Magnesian anorthositic granulites in lunar meteorites Allan Hills A81005 and Dhofar 309: Geochemistry and global significance


Corresponding author. E-mail:


Abstract– Fragments of magnesian anorthositic granulite are found in the lunar highlands meteorites Allan Hills (ALH) A81005 and Dhofar (Dho) 309. Five analyzed clasts of meteoritic magnesian anorthositic granulite have Mg′ [molar Mg/(Mg + Fe)] = 81–87; FeO ≈ 5% wt; Al2O3 ≈ 22% wt; rare earth elements abundances ≈ 0.5–2 × CI (except Eu ≈ 10 × CI); and low Ni and Co in a non-chondritic ratio. The clasts have nearly identical chemical compositions, even though their host meteorites formed at different places on the Moon. These magnesian anorthositic granulites are distinct from other highlands materials in their unique combination of mineral proportions, Mg′, REE abundances and patterns, Ti/Sm ratio, and Sc/Sm ratio. Their Mg′ is too high for a close relationship to ferroan anorthosites, or to have formed as flotation cumulates from the lunar magma ocean. Compositions of these magnesian anorthositic granulites cannot be modeled as mixtures of, or fractionates from, known lunar rocks. However, compositions of lunar highlands meteorites can be represented as mixtures of magnesian anorthositic granulite, ferroan anorthosite, mare basalt, and KREEP. Meteoritic magnesian anorthositic granulite is a good candidate for the magnesian highlands component inferred from Apollo highland impactites: magnesian, feldspathic, and REE-poor. Bulk compositions of meteorite magnesian anorthositic granulites are comparable to those inferred for parts of the lunar farside (the Feldspathic Highlands Terrane): ∼4.5 wt% FeO; ∼28 wt% Al2O3; and Th <1 ppm. Thus, magnesian anorthositic granulite may be a widespread and abundant component of the lunar highlands.


The meteorites Allan Hills (ALH) A81005 and Dhofar (Dho) 309 (and its paired specimens, including Dhofar 303 and 489) are feldspathic rocks derived from the lunar highlands, and both are relatively magnesian (i.e., with high Mg/Fe ratios) compared to most highland samples (Korotev et al. 1983a, 2006; Treiman and Drake 1983; Takeda et al. 2006, 2007). Since sample return missions visited only a limited portion of the Moon’s nearside, data provided by meteorites like these are invaluable for determining the composition of the bulk lunar crust. Remote sensing studies of the Moon by the Lunar Prospector and Clementine spacecraft show that, relative to the Apollo and Luna landing sites, most of the lunar surface has low abundances of FeO and incompatible elements (Jolliff et al. 2000; Gillis et al. 2004), with an average composition more like that of the feldspathic lunar meteorites (Korotev et al. 2003). Of particular interest are the magnesian granulitic lithologies, which are found as clasts in many feldspathic lunar meteorites. Magnesian granulitic rocks are compositionally distinct from Apollo and Luna pristine samples, and do not fit well in the context of simple models of lunar petrogenesis (Lindstrom and Lindstrom 1986; Korotev and Jolliff 2001; Korotev et al. 2003).

Lunar granulites are metamorphic rocks, and are generally interpreted as impactites—granulated, mixed, and possibly melted by impact events, and chemically equilibrated during prolonged cooling after the impacts (Simonds et al. 1974; Warner et al. 1977). Two-pyroxene thermometry of granulites generally gives equilibration temperatures of 1000–1150 °C (Cushing et al. 1999; Hudgins et al. 2008), which suggest formation either near impact craters of 30–90 km diameter (Cushing et al. 1999) or by burial metamorphism associated with basin-sized impacts (Hudgins et al. 2008). Many granulites are ancient, with radiometric ages of equilibration as old as 4.3 Ga (James and Hammarstrom 1977; Warner et al. 1977) and Ar-Ar ages to 4.1 Ga (Hudgins et al. 2008). Chemically, the Apollo granulites divide cleanly into magnesian and ferroan groups (Lindstrom and Lindstrom 1986), and contain little to no KREEP component (Warner et al. 1977; Lindstrom and Lindstrom 1986). The lunar highland meteorites contain abundant clasts of granulite, which are generally similar to the Apollo granulites. The chemical compositions of both the Apollo granulites and the lunar highland meteorites (from the limited data available) cannot be modeled as mixtures of, or fractionates from, known pristine lithologies (Lindstrom and Lindstrom 1986; Korotev and Jolliff 2001; Korotev et al. 2003), “Hence, the ALHA81005 granulitic clasts cannot be explained as a mixture of known pristine rock types and, although not themselves pristine, may represent an endogenous igneous rock type dissimilar to any currently recognized in the Apollo collection” (Korotev et al. 1983b).

