W ratio of the lunar mantle

Earth and Planetary Science Letters 404 (2014) 1–13

Contents lists available at ScienceDirect

Earth and Planetary Science Letters www.elsevier.com/locate/epsl

Redox controls on tungsten and uranium crystal/silicate melt partitioning and implications for the U/W and Th/W ratio of the lunar mantle Raúl O.C. Fonseca a,∗ , Guilherme Mallmann b , Peter Sprung c , Johanna E. Sommer a , Alexander Heuser a , Iris M. Speelmanns a , Henrik Blanchard a a b c

Steinmann-Institut, Universität Bonn, 53115 Bonn, Germany School of Earth Sciences, The University of Queensland, Brisbane QLD 4072, Australia Institut für Planetologie, Westfälische Wilhelms-Universität Münster, 48149 Münster, Germany

a r t i c l e

i n f o

Article history: Received 26 July 2013 Received in revised form 30 June 2014 Accepted 11 July 2014 Available online xxxx Editor: T. Elliott Keywords: Hf/W primitive mantle partition coefficient oxygen fugacity lunar magma ocean

a b s t r a c t The timing of core formation is essential for understanding the early differentiation history of the Earth and the Moon. Because Hf is lithophile and W is siderophile during metal–silicate segregation, the decay of 182 Hf to 182 W (half-life of 9 Ma) has proven to be a useful chronometer of core–mantle differentiation events. A key parameter for the interpretation of 182 Hf/182 W data is the Hf/W ratio of the primitive (i.e. undepleted) mantle. Since W is incompatible during mantle melting, its ratio relative to U and other similarly incompatible elements in basalts (e.g. Th, La) may be used as proxies for their mantle sources. However, the assumption that W and U are equally incompatible may be flawed for petrological systems that equilibrated over a large range of oxygen fugacity ( f O2 ). Although W is typically perceived as being homovalent, evidence suggests that U is heterovalent over the range of f O2 inferred for the silicate mantles of the Earth and the Moon. Here we report new partitioning data for W, U, high-field-strength elements (HFSE), and Th between clinopyroxene, orthopyroxene, olivine, plagioclase and silicate melt. In agreement with previous studies, we show that these elements behave as homovalent elements at f O2 characteristic of Earth’s upper mantle. However, both W and U become more compatible at low f O2 , indicating a change in their redox state, with W becoming more compatible at progressively lower f O2 . This result for W is particularly unexpected, because this element was thought to be hexavalent even at very low f O2 . The much higher compatibility of W4+ (the species inferred here at low f O2 ) relative to W6+ means that even a small fraction of W4+ will have a significant effect on the overall compatibility of W. Our results imply that over the range of reducing conditions in which lunar differentiation is thought to have taken place (i.e. ∼IW-2 to IW-0.5), W is likely to become fractionated from U. When our partitioning data are applied to model the fractional crystallization of a lunar magma ocean, lunar trends for U/W, Hf/W and Th/W are well reproduced. The result of this model carries with it the implication that the Hf/W of the bulk silicate fractions that comprise the Earth and the Moon are virtually identical. © 2014 Elsevier B.V. All rights reserved.

1. Introduction Because of the fundamental geochemical difference between W (siderophile) and Hf (lithophile), the short-lived 182 Hf–182 W decay system has been widely used to date metal–silicate differentiation events (e.g. core formation) in the Earth–Moon system (e.g. Kleine et al., 2002; Yin et al., 2002). However, for the 182 Hf–182 W de-

*

Corresponding author at: Steinmann Institut, Universität Bonn, Poppelsdorfer Schloss (Museum), 53115 Bonn, Germany. Tel.: +49 228 739782; fax: +49 228 9782. E-mail address: [email protected] (R.O.C. Fonseca). http://dx.doi.org/10.1016/j.epsl.2014.07.015 0012-821X/© 2014 Elsevier B.V. All rights reserved.

cay system to be successfully applied, both the ε 182 W (ε 182 W = [[182 W/184 W]sample /[182 W/184 W]standard − 1] × 104 ) and Hf/W of the bulk silicate Earth (BSE) and the Moon need to be known. The BSE and the Moon have recently been shown to have identical ε 182 W within ± 10 ppm, but distinct Hf/W values of 18.0 ± 5.2 and 26.5 ± 1.1, respectively (Touboul et al., 2007, 2009). These elemental ratios were estimated assuming U, Th, and W to be similarly incompatible during igneous processes (Newsom et al., 1996; Palme and Rammensee, 1981) and using average U/W values for the lunar mantle (Palme and Rammensee, 1981) and for the BSE (Newsom et al., 1996), in conjunction with a chondritic

2

R.O.C. Fonseca et al. / Earth and Planetary Science Letters 404 (2014) 1–13

Table 1 Summary of experimental run conditions. Experiment

Starting mixes

Time (h)1

T (◦ C)

log f O2

DQFM

CO:CO2 :O2

Exp1 Exp1b Exp2 Exp3 Exp3b Exp4 Exp5 Exp5b Exp6 Exp7 Exp8 Exp8b Exp9 Exp10 Exp11 Exp12 Exp12b Exp13 Exp13b Exp 142 SE1-1 SE1-2 SE1-3 SE1-4

W0, W1, W5 W1 W1, W5 W0, W1, AT W1 W0, W1, W5 W1, W5 W1 W1, W5 W1, W5 W1, W5 W1 W1, W5 W1, W5 W1 W1, W5 W1 W5 W1 W1, W5 SE-1 SE-1 SE-1 SE-1

3-72 3-96 3-72 3-72 3-96 3-72 3-72 3-96 3-72 3-72 3-72 3-96 3-72 3-72 3-72 3-72 3-96 3-72 3-96 3-72 3-72 3-72 3-72 3-72

1300 1305 1300 1300 1305 1300 1300 1305 1300 1300 1300 1305 1300 1300 1300 1300 1305 1300 1305 1300 1200 1200 1200 1200

− 0.7 − 0.7 −3.3 −5.3 −5.3 −7.2 −8.3 −8.3 −7.4 −6.8 −8.8 −9.3 −9.3 −9.8 −11.8 −11.4 −11.3 −10.8 −10.3 −9.2 − 0.7 −3.3 −5.3 −7.3

+6.6 6.6 +4.0 +2.0 +2.0 +0.1 −1.0 −1.0 −0.1 +0.5 −1.5 −2.0 −2.0 −2.5 −4.5 −4.1 −4.0 −3.5 −3.0 −1.9 +6.6 +4.0 +2.0 +0.0

air air 0:100:0 10:300:0.75 10:300:0.75 10:175:0 20:130:0 20:130:0 10:150:0 10:300:0 70:50:0 70:40:0 140:70:0 70:63:0 140:12:0 140:20:0 70:12:0 140:40:0 70:40:0 –2 air 0:100:0 10:300:4.1 10:175:0

See Mallmann and O’Neill (2009, 2013) for details. 1 Time at supraliquidus conditions plus time spent at the desired equilibrium temperature. 2 Air leaked into furnace so f O2 was calculated using V-in-olivine and V-in-cpx oxybarometers.

Hf/U value (Rocholl and Jochum, 1993). From the identical ε 182 W yet disparate Hf/W of the BSE and the lunar mantle, Touboul et al. (2007) constrained the Earth–Moon system to be at least 50 Myr younger than the oldest solar system solids (here given as the age of Ca, Al-rich inclusions – CAI). More recently, the Hf/W value of the BSE was revised to 25.8 on the basis that W was shown to behave more incompatibly than U and Th during melting of terrestrial mantle (König et al., 2011). This new estimate for the BSE overlaps, within uncertainties, both with the lunar estimate of Touboul et al. (2007) and with the slightly lower value of 24.9 proposed by Münker (2010) on the basis of a broader lunar database. If correct, this overlap makes the dating approach of Touboul et al. (2007) lack resolution, and could indicate that core–mantle equilibration on Earth and the Moonforming giant impact might have been linked (König et al., 2011; Münker, 2010). It has long been recognized that W behaved as an incompatible element during lunar silicate differentiation (Palme and Rammensee, 1981; Wänke et al., 1974; Wänke et al., 1975) and that ratios of W to other incompatible elements (e.g., U, La) show variations beyond analytical scatter (Münker, 2010). However, in contrast to the more incompatible behavior of W relative to U and Th on the Earth (König et al., 2011), W has been shown to behave less incompatible than these two elements but more incompatible than La during lunar silicate differentiation (Palme and Rammensee, 1981; Palme and Wänke, 1975). For instance, KREEP, the incompatible trace element enriched residual liquid of lunar magma ocean (LMO) crystallization, is characterized by U/W and W/La above the assumed average value of the lunar mantle (Palme and Rammensee, 1981; Warren and Wasson, 1979). Given that the lunar mantle’s La/U is thought to be chondritic (Palme and Rammensee, 1981), the same assumption may be valid for the bulk lunar Hf/W even if W and U are clearly fractionated in lunar rocks (Palme and Rammensee, 1981). One way to solve this issue is to investigate which processes may fractionate W from U in the context of LMO differentiation. Particularly, it is desirable to investigate the potential role of redox variations in the lunar mantle.

2. Experimental methods 2.1. One-atmosphere experiments One atmosphere (i.e. 1 bar) experiments were carried out in the CaO–MgO–Al2 O3 –SiO2 (CMAS) system doped with 0.5–1 wt.% of TiO2 , Fe2 O3 , P2 O5 (introduced as ammonium dihydrogenphosphate), and approximately 1000 ppm of W and 500 ppm of Nb, Ta, Zr, and Hf, all of which added as reagent-grade oxide powders. About 1000 ppm of U and Th were also added to the starting mixes both as ground natural pitchblende containing about 0.5 wt.% of ThO2 , as well using 1000 μg/ml PlasmaCal standard solutions. Starting compositions were selected based on the work of Mallmann and O’Neill (2009), and were planned to produce silicate melt (melt), olivine (olv) and an additional phase, either clinopyroxene (cpx) or orthopyroxene (opx). To investigate the partitioning of plagioclase (plg) relative to silicate melt, experiments were carried out with two extra compositions, one modeled after a natural basaltic glass from Hawaii (ALV-3352-7; Aigner-Torres et al., 2007), and the other after a picrite from the Solomon Islands (SE-1; Schuth et al., 2004). All starting compositions are summarized in Table S1 in the electronic supplement. Starting compositions were mixed with polyethylene oxide and water, and the resulting sludge mounted on loops of thin (0.1 mm thick) Pd wire. Wire loops were then suspended from a Pt frame (or “chandelier”) in a vertical tube furnace equipped for gas mixing and equilibrated at the temperature and oxygen fugacity of interest. Palladium wire was chosen since it displays only limited Fe uptake over the f O2 of interest (e.g. Laurenz et al., 2010). Only a few wt.% of Fe was detected in the Pd wire at the lowest f O2 used in this study. Samples were equilibrated under controlled atmospheres using various gas mixes (air or CO–CO2 –O2 ) to impose values of log f O2 between −0.7 (air) and −11.8 (IW-1) at 1300 ◦ C. Gas flows were measured using Mykrolis mass flow controllers with ranges of 5–500 SCCM (standard cubic centimeters per minute) for CO and CO2 and of 0.15–15 SCCM for O2 . The f O2 for each run was calculated from the thermodynamic data for gas species using an open-source Microsoft Excel spreadsheet (Kress et al., 2004). Oxygen fugacities set by gas mixing were reg-

