The water content and parental magma of the second chassignite NWA 2737: Clues from trapped melt inclusions in olivine


Corresponding author. E-mail:


NWA 2737, the second known chassignite, mainly consists of cumulate olivine crystals of homogeneous composition (Fo = 78.7 ± 0.9). These brown colored olivine grains exhibit two sets of perpendicular planar defects due to shock. Two forms of trapped liquids, interstitial melts and magmatic inclusions, have been examined. Mineral assemblages within the olivine-hosted magmatic inclusions include low-Ca pyroxene, augite, kaersutite, fluorapatite, biotite, chromite, sulfide, and feldspathic glass. The reconstructed parental magma composition (A#) of the NWA 2737 is basaltic and resembles both the experimentally constrained parental melt composition of chassiginites and the Gusev basalt Humphrey, albeit with lower Al contents. A# also broadly resembles the average of shergottite parent magmas or LAR 06319. However, we suggest that the mantle source for the chassignite parental magmas was distinct from that of the shergottite meteorites, particularly in CaO/Al2O3 ratio. In addition, based on the analysis of the volatile contents of kaersutite, we derived a water content of 0.48–0.67 wt% for the parental melt. Finally, our MELTS calculations suggest that moderate pressure (approximately 6.8 kb) came closest to reproducing the crystallized melt-inclusion assemblages.


More than any other planet, Mars has captured our attention because of its similarity to Earth and possible habitability. Although many Mars exploration missions have been conducted, Martian meteorites still provide the only opportunity to investigate the red planet using most modern laboratory analytical techniques. The SNC (shergottite, nakhlite, chassignite) meteorites are Martian igneous rocks and can be used to trace the composition and evolution of the Martian crust and mantle. SNCs consist of several different lithologies, including basalts and ultramafic cumulates (McSween 1994). The shergottites have been grouped into three lithologic groups based on bulk chemistry and petrology (basaltic, olivine-phyric basaltic, and lherzolitic). However, of the currently available samples, only the basaltic shergottite QUE 94201 and olivine-phyric shergottites Y 980459 were argued to represent plausible primary magma compositions (Kring et al. 2003; Greshake et al. 2004). Recent work by Filiberto and Dasgupta (2011) recalculated the olivine-melt KDFe-Mg to a value of 0.35, and suggested that two additional shergottites, NWA 5789 and NWA 2990, could represent magmas based on equilibrium between their olivines and the bulk rocks. Nakhlites and chassignites are cumulate rocks mainly consisting of clinopyroxenes and olivines, respectively. As a result of their cumulative nature, the whole-rock compositions of nakhlites and chassignites cannot represent the primary magmas from which they were derived (Treiman 1993). The cumulate crystals occasionally contain melt inclusions. These inclusions have been useful tools in retrieving magma compositions from rocks that have experienced complex igneous histories (Harvey and McSween 1992; Treiman 1993), and also can be used for understanding source heterogeneities and volatile contents of parental melts (Sobolev 1996; Kent 2008).

Many attempts have been made to calculate the primary magma compositions of shergottites (e.g., Longhi and Pan 1989; Harvey and McSween 1992; Goodrich 2003), Chassigny (Floran et al. 1978; Johnson et al. 1991; Nekvasil et al. 2007, 2009; Filiberto 2008) and nakhlites (e.g., Treiman 1993; Varela et al. 2001; Stockstill et al. 2005; Sautter et al. 2012) using melt inclusions. Other approaches have utilized melting experiments or mass-balance calculations (e.g., Harvey and McSween 1992; Hale et al. 1999; Danni et al. 2001). However, these studies have resulted in disparate liquid compositions, which could suggest either that some of these compositions do not represent primary liquids or that there are multiple independent mantle reservoirs on Mars.

Chassigny was the only member of its class until 2005, when NWA 2737 was recognized as the second chassignite (Beck et al. 2006). These two dunites are particularly important in providing information on the early, high-temperature stages of Martian magma crystallization. From its mineralogy and texture, it is apparent that NWA 2737 formed from a basaltic parental magma, as did Chassigny (Floran et al. 1978; Johnson et al. 1991; Wadhwa and Crozaz 1995). Further analysis demonstrated that this parental magma was strongly enriched in incompatible elements (Beck et al. 2006), similar to that of Chassigny. However, because the process of reconstructing melt-inclusion compositions is complex, the exact composition of the chassignite parental magmas is a matter of dispute, especially their Al and alkali contents (Longhi and Pan 1989; Johnson et al. 1991; Nekvasil et al. 2007; Filiberto 2008).

An early attempt to decipher Chassigny's melt-inclusion composition was provided by Floran et al. (1978), who proposed a low-Ca melt that is compositionally similar to LL chondrites. Johnson et al. (1991) proposed a composition designated A* which has been accepted as a parental source composition for Chassigny by several authors including Borg and Draper (2003). Recently, however, other source compositions have been proposed, such as a parental magma similar to the Adirondack-class basalts of Gusev crater (Nekvasil et al. 2007). Experiments by Filiberto (2008) indicated that a liquid with composition A* could not produce the cumulate or melt-inclusion assemblages seen in Chassigny and instead proposed parental composition A' (Mg- and Al-rich relative to A*), which may also be linked to the Adirondack-class basalts. Experimental data by Nekvasil et al. (2009) further suggested the rock Backstay, also in Gusev crater, could represent a liquid parental to Chassigny-like dunites.

NWA 2737 consists of cumulate olivine with devitrified melt inclusions similar to those found in Chassigny (Beck et al. 2006). Although extensive petrographic studies have been conducted on NWA 2737 (Beck et al. 2006; Treiman et al. 2007; Van de Moortèle et al. 2007b; McCubbin and Nekvasil 2008), no attempts have been made to extract the parental melt composition from its melt inclusions. As the only other member of the chassignites, information on the parental melt of NWA 2737 should address questions about whether the chassignites shared a common parental melt and mantle source. In this article, we report a petrologic study of olivine-hosted melt inclusions in NWA 2737 and try to construct its original melt composition. In addition, we estimate the water content of the parental magma for NWA 2737 using hydrous minerals found both within the melt inclusions and some intercumulus melts.

Analytical Methods

We analyzed nine melt inclusions in one thin section of NWA 2737 (Fig. 1). Backscattered electron (BSE) images were taken using scanning electron microscopes (SEM, JEOL-845, and Hitachi S-3400N) with an energy dispersive spectrometer (EDS) at the Purple Mountain Observatory, Nanjing. Inclusions were identified optically using optical microscopy and then examined by SEM.

Figure 1.

Backscattered electron (BSE) image of the NWA 2737 section examined in this study, showing melt inclusions (yellow boxes). Ol-Olivine, Aug-Augite, Pgn-Pigeonite, Crt-Chromite, Fld-Feldspar glass.

