PUBLICATIONS. Journal of Geophysical Research: Planets. Volcanic glass signatures in spectroscopic survey of newly proposed lunar pyroclastic deposits

Transcription

1 PUBLICATIONS Journal of Geophysical Research: Planets RESEARCH ARTICLE Key Points: Some pyroclastic deposits show mineralogy consistent with pyroclastic deposits Volcanic glasses are identified in three of the proposed deposits Birt E, Schluter, and Walther A appear to be comprised of glassy pyroclasts Correspondence to: S. Besse, sbesse@rssd.esa.int Citation: Besse, S., J. M. Sunshine, and L. R. Gaddis (2014), Volcanic glass signatures in spectroscopic survey of newly proposed lunar pyroclastic deposits, J. Geophys. Res. Planets, 119, doi: / 2013JE Received 25 SEP 2013 Accepted 14 JAN 2014 Accepted article online 22 JAN 2014 Volcanic glass signatures in spectroscopic survey of newly proposed lunar pyroclastic deposits S. Besse 1, J. M. Sunshine 2, and L. R. Gaddis 3 1 ESA/ESTEC, Noordwijk, Netherlands, 2 Department of Astronomy, University of Maryland, College Park, Maryland, USA, 3 Astrogeology Science Center, U.S. Geological Survey, Flagstaff, Arizona, USA Abstract Moon Mineralogy Mapper spectroscopic observations are used to assess the mineralogy of five sites that have recently been proposed to include lunar dark mantle deposits (DMDs). Volcanic glasses have, for the first time, clearly been identified at the location of three of the proposed pyroclastic deposits. This is the first time that volcanic glasses have been identified at such a small scale on the lunar surface from remote sensing observations. Deposits at Birt E, Schluter, and Walther A appear to be glassy DMDs. Deposits at Birt E and Schluter show (1) morphological evidence suggesting a likely vent and (2) mineralogical evidence indicative of the presence of volcanic glasses. The Walther A deposits, although they show no morphological evidence of vents, have the spectroscopic characteristics diagnostic of volcanic glasses. The deposits of the Freundlich-Sharonov basin are separated in two areas: (1) the Buys-Ballot deposits lack mineralogical and morphological evidence and thus are found to be associated with mare volcanism not with DMDs and (2) the Anderson crater deposits, which do not exhibit glassy DMD signatures, but they appear to be associated with possible vent structures and so may be classifiable as DMDs. Finally, dark deposits near the crater Kopff are found to be associated with likely mare volcanism and not associated with DMDs. The spectral identification of volcanic glass seen in many of the potential DMDs is a strong indicator of their pyroclastic origin. 1. Introduction Lunar pyroclastic deposits, also called dark mantle deposits (DMDs) [Head, 1974], have been observed over several locations of the lunar surface (Figure 1) and have been classified as either regional or localized based largely on their size [Gaddis et al., 2003]. The DMDs are thought to have been emplaced by explosive volcanic processes [e.g., McGetchin and Head, 1973; Wilson and Head, 1981]. The regional DMDs (>1000 km 2 ) are often observed as dark (low albedo) deposits in highlands near large maria, and their wide spatial distribution, mantling relationship to underlying terrain, and association with rilles and ventlike irregular depressions suggest an origin from an explosive volcanic eruptions [e.g., Gaddis et al., 1985; Head, 1974; Heiken et al., 1974]. At low magma ascent rates, Strombolian-style eruptions form, as exsolved gas coalesces into large bubbles inside the source vent. When the bubbles reach the surface, the resulting explosion can distribute particles far from the vent. At higher ascent rates, Hawaiian-style fire fountaining occurs [Wilson and Head, 1981] with lower content of gases. These eruptions usually start with crack formation fed by low-viscosity basaltic magmas. The compositions of regional deposits are dominated by varying ratios of iron-bearing glass to ilmenite and olivine [Gaddis et al., 1985; Pieters et al., 1974]. These compositions are interpreted to represent less and more crystalline varieties of the same melt, based on similarity to the ilmenite-bearing black beads at the Apollo 17 landing site [Adams, 1974]. The degree of crystallinity of lunar pyroclasts is thought to be determined by the density of the erupting plume [Head and Wilson, 1989], because melt droplets in denser plumes will experience slower cooling rates and will crystallize, while melt droplets in thin plumes and near the margins of plumes will experience rapid cooling and will become glassy [Weitz et al., 1998]. On the Moon, plume density will largely be determined by volatile content of the magma [Wilson and Head, 1981], so the crystallinity of pyroclastic deposits can be related directly to the relative volatile content of source magmas [Nicholis and Rutherford, 2009; Weitz et al., 1998, 1999]. By contrast, the smaller, localized DMDs are low-albedo deposits observed in association with noncircular volcanic vents typically located along fractures in crater floors [e.g., Head and Wilson, 1979]. These deposits, often called dark-halo craters, are thought to have formed as a result of intermittently explosive or Vulcanianstyle eruption in which gas in magma rising toward the surface in a dike builds up under a caprock and periodically explodes, entraining magmatic (juvenile) and/or local, country rock materials which may or may BESSE ET AL American Geophysical Union. All Rights Reserved. 1

