Dust Coverage and Surface Effects on APXS Analyses
 Rocks analyzed by the APXS exhibit varying amounts of dust/soil coverage. This coverage means that rock compositions contain contributions from both dust and the underlying rock. The amount of contamination by dust may be estimated using SO3 contents (2.5–4.1 wt %; Table 2): i.e., dust has high S contents, but the rock itself is expected to be relatively S poor [e.g., Yen et al., 2005; Blake et al., 2013]. Consistent with this, Jake_Matijevic has the cleanest surface in MAHLI images and the lowest SO3 of the rocks reported here (2.5 wt %). SO3 contents suggest Rocknest_3 and Et_Then are similarly dusty (~4.1 wt %) and that Bathurst_Inlet has an intermediate dust cover. Chlorine correlates with sulfur in other Martian surface materials [e.g., Gellert et al., 2006], and Cl contents reported here are also likely in part contributed by surficial dust. In addition, the X-ray for the lightest elements detected by the APXS (Na, Mg, and Al) are largely contributed by the outer 2–3 µm of the target and thus dust coatings can block and/or contribute X-rays that combine with those from the underlying rock.
 To quantify the effects that increasing dust thickness covering a rock substrate has on APXS analysis, we use a new code devised by author Campbell for modeling X-ray yields from defined samples. The model requires input of rock substrate and dust cover composition, and we have chosen a terrestrial mugearite substrate from St. Helena [Kawabata et al., 2011] and the Portage soil GUAPX concentrations (Table 4) to represent these. The actual composition of dust coatings on Martian rocks is likely to be different from Portage; we use it simply as an example of the kinds of effects that can be observed. Plots of the computed X-ray yields (i.e., the relative numbers of X-rays emerging from the two-layer sample in the direction of the X-ray detector) are presented in Figure 3. In reality, dust thicknesses are not constant, but if we regard the present results (Figure 3) as coming from successive values of mean thickness, these plots provide a sense of the uncertainties associated with unbrushed rock targets. At zero layer thickness, the yields are “pure” and solely derived from the mugearite. The light-element concentrations approach Portage soil concentrations (decreasing Na and Al and increasing Mg) as dust reaches ~2 µm thicknesses (Figure 3a). Yields for K and Ca with more penetrating PIXE-induced X-rays are shown in Figure 3b; about 10 µm of dust cover is now needed to complete the change from rock to Portage response. Finally, the iron signal is only slightly affected by a dust cover in the 1–10 µm range; the transition from the St. Helena mugearite signal to the full dust cover response requires thickness >100 µm; this is the case, indeed more so, for all heavier elements (e.g., Ni, Zn, and Br).
Figure 3. Relative X-ray yields calculated as a function of modeled Portage soil cover thickness on a St. Helena mugearite rock substrate [Kawabata et al., 2011] using the modeling code of Campbell. This code predicts the X-ray yields expected from a two-layer (dust on rock) target with given element concentrations in each layer and dust thickness. APXS yield provides essentially the reverse process to the spectrum analysis code GUAPX [Campbell et al., 2012], which is used for spectrum fitting and conversion of peak areas (X-ray yields) to concentrations. The calculations are subject to the same assumptions and approximations as the GUAPX code. X-ray yields are not adjusted to reflect detection efficiency or absorption in the Martian atmosphere, nor are the empirical correction factors discussed in Campbell et al.  employed. For these reasons, and also because of the closure rule (that analyses must add to 100%), these results do not map directly to concentrations. (a) X-ray yields of the lighter elements (Na, Mg, Al, Si, and S) are compared to those of (b) the heavier elements K, Ca, and Fe.
