The Moon Mineralogy Mapper (M3) acquired high spatial and spectral resolution data of the Aristarchus Plateau with 140 m/pixel in 85 spectral bands from 0.43 to 3.0 μm. The data were collected as radiance and converted to reflectance using the observational constraints and a solar spectrum scaled to the Moon-Sun distance. Summary spectral parameters for the area of mafic silicate 1 and 2 μm bands were calculated from the M3 data and used to map the distribution of key units that were then analyzed in detail with the spectral data. This analysis focuses on five key compositional units in the region. (1) The central peaks are shown to be strongly enriched in feldspar and are likely from the upper plagioclase-rich crust of the Moon. (2) The impact melt is compositionally diverse with clear signatures of feldspathic crust, olivine, and glass. (3) The crater walls and ejecta show a high degree of spatial heterogeneity and evidence for massive breccia blocks. (4) Olivine, strongly concentrated on the rim, wall, and exterior of the southeastern quadrant of the crater, is commonly associated the impact melt. (5) There are at least two types of glass deposits observed: pyroclastic glass and impact glass.
 The Aristarchus crater is a prominent Copernican crater on the western nearside of the Moon. Its prominence is partly due to its anomalous brightness relative to other craters of similar age and size making it of interest for early telescopic observations. But its complex and diverse geology has made it of interest as an important target of geologic investigation. The crater was formed at the contact between Procellarum basalts and an uplifted block of lunar crust that rises approximately 2 km above the volcanic plains [McEwen et al., 1994] (Figure 1). The uplifted crustal rocks are thought to have been emplaced at the time of the Imbrium impact [Moore, 1965; Guest, 1973; Zisk et al., 1977] as the crust readjusted following the formation of the Imbrium Basin. The block is dominated by noritic compositional signatures [Lucey et al., 1986; Chevrel et al., 2009]. The region is also host to a spectacular assemblage of sinuous rilles with the largest. Vallis Schröteri is on the plateau west of the Aristarchus crater with numerous rilles extending into the mare basalts to the east of the Aristarchus crater.
 In addition to the diversity of geologic deposits and processes, the Aristarchus region shows a wide range of compositions. Telescopic data indicate that pyroxene, olivine, and feldspathic compositions are present [Lucey et al., 1986; Pieters, 1986], but these data lack the spatial resolution to be precise about the geologic units that hosted the compositions. Clementine data with its visible and infrared multispectral imaging clearly defined olivine-rich units on the southeastern rim and ejecta of the crater [Le Mouélic et al., 1999], and the central peaks were dominantly feldspathic based on their anomalous brightness [McEwen et al., 1994]. Elsewhere on the crater walls the deposits were interpreted to be dominated by noritic compositions typical of the upper crust [McEwen et al., 1994] but that the ejecta of the crater was a complex combination of highland and mare compositions mixed with impact melt. However, the low spectral resolution of the data limited the fidelity of the composition analyses of the central peaks, walls, impact melt and ejecta deposits [Le Mouélic et al., 1999; Chevrel et al., 2009].
 The central peak of Aristarchus, that represents the deepest material excavated by the crater, is important for understanding compositional stratigraphy in the region. The first extensive telescopic spectra of the central peaks were interpreted to be a brecciated assemblage dominated by feldspar but containing pyroxene and olivine [Lucey et al., 1986]. The Clementine data were interpreted to clearly show a 1.25 μm anorthite absorption band showing the central peaks to be dominated by anorthosite [McEwen et al., 1994]. Chevrel et al.  interpreted the central peaks to be more similar to what Lucey et al.  proposed. Recent results from the Selenological and Engineering Explorer (SELENE) mission [Kato et al., 2008] reported the central peaks of Aristarchus to be >98% anorthite on the basis of a resolved 1.25 μm absorption band detected with the multispectral MultiBand Imager [Ohtake et al., 2009]. It should be noted that both McEwen et al.  and Chevrel et al.  used the telescopic data from Lucey et al.  to bootstrap the calibration of the Clementine data while the multiband imager used the Apollo 16 landing site and laboratory spectral of Apollo 16 soils to bootstrap their calibration.
 These analyses show that Aristarchus is compositionally diverse and that new data with both high spectral and spatial resolution could significantly improve our understanding of the geological evolution of this region. Here we report on the first results of compositional analysis of the region in the immediate vicinity of the Aristarchus crater with the Moon Mineralogy Mapper (M3). M3 was a NASA guest instrument on Chandrayaan-1, India's first mission to the Moon. The spacecraft was launched on 22 October 2008, and the M3 instrument operated until the end of the mission on 29 August 2009. The complete descriptions of the instrument, its calibration, and data collection are given by Pieters et al. [2009a], Boardman et al. , and R. O. Green et al. (The Moon Mineralogy Mapper (M3) imaging spectrometer for lunar science: Instrument, calibration, and on-orbit measurement performance, submitted to Journal of Geophysical Research, 2011). We address five key compositional aspects of the Aristarchus region: (1) the central peaks, (2) impact melt, (3) the crater walls and ejecta, (4) olivine-rich deposits, and (5) glass (both pyroclastic and impact melt).
