Composition of the Rheasilvia basin, a window into Vesta's interior
Abstract
[1] The estimated excavation depth of the huge Rheasilvia impact basin is nearly twice the likely thickness of the Vestan basaltic crust, so the mantle should be exposed. Spectral mapping by the Dawn spacecraft reveals orthopyroxene-rich materials, similar to diogenite meteorites, in the deepest parts of the basin and within its walls. Significant amounts of olivine are predicted for the mantles of bulk-chondritic bodies like Vesta, and its occurrence is demonstrated by some diogenites that are harzburgite and dunite. However, olivine has so far escaped detection by Dawn's instruments. Spectral detection of olivine in the presence of orthopyroxene is difficult in samples with <25% olivine, and olivine in Rheasilvia might have been diluted during impact mixing or covered by the collapse of basin walls. The distribution of diogenite inferred from its exposures in and around Rheasilvia provides a geologic context for the formation of these meteorites, but does not clearly distinguish between a magmatic cumulate versus partial melting restite origin for diogenites. The former is favored by geochemical arguments, and crystallization in either a magma ocean or multiple plutons emplaced near the crust-mantle boundary is permitted by Dawn observations.
Key Points
- Excavation of Rheasilvia has exposed Vesta's mantle
- Mantle consists of diogenite
- No olivine has been detected
1 Introduction
[2] The huge Rheasilvia impact basin (~500 km diameter, with a transient, pre-collapse diameter of 325–430 km) [Schenk et al., 2012] provides a unique perspective into the interior of asteroid 4 Vesta, potentially revealing rocks that formed during the body's early magmatic differentiation [Drake, 2001; Keil, 2002; McSween et al., 2011]. The depth of basin excavation is estimated at 30–45 km based on hydrocode simulations [Jutzi and Asphaug, 2011; Ivanov and Melosh, 2012]. The loss of this much material, when added to the earlier excavation caused by the underlying Veneneia basin and the erosion of Vesta's surface since its formation [Turrini, 2012], must have exposed the deep crust and possibly the mantle, based on crustal thickness estimates.
[3] A large south polar basin was first discovered using Hubble Space Telescope (HST) observations of Vesta, which formed the basis for a preliminary shape model [Thomas et al., 1997] and for spectroscopic interpretations of ultramafic (possibly mantle) rocks around or within the crater [Binzel et al., 1997; Li et al., 2010; Reddy et al., 2010]. The impact that formed this basin, now named Rheasilvia, is thought to have produced the Vestoids—smaller (≤10 km) asteroids dynamically linked to Vesta, in orbits between Vesta and the 3:1 Jovian and ν6 secular resonances [Binzel and Xu, 1993; Marzari et al., 1996]. Because these resonances act as escape hatches from the main belt and provide trajectories into the inner solar system, the Vestoids were likely the immediate parent bodies for most howardite-eucrite-diogenite (HED) meteorites [Migliorini et al., 1997]. Indeed, the reflectance spectra of Vesta [McCord et al., 1970; Gaffey, 1997] and of Vestoids [Binzel and Xu, 1993; Vilas et al., 2000; Burbine et al., 2001; Kelley et al., 2003; Shestopalov et al., 2008; Moskovitz et al., 2010; De Sanctis et al., 2011a; Mayne et al., 2011; Reddy et al., 2011] are very similar to those of howardites, eucrites, and diogenites. Thus, there is likely to be a corresponding linkage between the HED meteorites and Rheasilvia, a connection to be explored in this paper.
[4] The Dawn spacecraft carries three instruments that have provided compositional information on the Rheasilvia basin. The Framing Camera (FC) [Sierks et al., 2011] allowed for geological observations and limited (seven color filters between 0.4 and 1.0 µm) spectral measurements for mapping at high spatial resolution (61 m/pixel at the High-Altitude Mapping Orbit HAMO, ~16 m/pixel at the Low-Altitude Mapping Orbit LAMO). In this paper, an FC global map utilized data from HAMO. The Visible and Infrared Spectrometer (VIR) [De Sanctis et al., 2011b] measured reflectance spectra between 0.25 and 5.1 µm, a range that contains diagnostic mafic mineral absorption bands. Its nominal spatial resolution varies from 700 m/pixel during the Survey orbit, 200 m/pixel during HAMO, and 70 m/pixel during LAMO. The Gamma Ray and Neutron Detector (GRaND) [Prettyman et al., 2011] measured the leakage flux of neutrons during LAMO, with a mean altitude of 210 km [Prettyman et al., 2012]. At this altitude, the spatial resolution of GRaND was about 300 km full-width-at-half-maximum arc length on the surface. Thus GRaND is sensitive to compositional variations on a scale somewhat smaller than the Rheasilvia basin. Together, these instruments constrain the composition of the Rheasilvia basin and allow comparisons with the mineralogy and chemistry of HEDs.
