Distribution of iron on Vesta

Authors


Abstract

We have completed a mapping study of 7.6 MeV gamma rays produced by neutron capture by Fe at the surface of the main belt asteroid 4 Vesta as measured by the bismuth germanate scintillator of the Gamma Ray and Neutron Detector (GRaND) on the Dawn spacecraft. The procedures used to determine Fe counting rates are presented, along with a global map, constituting the necessary initial step to quantify Fe abundances. While the final calibration of orbital data to absolute concentrations has not been determined, the range of fully corrected Fe counting rates is compared with that of Fe in howardites. We find that the global distribution of corrected Fe counting rates is generally consistent with mineralogy and composition determined independently by other instruments on the Dawn spacecraft, including measurements of pyroxene absorption bands by the Visible and Infrared Spectrometer and Framing Camera, and an index of diogenitic materials provided by neutron absorption measurements by GRaND. In addition, there is a distinctive low Fe region in the western hemisphere that was not reported by reflectance or optical observations, possibly indicating the presence of a cumulate eucrite component in Vesta's regolith.

Introduction

Vesta has been proposed as the parent body of the howardite, eucrite, and diogenite (HED) meteorites for many years based on the observation of common spectral reflectance features produced by the absorption bands of ferrous iron in mafic pyroxene and olivine silicates (McCord et al. 1970) and finding of small, dynamically linked asteroids (“Vestoids”) that span between Vesta and the 3:1 resonance with Jupiter, where asteroids are perturbed into Earth-crossing orbits (Binzel and Xu 1993). Recent observations by the Dawn spacecraft also support Vesta as the source of the HEDs (e.g., McSween et al. 2013b), giving us an unprecedented view of a differentiated asteroid (e.g., McSween et al. 2011).

A primary objective of the Gamma Ray and Neutron Detector (GRaND) on NASA's Dawn spacecraft (Prettyman et al. 2011, 2012; Russell et al. 2012) is to characterize the elemental composition of Vesta and to provide geologic context for the HED meteorites. Of the elements measured by GRaND, Fe is among the most important. Fe is found in pyroxene and other mafic minerals in Vesta's regolith. The abundance of Fe varies predictably with rock type for the HED meteorites. Thus, Fe can be used along with other mineralogic and chemical information to classify the petrology of surface geologic units. Basaltic eucrites have the highest Fe abundance, whereas lower crustal and upper mantle materials (cumulate eucrites and diogenites) have the lowest. Howardites are intermediate with a range of about 12–17 wt% Fe (Prettyman et al. 2012). The abundance of Fe in olivine-rich rocks, which are a minor lithology in diogenites (Beck and McSween 2010), depends on the concentration of the forsterite component in olivine (e.g., Prettyman et al. 2013). High Fe abundances are expected to be found if Vesta is the parent body of mesosiderites, a class of meteorite in which a basaltic component is mixed with Fe-Ni metal.

To explore the elemental composition of planetary surfaces, nuclear spectroscopy with gamma rays and neutrons has been applied to a variety of solar system objects, including 433 Eros, 4 Vesta, Mars, Mercury, Venus, and the Moon (e.g., Metzger et al. 1973; Vinogradov et al. 1973; Feldman et al. 1998, 2002; Evans et al. 2001; Lawrence et al. 2002, 2013a; Mitrofanov et al. 2002; Prettyman et al. 2006, 2009, 2012; Boynton et al. 2007; Yamashita et al. 2010, 2012; Zhu et al. 2010; Peplowski et al. 2011). Nuclear spectroscopy has the potential to measure abundances of elements such as H, O, Mg, Al, Si, K, Ca, Ti, Fe, Th, and U (Reedy 1978). The specific elements that can be detected and quantified depends on the design of the detector and how it is deployed on the spacecraft, the intensity of background sources, and the mission parameters (e.g., observation time, pointing, distance from the planetary surface, and the flux of galactic cosmic rays, etc.).

A general introduction to planetary nuclear spectroscopy methods is presented by Prettyman (2007). A description of the GRaND instrument and data acquired during the Vesta encounter is given by Prettyman et al. (2011, 2012). In this paper, we present the analysis and interpretation of data acquired by GRaND's bismuth germanate (BGO) sensor, which measures the spectrum of gamma rays emitted by Vesta's surface. We focus specifically on Fe, which has a large cross section for radiative, thermal neutron capture that produces a characteristic gamma-ray doublet at 7.6 MeV, which was detected while in close proximity to Vesta (Prettyman et al. 2012). These gamma rays are produced steadily by galactic cosmic ray showers within Vesta's regolith. Gamma rays that escape Vesta to be detected by GRaND are produced within a meter of the surface (Prettyman et al. 2006). In the analysis presented here, we assume that Vesta's composition is uniform with depth within the upper meter.

Data acquired during several months in low altitude mapping orbit (LAMO) are analyzed (Prettyman 2011). The mean altitude of the Dawn spacecraft in LAMO was about 210 km, compared to Vesta's mean radius of 265 km (Prettyman et al. 2012). Because GRaND's sensitivity to gamma rays is omnidirectional (Prettyman et al. 2011), the spatial footprint depends on orbital altitude. The full width at half maximum (FWHM) of GRaND's footprint on Vesta's surface scales as about 1.5 times the orbital altitude (e.g., Reedy 1978), about 300 km. This is the scale at which spatial regions with distinct chemical compositions can be resolved and separately measured by GRaND. Consequently, large regions, such as the Rheasilvia basin, with a diameter of about 500 km (Schenk et al. 2012) can be characterized (Prettyman et al. 2012).