2. Regional Setting and Topography
 Aristarchus crater lies on the SE edge of the Aristarchus Plateau (Figure 1), a distinctive parallelogram-shaped plateau that rises 1–1.5 km above the surrounding Oceanus Procellarum. The Aristarchus Plateau lies west-southwest of the central part of Mare Imbrium and a few hundred kilometers outside of the extension into this region of the Apennine Mountain ring of the Imbrium impact basin. The Aristarchus Plateau and the Prinz crater uplands, about 100–150 km to the east of the plateau, therefore both lie just outside the main topographic ring of the Imbrium impact basin. These areas were thus heavily influenced by ejecta emplacement from the Imbrium basin-forming event. In addition to the 45 km diameter Copernican aged Aristarchus crater, Aristarchus Plateau is also the location of the 38 km diameter Imbrian aged Herodotus crater, the volcanic vent Cobra Head and Vallis Schröeter, a huge nested sinuous rille emanating from Cobra Head, extending across the plateau, and out into Oceanus Procellarum. Lunar Orbiter Laser Altimetry (LOLA) altimetry data [Smith et al., 2010] reveal the presence of a remnant heavily degraded ∼110 km diameter unnamed crater that lies just to the south of the Aristarchus Plateau (AP) (Figure 1). The rim crest of this crater, informally referred to as the South AP crater, cuts the southern margin of the plateau, forming a broad arch between the southeastern rim of Herodotus and the southwestern rim of Aristarchus. South AP crater is clearly pre-Imbrian in age, helps to form the distinctive scarp along the southern part of the Aristarchus Plateau, and its rim and ejecta are major contributors to the high topography that defines the high southern rim of the Aristarchus Plateau (Figure 1). Post-Imbrium basin mare basalts have extensively flooded South AP crater, filling it and embaying the rest of the plateau. Dark mantling deposits, associated with the formation of Cobra Head and the sinuous rilles on the plateau [Weitz et al., 1998], were likely emplaced during this time.
 This general geological setting provides a basic framework for the nature of the crustal target in which the Aristarchus impact crater formed (Figure 1). The Oceanus Procellarum was likely to have been a heavily cratered lowlands region prior to the formation of the Imbrium basin. Among the impact crater population was the 110 km diameter South AP crater, which appears to have impacted into a regional high that existed prior to the time of the Imbrium basin impact. Comparison of the crescentic South AP topographic crater rim crest with that of Prinz crater to the east (Figure 1) shows that when impact craters are embayed by mare basalts, their rim crests are preferentially preserved along high parts of preimpact topography. This suggests that at least part of the high topography associated with the Aristarchus Plateau existed prior to the formation of the Imbrium basin.
 Impact of the Imbrium basin projectile created a huge impact basin to the east and almost certainly delivered hundreds of meters of ejecta to the Aristarchus Plateau site, despite the asymmetric distribution of Imbrium secondaries toward the south of the basin [Spudis et al., 1994]. These ejecta mantled the plateau, modified and partially filled the South AP crater, and perhaps structurally modified the plateau margins. Subsequent to this time, and prior to the formation of Aristarchus, mare flooding of Oceanus Procellarum began and the Orientale impact basin formed to the south-southwest of Procellarum. Mare basalts continued to flood Oceanus Procellarum following the Orientale event, forming volcanic centers (Marius and Rumker hills) and extensive plains [Whitford Stark and Head, 1980; Hiesinger et al., 2000]. During this phase, Herodutus crater formed near the future site of Aristarchus crater and the plateau was blanketed with pyroclastic material associated with the eruptions that produced the sinuous rilles. Later emplacement of mare basalt units in the surrounding lowlands embayed this regional pyroclastic deposit.
 Examination of the location of Aristarchus crater in relation to the Aristarchus Plateau (Figure 1) shows that the impact occurred on the topographic margin of the plateau, with the southeastern part of the crater cavity probably sampling and excavating mare material, off of the edges of the plateau. As noted by Chevrel et al. , the impact crater formation in this region will lead to highly heterogeneous compositions in the ejecta.
 On the basis of this stratigraphic and topographic reconstruction, the target stratigraphy at the site of the Aristarchus impact should have consisted of the following components, in ascending stratigraphic order: (1) The first detectable event in this region was the South AP crater which impacted into already high topography and contributed ejecta to the target site, because Aristarchus crater lies directly on the crater rim. (2) Formation of the Imbrium basin emplaced as much as hundreds of meters of ejecta into the target site. (3) Following this event, possible cryptomare emplacement characterized this area, perhaps embaying the high plateau topography. (4) Formation of the Orientale basin may have contributed small amounts of ejecta to the region. (5) Emplacement of hundreds of meters of mare basalts surrounding the plateau followed the Orientale event but occurred off the margins of the topographically high Aristarchus Plateau; also during this time, formation of Herodotus crater on the plateau (and on the rim of the South AP crater) redistributed South AP crater ejecta into the Aristarchus target area to the east-northeast (Figure 1). (6) A tens of meters to locally several hundred meters thick layer of pyroclastics was emplaced on the plateau; in the surrounding mare, later basaltic flows embayed this pyroclastic layer.
