Meteoritics & Planetary Science 42, Nr 2, 255–269 (2007) Abstract available online at http://meteoritics.org

Geochemistry of 4 Vesta based on HED meteorites: Prospective study for interpretation of gamma ray and neutron spectra for the Dawn mission

Tomohiro USUI* and Harry Y. MCSWEEN, JR.

Planetary Geosciences Institute, Department of Earth and Planetary Sciences, University of Tennessee, Knoxville, Tennessee 37996, USA *Corresponding author. E-mail: [email protected] (Received 05 May 2006; revision accepted 05 December 2006)

Abstract–Asteroid 4 Vesta, believed to be the parent body of the howardite, eucrite, and diogenite (HED) meteorites, will be investigated by the Dawn orbiting spacecraft. Dawn carries a gamma ray and neutron detector (GRaND) that will measure and map some major- and trace-element abundances. Drawing on HED geochemistry, we propose a mixing model that uses element ratios appropriate for the interpretation of GRaND data. Because the spatial resolution of GRaND is relatively coarse, the analyzed chemical compositions on the surface of Vesta will likely reflect mixing of three endmember components: diogenite, cumulate eucrite, and basaltic eucrite. Reliability of the mixing model is statistically investigated based on published whole-rock data for HED meteorites. We demonstrate that the mixing model can accurately estimate the abundances of all the GRaND-analyzed major elements, as well as of minor elements (Na, Cr, and Mn) not analyzed by this instrument. We also show how a similar mixing model can determine the modal abundance of olivine, and we compare estimated and normative olivine data for olivine-bearing diogenites. By linking the compositions of well-analyzed HED meteorites with elemental mapping data from GRaND, this study may help constrain the geological context for HED meteorites and provide new insight into the magmatic evolution of Vesta.

INTRODUCTION roughly in proportion to orbiting altitude. GRaND analyses will be carried out at three different altitudes (survey, high- Dawn, the ninth NASA Discovery mission, will orbit and altitude mapping orbit [HAMO], and low-altitude mapping explore two of the largest Main Belt asteroids, 4 Vesta and orbit [LAMO]) (Fig. 1). Even at LAMO, GRaND’s spatial 1 Ceres, which are complementary protoplanets that have resolution (~300 km) (Prettyman et al. 2004) is larger than remained intact since their formation (Rayman et al. 2006; that of compositional heterogeneity of Vesta’s surface Russell et al. 2004). Vesta is highly differentiated and is (<50 km) as observed by the Hubble Space Telescope (Fig. 1) thought to be the parent body for the voluminous howardite, (Binzel et al. 1997). eucrite, and diogenite (HED) suites (Binzel and Xu 1993; Based on its spectral similarities to HED meteorites, Keil 2002; McCord et al. 1970). Ceres apparently Vesta is interpreted to have a basaltic surface (e.g., Binzel and incorporated water ice during accretion and may contain Xu 1993). Most eucrites and diogenites are brecciated and liquid water (McCord and Sotin 2005); its materials might be many are polymict rocks; the scale of compositional akin to carbonaceous chondrites. variability of HED meteorites (approximately centimeter- The Dawn spacecraft is equipped with a gamma ray and size) (Mittlefehldt et al. 1998) is much smaller than the spatial neutron detector (GRaND) (Prettyman et al. 2003). GRaND resolution of GRaND. Ground-based telescope and Hubble can directly measure the abundances and map the surface observations (Binzel et al. 1997; Gaffey 1997) suggest that distributions of some major (O, Si, Fe, Ti, Mg, Al, and Ca) the average surface of Vesta is analogous to polymict eucrite and trace (U, Th, H, and K) elements (Prettyman et al. 2006, and/or howardite (regolith breccias mainly composed of 2004). A similar analytical technique has been used eucrite and diogenite fragments) assemblages, although the successfully on lunar and Mars orbiters (e.g., Evans et al. diogenite spectral signature is also sporadically observed. 2004; Lawrence et al. 2000, 2002). The ability of GRaND to GRaND will therefore likely yield data that mimics the determine elemental abundances on Vesta is influenced by the mixing of HED meteorites. mission scenario; for example, its spatial resolution varies In this study, we compile 55 published whole-rock

255 © The Meteoritical Society, 2007. Printed in USA. 256 T. Usui and H. Y. McSween, Jr.

Fig. 1. a) An image of Vesta (computer reconstructed, after Zellner and Thomas 1997). b) A false-color (RGB) interpretive geological map of the near-equatorial region of Vesta from Hubble observations obtained during Vesta’s rotation (Binzel et al. 1997). The eastern hemisphere region shown in green may represent eucrite-rich rock, whereas the western hemisphere, shown as relatively “warm” colors, may represent diogenite-rich rock. Red circles show the size of spatial resolution of GRaND at LAMO, which is given by the full-width-at-half-maximum (FWHM) of the spatial response function (~300 km in diameter) of GRaND for 200 km altitude (Prettyman et al. 2004). Altitude for each orbit mode is given by Rayman et al. (2006). The scales of (a) and (b) are comparable. elemental compositions of 44 HED meteorites and propose a eucrites (Mittlefehldt 2003), GRaND analyses are expected to mixing model for interpreting GRaND data for the Dawn reflect the mixture of these three rock types. Mixing relations mission. In this model, we focus on element ratios that are of three endmember components can be displayed on sensitive to difference in mineral mode for HED meteorites. appropriate two-dimensional diagrams. Furthermore, we demonstrate that the HED mixing model can Many two-dimensional diagrams can be drawn by be used to estimate the abundances of some elements not selecting any two element ratios from 11 elements that will be analyzed by GRaND (if there is an adequate HED database), measured by GRaND analysis. However, most of them (e.g., as well as the modal abundances of olivine that occurs in Mg/Si versus Mg/Fe) are not diagnostic for displaying mixing some diogenites and is expected to comprise much of Vesta’s relations in the HED suite (Fig. 2). Selecting appropriate interior. element ratio pairs that are sensitive to differences in mineral modes for each meteorite suite is critical to proper A TWO-DIMENSIONAL DIAGRAM FOR A MIXING interpretation. In this paper, a pair of molar element ratios, MODEL WITH THREE ENDMEMBER (Mg + Fe)/Si and Al/Si (hereafter abbreviated as [M/Si]mol COMPONENTS and [Al/Si]mol, respectively), are employed (Fig. 3). HED meteorites mainly consist of low- and high-Ca pyroxenes and A mixing model is proposed by postulating that GRaND plagioclase. Both pyroxenes have compositions with variable analyses of the surface of Vesta are analogous to HED Mg/Fe ratios and little Al2O3 contents (<1 wt%). By meteorites. The HED suite contains three main igneous employing [M/Si]mol, we are able to cancel out such lithologies: basalt, cumulate gabbro, and orthopyroxenite, variability in Fe/Mg ratios; [M/Si]mol has a value of 1.0 for which correspond to basaltic eucrite, cumulate eucrite, and enstatite-ferrosilite solid solutions and 0.5 for diopside- diogenite, respectively (Mittlefehldt 2003). Since polymict hedenbergite solid solutions. Thus all pyroxenes in HED HED rocks are mixtures of diogenite and basaltic or cumulate meteorites have [M/Si]mol values from 0.5 to 1 with a constant Geochemistry of 4 Vesta based on HED meteorites 257

