H and Cl isotope systematics of apatite in brecciated lunar meteorites Northwest Africa 4472, Northwest Africa 773, Sayh al Uhaymir 169, and Kalahari 009

We have investigated the H and Cl systematics in apatite from four brecciated lunar meteorites. In Northwest Africa (NWA) 4472, most of the apatites contain ∼2000–6000 ppm H2O with δD between −200 and 0‰, except for one grain isolated in the matrix, which contains ∼6000 ppm H2O with δD of ∼500–900‰. This low‐δD apatite contains ∼2500–7500 ppm Cl associated with δ37Cl of ∼15–20‰, while the high‐δD grain contains ∼2500 ppm Cl with δ37Cl of ∼7–15‰. In NWA 773, apatites in a first group contain ∼700–2500 ppm H2O with δD values averaging around ∼0 ± 100‰, while apatites in a second group contain ∼5500–16500 ppm H2O with δD ∼250 ± 50‰. In Sayh al Uhaymir (SaU) 169 and Kalahari (Kal) 009, apatites are similar in terms of their H2O contents (∼600–3000 ppm) and δD values (−100 to 200‰). In SaU 169, apatites contain ∼6000–10,000 ppm Cl, characterized by δ37Cl of ∼5–12‰. Overall, most of the analyzed apatite grains have δD within the range reported for carbonaceous chondrites, similar to apatite analyzed in ancient (>3.9 Ga) lunar magmatic. One grain in NWA 4472 has H and Cl isotope compositions similar to apatite from mare basalts. With an age of 4.35 Ga, this grain could be a representative of the oldest known lunar volcanic activity. Finally, since numerous evolved clasts in NWA 773 formed through silicate liquid immiscibility, the apatite grains with extremely high H2O contents, reaching pure hydroxylapatite composition, could provide insights into the effects of such process on the evolution of volatiles in lunar magmas.


INTRODUCTION
Over the past decade, the long-standing paradigm of an anhydrous Moon has been challenged based on new measurements on lunar samples utilizing recent analytical developments. These measurements have revealed significant quantities of water, tens to thousands of parts per million (ppm) equivalent H 2 O, in lunar volcanic glasses, in olivine-hosted melt inclusions within these glasses (Saal et al. 2008(Saal et al. , 2013Hauri et al. 2011), in nominally anhydrous minerals from the lunar highlands (Hui et al. 2013), and in the mineral apatite from nearly all type of lunar lithologies (Boyce et al. 2010;McCubbin et al. 2010aMcCubbin et al. , 2010bMcCubbin et al. , 2011Greenwood et al. 2011;Barnes et al. 2013Barnes et al. , 2014Tart ese et al. 2013Tart ese et al. , 2014aTart ese et al. , 2014b. Although the utility of abundance measurements for OH, F, and Cl in lunar apatite for constraining the volatile abundances in their parental magmas is currently debated (e.g., Boyce et al. 2014), the H and Cl isotopic compositions of apatite still provide clues regarding the origin of lunar volatiles and the influence of any petrogenetic processes in modifying the original signatures. Therefore, a significant database for the H isotopic composition of lunar apatite has emerged recently (Greenwood et al. 2011;Barnes et al. 2013Barnes et al. , 2014Tart ese et al. 2013Tart ese et al. , 2014aTart ese et al. , 2014b, and a few studies have also initiated investigations of Cl isotopic composition of lunar apatite (e.g., Sharp et al. 2010a;Boyce et al. 2013;Treiman et al. 2014). Based on data available so far, the combined H and Cl isotopic compositions of lunar apatite indicate that apatite in mare basalts generally show dD>~500& and d 37 Cl % 0-15&, while apatite from the KREEP-rich Mg-and alkali-suite lithologies from the lunar highlands have comparatively lower dD values (<~300&), but elevated d 37 Cl values (>~25&).
Phosphates, especially apatite, have also been a target for U-Pb and Pb/Pb geochronological investigations in lunar meteorites, notably to constrain the crystallization ages of basaltic meteorites, where apatite is a common late-stage phase (Anand et al. 2003(Anand et al. , 2006Terada et al. 2005Terada et al. , 2007aTerada et al. , 2007bTerada et al. , 2008Joy et al. 2011Joy et al. , 2014. The wide range of crystallization ages, from approximately 4.35 Ga to approximately 2.9 Ga, obtained for lunar basaltic meteorites has expanded the range for the timing of basaltic volcanism on the Moon as previously determined on basaltic samples from the Apollo collection (approximately 4.2-3.1 Ga; see the review of Shearer et al. [2006] and references therein). Yet interpretation of apatite U-Pb and Pb/Pb dates can be equivocal; Pb isotopes can be disturbed in apatite at moderate temperatures of approximately 500°C (Cherniak et al. 1991), which makes the apatite U-Pb and Pb/Pb chronometers highly sensitive to reheating during impact events. As a result, the interpretation of dates obtained on isolated apatite grains found in the matrix of breccias might sometimes be complex (e.g., Joy et al. 2011Joy et al. , 2014 as they lack petrological context. In such cases, additional constraints are required to decipher the source lithology and the petrogenetic history of the dated mineral grains. In this study, we have investigated the H isotope systematics of apatite from four hot desert lunar meteorites (NWA 4472,NWA 773,SaU 169,and Kal 009). We have also measured the Cl isotopic composition of apatite from two of these meteorites (NWA 4472 and SaU 169). Two of these meteorites are regolith breccias (NWA 4472 and NWA 773), one is an impact-melt breccia (SaU 169), and the fourth one is a fragmental basaltic breccia (Kalahari 009). Geochronological data obtained on these meteorites cover the entire range of crystallization ages of basaltic meteorites (Fernandes et al. 2003;Gnos et al. 2004;Terada et al. 2007a;Sokol et al. 2008;Borg et al. 2009;Joy et al. 2011;Liu et al. 2012). Most of the studied apatite grains lack a full petrographic context (i.e., they are not associated with any other minerals), preventing us from relating them to different lithologies observed in these breccias. This study, therefore, aims to investigate how volatile abundances and the H and Cl isotope characteristics of apatite grains lacking petrographic context can be used to further decipher their source lithology. This approach has only become feasible because of the availability of a wealth of recent H and Cl data obtained on various lithologies represented in the Apollo sample collection (i.e., high-and low-Ti mare basalts, KREEP basalts, Mgand alkali-suite magmatic rocks). We also use the volatile abundances of the studied apatite grains, and their isotopic characteristics, to investigate the petrogenetic processes that could have affected their parental melt evolution (e.g., degassing of volatiles from magmas, silicate liquid immiscibility).

LUNAR METEORITES
The lunar meteorites investigated in this work have been described for their mineralogical and petrological characteristics in sufficient detail in several previous studies. Therefore, the following sections summarize their main petrographic and geochemical characteristics and refer to relevant previous work as appropriate.

Northwest Africa 4472
Northwest Africa 4472 is a $ 65 g polymict lunar meteorite paired with NWA 4485 (Connolly et al. 2007;Arai et al. 2009Arai et al. , 2010Korotev et al. 2009). In this study we have investigated a 7.1 9 7.5 mm polished section of NWA 4472 (Figs. 1a and 2a; see also Figs. S1 and S2), which is part of the larger slab studied by Joy et al. (2011). These authors described the mineralogical, petrographic, geochemical, and geochronological characteristics of NWA 4472. The sample is a KREEPrich regolith breccia comprising heterogeneous lithic clasts, mineral fragments, and spherules of impact glass, which have been linked to various lunar lithologies such as the Mg-suite, the alkali-suite, KREEP basalts, and mare basalts. Based on the elevated bulk-KREEP content of NWA 4472 and orbital remote sensing data obtained by the Lunar Prospector gamma-ray instrument, Joy et al. (2011) proposed that NWA 4472 was derived from the lunar regolith on the nearside of the Moon, in the vicinity of the Imbrium impact basin in the Procellarum KREEP Terrane (PKT).
Based on U-Pb and Pb/Pb dating of phosphates and zircons in NWA 4472 and NWA 4485 that yielded some old Pre-Nectarian (4.35 Ga) dates, it has been suggested that some fragments in this breccia could be related to early episodes of KREEP-driven magmatism, while younger U-Pb and Pb/Pb dates around 3.9-4.0 Ga could either date younger episodes of KREEPy magmatism or represent resetting dates related to impact events (Arai et al. 2010;Joy et al. 2011), possibly the Imbrium basin-forming event itself.

