Near-liquidus phase equilibria experiments have been conducted on a synthetic Fastball basalt composition, as analyzed at Home Plate plateau of Mars (Gusev Crater), to test if it represents a primitive mantle derived melt and place constraints on the temperature of the ancient mantle and on the lithosphere-asthenosphere boundary of Mars. The Fastball basalt is multiply saturated with olivine and orthopyroxene at ∼1.2 GPa and 1430°C. Based on melting models, we predict that the Fastball composition could be produced by 13–23% equilibrium melting of the Martian mantle, with extraction of melt from the base of ∼105 km thick lithosphere. The multiple saturation for Fastball also constrains the potential temperature of the Martian mantle to be approximately 1480–1530°C with an initial melting pressure of 4.0–4.7 GPa. This potential temperature is much lower than that of the terrestrial mantle derived from similarly ancient magmas, i.e., komatiites.
Until recently, the SNC meteorites represented the only source of information about Martian igneous chemistry [McSween, 2002]. This changed with the Mars Exploration Rovers, which have analyzed basalts on the surface of Mars in Gusev Crater [e.g., Squyres et al., 2004]. Compared to the Martian meteorite basalts, the basalts in Gusev Crater are thought to be much older (∼3.65 Ga vs. 1.0–0.17 Ga) [e.g., Jones, 1986; Nyquist et al., 2001; Greeley et al., 2005; Jones, 2007] and have distinctly different chemistries [e.g., McSween et al., 2009]. Because of the differences in basalt chemistry, we can constrain how the chemical variations in the Martian mantle and crust may have changed through time. Previous experimental works on other Martian basalts (e.g., Adirondack-class basalt Humphrey) have tried to determine if they could have been primary mantle derived magmas and thus constrain temperatures of basalt formation in the ancient Martian mantle [Monders et al., 2007; Filiberto et al., 2008]. Other experimental studies have sought to explain the compositional differences between the Mars surface basalts and some of the cumulate SNC meteorites through crystallization and mineral accumulation models [Filiberto et al., 2006; Filiberto, 2008; Nekvasil et al., 2009]. To complement these models, we have experimentally investigated the near-liquidus phase equilibria of the Fastball basalt, analyzed on Mars at the Home Plate site in Gusev Crater. Our ultimate goal is to determine if it might be a primitive mantle derived melt and place further constraints on the temperature of basalt formation in the ancient Martian mantle.
Home Plate Outcrop. Home Plate is a low plateau in the Columbia Hills of Gusev Crater, consisting of a layered sequence of clastic partially altered rocks with alkali basalt compositions [Squyres et al., 2007; Schmidt et al., 2009]. The outcrop consists mostly of basaltic glass, with lesser proportions of pyroxene, olivine, plagioclase, nanophase Fe-oxide, and magnetite (from Mössbauer and Thermal Emission Spectroscopy) [Squyres et al., 2007; Morris et al., 2008]. The Home Plate outcrop is thought to represent a pyroclastic deposit based on its bulk chemistry, mineralogy, stratigraphy, structure, and sedimentology [Squyres et al., 2007; Lewis et al., 2008]. Of the many rocks analyzed at Home Plate, we chose the Fastball composition because 1) it has the highest Mg# (molar Mg/(Mg+Fe)) of all of the Home Plate rocks suggesting it is the most likely to represent a primitive magma and 2) Fastball has relatively low abundances of Cl and SO3, suggesting that it was not altered extensively [Squyres et al., 2007; Schmidt et al., 2008].
2. Experimental Approach
2.1. Inverse Experimental Modeling Approach
Near-liquidus phase relations can show whether or not a basalt could be a primitive mantle derived liquid. If a magma is co-saturated at a single pressure (P) and temperature (T) with expected mantle minerals (olivine + orthopyroxene ± cpx ± plagioclase/spinel/garnet) of appropriate chemical compositions, that magma could reasonably represent a mantle-derived melt [Asimow and Longhi, 2004]. This approach has been previously applied to other Martian basalts, including the meteorites Yamato 980459 [Musselwhite et al., 2006] and NWA 1068 [Filiberto et al., 2010] and to the Adirondack-class basalt Humphrey analyzed on Mars at Gusev Crater [Monders et al., 2007; Filiberto et al., 2008]. Here, we use the same approach to determine whether the Home Plate-class pyroclastic basalts could have been mantle derived liquids.
2.2. Starting Bulk Composition
The starting composition for the experiments (Table 1 of Text S1 of the auxiliary material) is based on the analysis of Squyres et al. . The synthetic starting material was made from reagent grade oxides and carbonates ground together under acetone in an automatic mortar and pestle for 15 minutes to ensure homogeneity. This powdered mix melted in an iron-saturated Pt crucible in a muffle furnace at 1500°C under air, quenched to a glass, and reground for 15 minutes.