Fig. 2. Whole-rock Mg/Si versus Mg/Fe weight ratios for the HED meteorites. Mineral compositions of the HED meteorites are also shown. Mg/Fe ratio in plagioclase is illustrated by an arrow, although this does not affect the Mg/Fe bulk-rock values. Abbreviations: B-Eucrite = basaltic eucrite; C-Eucrite = cumulate eucrite; P-Eucrite = polymict eucrite; Low-Ca Px = low-Ca pyroxene; High-Ca Px = high-Ca pyroxene; Plag = plagioclase. Data sources are given in Table 1.

[Al/Si]mol value of ~0. Conversely, plagioclase contains very total little MgO and FeO (<1 wt%) but has abundant Al2O3 (>30 wt%), resulting in high [Al/Si]mol with [M/Si]mol value of ~0. Therefore, a [M/Si]mol versus [Al/Si]mol molar ratio plot is expected to discriminate among the three endmember lithologies diogenite, basaltic eucrite, and cumulate eucrite.

The quality of GRaND data depends critically on the Fig. 3. Whole-rock [M/Si]mol versus [Al/Si]mol plot for HED orbital altitude at which the measurements are made and the meteorites. Mineral compositions of the HED meteorites are also duration of the measurements. The performance of GRaND shown. Thin lines with crosses show (a) binary mixing curves for during the Dawn mission was investigated based on results EM-I and (b) EM-II. The inset in (b) is an enlarged [M/Si]mol versus [Al/Si]mol plot showing the compositional difference between Nuevo from the Lunar Prospector Gamma Ray Spectrometer Laredo and Stannern. Abbreviations are same as in Fig. 2. Data (LPGRS), because LPGRS has a gamma ray sensor with sources are given in Table 1. approximately the same volume, detection efficiency, and pulse height resolution (Prettyman et al. 2004). On this basis, plot between basaltic eucrite, cumulate eucrite, and diogenite, Prettyman et al. (2004) concluded that the spectrum obtained which is consistent with previous geochemical and at LAMO is enough above background to map abundances of petrographical studies (e.g., Mittlefehldt 2003; Mittlefehldt major rock forming elements, radioactive elements, and light et al. 1998). Such characteristics were also observed in all elements such as H, C, and N at Vesta and Ceres. Therefore, other element ratio plots (Usui and McSween 2006). GRaND data sets should allow construction of the mixing To evaluate the contribution of the three rock types to diagram employing [M/Si]mol and [Al/Si]mol. GRaND spectra, we propose two endmember sets, each Figure 3 shows whole-rock compositions of howardites, consisting of three meteorites. One set consists of Shalka, eucrites, and diogenites plotted on the [M/Si]mol versus [Al/ Serra de Magé, and Nuevo Laredo as endmembers Si]mol diagram (meteorite names, classification, and source of representing diogenite, cumulate eucrite, and basaltic eucrite, chemical analyses are given in Table 1). Diogenites have respectively (hereafter referred to as endmember set I distinctly lower [Al/Si]mol and higher [M/Si]mol values than [EM-I]). The other substitutes Stannern instead of Nuevo those of basaltic and cumulate eucrites. Basaltic eucrites have Laredo as a basaltic eucrite endmember (hereafter referred to higher [Al/Si]mol and lower [M/Si]mol values than those of as endmember set II [EM-II]). The elemental compositions of cumulate eucrites. Diogenites and basaltic eucrites have these four endmembers are given in Table 2. Shalka is a relatively uniform compositions compared to the polymict typical diogenite consisting of >90 vol% coarse-grained rocks. The polymict rocks (howardite and polymict eucrite) cumulus orthopyroxene (Mittlefehldt 1994). The cumulate 258 T. Usui and H. Y. McSween, Jr.