Sayh al Uhaymir 169
Sayh al Uhaymir 169 is a $ 206 g lunar meteorite composed of two main lithologies; a regolith breccia   2. Composite X-ray element maps of the studied lunar meteorites. Color scheme: Al = white; Mg = green; Fe = red; Si = blue; Ca = yellow; Ti = pink; K = cyan. Generally, Mg-rich olivine and pyroxene appear green-brown, Fe-rich olivine appears red, plagioclase appears white, K-feldspar and K-rich glass appears cyan, silica appears blue, ilmenite appears pink, and Ca-phosphates appear yellow. Fractures also often appear yellow due to terrestrial filling by calcite. (see online version for color figure.) (approximately 13 vol%) and an impact-melt breccia (approximately 87 vol%), displaying sharp contacts and characterized by noticeably different compositions (Gnos et al. 2004;Al-Kathiri et al. 2007;Korotev et al. 2009;Lin et al. 2012). In this study, we have investigated a 17 9 15 mm polished section of the impact-melt breccia (IMB) lithology (Figs. 1b and 2b; see also Figs. S3 and S4). The IMB contains a large proportion of mineral and lithic fragments embedded in a fine-grained crystalline matrix. The IMB of SaU 169 has very high concentrations of incompatible trace elements (ITE), at approximately 300 to 1000 times CIchondrite abundances and $ 1.3-1.8 times higher than average ITE contents of Apollo 14 high-K KREEP (Warren and Wasson 1979;Warren 1989;Korotev et al. 2009). Lin et al. (2012) have recently described a new type of lithic clast in SaU 169 IMB that they termed VHK (very high potassium) KREEP, which comprises Ca-poor and Ca-rich pyroxenes, Ba-rich K-feldspar, phosphates (merrillite and apatite), ilmenite, and zircon. Together with the Apollo 12 high-Th IMB fragments described by Korotev et al. (2011), this VHK KREEP lithology in SaU 169 is the most ITE-rich lunar mafic lithology identified to date, being 2-3 times more enriched in KREEP component compared to the bulk IMB of SaU 169 (Lin et al. 2012). VHK KREEP clasts are easily identified in the section we investigated, as they appear as K-enriched cyan colored pixels (Fig. 2b). The SaU 169 IMB is also characterized by Th/REE and U/REE ratios of about 0.9 times those measured in the mafic Apollo 14 high-K KREEP IMB, and by Eu/Sm and Ba/Sm ratios of only approximately 0.8 and 0.6 times those of Apollo 14 high-K KREEP IMB (Korotev et al. 2009). According to Korotev et al. (2009), such differences indicate that the IMB in SaU 169 lacks granite/felsite components (high Th/REE and Ba/REE compared to KREEP) and alkali anorthosite components (high Eu/Sm compared to KREEP) that are found in Apollo 12 and 14 IMBs (e.g., Jolliff et al. 1991;Jolliff 1998).
Zircon U-Pb and Pb/Pb dating in the IMB of SaU 169 yielded a major age peak at $ 3.91-3.92 Ga (Gnos et al. 2004;Lin et al. 2012;Liu et al. 2012) that has been interpreted as the likely age of the Imbrium basinforming impact event (Gnos et al. 2004;Liu et al. 2012).

Northwest Africa 773
Northwest Africa 773 is a $ 633 g mafic lunar meteorite composed of two main lithologies; a dark regolith breccia and a greenish-olivine cumulate (OC) lithology, which is likely a large clast in the breccia (Fagan et al. 2003(Fagan et al. , 2014Jolliff et al. 2003 Figs. S5 and S6). In this study, we have investigated a 5.9 9 7.7 mm polished section of NWA 773 mostly composed of the brecciated lithology (Figs. 1d and 2d; see also Figs. S7 and S8). The OC lithology of NWA 773 displays coarse-grained (mm sized) olivine crystals, low-and high-Ca pyroxene, and plagioclase with minor amounts of ilmenite (Figs. 1c and 2c). Apatite grains analyzed by Tart ese et al. (2014a) were found in intercumulus pockets of latestage melt trapped between major mineral phases, easily identified in Fig. 2c by the ubiquitous presence of latecrystallized K-feldspar. The OC lithology also constitutes the top and bottom edges of the polished section of the brecciated part of NWA 773 we have studied (Figs. 1d and 2d). The polymict breccia lithology contains mineral and lithic clasts (Fig. 2d) and tends to be much more ferroan than the OC lithology. In addition to smaller OC clasts, the main lithic clasts contained within the brecciated lithology have been classified as pyroxene-gabbro clasts (PG), symplectite (hedenbergite + fayalite + SiO 2 ) clasts (S) and alkali-phase-ferroan clasts (AF) by Fagan et al. (2014), who have argued that these four types of clasts (OC, PG, S, and AF) are co-genetic, and that S and AF clasts could correspond to the Fe-rich and Si-rich liquid fractions, respectively, that formed after silicate liquid immiscibility (SLI). In addition, some exotic clasts have also been observed by Fagan et al. (2003Fagan et al. ( , 2014. The OC and the breccia lithology of NWA 773 are reported to be moderately enriched in ITE and display Apollo KREEP-like REE patterns, with La abundances of approximately 30-70 9 CI chondrites (Fagan et al. 2003).
While a precise crystallization age is still unclear, the OC lithology within NWA 773 is young. Ar-Ar dating indicated closure to most Ar diffusion at $ 2.91 Ga (Fernandes et al. 2003). Subsequent Sm-Nd dating yielded a whole-rock and mineral separate date of $ 3.0 Ga (Borg et al. 2009), and more recent in-situ Pb/Pb dating of baddeleyite yielded dates of $ 3.1 Ga (Zhang et al. 2010;Shaulis et al. 2013).
Phosphate U-Pb dating and Lu-Hf dating yielded dates of approximately 4.3 Ga, interpreted as the crystallization age of the basaltic component of Kal 009 (Terada et al. 2007a;Sokol et al. 2008), making Kal 009 one of the oldest known products of basaltic volcanism from the Moon.

