Preservation of primordial signatures of water in highly-shocked ancient lunar rocks

a School of Physical Sciences, The Open University, Walton Hall, Milton Keynes, MK7 6AA, United Kingdom b Centre for Applied Planetary Mineralogy, Department of Natural History, Royal Ontario Museum, Toronto, M5S 2C6, Canada c Department of Earth Sciences, University of Toronto, Toronto, M5S 3B1, Canada d Department of Earth Sciences, The Natural History Museum, London, SW7 5BD, United Kingdom e University of Portsmouth, School of the Environment, Geography and Geosciences, Burnaby Road, Portsmouth, PO1 3QL, United Kingdom

Water in apatite has been thoroughly investigated in most types of lunar basalts ranging from high-Ti and low-Ti basalts (e.g.  to high-Al and KREEP-rich varieties, where KREEP stands for potassium (K), the rare earth elements (REEs), and phosphorus (P) (e.g., Greenwood et al., 2011;Tartèse et al., 2014b;Treiman et al., 2016). However, apatite in most mare basaltic samples crystallized at a late stage (>95% crystallization), by which time their parental melts were likely to have undergone significant H 2 -degassing during ascent and upon eruption. This degassing could have induced isotopic fractionation of hydrogen (H) from deuterium (D) (e.g., . The primary lunar crust, formed directly through Lunar Magma Ocean (LMO) solidification, is predominantly composed of anorthosites that do not contain apatite. Subsequently, incompatible-element-rich and Mgrich magmas were emplaced into the overlying anorthositic lunar crust as deep crustal cumulates giving rise to Mg-suite rocks (summarized in Shearer et al., 2015). These rocks are thought to have better preserved the H isotopic composition of their parent melts (Barnes et al., , 2014. Thus, the Mg-suite rocks are ideal targets for ascertaining the H isotopic composition of indigenous lunar water. We remain cognisant of the fact that some models suggest an impact origin for at least a portion of Mg-suite rocks  and references therein), most recently for the troctolite 76535 (White et al., 2020). Altogether, the impact models suggest the cumulates could have crystallized in insulated, slowcooling large melt sheets or pooled beneath the anorthositic crust (summarized in Shearer et al., 2015). However, no research has yet been carried out to investigate the effect of the large-scale impact melting on indigenous lunar volatile signature, such as might be the case for Mg-suite rocks, if they indeed crystallized from large melt sheets.
Regardless of their petrogenetic mode of formation, another important aspect of Mg-suite samples is that they have experienced certain levels of shock deformation after their initial crystallization. Experimental studies at elevated temperatures have shown H diffusion in apatite to be much faster than for all other elements, including Pb (Higashi et al., 2017), which would be relevant for apatite exposed to shock metamorphism. Apart from investigation of water in variably shocked anorthosite (Hui et al., 2017), this important aspect of lunar samples' history remains underexplored. To our knowledge, no similar efforts to understand the effect of shock deformation on volatiles in lunar samples have yet been undertaken, even though a clear need for this approach has been advocated . This is reinforced by observations of impact-associated volatile mobility within martian (Howarth et al., 2015) and terrestrial (Kenny et al., 2020) apatite.
Prior to this study, apatite water contents and H-isotopic compositions have only been investigated in three samples of Mg-suite lithologies (Barnes et al., 2014). These authors report water contents from <100 ppm to ∼2000 ppm, associated with mostly terrestrial-like H isotopic composition. Based on this study, early-formed crustal rocks on the Moon were interpreted to contain water that shares a common origin with water on Earth. The potential sources of this water are further narrowed down to various types of chondritic materials. The most likely candidates are CM and CI-type carbonaceous chondrites, as they provide the closest fit to the Moon's bulk H and N compositions (Alexander et al., 2012;Alexander, 2017;Hallis, 2017, and references therein). Other constraints require that a significant proportion of the CI/CM material must have been accreted to Earth prior to the Moon-forming impact (Greenwood et al., 2018).