2.3. Volatile Content of the Starting Material
The pre-eruptive water content of Fastball and the Home Plate pyroclastic rocks is currently unknown. Home Plate is thought to be a pyroclastic deposit which should imply a volatile-rich magma [Wilson and Head, 1994; Schmincke, 2004]. However, it is expected that a pyroclastic eruption on Mars need not require magma with high volatile contents because of Mars' lower gravity and thinner atmosphere [Wilson and Head, 1994]. Similarly, the Martian meteorite basalts and their parent magmas contained very little water [e.g., Stolper and McSween, 1979; Watson et al., 1994; Filiberto and Treiman, 2009]. Therefore, it is reasonable to assume that Martian magmas, including those that produced the Home Plate outcrop, contained little water and to conduct experiments accordingly. We have run our experiments with very low water (analyses of experimental glasses yield ∼0.05 wt.% H2O) contents. If the Home Plate pyroclastic magmas contained more than 0.05 wt% water before eruption, then the temperature of multiple saturation, derived below, represents a maximum temperature.
3. Experimental and Analytical Methods
3.1. High Pressure Experiments
High pressure-temperature experiments were conducted in an endloaded piston-cylinder apparatus at Rice University, using a 12.7 mm assembly, comprising graphite sample capsules, straight graphite furnaces, BaCO3 sleeves, and crushable MgO spacers. Experimental run products were analyzed using a Cameca SX-100 electron microprobe at NASA, Johnson Space Center for major element abundances of the glass and crystal phases. Volatile-contamination of the experiments were measured by FTIR analysis (see auxiliary material for full details of experimental and analytical techniques).
3.2. Approach to Equilibrium
In order to determine mineral abundances, verify close system, and to evaluate approach to equilibrium in unreversed experiments, mass balance calculations were conducted using the average starting bulk composition (Table 1 of Text S1) and the compositions of the crystallized phases and residual glass (Table 2 of Text S1), using the least square computations of the IgPet software [Carr, 2000]. For all experiments the sum of the least-square residuals was less than 0.5. A close approach of our experiments towards equilibrium can be argued based on measured Fe-Mg KD for olivine-melt and opx-melt pairs, as the values obtained in our study (KDFe*−Mg (ol-melt) = 0.35–0.37 and KDFe*−Mg (opx-melt) = 0.28–0.32) are in perfect agreement with those observed in previous partial melting experiments on terrestrial and Martian mantle compositions (KDFe*−Mg (ol-melt) = 0.26–0.36 and KDFe*−Mg (opx-melt) = 0.28–0.35; [e.g., Bertka and Holloway, 1994b; Kushiro and Mysen, 2002]) Moreover, lower T runs show systematically lower melt fractions, validating expected shifts in phase assemblages with changing T.
Figure 1 shows the near–liquidus phase diagram for the nominally volatile free Fastball composition in a P-T space. The liquidus T increases from 1425–1450°C at 1.1 GPa to 1450–1470°C at 1.5 GPa. Experiments at 1.1 GPa have olivine (Fo77) on the liquidus whereas above 1.3 GPa, opx (En77Wo3) is the liquidus phase. This suggests that the Fastball liquid is multiply saturated with both olivine and orthopyroxene at ∼1.2 GPa and 1430°C.
If Fastball composition represents a primitive magma, unmodified since its formation in the mantle, then the compositions of olivine and pyroxene that crystallize from it at the multiple saturation condition should be those similar to the Martian mantle mineralogy. The compositions of olivine and orthopyroxene near the multiple-saturation point in our experiments, are close to those found for Dreibus and Wänke's  model mantle composition at similar P and T (Figure 2) [Bertka and Holloway, 1994b]. Further, the experiments of Bertka and Holloway [1994b] at similar pressures, temperatures, and melt percentages, contain olivine and orthopyrxene only, clinopyroxene melts out within 20°C of the solidus, which is consistent with our experiments having only olivine and orthopyroxene at the liquidus. Thus, the experimental results are consistent with the Fastball composition being a primitive mantle melt with average melting conditions of ∼1.2 GPa and 1430°C. This corresponds to a depth of ∼105 km in the Martian mantle.
