An ancient reservoir of volatiles in the Moon sampled by lunar meteorite Northwest Africa 10989

Northwest Africa (NWA) 10989 is a recently found lunar meteorite we used to elucidate the history of volatiles (H and Cl) in the Moon through analysis of its phosphates. The petrology, bulk geochemistry and mineralogy of NWA 10989 are consistent with it being a lunar meteorite with intermediate-iron bulk composition, composed of 40% of mare basaltic material and (cid:1) 60% non-mare material, but with no obvious KREEP-rich basaltic components. It is probable that the source region for this mete-orite resides near a mare–highlands boundary, possibly on the farside of the Moon. Analyses of chlorine and hydrogen abundances and isotopic composition in apatite and merrillite grains from NWA 10989 indicate sampling of at least two distinct reservoirs of volatiles, one being similar to those for known mare basalts from the Apollo collections, while the other potentially represents a yet unrecognized reservoir. In situ Th-U-Pb dating of phosphates reveal two distinct age clusters with one ranging from 3.98 ± 0.04 to 4.20 ± 0.02 Ga, similar to the ages of cryptomare material, and the other ranging from 3.32 ± 0.01 to 3.96 ± 0.03 Ga, closer to the ages of mare basalts known from the Apollo collections. This lunar breccia features mixing of material, among which a basaltic D-poor volatile reservoir which doesn’t appear to have been recorded by Apollo samples.


INTRODUCTION
The majority of lunar meteorites are thought to originate from areas on the Moon not sampled by the Apollo and Luna missions (Warren and Kallemeyn, 1991;Korotev, 2005), thereby providing a broader coverage across the Moon's surface of both the nearside and the farside.Therefore, lunar meteorites expand our knowledge of the surface composition of the Moon, and enhance our understanding of the bulk composition of the Moon and its evolution throughout its geological history.Brecciated meteorites contain a variety of clast types and geochemical information, providing a snapshot into the material mixing processes at the Moon's surface, which occur at local to regional scales, and into the diversity of material that is present at a single location.In terms of their bulk composition, lunar meteorites have been divided into three categories: basaltic (17-23% FeO, 8-12% Al 2 O 3 ), feldspathic (low FeO < 7%, Al 2 O 3 > 25%), and intermediate, which contain both basaltic and feldspathic components (Korotev et al., 2009).The intermediate meteorite group has bulk FeO contents between 7 and 17 wt.%,and Al 2 O 3 contents between 13 and 20 wt.%, and constitute roughly 25% of the lunar meteorites that exist currently in the worldwide collections (Korotev et al., 2009).These lunar meteorites with intermediate-iron bulk composition could help to understand the crustal evolution on the Moon and are thought either to originate from the boundary between the lunar highlands and mare regions, most likely from the lunar nearside, or to have sampled cryptomare material.The discovery and categorization of more lunar meteorites, particularly meteorites with intermediate-iron bulk composition, suggests that the canonical 'Apollo' model of the Moon, in which all rocks are explained as a mixture of Feldspathic Highland Terrane (FHT), Procellarum KREEP Terrane (PKT), where KREEP stands for potassium (K), rare earth elements (REE), and phosphorous (P) (Warren and Wasson, 1979), and Mare (Jolliff et al., 2000), may not be applicable to the entire lunar surface (Korotev et al., 2009).
In order to identify reservoirs of lunar materials that formed a breccia and possibly locate its source region(s), hydrogen and chlorine isotopic compositions in apatite grains can be useful proxies.As such, mare basalts, KREEP-rich basalts and Mg and Alkali suite materials from Apollo samples appear to have distinct dD-d 37 Cl systematics (Sharp et al., 2010;Greenwood et al., 2011;Barnes et al., 2013Barnes et al., , 2014Barnes et al., , 2016Barnes et al., , 2019;;Tarte `se et al., 2013;Robinson et al., 2016).Nevertheless, analyses of volatiles in new lunar meteorites continue to expand the dataset and to broaden our knowledge of the history of lunar volatiles.
North West Africa (NWA) 10989 is a lunar meteorite that was found near the Morocco-Algeria border in 2014 (Bouvier et al., 2017).This meteorite is a single roughly spherical stone with a dark brown fusion crust that has a diameter of $2 cm and a mass of 14.41 g.Based on an initial petrographic and geochemical investigation, Ashcroft et al. (2017) classified this meteorite as a mixed lunar fragmental breccia that has intermediate bulk iron content.It is composed of mare-and highlands-derived materials in roughly equal proportions.This study presents a comprehensive mineralogical and geochemical data on mineral, lithic and impact-melt clasts in this meteorite including its bulk-rock major-and traceelement composition, abundance and isotopic composition of chlorine and hydrogen, and U-Th-Pb dates in apatites and a merrillite.This new dataset is then utilized to perform comparisons with other lunar samples in order to gain insights into lunar petrogenetic processes and evaluate potential mantle source region(s) for lithologies contained in NWA 10989.

Petrography
A polished thin section ($1.2 cm Â 1.2 cm) of NWA 10989 was prepared at the Open University (OU) for study by optical and scanning electron microscopy (SEM) methods.Microscopic observations under transmitted and reflected light were used to determine the main mineral phases and study the textures of the breccia.Backscattered electron (BSE) and secondary electron (SE) images were collected using the FEI Quanta 3-D FIB-SEM at the OU using beam conditions of 20 kV and 0.6 nA.Qualitative energy-dispersive spectroscopic (EDS) analysis were used for rapid identification of mineral phases.Whole sample and smaller area BSE maps were used as a guide for detailed electron-probe and ion-probe work.In addition, elemental X-ray maps of Al, Ca, Cl, Cr, Fe, Mg, Mn, Na, Ni, K, P, S, Si and Ti were also collected in EDS mode using an Oxford Instruments 80 mm X-MAX energy-dispersive X-ray detector attached to the SEM.A 20 kV accelerating voltage and a 0.60 nA beam current were used during the elemental mapping, and Xray maps were acquired at 512 Â 448 pixels and a magnification of $300, with a dwell time of 100 ms.The X-ray maps were used to calculate modal mineral abundances in different lithologies/clasts using the ImageJ software.

Mineral chemistry
Quantitative phase analyses were conducted using a CAMECA SX-100 electron-probe micro-analyzer (EMPA) in wavelength-dispersive spectroscopy (WDS) mode at the OU.An accelerating voltage of 20 kV and a 20 nA beam current were used for silicate, oxide and mineral phases, whereas a 10 nA beam was used for plagioclase and glass analysis to minimize any migration or loss of volatiles.In apatite, F was analysed separately with a 5 mm spot size and a beam at 10 kV and 4 nA, and then the remaining elements were analysed with a 20 kV, 20 nA beam current and 5 mm spot size, using the protocol described by Barnes et al. (2014).Spot sizes of 1, 5 or 10 mm were used depending on crystal size.Counting times were typically 15-40 s on the peak, with background measurements made before and afterwards for half the counting time.Primary standards used for calibration were a range of natural crystals including feldspar (Al, Si, K), jadeite (Na), forsterite (Mg), hematite (Fe), bustamite (Ca, Mn), apatite (P), chromite (Cr), rutile (Ti) and barite (Ba, S) as well as nickel metal (Ni).Secondary pyroxene, olivine, feldspar and apatite standards were run at the start, middle and end of analytical sessions to check the calibration and ensure internal consistency (see Supplementary Table S1).