Scanning Electron Microscopy and Electron Probe Micro-Analysis
The studied polished block and sections were carbon coated and examined using a Quanta 3D focused ion beam scanning electron microscope (FIB SEM) at the Open University, fitted with an Oxford Instruments INCA energy dispersive X-ray detector. A 0.6 nA electron beam with a 20 kV accelerating voltage was used. To locate phosphate grains in these sections, BSE images and X-ray maps of the whole sections were obtained by energy dispersive spectroscopy (EDS). Phosphates were located using P X-ray maps of the sections. High magnification BSE imaging and quick acquisition of X-ray spectra for approximately 10 s permitted distinction between apatite and merrillite without generating electron beam-induced volatile mobility in apatite (e.g., Barnes et al. 2013). Semiquantitative major element compositions of selected mineral phases were measured for 30 s by EDS with an Oxford Instruments 80 mm X-Max detector (Table S1).
Chemical composition of apatite was determined by electron probe micro-analysis (EPMA) at the Open University for sample NWA 773 and at the Natural History Museum (NHM) in London for sample NWA 4472. At the Open University, apatite analyses were carried out using a Cameca SX100 electron microprobe using a 10 kV accelerating voltage, a 4 nA beam current, and a 5 lm focused beam for 60 s for F analysis and 20 kV accelerating voltage, a 20 nA beam current, and a 5 lm focused beam for 25-30 s for all other elements. Standardization of F and Cl were performed using a SrF 2 crystal and a tugtupite crystal, respectively, and were carefully checked against well calibrated secondary apatite standards (see details in Barnes et al. [2014] and Tart ese et al. [2013]). At the NHM, apatite analyses were carried out using a Cameca SX100 electron microprobe using a 15 kV accelerating voltage, a 20 nA beam current, and a 5 lm focused beam for 10-30 s per element, and 60 s for the REE analyses. The instrument was calibrated using well-characterized mineral standards (see details in Joy et al. 2011), and F and Cl were standardized using apatite and halite crystals, respectively. Isopropanol was subsequently used to remove carbon coating from the samples in preparation for ion-probe work.

H Isotope Analysis
The H 2 O content and H isotopic composition of apatite grains were measured using the Cameca NanoSIMS 50L at the Open University, following a well-established protocol described in detail in Barnes et al. (2013Barnes et al. ( , 2014 and Tart Table S2). In brief, analyses were carried out with large Cs + primary beams of approximately 260-400 pA current, with an accelerating voltage of 16 kV, after a 3 min presputter during which the beam is rastered on the sample surface over 12 9 12 lm areas to eliminate any surface contamination. Secondary ions of 1 H, 2 H, 12 C, and 18 O were collected simultaneously on electron multipliers for 2000 cycles (approximately 20 min) from the inner 25% area, using electronic gating, of 7 9 7 lm to 10 9 10 lm rasters, depending on the size of the crack-free areas suitable for analysis. An electron gun was used for charge compensation and tuned to minimize its contribution to the H background. The mass resolving power was set to approximately 4000 (Cameca definition), more than sufficient to readily resolve 2 H + from the interfering H 2 + species. Secondary ion images of 1 H and 12 C were monitored in real time during presputtering to ensure that the analyzed areas were free of any surficial contamination, cracks, or hotspots. Unfortunately, cracks hidden underneath the sample surface appeared during some analyses. In such cases, only portions of the secondary ion signals corresponding to analysis of pristine material were considered and further processing was performed using the NanoSIMS DataEditor software developed by Frank Gyngard (Washington University). Data inclusion was based on the 12 C signal, which is very low in lunar apatites but is several orders of magnitude higher for material filling the cracks. It must be pointed out that no discernible differences in D/H and 1 H/ 18 O ratios have been observed when isolating portions of the signals compared to the approximately 20 min integration for the reference apatites, which ensure that standardization to approximately 20 min long analyses remains valid. Further details, such as calculations of H 2 O contents, or correction of dD values for instrumental mass fractionation, can be found in the Supporting Information file.

Cl Isotope Analysis
The Cl content and Cl isotopic composition of apatite grains were also measured using the Cameca NanoSIMS 50L at the Open University. As for H isotopes, analyses were carried out with a Cs + primary beam, albeit at a lower current of approximately 40 pA, with an accelerating voltage of 16 kV. Each analysis was preceded by a 3 min presputter using a 150 pA primary beam rastered on the sample over 10 9 10 lm areas to eliminate any surface contamination. Secondary ions of 16 O 1 H, 18 O, 28 Si, 35 Cl, and 37 Cl were collected simultaneously on five electron multipliers for 500-1500 cycles (approximately 5-15 min) from the inner 25% area, using electronic gating, of 6 9 6 lm to 8 9 8 lm area rasters. An electron gun was used for charge compensation. The mass resolving power was set to approximately 10,000 (Cameca definition) in order to readily resolve isobaric interferences such as 17 O on the 16 O 1 H peak and 19 F 16 O on the 35 Cl peak. Secondary ion images of 16 O 1 H were monitored in real time during presputtering to ensure that the analyzed areas were free of any surficial contamination, cracks, or hotspots. The vacuum in the analysis chamber remained constant around 5 9 10 À10 torr.
Apatite Cl contents were calibrated using the measured 35 Cl/ 18 O ratios and the calibrations derived using terrestrial apatite standards (Durango, Crystal Lode Pegmatite Mine, and Lake Baikal apatites described in McCubbin et al. 2012), pressed in indium along with a dry San Carlos olivine crystal. The slopes of the calibration lines defined by apatite standards with varying Cl contents were used to calculate the Cl contents of apatites in NWA 4472 and SaU 169 (see Fig. S13). The reported uncertainties on the Cl contents of apatite combine the 2r uncertainty associated with the calibrations and the analytical uncertainties associated with each individual measurement. The dry San Carlos olivine was used to calculate the instrument background for Cl, which remained very low at approximately 0.1 ppm. To insure that this measure was adequate for epoxy-mounted samples such as SaU 169, analyses were also carried out under these analytical conditions on a plagioclase in an epoxy-mounted polished section of the meteorite Graves Nunataks 06128 that we also analyzed during the same analytical session.
These analyses yielded approximately 0.05 AE 0.01 ppm Cl, which is consistent with the background Cl content determined on the indiummounted San Carlos olivine. This background Cl was subtracted from the abundances measured in unknown apatites. Finally, reference apatite Ap005 was used to correct the measured 37 Cl/ 35 Cl ratios for instrumental mass fractionation. The accuracy of our d 37 Cl measurements was checked by repeat measurements on a second reference apatite, the Durango apatite, Ap003, which yielded a weighted average d 37 Cl value of À0.6 AE 0.6 & (2r; MSWD = 0.99; n = 18), consistent with the weighted average d 37 Cl value of 0.4 AE 0.6 & (2r; MSWD = 0.95; n = 12) reported by Treiman et al. (2014). Cl isotopic composition is reported using the standard delta (d) notation with respect to the 37 Cl/ 35 Cl ratio of the standard mean ocean chloride (SMOC). d 37 Cl values are reported with their associated 2r uncertainties, which combine the reproducibility of 37 Cl/ 35 Cl measurements on the reference apatite Ap005 and the internal precision of each analysis.

NWA 4472
Numerous analyses were carried out on large apatite grains that are part of a large lithic clast, which Joy et al. (2011) interpreted as a KREEP basalt assemblage (Fig. 3a). Apatite in this KREEP clast is associated with an intergrowth of coarse-grained ilmenite and hedenbergitic pyroxene (En 10 Fs 50 Wo 40 , Mg# 15-20 ; Fig. 4a). Pyroxene in this clast is highly ferroan and similar in composition to those in a silica and fayalite-rich clast (Fig. 4a). The other apatite grains investigated in this sample were isolated grains located throughout the matrix of NWA 4472 (Figs. 3b and 3c). They are generally surrounded by small grains of pyroxene and plagioclase. The composition of these pyroxene grains in the matrix ranges from $ En 65 Fs 30 Wo 5 to $ En 45 Fs 45 Wo 10 (Mg# 50-70 ), similar to the compositions of low-Ca pyroxene in an alkalisuite (HAS) granulite clast and numerous matrix pyroxene analyzed by Joy et al. (2011) (Fig. 4a). Plagioclase in the matrix of NWA 4472 is generally Carich (An 85-95 ).