Mineral chemistry was analyzed using a Cameca SX100 electron microprobe at the University of Tennessee. Both synthetic and natural mineral standards were used, and matrix corrections were based on ZAF (atomic number-absorption-fluorescence) procedures. Accelerating voltage was 15 kV with a beam current of 20 nA for silicate and oxide minerals, a 1–5 μm spot size, and 20 s counting time. A defocused beam (5–12 μm) and current of 10 nA were applied for feldspar and glasses. In addition, we produced a series of element X-ray maps for major elements, using a 15 kV, 40 nA beam.

Analyses of glasses in the inclusions in olivine were performed twice. The first time we used a feldspar standard and a defocused beam of 5 μm and 10 nA to assess whether glasses have maskelynite or rhyolitic compositions. The results consistently yielded totals of only approximately 94–98%, probably caused by Na loss. In spite of probable Na loss, we evaluated the analyses assuming feldspar stoichiometry by looking at (1) the cation sum, which, for feldspar, should be close to 5 per 8 O atoms; (2) the sum of Ca + Na + K, which should be near 1; (3) Si + Al + Fe, which should be near 4; and (4) Al + Fe, which should be close to Ca + 1. Constraints (1) and (2) are directly affected by alkali loss, whereas constrains (3) and (4) are only indirectly affected (McCubbin and Nekvasil 2008). The glass analyzed in melt inclusions could not satisfy constraints (3) and (4), with Si + Al + Fe near 4.2 and Al + Fe slightly higher than Ca + 1. Therefore, we concluded that no feldspathic glasses analyzed were stoichiometric maskelynite. The second analyses were performed using a glass standard and a broadened raster size of 8–10 μm, giving totals of approximately 97–99% and confirming alkali loss in the first analyses.

Element analyses (including F and Cl) for amphibole and apatite were conducted with an accelerating voltage of 15 keV, beam current of 20 nA, and beam size of 5 μm. A natural fluorite was used as the F standard and a natural halite was used as the Cl standard. For a detailed description of the F and Cl analyses by EPMA, see the discussion of McCubbin et al. (2010c) and Stormer et al. (1993). Due to the small size of the apatite in inclusions, microprobe analysis of apatite may result in overlap with adjacent phases and produce higher SiO2 values, and so we excluded analyses with elevated Si. Analytical errors (1σ) on all analyses were 1–2% (relative) for major elements (Si, Al, Mg, Fe, Ti) depending on the abundance of the element. Errors on F and Cl were within 8% (relative) and 3% (relative), respectively.

The OH (H2O) contents in amphibole and apatite were calculated based on stoichiometry. Amphibole has the chemical formula A0–1X2Y5T8O22 (OH, F, Cl, O), where A site can hold K+ and Na+; X site can hold Na+, Ca2+; and Mg2+; Y site can hold Fe2+, Mg2+, Mn2+, Ti4+, Al3+, and Cr3+; and T site can hold Si4+ and Al3+ (and rarely Fe3+). Kaersutite is a Ti-rich amphibole that typically contains at least 1 structural formula unit (sfu) of (OH + F + Cl) in the O(3) crystallographic site (Leake et al. 1997; McCubbin et al. 2010a). The chemical calculation of amphibole was described by McCubbin et al. (2010a).


Petrology of NWA 2737

The petrography of NWA 2737 has been described in detail by Beck et al. (2006) and Treiman et al. (2007). This olivine-rich igneous rock consists of 85% vol olivine, similar to Chassigny (Beck et al. 2006). NWA 2737 shows a typical cumulate texture consisting mainly of anhedral to subhedral olivine crystals, with rare poikilitic pyroxenes (Fig. 1). However, NWA 2737 has a somewhat higher Mg/Fe ratio than that of Chassigny for both whole rock and minerals (olivines and pyroxenes). Plagioclase is absent, but minor anorthoclase and/or Na-rich glass are present (Beck et al. 2006).

The most distinctive feature of NWA 2737 is the color of its olivine, which is dark brown in thin section (Fig. 2a) and almost black in hand sample. Some shocked olivine crystals are not uniformly brown, but contain light and dark bands and stripes (Fig. 2b). In BSE images, these stripes appear dark gray, indicating a compositional difference. Some olivine grains exhibit sets of subperpendicular planar defects (Fig. 2c). The coloration is interpreted as shock-induced because of the presence of Fe3+ formed along with solid-state precipitation of dispersed nanometer-sized particles of metallic iron (npFe0) (Treiman et al. 2007; Van de Moortèle et al. 2007b; Pieters et al. 2008). Treiman et al. (2007) suggested two shock events. One event, reaching extreme pressure (approximately 55 GPa), produced the dark brown color of olivine. The subsequent event produced the subperpendicular planar defects and presumably ejected the meteorite from Mars. Van de Moortèle et al. (2007a) reported a high-pressure polymorph of olivine (ζ-(Mg, Fe)2SiO4), along with nanophase metallic iron. A recent microstructural investigation of these strongly stained olivines supports the two-stage shock history (Bläß et al. 2010). The brown staining of olivine is consistent with dispersed metallic nanoparticles, an extreme high density and a high abundance of Fe3+ (Bläß et al. 2010). All these features are shock-induced. Some large magmatic inclusions have been observed within olivine grains (Fig. 2d).

Figure 2.

Photomicrographs of Martian meteorite NWA 2737. Images (a) and (f) are in transmitted plane-polarized light; (b), (c), (d), and (e) are under crossed-polarized light. Ol-olivine, Aug-augite, Pgn-pigeonite.

Pyroxenes, both pigeonite and augite, mainly occur as interstitial minerals (Fig. 2e). Rare poikilitic augites surround early cumulus olivine or chromite. Floran et al. (1978) also reported a single poikilitic pyroxene grain in one of the Chassigny sections. Pigeonites in some cases exhibit thin exsolved augite lamellae or blebs. Some augites surround pigeonite grains and the pigeonite has augite exsolution. All these features are consistent with slow crystallization under near-equilibrium conditions. Under cross-polarized light, pigeonites exhibit extinction (Fig. 2e). In some cases, augite and pigeonite show intergrowth textures (Fig. 2f), which may be attributed to the immiscibility gap between augite and low-Ca pyroxenes during the cooling and intercumulus crystallization in NWA 2737 (Reynard et al. 2006).