2 Diogenites
2.1 Petrology, Compositions, and Ages
[5] As documented below, Rheasilvia is dominated by diogenite, so here we summarize diogenite compositions (focusing on features that can affect Dawn spectra). Several classes of Vestan coarse-grained ultramafic rocks, collectively called diogenites, have been recognized (Figure 1). Orthopyroxenitic diogenites (the most common type) are composed of >90 vol.% orthopyroxene. Harzburgitic diogenites (rare) are mostly orthopyroxene but with >10 vol.% olivine [Beck and McSween, 2010] with most having ~10–25 vol.% olivine [A. Beck et al., 2012a]. A dunitic diogenite, Miller Range (MIL) 03443, contains >90% olivine [Beck et al., 2011a] and allows the possibility of the full range of orthopyroxene-olivine mixtures. Henceforth, we refer to these diogenite types simply as orthopyroxenite, harzburgite, and dunite. Minor or accessory minerals in the diogenites are high-Ca clinopyroxene, chromite, troilite, metal, silica, and plagioclase. Most diogenites are breccias, so some clinopyroxene and plagioclase in them could reflect admixture of eucrite (basalt). Brecciated mixtures of orthopyroxenite and harzburgite are also fairly common [Beck and McSween, 2010].
[6] Diogenitic pyroxenes show narrow ranges of composition, resulting from metamorphic equilibration [Mittlefehldt, 1994], but the pyroxene compositions differ from meteorite to meteorite. Pyroxenes in harzburgites are generally Mg-rich with Mg#s (molar Mg/[Mg + Fe]*100) = 75–78, relative to pyroxenes in orthopyroxenites with Mg# = 70–74 [Beck and McSween, 2010], although some overlap has been observed. Pyroxenes in harzburgites have equilibrated with the coexisting olivines (Fo62-74). The more ferroan pyroxenes in orthopyroxenites typically have higher concentrations of minor elements (Al, Cr, and Ti) than magnesian pyroxenes. The compositions of pyroxene (Mg# = 77) and olivine (Fo74) in dunite [A. Beck et al., 2011] are within the ranges for harzburgite.
[7] Reported element abundances in bulk diogenites [Mittlefehldt, 1994; Barrat et al., 2008; Warren et al., 2009; Mittlefehldt et al., 2012] should be viewed with some caution, as analyzed samples of coarse-grained meteorites may not always be representative of the bulk meteorite. However, elements can distinguish between some types of HEDs. In Figure 2, we show Fe abundances (meteorite data compiled in Supplementary Online Materials for Prettyman et al. [2012]. The average wt.% Fe content for diogenites (13.2 ± 1.5) is lower than for typical upper crustal rocks (14.8 ± 0.7 for basaltic eucrites, 14.5 ± 0.7 for polymict eucrites, 13.9 ± 0.9 for howardites), but comparable to rocks that formed deeper within the crust (12.5 ± 1.7 for cumulate eucrites). Within the diogenites, harzburgites and dimict breccias composed of harzburgite + orthopyroxenite are not distinguished in Fe content from the orthopyroxenites (histogram in Figure 2). The spread in olivine and pyroxene Mg#s in harzburgite clasts within diogenite breccias [Beck and McSween, 2010] indicates a considerable compositional range. The Fe content of the MIL 03443 dunite (19.7 wt.%) is significantly higher than for other diogenites (Figure 2).
[8] Analyses of the decay products of the short-lived radioisotopes 26Al [Schiller et al., 2011] and 53Mn [Trinquier et al., 2008; Day et al., 2012] indicate diogenites formed within the first 2–4 Myr of Solar System history. Early crystallization is consistent with constraints from other long-lived radionuclides for the rapid differentiation of Vesta [Wadhwa et al., 2006], although Schiller et al. [2011] suggested that such rapid differentiation and cooling were inconsistent with diogenite formation on an asteroid as large as Vesta.
2.2 Formation Mechanisms
[9] Two different mechanisms could plausibly have formed the diogenites (Figure 3a). They might be solid residues from partial melting of the mantle [Stolper, 1975; Sack et al., 1991] or cumulates formed by magma fractionation [Fowler et al., 1994; Mittlefehldt, 1994; Warren, 1997; Ruzicka et al., 1997; Barrat et al., 2008; Beck and McSween, 2010]. Let us consider each mechanism in turn.