SaU 169
In SaU 169, several analyses were carried out on a large apatite grain present in a lithic fragment located toward the edge of the investigated polished section (approximately 2 mm 9 2.5 mm; Figs. 1b and 2b). Apatite is intergrown with merrillite and this assemblage is surrounded by large grains of pyroxene, plagioclase, ilmenite, and K-feldspar (

NWA 773 (breccia)
Apatite analyzed in the breccia lithology of NWA 773 occurs in a variety of settings, either as isolated grains in a fine-grained matrix or part of larger lithic clasts. Apatite in the region of interest (ROI) 3 is one such example of an isolated grain. It is surrounded by pyroxene displaying a large range of compositions, from low-Ca pyroxene (En 61-66 Fs 25-26 Wo 8-14 ) typical of those found in the OC lithology of NWA 773 ( Fig. 4c; Fagan et al. 2003Fagan et al. , 2014, to ferroan compositions (En 7-20 Fs 51-66 Wo 25-30 ) more akin to some of the highly evolved clasts described by Fagan et al. (2003Fagan et al. ( , 2014 (Fig. 4c). A couple of olivine analyses in ROI3 confirm their Mg-rich (Mg# 63-68 ) nature compared with those in the ferroan lithologies (Fagan et al. 2003(Fagan et al. , 2014 and plagioclase is approximately An 95 ; these compositions are consistent with those of mineral phases in the OC lithology (Fagan et al. 2003). A few grains of silica and small patches of Krich glass (SiO 2~7 0 wt%, Al 2 O 3~1 5 wt%, and K 2 O~11 wt%) were also observed. The petrographic context in which apatite in ROI13 occurs is very similar; it is surrounded by pyroxene, olivine, plagioclase, and tiny silica fragments (Fig. 3g). Matrix pyroxene compositions in ROI13 display large variations from magnesian to ferroan (En 7-67 Fs 24-70 Wo 6-32 ) (Fig. 4c), while olivine (approximately Mg# 71 ) and plagioclase (An 83-95 ) display relatively restricted compositions similar to those in the OC lithology of NWA 773 (Fagan et al. 2003). In ROI17, apatite grains are associated with a plagioclase + silica assemblage surrounded by pyroxene and a few olivine grains (Fig. 3h). A large pyroxene grain below the plagioclase + silica assemblage (Fig. 3h) displays a homogeneous composition (En 15-20 Fs 50-58 Wo 27-30 ) while the smaller grains are characterized by a greater range of compositions (En 6-57 Fs 32-51 Wo 11-43 ) extending up toward the alkali-ferroan field defined by Fagan et al. (2014). The two olivine grains analyzed have different compositions (Mg# 54 and Mg# 71 ). Plagioclase composition within the plagioclase + silica assemblage seems to be quite variable and moderately Ca-rich (An 77-90 ) while the large plagioclase outside the assemblage is much more Ca-rich (An 96 ). Finally, apatite in ROI22 is associated with a large fayalitic olivine grain (Mg# 3 ) enclosing large inclusions of K-rich glass (SiO 2~7 8 wt%, Al 2 O 3~1 1.5 wt%, and K 2 O~7 wt%) (Fig. 3i). Once again, compositions of pyroxene grains dispersed around the fayalite + apatite assemblage show very large variations (En 3-62 Fs 14-78 Wo 9-39 ) from low-and high-Ca magnesian compositions typical of those of the OC lithology to highly ferroan compositions ( Fig. 4c; Fagan et al. 2003Fagan et al. , 2014. ROI22 also contains a large symplectite assemblage composed of hedenbergite + fayalite + silica and isolated olivine grains (Mg# 62-71 ) typical of olivine composition of the OC lithology (Fagan et al. 2003) (Fig. 3i).

Volatile Inventory of Apatites
All the apatite grains analyzed by EPMA in samples NWA 4472, NWA 773, and SaU 169 have less than 20 mol% Cl (equivalent to approximately 1.35 wt% Cl) in their volatile-bearing crystallographic X-site (Fig. 5), consistent with the Cl-poor nature of lunar apatites (e.g., Gancarz et al. 1971;Beaty et al. 1979;Taylor et al. 2004;Boyce et al. 2010;McCubbin et al. 2010aMcCubbin et al. , 2010bMcCubbin et al. , 2011Tart ese et al. 2013). In NWA 4472, apatite analyses in the KREEP clast ( Fig. 3a) cluster toward the F-apex of the F-Cl-OH ternary diagram (Fig. 5), containing $ 3.3 wt% F and $ 0.6 wt% Cl (Table 1; see also Table S3). Assuming that the crystallographic X-site only contains F, Cl, and OH (e.g., McCubbin et al. 2010aMcCubbin et al. , 2011, apatite grains in the KREEP clast (ROI1) in NWA 4472 could contain up to $ 0.25 wt% H 2 O. In NWA 4472, Matrix 1 apatite grain (Fig. 3c) contains $ 2.9 wt% F and $ 0.45 wt% Cl and up to $ 0.5 wt% H 2 O (Fig. 5), while other apatite grains in the matrix tend to contain more Cl (0.6-1.0 wt%) and up to approximately 0.6 wt% H 2 O (Fig. 5). Overall, H 2 O-poor and F-rich apatite grains in NWA 4472 appear to be similar to those in the crystalline impact-melt lithology 1 in other KREEPrich lunar meteorites Dhofar (Dho) 925/961, all of them plotting within or close to the overlapping fields for apatites in KREEP basalts and mare basalts (Fig. 5). Several analyses carried out in apatite grains in the matrix of NWA 4472 extend toward H 2 O-rich compositions more typical of mare-basalt apatites (Fig. 5). Analyses of apatite grains in the breccia lithology of NWA 773 cover the whole field defined by mare basalt apatite, yielding homogeneous Cl contents of around 0.4 wt% but highly variable F and H 2 O contents of $ 1.0-2.7 wt% and $ 0.35-1.25 wt%, respectively (Fig. 5). The few analyses of apatite from the OC lithology also seem to plot in the same field (Fig. 5)

Water Content and H Isotope Systematics of Apatite
In NWA 4472, six analyses were carried out in the large apatite grains that occur in the KREEP clast (Fig. 3a). These apatite display variable H 2 O contents between $ 2850 and 5450 ppm associated with dD values ranging between À103 AE 91& and 9 AE 92& (weighted average dD is À74 AE 33&, 2SD; Fig. 6; Table 2). These H 2 O abundances appear to be higher compared to those estimated from the F and Cl contents measured by EPMA (detection limits for F and Cl by EPMA dictates that the H 2 O contents, calculated by difference, are only reliable estimates for values greater than approximately 1400 ppm). In Matrix 2 apatite (Fig. 3b), an isolated grain lacking petrogenetic context, the measured H 2 O contents and D/H ratios are similar, ranging from $ 3400-5300 ppm H 2 O with dD values between À243 AE 150& and À12 AE 84& ( Fig. 6; Table 2). Another isolated apatite grain in the matrix of NWA 4472, Matrix 3, is also characterized by similar dD values (weighted average dD is À35 AE 81&, 2SD) but slightly lower H 2 O contents of $ 1800-2300 ppm ( Fig. 6; Table 2). However, Matrix 1 apatite, which also occurs as an isolated grain (Fig. 3c), has a very different dD signature with values ranging between 486 AE 62& and 904 AE 44&, for a narrow range (approximately 5300 to 6400 ppm) of H 2 O contents ( Fig. 6; Table 2). H 2 O abundances determined by SIMS for matrix apatites are slightly higher compared to those calculated by difference from F and Cl abundances measured by EPMA (see the Volatile Inventory of Apatites section).
In the brecciated lithology of NWA 773, we have carried out fourteen analyses in seven apatite grains (some of these are shown in Figs. 3g-i). The H 2 O contents and D/H characteristics of these apatite grains define two groups; the first group comprises analyses from three apatite grains (Ap#2, Ap#8, and Ap#13), which display moderate amounts of H 2 O between $ 700 and 2500 ppm and relatively restricted range in dD values between À105 AE 79& and 104 AE 148& (weighted average dD is À54 AE 82&, 2SD; Fig. 6; Table 2). Four other apatite grains (Ap#3, Ap#17, Ap#21, and Ap#22) constitute a second group, characterized by high to extremely-high H 2 O contents between $ 5400 and 16,700 ppm and comparatively elevated dD values ranging between 144 AE 63& and 319 AE 64& (weighted average dD is 255 AE 46&, 2SD; Fig. 6; Table 2). The H 2 O contents for apatite of the second group are consistent with those estimated from the F and Cl contents measured by EPMA.
In SaU 169, six analyses were performed on four apatite grains (Figs. 3d-f). These yielded relatively narrow ranges in H 2 O contents and D/H ratios, as H 2 O content varies between $ 1650 and 3420 ppm and dD values between À148 AE 91& and 201 AE 156& ( Fig. 6; Table 2). The results obtained for six apatite grains in Kal 009 (Figs. 3j-l) are comparable, as they gave $ 710-2390 ppm H 2 O with associated dD values ranging from À95 AE 118& to 149 AE 360&, except one analysis performed in Ap#4 in section 1 that yielded $ 530 ppm H 2 O with a slightly heavier dD value of 451 AE 213& (Figs. 6; Table 2).

