Thorium abundances of basalt ponds in South Pole Aitken basin: Insights into the composition and evolution of the far side lunar mantle

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 116,, doi: /2010je003723, 2011 Thorium abundances of basalt ponds in South Pole Aitken basin: Insights into the composition and evolution of the far side lunar mantle J. J. Hagerty, 1 D. J. Lawrence, 2 and B. R. Hawke 3 Received 26 August 2010; revised 6 December 2010; accepted 2 March 2011; published 3 June [1] Imbrian aged basalt ponds, located on the floor of South Pole Aitken (SPA) basin, are used to provide constraints on the composition and evolution of the far side lunar mantle. We use forward modeling of the Lunar Prospector Gamma Ray Spectrometer thorium data, to suggest that at least five different and distinct portions of the far side lunar mantle contain little or no thorium as of the Imbrian Period. We also use spatial correlations between local thorium enhancements and nonmare material on top of the basalt ponds to support previous assertions that lower crustal materials exposed in SPA basin have elevated thorium abundances, consistent with noritic to gabbronoritic lithologies. We suggest that the lower crust on the far side of the Moon experienced multiple intrusions of thorium rich basaltic magmas, prior to the formation of SPA basin. The fact that many of the ponds on the lunar far side have elevated titanium abundances indicates that the far side of the Moon experienced extensive fractional crystallization that likely led to the formation of a KREEP like component. However, because the Imbrian aged basalts contain no signs of elevated thorium, we propose that the SPA impact event triggered the transport of a KREEP like component from the lunar far side and concentrated it on the nearside of the Moon. Because of the correlation between basaltic ponds and basins within SPA, we suggest that Imbrian aged basaltic volcanism on the far side of the Moon was driven by basin induced decompressional melting. Citation: Hagerty, J. J., D. J. Lawrence, and B. R. Hawke (2011), Thorium abundances of basalt ponds in South Pole Aitken basin: Insights into the composition and evolution of the far side lunar mantle, J. Geophys. Res., 116,, doi: /2010je Introduction 1.1. South Pole Aitken Basin [2] South Pole Aitken (SPA) basin, one of the largest and oldest basins in the entire Solar System [e.g., Wilhelms, 1987; Pieters et al., 2001], has been and will continue to be an area of intense interest to lunar scientists. Hypervelocity impact models indicate that SPA basin may have accessed the lower crust and/or the upper mantle of the far side of the Moon [Warren, 1996; Schultz, 1997; Lucey et al., 1998; Pieters et al., 1997, 2001], and therefore has the potential to provide a rare glimpse into normally inaccessible portions of the Moon, thus providing important constraints on the formation and evolution of the Moon as a whole. The crust below SPA basin is relatively thin compared to the rest of the far side lunar crust [Zuber et al., 1 Astrogeology Science Center, U.S. Geological Survey, Flagstaff, Arizona, USA. 2 Johns Hopkins University Applied Physics Laboratory, Laurel, Maryland, USA. 3 Hawai i Institute of Geophysics and Planetology, University of Hawai i atmānoa, Honolulu, Hawaii, USA. Copyright 2011 by the American Geophysical Union /11/2010JE ; Smith et al., 1997; Wieczorek et al., 2006; Ishihara et al., 2009] and therefore, based on our knowledge of basaltic magmatism on the nearside of the Moon, SPA should contain an abundance of basaltic materials. Nevertheless, SPA basin is relatively lacking in basaltic eruptions compared to the nearside of the Moon. Given the correlation between basaltic magmatism and energy producing elements on the nearside of the Moon [Lawrence et al., 2003], some workers have used the relatively sparse occurrence of basaltic materials in SPA to imply a low abundance of energy producing elements on the lunar far side [e.g., Lawrence et al., 2003]. Even though the abundance of basaltic materials in SPA is limited relative to the lunar nearside, there are as many as 52 different and distinct basalt ponds that litter the floor of SPA basin [Yingst and Head, 1997, 1999]. Basalt ponds are simply accumulations of basaltic magmas that are typically confined within lunar basins and craters. The basaltic ponds within SPA basin are important to this study because we can use them to indirectly investigate various characteristics of the far side mantle Basalt Ponds in SPA Basin [3] Multiple studies have shown that basaltic lithologies on the lunar surface represent partial melts from the lunar 1of23

2 Figure 1. LP GRS Th map of the SPA basin (outlined in white), deconvolved using Pixon method described by Lawrence et al. [2007]. SPA basin, which has an average of 2 3 ppm Th, is enriched in Th compared to the surrounding feldspathic highlands terrain. mantle, and as such have the potential to provide valuable information about the lunar mantle [Papike et al., 1998; Shearer et al., 2006]. More specifically, it is possible to use compositional information derived from the basaltic materials to infer the composition of the mantle from which the basaltic magmas were derived, provided that the magmas experienced limited assimilation or contamination during ascent [e.g., Neal et al., 1988; Neal and Kramer, 2006]. Therefore, determining the thorium (Th) composition of the basalt ponds within SPA not only can provide information about the Th content of the underlying lunar mantle, but can also provide information about the Th abundance distribution within the basin and possibly for the entire Moon. The vast majority of basalt ponds within SPA are mapped as Imbrian in age [e.g., Stewart Alexander, 1978; Wilhelms et al., 1979; Yingst and Head, 1997, 1999] with a few ponds that may have erupted in the Eratosthenian [e.g., Haruyama et al., 2009]. The ages of these ponds are important because they can provide information about the lunar mantle for specific portions of lunar history Thorium Abundances in SPA Basin [4] Thorium has several characteristics that make it a critical element for understanding lunar geochemistry and evolution. For instance, Th is an energy producing element that is predicted to be enriched in a geochemically important, but hypothetical layer of melt sandwiched between the lunar crust and mantle, known as urkreep [Warren and Wasson, 1979]. The ur in urkreep refers to the primeval nature of the layer, whereas, K = potassium, REE = rare earth elements, and P = phosphorus [Warren and Wasson, 1979]. The primeval layer of melt is thought to contain an abundance of elements that are not compatible in the crystal structures of major minerals that crystallized to form the Moon. The incompatible elements in the urkreep layer, including Th and U, have very similar chemical affinities, which means that the abundances and distributions of these elements are well correlated with one another, making it possible to use Th as a proxy for urkreep abundance and distribution [Jolliff et al., 2000]. [5] Thorium, like other elements in the urkreep layer, is a radioactive element that releases energy upon decay. This decay can serve as a source of heat for the thermal evolution of the lunar mantle, particularly after primordial accretionary heat has been lost. The energy producing elements in the urkreep layer have been linked to the abundant basaltic volcanic activity on the nearside of the Moon [Jolliff et al., 2000; Wieczorek and Phillips, 2000]. [6] Thorium is also one of the few elements that can be easily and accurately detected by both remote sensing and sample analysis techniques, which means that there are comparable data sets of different scales that can be combined to provide a wealth of information about the Moon. The Th data set from the Lunar Prospector Gamma Ray Spectrometer (LP GRS) is considered robust, because Th decay on the lunar surface produces a gamma ray signal that is strong, has a large dynamic range reflective of compositional variability in surface abundances, and can be easily measured from space [Reedy, 1978; Lawrence et al., 2003]. [7] The LP GRS Th map for SPA basin (Figure 1) shows that the basin has an elevated Th abundance relative to the surrounding feldspathic highlands terrane [Jolliff et al., 2000]. The basin has an average Th content of 2 3 ppm, with a regional high approaching 6 ppm Th in the northwestern portion of the basin (Figure 1). Several previous studies have attempted to identify the source of the Th in the basin [e.g., Haskin et al., 2004; Garrick Bethell and Zuber, 2005] and have concluded that the Th in the basin cannot be antipodal ejecta from Th rich basins on the nearside of the Moon, and must therefore be inherent to the basin itself. [8] Because the footprint of the LP GRS is relatively large (80 km full width, half maximum [Lawrence et al., 2003]), it is difficult to determine what specific features and/or lithologies are directly responsible for the Th enhancement of SPA basin. Fortunately, we can use forward modeling of the LP GRS Th data to determine Th abundance distribution 2of23