Table 1. List of analyzed samples. Rock type Sample Referencea Rock type Sample Referencea Diogenite Jonstown Ab 8 Basaltic eucrite Béréba Ab 2 Diogenite Jonstown Bb 13 Basaltic eucrite Béréba Bb 3 Diogenite Shalka Ab 8 Basaltic eucrite Pasamonte 2 Diogenite Shalka Bb 13 Basaltic eucrite Juvinas Ab 2 Diogenite Y-75032 8 Basaltic eucrite Juvinas Bb 6 Diogenite Y-75032,79 4 Basaltic eucrite Cachari Ab 2 Diogenite Y-791199,79 4 Basaltic eucrite Cachari Bb 3 Diogenite Ellemeet 13 Basaltic eucrite Haraiya 2 Diogenite Garland 13 Basaltic eucrite Ibitira 9 Diogenite Ibbenbühren 13 Basaltic eucrite Sioux County Ab 2, 9 Diogenite Tatahouine 13 Basaltic eucrite Sioux County Bb 3 Diogenite Manegaon 14 Basaltic eucrite Chervony Kut 3 Basaltic eucrite Lakangaon Ab 7 Cumulate eucrite Moama 1 Basaltic eucrite Lakangaon Bb 7 Cumulate eucrite Moore County 2, 9 Basaltic eucrite Nuevo Laredo Ab 7 Cumulate eucrite Serra de Mangé Ab 2, 9 Basaltic eucrite Nuevo Laredo Bb 7 Cumulate eucrite Serra de Mangé Bb 3 Basaltic eucrite ALHA81001, 16 Ab 7 Cumulate eucrite Y-791195,75 4 Basaltic eucrite ALHA81001, 16 Bb 7 Cumulate eucrite Binda Ab 2, 9 Cumulate eucrite Binda Bb 3 Howardite Bholghati 8 Polymict eucrite Y-791192,95 4 Howardite Kapoeta 8 Polymict eucrite Y-82049,75 4 Howardite Y-70308 8 Polymict eucrite Macibini 2, 9 Howardite Malvern 3 Polymict eucrite Petersburg 9 Howardite Molteno 3 Polymict eucrite Y-74450 8 Howardite Le Teilleul 3 Polymict eucrite EETA79004.53 5 Howardite Frankfort 3 Polymict eucrite EETA79005.54 5 Polymict eucrite EETA79011.27 5 Olivine diogenite NWA 1459 10 Basaltic eucrite Y-82066,63 4 Olivine diogenite ALHA77256 11 Basaltic eucrite Y-793164,69 4 Olivine diogenite EETA79002 12 Basaltic eucrite Stannern 2 a1 = Lovering (1975); 2 = McCarthy et al. (1973); 3 = Palme et al. (1978); 4 = Mittlefehldt and Lindstrom (1993); 5 = Palme et al. (1983); 6 = Wänke et al. (1972); 7 = Warren and Jerde (1987); 8 = Mittlefehldt et al. (1998); 9 = Mittlefehldt (2003); 10 = Irving et al. (2003); 11 = Takeda (1979); 12 = Sack et al. (1991); 13 = Fredriksson et al. (1976); 14 = MacCarthy et al. (1974). bA and B are used to differentiate different data sources for the same meteorite.

Table 2. Whole-rock compositions of the endmember components for the mixing model. Shalka Aa Serra de Mangé Ba Stannerna Nuevo Laredo Aa

SiO2 51.6 48.5 49.7 49.4 TiO2 0.062 0.13 0.98 0.82 Al2O3 0.60 14.8 12.3 11.9 Cr2O3 2.4 0.63 0.34 0.28 FeOtotal 16.3 14.4 17.8 19.7 MnO 0.55 0.48 0.53 0.58 MgO 25.8 10.7 7.0 5.6 CaO 0.73 9.8 10.7 10.4 Na2O 0.040 0.25 0.62 0.51 K2O 0.002 0.012 0.066 0.048 Total 98.1 99.8 100.1 99.2

b [M/Si]mol 1.0 0.58 0.51 0.50 b [Al/Si]mol 0.014 0.36 0.29 0.28 b [Ca/Si]mol 0.015 0.22 0.23 0.22 aData sources are in Table 1. bDefinitions are described in the text. Geochemistry of 4 Vesta based on HED meteorites 259

Table 3. Accuracy and precision of the mixing calculations. EM-I Diogenite (n = 9) Howardite (n = 6) Eucrite (n = 34) Diff. wt% Ave. St. dev. (1σ) Ave. St. dev. (1σ) Ave. St. dev. (1σ)

ΔSiO2 −2.3 0.7 −0.39 0.71 0.07 0.44 ΔTiO2 −0.008 0.043 −0.11 0.04 −0.11 0.13 ΔAl2O3 0.45 0.28 0.01 0.22 −0.04 0.18 ΔCr2O3 1.4 0.3 0.61 0.20 0.18 0.20 ΔFeOtotal −0.40 0.89 −1.4 0.6 −0.72 0.85 ΔMnO −0.021 0.054 0.006 0.020 −0.002 0.033 ΔMgO −1.8 0.9 0.47 0.42 0.49 0.55 ΔCaO −0.06 0.28 −0.14 0.54 −0.28 0.36 ΔNa2O 0.061 0.014 −0.050 0.029 0.003 0.11 ΔK2O 0.002 0.007 −0.011 0.009 −0.006 0.013

ΔCr2O3(corr) 0.09 0.17 −0.16 0.14 0.004 0.063 EM-II Diogenite (n = 9) Howardite (n = 6) Eucrite (n = 34) Diff. wt% Ave. St. dev. (1σ) Ave. St. dev. (1σ) Ave. St. dev. (1σ)

ΔSiO2 −2.3 0.7 −0.32 0.69 0.31 0.43 ΔTiO2 0.017 0.068 −0.076 0.070 0.04 0.14 ΔAl2O3 0.60 0.37 0.03 0.24 0.05 0.18 ΔCr2O3 1.4 0.3 0.61 0.20 0.20 0.19 ΔFeOtotal −0.6 1.1 −1.7 0.5 −1.7 0.7 ΔMnO −0.032 0.061 −0.004 0.019 −0.033 0.029 ΔMgO −1.8 0.9 0.66 0.43 1.1 0.5 ΔCaO 0.06 0.31 −0.06 0.54 −0.03 0.41 ΔNa2O 0.079 0.033 −0.022 0.033 0.09 0.11 ΔK2O 0.005 0.007 −0.007 0.007 0.008 0.013 eucrite Serra de Magé is rich in coarse-grained plagioclase unique value of a mixing ratio for GRaND data (see Appendix with equant cumulate texture, resulting in high whole-rock Al for the mathematical formulation). Because these four content (Mittlefehldt et al. 1998). The basaltic eucrites Nuevo meteorites, used as endmember components, plot near the Laredo and Stannern have similar values of [M/Si]mol and [Al/ ends of the ranges of individual meteorite groups in Fig. 3 Si]mol (Table 2; Fig. 3b inset) but are otherwise geochemically (and also in any other compositional diagram), most chemical distinct. Previous studies showed that most basaltic eucrites compositions on the surface of Vesta could likely be were chemically classified into three types (the main group explained by the mixing of the components Shalka, Serra de eucrites and eucrites of the Nuevo Laredo and Stannern Magé, and either Nuevo Laredo or Stannern. trends), based on a plot of whole-rock data for Mg/(Mg + Fe) versus an incompatible element such as Ti (BVSP 1981; ACCURACY AND PRECISION Stolper 1977). The Nuevo Laredo and Stannern trends are OF THE MIXING MODEL defined as extending from the main group composition toward Nuevo Laredo and Stannern, respectively, in this plot. Once a mixing ratio of three endmember components is While the Nuevo Laredo trend is characterized by increasing determined by applying the mixing model to GRaND incompatible element abundances with decreasing Mg/(Mg + analysis, we can reconstruct the entire chemical composition Fe), the Stannern trend has increasing incompatible element of the analyzed area by multiplying chemical compositions abundances with constant Mg/(Mg + Fe). Although the nature of the three endmember components (Table 2) and the of the process governing these two trends remains mixing ratio. However, the mixing model is based only on controversial, both Nuevo Laredo and Stannern may represent [M/Si]mol and [Al/Si]mol ratios, not on other elements. the most differentiated rocks in the HED suite. Because Moreover, the surface of Vesta may not consist only of Nuevo Laredo and Stannern have distinct chemical features, materials having the same compositions as the three we propose two chemical endmember sets for the mixing endmember components; for example, it may include model and examine which one can better explain the surface basaltic eucrite with a composition similar to that of main compositions of Vesta (discussed later). Using the three group eucrites. Therefore, the reliability of the mixing model endmember components, we can delineate mixing lines in the proposed in this study must be examined in terms of its [M/Si]mol versus [Al/Si]mol plot (Fig. 3) and thus obtain a accuracy and precision. 260 T. Usui and H. Y. McSween, Jr.

