A new lunar rock type, rich in (Mg, Fe)Al spinel and lacking abundant olivine and pyroxene, was recently detected by near-infrared reflectance spectroscopy by the M3 instrument on the Chandrayaan-1 spacecraft. No such material has been described from lunar rocks, either returned samples or meteorites. Here we describe a fragment of rock containing ∼30% (Mg, Fe)Al spinel from the lunar meteorite ALHA 81005. Although the fragment is not identical to the material detected by M3 (it contains ∼20% olivine + pyroxene), both share the defining feature of an unusual enrichment in spinel. The fragment, 350 × 150 μm, is so fine grained that it reasonably could represent a larger rock body; it is not spinel-rich merely by chance incorporation of a few spinel grains. The fragment is so rich in spinel (and consequently in Al2O3) that it could not have formed by melting a peridotitic mantle or a basaltic lunar crust. The clast's small grain size and its apparent disequilibrium between spinel and pyroxene suggest fairly rapid crystallization at low pressure. It could have formed as a spinel cumulate from an impact melt of troctolitic composition or from a picritic magma body that assimilated crustal anorthosite on its margins. The latter mechanism is preferred because it not only explains the petrographic and chemical features of our clast but is also consistent with the regional setting of the Moscoviense spinel deposit. In that area, M3 spectra have defined areas rich in olivine and in orthopyroxene; these could represent igneous cumulate rocks formed during crystallization and differentiation of a picritic magma body and thus suggests a possible link between the analyzed clast and the observed spinels at Moscoviense.
 The lunar crust contains some of the most important and accessible clues to the history of the Moon's chemical evolution [e.g., Taylor, 1982; Demidova et al., 2007; Isaacson et al., 2011]. The discovery of spinel-rich deposits on the lunar surface measured by the M3 near-infrared (NIR) imaging spectrometer on the Chandrayaan-1 spacecraft [Pieters et al., 2009, 2010, 2011; Lal et al., 2011; Dhingra et al., 2011] is one of the more intriguing discoveries raising new questions about igneous processes in the lunar crust. The M3 spectrometer identified several areas on the rim of the farside Moscoviense basin that are rich in (Mg, Fe)Al2O4 spinel, and contain less than 5% mafic silicate minerals (olivine and pyroxene). Pieters et al.  suggested that these deposits represent a previously unknown lunar rock type, a spinel anorthosite, which might be an important component of the lunar crust. However, the origin and mechanism(s) of formation of such spinel-rich rocks are not clear. To form a lunar rock so rich in (Mg, Fe)Al spinel would require extensive magmatic fractionation that could potentially have occurred in a post-magma-ocean pluton [e.g., Prinz et al., 1973; Marvin et al., 1989]. Spinel-bearing lunar rocks are relatively rare and to date, no spinel-rich material has been described, either in returned samples or in recovered meteorites. In the absence of actual spinel-rich samples from the Moon, however, verification of the validity of any formation hypothesis of origin will remain speculative.
 Here we describe a spinel-rich rock fragment in the lunar highlands meteorite ALHA 81005 (Figure 1), the first spinel-rich lunar such sample to be described. The fragment contains ∼30 vol % (Mg, Fe)Al2O4 spinel, enough that an outcrop scale of it would be easily detected by an instrument like the M3 imaging spectrometer. The spinel-rich fragment here is not identical to the spinel-rich rock recognized by M3 as it contains ∼20 vol % mafic silicate minerals. However, this rock fragment is the most spinel-rich material reported from the Moon and thus it will provide valuable insights into the petrogenesis of spinel-rich lunar material. It will also enlarge the data set of “ground truth” objects for calibration and quantitative analysis of reflectance spectra for the spinel-rich materials observed by M3, and will provide a crucial link between the local perspective of individual lunar samples and fragments and the regional perspective provided by orbital remote sensing.