 The Aristarchus cratering event target point appears to have been on the plateau itself, based on extrapolation of plateau topographic trends into the area below the crater (Figure 1). Excavation directly below and to the northwest of the target point thus sampled target stratigraphy primarily consisting of the rim and ejecta deposit of the South AP crater, overlying Imbrium ejecta, and overlying pyroclastic deposits. Excavation in the southeasternmost part of the cavity first intersected hundreds of meters of mare basalts before reaching down to the Imbrium ejecta and the degraded rim of the South AP crater. Thus, target stratigraphy sampled by the Aristarchus cratering event (the upper four kilometers of this column) is dominated by the ejecta of the South AP crater (likely to be upper crustal materials) and Imbrium ejecta, and to the southeast off the plateau, mare basalts overlying the same target sequence.
 Also important in reconstruction of the cratering event and its aftermath is the fate of impact melt in the short-term modification stages of the event. Hawke and Head  showed that preexisting topography in the target region plays a major role in the emplacement of melt on the crater rim. The asymmetric topography of the Aristarchus crater rim crest (several hundreds of meters lower to the southeast in the direction off the plateau; Figure 1) suggests that impact melt emplaced in terminal phases of the cratering event (during cavity collapse [Hawke and Head, 1977]) would favor expulsion of impact melt generated by the event, preferentially in this direction.
3. Moon Mineralogy Mapper Data Collection, Reduction, and Analysis Methods
 M3 is a push-broom imaging spectrometer designed to acquire 260 spectral channels from 0.43 to 3.0 μm simultaneously for each of 600 cross-track spatial elements [Pieters et al., 2009b]. Spacecraft motion provides a second dimension of spatial information, building a three-dimensional cube of inherently coregistered spectra. The first of four planned optical periods of Chandrayaan-1 operation extended through February 2009 as described by Boardman et al. . Over this period M3 acquired near-infrared low-resolution spectra for ∼60% of the lunar nearside with 140 m/pixel in 85 spectral channels from 0.43 to 3.0 μm). With initial calibration, these data have proved to be of high quality and the instrument performed fully within specifications (R. O. Green et al., submitted manuscript, 2011). The instrument has a radiometric accuracy of better than 90% that results in an accuracy of determining reflectance to better than 3% for a surface with a 30% reflectivity. Second-order calibration steps, including in-flight calibrations and band-to-band corrections (including Apollo “ground truth”) are ongoing and will continue to be refined. Lunar coordinates were assigned to each pixel allowing the data to be accurately map projected [Boardman et al., 2011].
 The reduction of M3 data from instrument units to apparent surface reflectance involves several steps as documented by R. O. Green et al. (submitted manuscript, 2011). Briefly, the detector background is subtracted using systematic measurements of the dark current that accompany M3 scene measurements. These dark corrected data are then converted to radiance using both preflight and inflight calibration measurements. From these calibrated radiance data an estimate of reflectance is generated by dividing by a solar spectrum and the cosine of the incidence angle. The solar spectrum is convolved to M3's wavelengths and resolution measured preflight and scaled to the Moon's solar distance.
 The eastern portion of the Aristarchus region was well covered during the first optical period of M3 (OP1b) and the western portion during optical period OP2a. We generated a mosaic of the available data acquired during these periods, where the data are filtered to prefer the smallest phase angle data when multiple observations existed for a location. The phase angle defined relative to the geoid is relatively uniform ranging from 34 to 39°. An overview of the study region with the M3 data coverage is shown in Figure 2. The data in mosaic and used in this analysis were corrected for thermal emission using the methods of Clark et al. .
 A series of mineral indicator parameters have been developed by the M3 team as a guide to analyses and to provide products that capture the fundamental mineralogic properties of the surface. The parameters were developed to capture the dominant modes of spectral variance related to mafic silicates, soil maturity, and space weathering. Initial analyses have shown two parameters that summarize the integrated band depth of the crystal field absorptions at 1 and 2 μm combined with the reflectance measured at 1.58 μm provide a excellent summary of the mineral diversity of the lunar surface with M3 data. The algorithm for determining the integrated band depths for the 1 and 2 μm bands are given by
where R refers to the reflectance at a given wavelength, Rc is the continuum reflectance defined as a straight line across the absorption band, 789 and 1658 are the first wavelengths, in nanometers in a series to be integrated over, 20 and 40 specifies the wavelength interval in nanometers, and n is the number of channels to be integrated over. These three parameters are shown together as a color composite for the study region in Figure 3 with the integrated band depth at 1 μm in red, 2 μm in green, and reflectance at 1.58 μm in blue.