Fig. 4. Absolute differences between estimated and reference whole-rock elemental abundances (expressed as oxide wt%), based on the mixing model employing EM-I. Black diamonds show data for individual meteorites and open diamonds with error bars (standard deviation = 1σ) show the mean values of diogenites (n = 9), howardites (n = 6), and eucrites (n = 34). If the mixing calculation correctly estimates a whole-rock elemental abundance, the value must be zero within an error. The results are summarized in Table 3. Geochemistry of 4 Vesta based on HED meteorites 261

Fig. 4. Continued. Absolute differences between estimated and reference whole-rock elemental abundances (expressed as oxide wt%), based on the mixing model employing EM-I. Black diamonds show data for individual meteorites and open diamonds with error bars (standard deviation = 1σ) show the mean values of diogenites (n = 9), howardites (n = 6), and eucrites (n = 34). If the mixing calculation correctly estimates a whole-rock elemental abundance, the value must be zero within an error. The results are summarized in Table 3. 262 T. Usui and H. Y. McSween, Jr.

Fig. 5. Absolute differences between estimated and reference whole-rock elemental abundances (expressed as oxide wt%), based on the mixing model employing EM-II. Symbols are as in Fig. 4. The results are summarized in Table 3. The scale of ordinate axis and sample order are the same as in Fig. 4 for ease of comparison. Geochemistry of 4 Vesta based on HED meteorites 263

Fig. 5. Continued. Absolute differences between estimated and reference whole-rock elemental abundances (expressed as oxide wt%), based on the mixing model employing EM-II. Symbols are as in Fig. 4. The results are summarized in Table 3. The scale of ordinate axis and sample order are the same as in Fig. 4 for ease of comparison. 264 T. Usui and H. Y. McSween, Jr.

We will assess the accuracy and precision of the mixing members in the mixing model. Because Cr occurs as a minor model by assuming that HED meteorites not used as element in HED meteorites, the discrepancy between the endmember components represent some portion of Vesta’s calculated and reference values (e.g., 1.4 ± 0.3 wt% for surface analyzed by GRaND. First, a mixing ratio for each diogenite) constitutes a fatal error for the estimation of Cr HED meteorite is calculated by employing its [M/Si]mol and content. For example, if the actual whole-rock Cr2O3 content [Al/Si]mol ratios. Note that the mixing ratio should vary from of a diogenite is 1.0 wt%, the mixing model yields a Cr2O3 0 to 1 (see Appendix). However, some meteorites that plot content of 2.4 ± 0.3 wt%, which is more than twice the actual outside of the mixing triangle (Fig. 3) have negative values. amount. An empirical correction for Cr is discussed below. By considering that a goal of the Dawn mission is to Reliability of the mixing model employing EM-II is measure elemental abundances with a precision ≤20% assessed in Fig. 5. The data points in Fig. 5 are more scattered (Rayman et al. 2006), a negative mixing ratio is than those in Fig. 4; standard deviations of EM-I and EM-II compensated as follows. If a meteorite sample yields a values for diogenite are 0.043 and 0.068 for ΔTiO2, mixing ratio smaller than −0.2, this sample is omitted and respectively; 0.28 and 0.37 for ΔAl2O3; 0.89 and 1.1 for not employed for calculating the mean value. If a sample has ΔFeOtotal; 0.054 and 0.061 for ΔMnO; 0.28 and 0.31 for ΔCaO; a mixing ratio ranging from −0.2 to 0, the ratio is converted and 0.014 and 0.033 for ΔNa2O (Table 3). Furthermore, EM-II to 0 and the other two ratios are normalized so that their is more likely than EM-1 to underestimate FeOtotal content for summation should be 1. Second, a whole-rock composition howardites and eucrites. These observations indicate that of each meteorite (as wt% oxides) is calculated for the major EM-II is less reliable than EM-I. total elements (SiO2, Al2O3, FeO , MgO, and CaO) and minor It should be noted that the samples employed as the elements (TiO2 and K2O) analyzed by GRaND. We also endmember components do not necessarily represent their include several oxides not analyzed by GRaND (Cr2O3, meteorite suites; Nuevo Laredo, for example, is exceptionally MnO, and Na2O) in this step, and will consider the rich in incompatible elements compared to the main group reliabilities of their estimates in a later section. A eucrites that dominate the basaltic eucrite suite. The comparison of the modeled whole-rock composition with its endmember set is chosen only mathematically so that the reference value in the literature yields an uncertainly mixing model predicts the surface elemental compositions as involved in the mixing model. When these procedures are accurately and precisely as possible, and so that it covers as performed on all HED meteorites, we obtain a mean value broad a compositional range as possible. Moreover, because and variability for each meteorite class, diogenite, eucrite, the mixing model is proposed by a conclusively unproven and howardite (Table 3). Accuracy is expressed as the hypothesis that the surface of Vesta is analogous to HED difference between the mean value and the reference value, meteorites, the reliability of the mixing model must be further and precision is expressed as the variability of investigated assessed based on comprehensive data sets from other samples for each meteorite suite. onboard instruments such as framing camera (FC) and visible Figure 4 shows absolute differences between calculated and infrared mapping spectrometer (VIR). and reference whole-rock compositions by employing EM-I. This mixing model adequately estimates most surface ESTIMATION OF ELEMENTS elemental compositions for all three meteorite classes: NOT ANALYZED BY GRaND diogenite, howardite, and eucrite. For example, the mean difference value in CaO contents (wt%, hereafter referred as The elements that can be analyzed by GRaND are limited ΔCaO) for eucrites is −0.28 ± 0.36 wt% (Table 3). This (Prettyman et al. 2003). The mixing model can theoretically indicates that for an actual whole-rock CaO content of estimate not only major elements but also minor and trace 10.00 wt% (appropriate for a eucrite), the mixing model elements such as Ni, Cr, Mn, P, rare earth elements, and yields a CaO content of 9.72 ± 0.36 wt%, which is identical others, if they are given by the three endmember components within error to the actual value of 10.00 wt%. Some elements and a sufficient HED database exists. Here we demonstrate estimated by the mixing model do show some scatter (e.g., the reliability of estimating the abundances of the minor SiO2 and MgO in diogenites). However, Si and Mg are elements Na, Mn, and Cr in HED meteorites, which will not abundant elements in diogenites (>50 wt% for SiO2 and be analyzed by the Dawn mission. Estimation of these >20 wt% for MgO), so the observed discrepancies in elements is carried out by employing EM-I. As is seen in modeled oxide abundances (Fig. 4) would not be crucial. One Fig. 4f and Table 3, ΔMnO values for three meteorite types element for which the mixing model produces large errors is are equal to zero within errors, and these errors (0.054 for Cr (Figs. 4d and 5d). Cr2O3 contents are systematically diogenite, 0.020 for howardite, and 0.033 for eucrite) are overestimated in some eucrites, all howardites, and especially more than ten times smaller than typical MnO content in these in diogenites. This may be explained as an effect of variations meteorites (~0.5 wt%). This suggests that the mixing model in the abundance of chromite (the principal host for Cr), can estimate MnO contents of Vesta’s surface materials with which is apparently not adequately addressed by the end relative error of <10%. ΔNa2O values (Fig. 4i; Table 3) are Geochemistry of 4 Vesta based on HED meteorites 265

