Massive layer of pure anorthosite on the Moon


Corresponding author: S. Yamamoto, Center for Environmental Measurement and Analysis, National Institute for Environmental Studies, 16-2 Onogawa, Tsukuba, Ibaraki 305-8506, Japan. (


[1] We present a new global survey of the purest anorthosite (PAN) rock using the Spectral Profiler onboard Kaguya. We found that PAN rocks are widely distributed over the Moon, including the Feldspathic Highland Terrain and the south and north polar regions. All PAN sites are associated with huge impact structures with diameters larger than 100 km. Based on the global distributions of PAN and olivine-rich sites, we propose the existence of a massive PAN layer with a thickness of ∼50 km below an uppermost mafic-rich mixed layer with a thickness of ∼10 km. Below the PAN layer, a lower crustal layer with olivine-rich materials may be present on the nearside, but not on the far side of the Moon. The existence of a PAN layer with a thickness of ∼50 km suggests an Al2O3 abundance of 33 to 34 wt.% in the lunar crust, which is higher than previous estimates of <32 wt.%. Our data indicate the massive production event of PAN during the early stage of the formation of the Moon, supporting the lunar magma ocean scenario.

1. Introduction

[2] The lunar magma ocean (LMO) scenario proposes fractional crystallization of LMO-derived mafic cumulates to produce the mantle, and plagioclase floatation to produce the crust [Snyder et al., 1992; Wieczorek et al., 2006]. Several models have been proposed that describe the compositional and structural evolution of a crystallizing magma ocean, but there are still many uncertainties involved. For example, most investigations have suggested that a substantial portion of the LMO crystallized at equilibrium because of strong convection [e.g., Snyder et al., 1992; Elardo et al., 2011]. The global distributions of major lunar minerals can provide important information on this topic [Tompkins and Pieters, 1999]. Recent hyperspectral observations by the Spectral Profiler (SP) on board the Japanese lunar mission SELENE/Kaguya revealed the surface distribution of olivine-rich material (ORM) on the Moon, possibly from both the lower crust and the upper mantle [Yamamoto et al., 2010]. In addition, multispectral observations by the Multiband Imager (MI) on board Kaguya revealed the existence of the purest anorthosite (PAN) [Ohtake et al., 2009], providing important information on the primordial lunar crust. However, the global surface distribution of PAN is still unclear: to what depth and to what lateral extent does PAN exist? Furthermore, it is necessary to determine the relation between the global distributions of PAN and ORM in order to fully understand the compositional and structural evolution of the LMO. We thus conducted a new global survey to reveal the global distribution of PAN based on the SP data [Matsunaga et al., 2008; Yamamoto et al., 2011], using the same algorithm that was used to determine the ORM distribution.

2. SP Global Survey

[3] The SP recorded continuous spectral reflectance data for about 70 million points (0.5 by 0.5 km footprint) on the Moon in the wavelength range λ = 0.5−2.6 μm and with a spectral resolution of 6–8 nm during the mission from November 2007 to June 2009 [Matsunaga et al., 2008; Yamamoto et al., 2011]. As in the case of the global survey based on olivine-rich spectra [Yamamoto et al., 2010], we searched the entire SP data set (∼70 million spectra) focusing on the 1.25 μm diagnostic band for plagioclase as follows: Data in the SP Level 2B (L2B) [Yamamoto et al., 2011] were corrected using the reflectance spectra of the Apollo 16 landing site and the spectral reflectance of the Apollo 16 soil 62231 as measured in the laboratory [Pieters, 1999]. We then reject (1) data for which the radiance at a wavelength λ = 0.5126 μm is less than 23.3 W m−2 μm−1 sr−1 and (2) data for which the continuum-removed reflectance Rc does not contain an absorption band with Rc < 0.95 between λ = 0.7 and 1.6 μm. This is because these data cannot be used to examine whether a characteristic plagioclase band exists for λ = 1.25 μm. We next search for wavelengths at which Rc between λ = 0.7 and 1.6 μm are the first-, second-, and third-lowest values, and select data for the case in which these three wavelengths are between 1.20 and 1.35 μm. We also reject low signal-to-noise data (i.e., jagged spectral data) for which the difference in Rc between λ = 0.9446 and 1.4598 μm is equal to or larger than 0.019. Using this algorithm, we finally identified 564 observational points that show a clear plagioclase band around 1.25 μm, as shown in Figures S1–S6 in the auxiliary material. Ohtake et al. [2009] showed that a greater absorption depth at 1.25 μm than at other lunar mineral bands (e.g., 1.05 μm, 1.00 μm, or 0.95 μm) is a basic criterion for identifying the spectra with modal abundance of plagioclase ≥98 vol.%. We thus consider that these spectra found by our global survey indicate the existence of PAN.

