Linking the Chassigny Meteorite and the Martian Surface Rock Backstay: Insights Into Igneous Crustal Differentiation Processes on Mars

Meteoritics & Planetary Science 44, Nr 6, 853–869 (2009) Abstract available online at http://meteoritics.org

Linking the Chassigny meteorite and the Martian surface rock Backstay: Insights into igneous crustal differentiation processes on Mars

Hanna NEKVASIL*, Francis M. MCCUBBIN, Andrea HARRINGTON, Stephen ELARDO, and Donald H. LINDSLEY

Department of Geosciences, Stony Brook University, Stony Brook, New York 11794–2100, USA *Corresponding author. E-mail: [email protected] (Received 05 August 2008; revision accepted 18 March 2009)

Abstract–In order to use igneous surface lithologies to constrain Martian mantle characteristics, secondary processes that lead to compositional modification of primary mantle melts must be considered. Crystal fractionation of a mantle-derived magma at the base of the crust followed by separation and ascent of residual liquids to the surface is common in continental hotspot regions on Earth. The possibility that this process also takes place on Mars was investigated by experimentally determining whether a surface rock, specifically the hawaiite Backstay analyzed by the MER Spirit could produce a known cumulate lithology with a deep origin (namely the assemblages of the Chassigny meteorite) if trapped at the base of the Martian crust. Both the major cumulus and melt inclusion mineral assemblages of the Chassigny meteorite were produced experimentally by a liquid of Backstay composition within the pressure range 9.3 to 6.8 kbar with bulk water contents between 1.5 and 2.6 wt%. Experiments at 4.3 and 2.8 kbar did not produce the requisite assemblages. This agreement suggests that just as on Earth, Martian mantle-derived melts may rise to the surface or remain trapped at the base of the crust, fractionate, and lose their residual liquids. Efficient removal of these residual liquids at depth would yield a deep low-silica cumulate layer for higher magmatic water content; at lower magmatic water content this cumulate layer would be basaltic with shergottitic affinity.

INTRODUCTION to reach the surface or near-surface without compositional modification, they must have retained all minerals Igneous rocks on the Martian surface contain invaluable crystallizing during any protracted ponding stage. Secondary information on the compositional and thermal structure of igneous processes such as fractionation, in which residual the planet and on the changes in these over time. They provide melt separates from the crystallizing minerals, can induce information on the nature of the volatiles transported from the major changes in composition and greatly obscure mantle and contributed to the surface though volcanic information on the mantle source regions. emissions, the production of alteration assemblages, and the In continental intra-plate regions on Earth, secondary formation of hydrothermal deposits. The recent lander and fractionation processes produce compositionally diverse orbiter missions have greatly expanded our understanding of derivative magmas that are reflected in rocks ranging from igneous diversity on the Martian surface and, through this, gabbros and anorthosites to rhyolites and phonolites. expanded the potential for gaining new insights into igneous Experimental investigations (e.g., Thompson 1975; Scoates processes on Mars. et al. 1999; Filiberto and Nekvasil 2003; Nekvasil et al. 2004; Use of the varied compositional data on Martian igneous Whitaker et al. 2007) have shown that many of these diverse rocks to extract information on the mantle source regions and lithologies can be obtained by crystallization of associated volatile budgets is of major importance to the understanding continental tholeiitic magma with different bulk water of the evolutionary history of the planet. This, however, first contents at the base of a thick continental crust, and separation requires assessment of whether the rocks represent mantle- and ascent of these residual magmas to shallower levels derived magmas that are primary (that is, reflect the within the crust. These residual magmas on Earth are often composition of the source region), or represent compositions seen as surface lavas where they are exposed along with rocks modified by secondary processes. In order for primary melts of the parental tholeiite composition. The dominance of a

853 © The Meteoritical Society, 2009. Printed in USA. 854 H. Nekvasil et al. deep fractionation signature in such evolved intra-plate follows the Coombs trend of Miyashiro (1978). Furthermore, terrestrial lavas suggests that the terrestrial continental crust they inferred that this melt was trapped at pressures above provides a density barrier to ascending mantle-derived 4.3 kbar for several reasons. First, the Al/Ti ratios of the magmas. A significant density barrier would induce ponding augites are consistent with experimental data on augite of ascending magma at the base of the barrier. Only upon crystallizing from hawaiite above 4.3 kbar. Second, the crystallization of dense minerals would the residual liquids be cumulus mineral compositions, the main minerals within the able to overcome the density barrier and separate and ascend melt inclusion assemblage, and the residual rhyolitic glass towards the surface. Such a fractionation process could compositions, can all be produced experimentally from produce lavas (residual liquids) that are compositionally quite hawaiite only at pressures above 4.3 kbar (and specifically at distinct from the original parental magma. The retention of a 9.3 kbar). Third, the low temperatures of magmatic feldspar clear high-pressure signature in the terrestrial intraplate bulk pairs within the melt inclusions (McCubbin and Nekvasil lava compositions further suggests that if any additional 2008) attest to significant solidus temperature depression and ponding and crystallization occurs within the crust, it did not hence, high volatile solubility in the melt, attainable only at result in a sufficient loss of crystals to move the lavas away elevated pressure. The presence of kaersutite provides further from the high-pressure trend. The primary crust on Mars support for elevated pressures of crystallization of the melt (Norman 1999, 2002) may have similarly provided a density inclusion assemblage. Although fluor-kaersutite can form at barrier for Martian mantle derived magmas. But is there any 1 atm (McCubbin et al. 2007a), the low F contents of the melt evidence of deep ponding events that have produced inclusion kaersutite of the Chassigny meteorite are consistent derivative magmas on Mars? Ideally, such evidence would be with higher pressures. provided by demonstrating that a known surface lithology If the Chassigny dunite is the product of the representative of a liquid composition could, if ponded at accumulation of minerals crystallizing from a magma body at depth, give rise to a known cumulate lithology. This would be depth, then its highly cumulate character (with over 95% further strengthened if, as on Earth, examples of both the cumulus olivine) is a testament to efficient crystal parental magma and the high pressure derivatives were found fractionation through melt removal. Because of the density in close proximity. difference between the residual liquid and the cumulus Exposure of deep-seated cumulates residual to crystals, the melt residual to cumulus mineral crystallization fractionation at the base of a thick crust on the surface of a likely rose to some shallower level within the crust, or even to planet with no plate-margin tectonics and hence, no exposure the surface. However, is there any evidence on the Martian of deep layers during uplift and erosion is likely to be rare. surface of a magma that reflects the type of parental liquid For Mars, as on Earth in continental intra-plate and cratonic that could produce the Chassigny cumulus and melt inclusion regions (e.g., Féménias et al. 2003; Dessai et al. 2004), assemblages at depth? Is any liquid residual to cumulus evidence of deep lithologies is likely restricted to accidentally mineral formation reflected in any of the surface rocks included cognate mineral grains, inclusions brought up analyzed on Mars? by derivative liquids from depth, or xenoliths brought up by The paucity of compositional data on Martian igneous unrelated magma once extensive meteorite bombardment of lithologies severely limits the possibility that lavas the surface ceased. Recent mineralogical and experimental representing liquids parental to Chassigny-like dunites and data on potential liquids associated with the formation of the derivatives from deep fractionation were found. Rocks that Chassigny meteorite suggest that this meteorite may well unequivocally represent melt compositions are rare among have had a deep-seated origin (Nekvasil et al. 2007; the SNC meteorites. However, the MER Rover Spirit has McCubbin and Nekvasil 2008) and was brought to shallower provided information on fine-grained vesicular volcanic levels through one of such processes before excavation by lithologies that likely represent melt compositions (McSween meteorite impact. et al. 2006). Any lithology among these that could represent a The Chassigny meteorite is a dunite with cumulus Fo68 liquid parental to a cumulate of deep origin must satisfy olivine and chromite with interstitial pyroxenes, apatite, several criteria: (i) upon crystallization at the pressure of maskelynite, and oxides. The olivine hosts large polyphase formation of the cumulus minerals it must become saturated inclusions that have been interpreted as “melt” inclusions that with the cumulus assemblage, (ii) through further contain pyroxenes, kaersutite, rare Ti-biotite, maskelynite, crystallization it must produce the observed melt inclusion alkali maskelynite, apatite, chromite, and additional minor mineral assemblage, (iii) it must produce the melt inclusion phases (e.g., Floran et al. 1978; Johnson et al. 1991; Wadhwa assemblage while preserving the observed paragenetic and Crozaz 1995; Nekvasil et al. 2007). Nekvasil et al. (2007) sequence. Any lithologies that represent residual liquids must concluded, based on the mineralogy of the cumulus and melt be saturated at the pressure of formation with the assemblages inclusion phases, that the melt trapped by the growing olivine with which they were in equilibrium before separation and was alkalic in composition and specifically a hawaiite of the produce, upon cooling, the later-formed minerals of the melt silica-saturated alkalic variety (i.e., hy normative) that inclusions within the cumulus grains. If, as suggested by Linking the Chassigny meteorite and the Martian surface rock Backstay 855

Fig. 1. Harker variation diagrams of major oxides versus silica for the rock Backstay (black circle labeled B; McSween et al. 2006) and rhyolitic glass from olivine-hosted melt inclusions within the Chassigny meteorite (black squares labeled C; Johnson et al. 1991). Bulk lava compositions from the typical Coombs trend alkalic suite from the Nandewar volcano, NSW Australia (open circles; Abbott 1969; Stolz 1985) are plotted for comparison. Lines connecting Backstay to the Nandewar trend at the same silica content as Backstay are shown to facilitate comparison.