Download figure to PowerPoint
 These results suggest caution in drawing strong conclusions regarding the lighter elements (particularly Na, Mg, and Al) when there is a dust cover on the order of a few micrometers. To take an extreme example, the Na/Mg ratio for the example shown (Figure 3a) is inverted by just 3 µm of dust. We have deliberately chosen for illustrative purposes an example where the light-element concentrations of dust and substrate are very different, and in some combinations, (e.g., where the rock and dust compositions are more similar), the effects will be less dramatic. Without knowing surface layer thickness or rock composition, it is difficult to quantify the effects that dust coverage has on unbrushed rock surfaces. This problem is being addressed by ongoing and future studies [e.g., Berger et al., 2013].
 To correct for dust coverage, subsequent plots assume S- and Cl-free rock surfaces (renormalized to 100%) and, given the uncertainties involved, do not correct the light elements (Table 7). We acknowledge the limitations to this correction, particularly if the rocks themselves contain S or Cl or if the dust cover and rock are very different in light-element concentrations. In addition, we note that Martian magmas are relatively enriched in S and Cl relative to terrestrial ones [e.g., Dreibus and Wänke, 1985; Filiberto and Treiman, 2009; McCubbin et al., 2013; Righter et al., 2009]. Adjustments to whole-rock compositions made by this correction are relatively minor (factor of 1.04 to 1.1 increase).
Table 7. APXS Concentrations of Bradbury Assemblage Rocks Recalculated Volatile-Freea
| ||(wt %)|| || || || || || || || || || ||(ppm)|| || || |
|Jake_Matijevic|| || || || || || || || || || || || || || || |
|Bathurst_Inlet|| || || || || || || || || || || || || || || |
 Compositions of the first four rocks and one soil examined by the APXS in Gale Crater as determined by the empirical (Gellert method) and fundamental parameter (GUAPX) approaches to reducing the spectra are presented in Tables 2 and 4, respectively. Figure 4 compares the compositions determined for the nighttime JM2 analysis by the two methods. Differences in concentration are mostly less than 10% for good quality nighttime spectra (cold temperatures and long durations), such as JM2, with the exception of certain elements (e.g., Mg and P), which are the subject of ongoing investigation. Differences in the results for the two lightest elements (Na and Mg) become larger for poorer quality daytime spectra (short durations and high temperatures), reflecting the increasing peak overlap caused by worsening energy resolution. The closeness of most concentrations for the long-duration, nighttime measurements validates the two methods when the energy resolution is good.
Figure 4. Bar graph comparing the APXS JM2 overnight elemental abundances in weight percent oxide as determined by the Gellert and GUAPX methods [Gellert et al., 2006; Campbell et al., 2012]. Statistical error bars are less than the width of the line where not visible. Inset graph presents minor element concentrations at a different scale.
Download figure to PowerPoint
 For the purposes of plotting the APXS results, we use the compositions determined by the Gellert method in order to better compare with the MER APXS data sets. We emphasize that neither method has been demonstrated to be more accurate than the other (although precision is well understood) and that our geological interpretations do not depend on which data set is used. Given the high degree of precision for both methods, variability outside the stated errors (Tables 3 and 5) between APXS targets on the same rock (e.g., MnO in Bathurst_Inlet) reflects heterogeneities in or at the surface of the rock, such as dust cover, proportions of mafic/felsic minerals, and/or mineral effects [Campbell et al., 2012].
 Previous missions have used APXS results to develop classification schemes by grouping similar rock types [e.g., Squyres et al., 2006]. However, the rocks reported on here are geochemically diverse and sufficiently distinct from one another to warrant each defining their own class. Given the clastic textures and degree of uncertainty about the geologic context for these rocks, igneous classification schemes should be considered as providing only general guidance about the compositions. Accordingly, we plot volatile-free Bradbury assemblage compositions on a total alkali versus silica diagram (Figure 5) such as has become common for relating different Martian rock compositions [e.g., McSween et al., 2009]. Soil contamination would probably pull the rock compositions in the general direction of the Portage soil with relatively low SiO2 and total alkalis. Of the rocks examined, Jake_Matijevic is notably alkaline and evolved, plotting in the phonotephrite/tephriphonolite field in the total alkali versus silica diagram (Figure 5), but is classified as a mugearite because it is similar to terrestrial mugearites in most MgO and SiO2 variation diagrams [Stolper et al., 2013]. The other three rocks are also alkali-rich with Bathurst_Inlet, Et_Then, and Rocknest_3 plotting within the basalt and hawaiite fields and near alkali-rich basalts found in Gusev Crater (Humboldt_Peak and Wishstone). Apart from the alkalis and certain trace elements, these three rocks are otherwise generally similar to the SNC meteorites, particularly in having high Fe and low Al contents.