Olivines in NWA 2737 are chemically homogeneous (Fo = 78.7 ± 0.9). There are abundant magmatic inclusions in olivines (Fig. 2d) similar to those in other Martian meteorites. Compositions of pyroxenes define a tie-line (En80Wo2-En45Wo46) and anorthoclase compositions are An1–13Ab68–79Or15–23 (Beck et al. 2006). These minerals do not retain typical igneous zoning patterns, but are also chemically homogeneous. This equilibration could represent a later thermal event. Compared with Chassigny, the higher Mg content of olivines and the lower Ca content of feldspar suggest that it formed from a less differentiated or more primitive parental magma (Floran et al. 1978). The mesostasis in NWA 2737 mainly shows chaotic texture (Treiman et al. 2007) and does not resemble the coarse texture in Chassigny, but mostly contains small grains of low-Ca pyroxene and other minerals such as apatite that are angular fragments dispersed in the silicate glass. The silicate glasses are “granitic,” with compositions like a mixture of silica and alkali feldspar. Accessory ilmenite, pyrrhotite, and baddeleyite are also present.

Petrography of the Melt Inclusions

We located polyphase silicate melt inclusions within several of the highly equilibrated olivine grains. Figure 3 shows examples of these large melt inclusions. Their locations have been highlighted in Fig. 1. Some inclusions are near the center of grains, whereas others are close to the edges, although their host olivines have similar compositions. We identified eight phases within the melt inclusions, including olivine, low-Ca pyroxene, augite, kaersutitic amphibole, apatite, chromite, sulfide, and alkali feldspar-rich “granitic” glass. Some melt inclusions contain Ti-biotite.

Figure 3.

Representative BSE images or element X-ray mapping of melt inclusions in NWA 2737. Ol-olivine, LCP-low-Ca pyroxene, HCP-high-Ca pyroxene, Crt-chromite, Tr-troilite. Amphiboles are kaersutite. The dark phase filling part of the volume of the inclusions is feldspathic glass. The small bright crystals in the glass or on the surface of the pyroxene are sulfide crystals (in MI-2, MI-4, MI-8), and some sulfides form an internal boundary of the inclusions (MI-4, MI-9).

Several types of melt inclusions can be distinguished based on habits. Inclusions MI-1 and MI-2 are spherical or elliptical and large (approximately 100–200 μm in diameter). These inclusions generally have a crystallized wall of low-Ca pyroxene and contain coarse augite and kaersutitic amphibole crystals and glass, with minor chromite or ilmenite. MI-3, MI-4, MI-5, MI-6, and MI-7 are spherical or elliptical inclusions (50–150 μm across) that do not have low-Ca pyroxene inner walls. They contain prismatic low-Ca pyroxenes and glass with minor augite and chromite, while only a few contain kaersutite laths (MI-3). A third set of inclusions consists principally of olivine, pyroxenes, glass, and the olivine crystallized on the walls of the original inclusions. This type of inclusion (Fig. 3, MI-4, and MI-5) is rare. Finally, some inclusions show subangular to angular shapes (MI-8 and MI-9). These inclusions are very large (approximately 200–300 μm) and contain low-Ca pyroxenes and glass as essential constituents, with minor augite, chromite, kaersutite, sulfide, and phosphate. Similar-shaped inclusions were reported recently by Basu et al. (2011) and described as “mesostasis inclusions.” However, we note that these shapes may actually be unrepresentative of all possible slices; detailed analyses of other sections may be required to determine that.

Kaersutitic amphibole grains occur in both elliptical and angular shaped inclusions (MI-1, MI-2, MI-3, MI-8, and MI-9). Amphibole sizes vary from 30 to 50 μm, such as in MI-1, MI-8, and MI-9, and up to 100 μm in MI-2 and MI-3. They are euhedral to subhedral and in contact with glass and pyroxenes. Some amphibole grains contain anhedral low-Ca or high-Ca pyroxene in their centers (MI-3 and MI-2), suggesting a reaction relationship. The amphibole grain in MI-2 is intergrown with the low-Ca pyroxene. Commonly, glass patches in the inclusions are surrounded by void space and radial cracks in many cases. The glass has scalloped irregular edges against the void space, while the silicate minerals (pyroxene and amphiboles) maintain straight crystal boundaries. Apatite grains are usually small (5–20 μm) compared with amphibole. In some cases, sulfides occur as small blebs in the glass or on the surface of the pyroxene grains in MI-2, MI-4, and MI-8, and in two cases form an internal boundary within the inclusions (MI-4, MI-9).

The large magmatic inclusions are texturally similar to those in Chassigny (Floran et al. 1978; Johnson et al. 1991). They contain blocky crystals of pyroxene and amphibole set in “granitic composition” glass. In some inclusions, pyroxene crystals grew inward from the walls, comparable to textures of melt inclusions in Nakhla (Treiman 1993).

Smaller melt inclusions (a few tens of micrometers in diameter) and multiple inclusions within a single host grain are common (Fig. 4). They are round and consist mostly of silicate glass, pyroxene, and occasional chromite or apatite. However, some small inclusions do not contain pyroxenes. Some small inclusions have odd shapes that may be a result of length shearing at some point after trapping (Fig. 4). Chromites sometimes also contain melt inclusions, which are usually small in size (approximately 5–30 μm) and consist of alkali feldspar-rich glass with minor low-Ca pyroxene.

Figure 4.

Representative smaller magmatic inclusions. a, b) Multiple inclusions, a few tens of microns in size, are common within a single host grain. The large inclusion in the center of the image (b) is the only inclusion that contains biotite in our section. c) Some small inclusions with odd shapes that may be a result of lengthwise shearing at some point after solidification of the inclusion. d) Chromites sometimes also contain melt inclusions. Ol-olivine, Pgn-pigeonite, Crt-chromite.

In most large inclusions, all mineral phases including daughter minerals were large enough for analysis by electron probe. In this study, we analyzed all the phases in the large inclusions and estimated the parental melt compositions using these large inclusions.

Mineral Chemistry of the Melt Inclusions

All data for mineral and glass compositions in the nine large melt inclusions are shown in Table 1, S1, and Fig. 5. Both low-Ca and high-Ca pyroxene grains in different inclusions are indistinguishable in terms of major elements on the pyroxene quadrilateral. Like pyroxenes outside of inclusions, pyroxenes in melt inclusions are unzoned in Fe and Mg and have compositions similar to primary pyroxenes. All low-Ca pyroxenes within the inclusions are orthopyroxene, whereas the high-Ca pyroxenes are augites.