[10] Extraction of basaltic (eucrite) magma, having a eutectic composition (the apex of arrows in Figure 3b) [Stolper, 1975], from a chondritic mantle source (gray area in Figure 3b) can form harzburgite, orthopyroxenite, or dunite residua, depending on the degree of partial melting, and harzburgite is a common partial melting residue on Earth. However, forming both orthopyroxenite and dunite by this mechanism would require a heterogeneous mantle, because olivine and orthopyroxene cannot both be exhausted by continued melting of the same residue. Other arguments against the restite model are based on major element mass-balance [Warren, 1997], trace element abundances [Barrat et al., 2008], oxygen isotope homogeneity [Greenwood et al., 2005], and overlap in pyroxene Mg#s between diogenite lithologies [Beck and McSween, 2010]. The occurrence of live 26Al in diogenites [Schiller et al., 2011] may also argue for pervasive rather than limited partial melting. For all these reasons, this model has generally fallen out of favor, although some of these constraints might be mitigated in the case of a heterogeneous Vestan mantle.
[11] Fractionation crystallization of magma(s) (arrow in Figure 3c), as advocated by Ruzicka et al. (1997) and Warren (1997), can produce both dunite and orthopyroxenite cumulates. Harzburgite could form by crystallization of both minerals along the olivine-pyroxene cotectic, provided that the pressure was high enough [Bartels and Grove, 1991]. Righter and Drake (1997) favored equilibrium crystallization on Vesta, made possible by suspension of crystals in convecting magma, followed by fractional crystallization when convective lockup occurred once the crystal fraction reached a high value. Transitioning from equilibrium crystallization to fractionation to form olivine or orthopyroxene cumulates might be possible at a later stage. Forming diogenites as magmatic cumulates requires either ultramafic parent magma or the production of significant amounts of complementary gabbroic (plagioclase-bearing) rocks (which have not been found in the HED collection, although magmas complementary to the cumulates might have preferentially erupted).
[12] The diogenites could have crystallized in a magma ocean [Righter and Drake, 1997; Ruzicka et al., 1997; Takeda, 1997; Warren, 1997; Drake, 2001; Greenwood et al., 2005; Schiller et al., 2011] or in multiple, smaller magma chambers [Shearer et al., 1997; Barrat et al., 2008; Beck and McSween, 2010]. Calculations by Wilson and Keil [2012] indicate that magmas in the mantles of asteroid-size bodies would be removed very efficiently, so that only small amounts of melt would be present at any time, precluding a magma ocean. Moreover, the dikes that transport magmas to the surface would be unstable, leading to accumulation of magmas in chambers at the base of the crust [Wilson and Keil, 2012]. In a magma ocean, olivine should have crystallized before pyroxene, but some harzburgites have Mg#s within the range of orthopyroxenites [Beck and McSween, 2010], and based on their 26Al contents [Schiller et al., 2011], the harzburgites did not consistently crystallize earlier than orthopyroxenites. Moreover, trace element abundances in bulk diogenites cannot be explained by crystallization from a common magma composition [Mittlefehldt, 1994; Shearer et al., 1997; Barrat et al., 2008], suggesting multiple magma chambers having different compositions. Separate plutons are also supported by the wide range in the Mg#s of pyroxenes and olivines among different harzburgites [Beck and McSween, 2010]. Dunite shows strong mineralogical similarities to the harzburgites and has also been interpreted as an olivine cumulate [A. Beck et al., 2011]. Based on the magnitude of europium anomalies, Barrat et al. [2008, 2010] argued that diogenite formed from magmas generated by remelting lower crustal cumulates from a magma ocean, but Mittlefehldt et al. [2012] offered an alternative hypothesis based on later diffusive exchange of europium between minerals.
3 Predicted Crust and Mantle Stratigraphy and Compositions
[13] Vesta's upper crust is basaltic, as revealed by the dominant eucrite component in the parts of the global regolith not blanketed by Rheasilvia ejecta [De Sanctis et al., 2012]. In the magma ocean model, the upper crust would overlie a lower crust or mantle of diogenite. In the serial magmatism model, multiple diogenite plutons could have been emplaced within the lower crust or near the crust-mantle boundary. Some insight into the depth of diogenite crystallization is provided by ordering of Fe and Mg in the pyroxene crystal structure, which is affected by cooling rate and thus burial depth. Cooling rate estimates for diogenitic pyroxenes [Zema et al., 1997] are generally faster than that for a clearly crustal cumulate eucrite determined using the same method [Domeneghetti et al., 1995], suggesting a crustal origin for diogenites. However, Fe site occupancy data [Verma et al., 2008] may suggest even slower cooling for diogenites, perhaps at mantle depths. Study of lattice preferred orientations in diogenite olivines reveals plastic deformation like that of terrestrial mantle rocks, supporting their formation at mantle depths [Tkalcec et al., 2013].
