Melt inclusions in augite from the nakhlite meteorites: A reassessment of nakhlite parental melt and implications for petrogenesis


  • Violaine SAUTTER,

    Corresponding author
    1. Laboratoire de Minéralogie et Cosmochimie du Muséum, CNRS-UMR 7202, Muséum National d’Histoire Naturelle, 57 RueCuvier, CP52, 75231 Paris Cedex 05, France
    Search for more papers by this author
  • Michael J. TOPLIS,

    1. Institut de Recherche en Astrophysique et Planétologie (CNRS-UMR 5277), 14 Ave. E. Bélin, 31400 Toulouse, France
    2. Observatoire Midi-Pyrénées, Université de Toulouse III, 31400 France
    Search for more papers by this author
  • Jean-Pierre LORAND,

    1. Laboratoire de Planétologie et de Géodynamique, UMR 6112, Faculté des Sciences et des Techniques, Université de Nantes, 2 rue de la Houssinière, BP 92208-44322 Nantes, France
    Search for more papers by this author
  • Michele MACRI

    1. Università degli studi di Roma “La Sapienza,” P.le A. Moro, 5, 00185 Rome, Italy
    Search for more papers by this author

Corresponding author. E-mail:


Abstract– The nakhlites, a subgroup of eight clinopyroxenites thought to come from a single geological unit at the Martian surface, show melt inclusions in augite and olivine. In contrast to olivine-hosted melt inclusions, augite-hosted melt inclusions are not surrounded by fractures, and are thus considered preferential candidates for reconstructing parent liquid compositions. Furthermore, two types of augite-hosted melt inclusion have been defined and characterized in four different nakhlites (Northwest Africa [NWA] 817, Nakhla, Governador Valadares, and NWA 998): Type-I isolated inclusions in augite cores that contain euhedral to subhedral augite, Ti-magnetite, and pigeonite plus silica-rich glass and a gas bubble; Type-II microinclusions that form trails crosscutting host augite crystals. Fast-heating experiments were performed on selected pristine primary augite-hosted melt inclusions from these four samples. Of these, only data from Nakhla were considered robust for reconstruction of a nakhlite parental magma composition (NPM). Based upon careful petrographic selection and consideration of iron-magnesium ratios, our data are used to propose an NPM, which is basaltic (49.1 wt% SiO2), of high Ca/Al (1.95), and K2O-poor (0.32 wt%). Thermodynamic modeling at an oxygen fugacity one log unit below the QFM buffer using the MELTS and PETROLOG programs implies that Mg-rich olivine was not a liquidus phase for this composition. Our analysis is used to suggest that olivine megacrysts found in the nakhlites are unlikely to have coprecipitated with augite, and thus may have been introduced during or subsequent to accumulation in the magma chamber, possibly from more evolved portions of the same chamber.


SNC meteorites (shergottites, nakhlites, chassignites) are commonly thought to be igneous rocks from Mars (McSween 2002), currently providing the unique opportunity to study Martian rocks with terrestrial-based analytical methods. The nakhlite subgroup is of particular interest as it is composed of a set of eight meteorites with uniformly young crystallization ages of 1.3 Ga (Nyquist et al. 2001), and cosmic-ray exposure ages of 10–11 Ma (Eugster et al. 2002). This is taken to suggest that these samples all originate from a single geological site sampled by the same impact event. These samples are all medium-grained olivine-bearing clinopyroxenites. Textural and chemical analyses of the cumulus augite cores (Lentz et al. 1999; Day et al. 2006) indicate long residence times in a shallow level magma chamber or thick lava flow (Harvey and McSween 1992a; Treiman 2005). However, significant differences also exist among the eight available samples, in particular concerning the amount of intercumulus matrix, its mineralogy, and zoning patterns of olivine (Treiman 2005). These features indicate variable amounts of high-temperature annealing. Complementary estimates of cooling rate (Mikouchi et al. 2003, 2005a) have been used to indicate different depths of sampling, either in a single differentiated magmatic pile (a subsurface sill or a thick lava flow: Mikouchi et al. 2003; Day et al. 2006) or from several flows (Lentz et al. 2005; Shirai and Ebihara 2008). If all nakhlites do come from a single lava flow, then all should be related to a single parent magma, commonly referred to as the nakhlite parent melt (NPM).

One possible way to reconstruct the composition of the NPM is to study melt inclusions trapped in early crystallizing minerals. Melt inclusions are solidified droplets of melt less than 300 μm in diameter (Schiano 2003) entrapped during mineral growth (Sorby 1858). In optimal circumstances, the host mineral is able to isolate these liquids from subsequent magmatic processes, providing a record of liquid chemistry at the time of entrapment (Frezzotti 2001). Analysis of melt inclusions is particularly useful in phenocrysts from igneous cumulate rocks such as the nakhlites, the bulk composition of which does not represent the composition of the parent magma. Nakhla itself has been the subject of several such studies, using both olivine and augite as cumulate hosts, and using a variety of fast- and slow-heating techniques. Melt chemistries determined from melt inclusions are diverse, ranging from basalt to dacite. These variations are typically interpreted as the result of fractional crystallization of a basaltic parent melt (e.g., Stockstill et al. 2005; Treiman 2005), although Varela et al. (2001) interpreted them as the result of heterogeneously trapped liquids formed through nebular nonigneous process. Evidence in favor of a magmatic origin is provided by good agreement of the observed spread in liquid composition and that predicted by thermodynamic modeling (Stockstill et al. 2005). However, the exact nature of the NPM is still an open question. In general, preferred candidates are tholeiites, poor in Al and rich in Fe relative to terrestrial counterparts, although proposed parent melts differ significantly in their total alkali content, with melt inclusions in olivine having much higher K2O and K2O/Na2O than those in augite (Goodrich et al. 2010). The value of olivine-bearing magmatic inclusions has been debated in the Earth science literature, with questions arising concerning both the trapping of nonrepresentative boundary layer melts due to rapid crystal growth (Faure and Schiano 2005) and the modification of inclusion composition due to rapid intracrystalline diffusion (e.g., Danyushevsky et al. 2000, 2002; Gaetani and Watson 2000). For these reasons, clinopyroxene (Cpx) can be considered a more appropriate crystalline host, although even within the family of Nakhla Cpx-hosted inclusions, considerable compositional variability is present, and a choice must be made concerning how to select the most appropriate parental liquid. For example, among their fast-heating experiments, Stockstill et al. (2005) retained the composition containing the lowest silica content (NA03) as their preferred candidate for the NPM.