Table 1. Electron microprobe analyses of minerals and glasses (average values) in melt inclusions in NWA 2737
 Inclusion MI-1Inclusion MI-2
 Host Ol1AmpCrmLCPGlassMelt1Host Ol2AmpAptLCPHCPGlassMelt2
TiO2 4.090.700.19 0.61 6.49 0.461.84 2.82
Al2O3 13.3212.042.5220.7710.49 13.35 3.275.7921.539.26
V2O3  0.12          
Cr2O30.010.7348.010.26 0.630.010.21 0.080.20 0.13
MnO0.370.140.370.38 0.200.400.14 0.390.20 0.23
P2O5    0.060.02  39.21  0.110.32
Na2O 2.88 2.780.370.030.473.721.51
K2O 0.32  3.301.20 0.41   3.360.51
NiO0.12     0.14     0.00
Y2O3        0.01    
Ce2O3        0.16    
F 0.75   0.07 0.381.24   0.15
Cl 0.14   0.01 0.102.43   0.06
 Inclusion MI-3Inclusion MI-4Inclusion MI-5
 Host Ol3LCPHCPGlassMelt3Host Ol4LCPHCPGlassOlMelt4Host Ol5HCP
TiO2 0.391.03 1.39 0.441.84  0.57 0.27
Al2O3 2.025.0823.798.75 2.045.2923.09 3.56 0.98
V2O3          0.00  
Cr2O30.010.260.30 0.410.020.210.24
MnO0.400.380.22 0.240.380.370.23 0.380.330.380.38
P2O5   0.110.02   0.08 0.00  
Na2O 0.050.435.231.85 0.040.666.61 0.48 0.03
K2O  0.013.480.79   3.35 0.19  
NiO0.12   0.000.12 
F  0.14 0.10        
Cl  0.01 0.02        
 Inclusion MI-5Inclusion MI-6Inclusion MI-7
 AptGlassOlMelt5Host Ol6MicaGlassHost Ol7LCPHCPCrtGlassMelt7
TiO2   0.13 5.92  0.330.612.30 0.72
Al2O3 22.86 2.42 13.0621.84 2.192.2712.2522.9712.29
V2O3   0.00      0.22 0.05
Cr2O3  0.610.630.9444.99 11.13
MnO  0.390.350.390.04 0.390.360.190.49 0.24
P2O538.570.07 0.24  0.04    0.120.04
Na2O0.344.85 0.43 1.415.54 0.050.69 8.523.24
K2O 3.25 0.28 7.273.45    3.691.36
NiO  0.11    0.00
Y2O30.07  0.00         
Ce2O30.24  0.00         
F2.38  0.01         
Cl1.74  0.01 0.14       
 Inclusion MI-8Inclusion MI-9
 Host Ol8AmpLCPHCPGlassMelt8Host Ol9AmpLCPHCPCrtGlass TrMelt9
TiO2 5.070.391.15 1.00 3.490.271.281.97 Co0.250.70
Al2O3 12.612.194.4623.316.54 12.681.705.059.6923.23Ni6.629.74
V2O3     0.00    0.21 S33.610.02
Cr2O30.010.950.220.81 0.310.020.610.240.5949.41 Cr0.104.26
MnO0.400.140.390.21 0.300.390.100.370.180.48   0.21
FeO20.077.4012.385.560.349.8019.606.7811.806.2130.470.60  10.49
P2O5    0.060.01     0.07  0.02
Na2O 2.820.040.594.140.99 2.890.070.49 7.07  2.29
K2O 0.37  3.320.50     3.24  0.96
NiO0.12    0.000.13       0.28
F 0.48             
Cl 0.11             
Total99.0698.5399.3598.2297.5399.21100.497.60100.699.4098.0697.68  98.53
 Intercumulus melt
 Host Ol IMLCPAptGlass Tr IMIM Melt
  1. The samples of “s Ol” are compositions of olivines that surround the inclusions.

  2. IM is a typical intercumulate melt of NWA 2737 that has been analyzed for comparison with melt inclusions.

  3. Each analysis of these minerals and glasses is listed in Table S1.

TiO2 0.35  Co0.040.23
Al2O3 0.29 22.06Ni1.816.79
V2O3    S34.510.00
Cr2O30.010.07  Cr0.010.05
MnO0.400.40    0.26
FeO19.7512.340.660.34  8.28
P2O5  40.740.08  1.70
Na2O  2.44
K2O   3.17  0.95
NiO0.12     0.00
Y2O3  0.04    
Ce2O3  0.18    
F  2.11    
Cl  1.55    
Total100.1100.47101.0101.3  100.8
Figure 5.

Plots of CaO, TiO2, Cr2O3, Na2O, Fe/Mg, and Fe/Mn as a function of Al2O3 for low-Ca and high-Ca pyroxenes and amphiboles in inclusions, with primary augite and pigeonite shown for reference. The compositions of primary augite and pigeonite are average values after Beck et al. (2006).

Al2O3 contents of both low-Ca and high-Ca pyroxenes vary widely; however, they show small variations of CaO, although high-Ca pyroxenes generally have higher concentrations of Al2O3 (up to 7 wt%) than low-Ca pyroxenes (Fig. 5). The low-Ca pyroxenes show little correlation between Al2O3 and CaO, TiO2, or Na2O; while the high-Ca pyroxenes show slight correlations of Al2O3 with CaO, Na2O, and a well-defined trend of increasing TiO2. The variations of Cr2O3 with Al2O3 for both types of pyroxenes are more complex and show greater scatter than other elements. Both Fe/Mg and Fe/Mn ratios of the low-Ca pyroxenes show little variation, with average values close to those of the primary pigeonite (Fe/Mg = 0.243, Fe/Mn = 32). For the high-Ca pyroxenes, Fe/Mg and Fe/Mn ratios show small spreads around those of the primary augite, and Fe/Mg shows a positive correlation with Al2O3. While amphiboles do not have well-defined trends with Al2O3, they show somewhat greater spread than both types of pyroxenes in the other oxides. Overall, the low-Ca pyroxenes show little variation for most of the oxides and Fe/Mg and Fe/Mn, while high-Ca pyroxenes show a slightly broader distribution and amphiboles display the broadest distributions. These distributions imply low-Ca pyroxenes have undergone intensive equilibration with their host olivines whereas amphiboles have not.

Amphiboles in the melt inclusions are kaersutitic amphibole, with TiO2 ranging from 3.6 to 7.1 wt%. Amphiboles were analyzed in five melt inclusions and vary substantially in composition among different inclusions (Table 1). Apatites in melt inclusions are small grains, so zoning is difficult to determine. However, their compositions vary among different inclusions. In general, apatites contain more F than Cl than OH (Table 1), as discussed below. One of the large inclusions (MI-6) contains a rare occurrence of biotite, which displays heterogeneous composition and probably consists of composite grains rather than a single crystal. The analyses of the grains are reported in Table 1. They are TiO2 and MgO-enriched (average TiO2 = 5.92 wt% and MgO = 13.82 wt%). Glasses within each inclusion are not homogeneous (Table 1). They show a large range of Al2O3 contents (20–23.7 wt%) and do not show correlations between CaO and Al2O3.