R.O.C. Fonseca et al. / Earth and Planetary Science Letters 404 (2014) 1–13

3

ularly checked using a CaO stabilized ZrO2 solid electrolyte cell, and were found to be within 0.2 log units of the calculated values. The temperature was held constant to ±5 ◦ for the duration of each run using a central type B (PtRh6 –PtRh30 ) thermocouple external to the alumina tube. Temperature was also checked regularly using an internal type B thermocouple along a vertical profile of the alumina tube. Experiments were kept in the furnace at the target temperatures for periods that extended from 48 to 72 hours. Samples were first heated to 1400 ◦ C (i.e. above the liquidus temperature) and then cooled down to the target temperature at rates of 0.75◦ /min (for cpx and opx) and 0.4◦ /min for olivine. These cooling rates were chosen through trial and error as they generated the largest homogeneous (i.e. unzoned) crystals. A summary of experimental run conditions is shown in Table 1. 3. Analytical methods All liquidus phases and silicate glass recovered from experimental charges were analyzed for their major and minor elemental content (P, Si, Ti, Al, Cr, Fe, Mg and Ca) using a JEOL JXA 8900 electron microprobe (Steinmann-Institut, Universität Bonn) in wavelength dispersive mode (WDS), employing 15 kV acceleration voltage and 15 nA beam current. Calibrations were carried out on San Carlos olivine, Cr-Augite and VG2 basaltic glass by measuring the peak and background positions for 10 and 5 s, respectively. A set of natural and synthetic minerals and glasses were used as secondary standards to check for reproducibility and measurement accuracy. Moreover, mineral grains were evaluated via X-ray mapping (not shown) to check for zonation and to test for equilibrium between liquidus phases and melt. Trace element abundances were measured by LA-ICP-MS using a M50-E Resonetics 193 nm excimer laser attached to a Thermo Element XR single-collector ICP-MS (Steinmann-Institut, Universität Bonn). Count rates were normalized using 29 Si as the internal reference isotope and the NIST-SRM 612 silicate glass as external standard with concentrations given by Jochum et al. (2011). The spot size for all analyses was set between 55 and 75 μm depending on the size of the phase analyzed. Laser fluence was measured at ∼7 J/cm2 and the laser frequency set to 10 Hz. The isotopes 29 Si, 31 P, 43 Ca, 47 Ti, 51 V, 53 Cr, 57 Fe, 90 Zr, 93 Nb, 137 Ba, 177 Hf, 178 Hf, 180 Hf, 181 Ta, 182 W, 184 W, 186 W, 232 Th, and 238 U were monitored, where 137 Ba was used as tracer for melt contamination of crystalline phases (following Wood and Trigila, 2001). The natural abundances of elements with multiple isotopes were well reproduced showing that there were no significant isobaric or molecular interferences. Measurement accuracy was tested using NIST-SRM 610 as a secondary standard, and was found to be better than 15% of the preferred value for that standard. Determining the trace element contents of highly incompatible elements in experimental run products is challenging due to ubiquitous “contamination” of the crystal signal output during laser ablation by small melt inclusions or adjacent phases. Special care was taken to mitigate or avoid this problem. Fig. 1 is an example of typical laser signals encountered during measurement of W and U abundances in clinopyroxene. Whenever possible, a flat, inclusionand spike-free laser signal was obtained (Fig. 1a). However, due to small grain size of most phases (50–100 μm) the laser would frequently drill through the crystal after 5–20 seconds of ablation (Fig. 1b). In such instances, only the initial part of the laser signal was integrated to yield trace element concentrations with the later “melt” signal filtered out. If less than 10 seconds were meltfree, those data were not used towards the global concentration average. More problematic was the presence of numerous melt inclusions during the integration time, which is a common problem in this type of experimental work. Fig. 1c is a typical example of a melt inclusion-rich laser signal, where numerous W- and U-rich

Fig. 1. Examples of typical laser-ablation signals recorded during the analyses of clinopyroxenes. (a) Melt inclusion-free signal. (b) Laser signal for a drilled-through clinopyroxene, where count rates increase sharply as the glass is reached. (c) Unused signal of a clinopyroxene with multiple melt inclusions.

“nuggets” are visible in the time resolved signal. In these cases, the signal was not integrated at all and these data were discarded. In an initial survey, anywhere between 10 and 20 individual laser spots were carried out for every phase. Moreover, in the more reduced experiments W was likely lost to the wire leading to reduced concentrations. To mitigate this issue, measurements were made in the vicinity of crystals and as far from the wire as possible. Count rates were reduced to concentrations following Longerich et al. (1996). If a sample yielded less than five inclusion-free spots, it was re-polished to expose additional crystals for further

4

R.O.C. Fonseca et al. / Earth and Planetary Science Letters 404 (2014) 1–13

paign, in particular cases, especially when crystals were smaller than 60 μm, a slightly lower laser frequency (8 Hz) and fluence (∼6 J cm−2 ) were used to enhance the amount of crystal signal to be integrated. 4. Results 4.1. Run products Examples of typical experimental run products are shown in Fig. 2a–c. Despite the large range of redox conditions investigated here, there are no major changes in phase assemblages, only a small increase in the melt fraction with increasing f O2 . The silicate melt in all experiments quenched to a homogeneous glass (Fig. 2a-c) as illustrated by the small uncertainties in the major element composition of glasses and liquidus phases (see Tables S2 and S3 in the electronic supplement). For the most part, liquidus phases produced in each experiment were large enough to measure by LA-ICP-MS. However, plagioclase crystals tended to be acicular and therefore challenging to analyze by LA-ICP-MS, leading to a limited dataset reported for plagioclase. A summary of the mineral/melt partition coefficients derived from our experiments is given in Table 2. 4.2. Partitioning of W and U Mineral/melt partition coefficients (D crystal/melt ) of W and U are given in Table 2 and plotted in Fig. 3. The results of this study, particularly when combined to literature data, clearly indicate that the partitioning behavior of both W and U are dependent on oxycpx/melt

gen fugacity (Fig. 3). For tungsten, both D W olv/melt DW

opx/melt

, DW

, and

show a considerable increase (more than an order of magnitude) from about QFM towards more reducing conditions cpx/melt

opx/melt

(Fig. 3a–b). For uranium, both D U and D U also show a considerable increase (about one order of magnitude) from about QFM + 2 towards more reducing conditions (Fig. 3d–e). For olivine, the effect of oxygen fugacity on U partitioning is not as clear as in the case of W, although experiments at lower f O2 display higher cpx/melt

DU than their more reduced counterparts, albeit to a lesser degree than shown by the pyroxene dataset (Fig. 3f). 4.3. Partitioning of Nb, Ta, Hf, Zr, Ti and Th Mineral/melt partition coefficients for Hf, Zr, Th, Ti, Nb and Ta are given in Table 2. These elements are all homovalent elements in high-temperature magmatic systems, thus their partitioning behavior is expected to be insensitive to f O2 . Among these elements, Hf, Zr, Th and Ti are tetravalent over almost the entire range of f O2 investigated here, although trivalent Ti is expected to occur at f O2 more reduced than QFM-3 (Mallmann and O’Neill, 2009; Borisov, 2012), and Nb, Ta and P are pentavalent over the same range of f O2 (Burnham et al., 2012). As expected, these elements show little evidence for a change in their oxidation state over the range of f O2 reproduced in this experimental study. However, plg/melt

DM present some variation, which may be explained by different plagioclase compositions in each experiment (see Section 5.2). 5. Discussion Fig. 2. Backscatter electron images of typical run products, where clinopyroxene (a, c), olivine (a, c), orthopyroxene (b), and plagioclase (c) are common liquidus phases.

LA-ICP-MS analyses. If crystals were too inclusion-rich, or grains too skeletal to produce a “melt-free” signal, those experiments were repeated until large enough crystals were obtained. Although laser conditions were kept constant for most of our analytical cam-

5.1. Crystal chemistry crystal/melt

Mineral/melt partition coefficients (D M ) for 4+, 5+ and 6+ metal cations (M) are plotted as a function of ionic radius (rM ) in Fig. 4. Cations of identical valence were fit to the lattice strain model of Blundy and Wood (1994), according to the expression:

Table 2 Summary of crystal/melt partition coefficients determined as a function of oxygen fugacity.

QFM

P

Ti

Cr

Zr

Clinopyroxene/melt W1-1 W1-2 W1-3 W1-4 W1-5 W1-6 W1-7 W1-8 W1-9 W1-10 W1-11 W1-12 W1-14

6.6 4.0 2.0 0.1 −1.0 −0.1 0 .5 −2.0 −2.5 −3.0 −4.5 −4.1 −1.9

0.030 (13) 0.011 (3) 0.021 (5) 0.011 (4) 0.020 (20) 0.019 (7) 0.027 (16) 0.037 (16) 0.058 (23) 0.047 (19) 0.101 (24) 0.038 (9) 0.020 (4)

0.20 (3) 0.17 (1) 0.21 (2) 0.17 (2) 0.21 (4) 0.22 (2) 0.22 (2) 0.28 (3) 0.30 (5) 0.32 (7) 0.47 (3) 0.36 (2) 0.26 (3)

27 (7) 17 (3) 65 (13) 9 (2) 7 (2) 12 (2) 14 (2) 6 (2) 6 (2) 5 (3) 17 (3) 11 (1) 4 (1)

0.06 0.03 0.05 0.03 0.05 0.04 0.05 0.08 0.06 0.06 0.08 0.05 0.05

Orthopyroxene/melt W5-1 W5-2 W5-4 W5-5 W5-6 W5-7 W5-8 W5-9 W5-10 W5-12 W5-13 W5-14

6.6 4.0 0.1 −1.0 −0.1 0 .5 −2.0 −2.5 −3.0 −4.1 −3.5 −1.9

0.016 (3) 0.034 (4) 0.035 (3) 0.040 (10) 0.029 (7) 0.015 (2) 0.154 (60) 0.107 (20) 0.093 (20) 0.024 (4) 0.017 (9) 0.028 (8)