2. Sample and Methods
 Antarctic meteorite Allan Hills (ALH) A81005, found in 1982, is a polymict, anorthositic regolith breccia from the lunar highlands [Treiman and Drake, 1983; Korotev et al., 1983; Warren et al., 1983; Goodrich et al., 1984], possibly from the lunar farside [Korotev, 2005]. ALHA 81005 is composed of rock and mineral fragments (including anorthosites, granulites, isolated mineral fragments, impactites, impact glasses and mare basalts) in a glassy, agglutinitic matrix [Gross and Treiman, 2010]. Thin section ALHA 81005,9 contains a single fragment of rock rich in (Mg, Fe)Al spinel (Figure 1), and no similar fragments have been reported in other samples.
 Backscattered electron (BSE) images, elemental X-ray maps, and quantitative chemical analyses were obtained with the Cameca SX100 electron microprobe (EMP) at NASA Johnson Space Center (JSC). Quantitative analyses were obtained by wavelength dispersive spectrometry. Operating conditions were: 15 kV accelerating voltage, 20 nA beam current, focused electron beam (<1 μm) for analyses of olivine, pyroxene and spinel, and defocused beam (5 μm) for analyses of plagioclase. Peak and background counting times were 20–40 s per element. Analytical standards were well-characterized synthetic oxides and minerals including diopside (Si, Ca, Mg), oligoclase (Na, Al), hematite (Fe), rutile (Ti), cobalt metal (Co), chromite (Cr), NiO (Ni), rhodochrosite (Mn) and orthoclase (K). Data quality was ensured by analyzing the standards as unknowns.
 Mineral proportions were determined from BSE and element X-ray maps, using methods similar to those of Maloy and Treiman . Uncertainties in mineral proportions are proportional to the square root of the number of grains in the clast. A rough count of grains gives ∼150 spinels, ∼75 olivines and plagioclases, and three pyroxenes, implying uncertainties of ∼8%, 12%, 12%, and 60%, respectively. Thus, volume proportions of minerals in the clast are: 49 ± 6% plagioclase, 30 ± 3% spinel, 15 ± 2% olivine, and 6 ± 4% pyroxene.
3. Bulk Clast Composition
 To derive a bulk chemical composition for the clast (Table 1), it was assumed that the analyzed area proportions are equal to volume proportions; volume proportions of minerals were converted to mass proportions using mineral densities interpolated from literature values [Robie and Hemingway, 1995; Smyth and McCormick, 1995; Mellin et al., 2008; Treiman et al., 2010]. The bulk composition was then calculated as the sum of average mineral compositions weighted by their mass proportions. Because the clast contains so much spinel, its bulk composition shows 35% Al2O3 and only 29% SiO2 (Table 1), and is clearly outside the range of normal basaltic rocks (its CIPW norm [e.g., Morse, 1994] has 41% plagioclase, 38% olivine, and 20% corundum).
Table 1. Representative Mineral Analyses and Calculated Bulk Clast Composition
Spinel (wt %)
Olivine (wt %)
Plagioclase (wt %)
Pyroxene (wt %)
Bulk Clast (wt %)
4. Petrography and Mineral Chemistry
 The spinel-rich rock fragment (called clast 2) in this study is readily recognized by its lilac-colored spinels in plane-polarized light. The clast is a spinel-anorthositic-troctolite, 350 × 150 μm in size, with a holocrystalline, intergranular texture (Figure 1). Mineral grain sizes are fairly uniform, ∼30–50 μm in “diameter.”
 Plagioclase occurs as subhedral to anhedral grains up to 90 μm in length with a relatively constant composition of An94–97Or0–0.1 typical to lunar anorthosite compositions. Plagioclase grains are in contact (at one place or another) with all other minerals in the clast.
 Spinel is pale purple in plane-polarized light (Figure 1a) and occurs as euhedral to subhedral equant grains with a grain size ranging from 10 to 50 μm (Figure 1). In some areas spinel grains are clustered together. The spinel has a constant composition with Fe# [molar Fe/(Fe + Mg)] ∼0.35; Cr2O3/(Cr2O3 + Al2O3) = 0.04, and TiO2/(TiO2 + Al2O3) < 0.005 (Table 1). Spinel grains are in contact with olivine and plagioclase but not pyroxene.