Chlorine Content and Cl Isotope Systematics of Apatite
We have measured Cl contents and the associated Cl isotopic compositions of several apatite grains from two meteorites, NWA 4472 and SaU 169. The same apatite grains were analyzed for their H 2 O content and H isotopic compositions, as discussed in the preceding sections. In the case of NWA 4472, six analyses were carried out in the large apatite grains composing the KREEP clast (Fig. 3a), which gave Cl contents of $ 2510-4170 ppm and d 37 Cl values ranging between 17.2 AE 2.4& and 19.7 AE 2.5& (weighted average d 37 Cl is 18.5 AE 1.1&, 2SD; Fig. 7; Table 3). Such Cl contents of 2510-4170 ppm are lower than those measured by EPMA ($ 6100 AE 1000 ppm; Table 1). If Cl contents measured by EPMA were slightly overestimated, this likely explains why the H 2 O contents calculated by difference from EPMA analysis of F and Cl appear to be lower than those measured by SIMS. Also, the reproducibility of the Cl contents and d 37 Cl values measured four days apart is excellent; analyses 1#2 and 1#4 were carried out a few microns apart (Fig. 3a) and yielded almost identical results within uncertainties ( Table 3). One analysis carried out in an isolated apatite grain in the matrix yielded $ 5450 ppm Cl with a d 37 Cl value of 15.4 AE 2.4& ( Fig. 7; Table 3). Finally, four analyses were performed in the high-dD Matrix 1 apatite grain (Fig. 3c). These analyses yielded Cl contents of $ 2280-2990 ppm associated with highly variable d 37 Cl values ranging from 6.8 AE 3.3& up to 15.3 AE 3.3& ( Fig. 7; Table 3). Once again these Cl contents are slightly lower than those measured by EPMA (approximately 4400 AE 200 ppm; Table 1), which likely explains why the SIMS H 2 O contents measured in Matrix 1 apatite are slightly higher than those calculated by difference based on EPMA analyses of F and Cl. In SaU 169, four apatite grains were analyzed for their Cl contents and Cl isotopic compositions (Figs. 3d-f). Analyses carried out in the large apatite in ROI7 (Fig. 3d) show a relatively wide range of Cl contents between $ 6270 and 10050 ppm with d 37 Cl values ranging from 5.3 AE 2.2& to 9.2 AE 2.5& (weighted average d 37 Cl is 7.6 AE 2.1&, 2SD; Fig. 7; Table 3). The three other apatite grains analyzed are associated with two VHK KREEP clasts. They gave a similar range of Cl contents between $ 7480 and 10,010 ppm, with d 37 Cl values averaging $ 11.4 AE 1.2& (2SD; Fig. 7; Table 3). For SaU 169, Cl contents measured by SIMS and EPMA are in better agreement than for NWA 4472.

Are Measured Apatite H and Cl Isotopic Compositions Indigenous Lunar Signatures?
Several secondary processes could have potentially modified the original isotopic compositions of lunar apatite after their magmatic crystallization. Two obvious processes affecting the material present at or near the surface of the Moon are the implantation of solar wind (SW) hydrogen and spallation from galactic cosmic ray (GCR). The H 2 O and H measured in lunar samples have long been interpreted as produced by either terrestrial contamination or SW implantation of hydrogen in the outer part of regolith grains (e.g., Epstein and Taylor 1973). In-situ analyses have subsequently demonstrated that SW implantation of Li, H, N, or C indeed occurred on the lunar surface, and could reach a depth of approximately 100 nm from the surface of the grains for H (e.g., Chaussidon and Robert 1999;Hashizume et al. 2000Hashizume et al. , 2004St ephant and Robert 2014). The depth over which SW hydrogen can be implanted in samples residing on the lunar surface is so small that it is unlikely to have affected the H 2 O contents and D/H ratios measured in apatite in the investigated lunar meteorites (contamination by SW hydrogen could only have occurred in apatites if they directly crystallized from an impact melt which itself had inherited a SW-hydrogen component from the regolith).
Spallation processes induced by exposure to GCR can greatly affect the H 2 O contents and D/H ratios measured in lunar samples, especially at low H 2 O contents (e.g., Saal et al. 2013;Barnes et al. 2014), as cosmic ray spallation can affect materials residing at much deeper levels (down to approximately 1-2 m depth; e.g., Williams and Gold 1975) than SW implantation. H and D are produced on the Moon by cosmic ray spallation (Merlivat et al. 1976), their amount depending on the cosmic ray exposure (CRE) time in the lunar regolith and the shielding depth of a given rock. For meteorites, it is generally assumed that CRE starts when they are ejected from their parent body, meteorites being buried at least several meters deep into their parent body before ejection (Eugster 2003). However, the meteorites investigated here are breccias and their constituent clasts could have been exposed to cosmic rays in the lunar regolith for prolonged periods of time during their geological history on the Moon. Lorenzetti et al. (2005) have  Fig. 8 for the brecciated lithology of NWA 773. We used the production rates for H (2 9 10 À10 mol H 2 g À1 Ma À1 ) and D (0.5 9 10 À12 mol D 2 g À1 Ma À1 ) determined by Merlivat et al. (1976) and assigned them a 50% uncertainty (see detailed discussion in Saal et al. 2013). With these production rates, exposure to GCR for 1 Ga results in the production of approximately 2 ppm H 2 O and approximately 4.6 9 10 À3 ppm D 2 O. Correction of the measured D/H ratios for spallation production of H and D resulted in a maximum decrease of dD by $ 60& for the driest apatite analysis ($ 1775 ppm H 2 O) from NWA 4472, by $ 75& for the driest apatite analysis ($ 720 ppm H 2 O) from NWA 773, and by $ 7& for the driest apatite analysis ($ 1600 ppm H 2 O) from SaU 169 (Fig. 8). After correction for H and D production by GCR, all the data remain largely within uncertainties of the original measurements (Fig. 8). As a result, and considering the large uncertainties associated with CRE ages for samples such as NWA 773, or the lack of CRE age for NWA 4472, only the measured H 2 O contents and dD values are eventually considered in this contribution. Finally, alteration of the isotopic characteristics of apatite by terrestrial weathering processes must also be addressed, as the four samples investigated are all hot desert finds and, therefore, have resided for prolonged periods of time on Earth (e.g., $ 10 ka and $ 17 ka for SaU 169 and NWA 773, respectively; Gnos et al. 2004;Fig. 8. Diagram displaying the measured H 2 O contents and dD values for apatite in brecciated lunar meteorites NWA 4472, NWA 773, and SaU 169 compared with the data corrected for spallation production of H and D (see text for details). Nishiizumi et al. 2004). The majority of the apatite grains analyzed have dD values ranging between À200& and 100& (weighted average dD is À42 AE 37&, 2r-n = 30), with Matrix 1 grain in NWA 4472 and the high-H 2 O group of apatites in NWA 773 having higher dD values (Fig. 6). The H isotopic composition of meteoric water and groundwater in the central Saharan region varies between À70& and 20& (Saighi et al. 2001), which is close to the D/H ratios measured in the majority of apatites. Thus, in the absence of any other independent constraints, it could be argued that most of the analyses reflect cryptic alteration of original H 2 O contents and D/H signature by interaction with terrestrial waters. Yet, several features are clearly inconsistent with such a possibility. First apatite in NWA 4472 and NWA 773 contain water with both high-dD as well as Earth-like dD values (Fig. 6). There are no a priori reasons for terrestrial waters to have altered the D/H systematics of only some apatite grains in a rock and not in other apatite grains located a few millimeters away, unless permeability varies dramatically on such a small scale. Also, the group of high-H 2 O apatite in the brecciated lithology of NWA 773 displays large variations in H 2 O contents from $ 5400 ppm up to 16700 ppm with relatively restricted dD values of $ 200-300& (Fig. 6). If these very high H 2 O contents resulted from apatite alteration by terrestrial waters, one would expect a decrease of dD to values of approximately 50& for $ 16000 ppm H 2 O analysis, which is not observed. The large range in dD values between $ 500& and 900&, observed in Matrix 1 apatite in NWA 4472, is also inconsistent with modification by terrestrial waters. Adding around 1200 ppm of Saharan water to the low-H 2 O and high-dD analysis in this grain (5200 ppm H 2 O and dD of $ 900&) would only lower the dD by about 175&. Finally, the d 37 Cl measured in apatite range between $ 7& and 20& for NWA 4472 and $ 5& and 12& for SaU 169. These values are distinctly higher than the d 37 Cl of terrestrial materials, including most of the waters and hydrothermal fluids (the range of d 37 Cl Earth could be averaged to 0 AE 2&; e.g., Stewart and Spivack 2004;Sharp et al. 2013b). It is worth noting that two large SIMS pits were created on Matrix 1 apatite in NWA 4472 during the course of U-Pb dating by Joy et al. (2011). Therefore, if our NanoSIMS spots for H and Cl isotope analyses overlapped with these previously sputtered areas, it may not be possible to completely rule out the disturbance of the volatile H and Cl in the apatite lattice by previous SIMS analyses, which could account for some of the variability observed for dD and d 37 Cl values in this grain. On the other hand, Matrix 2 apatite in NWA 4472 (Fig. 3b) has also been dated by SIMS by Joy et al. (2011), who sputtered three large SIMS pits on this grain. In this grain, our NanoSIMS analyses yielded a range of H 2 O contents ($ 3400 to 5300 ppm H 2 O), associated with dD values consistent with each other considering uncertainties (À243 AE 150& to À12 AE 84&). From our data set, it is not possible to argue for the presence of an evident and systematic effect induced by previous SIMS analyses, an issue that would merit further investigation in a future study.
To summarize, considering all the above lines of evidence, we are confident that most if not all of the apatite grains we have analyzed in brecciated meteorites NWA 4472, NWA 773, SaU 169, and Kal 009 have retained their original lunar isotopic signatures.