In this work, we return to the question of melt inclusions in nakhlites to assess the most appropriate samples for study, the most appropriate inclusions in those samples for the reconstruction of the NPM, and the status of olivine in the parental liquid. In addition to Nakhla itself, we have looked at samples of Governador Valadares, Northwest Africa (NWA) 817, and NWA 998. These latter two samples have contrasting zoning patterns of olivine and very different proportions of mesostasis (e.g., Sautter et al. 2002; Treiman 2005), interpreted to indicate different cooling rates (Mikouchi and Miyamoto 2002) and possibly positions within a single thick lava flow (e.g., Mikouchi et al. 2003, 2005a). In this way, our aim is to assess if any evidence for variations in parent-liquid composition could be found.

Experimental and Analytical Techniques

Fragments of NWA 817, NWA 998, Governador Valadares, and Nakhla have been used to prepare doubly polished thin sections as well as separated single crystals of augite and olivine. The Nakhla and Governador Valadares samples belong to the meteoritical collection of the Museum National d’Histoire Naturelle of Paris. A petrographic thin section of NWA 817, previously studied by Sautter et al. (2002), was provided by the CNRS “Theodore Monod” consortium. A fragment of NWA 998 (total weight 0.2 g) was obtained by Michele Macri from Jim Strobe. The thickness of doubly polished thin sections varies from 150 to 200 μm for rock samples to 50–150 μm for single crystals (depending on the size of the grain). Doubly polished sections were prepared first using silicon carbide 320 and 600, and then diamond Dp-spray of 3 and 0.25 μm.

Melt inclusions were studied using a Supra TM 55VP Zeiss FEG-Scanning Electron Microscope (SEM) equipped with a backscattered electron (BSE) detector for imaging (Pierre and Marie Curie University, Paris VI, France). Heating experiments were performed at the Laboratory Pierre Süe, CEA-CNRS, Gif sur Yvette (France). Single augite and olivine grains were mechanically detached from crushed surfaces and separated under a binocular microscope. Doubly polished thin sections (40 and 100 μm thick) were prepared for each grain. Grains were selected containing melt inclusions suitable for the rehomogenization process i.e., devoid of visible fractures, rounded in shape, and large enough to allow microanalytical work (10–80 μm in diameter). As discussed in more detail in the following section, it was found that of our four samples, only Nakhla contained inclusions that were of use in reconstructing the NPM composition. Heating experiments were conducted in a modified Sobolev-type heating stage (Sobolev et al. 1983; Schiano 2003). Crystals were placed on a corundum grid in a ring-shaped Pt sample holder (2.5 mm in diameter and 6 mm high) hosted in a gas-tight sealed body cooled by water. The whole process was monitored by a camera connected to a petrographic microscope through two optical synthetic quartz windows allowing visual assessment of the degree of homogenization as the experiment took place. Temperature was continuously monitored by a Pt-Pt90Rh10 thermocouple in the vicinity of the sample. To avoid oxidation of Fe-bearing phases, the oxygen fugacity was buffered between 10−9 and 10−10 atm at 1150 °C (approximately 1 log unit below the Quartz-Fayalite-Magnetite buffer) by a He flux, which was dried on a zirconia bed at 700 °C before being introduced into the heating cell. Several experiments were dedicated to definition of the heating procedure, with the aim of ensuring complete homogenization of all phases in the inclusions, while avoiding fracturing and melting of the host phase. Optimal heating conditions were found to be a sequence of (1) heating from room temperature to 1000 °C at 16.6 °C min−1; (2) 4.3 °C min−1 up to 1100 °C; (3) 1.4 °C min−1 up to 1175 °C; (4) 30 min at 1175 °C to ensure homogenization and an approach to equilibrium of the melt inclusion. This heating procedure is similar to that of the slow-heating experiments of Varela et al. (2001). Samples were then rapidly quenched with a decrease from 1175 °C to room temperature in less than 10 s. Such rapid quenching is necessary to transform homogeneous melt into glass.

Backscattered electron images of melt inclusions were acquired using a JEOL 840-SEM at the University of Paris VI. Major element compositions of minerals in multiphase inclusions and homogenized vitreous melt inclusions were determined using the CAMECA SX-50 and CAMECA SX-100 Electron Micro-Probe (EMP) facilities at the University of Paris VI (CAMPARIS) operating with an accelerating voltage of 15 kV and a sample current of 20 nA. International melt inclusion and mineral standards were used for calibration. Homogenized melt inclusions were analyzed using a defocussed beam to minimize loss of volatiles such as sodium.

One unheated melt inclusion was also analyzed using laser ablation with a defocused beam in combination with plasma-source mass spectrometry (LA-ICPMS). The laser spot was 10–100 μm. The Nd:YAG laser was operated at 266 nm (fourth harmonic) in pulsed mode (10 Hz), with a pulse energy of 4 mJ, resulting in a fluence of approx. 50 J cm−2. The laser beam was focused with an f = 40 mm plano-convex lens. A pinhole was used to control the spot size, by imaging it onto the sample surface. The ablation cell was 1.4 cm3, with a 0.5 mm i.d. inlet nozzle to ensure a steady gas jet. The PVC transport tube had an i.d. of 1.4 mm and was 1.5 m long. Helium was used as the carrier gas at a flow rate of 0.9 L min−1. The ICP-MS was a quadrupole plasma–mass spectrometer (PE/Sciex Elan 6000) operated under standard conditions. Background signal was monitored at mass 220.5 (off-peak) for 20 s prior to operation of the laser on the analyzed mass (on-peak). For calibration, three standard reference materials (SRM) were used. Two of these were commercially available certified reference materials from the National Institute of Standards and Technology (NIST 610, NIST 612), while the third was an inhouse clinopyroxene that had been previously characterized by electron microprobe analysis. The quantification procedure for melt inclusions using LA-ICPMS has been discussed in detail in Bleiner et al. (2006).

