Petrography, mineral chemistry, and crystallization history of olivine-phyric shergottite NWA 6234: A new melt composition


Corresponding author. E-mail:


Knowledge of Martian igneous and mantle compositions is crucial for understanding Mars' mantle evolution, including early differentiation, mantle convection, and the chemical alteration at the surface. Primitive magmas provide the most direct information about their mantle source regions, but most Martian meteorites either contain cumulate olivine or crystallized from fractionated melts. The new Martian meteorite Northwest Africa (NWA) 6234 is an olivine-phyric shergottite. Its most magnesian olivine cores (Fo78) are in Mg-Fe equilibrium with a magma of the bulk rock composition, suggesting that it represents a melt composition. Thermochemical calculations show that NWA 6234 not only represents a melt composition but is a primitive melt derived from an approximately Fo80 mantle. Thus, NWA 6234 is similar to NWA 5789 and Y 980459 in the sense that all three are olivine-phyric shergottites and represent primitive magma compositions. However, NWA 6234 is of special significance because it represents the first olivine-phyric shergottite from a primitive ferroan magma. On the basis of Al/Ti ratio of pyroxenes in NWA 6234, the minor components in olivine and merrillite, and phosphorus zoning of olivine, we infer that the rock crystallized completely at pressures consistent with conditions in Mars' upper crust. The textural intergrowths of the two phosphates (merrillite and apatite) indicate that at a very last stage of crystallization, merrillite reacted with an OH-Cl-F-rich melt to form apatite. As this meteorite crystallized completely at depth and never erupted, it is likely that its apatite compositions represent snapshots of the volatile ratios of the source region without being affected by degassing processes, which contain high OH-F content.


The compositions of Martian basaltic magmas can provide crucial clues for understanding Mars' mantle evolution and its volatile budget. In particular, Martian basaltic shergottites have yielded many insights into the planet's bulk composition, its differentiation, the nature of distinct geochemical reservoirs, and the geologically recent ages of magmatic activities on Mars (e.g., Dreibus and Waenke 1982, 1985; Jones 1986; Waenke 1991; Borg and Draper 2003; Treiman 2003; Agee and Draper 2004; Musselwhite et al. 2006; Draper and Agee 2008; Usui et al. 2008). Olivine-phyric shergottites have been recognized as a significant and important subgroup of the Martian shergottites (Goodrich 2002). Their relatively high bulk rock and olivine core magnesium numbers (Mg# = molar Mg/[Mg+Fe]) suggests that they could represent primitive melts, i.e., unfractionated liquids formed by direct partial melting of the mantle. Primitive melts can provide direct information about their mantle source regions, including compositions and mineralogy (e.g., Langmuir et al. 1992; Asimow and Longhi 2004). However, most olivine-phyric shergottites have a bulk rock Mg# that is too high to have been in equilibrium with their most magnesian olivines (e.g., Northwest Africa (NWA) 1068, Dar al Gani (DaG) 476, Sayh al Uhaymir (SaU) 005, Elephant moraine (EET) A79001 lithology-A, Dhofar 019). Therefore, they may not represent magma compositions, but instead may contain cumulate olivine crystals and/or have been affected by magmatic contamination (e.g., McSween and Jarosewich 1983; Wadhwa et al. 2001; Barrat et al. 2002a, 2002b; Goodrich 2002, 2003; Taylor et al. 2002; Shearer et al. 2008; Papike et al. 2009; Filiberto et al. 2010b, 2012; Filiberto and Dasgupta 2011).

Extensive work on olivine-phyric shergottites has already given clues to their crystallization history, magma versus accumulation issues, volatile history, and oxygen fugacity of their source regions (e.g., Herd 2003, 2006; Filiberto and Treiman 2009; Filiberto et al. 2010b, 2012; Gross et al. 2011; McCubbin et al. 2012). For example, experimental and mineralogical studies have shown that some olivine-phyric shergottites represent magma compositions, while many others contain up to 30% cumulate (phenocrystic or xenolithic) material (Treiman et al. 1994; Musselwhite et al. 2006; Usui et al. 2008; Filiberto et al. 2010b; Filiberto and Dasgupta 2011; Gross et al. 2011). Based on the compositions of apatite in olivine-phyric shergottites, the magmas in equilibrium with the apatite are thought to be enriched in chlorine compared with terrestrial basalts, and to contain a range of water concentrations (Filiberto and Treiman 2009; Patiño Douce and Roden 2006; Patiño Douce et al. 2011; McCubbin et al. 2012; Gross et al., personal communication). Furthermore, studies of olivine-pyroxene-spinel equilibrium from olivine-phyric shergottites provide evidence for low oxygen fugacities (approximately 1–3 log units below the QFM buffer) in Martian basalts and their mantle source region; more oxidized groundmass assemblages in these meteorites may reflect derivation and mixing of distinct magmas from oxidized source regions, the effects of degassing or internal fractionation of ferric iron, or as well as a possible oxidizing agent within the Martian crust that has contaminated some magmas during eruption and/or emplacement (Herd et al. 2002; Herd 2003, 2006; Peslier et al. 2010).

Here, we describe the mineralogy, petrology, and mineral chemistry of the new olivine-phyric shergottite North West Africa 6234 (hereafter NWA 6234). From petrographic observations, microprobe analyses, and zonation patterns of minerals of a doubly polished thick section, we constrain various aspects of its petrologic history, including conditions of crystallization (pressure, temperature, and oxygen fugacity) as well as the crystallization sequence of this magma. We also constrain the pre-eruptive volatile history of NWA 6234 based on merrillite/apatite interaction with an OH-Cl-F-rich melt.

Sample and analytical technique

Meteorite NWA 6234 was found in 2009 at an undisclosed location in Mali and purchased by an anonymous collector in February 2010. It was a 55.7 g partly fusion-crusted stone, cross-cut on the interior by several thin shock veins. A 3.31 g slice of the meteorite, which included a possible shock vein, was purchased from Marmet Meteorites. The sample was split, and distributed to an international consortium team for investigations, including this study (Filiberto et al. 2011). Filiberto et al. (2012) reported on the bulk rock geochemistry of this meteorite. Analyses here are from two polished thick sections that contain a melt vein that cuts through the sections (Fig. 1).