Inferring Source Lithology From H and Cl Isotope Systematics of Apatite
In brecciated meteorites, the petrogenetic context within which an isolated mineral, or a small assemblage of minerals of interest, occurs can unfortunately be lacking. This may hamper interpretation of chemical or geochronological data, as deciphering the source lithology may not be possible. As the U-Pb geochronometer in phosphates is easily disturbed by thermal events above $ 500°C (Cherniak et al. 1991), disentangling crystallization and impact-resetting ages in phosphate U-Pb data can be challenging. Even when lithic clasts are preserved, the question of their representativeness is also an issue (e.g., Warren 2012). Abundances of certain elements such as Fe, Mg, or REEs in apatite do not seem to enable straightforward discrimination among the different lunar lithologies (Fig. 9). Apatite in rocks from the Mg-suite have low Fe contents, high Mg#, and moderate Ce contents, while apatite in low-Ti mare basalts is generally characterized by high Fe contents, low Mg#, and relatively high Ce contents (Fig. 9). In other lithologies such as the alkali-suite rocks, impact-melt rocks, KREEP basalts, and high-Ti mare basalts, apatite Fe, Mg, and Ce contents largely overlap (Fig. 9). This is precisely in this overlapping area where the majority of the apatite analyses carried out in NWA 4472, NWA 773, and SaU 169 plot. The Fe, Mg, and Ce contents of the apatites analyzed would indicate possible genetic relationships with the Mg-and alkali-suite or with KREEP basalts (Fig. 9).

NWA 4472
In  (Fig. 6), which is consistent with the KREEP-rich nature of NWA 4472 (Korotev et al. 2009;Joy et al. 2011). On the other hand, apatite in the KREEP clast and Matrix 3 apatite are characterized by $ 2000 to 5000 ppm Cl and d 37 Cl values below $ 20&, which is more consistent with data reported for apatite from "typical" mare basalts rather than from KREEP-rich lithologies such as KREEP basalt and highland samples from the Mg-and alkali-suites (Sharp et al. 2010a;Boyce et al. 2013;Treiman et al. 2014) (Fig. 7). Overall, these apatite grains are, therefore, characterized by unfractionated dD values compared to most carbonaceous chondrites, but have highly fractionated d 37 Cl values.
The isolated Matrix 1 apatite grain (Fig. 3c), is characterized by H 2 O-D/H and Cl-d 37 Cl signatures akin to "typical" mare basalts rather than to KREEP-rich lithologies (Figs. 6 and 7a). It displays relatively large ranges in dD and d 37 Cl values ($ 500& to $ 900& and $ 7& to $ 15&, respectively) with narrow ranges in H 2 O and Cl contents ($ 5300-6400 ppm and $ 2300-3000 ppm, respectively). In this grain, it is difficult to define any obvious relationship between the H 2 O and Cl contents and their corresponding dD and d 37 Cl signatures, or with the locations where analyses were carried out in the grain (Fig. 3c). Moreover, the grain might only be a fragment of an originally larger grain, hampering identification of core and rim regions. Generally, one might have invoked rapid crystallization of the grain during degassing of H-bearing species as H 2 in the vapor phase, as proposed to explain the large variations of dD and the narrow range of H 2 O contents observed in apatites in high-Ti mare basalts . However, such a mechanism seems to be incompatible with the d 37 Cl characteristics of this apatite grain. To explain the elevated d 37 Cl of most lunar samples compared to other planetary materials Sharp et al. (2010a) argued for kinetic fractionation of the Cl isotopes during degassing of metal-chloride species, such as FeCl 2 or ZnCl 2 . The strongest constraint associated with the mechanism proposed by Sharp et al. (2010a) is that it requires anhydrous melts to prevent formation of HCl, as they argue that HCl degassing does not fractionate Cl isotopes, despite fractionation of up to $ 10& in d 37 Cl have been observed in experiments involving H 2 O-rich systems (Sharp et al. 2010b). Sharp et al. (2013a) reconciled this requirement of anhydrous melts for Cl isotope fractionation and the growing evidence indicating that lunar basaltic melts can contain non-negligible amounts of water by proposing that H-bearing species degases earlier than Cl-bearing species in lunar magmas (Ustunisik et al. 2011;Sharp et al. 2013a). Yet in Matrix 1 apatite in NWA 4472, the variations of dD and d 37 Cl would require crystallization during degassing of both H-bearing and Cl-bearing species, pointing toward HCl degassing. Hence, there is a critical need for degassing experiments of both HCl and metal chlorides under lunar conditions (in vacuum and at low fO 2 ) in order to further assess the likely mechanisms responsible for elevated d 37 Cl signatures in mare basalts.
Diffusive gain or loss of OH and Cl can also kinetically fractionate H and Cl isotopes, provided that apatite resided at high temperatures for a prolonged period of time in order to enable diffusive mobility of the volatiles (e.g., Boyce and Hervig 2008;Anand et al. 2014). As Matrix 1 apatite lacks petrogenetic context, this possibility is difficult to assess; on the other hand, the Pb/Pb and U-Pb chronometers have not been disturbed in this grain (Joy et al. 2011), which likely rules out extensive isotopic mobility by hightemperature diffusion, since diffusion of OH and Cl is faster than diffusion of Pb in the apatite lattice (Cherniak et al. 1991;Brenan 1993). Whichever mechanism caused large variations of dD and d 37 Cl in Matrix 1 apatite, its H and Cl isotope characteristics are consistent with those observed in "typical" mare basalts (Figs. 6 and 7a). With a U-Pb crystallization age of approximately 4.35 Ga (Joy et al. 2011), this grain could, therefore, be a remnant of some of the oldest known basaltic volcanism on the Moon (e.g., Terada et al. 2007a).