Petrology of Analyzed Samples

Petrology of Studied Samples

Detailed petrographic descriptions of nakhlite meteorites can be found in Treiman (2005). The type specimen, Nakhla, is an unbrecciated, medium-grained olivine-bearing clinopyroxenite (Fig. 1a) containing approximately 80 vol% augite with minor olivine (approximately 12 vol%) set in a fine-grained mesostasis (approximately 8%). This mesostasis contains a silica-rich phase, plagioclase laths, pigeonite and fayalite needles, Fe-Ti oxides, phosphates and trace amounts of sulfides. Augites form elongated laths (up to 1 × 0.5 mm) that tend to be clustered (Treiman 1990). Olivine is variable in size (on average 1 × 0.6 mm and up to 3 × 2 mm), evenly distributed in a given section and showing poikilitic texture enclosing augite grains (Fig. 1b). Based on these microtextural relationships, it is generally considered that olivine appears after augite in the crystallization sequence (Day et al. 2006; Sautter et al. 2002; Treiman 2005). Olivine crystals are sometimes broken (Fig. 1c) and may form clusters. Low-Ca pyroxenes may occur as overgrowths on augite and as replacement of olivine. Minerals observed in the mesostasis represent late-stage crystallization products. Governador Valadares shares the same principal petrographic features as Nakhla, with marginally more mesostasis and a little less olivine. The texture of NWA 817 is very similar, except that it has very large volume of mesostasis (up to 20%) that displays spectacular skeletal titanomagnetite crystals, pigeonite, fayalite, and apatite needles in a very fine-grained feldspathic groundmass (Sautter et al. 2002). The overall texture of large augite and olivine in a fine-grained matrix is typical of porphyritic basalts, and the relative abundance of phenocrysts with respect to the mesostasis is interpreted as the result of settling of cumulus grains from a basaltic magma (Treiman 2005 and ref. therein). In terms of mineral chemistry, augite displays large homogeneous cores of constant composition (Table 1, Fig. 2, Mg# 63 where Mg# is the molar ratio Mg/(Mg + Fe): Wo42En38Fs20). This core composition is nearly identical in all nakhlites (figs. 2 and 3 in Treiman 2005) and suggests that “all nakhlites are comagmatic or cogenetic either from the same magma or from very similar magma emplaced nearly simultaneously” (Treiman 2005). These cores have a euhedral to subhedral outline, and are surrounded by a narrow more iron-rich rim (10–20 μm, Table 1, Fig. 2), reaching Wo42En22Fs36 (Mg# 46) adjacent to the mesostasis. The fact that the compositional boundary between the homogeneous core and the Fe-rich rim is sharp indicates that augite core composition has been preserved. It is therefore concluded that primary augites had an Mg# of 63, and that melt inclusions trapped in cores have the potential to record primary magma composition.

Figure 1.

 Backscattered electron micrograph of Nakhla section. Olivine: light gray; augite: gray; mesostasis and melt inclusions: dark gray; oxides: white. a) General view. b) Subhedral poikilitic olivine with lobate grain boundaries. c) Broken olivine megacryst.

Table 1.   Augite and olivine composition in Nakhla.
 Augite-coreAugite-rimAugite mesostasisOlivine coreOlivine rim
  1. Note: nd = not detected.

  2. Each mineral composition is averaged over 30 analyses.

Figure 2.

 Pyroxene and olivine composition in nakhlites. Pyroxene quadrilateral with pyroxene compositions of Nakhla, Governador Valadares, NW 817, Northwest Africa (NWA) 998, MIL 03346, and NWA 5790. Augite core compositions of all nakhlites have nearly constant composition (field 1). Variable Fe/Mg increase toward the rim with concomitant decrease in Ca (arrows) except for MIL 0346 and NWA 5790 (field 2 surrounding individual dots) with an evolution toward hedenbergite instead of pigeonite. Olivine compositions are also shown (Fa and Fo contents). More Mg cores correlate with the extent of zoning except for NWA 5790 (Jambon et al. 2010).

Figure 3.

 Optical microscope photomicrographs showing trapped melt in olivine and augite grains from Northwest Africa (NWA) 998, NWA 817, Governador Valadares, and Nakhla. a) Governador Valadares Photo 10X under reflected light; olivine-hosted melt inclusion: radial fractures surround the inclusion; silica-rich phase (darker gray); augite (lighter gray) Fe-Ti minerals (white spots). b) NWA 998: Photo 40X under transmitted light; subrounded olivine-hosted melt inclusion; inclusion diameter is about 100 mm; marked fractures penetrate the inclusion; the silica-rich phase is recrystallized. c) NWA 998: Photo 40X under transmitted light (crossed polaroids); Augite-hosted type I melt inclusion; silica-rich phase is recrystallized as shown by the anisotropic behavior; daughter minerals and a shrinkage bubble are also present: the shape is irregular and tiny fractures penetrate the inclusion d) NWA 817: Photo 40X under transmitted light of type I augite-hosted melt inclusion; irregular shape; transparent recrystallized silica-rich phase and daughter minerals; shrinkage bubble is absent or totally altered. e) Nakhla: Photo 40× under transmitted light: type I isolated augite-hosted melt inclusion; silica-rich glass; daughter minerals (blacks crystal inside the inclusions are Fe-Tioxides) and a single shrinkage bubble; inclusion is isolated. f) Nakhla: Photo 40X under transmitted light; type Ia augite-hosted melt inclusion consisting of silica-rich glass, daughter minerals (augite, pigeonite, Ti-magnetite), and a shrinkage bubble. The inclusion is isolated and not penetrated by fractures; the silica-rich phase is partially devitrified as shown by the typical wrinkles (thin and diffuse gray lines in the inclusion), daughter minerals (black crystals inside the inclusion are Fe-Ti oxides), and a single shrinkage bubble. g) Nakhla: Photo 40X under transmitted light type II augite-hosted melt inclusion. The yellowish-brown inclusions forming trails are secondary inclusions that are associated with type I melt inclusions. Augite-hosted melt inclusion type II reflected light. h) NWA 998: Photo 10X under transmitted light type II augite-hosted melt inclusion trails and group of secondary inclusions are widespread in the crystal. Ol = olivine; Aug = augite; MI = melt inclusion.