Oxygen isotopes
Oxygen isotope analysis was performed at the OU, using an infrared laser-assisted fluorination system (Miller et al., 1999;Greenwood et al., 2017).Oxygen was released from the samples (one analysis of two aliquots -approximate weight of $2 mg) by heating in the presence of BrF 5 .The released oxygen gas was purified by passing it through two cryogenic nitrogen traps and over a bed of heated KBr.Oxygen gas was analysed using a MAT 253 dual inlet mass spectrometer.Recent levels of precision obtained on the OU system, as demonstrated by 38 analyses of an inhouse obsidian standard, were as follows: ±0.053‰ for d 17 O; ±0.095‰ for d 18 O; ±0.018‰ for D 17 O (2 SD) (Starkey et al., 2016).Oxygen isotope analyses are reported in standard delta notation.The d 18 O value is calculated as The reference for both is Vienna Standard Mean Ocean Water, VSMOW.The deviation from the terrestrial fractionation line, has been calculated using the linearized format of Miller et al. (2002): where k is 0.5247.The sample was not pretreated with ethanolamine thioglycolate (EATG) solution (to remove terrestrial weathering products) and therefore the bulk-rock that was measured for O isotopes would have included some terrestrial material (mainly as fracture-filling veins as seen in optical microscopy).

Bulk chemistry
The bulk-rock concentrations of major and minor elements (except K 2 O) were determined using inductively coupled plasma optical emission spectrometry (ICP-OES; Thermo iCap 6500 Duo) and concentrations of trace elements and potassium were determined using inductively coupled plasma mass spectrometry (ICP-MS; Agilent 7700x) at the Natural History Museum, London.A 40 mg sub-split of homogenized sample powder was used for the ICP-OES analysis and a separate 54.8 mg sub-split was used for the ICP-MS analysis.Analytical details of the both types of analyses are similar to those described in detail by Calzada-Diaz et al. (2017).Quality control was performed by simultaneous analyses of certified reference materials (CRMs), see Supplementary Table S2.
In addition, selected elements were analysed in the bulkrock by instrumental neutron activation analysis (INAA) at Washington University, St Louis, following the procedure of Korotev et al. (2006).Subsamples were sealed into tubes of ultrapure fused silica (Heraeus-Amersil Suprasil Ò T21, 5mm outside diameter).The principal standards for lithophile elements were chips of two synthetic glasses, designated WU-A and WU-B, that were prepared and standardized against the National Bureau of Standards SRM (Standard Reference Material) 1633a, coal flyash (Korotev, 1987).Each batch contained a sample of SRM 1633a powder that was used as the standard for Sc, Sm, and Ta and a cross-check for the glass standards.Finally, each batch contained 2-3 samples of a synthetic chemical standard for Ir and Au.Three subsamples of $30 mg (32.97, 29.11, 29.11 mg) were analysed.The calculated mass-weighted means, as well as the standard deviation of the three subsamples, are the concentrations that would have been obtained if we had analysed one sample consisting of the entire mass (91.19 mg).

Chlorine abundance and isotopic measurements
Chlorine abundance and isotopic composition were measured using the Cameca NanoSIMS 50L at the OU using a protocol modified after Tarte `se et al. (2014) and Barnes et al. (2016).Analyses were carried out using a Cs + primary beam with a diameter of $1 mm and an accelerating voltage of $16 kV.Each area was pre-sputtered before analysis for a few minutes using a 150 pA probe current over a 7 mm Â 7 mm area to remove any contamination from the surface.Analyses were performed using a primary probe current of 30 pA for $5 minutes over $5 mm Â 5 mm areas.Secondary negative ions of 16 O 1 H, 18 O, 35 Cl, 37 Cl, 19 F and 40 Ca 19 F were collected simultaneously on electron multipliers.Apatite and merrillite were identified with 40 Ca 19 F on real-time secondary ion images during presputtering and to monitor F contents, but was not used for precise quantification of apatite F content, as 40 Ca 19 F has poor ionization efficiency.In order to resolve isobaric interferences a mass-resolving power of $8000 was used, especially for the 17 O and 16 O 1 H peaks.Samples were coated with $20 nm of carbon.Phosphate Cl contents were calibrated using the measured 35 Cl/ 18 O ratios and the known Cl contents of terrestrial apatite standards Atlas Mountain (Ap 004) and Colorado apatite (Ap 005) (McCubbin et al., 2010(McCubbin et al., , 2012) ) which were set in an indium mount.Data obtained for standards are provided in Supplementary Table S3.Uncertainties reported on Cl contents combine the expanded uncertainty (coverage factor k = 2) associated with the calibrations and the analytical uncertainties associated with each individual measurement.

Hydrogen abundance and isotopic measurements
Hydrogen isotopes (D/H) and H 2 O abundances were also measured using the Cameca NanoSIMS 50L at the OU.H, D and 18 O secondary negative ions were quantified using a Cs+ primary beam of $530 pA. 13 C was also collected simultaneously to monitor possible terrestrial contamination.The primary beam was rastered over a 10 mm Â 10 mm and blanking was used to collect signal from only the 5 mm Â 5 mm central area during a total analysis duration of $20 min.Prior to the analysis, the surface was pre-sputtered for 5 minutes using the same beam intensity on a 12 mm Â 12 mm surface area.A H À /O À vs. H 2 O calibration is used to calibrate H 2 O concentrations for apatites, based on Durango apatite (Ap 003) and Atlas Mountain apatite (Ap 004) (McCubbin et al., 2010(McCubbin et al., , 2012)).Background was monitored using 'dry' olivine San Carlos, and analyses were corrected accordingly.To avoid H contamination, analyses were performed with an analytical chamber vacuum of 4.5 Â 10 À10 Torr.The background for H 2 O was estimated to be 10 ppm, based on constant monitoring analyses of dry San Carlos olivine.For hydrogen isotopic measurements, the instrumental mass fractionation factor a, calculated on standard apatites is 1.23 ± 0.05 (n = 19, 2SD).Because the two standards have water contents corresponding to the lower and upper ranges of NWA 10989 apatites, and have a restricted range of instrumental mass fractionation, no matrix effects are expected as a function of apatite water contents.Data obtained for standards analysed are presented in Supplementary Table S4.

Uranium-Thorium-Lead analyses of phosphates
Uranium-Thorium-Lead (U-Th-Pb) dating of six apatites and one merrillite was carried out using the Cameca IMS 1280 at the Institute of Geology and Geophysics, Chinese Academy of Sciences (IGGCAS).The O 2 -primary ion beam was accelerated at À13.8 kV with a current of $10-12nA.The Gaussian illumination mode was used in order to evenly sputter material over the analytical area.The spot diameter was $10 mm Â 15 mm.Positive secondary ions were extracted with a 10 kV potential.Four magnetic field sequences were used to collect secondary ions 40   (Li et al., 2012).The 40 Ca 2 31 P 16 O 3 + peak was used as a reference peak for centering the secondary ion beam as well as for making energy and mass adjustments.NW-1 apatite standard (1160 ± 5 Ma) was used for U-Pb fractionation calibration (Li et al., 2012) and ages were calculated using Isoplot (Ludwig, 2012).Further technical details can be found in Li et al. (2012) and Zhou et al. (2013).Uranium decay constants for U-Pb geochronology were taken from Schoene et al. (2006).