Figure 5. Total alkali versus silica diagram with igneous classification scheme after Le Bas et al. . Compositions of Martian rocks (volatile-free) determined by APXS plotted include the first four rocks and Portage soil in Gale crater, the relatively unaltered basalts and hydrothermally altered Home Plate rocks from Gusev crater, the Shoemaker formation and Bounce Rock from Meridiani Planum, and the Pathfinder rocks [Ming et al., 2008; Gellert et al., 2006; Brückner et al., 2003; Squyres et al., 2012; Zipfel et al., 2011]. The Martian SNC meteorites are also shown [Meyer, 2013; Agee et al., 2013]. The 2σ standard errors (Table 3) are generally smaller than the symbols.
Download figure to PowerPoint
 Although only four rocks were examined as of sol 102, the Bradbury assemblage spans nearly the entire Martian meteorite and mission data set in abundances of certain elements, particularly Fe and Mn. The MSL APXS measured the highest Fe (26.3 wt % FeO* in Et_Then) and Mn (0.78 wt % MnO in BI_For_Real) thus far measured by APXS for Martian basaltic rocks (Figure 6) with a range of 28 to 68 FeO*/MnO. CaO/Al2O3 ratios of the Bradbury assemblage are subchondritic and are in this respect like most Gusev rocks, but the Bradbury rocks cover the whole range of Al2O3 found in relatively unaltered Gusev basalts (Adirondack, Backstay, Irvine, and Humboldt Peak; Figure 6b).
Figure 6. Geochemical variation diagrams for Martian rock compositions, including the Martian SNC meteorites and rock compositions determined by APXS from Gale crater (Jake_Matijevic, Bathurst, Et_Then, and Rocknest_3) and Gusev crater (Home Plate, Adirondack, Wishstone, Backstay, Irvine, and Humboldt Peak). The Portage soil from Gale Crater represents the Mars global soil [Blake et al., 2013]. Compositions are plotted as weight percent and assumed to be volatile-free, except where noted. (a) A plot of CaO versus Al2O3 with line denoting the chondritic ratio of CaO/Al2O3 [Sun and McDonough, 1989]. (b) MnO versus total Fe as FeO*. (c) K2O versus Na2O. Estimated Nahkla primary melt compositions are from Goodrich et al. [2012, 2013]. Estimated bulk Mars and bulk Mars crust compositions are from [Wänke and Dreibus, 1988] and [Taylor and McLennan, 2009], respectively. (d) Zn versus K2O. (e) Plot of K2O versus Cl (not volatile-free) includes the Adirondack class rock Mazatzal, whose KCl-rich rind was progressively abraded by the Rock Abrasion Tool [Ming et al., 2008; Gellert et al., 2006; Brückner et al., 2003; Squyres et al., 2012; Zipfel et al., 2011; Meyer, 2013; Agee et al., 2013].