Three chromite grains observed in the large inclusions (MI-1, MI-7, and MI-9) are similar in composition, with low Mg# (0.22–0.25) and high Cr/(Cr + Al) ratios (0.71–0.77). Mg# and Cr/(Cr + Al) values of chromites in the inclusions are lower than those of primary chromites (Mg# = 0.26–0.28, Cr/(Cr + Al) = 0.80–0.87, after Beck et al. (2006). Only one sulfide grain in MI9 is large enough for quantitative analysis (approximately 30 μm). It is pyrrhotite with Fe/S atomic values ranging from 0.82 to 1.03 (Table S2). The “troilite” analyzed by Floran et al. (1978) in Chassigny was also apparently pyrrhotite, (Fe0.87Ni0.01)S. Sulfide in inclusions is distinguished from interstitial sulfide by having higher Cr (0.1 wt%). Olivines have been observed in MI-4 and MI-5. They grew on the wall of the inclusion and their compositions are indistinguishable from the surrounding host olivine crystals (Fo = 78–79). However, these olivines are distinctive because they were apparently unaffected (not changed in color) by the shock event.

Water Content of the Parental Magma

The trapped melt inclusions can be used to infer the composition of the parental magma of their host phenocrysts. Several Martian meteorites have been reported to contain melt inclusions with hydrous secondary minerals, e.g., amphibole, biotite, and apatite (Floran et al. 1978; McSween and Harvey 1993; Watson et al. 1994; Mysen et al. 1998; McCubbin et al. 2010a). Their occurrences suggest that there are significant amounts of H2O in the pre-eruptive magmas (Floran et al. 1978; Johnson et al. 1991; McSween and Harvey 1993; Watson et al. 1994; McCubbin et al. 2010a). Water and other volatiles in Martian meteorites may provide clues to understanding the crystallization history of their parental magmas, and the eruption and degassing processes (McCubbin et al. 2010a).

The measured bulk contents of H2O in the SNC meteorites are low (e.g., 300 ppm for Shergotty; Lodders 1998). SIMS analysis and hydrogen isotope geochemistry of amphiboles in SNC meteorites suggested similarly low water contents (approximately 0.15 wt%, Watson et al. 1994; Leshin et al. 1996). Mysen et al. (1998) estimated the water content in Martian mantle was 1–35 ppm based on the analysis of water content in Martian kaersutites. Popp et al. (1995) noted that Ti-oxykaersutite in Chassigny kaersutite seems to be stable at 1 bar and inferred low water content. Boctor et al. (2003) measured directly apatite using SIMS that inferred possibly low water contents for the Chassigny magma. Similarly, low water contents were argued by Filiberto and Treiman (2009). On the other side, olivine-hosted melt inclusions in two primitive shergottites inferred dry mantle (Usui et al. 2012). It has to be noted that the partitioning of F, Cl, and OH between apatite and silicate melt has now proved to be much more complicated than previously recognized (i.e., Mathez and Webster 2005; Webster et al. 2009; McCubbin et al. 2010b), and those low OH abundances in Chassigny apatite (the OH contents of which were directly measured using SIMS by Boctor et al. 2003) are also consistent with water-rich magmas depending on the partition coefficients that are used.

Alternatively, several authors have argued for higher water contents in Martian parental magmas. Pyroxene chemical zoning patterns were interpreted to be consistent with large water contents for shergottite magmas (Lentz et al. 2001; McSween et al. 2001). Danni et al. (2001) and McSween et al. (2001) estimated a H2O-saturated melt (approximately 1.8 wt%) for Shergotty parental melt and a loss of water during the ascent of the magma. Melt inclusions within Chassigny olivine were used to calculate that the magma contained approximately 1.5 wt% H2O prior to kaersutite crystallization (Johnson et al. 1991). Treiman et al. (2007) analyzed kaersutite in NWA 2737 and found similar concentrations of F and Cl (0.09 wt% Cl, 0.45 wt% F) to those in Chassigny (Johnson et al. 1991), suggesting similar water contents. Furthermore, based on stoichiometric arguments for kaersutite, McCubbin et al. (2010a) reported that kaersutite in Chassigny melt inclusions contain higher abundances of water and lower abundances of F and Cl than previously measured, which suggested parental magma water contents between approximately 0.5 and 0.8 wt%. Balta and McSween (2012) argued for approximately 1.1 wt% H2O in the parental magma of olivine-phyric shergottite LAR 06319. The most recent study by McCubbin et al. (2012) implies elevated water contents similar to terrestrial mantle, which also provides strong support of hydrous magmatism on Mars, but lower magmatic water contents than the values above in this paragraph.

Given the disparate conclusions regarding the water content of the melt inclusion and parental magmas, additional data on volatile contents of chassignite melt inclusions are warranted. In this study, we calculated OH (by difference) in both kaersutite and apatite of NWA 2737 melt inclusions by anion deficiency. The apatite in intercumulate melt was used for comparison (shown in Table S3 and Fig. 6).

Figure 6.

Ternary plots of X-site components of melt-inclusion apatite and amphibole, and also interstitial apatite within NWA 2737.

The method of McCubbin et al. (2010a) estimates the minimum amount of OH in kaersutite, but the calculation requires knowing the Fe3+/ΣFe value of these amphiboles. Fe3+ content cannot be calculated from EMPA data alone, because EMPA cannot analyze for hydrogen. If we assume that the Fe-microXANES measurements of kaersutite in Chassigny by Monkawa et al. (2006) provide a good estimate for the NWA 2737 amphiboles (5% Fe3+), the calculated water contents of kaersutitic amphibole are as follows: For the inclusion MI-1 the calculated H2O content of the amphibole is 0.44 wt%. The TiO2 content of this amphibole is not high enough to be a kaersutite, and it indeed is a pargasite according to the nomenclature of Leake (1978) and Leake et al. (1997) with the chemical formula [A(K0.06Na0.48)0.54 B(Na0.32Ca1.68)2.00 C(Ca0.10Mg3.17Fe2+0.84Mn0.02Cr0.08VIAl0.35Ti0.44)5.0 T(Si6.10IVAl1.90)8.00 O22 (OH0.62,F0.34,Cl0.03,O1.01)2.00]. Different from the amphibole in MI-1, the amphibole in other melt inclusions is kaersutite (such as the amphibole in MI-2 has the chemical formula [A(K0.08Na0.64)0.72 B(Na0.15Ca1.85)2.00 C(Mg3.03Fe2+0.91Mn0.02Fe3+0.05Cr0.02VIAl0.26Ti0.71)5.00T(Si5.96IVAl2.04)8.00O22 (OH0.56,F0.18,Cl0.03,O1.25)2.00] with the calculated water content ranging from 0.57 wt% (MI-2 and MI-3) to 0.61wt% (MI-8)). Thus, the lower limit range for the water contents is 0.44–0.61 wt% H2O in the amphibole. This variation is similar to that in Chassigny (0.41–0.74 wt%) (McCubbin et al. 2010a). Studies of water partitioning between amphibole and silicate melt suggest that the water content of the melt must be greater than that of equilibrated amphibole by at least a factor of two (i.e., DH2O amphibole/melt ≤0.5 from Merzbacher and Eggler 1984; Johnson et al. 1991). Assuming that this partitioning coefficient is applicable to kaersutite, the range of magmatic water contents at the time of kaersutite crystallization was 0.88–1.22 wt%.