2 Figure 1. Global distribution of lunar pyroclastic deposits (white points) after Gaddis et al. [2003] and updated by Gustafson et al. [2012]. DMDs analyzed in this study are highlighted by the red star symbols. not be volcanic [e.g., Hawke et al., 1989]. Analyses of telescopic spectral data have indicated that these deposits were compositionally diverse and that several of the pyroclastic deposits may have olivine components in addition to pyroxene [Coombs et al., 1990; Hawke et al., 1989], supporting an iron-rich basaltic composition for these deposits. These authors classified localized pyroclastic deposits into three groups based on the continuum slope and the position, depth, and shape of the 1 μm band. The groups were thought to represent mixtures of highlands wall rock, basaltic plug rock, and juvenile basaltic magma at each site. More recent results based on Clementine ultraviolet-visible data were less conclusive, supporting the existence of compositional diversity among the localized pyroclastic deposits but not clearly identifying an olivine component [Gaddis et al., 2000, 2003]. As volatile and metallic element (i.e., Fe and Ti) enriched remnants of ancient volcanic eruptions on the Moon, pyroclastic deposits provide clues to the nature of the early lunar interior [e.g., Delano, 1986; Heiken et al., 1974; Papike et al., 1998; Shearer and Papike, 1993; Shearer et al., 2006] and to the distribution of potential resource materials for future exploitation [Duke et al., 2006; Hawke et al., 1990]. Pyroclastic glasses may be the best examples of primitive materials on the Moon, and thus, they are important in characterizing the lunar interior and as a starting place for understanding the origin and evolution of basaltic magmatism on the Moon. Recent observations of the lunar surface with improved spectral and spatial resolutions now allow the focus in more detail on the morphological and mineralogical characterization of the lunar DMDs and to address outstanding questions about their nature and origin. Such spectral data provide information on the shape, position, and strength of the 1 and 2 μm bands. Analyses of these data will provide better discrimination of the presence of the major lunar mafic minerals such as high- and low-calcium pyroxenes, olivine, and iron-rich glass in soils of lunar volcanic deposits. The Moon Mineralogy Mapper (M 3 ) instrument [Pieters et al., 2009], with a spectral range from 0.4 to 3 μm and a spatial resolution of ~140 m/pixel or ~280 m/pixel, is very powerful for characterizing mafic absorption bands in volcanic units at both 1 and 2 μm. For example, explosive and effusive volcanic deposits have been clearly identified on the Marius Hills volcanic complex using 1 and 2 μm band parameter maps [Besse et al., 2011]. The mineralogical variations of the Oceanus Procellarum basaltic flows have also been categorized using 1 and 2 μm absorption bands [Staid et al., 2011], which dramatically highlight the differences in olivine content. In this analysis, we use data from the M 3 instrument to derive similar spectral parameter maps that highlight mineralogical characteristics of lunar volcanic materials to examine five recently proposed sites at which possible DMDs have been identified [Gustafson et al., 2012]. The relatively small sizes of the five proposed DMDs suggest that they correspond to localized DMDs. Gustafson et al. [2012] used albedo and morphologic information from Lunar Reconnaissance Orbiter Camera (LROC) Narrow Angle Camera (NAC) images [Robinson et al., 2010] and color images derived from Clementine multispectral data to identify several previously unrecognized DMDs. The improved spectral resolution and wavelength range of M 3 are used in this study to infer the spectroscopic characteristics of these recently proposed pyroclastic deposits. The objectives of this BESSE ET AL American Geophysical Union. All Rights Reserved. 2

3 analysis are to characterize the spectroscopic properties of five of the six recently proposed DMDs, to determine their mineralogy, and to assess their origin as pyroclastic deposits on the basis of their composition and morphologic setting. Figure 2. Reflectance spectra of lunar samples measured in the laboratory. Major lunar minerals are displayed along with volcanic glasses. Reflectance Experiment LABoratory IDs are LR-CMP-014 (olivine with chromite), LS-CMP-004 (plagioclase), LS-CMP-009 (clinopyroxene), LS-CMP-012 (orthopyroxene), LR-CMP-052 (green glass), LR-CMP051 (orange glass), and LR-CMP-050 (black beads). 2. Background 2.1. Lunar Minerals and Spectroscopic Properties The spectral properties of major lunar minerals are presented in Figure 2. These minerals exhibit absorption bands that differ by their shape and position along the spectral domain. Pyroxenes have two absorption bands, one centered near 1 μm and another near 2 μm; these band centers move to longer wavelengths as Ca and Fe substitute for Mg [e.g., Adams, 1974; Cloutis and Gaffey, 1991; Hazen et al., 1978]. Olivine has a complex absorption band centered beyond 1.05 μm that moves as Fe substitutes for Mg [Burns, 1970, 1974]. Significant amounts of olivine in lunar volcanic deposits will broaden the pyroxene absorption at 1 μm and shift it to longer wavelengths [Singer, 1981], while the 2 μm band remains fixed. Because olivine lacks a band at 2 μm, the 1 μm absorption in olivine-rich lunar deposits will be strengthened relative to the 2 μm band. However, it has been shown that chromite-rich olivine, if present, exhibits absorption features near 2 μm that could diminish the strength of the 1 μm relative to the 2 μm band [Isaacson and Pieters, 2010]. The presence of Fe-rich volcanic glasses in lunar soils causes broad and shallow absorption bands because of the amorphous structure of the glasses [e.g., Bell et al., 1976]. The 1 μm band center of lunar glass is generally shifted to longer wavelengths when compared to pyroxene, and the 2 μm band center to shorter wavelengths. Thus, the 1 and 2 μm band center positions of lunar glasses will typically appear close together than those of pyroxenes. The lunar orange and green glasses shown in Figure 2 can be distinguished by these two band properties: the orange glass has a 1 μm band minimum shifted to longer wavelength relative to the green glass, which are both shifted to longer wavelength relative to pyroxenes, and the strength of the 1 μm band is notably stronger relative to the 2 μm band strength for the green glass. However, as noted further below (section 4.2), it is difficult to distinguish whether glass has a volcanic or impact origin. Tompkins and Pieters [2010] have shown that glasses with both origins have similar spectroscopic properties, and the crystallized material from that impact melt is nearly indistinguishable from igneous rock without geological context [see also Smrekar and Pieters, 1985]. Thus, other characteristics and criteria (e.g., overall morphology, presence of volcanic vents, and darker albedo) will help to distinguish between volcanic and impact glasses. As outlined here, the major mineralogy and internal structures (i.e., crystallized versus amorphous glass) associated with both regional and localized DMDs can be identified and characterized with appropriate remote sensing data covering both the 1 and 2 μm absorption features Previous Work Spectral analyses of pyroclastic deposits have shown that they are compositionally diverse [Gaddis et al., 2000, 1985; Hawke et al., 1989; Weitz et al., 1998]. These spectral classifications were largely based on the position and asymmetry of the 1 μm band, which was interpreted as variations in the presence and relative amounts of volcanic glass, olivine, and/or pyroxene. For these particular soil constituents, the use of the 1 and 2 μm band is very important since they have very different spectral behavior. For example, analysis of only BESSE ET AL American Geophysical Union. All Rights Reserved. 3