This paper provides mineralogical and chemical data on five clasts of magnesian anorthositic granulite from thin sections of lunar meteorites: two from ALHA81005, and three from Dhofar 309. Their bulk chemical compositions were obtained by combining mineral analyses (from electron microprobe [EMP] and secondary ion mass spectrometry [SIMS]) with mineral proportions determined by X-ray mapping and multispectral image processing (Maloy and Treiman 2007). This method is virtually non-destructive, making it useful for fragments found only in thin section and for other rare samples (Taylor et al. 1996, 2002; Bowman et al. 1997; Hicks et al. 2002; van Niekerk 2003; Anand et al. 2006). Magnesian anorthositic granulites in lunar meteorites have been partially characterized in several previous studies, though not including mineral, bulk, and trace element compositions (Goodrich et al. 1984; Fukuoka et al. 1986; Jolliff et al. 1991; Lindstrom et al. 1991; Warren and Kallemeyn 1991; Consolmagno et al. 2004). The magnesian anorthositic granulite clasts in ALHA81005 and Dhofar 309 have nearly identical chemical compositions, although their host meteorites did not come from the same lunar source crater. Granulites with similar Mg′ are found in nearly all feldspathic lunar meteorites, which suggests that magnesian anorthositic granulite is widespread across the Moon.

Analytical Methods

Mineral Compositions

Major and minor element compositions of mineral phases were analyzed with a Cameca SX 100 electron microprobe (EMP), in wavelength dispersive mode, at Johnson Space Center (Table 1). Analyses were made using an electron beam with 15 kV accelerating potential and 20 nA current (into a Faraday cup) focused on the sample. Standards included well-characterized natural mineral samples, including kaersutite, olivine, albite, adularia, chromite, and spessartine. There was no evidence of Na loss from plagioclase during analyses.

Table 1.   Average compositions of major minerals in clasts A1 and A2 in ALHA81005 and clasts D1, D2, and D3 in Dho 309.
 ALHA81005—Clast A1ALHA81005—Clast A3Dho 309—Clast D1
Mg* An97. 97.181.683.386.8 96.885.786.986.6 
 Sc 9.727   9.719   8.825  
 V 25120   2666   3681  
 Co 6830   3126   7729  
 Ni 8228   84110   17063  
 Sr 6.74.8   62a68a   98a41a  
 Y 0.546.4   1.84.9   0.968.7  
 La  0.02  0.29Bdl   0.82< 
 Ce  0.23  0.83Bdl 
 Nd  0.69  0.67Bdl   1.2Bdl1.22.5 
 Sm  0.62  0.26Bdl   0.44Bdl0.811.8 
 Eu  0.16  0.81Bdl   0.95< 
 Dy  1.31  0.100.51   0.42Bdl1.42.9 
 Er  0.96  0.050.32   0.26Bdl0.981.6 
 Yb  1.17  0.030.35   0.28<0.40.901.6 
 Dho 309—Clast D2Dho 309—Clast D3
PlagOlvOpxCpxSpinel ISpinel IIPlagOlvOpx
  1. Element oxides, in %, by EMP; trace elements, in ppm, by SIMS. Blank indicates no analysis; “Bdl” is “below detection limit.”

  2. aHigh Sr concentrations inferred to reflect terrestrial contamination.

  3. For plagioclase, An is molar Ca/(Ca + Na) in %; for olivine and pyroxenes, Mg* is molar Mg/(Mg + Fe) in %.

An Mg*96.383.284.883.6  97.287.988.8
 Sc 7.737    9.624
 V 23120    4491
 Co 7735    7132
 Ni 16078    140230
 Sr 110a94a    2400a300a
 Y 0.5813    2.06.7

Trace and rare earth element (REE) abundances in major mineral phases (Table 1) were obtained by secondary ion mass spectrometry (SIMS) using a Cameca IMS 4f instrument at the University of New Mexico (see Papike et al. 1994 for details of this method). The primary beam was of O ions, at 10 kV potential and 20 nA current. The ion spot size on the sample was ∼20 μm diameter. Analyses involved repeated cycles of counting peaks and backgrounds. Element concentrations in a mineral were calculated from empirical relationships of Element+/30Si+ ratios (normalized to the known SiO2 content of that mineral), which were based on multiple analyses of daily-calibrated glass standards. One-sigma uncertainties in calibration and counting statistics are estimated to be 20% for the REEs (±∼0.4 × CI for REEs measured here, with the exception of Eu at ±∼2 × CI). Transition element abundances in olivine and pyroxene are uncertain to about 4%, except for Sc which is uncertain to ∼10% (Karner et al. 2003).

