Water in type I chondrules of Paris CM chondrite

Abstract


Introduction
Some carbonaceous chondrites (CC) hold hydrated minerals and contain up to 10 wt.% H 2 O (Kerridge, 1985).They are the most plausible source of water delivery to the Earth (Marty, 2012;Walsh et al., 2011) as they display, on average, a similar D/H ratio as the Earth oceans, i.e. 155.76×10 -6 (Robert, 2006).Thus, the carbonaceous chondrites are the key samples to constrain the origin of water on Earth.CM carbonaceous chondrites constitute an interesting class for studying aqueous alteration.Indeed, the CM chondrite group embraces type 1 (i.e.heavily aqueously altered) and type 2 (i.e. less aqueously altered) chondrites.Manganese-chromium dating of carbonates in CM chondrites gave a reliable aqueous alteration age of 4563.4+0.4/−0.5 million years (Fujiya et al., 2012); younger age than previous estimations (de Leuw et al., 2009).This CM carbonate age suggests the late onset of aqueous activities in the Solar System and allows Fujiya et al. (2012) to estimate that CM parent body accreted roughly 3.5 Myr after the birth of the Solar System.Hydrated minerals and organic matter carry hydrogen in aqueously altered chondrites.Aqueous alteration of silicate minerals in these chondrites is evidenced by the occurrence of hydrous phyllosilicates, as serpentines, saponites, and smectites, which are the predominant phase in matrices of CCs (Garenne et al., 2014;Beck et al., 2010), often associated with carbonates and oxides (Howard et al., 2011) The petrologic chondrite classification reflects varying degrees of aqueous alteration in different groups of chondritic meteorites (Van Schmus and Wood, 1967).Classification of CMs is established based on a sequence of aqueous alteration (Alexander et al., 2013;Browning et al., 1996;Rubin et al., 2007).
Whether alteration on CM chondrite took place in the protosolar nebula or on the parent body is still a matter of debate: Ciesla et al. (2003) have argued for the hydration of silicate dust in the icy regions of the nebula during the passage of shock waves and Bishoff (1998) favors pre-accretionary alteration in the nebula, based on the relationship between fine-grained rims and their surrounding components.On the other hand, several observations support the hypothesis that the aqueous alteration occurred after the accretion onto the parent body (Brearley, 2003).For example, the existence of alteration zones, as Fe-rich aureoles around metal grains (Hanowski and Brearley, 2001), fluid inclusions (Fieni et al., 1978;Zolensky et al., 1999), supports this hypothesis.The oxygen isotopic data also indicate a parent body alteration: Young et al. (1999) explained the oxygen isotopic diversity of CC by simple down-temperature fluid flow within the parent body, based on the systematic relationships between the bulk oxygen isotopic composition of CC and the degree of aqueous alteration.
Chondrites show a systematic enrichment in deuterium relative to the molecular hydrogen of the protosolar nebula (Boato, 1954).The D/H ratio in matrix clays of chondrites has been assumed to reflect the D/H ratio of water that has circulated during the aqueous alteration of the parent body.Based on petrographic and mineralogical analyses in the heavily aqueously-altered Alan Hills 81002 CM2 chondrite, Hanowski and Brearley (2001) argued for the alteration of chondrules in a parent body environment.Nevertheless, hydrogen isotopic measurements performed on the Bishunpur and Semarkona LL3 ordinary chondrites and Renazzo CR2 carbonaceous chondrite showed large D/H heterogeneities among matrix clays and also between silicate minerals in chondrules (Deloule and Robert, 1995;Deloule et al., 1998).An isotopic reequilibration during hydrothermal alteration on the parent body by a homogenous aqueous reservoir cannot explain alone the heterogeneities of D/H ratio in these chondrites.D-enrichment of chondrite water could result from 2 scenarios (Bonal et al., 2013): (i) enrichment is an heritage of water formed in the outer solar system or presolar molecular clouds (Deloule and Robert, 1995;Deloule et al., 1998), variations observed in chondrites reflect different water reservoirs having different D/H ratios ; (ii) light water was formed in the warmer inner solar system and got enriched by secondary processes during aqueous alteration on the parent body including exchange with D-rich organics, oxidation of metal or other processes able to fractionate the initial D/H of water (Eiler and Kitchen, 2004;Alexander et al., 2012) The CM2 Paris is a fresh sample that suffered only limited parent-body transformations, as revealed by preservation of amorphous silicates in the matrix.It is considered as the least altered CM chondrite currently known, classified as a CM2.7 (Hewins et al., 2014;Marrocchi et al., 2014).This sample offers a unique opportunity to study the initial stages of aqueous alteration on the CM parent body by investigating well preserved type I chondrules, which are not prone to aqueous alteration and are the least altered component in CM chondrites.Moreover, Paris exhibits heterogeneous alteration areas (Hewins et al., 2014), which bring us a large set of chondrules ranging from low to high alteration rate.Here we investigated the water content and the D/H ratio in different phases (olivine, pyroxene and mesostasis) of 12 type I chondrules and the surrounding matrix from Paris.In comparison, 4 chondrules from Bishunpur (LL3) and 3 chondrules from Renazzo (CR2) were also analysed.Our aim was to determine if chondrules and matrix underwent the same aqueous alteration, by using the D/H ratio, or if they were subjected to different hydration episodes.The high lateral resolution of NanoSIMS allows measuring water content and hydrogen isotopic ratio in contiguous olivine, pyroxene and mesostasis of chondrules.