 Olivine occurs as equant crystals of 10–60 μm in size (see Figures 1 and 2). Some olivine show resorption features (Figure 2a) and have a composition of Fo74±0.4, others seem to have crystallized interstitially or subhedral. These grains have a slightly higher forsterite component of Fo76±0.8. All grains are in contact with plagioclase and spinel, and partially or completely surrounding the pyroxene grains.
 Pyroxene is present as three anhedral grains in the clast, one of 110 × 50 μm and the others ∼30 μm across. The largest grain is partially surrounded by a thin 2–3 μm wide rim of olivine between it and the adjacent plagioclase (Figure 2). One other pyroxene is surrounded by a rind of olivine 30 μm thick. The pyroxenes have complex internal structures. Their cores are finely exsolved, with flecks of high-Ca pyroxene in a low-Ca pyroxene host (Figure 2c). These regions of exsolved pyroxene are surrounded by rims and bands of pyroxene with nearly no Ca (black in Figure 2c). The Mg#s (molar Mg/[Mg + Fe]) of pyroxene, pure low-Ca pyroxene and mixed spots, are constant at ∼79. The average pyroxene composition is ∼Wo12En70Fs18, suggesting that the original pyroxene was pigeonite. Pyroxene is not in contact with spinel.
 Aluminous spinel is uncommon among lunar samples and meteorites, and is found most commonly in plutonic and cataclastic rocks of troctolitic composition (dominated by olivine + plagioclase). These spinels are important as possible indicators of high pressure or unusual igneous fractionations. Previous research has made inferences about the petrogenesis of lunar spinel-bearing rock types on the basis of their petrography and mineral chemistry [Prinz et al., 1973; Marvin et al., 1989]. Spinel-bearing rock fragments in Apollo samples retain excellent textural evidence of plutonic igneous crystallization with spinel as a cumulate phase [Prinz et al., 1973]. Such rock fragments suggested that the Moon contains spinel-bearing igneous cumulates near or below its crust-mantle boundary [e.g., Prinz et al., 1973; Bence et al., 1974; Dymek et al., 1976; Marvin et al., 1989; Snyder et al., 1998]. Thermobarometric calculations on the mineral assemblages in these spinel-bearing rock fragments are consistent with an origin in the lower crust or upper mantle [Herzberg, 1978; Herzberg and Baker, 1980; Baker and Herzberg, 1980].
 The recent observation of lunar regions rich in (Mg,Fe)Al spinel [Pieters et al., 2010, 2011; Lal et al., 2011] raises the importance of these spinels from local curiosities to regional significance. Thus, it seems important to account for the petrogenesis of spinel-rich and spinel-bearing rocks in the early crust [Longhi and Boudreau, 1979]. However, most of these rocks contain little spinel, at most 5–6 vol % (Table 2); the only sample with more spinel is troctolite fragment in 67435 with ∼13 vol % spinel [Prinz et al., 1973; Warner et al., 1976; Ma et al., 1981]. The unique spinel-anorthositic troctolite described here (Figure 1a) is the first spinel-rich lithic sample from the Moon [Gross et al., 2011], and thus provides an important opportunity to explore the origins and potential formation histories of spinel-rich rocks.
6. Is the Spinel-Rich Fragment Representative of a Larger Rock Body?
 Clast 2 is small, only 350 × 150 μm, and thus brings up the obvious concern of whether such a small object could be representative of a larger rock mass, such as have been detected by the M3 spectrometer. It seems reasonable that clast 2 might be representative of a larger body, because it is fine grained, and contains many grains of each species (except pyroxene). In this way, it is unlike the spinel-troctolites of Prinz et al.  and Marvin et al. , which contain only a few grains of each species and might reasonably be unrepresentative fragments of a larger rock body. Of course, there is no way to know if clast 2 is representative of anything larger than itself. However, we will proceed on the assumption that it is.