Fig. 7. Absolute differences between reference and corrected Cr2O3 contents (wt%) by Equation 2 (see text). Black diamonds represent Fig. 6. A plot of Cr value (mol fraction) versus [M/Si]mol for HED data for individual meteorites and open diamonds with error bars meteorites. The solid line is a linear least-square fit to all data points. (standard deviation = 1σ) show the mean values of diogenites (n = 9), The regression equation and coefficient are given. howardites (n = 6), and eucrites (n = 34). Uncorrected values for each of the data points, which are the same as in Fig. 4d, are shown for also considered to be zero within errors, although that of comparison. diogenite slightly departs from zero-level, suggesting that Na2O content can be adequately estimated. However, ESTIMATION OF MODAL OLIVINE ABUNDANCE contributions of these errors to the estimation of Na are larger than those for Mn, considering the low Na2O contents of HED Vesta has experienced significant impact events, one of meteorites (<0.1 wt% for diogenite, ~0.2 wt% for howardite, which excavated a huge crater (460 km in diameter and ~0.4 wt% for eucrite) (Mittlefehldt et al. 1998). 13 km in depth) near its south pole (Thomas et al. 1997). ΔCr2O3 obviously has positive values (Table 3), Spectral variations within this large crater demonstrate suggesting that the mixing model must be modified to compositional stratigraphy, probably reflecting a high-Ca correct its discrepancy. ΔCr2O3 values are evidently pyroxene-rich plutonic assemblage deep within the crust correlated with meteorite types, from high ΔCr2O3 in and/or the exposure of olivine present within the upper diogenite to moderate in howardite and low in eucrite mantle (Thomas et al. 1997). Models of a differentiated (Fig. 4d). This relationship is best explained by a Cr versus asteroid support the idea of an olivine-rich mantle (e.g., [M/Si]mol plot (Fig. 6). This figure shows a strong positive Jones 1984). However, these studies appear to be correlation between Cr and [M/Si]mol, from which we obtain inconsistent with the fact that few olivine-bearing meteorites the regression equation: occur in the HED suite (e.g., Mittlefehldt et al. 1998). Most telescopic observations have been limited to the equatorial ΔCr= 0.035× [] M⁄ Si mol – 0.018 (1) regions of Vesta (e.g., Binzel et al. 1997; Gaffey 1997), so the possibility that other olivine-bearing units are exposed in where ΔCr is difference in molar abundance of Cr between the smaller craters or as volcanic flows on the asteroid’s surface reference and calculated values. The Cr2O3 content can be cannot be ruled out. corrected as follows: In this section, we demonstrate that it is possible to estimate the abundance of olivine from elemental data ΔCr× M obtained by GRaND. Similar to the previously Cr O = Cr O – ------(2) 2 3(corr) 2 3(cal) 2 described mixing model, a molar element ratio plot is used. An [M/Si]mol versus [Ca/Si]mol plot (Fig. 8) yields an where Cr2O3(cal) and Cr2O3(corr) are the initially calculated approximately linear trend for HED meteorites, unlike the Cr2O3 concentration (wt%) by the mixing model and its [M/Si]mol versus [Al/Si]mol plot (cf. Fig. 3). This linearity may corrected value, respectively, and M is molecular weight of be due to the presence of high-Ca pyroxene. Since Cr2O3. Using this correction, ΔCr2O3 values are significantly high-Ca pyroxene does not affect [Al/Si]mol but does affect suppressed from 1.4 ± 0.3 to 0.09 ± 0.17 for diogenite, 0.61 ± [Ca/Si]mol, it obscures differences in the modal abundance of 0.20 to −0.16 ± 0.14 for howardite, and 0.18 ± 0.20 to 0.004 ± plagioclase in the HED classes. Because the [M/Si]mol versus 0.063 for eucrite (Table 3; Fig. 7). These values are [Ca/Si]mol plot discriminates two rather than three sufficiently accurate for estimating the Cr concentration of endmember components, it can be used to estimate a third, Vesta’s surface materials. olivine, which plots off the linear HED trend. From the [M/ 266 T. Usui and H. Y. McSween, Jr.