Objects of study
The Paris chondrite is a breccia with varying degrees of alteration and thus it contains both well preserved and altered chondrules.Three sections were analyzed in three sessions: one mounted in epoxy (i.e.Session I-chondrules 4, 6, 9, 19, 24 and 27) and two epoxy-free (i.e.Session II: chondrules 1 and 7; Session III: chondrules A, B, D and E).The latest were chosen to assess vacuum issues on the hydrogen contamination in addition to the effect of epoxy -expected to impregnate the sample at a fine scale-that could bias our measurements (Aubaud et al., 2007;Hauri et al., 2006;Stephant et al., 2014).The three sections of Paris were used for D/H analyses but only the one of session III were used for water content measurements.Renazzo and Bishunpur, mounted in epoxy, were also analysed as they exhibit heterogeneities of the D/H ratios in their chondrules, documented by previous studies (Deloule and Robert, 1995;Deloule et al., 1998) (i.e.Session IV -chondrules X, Y, Z and I, J, K, L).As such, they are used as references in this study.Renazzo is a CR2 chondrite which contains abundant type I chondrules (Weisberg et al., 1993) with intermediate level of alteration, evidenced by the replacement of chondrule mesostases by phyllosilicates.Bishunpur is a LL3.1 ordinary chondrite whose aqueous alteration was limited to the production of smectite in the matrix (Alexander et al., 1989).Type I chondrules were imaged using a JEOL JSM 840-A SEM installed at MNHN.Chemical characterization of olivine and pyroxene from 6 Paris chondrules (in the epoxy mount) were carried out using a CAMECA SX-100 EMPA at the Camparis facility, in Paris, France.
Paris contains markedly more type I than type II chondrules and their size ranges from ~100 to ~500 µ m.
In addition, type I chondrules in Paris have various petrologic characteristics.We selected 12 chondrules of different textural types and various degrees of aqueous alteration in order to survey the range of chondrules present in Paris.Backscattered electron images of these 12 chondrules are presented in Figure 1 to illustrate this diversity: barred olivines (chondrules 19, 24 and 1), radial pyroxene (chondrule 4), and porphyritic chondrules.Degree of chondrule aqueous alteration is established on the amount of metal globules, glassy stage of mesostasis and presence of alteration aureole.Moreover, various degrees of aqueous alteration are observed in chondrule mesostasis.For instance, chondrule 19 is a well-preserved object with glassy mesostasis and metal globules whereas chondrule 27 exhibits an altered mesostasis in its center and lacks of Fe-rich globules (Fig. 1).X-ray elemental maps obtained by EDX allow us to select the target areas for NanoSIMS analysis; we aimed at areas with contiguous occurrences of olivine, pyroxene and mesostasis.The EMPA measurements (Table 1) indicate that all olivines type I chondrules from Paris are magnesium-rich (Fa<2).Low-Ca pyroxenes show limited compositional range (Fs ranges from 0.7 to 17.5; Wo ranges from 0.4 to 5.9 with an exception at 38.3).We also chose 4 representative type I chondrules in Bishunpur and 3 in Renazzo.