To estimate the water contents in primary melts, we must determine the extent of crystallization prior to appearance of amphibole. Our approach is based on the experimental phase equilibrium data of the parental melt composition for Chassigny melt inclusions. Filiberto (2008) proposed that the parent composition of chassignite is similar to the Adirondack-class basalt Humphrey, which was measured by Mars Exploration Rover Spirit (Gellert et al. 2006). Furthermore, based on the phase equilibrium experiments on the Humphrey composition (in the presence of volatiles), McCubbin et al. (2008) stated that amphibole enters the phase assemblage after about 45% crystallization. Thus, if we assume that amphibole enters the phase assemblage after about 45% crystallization for the inclusions in NWA 2737, then the water contents of the melt parental to the melt inclusions ranged from approximately 0.48–0.67 wt% H2O. This water content is consistent with that estimated for the Chassigny parental melt by McCubbin et al. (2010a) (0.43–0.84 wt% H2O).

Reconstruction of Primary Melts

Compositional Reconstruction for the Inclusions

Olivine is generally the first mineral on the liquidus and its trapped melt inclusions record chemical information of the magma that olivine formed from. However, obtaining the compositional information of the magma is a challenging task because various phases formed in melt inclusion after the melt was trapped in the olivine (Roedder 1979; Kent 2008).

Many attempts have been made to retrieve parental magma compositions from highly crystallized melt inclusions, mostly using nondestructive techniques. Although melt inclusions have been extensively exploited to obtain parental magma compositions of terrestrial igneous rocks (e.g., Sobolev 1996), most of these studies rely on experimental rehomogenization techniques. For meteoritic melt inclusions, the bulk composition of multiphase inclusion was estimate by averaging electron microprobe analysis with rastered or defocused beams (Treiman 1993), or using the weighted average of calculated bulk from model analysis (Goodrich 2003). We used an average composition calculated from model abundances of polyphase inclusions, as proposed by Goodrich (2003), Peslier et al. (2010), and Basu et al. (2011).

The compositions of phases found in the olivine inclusions, electron microprobe element X-ray maps, and BSE images of the inclusions were used to estimate the modal abundances and to calculate the bulk composition of an inclusion. The image processing software ImageJ (ImageJ version 1.44) was used to obtain the abundance of phases in inclusions, using the detailed methods as outlined by Jerram et al. (2003). The present bulk composition of an inclusion is defined to be the bulk silicate composition of its presently visible portion. The relative abundance of each phase (mol%) and calculated present bulk composition for each inclusion are shown in Table S4 and Table 2, respectively.

Table 2. The calculated present bulk compositions (pbcs) of the inclusions for NWA 2737
  1. Melt7-2 is calculated by excluding the chromites from the inclusion.

NiO    0.05   0.28 
SO2        2.77 

Among the calculated present bulk compositions (pbcs), MI-4 and MI-5 inclusions have extremely high MgO contents and MI-7 has an extremely high Cr2O3 content. The melt inclusions contain highly evolved phases of the postentrapment crystallization sequence such as oxides or sulfide at the center, ensuring that the whole inclusion is represented at the thick section surface and not just an off-center cut (Peslier et al. 2010). However, MI-4 and MI-5 inclusions are probably slices of off-center cuts because they are mainly composed of olivine and low-Ca pyroxene, and have no evolved phases in their centers. Large euhedral chromites are sometimes present in the large melt inclusions (such as in MI-7). When located at the edge of the inclusions, chromites are often primary minerals, which may have acted as nucleation sites for the formation of the melt inclusion and have been co-trapped in the inclusion (Roedder 1979). Thus, we reconstruct bulk compositions of the inclusions excluding MI-4, MI-5, and MI-7.

Postentrapment effects (crystallization of host phase onto walls, and the Fe/Mg re-equilibration with the hosts) can impact the initial inclusion compositions (Basu et al. 2011). Thus, we have to reconstruct primary trapped liquid compositions by addition of a portion of secondary olivine and corrections for inclusion-host interactions.

In all but the most rapidly quenched cases, secondary minerals within inclusions quickly re-equilibrate Fe and Mg with their host olivines, and simultaneous re-equilibration of the residual melt in the inclusion severely depletes FeO (Danyushevsky et al. 2000; Gaetani and Watson 2002). Leaving out H2O loss, no other important exchange reactions occur, as olivine does not accommodate significant amounts of any other cations and other diffusion coefficients are generally low (Cherniak 2010). To recalculate the original compositions, we used the Fe-loss calculation routine of the FeO-EQ2.EXE software (Danyushevsky et al. 2000), which reconstructs the parental composition by adding small increments of olivine to the melt composition while simultaneously accounting for the magmatic oxidation state. The calculation proceeds in two steps. The first step includes calculation of the composition of olivine in equilibrium with the current bulk composition and the temperature of the equilibrium. This step is followed by a balance-mass addition where a 0.1 wt% increment of calculated equilibrium olivine is combined with the liquid composition. This incremental addition causes the continuous increase in Mg# value of the melt, and, consequently, increase in Mg# of olivine calculated to be in equilibrium. The olivine/liquid KD value is highly dependent on bulk composition and temperature (Takahashi 1978; Toplis 2005; Matzen et al. 2011). The recent work by Filiberto and Dasgupta (2011) argued that KDFe/Mg Ol/melt of 0.35 ± 0.01 is appropriate for Martian compositions and is thus adopted here.

The main uncertainty in this calculation is the FeOT (total Fe as FeO) of the parental magma, which is unknown for NWA 2737 as it is a cumulate rock. However, we can estimate this value based on other chassignite parental melts, which show only limited variation in FeO. The parental melt of Chassigny has 19.95 wt% FeOT content (A*, Johnson et al. 1991) or 20.33 wt% (A', Filiberto 2008). The Martian surface basalts Humphrey analyzed in Gusev crater by the MER Spirit have an average FeOT content about 18 wt% (after Gellert et al. 2006). However, Filiberto (2008) showed that the A* could not produce the phases observed in NWA 2737 melt inclusions. Thus, we used 19 wt% FeO (the average FeOT content of the A' and the surface basalts Humphrey) as the estimated FeOT. The reconstituted melt-inclusion compositions are shown in Table 3.

Table 3. The reconstructed compositions of the melt inclusions for NWA 2737
  1. The Fe-loss calculation of the PETROLOG software (Danyushevsky et al. 2000, FeO-EQ2.EXE) was used and initial FeO* was proposed to be 19 wt %.