Nekvasil et al. (2007), the type of melt trapped in the cumulus characteristics, may well represent a liquid composition. olivine grains was a silica-saturated hawaiite, then surface Based on the presence of hy in the CIPW norm it can be rocks with this characteristic would represent a reasonable further classified as a silica-saturated hawaiite. Backstay first focus for the investigation of potential parental liquids bears strong compositional similarities to the evolved and high pressure derivative liquids. hawaiitic liquid proposed by Nekvasil et al. (2007) to have Among the rocks of the Columbia Hills of Gusev crater been trapped by the olivine growing from the magma are alkalic rocks (Ruff et al. 2006). Of these rocks, the rock parental to the Chassigny dunite at a pressure close to that of Backstay has been identified as hawaiitic by McSween et al. the base of the Martian crust. As shown in Fig. 1, however, in (2006) and, with its fine-grained, unaltered textural spite of the general similarities, there are compositional 856 H. Nekvasil et al. differences between Backstay relative to hawaiites with the yield 4 wt% of the bulk into a large (0.5” diam.) graphite- same silica content from typical terrestrial Coombs trend lined Pt capsule and packing dried powder on top. The silica-saturated alkalic suites. The Backstay composition has capsule was welded shut and loaded into a talc sleeve along lower K, Al, Ca, and Ti contents, but higher Fe content. with a graphite furnace and dried mullite spacers and then These differences are sufficiently large to make it inserted into a large volume piston-cylinder apparatus. The questionable to rely on experimental data from a terrestrial sample was pressurized to 10 kbar (nominal) and heated to hawaiite. For this reason, phase equilibrium experiments at 1250 °C for ~12 h to ensure complete melting. The resulting elevated pressure were conducted on the Backstay hydrous glass was used as the source of water for the composition of McSween et al. (2006). These experiments crystallization experiments. were designed to determine (i) if the Backstay composition The starting material for the crystallization experiments could produce the cumulus and melt inclusion assemblage of was produced by grinding the hydrous glass and mixing with the Chassigny meteorite at elevated pressure and hence, appropriate aliquots of dried powder with the same anhydrous reflect the type of melt parental to the dunite (or more composition to yield 1.5 wt% and 2.6 wt% bulk water contents. specifically, to a dunite layer of Fo68 olivine of a larger The mixtures were loaded into small volume graphite capsules magmatic system) or (ii), if it was saturated with the minerals which were then sealed with graphite lids machined for a tight formed at some stage in the melt inclusion and could fit, placed into a graphite furnace along with dried mullite therefore reflect the type of liquid residual to cumulus phase spacers, and then into extra-dense BaCO3 assemblies. An crystallization that separated and rose to the surface. Figure 2 alumina disk separated the thermocouple from the graphite shows these possibilities schematically. capsule. Temperature was monitored with a Pt–Pt90Rh10 Additional experiments at lower pressures indicate the thermocouple. Temperature calibration for the assembly difference in mineral assemblages formed by Backstay yielded a 14 °C positive gradient from the thermocouple to the composition magmas at depth versus the assemblages formed center of the capsule (at the hotspot). Temperatures reported during ponding of a Backstay-like magma at shallow levels. are those at the hot spot. Pressure calibration of the cells using Unless fractionation occurs at low pressure and effectively the reaction Mg-cordierite = sapphirine + quartz (Newton et al. removes the low pressure liquid, these minerals are likely to 1974) at 7 kbar yielded a 0.7 kbar negative correction. It was be present in rocks of Backstay composition and therefore, assumed that this pressure correction remained robust at the may be seen via thermal emission spectroscopy and provide a pressures of the experiments. However, since this is not known, baseline for spectral deconvolution. the pressure is more uncertain than the 0.7 kbar friction correction at low pressures. All experiments were conducted EXPERIMENTAL PROCEDURE using the piston-out method, with over-pressurization to a minimum of two kbar before decompression to the pressure of Experimental Strategy interest. Each sample was melted at 1250 °C for a minimum of three hours and then the temperature was dropped rapidly Crystallization experiments were conducted at 9.3 kbar, (~500 °C/min) to the crystallization temperature where it 6.8 kbar, 4 kbar, and 3 kbar on the composition of brushed crystallized for a minimum of three days. Backstay formulated by McSween et al. (2006) but Use of graphite capsules for the piston-cylinder renormalized on a S-free basis (Table 1). Two bulk water experiments constrained the fO2 to be at or below the GCO contents were used for experiments at 9.3 kbar. The apparent [graphite-(C-O fluid)] buffer. Whenever an assemblage ease with which fluid escapes graphite capsules before the existed in the piston-cylinder experimental run products that starting material melts necessitated using hydrous glass as the permitted the fO2 to be computed using the QUILF method of starting material for the experiments rather than simply Andersen et al. (1993), the fO2 was between 1.5 and 2.5 log adding water to a powdered mixture of oxides. Mixtures of units below FMQ, with the lower values found for the lower powdered oxides and hydrated glass were pressurized, heated pressure experiments. above the liquidus, and then cooled to the temperature of Testing for equilibrium by paired crystallization and interest. These experiments yielded crystal + liquid synthesis experiments was not feasible because of the assemblages, thereby providing information on the nature of metastable nature of the starting material (glass + oxides). residual liquids and potential cumulus minerals as a function Therefore, only crystallization experiments (i.e., heating of temperature and pressure. above the liquidus before crystallization at the desired temperature) were conducted. The possibility of an Experimental Details equilibrium assemblage was assessed by checking for consistency in phase compositions in successively lower- A mixture of oxides of the composition of Backstay was temperature experiments, computational testing by using homogenized by grinding under alcohol in an agate mortar. QUILF (Andersen et al. 1993) to evaluate the possibility of This mixture was dried at 175 °C. Hydrous glass was disequilibrium between the ferromagnesian minerals, and synthesized by syringing sufficient water into the capsule to checking for zoning in feldspars and pyroxenes. Linking the Chassigny meteorite and the Martian surface rock Backstay 857

Analytical Techniques

The resulting glass + crystal assemblages were analyzed by electron microprobe (EMP) at SUNY Stony Brook for major element abundances. In order to minimize Na loss during analysis of glass and feldspar, the largest possible beam diameter was used. Furthermore, a time-zero correction (the “decay-curve method” of Nielsen and Sigurdsson 1981) was used for glass analysis. Mass-balance calculations were conducted for all crystalline phase-bearing run products using the software IgPet for Windows (Carr 2000). These calculations indicated the possible presence of phases not analyzed and presented a means of selecting the best analyses. Only mineral analyses that yielded good stoichiometries and phase compositions that yielded sums of the squares of the residuals of <0.5 were accepted and tabulated in this paper. Loss of Na during analysis was a problem in some glass analyses and was generally evidenced by calculated normative corundum (co) (based on the CIPW norm) and/or by mass balance computations. The amount of co was reduced when a time-zero correction was implemented, indicating that Na loss was analytically induced rather than experimentally-induced. The lack of minerals in the Fig. 2. Schematic showing two processes that appear to occur together in intra-plate regions on Earth. A mantle-derived magma (on assemblage with alkali/Al ratio higher than feldspar (which is right) makes its way directly to the surface perhaps undergoing required to produce a peraluminous melt in the absence of a shallow crystallization, but not fractionation, in the process, or alkali-bearing fluid phase) further supports the conclusion becomes trapped at the base of the crust, partly crystallizes, and the that the Na-loss was analytically-induced. Any normative co residual liquids separate and rise to the surface. In the case of remaining after zero-time correction was removed effective residual liquid removal, information on the nature of the residual liquid that departed from the cumulus grains may be only computationally from glass analyses by the addition of Na2O preserved in trapped melt inclusions within the cumulus grains. The to each analysis (prior to renormalization) until all remaining experiments conducted investigate the possibility that Backstay co was used to make normative albite. In some higher represents either the type of liquid parental to the dunite or residual to temperature experimental glasses, however, Na loss was cumulus phase formation at depth. noted by mass balance alone rather than by normative co. In these cases, Na2O was added to the analysis until proper Na Table 1. Composition of synthetic Backstay composition mass balance could be attained. Correction was completed compared with the target composition. before normative diopside appeared in lieu of normative Brushed Backstay* Synthetic composition** hypersthene. The assignment of excess alumina to Na rather SiO 50.16 50.75 + 2 than Na K loss was justified by the major improvement in TiO2 0.94 0.92 sodium mass balance after this correction and retention of Al2O3 13.45 13.29 good mass balance for K2O. FeOT 13.85 14.15 Water contents were determined by IR spectroscopic Cr2O3 0.15 0.12 measurements conducted in transmittance mode using a MnO 0.24 0.13 Nicolet 20SXB FTIR spectrometer attached to a Spectra Tech MgO 8.41 8.21 IR Plan microscope at the American Museum of Natural CaO 6.11 5.79 History. Total dissolved water in the glass of a doubly Na2O 4.20 4.13 K O 1.08 1.02 polished glass plus crystal wafer was determined from the 2 −1 P2O5 1.41 1.47 intensity of the broad band at 3570 cm ; 1024 scans were Total 100.00 100 performed for each IR spectrum acquired. Thicknesses of the *S-free renormalized composition from McSween et al. (2007). wafers were measured using a Mitutoyo digimatic indicator. **Normalized electron microprobe analysis of fused oxide mixture. Total water concentrations were calculated via the method of Dixon et al. (1995) using a molar absorptivity of 62 L/mol m. the 50 µm IR beam could pass through glass unimpeded by A more detailed discussion of the technique used is given by crystals. The water content of the residual glass at 1150 °C for Mandeville et al. (2002). the higher water experiments was determined to be 2.82 wt%. The bulk water content for each pressure was determined Mass balance indicated a crystallinity of 8.1 wt% (Table 2). for experiments in which the crystallinity was low enough that Because of the presence of only anhydrous mineral phases, 858 H. Nekvasil et al.