4 Table 1. Summary of the M 3 Observations Used in This Study for Each Individual Target a Target Freundlich- Sharonov Kopff Schluter Birt E Walther A Images Used M3G T203200_V01_L2 M3G T003411_V01_L2 M3G T043741_V01_L2 M3G T175144_V01_L2 M3G T220912_V01_L2 M3G T050712_V01_L2 M3G T093119_V01_L2 M3G T055240_V01_L2 M3G T095022_V01_L2 M3G T083142_V01_L2 M3G T125102_V01_L2 a The center of the M 3 image/mosaic is given. Average Phase Angle (degree) Optical Period Average Resolution (m/pixel) Centered Latitude/Longitude 37.8 OP2C / OP2C / OP2C / OP2C / OP2C /1.4 the 1 μm band could easily lead to confusion between olivine and volcanic glass because they both have an asymmetric band shifted to longer wavelengths with respect to the pyroxenes [Bell et al., 1976; Burns, 1993]. The limited spectral coverage of previous studies (e.g., 0.6 to 2.6 μm) [Gaddis et al., 2003; Hawke et al., 1989] has restricted the identification and characterization of certain mafic minerals and volcanic glass in particular. Although it is known that volcanic glasses are associated with pyroclastic deposits since the Apollo era [Pieters et al., 1974], definitive spectroscopic evidence has not yet been shown at longer wavelengths where more accurate characterization is possible. The limited spatial resolution of remote sensing instruments could also have prohibited a clear identification and characterization of lunar volcanic glass, especially for many of the localized deposits. However, with improved spatial and spectral resolution, the identification of lunar minerals and in particular the discrimination between olivine, volcanic glass, and pyroxene is now possible. For example, the M 3 data set is very powerful in identifying the mineralogy of lunar DMDs. Recent studies of DMDs near Lavoisier crater using M 3 data [Souchon et al., 2013] have shown their complexity, with volcanic units of similar mineralogical composition possibly having different degrees of compaction and/or porosity. However, no olivine detection was reported. Investigation of the 2 μm absorption bands of other DMDs indicates that they are mineralogically distinct and readily separable from the mare basalts that often surround them [Allen et al., 2013; Jawin et al., 2013]. Thus, the M 3 data set has the capability of resolving key issues of mineralogical identification of the lunar DMDs with spectral and spatial resolutions never obtained previously. 3. Data and Observations This investigation of lunar DMDs is supported by the use of data from two remote sensing instruments. (1) The M 3 instrument was a visible and near-infrared spectrometer covering the spectral range from 0.4 to 3 μmwitha spectral resolution up to 20 nm, and a spatial resolution of ~140 m/pixel or ~280 m/pixel. In this analysis, the Level 2 v1.0 data archived in Planetary Data System (PDS) [Malaretetal., 2011] are used. Level 2 calibrated data include radiometrically corrected data [Boardman et al., 2011; Green et al., 2011] with removal of thermal effects [Clark et al., 2011] and improved photometric correction at long wavelengths [Besse et al., 2013]. More recently available ground truth [Isaacson et al., 2013] and cross-track [Besse et al., 2013] corrections are not applied in this analysis because they do not improve the quality of the data available for the DMDs. It is noted that despite all the efforts made to correct the variation of brightness as a function of the observation conditions (i.e., photometric correction), residuals persist due to unplanned variations in detector temperature [Green et al., 2011]. As will be shown later (Figures 3c and 5a 5d), this results in vertical banding in the M 3 images and derived products, and consequently longitudinal variations of absolute reflectance of presumed identical terrains. (2) LROC Calibrated Data Records as delivered to the PDS are used in this analysis also. The NAC data provide very high spatial resolution views of ~0.5 m/pixel [Robinson et al., 2010] of the lunar surface, and the monochrome observations from the Wide Angle Camera (WAC) at a resolution of ~100 m/pixel [Henriksen et al., 2013]. Together, the LROC data provide geological context that facilitates the interpretation of spectral information from the M 3. The DMD areas of interest are distributed broadly across the lunar surface (Figure 1). They include the floorfractured craters Kopff and Anderson E and F, the mare-flooded craters Buys-Ballot and Schluter, the crater BESSE ET AL American Geophysical Union. All Rights Reserved. 4