Melt 346.780.945.892.7222.050.2015.903.401.250.530.010.33142878.6−37.3137.3
A#49.37 1.98 8.44 2.11 17.09 0.22 12.11 6.06 1.57 0.61 0.16 0.26     

In melt-inclusion studies, the most primitive inclusion composition (low Si and high Mg) is normally chosen to represent the parental melt composition. Based on this standard, we would conclude that Melt2 and Melt3 (Table 3) most likely represent the primary melt composition of NWA 2737. Thus, the average of retrieved Melt2 and Melt3 may be the best candidate for the primary melt of NWA 2737 (A# was used for short, see Table 3).

MELTS Modeling of the Crystallization Sequence

For comparison with our inferred crystallization history, we also modeled the NWA 2737 crystallization sequence using the MELTS algorithm (Ghiorso and Sack 1995; Asimow and Ghiorso 1998; Smith and Asimow 2005). This algorithm has been used with some success to model the crystallization of Martian magmas at low pressure, and is therefore currently the best method available for simulating crystallization at modest pressure. Given the limited information available regarding trapping of these inclusions, we performed isobaric MELTS calculations and attempted to constrain reasonable trapping pressures for these inclusions, neglecting pressure changes that may have occurred during crystallization (although we can rule out extremely low pressures due to the presence of amphibole and the lack of obvious shrinkage/vapor bubbles). Nekvasil et al. (2007, 2009) inferred that Chassigny inclusions were trapped at pressures greater than 4.3 kb, and Nekvasil et al. (2009) suggested that the likely trapping pressure was between 6.8 and 9.3 kb. Similarly, McCubbin et al. (2008) estimated that approximately 9.3 kb was the closest pressure to reproduce the Chassigny inclusion crystallization path. Based on these constraints, we performed a series of isobaric crystallization calculations for our estimated primary melt composition, using the retrieved primary composition A# (Table 3), at pressures from 1 to 10 kb, but focused at 6.8 and 9.3 kb, with bulk water contents of 0.5 wt% as per our estimates of inclusion water content. We began all calculations at QFM-2, but then removed the oxygen buffer such that the Fe2+/Fe3+ ratio could be adjusted due to crystallization (which also stabilizes the MELTS calculation routines). We ran the calculation in several modes, testing both batch and fractional crystallization and both the MELTS calibration of Ghiorso and Sack (1995) and the pMELTS calibration of Ghiorso et al. (2002) at each pressure. All calculations started well above the liquidus and proceeded either to the point where calculation became unstable or 750 °C (outside the well-calibrated range), whichever was reached first. Calculated crystallization orders are displayed graphically for selected calculations, which neared completion as in Fig. 7.

Figure 7.

Orders of crystallization and minerals calculated using the MELTS algorithm. Modes are given in the figure, minerals are discussed in text, temperatures are given in degrees (C). Dashed line in (c) shows cut-off beyond which calculation could not be completed. Other calculations (not shown) in MELTS setting similarly did not run to completion. OPX = orthopyroxene.

The calculations reproduce some features of the inclusions. Pyroxene is the most abundant phase in all of the calculations, representing 60–75% of the calculated solid mass, with multiple pyroxenes present including orthopyroxene, augite, pigeonite, and an unobserved high-Al2O3, high-TiO2, low-SiO2 pyroxene. Analyzed pyroxene compositions generally fall within the range of calculated compositions with some exceptions (measured Al2O3 contents are higher than calculated, and measured FeO contents are lower than calculated in clinopyroxene, which could be attributed to the interaction with the host olivine in the natural samples). Although the mineral compositions are generally consistent with measurements, the crystallization order and mineral abundances show important differences from the actual inclusions. First, while using the original MELTS calibration, we found that orthopyroxene was the liquidus phase under all but the lowest pressures, while olivine saturation was not achieved (Fig. 7). Although there is substantial orthopyroxene in the inclusions, the fact that these grains are olivine-hosted and continued crystallizing olivine on the walls argues that olivine was a liquidus phase at the time of trapping. For this reason, we also chose to apply the pMELTS calibration, which puts olivine as the liquidus phase across this pressure range, although only a small amount is calculated to crystallize before pyroxene saturation at >9 kbar. Current work suggests that the MELTS algorithm overestimates the stability of pyroxene relative to olivine in Martian liquids, and as such the pMELTS calibration is likely closer to correct on this point (Gross et al. 2011; Balta and McSween 2012). Secondly, garnet appears as a stable phase in the fractional crystallization calculations at and above 6.8 kbar, which may suggest that it could have been a stable phase if the melts were trapped at high pressure, but even if a small amount crystallized, it could have reacted out as the inclusion cooled or if pressure dropped due to thermal contraction. Apatite is calculated to crystallize at all pressures. Finally, although both biotite and amphibole are observed in inclusions, biotite is the only hydrous phase to appear in the calculated crystallization path, although amphibole was calculated to appear well below the solidus (indicating that the calculation has approached but not reached its saturation curve at the low-temperature end of the calculation). Although the MELTS algorithm does calculate solubility of amphiboles, because of their complex chemistry, their saturation surface is poorly constrained by the program, and consequently, no significant conclusions can be drawn from the lack of appearance of amphibole in the calculations (Asimow and Ghiorso 1998; Asimow et al. 2004). Instead, it appears that the unobserved orthopyroxene phase has accommodated some of the components that would otherwise have driven amphibole crystallization, leaving biotite as the only calculated hydrous phase. The pMELTS calculations come closest to fitting the observed crystallization sequence at moderate pressures close to approximately 6.8 kbar. At 9.3 kbar, olivine is not abundant nor is it the major liquidus phase, suggesting that 9.3 kbar is too high for a potential trapping pressure. The MELTS calculations do not put a strong lower bound on the trapping pressure, although the lack of vapor bubbles argues against low-pressure trapping.

Comparison with Chassigny and Other Martian melts

Comparison with Chassigny

NWA 2737 differs in some aspects from Chassigny, and these differences might be important in understanding the petrogenesis of chassignites. Although the two meteorites have similar mineralogy, the Mg# is distinct (NWA 2737 has higher Mg# than Chassigny). The REE abundances in mineral phases indicate that the two chassignites formed from melts displaying a similar degree of LREE enrichment (Beck et al. 2006). Nd isotopic systematics are compatible with the proposed petrogenetic relationship between these two chassignites (Misawa et al. 2005), possibly through olivine fractionation from the parental melt. Unlike Chassigny, anorthitic plagioclase is absent from NWA 2737 and only albite is observed. The absence of plagioclase in NWA 2737 is consistent with a slightly lower Ca content of the parental melt for NWA 2737.