0.066 (1) 0.073 (1) 0.069 (1) 0.073 (1) 0.079 (1) 0.069 (1) 0.074 (1) 0.091 (2) 0.093 (1) 0.100 (3) 0.094 (2) 0.091 (1)

110 (38) 8.6 (20) 2.9 (5) 2.8 (7) 3.0 (4) 2.1 (1) 2.0 (2) 1.8 (6) 2.0 (4) 1.2 (2) 1.8 (7) 2.8 (13)

0.003 0.004 0.003 0.004 0.004 0.003 0.006 0.006 0.011 0.004 0.008 0.004

Olivine/melt W1-2 W1-3 W1-4 W1-5 W1-13 W1-14 SE1-4b W1-OLV-1b W1-OLV-3b W1-OLV-5b W1-OLV-8b W1-OLV-13b W1-OLV-12b

4.0 2.0 0.1 −1.0 −3.5 −1.9 0 .1 6.6 2.0 −1.0 −2.0 −3.0 −4.0

0.35 (5) 0.04 (6) 0.11 (1) 0.34 (8) 1.04 (16) 0.92 (37) 0.80 (29) 0.12 (1) 0.12 (3) 0.07 (2) 0.11 (3) 0.11 (3) 0.12 (3)

0.005 0.006 0.005 0.005 0.007 0.009 0.014 0.003 0.004 0.003 0.004 0.005 0.008

(3) (3) (1) (2) (1) (2) (4) (1) (0.5) (1) (1) (1) (2)

2.7 (7) 8.0 (19) 1.5 (3) 1.2 (3) 1.3 (1) 1.4 (3) 1.9 (2) 2.1 (1) 1.2 (2) 1.0 (1) 0.8 (1) 1.4 (1) 1.4 (1)

0.0006 0.0010 0.0007 0.0004 0.0008 0.0004 bdl 0.0008 0.0008 0.0005 0.0010 0.0010 0.0010

bdl 0.0409 (3) 0.0506 (7) bdl -

0.054 0.042 0.047 0.024

(9) (5) (4) (9)

bdl bdl bdl bdl

0.0038 (5) bdl bdl 0.0005 (2)

Plagioclase/melt AT-3 SE-1c SE1-3b W0-4a

2.0 6.6 2.0 0 .1

-

Nb (2) (1) (1) (1) (3) (1) (2) (1) (1) (2) (1) (2) (1)

(1) (1) (1) (2) (1) (1) (2) (1) (5) (2) (10) (1)

(1) (4) (2) (1) (1) (1) (2) (2) (2) (1) (2) (4)

Hf

Ta

W

0.0028 0.0019 0.0025 0.0016 0.0027 0.0019 0.0069 0.0054 0.0101 0.0097 0.0283 0.0111 0.0039

(10) (5) (10) (10) (20) (10) (10) (20) (14) (60) (60) (40) (10)

0.10 (4) 0.08 (2) 0.12 (2) 0.07 (2) 0.11 (5) 0.10 (3) 0.11 (4) 0.16 (3) 0.12 (3) 0.15 (5) 0.19 (2) 0.11 (4) 0.10 (3)

0.008 0.003 0.005 0.003 0.006 0.004 0.009 0.010 0.010 0.011 0.020 0.009 0.004

0.0014 0.0021 0.0015 0.0017 0.0022 0.0020 0.0034 0.0052 0.0084 0.0021 0.0062 0.0051

(3) (5) (6) (6) (4) (1) (23) (16) (48) (11) (86) (30)

0.007 0.008 0.006 0.009 0.010 0.007 0.008 0.009 0.014 0.006 0.011 0.006

0.0012 0.0020 0.0014 0.0013 0.0016 0.0010 0.0021 0.0035 0.0070 0.0045 0.0074 0.0149

0.00039 0.00047 0.00010 0.00016 0.00012 0.00014 bdl 0.00008 0.00006 0.00007 0.00011 0.00010 0.00007

(27) (24) (5) (11) (2) (2) (5) (1) (3) (1) (1) (3)

0.0075 (4) bdl bdl 0.0007 (3)

(2) (3) (2) (4) (3) (1) (3) (5) (5) (3) (10) (1)

0.0013 0.0013 0.0012 0.0009 0.0014 0.0009 bdl 0.0020 0.0015 0.0008 0.0021 0.0021 0.0022

(3) (4) (4) (3) (2) (2) (6) (4) (3) (9) (6) (10)

0.003 (2) 0.0981 (30) 0.042 (76) bdl -

(3) (1) (2) (2) (4) (2) (10) (3) (4) (6) (4) (4) (2)

(4) (7) (4) (6) (2) (2) (8) (11) (45) (22) (55) (12)

0.00015 0.00038 0.00005 0.00027 0.00009 0.00009 bdl 0.00014 0.00004 0.00009 0.00013 0.00013 0.00017

(14) (32) (1) (15) (4) (4) (1) (1) (5) (2) (1) (1)

0.0058 (4) bdl bdl bdl -

Th

U

0.00008 0.00007 0.00007 0.00005 0.00007 0.00006 0.00007 0.00076 0.00204 0.00299 0.00632 0.01164 0.00092

(4) (4) (2) (1) (3) (2) (1) (54) (113) (162) (352) (510) (16)

0.005 0.003 0.003 0.002 0.004 0.003 0.010 0.010 0.007 0.007 0.008 0.007 0.003

(3) (1) (2) (1) (2) (2) (10) (6) (4) (3) (1) (1) (1)

0.00069 0.00070 0.00159 0.00155 0.00126 0.00104 0.00210 0.00500 0.01138 0.00150 0.00540 0.00250

(18) (1) (6) (5) (3) (10) (70) (121) (655) (83) (92) (144)

0.0031 0.0042 0.0022 0.0014 0.0018 bdl 0.0031 0.0030 0.0069 0.0038 0.0026 0.0020

0.00003 0.00013 0.00003 0.00015 0.00128 0.00050 0.00003 0.00003 0.00003 0.00003 0.00026 0.00053 0.00375

(1) (2) (2) (8) (4) (21) (1) (1) (1) (1) (13) (26) (12)

bdl bdl bdl bdl bdl bdl bdl 0.0000012 0.0000023 0.0000020 0.0000101 0.0000030 0.0000043

0.00106 0.00050 0.00006 0.00059

(3) (12) (1) (34)

0.006 0.035 0.016 0.004

(3) (20) (7) (13) (3) (29) (2) (42) (16) (18) (11)

(1) (12) (5) (3)

(8) (7) (10) (14) (1) (1)

0.0006 0.0004 0.0005 0.0010 0.0023 0.0016 0.0013 0.0043 0.0054 0.0059 0.0056 0.0059 0.0027

(2) (1) (1) (2) (8) (4) (3) (11) (24) (23) (5) (7) (6)

0.0002 0.0003 0.0003 0.0002 0.0001 bdl 0.0014 0.0037 0.0046 0.0043 0.0024 0.0029

(1) (1) (2) (1) (0.6) (8) (13) (3) (19) (3) (17)

bdl bdl bdl bdl bdl bdl bdl 0.0000013 0.0000024 0.0000036 0.0000104 0.0000074 0.0000095

0.00113 0.00011 0.00002 0.00042

R.O.C. Fonseca et al. / Earth and Planetary Science Letters 404 (2014) 1–13

Run

(1) (1) (2) (7) (2) (2)

(50) (5) (1) (1)

Uncertainties are 1 s standard deviation and given in parentheses as the last significant digits of the mean value.

5

6

R.O.C. Fonseca et al. / Earth and Planetary Science Letters 404 (2014) 1–13

Table 3 Results for non-linear least squares fitting of the crystal/melt partitioning data to Eq. (1). Valence

Cations*

Site

r0

1σ

E (GPa)

1σ

D0

1σ

χv2

Clinopyroxene 4+

Hf, Ti, Zr, Re

M1(VI-fold)

0.6595

0.0015

6740

745

7.65

2.52

7.4

Orthopyroxene 4+

Hf, Ti, Zr, Re

M1(VI-fold)

0.6526

0.0020

7585

835

1.57

0.48

1.9

Olivine 4+

Hf, Ti, Zr, Re

M1(VI-fold)

0.6566

0.0015

8612

919

0.27

0.12

4.0

*

Re partitioning data from Mallmann and O’Neill (2007). Regressions were weighted with uncertainties in D as measured or 25% and 0.002 A in ionic radius (from Shannon, 1976).

crystal/melt

DM

  4π N A E r 0 = D 0 exp − (rM − r0 )2 RT 2  1 3 + (rM − r0 )

Table 4 Non-linear least squares fits to Equation (2).

(1)

3

where r0 is the characteristic size of the lattice site at which lattice strain is minimized, E is the apparent Young’s modulus of the site occurring when rM = r0 , N A is the Avogadro’s constant, R is the universal gas constant, T is temperature, and D 0 is the partition coefficient of a cation with ionic radius r0 . Summary of fitting results are given in Table 3. With the exception of P5+ , U4+ and Th4+ , all other 4+, 5+ and 6+ cations appear to substitute into the octahedral M1 sites of pyroxenes. Tetravalent cations are noticeably more compatible than pentavalent and hexavalent cations. Because of its much smaller ionic radius, P5+ likely substitutes for Si in tetrahedral sites. U4+ and Th4+ are considerably larger than the other tetravalent cations and, therefore, may substitute along with the REEs in M2 sites. Partitioning of HFSE between olivine and melt is overall very similar to that of pyroxenes, with most HFSE cations entering the smaller octahedral M1 site of olivine. Exceptions are again P5+ , U4+ and Th4+ , with P5+ most likely substituting for Si in tetrahedral sites (e.g. Mallmann et al., 2009), and U4+ and Th4+ entering the slightly larger octahedral M2 site of olivine. It should be noted that, because there are not sufficient constraints to fit all three unknowns in Eq. (1) for pentavalent and hexavalent cations in pyroxenes and olivine, we have not been able to constrain E for hexavalent and pentavalent cations in the M1 site. However, Hill et al. (2011, 2012) have recently shown that E should increase in the M1 site as a result of increasing cation charge. Unfortunately, to the best of our knowledge, there is no study constraining the lattice strain parameters of pentavalent and hexavalent elements in silicates. This is likely due to the simple fact that there are not many pentavalent and hexavalent elements present in the most common rock-forming minerals. The parabolae obtained for tetravalent cations enable partition coefficients of W4+ to be predicted. Taking the ionic radius of W4+ in VI-fold coordination to be 0.66 (Shannon, 1976), the following cpx/m D W 4+

opx/m D W 4+

ol/m D W 4+

values are obtained: = 7.4, = 1.4, and = 0.26. These values are much higher than those determined experimencpx/melt

tally (see Fig. 3). For example, all D W

presented here and in cpx/m

the literature are far lower than the expected D W4+

of 7.4. The

implication of this is that only a small proportion of W4+ is actually present at QFM and that the bulk of the W is in a hexavalent state in the melt in agreement with previous studies (e.g. O’Neill et al., 2008, Wade et al., 2012, 2013). Given these constraints, we fit the partitioning data for U and W to the following expression: crystal/melt

DM

=

crystal/melt D M4+ K ( f O2 )−1/2

1 + K ( f O2

crystal/melt + D M6+ )−1/2

(2)

where K is the equilibrium constant for the redox reaction M6+ O3 = M4+ O2 + (1/2)O2 and M is either W or U (see

K

D 4+

D 6+

n

χv2

Tungsten cpx opx olv

1.02E-09 4.60E-08 1.87E-08

7.4a 1.4a 0.26a

0.00004b 0.00070 0.00003

8 11 9

1.94 1.75 0.62

Uranium cpx opx olv

2.19E-05 1.95E-05 2.35E-05

0.0099 0.0040 0.000010

0.00037 0.00020 0.0000017

8 9 6

0.64 0.52 0.25

a b

Constrained from the lattice strain fit of tetravalent cations. Defined from average values at the oxidizing plateau.