[14] Estimates of the thickness of the Vestan crust assume that it has a basaltic composition. These estimates are not well constrained, but are generally less than Rheasilvia's excavation depth. A chondritic bulk asteroid would, if fully melted, produce a eucritic crust ~15–20 km thick if porosity-free [Toplis et al., 2012]. Magma ocean models based on HED meteorites posit eucritic crusts ranging in depth from 10–15 km [Righter and Drake, 1997] to 23–42 km [Ruzicka et al., 1997]; the highest values are for a bulk Vesta composition (enstatite chondrite) that is probably unreasonable [Toplis et al., 2012]. Miyamoto and Takeda [1994] estimated a minimum crustal thickness of ~15 km, from cooling rates based on exsolution lamellae in eucrite pyroxenes. Spectral studies of Vestoids have identified eucrite chunks as large as 6–10 km [Binzel and Xu, 1993; Mayne et al., 2011]. Taking all these constraints into account, we suggest that a plausible thickness of the Vestan crust is 15–20 km. This is about half the estimated 30–45 km excavation depth of Rheasilvia [Jutzi and Asphaug, 2011; Ivanov and Melosh, 2012], implying that the mantle should be exposed.
[15] Studies of HEDs indicate that the bulk composition of Vesta is approximately chondritic [Dreibus and Wanke, 1980; Warren, 1983; McSween et al., 2011]. Petrogenetic models for Vesta typically invoke large amounts of olivine in the bulk asteroid, because plausible chondritic precursors contain high abundances of that phase. Carbonaceous chondrite bulk models are dominated by olivine (>85%), and ordinary chondrite models have roughly equal amounts of olivine and orthopyroxene [Toplis et al., 2012]. An HED model based on mixing carbonaceous and ordinary chondrites [Righter and Drake, 1997] satisfies most geochemical constraints. Consequently, the mantle of differentiated Vesta should contain significant (>50%) olivine.
[16] Previous possible interpretations of olivine in Vesta's interior were based on a 1 µm band (henceforth called BI) that is broader and deeper than for the rest of Vesta in terrains interpreted to be Rheasilvia ejecta [Binzel et al., 1997; Gaffey, 1997; Thomas et al., 1997]. High-Ca pyroxene might make the 1 µm band appear broader, but significant amounts of high-Ca pyroxene have not been found in harzburgites, and none of the spectra of 15 Vestoids examined by Mayne et al. [2011] show 1 µm broadening. Harzburgite and dunite in the HED collection confirm the occurrence of olivine, usually in proportions varying from <30 to >90 vol.%.
4 Geologic Mapping of Rheasilvia From FC Data
[17] Eucrites and diogenites can be distinguished in the 0.44–0.98 µm spectral range by the approximate depth of the BI low-calcium absorption feature; the orthopyroxene-rich diogenites have a deeper BI feature than eucrites [Reddy et al., 2011]. In order to identify diogenite- and eucrite-rich terrains using HAMO FC data, color images were processed using methods outlined by Reddy et al. [2012], and then color data were used to calculate Clementine color ratios where red = 0.75/0.45 µm, green = 0.75/0.92 µm, and blue = 0.45/0.75 µm (Figure 4). Green areas comprised material with deeper 1 µm absorption features and are likely more diogenitic. Although the relative abundance of eucritic material cannot be estimated directly in this narrow spectral range, given that Vestan geology broadly consists of only two end-member components, we can assume areas that appear less diogenitic are more eucritic. The general correlation between diogenite- and eucrite-rich areas identified by FC Clementine color ratios in previous orbital phases [Reddy et al., 2012] with those identified using hyperspectral VIR data [DeSanctis et al., 2012] further validates this method for mapping compositional terrains on Vesta.
[18] The base of the central uplift in Rheasilvia appears to be exposed at its greatest depth in several locations. The first, which also corresponds to the deepest portion of the basin, occurs on the southern flank of the uplift between ~60°–80°S, 100°–225°E, and is ~22 km deep, relative to a 285 × 229 km ellipsoid. A second slightly shallower exposure ~20 km deep occurs on the northern flank of the uplift at ~60°–80°S, 300°–330°E (Figure 4a). These areas are also distinct in that they have abundant material with deeper 1 µm absorption features, as defined by the 0.75/0.92 µm ratio, an indication that they are diogenite-rich (Figure 4b). Other less well-exposed portions of the base of the uplift show shallower ratios. Stratigraphically higher sections of the central uplift appear to be mixtures of material with 0.45/0.75 µm signals and significantly less diogenite than is observed at the deeply excised base locations. Localized areas on the crater floor and wall appear enriched in diogenite as well. The highest concentration of diogenite material along the wall corresponds to the section of the wall with the highest relief, Matronalia Rupes, located in the upper right quadrant of Figure 4a at ~50°S, 45°–90°E. Note that the high-relief terrain in the lower left quadrant of Figure 4a may be ejecta material [Marchi et al., 2012] or may represent a part of the contour of the basin rim associated with ejecta deposits. Because the portion of the wall near Matronalia Rupes displays the highest relief, it likely preserves exposures to the greatest relative depth.