If Fastball is derived from a Martian mantle similar to that proposed by Dreibus and Wänke , we can use the phase equilibrium melting experiments of Bertka and Holloway [1994b] to estimate the proportion of melting involved in forming the Fastball magma. (Figure 3, insert). This yields a melt fraction of 15–30% for Fastball, if it is produced at a nominal pressure of 1.5 GPa. To further refine the melt fraction estimate for Fastball, we used the constraint from Ti partitioning during mantle melting (Figure 3). For this calculation, we assume: (1) that Fastball is a mantle-derived primary magma, (2) that Fastball is a batch or aggregate partial melt, (3) that its parental mantle has Dreibus and Wänke's  composition (0.14 wt% TiO2), and (4) a bulk partition coefficient, DTimantle/melt, of 0.03–0.07 (based on experimentally determined mantle mineralogy [Bertka and Holloway, 1994b]). This calculation suggests that Fastball formed as a partial melt of 13–17 wt.% from a Dreibus-Wänke mantle.
We have also performed mass balance calculations for all major elements using Dreibus and Wänke's  and Agee and Draper's  bulk Martian mantle compositions, the olivine and orthopyroxene near the multiple saturation pressure from our experiments, and the Fastball bulk composition. The calculations produce close convergence and yield: 50 wt% olivine, 27 wt% orthopyroxene, and 23 wt% melt of the Fastball composition for Dreibus and Wänke's  mantle composition, and 34 wt% olivine, 50 wt% orthopyroxene, and 16 wt% melt of the Fastball composition for an Agee and Draper  mantle Based on the range from the three approaches, we suggest that Fastball formed as a 13–23% melt of the Martian mantle. The multiple saturation temperature of 1430°C is about 180°C above the 1.2 GPa solidus of Bertka and Holloway [1994b], which is also consistent with Fastball having a high partial melt fraction.
So, if Fastball represents an average, aggregate melt and if we assume a melt production rate of 10%/1 GPa as estimated for mid-oceanic ridge basalt generation on Earth [Langmuir et al., 1992], we would expect one-dimensional melting interval to be 3.2–2.7 GPa, i.e., 270–230 km long decompression-melting column. If the multiple saturation point determined in this study roughly coincides with the mean pressure of melting column, then this predicts the top of the melting column to extend to negative pressure. This is inconsistent with the estimated crustal thickness of Mars and rules out any presence of mantle lithosphere. Hence we predict that melt production in the Martian mantle, at least for Fastball, was not a fractional processes. However, melt should separate from the mantle source region and ascend to the surface well before 13 wt% melt fraction, therefore, melt production was most likely an incremental process but equilibrium between migrating melts and residual solids appears to have been maintained until a depth of ∼105 km, at the base of lithosphere.
The olivine + opx + melt multiple saturation for Fastball, at ∼1430°C and ∼1.2 GPa, also constrains the potential temperature of the Martian mantle and initial melting pressure. We correct the MSP to zero pressure using a mantle adiabatic gradient of 0.18°C/km [Kiefer, 2003]. We correct for the effect of latent heat of melting on the temperature using the expression: ΔTfus = F(Hfus/Cp) where F is the melt fraction, Hfus (6.4 × 105 JK −1kg−1 [Kiefer, 2003]) is the heat of fusion and Cp (1200 JK−1kg−1 [Kiefer, 2003]) is the heat capacity (see Langmuir et al. [1992 and Putirka  for a full description of the method). We calculate an initial melting pressure by extrapolating the mantle potential temperature to the mantle solidus [Bertka and Holloway, 1994a] using the adiabat gradient (0.18°C/km). For 13–23% partial melting, the Fastball results imply a potential temperature of 1480–1530°C and initial melting depth of 4.0–4.7 GPa (325–380 km). This exceeds the mantle potential temperatures estimated for present-day terrestrial mid-ocean ridges (1315–1450°C [McKenzie et al., 2005; Putirka, 2005; Courtier et al., 2007]) and is similar to estimates for modern terrestrial oceanic island basalts (1450–1600°C [Putirka, 2005; Courtier et al., 2007; Herzberg and Asimow, 2008]). However, Gusev basalts are 3.65 b.y. old [Greeley et al., 2005] and if we compare our estimate of ancient Martian mantle potential temperature, as derived from Fastball, with the terrestrial mantle potential temperatures based on archean komatiites (≥1700°C [Lee et al., 2010]), we observe that the Martian mantle was actually colder than the terrestrial mantle at the same point of time. Colder mantle temperature of Mars compared to the Earth's mantle of similar age is consistent with cooling of a planetary body of a smaller size and may be indicative of efficient, convective-cooling of an ancient Mars.
We are grateful to P. Luffi for help with the FTIR and L. Le for help with the microprobe. We thank two anonymous reviewers for helpful comments and J. Longhi for discussion. This work was supported by NASA MFR grant NNX09AL25G to A.H.T. and J.F., MFR grant NNX09AL23G to W.S.K., and Rice University start-up grant to R.D. LPI contribution 1560.