One of the differences between Vesta and planetary surfaces previously examined by gamma rays and neutrons is the dynamic range of elemental compositions. Fe on the surface of the Moon is known to vary by about a factor of 9 (Lawrence et al. 2002; Prettyman et al. 2006) because of the large-scale differentiation that resulted in an Fe-poor anorthositic crust (e.g., Jolliff et al. 2000) onto which flowed Fe-rich mare basalts, while that in howardites varies by only 40% (Prettyman et al. 2012). Therefore, variations of Fe signals from Vesta are expected to be small, which can present challenges for data acquisition and analysis. However, the determination of the distribution of elemental Fe abundance at the surface of Vesta would provide important independent information to supplement or confirm inferences made based on mineralogic and/or optical observations made by the other Dawn instruments.

Grand Measurements

Instrument

The BGO scintillator measures gamma rays emitted from the surface with energies from approximately 0.3–9 MeV (Prettyman et al. 2011). Energies of characteristic gamma rays are unique to the source nuclide from which the gamma rays are emitted, and their intensities are proportional to the abundance of the source nuclide after appropriate corrections are made (e.g., Evans et al. 1993; Lawrence et al. 2002; Prettyman et al. 2006; Prettyman 2007). In this work, the neutron-capture gamma rays from 56Fe at 7.6 MeV, which appear as a distinct peak in the BGO pulse height spectrum, were used to study the surface distribution of Fe on Vesta.

The depth sensitivity of the measurement depends on the depth profile and energy distribution of the interrogating neutrons, which is strongly influenced by the abundance of H, and the gamma ray attenuation coefficient of the regolith. For H-free lunar soils that are similar to HED materials, Prettyman et al. (2006) showed that capture gamma rays (at 3 MeV) were sensitive to depths of about 100 g cm−2, where depth is presented as the product of geometric thickness (cm) and bulk soil density (g cm−3). Because the attenuation coefficient of 7.6 MeV gamma rays is somewhat lower than that of 3 MeV gamma rays, greater depths are probed by the 7.6 MeV gamma rays. The addition of H pushes the neutron number density profile, and, therefore, the production of gamma rays, closer to the surface (e.g., Feldman et al. 2000; Prettyman et al. 2011); however, given the variation of H on Vesta (about 400 μg g−1) (Prettyman et al. 2012), the effect of H on the depth sensitivity of gamma ray measurements is expected to be small (see fig. 3 of Prettyman et al. 2011).

Observations

The gamma-ray data for mapping were acquired during LAMO phase, which took place from December 2011 to May 2012 (Prettyman et al. 2012). To determine the background contribution from the spacecraft, data acquired during the Vesta Survey period, which took place at more than 2700 km away from Vesta's surface, were analyzed. The total spectra obtained during the two observation campaigns are shown in Fig. 1 (Prettyman et al. 2011, 2012).

Figure 1.

Energy spectra of gamma rays obtained by Gamma Ray and Neutron Detector's bismuth germanate (BGO) scintillator and accumulated over the whole Vesta surface during LAMO and Survey. Spectra obtained in single interaction events (CAT9) range up to about 9 MeV, while that obtained in coincidence events (−Z CAT2-BGO, BGO spectra obtained during a prompt coincidence between the BGO and −Z boron loaded plastic scintillators, following the neutron capture reaction by 10B) ranges up to about 2 MeV. See Prettyman et al. (2011) for similar spectra acquired in the laboratory and a description of event categories. Distinct peaks used in gain correction are labeled with the source element and energies. The peaks at 4.4 MeV are attributed to the noryl canister and cosmogenic 12C produced from 16O.

Spectral Analysis

The amplitudes of output pulses from the BGO sensor's front-end electronics are digitized to 1024-channel histograms by the analog-to-digital converter (ADC). Raw histograms produced by the GRaND's ADC have high-frequency variations in counts with an odd-even-channel pattern that repeats itself every 32 channels. Such artifacts are produced by differential nonlinearity in energy widths of the ADC channels, independent of time (Prettyman et al. 2004a, 2004b, 2011, 2012; Maurice et al. 2011; Prettyman and Feldman 2012). To remove such effects, a correction pattern was produced by taking the average of 14 32-ch patterns retrieved from channels 192–639 of the smoothed (boxcar for 10 channels) total spectrum and applied to all time-series spectra as a primary correction. The process is described by Prettyman et al. (2004a, 2004b) and Prettyman and Feldman (2012).