5 Mineralogy of Rheasilvia From VIR Data
[19] The VIR spectra of Vesta's surface show ubiquitous absorption bands at both 0.9 (BI) and 1.9 µm (BII), confirming the widespread occurrence of iron-bearing low-calcium pyroxenes [De Sanctis et al., 2012]. Band depths and centers have been computed for BI and BII. The variations of these parameters and their distributions across Vesta allow the identification and localization of compositional units. The earliest spectroscopic observations by Dawn showed a clear dichotomy: within the Rheasilvia basin, the pyroxene bands are, on average, deeper than in the equatorial region. BI depths within Rheasilvia are commonly 0.45–0.55 (relative to the continuum, as defined by Clark and Rousch [1984]) compared to 0.35–0.40 in the equatorial region, and BII depths in Rheasilvia are 0.25–0.30 compared to 0.15–0.20 at the equator [De Sanctis et al., 2012]. These observations demonstrate that the composition/abundance of pyroxene in the Rheasilvia basin is different from the equatorial region, and provide a means of comparing spectra from FC and from laboratory measurements of HEDs.
[20] The VIR BI and BII center positions (Figure 5) are especially useful in identifying diogenite and eucrite. Some localized regions in Figure 5 have BI and BII centers shifted to shorter wavelengths with respect to the average values. This implies the presence of pyroxenes with higher Mg# [Klima et al., 2007]. As noted in 6, two of these regions are located well inside the Rheasilvia basin and are associated with some of the lowest topography on the Vestan surface [Jaumann et al., 2012]. These are presumably outcrops of the deep crust or mantle exposed at the base of Rheasilvia's central uplift [Reddy et al., 2012]. A third, spectrally similar region is located on Rheasilvia's rim, in a topographically higher region, suggesting the occurrence of rocks excavated from depth during the impact event. Thus, there is excellent agreement between diogenite occurrences mapped by VIR and FC.
[21] As noted in 2, diogenites have pyroxenes with higher Mg#s than eucrites. The resulting difference in spectra can be quantified using a scatterplot of BI center versus BII center positions (Figure 6a). We used spectra in the RELAB database to define the different HED spectral areas. In this diagram, diogenites and eucrites populate distinct areas because both BI and BII center positions are sensitive to the pyroxene compositions. Howardites, which are physical mixtures of diogenite and eucrite, sensibly plot between and partly overlap these fields. Superimposed on the HED fields in Figure 6a is a cloud of Survey-orbit VIR spectra from the southern hemisphere below latitude −30°. Associating a color indication of composition for every region in the scatterplot (red for diogenite, green for howardite, purple for eucrite, with overlapping fields of yellow for diogenite-howardite and cyan for eucrite-howardite), we constructed the plot in Figure 6b and the correspondence map using the same color scheme in Figure 6c. From this map, it is clear that all the regions with high-Mg# pyroxenes (red and yellow) correspond to a diogenite-dominated lithology within Rheasilvia and that regions outside Rheasilvia generally have pyroxene compositions that are eucrite-dominated. One isolated diogenitic area lies outside the rim near 40°S latitude, 90° longitude; this area, probably ejecta, was not recognized in the FC map (Figure 4b). Black points in Figure 6b are unclassified; these few points are distributed near 30°S latitude, 90°–135° longitude and correspond to low signal/noise ratio spectra.
[22] The VIR has also mapped diogenite at higher spatial resolution. A VIR image of Matronalia Rupes (Figure 7) was acquired during HAMO with a nominal resolution of 200 m/pixel. This image is an RGB combination of three wavelengths: 1.23, 1.91, and 1.96 µm. In this combination, BII center variations are enhanced, and high and low Mg# areas (diogenite and howardite) are indicated by the violet and yellow colors, respectively.
[23] Olivine spectral shape is characterized by a strong absorption band at 1 µm (Figure 8) due to crystal field transition of Fe2+ ions at non-centrosymmetric M2 sites [Burns, 1970; Adams, 1975]. Minimum position and shape of this band depend on the olivine composition (i.e., Fe content). Due to nearly overlapping bands at 1 µm in an olivine/pyroxene mixture (Figure 8), the olivine spectral characteristics tend to be masked as the olivine abundance in the mixture decreases, making the detection of olivine very challenging. Beck et al. [2011b] suggested that the ratio of the areas of BI and BII, the so-called BAR index [Cloutis et al., 1986], might distinguish olivine, but analysis so far does not show an effect on Vesta. An additional effect should be a shift of the orthopyroxene 1 µm band to longer wavelength (Figure 8). Other well-specified spectral indices [Pelkey et al., 2007; Poulet et al., 2007] might be useful in constraining the presence of olivine-rich materials on Vesta. These indices were fruitfully applied to Mars to detect olivine-rich deposits in the Nili Fossae/Syrtis Major region, which contains up to 40 vol.% olivine [Hoefen et al., 2003]. At a global scale, this analysis did not show evidence of olivine because the strongest olivine signatures were found exclusively in small, localized regions [Pelkey et al., 2007]. Discrimination of olivine abundances as low as 10 vol.% seems to be possible, as found in the compositional and mapping analysis of Tyrrhena Terra [Poulet et al., 2009]. However, the compositions of the Vestan and Martian surfaces are very different, and this could reduce the efficiency of the olivine spectral indexes when applied to Vesta. We made a first attempt to apply these olivine indexes to VIR spectra acquired during Approach, Survey, and HAMO, covering about 65% of the surface with spatial resolution from 1.3 km down to 200 m/pixel. We did not obtain positive results, suggesting that olivine-rich deposits are not exposed as large coherent units on the surface. This conclusion supports a similar finding from spectral studies of olivine-bearing diogenites [P. Beck et al., 2011; A. Beck et al., 2012a]. To address this question more reliably, customized Vestan olivine spectral indices are under development and testing.