SaU 169
In the IMB lithology of SaU 169, the H 2 O-D/H systematics of the analyzed apatite grains are also consistent with the H 2 O-D/H characteristics of apatite rich in H 2 O in KREEP basalts (Tart ese et al. 2014a) and in the Mg-and alkali-suite rocks   (Fig. 6), in good agreement with the KREEP-rich nature of SaU 169 (Gnos et al. 2004;Korotev et al. 2009;Lin et al. 2012). Combined together, the Cl contents and Cl isotope characteristics of apatites in the IMB lithology of SaU 169 differ from those reported so far for apatite in KREEP-poor mare basalts and KREEP-rich rocks of the lunar highlands (breccias and Mg-suite rocks). Apatite in SaU 169 have d 37 Cl values similar to those measured in apatite in mare basalts (weighted average d 37 Cl is 7.6 AE 2.1&, 2r; Fig. 7), but contain much higher Cl contents. The IMB lithology of SaU 169 is geochemically unique, containing $ 300-1000 times CI chondrite abundances of ITE (Gnos et al. 2004). Most of the apatite grains we analyzed are associated with even more evolved lithic clasts (i.e., VHK KREEP clasts; Lin et al. 2012), which are characterized by abundances of KREEP at $ 1500 9 CI chondrites. Apatite in this unique lithology appear to be characterized by a distinct Cl-d 37 Cl systematics compared with the limited amount of apatite data currently available for other lunar lithologies. More importantly, the Cl-d 37 Cl systematics of apatite in the KREEP-rich IMB lithology of SaU 169 seems to break down the dichotomy of "low-d 37 Cl/ low KREEP content vs. high-d 37 Cl/high KREEP content" observed from previously available data (Fig. 7). Considering that almost the entire spread of d 37 Cl values observed in solar system materials has thus far been observed only in lunar samples, it is likely that Cl isotopes have been fractionated by some processes specific to the Moon (e.g., low-fO 2 degassing of HCl or of metal chlorides; Sharp et al. 2010aSharp et al. , 2010b, but such process(es) have yet to be identified and proven experimentally. In addition, several Cl reservoirs with different isotopic characteristics might occur in and on the Moon, opening the possibility that measured d 37 Cl values correspond to mixing signatures. These observations lead us to conclude that our current understanding of Cl isotopes in lunar materials is far from being complete. Further Cl and d 37 Cl analyses of apatite in a broad range of lunar lithologies, coupled with specific experimental investigation of Cl isotope fractionation processes at lunar conditions, are required in order to improve our understanding of the Cl geochemical cycle in and on the Moon.

NWA 773 (breccia)
In the brecciated lithology of NWA 773, the H 2 O-D/H systematics of apatite defines two groups. In the first group, apatite grains contain moderate H 2 O contents between $ 700 and $ 2500 ppm associated with chondrite-like D/H ratios (weighted average dD = À54 AE 82&) (Fig. 6). The H 2 O-D/H systematics of this first group of apatite analyses is consistent with results obtained on apatite grains analyzed in the OC lithology (Tart ese et al. 2014a), indicating that they likely originated from a similar lithological source. In the second group, apatite grains are characterized by high to very high H 2 O contents ($ 5400-16,700 ppm H 2 O) and dD values of approximately 250 AE 50& (Fig. 6). Such characteristics are unique among apatite in all the lunar samples analyzed so far. This suggests that these apatite grains are either related to a unique, unknown, source lithology or that their parent rock(s) were affected by specific petrogenetic processes. We favor the second possibility as discussed in the Behavior of Volatiles During Silicate Liquid Immiscibility section.

Kalahari 009
Compared with NWA 4472, NWA 773, and SaU 169, which all contain a significant KREEP component, Kal 009 has very low bulk-ITE abundances (Sokol et al. 2008 Neal and Kramer 2006). However, the low dD values measured in apatites in 14053, which has a unique petrologic history (e.g., Taylor et al. 2004), could have resulted from interaction with solar wind-implanted hydrogen in the lunar regolith during an impact-heating event (Greenwood et al. 2011 Fig. 6), but display a similar range of dD values (À95 AE 118& to 451 AE 213&; Fig. 6). The crystallization age of Kal 009 of $ 4.35 Ga (Terada et al. 2007a;Sokol et al. 2008) and bulk-rock composition of $ 13 wt% Al 2 O 3 is also consistent with the range of crystallization ages ( $ 3.9 to $ 4.3 Ga) and bulk chemistry of Apollo 14 high-Al (>11 wt% Al 2 O 3 ) mare basalts (Neal and Kramer [2006] and references therein). Therefore, the H 2 O-D/H systematics and other geochemical similarities between Kal 009 and high-Al mare basalts (sampled at the Apollo 14 landing site) indicate a possible genetic link between them.