The chemistry of olivine is more complex, although in general some degree of core to rim iron-enrichment is observed, except where olivine encloses augite. For example, in the case of Nakhla, olivine shows smooth core-rim variations from Mg# 36 to Mg# 26 (Table 1). Intriguingly, this core composition is much richer in Fe than that predicted to be in equilibrium with augite of Mg# 63 (olivine should have an Mg# of 53 in that case: Treiman 2005). The same is true for all other nakhlites, which have olivine core compositions from Mg# 43 to Mg# 34 (see figs. 2 and 3 in Treiman 2005). Given that olivine core composition is correlated with the degree of core-rim zoning, many researchers have concluded that the difference in Mg# between olivine cores from different samples results from late-stage diffusional reequilibration, which has modified original chemical profiles (Treiman 2005 and ref. therein). This idea has been used to calculate the near-solidus cooling rate assuming an initial olivine composition of Mg# 46 (Mikouchi and Miyamoto 2002; Mikouchi et al. 2005a): NWA 817 (2.2 °C h−1); Nakhla (0.04 °C h−1); GV (0.08 °C h−1); Lafayette (0.015 °C h−1). Later work by Hammer (2009) based upon pyroxene morphology in the groundmass of mesostatis-rich MIL 03346 suggests a cooling rate of 20 °C h−1 for that sample, broadly consistent with expectations based upon the results of Mikouchi and Miyamoto (2002). The systematic variations in groundmass proportion, olivine zoning, and derived cooling rates have been used to suggest a stratigraphic order within a single cooling pile (Mikouchi et al. 2003, 2005a), although consideration of olivine grain size and number density has been used to suggest provenance from multiple flows (Lentz et al. 2005). In either case, sample NWA 817 would have been located close to the surface, Nakhla/Governador Valadares from some intermediate depth, and NWA 998 from the greatest depth sampled by the nakhlite clan (Mikouchi et al. 2005a). Surprisingly, olivine found in NWA 5790, the least equilibrated of all nakhlites (Jambon et al. 2010), has a more Fe-rich (Mg# 35) core than olivine found in the other nakhlites close to surface (NWA 817 and MIL 03346).

Typology and Composition of Unmelted Melt Inclusions

Although melt inclusions occur in both olivine and augite, we have chosen to study only those occurring in pyroxene. As argued by Stockstill et al. (2005), the latter mineral is the only phase of indisputable cumulus origin, and it is a host that more efficiently protects inclusions from postentrapment modification (Danyushevsky et al. 2000, 2002; Gaetani and Watson 2000). Indeed, olivine-bearing inclusions are often connected to intercumulus matrix via radial fractures (Figs. 3a and 3b), allowing cation exchange between the melt inclusion and the mesostasis, compounded by the fact that Fe-Mg volume diffusion through olivine is much faster than through pyroxene (Dimanov and Sautter 2000; Gaetani and Watson 2000).

Augite-hosted melt inclusions in the studied samples form two different types: Type I melt inclusions are 10–100 μm in diameter (Figs. 3c–h), colorless to pale yellow, usually occurring as isolated or groups of ten or so inclusions within a single grain, randomly distributed but statistically concentrated in the augite core. Type II melt inclusions, less than a few μm in their maximum dimension (Figs. 3g and 3h), are the smallest inclusions detectable by standard optical microscopy. Colorless to dark brown, they are scattered in almost all augite grains, forming trails or groups with graphic textures. They are so small that their composition cannot be reliably measured by electron microprobe.

Type I melt inclusions are generally free of radial fractures (Figs. 3c–f). Their shape varies from rounded to subrounded (Nakhla: Fig. 3e; NWA 998: Fig. 3c) to irregular (NWA 817, Fig. 3d), sometimes having negative crystal shape in Nakhla (Fig. 3f). Unlike olivine-hosted melt inclusions (Harvey and McSween 1992a; Treiman 1993), spherical shrinkage bubbles (one bubble in a single melt inclusion, Figs. 3e and 3f) are observed in all the studied augite-hosted melt inclusions except in NWA 817, where melt inclusions are altered (Fig. 3d). A translucent silica-rich phase, glassy (Nakhla, Figs. 3e and 3f) or devitrified (NWA 817, Fig. 3d; NWA 998, Fig. 3c), coexists with μm-sized daughter minerals of anhedral to subhedral augite, titanomagnetite, pigeonite, chlor-apatite, and Fe-sulfide. We note that forsteritic olivine is never observed in these augite-hosted primary melt inclusions, although fayalite and hedenbergite are present in type I melt inclusions of NWA 817 (Bleiner et al. 2006). Type I melt inclusions can be further subdivided into type Ia, free of surrounding fractures and type Ib, where tiny fractures can reach and penetrate the inclusion. Type Ib melt inclusions appear darker in reflected light compared with type Ia melt inclusions.

Mineral and glass compositions of unheated melt inclusions have been measured (Table 2 and tables 1–3 in Bleiner et al. 2006). The augite host is enriched in Fe and depleted in Ca over a few micrometers at the contact with type I melt inclusions. This indicates some Fe loss from the inclusion to the host mineral probably balanced by complementary enrichment in Mg in the inclusion. Glass composition in both type Ia and Ib inclusions approaches that of alkali feldspar, with SiO2 ranging from 76.2 to 62 wt% (Table 2). In detail, the glass from type Ia inclusions is richer in Fe and Ca, and poorer in Si and Al, compared with average type Ib glass composition. Glass from type Ib melt inclusions is closer in composition to that described from olivine-hosted melt inclusions (Harvey and McSween 1992b; Treiman 1993), being richer in Si and poorer in Ca than type Ia melt inclusions. Daughter augite in type Ia inclusions is richer in Al, Fe, and Ti compared with the core of the large augite host (Mg# 64-76 against 43, tables 1, 2, and 3 in Bleiner et al. 2006). Daughter pigeonite is richer in Al, Mg, and poorer in Fe compared with pigeonite from the mesostasis (Treiman 2005). Daughter titano-magnetite is more Cr-rich (table 1 in Bleiner et al. 2006).

Table 2.   Silica-rich phase in Nakhla.
 In the mesostasisaIn augite hosted type Ia MIIn augite hosted type Ib MIIn olivine hosted MI
  1. Note: nd = not detected.

  2. aSilica-rich phase, partially devitrified in the mesostasis, is feldspatic.