3 Figure 2. CSR albedo map (natural color) of the SPA basin, showing the location of all of the basalt ponds investigated in this study. Image was created using the USGS map a planet software. at a finer scale [e.g., Hagerty et al., 2006a, 2009; Lawrence et al., 2003, 2007]. The goal of this study is to determine the Th content of basaltic ponds in SPA basin using forward modeling and in turn place constraints on the geochemical evolution of the Moon. 2. Identifying Regions of Interest [9] As part of a comprehensive survey of multiple basalt ponds in SPA we used LP GRS, LP neutron spectrometer, and Clementine Spectral Reflectance (CSR) data to determine if there are large expanses of uncontaminated basalt that can be reliably used to obtain compositional information about the underlying mantle sources. We considered thirteen different ponds for this study (Figure 2), but for the sake of brevity we only present detailed discussion of regions that were good candidates for the forward modeling process. In sections we provide background information and present the rationale for selecting specific regions of interest Apollo Basin [10] Apollo basin, located in the northeastern portion of SPA (i.e., 36.1 S, W), is a 505 km diameter, pre Nectarian impact basin [Yingst and Head, 1997, 1999] (Figure 2). Apollo basin is the largest impact structure within SPA and therefore represents the deepest penetration into the far side lunar crust [Petro and Pieters, 2004, 2008; Petro et al., 2010]. The penetration depth of Apollo basin is important for several reasons. First, Apollo basin and its associated ejecta provide a window into the composition and structure of the far side lunar crust. Second, Apollo basin provides the clearest, easiest route for ascension of basaltic magmas within SPA. We refer to these basaltic ponds as Apollo north, Apollo south, and Apollo west (Figure 2). A fourth pond is also present in the eastern portion of Apollo basin but this pond is too small to be effectively investigated in this study and therefore will not be discussed further. Below, we provide brief geologic descriptions of each of the basalt ponds in Apollo basin Apollo North Pond [11] The basaltic pond located in the center of Apollo basin (36 S, 152 W), consists of a single, Imbrian aged pond that has a surface area of 8,335 km 2, an average thickness of km, and a volume of 3,125 km 3 [Yingst and Head, 1997, 1999]. The crust below the pond is estimated to be only 20 km thick [Zuber et al., 1994; Wieczorek et al., 2006], making it the thinnest portion of crust on the far side of the Moon [Petro and Pieters, 2004, 2008]. The geologic map for this region (Figure 3a) shows the Imbrianaged basalt pond as a single homogenous unit (e.g., unit Im in Figure 3a). The pond is surrounded by Nectarian aged rolling terra (Nt) and light plains material (INp) that make up the floor of Apollo basin. The region also contains Imbrian aged craters (e.g., Ic, Ic 1, and Ic 2 ) as well as a few Nectarian aged craters (Nc). [12] The CSR albedo map of this pond (Figure 3b) indicates that the pond is quite dark (i.e., low albedo) and relatively free of nonmare cover, especially in the southwestern portion of the pond (Figure 3b). However, craters to the north and northeast of the pond have ejected nonmare material onto the pond. There are also several small craters on the south side of the pond that have also ejected nonmare material onto the pond (Figure 3b). The CSR FeO map (Figure 3c) confirms the presence of low FeO (i.e., nonmare) materials on the pond. The black outline in Figure 3c represents the area of the pond that contains little or no nonmare ejecta. This is the area from which we will derive the most realistic estimate of thorium, which is likely to represent the thorium content for entire pond. [13] One of the most useful tools for evaluating each of the basaltic ponds in this study is the Clementine color ratio ( false color ) map (Figure 3d), which was created by calculating ratio images using three of the five Clementine ultraviolet visible (UVVIS) camera bands and combining these into the red (750 nm/415 nm), green (750 nm/950 nm), and blue (415 nm/750 nm) channels of a color image [e.g., Pieters et al., 1994]. The color ratio technique serves to reduce the brightness variations of the scene and enhances color differences related to soil mineralogy and maturity [Pieters et al., 1994]. For instance, anorthositic lithologies in this map are depicted in shades of red (more mature) and blue (less mature), while the mare basalts are portrayed in shades of yellow/orange (iron rich, low titanium) and blue/ purple (iron rich, higher titanium) [Pieters et al., 1994]. The blue/purple color of the uncontaminated portion of the Apollo north pond (Figure 3d) reflects the elevated iron and titanium abundances for the basalt in that region. [14] Posteruption craters on the pond have not excavated high albedo material and therefore did not excavate below the basalt pond, indicating that the uncontaminated portion of the pond shows little or no vertical or horizontal variation. Based on the above information we determined that the southwestern portion of the pond (i.e., black outline in Figure 3c) provides the most representative composition for the basaltic pond as a whole. However, it is important to note that the Th compositions of the nonmare materials (i.e., the overlying material) in the northeastern portion of the pond can also be modeled to obtain information about the 3of23

4 Figure 3. (a) Digitized geologic map for the region including the Apollo north basalt pond [from Stewart Alexander, 1978]. Units of the same color are mapped as the same geologic unit by Stewart Alexander [1978]. (b) CSR albedo map for the Apollo north basalt pond showing the nonmare ejecta on top of the pond and potential source craters. (c) CSR FeO map for the Apollo north basalt pond. The uncontaminated portion of the basalt pond is outlined in black. (d) Color ratio map for the Apollo north pond. The blue/purple color of the uncontaminated portion of the pond reflects the elevated iron and titanium contents in the pond. overlying nonmare lithologies (see section 3 for more details) Apollo West Pond [15] The westernmost basalt pond within Apollo basin is located at 37 S 160 W (Figure 2). This pond, referred to as Apollo west, consists of a single, Imbrian aged basalt pond, with a surface area of 5,430 km 2, an average thickness of km, and a volume of 2,035 km 3 [Yingst and Head, 1997, 1999]. Zuber et al. [1994] estimated that the crust below this pond is approximately 35 km thick, which is thicker than the crust below Apollo north and more similar to crustal thicknesses on the nearside of the Moon [Wieczorek et al., 2006]. The geologic map of the region (Figure 4a) indicates that the pond is a single, homogenous unit. Note that we use the geologic maps only as a basis of preliminary comparison and initial orientation. Subsequent identification of individual geologic features and/or lithologies is made based on compositional and morphologic data derived from the Lunar Prospector and Clementine missions. [16] Apollo west is located just inside the pre Nectarian rim of Apollo basin (e.g., pnb) and is largely surrounded by light plains material (e.g., Ip, INp) thought to have been derived from an Imbrian or Nectarian aged basin. The region of interest also contains Nectarian, Imbrian, and Eratosthenian craters (Nc, Ic, and Ec, respectively). [17] The CSR albedo map for this region (Figure 4b) shows that the Apollo west pond is relatively small and complex with at least two observable posteruption impact craters. The albedo of the interior of largest posteruption crater is high compared to the basalt; however, the fact that there are no high albedo ejecta deposits around the crater indicates that the apparent albedo difference is due to a lighting effect and only basaltic material was ejected. Conversely, a small posteruption crater in the southern portion of the pond appears to have some high albedo, nonmare ejecta associated with it. The posteruption craters and the nonmare materials can be seen in the CSR FeO map of the 4of23