Table 2. Experimental phase abundances (wt%) calculated by mass balance. Temperature Phase abundance (wt% of system) (°C) Chr Ol Pig Aug Opx Pl Ti-amph Ap Ilm Ti-mgt Gl s.s.r. 9.3 kbar 2.6 wt% water 1330 0 0 0 0 0 0 0 0 0 0 100 0 1150 0.1 8.1 0 0 0 0 0 0 0 0 91.8 0.02 1100 tr. 8.5 1.4 0 7.9 0 0 0 0 0 82.2 0.02 1050 tr. 13.3 0 6.1 5.1 0 0 tr. 0 0 75.6 0.04 1000 0 12.2 0 0 13.8 4.4 12.6 2.3 0 0 54.7 0.08 980 0 7.3 0 0 21.5 7.5 17.4 2.8 0 0 43.5 0.06 9.3 kbar 1.5 wt% water 1200 tr. 2.5 0 0 0 0 0 0 0 0 97.6 0.03 1125 tr. 6.8 3.4 0 7.1 0 0 0 0 0 82.7 0.09 1080 0 6.9 0 6.3 16.9 6.4 0 0.2 0 0 63.3 0.05 1040 0 11.2 0 7 16.4 16.2 0 1.2 0 tr 48 0 6.8 kbar 2.6 wt% water 980 0 21.2 0 0 7.6 19.8 7.1 2.8 0.1 0.7 40.6 0.07 4 kbar 2.6 wt% water 980 0 20.4 0 3.4 14.1 40.6 0 2.7 tr. 1.8 17.1 0.01 3 kbar 2.6 wt% water 1080* 0 30.6 0 0.1 0 40.2 0 2.6 0.3 1.1 25.1 0.5 1080** 0 30 0 1.3 0 37.1 0 2.5 0.1 1.3 27.6 0.1 Symbols: Chr-chromite; Ol-olivine; Pig-pigeonite; Aug-augite; Opx-orthopyroxene; Pl-plagioclase; Ti-amph-Ti-amphibole; Ap-apatite; Ilm-ilmenite; Ti-mgt-Ti-magnetite; Gl-glass; tr.-seen in thin section and analyzed by EDS, but only trace amounts computed by least squares.s.s.r.-sum of the squares of the residuals. *Assumes assemblage formed post melt dehydration upon fluid loss and that Na loss is via fluid loss. **Assumes assemblage formed prior to melt dehydration, Na-corrected for presumed loss during analysis. this indicates a bulk water content of 2.6 wt% ± 0.05. (This higher water content; pyroxenes and plagioclase follow them uncertainly arose primarily from the variation in thickness of at the lower water content. Pigeonite reaches higher the polished wafer, rather than variations in IR peak heights.) abundances at the lower bulk water content. The melt reaches The water content of the low water experiments at 1200 °C saturation with plagioclase at higher temperature at the lower was determined to be 1.55 wt% and with 2.5 wt% crystallinity, bulk water content. Amphibole stability decreases with this yields a bulk water content of 1.5 ± 0.05 wt%. For each bulk decreasing pressure at constant temperature; amphibole was water content an additional measurement was made on glass at only observed at pressures above 4.3 kbar. At 2.8 kbar, high a lower temperature. The consistency of the bulk water electron microprobe totals (close to 100%) indicated water content of both temperatures indicated that water loss was not loss from the sample. This is consistent with the formation of an inherent problem with the technique and did not occur prior a fluid phase, its escape from the graphite capsule, and to melting. This comparison was repeated for each pressure progressive dehydration of the melt during crystallization. investigated and yielded the same results. Thus, the water loss For this reason, it is not known whether the Na-loss observed of the 3 kbar experiment suggests fluid saturation and in 2.8 kbar experiments was analytical or real, that is, mobilization instead of experimental failure by loss of water occurring during fluid loss from the graphite capsule. Table 2 during heating. shows mass balance implications for both possibilities. This experiment simulates the natural process of fluid loss after EXPERIMENTAL RESULTS ascent to shallow levels. Tables 3–5 show the characteristic compositions of Each experiment yielded an assemblage of homogenous phases in the assemblages of Table 2 for each temperature glass with or without crystals. Crystals were well-formed, and pressure investigated. Figure 3 shows the down- often euhedral, and readily identifiable by optical and temperature evolution of the ferromagnesian phases. Such a electron microprobe analysis. Table 2 indicates the phase trend is expected to arise during crystallization in the main assemblage at each temperature investigated for both bulk magma chamber (without the Fe:Mg exchange imposed by water contents and several pressures. Tables 3–5 indicate the the hosting olivine on melt inclusion minerals). It shows the compositions of the phases of Table 2. Chromite and olivine change in assemblage from olivine to olivine + pigeonite + are the phases found at the highest temperatures investigated orthopyroxene, the loss of pigeonite and crystallization of at 9.3 kbar. Pyroxenes and amphibole follow them at the the assemblage olivine + orthopyroxene + augite, and Linking the Chassigny meteorite and the Martian surface rock Backstay 859

Fig. 3. Ferromagnesian phases (compositions from projections using the QUILF projection scheme of Andersen et al. 1993) from experiments on Backstay at 9.3 kbar with 2.6 wt% bulk water (black circles) and 1.5 wt% bulk water (gray circles). Tie-lines connect co- Fig. 5. Variation of total alkalis versus silica in liquids residual to existing phases at each experimental temperature. Experimental crystallization of liquid of Backstay composition at 9.3 kbar and temperatures are indicated in °C. Phase symbols: Aug (augite); Pig 2.6 wt% bulk water content (black circles) and 1.5 wt% water (gray (pigeonite); Opx (orthopyroxene); Ol (olivine) Component symbols: circles). Arrows point in the down-temperature direction. The curve En (Enstatite); Fo (Forsterite); Fs (Ferrosilite); Fa (Fayalite); Wo IB is the alkalic/subalkalic boundary of Irvine and Baragar (1971). (Wollastonite). Rhyolite glass in melt inclusions from the Chassigny meteorite analyzed by Johnson et al. (1991) and Varela et al. (2000) are indicated by Ch.