Table 3.   NPM from rehomogenization of type Ia and type Ib augite-hosted melt inclusion (this study). NMP-Calc is the LA-ICPMS analyses of an unheated melt inclusion in Nakhla.
  1. Note: nd = not detected.

Na2O + K2O1.341.021.924.172.120.794.955.01

An attempt to quantify the bulk chemical composition of an unheated melt inclusion was made using LA-ICPMS measurements performed on a doubly polished section of Nakhla in a selected type Ia inclusion of 53 μm diameter. A laser beam 70 μm in diameter was used to ablate both the inclusion and augite host. The contribution from the host augite was assessed using the procedure described in Bleiner et al. 2006 (page 1630, fig. 7), but the resulting composition (“NPMcalc” in Table 3) appeared too rich in Mg and Ca, suggesting that correction for the augite host is not satisfactory. For this reason, we concentrated on the results of homogenization experiments, as described below.

Homogenization Experiments/Results

Selection of Suitable Melt Inclusions for Heating Experiments

To retrieve the most primitive melt from homogenization experiments, and potentially describe a range of composition reflecting magma evolution, care must be taken to select the most appropriate melt inclusions. In this respect, type II inclusions forming trails crosscutting the whole augite crystals were not considered, because they were probably trapped late in the sample history. Among type I inclusions, care was taken to avoid those showing traces of alteration or significant Fe-Mg re-equilibration with intercumulus liquid. Unfortunately, these criteria led us to eliminate melt inclusions from Governador Valadares, NWA 817, and NWA 998, as these are systematically surrounded by small fractures connecting them to the mesostasis. In Nakhla, we selected type Ia unaltered inclusions, as well as two pristine-looking type Ib inclusions, although some fractures do occur.

Chemical Composition of Heated Melt Inclusions in Nakhla

Type I inclusions were heated and homogenized using a slow-heating procedure akin to that used in Varela et al. (2001) and described above. EMP analyses of six homogenized (four type Ia and two type Ib) melt inclusions are given in Table 3. The four type Ia melt inclusions span a narrow range of basaltic composition (SiO2 from 47.8 to 50.6 wt%), significantly more restricted than that described by Varela et al. (2001) (from 48 to 60 wt%) and Stockstill et al. (2005) (from 47 to 64 wt%). Alkali (Na2O + K2O) contents are low (less than 2.1 wt%; Table 3) except the inclusion NPM-04-Ia. On the other hand, we find that rehomogenized type Ib inclusions correspond to more evolved andesitic to dacitic liquids (enriched in SiO2, Al2O3, and K2O). Taken together, type Ia and Ib compositions span a range of SiO2 values (47.8–65.9) in excellent agreement with the fast-heating experiments of Stockstill et al. (2005) (Table 4).

Table 4.   NPM from rehomogenization of augite-hosted melt inclusions (MI) in Nakhla.
  1. Note: nd = not detected.

  2. GHa,b,c,d are rehomogenized individual MI from Varela et al. (2001) with GH(6) being an average between 6 heated MI. S stands for Stockstill et al. (2005) data: S1b, S1a, and S2 represent, respectively, NA01b, NA01a, and NA02 from rehomogenized MI by slow-heating experiments; S3 and S4 represent, respectively, NA03 and NA04 from rehomogenized MI by fast-heating experiments.

Na2O + K2O2.1236.

Selection of the Most Appropriate Composition for NPM

Based upon the available textural and geochemical criteria, we conclude that homogenized type Ia inclusions are the best potential representatives of the nakhlite parent melt. To further constrain the most appropriate primary composition, we have chosen to focus on the Mg# of the liquids produced. Indeed, thermodynamic and experimental considerations may be used to show that cumulus augite of Mg# 63 should be in equilibrium with a liquid of Mg# 27 (Longhi and Pan 1989; fig. 6 adapted from Toplis and Carroll 1995; Treiman 1986, 1993). In this respect, we note that the Mg# of our four type Ia inclusions varies from 43 (NPM02) to 19 (NPM03), with an average value (NPM05) of 29. While the average value is close to that expected, we feel that such a large spread in Mg# is unlikely to represent a primary feature of the trapped melts. Alternative explanations include chemical interaction with the host augite (affecting Fe, Mg, Ca) and/or alkali loss during the slow-heating procedure. To assess these possibilities, the variations in major elements (SiO2, Al2O3, MgO, FeO, CaO) and minor elements (TiO2, Na2O, K2O) are considered as a function of Mg# (Fig. 4). The two low-silica liquids (<50 wt% SiO2) measured by Stockstill et al. (2005) have also been added. The core and rim compositions of the Nakhla augite host are also shown. For the sake of comparison, we include spatially dominant basalt compositions from Gusev crater as measured by the MER rover Spirit (McSween et al. 2006) and residual liquid compositions related to those basalts derived from experimental studies at low pressure (Monders et al. 2007; Filiberto et al. 2008). For virtually all the elements (Fig. 4) except K2O (Fig. 4f), compositions of the melt inclusions rehomogeneized in the present study appear to form a trend that passes through first the augite rim, then the augite core composition. Thus, variations in melt composition are clearly explained by augite addition (NPM 02) or augite subtraction (NPM 03) during heating experiments. Moreover, augite composition rimming the melt inclusion seems to have formed by reaction between augite core and interstitial melt trapped as inclusions. Note that CaO data (Fig. 4d) would argue against magmatic differentiation as the origin for these variations, the CaO content of NPM02 being unusually high for Martian basalts (cf. data for Gusev samples on Fig. 4d). For these reasons, we do not directly consider liquids with extreme values of Mg# (e.g., NPM02 with Mg# = 44 and NPM03 with Mg# approximately 19). Similarly, we rejected the NPM composition obtained from LA-ICPMS on an unheated type Ia melt inclusion in augite. On the other hand, we note that compositional trends systematically cross the line Mg# = 27 with relatively well-defined values (Fig. 4) (1) SiO2 approximately 50 wt%; (2) Al2O3 approximately 7 wt%; (3) FeO approximately 22 wt%; (4) CaO approximately 12 wt%; (5) MgO approximately 5 wt%; (6) Na2O approximately 1.5 wt%; (7) K2O approximately 0.5 wt%; (8) TiO2 approximately 1 wt%. We note that in agreement with previous studies, our NPM obtained from melt inclusions has superchondritic CaO/Al2O3 (apparently due to low Al2O3 abundance), in contrast to Gusev basalts (Fig. 4e). No amount of low-pressure crystallization of Gusev basalts would appear to modify the CaO/Al2O3 ratio sufficiently to produce liquids resembling our NPM composition. As discussed in the literature (McSween et al. [2009] and references therein), nakhlites and Gusev basalts thus appear to have originated from separate mantle reservoirs, possibly due to earlier periods of partial melting, as recently suggested based on the variation in Th content of volcanic regions at the surface of Mars (Baratoux et al. 2011).