To investigate the influence of shock-induced deformation, metamorphism and impact-melting on apatite water abundance and H isotopic composition, we conducted a study on a suite of variably shocked rock-fragments and impact-melt breccias from the Apollo 17 collection. The texture, microstructure and petrological context of apatite is a critical parameter in understanding the severity of shock deformation and were characterized by detailed scanning electron microscope (SEM) imaging and electron backscatter diffraction (EBSD) analysis of each grain prior to measuring its water abundance and D/H ratio using a NanoSIMS.

Sample selection
The criteria for sample selection included geochemical pristinity, paucity of data on volatiles in Mg-suite apatites, and a sample set covering a range in level of shock deformation. The pristinity level of the samples (Warren, 1993) was taken into account with the aim of measuring the water content incorporated into the apatite lattice during its crystallization. All the selected samples have been previously characterized as compositionally pristine (Warren, 1993), meaning their bulk compositions and, in particular, highly siderophile elements (HSE) contents are representative of individual, unmixed, endogenous igneous rocks. At the same time, however, we targeted lithologies exposed to different levels of shock metamorphism.
Seven different lunar samples (nine specific thin-sections) of Mg-suite norite, troctolite and gabbro from the Apollo 17 collection were selected for this study. We previously determined the shock stage of some of the samples based on plagioclase shock-barometry (Černok et al., 2019 and references therein) and the shock microstructures of individual apatite grains using EBSD. These samples include troctolite 76535,51 (shock stage 1, S1), anorthositic troctolite 76335,60 (S2), Civet Cat norite fragment from the 72255,100 impact-melt breccia (S3-4), and heavily shocked norites 78235,43 and 78236,22 (S5 and S6). It is worth noting that the cooling history of the accessory mineral baddeleyite within the troctolite 76535 has recently been interpreted in the context of a large impact-melt sheet (White et al., 2020), despite not showing any compositional evidence for projectile contamination (Warren, 1993). The troctolite, however, does not display any shock features formed and has thus been assigned the deformation stage S1.
By applying the same criteria as in Č ernok et al. (2019) we here assess shock-stage deformation of apatite in troctolite 73235,136, as well as in three sections containing different lithologies inside an impact-melt breccia 76255 (norite 76255,68; gabbro 76255,71 and troctolite 76255,71). EBSD data acquired from these apatites can be found in Supplementary Material (SM, Supplementary Fig. 1). A list of studied samples, their cosmic ray exposure (CRE) ages, and all analyzed apatite grains with their interpreted shock-deformation stages is provided in Table 1. Thus, the selection of samples covers all representative lunar Mg-suite lithologies and shock histories, including coherent rock specimens, lithic fragments included within polymict impact breccia, and impactmelt matrix. Detailed sample descriptions, including associated

Analytical techniques
Polished thin-sections of Apollo samples were used in this study. All sections were prepared at NASA Johnson Space Center using a water-free medium, and mounted onto a glass slide using araldite epoxy (see  for further details).

Scanning Electron Microscopy and Electron Backscatter Diffraction (EBSD)
The studied polished thin-sections were carbon-coated and examined using a Quanta 3D Focused Ion Beam Scanning Electron Microscope (FIB-SEM) at The Open University, equipped with an Oxford Instruments INCA energy dispersive X-ray detector. The electron beam was generated with an acceleration voltage of 15-20 kV and a current of 0.52 to 0.62 nA. A working distance of 15 mm was used for the generation of secondary electron (SE) and backscatter electron (BSE) images, as well for the acquisition of energy dispersive spectra (EDS). Micro-to nano-scale structural analysis was conducted by EBSD using an Oxford Instruments Nordlys EBSD detector mounted on a Zeiss EVO MA10 LaB 6 -SEM housed at the University of Portsmouth (see SM and Supplementary Table 1), following earlier protocols (Darling et al., 2016;Č ernok et al., 2019).