Fig. 8. Whole-rock [M/Si]mol versus [Ca/Si]mol for all HED meteorites, including three olivine-bearing diogenites: ALHA77256 (Takeda 1979), EETA79002 (Sack et al. 1991), and NWA 1459 (Irving et al. 2003). Note that whole-rock composition of EETA79002 was determined from a thin section by averaging a series of electron microprobe analyses of minerals, and that of NWA 1459 was calculated by multiplying representative major element compositions of minerals and their density-weighed modal abundances. The density-weighed modal abundances were calculated by multiplying the modes in vol% (Irving et al. 2003) by densities: 3.8 for olivine, 3.4 for low-Ca pyroxene, 2.7 for plagioclase, and 5.0 for chromite. The composition of pure olivine is shown at the upper left. The broken line is a linear least-square fit to all the data points for the HED meteorites except the olivine-bearing diogenites. Solid lines with ticks show binary mixing curves between olivine, Shalka, and Nuevo Laredo. Normative modes of olivine (mol%) in the olivine-diogenites are also shown. Abbreviations: B-Eucrite = basaltic eucrite; C-Eucrite = cumulate eucrite; P-Eucrite = polymict eucrite; Ol-diogenite = olivine diogenite; Low-Ca Px = low-Ca pyroxene; High- Ca Px = high-Ca pyroxene; Plag = plagioclase.

Si]mol versus [Ca/Si]mol plot (Fig. 8), the following regression calculated the CIPW norms for these compositions and equation is obtained: recalculated the normative proportion of olivine as mol% for comparison. NWA 1459, ALHA77256, and EETA79002 have []MSi⁄ mol = –CaSi2.4 × []⁄ mol + 1.1 (3) 42 mol%, 17 mol%, and 18 mol% of normative olivine, respectively. These values are consistent with the estimation If olivine is present in units exposed on the surface of of the mixing model. Vesta, the GRaND [M/Si]mol and [Ca/Si]mol values for these In addition to GRaND, the Dawn spacecraft is equipped units should plot above this regression line in Fig. 8. To with a VIR, which can identify some minerals on Vesta with evaluate the contribution of olivine to HED meteorites, we higher spatial resolution than that of GRaND (Russell et al. propose two endmember components from EM-I: Shalka as 2004). In particular, diagnostic absorption bands for olivine the diogenite and Nuevo Laredo as eucrite. The [Ca/Si]mol and pyroxenes occur in the visible and near-infrared regions. values of the diogenite and eucrite endmember components Experiments by Cloutis et al. (1986) showed that the ratio of are given in Table 2, and their [M/Si]mol values are derived by the area of an absorption band near 2 μm (band II) relative to substitution of their [Ca/Si]mol values into Equation 3 so that the area of an absorption band near 1 μm (band I) has a strong these endmember components plot exactly on the regression linear correlation with the abundance ratio of olivine to line. The olivine component in Fig. 8 is represented by the orthopyroxene. The abundance of olivine also shows a composition of the forsterite-fayalite solid solution correlation with [M/Si]mol measured by GRaND. Thus, data (Mg,Fe)2SiO4, resulting in [M/Si]mol = 2 and [Ca/Si]mol = 0. from GRaND and VIR complement each other, and To estimate the proportions of these three endmembers, linear combining these comprehensive data sets can strengthen mixing lines are delineated (Fig. 8). Although olivine is mineralogical interpretations and yield new insights into generally absent from HED meteorites, some recently mapped surface lithologies on Vesta. recovered northern African olivine diogenites contain >40 vol% olivine (Irving et al. 2005, 2003). To evaluate the CONCLUSIONS olivine mixing model, the whole-rock compositions of NWA 1459, ALHA77256, and EETA79002 (Irving et al. 2003; Sack We propose a quantitative mixing model for geochemical et al. 1991; Takeda 1979) are plotted in Fig. 8. We also data of asteroid 4 Vesta to be obtained by the GRaND Geochemistry of 4 Vesta based on HED meteorites 267 spectrometer on the Dawn asteroid orbiter, based on the context for HED meteorites and provide significant new chemical compositions of HED meteorites. As inferred from insights into the structure and igneous evolution of one of the telescopic spectra, the surface of Vesta mainly consists of few surviving protoplanets in the solar system. polymict rocks that are mixtures of diogenites and basaltic and cumulate eucrites. The spatial resolution of GRaND Acknowledgments–We thank David W. Mittlefehldt, Carlé M. analyses dictates that they will almost certainly reflect Pieters, and an anonymous reviewer for constructive reviews, mixtures of these three rock types. We have employed a two- and Timothy J. McCoy, Thomas H. Prettyman, Carol A. dimensional element ratio plot (molar ratio of (Mg + Fe)/Si Raymond, and Robert C. Reedy for thoughtful discussions of versus Al/Si) to resolve and model the three endmember this research. This work was supported by NASA components. Because these element ratios amplify the cosmochemistry grant NNG06GG36G and UCLA Dawn difference in modal abundances of plagioclase and low- and co-investigator subcontract 2090GFC199 to H. Y. M. high-Ca pyroxene (the major phases in HED meteorites), this plot discriminates the three prospective endmember Editorial Handling—Dr. Carlé Pieters lithologies, diogenite, basaltic, and cumulate eucrites. To evaluate the contribution of the three rock types to REFERENCES GRaND spectra, we proposed two endmember sets (EM-I and EM-II). EM-I consists of Shalka, Serra de Magé, and Nuevo Binzel R. P., Gaffey M. J., Thomas P., Zellner B. H., Storrs A. D., and Laredo as endmembers representing diogenite, cumulate Wells E. N. 1997. Geologic mapping of Vesta from 1994 Hubble eucrite, and basaltic eucrite, respectively. EM-II substitutes Space Telescope images. Icarus 128:95–103. Stannern for Nuevo Laredo as the basaltic eucrite endmember Binzel R. P. and Xu S. 1993. Chips off asteroid 4 Vesta: Evidence for component. Because these endmember meteorites plot near the parent body of basaltic achondrite meteorites. Science 260: 186–191. the ends of the ranges of HED classes in the [M/Si]mol versus Basaltic Volcanism Study Project. 1981. Basaltic volcanism on the [Al/Si]mol plot, most chemical compositions of the surface of terrestrial planets. New York: Pergamon Press. 1286 p. Vesta could likely be explained by the simple mixing of these Cloutis E. A., Gaffey M. J., Jackowski T. L., and Reed K. L. 1986. three components. Calibrations of phase abundance, composition, and particle size Once a mixing ratio of three end-components is distribution for olivine-orthopyroxene mixtures from reflectance spectra. Journal of Geophysical Research 91: determined for an HED meteorite, its whole-rock 11,641–11,653. composition can be estimated from the compositions of the Evans L. G., Starr R. D., Reedy R. C., and Boynton W. V. 2004. endmembers. By comparing the estimated whole-rock data Elemental composition variations for large dusty and rocky for all the HED meteorites with their published chemical regions on Mars using gamma ray data from the Mars Odyssey analyses, the reliability of the mixing model was investigated. Gamma Ray Spectrometer. (abstract #1258). 35th Lunar and Planetary Science Conference. CD-ROM. The mixing model employing EM-I more accurately and Fredriksson K., Noonan A., Brenner P., and Sudre C. 1976. Bulk and precisely estimates HED compositions than that employing major phase composition of eight hypersthene achondrites. EM-II. Furthermore, assuming the HED meteorites used in Meteoritics 11:278–280. this study are fully representative of Vesta’s surface, we Gaffey M. J. 1997. Surface lithologic heterogeneity of asteroid 4 demonstrated that the mixing model can be used to estimate Vesta. Icarus 127:130–157. Irving A. J., Kuehner S. M., Carlson R. W., Rumble D. I., Hupe A. C., elements that will not be analyzed by GRaND, such as Na, and Hupe G. M. 2005. Petrology and multi-isotopic composition Mn, and Cr, although estimation of Cr requires an empirical of olivine diogenite NWA 1877: A mantle peridotite in the correction because variations in chromite abundance are proposed HEDO group of meteorite (abstract #2188). 36th Lunar apparently not adequately addressed by the endmembers in and Planetary Science Conference. CD-ROM. the mixing model. Irving A. J., Kuehner S. M., Rumble D. I., Hupe A. C., and Hupe G. M. 2003. Olivine diogenite NWA 1459: Plumbing the depths Telescopic observations and theoretical models of Vesta of 4 Vesta (abstract #1502). 34th Lunar and Planetary Science suggest an olivine-rich mantle or lower crust, but olivine is Conference. CD-ROM. absent or occurs only as a minor constituent mineral in most Jarosewich E. 1990. Chemical analyses of meteorites: A compilation HED meteorites. We also propose another mixing model of stony and iron meteorite analyses. Meteoritics 25:323–337. employing GRaND data, by which we can estimate the modal Jones J. H. 1984. The composition of the mantle of the eucrite parent body and the origin of eucrites. Geochimica et Cosmochimica abundance of olivine in HED meteorites and thus surface Acta 48:641–648. units on Vesta. GRaND-derived mineral abundances will Keil K. 2002. Geological history of asteroid 4 Vesta: The “smallest complement data from other Dawn spectrometers. terrestrial planet.” In Asteroids III, edited by Bottke W., This study provides a way to leverage the large Cellino A., Paolicchoi P., and Binzel R. P. Tucson, Arizona: The geochemical and mineralogical database on HED meteorites University of Arizona Press. pp. 573–584. Lawrence D. J., Feldman W. C., Barraclough B. L., Binder A. B., as a tool for interpreting chemical analyses by GRaND of Elphic R. C., Maurice S., Miller C. F., and Prettyman T. H. 2000. mapped units on the surface of Vesta. We expect that the new Thorium abundances on the lunar surface. Journal of data sets to be provided by Dawn will constrain the geologic Geophysical Research 105:20,307–20,331. 268 T. Usui and H. Y. McSween, Jr.