Figure 4.

 Concentration of various oxides as a function of Mg number (Mg#). Symbols: black dot: type Ia rehomogenized MI in Nakhla of the present study (for number see Table 3); circle: S3 and S4 type I augite hosted MI from Stockstill et al. 2005; square: average analysis GH6 of Varela et al. 2001; star: augite core; empty star: augite rim; fields represent Gusev and Gusev experimental analog data points from Monders et al. (2007); Filiberto et al. (2008) on Gusev basalts (McSween et al. 2006). Note that trends for all oxides except K2O (f) point to augite addition and subtraction when establishing melt compositions. Vertical dashed line is Mg# 27 of liquid in equilibrium with cumulus augite.

In detail, we note that of our rehomogenized melt inclusions, NPM01 and NPM04 are of reasonable Mg# (Table 3). On the other hand, minor element contents of NPM04 (in particular Na2O and TiO2) are high relative to the other three inclusions of this study and to homogenized inclusions of Stockstill et al. (2005) (Figs. 4b and 4h). The case of K2O (Fig. 4f) is also of note as values derived from melt inclusion studies are extremely variable, even when only type Ia inclusions in augite are considered. Here again, comparison with data of Stockstill et al. (2005) leads us to prefer a value for the NPM of <0.5 wt% K2O. Based upon these criteria, NPM01 would appear to be our most direct representative of the liquid from which Nakhla’s augite initially crystallized, except that it is nearly devoid of K2O. For K2O, a value of 0.32 wt% has therefore been adopted, calculated from the average of the three other analyses. Based upon all of these criteria, we define our preferred nakhlite parent melt composition, shown in Tables 3 and 4 as NPM05.


Comparison with Other Studies on Augite-Hosted and Olivine-Hosted Melt Inclusions

The six augite-hosted melt inclusions from Nakhla described by Varela et al. (2001) were rehomogenized using a slow-heating procedure, similar to that used here. However, Varela et al. (2001) did not distinguish different types of inclusion and their average composition, GH06, is andesitic. It is the only quartz normative NPM composition in the literature and it is Al-, Na-, and Fe-poor compared with our NPM05 composition. The composition of GH06 most probably includes type Ia inclusions and contaminated type Ib inclusions connected with the chemically evolved mesostasis. Stockstill et al. (2005) performed both slow- and fast-heating experiments on eight melt inclusions in augite from Nakhla. For both heating procedures, they found heterogeneous compositions (cf. table 1, page 281 of Stockstill et al. 2005), which range from basalt to dacite, spanning a range of SiO2 (Fig. 5) similar to that described by Varela et al. (2001). Stockstill et al. (2005) opted to retain the most silica-poor liquid obtained from fast-heating experiments as their preferred NPM composition (S3; see Table 4 and Fig. 5) because this liquid was considered to be the most “primitive.” Using the MELTS thermodynamic routine (Ghiorso and Sack 1995), Stockstill et al. (2005) showed that predicted phase equilibria and mineral chemistry for S3 are broadly consistent with petrographic observations in nakhlites. Stockstill et al. (2005), nevertheless, concluded that augite and olivine are comagmatic in Nakhla. However, the first olivine to crystallize in their model is iron-rich (Fo34 in Stockstill et al. 2005table 2 page 384) compared with the expected olivine (Fo53) in equilibrium with augite cores for a liquid Mg# 27 (Fig. 6). Interestingly, reconsideration of the data of Stockstill et al. (2005) in the light of our own data shows that their composition S4 is remarkably close to that of NPM01 measured here. Furthermore, the Mg# (29) of S4 makes it a more than valid candidate for a primary nakhlite liquid, and we suggest that S4 rather than S3 (Mg# 23) is to be preferred as an estimate of the NPM. In summary, these different studies on augite-hosted melt inclusions show that (1) first and foremost, it is crucial to carefully select pristine primary melt inclusions on the basis of petrographic criteria. (2) The heating procedure (slow versus fast) would not appear critical for the reconstruction of primitive rather than evolved melts. (3) Comparison within and between studies, including other rocks from Mars, may help constrain acceptable compositional fields.

Figure 5.

 Calculated evolutionary paths of liquid composition produced from equilibrium crystallization modeling in MELTS plotted on the total alkali silica (TAS) diagram. Our data (points 1, 2, 3, 4, and 5) and published estimates of the nakhlite parent melt (noted S from Stockstill et al. [2005] and GH6 from Varela et al. [2001]) are plotted. S1a and S1b are from slow-heating experiments; S3, S4, S5, S6, and S7 are from fast-heating experiments. Predicted liquid lines of descent starting from: NPM05 (5) (black line); S3 (dashed gray line); S4 (dashed black line).

Figure 6.

 Variation in the Mg number of olivine (stars) and augite (squares) on vertical axis as a function of the coexisting liquid (horizontal axis) in iron-rich systems (data from Toplis and Carroll 1995). The white rectangle (1) on the olivine curve represents the range of olivine core composition (from Mg# 35 to Mg# 43) analyzed in all nakhlites (see Fig. 2). The white square (2) represents the expected but never encountered primary olivine core composition Mg# 53 that should be in equilibrium with primary augite core composition analyzed in all nakhlites (Mg# 63, see Fig. 2) represented by white square (3) for liquid with Mg# 27 (the vertical dashed line).

NPMs derived from olivine-hosted melt inclusions (NK3: Harvey and McSween 1992b, and NK93, NK93’, NK01: Treiman and Goodrich 2001) are potassium-rich compared with augite-hosted melt inclusions. This enrichment may be due to a difference of growth rate between pyroxene and olivine (Goodrich et al. 2010), with fast olivine growth trapping a boundary layer unrepresentative of NPM.