produces plagioclase earlier than at higher water and this plagioclase is slightly more ternary than that produced at the higher bulk water content. This figure shows several feldspar solvus sections computed for 5 kbar (from McCubbin and Nekvasil 2008). As seen by the experimental data plotted, the solvus would not be intersected by the plagioclase solidus until temperatures much lower than those of the experiments, which is consistent with the lack of alkali feldspar in the experimental assemblages. The small pressure dependence of the ternary feldspar solvus (e.g., Wen and Nekvasil 1994) indicates that this would hold true also at the higher pressure of the experiments. Figure 5 shows the total alkalis versus silica variation in liquids residual to crystallization of Backstay composition liquid at the temperatures of the experiments at 9.3 kbar and for both 2.6 and 1.5 wt% water. Both water contents produce an increase in silica and total alkali contents in the residual Fig. 4. Plagioclase compositions crystallized at 9.3 kbar from liquids at lower temperatures. This is due to olivine % % Backstay liquid with 2.6 wt bulk water (black squares) and 1.5 wt crystallization in the drier melts and amphibole in the wetter water (gray squares). Experimental temperatures are indicated in °C. Melt inclusion maskelynite and alkali maskelynite compositions melts. The early liquids become increasingly alkalic, (open squares) from the Chassigny meteorite (McCubbin and following the Coombs trend of silica-saturated alkalic melts Nekvasil 2008) are shown for comparison. Isothermal ternary of Miyashiro (1978) in which the alkalic liquids remain hy- feldspar solvus sections (shown for 800 and 600 °C at 5 kbar) and a normative (silica-saturated) with increasing alkalinity rather likely feldspar evolutionary path for melt inclusion feldspar glass than ne-normative (silica-undersaturated). As seen in Fig. 5, with dropping temperature (heavy gray line) are adopted from McCubbin and Nekvasil (2008); details regarding these and the melt at any given silica content along the liquid lines of evolution, inclusion glass compositions are given in this reference. the residual liquids produced at lower water content are more alkalic (i.e., higher in total alkalis) than those produced at finally, as Ti-amphibole appears, the change to olivine + higher water content. This is due to the greater extent of orthopyroxene. Figure 4 shows the composition of orthopyroxene crystallization (and hence a more silica-rich plagioclase produced at 9.3 kbar for both 2.6 wt% bulk crystalline assemblage) at the expense of olivine at lower water and 1.5 wt% bulk water. The lower bulk water content water contents. This inhibits the increase in silica that is 860 H. Nekvasil et al. T Solid 33 Fs 4 Wo 63 En 22 Fs 28 Wo 50 En balance results in Table 2. balance results in Table T 40 Fa 60 0.01 0.12 0.51 0.05 0.15 Fo ) were computed using the mass T 1050 Glass Olivine Chromite Augite Opx (5.25)** 2 Or 63 T Ab 35 An Solid 28 40 Fs Fs 5 4 ns of total solids (Solids Wo Wo 67 56 En En 18.31 24.45 9.94 34.16 34.19 12.38 19.82 26.10 28 Fs 13 Wo 9 99.25 100.0 96.41 99.18 98.53 98.50 98.85 100.0 59 0.06 0.10 7.11 2.70 0.10 1.84 En bulk water content. % (6.65)** Glass Olivine Opx Plagioclase Amphibole Apatite Solid 36 51 T Fa Fa 1992). Subscript T refers to total. Compositio 64 49 0.01 0.00 0.26 0.06 0.05 4.83 Fo Fo the norm (corrected value in parenthesis). 2.89 1.59 5.83 Glass Olivine Chromite Pigeonite Opx (4.79)* 2 Or 62 LF (Anderson and Lindsley T Ab til no Co remained in of phases at 9.3 kbar and 2.6 wt 37 (6.95)* An Solid lue in parenthesis). 37 Fs 5 Wo 58 projected through QUI En to the CIPW norm un 31 46 Fa Fa 69 54 ss balance (corrected va 0.03 0.02 0.03 4.64 0.04 0.08 6.55 Fo Fo 0.960.03 0.00 0.07 1.71 41.12 0.021.48 0.49 1.06 0.14 0.05 0.01 0.020.74 0.07 2.15 0.140.01 33.30 0.047.79 0.32 0.02 0.33 0.99 1.65 37.96 0.18 21.98 0.29 0.14 0.59 0.050.96 0.01 0.01 0.30 0.16 3.92 0.45 0.38 0.12 13.77 0.53 0.09 1.09 1.27 0.01 21.36 0.03 0.22 0.31 0.07 0.14 6.55 0.07 0.02 3.55 0.05 32.43 0.09 41.40 0.31 0.04 0.68 0.87 0.24 1.78 24.22 1.90 0.24 0.30 0.48 0.65 0.29 0.02 0.74 0.01 0.02 0.36 0.27 0.36 3.34 15.17 0.29 0.36 0.08 0.06 1.81 0.07 0.02 19.84 1.17 0.14 0.22 0.25 0.41 40.59 2.26 (4.3)* (6.21)** 49.92 36.9413.8212.52 0.00 0.04 27.37 18.19 28.20 37.13 0.23 50.56 27.80 15.56 36.15 11.62 0.0054.14 0.06 30.65 51.48 23.05 34.8017.74 31.31 51.84 0.04 3.05 17.71 51.75 3.44 58.46 2.83 44.67 25.83 40.72 51.47 1.53 12.45 43.48 35.30 16.75 56.08 7.37 1.09 33.83 18.45 0.04 49.64 20.82 49.43 0.03 51.12 4.05 58.36 3.64 42.52 3.32 41.38 25.37 1.72 5.77 12.40 44.65 1.50 8.76 3 3 3 3 T T C) 1150 1100 5 5 C) 1000 980 2 2 2 2 O4.21 O4.21 O5.29 O O O O ° ° 2 2 O 1.08 0.00 0.02 0.00 1.14O 0.00 1.55 0.03 0.00 0.01 0.01 0.00 0.28 0.00 0.37 1.27 0.14 2.06 0.02 0.06 0.02 0.00 0.00 0.00 0.30 0.01 0.35 0.11 0.16 2 2 2 2 O O 2 2 Na loss corrected by adding 2 2 Na loss corrected by ma Pyroxene and olivine compositions Al FeO Total 95.86 98.49 96.89Al FeO 100.0MgOCaONa 1.89 97.00 4.48 24.27 0.25 98.64 19.33Total 97.20 2.24 94.67 99.5 0.05 98.15 8.01 99.76 11.47 9.88 16.00 100.1 1.35 6.48 96.43 22.12 3.60 100.0 0.24 18.48 94.97 98.48 1.86 0.07 98.58 11.41 7.44 99.57 0.53 9.34 13.73 96.99 50.12 100.7 7.21 100.0 * K Phase Glass Olivine Chromite T ( SiO T ( PhaseSiO Glass Olivine Opx Plagioclase Amphibole Solid ** TiO Cr TiO Cr MgOCaONa 5.64 6.07 33.50 0.16 7.33 0.13 33.75 0.16 4.04 6.55 31.03 7.06 0.26 0.13 20.70 4.69 23.13 2.07 27.02 3.10 1.42 28.75 6.05 5.77 0.21 14.74 0.32 21.45 15.16 24.02 1.98 4.36 MnOP 0.13 0.24 0.16 0.24 0.13MnO 0.25P 0.09 0.16 0.28 0.26 0.26 0.19 0.02 0.23 0.12 0.12 0.20 0.07 0.26 0.16 0.33 0.18 0.29 0.26 0.01 0.24 0.13 0.04 0.20 K Table 3. Representative compositions Table Linking the Chassigny meteorite and the Martian surface rock Backstay 861 T Solid 3 Or 59 Ab 38 An 34 Fs 4 Wo 62 En 25 Fs 28 Wo 47 2.31 20.33 0.36 19.96 En 0.02 0.78 0.12 6.82 0.58 43 Fa 57 (5.4) Glass Olivine Augite Opx Plag Fo T Solid 29 Fs 4 9 46.02 50.63 35.60 50.48 51.04 58.85 48.43 2 0.46 0.05 0.05 0.59 0.63 0.02 0.49 mineral compositions. Wo 32 22.32 11.53 36.36 1 67 En 26 Fs 14 Wo sults in Table 2 and sults in Table bulk water. 60 % En 33 T Fa 67 Fo in the CIPW norm. 38 Fs 4 Wo 58 Glass Olivine Pigeonite Opx En 23 nderson and Lindsley 1992). Fs ) were computed using the mass balance re no corundum remained 30 T T Wo 47 Solid En s of phases at 9.3 kbar and 1.5 wt 3 l solids (Solids Or value in parenthesis). 60 in parenthesis until Ab 37 An tries from QUILF projection (A tries from QUILF projection 46 mpositions of tota Fa 54 0.03 0.02 0.03 4.76 0.02 0.320.02 6.80 0.12 0.797 0.12 5.33 0.11 2.30 ss balance (corrected ss balance (corrected Fo adding Na to value 0.920.09 0.03 0.12 1.18 42.87 0.031.40 0.12 0.55 1.06 0.01 0.02 0.041.54 0.22 0.540.00 0.34 0.04 0.78 0.08 0.08 1.61 0.00 0.23 0.5 0.917 0.171.82 0.18 0.623 0.12 0.27 1.24 0.50 0.16 0.45 0.05 0.32 0.03 0.72 0.25 0.243 0.10 0.39 2.01 0.06 0.56 0.03 1.16 0.25 0.39 0.04 0.21 0.18 (4.3)* (6.01)* 49.1813.12 37.9213.16 0.52 0.05 23.49 19.58 24.78 37.23 0.05 49.56 23.07 15.56 11.83 37.3152.22 50.60 0.14 29.3416.55 34.94 3.35 17.0110.14 57.51 0.04 52.1 38.24 26.05 47.4 3.06 18. 