 While mineral indicator parameters are useful guides, they can be equivocal and nonunique. For example several minerals have a strong 1 μm band (e.g., olivine, pyroxene) and the exact strength, shape, and position of a given absorption provides greater diagnostic discrimination. Thus, regions of interest based on the mineral indicators/spectral parameters are systematically investigated in detail with spectral analysis to verify and validate the presence of mineral spectral signatures. This requires extraction of full spectral resolution data and comparison of these data to spectral libraries. To enhance the expression of spectral absorption features, we also employ ratioing techniques. This technique is commonly used in the analysis of imaging spectrometer data from Mars [e.g., Mustard et al., 2005, 2008], where ratioing suppresses artifacts due to residual and systematic instrumental errors common to all spectra in a data set. Because many of these artifacts are multiplicative in nature, ratioing two spectra, where the numerator is extracted from an area of interest and the denominator from a region exhibiting a neutral or unremarkable spectral character, typically suppresses these artifacts. Due to the design of M3 as a push broom array, many artifacts are associated along the columns of the detector, and therefore as columns in the M3 imaging data. Thus, in calculating ratios it is beneficial that the numerator and denominator are extracted from the same column or columns in the case of area averages. Typically, we use at least 3 × 3 averages to increase the signal to noise in the resulting ratio spectra.
4.1. Aristarchus Central Peak Materials
 The central peak of Aristarchus is relatively small and compact (Figure 4a). As observed in recent high-resolution imaging data (Lunar Reconnaissance Orbiter Camera (LROC) [Robinson et al., 2010] and the Terrain Mapping Camera (TMC)) the central peak can be defined as a mound approximately 1 × 3 km in size. The peak has three distinct zones from north to south defined by apparent albedo relative to the floor materials, where the northern section is bright, the middle section is similar in albedo to the crater floor, and the southern section is intermediate in albedo between the northern and middle sections, and somewhat mottled in tone. There are abundant blocks at the surface on the northeast and southeast quadrants of the mound. The central peak mound appears to be embayed by floor material. The floor exhibits a hummocky morphology where the hummocks exhibit a range in albedos but each hummock has a characteristic albedo. The floor between the hummocks is rough textured and exhibits fractures suggestive of cooling cracks.
 The M3 reflectance spectra of the central peak material are shown in Figure 4b along with a spectrum of that is representative of typical mature highland soils. The mature highland soil spectrum was taken from the same M3 observation 560 km north of the central peaks. The central peak spectra are very bright and characterized by a distinct lack of diagnostic mafic mineral absorptions. The small apparent feature near 1.350 μm is likely a residual artifact of the data processing and calibration. This artifact is confirmed by the spectrum shown in Figure 4c that is a ratio of spectrum B to the highland spectrum (the spectra from other regions of the central peak show essentially identical ratio spectra). The ratio spectrum is featureless except for a negative slope, a small positive feature near 0.90 μm, and a sharp positive feature centered at 0.550 μm. The central peak spectra do express the distinct drop in reflectance between 2.6 μm and 2.9 μm that is characteristic of OH/H2O absorption [Pieters et al., 2009b; Clark, 2009; McCord et al., 2011].
 The M3 spectra show the materials in the central peaks to be very bright and lacking any distinctive Fe-related absorptions. This is consistent with a high abundance of feldspathic components. However, the lack of absorptions diagnostic of crystalline anorthosite shows that the rocks are extremely iron poor or have been shock processed to the point of removing any indications of small Fe absorptions typical of anorthosite [Bell and Mao, 1973; Adams and Goullaud, 1978; Pieters, 1986]. Previous analyses of the central peaks of Aristarchus with Clementine data [McEwen et al., 1994; Le Mouélic et al., 1999] concluded that the central peaks were rich in feldspar likely anorthite. Telescopic data [Lucey et al., 1986] and a reexamination of Clementine data [Chevrel et al., 2009] suggested that peaks were dominated by feldspar but with mafic components (pyroxene, olivine) contributing to a composite 1 μm band. Multispectral data from the Multiband Imager instrument on the SELENE mission indicated a 1.25 μm absorption diagnostic of Fe-bearing crystalline plagioclase [Ohtake et al., 2009]. In contrast the spectral properties of the central peaks as shown by M3 data are characterized by a lack of any definitive ferrous absorptions and a high albedo. Thus, there are no diagnostic absorption features, but the mineral that likely contributes to the high albedo is feldspar.
4.2. Impact Melt: Crater Floor, Exterior Regions
 The parameter mapping with M3 in Figure 3 show a distinctive spectral unit on the crater floor and on the eastern, southern, and western crater rim that appears blue in the parameter composite. This color in the ratio composite indicates that these regions have weak to absent 1 and 2 μm absorptions. We have extracted M3 spectra for the deposits that share these characteristics from the floor, the eastern rim, and the deposits exterior to the rim, also in the eastern region. These spectra are shown in Figure 5. Like the spectra of the central peaks, these reflectance spectra are relatively featureless, with a moderate spectral slope and a lack of 1 and 2 μm mafic mineral absorptions. However, compared to the central peak spectra they are a factor of 2–3 less bright. The ratio spectra confirm these properties and show that relative to typical highland soil, there are no distinct mineralogic absorptions. There is a sharp rise in reflectance between 0.45 and 0.8 μm that is not typical of the highland soils. However, the possibility of residual scattered light within the instrument optics in this wavelength region (R. O. Green et al., submitted manuscript, 2011) leads us to be cautious about interpretation of these properties. All the spectra of this unit are quite similar, and appear to be offset from each other by a simple linear factor.