Mallmann and O’Neill, 2007, 2009 for details). The results of the fit are summarized in Table 4 and are plotted as full lines in Fig. 3. 5.2. Comparison with previous work O’Neill et al. (2008) have shown using XANES spectroscopy that W in silicate melt is predominantly hexavalent over the range of f O2 reported for solar system materials (i.e. more oxidized than IW-5; Wadhwa, 2008). The results of O’Neill et al. (2008) have since been supported by similar observations by Cottrell et al. (2009) and Wade et al. (2012, 2013). However, from our results, the systematic variability of W partition coefficients with f O2 is consistent with a minor, albeit important, contribution of W4+ to the bulk W content of the silicate melt. Indeed, the large differences between the partition coefficients determined here for clinopyroxene, orthopyroxene and olivine and those obtained by Righter and Shearer (2003), Adam and Green (2006), Frei et al. (2009), van Kan Parker et al. (2010) and Dygert et al., 2014 are reconciled once f O2 is taken into account, despite including data from experiments where clinopyroxene and orthopyroxene have higher

IV

cpx/melt

Al contents (Fig. 3a-c). Furthermore, D W

olv/melt DW

opx/melt

, DW

and are nearly constant over the range of f O2 where W is exclusively hexavalent (i.e. above QFM), in agreement with Bali et al. (2012), irrespective of the concentration of W in the melt (0.067 to 4.4 wt.%) suggesting that Henry’s law is obeyed. There are plg/melt

few D W

data available in the literature. Righter and Shearer plg/melt

were de(2003) reported only two experiments where D W termined, with values ranging between 0.003 (run #9b) and 0.028 (run #133). Our own values are lower than those reported by these authors and span a range between 6 × 10−5 (SE1-3b) and 0.0016 (AT-3) and do not correlate with f O2 . It is possible, however, plg /melt

that the discrepancy between our D W and those reported by Righter and Shearer (2003) could be explained by variations in f O2 , as Righter and Shearer’s experiments were qualitatively more reduced than those reported here. Additional experimental data are needed to fully explore this possibility to the same extent as for clinopyroxene, orthopyroxene and olivine.

R.O.C. Fonseca et al. / Earth and Planetary Science Letters 404 (2014) 1–13

7

Fig. 3. Partitioning of W and U as a function of f O2 (given relative to the QFM redox equilibrium) between (a, d) clinopyroxene, (b, e) orthopyroxene, (c, f) olivine, and haplobasaltic melt. Also shown are the best fits of the partitioning data to equation (2). Only data that was fit to equation (2) are shown. Fitted parameters are given in Table 4. Error bars are 1σ standard deviation. Selected literature partitioning data for W (Righter and Shearer, 2003; Adam and Green, 2006; and van Kan Parker et al., 2011), and U (LaTourrette and Burnett, 1992; Kennedy et al., 1993; Beattie, 1993; Dunn and Sen, 1994; Lundstrom et al., 1994; Wood et al., 1999; McDade et al., 2003; Adam and Green, 2006; van Kan Parker et al., 2012; and Dygert et al., 2014) are shown for comparison. Whenever data from previous studies was produced at unconstrained cpx/melt

f O2 , this parameter was calculated based on D V O’Neill (2009, 2013).

cpx/melt

measured in the experiments using the relationship between D V

Due to interest in interpreting U/Th disequilibrium series during partial melting, a wealth of data exists on the partitioning of U and Th between liquidus phases and silicate melt (LaTourrette and Burnett, 1992; Beattie, 1993; Kennedy et al., 1993; Dunn and

olv/melt

, DV

and f O2 calibrated by Mallmann and

Sen, 1994; Lundstrom et al., 1994; Wood et al., 1999; Wood and Trigila, 2001; Blundy and Wood, 2003; Adam and Green, 2006; Frei et al., 2009; van Kan Parker et al., 2012). Unlike Th, which is always tetravalent, U changes oxidation state as a function of

8

R.O.C. Fonseca et al. / Earth and Planetary Science Letters 404 (2014) 1–13

f O2 in controlling the partitioning behavior of U during mantle melting (e.g. LaTourrette and Burnett, 1992; Lundstrom et al., 1994), there has not been a systematic study on the partitioning of U between silicates and melt covering the range of f O2 over which U changes its oxidation state. Our results, however, show that a clear redox transition from U6+ to U4+ is presented mineral/melt

around QFM, even if D U are still well below unity (i.e. ∼10−6 –10−3 ). As a consequence, D U / D Th changes as a function of f O2 from ∼2 when U4+ is present to ∼0.1 when U6+ is predominant. The former value is similar to those reported in previous cpx/melt

cpx/melt

studies for D U4+ / D Th within uncertainties (see Blundy and Wood, 2003), probably because the bulk of experimental partitioning studies of Fe-bearing compositions have been done in Pt-lined graphite capsules which limits the f O2 of the system to lower

values where U4+ is dominant. However, our D U / D Th under oxidizing conditions are lower than literature values, confirming that in previous experimental work U4+ is prevalent. It is not surprising then that once f O2 is considered, our U4+ and Th partitioning data for clinopyroxene and orthopyroxene show good agreement with values reported in previous work (see also cpx/melt

cpx/melt

olv/melt

Fig. 3d to f). There are scarce D U experimental data available in the literature and none span a large range in f O2 . Previous studies, namely where an ion probe was used to measure U and Th abundances in olivine (Beattie, 1993; Kennedy et al., 1993; Wood et al., 1999; McDade et al., 2003) report values lower than 10−4 –10−5 , in agreement with our own results (Fig. 3f). However, olv/melt

other studies feature D U as high as 0.01 (Dunn and Sen, 1994; Adam and Green, 2006). The latter studies employed LA-ICP-MS to olv/melt

measure U concentrations in olivine, thus the higher D U

reflect contamination by melt inclusions. There are few plg /melt

and D Th

may

plg/melt DU

data reported in the literature. Bindeman and Davis plg/melt

, for plagioclase with ∼An70 , of around (2000) obtained D U 0.011. These results are higher than the value of 0.00042 for the most reducing experiment reported here. On the other hand, plg /melt

De Vries et al. (2012) constrained D U4+ at ∼0.00103 using Zr, Hf and Th to define lattice strain parameters, which is about a factor of two larger than our preferred value of 0.00042 (at IW + 3.5), and identical within error to the aggregate value of 0.0006 from the compilation of Blundy and Wood (2003). There are even fewer plg/melt

D Th in the literature, with only one recent measurement by De Vries et al. (2012) who obtained a value of ∼0.001. This value plg /melt

is slightly lower than the D Th of 0.004 of experiment W0-4c, as well as the aggregate value of ∼0.0034 of Blundy and Wood plg /melt

plg /melt

(2003). However, variations of D U and D Th between each dataset can be explained by differences in plagioclase composition. Tepley et al. (2010) have shown that trace elements become substantially less compatible in plagioclase for compositions >An75 , in agreement with previous work (e.g. Blundy and Wood, 1991; plg/melt

Fig. 4. Average clinopyroxene/melt (a), orthopyroxene/melt (b), and olivine/melt (c) partition coefficients for 4+, 5+ and 6+ cations plotted against ionic radius in VI-fold coordination (except for U4+ and Th4+ in pyroxenes, which are given in VIIIfold coordination). Ionic radii taken from Shannon (1976) with assumed uncertainty of 0.002Å. Lines represent least-squares fits of the data to the lattice strain model of Blundy and Wood (1994), and Wood and Blundy (2001) as expressed in Eq. (1). The dashed line in (c) represents the fit to trivalent cations entering the M2 site defined by Evans et al. (2008). Lattice strain parameter for hexavalent and pentavalent cations are unconstrained, thus parabolas are shown as dashed lines. Details are given in Table 4. Partitioning data for Re and V are taken from Mallmann and O’Neill (2007) and from Mallmann and O’Neill (2009), respectively.

f O2 , from U4+ under reducing conditions, to U6+ at high f O2 , and possibly, U5+ at intermediate f O2 (Calas, 1979; Berry et al., 2008). Although a few studies recognized the importance of

Bindeman, 2007). For example, the D U values reported by Bindeman and Davis (2000) have compositions of An70 and below. On the other hand, the plagioclase from experiment W0-4c, and those described in De Vries et al. (2012), consist of almost plg /melt

plg /melt

pure anorthite, which would drive D U and D Th to lower values. Overall, our average mineral/melt partition coefficients for Nb, Ta, Hf, Zr, and Ti are well within error to those reported previously by Mallmann and O’Neill (2009) for clinopyroxene, orthopyroxene and olivine, as well as those reported recently by Dygert et al. (2014) for clinopyroxenes of similar composition (i.e. Mgrich). Moreover, partition coefficients for Hf, Zr, Ti, Nb and Ta follow the normal trend of incompatible elements, with D cpx/melt > D opx/melt > D olv/melt (Table 2). Partition coefficients obtained for plagioclase are broadly consistent with the results presented by

R.O.C. Fonseca et al. / Earth and Planetary Science Letters 404 (2014) 1–13

Tepley et al. (2010) if the An content of plagioclase is taken into plg/melt

account. For example, our D Nb plg/melt

plg /melt

(0.0075), D Ta

(0.0058), and

D Ti (0.054) are comparable with values reported by Tepley et al. (2010) for plagioclase with similar composition (i.e. An77–78 ; plg /melt

plg/melt

0.0054, 0.0020, 0.045 respectively). For D Zr , and D Hf the agreement is not as good, but cover a similar range of values (∼10−3 to 10−2 ) to those reported by Blundy et al. (1998), Tepley et al. (2010) and De Vries et al. (2012). 5.3. Factors controlling the partitioning behavior W, U, Th and the HFSE The behavior of W, U, Th and the HFSE during partial melting is expected to depend on melt composition (e.g. O’Neill and Eggins, 2002; O’Neill et al., 2008; Dygert et al., 2013) as well as on crystal chemistry (e.g. Hill et al., 2000), pressure and especially temperature. Hill et al. (2000) argued that the partitioning of W between clinopyroxene and melt is inversely correlated with the IV Al content of the clinopyroxene in their experiments, while the HFSE show the opposite behavior. This observation is at odds with our cpx/melt