5 Figure 4. (a) Digitized geologic map for the region including the Apollo west basalt pond [from Stewart Alexander, 1978]. Units of the same color are mapped as the same geologic unit by Stewart Alexander [1978]. (b) CSR albedo map for the Apollo west basalt pond showing posteruption modification of the basalt pond. (c) CSR FeO map for the Apollo west basalt pond. For clarity we have divided the pond into two portions, a northern portion and a southern portion. (d) Color ratio map for the Apollo west pond. The yellow/orange color of the uncovered portion of the pond reflects the low titanium content of the pond. region (Figure 4c). The black outline in Figure 4c represents the total areal extent of the pond. [18] The CSR color ratio map for the Apollo west pond (Figure 4d) shows yellow/orange exposures of basalt that are consistent with the low to intermediate Ti content of the pond. The color ratio map also shows that the southern portion of the pond almost blends in with the surrounding highland terrain, indicating that this portion of the pond has seen abundant mixing with highland materials. Nevertheless, we find that the northern part of the pond, just above the posteruption impact crater (Figures 4c and 4d), has a basaltic signature and is distinct from the surrounding terrain, indicating that this portion of the pond is relatively free of nonmare cover. The relatively uncontaminated portion of the pond provides the most representative compositional information for the Apollo west pond Apollo South Pond [19] The largest basalt pond in Apollo basin, referred to as Apollo south, is located at 42 S, 155 W (Figure 2). Yingst and Head [1997, 1999] demonstrated that the southern Apollo pond consists of multiple Imbrian eruptions that cover a total surface area of 11,445 km 2, have an average thickness of km, and a total volume of 4,290 km 3. Zuber et al. [1994] indicated that the crust below this pond is approximately 30 km thick, which is consistent with crustal thicknesses on the nearside of the Moon [Wieczorek et al., 2006]. The geologic map of the area (Figure 5a) shows the basaltic pond as a very large, continuous/ homogenous unit. The entire Apollo south pond is bounded 5of23

6 Figure 5. (a) Digitized geologic map for the region including the Apollo south basalt pond [from Stewart Alexander, 1978]. Units of the same color are mapped as the same geologic unit by Stewart Alexander [1978]. (b) CSR albedo map for the Apollo south basalt pond showing posteruption modification of the basalt pond, including ejecta from Chaffee S crater. (c) CSR FeO map for the Apollo south basalt pond. For clarity we have divided the pond into three portions, a western portion, a central portion, and an eastern portion. (d) Color ratio map of the Apollo south pond. The blue/purple color of the uncovered portion of the pond reflects the elevated iron and titanium contents in the pond. on the south by pre Nectarian basin material (e.g., pnbr and pnbm) and on the north by light plains materials from an Imbrian basin (e.g., INp) [Stewart Alexander, 1978]. The region also contains Nectarian terra (Nt) as well as Nectarian and Eratosthenian craters (Nc and Ec, respectively). The western portion of the Apollo south pond appears to be partially contaminated by ejecta from Chaffee S crater, labeled as Ec in the geologic map (Figure 5a). [20] The CSR albedo map for this region (Figure 5b) shows that this pond has seen much more posteruption modification than the other Apollo ponds (i.e., primary and secondary impact cratering and mass wasting from crater walls). The northern portion of south pond appears to have been contaminated by ejecta from the relatively recent Chaffee S crater (Figure 5b). The eastern portion of the pond is not well exposed and appears to have seen significant posteruption modification. The CSR FeO map for the region (Figure 5c) shows how much nonmare material actually covers the pond. The black outlines in Figure 5c represent the total physical extent of the pond as derived from the CSR albedo map (Figure 5c). The blue/purple color of the central portion of the pond in the color ratio map (Figure 5d) is consistent with elevated Ti abundances. In summary, the central portion of the pond exhibits the lowest albedo and the highest FeO and TiO 2 concentrations, which indicates that the central portion of the pond provides the most representative compositional information for this mare deposit Rumford Crater [21] Rumford crater is a 64 km diameter, Imbrian aged crater, located in the northern portion of SPA basin (29 S, 170 W) (Figure 2). The basalt pond on the floor of Rumford crater has a surface area of 895 km 2, an average thickness of 0.46 km, and a volume of 415 km 3 [Yingst and Head, 1997, 1999]. The inferred crustal thickness for this area is 40 km [Zuber et al., 1994]. The pond is completely enclosed by Rumford crater (Ic 1 ) (Figure 6a). Rumford crater itself is surrounded by Nectarian and pre Nectarian terra (e.g., pnt, NpNt, and Nt) (Figure 6a). [22] The CSR albedo map for Rumford crater (Figure 6b) shows that the basaltic pond has seen relatively little posteruption modification; however, there are several interesting characteristics to address. For example, there is a feature in the approximate center of the pond that is reminiscent of an embayed central peak. Regardless of origin, the peak like feature is too small to contribute significantly to the composition of the pond, at least as seen from LP GRS or CSR. It should also be noted that the crater rim has experienced some mass wasting (Figure 6b). Fortunately, the products of the mass wasting appear to have been confined to the edges of the basalt pond, leaving the bulk of the pond s surface 6of23

7 Figure 6. (a) Digitized geologic map for the region including the basalt pond inside of Rumford crater [from Stewart Alexander, 1978]. Units of the same color are mapped as the same geologic unit by Stewart Alexander [1978]. (b) CSR albedo map for the pond in Rumford crater showing posteruption modification of the basalt pond, as well as the presence of a central peak. (c) CSR FeO map for the basalt pond in Rumford crater. The entire pond (outlined in black) appears to have little or no nonmare contamination. (d) Color ratio map for the pond in Rumford crater. The yellow/orange color of the exposed basalts is consistent with the low Ti content of the pond. exposed. The CSR FeO map (Figure 6c) reveals that the basalt pond in Rumford crater has little nonmare cover, while the CSR color ratio map (Figure 6d) shows that the pond is distinctly different in composition from the surrounding material and has a yellow/orange color, consistent with lower Ti abundances than were seen in the Apollo ponds. The CSR data indicate that the pond in Rumford crater is relatively uncontaminated and provides clear compositional information Chrétien Crater [23] Chrétien crater is an 80 km diameter, heavily degraded, Nectarian crater located at 46.5 S and 163 E (Figure 2). The basaltic material on the floor of Chrétien crater appears to consist of Imbrian aged basalt eruptions that spread out into intercrater terrain in the north (Figure 7a). In total, the surface area of basaltic material is 4,565 km 2, with an average thickness of 0.4 km, and a total volume of 1,795 km 3 [Yingst and Head, 1997, 1999]. The crust below the basaltic material is relatively thick (i.e., 35 km) for the SPA basin [Zuber et al., 1994; Wieczorek et al., 2006]. The geologic map for the Chrétien region (Figure 7a) indicates that the materials surrounding Chrétien crater consist mostly of ancient, highly cratered, Nectarian and pre Nectarian terrain, as well as a significant number of Imbrian aged craters (Figure 7a). The geologic map also reveals the presence of the Imbrian grooved terrain (Ig), also identified in the Ingenii, Von Kármán, and Leibnitz regions. [24] The CSR albedo map for Chrétien crater (Figure 7b) shows that there is a large pond in this area. The northernmost 7of23