Behavior of Volatiles During Silicate Liquid Immiscibility
Soon after initial studies of Apollo samples it was argued that some rock types, such as the felsites, could have formed due to silicate liquid immiscibility (SLI). This process is likely a common process and a result of extensive low-P fractional crystallization of many lunar volcanic products, producing a Fe-rich basaltic melt enriched in REE and P and a Si-rich rhyolitic melt enriched in K (e.g., Roedder andWeiblin 1970, 1971;Rutherford et al. 1974;Hess et al. 1975;Philpotts 1982;Roedder 1984;Longhi 1990). Evidence for SLI has been recognized, for example, in melt inclusions from mare basalts (e.g., Roedder 1984), within late-stage melt pockets (mesostasis) in mare basalts, and in some plutonic rocks that crystallized at shallow depths in the lunar upper crust (e.g., Jolliff 1991Jolliff , 1998. To comprehensively characterize the volatile inventory of lunar magmatic products, the effects of SLI on the partitioning behavior of volatiles between immiscible melts, and on the extent of isotopic fractionation of volatiles such as H and Cl, if any, needs to be investigated (e.g., Pernet-Fisher et al. 2014). It should be pointed out, however, that due to the lack of relevant data, notably experimental, this section of the discussion remains speculative. Fagan et al. (2014) have argued that the highly ferroan lithic clasts (Fig. 4c) in the brecciated lithology of NWA 773 were likely formed after SLI occurred, the symplectite (SP) assemblages representing the Fe-rich fraction while the alkali-ferroan (AF) clasts correspond to the Si-rich liquids. It is, therefore, a possibility that the unusual H 2 O-dD characteristics of some apatites analyzed in the brecciated lithology of NWA 773 might be related to the effect of SLI. Ap#22 for example is clearly associated with an AF clast comprising fayalite + K-rich glass (compare Fig. 3i with the fig. 7d of Fagan et al. 2014). Ap#17 is associated with an unusual plagioclase + silica assemblage (Fig. 3h) surrounded by some ferroan pyroxene (Fig. 4c). These apatite grains contain very high amounts of H 2 O ($ 5400-16,700 ppm) reaching almost pure hydroxylapatite composition (17900 ppm H 2 O). As discussed by Boyce et al. (2014), apatite H 2 O content is not solely related to the H 2 O content of its parental melt. This is because water is as an essential structural constituent (as OH -) of the apatite X-site, and as such its incorporation depends on stoichiometric constraints; F, Cl, and OH all reside in this X-site, and are competing for it, implying that the abundance of any of these elements is inevitably related to the abundance of the other ones. These H 2 O-rich apatite contain approximately 2-7 times more H 2 O compared to those of the OC lithology (Fig. 6), and seem to have crystallized from the highly ferroan alkaline and Si-rich melt fraction formed after SLI. Again, this does not necessarily imply that this Si-rich melt fraction was enriched in H 2 O compared to residual melt from which apatite crystallized in the OC lithology. This could indicate instead that F and/or Cl were depleted in the Si-rich melt fraction during SLI, resulting in higher H 2 O/F and H 2 O/Cl ratios. Unfortunately data on partitioning of F and Cl during SLI are extremely scarce. There is no measurement of F and Cl in coexisting Fe-rich and Si-rich melt inclusions found to have formed from SLI in lunar basalts. Lester et al. (2013) have investigated experimentally element partitioning during SLI for volatile-rich systems relevant to terrestrial conditions (fO 2 at or above QFM). Lester et al. (2013) found that F behavior is quite unpredictable as it can be concentrated either in the Fe-rich melt fraction or in the Si-rich one (average D F Fe-rich/Si-rich $ 1 AE 0.6), in contrast to sulphur, for example, which is strongly partitioned into the Fe-rich fraction. We propose that the very high H 2 O contents measured in some apatites in the brecciated lithology of NWA 773 might be a direct result of SLI, which would have depleted the Si-rich melt fraction in F and Cl, and in some extreme cases left H 2 O as the only competitor for the apatite X-site. Further constraints, analytical or experimental, on the behavior of F and Cl during SLI in lunar magmas would be required to investigate this issue further.
Besides being H 2 O-rich, these apatite grains are characterized by homogeneous dD values with a weighted average of $ 255 AE 46&. These dD values are $ 200-300& higher than those measured in apatite in the OC lithology in NWA 773. Following the discussion above on the possible role of SLI in producing Si-rich melt with elevated H 2 O/F and/or H 2 O/Cl ratios, which would have resulted in the crystallization of H 2 O-rich apatite, SLI could have also resulted in an increase of dD by $ 200-300&. It has been shown that SLI induces fractionation of O isotopes, the d 18 O in the Si-rich melt fraction increasing by approximately 0.6& compared to the Fe-rich melt fraction (Kyser et al. 1998). Kyser et al. (1998) also showed that thermal gradients induce isotopic fractionation in silicate melts, the lighter isotope getting enriched toward hotter regions; they measured fractionation of d 18 O of 1.2-1.4& per 100°C. More recent studies added Mg, Si, Ca, and Fe to the list of isotopic systems fractionated due to thermal gradients in silicate melts (Richter et al. 2008(Richter et al. , 2009Huang et al. 2010). Bindeman et al. (2013) recently demonstrated that the presence of H 2 O in the melt greatly enhances thermal isotopic fractionation, and that H isotopes of the water itself are fractionated, dD increasing by approximately 45& per 100°C gradient in their experiments. The absence of data on the possible effect of SLI sensu stricto on fractionation of H isotopes prevents us from being certain about SLI to be the cause of the increase in dD by $ 200-300&. Nevertheless, on balance, the existing data and observations point toward a potential role of SLI in modifying dD signatures of lunar melts and further work is necessary to confirm this hypothesis both through actual measurements of appropriate samples and experimental work.

SUMMARY
In this study, we have investigated the H isotope systematics of apatite from four brecciated lunar meteorites (NWA 4472, NWA 773, SaU 169, and Kal 009), and measured the Cl isotopic composition of apatite in two of these meteorites (NWA 4472 and SaU 169) using secondary ion mass spectrometry. Most of the apatite analyses carried out in the regolith breccia NWA 4472 yielded $ 2000-6000 ppm H 2 O, with associated dD values between À200 and 0&. One isolated apatite grain located in the matrix of the breccia is completely different as it contains $ 6000 ppm H 2 O with associated dD values of $ 500-900&. In NWA 4472, the low-dD apatite grains also contain $ 2500-7500 ppm Cl associated with d 37 Cl values of $ 15-20&, while the high-dD grain isolated in the matrix yielded $ 2500 ppm Cl and heterogeneous d 37 Cl values of $ 7-15&. In the regolith breccia part of NWA 773, apatite grains we analyzed define 2 groups. In the first one, apatite grains contain 700 to 2500 ppm H 2 O, associated with homogeneous dD values around $ 0 AE 100&. In the second group, apatite analyses yielded elevated H 2 O abundances ($ 5500-16,500 ppm), associated with dD values $ 250 AE 50&. Apatite in the impact-melt breccia SaU 169 and the fragmental basaltic breccia Kal 009 yielded similar H 2 O contents ($ 600-3000 ppm) and dD values (À100 to 200&). In SaU 169, these apatite grains contain $ 6000-10,000 ppm Cl, characterized by d 37 Cl values of $ 5-12&.
As the d 37 Cl of terrestrial materials cluster around 0&, the elevated d 37 Cl values above 5& obtained on apatite in NWA 4472 and SaU 169 suggest that apatite in these hot-desert meteorites have retained their original lunar isotopic signatures. Regarding H isotopes, most of the analyzed apatite grains have dD values within the range reported for carbonaceous chondrites, and are consistent with compositions measured in H 2 Orich apatite analyzed in ancient (>3.9 Ga) magmatic rocks from the Mg-and alkali-suites, as well as that of KREEP basalts. On the other hand, one apatite grain in NWA 4472 is characterized by H and Cl isotope compositions similar to those reported for apatite from typical mare basalts. This grain could represent a remnant of some of the oldest known lunar volcanic activity as it has previously been dated at 4.35 Ga. Lastly, we found a group of apatite in the brecciated lithology of NWA 773 characterized by high to extremely-high H 2 O contents, reaching pure hydroxylapatite composition. As it has been argued that numerous evolved clasts in NWA 773 formed through silicate liquid immiscibility of their parent melts, these H 2 O-rich apatites could provide important information regarding the potential effects of such a process on the evolution of volatiles in lunar magmas.
were extensively tentbound during the 2012 ANSMET season.

SUPPORTING INFORMATION
Additional supporting information may be found in the online version of this article: Data S1. The supporting information file contains twelve figures (Figs. S1-S12) displaying backscattered electron images and composite X-ray element maps of the studied polished sections of lunar meteorites. It also contains further details regarding secondary ion mass spectrometry protocols for H and Cl isotope analyses, as well as a figure displaying an example of calibration obtained for Cl content measurements on terrestrial apatite standards (Fig. S13). Table S1: Semi-quantitative EDS analyses of selected silicates. Table S2: Beam conditions and calibration parameters for the D/H analytical sessions. Table S3: Electron microprobe analyses of apatite in NWA 4472 and NWA 773.