DW , which are always lower than those reported by Hill et al. (2000) even though the IV Al in the clinopyroxene of our expercpx/melt

iments is lower than theirs. However, it is possible that D W reported by Hill et al. (2000) are affected by the CaO content of their melts, which varies from 17 to 30 wt.% and is inversely correcpx/melt

lated with D W (O’Neill and Eggins, 2002). O’Neill et al. (2008), in their study of the solubility of W in silicate melts, found that the XANES spectra obtained from their experimental glasses indicates that W6+ displays tetrahedral coordination. They argued that W is likely associated with CaO forming stable CaWO4 melt species. O’Neill et al. (2008) interpreted this to indicate that CaO (and to a lesser extent MgO) enhance the solubility of W in a silicate melt, and change the mineral/melt partitioning behavior of W. It is also possible that variations in the redox conditions operating in the experiments by Hill et al. (2000) might have affected the partitioning behavior of W, in addition to changes in melt composition and IV Al in clinopyroxene. Hill et al. (2000) carried out their experiments in sealed Pt capsules, so f O2 is unconstrained. It is important to point out that the effect of CaO in the silicate melt may have further effects on the partitioning behavior of W. For example, O’Neill and Eggins (2002) have shown that the Mo6+ –Mo4+ transition in silicate melts (a useful proxy for W) is not only f O2 -dependent, but also depends on the CaO content of the melt. O’Neill and Eggins (2002) compared the position of the Mo6+ –Mo4+ redox transition between two different melt compositions, one CaO-free and MgO-rich (MAS system) and the other CaO-rich and MgO-free (CAS system). They found that the Mo6+ –Mo4+ redox transition was at an f O2 almost two orders of magnitude higher for the CaO-free melt composition when compared to the CaO-rich melt composition. The implication of these results is that the redox state of Mo in silicate melts at hightemperature is affected by the Ca/Mg ratio of the melt. The same is also likely true for the W6+ –W4+ redox transition (O’Neill et al., 2008). Natural melts do not display such variations in Ca/Mg, so that it is unlikely that the W6+ –W4+ transition would shift by two orders of magnitude in f O2 . Nevertheless, given that the compatibility of W changes by several orders of magnitude over this redox-transition, even a small shift in its position will have a significant effect on the bulk partitioning behavior of W during partial melting. Considering that natural basalts have variable CaO contents, it is conceivable that in natural systems the W6+ –W4+ redox transition may shift to different f O2 , which merits further investigation. Fe-Ti-ox/melt Recently, Dygert et al. (2013) showed that D M for the HFSE, Cr and V decrease as TiOmelt increases. This was inter2

9

preted to reflect associations between these elements and pseudobrookite (FeTi2 O5 ) melt complexes that become more abundant as TiOmelt increases. Dygert et al. (2013) argued that the presence 2 of these Fe–Ti–O complexes lowers the activity the HFSE, Cr and V in the melt. Because these elements substitute into the same crystal site as Ti in Fe–Ti oxides (i.e. the HFSE, Cr and V), their Fe-Ti-ox/melt DM become lower as a result. It is therefore possible that any cation which assumes the same lattice site as Ti (i.e. W4+ ) in a liquidus phase may be similarly affected. For terrestrial basalts, where TiOmelt is seldom higher than 4 wt.% (Bence et al., 1980), 2 this compositional effect is likely unimportant. However, for lunar basalts, which have TiO2 contents ranging between 0.2 and 17 wt.% crystal/melt

(e.g. Delano, 1986), it is possible that D W

may be also low-

ered by TiOmelt in similar fashion to the HFSE, Cr and V in Fe–Ti 2 oxides. It should be pointed out, however, that O’Neill et al. (2008) did not observe any effect of TiO2 on W solubility in silicate melts past a simple dilution effect. As seen from the discussion above, the competing effects of melt composition, crystal chemistry and f O2 on the partitioning behavior of trace elements like W make it challenging to apply experimental data directly to natural settings. However, our results clearly show that, at least to some extent, differences between different datasets can be reconciled once the effect of f O2 is taken into account. Nevertheless, the effect of melt composition, particularly the way melt complexes (i.e. CaWO4 , FeTiO5 ) affect trace element partitioning, requires additional study. 5.4. Redox-dependent fractionation of W from U, Th, and the HFSE during lunar magma ocean crystallization and subsequent lunar mantle melting Much of the interest in unraveling the behavior of W during mantle melting hinges on knowing how W fractionates from Hf throughout planetary differentiation. Our results show that W, Th, and U are equally incompatible during magmatic differentiation at f O2 close to IW + 4 (i.e. around QFM). This observation is consistent with the use of average mantle values of U/W or Th/W that have commonly been employed in conjunction with chondritic Hf/Th or Hf/U to estimate the bulk Hf/W composition of a planetary body (e.g. Touboul et al., 2007; Münker, 2010). However, the partitioning of W and U during mantle melting is shown to be redox sensitive. It is thus important to readdress the fractionation of these elements during LMO crystallization, which took place under more reducing conditions than on Earth (IW-2 to IW-0.5 — Stanin and Taylor, 1980; O’Neill, 1991; Righter and Drake, 1996). The fractionation of W from the HFSE, U, and Th during lunar silicate differentiation has long been recognized (e.g., Münker, 2010; Palme and Rammensee, 1981). The most important lunar silicate differentiation event is thought to have been the crystallization of the LMO (Smith et al., 1970; Snyder et al., 1992; Warren, 1985; Wood et al., 1970). Snyder et al. (1992) developed an LMO crystallization model that accounts for major- and trace-element and Nd, Sr, and Hf isotope features of the lunar sample suite (e.g., Münker, 2010; Snyder et al., 2000; Snyder et al., 1994; Snyder et al., 1997; Sprung et al., 2013). This crystallization model has five steps: (i) equilibrium crystallization of olivine (until 40% solidification), (ii) appearance of orthopyroxene (until 78% solidification), (iii) fractional crystallization of plagioclase, olivine, and pigeonite (until 86% solidification), and (iv) fractional crystallization of clinopyroxene, plagioclase and pigeonite (until 95% solidification), and finally (v) crystallization of pigeonite, plagioclase, clinopyroxene, and ilmenite (until 99.5% solidification). In addition, it is assumed that various amounts of trapped instantaneous liquid (TIL, i.e. coexisting melt at the time of crystallization) are part of lunar mantle cumulates, and that 98% of the crystallizing plagioclase sustained flotation to the uppermost portion of the LMO.

10

R.O.C. Fonseca et al. / Earth and Planetary Science Letters 404 (2014) 1–13

experiments. Values for orthopyroxene were taken as proxies for those of pigeonite (following Münker, 2010). Plagioclase/melt parplg/melt

tition coefficients for W are taken from run W0-4a (D W of 0.0006), and those for Nb, Zr, Hf, Ta, and Th are average values from runs W0-4a and AT-3 (as they cover the suitable An-content ilm/melt

of lunar plagioclase; Table S2). For ilmenite, D W was estimated from the lattice strain parameters of van Kan Parker et al. (2011) combined with the expected W4+ /W6+ at IW to IW-2 as obtained from our partition coefficients (see Table S4 in supplementary section). All other values for ilmenite are from run HD37 of Klemme et al. (2006). Following Albarède (1976), TIL was crystal/melt

treated as a crystallizing phase having a D M of 1. The bulk LMO composition is taken from Münker (2010), and is extended to U and Th, using chondritic Th/Hf from Dauphas and Pourmand (2011) and associated U/Hf ratio of 0.0765 (defined by all meteorite falls except the strongly deviating CV3 Grosnaja). Fig. 5 displays modeled U/W, Hf/W, Th/W, and Ta/W, evolutions of the crystallizing LMO at f O2 from IW to IW-2 and an assumed average of 3% TIL. At IW-1, modeled U/W, Hf/W, Th/W of the residual LMO after 99.5% solidification agrees well with the composition of KREEP (2.210 ± 0.244, 19.27 ± 3.82, and 8.58 ± 1.85, respectively at 95% confidence, Fig. 5) and the f O2 -insensitive result for Zr/Hf is only 1.8% below the KREEP value (40.5 ± 0.3). For Nb/Ta and Ta/W, the results are ca. 11% and 30% lower than KREEP estimates (22.2 ± 0.5, 2.09 ± 0.33), respectively. A possible explanation is that the model assumes slightly erroneous volumes and mixtures of crystallizing ilmenite/armalcolite-assemblages (cf. Elardo et al., 2011), in which Nb and Ta are particularly compatible. We have also applied our experimental results to LMO evolution models by Elardo et al. (2011); lunar primitive upper mantle composition) and Elkins-Tanton et al. (2011); for further details see Sprung et al. (2013). The three models differ little with respect to the Th/W and U/W (Fig. 5a), as well as the Hf/W (not shown) compositional evolution of the LMO, and KREEP is always well reproduced regardless of the model used. Moreover, the good agreement between the LMO crystallization model of Snyder et al. (1992) with and recent experimental study by Rapp and Draper (2014) on LMO crystallization reinforces our approach. However, the biggest differences are obtained for Nb/Ta and Zr/Hf, namely when using the model by Elardo et al. (2011). Moreover, KREEP Nb/Ta and Zr/Hf are better ilm/melt

Fig. 5. Compositional evolution of the crystallizing lunar magma ocean, assuming 3% of trapped instantaneous liquid in the cumulate sources. (a) Direct comparison of the different LMO fractionation models by Snyder et al. (1992), Elardo et al. (2011) and Elkins-Tanton et al. (2011) with respect to the U/W and Th/W compositional evolution of the LMO at IW-1 (a). (b) Hf/W and U/W, c) Th/W and Ta/W, LMO composition as a function of the degree of crystallization at f O2 between IW and IW-2. ‘KREEP’ defined by samples 14 163, 14 310, 15 382, 15 386, 62 235, and 65 015, using isotope dilution data for Hf, W, Zr, and Ta as well as associated Nb data (Kleine et al., 2005; Lee et al., 1997; Münker, 2010; Münker et al., 2003), U and Th data determined by various methods (Meyer, 2012; Wänke et al., 1977; Warren and Wasson, 1978). See text for discussion. All data included in model curves can be seen in Table S5.