8 Figure 7. (a) Digitized geologic map for the basalt ponds in and around Chrétien crater [from Wilhelms et al., 1979]. Units of the same color are mapped as the same geologic unit by Wilhelms et al. [1979]. (b) CSR albedo map for the ponds in and around Chrétien crater showing posteruption modification of the basalt pond. (c) CSR FeO map for the basalt ponds in and around Chrétien crater. The area outlined in black appears to have little or no nonmare cover. (d) Color ratio map for the ponds in and around Chrétien crater shows that the portion of basalt in the center of the image has a yellow/orange color consistent with an exposure of low Ti basalt. portion of the pond exhibits a significant amount of posteruption modification including a relatively fresh crater in the northeastern portion of the pond, several mass wasting debris flows, and a crater ray that cuts diagonally across the pond, all of which obscure the surficial expression of the basaltic pond. The portion of the pond, contained entirely within Chrétien crater, provides the best potential for finding relatively uncontaminated exposure of basalt, but it too has seen posteruption modification (Figure 7b), mostly in the form of cross boundary mixing [e.g., Li and Mustard, 2005], mass wasting, and posteruption cratering. The southern and northernmost parts of the pond are almost entirely obscured by nonmare materials. [25] The CSR FeO map (Figure 7c) shows that the eastern portion of the pond is not a viable candidate for finding uncovered basalt, whereas the northern and cratercentered ponds show some potential (area outlined in black) due to the lack of nonmare cover. The CSR color ratio map (Figure 7d) confirms that the northern portions of the pond have significant nonmare cover. The north central portion of the pond within Chrétien crater has a yellow/orange color distinct from the surrounding terrain suggesting that this portion of the pond is consistent with an Fe rich, low Ti basalt and therefore provides a relatively clear compositional signature All Other Regions of Interest [26] All other regions that we investigated (i.e., Maksutov crater, Bose crater region, Leibnitz crater, Von Kármán crater, Mare Ingenii, Poincaré crater, and Jules Verne crater) 8of23

9 have all been significantly influenced by one or all of the following: posteruption impact craters, mass wasting debris, ejecta from adjacent craters, and/or lateral mixing with nonmare lithologies [e.g., Li and Mustard, 2005]. These effects essentially obscure the basalts and therefore, no clear compositional signature could be obtained any of the basaltic ponds in these regions. Even though the details are not presented here, it is important to reiterate that we did construct forward models for each of these regions and found that we were able to obtain useful information about the nonmare materials in these regions (see section 4.4 for details). 3. Forward Modeling 3.1. Background [27] Previous work [e.g., Hagerty et al., 2006a, 2009; Lawrence et al., 2003, 2005, 2007] has shown that it is possible to use forward modeling and deconvolution of Th data from the LP GRS to estimate the Th abundances of specific features on the lunar surface. In areas that have high Th abundances, and thus high signal to noise ratios, the previous methodology works well. However, in areas with low Th abundances, the signal to noise ratio is much lower, thus increasing the probability of having noise related artifacts in the LP GRS Th maps. The presence of significant noiserelated artifacts can make it difficult to accurately evaluate the validity of a given forward model. Therefore, in an effort to significantly reduce the amount of noise in areas containing low Th abundances, we have employed the Pixon spatial deconvolution method described by Lawrence et al. [2007] The Pixon Method [28] The spatially adaptive image restriction method known as Pixon [Pina and Puetter, 1993; Puetter, 1995; Puetter and Yahil, 1999], is a deconvolution technique that seeks the smoothest possible image as constrained by both the original data and the data uncertainty (or noise). A unique feature of the Pixon method is that it uses variably sized smooth patches, or Pixon elements, within the image in order to express the information content of the image. For example, if the variation within a large portion of the image can be attributed solely to noise, then the size of the Pixon element in this region would be relatively large. In contrast, regions containing statistically significant smallscale spatial variations will have smaller Pixon elements that capture larger amounts of information. In general, a set of Pixon elements for a given image represents the minimum set required to describe the image information content within the limits allowed by the noise [Puetter, 1995]. [29] The use of the Pixon processed maps is advantageous for this study as the Pixon algorithm reduces the statistical uncertainties for regions with low Th concentrations. The regions being investigated in this study have characteristically low Th concentrations and therefore have larger statistical uncertainties than regions with larger Th concentrations. Due to the spatially variable nature of the Pixon algorithm, the spatial response function has not been directly calculated for the Pixon derived Th maps of the Moon. Therefore, in this study we are making the approximation that the spatial footprint used in previous studies [Lawrence et al., 2005; Hagerty et al., 2006a, 2006b] is valid here. Due to the slowly varying nature of these response functions (i.e., they are not step functions), this is likely a good approximation for the purposes of this study. Early analyses of SPA mare basalt ponds using the non Pixon data of Lawrence et al. [2003] showed similar qualitative results as is given here. The validity of this methodology is further enhanced by the fact that for other portions of the Moon, the Pixon processed maps match astonishingly well with and reinforce our forward modeling results [e.g., Lawrence et al., 2007] Forward Modeling Methodology [30] We begin the forward modeling process by delineating specific geologic features and/or large expanses of single lithologic units in the region of interest. Specific features were selected based on information from geologic maps [e.g., Stewart Alexander, 1978; Wilhelms et al., 1979], shaded relief maps [e.g., Rosiek and Aeschliman, 2001], high resolution CSR data (e.g., FeO, TiO 2, color, and albedo) [e.g., Lucey et al., 2000], and neutron and gammaray data from Lunar Prospector (e.g., Th, FeO, TiO 2,Sm Gd, and epithermal neutron data) [Elphic et al., 2002; Lawrence et al., 2003; Prettyman et al., 2006]. However, as was discussed earlier, one of the most useful tools for selecting the various lithologies in SPA basin was the Clementine ratio ( false color ) data. The color ratio image product serves to reduce the brightness variations of the scene and enhances color differences related to soil mineralogy and maturity [Pieters et al., 1994]. These data helped us to delineate the boundaries between volcanic and impactgenerated lithologies. [31] The next step in the forward modeling process is to estimate Th abundances for each of the geologic features in the region of interest. We use LP GRS data from Lawrence et al. [2003] as an initial lower bound on our estimates and Th data from the lunar sample suite [e.g., Jolliff, 1991, 1998; Korotev, 1998; Papike et al., 1998] as an upper bound on our estimates. For example, we do not allow geologic features to have Th values higher than what have been measured in the lunar sample suite. However, it is important to note that we make no a priori assumptions about the composition of any of the features in the forward model. [32] Once we have an estimated Th abundance for each feature, we propagate the expected gamma ray flux from our estimated abundance through the entire instrument response function of the LP GRS. The resultant data are then compared to the Pixon processed data and, if needed, the modeled abundance distribution is iteratively modified until a match is achieved. [33] We recognize that this procedure gives a nonunique result; however, the assigned Th values can be constrained in two ways. First, the Pixon Th map itself provides a check on the modeled Th abundances. For instance, in some circumstances changing a modeled Th value by as little as 1 ppm can cause the modeled abundance distribution to deviate from the Th distribution in the Pixon map, thus invalidating the selection of that Th value. For some portions of the mapped regions, we can obtain quantitative estimates and uncertainties of surface abundances using a chi square (c 2 ) minimization technique [e.g., Press et al., 1988]. More specifically, the optimum Th value for a 9of23