0.49 4.851 1.98 13.5 50.88 16.46 3.38 48.13 0.07 22.90 9.89 5.53 17.55 4.91 26.34 5.27 25 3 3 3 3 Fa T T C) 1200 1125 1080 5 5 C) 1040 2 2 2 2 O3.74 O3.74 O5.02 O O O O ° ° 2 2 O 1.03 0.01 0.02 0.01 1.25O 0.00 1.92 0.01 0.00 0.45 0.00 0.031 0.00 1.54 0.01 0.04 0.15 0.01 0.00 0.42 0.03 75 2 2 2 2 O O 2 2 2 2 K K Pyroxene and olivine stoichiome TiO Al TotalFo 95.56 101.8 98.66Al FeO 100.0 96.59 101.10Total 98.74 96.67 99.49 100.0 99.41 100.0 96.97 97.96 100.5 100.3 100.4 99.96 100.5 101.35 99.90 FeO T ( PhaseSiO Glass Olivine Chromite T ( PhaseSiO Glass Olivine Plag Aug Opx corrected by ma *Na loss Solid **Na loss corrected by Subscript T refers to total. Co Cr TiO Cr MgOCaONa 2.41 4.92 25.19 0.19 0.02 7.78 12.69 15.6 19.87 1.78 13.48 6.47 MgOCaONa 7.09 5.62 39.18 0.15 9.27 0.13 38.47 0.15 4.31 6.39 33.33 20.82 0.19 4.99 23.27 1.92 26.66 3.10 1.86 27.13 5.93 13.37 0.28 20.73 16.02 0.02 1.93 18.89 8.25 5.31 MnOP 0.20 0.32 0.27 0.32 0.25MnOP 0.34 0.14 0.38 0.48 0.00 0.37 0.32 0.36 0.14 0.35 0.36 0.26 0.31 0.42 0.02 0.36 Table 4. Representative composition Table 862 H. Nekvasil et al. ) T T Solids 4 Or 64 Ab 32 An tions of total solids (Solids 19 Fs 40 Wo 41 En t T refers to total. Composi 40 Fs 6 Wo 54 En 49 Fa 51 Fo 1.180 0.184.207 39.60 0.38 0.01 23.86 0.25 0.970.46 10.76 0.32 0.56 27.18 55.26 0.16 0.14 1.07 7.42 0.04 0.36 0.84 16.23 0.05 0.25 0.35 1.64 64.096 33.7216.099 0.11 51.06 1.51 51.08 2.20 0.55 58.08 2.20 24.88 48.09 12.74 3 3 T 5 2 2 O 6.707 0.07 0.08 0.42 0.15 7.13 3.57 O O 2 O 4.138 0.03 0.01 0.04 0.08 0.75 0.39 2 2 O 2 2 Pressure (kbar)Bulk water (wt%) 4.0 2.6 Temperature 980 Phase Glass Olivine Opx Augite Mgt Plagioclase mpositions as QUILF projections (Andersen et al. 1991). Subscrip T Solids 100.2 Total 99.25 98.80 99.18 99.96 97.00 99.38 100.2 2 Or 56 Ab 42 An T 3 Or 56 Ab 41 ined (corrected value in parenthesis). Pyroxene and olivine co An 19 35 Fs Fs 5 34 Wo Wo 60 47 En En ositions of phases at selected pressures below 9.3 kbar. 47 46 Fa Fa 53 54 0.04 0.33 6.49 3.44 Fo 1.870.02 0.134.63 0.00 0.91 37.39 0.93 10.40 0.070.90 0.02 1.07 0.55 0.55 0.27 17.15 0.31 0.17 1.72 0.970.00 0.005.80 0.00 0.40 38.50 0.21 23.000.57 4.12 14.29 0.38 0.30 51.32 0.06 16.00 40.84 0.94 47.57 19.36 0.0667.62 0.37 0.58 0.00 33.6014.24 0.88 51.27 0.08 0.00 TiO 19.52 0.32 FeO 0.02 Cr 3.15 58.97 0.16 26.24 45.31 2.12 P 13.99 58.4316.53 34.86 50.17 0.02 2.22 40.55 13.20 0.22 1.37 0.42 57.82 7.11 26.42 43.56 SiO 10.88 Al (5.94)* Fo 3 3 3 3 T T C) 980 5 5 2 2 2 2 O 6.28 0.02 0.05 2.91 0.05O 0.18 6.75 3.97 2.63 Na O O O O ° 2 2 O 2.28 0.01 0.00 0.36 0.05O 0.10 0.27 3.31 0.14 0.01 K 0.02 0.51 0.28 2 2 2 2 O O 2 2 were computed using the mass balance results in Table 2 and mineral compositions. were computed using the mass balance results in Table 2 2 MnOP 0.10 0.59 0.51 0.16 0.51 0.46 0.00 0.31 MnO 0.101 0.69 0.43 0.25 0.58 0.02 0.28 TiO Cr Al FeO MgOCaONa K 1.22Total 2.63 24.76 0.41 14.45 17.79 100.54 97.66 0.34 99.77 8.67 10.63 102.5 6.49 100.2 Bulk water (wt%) 2.6 T ( Phase Glass Olivine Opx Amph Ilm Spinel Plagioclase Pressure (kbar) 6.8 SiO Pressure (kbar)Bulk water (wt%) 2.6 3.0 TemperaturePhaseSiO 1080 1080 Glass 1080 Olivine Augite 1080 Plagioclase 1080 Solids *Na loss corrected by adding Na to the CIPW norm until no Co rema TiO Cr Al FeO MgOCaONa 1.34 3.34Total 24.83 0.22 20.05 95.64 2.17 11.38 99.35 9.60 98.84 3.88 3.46 97.32 0.20 0.30 0.03 98.43 8.82 95.93 13.00 100.9 MgO 6.84 CaO 0.676 23.69 1.585 18.48 0.39 13.16 2.96 20.16 3.36 0.09 0.18 6.83 9.75 6.46 K MnOP 0.14 0.68 0.25 0.00 0.29 Table 5. Representative comp Table Linking the Chassigny meteorite and the Martian surface rock Backstay 863 induced by olivine crystallization and leads to a greater If B1150 emulates the type of liquid trapped by the change in combined (Fig. 5) and individual alkalis (Fig. 6) cumulus olivine of the Chassigny meteorite, then Backstay relative to silica during liquid evolution. The extent of silica should also produce the mineral assemblage of the melt enrichment relative to alkalis increases with decreasing inclusions at the pressure of formation of the olivine grains. pressure and decreases with decreasing water content (Figs. 5 However, the compositions of the melt inclusion minerals and 6). The differences in mineralogy of the crystallizing (primarily the ferromagnesian minerals) in the Chassigny phases as a function of water content and pressure also affect meteorite were likely affected by Fe:Mg exchange with the other major elements. Figure 6 shows the composition of olivine host during cooling of the melt inclusions. This residual liquids at 9.3 kbar and 2.6 and 1.5 bulk water content. exchange would inhibit the down-temperature evolution of For any given silica content, K2O, TiO2, and P2O5 tend to be pyroxene to lower Mg#s and increase the Ca content of the higher in the low water melts. Pressure strongly affects the augite (e.g., Nekvasil et al. 2007). Such modification restricts titania content, with higher titania contents retained in the useful comparison of experimental and natural mineral melt to higher silica contents at the lower pressures. Alumina compositions to phases that formed close to the temperature contents decrease more strongly with increasing silica content of crystallization of the olivine host. These likely remained at lower pressures (Fig. 6) because of the crystallization of relatively unperturbed in composition because of the small plagioclase. The evolution of the oxides constituents CaO, compositional changes required to equilibrate them with the MgO, and FeOT of the residual melts are affected only host olivine. Texturally, both pigeonite and orthopyroxene slightly by pressure and water content. appear to have crystallized early in the Chassigny melt inclusions (Nekvasil et al. 2007). The composition of the ° % DISCUSSION orthopyroxene at 1100 C (En67Wo5Fs28) for 2.6 wt water agrees well with that in such inclusions (En68Wo4Fs28; Backstay Composition: A Proxy for the Magma Parental Nekvasil et al. 2007). Comparison of pigeonite analysis in the to the Chassigny Dunite? melt inclusion (En63Wo13Fs24; Nekvasil et al. 2007) with that ° crystallized at 1100 C (En59Wo13Fs28; Table 3, Fig. 2) If ponded at ~70 km depth (9.3 kbar pressure), a magma suggests a slightly higher temperature of crystallization for of Backstay composition with 2.6 wt% bulk water content the former. Lower temperature re-equilibration with the produces the Chassigny cumulus assemblage of Fo68 olivine olivine host could explain the difference between the Ca ° % and chromite at 1150 C after 8 crystallization. This content of the augite of the melt inclusions (En50Wo34Fs16; indicates that the Backstay composition cannot represent the Nekvasil et al. 2007) and that of Table 3 at 1050 °C type of melt residual to crystallization of the cumulus (En50Wo28Fs22). assemblage of the Chassigny meteorite, but has the potential Although the lower temperature ferromagnesian mineral to be representative of the type of melt parental to dunitic compositions of the melt inclusion are expected to differ from layers with Fo68 olivine and chromite as well as to cumulates those experimentally obtained because of the equilibration/re- of olivine of more forsteritic composition. Pigeonite and equilibration with the host, the igneous paragenetic sequence orthopyroxene appear experimentally at lower temperatures of melt inclusion minerals may well be comparable with the than olivine of the Chassigny cumulus olivine composition experimental mineral crystallization sequence. Based on (Fo68) consistent with their absence from the Chassigny petrography, Nekvasil et al. (2007) and McCubbin et al. cumulus assemblage. Since the absence of cumulus pyroxene (2008a) concluded that feldspar (now maskelynite and minor in the Chassigny dunite likely suggests that the residual alkali maskelynite) appears to be a late major phase appearing liquids separated and ascended prior to pyroxene in the crystallization sequence of the Chassigny melt crystallization, any melt trapped in the growing olivine inclusions, and its appearance post-dates the appearance of would most likely reflect the composition of evolved liquid Ti-amphibole (maskelynite is clearly interstitial to subhedral prior to pyroxene saturation but after saturation with the amphibole and neighboring pyroxene). Apatite also appears cumulus olivine and chromite. Table 6 shows the earlier than feldspar as indicated by the presence of apatite composition of the derivative liquid from Backstay saturated within maskelynite (McCubbin and Nekvasil 2008). This ° with Fo68 and chromite at 1150 C that simulates the liquid sequence of appearance of minerals is consistent with that trapped by growing olivine. Table 6 also shows the derivative obtained experimentally at 9.3 kbar and 2.6 wt% water (Table 2). liquid from a terrestrial hawaiite proposed by Nekvasil et al. Although, the exact sequence of appearance of feldspar (2007) to simulate the trapped liquid of the Chassigny melt relative to amphibole was not explicitly ascertained in the inclusion. For the most part, the differences between the experiments, the phase abundances of these at 1000 °C starting compositions (Backstay and Nandewar hawaiite are consistent with the amphibole appearing before 49000) are retained in the evolving liquids, although the CaO plagioclase based on previous experiments on kaersutite and contents of the evolved liquids become more similar to each plagioclase bearing assemblages in alkalic and tholeiitic other and the MgO more dissimilar (Fig. 6; Table 6). parental compositions (e.g., Nekvasil et al. 2004). The 864 H. Nekvasil et al.