7. Summary of Critical Petrographic Observations
 These petrographic and mineralogical data suggest five critical inferences about the clast that must be satisfied by any hypothesis of its origin.
 1. The euhedral shapes of spinel and plagioclase grains suggest crystallization from a melt. Rapid crystallization is suggested by the minerals' small grain sizes (30–50 μm); a slow cooling regime would tend to produce larger grains but fewer grains of each species.
 2. The chemical homogeneity of the fragment's minerals, and their consistency with Fe-Mg equilibria among mafic minerals suggests significant duration of subsolidus cooling and equilibration, consistent with the thermal regimes of large igneous bodies and of some impact ejecta and melt sheets [Hudgins et al., 2008].
 3. The apparent disequilibrium (lack of physical contact; see Figure 2b) between spinel and pyroxene suggest that crystallization was not deep in the crust or mantle (Figure 3); (Mg, Fe)Al spinel and pyroxene cannot be in equilibrium at pressures below 3–5 kbar (∼60–100 km) [see Herzberg and Baker, 1980; Morse, 1994].
 4. The composition of olivine grains fall into two groups, correlated with their shapes. Subhedral olivine grains, occurring interstitally between spinel and plagioclase, have a composition of Fo76. The irregular, embayed olivine grains that show resorbtion features (near the largest pyroxene; see Figure 2) are slightly more ferroan at Fo74.
 5. The pyroxene grains are surrounded by olivine, partially or completely. This texture is not expected in igneous crystallization, at either high or low pressure. At low pressure, crystallization along the olivine-pyroxene peritectic curve should produce low-Ca pyroxene rims around olivine (Figure 3, left) as is commonly seen in basaltic rocks owing to the expected crystallization sequence. At high pressure, olivine and low-Ca pyroxene crystallize together, and should not produce olivine rims on pyroxene.
7.1. Generation of Bulk Composition
 In basaltic and peridotitic systems like the Moon, spinel-bearing rocks (i.e., Al2O3-rich compositions) cannot represent primary liquids, because basaltic and corundum-normative (including spinel forming) melts are separated by a thermal divide [e.g., Walker et al., 1973a; Longhi, 1978; Soulard et al., 1994]. To form a lunar rock as rich in (Mg, Fe)Al spinel would require extensive magmatic fractionation, as might have occurred in a post-magma-ocean pluton [Prinz et al., 1973; Marvin et al., 1989]. In the absence of spinel-rich samples from the Moon, however, these ideas have been highly speculative. In the following paragraphs we will discuss different origins, bulk compositions, precursor materials and formation histories that can lead to potential spinel-bearing and spinel-rich samples: (1) exogenic origin and (2) endogenic origin, including: peridotite melts, troctolitic melt, and assimilation processes.
7.2. Exogenic Origin
 It is possible that the spinel-rich areas on the Moon represent meteoritic (exogenic) impactors, because (Mg, Fe)Al spinel is abundant in some chondritic materials, like calcium-aluminum inclusions (CAIs) and Al-rich chondrules [Brearley and Jones, 1998], and some achondrites like angrites [Mittlefehldt et al., 1998]. Thus, the spinel-rich areas measured by M3 could be explained as asteroidal fragments, possibly from the disruption of a rubble pile asteroid [Pieters et al., 2011].
 However, this scenario does not seem reasonable because the spinel-rich areas are not associated with any obvious impact crater and appear to have remained in place undisturbed since the end of the modification stage following the Moscoviense basin forming event. It would require special circumstances to allow the impactor material to be deposited in uniform outcrop-scale areas [see Sunshine et al., 2008; Pieters et al., 2011]. So an exogenic origin is thus not favored [Pieters et al., 2011].