 LROC and TMC data of these deposits show clear evidence of flow of a viscous material, cracking interpreted to be due to cooling (Figure 4), and pooling into ponds. These characteristics are typical of impact melt. Figure 6 shows an overview of the Aristarchus crater imaged by the TMC. A full resolution subset is shown in Figure 7 that includes the melt deposit on the floor as well as the crater wall. The texture of the deposits of the floor is hummocky with abundant mounds that range in size from the limit of resolution for the camera to 10s of meters (see also Figure 4). Individual mounds are relatively uniform in brightness but the collection of mounds show a range in albedo while the floor deposits between the mounds are typically uniform in albedo. Throughout the flat regions of the floor deposits are cracks and fractures suggestive of cooling fractures (Figures 4 and 7). Small, smooth and apparently flat deposits occur frequently on the walls on terraces or in small topographic depressions (Figure 7). These appear to be melt deposits that have accumulated on the walls. The walls also show numerous channels perpendicular to the wall slope and appear to have been pathways of drainage for impact melt (Figure 7). Exterior to the crater rim, a similar collection of morphologic features are commonly observed, including melt ponds and channels that drained accumulated melt (Figure 8).
 The spectroscopic signatures combined with the morphologic evidence lead to the interpretation that at the deposits are impact melt. The spectral signatures of the impact melt deposits are not confined to just the regions that have morphologic features that are interpreted as melt. They also occur across broader regions suggestive of a possible veneer of melt that has modified the spectral properties of the surface to be consistent with melt but that in high resolution LROC images but no morphologic expression. This is best illustrated in Figure 8. There are small ponds of melt indicated by the white arrows but the regions between the ponds are characterized by the spectral signatures of melt yet there is not clear evidence of ponding or flow of melt.
 The LROC camera provides extraordinary high-resolution images that allow further insight into the melt. LROC data for the northern rim of Aristarchus are shown in Figure 9. Note the small linear depression on the north rim that trends downslope to the crater interior (Figure 9a). In the depression is a zone of fractured rock that shows the characteristic of melt having filled the depression. At the highest resolution (Figure 9b) small bright blocks are resolved in the fractured melt suggesting this is a melt breccia perhaps analogous to the suevite of the Reis crater. However, within the same depression are small lobate deposits (Figure 9c) that may be zones of more discrete melt lacking breccia blocks.
4.3. Crater Walls and Ejecta
 Parameter mapping of the M3 data shows that the Aristarchus crater has a diverse suite of compositions represented in the walls and ejecta of the eastern portion of the crater (Figure 10). The crater shows a distinct difference between the northern and southern walls. The northern wall appears green in the parameter mapping indicating a strong 2 μm band compared to other regions. The southern wall shows as blue, indicating a lack of mafic absorptions consistent with impact melt, while the southeastern region of the rim is distinctly red indicating a strong 1 μm band but no 2 μm band consistent with the mineral olivine. The olivine deposit is discussed further below. The base of the eastern wall of Aristarchus shows the compositional diversity very well where discrete mounds are resolved that show differences in apparent composition based on the parameter mapping. We have draped the parameter mapping onto an LROC image to better resolve the morphology of the surface related to the compositional changes.
 We have extracted M3 spectra for four discrete mounds, and the reflectance and a continuum-removed reflectance are shown in Figure 10 (bottom right). (Continuum removal is performed by fitting a straight line between the reflectance at 0.73 and 1.6 μm and dividing the spectrum by the line defined by these points.) The more northerly mounds (A and B) are clearly green in the color composite parameter mapping and thus have extremely well developed 1 and 2 μm bands. The position of the 1 μm band is at 0.94–0.95 nm, and the 2 μm band is near 2.0 μm. The 2 μm band is a little more difficult to identify due to an unresolved component of thermal emission that is clearly evident beyond 2.2 μm. Regardless the position of the absorptions indicates a strong contribution from low-Ca pyroxene (LCP) and the presence of abundant LCP is typical of a dominantly noritic composition. However, the width of the absorptions and their band positions indicate that these is come component of high-Ca pyroxene present in these rocks as suggested by Chevrel et al. . The more southerly mounds (C and D) exhibit weak 1 μm absorptions near 0.95 μm and a weak to absent 2 μm band that, when present, is centered at or longer than 2.2 μm. The weak LCP bands or overall lack of absorptions for these mounds implies a strongly feldspathic composition but the moderate albedo indicates these are not as feldspathic as the central peaks. The association of distinct spectral signatures with individual mounds suggests that these are large blocks of relatively uniform composition. We interpret these distinct mounds here and on the floor as large breccia blocks emplaced by the impact process.