Fig. 6. Variation of major oxides with silica in liquids residual to crystallization of Backstay composition (B) at 9.3 kbar and 2.6 wt% bulk water content (black circles) and 1.5 wt% water (gray circles). B1150 (black circle with gray outline) is the liquid residual to crystallization of Backstay composition at 9.3 kbar and 2.6 wt% bulk water content. Gray diamonds show residual liquids from lower pressure experiments (pressures labeled in italics). Numbers in (a) refer to the amount of liquid remaining (wt%) for liquids with 2.6 wt% bulk water. Temperatures of experiments with 2.6 wt% water are indicated in (c). Compositions of the Gusev crater rocks Humphrey (H; black square) and Irvine (I; black triangle) are indicated. The Nandewar hawaiite, used experimentally by Nekvasil et al. (2007), is indicated by NH and gray squares show the liquids residual to its crystallization at 9.3 kbar and 2 wt% water. NH1050 is the liquid residual to NH crystallization proposed by Nekvasil et al. (2007) to represent the type of liquid trapped in the cumulus olivine of the Chassigny meteorite. The open circle indicates the bulk olivine + chromite composition from the highest temperature experiments at 2.6 wt% water. The arrow indicates the direction that liquid would evolve should such an assemblage crystallize from Humphrey and Irvine composition liquids. Rhyolite glass in melt inclusions from the Chassigny meteorite (Johnson et al. 1991; Varela et al. 2000) is indicated by Ch (open triangles). feldspar compositions lie on the plagioclase evolutionary path found in the Chassigny melt inclusions. These include for melt inclusions of the Chassigny meteorite (Fig. 3) as pigeonite and chromite, two phases produced by the Backstay indicated by McCubbin and Nekvasil (2008). composition but not by the terrestrial hy-normative hawaiite The experimental mineral assemblages arising from the proposed by Nekvasil et al. (2007). Nonetheless, there remain Backstay composition contain most of the mineral phases some differences between the observed melt inclusion Linking the Chassigny meteorite and the Martian surface rock Backstay 865