Petrography
NWA 10989 is a fragmental polymict breccia composed of minerals and mineral fragments, various lithic and impact-melt clasts, embedded in a dark-brown glassy matrix (Fig. 1).The mineral components constituting this breccia include feldspar, olivine (forsteritic as well as fayalitic), pyroxene (opx, low-Ca and high-Ca), and trace amounts of silica, ilmenite, troilite, apatite, merrilite, chromite, schreibersite and kamacite.Despite the meteorite being a find, the terrestrial weathering is limited to minor carbonate veining along cracks and fractures.Evidence for shock recrystallization, where present, can be observed as replacement of millimetre-sized feldspar clasts by submillimetre feldspar crystals.Based on texture and mineral abundances, the polished section of the meteorite was broadly divided into four areas corresponding to distinct lithologies (referred in the text as Areas 1, 2, 3 and 4).Higher magnification BSE images of these four areas are shown in Fig. 1(1c-1f).Among these four areas, a range of clast types set in a fine-grained matrix have been identified, and these can be divided into two categories -lithic fragments and impact-melt clasts, highlighted in Fig. 1.Back-scattered electron images of notable lithic clasts and impact melt clasts, are presented in Figures S1 and S2, respectively, in order to highlight the range of lithologies present.Lithic and impact-melt clasts are further subdivided into different groups based on their texture and com-position.Lithic clasts consist of anorthositic troctolite, gabbroic anorthosite, mafic to ultramafic cumulate, granulite and evolved basaltic clasts, based on their relative modal abundances of olivine, pyroxene and plagioclase (Figure S1a), revealing a dominantly feldspathic components.It should be noted that because of the relatively small sizes of these clasts, the observed lithological classification may not truly reflect the parent lithologies from which the clasts were derived (Warren, 2012).Detailed petrography and compositional data of each clast are provided in supplementary information.The representative mineral EMPA compositional data from different clasts are presented in Table S5.The EMPA data for each lithic clast can be found in Supplementary Tables S6 to S9. Impact melt clasts represent both highlands and mare-derived materials.Details on impact melts and matrix glasses are provided in supplementary information, and EMPA data for these are presented in Supplementary Tables S10-S12.Within the studied polished section, we estimate that mafic and feldspathic (non-mare) materials are represented in a ratio of $40:60, characteristic of a lunar fragmental breccia with intermediate-iron bulk composition.Area 1 ($9% area of the polished section) is a glassy area of impact melt, poorly crystallized in some places, containing large but rare vesicles and mineral fragments, some of which are partly resorbed (Fig. 1c).Area 2 is a crystal-dominated breccia with a cataclastic, seriate texture, with minerals and mineral fragments ranging from $1 to 100 s mm in size.Impact melt and lithic fragments up to 1 mm in size are rare (Fig. 1d).Area 3 ($64% of the polished section) is a melt-supported breccia containing almost all impact-melt clasts ($8% of the whole section), lithic clasts ($7% of the whole section) and minerals, up to 1 mm in size (Fig. 1e).There is one large (2 mm Â 4 mm) clast of a crystallized impact-melt.The surrounding melt is dark brown in colour -when viewed under plane polarized light (PPL) -and contains mineral fragments and vesicles.Area 4 contains mineral fragments similar to Area 2, however, there is a higher proportion of matrix, and some lithic clasts are also present (Fig. 1f).Areas 2 and 4 account for 27% of the whole section.The boundaries between Areas 2, 3 and 4 are gradational.
The lunar meteorite NWA 10989 shows textural similarities to other lunar meteorites with intermediate-iron bulk concentration in particular to those that are thought to be binary mixtures.These types of lunar meteorites are defined by Korotev et al. (2009) as mixtures of brecciated anorthosite from Feldspathic Highland Terrane and volcanic basalt from maria, with low contents of incompatible trace elements such as Sm (1-3 lg/g).Currently, only a few meteorites are part of this group: Meteorite Hills (MET) 01210 (Day et al., 2006); the YQN launch pair, that is, Yamato (Y) 793274/981031, Queen Alexandra Range (QUE) 94281, NWA 4884 and Elephant Moraine (EET) 87521/96008), Dhofar 1180 and NWA 3136 (Arai and Warren, 1999;Anand et al., 2003a;Korotev et al., 2003, Korotev andZeigler, 2014).Mount DeWitt (DEW) 12007 has also been classified as a binary regolith breccia, containing Very Low Titanium (VLT) materials and magnesian feldspathic materials, with a suggestion of launch-paring with the YQN group (Collareta et al., 2016).Newly classified lunar breccias NWA 7611 and NWA 7834 could also be added to this binary group (Ruzicka et al., 2015).NWA 10989 contains a higher abundance of feldspathic material than most of the YQN clan, but closer to Y-793274 (Warren and Kallemeyn, 1991;Arai and Warren, 1999), NWA 7611 and NWA 7834 (Ruzicka et al., 2015).S1).On the BSE image, the four areas are highlighted (orange outlines) and are also presented in higher magnification below: 1c -Area 1; 1d -Area 2; 1e -Area 3; 1f -Area 4. Abbreviations stand for minerals: Ol-olivine, Px-pyroxene, Pl-plagioclase.Details are given in the text.(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)the YQN clan, as indicated by the presence of vesicles (potentially relicts of agglutinates; cf.Fig. S2c).