Mineral Modes

To determine mineral proportions in each clast (Table 2), a representative area was selected by optical microscopy. An X-ray map of this area was acquired using the Cameca SX100 electron microprobe. The thin section was moved beneath the static electron beam in a grid of 1.0–2.5 μm steps. At each point, X-ray counts were acquired for 65 milliseconds at the characteristic Kα energies for Al, Ca, Cr, Fe, Mg, Ni, P, S, Si, and Ti. Raw X-ray counts in each X-ray image were scaled to fall in the range of 1 to 255 for use in multispectral image processing software.

Table 2.   Area percentages of minerals in magnesian granulite clasts.

We retrieved the modal mineralogy of magnesian granulite clasts using image classification software designed for interpretation of remote sensing data (Maloy and Treiman 2007). To determine the proportions of plagioclase, olivine, pyroxenes, and spinel, we used X-ray maps of Al, Ca, and Mg and classified their pixels using the computer software package Erdas Imagine 8.7© (Leica Geosystems 2003). Pixels were assigned to classes by an unsupervised ISODATA algorithm. The number of target classes in the classification was varied until results were consistent with visually observed boundaries among minerals. Typically, a mineral was represented by several classes, which were manually combined to give the total area of the mineral. The ISODATA classification did not return consistent results for area proportions of augite versus orthopyroxene, so that ratio was determined manually from a histogram of Ca X-ray counts for all pixels (Maloy and Treiman 2007), on which each Ca-bearing mineral is represented by a distinct peak. In order of decreasing Ca-counts, these peaks were assigned to merrillite, plagioclase, augite, orthopyroxene, and olivine. Area proportions of several minor phases were obtained from EMP X-ray maps of single elements: chromite from Cr; ilmenite from Ti; troilite from S; and merrillite from P.

The major source of uncertainty in modal mineral abundances is from counting statistics on the number of mineral grains. Solomon (1963) provides a measure of this variance σ in point counting as


where p is the volume percentage of a particular mineral in the rock, a is the distance between analytical spots, R is the average radius of grains of that mineral, and A is the total measurement area. From this relationship, the most certain mineral abundance in the clasts is plagioclase in D3, at 72 ± 2% 1 σ by volume (for = 2.5 μm, = 2,250,000 μm2, and R(pl) ≈ 25 μm; Table 2), and this level of uncertainty would apply to the elements sited in plagioclase (Ca, Al, Eu, Sr). Uncertainties in abundances of trace minerals are near their levels of abundance (e.g., merrillite in clast A3 is present at 0.001 ± 0.001%).

Bulk Composition from Mineral Modes

The bulk chemical compositions of the mapped areas of each clast, Table 3, were derived from the mineral analyses (major and trace elements), mineral proportions, and mineral densities. It was assumed that the mapped and analyzed area proportions are equal to volume proportions because mineral distributions within the clasts are isotropic. Volume proportions of minerals were then scaled to their mass proportions by mineral densities, either from literature compilations or calculated based on Fe/Mg (for mafic minerals) or Na/Ca (for plagioclase). Uncertainties in derived bulk compositions are shown in Table 3, including the independent uncertainties in element abundances in minerals (EMP and SIMS), and in mineral mass proportions (by element X-ray mapping).

Table 3.   Calculated compositions of magnesian granulite clasts.
  1. Uncertainties are estimated 2σ, in percentage of mass present. Uncertainties for major/minor element abundances are dominated by those of mineral proportions, notably of ilmenite (Ti) and chromite (Cr). Uncertainties in trace element abundances are dominated by those of mineral analysis by SIMS.

  2. aBulk Sc values include calculated contributions from clinopyroxene using ScDcpx/opx = 3. Bulk values for other transition metals were calculated from olivine and opx data alone, assuming that contributions from plagioclase, spinel, and augite are small. Values in parentheses include some calculated mineral abundances. Values in brackets were affected by terrestrial weathering.

 La (0.21)0.520.43 40
 Ce (0.62)1.20.95 40
 Nd (0.60)0.850.84 40
 Sm(0.37)(0.33)0.360.43 40
 Eu (0.54)0.580.56 40
 Dy(0.44)(0.54)0.410.41 40
 Er(0.34)(0.33)0.270.34 40
 Yb(0.41)(0.24)0.270.40 40

Determining bulk composition via mineral modes and spot analyses is advantageous in that compositions can be obtained for objects or areas in situ in thin sections or slabs, that all major elements can be analyzed (unlike, for example, Si by INAA or ICP-MS), and that it is non-destructive (except for SIMS pits). The foremost weakness of this method is the necessary assumption that the mapped area is representative of the clast and a larger rock body that the clast came from. We chose map areas that appeared to be representative in thin section, in terms of textures and proportions of abundant minerals. We do not know if the map areas were representative in their proportions of trace minerals, like augite or merrillite. For small objects (like the granulite clasts here) reconstructed bulk compositions may not be good representations of the parental lithology. This uncertainty is difficult to quantify, but varies inversely with the number of discrete mineral grains in the analyzed object.