Standards
Silicate standards were used for calibration of water contents and for instrumental mass fractionation on D/H ratio.These standards match different phases analyzed in chondrules: pyroxene KBH1, olivine KLV-23, glass DR15-1-5, glass DR20-1-1, amphibole Mont Emma and amphibole Kipawa (Clog, 2010;Deloule, 1991;Koga et al., 2003).Their chemical compositions are reported in Table 2. Samples were mounted in such a manner to reduce a contribution of epoxy to the OH signal while having a nice polished surface: holes were drilled in a 10 mm aluminium disk with a 2 mm diameter drill bit.In each hole, samples were mounted individually with epoxy and then polished down to 0.25 µm diamond paste.
Mounts were carbon-coated, thickness 20 nm.Chondrites and standards were outgassed for one week in the NanoSIMS airlock before each analytical session.

Quantification of water content
Elemental measurements were performed with the Cameca NanoSIMS 50 at MNHN, Paris. 16OH -, 28 Si -, 24 Mg 16 O -and 27 Al 16 O -secondary ions were imaged and quantified using a Cs + primary beam.Special care was dedicated to correct the contribution of water contamination: we analyzed consecutively the same areas with two different primary beam intensities, 1.2pA and then 15pA, in order to estimate the part of OH contamination (Stephant et al., 2014).Indeed, using a correlation between the primary beam intensity and the OH -/Si -ratio, we calculated the OH -/Si -at the "infinite" primary beam intensity, supposed free of any OH contamination (see supplementary information for raw data).The primary beam was rastered over a 20×20µm² surface area divided in 256×256 pixels with a 200µm aperture diaphragm for 60 cycles.The counting time was set to 1ms/px.Prior to the analysis, a 25×25µm² surface area was presputtered for 20 minutes using a high primary beam current (i.e.280pA and 230pA for the session III and IV, respectively) and a 750µm aperture diaphragm.The mass resolution power was ~6100 for 16 OH -and ~7500 for 28 Si - using the entrance slit ES3 and the aperture slit AS2.39 areas were imaged in type I chondrules from Paris during session III, 14 from Bishunpur and 11 from Renazzo during session IV.Vacuum in the analytical chamber was 9.3×10 -10 Torr for session III and 1.1×10 -9 Torr for session IV.
Water concentrations in the different phases of chondrules were obtained using a calibration of OH -/Si - versus water concentration; water concentration is reported as [H 2 O] and represents H 2 O ppm /SiO 2 wt.%.
The calibration lines were forced to the origin: using our method to estimate the OH -/Si -ratio at infinite primary beam intensity, the surface contamination contribution on the OH -/Si -value is withdrawn (see the discussion of this calibration in Stephant et al., 2014).Different calibrations were used depending on the silicate mineralogy (Tenner et al., 2009) (Fig. 2): matrix effects are so strong that ferromagnesian minerals and amphiboles show similar OH -/Si -ratios for very distinct water concentrations (less than 200 ppm up for the former to compare with 1 wt.% for the latter).Thus it is crucial to precisely identify each analyzed mineral to accurately quantify its water content.To this end, 24 Mg 16 O -/ 28 Si -and 27 Al 16 O -/ 28 Si -are used to decipher between olivine, pyroxene and mesostasis (see supplementary information).Standard measurements for each session of analysis (i.e.III and IV) are reported in supplementary information.
Uncertainties for each calibration line were evaluated using a code written on the R program (Thomen et al., 2014).The reported errors on meteorite measurements are dominated by uncertainties on the calibration slopes.The counting statistics on the OH -/Si -ratio are negligible compared to the calibration slope uncertainties.