10 Figure 8. CSR TiO 2 map [Lucey et al., 2000] for the SPA basin (outlined in white). From this image only the Apollo ponds, Rumford, Maksutov, Bose, and most of Chrétien show up well. Portions of Leibnitz, Ingenii, and Jules Verne show up but they are not bright. given geologic feature is the one that minimizes the c 2 value as defined by: 2 ð½thšþ ¼ XN i¼0 2 C new ð½thšþ C reported : Here, C new ([Th]) is the Th concentration being modeled and C reported is the reported Th concentration derived from the Pixon derived data. The symbol, n, represents the degrees of freedom in the fit and N represents the total number of pixels in the modeled region. The symbol, s i, is the measured standard deviation of the counting rate in each pixel (i.e., the uncertainty of the Th measurements). See Hagerty et al. [2006a] for additional details. [34] The variable c 2 is a statistic that characterizes the way in which observed frequencies differ from expected frequencies [Bevington and Robinson, 2003]. If the observed and predicted frequencies agree exactly, the c 2 value should be zero [Bevington and Robinson, 2003]. However, in practice the minimum c 2 value is generally equal to or greater than the degrees of freedom (n) in the fit [Bevington and Robinson, 2003]. In our case, n = N 1, since we are optimizing only the ([Th]) parameter. If the value of c 2 is much larger than one, some aspect of the model is not entirely correct. For example, complications can arise if a region in the forward model is too close to the boundaries defined by the model, which in turn can give spurious results when this region is smoothed by the response function. This complication can be minimized by conducting the c 2 minimization on areas far from model boundaries (i.e., by placing the region of interest in the center of the model boundaries). Also note that the edges of the modeled regions were not considered in the error analysis and they have little effect on the features of interest. Additional complications can occur when trying to model a s 2 i compositionally diverse region. We account for this complication by using sample data, geologic maps, and remote sensing data to constrain the composition and areal extent of each lithology in the modeled region. [35] Finally, we can determine the standard deviation associated with the c 2 minimization by increasing the minimum c 2 value by 1 [see Press et al., 1988]. Once we have determined the optimum Th values and the associated errors for a given region, we incorporate those values into a final forward model. We then compare the results from the forward model to the Pixon derived Th map and check for completeness of the modeled Th distribution. 4. Results 4.1. Apollo Basin [36] All of the ponds in Apollo basin have significant exposures of basalt, each of which have titanium contents ranging from 4.5 to 10 wt.% (Figure 8). In sections we describe the results of each forward model for each region of interest Apollo North Pond [37] The Pixon Th map (Figure 9a) shows that the area surrounding the Apollo north basalt pond contains several Th enhancements. For example, there are slight Th enhancements in the north central and southeastern parts of the scene and both enhancements appear to be associated with INp units [e.g., Stewart Alexander, 1978]. Stewart Alexander [1978] refers to the INp unit as light plains material related to Imbrian and/or Nectarian basins. Conversely, the southwestern portion of the pond, which coincides with the clearest basalt exposure, appears to have low Th abundances (Figure 9a). Taking into account all of the available information, we defined 21 different units that surround and include the Apollo north basalt pond (Figure 9b). It is 10 of 23

11 Figure 9. (a) Pixon processed Th map for the Apollo north basalt pond. The large area outlined in white represents the total areal expression of the Apollo north pond, as defined by the CSR FeO and albedo maps. (b) Forward model that was constructed based on information from geologic maps, and CSR and LP data sets. (c) The end result of the forward modeling process. We obtain a close match with the Pixon Th map when the area outlined in white (i.e., the uncontaminated basalt) has 0.00 ppm Th. (d) End result of forward modeling if the uncovered portion of basalt has as little as 1.00 ppm Th. (e) Chi square analysis of the area surrounding the relatively uncontaminated exposure of basalt. This analysis indicates that the exposure of basalt has an optimum Th value of 0.0 ± 0.9 ppm. No negative values were considered in the modeling process. 11 of 23

12 Table 1. Compositional Data for Uncovered Portions of Basaltic Ponds CSR FeO (wt.%) CSR TiO 2 (wt.%) LP GRS Th (ppm) Modeled Th (ppm) Apollo north ± 0.90 Apollo south ± 0.14 Apollo west ± 0.29 Rumford ± 0.22 Chrétien ± 0.27 important to note that adjacent units containing the same Th values (within error) will appear as one large unit. While many of our selected units coincide with features identified in the geologic map (Figure 3a), we used all of the available resources to select several new units for forward modeling analysis (Figure 9b). [38] After iterating on a variety of Th abundance distribution permutations, we find that our best approximation of the Th abundance distribution in the Apollo north region (Figure 9c) occurs when the uncovered portion of the pond has 0.00 ppm Th. We find that if the uncovered portion of the pond has 1.0 ppm or more Th (Figure 9d), we cannot reproduce a match with the Pixon Th map (Figure 9a). Chisquare statistical analysis of the pixels in the center of the scene, indicate that the optimum Th value for the representative exposure of basalt is 0.00 ± 0.9 ppm (Figure 9e). Because we are dealing with very low Th values and with a relatively large error, all that we can definitively say is that the uncontaminated portion of the basalt pond is consistent with having less than 1 ppm Th. [39] To reconstruct the Pixon Th map accurately, we had to incorporate unrealistically high Th values (e.g., >11 ppm) for the Th enhancements at the edges of the scene. The high Th values are an artifact affiliated with the edge effects of the modeling process. As a result of selecting a region of interest that extends beyond the basalt pond, we effectively cut off portions of the surrounding terrain, artificially reducing the size of geologic features that could actually have a much larger areal extent. Nevertheless, the center of the scene (i.e., the basalt pond) is the area in which we are truly interested. The values in the center of the modeled region are much more accurate and reflect the compositions of entire lithologic units and/or geologic features. The subsequent statistical analysis was limited to the pixels in the center of the scene that include both the uncontaminated portion of the basaltic pond and the nonmare cover on top of the pond. [40] Note that Figure 9b shows that there is a relatively high Th value (i.e., 8 ppm Th) within the basalt pond, adjacent to the uncovered portion of basalt. This Th enhancement appears to be associated with overlapping ejecta deposits from the crater northeast of the pond and from several small craters to the southeast of the pond, thus indicating that the overlapping ejecta deposits are responsible for the apparent Th enhancement. The region of Th enhancement coincides with a region of moderate FeO (9 12 wt.%) and TiO 2 (2 6 wt.%) abundances (Table 1). Other areas of nonmare contamination on top of the basalt pond also appear to have between 1 and 3 ppm Th (Figure 9b), which is consistent with the average Th composition of the SPA floor. The combination of the Th, FeO, and TiO 2 data for the nonmare cover are consistent with the composition of noritic to gabbronoritic materials [e.g., Papike et al., 1998] that appear to dominate the floor of SPA basin [Pieters et al., 2001; Isaacson and Pieters, 2009;Nakamura et al., 2009] Apollo West Pond [41] The Pixon Th map for Apollo west (Figure 10a) indicates that the northern portion of the pond is associated with a localized Th low and that the southern portion has slightly higher Th values. Figure 10a also shows that there is a Th enhancement in the northwestern portion of the scene that is associated with the edge of the western rim of Apollo basin (labeled as Nc in Figure 4a). The estimated Th abundance of this possible rim feature will be artificially high due to edge effects associated with the modeling process; however, it is not unreasonable for this feature to have elevated Th abundances (see discussion of lower crust below, section 5.2.2). Taking into account all of the available information, we defined 14 different units that surround and include the Apollo west basalt pond (Figure 10b). While many of our selected regions coincide with regions identified in the geologic map (Figure 4a), we used all of the available resources to select several new regions for forward modeling analysis (Figure 10d). [42] After multiple iterations of the forward model, we find that our best approximation for the Th abundance distribution (Figure 10c) occurs when the northern portion of the pond has 0.00 ppm Th. Adding even a small amount of Th (i.e., 1.0 ppm Th) to the northern part of the pond completely disrupts the map (Figure 10d), thus preventing us from achieving a match with the measured Th map (Figure 10a). The chi square analysis of the region (Figure 10e) confirms that 0.00 ± 0.29 ppm is the optimum Th value for the relatively uncontaminated exposure of basalt. However, because we are dealing with very low Th values and with a relatively large error, all that we can definitively say is that the northern portion of the basalt pond is consistent with having less than 1 ppm Th. [43] The Th abundance distribution for the southern portion of the Apollo west basalt pond is consistent with a Th abundance of 2.7 ppm. Because the southern portion of the pond is almost entirely obscured by nonmare contamination, we infer that the Th enhancement is associated with the nonmare cover, which also appears to have moderate FeO and TiO 2 abundances. The relatively high Th abundance, in conjunction with the LP FeO data, the CSR albedo, and the CSR FeO data, all indicate that the compositional data from the southern portion of the pond are influenced by nonmare cover and that the overlying material has a composition consistent with noritic to gabbronoritic lithologies [e.g., Papike et al., 1998] Apollo South Pond [44] The Pixon Th map (Figure 11a) shows that there is a significant Th low associated with the central portion of the Apollo south pond and slight Th enhancements associated with the eastern and western portions of the pond. The highest Th enhancement on the left side of the scene appears to be associated with material mapped as Ip by Stewart Alexander [1978]. The geologic unit, Ip, is interpreted as smooth light plains material related to Imbrian aged basins. The estimated Th abundances for the western enhancement 12 of 23