Major silicate minerals
Olivine occurs as euhedral grains which generally appear brown under PPL.Olivines share same characteristic as pyroxenes, i.e. most of the crystals are unzoned and are present throughout the thin section of NWA 10989, with a predominance in Areas 2 and 4. In total, 170 olivine analyses were collected from single mineral grains (n = 134) from the matrix, from lithic clasts (n = 27) and from impact melts (n = 9) across the studied thin section (available in Supplementary Table S13).Olivine Fo (Fo = molar (Mg/ (Mg + Fe)) * 100) content varies between Fo 5 -Fo 76 with an average of Fo 58±15 .Figure S3a shows a histogram of olivine Fo content, and it is apparent that there are 3 groups of olivines: Fo 5-15 , Fo 31-50 , Fo 59-76 ; the majority of olivines (n = 139) being relatively Mg-rich.The few fayalitic olivines (Fo 5 -Fo 15 ) found in the section all belong to the evolved basaltic clasts, while the most Mg-rich olivines are found in the troctolitic clasts (Fo 74 on average).The average Fe/Mn in olivine is 104 ± 11.
Pyroxene crystals occur as blocky crystals.Under PPL they appear white to light brown or green in colour.Under cross-polarised light (XPL), evidence for shock is seen in terms of undulose extinction and sub domains within grains, which would suggest a shock stage S3 based on the classification of Rubin et al. (1997).Most pyroxene grains exhibit exsolution lamellae which are 1-2 mm in width, suggesting slow cooling (Grove, 1982).Similar exsolution lamellae in pyroxene have been observed in VLT basalt clasts in EET 87521/96008, Y-793274, QUE 94281 as well as MET 01210 basaltic breccias, implying this feature, although rare, is common to VLT basalts (Jolliff et al., 1998;Anand et al., 2003a, Arai andWarren, 1999;Day et al., 2006).Pyroxenes occur throughout the thin section of NWA 10989, however, they are more common in Areas 2 and 4. In total, 196 pyroxene analyses were collected from single mineral grains from the matrix (n = 159), from lithic clasts (n = 20) and from impact melt clasts (n = 17) across the studied polished thin section (cf.Supplementary Table S14).Pyroxenes in NWA 10989 show an extensive variation in term of Ca-Mg-Fe composition (cf.Fig. 2a).Orthopyroxene/enstatite, pigeonite, augite and ferroaugite are all observed within our thin section with the total range in pyroxene composition varying between Wo 3-43 En 5-75 Fs 15-64 .Mg-rich orthopyroxene and pigeonite (Wo 7 En 65 Fs 25 -Wo 37 En 41 Fs 22 ) are dominant in feldspathic lithic clasts, revealing a ferroan anorthosite origin of these clasts (James et al., 1989), while Fe-rich augite (Wo 26 En 27 -Fs 47 -Wo 41 En 19 Fs 40 ) are typically found in basaltic evolved clasts (Fig. 2b).Crystallization trends in the Ca-Mg-Fe diagram are typical of mare basalt, with crystallization of Alrich pyroxenes, followed by pigeonites and augites (Fig. 2a).Al/Si vs Fe/Mg plot (Fig. 2c) shows that firstformed orthopyroxenes are Al enriched, while the pigeonite and augite have a constant low Al/Si ratio, due to plagioclase crystallization (Bence and Papike, 1972).Some pyroxenes show core to rim igneous zoning; for example, two pyroxenes in Area 2 have compositions which vary from Wo 8 En 42 Fs 50 to Wo 35 En 40 Fs 25 and Wo 5 En 64 Fs 30 to Wo 12 -En 60 Fs 29 .However, the majority of pyroxene crystals are not zoned, which suggests sub-solidus re-equilibration.The average Fe/Mn in pyroxene is 64 ± 11.
Plagioclase feldspar occurs as both small and large grains (up to 1 mm in size), some are euhedral with 2:1 aspect ratios.Other grains are more rounded or irregular in shape suggesting that they have undergone some physical comminution.Although feldspar crystals occur throughout NWA 10989, they tend to be larger and more abundant in Area 3 of the breccia.Under PPL, feldspars appear as white to beige in colour, often with a dusty or speckled appearance.Small surface cracks are commonly seen across the crystal surface.Under XPL, most of large feldspar crystals show evidence for shock, such as partial maskelynitisation of the crystal, as well as the presence of impact melt within the crystals.Primary twinning is only observed in a few small crystals.According to the shock stage classification (Rubin et al., 1997), these features are consistent with the S3/S4 stage.Some of the larger (mm) feldspar crystals show evidence for fine-grained recrystallization and undulose or wavy extinction.In total 222 distinct EMPA data points were collected on feldspar grains, among which 178 were in the matrix, 24 in the lithic clasts and 20 in the impactmelt clasts (cf.Table S15).The variation in An component (An# = molar (Ca/(Ca + Na + K)) * 100) of these feldspar grains are displayed in an histogram in Figure S3b and in general, the measured An component ranges from An 80 to An 99 .This wide range of plagioclase composition in NWA 10989 overlaps with most of the compositional variations observed in felsdpathic and mare-basalt materials (Papike et al., 1991).Several crystals were observed to have lower An contents, including a small ($5 mm Â 5 mm) crystal in Area 4 with An 82 , and a $0.5 mm crystal within Area 3 with An 80 .There is generally no systematic variation in crystal composition across the sample, with most feldspars having a near average composition.All feldspar grains contain a low orthoclase (Or) component (molar (K/(Ca + K + Na)) * 100), with the maximum Or-component recorded as 2.5.

Minor phases
Oxide phases such as Al-Cr-rich spinels and ilmenite occur as sub-angular to sub-rounded grains within lithic clasts or as single mineral grains in the matrix.Ilmenite compositions do not show much variation throughout the sample with 40-45 wt.% FeO, 52-55 wt.% TiO 2 , 0-3.5 wt.% MgO.All of the spinels are chromite, a few of which can be categorized as aluminium chromite with compositions varying between 35-45 wt.% Cr 2 O 3 , 7-17 wt.% Al 2 O 3 , 1-10 wt.% TiO 2 , 2-5 wt.% MgO and 2-36 wt.% FeO. Figure S4 shows the 2Ti-Al-Cr ternary diagram, the Ti# vs. Fe# and Cr# vs. Fe# plots with a comparison with spinels from Apollo basalts and highlands samples.The spinel compositions bear a resemblance to both the Mg-Suite spinels as well as basaltic spinels (Papike et al., 1991).
Apatite is the only major phosphate mineral in this meteorite followed by minor merrillite.Apatite generally occurs as euhedral or blocky isolated mineral grains in the matrix, and has a size range of $15-50 mm.The exceptions are an apatite crystal that occurs in association with a fayalite-hedenbergite-silica symplectite (cf. Figure S1b), and two apatites that are intergrown with merrillite at the crystal edges.These apatites are not associated with an impact melt or any other shocked phases, and as such they appear to display primary igneous textures with no evidence for recrystallization.Chemical composition of apatites as measured by EMPA is presented in Table 1.All apatites have a similar chemical composition, although apatite #10 is depleted in F and Cl compared to the other grains.SEM images of all apatites are presented in figure S5.One merrillite crystal was also found (Figure S5).