H measurements using NanoSIMS
Following EBSD analyses, apatites were selected based on their textural context and microstructural features for in-situ measurements of H content (expressed as equivalent H 2 O) and associated H isotopic composition (expressed using δD notation). We used a CAMECA NanoSIMS 50L at The Open University, following the protocol reported by Barnes et al. (2013) and applied in several subsequent studies (Barnes et al., 2014;Barrett et al., 2016;. Thin-sections of the samples were covered by a 20-30 nm thick layer of carbon coat, kept for approximately one week in a vacuum oven at 40 • C, and then placed into the instrument airlock under the vacuum of ∼8 × 10 −8 Torr or better for 2-3 days before the start of each analytical session. An indium block containing the apatite standards Ap003, Ap004 and Ap018 (McCubbin et al., 2012) and the nominally anhydrous San Carlos (SC) olivine, were simultaneously carbon coated with the samples and kept in the vacuum along with the samples. The apatite standards and the SC olivine are used for calibration and background corrections, respectively. Details on instrumental parameters, data reduction, analytical background and spallation corrections are provided in the SM, Supplementary

Electron Probe Microanalysis
Electron probe microanalysis (EPMA) was performed on all apatite grains after analysis by NanoSIMS, to avoid any volatile loss by exposure to intense electron beam of the EPMA. Analyses were undertaken using the CAMECA SX100 instrument at The Open University, with protocols that were slightly modified after previous studies in the same laboratory (e.g., Barnes et al., 2013;. Analytical conditions are documented in SM and in Supplementary Table 4.

Textural context and shock-induced microtexture of apatite
In the studied selection of coherent rock fragments and impactmelt breccias, we observed two distinct types of apatite based on their nanostructure. The primary apatites show variable degrees of internal nanostructural complexities induced by shock-induced deformation, ranging from stage S1 to S6 (Fig. 1). They were found in all coherent rocks and within lithic clasts inside the breccia, in contact with another primary, magmatic mineral assemblage (Fig. 1). The secondary apatites show no nanostructural complexities and were exclusively associated with impact melt in the matrix of all three 76255 breccia samples (Fig. 2).
Primary apatite is less abundant (a few grains per thin section) and is comparably small, with the largest grains reaching up to ∼50 to 100 μm in their longest dimension. The microtexture and nanostructure of apatite from samples 76535, 76335, 72255, 78235 and 78236 were reported in a previous study (Černok et al., 2019). Following the criteria used in Č ernok et al. (2019), we assess the extent of shock-deformation stages in all other primary apatite grains in samples 76255 and 73235 based upon nanostructural data presented in Supplementary Fig. 1. In brief, the criteria used include the extent of subgrain size (defined by shape or low angle grain boundaries of <20 • ), their density and total misorientation across the grain surface, which is a useful measure of crystal-plastic deformation (CPD). The observed features were derived from band contrast (BC) images, texture component (TC) maps and associated pole figures ( Supplementary Fig. 1). We previously established that the internal nanostructure of primary apatite shows progressively higher levels of deformation with increasing shock level, ranging from no discernible nanostructural complexities at S1, to the maximum observed CPD at S5 as recorded by ∼25 • of total misorientation across apatite grains (Černok et al., 2019). Apatite within lithic fragments in all 76255 sections typically contain large subgrains and low degrees of CPD, which is comparable to apatite from S2 stage of deformation (Černok et al., 2019). More complex subgrain formation, along with 15-20 • of total misorientation, was observed in apatite from troctolite 73235,136 and one apatite grain in 76255,71 (Ap10, Fig. 1 and Supplementary Fig. 1) that occurs at the contact of a large pyroxene clast and the fine-grained impact melt. These features are comparable to apatite from S3-S4 stages of shock deformation (Černok et al., 2019).