Hydrogen isotopic ratios
Hydrogen isotopic measurements were performed by imaging H -and D -over the same areas studied for elemental analyses to provide both isotopic and elemental data at exactly the same location.Thereby, D/H images could be superimposed on elemental images.The primary beam was rastered over a 20×20µm² surface area divided in 256×256 pixels with a 300µm aperture diaphragm for 60 cycles with a counting time set to 1ms/px.The primary beam intensities was ~200-240pA for the three sessions on Paris chondrite and 29pA for the session IV (i.e.Bishunpur and Renazzo).Prior to each analysis, a 25×25µm² surface area was presputtered for 5 minutes using a high primary beam current (i.e.230-280pA) with the 750µm aperture diaphragm.The entrance and the aperture slits were both set at the position 1.The instrumental isotopic fractionation factor for hydrogen was determined with the set of silicate standards previously described at each session (supplementary information).During sessions I, II and III, no matrix effect was noticed for the D/H ratio using NanoSIMS at the precision of our measurements (Hu et al., 2015).However, in the session IV, a significant difference in the instrumental isotopic fractionation was observed for the amphibole and the glass standards.That can be due to the lower primary beam intensity during this session compared to the three prior sessions.The instrumental isotopic fractionation factor was also calculated as the average of these standards.The uncertainty on the instrumental isotopic fractionation corresponds to the standard deviation of terrestrial standards (see supplementary information).Errors at the 2σ level (quadratic sum) combine the uncertainty on the instrumental mass fractionation factor and the counting statistic error on each measurement (dominated by the total number of counts detected for D -).The D/H ratios are expressed in δ units (where δD=[(D/H sample )/(D/H SMOW )-1] ×1000).
In olivine, water concentration varies markedly between Renazzo, Bishunpur and Paris: 156±44 ppm (1SD, n=7), 222±123 ppm (1SD, n=7) and 860±132 ppm (1SD, n=14), respectively.Relative uncertainties for these measurements are ±25% for Paris NAMs and ±50% for Bishunpur and Renazzo NAMs (supplementary information).It must be noted that significant uncertainties in the determinations of the water content in NAMs are due to the fact that our standards -being terrestrial nominally anhydrous minerals -have much lower water concentrations than those measured in Paris chondrules.This implies a large extrapolation towards large values hence larger uncertainties.Using the uncertainties on the NAMs water calibration line, the lowest estimations for pyroxene mean water concentrations become 447±318 ppm, 439±268 ppm and 557±147 ppm for Renazzo, Bishunpur and Paris, respectively.Regarding the olivines, the lowest estimations for mean water concentrations are 78±22 ppm, 111±61 ppm and 645±99 ppm, for Renazzo, Bishunpur and Paris, respectively.Therefore, Paris olivines appear to be more hydrated than those in Bishunpur and Renazzo chondrules and do contain a significant amount of water for a nominally anhydrous mineral.For pyroxenes, Bishunpur and Renazzo exhibit a more scattered distribution than Paris.The small numbers of chondrule analyzed for Bishunpur (i.e. 4) and Renazzo (i.e. 3) compared to Paris doesn't allow us to make a robust statement on the water content of these two meteorites.Moreover, water concentrations in Paris chondrules were determined in epoxy-free sections (session III), allowing us to rule out a significant contribution from the epoxy in the OH -signal and a subsequent bias in the determination of the water content (Aubaud et al., 2007).3) are shown in Figure 4.The mean δD value for these three meteorites fall within the terrestrial domain (Lécuyer et al., 1998).Paris exhibits a wide spread of δD from -398±23‰ to 366±35‰ (i.e.covering a range of 764‰) with a mean value of -88±141‰ (2SD, n=70).In Paris, the δD of the surrounding matrix is -69±32‰.The possible contamination of the Paris samples by terrestrial moisture can be ruled out since chondrules show both enrichment and depletion of deuterium relative to the terrestrial value.The Paris mounts without epoxy (i.e.session II and III; chondrules A, B, D, E, 1 and 7) shows the largest range of δD values.In Renazzo, the δD in olivines, pyroxenes and mesostases range from -212±125 to 15±156 ‰ with a mean value of -131±115‰ (2SD, n=21).An image recorded in the matrix gives a value of 316±202‰.These determinations are consistent with previous studies in Renazzo (Deloule and Robert, 1995) showing that the matrix is enriched in deuterium (-37 to 733‰) relative to chondrules (-294 to 284‰).Bishunpur minerals range from -166±133‰ to 137±176‰ with a mean value of -66±148‰ (2SD, n=16).This range is significantly smaller compared to previous studies on Bishunpur chondrules: -371‰ < δD <1934‰ (Deloule and Robert, 1995;Deloule et al., 1998).We will focus our attention in the upcoming discussion on the emblematic case of Paris for which the δD variations reach 764‰.