Bulk composition
The whole-rock major-and trace-element composition of NWA 10989 as determined by INAA and ICP-OES/ ICP-MS are presented in Table 2. Bulk-rock chemistry of NWA 10989 supports the intermediate nature of this lunar breccia, in agreement with petrologic observations of a roughly equal proportion of feldspathic and basaltic/mafic materials (e.g.60:40).Firstly, FeO content (i.e.12.6 ± 0.8 wt.% -INAA) and Al 2 O 3 content (18.5 ± 0.3 wt.%) of NWA 10989 fall in the compositional criteria established by Korotev et al. (2009) (7-17% FeO, 13-20% Al 2 O 3 ) for lunar breccia of intermediate-iron concentration.Secondly, the whole-rock chondrite-normalized rare earth element (REE) abundances (Fig. 3) show a typical signature of feldspathic-mare mixture with a slight negative Eu anomaly (Eu/Eu* = 0.67), and LREE enrichment compared to HREEs (e.g.$ 39 Â CI La and $ 37 Â Lu, La/Yb (CInormalized) = 2.05).Finally, Fig. 4 shows Sm vs Sc plot that also positions NWA 10989 as a lunar meteorite of intermediate-iron composition.For the bulk-rock, the low Na 2 O (i.e.0.38-0.42wt.%),K 2 O (0.10 wt.%) and TiO 2 (0.59 wt.%) contents, as well as the low/medium Th content of 1.04 lg/g suggest a relatively minor contribution of KREEP-rich material to NWA 10989, consistent with petrographic observations.This low bulk-rock TiO 2 content also suggests that Ti-rich basaltic material is absent from this meteorite, again consistent with petrographic observations.Using the 'Apollo model' compositional end members from Korotev et al. (2009), a simple mass balance calculation suggested NWA 10989 to be a 45:52:3 mixture of mare:feldspathic:KREEP material.This is broadly comparable to the estimate made from petrographic observations ($40:60 ratio of mafic and feldspathic material).As such, it would appear that we have studied a polished section which is representative of the bulk sample.The bulkrock Ir (4.3 ng/g -equivalent to 0.7% CM chondrite (Anders and Grevesse, 1989)) and Ni (189 lg/g) contents are low, suggesting a minimal contribution by meteoritic material.
The lunar meteorite NWA 10989 shows geochemical similarities to the anorthosite-basalt mixtures (i.e. the YQN launch pairs, Dhofar 1180 and NWA 3136, DEW 12007, NWA 7611 clan, NWA 7834 clan; Korotev et al., 2009;Ruzicka et al., 2015).They contain very low incompatible trace elements such as Sm (1-3 lg/g) and thus do not contain a substantial KREEP component; in the case of NWA 10989 the estimated KREEP component is less than 3%.The YQN group that consists of VLT and magnesian felsdpathic components, are chemically very close to NWA 10989 (Korotev and Zeigler, 2014).Indeed, these meteorites plot on a mixing trend between VLT basalt and a nonmare ($feldspathic) component, more magnesian than any felsdpathic lunar meteorites (Korotev et al., 2003).In particular, NWA 10989 is extremely close to QUE 94281 which has a FeO content of $13-14 wt.% and a Al 2 O 3 content of 15.8 wt.% (Jolliff et al., 1998;Arai and Warren, 1999;Korotev et al., 2003), as well as its launch pair Y-793274/981031 because of the higher felsdpathic content of this meteorite compared to the other ones of the YQN group (Warren and Kallemeyn, 1991;Jolliff et al., 1998;Arai and Warren, 1999) .Interestingly, it is also close to NWA 5651 (FeO = 12.7 wt.% and Al 2 O 3 = 17.3 wt.%; Korotev et al., 2009) which is a fragmental to melt-matrix breccia (Korotev et al., 2009).Two other clans of lunar breccias, the NWA 7611 clan and particularly the NWA 7834 clan have composition closer to NWA 10989 (cf.Fig. 4, Ruzicka et al., 2015) than the YQN clan.For instance, NWA 7834 have a FeO content of 12.9 wt.% (compared to 12.6 ± 0.08 wt.% for NWA 10989), a Sc content of 25.0 lg/g (compared to 22.5-25.5 lg/g), Sm content of 3.1 lg/g (compared to 3.5-4.2lg/g for NWA 10989) and a Th content of 0.9 lg/g (compared to 1.04-1.57lg/g).This meteorite and its launch pairs are also feldspathic breccias, thus with lower mare material, suggesting that NWA 10989 could be another paired meteorite of NWA 7834 clan.
In the following sections, we discuss composition of the basaltic material, predominantly in Areas 2 and 4; and composition of felsdpathic material of NWA 10989, mainly represented in Areas 3 and 4 (cf.Fig. 1) by impact melts, lithic clasts and matrix components.
The Fe# vs. Ti# of pyroxenes characterize trends indicative of parent magma composition (Nielsen and Drake, 1978).Two major trends on Fe# vs Ti# plot are observed among the pyroxene fragments analysed in NWA 10989 matrix, as well as the pyroxene grains within lithic clasts and impact-melt clasts (Fig. 6).The main trend defines a correlation between Ti# and Fe#, typical of basaltic magmas, and is similar to that seen for VLT basalts as meteorite data plot close to the trend defined by Apollo 17 (Vaniman and Papike, 1977) and Luna 24 (Nielsen and Drake, 1978;Meyer et al., 1978) VLT basalts.Pyroxenes in NWA 10989 that fall on the VLT trend show also low alkali contents typical of Apollo 17 VLT pyroxenes (Na 2 O + K 2 O < 0.2 wt.%; Vaniman and Papike, 1977).However, Ti# and Fe# of NWA 10989 pyroxenes not define a single magmatic field but the data are scattered across fields defined by the mare basalts.These basaltic pyroxenes can be separated in two sub-trends, one of more mafic (i.e.Fe# < 0.5) and one of more ferroan (Fe# > 0.6) types.These two subtrends have been observed previously in the basaltic polymict breccia Y-793274 (Arai et al., 1996).Indeed, these trends are presented for all breccias of the YQN clan (Arai and Warren, 1999;Anand et al., 2003a;Korotev et al., 2003, Korotev andZeigler, 2014), as well as presented in other intermediate-iron bulk composition breccias such as MET 01210 (Day et al., 2006).Mineralogically, the composition of pyroxenes from NWA 10989 show similarity with VLT materials of EET98/96 and Y-793274 from the YQN clan, albeit pyroxenes in NWA 10989 contain $1 wt.% less Al 2 O 3 .Moreover, the relatively thick exsolution lamellae ($1 mm) in pyroxenes are also observed in basaltic clasts of EET 87521/96008, Y-793274, QUE 94281 as well as MET 01210 (Jolliff et al., 1998;Anand et al., 2003a, Arai andWarren, 1999;Day et al., 2006;Arai et al., 2010).
Using the correlation between bulk-rock TiO 2 content of low-Ti mare basalt and pyroxene Ti# 0.5 (i.e. the Ti# of pyroxene at fixed Fe# = 0.5) defined by Arai et al. (1996), we estimated the bulk TiO 2 content of NWA 10989 basaltic source.As shown in Fig. 7, pyroxenes of NWA 10989 predict a parent magma composition containing 1.0 ± 0.2 wt.% TiO 2 .This estimated value is higher than the bulk-rock TiO 2 content determined by ICP-OES (0.6 wt.%;Table 2).This relatively large difference is almost certainly a result of the intermediate nature of NWA 10989, and consequently to the large amount of felsdpathic material that was represented in the measured bulk composition.Moreover, this estimated bulk magma TiO 2 content is similar to the average bulk TiO 2 content of the impact melt glass measured in Area 1 (i.e.1.15 ± 0.07 wt.% TiO 2 ), as well as most of glasses in Area 2. The Area 1 impact melt is thus postulated to be representative of the bulk composition of the VLT basalt component in this meteorite (Table S5), characterized by TiO 2 content of $1 wt.%.This number is comparable to estimation made for the YQN basaltic meteorite group (Arai et al., 1996;Arai and Warren, 1999;Anand et al., 2003a), confirming NWA 10989 being a mixture of mare basalts, with an important contribution derived from melting of VLT basaltic reservoir.

Feldspathic materials
The felsdpathic composition is characterized by impactmelt clasts, glasses, mineral fragments and also most of lithic clasts.From the EMPA analyses on the matrix glasses (cf.Table S5) in Area 3, which constitutes $64% of the section, and Area 4 account for majority of the feldspathic components in the meteorite.They define a linear contin-Fig.6. Plot of #Fe versus #Ti for pyroxenes in the evolved basaltic clasts, impact-melt class and matrix.Area representing typical pyroxenes compositions for VLT, low-Ti and high-Ti lithologies are shown as envelopes using different shades of grey.The VLT field outlined includes the data for Apollo 17 VLT pyroxenes (Vaniman and Papike, 1977), Luna 24 pyroxenes (Meyer et al., 1978) and pyroxenes from several lunar meteorites (Arai et al., 1996;Anand et al., 2003a).
Several measurements made across a glassy impact clast (up to 750 lm; Figure S2), give an average composition (N = 14) of 19.5% Al 2 O 3 and 10.3% FeO, similar to the noritic end of glass composition.This mafic end member of the linear trend does not correlate with any known Apollo glass composition.However, it is compositionally similar to the noritic glasses reported in PCA 02007 (i.e.$18 wt.% Al 2 O 3 , $14 wt.% FeO; Day et al., 2006;Joy et al., 2010), the norite composition in 77215 (Chao et al., 1976) and the glass vein found in QUE 94281 (i.e.20 wt.% Al 2 O 3 , 11 wt.%FeO; Jolliff et al. 1998).
The average An# of plagioclase is plotted as a function of the average Mg# of associated mafic minerals (i.e., olivine and pyroxene) for each feldspathic clast and impactmelt clast (Fig. 8).The four gabbroic anorthosites and anorthositic gabbros show an affinity to Ferroan Anorthositic (FAN) composition, typical of Apollo 16 samples (McGee, 1993) but rarer for lunar felsdpathic breccia (Gross et al., 2014).The two anorthositic troctolites fall in between the FAN suite and the Mg-suite, as seen among many lunar meteorites (Gross et al., 2014).Indeed, it has been observed that mixed meteorites contain more sodic plagioclase and less mafic silicates than pristine highland Mg-suite rocks and FAN rocks (Demidova et al., 2007).The large impact melt has a similar FAN composition than the gabbroic anorthosites and anorthositic gabbros, and is probably derived from a similar FAN reservoir.Finally, for the impact-melt clasts, one plot along the Mg-suite trend, due to the more sodic nature ($An 90 ) of their feldspars, which is again not so common among felsdpathic meteorites (Gross et al., 2014).The others are like the troctolites, in between the range of Mg-suite and FAN suite.The feldspathic material in NWA 10989 tends to be associated with slightly more mafic minerals than typical Apollo 16 anorthositic highlands material, similar to the highly magnesian anorthositic trend identified in the majority of feldspathic meteorites (Gross et al., 2014).Thus, the majority of the felsdpathic material from NWA 10989 appears to be derived from two sources, one being derived from FAN material while the other plots in the mafic range of Apollo 16 FAN material.