5 Table 2. Summary of Spectral Parameters Used in the Analysis of the Lunar DMDs a Name Wavelengths Integrated (nm) Continuum Boundaries (nm) IBD IBD BD BD BD BD BD R N/A a IBD corresponds to the integrated band depth, BD to the band depth, and R to the reflectance. The reflectance at 1580 is used as a neutral reflectance where absorptions are limited. Walther A, and the mare crater Birt E (also summarized in detail in Gustafson et al. [2012, Table 1]). The DMD at Mount Carpatus described in Gustafson et al. [2012] is not investigated with the M 3 due to unfavorable illumination conditions and limited spatial coverage. Other possible DMDs proposed by Gustafson et al. [2012, Table 2] are not investigated here because of their low likelihood of being DMDs (as acknowledged by Gustafson et al. [2012]) and their smaller sizes, which are not readily observable with the spatial resolution of the M 3. The M 3 data used for this analysis (Table 1) are divided into Optical Periods (OPs) that differ in terms of detector temperature and altitude of the spacecraft [Boardman et al., 2011]. To maximize the comparison of the DMDs compositions, observations from the same OP2C (Table 1) have been used for all analyzed targets. To emphasize the mineralogical properties of the DMDs in a regional context, both individual spectral bands (e.g., reflectance at 0.75 and 2.94 μm to highlight albedo) and color parameter maps are used based on the spectral characteristics of mafic minerals. The primary parameter map used here is the Integrated Band Depth (IBD), which is used to highlight and identify major lunar minerals, including the chromium spinel-rich composition of the Sinus Aestuum pyroclastic deposits [Sunshine et al., 2010] and the olivine and pyroxene content of lunar soils [e.g., Besse et al., 2011]. The mathematical expression of the IBD at 1 μm is given in equation (1): IBD1000 ¼ 26 R C1 ð789 þ 20nÞ n ¼ 0 (1) Rð789 þ 20nÞ where R C1 is the 1 μm continuum approximated by a straight line between 0.75 and 1.58 μm andr is the reflectance between 0.79 and 1.31 μm. Band depths (BD) are calculated for only one specific wavelength (i.e., not integrated over the whole absorption band). Reflectance (R) at1.58μm isusedasa neutral reflectance where absorptions are limited. A summary of the parameters IBD, BD, and their respective continuum boundaries are presented in Table 2. In addition, continuum-removed spectra are used to emphasize the shape and position of absorption bands at 1 and 2 μm and to improve comparisons between units (Figures 3f, 4f, 5f, 6f, 7f, and 8b). The continuum is approximated by a straight line and divided from the original spectrum. The straight line is defined between 0.73 and 1.62 μmforthe1μm band and between 1.62 and 2.58 μm forthe2μm band. 4. Results From Sites of Proposed DMDs In this section, the spectral characteristics of each of the five investigated sites are discussed on the basis of derived spectral parameter maps and continuum-removed spectra extracted from the M 3 data (Table 2) Anderson E and F, and Buys-Ballot Anderson E and F are two impact craters (28 km and 49 km diameter, respectively) located near the center of the farside Freundlich-Sharonov basin. The floors of both craters are bisected by fractures, both have dark deposits that appear to be volcanic in origin, and these dark units are proposed to be DMDs [Gustafson et al., 2012]. Buys-Ballot is an elongated pear-shaped crater (diameter of 55 km) with a mare-like flat floor that may have been formed as a result of an oblique impact by a mafic-rich impactor [Schultz et al., 1998]. Buys-Ballot crater is located 100 km north of the Anderson craters. Mare-like deposits extend further south into the Lacus Luxuriae maria, and Gustafson et al. [2012] suggest that DMDs mantling local highlands are associated with these mare deposits. The Lacus Luxuriae DMDs of Gustafson et al. [2012] are outlined in Figures 3a, 3b, and 3d by a white line. Together with the deposits at Anderson E and F, the association of low-albedo mafic deposits suggests also a volcanic origin for all these deposits as indicated by Gustafson et al. [2012]. Figure 3 shows M 3 albedo and parameter maps of these three deposits altogether. The IBD1000, shown in Figure 3b, is a measure of the strength of the 1 μm absorption and is a very good proxy for the mafic content BESSE ET AL American Geophysical Union. All Rights Reserved. 5

6 Figure 3. The Freundlich-Sharonov basin region as viewed by M 3.(a)Reflectance at 0.75 μm, (b) 1 μm integrated band depth (IBD1000), (c) color composite showing the mineralogical variability of the region with an emphasis on the 2μm absorption band (R=BD1900, G =BD2300, B= R1580; see Table 1), (d) reflectance at 2.94 μm, (e) spectra of locations given in Figures 3a and 3c, and (f) same spectra after continuum removal. The vertical banding (Figure 3c) is due to uncorrected photometric properties of the data and a residual cross-track gradient [Besse et al., 2013]. The proposed DMDs by Gustafson et al. [2012] are outlined in white in Figures 3a, 3b, and 3d. The two white circles in Figure 3b correspond to craters Anderson E and F. Figure 3c uses a different color scheme than the other figures in order to highlight the 2 μm absorption band variations (i.e., olivine) that are characteristic of the region. In Figure 3c, mineralogical variations are mapped into two areas outlined in red (Area 1) with orange hues corresponding to olivine-richer deposit then Area 2 outlined in green. This is confirmed by Figure 3f, with spectra showing band location consistent with pyroxene/ olivine mixture rather than glass. of the surface. Mafic signatures (shown as bright areas on Figure 3b) are found in Buys-Ballot, Lacus Luxuriae, Anderson E and F, and also in Buys-Ballot Q on the western part of the image. The proposed DMD unit in Lacus Luxuriae clearly overlaps mare units with at least two different intensities of mafic absorptionat1μm. The northern part (Area 1 on Figure 3c) is darker and has a weaker 1 μm absorption than the southern part. In Figure 3c, in which the strength and position of the 2 μm absorption bands are highlighted with a color combination that separates the weak absorptions (i.e., orange color and Area 1) and the stronger absorptions (i.e., green color and Area 2), a similar observation can be made. It appears that the proposed DMDs in Lacus BESSE ET AL American Geophysical Union. All Rights Reserved. 6