However, if grains are too small they cannot be analyzed well by the available techniques. In these granulites, the 1 μm spot size of EMP allows analyses for major and minor elements in all mineral phases. Nevertheless, the diameter of the Cameca 4f SIMS primary ion beam is larger than several types of mineral grains, notably merrillite and augite. In those cases, trace element compositions of small mineral grains are calculated from analyses of other minerals and mineral-mineral distribution coefficients (D) measured elsewhere for comparable lunar granulites or anorthosites (Treiman 1996; Floss et al. 1998; Cahill et al. 2004).

Similarly, abundances of elements sited primarily in rare mineral grains (e.g., P in phosphate, Ti in ilmenite) will be uncertain. For abundances of these elements, uncertainties scale as one over the square root of the number of grains present, and so may be very large. For the granulites in this study, abundances of most REEs are not strongly affected by this uncertainty. The bulk of their REEs reside in pyroxenes and plagioclase, as calculated from mineral abundances (Table 4) and mineral-mineral partition coefficients (Treiman 1996). A related issue is that highly incompatible elements might be concentrated along grain boundaries and not in minerals, e.g., like La and Th in dunite (Hiraga et al. 2007). In the granulites analyzed here, REE abundances are low (<5 × CI, except Eu) and it seems likely that these small abundances could be accommodated into the clasts’ plagioclase and pyroxenes.

Table 4.   Clast D1: percentage contributions of La, Eu, and Yb in the bulk composition from major and REE-rich minerals.
  1. La and Eu in olivine assumed to be zero. Yb in olivine calculated Yb in orthopyroxene (Table 1) and Dolv/opx from analyses of clast D2. REE in merrillite calculated from REE in augite (Table 1) and Dmer/aug from comparable lunar rocks (Treiman 1996; Floss et al. 1998; Cahill et al. 2004).


Petrography and Chemistry of Analyzed Clasts

Allan Hills A81005

Allan Hills A81005 is a polymict regolith breccia (Antarctic Meteorite Newsletter 1983) with bulk composition of a noritic anorthosite (Korotev et al. 1983a; Treiman and Drake 1983). Among its many varieties of lithic and mineral clasts is a matrix of glassy agglutinates and broken mineral fragments, with rare spheres of impact melt–typical for a regolith breccia. We analyzed two magnesian anorthositic granulite clasts from thin section ALHA81005, 48–A1 and A3, which are clasts 48–2,b and 48–3 of Goodrich et al. (1984). The clasts have identical granoblastic textures (Figs. 1 and 2) and identical minerals (plagioclase, olivine, orthopyroxene, clinopyroxene, and traces of spinel and troilite) (Table 2). Minerals in both clasts are chemically homogeneous, but do have slightly different compositions (Table 1) suggesting that they are not pieces of a single larger clast (see Goodrich et al. 1984).

Figure 1.

 Reflected light images (mosaics of microphotographs) with portions of analyzed thin sections. Images not at same scale. a) Portion of ALHA81005,48 showing granulite clast A1 and surrounding regolith. Analyzed area shown as box (500 μm × 300 μm). b) Portion of ALHA81005,48 with two granulite clasts and intervening regolith matrix. Clast A3 is labeled; analyzed area shown as box (500 μm × 300 μm). c) Portion of Dho309 #1057 with three larger clasts of granulite. Clasts D1, D2, and D3 are labeled, with areas analyzed here shown as insets (D1: 2500 μm × 2500 μm; D2: 1000 μm × 500 μm).

Figure 2.

 Maps of characteristic X-ray emission under EMP of analyzed clasts. a) A1, b) A3, c) D1, and d) D2. Fields correspond to insets in Fig. 1. False colors represent intensities of emissions of characteristic X-rays: Al in blue, Ca in green, and Mg in red. In this color coding, plagioclase is greenish blue (Al + Ca), olivine is bright red (Mg), orthopyroxene is dark red (less Mg), augite is muddy green (Ca + Mg), spinel is purplish blue (Al + Mg), calcite is bright green (Ca), and cracks and epoxy are black. Calcite in D1 and D2 (bright green) is found only in cracks, and is interpreted as a terrestrial deposit.