All secondary apatites were found in sample 76255, in impactmelt portions of all three thin sections (68, 71 and 75). This textural type is very abundant (several dozen per thin section) and forms, on average, larger grains that reach up to ∼200 μm in the longest dimension. Most secondary apatites appear to be individual, anhedral (A) mineral grains within an impact melt with scalloped grain outlines ( Fig. 2A and B), but a few euhedral (E) crystals were also observed (Fig. 2D). All secondary apatites, regardless of their shape, show strong diffraction patterns with unique orientation throughout the interior that are easily indexed with a good fit to the predefined apatite structure. They yield band contrast images (Fig. 2B) and inverse pole figures representative of single, undeformed crystals (Fig. 2C). They demonstrate a lack of sub-grain formation or any other evidence of internal structural complexities (Fig. 2B, C). The anhedral grains, with scalloped outlines, are possibly xenocrysts that recrystallized in contact with the impact melt ( Fig. 2A, B). However, considering the sub-euhedral to anhedral shape of fine-grained plagioclase and pyroxene from the impact melt in contact with this apatite ( Fig. 2A, B, D), we cannot exclude a possibility that these anhedral grains crystallized directly from the melt. The euhedral apatite unambiguously grew directly from the impact melt (Fig. 2D). In summary, NanoSIMS analyses were performed on twenty-eight individual apatite grains previously characterized by EBSD. Eighteen apatite grains are of primary origin (S1-S6) while another ten are secondary in origin, of either anhedral (A) or euhedral (E) shape.

Composition of Mg-suite apatite
The measured composition of Mg-suite apatite studied here (Fig. 3) is comparable to that of previously reported data from highlands apatites (e.g. Barnes et al., 2014;. All grains are fluorapatite (F > Cl > OH), most of which have Cl-content varying from 0.5 to 1.2 wt%. Apatites in troctolites 76535 and 76335, as well as in the norites 78236 and 72255 show the highest Cl abundance (>0.8 wt%). All studied grains within breccia 76255 contain ∼0.6 wt% Cl, except for the Ap3 in 76255,75 which contains only 0.2 wt% Cl. Similarly, the sole apatite grain in 73235 troctolite contains 0.2 wt% Cl. Due to unreliable F-measurement using EPMA protocol without TDI-correction, the F-content (Fig. 3) was calculated by subtracting OH content, determined by NanoSIMS, and Cl content, determined by EPMA (see Supplementary Table 5 for exact composition and stoichiometry). The obtained difference between the estimated and measured F content do not show significant discrepancies for most of the studied grains (Supplementary Table 5). The most common impu- rities in all studied apatites are Fe, Mg, and Si. The Cl-rich apatites (troctolites 76535 and 76335; norites 78235 and 78236) contain a negligible amount of Ce, Nd, and Y compared with the Cl-poor apatite in impact melt breccia 76255, Civet Cat norite 72255, and troctolite 73235. No obvious difference in major element composition is observed between primary and secondary apatite in the breccia 76255.

H 2 O abundance and δD composition
Apatite in shocked Apollo 17 rocks shows a wide range in water concentrations and in δD values (Table 1 and Fig. 4). As a number of apatite grains contain low water abundance (i.e., <100 ppm H 2 O), consideration must be given to spallogenic D and H that significantly contribute towards the measured H abundance, and more drastically, to D/H isotopic ratios (e.g., Füri et al., 2017;Saal et al., 2013). We report spallation corrected data using both available D-production rates (  (76255, 68; 76255,71 and 76255,75) contain between 36 ppm and 964 ppm H 2 O, with δD values varying from −475 ± 144 to −59 ± 175 . Weighted average δD value of 20 apatites from all three sections is −218 ± 45 , with that of 10 secondary apatites being indistinguishable from it (−217 ± 44 ). Across the individual sections, grains in the troctolitebearing (76255,75) clast have the highest water content of all analyzed apatites (533-964 ppm) but show similar δD values (−177 ± 45 ). The section with noritic clasts in impact melt (76255,68) contain less than ∼270 ppm H 2 O, with an exception of an individual spot (Ap10) that has ∼500 ppm water, and a weighted δD value of −248 ± 84 . Apatite in the section containing the gabbro clast (76255,71) has among the lowest water contents, with less than 90 ppm H 2 O, apart from one individual spot (Ap10, 679 ppm) and an associated δD value −230 ± 77 that is comparable to other samples. In summary, in the breccia 76255 no obvious difference in water content or the associated δD is observed between clast-hosted (primary) and melt-hosted (secondary, i.e., either recrystallized or grown from impact melt) apatites.  (Robinson et al., 2016;Barnes et al., 2014), we exclude this sample from any further discussions on interpreting the volatiles data on lunar crustal samples and consider it as an 'anomalous' sample in this context (see further discussion in SM). However, a good internal consistency of D/H measured in control sample 78235 (Barnes et al., 2014) provides confidence in the reproducibility of our results.