Figure 1.

BSE images of NWA 6234. Larger image shows the overall fine-grained texture, which is composed of olivine (Ol) crystals (0.1–1.0 mm diameter), set in a finer grained groundmass (see detailed inset) of pyroxene (Pyx), maskelynite (Mask), ferroan olivine (Ol), oxides (Ox), merrillite (Merr), and apatite (Ap). The arrow points to the melt vein in the rock that runs almost horizontally to the right.

Backscattered electron (BSE) images were taken with the Cameca SX100 electron microprobes (EMP) at NASA Johnson Space Center (JSC) and at the American Museum of Natural History (AMNH). These images were used to determine the textural characteristics and the modal mineral abundance using techniques described by Maloy and Treiman (2007).

Mineral chemical compositions were obtained with the Cameca SX100s at NASA JSC and at the AMNH. Operating conditions for minerals other than apatite (see below) were: 15 kV accelerating voltage, 15–20 nA beam current, focused electron beam (1 μm in size), and peak and background counting times of 20–40 s per element. Analytical standards were well-characterized synthetic oxides and natural minerals including diopside (Si, Ca, Mg,), olivine (Si, Mg, Fe), oligoclase, albite, jade (Na), hematite (Fe), rutile (Ti), corundum, augite (Al), chromite (Cr), Ni-diopside (Ni), rhodochrosite (Mn), orthoclase (K), and apatite (P). Data quality was ensured by analyzing standard materials as unknowns.

Apatite analyses were undertaken using the method of Goldoff et al. (2012) and Webster et al. (2009) to minimize F, Cl, and Na loss during analysis. Goldoff et al. (2012) and Webster et al. (2009) showed that the best apatite analyses are obtained if Na, Cl, and F are analyzed first with an acceleration voltage of 10 kV and 4 nA beam current. Other elements (P, Si, Fe, Mg, Al, Mn, Ti, Ca, K, S, and Ce) can be analyzed thereafter with an acceleration voltage of 15 kV and 20 nA beam current. All apatite analyses were obtained (and calibrated) with a defocused electron beam (10 μm in diameter). Peak and background counting times were 30 s and 15 s per element. Analytical standards were well-characterized synthetic and natural oxides and minerals including diopside (Si, Mg), K-feldspar (Al, K), wollastonite (Ca), olivine (Fe), jade (Na), rutile (Ti), MgF2 (F), rhodochrosite (Mn), boracite (Cl), troilite (S), CePO4 (Ce), and berlinite (P). Data quality was ensured by analyzing standard materials as unknowns. Hydroxyl in apatite cannot be measured directly by EPMA; however, the missing component in the X-site of the apatite was calculated from stoichiometry as OH- (assuming no vacancies or O2− substitutions), using the method of McCubbin et al. (2011).

Kα X-ray intensity maps for P, Al, Ca, Cr, Fe, Mg, and Ti were obtained using the SX100 at the AMNH with an accelerating voltage of 15 kV, beam current between 40 and 100 nA, beam diameter of 1 μm, and pixel spacing between 1 and 2 μm. Five images were acquired simultaneously. With two cycles per image, 10 elements could be mapped. Phosphorus was mapped on three spectrometers simultaneously (a PET and TAP crystal and a large PET [LPET] crystal) to improve the counting statistics and the signal-to-noise ratio. Dwell times ranging from 100ms to 260ms per point were sufficient to detect phosphorus zoning.

Petrology and mineralogy

NWA 6234 is an olivine-phyric shergottite with an unusual texture compared with other olivine-phyric shergottites. It is relatively unaltered, fine-grained, and composed of olivine crystals (0.1–1.0 mm diameter) set in a finer grained groundmass of pyroxene, maskelynite, ferroan olivine, spinel, ilmenite, merrillite, apatite, and accessory Fe-sulfide (Fig. 1; inset). The modal mineralogy, by volume, of the rock was determined by image analysis of a BSE image mosaic of a whole slab face (135 mm2) using a technique similar to that of Maloy and Treiman (2007) resulting in 35% olivine; 38% pyroxene; 24% plagioclase; and 3% oxides, phosphates, and accessories.


Olivine grains have euhedral to subhedral shapes. Their size distribution is seriate, i.e., grains range more or less continuously in size, from approximately 10–15 μm in length up to approximately 1.0 mm (Fig. 1). The largest olivine grains contain inclusions of chromite and of partially crystallized melts. The melt inclusions are commonly composed of glassy material with tiny dendritic crystals of pyroxene and/or olivine and/or plagioclase. The host olivine rarely shows Fe/Mg zoning at the contact with the melt inclusion.

Olivine grains larger than 200 μm in size are zoned. The cores are as magnesian as Fo78 (Table 1; Figs. 2 and 3) and are homogeneous, with consistently higher Fo content than the mantle areas (Figs. 2 and 3). Outside the core zone, the Fo content slowly decreases (Fo76–62) with slowly increasing CaO content in the wide mantle area toward the rim. The rim has a Fo content from Fo75–68 (Figs. 2 and 3). Because olivine shows a seriate texture, the cores of smaller grains have a Fo content equal to that of the mantle of slightly larger grains (Fig. 2). Olivine grains smaller than 200 μm are not zoned and are iron-rich ranging from Fo55 to Fo43 in very small grains (Fig. 2). Minor elements in olivine have almost no relationship with Fo content, as Fo decreases only MnO increases significantly, while CaO increases slightly (Fig. 3). NiO and Cr2O3 show no correlation.

Table 1. Representative microprobe analyses of olivine
 Large olivineMedium olivineMatrix olivine
  1. na = not analyzed; bd = below detection limit.

Oxide wt%
Normalized to 8 oxygen
V  0.000.00   
Figure 2.

Calcium abundance in large olivine (>600 μm; circles), smaller sized olivine (<600 μm; squares), and the Fe-rich olivine in the groundmass (<200 μm; diamonds). Note the overlap of the composition, suggesting that the olivines are cogenetic and seriate in nature.