13 Figure of 23

14 in Figure 11b are artificially high due to the edge effects of the modeling process. Using the LP GRS Th map, along with all of the other data sets, we were able to define 16 distinct regions, similar to the 16 regions identified in the geologic map (Figure 5a). The 16 regions that we identified comprised our forward model for the region (Figure 11b). [45] After multiple iterations we find that our best approximation for the Th abundance distribution (Figure 11c) occurs when the central portion of the pond has 0.0 ppm Th. Adding even a small amount of Th (i.e., 1.0 ppm Th) to the central part of the pond completely disrupts the abundance distribution (Figure 11d), thus preventing us from achieving a match with the measured abundance map (Figure 11a). Chi square analysis of the region (Figure 11e) shows us that the optimum Th value for our region of representative basalt is 0.02 ± 0.14 ppm. Because we are dealing with very low Th values and with a relatively large error, all that we can definitively say is that the central portion of the basalt pond is consistent with having less than 1 ppm Th. Conversely, the Th distribution for this basalt pond as a whole is consistent with the eastern portion of the pond having as much as 3.0 ppm Th. The relatively high Th abundance in the eastern portion of the pond, in conjunction with the LP FeO data, the CSR albedo, and the CSR FeO data, are all consistent with nonmare material similar in composition to norite or gabbronorite Rumford Crater [46] The Pixon Th map for Rumford crater (Figure 12a) shows that there is a Th enhancement in the northwest portion of the scene; however, this enhancement does not appear to be associated with the basalt pond, which is located on the edge of the Th enhancement. Our forward model for this region (Figure 12b) requires high Th abundances in the northeastern and southeastern portions of the scene in order to explain the slight Th enhancements. However, as was noted earlier, the high Th values are artificially high due to edge effects associated with the forward modeling process. No other compositional or morphologic parameters could be used to argue for the presence of high Th abundances in the northeastern and southeastern portions of the scene. [47] Our best approximation of the Th distribution (Figure 12c) occurs when the basaltic pond has 0.0 ppm Th. We find that if the uncontaminated portion of the pond has more than 1.0 ppm Th, we cannot reproduce a match with the LP GRS Th map (Figure 12d). Chi square analysis of the region (Figure 12e) shows that the optimum Th value for the pond is 0.00 ± 0.22 ppm. Because we are dealing with very low Th values and with a relatively large error, all that we can definitively say is that the uncontaminated portion of the basalt pond is consistent with having less than 1 ppm Th. The Th enhancement northeast of the pond appears to be associated with an older, degraded crater and can be explained by lithologies that have between 4 and 6 ppm Th, as well as 1 2 wt.% TiO 2 and 8 11 wt.% FeO. These compositions are consistent with noritic to gabbronoritic lunar lithologies [e.g., Papike et al., 1998] Chrétien Crater [48] The Pixon Th map for Chrétien crater (Figure 13a) shows an abundance of compositional variation. There are several local Th enhancements and several local Th depletions. The Th enhancement to the southwest of the selected basalt exposure appears to be associated with the INp unit in the geologic map (Figure 7a). The INp unit is supposedly a light plains unit derived from an Imbrian aged or Nectarianaged basin; however, the localized Th anomaly associated with this unit is associated with a specific crater within the region of interest, thus implying a local origin [i.e., Garrick Bethell and Zuber, 2005]. The large Th enhancement in the northeastern part of the scene is affiliated with the regional Th enhancement in the northwestern portion of SPA basin (Figure 1). The large regional Th enhancement may have a significant influence on the Th abundance distribution in the region we selected, thus making it difficult to model effectively. Nevertheless, we collected all of the available data and produced a forward model for the region (Figure 13b). We find that our best approximation (Figure 13c) occurs when the selected exposure of basalt has a value of 0.0 ppm Th. Adding Th from the selected area completely disrupts the map (Figures 13d), thus preventing us from achieving a match with the Pixon Th map. Chi square analysis of the region (Figure 13e) shows that the optimum Th value for the pond is 0.08 ± 0.27 ppm Other Regions of Interest [49] As was discussed in section 2, there are several basaltic ponds within SPA basin that are either too small to model or do not contain large expanses of uncovered basalt. Nevertheless, forward modeling of these other ponds has provided some interesting results. For instance, the basalt ponds within Bose, Ingenii, Leibnitz, and Von Kármán craters all have large areas that have been covered by nonmare materials. The nonmare materials covering the ponds are likely ejecta deposits related to crater formation in the adjacent terrain. The compositions of the nonmare materials are strikingly similar (see Table 2) even though their loca- Figure 10. (a) Pixon processed Th map for the Apollo west basalt pond. The area outlined in black represents the total areal expression of the Apollo west pond, as defined by the CSR FeO and albedo maps. The uncontaminated exposure of basalt appears to be spatially correlated with a local Th low; whereas the covered portion of the pond coincides with slightly higher Th values. (b) Forward model that was constructed based on information from geologic maps, and CSR and LP data sets. (c) The end result of the forward modeling process. We obtain a close match with the Pixon Th map when the northern portion of the area outlined in black (i.e., the uncontaminated basalt) has 0.00 ppm Th. (d) End result of forward modeling if the uncovered portion of basalt has as little as 1.00 ppm Th. (e) Chi square analysis of the area surrounding the relatively uncontaminated exposure of basalt. This analysis indicates that the exposure of basalt has an optimum Th value of 0.0 ± 0.29 ppm. No negative values were considered in the modeling process. 14 of 23

15 Figure 11. (a) Pixon processed Th map for the Apollo south basalt pond. The areas outlined in white represents the total areal expression of the Apollo west pond, as defined by the CSR FeO and albedo maps. The uncontaminated exposure of basalt (central portion) appears to be spatially correlated with a local Th low; whereas the covered portions of the pond coincide with slightly higher Th values. (b) Forward model that was constructed based on information from geologic maps, and CSR and LP data sets. (c) The end result of the forward modeling process. We obtain a close match with the Pixon Th map when the central portion of the area outlined in white (i.e., the uncontaminated basalt) has 0.0 ppm Th. (d) End result of forward modeling if the uncovered portion of basalt has as little as 1.00 ppm Th. (e) Chi square analysis of the area surrounding the relatively uncontaminated exposure of basalt. This analysis indicates that the exposure of basalt has an optimum Th value of 0.02 ± 0.14 ppm. No negative values were considered in the modeling process. 15 of 23