Intra-chondrule variations in Paris
In addition to the broad range of δD observed among chondrules, intra-chondrule heterogeneities are also pronounced in Paris.Indeed, an isotopic variation of 658‰ is observed between the different phases of the chondrule E and up to 664‰ for the chondrule A. More surprisingly, isotopic variation among olivines in chondrule A reaches 454‰ in less than ~200 µm.In Figure 5, NanoSIMS images of AlO -/Si -and D/H of a 20×20µm² area in the chondrule A are shown.Olivine and mesostasis that are contiguous exhibit a difference in δD of 317‰ with no visible gradient (the NanoSIMS resolution does not exceed ~500nm).In most cases, the mesostasis is enriched in deuterium compared to olivine and pyroxene (chondrules A, B, D, E, A, 7, 9) with the exception of chondrules 24, 6 and 27.As such, no systematic between mesostases δD and silicates δD is observed, suggesting there is no equilibrium between mesostasis and silicates minerals.Futhermore, no correlation is observed between the degree of aqueous alteration experienced by the chondrule, as revealed by its petrology, and the δD distribution of the chondrule.In the less altered chondrule (19), there is no variation, within errors, of the δD value between the mesostasis and olivine and pyroxene (i.e.-34±79‰ on average, 2SD).However, chondrules A, E and 7 exhibit a low degree of aqueous alteration illustrated by abundant metal grains and they show a large isotopic heterogeneity between their minerals.In addition, chondrule 27 is heavily altered (its mesostasis is altered into phyllosilicates) while δD variations do not exceed 232‰ between mesostasis and minerals.An important point to notice is that chondrules A, E and 1, showing the largest range of δD values, were all measured in epoxy-free mounts.As a conclusion of this isotopic survey, δD variations are intrinsic to each chondrule but not related to their degree of aqueous alteration.

δD fractionation among coexisting minerals
The isotopic fractionation coefficient α between two phases corresponds to the ratio of their D/H values.
Figure 6 reports the isotopic fractionation coefficients α between olivine and mesostasis versus those between pyroxene and mesostasis, reported as 10 3 ×lnα in per mil.Fractionation coefficients are calculated for adjacent phases (see supplementary information).Therefore, each point corresponds to a NanoSIMS image containing the 3 phases in a 20×20µm² surface area.A trend is observed with both positive and negative values.The slope of the correlation is not far from unity, indicating that pyroxene and olivine have similar δD.The important point to notice is the large range of fractionation coefficient values (i.e. 10 3 ×lnα ranging for -431‰ to +283‰).This correlation requires a process that would generate a large fractionation factors between minerals during chondrule history.Furthermore, this correlation has no link to the degree of aqueous alteration and is not simply related to hydrothermal alteration, which may imply that chondrules followed different fractionation history.

Low-temperature and hydrothermal alteration on the parent body under isotopic equilibrium conditions
The observed fractionation factors are too large to account for an isotopic fractionation at equilibrium during H isotopic exchange between H-bearing minerals and water.Significant intra-chondrules heterogeneities among the same phases are up to 764 ‰.In terrestrial environments, such variations are never observed among alteration products and fractionation factors between minerals and water are restricted to 60‰ (Saccocia et al., 2009).This should be regarded as clear evidence against a simple equilibrium isotope fractionation between phases of the same chondrule.

Modelling the chondrule alteration by an isotopically homogeneous fluid.
NAMs (i.e.olivine and pyroxenes) are either enriched or depleted in D relative to the mesostasis (Fig. 6).
This suggests there is no equilibrium between mesostasis and these minerals.This results in a crossover rarely observed in natural environment, which should be regarded as a dependable constraint for chondrules formation models.We first discuss common geochemical situations that cannot yield such a crossover.
The progressive isotopic re-equilibration of the minerals with the mesostasis by diffusion of an external fluid to the chondrule cannot account for the isotopic crossover observed in the Figure 6 since isotopic fractionation will be unidirectional (Kyser and O'Neil, 1984).In other terms, the expected correlation for diffusion in the Figure 6 is either enriched or depleted relative to 0 but never on both sides.However, we considered here diffusion by a reservoir of water with homogeneous isotope composition.Bonal et al. (2013) have shown that the δD of the matrix water is highly heterogeneous at the micrometer scale in CR chondrite.Piani et al. (2015) have shown that water-ice precursors with a large range of δD compositions could have accreted into the parent body and been responsible for matrix δD heterogeneities at the micron scale.Secondary processes on the parent body have also been suggested to explain this spatial variability (Alexander et al., 2010;Alexander et al., 2012).This δD spatial heterogeneity at a small scale could potentially have induced the heterogeneities observed among single chondrules.However, this model holds three issues.(1) The fractionation coefficients between ferromagnesians (i.e.olivine and pyroxene) and mesostasis show significant variations (cf.Fig. 6).This means that the δD of the mesostasis can be distinct from that in the pyroxene or olivine next to this mesostasis.Even if the aqueous alteration occurs at the micrometer scale, phases having common boundaries should record a similar δD value; plus or less the isotopic fractionation at equilibrium between phases (Saccocia et al., 2009).( 2) The δD measured in the matrix covers a more restricted range (i.e.-66±70‰; 2SD) compared to the broad distribution of δD in chondrule phases; (3) This model does not account for the high water content in olivines, as it seems quite unrealistic to incorporate a minimum of 645 ppm H 2 O inside an olivine at chondritic parent body conditions of pressure and temperature.This problem will be discussed in the next section.
Hence, hydration of the chondrule by parent body water does not fit the requirements to explain the isotopic crossover observed in the Figure 6.However, as shown in the next section, it may be explained by the evaporation of H before the formation of the parent body, provided that this H was present in chondrules at the moment of their formation.