7.3. Endogenic Origin
 On Earth, spinel is found in metamorphosed pelitic rocks, where clay-rich (and therefore Al-rich) protoliths were metamorphosed at high temperature. However, this process is unlikely for the Moon since it never had a hydrosphere and could not have generated clay-rich rocks [Heiken et al., 1991]
 There are three hypotheses for the endogenic origin of lunar spinel-bearing lithologies, within the context of a subchondritic bulk Moon and a crust formed by basaltic igneous processes (e.g., a lunar magma ocean and basaltic plutonism). First, lunar spinel-bearing rocks could have formed as cumulates or restites from basaltic systems at high pressures [e.g., Prinz et al., 1973; Marvin et al., 1989], similar to spinel-peridotites in Earth's mantle formed at pressures between 1 and 2–3 GPa). Basaltic magmas at such high pressures can produce phenocrysts of spinel, which are rarely found at Earth's surface. Second, lunar spinel-bearing rocks could have formed from troctolitic melts at relatively low pressures [Walker et al., 1973a; Marvin and Walker, 1985]. And third, spinel-rich rocks could form by the assimilation of anorthosite into olivine saturated (picritic) melts to form troctolitic melt compositions [Morgan et al., 2006]. The next paragraphs will discuss these scenarios and compare them to the critical petrographic observations from our spinel-rich clast.
7.3.1. Peridotite Precursor
 Melting and crystallizing a peridotite precursor can be explained, and its phase equilibria illustrated, by projections of liquidus equilibria in the system olivine-plagioclase-silica (Figure 3) [Walker et al., 1972, 1973a]. The crucial feature of those diagrams is the spinel liquidus field near the olivine-plagioclase join. A composition for hypothesis 1 representing a “peridotitic” bulk Moon (“M” in Figure 3) [e.g., Longhi, 1978], illustrates that melting a mantle-like lunar composition at low pressure would not yield spinel, either in partial melting or in crystallization. A melt with such a composition (M) will first crystallize olivine during cooling and the melt will evolve toward the olivine + pyroxene peritectic, at which pyroxene starts to crystallize. In an olivine-rich system like a peridotite or picrite, the melt will progress toward the olivine-pyroxene-plagioclase peritectic and will not encounter the spinel liquidus field. Thus rocks that crystallized at low pressure from a melt composition similar to M cannot produce spinel.
 At high pressures, these phase relations change. The olivine-pyroxene peritectic moves toward olivine-rich compositions (Figure 3, right), and changes to cotectic. The size of the spinel liquidus field also increases significantly. The combination of these changes means that spinel + pyroxene becomes stable on the liquidus, and olivine + plagioclase becomes unstable. Thus, high pressure cooling and crystallization of composition M will (as before) produce olivine first and evolve to the olivine-pyroxene cotectic. From here, the melt will move along the cotectic to the olivine-pyroxene-spinel eutectic. Thus a rock that crystallized at high pressure (>5 kbar) [Morse, 1994] from a melt composition similar to M can yield spinel.
 An interesting consequence of these phase relations provides an explanation of observation #5, olivine surrounding pyroxene. If the melt has a lower normative olivine content than composition M, pyroxene could be the liquidus phase at high pressures. A magma with such a bulk composition could be in the olivine field after a sudden drop in pressure [Goodrich et al., 1985]. Pyroxene that had already crystallized would become unstable and partly or wholly resorbed and olivine could crystallize as rims around pyroxene, using the pyroxene grains as nucleation grains. However, a melt with lower normative olivine will not yield spinel, the most important mineral in our clast, even with a sudden drop in pressure. Thus a scenario like this seems unlikely in the origin of the spinel-rich clast.
7.3.2. Troctolitic Precursor
 Lunar spinel-bearing rocks could have formed from troctolitic melts at relatively low pressures [Walker et al., 1973a; Marvin and Walker, 1985]. Considering a melt of troctolite composition, like “S” in Figure 3, any pressure (low and high) will lead to spinel as the liquidus phase and thus can produce spinel-bearing rocks.
 The difference between low- and high-pressure crystallization of composition S is in the minerals that accompany spinel. At low pressure, melt of composition S will evolve toward the plagioclase-spinel cotectic, where spinel and melt will react to produce plagioclase. Continued cooling and crystallization will drive the melt to the spinel-plagioclase-olivine peritectic, where spinel and melt will react to produce plagioclase + olivine.
