20cm CT	1.032	INB	20.0	48.40	0.23	41.46	2.97	88.15	2.02	111.71	5.31	73.28	8.74	4.54	0.53	4.22	0.452	0.007	0.12888	0.00004	−1.7
12038, 258	10cm CT	0.854	FNB	6.8	10.57	0.25	0.39	0.30	1.10	0.77	21.86	5.83	2.20	1.45	0.60	0.36	1.79	0.275	0.004	0.15148	0.00186	29.3
12038, 258	HPA	1.076	FNB	6.8	3.44	0.19	0.12	0.11	0.74	0.57	6.63	1.39	1.34	0.93	0.86	0.47	0.31	1.203	0.018	0.12982	0.00048	−41.2
15555, 950	10cm CT	0.589	ONB	11.3	43.00	0.37	47.46	2.73	33.20	2.31	152.70	31.85	24.08	2.88	1.30	0.60	2.66	0.146	0.002	0.12196	0.00004	8.2
15555, 955	10cm CT	0.255	ONB	11.7	12.43	0.67	25.40	2.45	28.53	2.32	97.29	31.11	32.52	23.50	1.46	1.08	1.07	0.567	0.009	0.12854	0.00035	−8.2
15555, 958	10cm CT	0.259	ONB	12.8	8.07	0.64	25.05	2.41	17.10	2.17	71.14	27.56	35.42	24.86	1.31	0.99	0.54	0.782	0.012	0.12349	0.00026	−24.6
70135, 98	10cm CT	0.801	IImB	9.8	4.12	0.26	3.69	1.35	4.27	1.28	22.66	12.84	11.91	2.00	0.62	0.36	0.31	0.729	0.011	0.12548	0.00009	−22.7
MIL 05035, 20	10cm CT	0.593	Low-Ti Met	7.4	0.48	0.21	0.45	0.39	0.45	0.38	31.97	17.77	7.24	2.24	2.52	0.77	0.03	25.2	0.4	0.12437	0.00021	−1187.5
LAP 04841,6	10cm CT	0.512	Low-Ti Met	7.3	5.42	0.40	1.58	1.07	3.51	1.57	49.99	24.04	20.51	3.18	2.58	0.86	0.63	2.297	0.034	0.13615	0.00023	−82.8

Abundance uncertainties, shown in parentheses, reflect propogated blank correction errors relating to the measured blank in the appropriate sample cohort (Table S1). Because of the range in the quantitites of HSE present among samples,uncertainties in the abundance data are highly variable, with the magnitude of uncertainties largely reflecting blank/sample ratios. Uncertainty in ¹⁸⁷Re/¹⁸⁸Os and γ_Os ratios reflects the propogated errors of the Os and Re abundance uncertainties. Re* is the concentration of Re calculated using¹⁸⁷Os/¹⁸⁸Os and Os concentration, assuming chondritic¹⁸⁷Os/¹⁸⁸Os at the time of sample formation. Ages for samples (in Ga) are listed in Table S2. CT = Carius tube digestion; HPA = high-pressure asher digestion; ONB = olivine-normative basalt; PNB = pigeonite-normative basalt; INB = ilmenite-normative basalt; FNB = feldspathic-normative basalt; HmB = ilmenite-rich basalt; Low-Ti Met = low titanium mare basalt meteorite.

The ‘missing HSE’ cannot be reconciled by their sequestration into a small lunar core, which would act to fractionate inter-element ratios at low lunar interior pressures (<4 GPa). Generation of a thick crust early in the history of the Moon that acted as a ‘shield’ to HSE-rich late accreted materials can potentially explain low HSE abundances in the lunar mantle (Walker et al., 2004; Day et al., 2007). In this scenario, the lunar crust would have, on average, ~10 to 80 ng g⁻¹ of Ir or Os, and approximately 3.7×10²⁰ kg of impactor material (Day et al., 2010). Such high HSE abundances, however, are rarely approached in lunar regolith or impact melt breccias (e.g., Chen et al., 2002; Norman et al., 2002; Puchtel et al., 2008; Fischer-Godde & Becker, 2012) and, instead, a conservative estimate is that only 1×10¹⁹ kg of impactor material is present in the lunar crust (Day et al., 2010). The solution to the conundrum of lower HSE abundances in the lunar mantle, relative to the terrestrial or martian mantles, may lie in the stochastic delivery of leftover planetesimal populations dominated by massive projectiles (Bottke et al., 2010). These authors showed that, combined with the Earth/Moon gravitational flux, massive projectiles (2,500–3,000 km in diameter) would be delivered to Earth, while fewer and smaller (~250–300 km in diameter) projectiles would strike the Moon.

The homogeneity of HSE, in chondritic-relative proportions, estimated for lunar mare basalt sources indicate efficient mixing of the HSE into the lunar interior, and even distribution within low- and high-Ti mare basalt mantle sources, indicating ~0.02 wt.% late accretion addition (versus ~0.5–0.8 wt.% for Earth). The obvious time for mixing of chondritic HSE to have occurred into the lunar mantle is during large-scale differentiation of the Moon, and prior to the formation of a thick lunar crust. Studies of long-lived lithophile radiogenic isotopes (Nd, Hf) indicate that differentiation of the lunar interior was completed by >4.3 Ga (e.g., McLeod et al., 2014) to 4.37 Ga (e.g., Gaffney & Borg, 2014), which is sometime after the generation of the oldest lunar crustal anorthosites (Norman et al., 2003). The timing of late accretion additions to the Moon is not as robust as for similar additions to smaller planetary bodies (e.g., Day et al., 2012), but is consistent with late accretion representing a natural continuum of accretion to planetary bodies after the major phase of metal-silicate equilibrium and planetary growth. Chondritic-relative abundances of the HSE in the bulk silicate Moon strongly imply delivery of the HSE after formation of the Moon, and not from the Earth during a giant impact scenario. This is because of the obvious difficulty of maintaining homogeneous and chondritic-relative abundances of these elements during formation, especially with the development of a metallic lunar core (Weber et al., 2011), which would likely draw-down and fractionate the HSE.

Given the large difference in late-accreted mass to the Earth (~0.5 to 0.8 wt.%) compared with the Moon (~0.02 wt. %), calculations predict that W isotopic composition of the lunar mantle should be more radiogenic than Earth by approximately 10 to 45 ppm (e.g., Halliday, 2007; Walker, 2014). This is due to the mass balance of mixing a relatively low ¹⁸²W/¹⁸⁴W chondritic source with silicate mantles that experienced early metal-silicate differentiation and evolved to high ¹⁸²W/¹⁸⁴W through the decay of ¹⁸²Hf, leading to an approximately −200 ppm difference in ¹⁸²W/¹⁸⁴W between chondrites and bulk silicate Earth. With recent evidence from W isotopes suggesting that the bulk silicate Moon has a ~+20 ppm ¹⁸²W/¹⁸⁴W value relative to the bulk Silicate Earth (Kleine et al., 2014; Touboul et al., in review), the new results support the delivery of ~1.47 × 10¹⁹ kg of late accretion to the Moon, relative to ~2000 to 3000 times more mass delivered to Earth (3 to 4.8 ×10²² kg).