16 Figure 12. (a) Pixon processed Th map for the basalt pond in Rumford crater. The area outlined in black represents the total areal expression of the Maksutov pond, as defined by the CSR FeO and albedo maps. (b) Forward model that was constructed based on information from geologic maps, and CSR and LP data sets. (c) The end result of the forward modeling process. We obtain a close match with the Pixon Th map when the area outlined in black (i.e., the uncontaminated basalt) has 0.0 ppm Th. (d) End result of forward modeling if the uncovered portion of basalt has as little as 1.0 ppm Th. (e) Chi square analysis of the area surrounding the relatively uncontaminated exposure of basalt. This analysis indicates that the exposure of basalt has an optimum Th value of 0.00 ± 0.22 ppm. No negative values were considered in the modeling process. 16 of 23

17 Figure of 23

18 Table 2. Compositional Data for Nonmare Materials Covering Basaltic Ponds CSR FeO (wt.%) CSR TiO 2 (wt.%) tions are widely separated within SPA basin. The nonmare compositions listed in Table 2 are consistent with noritic to gabbronoritic lunar lithologies [e.g., Papike et al., 1998], which in turn are consistent with mineralogic evidence that the floor of SPA basin is dominated by noritic lithologies [e.g., Pieters et al., 2001; Isaacson and Pieters, 2009; Nakamura et al., 2009]. 5. Discussion and Conclusions LP GRS Th (ppm) Modeled Th (ppm) Bose Ingenii Leibnitz Von Kármán Average norite [Papike et al., 1998] Basalt Ponds Thorium Content [50] Our results indicate that the basalt ponds in SPA basin contain little or no thorium. If we use the basalt ponds as probes into the far side lunar mantle, we can infer that as of the Imbrian Period, there was little or no Th in the mantle under SPA basin. However, there is at least one major caveat to consider. What if the mantle under SPA is heterogeneous? Is it possible that our limited sample set (i.e., five basalt ponds), does not provide an accurate representation of the entire region? In other words, could there be Th rich regions of the mantle that were simply not sampled by the SPA basalt ponds? While this is a plausible scenario, there are two other ways to address the question of mantle heterogeneity. First, the fact that the basalt ponds are widely separated (i.e., hundreds of kilometers), have different major element compositions, have different (relative) ages, and are likely derived from different depths, indicates that the ponds represent significantly different portions of the underlying mantle [e.g., Shearer et al., 2006]. Second, if there were regions of the underlying mantle that had significant abundances of energy producing elements, we would expect to see multiple examples of expansive, Th rich basalt, similar to the Procellarum KREEP Terrane (PKT) on the nearside of the Moon [i.e., Jolliff et al., 2000]. The relative lack of basaltic magmatism compared to the nearside of the Moon, in conjunction with the miniscule Th content of the SPA basalt ponds, leads us to conclude that the mantle under SPA basin contained very little Th as of the Imbrian, which in turn means that the decay of energy producing elements was not a thermal driver for basaltic volcanism in SPA. The only other option for initiating basaltic magmatism is decompressional melting, which is supported by the fact that all of the ponds are colocated within craters and basins. It is important to note that a direct temporal relationship between impact cratering and eruption of basaltic magma is not explicitly required. In fact, smallscale melting induced by impact events can extend for a period of 350 million years [e.g., Elkins Tanton et al., 2004] Ti Bearing Basalts [51] The observation that the uncovered portions of basalt ponds in SPA basin have TiO 2 abundances as high as 9 10 wt.%, implies that a Ti component was involved in the petrogenesis of these basalts [e.g., Shearer et al., 2006]. Petrogenetic models for the Ti bearing basalts on the nearside of the Moon necessarily require that a significant ilmenite component be present in the source region [Papike et al., 1998; Shearer et al., 2006]. If the Snyder et al. [1992] model for lunar magma ocean (LMO) crystallization is correct in stating that significant amounts of ilmenite crystallize from the LMO between 95 and 99% fractional crystallization, we can infer that the far side mantle must have seen similar degrees of fractional crystallization in order to produce the ilmenite that is required for the source regions of the Ti rich basalts. The presence of high Ti basalts, in conjunction with the large regional enrichments of Th inherent to SPA basin, is strong evidence that the far side of the Moon did see extensive fractional crystallization. However, the observation that the Imbrian aged basalts are Th poor indicates that, as of the Imbrian, there was very little Th in the mantle underlying SPA basin. If that was indeed the case, a heating mechanism, other than decay of radioactive elements, would be required to generate the melts that formed the basaltic ponds (i.e., impactinduced decompressional melting) Implications for the Far Side Mantle [52] There are least four different models for basalt petrogenesis that could explain our observations. In the following paragraphs we will briefly discuss each model, list any relevant pros or cons, and discuss the potential implications for the strongest models. [53] 1. The first model to consider is that the far side mantle experienced extensive fractional crystallization and remained stratified whereby Th rich materials were sandwiched between the lower crust and the upper mantle [e.g., Snyder et al., 1992, 1995]. In this model no Th would have existed at depth in the mantle making it possible to generate basaltic ponds from Th poor source regions. However, a nonradiogenic source of energy is required to produce basaltic magmas and any basaltic magmas that were generated would need to avoid assimilating any of the Th rich lithologies that would necessarily be concentrated at the Figure 13. (a) Pixon processed Th map for the basalt pond in Chrétien crater. The area outlined in black represents the area we selected as a representative exposure of basalt. (b) Forward model that was constructed based on information from geologic maps, and CSR and LP data sets. (c) The end result of the forward modeling process. We obtain a close match with the Pixon Th map when the area outlined in black (i.e., the uncontaminated basalt) has 0.0 ppm Th. (d) End result of forward modeling if the uncovered portion of basalt has as little as 1.0 ppm Th. (e) Chi square analysis of the area surrounding the relatively uncontaminated exposure of basalt. This analysis indicates that the exposure of basalt has an optimum Th value of 0.00 ± 0.14 ppm. No negative values were considered in the modeling process. 18 of 23