We have repeated this model using our new set of partition coefficients at f O2 from IW to IW-2 (Fig. 5), thereby covering an appropriate range of oxygen fugacities for the lunar mantle (IW-2 to IW-0.5 — Stanin and Taylor, 1980; O’Neill, 1991; Righter and Drake, 1996). Partition coefficients for homovalent trace elements (i.e. HFSE and Th) are average values of all our

reproduced when the D i of Klemme et al. (2006) are used instead of those by Dygert et al. (2013). These results illustrate that any ratios involving the HFSE are model dependent as they are strongly affected by the crystallization of oxide phases such as ilmenite or chromite. Ratios between W, U and Th are thus better proxies to evaluate the W abundance in the bulk silicate moon since they are the least model sensitive. Using our new partition coefficients, the less incompatible behavior of W relative to U and Th on the Moon (e.g., Palme and Rammensee, 1981) is convincingly reproduced for the first time (Fig. 5b, c). Our model results are entirely consistent with lunar data when the identical (with error) bulk lunar mantle Hf/W from Münker (2010) and Touboul et al. (2007) are used (24.9 ± 9.3 and 26.5 ± 1.1 respectively). To substantiate our conclusion, modeled U/W and Hf/W of modal aggregate fractional melts of cumulates (Fig. 6) are compared to lunar sample compositions. Early-formed LMO cumulates (after Snyder et al., 1992) whose compositions match the Hf–Nd isotope systematics of low-Ti mare basalts (Sprung et al., 2013) yield melts consistent with the U/W and Hf/W of Apollo 12 olivine and pigeonite basalts (Fig. 6, trend A). Furthermore, assuming that LMO cumulates lost a tiny melt fraction (e.g. ∼0.1%) prior to the main melting event, modelled melt compositions are strikingly similar to Apollo 12 and 15 low-Ti mare basalts (Fig. 6, trend B). Some of the major competing ideas concerning the petrogenesis of high-Ti mare basalts in the context of a presumed LMO

R.O.C. Fonseca et al. / Earth and Planetary Science Letters 404 (2014) 1–13

11

assimilation of ilmenite and clinopyroxene have been discussed in proportions of 3:1 (Wagner and Grove, 1997) or 1:2 (Van Orman et al., 2002). Such scenarios (e.g. Fig. 6 trend F) appear inconsistent with the prominent trend of most Apollo 17 high-Ti mare basalts (Fig. 6). In contrast, melts from a hybrid source (early-formed olivine–orthopyroxene and 5% ilmenite-bearing cumulate) with either residual metal, and/or having equilibrated at −2 > IW > −1.5 can reproduce the Apollo 17 trends (Fig. 6, trends C, D, E, respectively) and reasonably satisfy the Zr/Hf and Nb/Ta systematics of these rocks. Münker (2010) excluded the possibility that residual metal had limited effect on the behaviour of W in the sources of high-Ti basalts based on positively correlated Sm/Nd and Ta/W and instead favoured a partial assimilation of an ilmenite-rich cumulates. However, Münker (2010) did not consider the possibility that W may decouple from Ta due to variations in f O2 . Both the presence of residual metal and lower f O2 appear reasonable given the evidence for metallic Fe in lunar mantle sources (Delano, 1990; Jones, 2004), as well as the fact that olivine from high-Ti basalts is so much more depleted in Ni when compared to olivine from lowTi basalts (∼20 and ∼400 ppm respectively; Karner et al., 2006). For all three melting scenarios above, the hybrid source of trends C, D, and E (Fig. 6b) provides a good match to the Hf–Nd systematics of Apollo 17 high-Ti mare basalts (see Table S3 of Sprung et al., 2013), bolstering the proposed scenario. 6. Concluding remarks

Fig. 6. (a) Modelled U/W (a) and Th/W (b) versus the Hf/W composition of lunar mare basalts, compared to lunar data. Whenever available, isotope dilution data for Hf and W (Kleine et al., 2005; Lee et al., 1997; Münker, 2010; Münker et al., 2003) or neutron activation data (Wänke et al., 1975; Wänke et al., 1974; Wänke et al., 1970; Wänke et al., 1971) were used. All other data were determined by various techniques (Meyer, 2012; Neal, 2001). Lunar cumulate compositions taken from Snyder et al. (1992). Trajectories A–E: composition of aggregate modal fractional melts for degrees of melting given by tick marks and indicated in %. Source compositions: A: olv–opx–3%TIL + plg–olv–pig–3%TIL cumulates (1:1); B: plg–olv–pig–15%TIL cumulate after having lost 0.1% of melt (QNB stands for quartznormative Apollo 15 low-Ti mare basalts); C–E: olv–opx–1%TIL + pig–plg–ilm–1%TIL cumulates (9:1) at C: IW-1 and 1% residual metal using the average value of W metal–silicate partition coefficients conducted in MgO-capsules (6.2) from Righter et al. (2010); D: IW-1.5 without residual metal; and E: IW-2 without residual metal. F: mixing of ilm–cpx (1:2) from a pig–plg–ilm–3%TIL cumulate with a 5% modal aggregate fractional melt of an olv–opx–3% cumulate illustrating assimilation scenarios. The fraction of assimilated ilm–cpx ranges between 0 and 20%. Using ilm–cpx (3:1) does not change the trajectory significantly as it produces a trend that almost completely overlaps with F. Lunar data and calculated model curves can be seen in Table S5.

(cf. Shearer and Papike, 1999) consist of: (1) The assimilation of late-stage, ilmenite-bearing LMO cumulates by melts similar to low-Ti mare basalts (e.g. Dygert et al., 2013; Münker, 2010; Wagner and Grove, 1997); (2) Melting of mixed cumulate sources (e.g., Beard et al., 1998) juxtaposed by local (Snyder et al., 1992) or global-scale mantle overturn (e.g., Hess and Parmentier, 1995). The

As demonstrated, LMO crystallization and basic melting models reproduce the Hf-U-W-budget of key lunar lithological units, once the f O2 -sensitive partitioning behavior of W and U is accounted for. In light of these findings, our verification of lunar Hf/W estimates gains significance. The silicate Moon and BSE consequently have Hf/W values within <10%. This qualitatively limits the amount of W that could have been sequestered into the cores of both the Earth and the Moon after the separation of the two bodies, likely indicating that the BSE did not fully equilibrate with the Earth’s core during and/or after the giant moon-forming impact. Similar Hf/W values for BSE and the silicate Moon are supported by recent Ti and H isotopic evidence (Zhang et al., 2012). These new data hint that the proto-Earth was a likely source of lunar material (Zhang et al., 2012) and that both bodies likely share the same chondritic heritage (Saal et al., 2013) despite some differences in the 17 O of both bodies (Herwartz et al., 2014). It should be pointed out, however, that the scarcity of highprecision HFSE, W, Th and U data for lunar lithologies limits our current ability to combine these data with the experimental constraints depicted here and elsewhere. In particular, the lack of high-precision isotope dilution data for lunar samples with low W abundances likely obscures the full extent to which W may fractionate from Th, U and the HFSE. Because of this, it is considerably challenging to unravel the petrogenetic processes taking place during the crystallization of the LMO and the subsequent formation of younger lunar lithologies. More high-precision isotope dilution data for lunar materials are thus needed to further constrain the early history and evolution of the Earth–Moon system and the magmatic evolution of the Moon. Acknowledgements We thank Nils Jung, Thomas Schulz, Herbert Phiesel and Dieter Lülsdorf for their outstanding technical support at the Universität Bonn. We are particularly grateful to Renate Schumacher (Mineralogisches Museum, Universität Bonn) for generously providing the natural pitchblende. We are indebted to Hugh O’Neill, Herbert Palme, Chris Ballhaus, and Stefan Peters for their insightful comments and suggestions. We are most grateful to Carsten Münker

12

R.O.C. Fonseca et al. / Earth and Planetary Science Letters 404 (2014) 1–13

for fruitful discussions which helped to shape our modeling. We are thankful for the constructive and helpful comments made by Bernie Wood, Nick Dygert and an anonymous reviewer, as well as input given by two anonymous reviewers on a previous version of this manuscript. We are also grateful to Tim Elliott for his editorial handling. R.F. acknowledges financial support from the Deutsche Forschungsgemeinschaft (via DFG grant FO 698/1-1). G.M. acknowledges financial support from FAPESP (grant 2010/05512-1). P.S. aknowledges funding from the Deutsche Forschungsgemeinschaft via DFG grant SP 1385/1-1. This manuscript is contribution n◦ 11 of the LA-ICPMS laboratory of the Steinmann Institute for Geosciences, University of Bonn. Appendix A. Supplementary material Supplementary material related to this article can be found online at http://dx.doi.org/10.1016/j.epsl.2014.07.015. References Adam, J., Green, T., 2006. Trace element partitioning between mica- and amphibolebearing garnet lherzolite and hydrous basanitic melt: 1. Experimental results and the investigation of controls on partitioning behaviour. Contrib. Mineral. Petrol. 152, 1–17. Aigner-Torres, M., Blundy, J., Ulmer, P., Pettke, T., 2007. Laser ablation ICPMS study of trace element partitioning between plagioclase and basaltic melts: an experimental approach. Contrib. Mineral. Petrol. 153, 647–667. Albaréde, F., 1976. Some trace element relationships among liquid and solid phases in the course of the fractional crystallization of magmas. Geochim. Cosmochim. Acta 40, 667–673. Bali, E., Keppler, H., Audetat, A., 2012. The mobility of W and Mo in subduction zone fluids and the Mo-W-Th-U systematics of island arc magmas. Earth Planet. Sci. Lett. 352, 195–207. Beard, B.L., Taylor, L.A., Scherer, E.E., Johnson, C.M., Snyder, G.A., 1998. The source region and melting mineralogy of high-titanium and low-titanium lunar basalts deduced from Lu–Hf isotope data. Geochim. Cosmochim. Acta 62, 525–544. Beattie, P., 1993. The generation of uranium series disequilibria by partial melting of spinel peridotite: constraints from partitioning studies. Earth Planet. Sci. Lett. 117, 379–391. Bence, A.E., Grove, T.L., Papike, J.J., 1980. Basalts as probes of planetary interiors: constraints on the chemistry and mineralogy of their source regions. Precambrian Res. 10, 249–279. Berry, A.J., O’Neill, H.St.C., Foran, G.J., 2008. The oxidation state of uranium in mantle melts. Geochim. Cosmochim. Acta 72, A79. Bindeman, I., 2007. Erratum to I.N. Bindeman, A.M. Davis, and M.J. Drake (1998), “Ion microprobe study of plagioclase–basalt partitioning experiments at natural concentration levels of trace elements. Geochim. Cosmochim. Acta. 62, 1174–1192”. Geochim. Cosmochim. Acta 71, 2414. Bindeman, I.N., Davis, A.M., 2000. Trace element partitioning between plagioclase and melt: investigation of dopant influence on partitioning behavior. Geochim. Cosmochim. Acta 64, 2863–2878. Blundy, J., Wood, B.J., 1991. Crystal-chemical controls on the partitioning of Sr and Ba between plagioclase feldspar, silicate melts and hydrothermal solutions. Geochim. Cosmochim. Acta 55, 193–209. Blundy, J., Wood, B.J., 1994. Prediction of crystal–melt partition coefficients from elastic moduli. Nature 372, 452–454. Blundy, J., Wood, B.J., 2003. Mineral–melt partitioning of uranium, thorium and their daughters. Rev. Mineral. Geochem. 52, 59–123. Blundy, J.D., Robinson, J.A.C., Wood, B.J., 1998. Heavy REE are compatible in clinopyroxene on the spinel lherzolite solidus. Earth Planet. Sci. Lett. 160, 493–504. Borisov, A., 2012. The Ti4+ /Ti3+ ratio of magmatic melts: application to the problem of reduction of Lunar Basalts. Petrology 20, 429–436. Burnham, A.D., Berry, A.J., Wood, B.J., Cibin, G., 2012. The oxidation states of niobium and tantalum in mantle melts. Chem. Geol. 331, 228–232. Calas, G., 1979. Etude Expérimentale du comportement de l’uranium dans les magmas. États d’oxydation et coordinances. Geochim. Cosmochim. Acta 43, 1521–1531. Cottrell, E., Walter, M.J., Walker, D., 2009. Metal–silicate partitioning of tungsten at high pressure and temperature: implications for equilibrium core formation in Earth. Earth Planet. Sci. Lett. 281, 275–287. Dauphas, N., Pourmand, A., 2011. Hf–W–Th evidence for rapid growth of Mars and its status as a planetary embryo. Nature 473, 489–492. De Vries, J., van Westrenen, W., van den Berg, A., 2012. Radiogenic heat production in the Moon: constraints from plagioclase–melt trace element partitioning experiments. Proc. Lunar Planet. Sci. Conf. 43, 1737. Delano, J.W., 1986. Pristine lunar glasses: criteria, data, and implications. J. Geophys. Res., Solid Earth 91 (B4), 201–213.