Chlorine abundance and isotopic composition
In total nine measurements were made across six apatite grains from this meteorite.The chlorine isotope measurements and abundances are plotted in Fig. 9, and presented in Table 3.All apatite occur as isolated grains in the matrix, with no definitive petrographic context, except for apatite 9 which was located inside the symplectite clast (cf. Figure S5).The matrix apatite show a limited range in d 37 Cl between +14.4 and +18.7‰ with an average value of +15.9 ± 1.5‰ (2SD).Chlorine contents vary between 1270 and 4880 ppm with an average value of 2700 ± 1540 ppm.The apatite grain 9 in the symplectite clast displayed a range in d 37 Cl and Cl contents.The core of the crystal exhibits d 37 Cl of 11.9 ± 0.9‰ and 2700 ± 200 ppm Cl (average of 9a and 9c -Table 3), while the rim is more Cl rich (i.e.6050 ± 630 ppm Cl), associated with heavier d 37 Cl value of 16.8 ± 1.5‰.

Hydrogen abundance and isotopic composition
Hydrogen isotopic ratio and concentration were also measured in eight apatites, among which six of them were also analysed for chlorine (cf.Table 3), and one merrillite (#11), as shown in Fig. 10.In total, seven out of nine phosphates exhibit low dD signatures, with values ranging from À420 ± 90 to +210 ± 90‰.Water content associated with these apatites range from +260 ± 20 to +1580 ± 120 ppm H 2 O. On the other hand, apatites #1 and #10, show high water contents of 0.3 and 0.9-1 wt.% H 2 O, respectively, along with one of the highest ever measured dD in an apatite from a lunar meteorite apatite (from +410 ± 70 to +1010 ± 90‰) (Tarte `se et al., 2013).It should be noted that these H 2 O measurements are not similar to the OH determined by stoichiometry in Table 1, which is most likely because of overestimation of F by EMPA.No spallation correction has been made on these analyses as the CRE age of NWA 10989 is currently unknown.However, Tarte `se et al. ( 2014) estimated that based on the older CRE age of lunar polymict breccia NWA 4472 (i.e. 2 ± 0.2 Ga, Lorenzetti, 2005), for an apatite containing $1800 ppm H 2 O, the maximum correction for dD is $60‰.Although in some cases, the measured water contents are lower than 1800 ppm, we assume that the spallation correction is within our stated uncertainties of each Fig. 7. Plot of Fe#-normalized Ti# versus bulk-rock TiO 2 content for Apollo mare basalts and basaltic meteorites (modified after Arai et al., 1996;Anand et al., 2003a).measurement.In any case, we recommend that the reported dD values for low-water apatite be treated as the upper limit as the corrected dD values will be displaced towards lighter values.Lunar meteorites are susceptible to terrestrial contamination and therefore often display dD values closer to terrestrial water (Tarte `se et al., 2013).However, (i) the relatively high water concentrations in apatites (Stephant et al., 2018), (ii) the high dD values of phosphates compared to terrestrial dD value, as well as (iii) the lack of any appar-ent terrestrial contamination trend on the dD vs H 2 O graph, allows us to rule out any major terrestrial contamination affecting the H isotope systematics of phosphates this sample.

Lead-lead (Pb-Pb) age of apatites and merrillite
Ages of apatites was carried out in this study using Pb-Pb method (Table 4).The concentrations of U and Th in 8. Average An# in plagioclases as a function of average Mg# of associated mafic minerals (olivines and pyroxenes) in impact-melt clasts and feldspathic lithic clasts.Each marker represents a single clast, except for the large impact-melts where markers represent two areas.Errors associated are standard deviation of both plagioclase An# and mafic mineral Mg# in the clast.The areas of FAN and Mg suite are delineated with grey dashed lines and are from Papike et al. (1998).The grey area represents meteorite anorthositic clasts and are from Gross et al. (2014).Fig. 9. Chlorine isotopic composition (‰) and (ppm) in NWA 10989 compared lunar meteorite NWA 4472 and to Apollo samples.Red symbols are for NWA 10989 apatites with ages ranging from 3.98 to 4.20 Ga while orange symbols are for apatites with ages ranging from 3.32 to 3.96 Ga.Squares represent NWA 10989 matrix apatites, while triangles represent the symplectite apatite.Lunar meteorite data are from Tarte `se et al. (2014); Apollo data are from Boyce et al. (2015), Barnes et al. (2016), Potts et al. (2018).
Table 3 Chlorine and hydrogen isotopic ratios expressed in delta values (‰), along with Cl, F, H 2 O ppm abundances and the Pb-Pb dates of 9 apatites from NWA 10989.Apatite 9 is from a symplectite (Figure S1b) while the other apatites are from the matrix.d 37 Cl, Cl content, F content and dD, H 2 O content were measured in 2 different NanoSIMS sessions (see methodology for details).

Phosphates
Age (Ga)   (2014) and data for KREEP-rich evolved rocks are from Robinson et al. (2016).The legend is same as in Fig. 9. apatite range from 12 to 187 ppm and 16 to 406 ppm, respectively, with Th/U ratios in the range of 1.3-2.5.The single merrillite has a U and Th concentrations of 40 and 640 ppm, respectively, which gives a Th/U ratio of 15.8.The Pb-Pb ages are discordant from the concordia curve and apatites do not define one discordia line.As such, these phosphates probably have experienced some Pb loss due to a shock event and do not seem to have a single crystallization age.If Pb was lost from apatites, their true crystallization age will be higher than the recorded Pb-Pb age.
Apatites #1 and #10, defined as ''typical" mare basalt apatites based on dD-d 37 Cl systematics (Fig. 11), have older Pb-Pb age than the rest of the apatites at 4200 ± 13 and 3984 ± 42 Ma, respectively.The rest of NWA 10989 apatites and the merrillite, similar to typical Mg and Alkali suite in terms of dD-H 2 O systematics, record younger Pb-Pb ages from 3323 ± 9 to 3962 ± 30 Ma.