Figure 3.

Chemical zoning profile across a large olivine crystal from core to rim. “Fo” is forsterite content (molar Mg/(Mg+Fe) × 100), left y-axis; other elements are in wt%, right y-axis. For details, see text.

X-ray maps of phosphorus in the olivine (Fig. 4) reveal oscillatory zoning patterns mostly parallel to the olivine edges, of thin (<10 μm) phosphorus-rich zones alternating with wider (50 to >150 μm) phosphorus-poor zones. The phosphorus zoning does not correlate with that of any other element. The larger grains typically have a thin phosphorus-rich core region followed by a very wide (>150 μm) phosphorus-poor region, and strong oscillatory zoning toward the rim. As the olivine grains show seriate texture, it is not surprising that some of the smaller grains have strong oscillatory phosphorus zoning starting from the core toward the rim. The very small olivine grains with Fo55–43 appear to be homogeneous in terms of phosphorus and other elements. Areas immediately around melt inclusions and chromite inclusions show neither enrichment nor depletion in phosphorus.

Figure 4.

Element map of phosphorous in olivine. The zoning has an oscillatory pattern mostly parallel to the olivine edges, with thin (<10 μm) phosphorous-rich zones alternating with wider (50 to >150 μm) phosphorous-poor ones. The bright white spots in the groundmass are merrillite and apatite crystals.


Pyroxene occurs in the groundmass as subhedral to anhedral grains clustered together. Single grains are rare, the largest subhedral prismatic grain measures approximately 300 μm in length (Fig. 1). The cores are Mg-rich orthopyroxene (En70Fs26Wo4), which zone outward by continuously increasing Ca and Fe contents to pigeonite (En46-49Fs29–43Wo11–19) (Table 2; Fig. 5). A second type of pyroxene found interstitially is a Ca-rich augite (En37Fs26W37) (Table 2; Fig. 5). Glass or melt inclusions appear to be absent from pyroxene grains.

Table 2. Representative microprobe analyses of pyroxene
 Core Mantle-RimHigh Ca
  1. na = not analyzed; bd = below detection limit.

Oxide wt%
Normalized to 12 oxygen
Cations total 4.013.984.
Figure 5.

Pyroxene quadrilateral for NWA 6234. Pyroxene shows systematic chemical zoning, core to rim, from magnesian orthopyroxene (enstatite) to pigeonite. A second type of pyroxene found interstitially is Ca-rich augite. The range of olivine compositions is also shown at the bottom of the pyroxene quadrilateral. The cores of large olivine grains (brown circles) have slightly higher Mg# than the core pyroxene (orange circles); olivine mantles (green circles) have similar Mg# to intermediate pyroxene compositions (blue circles). For detailed explanation, see text.


The oxide minerals observed are ilmenite and spinel ranging in composition from chromite to titanomagnetite. They are present as small grains and inclusions in other minerals throughout NWA 6234. Olivine crystals and their melt inclusions contain euhedral chromite grains (Fig. 1). Subhedral to anhedral grains of chromite, ilmenite, and titanomagnetite occur in the groundmass, typically as larger grains cored by chromite with titanomagnetite rims or as ilmenite-titanomagnetite intergrowths interstitial to pyroxene and plagioclase. Chemical analyses of oxide minerals are given in Table 3. Ferric iron (i.e., magnetite component) is calculated from charge balance, assuming that the oxides contain no cation vacancies.

Table 3. Representative microprobe analyses of oxides in NWA 6234
 Chromite TitanomagnetiteIlmenite
  1. a

    Calculated from stoichiometry.

  2. b

    Chr, chromite; Mt, magnetite; Sp, spinel; Usp, ulvöspinel.

  3. c

    Relative to the Iron-Wüstite (IW) buffer as defined by Herd (2008) using the data of O'Neill and Pownceby (1993).

  4. d

    Relative to the Quartz-Fayalite-Magnetite buffer as defined by Wones and Gilbert (1969).

  5. e

    Calculated according to the CT Server Ol-Px-Sp model ( using pyroxene Wo2.4En68.9Fs28.7 and olivine Fo67 at P = 1 bar

  6. f

    Calculated according to the CT Server Ol-Px-Sp model ( using pyroxene Wo10.9En47.5Fs41.6 and olivine Fo49 at P = 1 bar

  7. g

    Calculated according to the Fe-Ti oxide oxythermobarometer of Ghiorso and Evans (2008), using the oxide pair given in the table.

Oxide wt%
Total cation 2.993.013.022.02
Molar composition b
fO2 estimates
T (°C)1103922674843
PhasesOl, PxeOl, PxeOl, PxfOxidesg

The chromite grains vary in composition according to their setting. Chromite inclusions in olivine and cores of groundmass grains are the most magnesian and Cr-rich, with 79–83% chromite (Chr = molar Cr/[2Ti+Cr+Al+Fe3+]) and 11–12% spinel (Sp = molar Al/[2Ti+Cr+Al+Fe3+]) with 4–5% ulvöspinel component (Usp = molar 2Ti/[2Ti+Cr+Al+Fe3+]) and 1–5% magnetite component (Mt = molar Fe3+/[2Ti+ Cr+Al+Fe3+]). Several larger grains in the groundmass are zoned, with rims as Ti-rich as Chr29Sp7Usp48Mt16. Fe-Ti oxide pairs in the groundmass consist of titanomagnetite of composition Chr2–3Sp3–5Usp64–74Mt21–28 and ilmenite with 3–8% hematite (Fe2O3), 8–14% geikielite (MgTiO3), and approximately 1.5% pyrophanite (MnTiO3) components. The abundance of the magnetite component increases with increasing ulvöspinel, as shown in Fig. 6, a plot of Fe3+/(Fe3++2Ti+Cr+Al) [atomic%] versus 2Ti/(2Ti+Cr+Al) [atom%]. Also notable is a gap in ulvöspinel content and a concomitant jump in magnetite content.

Figure 6.