6 Geochemistry of Rheasilvia From GRaND Data
[24] Neutron counting rates in the thermal to epithermal range (neutrons with kinetic energies <0.7 MeV) were measured by GRaND's lithium-loaded glass (LiG) scintillator. Neutrons in the epithermal range (0.1 eV to 0.7 MeV) were separately measured by a boron-loaded plastic (BLP) scintillator. The leakage flux of thermal neutrons (<0.1 eV) is sensitive to absorption of neutrons by radiative capture, whereas the flux of epithermal neutrons is primarily sensitive to elastic scattering by hydrogen [Prettyman et al., 2011]. Based on modeling, counting rates for the LiG and BLP were found to follow a linear trend as hydrogen was added to any selected HED whole-rock composition. Mapped counting rates did not fall on the same trend line, which indicates that neutron absorption varies over Vesta's surface [Prettyman et al., 2012].
[25] The contribution of neutron absorption to changes in the LiG counting rate (∆C┴) was determined from BLP and LiG counting data. A map of ∆C┴ in the southern hemisphere of Vesta is shown in Figure 9. ∆C┴ varies with the neutron absorption cross section of the regolith. As shown in Figure 9, the measured range of ∆C┴ is within the full range expected for HED whole-rock compositions.
[26] The Rheasilvia basin has low absorption cross sections (low values of ∆C┴) in comparison to the Vesta average (Figure 9). If the composition corresponding to ∆C┴ = 0 is assumed to be that of the average howardite, then the Rheasilvia basin would have similar absorption properties to diogenites and cumulate eucrites. Thus, the neutron measurements show that Rheasilvia's composition is consistent with that of lower crustal or mantle materials.
[27] Significant variations in neutron absorption among HEDs result from differences in the abundance of Fe, Ca, Al, Ti, and Mg [Prettyman et al., 2011, 2012]. For H-free materials, diogenites, with relatively low Fe, Ca, and Al contents, produce higher LiG counting rates than basaltic eucrites, which have higher abundances of these elements. Cumulate eucrites have lower Fe and Mg contents than diogenites, but higher Ca and Al contents that produce ∆C┴ values that overlap the range of counting rates for diogenites (Figure 9). Thus, the GRaND neutron counting rates are consistent with diogenite, although cumulate eucrite may also occur. The region of lowest ∆C┴ in Figure 9 includes Rheasilvia's central uplift. In addition, although smaller than the scale of the basin, the broad spatial resolution of GRaND must be considered in comparing data sets from different instruments (c.f. Figures 6c and 9).
7 Discussion
7.1 Where Is the Olivine?
[28] Dawn instruments have so far shown no indication of the olivine that was inferred from the limited filter coverage of previous HST spectra and documented by olivine-bearing diogenites. We will explore several possible explanations. Some collapse of the walls of Rheasilvia has clearly occurred (Jaumann et al. 2012) and covered much of the basin floor, but the presence of diogenite without apparent olivine still requires explanation.
[29] First, the estimated crustal depth or the calculated excavation depth of Rheasilvia, which when compared imply exposure of the Vestan mantle, might be in error. Our crustal depth is based on its having a basaltic composition, but it could be thicker if diogenite plutons were emplaced within it. However, the hydrocode calculations of Jutzi and Asphaug [2011] and Ivanov and Melosh [2012], on which the excavation depth was based, accurately predicted the basin topography, so it seems unlikely that the Rheasilvia impact, when added to the excavation of underlying Veneneia, would not have exposed mantle material.
[30] Second, it is conceivable that the mantle rocks in Rheasilvia experienced impact melting, so that crystalline olivine is no longer present. A. Beck et al. [2012b] described howardites containing abundant olivine-rich glasses, which they interpreted as impact melt clasts of harzburgite and dunite target rocks. Although relic olivine was present in these glasses, it is possible that completely melted glasses could have formed in large impacts. This is not an attractive explanation for the apparent absence of olivine in Rheasilvia, however, because complete and widespread melting would also have affected orthopyroxene, whose presence is clearly documented by Dawn instruments. Also, no ponds of impact melt are apparent on the floor of Rheasilvia [Schenk et al., 2012]. Dilution of olivine-bearing mantle rocks by impact mixing might have occurred, however.