Download figure to PowerPoint
 While diverse in many respects, the Bradbury assemblage rocks also share certain characteristics, in particular high abundances of potassium (1.6–2.9 wt %; Figure 6c). Moderately volatile metals (Zn and Ge) are also high in some targets; 1210 to 1330 ppm Zn is found among the Bathurst_Inlet targets (Figures 6d and 7c), and 102 ppm Ge is found in Rocknest_3 (Table 2). Concentrations of K and Zn do not correlate with halogen or S contents, as illustrated in a plot of K2O versus Cl (Figure 6e), suggesting that these moderately volatile metals are not associated with salts, in contrast to some rock analyses from other landing sites (e.g., Mazatzal [Squyres et al., 2004b; Morris et al., 2008]). Individual rocks from the Home Plate region of Gusev Crater region that are similarly enriched in K and Zn (e.g., FuzzySmith and Montalva [Ming et al., 2008]) show evidence of being hydrothermally altered; FuzzySmith has high SiO2 concentrations and Mössbauer spectrometer observations suggest that it likely contains marcasite or pyrite, and Montalva has very high Fe3+/FeT (0.93 [Morris et al., 2008]).
Figure 7. Geochemical variation diagrams for the Bradbury assemblage and Portage soil determined by APXS. Concentrations are in weight percent, unless noted. (a) Al/Si versus (Fe + Mg)/Si. (b) MgO versus FeO*. (c) Zn versus SiO2. (d) Cr2O3 versus FeO*.
Download figure to PowerPoint
 The rock Jake_Matijevic is the most likely igneous rock examined so far in Gale Crater, and its chemistry and a potential scenario for its petrogenesis are described in detail by Stolper et al. . Jake_Matijevic is more fractionated than most known Martian igneous rocks, having low Mg and Mg# of 44.6–47.4 (Mg# is molar MgO/(FeO + MgO) × 100; assuming Fe3+/FeTotal = 0.15). Its Ni and Cr contents are also low (below detection to 59 ppm Ni and 225–583 ppm Cr; Figure 7d); the Ni is, in fact, the lowest found to date for an unbrushed rock target on the surface of Mars. The high Na2O concentrations (6.6–7.1 wt %) of Jake_Matijevic are also remarkable relative to other Martian compositions, particularly given its dusty surface. The high alumina and alkali concentrations make Jake_Matijevic a truly alkaline composition rock with 16% normative nepheline (assuming Fe3+/FeT = 0.15; Figure 8 and Table 8).
Figure 8. Bar graph portraying the CIPW normative mineral contents in weight percent of representative Bradbury assemblage APXS analyses. (pl = plagioclase, or = orthoclase, ne = nepheline, di = diopside, hy = hypersthene, ol = olivine, il = ilmenite, mt = magnetite, ap = apatite, and cm = chromite).
Download figure to PowerPoint
Table 8. CIPW Normative Mineralogy of the Bradbury Assemblagea
| ||Jake_Matijevic (JM2) c||BI_For_Real||ET_Then||Rocknest_3|
|Nepheline||16.0||15.2|| || || || ||2.8||1.3|
|Hypersthene|| || ||0.2||0.2||22.0||28.8|| || |
|100 × Mg/(Mg + Fe2+) in rock||47||50||45||48||25||27||38||41|
|100 × Mg/(Mg + Fe2+) in silicates||51||56||48||54||27||32||41||46.|
|100 × Ca/(Ca + Na) in rock||36||36||61||61||45||45||46||46|
|Plagioclase An content||14||14||20||20||16||16||14||13|
 In contrast, Bathurst_Inlet is more mafic and is much richer in Fe (20.8–21.9 wt % FeO*) than Jake_Matijevic. Bathurst_Inlet ranges to the highest K2O of the Bradbury assemblage, and it is relatively Al poor (7.9–8.0 wt % Al2O3) and slightly hypersthene normative (Figure 8 and Table 8). There is uncertainty as to whether Bathurst_Inlet represents an igneous composition, but even if so, its relatively low Mg# (45.2–46.9) and silica saturation with respect to normative mineralogy indicate that it could not be parental to Jake_Matijevic by fractional crystallization at low pressures. Slightly higher FeO* found in the BI_For_Real than in BI_Top correlates with higher concentrations of transition metals (beyond stated errors), including Ti, Cr, Ni, Zn, and especially Mn (Figures 6b and 7d). These elements are compatible in oxide minerals, such as magnetite, and may suggest one of these is present in this rock.