Temporal Gain Shift Correction

The output pulse height of the BGO detector can vary with temperature and various electronic conditions (e.g., magnetic fields produced by nearby equipment). As elemental observations by gamma-ray spectroscopy require accumulation of many spectra over time, it is important to correct all of them to the same gain to achieve optimal energy resolution and signal-to-noise ratios. Therefore, we corrected each time-series gamma-ray spectrum for gain. First, centroids for distinctive Gaussian peaks in gamma-ray spectra were determined using a simple, robust peak identification algorithm (Mariscotti 1967). These peaks include positron annihilation at 0.511 MeV, 27Al(n,n'γ) at 2.211 MeV, 12C* at 4.438 MeV, and 16O(n,n'γ) at 6.129 MeV in the BGO single event spectrum (CAT9), and 10B at 0.478 MeV in the BGO coincidence event spectrum (−Z CAT2-BGO). The 0.478 MeV peak was observed by a prompt coincidence between the BGO and −Z boron loaded plastic scintillator following the neutron capture reaction with 10B (see Prettyman [2011] and Prettyman et al. [2011] for descriptions of event categories). Second, the linear regression between the derived peak centroid channels and gamma-ray reference energies was taken to establish the relationship between the channel number and gamma-ray energy for each spectrum. Then, using the regression line, the uncorrected channel counts were converted to corrected channel counts with each new pulse height spectrum having equivalent energies of 8.9 × channel number (keV) (Lawrence et al. 2004; Prettyman et al. 2004a, 2004b, 2011; Prettyman and Feldman 2012). The cumulative energy spectra for LAMO and Survey corrected for gain are shown in Fig. 1 with key peaks identified. For LAMO, the peak widths in the cumulative spectrum were reduced by 16% at 4.4 MeV and 10% at 6.1 MeV following the gain correction, which contributed to an increase in signal-to-noise ratios of net counts of peaks of interest. Peak centroids were well calibrated by this method within approximately 3% of the reference energies at most, providing reliability for peak identification. Net counts of peaks were preserved by the gain correction algorithm.

Determination of Peak Areas

The net counting rate of the Fe gamma ray peak was determined for all the time-ordered records by subtracting background counts from the gross counts within the 7.6 MeV peak region of interest (ROI) as illustrated in Fig. 2.

Figure 2.

Energy spectra of gamma rays from 6.2 to 9.1 MeV during LAMO and Survey. The baselines for the Fe peak (shown by the gray lines) were determined by fitting ROIbase1 and ROIbase2 (defined by dotted lines) by a single exponential curve to derive the net counts.

The ROI was determined based on peak centroid energies of Fe and surrounding gamma ray peaks. More specifically, the ones at 7.631 and 7.646 MeV have the highest intensities among many neutron-capture gamma rays emitted from Fe, and have energies higher than gamma rays produced by most other elements (Reedy 1978; Prettyman et al. 2006). The BGO sensor on GRaND had a prelaunch energy resolution of approximately 10% at 0.662 MeV at room temperature (Prettyman et al. 2011), which is insufficient to resolve the two Fe capture gamma rays at 7.6 MeV. We, therefore, treat the combined response of the two gamma rays as one peak. We take the peak ROI, denoted as ROIFe in Fig. 2, from 7.36 to 7.97 MeV to maximize the statistics and to minimize the contribution from the nearby peak centered at approximately 7.1 MeV, which is created by inelastic scattering of neutrons with O and single escape peaks of the 7.6 MeV Fe gamma rays.

Based on the peak-fitting analysis of other distinctive peaks in spectrum shown in Fig. 1, namely Al at 2.211 MeV, C at 4.438 MeV, and O at 6.129 MeV, the BGO detector has an energy resolution of 5.1% or 390 keV FWHM at 7.638 MeV (mean of the two Fe gamma rays) for the total LAMO spectrum. The peak ROI was determined such that the upper bound is 327 keV above the peak centroid and the lower bound is 278 keV below the centroid. Assuming the Fe peak has a Gaussian shape (Knoll 2010; Prettyman et al. 2011), these correspond to 1.97 sigma for the higher bound and 1.67 sigma for the lower bound or approximately 93% of Fe counts covered by this peak ROI. The lower bound was set to be closer to the centroid because of the adjacent 7.1 MeV peak.

The possible effect of Al gamma rays at 7.7 MeV is discussed in the Appendix. All the gamma ray events that had pulse heights from the BGO detector within the peak ROI were assumed to originate from Fe or were part of the underlying gamma ray continuum.

Background Determination

Gamma rays and other photons with energies higher than the peak form a continuum background, on which the Fe peak resides. Therefore, the background counts within the peak ROI must be subtracted from the gross Fe counts to determine net Fe counts.

The background for the Fe peak was determined using the higher energy portions of the spectrum, where a few weak peaks exist, as a baseline and fitted to a curve for extrapolation to lower energies (e.g., see Lawrence et al. 2002). In the absence of Fe gamma rays, the Survey spectrum indicates that this region has the form of an exponential, as shown in Fig. 2.

ROIbase1 ranges from 8.05 to 8.37 MeV and ROIbase2 ranges from 8.90 to 9.03 MeV. The region between 8.37 and 8.90 MeV was excluded from the modeling of the baseline because of the presence of a weak structure in the spectra, which can be attributed to one or more peaks that cause the region to deviate from an exponential (Fig. 2). Peaks around 8.6 MeV are also observed by other planetary gamma-ray spectrometers, such as those on Lunar Prospector (Feldman et al. 2004; Lawrence et al. 2004; Prettyman et al. 2006), Mars Odyssey (Boynton et al. 2004; Evans et al. 2006), and SELENE (Kaguya) (Hasebe et al. 2008; Yamashita et al. 2009) spacecraft. All the counts below the baseline within the peak ROI were treated as background.