H 2 O and δD of Mg-suite apatite in the context of its nanostructure and shock-deformation
The measured water contents are well within the range of values previously reported for apatite in different highlands lithologies (Fig. 5), including evolved rocks such as felsite, quartzmonzodiorites, and granite clasts in breccias, but also the Mgand alkali-suite samples (Greenwood et al., 2011;Barnes et al., 2014;Robinson et al., 2016). Similar abundances were measured in apatite from KREEP and high-Al basalts (Greenwood et al., 2011;Tartèse et al., 2014a,b). The water-rich apatites measured here contain water in abundances also comparable to that of the lower-end values reported for apatite in low-Ti and high-Ti basalts ( Fig. 5; Barnes et al., 2013;Tartèse et al., , 2014aPernet-Fisher et al., 2014;Treiman et al., 2016). Our measurements do not show any particular lithology-specific variabilities in water abundances, as recorded in apatite, but do show that not all troctolite-hosted apatite grains are water-poor, as previously inferred (e.g., Barnes et al., 2014). δD values of the Mg-suite apatite, ranging from −535 ± 56 to +147 ± 224 (2σ ), reveal a pattern in which mostly the water-poor apatite (<100 ppm H 2 O) are depleted in D relative to the water-rich apatite. This range of δD values is within that mea-   of all analyzed spots. B) Different shock stages of primary (S1-S6), secondary anhedral (A) and secondary euhedral (E) apatite, using correction after Merlivat et al. (1976). The overall H 2 O-δD trend observed in this study does not seem to correlate with the level of shock deformation experienced by individual apatite grains (Fig. 4b). For instance, apatites in the two most heavily-deformed norites, 72255 Civet Cat and 78236, plot at the opposite ends of the observed H 2 O-δD range. Results from impact-melt breccia 76255, which contains primary apatite inside lithic clasts (norite, gabbro and troctolite) and secondary apatite in the surrounding fine-grained impact melt, span across the entire H 2 O-δD range seen in this study. However, apatites from each individual clast and the surrounding impact melt have distinct H 2 O-δD values. For example, secondary apatite in 76255,68 contains similar amount of water as primary apatite inside the norite clast (Table 1), but in most cases the content is different than in the troctolite and the gabbro sections. This could indicate that the water abundance in the impact melt was not well homogenized or that the variable water abundances in the crystallizing apatite are being dictated by complex partitioning behavior of F, Cl and OH in the melt. Lithic clasts incorporated in the impact melt are shockdeformed at pressures and temperatures (S2) insufficient for their incipient melting, so the impact melt was not derived from those exact clasts. Instead, they were incorporated as coherent, unmelted clasts inside an already formed impact melt. However, the source rock must have had similar composition and contained apatites of comparable volatile signature. The source rock apatites that recrystallized in contact with impact melt, possibly acted as nuclei for further growth of secondary, anhedral apatite. Source apatite that melted in contact with impact melt was the main phosphorous and volatile source to form euhedral secondary apatite within the impact melt. Indistinguishable mineral chemistry of primary and secondary apatite associated with indistinguishable δD values, despite their obvious textural and microstructural differences, further strengthens the hypothesis of similarity with the source rock. These results strongly suggest that no discernible water loss and no preferential loss of H over D is recorded in apatite as a result of the impact.