3. Results

[4] In Figure 1, we plot the 564 PAN points on a lunar crustal thickness map. Note that 561 of the points appear to be grouped into major lunar basins or craters with diameters larger than ∼100 km. On the nearside of the Moon, PAN points are found around the (A) Orientale, (B) Grimaldi, (C) Serenitatis, (D) Crisium, (E) Nectaris, (F) Humorum, (G) Humboldtianum, (H) Mendel-Rydberg, (I) Australe, and (J) Schiller-Zucchius basins. In the Farside Highland Terrain (FHT), PAN sites are found around the (K) Hertzsprung, (L) Korolev, (M) Dirichlet-Jackson, (N) Freundlich-Sharonov, (O) Moscoviense, (P) Mendeleev, (Q) Coulomb-Sarton, (R) Birkhoff, and (S) Lorentz basins. PAN points are also found in the outer region of the South Pole-Aitken (SPA) basin and around the north polar region (Figures 1 and 2). Thus, PAN rocks are widely distributed over the Moon, not only in the FHT but also in the north and south polar regions.

Figure 1.

Locations of PAN points (white circles) on the Moon plotted on a total lunar crustal thickness map (crustal materials and mare basalt fills) based on SELENE gravity measurements and a topographic model produced by the Kaguya explorer [Ishihara et al., 2009; Araki et al., 2009; Namiki et al., 2009; Wessel and Smith, 1991]. Red circles indicate ORM points found by the SP survey [Yamamoto et al., 2010]. The names of PAN-bearing impact basins or craters are listed in Tables S1–S3 of the auxiliary material.

Figure 2.

Local distribution of PAN sites (white circles) around different basins and large craters plotted on the surface topography obtained by the Kaguya mission [Araki et al., 2009]. The names of PAN-bearing impact basins or craters are listed in Tables S1–S3 of the auxiliary material. Photos N1, K1, B1, A5, and V1 are close-up images of PAN sites taken by MI or the Terrain Camera (TC) onboard Kaguya [Ohtake et al., 2009; Haruyama et al., 2008], where PAN points are plotted as red rectangles with white 5 km scale bars. The size of the rectangle corresponds to 500 m × 500 m. The accompanying spectra figures show the continuum-removed reflectance spectra Rc at the location of the yellow square on each close-up image. Note that the PAN points and close-up images for each site are not always recorded simultaneously, but during different orbits of Kaguya.

[5] On a localized scale, all regions with multiple PAN points are associated with small fresh craters or the slopes of peaks. Figure 2 shows close-up images of several PAN points. For example, in the Hertzsprung basin, clear PAN spectra are found on the rim and in the ejecta region of a 3-km-wide crater on the inner ring (K1 in Figure 2). At the A5 site in the Orientale basin, PAN points are found on the slopes of peaks (A5 in Figure 2). In the outer region of the Amundsen-Ganswindt basin, many PAN points are found only on the wall of the Shackleton crater (V1 in Figure 2). Thus, all PAN points are located on fresh craters or the slopes of peaks associated with large impact structures. On the other hand, the spectral features for areas outside these PAN exposures are too unclear to reliably interpret their mineral composition. This is because most of the lunar surface is covered with mixtures of different minerals and space weathering obscures the spectral features [Hapke, 2001]. However, recent impacts or landslides on the slopes of peaks have exposed PAN materials, allowing them to be identified by spectral remote sensing. The occurrence of multiple PAN points on fresh craters or peak slopes is similar to the case for ORM [Yamamoto et al., 2010, 2012]. Note that outside these large impact basins/craters, PAN spectra are not detected even at fresh craters or peak slopes. Therefore, based on the local geological context, we divided the 564 PAN points into 134 PAN sites in 23 impact basins and 26 large craters (Tables S1–S3 in the auxiliary material).