After the background subtraction, the net counts were corrected for variations in live times, solid angles of Vesta, and the intensity of galactic cosmic rays. These corrections followed the procedures described by Prettyman et al. (2012). The largest correction was for solid angle. The solid angles subtended by Vesta at the spacecraft were calculated for each measurement location from spacecraft ephemeris data, as described by Prettyman et al. (2011, 2012). The Fe net counting rates were found to vary in proportion to the solid angle of Vesta (Fig. 3). Using the proportionality relationship, the Fe time-series counting rates were normalized to an equivalent altitude of 210 km, roughly the average distance from the surface of Vesta during LAMO. This was accomplished by dividing the counting rate for each measurement by the solid angle (steradians) of Vesta subtended by the spacecraft at the midpoint location of the measurement and multiplying the result by 1.07 steradians, which is the solid angle of a 265 km radius sphere at an altitude of 210 km.

Figure 3.

Fe counting rates measured during LAMO as a function of solid angles subtended by Vesta at the spacecraft. The Fe count data used in this plot are corrected for differential nonlinearity and accumulation time, but not for variations in the cosmic ray flux and solid angle. The correlation coefficient R is 0.957 with a slope of 0.0695 and a very small offset of −0.00144, indicating that the signal is from Vesta and not from the spacecraft.

The temporal variation in the flux of galactic cosmic rays was normalized to its mean value in LAMO using the counting rate for coincidences with three or more sensors of GRaND. Such coincidence events were interpreted to occur when a high-energy charged particle penetrated through the GRaND sensors and therefore provide a good proxy for galactic cosmic rays (Prettyman et al. 2011).

The gamma-ray data acquired when the spacecraft is nadir pointing (<5° between the +Z axis of the spacecraft and the center of Vesta) were used for mapping. The average Fe-ROI net counting rate was approximately 0.074 counts s−1. The absence of the 7.6 MeV peak in a spectrum acquired far from Vesta (in Survey, about 2700 km altitude) in Fig. 2 and the very small offset (−0.00144) in Fig. 3 show that the 7.6 MeV signature is of vestan origin, and not due to cosmic-ray interactions with the spacecraft.

Neutron Number Density Correction

As the Fe gamma rays measured by GRaND are produced by neutron capture reactions, the Fe counting rate (CFe) is proportional to the abundance of Fe (wFe) as well as the number density of neutrons slowing down in the regolith (Feldman et al. 2000; Prettyman et al. 2006, 2013), i.e., CFe ∝ wFeN, where N is the neutron number density. Thus, Fe abundance is not strictly proportional to the raw counting rate of gamma rays at 7.6 MeV. This was first demonstrated by Lawrence et al. (2002) for measurements of lunar Fe; however, the algorithms developed for lunar measurements cannot easily be extended to Vesta. With the exception of high latitudes, the lunar surface has very low (approximately 50 ppm) abundances of hydrogen. Consequently, the effect of hydrogen was neglected in determining lunar neutron number density. The abundance and variability of hydrogen on Vesta (Prettyman et al. 2012) strongly influence the near-surface energy distribution of neutrons. Thus, number density corrections must account both for the effect of neutron absorption (e.g., by Fe, Ca, Al, Ti, and Gd) as well as moderation by hydrogen.

Using numerical simulations, an algorithm was developed to establish the relationship between neutron counting rates and neutron number density (see Prettyman et al. 2013 for a detailed description and derivation of this algorithm). The number density was found to vary as a power law with the “thermal plus epithermal” counting rate measured by GRaND's +Z phoswich. The coefficients of the power law depend on hydrogen content, which was determined from measurements of epithermal neutrons (Prettyman et al. 2012). Measurements of thermal and epithermal neutrons by GRaND were used to develop a map of neutron number density (Fig. 4), which was used to correct a map of raw Fe counting rates to determine the relative variation of Fe abundances on Vesta.

Figure 4.

A map of neutron number density on the surface of Vesta derived from the measurements of thermal and epithermal neutrons by GRaND (Prettyman et al. 2013). Approximate boundaries of the Rheasilvia (R.S.) and Veneneia (V.) impact basins are indicated by white broken curves. The maps presented in this work are in the Claudia coordinate system with the east longitude convention and the prime meridian at the edge of the maps (Russell et al. 2012).

The correction was carried out by dividing a map of raw 7.6 MeV net counting rates by the neutron number density map. The mapping process, which involves spatial gridding of the Fe time-series counting data and smoothing, is described in the 'Results' section. Once corrected for number density, the mapped counting rates are directly proportional to Fe abundances on Vesta.

Results

The time-series counting rates for the Fe gamma rays were first gridded spatially on a 5° quasi-equal-area map (Lawrence et al. 2003). As the Fe counting rate measured by GRaND is as low as one-tenth of fast neutrons (Prettyman et al. 2012; Lawrence et al. 2013b) and one-hundredth of thermal plus epithermal neutrons (Prettyman et al. 2012, 2013), the statistical precision of the 5° pixels is very poor. To improve the statistical precision, the pixels were averaged over the neighboring pixels within the radius of 30° with no weight (i.e., boxcar filtering). The averaged pixels were then rebinned into 30° equal-area pixels to avoid autocorrelation. Finally, the 30° pixels were smoothed using the spline technique (Smith and Wessel 1990).