Table 6. Composition of proposed liquid trapped in In spite of these deviations between the melt inclusion Chassigny melt inclusions (B1150: Backstay residual assemblages and those determined experimentally for the liquid at 1150 °C, 9.3 kb, 2.6 wt% bulk water; NH1050: Backstay composition, the similarity of the major minerals Nandewar hawaiite 1050 °C, 2 wt% water, 9.3 kbar) and shows that magma with composition similar to Backstay parental liquids. could have played an important role in the formation of Backstay* B1150 NH** NH1050 dunite at the base of the Martian crust. The phase relations SiO 50.16 52.08 48.61 52.59 for the terrestrial hawaiite were sensitive to pressure and 2 indicated a need for deep crystallization to produce some of TiO2 0.94 1.00 2.74 1.97 the key phases of the Chassigny melt inclusion assemblage Al2O3 13.45 14.42 15.05 18.48 FeOT 13.85 13.06 10.85 10.02 (Nekvasil et al. 2007), but are the assemblage of Backstay Cr2O3 0.15 0.03 0.00 NA equally sensitive? How sensitive are the results to the bulk MnO 0.24 0.14 0.13 0.13 water content and the parental magma composition? Could MgO 8.41 5.88 8.46 3.24 magmas with compositions of any of the other Gusev crater CaO 6.11 6.33 9.28 6.48 rocks such as Irvine or Humphrey produce similar Na2O 4.20 4.49 2.90 3.92 assemblages? K2O 1.08 1.13 1.48 2.38 P O 1.41 1.54 0.49 0.78 2 5 Constraints on the Pressure of Crystallization Total*** 100.0 100.00 100.0 100.0 The pressure of the first set of experiments was chosen to *S-free composition from McSween et al. (2007). **Nandewar hawaiite 49000 (Stolz 1985). facilitate comparison with the results for the terrestrial ***Renormalized to an anhydrous basis. hawaiite proposed by Nekvasil et al. (2007) as a possible example of the type of liquid trapped as Chassigny melt assemblages and those produced by Backstay at 9.3 kbar and inclusions. Several additional experiments were conducted on 2.6 wt% bulk water. Backstay under these conditions the Backstay composition with 2.6 wt% bulk water content at produces a Ti-amphibole much lower in Ti than the kaersutite 6.8, 4, and 3 kbar pressure in order to constrain the pressure found in the Chassigny melt inclusions. This suggests that the range over which a Backstay-like liquid could produce the Backstay bulk composition is either too low in Ti compared to minerals and paragenetic mineral sequence of the melt the actual trapped melt, or that an additional fractionation inclusions. stage for Backstay liquid at low pressure involving the loss of At 6.8 kbar for 2.6 wt% bulk water content, the Ti- a Ti-bearing phase (e.g., ilmenite) took place. amphibole is kaersutite (Table 5). This result suggests that A mineral found in the melt inclusions and not seen in Backstay crystallizing at 6.8 rather than 9.3 kbar may provide the experimental assemblage at 9.3 kbar and 2.6 wt% bulk an even better fit to the observed paragenetic sequence of the water content is Ti-biotite. Ti-rich biotite is found in a melt inclusions. [However, although kaersutite is produced at variety of alkalic rocks on Earth (e.g., Bachinski and this pressure, the low Ti content of the amphibole relative to Simpson 1984) and appears to be stable at higher that within the Chassigny melt inclusions (and relative to temperatures than low-Ti biotite (e.g., Zhou 1994). Nekvasil what was produced at this pressure from a terrestrial hawaiite) et al. (2004) determined that for the Nandewar hawaiite, still suggests that the Ti content of the Backstay composition Ti-biotite appears before kaersutite along the experimental is lower than that of the magma parental to the Chassigny fractionation path at bulk water contents lower than 2 wt%, dunite.] At 4 kbar and 3 kbar, plagioclase appears before % yet above 0.4 wt H2O. In addition to the limited water amphibole, which disagrees with the observed paragenetic contents over which it is stable, Nekvasil et al. (2004) sequence of the meteorite (Nekvasil et al. 2007; McCubbin observed a limited thermal stability of Ti-biotite as it and Nekvasil 2008). These results for the crystalline undergoes reaction with melt to produce kaersutite soon assemblages indicate that crystallization of Backstay after forming. Its limited stability may explain why it composition liquid at depths of roughly 50 km or higher with appears to be so rare in Chassigny melt inclusions and so bulk water contents of 2.6 wt% could produce the Chassigny difficult to obtain experimentally. Alternatively, its absence dunite and melt inclusion assemblages, but crystallization at may be related to a Ti or K content of Backstay that is lower shallower levels would not. than a true parental liquid of the Chassigny dunite. Comparison of the liquid lines-of-descent at high silica Alkali feldspar (alkali maskelynite in the Chassigny melt contents with the sodic high alkali rhyolitic glass of the inclusions, McCubbin and Nekvasil 2008) is the other mineral Chassigny melt inclusions (e.g., Johnson et al. 1991; Varela not found in the experimental assemblage. This is a very late et al. 2000) provides further constraints on pressure of phase in the Chassigny melt inclusions and its absence in the crystallization. The composition of this rhyolitic glass is experimental assemblages may simply reflect experimental similar to rhyolite in silica-saturated magmatic suites in temperatures that did not extend close enough to the solidus continental hotspots (e.g., Nandewar volcano of NSW temperature for the feldspar solidus to intersect the solvus (as Australia, Stolz 1985) in which the mafic and intermediate suggested by the feldspar solvus in Fig. 4). lavas lie along an evolutionary trend consistent with 866 H. Nekvasil et al. fractionation of a water-“rich” tholeiite at the base of the Role of Bulk Composition continental crust. These rhyolites differ distinctly from those Beyond the Backstay composition proposed here and the found in low-pressure hotspot suites (e.g., Thingmuli, terrestrial tholeiite proposed by Nekvasil et al. (2007), several Iceland, Carmichael 1964) or crystallizing from “dry” compositions have been proposed in the literature to basaltic magma. Figure 6 shows that the liquid produced at characterize the melt parental to the Chassigny dunite (e.g., 980 °C and 6.8 kbar is consistent with a trend that would Johnson et al. 1991; Nekvasil et al. 2007; Filiberto 2008). reach the Chassigny rhyolite. Mass balance calculations Such a melt need not be saturated with the cumulus phases but indicate that this could occur with the minerals precipitating should eventually produce not only the cumulus phases but at 980 °C. The liquid at 4 kbar deviates sufficiently in Ti and also the melt inclusion assemblages. Proposed parental Al that it cannot lead to the Chassigny rhyolite (Fig. 6) either compositions for the Chassigny dunite range from low-Al, with the assemblage precipitating or through a variety of high Fe compositions (e.g., Johnson et al. 1991) to the reasonable mineral assemblages tested by mass balance. At picrobasalts found at Gusev crater (e.g., Filiberto 2008). 3 kbar, fluid saturation was reached significantly above the These compositions provide a means of assessing the solidus temperature, and at least by the stage of 75% importance of bulk composition in producing the Chassigny crystallinity. The residual silicate melt produced at this assemblages. pressure also deviates from the rhyolitic glass seen in melt Filiberto (2008) conducted experiments on the low-Al inclusions of the Chassigny meteorite (Fig. 6) particularly in high-Fe liquid proposed by Johnson et al. (1991) as parental alumina and magnesia; upon further crystallization of the to the Chassigny meteorite. The experimental results obtained observed assemblage it would not yield the rhyolitic glass coupled with those of Minetti et al. (2000) indicate that such compositions. melt (or melt of similar composition) does not produce the cumulus mineral assemblage of the Chassigny meteorite, nor Constraints on Bulk Water Content can it produce the melt inclusion assemblages at the varied The absence of Ti-biotite and the low Ti-content of the pressures and water contents investigated. amphibole may indicate that the bulk water content (2.6 wt%) Filiberto (2008) proposed that magma similar in chosen for the first set of experiments is too high. For this composition to the Gusev basalt Humphrey may produce the reason, additional experiments were conducted at 1.5 wt% requisite assemblages upon cooling. A variety of bulk water. Table 2 shows the mineral assemblages obtained; experimental phase equilibrium studies have been conducted phase compositions are given in Table 4. The experimental on the Gusev crater rock Humphrey (e.g., Monders et al. results indicate that this bulk water content is too low to 2007; Filiberto 2008; McCubbin et al. 2008). Subliquidus produce the paragenetic sequence of minerals found in the phase assemblages on the F-, Cl- and Cr-bearing Humphrey Chassigny melt inclusions for several reasons. First, the composition at 9.3 are given in McCubbin et al. (2008). At composition of Chassigny cumulus olivine is not attained this pressure, they show early saturation with olivine and until the melt is also saturated in pigeonite and chromite; however, the assemblages forming from the orthopyroxene. These pyroxene phases, however, do not hydrous melt do not include kaersutite but instead include appear in the cumulus assemblage of the Chassigny dunite; basaltic hornblende. Furthermore, the very low K2O content instead, they are restricted to the regions interstitial to the suggests that Ti-biotite cannot be stable at any temperature cumulus grains (Nekvasil et al. 2007). Second, plagioclase and the residual liquid would produce neither alkali feldspar appears prior to amphibole unlike in the observed paragenetic nor the rhyolite glass found in the melt inclusions (e.g., sequence of the melt inclusion assemblage. Since dissolved Johnson et al. 1991). water suppresses plagioclase crystallization, the early Irvine is another Gusev basalt (McSween et al. 2006), but appearance of plagioclase indicates that the dissolved water with characteristics more similar to Backstay. Experiments on content is too low. Furthermore, its compositional evolution the Irvine composition by Harrington et al. (2008 and (Fig. 4) suggests that at lower temperatures it will become too personal communication) at 1.9 wt% bulk water and 9.3 kbar ternary in character to be consistent with the analyzed melt produce a mineral assemblage very similar to that of Backstay inclusion maskelynite (McCubbin and Nekvasil 2008). and the requisite silica-enrichment during crystallization. Finally, apatite appears after plagioclase, unlike what is seen However, its bulk K content is too low to produce the rhyolite in the melt inclusions. Taken together, these results suggest glass composition of the Chassigny melt inclusions. that although 2.6 wt% bulk water may be too high, the bulk The terrestrial hy-normative hawaiite (H1050, Table 6) water content must be above 1.5 wt% to obtain the correct proposed by Nekvasil et al. (2007) to approximate the type of paragenetic sequence. The conclusion that the melt parental liquid that could have been trapped by the olivine grains, to the Chassigny meteorite contained significant water is produced many of the mineralogical characteristics of the consistent with the results of McCubbin and Nekvasil (2008) cumulus and melt inclusion assemblages of the Chassigny who looked at apatite and maskelynite relations to constrain meteorite but did not produce pigeonite and chromite, two fluid history in the Chassigny meteorite. phases produced by the Backstay composition. It further Linking the Chassigny meteorite and the Martian surface rock Backstay 867 produced earlier saturation with augite than the cumulus any given silica content along the path. For the same reason, mineralogy suggests (Nekvasil et al. 2007). These at any given extent of crystallization, the lower water content characteristics likely reflect the lower chromium content and liquids will be more alkalic. higher calcium content of the terrestrial hawaiite bulk Table 3 shows the bulk solids crystallizing at any one composition (Table 6 and Fig. 6). time from Backstay composition liquid at 9.3 kbar and The suspected inability of magmas reflecting the 2.6 wt% bulk water content. If the residual liquid leaves the Humphrey, Irvine, or terrestrial hawaiite composition to system and ascends towards the surface, the base of the crust produce the Chassigny cumulus and melt inclusion and the lower crust will be enriched in the accumulated assemblages is consistent with their inability to produce a minerals. Early separation leaves behind cumulus material B1150 liquid (Table 6) by crystallization of the olivine (and that is low in silica and high in MgO with a much higher Mg# chromite) cumulus phases of the Chassigny dunite. This is then the departing liquid. This reflects the early dominance of demonstrated in Fig. 6 by the tielines from olivine through olivine in the cumulus assemblage. Once pyroxene joins the these bulk compositions. None of the tielines intersect this assemblage, the cumulus material left behind becomes more composition for all oxides. Furthermore, since none of these silica-rich, with increasing amounts of alumina as the rocks lie compositionally along the liquid lines-of-descent of FeO+MgO is reduced. The change to augite increases the the Backstay composition (Fig. 6), none of these rock CaO content of the cumulus material but drops the silica compositions could represent liquids residual to partial content slightly. The cumulus assemblage remains crystallization of a Backstay-like magma. characterized by a low silica content through 50 wt% crystallinity. Implications for Martian Crustal Stratigraphy For lower bulk water contents, the Backstay solid assemblage left behind in the lower crust at 9.3 kbar will show The experimental data suggest that Backstay may have a different compositional evolution. Table 4 shows that the many of the characteristics of the type of magma trapped at bulk solids take on a basaltic silica content with depth that gave rise to the Chassigny dunite. Yet its compositional characteristics more similar to the basaltic presence on the surface indicates that such magma can also shergottites than to a terrestrial basalt. The compositions rise to the surface. If Backstay liquid with 2.6 wt% water show low alumina and high FeO contents. Once plagioclase were to remain at depth and continue to cool and crystallize joins the assemblage, the alumina and sodium content of the beyond 1150 °C (the temperature of production of the bulk solids increase while the potassium content remains low. cumulus assemblage of the Chassigny dunite), residual At lower pressures and at the temperatures of the liquids would be produced that have high silica and alkali experiments, the bulk solids are characterized by high sodium contents (Fig. 6). At any stage of crystallization, such and low or basaltic silica contents. residual liquids could separate from the cumulus These experiments indicate a marked difference between assemblage and rise to the surface, leading to a range of the composition of the material remaining and the residual alkali-rich intermediate compositions. These compositions liquids. This suggests that fractionation near the base of the differ from the calc-alkaline andesite so common on Earth crust can produce significant compositional variation between because of their higher alkali contents and lower Ca the upper and lower crusts. Should such fractionation contents, and are termed hawaiite, trachyandesite (or processes be common on Mars, the igneous surface rocks mugearite), tristanite (or benmoreite), and trachyte with would likely predominantly reflect the compositions of the increasing silica content. Importantly, these alkalic liquids evolved liquids residual to fractionation. They would not are not silica-undersaturated in that they contain no represent primary mantle melts and in fact, may have retained feldspathoids in the norm. With increasing silica, these little information about the chemical characteristics of the liquids become more Fe-poor and Al-rich (until highly source region of the parental magma. silicic compositions are attained). These compositions differ from evolved liquids of terrestrial silica-saturated hawaiite Acknowledgments–This manuscript benefitted greatly from with higher Na and P, and lower K and Al (Fig. 6); detailed reviews by E. Medard, J. Hammer, and C. Floss. however, some of these differences diminish for the more Financial support for this work was provided by NASA MFR evolved silicic compositions. grant NNG04GM79G to H.N. and a Department of Education Although bulk water content (within the range studied GAAN grant for F. McCubbin. here) has a significant impact on mineral paragenetic history, it has only a minor impact on the overall liquid line-of- Editorial Handling—Dr. Christine Floss descent (Fig. 6). However, the higher proportion of pigeonite crystallizing out early at lower water contents inhibits the REFERENCES increase in silica content of the evolving liquid. Therefore, as seen in Fig. 6, less liquid remains at lower water content for Abbott M. J. 1969. Petrology of the Nandewar Volcano, N.S.W., 868 H. Nekvasil et al.