[31] A third possibility is that the Vestan mantle, or at least the upper mantle, is composed primarily of orthopyroxene rather than olivine. We note that olivine-bearing diogenites are much less common than orthopyroxenites, perhaps reinforcing the conclusion that olivine is not abundant. Even if harzburgitic diogenite, containing 10–25 vol.% olivine, constitutes the primary upper mantle lithology, it would be difficult to recognize that modest proportion of olivine in VIR spectra. Spectral measurements of olivine-bearing diogenites by P. Beck et al. [2011] and A. Beck et al. [2012a] showed no evidence of olivine in the VIR spectral range. Rather, the spectra of olivine-free orthopyroxenite and harzburgites containing 10 and 25 vol.% olivine were indistinguishable, consistent with other work on binary mixtures of olivine and pyroxene [Cloutis et al., 1986; Sunshine et al., 2004]. A greater proportion of olivine should be spectrally identifiable on Vesta, but it would have to occur at an abundance >25% over the scale of a VIR footprint.
[32] Given the considerable landslide cover within Rheasilvia [Jaumann et al., 2012], the Vestoids might seem to be a more fruitful place to search for olivine. Indeed, analyses of faint absorption bands in the visible range of some Vestoid spectra have been interpreted as indications of 6–12 vol.% olivine mixed with orthopyroxene [Shestopalov et al., 2008]. This same study also found some Vestoids with surfaces containing 12–30 vol.% chromite, which greatly exceeds the chromite contents of any diogenites and suggests mineral sorting (which might also have affected olivine).
7.2 What Is the Composition of the Central Peak?
[33] Models of the formation of large basins with central peaks formed at low impact velocities (<12 km/s) suggest that much of the projectile survives, albeit as crushed and strongly deformed material (Z. Yue, personal communication, 2012). In these models, projectile debris is swept back into the central peak by the collapse flow. Impact velocities onto Vesta are generally low (~5 km/s) [Bottke et al., 1994; O'Brien and Sykes, 2011], and surviving carbonaceous chondrite impactor debris occurs in howardites [Gounelle et al., 2003] and in dark regions on Vesta's surface [McCord et al., 2012; Prettyman et al., 2012]. Thus, it seems likely that Rheasilvia's central peak should contain significant projectile materials.
[34] The VIR spectra of the central peak itself, obtained during LAMO, are consistent with howardite. Perhaps the impactor was a differentiated asteroid similar in composition to Vesta, or if chondritic, its debris was diluted with howardite target materials.
7.3 What Can Be Inferred About Stratigraphy and Origin of the Vestan Interior?
[35] If diogenites formed as residues from partial melting, we might expect their distributions in the Vestan mantle to be discontinuous and associated with unmelted chondritic protolith. The global Fe/O and Fe/Si mass ratios for Vesta measured by GRaND are lower than those ratios for chondrites [Prettyman et al., 2012], effectively ruling out a significant component of unmelted chondritic mantle in the regolith. The Fe content of the diogenite-rich regolith around Rheasilvia would be even lower than the global value. The apparent absence of excavated chondritic protolith within the basin or its ejecta blanket would seem to provide an additional argument against the restite model, to be added to other lines of evidence favoring a cumulate origin for diogenites [Warren, 1997; Greenwood et al., 2005; Beck and McSween, 2010].
[36] The connection between depth of exposure and the occurrence of diogenite, as revealed by FC and VIR observations of Rheasilvia, has further implications for diogenite petrogenesis. It is likely that the localized areas with strong 1 µm absorption features at the base of the central uplift and along Matronalia Rupes are deep-seated diogenites that have been brought to the surface via impact rebound, faulting, and collapse. The distribution of these materials may be an indication of the style of differentiation that formed the diogenites. Diogenite-rich spectral signatures along the northern and southern base of the uplift are observed at approximately equal depths, although 200 km apart. This suggests that diogenite may be uniformly distributed below the surface at a scale that is ≥200 km. Furthermore, if the exposed diogenite-rich material associated with Matronalia Rupes were sampled from a similar depth, which given the high relief seems probable, this would suggest that diogenite may be uniformly distributed at depth on a larger scale. Such a uniform distribution at depth over a large area would seem to favor a magma ocean model for diogenites versus formation in multiple smaller diogenite plutons, unless pluton depth is controlled by ponding of magmas at the base of a eucritic crust.