This result is surprising, taking into account very high Hdiffusivity in apatite, which is expected to be even higher at temperatures relevant to impact melt (e.g. Higashi et al., 2017). Nevertheless, shock-experimental study on plagioclase (Minitti et al., 2008a) and in situ studies of Apollo highlands samples (Hui et al., 2017) have reported that this nominally anhydrous mineral behaves in a similar way to apatite, as observed in this study, despite an expected increase in H-diffusion at high temperatures (Johnson and Rossman, 2013). However, other minerals show very different behavior. For example, meteorite-relevant amphiboles (Minitti et al., 2008b(Minitti et al., , 2008a experience slight water-enrichment, followed by an increase in δD when studied in the same shock-experiment as plagioclase. On the other hand, D-heavy signature of martian atmosphere is seen in numerous shock-affected minerals found in martian meteorites, implying that shock-induced textures can enhance mixing of different reservoirs. Furthermore, individual grains of martian phosphates have been observed to undergo devolatalization during an impact event (Howarth et al., 2015) or due to shock-induced phase transitions (Adcock et al., 2017). In contrast, ringwoodite, formed as a high-pressure polymorph of water-poor olivine in martian meteorite Tissint , was observed to incorporate more water, aided by the shock-induced phase transition. This resulted in a considerably higher δD signature of ringwoodite, arising from incorporation of D-heavy martian atmosphere into the mineral.
In summary, the current understanding of shock-induced changes in volatile signatures is seemingly ambiguous, given that both water-bearing and nominally anhydrous minerals have been observed to be affected but also entirely unaltered by shockdeformation processes. Based on this comprehensive study, reporting the largest dataset currently available, of water abundance and H isotopic composition from systematically analyzed shocked apatite in lunar samples, we conclude that relatively water-rich apatite in Mg-suite rocks are not affected by shock processes in terms of their water inventory and δD signatures -a finding that is likely to be of relevance to other shock-related apatite occurrences, particularly originating from airless bodies (e.g., 4Vesta).
Grains Ap3 and Ap4 in 76255 show a large intragranular inhomogeneity, with both grains showing spatially resolved, distinct δD values. Due to this fact, we exclude the possibility that our samples record a primitive low-δD reservoir, as inferred by previous studies (Robinson et al., 2016), and conclude that the source of low-δD must be of secondary origin. Direct implantation of solar hydrogen into apatite, found in these samples, is unlikely because solar wind can only penetrate <1 μm under the surface of directly exposed grains (e.g., Robinson et al., 2016). The main low-δD (i.e., H-rich) reservoir on the Moon is the surface regolith (e.g., Liu et al., 2012;Treiman et al., 2016), containing fine grains with long exposure to surface implantation of solar-wind hydrogen. Treiman et al. (2016) suggested that the D-poor signature of apatite they measured in basalts, reflects mixing with a regolith component. The physical mechanism for introduction of hydrogen into apatite is likely to be thermal degassing of the D-poor regolith, induced by the heat of a nearby impact or, in case of mare basalts, the heat of basalts themselves (Treiman et al., 2016 and references therein). Similar regolith-interaction processes were recognized in high-Al basalt 14053 (Greenwood et al., 2011;Taylor et al., 2004) and olivine-hosted melt inclusions in basalts (Singer et al., 2017). Traces of trapped solar noble gasses within the sample 76255 were reported early on (Bogard et al., 1975, and references therein), however, they seem to be absent in 72255 breccia (Leich et al., 1975). In addition, impact-melt breccia 15405 was reported to contain apatite with variable δD composition (Barnes et al., 2014). Robinson et al. (2016) excluded regolith as a source of D-poor water following a different line of arguments; they calculated that even the maximum amount of regolith which could have been assimilated by the impact melt, produced by the formation of Aristillus crater, the possible source of the Apollo 15 samples they studied, could not have supplied enough hydrogen to alter the D/H ratio of the most water-poor apatite. Their reasoning is further strengthened by the fact that all apatite analyzed in those samples had similarly low-δD values. However, in our study we observe that low-δD apatite is heterogeneous and is either directly affected by impact melt (secondary apatite) or is located inside a small clast (primary apatite) surrounded by impact melt, implying that they most certainly experienced thermal processing. Furthermore, all primary low-δD apatites also show very complex nanostructures with sub-grain formation as a result of shock-deformation. Due to the textural context and the nanostructure of the low-δD apatite observed in our study, we conclude that networks of microto nanoscale grain boundaries could have acted as open pathways to enhance diffusion of