Olivine Status in Nakhla

As alluded to above, debate surrounds the question of whether the olivine megacrysts observed in Nakhla (and nakhlites as a whole) are phenocrysts that crystallized in situ or xenocrysts coming from elsewhere. The model that has current support is that the nakhlite parent magma was at or close to cosaturation with olivine, the observed offset of olivine Mg# being the result of late re-equilibration (e.g., Treiman 2005 and references therein). However, from a textural point of view, olivine megacrysts in Nakhla (and all nakhlites) crystallized after cumulus augite, the lobate shape of olivine grain boundaries (Figs. 1a and 1b) clearly pointing to disequilibrium with melt that precipitated augite mush. Furthermore, the lack of Mg-rich daughter olivine in augite-hosted melt inclusions would also argue against imminent olivine saturation of these trapped liquids. In addition, the idea that diffusive re-equilibration is responsible for apparent augite-olivine disequilibrium in terms of Fe-Mg partitioning also requires more critical assessment. This hypothesis requires a minimum time span for cation diffusion to propagate over at least one hundred microns in olivine megacrysts (average half diameter >200 μm). While this may be the case for slowly cooled samples such as Nakhla, the case for rapidly cooled samples such as NWA 817, NWA 5490, and MIL 03346 is not so straightforward. For example, it has been shown (Sautter et al. 2002) that Fe-Mg interdiffusion is not fast enough in NWA 817 to propagate iron enrichment to the olivine core. In this case, one concludes that the core composition of olivine megacrysts from rapidly cooled samples preserves original primary magmatic compositions (NWA 817: Mg# approximately 41; MIL 00336: Mg# approximately 43), clearly in disequilibrium with cumulus augite, and thus with our proposed composition NPM05. Using experimentally based correlations between the Mg# of coexisiting liquid, olivine, and augite (e.g., Fig. 6), we infer that olivine compositions such as those found in the cores of rapidly cooled nakhlites would have been in equilibrium with an augite of Mg# approximately 50 and a liquid of Mg# approximately 20. In light of these considerations, we aim to quantify if and with what composition our liquid NPM05 would have been saturated in olivine, with a view to providing a self-consistent picture of the status of olivine in the nakhlite clan of samples as a whole.

Constraints from Thermodynamic Modeling of NPM05

To provide insight into the crystallization sequence of NPM05, thermodynamic models have been used to predict phase equilibria and mineral/liquid compositions as a function of temperature, which complement earlier studies (Kaneda et al. 1998; Longhi and Pan 1989). Several models are available, such as MELTS (Ghiorso and Sack 1995), MAGPOX (Longhi 1991), and PETROLOG (Danyushevsky 2001). Work of Thompson et al. (2003) has compared application of these models with nakhlites, and in light of those results, only MELTS and PETROLOG will be considered here. Following Stockstill et al. (2005), these programs were run using a starting temperature of 1200 °C, an oxygen fugacity one log unit below the QFM buffer, and temperature steps of 10 °C down to 950 °C. Applying MELTS and performing calculations for equilibrium crystallization yield (Fig. 7a) clinopyroxene as the first liquidus phase appearing at 1170 °C, with a composition Wo43En41Fe16 (Table 5). Approximately 100 °C lower, plagioclase (An51Ab49) and olivine (Fo26) are predicted to saturate, when the system consists of 55% liquid and 45% augite. Spinel and ilmenite are predicted to appear at 1050 °C and orthopyroxene at 970 °C, just before whitlockite. At the solidus (T = 950 °C), olivine is Fo15, Cpx is Wo37En23Fs39, and plagioclase is An28Ab67. These results are in broad agreement with those obtained for the S3 composition modeled by Stockstill et al. (2005), differences in the temperatures of phase appearance being generally less than 10 °C. In the simulations of Stockstill et al. (2005), olivine saturates at 1080 °C, when the liquid has crystallized 25% augite, and the liquidus olivine has a composition of Fo29. We have also tested crystallization of composition S4 of Stockstill et al. (2005), as this liquid may also be a plausible parental liquid composition of the nakhlites, as discussed above. Those simulations, illustrated in Fig. 7b, are almost identical to results for NPM05, with saturation of an iron-rich olivine (approximately Fo31) well below the appearance temperature of Cpx. Simulations of fractional crystallization do not result in earlier saturation of olivine, and olivine compositions at saturation are, as expected, more iron-rich than those predicted for equilibrium crystallization. We note that for both equilibrium and fractional crystallization, augite is saturated at all temperatures, with compositions that fall within the range measured in nakhlites. Results of simulations using the PETROLOG program provide comparable results to MELTS, with the notable difference that an orthopyroxene (Mg# 60) is predicted to cosaturate with Cpx (Mg#60) at the liquidus temperature (1169 °C), and that no subsequent olivine saturation is predicted.

Figure 7.

 Calculated equilibrium crystallization sequence using MELTS program for magma having the bulk composition (NPM 05 [this study] and S4 [Stockstill et al. 2005]) at oxidation state corresponding to QFM-1Runs were performed from the liquidus temperature to 900 °C for all conditions. Dashed horizontal lines represent temperatures at which there was x% liquid remaining. The compositions of plagioclase (mole% anorthite) and olivine (mole% forsterite) during magma crystallization are given on the respective bars, and the total cumulative amounts of each phase crystallized (g of solid produced per 100 g of magma) are given at the bottom of the figure.

Table 5.   Proposed parent magma compositions for nakhlite.
  1. Note: nd = not detected.

  2. NPM05: this work; GH(6): Varela et al. 2001; S04: Stockstill et al. 2005; NK3: Harvey and McSween 1992a; NK01: Treiman and Goodrich 2001.

Na2O + K2O2.123.82.692.23.5

Taken together, these simulations indicate that olivine is not close to being a liquidus phase in liquids represented by primitive melt inclusions in augite from nakhlites. In this respect, it is of interest to note that recent crystallization experiments of a potential nakhlite parental melt performed by Imae and Ikeda (2008) produced augite with composition identical to those of phenocrystic core augite, but no olivine, although those authors interpret the lack of olivine as a kinetic artifact. Coming back to the results of the modeling, we note that when olivine does appear, approximately 100 °C lower than the liquidus temperature of cumulus Cpx, it is typically richer in iron than the core of olivine from the least reequilibrated nakhlites (NWA 817 and MIL 03346). On the other hand, this Fe-rich olivine is similar in composition to overgrowths rimming olivine megacrysts, and to grains associated with plagioclase (Ab48An51) found within the fine-grained matrix that precipitated from evolved intercumulus liquid (Table 5).