7 Luxuriae are superimposed on the northern portions of both Areas 1 and 2. Given the depth of the largest crater in Lacus Luxuriae (LL-C in Figure 3a, left side of Area 1), it is most likely that the crater has punched through the older volcanic layer, which exhibits a different mineralogy and especially a weaker 2 μmband (Area 1). The ejecta from the impact cover a portion of Area 2 that extends further east. They define Area 1 and have the same composition as a portion of the proposed DMDs covering the southern part of Buys- Ballot (orange color, Area 1) and could therefore explain the mantling description given by Gustafson et al. [2012] if they correspond to ejecta. The deposits extending further southwest have stronger absorption bands that are consistent with small spots within Anderson E and F (green color, Area 2). It is noted that the deposits in Buys-Ballot Q also exhibit two different compositional units, which are consistent with Area 1 and also are associated with a crater that could have excavated a deeper layer with different mineralogy. Individual spectra, specifically LL-Ce and LL-Ne, (Figures 3e and 3f), confirm that the Area 1 unit has weak absorptions, characteristic of mature soils. The spectrum LL-C has stronger absorptions because it is associated with a fresh crater. The 2 μm absorption band of LL-C is shallow, but the 1 μm absorption band is quite strong, broad, and characteristic of an olivine-rich composition. The LL-C spectrum can be compared with other fresh craters in the Area 2: LL-S, LL-Nw, and BB-C, which all have strong 1 and 2 μm absorptions indicative of the presence of a mafic mineral such as high-calcium pyroxene, supporting a mare volcanic origin for many of these deposits. However, the 1 μm absorption in the spectrum of LL-C is both broad and occurs at longer wavelengths than the other fresher craters indicating a strong enrichment of olivine in Area 1. Thus, olivine content likely is the main difference between the two clearly defined mafic units within the proposed DMD of Lacus Luxuriae. From Figures 3e and 3f, it can be seen that the mafic signatures within Anderson E and F are more consistent with an olivine-poor mafic composition as seen also in the spectra of Ae-S and Ae-N. None of the individual spectra exhibit band position and asymmetry that could be associated with volcanic glass. Based on the mineralogy, if the deposits at Lacus Luxuriae are pyroclastic in origin, they must be divided into two distinct deposits of roughly equal size. This would imply that two distinct volcanic episodes have emplaced pyroclastic deposits adjacent to each other. Another explanation could be that part of the DMDs is covered by the ejecta of the fresh crater in Lacus Luxuriae; however, this does not explain why the ejecta have the same albedo. From the albedo maps at 0.75 μm (Figure 3a) and at 2.94 μm (Figure 3d), the proposed DMDs do not appear darker in the M 3 images as is typical of other likely DMDs (see sections 4.3 and 4.5). Together with the spectral characteristics, this makes the Buys-Ballot dark deposits less likely than Anderson E and F craters to be a DMD because its emplacement would need to have been very complicated. We propose a basaltic origin of these deposits that is comparable to that of other basaltic deposits in Freundlich-Sharonov basin as proposed by Gustafson et al. [2012]. Anderson E and F craters do not show specific mineralogical signatures (e.g., volcanic glass) that would clearly indicate pyroclastic origin. However, morphological evidence from Gustafson et al. [2012], including the presence of possible low-relief vents in association with floor fractures, still argues in favor of a pyroclastic origin for the dark deposits in Anderson craters E and F Kopff Kopff is a crater (41 km diameter) located in the eastern portion of the Orientale basin. The floor of the crater is dominated by numerous fractures, which are more concentrated in the southern part. The northern part of the crater is slightly elevated with isolated massifs. Gustafson et al. [2012] proposed that the entire floor of Kopff and a small extent to the south (i.e., spectra K-S in Figure 4) are covered with possible DMDs. From Figure 4b, a map of the strength of absorption bands at 2 μm, it appears that most of the crater floor shows a general mafic enhancement, but that fresh mafic signatures (with stronger and deeper absorption bands) are limited to small features that are either fractures or impact craters. A limited number of these mafic signatures are also located in the walls of the crater. Note that some portions of the floor show very weak mafic signature; they are outlined in white for the Area 1 in Figures 4a 4d and correspond to bright crater rays derived from large craters outside the scene and superposed on the floor. Units with low reflectance at 2.94 μm (arrows in Figure 4d), located close to the inner rim, are also associated with strong mafic absorptions. Figure 4c shows similar results, with strong absorptions (i.e., bright features) associated with small, fresh craters, fractures, and low reflectance at 2.94 μm. The black outline in each image of Figure 4 corresponds to the spatial extent of the ejecta of the larger fresh crater on the eastern portion of the floor. The proposed DMDs that extend south of the crater are slightly distinguishable in Figure 4c with a bluer color. BESSE ET AL American Geophysical Union. All Rights Reserved. 7

8 Figure 4. Kopff crater as viewed by M 3. (a) Reflectance at 0.75 μm, (b) 2 μm integrated band depth (IBD2000), (c) color composite exhibiting the mineralogical variability of the region with an emphasis on the 1 μm absorption band (R = BD950, G = BD1050, B = R1250; see Table 1), (d) reflectance at 2.94 μm, (e) spectra of locations given in Figure 4a, and (f) same spectra after continuum removal. The white outlines correspond to Area 1 with almost no mafic absorptions; the black outline corresponds to a fresh crater and its ejecta. Arrows in Figure 4d point to low-reflectance areas at 2.94 μm that show mafic signature on Figures 4b and 4c. The spectra are consistent with pyroxene/olivine mixture. From Figures 4e to 4f, spectra showing mafic signatures within Kopff have absorptions centered at 0.98 to 1.0 μm and 2.2 μm (e.g., see spectra K-N, K-W, and K-M). This trend could be consistent with a higher concentration of high-ca pyroxenes in the floor materials, with K-M being one example of this high-ca pyroxene end-member. Note that K-N corresponds to a small fresh crater located outside of Kopff that clearly exhibits a mafic signature (see Figures 4b and 4c) comparable to that of K-W and K-M. The southern deposits outside of the crater (K-S and K-Se, and located in the proposed DMDs) do not have discernible mafic absorption bands. They have, however, a redder slope than the other deposits, which could explain the color variation on Figure 4c (i.e., bluer hue). However, the association of the southern deposits with mafic or iron-bearing minerals as proposed by Gustafson et al. [2012] is not supported by the M 3 data. In the case of Kopff, the association of the mafic signatures with fractures could be consistent with DMDs. However, the fact that small craters and outcrops on the inner rim also exhibit the signature is puzzling. This suggests that there is an underlying layer of mare-like material. Since the craters that exhibit these mafic signatures are rather small, smaller than the fresh crater outlined in black, this implies that the mafic unit is very thin. This could be potentially linked to crystallization of impact melt residual from the impact of Kopff (and/or Orientale basin) or underlying mare basalts from Mare Orientale. The case of crystallization of residual of impact melt is particularly interesting; it could easily explain the various pyroxene/olivine mixtures as products of the crystallization and differentiation mechanisms. This hypothesis could easily explain variations of spectra like K-M and K-W, and the dark albedo mafic-rich features located at the edge of the rim. Tompkins and Pieters [2010] have shown that it is impossible to spectrally distinguish mafic signature of recrystallized impact melt from igneous rock; thus, the hypothesis of impact melt differentiation is possible. In the case of Kopff, no spectral signature of volcanic glass could be seen, and the 2 μm band center locates at wavelengths well beyond 2 μm and is thus inconsistent with the presence of volcanic glass. While a volcanic origin is likely for the dark floor deposits of Kopff crater, the lack of observed volcanic glass, coupled with BESSE ET AL American Geophysical Union. All Rights Reserved. 8