[37] Magma ocean-style fractionation can be further tested by comparing depths and thickness of diogenite layers predicted through modeling to those based on mapping. One model [Ruzicka et al., 1997] based on a CI chondrite bulk composition predicts a 14 km thick diogenite layer at a depth of 26 km. A much thicker (42 km) layer below a 41 km basaltic crust would be produced from an enstatite chondrite bulk composition [Ruzicka et al., 1997], a model that is unlikely [Toplis et al., 2012]. A model based on a mixture of two chondritic end-members [Righter and Drake, 1997] predicts a ~100 km thick mantle composed of pyroxene and olivine, below a 10–15 km thick basaltic crust.
[38] We can attempt to constrain the depth of diogenite units on Vesta through examining relationships in the central uplift. The greatest depth of exposed diogenite is ~20–22 km, relative to the 285 × 285 × 229 km ellipsoid, but these units have been uplifted so this is a minimum depth. This depth is less than that to the top of the diogenite layer predicted for a carbonaceous chondrite bulk asteroid (26 km) [Ruzicka et al., 1997], but perhaps marginally compatible if some additional uplift is assumed. On the other hand, the observed depth of diogenite is consistent with the eucrite-diogenite interface predicted by the two-chondrite bulk model (15 km) [Righter and Drake, 1997] and much shallower than the enstatite chondrite bulk model (41 km) [Ruzicka et al., 1997]. The latter model can thus be eliminated, since it is unlikely that the base of a central uplift would be brought up ~20 km (the height of the central uplift itself is ~20 km). Constraining the thickness of the diogenite layer might be possible by examining cliffs along Rheasilvia's rim. The ~23 km thick diogenitic unit exposed along Matronalia Rupes (Figure 4b) appears to be in disagreement with the 14 km thickness predicted from the CI chondrite model, but could be a slice of the much thicker diogenetic mantle of the two-chondrite model. However, Matronalia Rupes has undergone mass wasting [Jaumann et al., 2012], so diogenitic debris may have collected at the base of the ridge giving the appearance of a thicker overall unit. On the other hand, it is also possible that this unit extends to greater depths. Some magma ocean models predict depths and thicknesses of diogenite layers of the right order of magnitude, but none correlate consistently with both constraints. The VIR high spatial resolution map of Matronalia Rupes (Figure 7) provides a further test of these hypotheses. The diogenite unit constitutes only the upper 10–12 km of the ridge, while the lower portions show an increasingly howarditic composition (Figure 6c). This supports the interpretation that the upper 10–12 km is exposed diogenite and the lower portions of the slope are debris from mass wasting. This observed thickness is in line with that predicted from the CI chondrite model, but we cannot rule out the possibility that the diogenite extends to greater depths and is now covered by debris.
8 Summary and Conclusions
- [40]
The FC mapping using color ratios reveals the occurrence of orthopyroxene-rich rocks similar to diogenite meteorites at three localized regions in Rheasilvia. Two regions are topographically low areas at the base of the central uplift, interpreted as deep exposures of mantle rocks. A third occurrence on the crater wall is interpreted as deep materials emplaced during impact.
- [41]
The VIR spectra confirm the diogenite occurrences by documenting their more magnesian pyroxene compositions, relative to eucritic rocks outside the basin.
- [42]
Olivine, which is predicted in the mantle of a bulk-chondritic asteroid and expected from the occurrence of olivine-bearing diogenites, has so far evaded detection in VIR spectra. However, olivine abundances of 10–25 vol.%, as appropriate for harzburgitic diogenites, are spectrally obscured by pyroxene and are likely below the detection limit for VIR.
- [43]
The GRaND-measured neutron counting rates in Rheasilvia are lower than outside the basin, consistent with the chemical composition of diogenite.
- [44]
Fe/O and Fe/Si ratios in the regolith measured by GRaND argue against excavation of a significant chondrite component from the mantle, providing evidence that diogenites are not residues after extraction of eucritic partial melts. This observation complements other lines of evidence that support interpretation of diogenites as magmatic cumulates.
- [45]
The occurrence of diogenite at similar depths in areas separated by several hundred km in Rheasilvia may suggest a widespread layer of at least that horizontal extent, but conclusions about the lateral extent of diogenite are complicated by mass wasting within the basin.
- [46]
Dawn observations provide a valuable geologic context for diogenites that is broadly consistent with properties of the meteorites themselves. Unfortunately, the data do not provide unambiguous support for distinguishing whether diogenites crystallized in a magma ocean or in multiple plutons near the crust-mantle boundary.
Acknowledgments
[47] This work was funded by NASA's Discovery Program through a contract to the University of California, Los Angeles, by NASA's Dawn at Vesta Participating Scientists Program, by the Italian Space Agency, by the Max Planck Society and German Space Agency (DLR), and by the Planetary Science Institute under contract with the Jet Propulsion Laboratory, California Institute of Technology. We appreciate reviews by K. Keil, K. Righter, and an unnamed reviewer.