The global Fe counting rate map resulting from our data processing procedures is shown in Fig. 5. The statistical uncertainty of the mapped data following smoothing with the boxcar filter was evaluated by deriving sample standard deviations for each pixel, which varied from 0.0005 to 0.0023 counts s−1. The Fe counting rate varied from 0.071 to 0.076 counts s−1. The derived deviation relative to the Fe counting rate, σμ, was smoothed for display in the same way as in Fig. 5 and mapped (Fig. 6). The uncertainty of the Fe counting rate was principally driven by accumulation time, which was longest near the poles and shortest near the equator. Thus, the latitude variation of the uncertainty is much larger than that of longitude (Fig. 6).

Figure 5.

A map of Fe counting rates on the surface of Vesta corrected for variations of live times, solid angles of Vesta, and intensity of galactic cosmic rays, but before neutron number density correction is applied. Approximate boundaries of the Rheasilvia (R.S.) and Veneneia (V.) impact basins are indicated by white broken curves.

Figure 6.

Statistical uncertainty (1-sigma) of Fe counting rates before neutron correction. Approximate boundaries of the Rheasilvia (R.S.) and Veneneia (V.) impact basins are indicated by white broken curves.

The counting rate map was then divided by neutron number density to determine a fully corrected map that is proportional to the abundance of Fe shown in Fig. 7. The corrected map exhibits the highest Fe abundance in the northern hemisphere. The hydrogen-rich regions discovered by neutron observations near the equator (Prettyman et al. 2012) have, in general, high Fe abundances, as does the portion of the Veneneia basin remaining after formation of Rheasilvia. The Fe counting rate is lowest in Rheasilvia in the south polar region. Two low-Fe regions extend from the Rheasilvia rim northward as two lobes across the equator; one at 60° lon. and the other at 300° near a topographic low containing the Oppia crater. The two lobes forming a V-shaped distribution near the south pole are clearly seen in the polar map Fig. 8.

Figure 7.

Global distribution of Fe counting rates on the surface of Vesta. The gamma-ray counts were fully corrected for variations of neutron number density and abundance of hydrogen. The map has been smoothed with a boxcar filter having a radius of 30°. The Rheasilvia and Veneneia basins are indicated by white broken curves. Eastern and western low-Fe lobes are delineated with yellow dots. H-rich regions are shown with blue dots. These approximate the 320 μg g−1 H contours from Prettyman et al. (2012). The low-Fe, intermediate-absorption region, indicative of cumulate-eucritic materials (see text), is presented with a red broken circle. The map is drawn using a simple cylindrical projection.

Figure 8.

Distribution of Fe counting rates in the southern hemisphere from −30° to −90° in latitude with stereographic projection. The same color bar and smoothing technique were used as those in Fig. 7. The Rheasilvia and Veneneia basins, and the central massif of Rheasilvia are indicated by white broken lines, with a cross showing the center of Rheasilvia (Schenk et al. 2012). Eastern and western low-Fe lobes are delineated with yellow dots. The map is drawn using a stereographic projection.

Discussion

The relative variation of corrected Fe counting rates was approximately ±6% from the mean value after smoothing, which is very small compared with that of the Moon (approximately 9×; Lawrence et al. 2002; Prettyman et al. 2006) or Mars (approximately 2×; Evans et al. 2006). The small range of Fe observed at Vesta is consistent with HED compositions that show that the Fe abundance in howardites varies by only 40% (Prettyman et al. 2012).

Based on the agreement between the observed elemental composition of Vesta and HED meteorites (Prettyman et al. 2012), and the fact that the background interferences (e.g., Al) were found to be negligibly small (see Fig. 2 and Appendix), we scaled the average counting rate to the mean Fe content of 13.8 wt% of howardites. This scaling is justified, given that the global Fe/O ratio determined by GRaND is within the range of howardites (Prettyman et al. 2012). The corresponding Fe values are compared with those of the HED clan (see Supplementary Material of Prettyman et al. 2012) in Fig. 9 with their mean values, the full ranges, and standard deviations. The GRaND measurements do not span the full range of HEDs; however, with a spatial resolution of 300 km (Prettyman et al. 2011) compared with the decimeter sizes typical of meteorites, it is expected that the full range of HED endmembers would not be sampled.

Figure 9.

Comparison of ranges for Fe abundances in howardite, eucrite, and diogenite (HED) meteorites and of GRaND observation. The circle and diamonds indicate the mean values, and the solid and dotted error bars denote standard deviations (1σ) and full ranges, respectively. The abbreviations CE, D, H, PE, and BE stand for cumulate eucrites, diogenites, howardites, polymict eucrites, and basaltic eucrites. The HED Fe abundances were derived from the Supplementary Material for Prettyman et al. (2012).