Lunar origin
Several lines of evidence have been used to confirm a lunar origin for NWA 10989.First of all, the Fe/Mn ratios of mafic minerals, i.e. olivines and pyroxenes, commonly used to discriminate between different planetary sources (Papike et al., 1998), are consistent with the known lunar trends (Karner et al., 2003, Karner et al., 2006) (Figure S6).Secondly, bulk-rock oxygen isotope analysis of this meteorite yielded the following values: d 17 O 3.42‰, d 18 O 6.51‰, D 17 O 0.01‰.The oxygen isotopic composition of NWA 10989 plots on the terrestrial fractional line (Fig. 12), although the d 17 O and d 18 O values are shifted towards slightly higher values compared to those for Apollo samples (Hallis et al., 2010;Wiechert et al., 2001;Spicuzza et al., 2007;Herwartz et al., 2014;Young et al., 2016).Indeed, the oxygen isotopic composition of NWA 10989 is in the range of lunar meteorites (Fig. 12), in which the elevated oxygen isotopic values could be ascribed to presence of terrestrial weathering products formed in hotdesert environments (Stelzer et al., 1999;Anand et al., 2003b;Pillinger et al., 2013); NWA 10989 is also a hotdesert meteorite but its terrestrial residence time is currently unknown.Finally, the similarity between the mineralogy and geochemistry of minerals and lithic fragments seen in NWA10989 to Apollo and other lunar meteorite samples provides further evidence for a lunar origin.
The increase of Cl content from core to rim of the apatite in the symplectite appears consistent with fractional crystallization.The slight increase of d 37 Cl observed between core and rim could be explained by magmatic degassing of metal chlorides, although the difference remains within the variation observed among all the apatites in this meteorite.The reason for heavy lunar d 37 Cl compared to terrestrial value was first suggested to be due to degassing of the lunar magmas during and upon eruption (Sharp et al., 2010), which would be consistent with slight increase of d 37 Cl observed between core and rim of apatite 9.More recent studies have advocated for a mixing between two isotopic reservoirs: one with an elevated signature (i.e.+25 to +30‰) originally associated with the urKREEP reservoir and another with a lighter signature ($0‰) representative of mare-basalt source regions (Boyce et al., 2015;Barnes et al., 2016).
Apatites in NWA 10989 form two clusters in term of their H 2 O-dD systematics (Fig. 10).These clusters may appear at first as a positive trend, implying that a magmatic or secondary process affected apatites from a single volatile reservoir.However, H 2 degassing of melts from a single reservoir would induce a negative trend and no known magmatic process could explain this trend.Moreover, no such trend is observed for chlorine systematics, which clearly ruled out any type of magmatic degassing.As stated previously, we do not expect a spallation contribution higher than 60‰ suggesting that these two clusters could in fact be signatures of two distinct volatile reservoirs.Finally, implantation of solar wind may play a role (Treiman et al., 2016) in lowering dD signatures of a basaltic melt, but tin such case, we would expect all apatites to have being affected.As such, we propose that these two clusters are indeed a signature of two distinct volatile reservoirs.
The dD-H 2 O systematics of apatites #1 and #10 are in the range of what has been measured for Apollo mare basalts, i.e. high dD (>$+500‰) (Tarte `se et al., 2013;Barnes et al., 2013), in agreement with their chlorine isotope signature (Fig. 11).However, the other six apatites and the merrillite with lower dD values (i.e.<$+200‰) and lower H 2 O contents (i.e.< $1600 ppm) fall in the dD-H 2 O systematic range of values measured previously for Mg-and alkali-suites (Barnes et al., 2014;Greenwood et al., 2011) (Fig. 10), i.e. a low-dD reservoir, while their chlorine signature indicate a basaltic reservoir.Indeed, their chlorine isotope signature are similar to ''typical" Apollo mare basalts (Sharp et al., 2010;Boyce et al., 2015;Barnes et al., 2016), and no characteristic d 37 Cl signature associated with highland material (e.g.$+30‰, has been recorded in dD-poor apatites from Apollo samples).Moreover, as stated previously, it is unlikely that phosphates sampled a feldspathic reservoir.Indeed one of these apatites is located in the symplectite, which represents an evolved basaltic clast.It should be noted that three apatites in low-Ti mare basalt 12040 have been reported (Boyce et al., 2015;Treiman et al., 2016) to have lower dD values than seen in typical mare basalt.However, because of extremely low water contents in these apatites, only one analysis could be considered significant, with a dD value of À20 ± 200‰.At this stage, it is tenuous at best to make any detailed comparisons with this single data point, and additional data on this sample are required.Regardless, 12040 most probably sampled a different low-dD reservoir than NWA 10989, as the lack of KREEP component places the source region for this lunar meteorite far from Apollo landing sites.However, two basaltic meteorites, the monomict VLT breccia Kalahari 009 and the KREEP-rich breccia NWA 4472 have similar dD-H 2 O systematics (Tarte `se et al, 2014).Interestingly, the polymict lunar breccia NWA 4472 also displays similar clustering in Cl and H isotope space (Tarte `se et al., 2014) (cf. Fig. 11).These two Fig. 12.The oxygen isotopic composition NWA 10989 plots along the terrestrial fractionation line, also defined by other lunar meteorites (cf.Meteoritical Bulletin) and Apollo samples (Wiechert et al, 2001).NWA 10989 has slightly heavier isotopic values than other lunar meteorites, comparable to Dhofar 287A (Anand et al., 2003b).
clusters have been associated with two different reservoirs in the Moon, i.e. mare-basalt source regions for which the elevated dD values in apatites are thought to be a result of magmatic degassing and KREEP-rich basaltic reservoir associated with low-dD apatites.Therefore, at first, it would seem likely that apatites from both NWA 10989 and NWA 4472 sampled similar reservoirs for Cl and H.However, as stated before, NWA 4472 is a KREEP-rich breccia (Joy et al., 2011) while NWA 10989 has relatively low KREEP component (<3%).Moreover, apatites from KREEP-rich lithologies have even lower H 2 O content than what is recorded in NWA 10989 apatites (Robinson et al., 2016;cf. Fig. 10), which are more similar to apatite water contents in Mg and alkali suite lithologies (Barnes et al., 2014).

Combination of phosphate volatile compositions and age dating
Age dating of apatites and merrillite can help us investigate further the potential lunar reservoirs sampled by NWA 10989 (Table 4), although these Pb-Pb ages have to be interpreted with caution as they may not represent true crystallization ages.In case of NWA 10989 apatites, there is no obvious evidence for shock (e.g.no association with other shocked phases or impact melt) which could have reset the Pb-Pb ages.Apatites #1 and #10, defined as ''typical" mare basalt apatites based on dD-d 37 Cl systematics (Fig. 11), have older Pb-Pb ages compared to the rest of the apatites at 4.20 and 3.98 Ga, respectively.Thus, they are considered relics of old basaltic volcanism (cryptomare) on the Moon, as crystallization ages for majority of marebasalts range from 2.9 to 3.9 Ga; the oldest mare basalt (thought to be a sample of cryptomare) sample being recorded by the monomict VLT breccia Kalahari 009 (Terada et al., 2007;Snape et al., 2018) at 4.35 Ga.The ''typical" mare basalt apatites in NWA 4472 has also recorded an old crystallization age of 4.35 Ga similar to Kal 009 (Joy et al., 2011;Tarte `se et al., 2014).However, while NWA 4472 and NWA apatites #1 and #10 have similar dD-d 37 Cl systematics, apatites in Kal 009 have a low dD signature, suggesting that these three meteorites (i.e.Kal 009, NWA 4472 and NWA 10989) contain components of ancient volcanic materials on the Moon that either had a distinct source region characteristic for H and Cl isotopes or post-magmatic processes affecting isotope fractionation were different than envisaged for Apollo samples.
The rest of NWA 10989 apatites and merrillite with low dD record younger ages from 3.32 to 3.96 Ga.These ages correspond to the main phase of mare-basalt magmatism, as Mg and alkali-suite samples generally have crystallization ages older than 4.17 Ga (Shearer et al., 2015).These Pb-Pb ages are also younger than ages reported for KREEP-rich basaltic components in NWA 4472 (i.e. 3.93-4.07 Ga;Joy et al., 2011), yet they have similar H and Cl isotope composition.Moreover, the average Th and U contents of NWA 10989 apatites are 57 and 183 lg/g, respectively, similar to apatite in lunar basaltic meteorites EET96008 and LAP 02205 (Anand et al., 2003a;Anand et al. (2006)), which are higher than in apatite from anorthositic material (Norman and Nemchin, 2014).Thus, the Th and U contents, added to the chlorine isotopic composition and the ages of these seven apatites, are in agreement with a mare heritage.As such, if these apatites are indeed from a mare source as most pieces of evidence seem to indicate, it highlights the existence of a mare reservoir which is depleted in D relative to typical mare basalts from Apollo collections, and even lower than what have been measured for Apollo low-Ti basalt 12040 (Boyce et al., 2015;Treiman et al., 2016) and Kalahari 009 (Tartese et al., 2014).Indeed, if our inference is correct of a mare heritage for apatite in NWA 10989, data presented here would represent the lowest dD ever recorded in a mare basalt sample (i.e.$À400‰).Our work also shows that to infer the origin of lunar breccias, and the multiple reservoirs of volatiles that their individual components may have sampled, is not trivial.A combined approach based on H, Cl and age data on apatite is preferable, especially for apatites lacking a petrologic context.
Apatites #1 and #10 display signatures typical of mare basalts but their crystallization ages at $4-4.2 Ga makes them some of the oldest examples of volcanic products on the Moon (Terada et al., 2007;Snape et al., 2018).The rest of the apatites seem to display signatures more akin to KREEP-rich basalts rather than Mg or alkali-suite as the chlorine isotopes are much less fractionated than is the case for typical highlands samples, as well as the younger ages of these apatites (i.e.3.3-3.9).However, no obvious KREEPrich basaltic component has been found in NWA 10989.As a result, it seems likely that NWA 10989 has sampled a mare basalt reservoir that has not been identified for existing lunar samples.
To account for the differences in these low dD signatures in some mare basalts compared to typical mare basalt dD signature, several hypotheses have been put forward.Elevated dD in lunar basaltic apatites have been argued to have resulted from magma degassing (Fu ¨ri et al., 2014;Tarte `se et al., 2014).On the other hand, assimilation of regolith containing solar wind H has been argued as a possibility to account for D-depleted signatures in mare basalts (Treiman et al., 2016).