9 Figure 5. Schluter crater as viewed by M 3. (a) Reflectance at 0.75 μm, (b) IBD2000, (c) color composite (R = BD950, G = BD1050, B = R1250), (d) reflectance at 2.94 μm, (e) spectra of locations given in Figure 5a, and (f) same spectra after continuum removal. The sharp boundary (Figures 5a and 5d) is due to the incompleteness of the photometric correction. Arrow in Figure 5d points to low-albedo regions at 2.94 μm of one potential DMD. The two outlined regions show variations in the spectral properties of the two units with 1 and 2 μm bands shifted to longer wavelengths for Area 2. Continuum-removed spectrum of S-M in Figure 5f highlights the shifted 1 μm band to longer wavelength attributed to the contribution of volcanic glass. the distribution of dark material and the mafic spectral signature exposed by small craters, suggests that it is unlikely that these dark deposits have a pyroclastic origin Schluter Schluter is an impact crater (diameter of 89 km) located northeast of Orientale basin (~300 km north of Kopff crater), and it has a central peak and floor fractures. Dark units covering fractures and hummocks are visible in the northern and west central part of the crater floor. In association with these likely mare deposits, Gustafson et al. [2012] identified three areas of possible pyroclastic deposits within Schluter: two associated with the mare materials in the eastern and northeastern portions of the crater floor and a third possible deposit just west of the central peak. The mare-like dark units in the northeast floor of Schluter crater have two separable units (Areas 1 and 2, Figures 5a 5d) that differ in the strength of the absorption bands at 2 μm (Figure 5b, bright areas have a strong absorption band) and the shape of the 1 μm absorption band (Figure 5c, bluer areas have band centers shifted to longer wavelengths). Note that for the M 3 data of Schluter, only latitudinal variations are considered because of the longitudinal variation in brightness as seen in the 750 nm albedo map. This is due to incomplete photometric correction that could alter the spectra (e.g., absolute reflectance and spectral slope). The proposed DMDs of Schluter [see Gustafson et al., 2012, Figure 11] are characterized by the spectra S-E and S-N (from Area 1) and S-Se (from Area 2) in Figures 5e and 5f. The comparison of the two spectra S-N and S-Se shows that they are quite different in terms of position of the absorption bands (i.e., band center at 1.0 and 1.03 μm, respectively) with Area 2 having bands shifted to longer wavelengths. These variations could be caused by a higher concentration of high-ca pyroxene for S-Se, which would be consistent with a mare origin for this unit. If there are pyroclastic deposits associated with either Area 1 or Area 2, they are indistinguishable from the maria with no increase in the olivine and/or volcanic glass content. Nevertheless, both areas are consistent with pyroxene-dominated mineralogy with different composition of the pyroxenes. BESSE ET AL American Geophysical Union. All Rights Reserved. 9

10 Figure 6. Birt E as viewed by M 3.(a)Reflectance at 0.75 μm, (b) IBD1000, (c) color composite (R = BD950, G = BD1050, B = R1250; see Table 1), (d) reflectance at 2.94 μm, (e) spectra of locations given in Figure 6a, and (f) same spectra after continuum removal. The spectra DMDs Roi correspond to the entire area outlined in white, and it corresponds to the extent of the DMDs. The yellow and black lines correspond to boundaries on the mare basalt identified, respectively, in Figures 6b and 6d. In Figure 6f, the 1 and 2 μm band locations of the DMDs Roi are shifted to longer and shorter wavelengths, respectively, and attributed to volcanic glass contribution. Gustafson et al. [2012] also mapped possible DMDs on the western side of the central peak, and the spectrum S-M represents this unit (Figures 5a, 5e, and 5f). This spectrum has a wider, asymmetric 1 μm absorption shifted to longer wavelengths when compared to S-N and S-Se. However, the 2 μm absorption is shifted to a shorter wavelength. This characteristic distinguishes the western unit (represented by S-M) from the northeastern deposits. The characteristics of the absorption bands are consistent with the presence of volcanic glasses that shift the center of the 1 μm absorption to longer wavelengths and shorten the 2 μm absorption [Bell et al., 1976; Gaddis et al., 2003], and specifically more consistent with orange glass. The western unit (S-M) also has a low reflectance at 2.94 μm as seen in Figure 5d. High-resolution NAC images reveal an irregular depression that could be a vent in the western area. However, the extent of the low reflectance is not limited to the possible vent structure only and therefore could be associated with DMDs around the vent. Thus, the western deposit on the floor of crater Schluter (S-M) shows spectroscopic (i.e., glass signature and low albedo) and morphologic characteristics (i.e., possible vent structure) that make it a likely candidate for a pyroclastic deposit. The mafic mare-like spectral characteristics of the northeastern floor deposits of Schluter are likely to be associated with effusive volcanic deposits with varying mineralogy. However, their origin as pyroclastic cannot be completely discounted. None of the Schluter floor deposits exhibit evidence of the presence of olivine-rich minerals as previously suggested for some lunar DMDs [Hawke et al., 1989] Birt E Birt E is a noncircular feature mapped as a small crater (diameter of 5 km) that has been suggested to be a vent and the source of a small DMD in this area [Gustafson et al., 2012]. The Birt E feature is located in Mare Nubium and at the termination of the Rima Birt I linear rille. Another linear rille runs parallel to and west of Birt E, and this is also mentioned as a potential source for the dark albedo deposits. Among all the possible DMDs BESSE ET AL American Geophysical Union. All Rights Reserved. 10