Magnetite versus ulvöspinel content for spinels in NWA 6234. Solid symbols are compositions of spinels enclosed in olivine; open symbols are from grains in the groundmass.


Two phosphates (merrillite and apatite) are present in the groundmass of NWA 6234, with merrillite being more abundant than apatite. Both phosphates occur as intergrowths with each other. The grain size ranges from 10 to 100 μm in length (Fig. 7). They are in textural equilibrium with pyroxene and plagioclase. Typical analyses of merrillite and apatite are presented in Table 4. The Mg# of merrillite ranges from 80.6 to 69.3. Na2O content increases slightly with decreasing Mg# from 1 to 1.5 wt% (Fig. 8). Apatite is halogen-rich, with up to 1.9 wt% F and 3.4 wt% Cl.

Table 4. Representative microprobe analyses of merrillite and apatite
  1. na = not analyzed; bd = below detection limit.

Oxide wt% Oxide wt%
Normalized to 56 oxygens Normalized to 8 cation
Mn0.0290.0280.030Ce 0.0010.0010.001
Cr0.0000.0060.000sum A5.0715.0465.0215.032
P14.03113.78013.788S 0.0010.0000.001
Si0.0720.0900.078sum B2.9262.9512.9742.965
Cation total35.34935.74435.608F0.5030.4200.3230.165
    sum X1.0001.0001.0001.000
    Cation total7.9967.9977.9947.997
Figure 7.

Schematic sketch of the intergrowth relationship between merrillite and apatite: merrillite is the primary magmatic phase that is replaced by apatite through the interaction with a Cl-F-OH-rich fluid derived from the crystallized NWA 6234 magma (after Shearer et al. 2011). ap = apatite; mer = merrillite; pyx = pyroxene; plag = plagioclase/maskelynite.

Figure 8.

Na (afu = atoms per formula unit) content versus Mg# of merrillite in NWA 6234 compared with other Martian meteorites (after Burger et al. 2012). a) The high Na content in merrillite in NWA 6234 suggests that plagioclase began crystallizing at approximately the same time or shortly after merrillite. b) The Na content in NWA 6234 slightly increases with decreasing Mg# in merrillite consistent with plagioclase cocrystallization. For further explanations, see text.

Water content of the apatite was determined indirectly from our EMP analyses of F and Cl, and the assumption of stoichiometry that F+Cl+OH = 1.00 structural formula unit (sfu) (see McCubbin et al. 2011). This calculation implies that the apatite has up to 0.6sfu OH component (i.e., hydroxyl-apatite). This inferred OH content represents a maximum value, in that it does not consider unmeasured X-site components like S2−, O2−, carbonate, and vacancies. However, it could also represent a minimum value as F and Cl abundances are often overestimated in EMP analyses (Goldoff et al. 2012). As outlined in the Methods section, we used a special routine to analyze apatite accurately, and we are therefore confident that F and Cl are not overstated in our analyses. Within the Cl–F–OH ternary plot, apatite compositions plot toward the OH component (Fig. 9).

Figure 9.

Ternary plots of apatite X-site occupancy (mol%) from NWA 6234 using electron probe microanalysis. For EPMA data, OH was calculated by stoichiometry based on 13 anions and assuming 1−F−Cl = OH.

Other Phases

Plagioclase occurs in the groundmass and ranges in composition from An62–51 with up to 0.37 wt% K2O (Or2). It has been completely transformed by shock to maskelynite (Irving et al. 2011). Iron sulfide is present in the groundmass; its S content ranges from 37.5 element-wt% to 38.8 element-wt%.

Melt Vein

The melt vein in NWA 6234 cuts as a straight line through the rock and ranges from 10 μm to about 100 μm in thickness. The vein contains melt with schlieren, and fragments of crystals from the surrounding rock. Some of these crystals show rounded edges; others have clearly not reacted with the melt and have sharp edges. In some areas, the melt vein cuts through olivine and part of the crystal can be found in the melt (Fig. 10). Although there are schlieren visible in the vein, the vein's chemical composition is relatively homogeneous, and almost identical to the bulk composition of the meteorite (Table 5) (Filiberto et al. 2012). The FeO content of the vein is slightly higher than in the bulk rock (23.8 wt% versus 21.3 wt%) and it is richer in TiO2 by approximately 0.3 wt%. A detailed investigation of the melt vein texture and mineralogy will be presented elsewhere; however, we note that it is similar to the vein reported by Walton et al. (2012) in NWA 4797.

Table 5. Composition of the bulk rock of NWA 6234 compared with the composition of the melt vein
 NWA 6234 bulk rock compositionaMelt vein (this study)
  1. a

    Filiberto et al. (2012).

Oxide wt%  
Figure 10.

BSE image of the melt vein cross cutting NWA 6234. Note the schlieren and fragments of crystals in the vein. Ol = olivine; Pyx = pyroxene; Mask = maskelynite.

Does NWA 6234 represent a magma composition?

For any igneous rock, it is important to know whether its bulk composition is that of a pure magma, or whether it includes excess components such as cumulus crystals, xenocrysts, and/or assimilated material. This question is particularly important for the Martian basalts, as magma compositions are required to derive constraints on Martian mantle compositions and magma generation (e.g., Musselwhite et al. 2006; Filiberto et al. 2010a; Filiberto and Dasgupta, 2011; Gross et al. 2011). However, many Martian basalts are inferred to contain cumulus or xenocrystic material (e.g., Stolper and McSween 1979; McSween 1994, 2002; Papike et al. 2009; Filiberto and Dasgupta 2011).