The regions of high Fe counting rate extending from the north pole, through the Marcia crater, and further southward to part of the remnant of the Veneneia basin are consistent with more basaltic-eucrite-rich materials (Fig. 7). The high Fe region in the equatorial latitudes coincides with an H-rich region that is associated with the exogenic accumulation of carbonaceous chondrites (Denevi et al. 2012; De Sanctis et al. 2012b; McCord et al. 2012; Prettyman et al. 2012; Reddy et al. 2012a). The presence of relatively high abundances of Fe in this region supports the idea that its surface was not blanketed by ejecta from large impacts that might have distributed diogenitic materials (Prettyman et al. 2013). As the infall of carbonaceous chondrites that delivered 400 μg g−1 of hydrogen would only increase the Fe abundance in this region by less than half a percent (Prettyman et al. 2013), our interpretation that the high Fe content results from the presence of basaltic-eucrite-rich materials is consistent with the interpretation of H delivery (Prettyman et al. 2012). Framing Camera (FC) color ratio maps and Visible and Infrared Spectrometer (VIR) pyroxene band centers and depths in this region are consistent with high abundances of FeO-rich pyroxene in eucrite and thus further support this interpretation (De Sanctis et al. 2012a; Reddy et al. 2012b; McSween et al. 2013a; Prettyman et al. 2013).

The lowest Fe counting rate occurs near the center of Rheasilvia basin as illustrated in Figs. 7 and 8. In addition, there are two notable low-Fe lobes that extend northward from the Rheasilvia basin. Because diogenites and cumulate eucrites contain less Fe than many basaltic/polymict eucrites or howardites (Supplementary Material of Prettyman et al. 2012), this Fe observation generally supports the characterization of these surfaces as being rich in one or both of these lithologic types. The distribution of such low Fe in the lobe covering 15–75° lon. combined with the eastern portion of Rheasilvia is very similar to other observations by Dawn, such as neutron absorption by GRaND's neutron spectrometer (Prettyman et al. 2013), the pyroxene BII center measured by VIR (McSween et al. 2013a), and the R(0.98)/R(0.92) ratio by FC (Reddy et al. 2012b), all of which describe these regions as diogenite-rich. Combined with observations from the HED suite that diogenites are more numerous than cumulate eucrites, and that cumulate eucrite material is rarely found in howardites (McSween et al. 2011; Mittlefehldt et al. 2013), we favor exposure and/or deposition of diogenitic lithologies as the cause of the low-Fe regions in the Rheasilvia basin and the lobe at 60° lon.

Differences arise in the other lobe of low Fe covering 225–330° lon. in the western hemisphere. This lobe originates near the center of Rheasilvia located at −75° lat. and 301° lon. (Schenk et al. 2012) and its tip extends to the geographic depression located at approximately 15–45° lat. and 225–270° lon., although there is a slight increase in counting rates near the equator. The region of the lowest Fe counting rates shown in purple in Fig. 8 coincides with the central massif of Rheasilvia (180 km or approximately 40° wide, Schenk et al. 2012). The low-Fe lobe centered at 300° crosses the basin's border to the west of the Oppia crater. The lobe contains regions mapped as rich in eucrite by FC and VIR (270–330° lon., 0 to −60° lat.). This region has relatively low-to-intermediate values for neutron absorption and the compositional parameter derived from high-energy gamma rays, inconsistent with high abundances of basaltic eucrite (Peplowski et al. 2013; Prettyman et al. 2013). A partial explanation for the discrepancy between nuclear spectroscopy and reflectance spectroscopy is the difference in the spatial resolution (300 km FWHM versus <100 m per pixel in LAMO). For example, instrument spatial mixing of diogenitic materials within the Rheasilvia basin with materials outside the basin may suppress Fe counting rates near the boundary, where pyroxene-rich materials are observed by VIR and FC.

Another possibility for explaining the apparent discrepancy in observations is differences in mineralogy, mineral abundances, or mineral composition. For example, cumulate eucrites may be a major component in the low-Fe region in the western lobe. Due to their low Fe contents, cumulate eucrites are indistinguishable from diogenites, but are readily separated from basaltic and polymict eucrites, which have relatively high abundances of Fe (Fig. 9). In contrast, the band centers for 1 and 2 μm absorption features in cumulate eucrites overlap with those in other eucrites (e.g., McSween et al. 2013b), but can be distinguished from the shorter wavelength band centers of diogenites. The differences in sensitivity between nuclear and reflectance spectroscopy provide opportunities to distinguish between cumulate and basaltic eucrite materials. As much of the western lobe maps as low Fe, but also contains regions mapped as rich in pyroxene with band centers at longer wavelength than those of diogenites, the lobe may contain a cumulate eucrite component.