Despite these results, the fact remains that nakhlites typically contain approximately 10% of relatively Mg-rich olivine megacrysts (Figs. 1 and 2). As demonstrated just above, these crystals are unlikely to have been in direct chemical equilibrium with the liquids that precipitated cumulus clinopyroxene (e.g., NPM05) or their differentiation products, but textural relationships suggest that they reached some sort of equilibrium with the interstitial liquid in the nakhlite pile. We have therefore used MELTS to explore the petrologic and chemical consequences of adding various amounts of olivine to NPM05, with the aim of constraining the conditions necessary to saturate that liquid in olivine. Conceptually, we imagine that olivine was added from some external source to a melt + augite mush. We consider that diffusive re-equilibration of Cpx is negligible (as inferred by the fact that cumulus cores of augite in the nakhlites retain their original composition), such that the modeled system is a mechanical mixture of NPM05 and olivine, corresponding to the interstitial material in the nakhlite samples. Based upon petrographically determined modal proportions (e.g., Treiman 2005), the relative proportion of added olivine (i.e., olivine/[olivine + liquid]) was varied from 35 to 65% in steps of 10%. Composition of the added olivine is also a critical variable, but no clear constraint on this value is available. Values for XMg were explored from an upper limit of Fo70 taken from SNC meteorites in general (e.g., Chassigny), to a lower limit of Fo30 corresponding to olivine core composition of the least equilibrated nakhlite NWA 5790 (Jambon et al. 2010) and the olivine composition predicted to eventually saturate from NPM05 (see above).

Results of this modeling show that no addition of magnesian olivine (>Fo50) is capable of saturating NPM05 in olivine, the liquid becoming saturated in a low-Ca pyroxene. Furthermore, augite crystals in these models have an Mg# >63, suggesting that cumulus augite grains would have had Mg-rich rims, in contrast to petrographic observations. On the other hand, addition of Fe-rich olivine (<Fo45) does lead to saturation in olivine; the more iron-rich the olivine, the easier it is to saturate the liquid (i.e., the added proportion can be lower when olivine is Fe-rich). As an example, a mixture of 55% Fo30 and 45% NPM05 (Table 5) is predicted to be saturated in olivine (Fo53) at 1170 °C (i.e., the initial temperature of the interstitial liquid, corresponding to the liquidus of augite cores). At 1130 °C, equilibrium olivine has a composition of Fo43 (corresponding to that observed in the cores of rapidly cooled nakhlites MIL 03346 and NW A817). This observation is qualitatively consistent with the idea of adding cooler Fe-rich olivine, which comes to equilibrium with NPM05 at a temperature <1170 °C. The solidus olivine in this case is Fo25, slightly iron-rich compared with olivines observed in equilibrated nakhlites (NWA 998), although in detail the value of this parameter is a function of the amount and composition of added olivine.

These models provide qualitative insights that may be used to constrain the magmatic history of the nakhlite pile. In a first stage, we infer that augite crystals grew alone from a relatively evolved Al-poor mafic melt, forming a cumulate, as indicated by the close association of these minerals in clusters (Sautter et al. 2002, 2006; Treiman 1990). While still present as a mush, and possibly during emplacement, Fe-rich olivines were added, disrupting the crystal mush (explaining the presence of broken crystals in Fig. 1c), trapping augite (see olivine crystal in the top of Fig. 1a) and reacting with the melt to reequilibrate its Mg number (see lobate olivine rims in Figs. 1a and 1b). While these olivines may have come from a truly external source such as a placer deposit, similar to that inferred at Nili Fossae (Mustard et al. 2007), they may also simply have come from a cooler part of the nakhlite magma chamber itself, where an olivine of approximately Fo30 was on the liquidus. This latter scenario may explain why olivine-hosted melt inclusions are substantially more potassium- and titanium-rich compared with rehomogeneized augite-hosted melt inclusions (e.g., Goodrich et al. 2010). Further evidence for interaction of interstitial liquid and added Fe-rich olivine may be found in certain zoning profiles of augite. For example, Imae et al. (2005) and Imae and Ikeda (2007) show that the cores of cumulus augites in Yamato nakhlites (e.g., Y-000593) and MIL 03346 are surrounded by an inner rim characterized by rapid changes in Mg#, Al, and Ti, then an outer rim more characteristic of crystallization from an interstitial melt upon cooling.


Careful petrographic selection and high-temperature rehomogenization of melt inclusions from augite crystals in nakhlites lead to definition of a proposed nakhlite parent melt (NPM05) found to be very close to the composition S4 of Stockstill et al. (2005), but significantly different from olivine-hosted melt inclusions in terms of K2O content (Goodrich et al. 2010; Harvey and McSween 1992a; Treiman 1993). NPM05 cannot result from differentiation of a Gusev-type primitive picritic basalt, suggesting a difference in source composition between Hesperian basaltic magmatism at Gusev and Amazonian magmatism sampled by the nakhlites. Thermodynamic modeling suggests that NPM05 was not at or close to olivine saturation when augite cores precipitated. However, addition of Fe-rich olivine from an external source, possibly from colder regions of the nakhlite magma chamber, can account for ambiguous textural features in olivine (see figs. 1 and 5 of Sautter et al. 2002) and explain apparent Fe–Mg disequilibria between augites and coexisting olivine.


Acknowledgments–– We thank Omar Boudouma for SEM investigation in Paris, and F. Alvarez and C. Bottalli for support on the laser ablation facility in Zurich. J. Strobe is acknowledged for providing section of NWA 998 and B. Zanda at MNHN for providing the other nakhlite sections used in this study. D. Brunelli is acknowledged for helpful discussion. The present manuscript benefitted from helpful comments of the three reviewers M. McCanta, T. Mikouchi, and A. Ruzicka.

Editorial Handling–– Dr. Alex Ruzicka