To determine whether NWA 6234 represents a true magma composition, or contains cumulate grains or xenocrystic material, the chemical compositions of the most magnesian crystals (in this case olivine) in the rock were analyzed. These crystals were the first to crystallize and have been in equilibrium with a melt of initial magma composition. If that initial magma composition is equal to the measured bulk rock composition, this bulk rock composition would represent a magma composition (e.g., Shearer et al. 2008; Filiberto et al. 2009; Gross et al. 2011). By comparing the chemistry of olivine with the bulk rock composition, one can thus determine if the olivine represents true phenocrysts. In Fig. 11, the Mg# of the olivine cores are compared with the bulk rock Mg#. The blue dotted line (from Filiberto et al. 2012) represents the experimentally constrained Fe-Mg equilibrium between olivine and basaltic magma, and reflects an equilibrium KDFe/Mg Olivine/Melt of 0.355 ± 0.01 as determined from experiments on Martian basaltic compositions (both surface rocks and meteorites) by Filiberto and Dasgupta (2011). The NWA 6234 olivine cores are within uncertainty of the calculated Mg# equilibrium of the melt's bulk composition from which they crystallized (Fig. 11). Hence, it is plausible that NWA 6234 represents a melt composition.

Figure 11.

Mg# (molar Mg/[Mg+Fe] × 100) of bulk rock versus Mg# in olivine cores, for selected Martian meteorites (after Papike et al. 2009). Blue dotted line represents the calculated Mg# of melt in equilibrium with olivine cores (Filiberto et al. 2012); blue fine dashed lines represent the error of the calculated Mg# of melt.

We can further test whether NWA 6234 represents a primary mantle or fractionated melt using the inverse experimental approach—i.e., determine if a melt of its bulk composition would be cosaturated with expected mantle minerals (olivine, orthopyroxene, spinel, etc.) of likely chemical compositions at a reasonable mantle pressure and temperature (e.g., Asimow and Longhi 2004). Numerous experimental studies have used this technique to constrain the pressures and temperatures of basalt formation, and mantle potential temperatures for Mars (e.g., Musselwhite et al. 2006; Monders et al. 2007; Filiberto et al. 2008 2010a 2010b), the Moon (e.g., Grove and Vaniman 1978; Delano 1979, 1980; Elkins-Tanton et al. 2003; Draper et al. 2006), and the Earth (e.g., Basaltic Volcanism Study Project, 1981). However, experiments have not yet been conducted on the NWA 6234 composition.

We calculate the temperature and pressure of formation of NWA 6234 from olivine-melt Mg-exchange thermometry (Putirka 2005; Lee et al. 2009) and melt silica activity barometry (Albarede 1992; Lee et al. 2009). We chose this geothermobarometer (Putirka 2005; Lee et al. 2009) because it has been shown to reproduce experimental results for Martian basalts within uncertainty (Filiberto and Dasgupta 2011). We have chosen a Fe-Mg Ol-Melt KD of 0.355 (Filiberto and Dasgupta 2011) and a mantle composition of Mg# = 80 (Agee and Draper 2004). Using these values, the NWA 6234 bulk composition is in equilibrium with a Martian mantle of Fo80, and thus most likely represent a primitive, mantle-derived melt. The calculated average equilibration pressure (with olivine + orthopyroxene bearing mantle) is 2.7 GPa with a temperature of approximately 1600 °C. This is significantly higher pressure than any other estimates for Martian basalt formation (<2 GPa and <1550 °C; e.g., Musselwhite et al. 2006; Monders et al. 2007; Filiberto et al. 2008; Lee et al. 2009; Filiberto et al. 2010a, 2010b; Filiberto and Dasgupta 2011; Gross et al. 2011). These results suggest that NWA 6234 was derived from a unique source region deeper within Mars than any other Martian basalt. Experiments are needed on the NWA 6234 composition to fully constrain the temperatures and pressure of formation of this basalt and the mantle potential temperature of its source region.

Crystallization history

The compositions and zoning patterns in minerals of NWA 6234 preserve a record of its igneous crystallization history. In the next paragraphs, we place constraints on the formation conditions during crystallization and the crystallization sequence of NWA 6234 based on the mineral chemistry, petrography, and textural features and relations.

Crystallization Sequence

Fe/Mg Ratios

The Fe/Mg ratios of olivines and pyroxenes in NWA 6234 can indicate which portions of these minerals crystallized simultaneously. The compositional relationships between olivine and pyroxene are shown in Fig. 5. The cores of orthopyroxene grains have a composition of En70Fs26Wo4, and olivine cores are as magnesian as Fo78. The observed KDMg/FeOl/low-Ca-px for this pair (=0.73) is inconsistent with equilibrium values from shergottite melts, where KDMg/FeOl/low-Ca-px = 1.2 (Longhi and Pan 1989). Thus, the most magnesian olivine in NWA 6234 is too magnesian to have formed in equilibrium with the most magnesian orthopyroxene. Based on a KDMg/FeOl/low-Ca-px of 1.2, olivine of composition Fo67 would be in equilibrium with the most magnesian orthopyroxene. Therefore, the cores of pyroxene grains must have crystallized after the first olivine grains. The mantle compositions of the largest olivines (and core composition of slightly smaller olivine grains) are consistent with Mg-Fe-equilibrium with the early pyroxene core composition and probably crystallized at approximately the same time. The same relationship holds for the early pyroxene rims, the early Ca-rich augites, and the olivine rims. The smallest most Fe-rich olivine grains in the groundmass are, based on their KD, in equilibrium with the latest Ca-rich augites and the latest low-Ca pyroxene rims.


Major and minor elements in merrillite, especially Mg, Fe, and Na, can be used to place constraints on the crystallization sequence of plagioclase and merrillite. High Na contents of merrillite in Martian meteorites are thought to imply that it crystallized before plagioclase (Burger et al. 2012). In Fig. 8a, the Na content of merrillite in NWA 6234 is compared with those of other Martian meteorites. In NWA 6234, merrillite has an Na2O content > approximately 1.1 wt%, which suggests that plagioclase began crystallizing after merrillite. However, plagioclase has a higher Ca/Na ratio than the liquid from which it forms, and crystallization of plagioclase should slightly enrich the residual magma in Na. Thus, the later-formed merrillite should also be slightly richer in Na, which is consistent with the slight increase in Na content with decreasing Mg# in merrillite (Fig. 8b).