 In some very small locations on the crater wall (500 × 500 m) we observe distinct spectral signatures typical of lithologies more enriched in high-Ca pyroxene (Figure 11). The center of the 1 μm band is shifted to near 1.0 μm, and the 2 μm band to 2.2 μm. Interestingly these are very small and thus far only observed in the wall slopes. These spectral signatures are similar to that expected for a gabbroic lithology or for mare basalts.
4.4. Olivine-Rich Zones
 One of the more striking aspects of the mineral indicator mapping in Figure 3 is the high concentration of materials that have a strong 1 μm band but no 2 μm band; these are expressed as red in Figure 3. This combination of parameters is typical of regions rich in the mineral olivine, but lacking other mafic silicates. A number of spectra have been extracted from the regions appearing as red in Figure 3, and representative spectra are shown in Figure 12a. These spectra show a distinct absorption band between 0.8 and 1.7 μm with a weak to absent absorption near 2 μm. To better show the character of these absorptions, the reflectance values are ratioed to typical highland soils acquired from the same observation and same column (Figure 12b).
 The relative reflectance spectra clearly show the composite crystal field absorption band between 0.8 and 1.7 μm characteristic of olivine [Burns, 1993; Cloutis et al., 1986; Sunshine and Pieters, 1998; Isaacson et al., 2011]. The relative reflectance spectra have a clearly resolved band minimum near 1.1 μm and a weak inflection near 0.9 μm and a distinct inflection at 1.3 μm that collectively are characteristic of the diagnostic olivine crystal field bands. In some of the spectra in Aristarchus there is a well-resolved absorption centered near 2.2–2.3 μm, and example of which is shown in top spectrum in Figure 12b. Lunar olivines typically contain a few % of spinel (typically chrome spinel) [e.g., Isaacson et al., 2011] and a band near 2.2 μm is typically attributed to chrome-rich spinel, based on analyses of spinel spectra [Cloutis et al., 2004]. However, the 2 μm band could also be due to pyroxene, and given the position near 2.2 μm this band position would be expected for a high-Ca pyroxene [e.g., Cloutis and Gaffey, 1991].
 The positions and shapes of the crystal field absorptions are a function of the solid-solution chemistry of olivine where the band centers shift to longer wavelengths with increasing iron content [Burns, 1993; King and Ridley, 1987]. Sunshine and Pieters  applied the modified Gaussian model (MGM) [Sunshine and Pieters, 1993] to solid solution series of olivine to quantify the shift in absorptions as a function of iron content. Isaacson et al.  validated the application of the MGM to spectra of carefully prepared chromite-bearing lunar olivine mineral separates and showed that the magnesium number (Mg # = Mg/(Fe + Mg)) could be derived with a modified MGM-based approach. Isaacson et al.  have applied this approach to M3 data of regions with high olivine concentrations, including spectra for the Aristarchus region. The possible presence of minor pyroxene contamination in the Aristarchus olivine spectra makes the derivation of a specific Mg # relatively uncertain, but it appears that the Aristarchus olivine is more iron-rich than the regions in Moscoviense and Copernicus analyzed with the same approach [Isaacson et al., 2011].
 High resolution imaging of the olivine-rich regions shows that these deposits are not uniquely associated with a particular morphology or type of deposit (Figure 13). The olivine signatures are most strongly concentrated immediately along the southeastern rim of Aristarchus. In this area are small melt ponds as observed elsewhere along the rim (e.g., Figures 7 and 8). Just exterior to the crater we observe some differences in tonality in this area but these are not distinctly associated with changes in composition represented by the parameter mapping. In one location however is a collection of boulders that are arranged radial to the crater and are slightly lighter in tone. This collection of boulders appears to be enriched in low-Ca pyroxene (Figure 13).
 There is a scarp wall exposed interior to the crater rim shown in Figure 13 where it is possible to resolve slight differences in tone and texture. These units appear to be the size and shape of breccia blocks perhaps emplaced in the proximal ejecta deposit of Aristarchus or were present in the target rocks prior to impact. The scarp wall is not uniquely olivine rich but grades from olivine rich near the crater rim to more noritic lower in elevation along the crater wall. In the near-rim ejecta it is possible to identify low-Ca pyroxene rich bounders (shown by the yellow arrows in Figure 13). Elsewhere the ejecta deposits consists of ponded melt or changes in tonality related to a thin veneer of melt. We do not see olivine associated with specific ejecta blocks or breccia units. The observations suggest olivine is a key component in the impact melt and ejecta rather than as discrete blocks or lithologies excavated from deeper preimpact lithologies.