A cumulate eucrite signature can also be detected by comparing nuclear spectroscopy measurements. For example, neutron absorption for diogenites is expected to be low, whereas intermediate values for neutron absorption are expected for cumulate eucrites (see Prettyman et al. 2013). Therefore, neutron absorption may help separate cumulate eucrites from diogenites when combined with Fe observations. As described by Prettyman et al. (2013), the neutron absorption map follows a similar trend as the Fe map, except for that the western lobe does not appear prominently in the absorption map. When spatially smoothed to the same degree as the Fe map (see the 'Results' section) and rebinned to the quasi-equal area pixels of 15° × 15°, values of neutron absorption are generally linearly correlated with values of Fe; however, a spur extending downward at low Fe and intermediate-absorption values is evident in this scatterplot (Fig. 10). These outlier values correspond to the tip of the western Fe lobe approximately from 8 to 53° lat. (circled in red in Fig. 7). A similar pattern is observed for the HED meteorites (Fig. 11) (Usui and McSween 2007; Usui et al. 2010), which show a low-Fe, intermediate-absorption spur that corresponds to cumulate eucrite compositions. The fact that the absolute values are shifted may be attributed to the instrument spatial mixing. Note that the tip of the western lobe is not necessarily the only possible location for a cumulate eucrite component; however, this component is distinguished from other HED types using Fe and neutron absorption only in that region (Fig. 11). The low-Fe region near the western rim of Rheasilvia (approximately 270–330° lon., −15° lat.) could also be explained by cumulate eucrites based on VIR/FC observations of eucrite-rich material in this region. Variations in abundances of phases other than pyroxene (e.g., Fe, Ti-oxides, plagioclase) would affect total Fe as measured by GRaND without changing the concentration of ferrosilite in pyroxene as measured by VIR (Frigeri et al. 2013).

Figure 10.

Comparison of fully corrected Fe counting rates (this work) and neutron absorption (Prettyman et al. 2013) observed by GRaND. Each marker represents a quasi-equal area pixel equivalent to 15° × 15° at the equator of Vesta. The Fe data presented in Fig. 7 were rebinned to the 15° pixels to improve statistics. The neutron absorption data were smoothed to the same degree as Fe (see the 'Results' section) and then rebinned to the 15° pixels to compare with the Fe data. A linear major trend was found between the two datasets with a correlation coefficient R of 0.81, while a spur extending downward at low Fe and intermediate absorption values is evident, which corresponds to the tip of the western Fe lobe circled in red in Fig. 7.

Figure 11.

Comparison of Fe abundance and neutron absorption for howardite, eucrite, and diogenite meteorites (Usui and McSween 2007; Usui et al. 2010). A spur from the major linear trend at low Fe and intermediate, which is also seen for GRaND data (Fig. 10), corresponds to cumulate eucrite compositions.

Conclusions

We have observed Fe gamma-ray signals from Vesta using the BGO detector of GRaND on Dawn and developed an analysis method to derive a global map of Fe counting rates that are directly proportional to elemental abundances of Fe. The counting rates were fully corrected for temporal and spatial variations of galactic cosmic rays, solid angles of Vesta, live times, and neutron number density. The potential Al contamination to the Fe signals was shown to be negligible. The resultant Figs. 7 and 8 indicate nonuniformity of the elemental Fe distribution in the regolith of Vesta that is correlated with regional geographic features such as the Rheasilvia basin.

The distribution of Fe can be used as an indicator of the lithology of regolith materials, and provides independent information to supplement or confirm mineralogic and other nuclear spectroscopy observations. The GRaND observation revealed two lobes of low Fe extending from the Rheasilvia basin. The one in the east is consistent with reflectance and neutron absorption measurements indicating the presence of diogenite, while the swath of low Fe concentrations in the west has only been observed by nuclear spectroscopy and is possibly indicative of cumulate eucrites.

Acknowledgments

We thank T. Usui and two anonymous reviewers for their valuable comments and suggestions that led to improvements in this manuscript. The Dawn mission is led by the University of California, Los Angeles, and managed by Jet Propulsion Laboratory (JPL) under the auspices of the NASA Discovery Program Office. A portion of this work was carried out by the Planetary Science Institute under contract with JPL, by JPL under contract with NASA, and by NASA's Dawn at Vesta Participating Scientist program. Science experimental data records and housekeeping data acquired by GRaND during Vesta encounter and used in this study are available from NASA's Planetary Data System.

Editorial Handling

Dr. Hap McSween

Appendix

Aluminum Interferences

Given the ultramafic to basaltic nature of howardite, eucrite, and diogenite (HED) meteorites, Al is expected to be low in abundance on Vesta compared with the Moon, where anorthositic rocks dominate much of the surface (Prettyman et al. 2006, 2012; Usui and McSween 2007; Usui et al. 2010). However, the dynamic range of Al is quite large for the HEDs and the Al content of eucrites is far from negligible in terms of wt%, such that it is of interest to quantify the effect of Al content on intensity in the “Fe” peak region of the spectrum. In the vicinity of the Fe gamma rays at 7.6 MeV, two 27Al gamma rays exist, which could potentially interfere with the Fe gamma ray counting rates depending on the elemental abundances. As the BGO detector cannot fully resolve the Al and Fe gamma rays, the analytical and numerical methods we employed to reject Al interference contributions to Fe are described below.

Aluminum on Vesta

The Al gamma rays in this energy region are created by neutron capture reactions, as are the Fe gamma rays, and have energies of 7.693 and 7.724 MeV, the latter being more intense. The emission rate of neutron-capture gamma rays from planetary surfaces is proportional to the product of: (1) microscopic cross sections, (2) the number density of the target element, and (3) the neutron number density (Lawrence et al. 2002). Since the 27Al and 56Fe nuclides are exposed to the same flux of neutrons in the subsurface of Vesta, we only need to calculate the product of (1) and (2) microscopic cross sections and the neutron number density or macroscopic cross sections to estimate their relative intensities.