Delano, J.W., 1990. Experimental constraints on the oxidation state of the lunar mantle. In: Abstracts of the Lunar and Planetary Science Conference, vol. 20, pp. 278–279. Dunn, T., Sen, C., 1994. Mineral/matrix partition coefficients for orthopyroxene, plagioclase, and olivine in basaltic to andesitic systems: a combined analytical and experimental study. Geochim. Cosmochim. Acta 58, 717–733. Dygert, N., Liang, Y., Hess, P., 2013. The importance of TiO2 in affecting major and trace element partitioning between Fe–Ti oxides and lunar picritic glass melts. Geochim. Cosmochim. Acta 106, 134–151. Dygert, N., Liang, Y., Sun, C., Hess, P., 2014. An experimental study of trace element partitioning between augite and Fe-rich basalts. Geochim. Cosmochim. Acta 132, 170–186. Elardo, S.M., Draper, D.S., Shearer, C.K., 2011. Lunar magma ocean crystallization revisited: bulk composition, early cumulate mineralogy, and the source regions of the highlands Mg-suite. Geochim. Cosmochim. Acta 75, 3024–3045. Elkins-Tanton, L.T., Burgess, S., Yin, Q.Z., 2011. The lunar magma ocean: reconciling the solidification process with lunar petrology and geochronology. Earth Planet. Sci. Lett. 304, 326–336. Evans, T.M., O’Neill, H.St.C., Tuff, J., 2008. The influence of melt composition on the partitioning of REEs, Y, Sc, Zr and Al between forsterite and melt in the system CMAS. Geochim. Cosmochim. Acta 72, 5708–5721. Frei, D., Liebscher, A., Franz, G., Wunder, B., Klemme, S., Blundy, J., 2009. Trace element partitioning between orthopyroxene and anhydrous silicate melt on the lherzolite solidus from 1.1 to 3.2 GPa and 1230 to 1535 ◦ C in the model system Na2 O–CaO–MgO–Al2 O3 –SiO2 . Contrib. Mineral. Petrol. 157, 473–490. Herwartz, D., Pack, A., Friedrichs, B., Bischoff, A., 2014. Identification of the giant impactor Theia in lunar rocks. Science 344, 1146–1150. Hess, P.C., Parmentier, E.M., 1995. A model for the thermal and chemical evolution of the Moon’s interior: implications for the onset of mare volcanism. Earth Planet. Sci. Lett. 134, 501–514. Hill, E., Wood, B.J., Blundy, J.D., 2000. The effect of Ca-Tschermarks component on trace element partitioning between clinopyroxene and silicate melt. Lithos 53, 203–215. Hill, E., Blundy, J.D., Wood, B.J., 2011. Clinopyroxene–melt trace element partitioning and the development of a predictive model for HFSE and Sc. Contrib. Mineral. Petrol. 161, 423–438. Hill, E., Blundy, J.D., Wood, B.J., 2012. Erratum to: Clinopyroxene–melt trace element partitioning and the development of a predictive model for HFSE and Sc. Contrib. Mineral. Petrol. 163, 563–565. Jochum, K.P., Weis, U., Stoll, B., Kuzmin, D., Yang, Q., Raczek, I., Jacob, D.E., Stracke, A., Birbaum, K., Frick, D.A., Günther, D., Enzweiler, J., 2011. Determination of reference values for NIST SRM 610-617 glasses following ISO guidelines. Geostand. Geoanal. Res. 35, 397–429. Jones, J.H., 2004. Redox conditions among the terrestrial planets. In: 35th Lunar and Planetary Science Conference. Abstract #1264. Karner, J.M., Sutton, S.R., Papike, J.J., Shearer, C.K., Jones, J.H., Newville, M., 2006. Application of a new vanadium valence oxybarometer to basaltic glasses from the Earth, Moon, and Mars. Am. Mineral. 91, 270–277. Kennedy, A.K., Logren, G.E., Wasserburg, G.J., 1993. An experimental study of trace element partitioning between olivine, orthopyroxene and melt in chondrules: equilibrium values and kinetic effects. Earth Planet. Sci. Lett. 115, 177–195. Kleine, T., Münker, C., Mezger, K., Palme, H., 2002. Rapid accretion and early core formation on asteroids and early core formation on asteroids and the terrestrial planets from Hf–W chronometry. Nature 418, 952–955. Kleine, T., Palme, H., Mezger, K., Halliday, A.N., 2005. Hf–W chronometry of lunar metals and the age and early differentiation of the Moon. Science 310, 1671–1674. Klemme, S., Gunther, D., Hametner, K., Prowatke, S., Zack, T., 2006. The partitioning of trace elements between ilmenite, ulvospinel, armalcolite and silicate melts with implications for the early differentiation of the moon. Chem. Geol. 234, 251–263. König, S., Münker, C., Hohl, S., Paulick, H., Barth, A.R., Lagos, M., Pfänder, J., Büchl, A., 2011. The Earth’s tungsten budget during mantle melting and crust formation. Geochim. Cosmochim. Acta 75, 2119–2136. Kress, V., Ghiorso, M., Lastuka, C., 2004. Microsoft EXCEL spreadsheet-based program for calculating equilibrium gas speciation in the C–O–H–S–Cl–F system. Comput. Geosci. 30, 211–214. LaTourrette, T.Z., Burnett, D.S., 1992. Experimental determination of U and Th partitioning between clinopyroxene and natural and synthetic basaltic liquid. Earth Planet. Sci. Lett. 110, 227–244. Laurenz, V., Fonseca, R.O.C., Ballhaus, C., Sylvester, P.J., 2010. Solubility of palladium in picritic melts: 1. The effect of iron. Geochim. Cosmochim. Acta 74, 2989–2998. Lee, D.C., Halliday, A.N., Snyder, G.A., Taylor, L.A., 1997. Age and origin of the Moon. Science 278, 1098–1103. Longerich, H.P., Jackson, S.E., Günther, D., 1996. Laser ablation inductively coupled plasma mass spectrometric transient signal data acquisition and analyte concentration calculation. J. Anal. At. Spectrom. 11, 899–904. Lundstrom, C.C., Shaw, H.F., Ryerson, F.J., Phinney, D.L., Gill, J.B., Williams, Q., 1994. Compositional controls on the partitioning of U, Th, Ba, Pb, Sr and Zr between