4.5. Glass: Impact Melt and Pyroclastic
 A distinctive spectroscopic signature is observed in the region both in the plains surrounding Aristarchus, that have been previously shown to covered to a depth of up to 30 m by pyroclastic deposits [e.g., Lucey et al., 1986; McEwen et al., 1994], and in close proximity to the crater rim. The spectral characteristics of these deposits are shown in Figure 14. The reflectance spectra are relatively featureless except for a prominent reflectance maximum at 0.75 μm. They also show a broad, poorly defined 1 μm band and a very weak 2 μm band. The spectral properties are better defined in the relative reflectance spectra also shown in Figure 14. Here the prominent reflectance maximum at 0.75 μm is clearly defined as well as distinct 1 and 2 μm bands separated by a local reflectance maximum at 1.6 μm. The width, center, and shape of the 1 and 2 μm bands are typical of that expected for synthetic glasses as well as some lunar pyroclastic glasses [Bell et al., 1976]. A recent study of the spectra of lunar and prepared glasses [Tompkins and Pieters, 2010] showed that many of the melt rocks had recrystalized resulting in spectra that were similar to igneous crystalline lithologies. Of the glasses studied by Tompkins and Pieters , the prepared glasses from Apollo 17 soils are the most similar to those observed in the Aristarchus region, either pyroclastic or of impact melt.
 It is difficult to uniquely separate the distribution of the glass deposits on the basis of the M3 spectral parameters. The spectral properties of the melt deposits are similar to deposits rich in olivine or other mafic silicates using the simple spectral parameters. To better isolate some of the melt deposits, we applied the Spectral Angle Mapper (SAM) [Kruse et al., 1993; Mustard and Sunshine, 1999] to M3 observation M3G20090209T054030. These spectral data were converted to reflectance relative to mature highland soil. The mature highland soil spectra used to ratio the reflectance data were acquired from the same columns and averaged over many lines. Two spectral types are mapped by the SAM algorithm: A spectrum enriched in olivine and one typical of the glassy spectra shown in Figure 14. The end-members are shown in Figure 15. We use the more restricted wavelength range 750–2600 nm to focus on the reflectance maximum at 0.75 μm and the diagnostic 1 and 2 μm bands.
 The results of applying the SAM approach to these type spectra types are shown in Figure 15. The method clearly distinguishes the olivine-rich and glass-rich deposits from the other surface compositions as well as from each other. A close-up of the spatial relationships between the glass and olivine-rich deposits shown in Figure 15 shows the compositional units are very well defined but that there are very sharp changes in composition over short spatial scales. The spectral character of these transitions has been verified by examining single pixel spectra moving across the transitions and the mapping is quite accurate. Examination of these deposits with LROC data clearly shows evidence of melt accumulations and flow (Figures 7, 8, and 9), cooling fractures, and ponds. Thus the melt generated by Aristarchus has highly heterogeneous compositions over short spatial scales.
 The high spatial and spectral resolution data from the M3 instrument provide important new insights into the composition of key units surrounding Aristarchus. There has been a diversity of compositional interpretations for the central peaks of Aristarchus. The M3 instrument clearly resolves the central peaks, unmixed with other units such as the floor deposits. The spectral properties are defined by a high albedo, much higher than any other units or deposits in Aristarchus, and a notable lack of any absorptions. Compared to mature highland soils the central peaks have a shallower continuum slope. The lack of diagnostic absorptions including any related to mafic minerals or feldspars is in contrast to the interpretations of McEwen et al. , Chevrel et al. , Lucey et al. , and Ohtake et al. . The telescopic spectra of Lucey et al.  may have included wall or floor materials that had mafic mineral absorptions. Chevrel et al.  used telescopic spectra to aid in the calibration of Clementine NIR data and mafic signatures from these spectra may have affected the reprocessed data. The M3 calibration is independent of any previous observations, relying on preflight and in-flight calibration measurements. These are the most completely calibrated data. The lack of mafic and plagioclase absorptions indicates that the central peak materials are dominated by low-Fe feldspathic materials. The central peaks for this 42 km diameter crater were uplifted from a depth of 3–5 km and the rocks exposed were sourced from the upper crustal, anorthositic layer of the Moon.
 Impact melt products are widely distributed in different settings including in a thick sheet on the floor, melt ponds and deposits associated with the crater walls, and extensive regions exterior to the crater as thin deposits and melt ponds. The spectral properties of the melt are characterized by a lower spectral slope than mature highland soils and a lack of any mafic absorption bands. Some deposits that have clear morphologic indications of melt (e.g., Figures 8, 9, and 13) also show the spectral properties diagnostic of glass (Figures 14 and 15) that are different than the spectrally unremarkable character of the majority of the melt deposits (Figure 5). The highest resolution imaging show that some of the melt deposits are loaded with breccia blocks as expected for an impact melt breccia (Figure 9b) and others show evidence for more a more fluid character (Figure 9c) suggesting segregation of highly fluid impact products from the more viscous constituents. The parameter mapping in Figure 3 shows that melt products are widely distributed outside the crater and that there is a high degree of apparent variability from olivine- and glass-rich to melt products with no mafic mineral signatures variation. However, when examined at high resolution many of these deposits lack distinct morphologic character suggesting the melt is present as a thin veneer (e.g., Figure 13).