The microscopic cross sections σγ of Al for producing 7.693 and 7.724 MeV gamma rays are 0.0081 and 0.0493 barns, respectively, or 0.0574 barns as a total contribution to the 7.6 MeV peak in the BGO spectra, while those of Fe for 7.631 and 7.646 MeV are 0.653 and 0.549 barns or 1.20 barns as a total (Firestone 2007). These cross sections were then multiplied by the elemental number densities of two endmembers of HED meteorites, a diogenite Shalka (with the lowest Al/Fe ratio) and a cumulate eucrite Serra de Magé (with the highest Al/Fe ratio) shown in Table AA1 (derived from table 2 of Usui and McSween 2007 and references therein).

As a result, the macroscopic cross sections Σγ of Al and Fe in the two meteorites were calculated and shown in Table AA1. Note that all the cross sections discussed here are for gamma-ray emission near 7.6–7.7 MeV and differ from elemental totals or neutron absorption cross sections. The ratio of the macroscopic cross section of Al to that of Fe is 0.0025 for Shalka and 0.062 for Serra de Magé. In other words, the Al contribution to the 7.6 MeV gamma-ray peak is 0.25% for the diogenitic surface and 6.2% at most for the cumulate-eucritic surface compared to Fe. As most howardites have intermediate Al/Fe ratios with the ratios of macroscopic cross sections of approximately 3%, a typical interference from Al on Vesta is expected to be between the two extreme cases. Because the energies of the Al and Fe gamma rays are so close, the attenuation within the regolith and the detection efficiency would be essentially the same and cancel out when the ratios are taken.

This estimate is supported by numerical simulations calculated for the Moon. Apollo 11 samples have an Al/Fe ratio similar to cumulate eucrites (Prettyman et al. 2006; Usui and McSween 2007). The simulated gamma-ray spectrum for Apollo 11 material shows that the Al counting rate near 7.6 MeV is approximately 5% of Fe (fig. 17 of Prettyman et al. 2006). Therefore, the contribution of Al from Vesta to the Fe counting rate is considered to be negligible.

Aluminum on Spacecraft

Aluminum on the Dawn spacecraft could also interfere with the 7.6 MeV peak. Such interferences have already been minimized by the design concept of GRaND, which employed a noryl canister for detector housings rather than Al metal. Emission of the 7.7 MeV Al gamma rays from Dawn is induced by either low-energy neutrons directly created within the spacecraft by galactic cosmic rays, or neutrons leaking from Vesta's surface. These two cases are evaluated experimentally as follows.

The former interference case was evaluated by the Survey spectrum for which the intensity of GCR was at a maximum because Dawn was not shielded from GCRs by Vesta while the Fe signal is negligible. It turned out that the Survey spectrum contained no signature for Al (or Fe) within the 7.6 MeV ROI (Fig. 2).

The latter case, leakage neutrons, was evaluated spectroscopically. If the Al contribution was a large factor in the 7.6 MeV peak, then its peak centroid would shift toward 7.724 MeV, an energy of the major Al gamma ray, when the leakage neutron flux is higher. We test this hypothesis by comparing gamma-ray spectra accumulated over high and low neutron number density regions (Prettyman et al. 2013) as shown in Fig. A1. However, we did not observe any peak shift for the two extreme neutron leakage conditions.

Moreover, raw neutron counting rates were higher in the southern hemisphere than in the equatorial region (see Supplementary Material for Prettyman et al. 2012). If the leakage neutrons induced significant amount of Al gamma rays in GRaND surroundings, then higher 7.6–7.7 MeV counting rates would be expected for the south pole than the equator. However, such spatial signature for the contamination has not been observed either. In addition, MCNPX simulations of gamma rays produced by vestan neutrons in the instrument and spacecraft structure at LAMO altitudes indicate that this interference source is small in comparison to the gamma ray signal from Vesta (Prettyman et al. 2012).

It would be extremely difficult to quantitatively assess the contribution of local Al any further at this time. Some 7.6 MeV gamma rays could be of local origin; however, the lack of above-mentioned spatial signatures for the contamination strongly suggests that at the very least Al cannot be the dominant contributor. Thus, we conclude that Al contribution from the spacecraft can be treated as negligible to the observation of Fe.

Table A1. Compositions and emission cross sections for the end members of the HED meteorites
 σγ (barn)DiogeniteCumulate eucrite
ShalkaSerra de Magé
wt%γ (g cm−2)wt%γ (g cm−2)
  1. σγ: microscopic cross sections for emitting gamma rays near 7.6–7.7 MeV (see text).

  2. γ: macroscopic cross sections for emitting gamma rays near 7.6–7.7 MeV for each element.

  3. wt%: Derived from table 2 of Usui and McSween (2007).

Al0.05740.324.07 × 10−64.726.04 × 10−5
Fe1.2012.71.64 × 10−37.569.79 × 10−4
Ratio of ∑γ  0.25% 6.2%
Figure A1.

Comparison of the Fe peaks accumulated over high and low neutron number density (NND) regions. The ordinates are normalized at ROIbase1. If the Al gamma rays induced by leakage neutrons contributed significantly to the 7.6 MeV peak, then its peak centroid should have been shifted toward the high-energy side in the high-NND spectrum.

Ancillary