Lawrence D. J., Feldman W. C., Elphic R. C., Little R. C., Prettyman neutron spectrometer for Dawn (abstract #2231). 37th Lunar and T. H., Maurice S., Lucey P. G., and Binder A. B. 2002. Iron Planetary Science Conference. CD-ROM. abundances on the lunar surface as measured by the Lunar Prettyman T. H., Feldman W. C., Ameduri F. P., Barraclough B. L., Prospector gamma ray and neutron spectrometers. Journal of Cascio E. W., Fuller K. R., Funsten H. O., Lawrence D. J., Geophysical Research 107, doi:10.1029/2001JE001530. McKinney G. W., Russell C. T., Soldner S. A., Storms S. A., McCarthy T. S., Erlank A. J., and Willis J. P. 1973. On the origin of Szeles C., and Tokar R. L. 2003. Gamma ray and neutron eucrites and diogenites. Earth and Planetary Science Letters 18: spectrometer for the Dawn mission to 1 Ceres and 4 Vesta. IEEE 433–442. Transactions on Nuclear Science 50:1190–1197. McCarthy T. S., Erlank A. J., Willis J. P., and Ahrens L. H. 1974. New Prettyman T. H., Feldman W. C., Barraclough B. L., Capria M. T., chemical analyses of six achondrites and one chondrite. Coradini A., Enemark D. C., Fuller K. R., Lawrence D. J., Patrick Meteoritics 9:215–221. D. E., Raymond C. A., Storms S. A., and Williforda R. L. 2004. McCord T. B., Adam J. B., and Johnson T. V. 1970. Asteroid Vesta: Mapping the elemental composition of Ceres and Vesta: Dawn’s Spectral reflectivity and compositional implications. Science gamma ray and neutron detector. Proceedings of the Society of 168:1445–1447. Photo-Optical Instrumentation Engineers 5660:107–116. McCord T. B. and Sotin C. 2005. Ceres: Evolution and current state. Rayman M. D., Fraschetti T. C., Raymond C. A., and Russell C. T. Journal of Geophysical Research 110, doi:10.1029/ 2006. Dawn: A mission in development for exploration of Main 2004JE002244. Belt asteroids Vesta and Ceres. Acta Astronautica 58:605–616. McCoy T. J., Burbine T. H., McFadden L. A., Starr R. D., Gaffey Russell C. T., Coradini A., Christensen U., De Sanctis M. C., M. J., Nittler L. R., Evans L. G., Izenberg N., Lucey P. G., Feldman W. C., Jaumann R., Keller H. U., Konopliv A. S., Trombka J. I., Bell J. F. III, Clark B. C., Clark P. E., Squyres McCord T. B., McFadden L. A., McSween H. Y., Jr., Mottola S., S. W., Chapman C. R., Boynton W. V., and Veverka J. 2001. The Neukum G., Pieters C. M., Prettyman T. H., Raymond C. A., composition of 433 Eros: A mineralogical-chemical synthesis. Smith D. E., Sykes M. V., Williams B. G., Wise J., and Zuber Meteoritics & Planetary Science 36:1661–1672. M. T. 2004. Dawn: A journey in space and time. Planetary and Mittlefehldt D. W. 1994. The genesis of diogenites and HED parent Space Science 52:465–489. body petrogenesis. Geochimica et Cosmochimica Acta 58:1537– Sack R. O., Azeredo W. J., and Lipschutz M. E. 1991. Olivine 1552. diogenites: The mantle of the eucrite parent body. Geochimica et Mittlefehldt D. W. 2003. Achondrites. In Meteorites, comets, and Cosmochimica Acta 55:1111–1120. planets, edited by Davis A. M. Oxford: Elsevier-Pergamon. Stolper E. 1977. Experimental petrology of eucrite meteorites. pp. 291–324. Geochimica et Cosmochimica Acta 41:587–611. Mittlefehldt D. W. and Lundstrom M. M. 1993. Geochemistry and Takeda H. 1979. A layered-crust model of a howardite parent body. petrology of a suite of ten Yamato HED meteorites. Proceedings Icarus 40:455–470. of the NIPR Symposium of Antarctic Meteorites 6:268–292. Thomas P. C., Binzel R. P., Gaffey M. J., Storrs A. D., Wells E. N., Mittlefehldt D. W., McCoy T. J., Goodrich C., and Kracher A. 1998. and Zellner B. H. 1997. Impact excavation on asteroid 4 Vesta: Non-chondritic meteorites from asteroidal bodies. In Planetary Hubble Space Telescope results. Science 277:1492–1495. materials, edited by Papike J. J. Washington, D.C.: Usui T. and McSween H. Y., Jr. 2006. Characterizing the surface Mineralogical Society of America. pp. 1–195. elemental composition of 4 Vesta based on HED meteorites: Palme H., Baddenhausen H., Blum K., Cendales M., Dreibus G., Prospective study of gamma ray and neutron spectrometer for the Hofmeister H., Kruse H., Palme C., Spettel B., Vilcsek E., and Dawn mission (abstract #1407). 37th Lunar and Planetary Wänke H. 1978. New data on lunar samples and achondrites and Science Conference. CD-ROM. a comparison of the least fractionated samples from the Earth, the Wänke H., Balacescu A., Teschke F., Spettel B., Dreibus G., Moon, and the eucrite parent body. Proceedings, 9th Lunar and Palme H., Quijano-Rico M., Kruse H., Wlotzka F., and Planetary Science Conference. pp. 25–57. Begemann F. 1972. Multielement analyses of lunar samples and Palme H., Spettel B., Burghele G., Weckwerth G., and Wänke H. some implications of the results. Proceedings, 2nd Lunar Science 1983. Elephant Moraine polymict eucrites: A eucrite-howardite Conference. pp. 1251–1268. compositional link (abstract). 14th Lunar and Planetary Science Warren P. H. and Jerde E. A. 1987. Composition and origin of Nuevo Conference. pp. 590–591. Laredo trend eucrites. Geochimica et Cosmochimica Acta 51: Prettyman T. H., Baldonado J. R., Bernardin J. D., Dingler R. D., 713–725. Enemark D. C., Little R. C., Miller E. A., Patrick D. E., Pavri B., Zellner B. H. and Thomas P. C. 1997. Hubble reveals huge crater on Raymond C. A., Russell C. T., Storms S. A., Sweet M. R., the surface of the asteroid Vesta. September 4, 1997. Space Williforda R. L., and Wong-Swanson B. 2006. Gamma ray and Telescope Science Institute Press Release #1997–27.