Several compositions have been proposed in the literature to represent the melt parental to Chassigny (e.g., low-Al and low-Mg liquid A*, Mg- and Al-rich composition A', and basalts Humphrey). Filiberto (2008) argued that A* failed to crystallize the Chassigny mineral assemblages. Our proposed parent composition for NWA 2737 (which we will call A#) has higher MgO content and lower CaO and CaO/Al2O3 comparing to A* (Table 4; Fig. 8). A# resembles A' and the Martian surface basalt Humphrey, albeit with a lower Al2O3 content (Table 4; Fig. 8).

Table 4. Summary of proposed primary Martian magma compositions





  1. 1. From this study; 2. Johnson et al. (1991); 3. Filiberto (2008); 4. The composition of the Martian surface basalt Humphrey (use the average composition of brush, RAT1, and RAT2) was after Gellert et al. (2006); 5. Hale et al. (1999); 6. Goodrich (2003); 7. Basu et al. (2011); 8. Greshake et al. (2004); 9. Bunch et al. (2009); 10. Gross et al. (2011); 11. Harvey (1993); 12. Sautter et al. (2012); 13. Stockstill et al. (2005); 14. Treiman (1993).

NWA 2737





2C50.531.768.19 7.428.990.5219.951.720.430.



3C45.551.3511.43 11.866.940.520.331.320.330.390.611.650.51



4 50.160.9413.45 8.416.110.2413.854.21.081.410.455.280.52
SaU 0056Ol-S51.7111.10.075.811.70.1416.
LAR 0637Ol-S49.40.797.010.6312.47.950.3118.91.330.490.481.131.820.54
NWA 29909Ol-S51.080.629.240.118.0611.670.4416.421.740.160.491.261.90.47
NWA 578910Ol-S49.240.485.650.4617.746.480.4716.610.430.020.381.150.450.66
LEW 8851611L-S49.470.616.88 14.878 16.941.270.06 1.161.330.61






13N51.20.847.1 4.611.20.5619.



Figure 8.

Melt compositions for the proposed SNC parent melts.

Comparison with Other Parental Melts

Parental melts for Martian meteorites have been estimated by many authors, as shown in Table 4 and Fig. 8. Among these compositions, the parental melts of Y 980459 and NWA 5789 (from Greshake et al. 2004; Gross et al. 2011) appear to be the most primitive based on their very high MgO compositions, while the parental magmas of nakhlites have the lowest Mg# value (Treiman 1993; Stockstill et al. 2005; Sautter et al. 2012). These proposed parents could also be grouped based on their CaO/Al2O3 ratios (see Fig. 8). Shergottite proposed parents have CaO/Al2O3 values between 1 and 1.5, which is often noted to be superchondritic. Nakhlite parents have much higher CaO/Al2O3. For chassignites, A' (proposed by Filiberto 2008 for Chassigny) and the NWA 2737 parent (this study) are similar to Martian surface rocks, which have lower CaO/Al2O3 (<1).

There is a large composition variation for shergottite parent magmas (Fig. 8). Some shergottite parents have much lower Mg# and higher CaO and Al2O3 contents, such as SaU 005, Shergotty, and NWA 2990 (Hale et al. 1999; Goodrich 2003; Bunch et al. 2009). Both Shergotty and NWA 2990 (olivine-phyric shergottites) are “enriched” shergottites (with relatively elevated incompatible elements abundances, Hale et al. 1999; Bunch et al. 2009), while SaU 005 is a LREE-depleted sample. The NWA 2737 parent resembles the average of shergottite parents or LAR 06319 (Basu et al. 2011), although they have very different CaO/Al2O3.

Based on the basaltic meteorites, Martian magma ocean models have been proposed by a number of authors including Borg and Draper (2003). Martian magma ocean models require all Martian basalt source regions to be superchondritic with respect to CaO/Al2O3, which is accomplished by garnet fractionation as the ocean cools. However, results from the Mars Exploration Rovers suggest that not all Martian rocks have high CaO/Al2O3. These rocks might represent magmas formed in portions of the mantle that were not depleted in Al. The Martian surface basalts from Gusev crater are estimated to have formed at approximately 3.6 Ga, much earlier than most of Martian meteorites (<1.3 Ga, Nyquist et al. 2001), and as such could be proposed to represent a limited or ancient mantle reservoir. In this study, we suggest that the parental melts of some young Martian meteorites (e.g., NWA 2737) have low CaO/Al2O3 ratios similar to Martian surface rocks. If this is the case, it suggests the preservation of two fundamentally different mantle reservoirs through time. One possible explanation for this preservation is that the global magma ocean model may simply be incorrect, and the Martian mantle never underwent significant homogenization. Alternatively, these basalts could represent sampling of a reservoir that was enriched in Al2O3 during differentiation, which would require melting either of a garnet-rich layer at high pressure similar to those discussed by Agee and Draper (2005) or by convective mixing of this Al2O3-rich layer to shallow levels where its melts would more easily reach the surface. Thus, it appears that the Martian mantle is indeed complex and has preserved distinct reservoirs throughout geologic time, one with a subchondritic CaO/Al2O3 ratio and another with a superchondritic CaO/Al2O3 ratio.


We report analyses of polyphase silicate melt inclusions from the second chassignite, NWA 2737. Inclusions differ in size, shape, and mineralogy. Both large rounded and angular inclusions have similar mineral assemblages, including olivine, low-Ca pyroxene, augite, kaersutite amphibole, apatite, sulfide, and alkali feldspar-rich “granitic” glass. Some melt inclusions contain Ti-biotite.

Olivine-hosted melt inclusions are useful tools in retrieving magma compositions for cumulate rocks. Based on the analysis of the volatile contents of kaersutite, we estimated that the parental melts have water contents ranging from approximately 0.48-0.67 wt% H2O. We also estimated the NWA 2737 parental magma composition (A#). The pMELTS calculations come closest to fitting the observed crystallization sequence at moderate pressures close to approximately 6.8 kbar, which suggests 9.3 kbar is too high of a potential trapping pressure. The A# composition resembles that of A' (experimentally constrained composition by Filiberto 2008), albeit with a lower Ca content. The estimate also has a low CaO/Al2O3 ratio that is similar to the Martian Gusev basalts, suggesting preservation of a low CaO/Al2O3 mantle reservoir through time.


This study was supported by Natural Science Foundation of China (41072045, 41173076, 41273079), Doctoral Fund of Ministry of Education of China (20090145110001) and the Scientific Research Foundation for the Returned Overseas Chinese Scholars, State Education Ministry (2010012009) and the Minor Planet Foundation of China. We thank Allan Patchen (University of Tennessee) for assistance during the electron microprobe analysis. We also would like to express our gratitude to Dr. Francis McCubbin and Dr. Jean-Alix Barrat for their constructive and helpful comments. Dr. Randy Korotev is acknowledged for handling and editing the manuscript.

Editorial Handling

Dr. Randy Korotev