 Rocknest_3 is intermediate in composition between Jake_Matijevic and Bathurst_Inlet in most variation diagrams (Figures 6 and 7). Although not as extreme as Jake_Matijevic, Rocknest_3 is silica undersaturated with ~3% normative nepheline (Figure 8). In most element-element variation diagrams as well as in a plot of Al/Si versus (Fe + Mg)/Si, Rocknest_3 plots along a line joining Jake_Matijevic and Bathurst Inlet, suggestive of a two-component mixing relationship (Figures 6 and 7) that likely reflects mechanical mixing by sediment transport. Exceptions from the mixing trend include Cl, P, Ni, Ge, and MgO versus FeO (Figure 7b) and may imply that multiple processes (e.g., dust coverage effects on light-element concentrations or dissolution of olivine or other mineral phases) have contributed to the Rocknest_3 composition.
 The float rock Et_Then is distinct from the other rocks examined by the APXS, having very high FeO* (26.3 wt %) and low MgO (4.2 wt %) contents (Mg# = 25; Figure 7b). The high FeO* in Et_Then does not correlate with elevated Mn, Ti, or Cr, in contrast to Bathurst_Inlet. Et_Then is the only rock of the Bradbury assemblage that is strongly hypersthene normative (22%; Figure 8), although we note that the high Fe with low Mn and Ti are likely not reflective of an igneous bulk composition. More probably, aqueous alteration involving the addition of Fe contributed to the Et_Then composition.
 In order to evaluate the relative proportions of primary phases in the Bradbury assemblage rocks, we plot molar proportions of FeO* + MgO, Al2O3, and CaO + Na2O + K2O on a ternary diagram (Figure 9). Felsic and mafic igneous minerals plot at distinct locations on this diagram. The position of Jake_Matijevic is consistent with a composition that is relatively enriched in aluminosilicate minerals, while Bathurst is consistent with a composition that is more enriched in ferromagnesian minerals. Rocknest_3 plots at an intermediate position and supports the suggested mixing relationships.
Figure 9. Ternary diagram representing the molar proportions of Al2O3, CaO + Na2O + K2O, and MgO + FeO* of the Bradbury assemblage. The important mineral (feldspar, pyroxene, olivine, and Fe oxide) groups and the direction that terrestrial weathering pulls compositions are indicated for reference. See text for discussion. The field for Home Plate (Gusev) includes all basaltic layered rocks in the vicinity of the Home Plate outcrop. Data are from Ming et al.  and the Planetary Data System.
Download figure to PowerPoint
 Figure 9 is also useful for examining the effects of moderate pH subaerial weathering [Nesbitt and Wilson, 1992]: as igneous minerals are chemically weathered, they tend to lose soluble cations (Na, K, Ca, and Mg) in preference to less soluble cations (Al and ferric iron), causing progressive enrichment in Al and Fe, evolving on a trend perpendicular to the tie line between feldspar and the FeO* + MgO apex. Previous in situ and experimental studies have thus used this diagram to distinguish between terrestrial and Martian-style chemical weathering, which is dominated by olivine dissolution processes [e.g., Hurowitz et al., 2006]. By this approach, the Bradbury assemblage rocks do not display any noticeable terrestrial weathering trend. Chemical or diagenetic processes such as olivine dissolution or the addition of Fe (or Mn-rich) oxides as a coating or cement would lead to trends away from or toward the FeO* + MgO apex and are consistent with observed variations. Because compositions of the Bradbury assemblage can be interpreted as being essentially mixtures of igneous minerals, these rocks are interpreted to be of volcanic/volcaniclastic origin with little subsequent leaching of soluble cations by water, or if they are sedimentary rocks, their sources underwent at most only incipient subaerial weathering with little attendant modification of primary bulk igneous compositions [McLennan et al., 2013].