[6] Our global survey did not identify PAN exposure in the center region of the Procellarum KREEP Terrane (PKT) or in the center region of the SPA terrain (SPAT). For example, no PAN was detected in the Aristarchus crater, where plagioclase spectra were previously reported. One reason for this is probably the incomplete coverage of the SP footprint. However, for each PAN site, most of the spectra were measured for several consecutive SP footprints (Figure 2), indicating that the PAN exposures extended over several kilometers. Since the average interval between SP footprints is ∼1.7 km at the equator, the SP survey could have detected large PAN exposures with widths of several kilometers if they were present in the center region of PKT or SPAT. Indeed, Figure 3 shows the Multiband Imager (MI) image for central peak of the Aristarchus crater, where the area of PAN exposures with a strong 1250 nm band (locations marked B and D) in this crater extends over only a few hundred meters scale, which is smaller than the SP footprint size. Thus, although we cannot rule out the existence of smaller PAN exposures, it is likely that there are no large PAN exposures in the center region of PKT or SPAT of the type identified in other terrains by the SP survey [see also Ohtake et al., 2009; Mustard et al., 2011].

Figure 3.

Color-composite image map of the central peak of the Aristarchus crater taken by MI. All reflectance spectra are given as the average of a 120 m by 120 m area to remove spatial variation. The size of the rectangles in Figure 3a corresponds to 120 m by 120 m. The small light blue areas at the locations of B and D show a 1250 nm band of PAN. Outside these areas, we did not find PAN.

[7] We next examined the relation between the PAN and ORM sites. In Figure 1, the distribution of ORM sites by Yamamoto et al. [2010, 2012] is also plotted. Most are found around large impact basins located in thinner crustal regions on the nearside of the Moon. Even in the farside region, the ORM sites are located on the thinnest crustal regions in the Moscoviense [Ishihara et al., 2009] and SPA basins. On the other hand, many PAN sites are found in thicker crustal regions, especially in FHT. Some basins in thinner crustal regions possess both PAN sites and ORM sites. The Imbrium basin, located in a thinner crustal region, has many ORM sites but no PAN site. Thus, the ORM sites are found only in thinner crustal regions, while the PAN sites are located mainly in thicker crustal regions but also occur with ORM sites in thinner crustal regions.

4. Discussion

[8] How do we interpret these global distributions? An important result is that PAN sites are always associated with large impact structures wider than ∼100 km. This indicates that the PAN materials come from regions deeper below the surface, and the near-surface layer is not dominated by them. In addition, the fact that the PAN sites are widely distributed over the Moon suggests that PAN materials originally formed a continuous layer, and are not small portions of discrete plutons. Therefore, our data suggest the existence of an extensive PAN layer below the surface layer which is poorer in PAN. This is consistent with the crustal model by Hawke et al. [2003], who proposed the existence of a mafic-rich mixed layer above the plagioclase-rich layer. Furthermore, the fact that the ORM sites are found only in large impact basins in thinner crustal region indicates that ORM originates from much deeper regions than PAN does.