Loss of H from chondrules during their cooling in the gas phase.
Here we propose that distillation of H took place during the melting and the cooling of the chondrule in the gas phase.This outgassing model has two implications: (1) the initial [H] concentration and (2) the degree of distillation caused by H 2 escape can vary from chondrule to chondrule.As the analyzed chondrules show various textural and petrological features, there is no reason that they share common initial water content and that they underwent the same degree of distillation.
Rayleigh distillation induced by the loss of H 2 from melted chondrules can be written as: With D/H(0) the initial hydrogen isotope ratio of the mineral; D/H(t) the hydrogen isotopic ratio at instant t and ݂ ௦ is the remaining fraction of H in the mineral (i.e. from 1 to 0); ߙ ுଶିுଶை is the isotopic fractionation between H 2 and H 2 O in phases.The lowest measured value on chondrule phases i.e. for δD= -398‰ is used to set ‫ܦ‬ ‫)0(ܪ‬ ⁄ ௦ .We assume that each phase follows its own f Rayleigh coefficient i.e. that NAMs and mesostasis evolve independently since they do not share the same level of advancement of the diffusion for H species (Ingrin and Blanchard, 2006) and initial water concentration.Indeed, mesostasis can be either depleted or enriched relative to olivine and pyroxene, function of their relative degree of degassing.Calculations are reported in Figure 6 for ݂ ௦ values randomly generated for mesostasis and for NAMs.These calculations reproduce the wide range of isotopic fractionation variations with ߙ ுଶିுଶை value of ≈0.9 for olivine and pyroxene and ≈0.85 for mesostasis.These ߙ values are markedly different from unity and therefore consistent with the large H 2 -H 2 O isotopic fractionation factor (Simon et al., 2011).From these ߙ values, we calculate a reducing temperature of the system during the isotope fractionation around 900-1100 °C (Simon et al., 2011).
An alternative scenario could involve the reduction of water during parent body processes (Alexander et al., 2010).Lécluse and Robert (1994) showed that the isotope exchange between H 2 O and H 2 is still fast down to 200-300°C, i.e. at hydrothermalism temperatures.If we consider that a distillation took place around 150°C in the parent body (Clayton and Mayeda, 1999;Guo and Eiler, 2007), the fractionation factor between H 2 and H 2 O should be far (ߙ ுଶିுଶை ≈0.3, by extrapolating from Simon et al., 2011) from our calculated value (ߙ ுଶିுଶை ≈0.9).In other terms, it seems unrealistic to argue that water reduction took place on the parent body while the hydrogen isotopic system registered temperatures around 900-1100 °C.
We thus propose that chondrules have degassed their juvenile water prior to their incorporation in their parent body, presumably in the nebula, following the reduction of their water at temperature around 900-1100°C.
The lack of correlation between water content and the δD demonstrates that each chondrules contained heterogeneous initial water concentrations or δD or both.As analyzed chondrules are of different textural types, we can assume that each chondrule gets its own initial water content.