11 analyzed here, Birt E has morphological characteristics that strongly suggest a pyroclastic origin: (1) irregular depression that could be a vent, (2) almost circular pattern of the dark albedo deposit that is consistent with Birt E as the source, and (3) a low albedo at 0.75 μm that is readily visible in the M 3 images (Figure 6a). Although mare units west and east of Birt E are readily separated in two units based on the IBD1000 (Figure 6b, yellow outline), the DMD resembles the western mare unit and this parameter alone does not serve to isolate the DMDs. However, Figure 6c clearly captures the extent of the DMDs by mapping the band depth at 1.25 μm (i.e., the blue channel of the color map), and this suggests that the 1 μm band center of the DMD is near that longer wavelength. As it can be seen in Figure 6f, the 1 μm band for the DMD is also slightly wider than in the mare unit spectra, suggesting that a component such as volcanic glass or olivine may be present. The spectra in Figures 6e 6f do not show strong variations of the band position at 2 μm but do show band centers near 2.1 μm (although the one from the DMD is shifted to slightly shorter wavelengths). It is clearly the wider 1 μm absorption band that characterizes the uniqueness of possible Birt E DMDs. Color ratio (i.e., ultraviolet-visible slope) using Clementine observations can also isolate the dark deposits suggesting a lower titanium concentration than the surrounding mare basalts [see Gustafson et al., 2012, Figure 14b]. It is not unusual to have strong variations of the titanium level in the DMDs [Gaddis et al., 2003], and in mare basalts as well [Staid et al., 2011]. Although the shape and position of the 1 μm absorption band of the DMDs could be consistent with olivine, the position of the 2 μm band and its strength relative to the 1 μm band is more consistent with the presence of volcanic glass, and in particular orange glass. From the M 3 observations, the Birt E DMDs candidates are not only distinguishable from the surrounding mare basalts in terms of albedo but they also exhibit different mineralogy. The change in position and shape of the absorption is also consistent with the presence of volcanic glass. Coupled with the morphological evidence, Birt E could be considered as a likely DMD. It is noted that the DMDs do not show darker reflectance at 2.94 μm as seen for Schluter. This could be due to the presence of mare basalt materials that are already very dark, thus making the identification of DMDs more difficult at this location and at the M 3 wavelengths Walther A Walther A is a small crater (diameter of 11 km) located within the crater Walther (135 km diameter), itself located in the highlands ~450 km northeast of Tycho and southeast of Mare Nubium (Figure 1). The dark deposits associated with possible DMDs are located north of Walther A, which is itself north of the central peak of Walther. The terrain in the crater floor around Walther A appears rougher with some topographic features that could be part of a central peak complex. From Figure 7a, several deposits appear darker in the region. However, only the deposits around Walther A display 1 μm mafic absorption bands (Figures 7b and 7c) and lower reflectance at 2.94 μm (Figure 7d). The mafic signatures match spatially with the lower reflectance of Figure 7d and the outline of the DMDs proposed by Gustafson et al. [2012]. The low-albedo deposits are mainly concentrated around the impact crater Walther A and are potentially related to ejected materials from the impact itself. However, some of the dark material is found inside the Walther A impact crater. Examination of LROC-WAC data (Figure 7e) shows that the dark deposit inside the crater (highlighted by black arrows) could also be related to ejecta from a smaller impact crater (highlighted by a white arrow). The low-albedo feature south of Walther A at 0.75 μm (i.e., location of spectrum W-S) still appears darker in the albedo map at 2.94 μm (Figure 7d), although it is lighter than the deposits around Walther A. However, as seen in Figures 7b and 7c, and also in the spectra of Figures 7e and 7f, these deposits do not exhibit mafic absorptions. Investigations of spectral characteristics as a function of distance from Walther A are plotted in Figure 7f with a continuum removal (W-Cs, W-Cm, and W- Cn). The only notable spectral variation observed is a change in the depth of the absorption for the 1 and 2 μm bands. No shift of the band position, as was seen in the case of Alphonsus where the band position changed with distance from the source [Jawin et al., 2013], was observed. This change in band depth is more likely related to the thickness of the deposits as a function of distance from the impact crater and/or a mixing with more feldspathic material from the floor. However, like the spectra from Schluter and Birt E, the band positions at 1 μm are shifted to longer wavelengths and the band position at 2 μm to shorter wavelengths relative to pyroxenes. This also indicates the presence of volcanic glass, which is a common component of DMDs [e.g., Gaddis et al., 1985]. The position of the 1 μm band and the relative intensity between the two bands is more consistent with the presence of orange glass. In the case of the Walther A dark deposit, no olivine signature could be observed in the deposits associated with the DMD. BESSE ET AL American Geophysical Union. All Rights Reserved. 11

12 Figure 7. Walther A as viewed by M 3.(a)Reflectance at 0.75 μm, (b) IBD1000, (c) color composite (R = BD950, G = BD1050, B = R1250; see Table 1), (d) reflectance at 2.94 μm with the proposed outline of the DMDs by Gustafson et al. [2012], (e) the spectra of locations given in Figure 7a, and (f) the same spectra with a continuum removal applied. (g) An LROC-WAC image of the crater Walther A. White arrow points to a circular depression (crater or vent) that exhibits strong mafic absorptions, as well as the black arrows that mantle the walls of the crater Walther A. In Figure 7f, the continuum removal shows that the 1 μm band of W-Cs, W-Cn, and W-Cm are shifted to longer wavelengths, a characteristic of volcanic glass. In Figure 7b, dark spots in the image are locations where no 1 μm absorptions are observed. Spectrum W- Spinel in Figures 7e and 7f exemplifies this observation and indicates that the deposit has a very strong 2 μm absorption and a weak-to-absent 1 μm absorption that is recognized as characteristic of spinel-rich materials [Cloutis et al., 2004]. These deposits are located along walls of the central peak complex of Walther. Of the two types of spinel discovered on the Moon by the M 3, this one more closely resembles the Mg-spinel first seen at Moscoviense [Pieters et al., 2011], rather than the regionally extensive very dark chromite-spinel deposits unique to Sinus Aestuum [Sunshine et al., 2010; Yamamoto et al., 2013]. We note that the Walther spinel deposits are spatially unrelated to the DMDs of Walther A and likely not associated with a pyroclastic origin. 5. Spectroscopic Properties, Classification, and Comparisons of DMDs Candidates 5.1. Contribution of Volcanic Glasses The ability to detect the presence or absence of volcanic glass with M 3 provides a new tool for assessing a possible explosive origin of some small, dark, likely volcanic deposits on the Moon. Figure 8 directly compares the spectra of the DMDs candidates studied here, before (Figure 8a) and after continuum removal (Figure 8b), to highlight the variation in volcanic glass content of the different sites analyzed. In Figure 8a, the wavelength range extends to 2.9 μm to highlight the quality of the thermal correction, which shows no reddening of the slope. From Figure 8a, it is noted that most M 3 spectra have relatively similar slopes. This characteristic is comparable to the slopes of laboratory spectra of volcanic glasses, in particular the green and orange glass sampled, respectively, at the Apollo 15 and 17 landing sites. Such spectral slopes could be influenced by variations in BESSE ET AL American Geophysical Union. All Rights Reserved. 12