APPENDIX abscissa and ordinate, respectively; xi, yi = coordinates of data point Ci (note that subscript i is for all three endmember Here we present a general method for calculating a components, C1, C2, and C3); nxi, nyi = numerator of xi and yi; mixing ratio based on a three-component mixing model that dxi, dyi = denominator of xi and yi; xm, ym = coordinates of data uses a two-dimensional ratio-ratio plot (Fig. A1). We assume point Cm. xm and ym are given by measurements (e.g., HED that component Cm is obtained by mixing three endmember analyses or a GRaND analysis). Simple mass balance components, C1, C2, and C3 with a mixing ratio of c1, c2, and calculation gives the following equations: c3, which satisfy the equation: 3 3 c1 + c2 + c3 = 1 (A1) xm = ∑ cinxi ⁄ ∑ cidxi Symbols used are: x, y = general variables along the i = 1 i = 1 Geochemistry of 4 Vesta based on HED meteorites 269

3 3

ym = ∑ cinyi ⁄ ∑ cidyi (A2) i = 1 i = 1 From Equations A1 and A2, we obtain the matrix product:

X1 X2 X3 c1 0 ⋅ = (A3) Y1 Y2 Y3 c2 0

111 c3 1 where Xi = xmdxi − nxi and Yi = ymdyi − nyi. Solving Equation A3, we obtain the mixing ratios c1, c2, and c3 as follows: α c = ------1 αβγ++

β c = ------2 αβγ++ Fig. A1. A schematic plot of two element ratios for the mixing of γ three endmember components. c = ------(A4) 3 αβγ++ This formulation explains how to calculate the plot of where any two element ratios. For example, to construct Fig. 3, x and α = Y2X3 − X2Y3 y correspond to the molar abundance ratios of (Mg + Fe)/Si and Al/Si, respectively. nxi and nyi are the molar abundances β = Y3X1 − X3Y1 of Al and (Mg + Fe), and both dxi and dyi represent the molar γ = X2Y1 − Y2X1 (A5) abundances of Si.