The textural relationships, intergrowth between merrillite and apatite, and phosphorus content of the melt yield insights into the igneous crystallization sequence of the phosphates themselves (Jolliff et al. 1993; McCubbin et al. 2011). The bulk composition of NWA 6234 is rich in phosphorus (Filiberto et al. 2012) and fractional crystallization would have raised the activities of P2O5 until merrillite became stable. Further fractional crystallization will increase the volatile (F2, Cl2, H2O) activity in the melt while the P2O5 fugacity is buffered by merrillite crystallization (Patiño Douce and Roden 2006). At some point, the melt became saturated in apatite, at which time it cocrystallized with merrillite. The textural relationship of apatite and merrillite is consistent with this interpretation. Thus, the intergrowth of merrillite and apatite in NWA 6234 can be interpreted to represent a chemical reaction between magmatic merrillite and an OH-Cl-rich melt. This interpretation is not uncommon in Martian meteorites (Greenwood et al. 2003; Shearer et al. 2011). A detailed investigation of the volatile fugacity ratios and pre-eruptive volatile history of NWA 6234 is presented elsewhere (Gross et al. 2012; Gross et al., personal communication).

Phosphorus Zoning in Olivine

Olivine is the liquidus phase in NWA 6234, the first phase to crystallize from the magma. Therefore, its chemical composition can potentially record the evolution of mantle-derived magmas indicating magmatic conditions and timescales of crystallization. Recent studies have shown that phosphorus zonation is common in igneous olivine in many types of igneous rocks (komatiites; basalts; andesites; dacites from Earth, Moon, Mars; meteorite suites; e.g., Boesenberg et al. 2004; Milman-Barris et al. 2008; Stolper et al. 2009; Qian et al. 2010; Spandler and O'Neill 2010; Peslier et al. 2010). Due to its incompatible nature, the distribution of phosphorus in olivine can record the kinetic, dynamic, and geochemical states of the magmatic system (Boesenberg et al. 2004; Milman-Barris et al. 2008; Stolper et al. 2009; Qian et al. 2010; Spandler and O'Neill 2010) and also may preserve a record of the crystal growth rate variations because it diffuses more slowly through olivine than other elements. Dynamic crystallization experiments on basaltic compositions showed that the phosphorus variation in olivine is not caused by melt composition variation resulting from magma mixing or transport of olivine as xenocrysts (Buseck and Clark 1984; Peslier et al. 2010). The most likely cause of the phosphorus oscillatory zoning is a variation in the crystal growth rate of olivine (Milman-Barris et al. 2008). High phosphorus zones record rapid olivine growth, the olivine component being actively depleted from the melt. If the phosphorus diffusion rate in the melt was equal to or slower than the growth rate of the olivine, then phosphorus is preferentially incorporated into the olivine over the melt (Peslier et al. 2008). Thus, areas of the olivine with low phosphorus are the result of slow growth as the boundary layer between olivine and melt is temporarily depleted in olivine component. This may reflect equilibrium crystallization conditions in which phosphorus behaves as an incompatible element and goes preferentially into the melt.

The high phosphorus zoning in the olivine cores in NWA 6234 reflects initial undercooling followed by nucleation and a pulse of rapid crystal growth. The large mantle area that shows low phosphorus concentration (Fig. 4) reflects the slow equilibrium growth of the olivine with the melt. During crystallization toward the end, alternating slow and fast growth rates must have occurred to explain the oscillatory zoning in phosphorus. This change in growth rate seems to coincide with the change in fO2 from iron-wüstite (IW) + 1, at the time the olivine were in equilibrium with the earliest pyroxene, to the time where the groundmass crystallized (approximately IW +4 to +5); see the 'Temperature and Oxidation State' section below. High-temperature annealing must have occurred for long enough to let other minor elements such as Ca, Mn, Al, Cr, and Ni homogenize via diffusion (Clark et al. 1986; Milman-Barris et al. 2008).


Minor elements in pyroxenes, especially Al and Ti, can be used to place constraints on the pressure conditions at which the mineral crystallized. The Al/Ti ratios for pyroxenes from the groundmass and those from the mesostasis are compared in Fig.  12 with the model of Nekvasil et al. (2004, 2007) modified for Martian compositions (Filiberto et al. 2012). This model illustrates the pressure dependence of the Al/Ti ratio for pyroxenes in equilibrium with a basaltic magma and is calibrated from experiments on a suite of terrestrial alkalic basalts. Nekvasil et al. (2004) showed that pyroxenes crystallized from basaltic melts had higher Al/Ti ratios when grown at high pressures. In crystal-chemical terms, this means that high-pressure pyroxenes have greater proportions of octahedral Al in the Tschermak's component molecules MgAl(AlSi)O6 and CaAl(AlSi)O6. This result has been confirmed for Martian magma compositions, after correction for the different Al/Ti ratio of the parental magma (Filiberto et al. 2012), and is consistent with experimental results on model chemical systems (e.g., Gasparik 2000).

Figure 12.

Ti versus Al (afu = atoms per formula unit) for a) pyroxenes in NWA 6234 (core = orange circles; mantle-rim = blue circles, second-generation Ca-rich pyroxenes = green squares); and b) pyroxenes in NWA 5789 (core = dark blue circles; rim = light blue circles, mesostasis = green circles); in Y98 (experimental pyroxene; white circles); in NWA 1068 (experimental pyroxene, small red circles). Also, shown for comparison are simplified model depths of formation (modified from Nekvasil et al. 2004).

As this model is not calibrated for the exact bulk rock composition of NWA 6234, it will only yield approximate pressures of crystallization. The pyroxenes in NWA 6234 have a constant Al/Ti ratio (Fig. 12a) from core to rim to Ca-rich augites, suggesting that they formed at constant pressure, consistent with upper crustal conditions (i.e., near the surface of Mars). Thus, one can infer that the parent magma of NWA 6234 crystallized completely at shallow depth inside Mars and never erupted. This history is unlike those of NWA 5789 and Yamato-980459 (Y98 hereafter), which both crystallized deeper within the crust and then erupted onto the Martian surface (Fig. 12b). These differing histories are consistent with the petrographies of the meteorites: NWA 6234 has a fine, seriate grain size distribution, whereas NWA 5789 and Y98 show a strong bimodal grain size distribution (Usui et al. 2008; Gross et al. 2011).