Data on Earth
It is now established that nominally anhydrous minerals are a major host for water in terrestrial mantle (Bell and Rossman, 1992;Bolfan-Casanova, 2005;Hirschmann et al., 2005;Peslier, 2010).Martin and Donnay (1972) were the first to propose a water incorporation model based on OH ions replacing O in the structure of pyroxenes.Water storage has been extensively studied in NAMs.Natural pyroxene water concentrations range from few 10s up to ~1100 ppm of H 2 O, depending on the geological occurrence (Rauch and Keppler, 2002;Skogby, 2006;Sundvall and Stalder, 2011).Natural olivine contains significantly less water than pyroxene with contents ranging roughly between few and 400 ppm of H 2 O (Bell and Rossman, 2003;Beran and Libowitzky, 2006).Measured water concentrations in NAMs of Paris chondrules (mostly olivines) are markedly higher than those in terrestrial ones.Thus water contents measured in olivine seem too high to come only from water in olivine.
However, several experiments show that olivine can incorporate significant amount of water at high pressure, high temperature (Kohlstedt et al., 1996;Mosenfelder et al., 2011;Ohtani et al., 2001).Férot and Bolfan-Casanova (2012) measured olivine water contents of 4690±1655 ppm for experiments performed at 9 GPa and 1175 °C.Smyth et al. (2006) found water solubility in olivines up to 8900 ppm at 12 GPa and 1250°C.
OH-rich inclusions in olivines and pyroxenes of magmatic origin could account for the large content we determine.For instance, Khisina et al. (2001) have shown that hydrogen incorporation in olivine exists in two modes: intrinsic mode into olivine structure (Wright, 2006) and extrinsic incorporation as nanometersized inclusions of OH bearing silicates.In pyroxene, hydroxyl can occur in the form of narrow lamellae of amphibole (Skogby et al., 1990;Veblen and Buseck, 1981).

Implications for chondrule olivines of Paris
From the experimental data on Earth's olivine, the striking water contents measured in Paris olivines could suggest that olivines were formed at high-pressure, high-temperature conditions in order to incorporate such amount of water; assuming that this water is under the form of hydroxylated groups in the structure of olivines.The model proposed by Libourel and Krot (2007) suggested that olivines are inherited from mantle planetesimals fragmented by impacts.In this scenario, mantle olivine can incorporate large amounts of water under conditions of high pressure and high temperature.As suggested by Libourel and Krot, pyroxenes could result from the destabilization of olivine caused by the incorporation of SiO diffusing from the gas phase.Qualitatively, this model accounts for the similar δD values in the pyroxene and in its parent olivine.Nevertheless, there are restrictions to this hypothesis.
First, this scenario by Libourel and Krot is based on chondrules having a granoblastic texture that is not found in all our studied chondrules.Secondly, the kind of high pressure needed to incorporate a minimum of 645 ppm of water in olivine seems difficult to reconcile with a small parent body.
The speciation of hydrogen is not known through NanoSIMS analyses and it is limited by its spatial resolution (100nm).Nanometer-sized inclusions cannot be identified by NanoSIMS imaging.We assumed that the hydrogen occurs under the form of hydroxylated groups in the chondrule olivines.However, glass inclusions in olivines can be found in any type of chondrules (Varela et al., 2002;Florentin et al., 2016).
These inclusions occur at the micrometer scale and their water contents should be more similar to the mesostasis one, in the range of few wt.%H 2 O. Indeed, such inclusions should appear on our NanoSIMS images of AlO -/Si -, MgO -/Si -and OH -/Si -, as we have a resolution down to 100 nm at 1.2 pA.We do not observe any variations in these three ratios during the duration of our analyses, meaning that we do not cross any glass inclusions of this type.
We propose that type I chondrules precursors could be composed of glassy grains, clays minerals or altered ferromagnesians, i.e. hydrous matrix-like material.These precursors would have been incompletely degassed after chondrule cooling and before incorporation into the parent body.Water should have been incorporated into chondrules before their accretion into the parent body, as it seems unrealistic to add a large amount of water in chondrule olivines at parent body conditions.