Australia. Contributions to Mineralogy and Petrology 20:115– Lincei, Mineralogial Society of America, and Geochemical 134. Society Short Course on Amphiboles: Abstracts 26–27. Andersen D. J., Lindsley D. H., and Davidson P. M. 1993. QUILF— McCubbin F. M., Nekvasil H., Harrington A., Elardo S., and Lindsley A Pascal program to assess equilibria among Fe-Mg-Mn-Ti D. H. 2007b. Experimental crystallization of dry and wet oxides, pyroxenes, olivine, and quartz. Computers and Humphrey at 9.3 kbar: Implications for compositional diversity Geosciences 19(9):1333–1350. of the Martian crust. AGU Fall Meeting, San Francisco, CA. Bachinski S. W. and Simpson E. L. 1984. Ti-phlogopites of the Shaw Abstract #P13E-07. Cove minette—A comparison with micas of other lamprophyres, McCubbin F. M., Nekvasil H., Harrington A., Elardo S., and Lindsley potassic rocks, kimberlites, and mantle xenoliths. American D. H. 2008. Compositional diversity and stratification of the Mineralogist 69(1–2):41–56. Martian crust: Inferences from crystallization experiments on the Carmichael I. S. E. 1964. The petrology of Thingmuli, a Tertiary picrobasalt Humphrey from Gusev crater, Mars. Journal of volcano in eastern Iceland. American Mineralogist 52:1815–1841. Geophysical Research 113, doi:10.1029/2008JE003165. Carr M. E. 2000. IgPet for Windows. Somerset, NJ: Terra Softa Inc. McSween H. Y., Ruff S. W., Morris R. V., Bell J. F. III, Herkenhoff Dann J. C., Holzheid A. H., Grove T. L., and McSween H. Y. Jr. 2001. K. E., Gellert R., Stockstill K. R., Tornabene L. L., Squyres S. W., Phase equilibria of the Shergotty meteorite: Constraints on pre- Crisp J. A., Christensen P. R., McCoy T. J., Mittlefehldt D. W., eruptive water contents of Martian magmas and fractional and Schmidt M. 2006. Alkaline volcanic rocks from the crystallization under hydrous conditions. Meteoritics & Columbia Hills, Gusev crater, Mars. Journal of Geophysical Planetary Science 36:793–806. Research-Planets 111:E09S91. Dessai A. G., Markwick A., Vaselli O., and Downes H. 2004. Minitti M. E. and Rutherford M. J. 2000. Genesis of the Mars Granulite and pyroxenite xenoliths from the Deccan Trap: Insight Pathfinder “sulfur-free” rock from SNC parental liquids. into the nature and composition of the lower lithosphere beneath Geochimica et Cosmochimica Acta 64:2535–2547. cratonic India. Lithos 78:263–290. Miyashiro A. 1978. Nature of alkalic volcanic rock series. Dixon J. E., Stolper E. M., and Holloway J. R. 1995. An experimental Contributions to Mineralogy and Petrology 66:91–104. study of water and carbon dioxide solubilities in mid-ocean ridge Monders, A. G. et al. 2007. Phase equilibrium investigations of the basaltic liquids. Part I: Calibration and solubility models. Journal Adirondack class basalts from the Gusev plains, Gusev crater, of Petrology 36:1607–1631. Mars Meteoritics and Planetary Science 42:131–148. Féménias O., Coussart N., Bingen B., Whitehouse M., Mercier J. C., Nielsen C. H. and Sigurdsson H. 1981. Quantitative methods of and Demaiffe D. 2003. A Permian underplating event in late- to electron microprobe analysis of sodium in natural and synthetic post-orogenic tectonic setting. Evidence from the mafic- glasses. American Mineralogist 66:547–552. ultramafic layered xenoliths from Beaunit (French Massif Nekvasil H., Filiberto J., McCubbin F. M., and Lindsley D. H. 2007. Central). Chemical Geology 199(3–4):293–315. Alkalic parental magmas for chassignites? Meteoritics & Filiberto J. 2008. Experimental constraints on the parental liquid of Planetary Science 42:979–992. the Chassigny meteorite: A possible link between the Chassigny Nekvasil H., Dondolini A., Horn J., Filiberto J., Long H., and meteorite and a Martian Gusev basalt. Geochimica et Lindsley D. H. 2004. The origin and evolution of silica-saturated Cosmochimica Acta 72:690–701. alkalic suites: An experimental study. Journal of Petrology 45: Filiberto J. and Nekvasil H. 2003. Linking tholeiites and silica- 693–721. undersaturated alkalic rocks: An experimental study. GSA Newton R. C., Charlu T. V., and Kleppa O. J. 1974. Calorimetric Abstracts with Program 34:632. investigation of stability of anhydrous magnesium cordierite Floran R. J., Prinz M., Hlava P. F., Keil K., Nehru C. E., and with application to granulite facies metamorphism. Hinthorne J. R. 1978. Chassigny meteorite—Cumulate dunite Contributions to Mineralogy and Petrology 44:295–311. with hydrous amphibole-bearing melt inclusions. Geochimica et Norman M. D. 1999. The composition and thickness of the crust of Cosmochimica Acta 42:1213–1229. Mars estimated from rare earth elements and neodymium- Harrington A., McCubbin F. M., Elardo S., Nekvasil H., and Lindsley isotopics of Martian meteorites. Geochimica et Cosmochimica D. H. 2008. Irvine: A parent to Backstay? (abstract #2219). Acta 34:439–449. Proceedings, 39th Lunar and Planetary Science Conference. Norman M. D. 2002. Thickness and composition of the Martian crust CD-ROM. revisited: Implications of an undepleted mantle with a Nd Irvine T. N. and Baragar W. R. A. 1971. A guide to the chemical isotopic composition like that of QUE 94201 (abstract #1175). classification of the common volcanic rocks. Canadian Journal 33rd Lunar and Planetary Planet Science Conference. CD-ROM. of Earth Science 8:523–548. Ruff S. W., Christensen P. R., Blaney D. L., Farrand W. H., Johnson Johnson M. C., Rutherford M. J., and Hess P. C. 1991. Chassigny J. R., Michalski J. R., Moersch J. E., Wright S. P., Squyres, S. W. petrogenesis—Melt compositions, intensive parameters, and 2006. The rocks of Gusev crater as viewed by the Mini-TES water contents of Martian (?) magmas. Geochimica et instrument. Journal of Geophysical Research—Planets 111: Cosmochimica Acta 55:349–366. E12S18. Mandeville C. W., Webster J. D., Rutherford M. J., Taylor B. E., Scoates J. S. et al. 1999. Fractional crystallization experiments on a Timbal A., and Faure K. 2002. Determination of molar candidate parental magma to anorthosite (abstract #F1096), EOS absorptivities for infrared absorption bands of H2O in andesitic Transactions of the American Geophysical Union 80. glass. American Mineralogist 87:813–821. Stolz A. J. 1985. The role of fractional crystallization in the evolution McCubbin F. M. and Nekvasil H. 2008. Maskelynite-hosted apatite of the Nandewar Volcano, Northeastern New South Wales, in the Chassigny meteorite: Insights into late-stage magmatic Australia. Journal of Petrology 26(4):1002–1026. volatile evolution in Martian magmas. American Mineralogist Thompson R. N. 1975. Primary Basalts and Magma Genesis 2. Snake 93:676–684. River Plain, Idaho, USA, Contributions to Mineralogy and McCubbin F. M., Nekvasil H., and Lindsley D.H. 2007a. O(3) Petrology 52:213–232. occupancy of kaersutite crystallized at 1 atm and low oxygen Varela M. E., Kurat G., Bonin-Mosbah M., Clocchiatti R., and fugacity in a dry, F-bearing system. Accademia Nazionale dei Massare D. 2000. Glass-bearing inclusions in olivine of the Linking the Chassigny meteorite and the Martian surface rock Backstay 869

Chassigny achondrite: Heterogeneous trapping at sub-igneous Whitaker M. L., Nekvasil H., Lindsley D. H., and Difrancesco temperatures. Meteoritics & Planetary Science 35:39–52. N. J. 2007. The role of pressure in producing compositional Wadhwa M. and Crozaz G. 1995. Trace and minor elements in diversity in intraplate basaltic magmas. Journal of Petrology 48: minerals of nakhlites and Chassigny—Clues to their 365–393. petrogenesis. Geochimica et Cosmochimica Acta 59(17):3629– Zhou F. 1994. Ti-Mg-Fe biotites: Formation, substitution and 3645. thermodynamic properties at 650 °C to 900 °C and 1.1 kbar with

Wen S. and Nekvasil H. 1994. SOLVCALC: An interactive graphics fO2 defined by the Ch4-G buffer. Ph.D. thesis, State University of package for ternary feldspar solvus calculations and New York, Stony Brook, NY, USA. geothermometry. Computers in Geoscience 20(6):1025–1040.