hydrogen from D-poor regolith into the apatite grain, lowering its overall δD value, and resulting in its heterogeneous distribution. The regolith could have been either incorporated directly into the impact melt, or experienced thermal degassing caused by the heat of the impact melt, or both. Unfortunately, we cannot estimate the amount of regolith that was assimilated in the impact melts, produced during large basinforming impacts (e.g. Imbrium or Serenitatis), which are thought to be the likely sources of the breccias 72255 and 76255 (Thiessen et al., 2017). Although it is expected that the large basin-forming impacts would vaporize much of the regolith (e.g., Cintala, 1992), these breccias could have been formed as ejecta material further away from the impact crater where they incorporated or affected the local regolith. No further thermal processing of these samples had taken place after the impact event by which the breccias were formed. Furthermore, this study also reveals that the extent to which δD values can be affected during an impact largely depends on the initial water abundance of the host apatite. Regardless of the level of shock, only the water-poor apatite, exposed to shock deformation, appears to show extremely low-δD values. The apatites with extremely low-δD values are found within the same samples that also contain apatite with higher-δD values, with the overall microtextural setting strongly suggesting that they are of coeval origin, and thus unlikely to have sampled two separate δD reservoirs. On the other hand, apatite that contains more than ∼100 ppm H 2 O is unlikely to change its isotopic composition significantly, thus more likely to have preserved its primordial H isotopic composition.

Implications for source and timing of delivery of lunar water
Numerous petrogenetic models have been proposed to account for the contrasting primitive and evolved characteristics of the Mgsuite lithologies (summarized in Shearer et al., 2015). In brief, most models hypothesize partial melts of early LMO cumulates intruding into the overlying anorthositic crust, with a KREEP component either present in their source regions at the time of melting or being assimilated during their ascent. On the other hand, a few models have suggested involvement of large-scale (i.e., basin forming) impact events in producing at least a portion of the Mg-suite rocks (Shearer et al., 2015 and references therein), most recently of the troctolite 76535 (White et al., 2020). New research would be needed to understand the effect of the large-scale impact melting on volatile degassing of a large melt sheet and, consequently, on indigenous lunar volatile signature in cumulates crystalized within it. However, the impact-related models infer that the cumulates crystallized in insulated, slow-cooling large melt sheets or pooled beneath the anorthositic crust (summarized in Shearer et al., 2015), thus experiencing minimal degassing. Regardless of the exact mode of formation of Mg-suite rocks, their 'deep-seated' origin is universally accepted, suggesting that apatite crystallized in most of Mg-suite rocks would have experienced comparatively minimal, if any, H-isotope fractionation through degassing compared to those formed in mare basalts (e.g., Tartèse et al., 2014a,b). Thus most Mg-suite rocks are believed to directly record the primordial H isotopic composition of the lunar water (Barnes et al., , 2014. The lunar mantle derivatives seem to record variable δD values. The weighted average δD value of apatite in the Mg-suite obtained in this study (−209 ± 47 ) coincides with the weighted aver-age δD value estimated for lunar KREEPy samples (McCubbin and Barnes, 2019). In particular, the obtained δD value is similar to that of apatite in other plutonic rocks (Barnes et al., 2014), but also to that of basaltic lithologies in which apatite crystallized before any substantial degassing, such as in KREEP basalts (Tartèse et al., 2014a,b). Furthermore, it closely matches the values recently obtained for olivine-hosted melt inclusions in mare basalts (Singer et al., 2017). In contrast, anorthositic rocks (Hui et al., 2017;Greenwood et al., 2011) and olivine-hosted melt inclusions in volcanic glasses (Saal et al., 2013) record δD values of up to +350 . Hui et al. (2017) attempted to explain the difference in the observed δD values between anorthosites and Mg-suite rocks by using a mixing model. However, the mixing model failed to reconcile the two different values and instead it was suggested that the two distinct sources could have originated from a layered LMO, in which surface anorthosites were affected by LMO degassing and, thus, recorded increased δD-values. The cumulates in the lower portion of the LMO, such as Mg-suite rocks, sample primordial δD that was largely unaffected by magmatic degassing. The δD value obtained in the current study (−209 ± 47 ), indistinguishable from other KREEPy lunar rocks (McCubbin and Barnes, 2019), suggests that early lunar mantle, or deeper layers of LMO, have δD strikingly similar to terrestrial deep mantle (Hallis et al., 2015). This further strengthens the existing hypothesis of a common origin of water on Earth and Moon.