13 Figure 8. (a and b) Spectral comparison of the newly proposed DMDs. Only a representative sample of the spectra from previous locations is plotted here to allow comparison of the band positions. At Figure 8b, the same continuum removal as in Figures 3 7 is used for both the M 3 and laboratory spectra. O glass corresponds to the laboratory spectrum of lunar pyroclastic orange spheres from Apollo 17 sample 74220, and G glass is a laboratory spectrum for green glass from sample from Apollo 15 [Adams, 1974; Pieters et al., 2000]. Spectra labeled W-Cs, W-Cm, W-Cn, Birt E, and S-M have band shapes and positions that more closely resemble those of orange volcanic glass. roughness, space weathering effects, and optical maturity; for example, for more mature surfaces, spectral slopes typically increase at longer wavelengths [Hapke, 2001; Noble et al., 2001; Taylor et al., 2001]. The observed DMD spectral slopes indicate that the deposits have comparable levels of maturity. However, the slope of the Schluter-M spectrum appears shallower than the other spectra. Along with Birt E, Schluter-M is the only possible vent identified within the DMDs candidates, and the topography of the vent itself could explain the variation in spectral slope. In Figure 8b, the removal of a linear continuum from the spectra highlights the variations of shape and positions of the absorption bands. The population of spectra can readily be separated into two groups: (1) spectra with a 1 μm absorption centered at 1.0 μm and a 2 μm absorption centered at 2.2 μm and (2) those having asymmetric 1 μm absorption bands centered at longer wavelengths (e.g., μm) and 2 μm absorption band centers shifted to shorter wavelength (e.g., 2.05 μm). Spectra S-M, W-Cs, W-Cm, W-Cn, and Birt E correspond to the second category. For comparison, continuum-removed laboratory spectra of green and orange lunar volcanic glass samples are also shown. The spectra of both volcanic glasses exhibit the same characteristic as the second group of in situ measurements with shifted absorption bands. The presence of volcanic glasses-rich material in spectroscopic signatures of lunar DMDs has been discussed previously [Gaddis et al., 1985, 2003; Weitz et al., 1998], but the higher spectral resolution at 1 and 2 μm absorptions in the M 3 data supports this more detailed and definitive identification and analysis of glasses in the small DMDs studied here. Although the wider and shifted 1 μm absorptions could also be attributed to the presence of olivine in some of these deposits, the shift to shorter wavelengths of the 2 μm absorptions and the relative strength of the 2 μm band are indicative of the presence of volcanic glasses and cannot be confused with olivine. Previous analyses [Hawke et al., 1989] may have overestimated the contribution of olivine in BESSE ET AL American Geophysical Union. All Rights Reserved. 13

14 Figure 9. Diagram of the band position at 1 and 2 μm. The positions are extracted from the continuum-removed spectrum of Figure 8b. The band position of old Procellarum low-ti and young Procellarum high-ti mare basalt are extracted from Staid et al.[2011,figure6].theycorrespondto end-members of basaltic flows within Oceanus Procellarum. A cluster of possible DMDs with band positions very close to the orange glass is used as a criterion to characterize them as DMDs. Birt E, which is a DMD located within a mare basalt unit, is plotted halfway between the two populations, probably because of a lower glass contribution. Note that titanium is not controlling the band position at 1 and 2 μmofthemarebasalts. many of the lunar DMDs because of the truncated spectra that did not include the full long-wavelength extent of the 2 μm band. Figure 9 highlights the two different populations of spectra in another way, by plotting the band center position of the 1 and 2 μm absorptions from the continuum-removed spectra of Figure 8b. The band position is estimated by the reflectance minimum in the band range of the 1 and 2 μm absorptions. Similar plots have been especially useful in analyzing the variability of pyroxene composition [e.g., Adams, 1974; Klima et al., 2007, Figure 10]. The band centers in the spectra plot in two clusters, one with band position consistent with orange glasses and thus interpreted as DMDs, and a second group associated with mare basalt signatures and not necessarily associated with DMDs. For comparison, data for both young high-ti and old low-ti mare basalts obtained with the M 3 in Oceanus Procellarum [Staid et al., 2011] are plotted. Similarity of the band centers of the second group of deposits with classical mare basalts is clearly shown. On the other hand, some DMDs are also well correlated with orange glasses. It is noted that S-M is on the right side of the orange glasses, and Birt E is halfway between the volcanic glasses and the mare basalts. In fact, S-M and Birt E are located within mare basalt material that is likely mixed with volcanic glasses. This trend could also be characteristic of variations of the juvenile component, with less juvenile glass. It is important to note it is not just the lack of a volcanic glass signature and the spectral resemblance to mare basalts that motivated the classification of some DMD candidates as unlikely. Hawke et al. [1989] have shown that compositions of a group of small pyroclastic deposits are clearly dominated by iron-rich, mare basalt-like compositions. Indeed, two of the proposed DMDs studied here (e.g., Lacus Luxuriae and Kopff) have mafic but not volcanic glass spectral signatures and also show morphologic characteristics (e.g., absence of volcanic vents and variable albedo) that make a pyroclastic origin unlikely. While the presence of volcanic glass is used here as an important distinguishing characteristic to assess a possible pyroclastic origin for any given DMD, it is acknowledged that the detection of volcanic glass is only the first step toward the full characterization of lunar pyroclastic deposits that is possible with data such as those from the M 3. An extension of this analysis to the list of nearly 100 DMDs described by Gaddis et al. [2003] and updated with any newly identified DMDs is needed, although this is beyond the scope of this paper Are the DMDs Candidates Pyroclastic in Origin? This analysis provides evidence on the likelihood of a pyroclastic origin of possible DMDs studied by Gustafson et al. [2012]. The evidence summarized in Table 3 and below can be separated into four main characteristics of pyroclastic DMDs: morphology (i.e., presence of vents and association with fractures and/or rilles), low albedo BESSE ET AL American Geophysical Union. All Rights Reserved. 14