 At high pressures, a melt of composition S will evolve toward the spinel-plagioclase-pyroxene eutectic point. Thus, rocks that crystallized from a melt like S will have spinel in equilibrium with pyroxene at high pressures (similar to melts with a bulk composition like M). However, this inference is not consistent with our observation of textural disequilibrium between spinel and pyroxene, which points to crystallization at low pressures. It also does not explain the critical observation of the two different olivine compositions and their different textures nor does it explain the rim of olivine around pyroxene.
 Another question that arises regards the origin of a troctolitic melt composition. Such a melt cannot be derived from a chondritic (or nearly chondritic) source by fractional crystallization at low or high pressure. The source of a troctolitic melt must be enriched in plagioclase relative to a “chondritic” or picritic source (it must have superchondritic Al/Mg). After such a differentiated source region is created, it must then be melted, which requires a significant addition of heat. The melting temperatures of troctolitic compositions are above 1250°C [Walker et al., 1973b], which is higher than the melting temperatures of most basalts. The required heat source could come from a meteorite impact event, or from a very hot intrusive body (like a picrite or anhydrous komatiite).
 These scenarios involving troctolitic source regions can accommodate most of the critical observations above: rapid igneous cooling, extensive postigneous equilibration, apparent disequilibrium between spinel and pyroxene. The olivine rims around pyroxenes could arise also if the pyroxene were relict, unmelted grains from the original troctolite. However, they cannot explain the resorption features of the olivine grains.
 A third scenario for generation of spinel-bearing magmas and rocks is assimilation of anorthite into high-temperature (picritic) basaltic magmas. Anorthosites are abundant in the lunar crust, picritic magmas have passed through the crust en route to eruption, and so it seems reasonable that picritic magmas might have reacted chemically reacted with the crust. In fact, laboratory experiments show that this mechanism can produce spinel-bearing silicate melts [Finnila et al., 1994; Morgan et al., 2006]. The effects of anorthite assimilation by picritic magma may vary depending on the geometry and extent of the chemical interaction, and the temperature and pressure of interaction (high level versus deep crustal).
Figures 4 and 5 show a simple geologic and petrologic model for this assimilation scenario: a magma chamber of olivine-saturated picritic melt emplaced into anorthositic crust (Figure 4) [after Hiesinger and Head, 2006]. Relevant phase relations (Figures 5a and 5b) are simplified and projected into the systems olivine-anorthite-silica and olivine-anorthite [Walker et al., 1972, 1973a; Morgan et al., 2006]. Compositions of spinel and spinel-bearing melts fall outside these systems. Chemical interactions between picritic magma and wall rock will produce reaction zones along their contacts (zones 1–5; see Figures 5c and 5d) [see Morgan et al., 2006]. As anorthosite (zone 5) dissolves, the magma (zone 1) temperatures will decrease because of the energy required for dissolution [Morgan et al., 2006]. However, the magma composition will move toward that of anorthite, into zone 2 (Figure 5b), which will lower its liquidus temperature, and will cause any previously crystallized olivine to dissolve back into the melt. Continued assimilation will drive the melt compositions farther toward anorthite, first into the spinel + liquid field (zone 3; see Figures 5a and 5b) and (near the contact with anorthosite) into the anorthite + spinel + liquid field (zone 4; see Figures 5a and 5b). These reaction zones may ultimately be preserved as a range of rock types in close spatial association (Figures 5c and 5d) along the walls of the magma chamber.
 Near the upper contact (Figure 5d), melt of zone 4 will crystallize plagioclase and then plagioclase + spinel. With continued cooling, its melt composition (in this projection) will move down the spinel-anorthite reaction curve to the spinel-plagioclase-olivine peritectic point. There, the remaining spinel will react with melt to produce olivine, ultimately forming a troctolite rock (possibly with relict spinel grains).