Based on recent FTIR measurements in ferroan anorthosite plagioclase (Hui et al., 2013(Hui et al., , 2017 and some other studies (Chen et al., 2015; it has been estimated that the LMO contained ∼136 ppm of water. Models suggest that this amount of water was likely to be too high to survive the Moonforming impact (Elkins-Tanton and Grove, 2011). Instead, most of the water in lunar interior was suggested to have accumulated during the protracted evolution of the LMO sometime between 10 and 200 Ma after the formation of the Moon (Tartèse and Anand, 2013 and references therein), with only up to 25% of terrestrial water contributing towards the entire lunar budget . In a recent study, Greenwood et al. (2018) suggested that most of the terrestrial water was formed during planetary accretion, before the Moon-forming impact occurred. They further set the constraints on composition of the Late Veneer based on the oxygen isotopes, that most likely CM and water-poor enstatite chondrites delivered water to the Earth. On the other hand, the composition of the Late Veneer determined by the water abundance in lunar samples concluded that other types of carbonaceous chondrites in addition to CM -CO, CI, and possibly CV -acted as a source of water . The similarities in the mantle values of terrestrial and lunar water, followed by the assumption that most of terrestrial water was delivered before Moon formed, dictate that the chondrites accreted on Earth prior to the Giant Impact must have been of comparable composition to those accreted on both planets during the Late Veneer, and, thus, during the time when most of Mg-suite apatite crystallized (10-200 Ma after the Giant Impact). Furthermore, the current study reveals that the H isotopic composition of apatites within the early lunar crustal rocks (≥4.2 Ga; 76335; 78235, 78236) and impact-melt breccias formed at ∼3.9 Ga (72255, 73235 and 76255) are similar. Based on this fact, it can be hypothesized that throughout the early bombardment history of the Moon, impactors delivered water of the composition similar to indigenous lunar and terrestrial water (i.e., impactors did not vary in composition significantly, at least in their H isotopic composition; e.g., CM, CO, CI, and possibly CV) or that they had low water abundances (e.g., enstatite chondrites) and hence no capacity to affect the H isotopic composition of pre-existing water in the primary lunar rocks.

Conclusions
This study reports on the chemistry, water abundance and H isotopic composition of Mg-suite apatite from troctolite, gabbro and norite lithologies that show variable degrees of shockdeformation. Petrographic context, microtextural and nanostructural inspection reveal two distinct types of apatite: primary (with shock stages S1-S6) and secondary (recrystallized or grown from impact melt). No obvious difference in mineral composition, H 2 O abundance and δD was observed between the two textural types. Mg-suite apatite studied here contains up to ∼960 ppm of H 2 O and shows δD values (−535 ± 134 to +147 ± 194 ) within the range of terrestrial and carbonaceous chondrite values, with the weighted average δD values of −209 ± 47 being indistinguishable from the Earth's deeper mantle. H 2 O-δD systematics of apatite do not correlate with the level of shock each apatite grain has experienced, but it is apparent that individual water-poor apatite grains were influenced by a process, which resulted in the lowering of their original δD values. The grains with the lowest δD values are either in contact with impact melt or have severely complex nanostructures induced by shock-deformation. Locally, on a level of an individual grain, the network of grain boundaries could enhance reaction with D-poor regolith, which is thermally processed by a hot impact-melt, resulting in D-poor signature of apatite. Water-rich primary apatites (above ∼100 ppm H 2 O) that contain complex microtextures do not show any evidence of significant water-loss or alteration in their δD value. Thus, water-rich apatites in some of the oldest lunar crustal materials act as a reliable recorder of the primordial H isotopic signature of the Moon, despite experiencing variable degrees of shock-deformation at a later stage in lunar history.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.