[9] A more quantitative analysis is shown in Figure 4, which plots the depth of origin (H) of PAN or ORM against the proximity to the crust-mantle boundary (P = TH) [Cahill et al., 2009] for each impact basin/crater, where T is the crustal thickness before impact. (The estimates of H and T are described in Ishihara et al. [2010] and Text S1 in the auxiliary material.) From this figure, three conclusions can be formed. First, it is clear that there is no excavation of PAN materials for H < ∼10 km. This indicates that the uppermost layer with a thickness of ∼10 km is not dominated by PAN. Second, exposures of PAN without ORM can be observed up to H ∼ 60 km. This suggests that the layer of PAN rocks extends to a depth of ∼60 km. In other words, there is a massive PAN layer with a thickness of up to ∼50 km below the uppermost mixed layer. Third, in the five basin formations in group B (dashed circle), the ORM could have been excavated from layers with P ∼ 7–16 km. This may indicate that the thickness of the ORM-containing lower crust is ∼16 km. On the other hand, in spite of having similar values of P, in the four basin formations in group A (dashed ellipse), ORM was not excavated. The group B basins are located in the longitude ranging from −60° to +120°, while the group A basins are located outside this longitude range. This may therefore indicate a difference in the composition or the thickness of the lower crust between one hemisphere and the other of the Moon, i.e., a dichotomy of the lunar lower crust. This may be due to asymmetric crystallization of a primordial magma ocean [Arai et al., 2008].

Figure 4.

Depth of origin (H) for PAN-bearing impact basins or craters vs. proximity to the crust-mantle boundary (P = TH), where T is the crustal thickness before impact. Here we do not show the Schrödinger and Ingenii basins within the SPA basin. Group A (dashed ellipse) includes the Orientale, Grimaldi, Mendel-Rydberg, and Freundlich-Sharonov basins. Group B (dashed circle) includes the Serenitatis, Humboldtianum, Nectaris, Humorum, and Crisium basins.

[10] The absence of large PAN exposures in the center region of SPAT and PKT may be explained as follows. The impact that formed the SPA basin could have blasted away most of the primordial PAN layer. In addition, it has been proposed that a huge impact (putative Procellarum impact) removed a large part of the primordial crust in PKT [Wilhelms, 1987; Hawke et al., 2003]. Although some crustal components might have been left, the center regions of the basins were filled with a melt composed of crustal and mantle materials [Nakamura et al., 2009; Ivanov et al., 2010; Yamamoto et al., 2012]. Local differentiation of the melt might have produced small portions of plagioclase-rich materials, which may account for the small exposures of PAN at Aristarchus crater as shown in Figure 3. Local differentiation of melted materials may also explain the existence of lower density materials in the center of SPAT and PKT [Wieczorek and Phillips, 1999], and the isotopic composition of the feldspathic sample in PKT [Arai et al., 2008]. Thus, smaller exposures of PAN in PKT (and SPAT) would originate from impact melt or discrete pluton, while the globally distributed PAN rocks identified in this study originate from the LMO.

[11] What do our data imply for the evolution and the composition of the Moon? The most important result is the existence of a massive PAN layer with a thickness of ∼50 km. This indicates the massive production event of PAN during the early stage of the formation of the Moon, supporting the lunar magma ocean scenario, which has been questioned by several previous authors [e.g., Borg et al., 2011]. In addition, following a previously reported calculation approach [Wieczorek and Zuber, 2001], we estimated the Al2O3 abundance in the lunar crust as follows: First, the PAN layer is assumed to contain 35% Al2O3 and have a thickness of 50 km, with an An96 composition [Warren, 1990] and a purity of 98% plagioclase [Ohtake et al., 2009]. We then assume that the lower crust is composed of 24–29% Al2O3 [Wieczorek and Zuber, 2001] with a thickness of 10–20 km. In this case, the overall Al2O3 abundance in the lunar crust is estimated to be 33–34 wt.%. (Here we neglect the contribution of the uppermost mixed layer). This value is larger than the previous estimates of ∼26–32 wt.% [Lucey et al., 1995; Wieczorek and Zuber, 2001; Longhi, 2006]. Therefore, our data indicate that the lunar crust is richer in refractory elements (Al or Ca) than previously thought. Our data may provide a new constraint on the modeling of the formation of the Moon [e.g., Taylor et al., 2006].


[12] We thank all the members and contributors of the SELENE (Kaguya) project and the Lunar Imager/Spectrometer (LISM) team for their help with the development, operation, and data processing of SELENE and LISM. We also thank J. Taylor and anonymous reviewer for helpful comments. The present study was supported in part by a grant-in-aid for scientific research (C) (23540502) (S.Y.) from Japan Society for the Promotion of Science (JSPS).

[13] The Editor thanks the two anonymous reviewers.