 Melt of zone 3 falls in the liquidus field of spinel (Figures 5a and 5b) [see Morgan et al., 2006, Figure 2]. As this melt cools, spinel will continue to crystallize, followed by plagioclase + spinel. When the melt reaches the spinel-plagioclase-olivine peritectic point, olivine will crystallize at the expense of spinel, and eventually yield a spinel-rich troctolite (Figure 5d).
 Melt of zone 2 will crystallize olivine first after some cooling, and will move to the olivine-spinel reaction curve. Crystallization then of olivine + spinel will move the melt composition to the olivine-spinel-plagioclase peritectic, where spinel will react with the melt to form plagioclase + olivine, yielding again a spinel-bearing troctolite (Figure 5d).
 Melt of zone 1, the main magma chamber, will yield a typical layered mafic intrusion, consisting of (bottom to top): cumulate dunite (olivine-rich), cumulate peridotite (olivine-pyroxene), cumulate pyroxenite, and gabbros (Figures 5c and 5d). Such rocks are known from the Moon as “magnesian suite plutonics,” and are inferred to have formed in post-magma-ocean basaltic (and picritic) intrusions into the lunar crust [e.g., Raedeke and McCallum, 1979, 1980; James, 1980].
 Crystallization at the lower contact (Figure 5d) will yield a sequence of rocks similar to that of the upper contact, but probably thicker because crystals will settle rapidly in fluid lunar magmas [Taylor and Lu, 1991]. Spinel-rich cumulates are likely to form as spinel-rich melts descend from the upper contact [see Morgan et al., 2006, Figure 3]. One should also expect cumulates rich in olivine and spinel.
 This model can explain all of the critical inferences about clast 2: the rapid cooling which yields smaller grain sizes can be explained, as the crust is much colder than the intrusion; the textural disequilibrium between spinel and pyroxene; the resorbed olivine grains represent the original olivine crystallized from the picritic melt, and the interstitial olivine grains represent late stage crystallization at the olivine-spinel-plagioclase peritectic. The late stage olivine should be more magnesian than the olivine that crystallized from the picritic melt as FeO partitions preferentially into spinel [Herzberg and Baker, 1980]. This could explain why the olivine grains that show strong resorbtion features are slightly less magnesian than later crystallized olivines. The pyroxenes surrounded by olivine would represent relict low-Ca pyroxene from the anorthosite that got intruded; they were not stable in such a melt but were not fully melted/resorbed.
 All of the critical petrologic inferences about clast 2 can be explained by a system where anorthosite is assimilated by a picritic magma. In this case, the spinel-rich clast most likely originated from the zone 3, the spinel-rich cumulate layer.
 This model not only explains the petrographic and chemical features of our clast, but is also consistent with the regional setting of the spinel-rich regions identified by the M3 reflectance spectroscopy. The M3 spinel-rich area in the Moscoviense region is spatially associated with two other areas of unusual mineral compositions, one rich in olivine, and the other rich in orthopyroxene (the “OOS” association) [Pieters et al., 2011]. A picritic intrusion into the anorthositic crust, as suggested here, could produce such rocks as cumulate dunite and pyroxenite, in addition to its spinel-rich cumulates. All of these rocks, igneous and metasomatic, could have been brought to the Moon's surface in the impact that produced the Moscoviense basin. Other spinel-rich rocks of such a picritic intrusion, like spinel-olivine cumulates and spinel-rich troctolites may be present but not yet identified. With the advent of mapping spectrometers of high spectral and spatial resolutions and a model of what to look it seems likely exposures of spinel-rich materials might be found elsewhere in the lunar highlands [Pieters et al., 2011; Dhingra et al., 2011; Lal et al., 2011].
 We are grateful to L. Le and A. Peslier for assistance with the EMP analyses and to J. Filiberto for fruitful discussions and suggestions. We thank P. Isaacson and an anonymous reviewer for helpful reviews and comments, as well as M. Wieczorek for his editorial handling of this manuscript. This paper is partially supported by NASA Cosmochemistry grant NNX08AH78G, LPI contribution 1642.