Open questions
The Rayleigh distillation model seems to indicate that chondrules would have degassed their primordial hydrogen, distinct from the alteration water, resulting from reduction of water into H 2 at ⋍ 900°C.Still, there are some interesting unsolved issues: (i) The primordial source of hydrogen in chondrules is unknown.The hydrogen isotopic ratio of this source could be as low as -398‰, i.e. the lower δD value measured in chondrule phases.This value is markedly distinct from both the nebula H 2 and the mean chondrite hydrogen isotopic signature.There are few components with such low δD among solar system objects.The unique organic matter contained in the Abee meteorite in one of them (Remusat et al. 2012;Yang and Epstein 1983).This material signs the occurrence, in the early solar system, of a H-reservoir that was at isotopic equilibrium with molecular H 2 in the neutral environment (Lécluse and Robert, 1994).This source of H could also be responsible of the end-member δD value advocated to interpret the linear variations observed at the scale of the bulk CCs in a D/H vs. C/H diagram: at C/H=0, δD= -444±23‰ (Alexander et al., 2012).According to the present study, in bulk CCs, organic free minerals (chondrules) could be the carriers of a H-reservoir with a low δD.
(ii) This model implies that NAMs had a higher water concentration before the distillation since H 2 was lost from chondrule.Therefore chondrule precursors were hydrated before their melting/crystallization into the form of chondrules and were incompletely outgassed during the rapid cooling of the chondrules.

4. 2
Hydrogen isotopic ratio in chondrules 4.2.1 Comparison between Paris, Renazzo and Bishunpur Distribution of the hydrogen isotopic composition of minerals in chondrules in Paris, Renazzo and Bishunpur (Table

Figure 1 :
Figure 1: SEM images of 12 type I chondrules in Paris from the section with epoxy -Chondrules 1, 19 and 27: barred olivine.Chondrule 19 exhibits a well-preserved glassy mesostasis.Chondrule 4: radial pyroxene.Others chondrules are porphyritics.They presented various stages of aqueous alteration.Some of them like chondrule 6 or D exhibit well-preserved Fe grains, sign of low aqueous alteration.In the other hand, chondrule 27 has its mesostasis transformed into phyllosilicates and an aqueous alteration rim grew around him.

Figure 2 :
Figure 2: OH -/Si -vs H 2 O/SiO 2 (ppm/wt.%)calibration for the quantification of water contents.Dashed lines are regression lines calculated with the R program.

Figure 3 :
Figure 3: Water content distribution in olivine and pyroxene chondrule minerals of Renazzo, Bishunpur and Paris chondrites.Y axis represents the cumulate number of analyses.

Figure 5 :
Figure 5: NanoSIMS images of an area (20×20µm²) of Paris chondrule A. (a) 27 Al 16 O -image permits to distinguish mesostasis (Mes) and olivine (Ol).(b) D -/H -ratio image.We defined ROIs for each phase (blue lines) that fit in both images.The ROI is chosen smaller than the phases due to the different image resolution and the possible shifting occurring between the elemental images and the isotopic image.

Figure 6 :
Figure 6 : Fractionation coefficients between olivine and mesostasis versus pyroxene and mesostasis reported in per mil.Black square symbols represent our analyses and gray square symbols stand for the distillation numerical simulation explained in section 4.2.3.Arrows indicate direction of fractionation for hydrous alteration of the Paris chondrules, which is unidirectional.The direction will depend on the alteration parameters.
Incorporation of high water content could have been achieved because chondrule precursors included hydroxylated minerals.It should be emphasized that the presently proposed distillation model does not preclude a late alteration (hydration) of chondrules by circulating parent body water causing a limited homogenization of the initial chondrule δD values.Indeed, -if the fluid was isotopically homogenous -the isotope fractionation between minerals and water is limited to 50 ‰ and thus should not have significantly altered the observed chondrule δD distribution that spars over 764‰.Zolensky M.E., Bodnar R.J., Gibson E.K., Nyquist L.E., Reese Y., Shih C.-Y. and Wiesmann H. (1999) Asteroidal Water Within Fluid Inclusion-Bearing Halite in an H5 Chondrite, Monahans.Science 285,

Table captions Table 1 -
EMPA analyses of olivines and pyroxenes from 6 Paris chondrules.

Table 3 -
NanoSIMS data of water concentrations and hydrogen isotopic ratios (expressed in δD) of mesostases, olivines and pyroxenes from 3 Renazzo chondrules (session IV), 4 Bishunpur chondrules (session IV) and 12 Paris chondrules (session III).For sessions II and III, the water concentrations are not reported in the table because the data were obtained at low primary beam intensity and background 636 correction could not be done.