Temperature and Oxidation State

The temperature and redox conditions of crystallization of NWA 6234 can be calculated via mineral thermometers and oxybarometers, particularly from appropriate mineral assemblages of olivine-pyroxene-spinel (Ol-Px-Sp), and from the ferric iron contents of early- and late-formed oxides. The most magnesian olivine in NWA 6234 crystallized prior to the most magnesian orthopyroxene, based on Mg-Fe exchange KD; therefore, we can only study the redox conditions once orthopyroxene started to crystallize, assuming no subsolidus equilibration. We used the composition of the most magnesian orthopyroxene we analyzed, Wo2.4En68.9Fs28.7, olivine in Fe/Mg equilibrium with orthopyroxene (Fo67), and chromite compositions. We selected four compositions from the chromite grains enclosed in olivine, favoring those with high Cr# (molar Cr/[Cr+Al]) and low Fe# (molar Fe/[Fe+Mg]), following the criteria of Goodrich et al. (2003). Oxygen fugacity and temperature were calculated from the Ol-Opx-Sp oxybarometer, implemented in the online calculator on the CT Server (; this calculator is based on the thermodynamic models of Sack and Ghiorso (1989 1991a 1994a 1994c). Calculations were made for both 1 bar and 10 kbar pressure, and results are shown in Table 3. Temperatures from olivine-chromite thermometry (Sack and Ghiorso 1991a) range from 900 to 1185 °C. The highest temperature comes from the lowest Fe# chromite (Table 3). The lowest temperatures are given by the highest-Fe# chromites, which suggest subsolidus Fe-Mg exchange with olivine hosts. For both sets of chromites, olivine-spinel oxybarometry gives oxygen fugacities approximately 1 log unit above the iron-wüstite (IW) buffer. Pressure has no effect within uncertainties (Table 6), and subsolidus equilibration has apparently had little effect on the oxygen fugacity recorded by the chromites.

Table 6. Summary of oxygen fugacity estimates for NWA 6234 early phases using the Ol-Px-Sp oxybarometer
 P(kbar)T(°C)aLog10fO2Log10fO2 (IWb)Log10fO2 (QFMc)Fe#
  1. All calculations involved Fo67 olivine, and Wo2.4En68.9Fs28.7 orthopyroxene. Uncertainties on the average are 2 sigma standard deviation of the mean.

  2. a

    Temperature calculated from the CT Server Ol-Px-Sp oxybarometer, which uses the olivine-spinel geothermometer of Ghiorso and Sack (1991).

  3. b

    Relative to the Iron-Wüstite (IW) buffer as defined by Herd (2008) using the data of O'Neill and Pownceby (1993).

  4. c

    Relative to the Quartz-Fayalite-Magnetite buffer as defined by Wones and Gilbert (1969).

Average0.001940 ± 114 0.9 ± 0.2−2.8 ± 0.30.847 ± 0.032
Average101014 ± 120 1.2 ± 0.2−2.5 ± 0.30.847 ± 0.032

Ol-Opx-Sp thermombarometry on the groundmass minerals yields significantly lower temperatures and higher oxygen fugacities. Regardless of whether spinels on the lower Ti, Fe3+ or higher Ti, Fe3+ side of the gap were used, the oxygen fugacity results are the same, approximately IW +5. Temperatures obtained from spinels on the low-Ti, Fe3+ side of the gap yield higher temperatures (approximately 870 versus approximately 720 °C), suggesting that these spinels are closer to equilibrium with the olivine and low-Ca pyroxene than the higher Ti, Fe3+ spinels. The choice of pyroxene–olivine pair does not significantly affect estimates. Our best estimate for this stage of crystallization is fO2 = IW +4.8 ± 0.1, T = 873 ± 64 °C (n = 7, 2σ deviation of the mean).

The most Ti, Fe3+-rich titanomagnetite and ilmenite are present in the groundmass. The application of the Fe-Ti oxide oxybarometer of Ghiorso and Evans (2008) to these pairs yields variable results, from as low as IW +2.3 to IW +4.1. This variability may be attributed to disequilibrium between oxides, likely as a result of subsolidus re-equilibration. Selection of the most Fe3+-rich ilmenite with a Fe3+-rich titanomagnetite (from above the gap in Fig. 6) yields IW +4.1 and T = 1040 °C. Application of the Ca-QUIlF model (Andersen et al. 1993) to the same pair yields IW +4.1 and = 843 °C (Table 3), in agreement with the results from the Ghiorso-Evans model; a good indication that these oxide compositions are in equilibrium. Inclusion of this same oxide pair with Fe-rich olivine and low-Ca pyroxene (as above) yields IW +4.9 and T = 781 °C using Ca-QUIlF, and although the temperature is lower, it is within uncertainty of the two thermodynamic models. This suggests that all of these silicate and oxide compositions were in equilibrium during the final stages of crystallization.

The results from late-stage oxides and oxide+silicate phases suggest that oxygen fugacity increased approximately 4 log units between crystallization of the earliest olivine (IW +1) and the groundmass (IW +5). A similar increase is observed in NWA 1068/1110, from IW +1 to IW +4.5 (Herd 2006). Although the most Fe3+-rich titanomagnetites in NWA 6234 are not as enriched in magnetite (Fe3+/(Fe3++2Ti+Cr+Al) to 42%; Fig. 6) as those in NWA 1068/1110 (Fe3+/[Fe3++2Ti+Cr+Al] up to 66%), the latest oxides in NWA 1068/1110 and NWA 6234 (i.e., with 2Ti/[2Ti+Cr+Al] = 80–90%) do show some overlap, and agree in oxygen fugacity within uncertainties (IW +4.8 ± 0.3 for NWA 1068/1110 using Ca-QUIlF; Herd 2006 compared with IW +4.8 ± 0.1 for NWA 6234 using the same method). NWA 6234 crystallized in a closed system with respect to volatile abundance (see the 'Volatile History of NWA 6234 during Late Crystallization' section) and the increase in oxygen fugacity between early and late phases in this closed system may be explained by buildup of ferric iron as crystallization proceeds, similar to that observed in the LAR 06319 olivine-phyric shergottite (Peslier et al. 2010).