 The different toned mounds observed on the floor (Figure 7) and mounds showing different compositions observed at the base of the crater walls (Figure 10) are evidence of large blocks of lunar crust that have been mobilized and redistributed by the impact processes. It is interesting to note that the blocks are relatively uniform in composition that we interpret to show that the impact ejecta is not completely homogenized by the process. The walls of the crater are relatively well covered by debris and ejecta such that clear indications of crustal composition and stratigraphy in the walls is not exposed. However, in the upper scarp face we do see evidence for that the upper crust is composed of a complex breccia (Figure 13). This provides a window into the composition and structure of the Aristarchus plateau.
 The olivine deposit first identified by telescopic spectra [Lucey et al., 1986], clearly defined by Clementine data [Le Mouélic et al., 1999; Chevrel et al., 2009], and recently observed by the Spectral Profiler on SELENE [Yamamoto et al., 2010] is well resolved in the M3 data, and the results strongly support the distribution and composition of previous work. An imp77ortant aspect of the composition is the spectral dominance by olivine with no diagnostic evidence for pyroxene or spinel. This was also noted by Le Mouélic et al. . The olivine-rich regions show clear evidence for impact melt in high-resolution imaging (Figure 13) and there is a strong association with impact glass (Figure 15). Yet there is little evidence for olivine-rich lithologies in the crater walls, floor, or blocks of crustal material in the ejecta on the floor, walls, and exterior to the crater.
 These observations provide important constraints on the fundamental question of the origin of the olivine-rich deposits. Le Mouélic et al.  concluded the olivine was sourced from shallow depths to explain its location on the southeast rim and thus was derived from a shallow pluton. Chevrel et al.  concluded the olivine-rich materials were excavated from an olivine-rich layer that sits above an anorthositic crust but beneath the noritic Imbrium ejecta. Global olivine detections with data from the SELENE spacecraft [Yamamoto et al., 2010] appear to be clustered around impact bains from which they conclude at least some of the olivine originates from the mantle, though this paper is not specific as to the origin of the olivine in Aristarchus.
 We conclude the olivine was excavated from either (1) a shallow pluton or (2) an olivine-rich region of the Imbrium ejecta or (3) derived through melting and excavation of olivine-rich Procellarum basalts. While a mantle source is possible, the olivine deposits are on the crater rim and thus excavated from a relatively shallow source. The Aristarchus crater sits on the rim of South AP crater (Figure 1) that may have uplifted deep seated crustal material including shallow plutons. While we cannot exclude a shallow pluton source or unique compositional layer at depth, there is little evidence to support this conclusion. The Aristarchus impact was formed at the boundary between mare and a block of highland crust. The mare in this region are olivine-bearing [Staid and Pieters, 2001; Staid et al., 2011] and the location of the olivine deposit on the southeast rim is where the basalt would be expected to be deposited given the reconstructed preimpact geology [Zisk et al., 1977; McEwen et al., 1994; Chevrel et al., 2009]. The energy of impact would be more than sufficient to efficiently melt mare basalt and emplacement of the melt would rapidly crystallize olivine. The intimate association of olivine with impact glass (Figure 15) is supporting evidence for emplacement of a melt that is olivine rich. It is also possible that some of the olivine could be olivine crystals that were present in the basalts were not melted by the impact process and survived in the melt.
 Analysis of data from the M3 instrument has confirmed the broad results of previous analyses that the Aristarchus plateau and crater have an extraordinary diversity of geologic landforms and compositions. However, the spectral and spatial resolution of M3 has opened a new depth of understanding that was not previously possible. The central peaks of the crater are shown to be devoid of any mafic silicate absorptions. While this is not definitive evidence of an enrichment in plagioclase, the extreme brightness and lack of mafic silicates indicate the peaks are likely sourced from the upper anorthositic crust of the Moon.
 There are abundant impact melt deposits in an around the crater that are well resolved by the new lunar imaging data sets and M3. Most easily recognized by their morphology and lack of mafic absorptions, the melt spectral properties range from essentially identical to highland soils to those where olivine is the only spectral signature that is identified to those where the compositional signatures are dominated by glass. The olivine-rich areas are particularly interesting. No evidence of a deep olivine source is observed in the crater lithologic units (e.g., outcrops in the crater floor or walls). The most abundant proximal source of olivine is the Procellarum basalts that were present prior to the impact.
 This first analysis of this compelling region has only just touched on the richness of compositional information available with the M3 data. For example these data have clearly resolved the spectral signatures of pyroclastic glass but the details of this compositional variability and distribution across the plateau are unknown and will be the subject of future work. The extraordinary diversity of compositions exposed in the walls and impact ejecta of Aristarchus, and their implications for the extensively brecciated upper crust of the Moon will also be the subject of future work.
 The M3 instrument was funded as a mission of opportunity through the NASA Discovery program. M3 science validation is supported through NASA contract NNM05AB26C. The M3 team is grateful to ISRO for the opportunity to fly as a guest instrument on Chandrayaan-1. We acknowledge the insight and careful comments of a reviewer.