R.O.C. Fonseca et al. / Earth and Planetary Science Letters 404 (2014) 1–13

clinopyroxene and haplobasaltic melts: implication for uranium series disequilibria in basalts. Earth Planet. Sci. Lett. 128, 407–423. Mallmann, G., O’Neill, H.St.C., 2007. The effect of oxygen fugacity on the partitioning of Re between crystals and silicate melt during mantle melting. Geochim. Cosmochim. Acta 71, 2837–2857. Mallmann, G., O’Neill, H.St.C., 2009. The crystal/melt partitioning of V during mantle melting as a function of oxygen fugacity compared with some other elements (Al, P, Ca, Sc, Ti, Cr, Fe, Ga, Y, Zr and Nb). J. Petrol. 50, 1765–1794. Mallmann, G., O’Neill, H.St.C., 2013. Calibration of an empirical thermometer and oxybarometer based on the Partitioning of Sc, Y and V between olivine and silicate melt. J. Petrol. 54, 933–949. Mallmann, G., O’Neill, H.St.C., Klemme, S., 2009. Heterogeneous distribution of phosphorus in olivine from otherwise well-equilibrated spinel peridotites xenoliths and its implications for the mantle geochemistry of lithium. Contrib. Mineral. Petrol. 158, 485–504. McDade, P., Blundy, J.D., Wood, B.J., 2003. Trace element partitioning on the Tinaquillo Lherzolite solidus at 1.5 GPa. Phys. Earth Planet. Inter. 139, 129–147. Meyer, C., 2012. Lunar sample compendium. Available at http://curator.jsc.nasa.gov/ lunar/compendium.cfm. Münker, C., 2010. A high field strength element perspective on early lunar differentiation. Geochim. Cosmochim. Acta 74, 7340–7361. Münker, C., Pfänder, J.A., Weyer, S., Büchl, A., Kleine, T., Mezger, K., 2003. Evolution of planetary cores and the Earth–Moon system from Nb/Ta systematics. Science 301, 84–87. Neal, C.R., 2001. Interior of the Moon: the presence of garnet in the primitive deep lunar mantle. J. Geophys. Res. 106, 27865–27885. Newson, H.E., Sims, K.W., Noll Jr., P.-D., Jaeger, W.L., Maehr, S.A., Beserra, T.B., 1996. The depletion of tungsten in the bulk silicate earth: constraints on core formation. Geochim. Cosmochim. Acta 60, 1155–1169. O’Neill, H.St.C., 1991. The origin of the Moon and the early history of the Earth – a chemical model. Part 1: The Moon. Geochim. Cosmochim. Acta 55, 1135–1157. O’Neill, H.St.C., Eggins, S.M., 2002. The effect of melt composition on trace element partitioning: an experimental investigation on the activity coefficients of FeO, NiO, CoO, MoO2 and MoO3 . Chem. Geol. 186, 151–181. O’Neill, H.St.C., Berry, A.J., Eggins, S.M., 2008. The solubility and oxidation state of tungsten in silicate melts: implications for the comparative chemistry of W and Mo in planetary differentiation processes. Chem. Geol. 255, 346–359. Palme, H., Rammensee, W., 1981. The significance of W in planetary differentiation processes: evidence from new data on eucrites. Proc. Lunar Planet. Sci. Conf. 12, 949–964. Palme, H., Wänke, H., 1975. A unified trace-element model for the evolution of the lunar crust and mantle. In: Proc. Lunar Sci. Conf. 6th, pp. 1179–1202. Rapp, J.P., Draper, D.S., 2014. The lunar magma ocean: sharpening the focus on process and composition. In: Proc. Lunar Planet. Sci. Conf. 45 Abstract #1527. Righter, K., Drake, M.J., 1996. Core formation in Earth’s Moon, Mars, and Vesta. Icarus 124, 512–529. Righter, K., Shearer, C.K., 2003. Magmatic fractionation of Hf and W: constraints on the timing of core formation and differentiation in the Moon and Mars. Geochim. Cosmochim. Acta 67, 2497–2507. Righter, K., Pando, K.M., Danielson, L., Lee, C.T., 2010. Partitioning of Mo, P and other siderophile elements (Cu, Ga, Sn, Ni, Co, Cr, Mn, V, and W) between metal and silicate melt as a function of temperature and silicate melt composition. Earth Planet. Sci. Lett. 291, 1–9. Rocholl, A., Jochum, K.P., 1993. Th, U and other trace elements in carbonaceous chondrites: implications for the terrestrial and solar-system Th/U ratios. Earth Planet. Sci. Lett. 117, 265–278. Saal, A.E., Hauri, E.H., Van Orman, J.A., Rutherford, M.J., 2013. Hydrogen isotopes in lunar volcanic glasses and melt inclusions reveal a carbonaceous chondrite heritage. Science 340, 1317–1320. Schuth, S., Rohrbach, A., Münker, C., Ballhaus, C., Garbe-Schönberg, C., Qopoto, C., 2004. Geochemical constraints on the petrogenesis of arc picrites and basalts, New Georgia Group, Solomon Islands. Contrib. Mineral. Petrol. 148, 288–304. Shannon, R.D., 1976. Revised effective ionic radii and systematic studies of interatomic distances in halides and chalcogenides. Acta Crystallogr. A, Found. Crystallogr. 32, 751–767. Shearer, C.K., Papike, J.J., 1999. Magmatic evolution of the Moon. Am. Mineral. 84, 1469–1494. Smith, J.V., Anderson, A.T., Newton, R.C., Olsen, E.J., Wyllie, P.J., 1970. A petrologic model for Moon based on petrogenesis, experimental petrology, and physical properties. J. Geol. 78, 381–405. Snyder, G.A., Taylor, L.A., Neal, C.R., 1992. A chemical-model for generating the sources of mare basalts – combined equilibrium and fractional crystallization of the lunar magmasphere. Geochim. Cosmochim. Acta 56, 3809–3823. Snyder, G.A., Lee, D.C., Taylor, L.A., Halliday, A.N., Jerde, E.A., 1994. Evolution of the upper-mantle of the Earth’s Moon – neodymium and strontium isotopic constraints from high-Ti mare basalts. Geochim. Cosmochim. Acta 58, 4795–4808. Snyder, G.A., Neal, C.R., Taylor, L.A., Halliday, A.N., 1997. Anatexis of lunar cumulate mantle in time and space: clues from trace-element, strontium, and neodymium isotopic chemistry of parental Apollo 12 basalts. Geochim. Cosmochim. Acta 61, 2731–2747.

13

Snyder, G.A., Borg, L.E., Nyquist, L.E., Taylor, L.A., 2000. Chronology and isotopic constraints on lunar evolution. In: Canup, R., Righter, K. (Eds.), Origin of the Earth and Moon. University of Arizona Press, pp. 361–398. Sprung, P., Kleine, T., Scherer, E.E., 2013. Isotopic evidence for chondritic Lu/Hf and Sm/Nd of the Moon. Earth Planet. Sci. Lett. 380, 77–87. Stanin, F.T., Taylor, L.A., 1980. Armalcolite: an oxygen fugacity indicator. Proc. Lunar Planet. Sci. Conf. 11, 117–124. Tepley, F.J., Lundstrom, C.C., McDonough, W.F., Thompson, A., 2010. Trace element partitioning between high-An plagioclase and basaltic to basaltic andesite melt at 1 atmosphere pressure. Lithos 118, 82–94. Touboul, M., Kleine, T., Bourdon, B., Palme, H., Wieler, R., 2007. Late formation and prolonged differentiation of the Moon inferred from W isotopes in lunar metals. Nature 450, 1206–1209. Touboul, M., Kleine, T., Bourdon, B., Palme, H., Wieler, R., 2009. Tungsten isotopes in ferroan anorthosites: Implications for the age of the Moon and lifetime of its magma ocean. Icarus 199, 245–249. van Kan Parker, M., Liebscher, A., Frei, D., van Sikl, J., van Westrenen, W., Blundy, J., Franz, G., 2010. Experimental and computational study of trace element distribution between orthopyroxene and anhydrous silicate melt: substitution mechanisms and the effect of iron. Contrib. Mineral. Petrol. 159, 459–473. van Kan Parker, M., Mason, P.R.D., van Westrenenen, W., 2011. Trace element partitioning between ilmenite, armalcolite and anhydrous silicate melt: implications for the formation of lunar high-Ti mare basalts. Geochim. Cosmochim. Acta 75, 4179–4193. van Kan Parker, M., Mason, P.R.D., van Westrenenen, W., 2012. Experimental study of trace element partitioning between lunar orthopyroxene and anhydrous silicate melt: effects of lithium and iron. Chem. Geol. 285, 1–14. Van Orman, J.A., Grove, T.L., Shimizu, N., Layne, G.D., 2002. Rare earth element diffusion in a natural pyrope single crystal at 2.8 GPa. Contrib. Mineral. Petrol. 142, 416–424. Wade, J., Wood, B.J., Tuff, J., 2012. Metal–silicate partitioning of Mo and W at high pressures and temperatures: evidence for late accretion of sulphur to the Earth. Geochim. Cosmochim. Acta 85, 58–74. Wade, J., Wood, B.J., Norris, A., 2013. The oxidation state of tungsten in silicate melt at high pressures and temperatures. Chem. Geol. 335, 189–193. Wadhwa, M., 2008. Redox conditions on small bodies, the Moon and Mars. Rev. Mineral. Geochem. 68, 493–506. Wagner, T.P., Grove, T.L., 1997. Experimental constraints on the origin of lunar highTi ultramafic glasses. Geochim. Cosmochim. Acta 61, 1315–1327. Wänke, H., Rieder, R., Baddenhausen, H., Spettel, B., Teschke, F., Quijano-Rico, M., Balacescu, A., 1970. Major and trace elements in lunar material. In: Proceedings of the Apollo 11 Lunar Sci. Conf., vol. 2, pp. 1719–1727. Wänke, H., Wlotzka, F., Baddenhausen, H., Balacescu, A., Spettel, B., Teschke, F., Jagoutz, E., Kruse, H., 1971. Apollo 12 samples: chemical composition and its relation to sample locations and exposure ages, the two-component origin of the various soil samples and studies on lunar metallic particles. In: Proc. Lunar Planet. Sci. Conf., vol. 2, pp. 1187–1208. Wänke, H., Palme, H., Baddenhausen, H., Dreibus, G., Jagoutz, E., Kruse, H., Spettel, B., Teschke, F., Thacker, R., 1974. Chemistry of Apollo 16 and 17 samples: Bulk composition, late stage accumulation and early differentiation of the moon. In: Proc. Lunar Sci. Conf., vol. 5, pp. 1307–1335. Wänke, H., Palme, H., Baddenhausen, H., Dreibus, G., Jagoutz, E., Kruse, H., Palme, C., Spettel, B., Teschke, F., Thacker, R., 1975. New data on the chemistry of lunar samples: primary matter in the lunar highlands and the bulk composition of the moon. In: Proc. Lunar Planet. Sci. Conf., vol. 6, pp. 1313–1340. Wänke, H., Kruse, H., Palme, H., 1977. Instrumental analysis of lunar-samples and identification of primary matter in lunar highlands. J. Radioanal. Chem. 38, 391–403. Warren, P.H., 1985. The magma ocean concept and lunar evolution. Annu. Rev. Earth Planet. Sci. 13, 201–240. Warren, H.P., Wasson, J.T., 1978. Compositional petrographic investigation of pristine nonmare rocks. In: Proc. Lunar Planet. Sci. Conf., vol. 9, pp. 185–217. Warren, P.H., Wasson, J.T., 1979. The origin of KREEP. Rev. Geophys. 17, 73–88. Wood, B.J., Blundy, J.D., 2001. The effect of cation charge on crystal/melt partitioning of trace elements. Earth Planet. Sci. Lett. 188, 59–71. Wood, B.J., Trigila, R., 2001. Experimental determination of aluminous clinopyroxene-melt partition coefficients for potassic liquids, with application to the evolution of the Roman province potassic magmas. Chem. Geol. 172, 213–223. Wood, J.A., Dickey, J.S., Marvin, U.B., Powell, B.N., 1970. Lunar anorthosites and a geophysical model of the Moon. In: Proceedings of the Apollo 11 Lunar Science Conference, vol. 1, pp. 965–988. Wood, B.J., Blundy, J.D., Robinson, J.A.R., 1999. The role of pyroxene in generating U-series disequilibrium during mantle melting. Geochim. Cosmochim. Acta 63, 1613–1620. Yin, Q., Jacobsen, S.B., Yamashita, K., Blichert-toft, J., Télouk, P., Albarède, F., 2002. A short timescale for terrestrial planet formation from Hf–W chronometry of meteorites. Nature 418, 949–952. Zhang, J., Dauphas, N., Davis, A.M., Leya, I., Fedkin, A., 2012. The proto-Earth as a significant source of lunar material. Nature Geosci. 5, 251–255.