Possible source regions of NWA 10989
The lunar breccia NWA 10989 is a lunar meteorite of intermediate-iron bulk composition composed of two distinct compositional sources -a magnesian highlands reservoir, mostly similar to Apollo FAN material, although more mafic, and a mixture of mare basalt reservoirs, one being similar to VLT basalts.As implied by its binary nature, NWA 10989 probably originated like other felsdpathicmare breccias of intermediate-iron concentration from a region of the Moon where mare basalt has mixed with felsdpathic highland terrane (Korotev et al., 2009;Jolliff et al., 2000).Mare-highland boundaries do exist on the lunar nearside, potentially close to the Luna 24 or Apollo 17 landing sites where VLT basalts and FAN materials, similar to Apollo 16 materials, have been found.However, in NWA 10989, there is no substantial lithological evidence for KREEP-rich basalt material, despite 'Apollo Model' (Korotev et al., 2009) mass balance suggesting up to 3% KREEP, with 45% mare material and 52% feldspathic material.Indeed, all Apollo breccias consisting mainly of anorthositic material and basalt also contain a small KREEP component.The 'Apollo model' does not distinguish between mare basalts and mafic lithologies which may occur within highlands rocks.As such, in NWA 10989, the incompatible elements may be supplied by this mafic non-mare component (cf.Fig. 9), supported by the presence of ultramafic cumulate clasts; component which doesn't exist in the Apollo samples.As a result, we infer that NWA 10989 was derived from a mare-highland boundary region distant from the Procellarum KREEP Terrane on the Moon, similar to other binary brecciated anorthositic and mare-basalt breccias (Korotev et al., 2009).
Hydrogen and chlorine isotopic compositions of NWA 10989 apatites show a strong resemblance with the KREEP-rich breccia NW4472 (cf.Fig. 11).However, as stated previously, NWA 10989 does not contain any significant KREEP-rich basaltic material.Combined analyses of dD-d 37 Cl with age dating of apatites highlight two groups of apatites: one group which sampled a typical mare basalt-type source (i.e.dD > +400‰, d 37 Cl $ 16‰), while the other group sampled an unusual basaltic reservoir, with dD values as low as À433 ± 88‰, similar to the older Mgsuite rocks from Apollo collections, but with chlorine isotope signature more akin to younger mare basalts.This evidence strongly argues for the presence of material in this breccia, which is either not yet identified in the Apollo collections or more likely represents an area of the Moon not sampled by the Apollo missions.As a result, a more prob-able source location for this breccia would be on a mariahighlands boundary on the lunar farside.

CONCLUSION
NWA 10989 is a lunar meteorite of intermediate-iron bulk composition with a mixture of $40% mare basaltic material and $60% feldspathic material.This meteorite does not contain any significant KREEP component which is in contradiction with estimation of a KREEP component based on the canonical 'Apollo Model', suggesting that this model may not be universally applicable to all lunar breccias.Based on hydrogen and chlorine isotopes, as well as age dating of apatite grains, NWA 10989 appears to have sampled at least two distinct mare basalt reservoirs on the Moon.Some apatites come from an ancient cryptomare reservoir, while the other apatites come from a Ddepleted mare basalt reservoir relative to the source regions of typical mare basalts from Apollo collections, which has not been yet identified in lunar sample collection.These new data on lunar volatiles strongly favour the hypothesis that multiple volatile reservoirs are present on the Moon.NWA 10989 is thus an important lunar breccia, which most probably sampled a region different from any of the Apollo landing sites, possibly near a mare-feldspathic boundary on the farside of the Moon.

Fig. 1 .
Fig.1.Optical plane-polarized light (PPL) (1a) and BSE (1b) images of NWA 10989.On the optical image, lithic and impact melt clasts have been marked, each colour corresponds to a type of clast: white -impact melt clasts, red -mafic cumulate clasts, orange -evolved basaltic clasts, green -troctolite clasts, blue -norite clasts and yellow -granulite clast (higher magnification images of noteworthy clasts are shown in Fig.S1).On the BSE image, the four areas are highlighted (orange outlines) and are also presented in higher magnification below: 1c -Area 1; 1d -Area 2; 1e -Area 3; 1f -Area 4. Abbreviations stand for minerals: Ol-olivine, Px-pyroxene, Pl-plagioclase.Details are given in the text.(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 4 .
Fig. 4. Bulk-rock Sm versus Sc contents of NWA 10989.Other lunar meteorites have been plotted for comparison including the other meteorites of intermediate-iron composition (modified after Korotev et al., 2009).

Fig. 5 .
Fig. 5. Plot of Al 2 O 3 versus FeO for matrix glasses and the large glassy impact melt from the 4 different areas of NWA 10989.The glasses exhibit a continuum from pure anorthositic to basaltic composition.The bulk-rock composition of NWA 10989 is also plotted.

Table 1
(Korotev et al., 2009)patite analysed in NWA 10989.OH is calculated by stoichiometry.Bulk-rock major-and trace-element composition of NWA 10989.Data from ICP-OES/ ICP-MS and INAA are based on 3 replicates and 3 subsamples, respectively.Major-element oxide (except K 2 O) as well as Cr concentrations were determined by ICP-OES, the rest of the elements, by ICP-MS.Mass-weighted data are reported for INAA.Variations in INAA and ICP-MS data are likely to be because of difference in relative proportion of felsdpathic vs basaltic clasts in subsamples, due to the intermediate nature of the sample(Korotev et al., 2009).