Volatile History of NWA 6234 during Late Crystallization

Despite the importance of water to Martian geology and its potential in affecting the course of igneous petrogenesis, Martian meteorites are rather dry: approximately 50–150 ppm H2O (Leshin et al. 1996; Leshin 2000; Leshin and Vicenzi 2006). However, they could have degassed upon eruption and thus no longer record the original volatile concentrations in their parental, pre-eruption magmas. There is significant disagreement on the pre-eruption volatile contents of Martian basaltic magmas, with estimates ranging from nearly anhydrous (based on the bulk concentrations, experimental petrology, and mineral chemistry), through nearly 2 wt% H2O (based on experimental petrology, crystallization temperatures, and mineral chemistry); see McCoy et al. (2011) and references therein.

Apatite records the volatile contents (OH, F, Cl) of its parental magma and thus has been used in the past to evaluate the pre-eruptive volatile concentrations of terrestrial, Martian, and lunar magmas (Stormer and Carmichael 1971; Westrich 1982; Mathez and Webster 2005; Patiño Douce and Roden 2006; Filiberto and Treiman 2009; Boyce et al. 2010; McCubbin et al. 2010, 2011, 2012; Patiño Douce et al. 2011). In the absence of postcrystallization secondary processes that can affect apatite volatile contents, apatite compositions can be used to determine the magmatic volatile abundances at the time of apatite crystallization. However, the volatile abundances of the parental liquid (and from these, volatile abundances in the magmatic source region) are difficult to constrain from apatite analyses due to the possibility of secondary magmatic processes that could disturb the parental magmatic volatile abundances prior to apatite crystallization (i.e., precrystallization secondary processes). The precrystallization secondary processes can include (but are not limited to) degassing/fluid loss from the melt and/or assimilation/mixing with other lithologic components/fluids (McCubbin et al. 2011). The effect of shock on volatile gain or loss has not been studied for Martian apatites. However, Ostertag et al. (1985) studied terrestrial apatites from Haughton Impact crater and showed that their chemistry is not affected by the impact and shock process. The solubility of H species in silicate melts is strongly pressure-dependent and any melt subjected to low pressures (including those at the Martian surface or in shallow magma chambers) will have undergone magmatic degassing even at very low hydrogen concentrations (as summarized by Burnham 1994; McMillan 1994).

NWA 6234 crystallized under upper crustal pressure conditions and never erupted onto the surface, thus a degassing/fluid loss event seems unlikely under these conditions. Furthermore, degassing should cause the melt to lose significant proportions of its fluid-soluble elements (notably H, Cl, and Li), but apatite in NWA 6234 is rich in H (up to 60% hydroxyl component), which is inconsistent with significant degassing. Similarly, the variation and abundance of Li-isotopes in NWA 6234 suggest that the rock was never affected by degassing or interaction with a fluid (Filiberto et al. 2012). In addition, NWA 6234 shows no evidence of late aqueous fluids that might have affected the composition of apatite; the meteorite contains no hydrous silicate phases (e.g., serpentine, smectite) such as might form during aqueous alteration. Thus, it seems reasonable to infer that NWA 6234 crystallized in a closed system with respect to volatile abundance, and the apatite composition (rich in OH) reflects that of a late-stage melt, which in turn reflects the proportions and composition of volatile species in the parental mantle-generated melt. With the OH-F contents of the apatites in NWA 6234 being among the highest values measured for any Martian meteorite, it is reasonable to infer that the actual content of fluorine, chlorine, and water of the Martian mantle, parental to NWA 6234, may be higher than previously thought estimates based on other Martian basalts. A detailed investigation of the volatile fugacity ratios and pre-eruptive volatile history of NWA 6234 is presented elsewhere (Gross et al. 2012; Gross et al., personal communication).


The Martian meteorite NWA 6234 is an olivine-phyric shergottite with an unusual texture compared with other olivine-phyric shergottites. It is relatively unaltered, fine-grained, and composed of a seriate olivine texture set in an even finer grained groundmass of pyroxene, maskelynite, ferroan olivine, spinel, ilmenite, merrillite, apatite, and accessory Fe-sulfide. The olivine core compositions with Fo78 are in equilibrium with the bulk rock Mg# of 59. Olivine-melt Mg-exchange thermometry, and silica activity in the melt barometry calculations show that NWA 6234 not only represents a melt composition but is a primitive melt derived from an approximately Fo80 mantle. In this sense, it is similar to the other olivine-phyric shergottites NWA 5789 and Y98, which also represent magma compositions. However, the calculated pressures of mantle melt formation for NWA 6234 may be higher than previous estimates for Y98, NWA 5789, and surface basalts.

Based on textural relationships, minor element mineral chemistry, and zoning patterns NWA 6234 crystallized fully within the lower crust at shallow depth, and never erupted onto the surface. Thus, the volatile content of the apatite composition most likely represents a true snapshot of the primitive original volatile content of the melt from which the meteorite crystallized. The Cl/F/OH ratios in apatite suggest that the H2O content of the melt in the upper crust is higher than previously thought. These results, the higher pressure of the mantle melt formation, and the ferroan nature of its composition, suggest that NWA 6234 may represent a magma from a unique, previously untapped source region (at least in terms of olivine-phyric shergottites) with unaltered volatile composition.


We are grateful to A. Peslier for assistance with the EMP analyses at NASA JSC, and C. Bryden for help with oxygen fugacity calculations. We thank Drs. L. Hallis, S. Symes, and M. McCanta for helpful reviews and comments on this manuscript, as well as editor C. Floss for handling this manuscript and helpful thoughts and comments. This study was supported by NASA MFR grant NNX09AL25G, Natural Sciences and Engineering Research Council of Canada Discovery Grant 261740-08, and an SIU-C startup package.

Editorial Handling

Dr. Christine Floss