19 base of the lunar crust. One positive aspect of this model is that a Th rich, KREEP like material would exist at the crust mantle boundary and could explain the presence of Th rich materials exposed by the excavation of SPA basin. [54] 2. Perhaps the far side mantle is heterogeneous with respect to Th and the basalts investigated in this study simply did not sample any Th rich portions of the mantle. In this model, the far side mantle would have evolved in a manner similar to what has been described for the nearside mantle (i.e., extensive crystallization, density driven convective overturn, and compositional hybridization of the mantle). However, as discussed above, our results indicate that this model is unlikely (i.e., multiple, widely separated ponds of different ages should exhibit compositional diversity if it existed). [55] 3. Perhaps no significant source of Th ever existed on the far side of the Moon. In this model the presence of Th within SPA would necessarily be exogenous and the only way to explain exogenous Th would be Th rich antipodal ejecta from a nearside basin forming event. As noted previously, several recent studies have shown that the Th in SPA is not from antipodal ejecta and is indeed inherent to the basin itself. [56] 4. A Th rich, KREEP like layer once existed on the far side of the Moon but was mobilized prior to the formation of the source regions for the basalt ponds [e.g., Arkani Hamed and Pentecost, 2001]. At first glance this model may seem unrealistic; however, there are two pieces of evidence that add credence to this model. The presence of Ti rich basalts, in conjunction with the elevated Th abundances inherent to the SPA basin, indicate that the far side mantle underwent extensive fractional crystallization likely resulting in the production of a KREEP like component (i.e., there was a global layer of urkreep) Nonmare Mafic Lithologies Distribution [57] While our focus was the basalt ponds, the LP GRS Th data do indicate that Th rich nonmare lithologies exist within SPA basin. Widespread abundances of noritic to gabbronoritic material thought to be inherent to SPA basin [e.g., Pieters et al., 2001; Isaacson and Pieters, 2009; Nakamura et al., 2009], the presence of indigenous regional Th enhancements (Figure 1), and the direct correlation between local Th enhancements and nonmare mafic lithologies identified in this region [e.g., Blewett et al., 2000; Haskin et al., 2004; Garrick Bethell and Zuber, 2005], lead us to conclude that norites and/or gabbronorites are the source of the regional Th enhancements within the SPA basin. Such an assertion is supported by the Th data from norites/gabbronorites in the lunar sample suite (see Table 2), which indicate that norites and gabbronorites have a range of Th abundances with a mean value of 2 3 ppm Th, similar to the average Th composition of the SPA basin floor (Figure 1). [58] Detailed studies of the NW thorium anomaly within SPA basin [Haskin et al., 2004; Garrick Bethell and Zuber, 2005] not only showed that the anomaly was not the result of antipodal ejecta from Imbrium [i.e., Wieczorek and Zuber, 2001], but also lend support to our assertion that the source of Th was in place prior to the formation of SPA basin. For instance, Garrick Bethell and Zuber [2005] used compositional and topographic data to show that the NW Th anomaly is a province of elevated Th exposed by the SPA impact event. They proposed that subsequent periods of meteoroid bombardment likely buried the province only to be exposed by younger, Eratosthenian craters, thus producing the crater centered lobes of the anomaly [Garrick Bethell and Zuber, 2005]. It is possible that the regional Th enhancements identified by Lawrence et al. [2003] and discussed by Garrick Bethell and Zuber [2005] represent excavated plutons of Mg suite lithologies. Given this information, it is logical to assume that there must have been a source of Th on the far side of the Moon at some point between the initial formation of the lunar crust and the formation of the SPA basin Implications for the Lower Crust [59] Because norites and gabbronorites are the intrusive equivalents of basalts, we know that they must have formed as a result of ancient intrusions of basaltic magma into preexisting ferroan anorthositic crust [e.g., Shearer et al., 2006]. The observation that the norites in SPA basin are the source of the regional Th enhancement in the basin has important implications for the formation of the lower crust on the far side of the Moon. More specifically, Th rich norites must have crystallized from Th rich basaltic parent melts [e.g., Snyder et al., 1995]. How the Th was incorporated into the basaltic melt is an open ended question. However, the fact that the intruded basaltic parent magmas must have had elevated Th abundances, relative to the surrounding terrain, requires that the basaltic magmas interacted with a Th rich component on the far side of the Moon [e.g., Neal et al., 1988; Snyder et al., 1995; Shearer et al., 2006]. [60] Recent deconvolution of the LP GRS Th data [Lawrence et al., 2007] has shown that SPA has a regional Th high of approximately 6 ppm. Given billions of years of regolith processing and the observation that the floor of SPA has an average of 2 3 ppm Th, it is logical to assume that the initial sources of Th contained at least 6 ppm Th, with higher abundances likely. In fact, forward modeling of numerous regions in this study indicate that individual lithologic units can have as much as 8 10 ppm Th. A regional Th value of 6 ppm is lower than Th values associated with the Procellarum KREEP Terrane (PKT) [Jolliff et al., 2000; Haskin et al., 2004] on the nearside of the Moon; however, the only way to obtain large, regional Th enhancement of at least 6 ppm Th is through large scale magmatic fractionation, similar to what has been proposed for the nearside of the Moon. The method of incorporating Th into the norites is debatable, but it is clear that a source of Th did exist in this part of the Moon at the time of lower crust formation Revised Model of Lunar Evolution [61] Given the information presented above, there are several critical observations that must be explained in any model for lunar evolution: (1) A Th rich noritic to gabbronoritic crust must be produced on the far side of the Moon prior to the SPA impact event. (2) Regional Th abundances of at least 6 ppm and local abundances of 8 10 ppm Th must be explained. (3) Ti rich, Th poor 19 of 23

20 Figure 14. Hypothetical model for the formation of the far side lunar crust. (top) The early anorthosite crust has been intruded by numerous basaltic magmas that likely passed through a superheated, urkreep layer. In some instances, the intrusions and their affiliated latent heats of crystallization may have led to partial melting of the existing crust [e.g., Hagerty et al., 2006a, 2006b]. If it is assumed that the crust preceding the SPA impact event was as thick as the current far side crust is, it is likely that few if any basaltic magmas would have made it to the lunar surface. (bottom) The aftermath of the SPA impact event, which excavated and melted a significant portion of the crust, instigated the transport of urkreep to the lunar nearside, and initiated sinking of ilmenite pods on the far side. The subsequent formation of large craters such as Apollo may have led to decompressional melting in the underlying mantle and eruption of Ti bearing, Th poor basalts. basaltic source regions must be produced in the far side mantle. (4) Ti rich basalts must be generated by 3.9 Ga without significant energy input from the decay of radioactive elements. [62] Based on the results of our study and the potential explanations for our observations, we have arrived at a comprehensive model that explains the distribution of thorium on the Moon (Figures 14a and 14b). Similar to other models for lunar evolution, we suggest that the entire Moon experienced extensive fractional crystallization resulting in the early formation of a primordial ferroan anorthosite crust, an initially stratified lunar mantle, and a urkreep layer being sandwiched between the crust and mantle [e.g., Warren and Wasson, 1979; Snyder et al., 1992]. The initial stage of primordial crust formation was followed by an early stage of basaltic volcanism, some of which reached the lunar surface, while some of the basaltic magma was intruded into the early lunar crust [e.g., Shearer et al., 2006]. Studies of ancient aluminous lunar basalts show that the lunar mantle produced basaltic magmas that were emplaced in the lunar crust as early as 4.45 Ga [Shearer et al., 2006]. In our revised model, early basaltic magmas would have assimilated a still molten (and/or superheated) urkreep component during their ascent [e.g., Neal et al., 1988; Snyder et al., 1995; Neal and Kramer, 2006], which is certainly plausible given the results of thermal modeling [e.g., Wieczorek and Phillips, 2000] which show that the urkreep layer could have remained molten for a billion years. Assimilation of a molten, KREEP like component appears to be more likely than derivation of Th rich lithologies from melting of a Th rich hybridized mantle given that the younger, Imbrianaged basalt ponds show no evidence of Th at depth in the far side mantle. The intruded Th rich basaltic magmas would have crystallized in situ, resulting in the formation of Th rich norites and/or gabbronorites [e.g., Snyder et al., 1995; McCallum and Schwartz, 2001; Shearer et al., 2006]. [63] During, or shortly after the formation of the noritic lower crust, the SPA impact event would have occurred, excavating and/or melting the Th rich noritic lower crust and mobilizing the molten urkreep component, concentrating the urkreep on the nearside of the Moon [e.g., Arkani Hamed and Pentecost, 2001]. This model, initially proposed by Arkani Hamed and Pentecost [2001], suggests that energy from the SPA impact event acted as a thermal and physical perturbation to the early Moon, effectively fluidizing the upper portions of the lunar mantle. Fluid dynamics equations and modified viscosity models for planetary interiors were used to suggest that mantle rebound after the impact event initiated vigorous mantle circulation. 20 of 23