The REE abundances of clast A3, reconstructed as above, show a flat pattern at ∼1.5 × CI, and a positive Eu anomaly of ∼10 × CI (Table 3; Figs. 4 and 5). Clast A3 is slightly depleted in light REEs, nearly within analytical uncertainties. This depletion might be real or it could be an artifact from using partition coefficients (D values) to calculate the REE contents of pyroxenes. On the other hand, the bulk composition of ALHA 81005 is slightly enriched in the LREE (Fig. 5), and it is possible that some LREE-enriched matrix or other material was included in the clasts analyzed by Goodrich et al. (1984).

Figure 4.

 Reconstructed REE abundances of bulk clasts A1, A3, D1, and D2; linear REE scaling, CI-normalized (Anders and Grevesse 1989). One σ uncertainties are approximately 30% of the graphed abundances (of which ∼20% derives from uncertainties in SIMS analyses, and ∼10% from uncertainties in mineral proportions).

Figure 5.

 Rare earth element abundances of meteoritic magnesian anorthositic granulite and other lunar materials; linear REE scaling, CI-normalized (Anders and Grevesse 1989). Light gray range is for magnesian anorthositic granulite clasts analyzed here (Fig. 4); dark gray range is for magnesian granulitic clasts in ALHA81005 from Goodrich et al. (1984). Bulk ALHA81005 in filled circles (Palme et al. 1983); bulk Dhofar 489 (paired with Dhofar 309) in open circles (Takeda et al. 2006). Apollo magnesian granulitic breccias, open circles, include: 67415 (average of 33A and 33B); 67955 (average of 74A and 74B); 76230; average 79215; 76503,7021; 76503,7052; 72275,439/495; and average 60035 (Ma and Schmitt 1982; Lindstrom and Lindstrom 1986; Salpas et al. 1988; Jolliff et al. 1996; Hudgins et al. 2008).

Allan Hills A81005—Clast A1

Texturally, clast A1 is similar to A3. Plagioclase occurs as larger equant grains (up to ∼200 μm across), set in a matrix of smaller equant grains (5–30 μm across) of plagioclase, olivine, orthopyroxene, and augite. Chromite, ilmenite, troilite, and merrillite are present but rare (Table 2).

Clast A1 includes 63% plagioclase (An97.0), 20% olivine (Fo82.2), 14% orthopyroxene, 2% augite, and a small proportion of aluminous chromite; all minerals are of constant composition (Table 1, Fig. 3). SIMS data for orthopyroxene and olivine are given in Table 1; data for plagioclase are not available. REE abundances in augite and plagioclase were calculated from the orthopyroxene analyses using mineral-mineral distribution coefficients from a granulite with similar major-element mineral compositions (Dhofar 025; Cahill et al. 2004). Orthopyroxene is the principal host of the heavy REE in clast A1, so their calculated bulk abundances are probably fairly precise; calculated bulk abundances of Eu and the light LREE and Eu are quite imprecise as little of these elements are sited in the orthopyroxene; these values are not listed in Table 3.

Figure 3.

 Mg′ [molar Mg/(Mg + Fe)] versus plagioclase composition [An = molar Ca/(Ca + Na)] for magnesian anorthositic granulite clasts analyzed here, large filled circles (Table 1); magnesian anorthositic granulite clasts analyzed by Goodrich et al. (1984), small filled circles; and Apollo magnesian granulites, open circles (data from Ma and Schmitt 1982; Lindstrom and Lindstrom 1986; Salpas et al. 1988; Hudgins et al. 2008). Fields for pristine magnesian-suite rocks and ferroan anorthosites are after Goodrich et al. (1984).

The reconstructed bulk composition of clast A1 is given in Table 3. Clast A1 has Mg′ = 82 (forced by its olivine), and HREE abundances are at ∼2–3 × CI. For major, minor, and heavy rare earth elements, the composition of clast A3 is very similar to that of A1 (other REE abundances cannot be calculated with any certainty), although the compositions of their chromites are distinct (Table 1).

Dhofar 309

Dhofar 309 is an anorthositic impact melt breccia (Russell et al. 2003) and is paired with 14 other stones including Dhofar 303 and Dhofar 489 (Korotev et al. 2006; Takeda et al. 2006, 2007). Cracks in Dhofar 309 contain deposits of calcium carbonate, calcium sulfate, iron oxy-hydroxide(s), and other minerals acquired during residence on Earth. In this study, a thin section provided by the Vernadsky Institute, Russia (their section Dhofar 309#1057; Fig. 2b) was examined.

Consistent with other descriptions of Dhofar 309 and its pairs, the matrix of the analyzed section consists of finely crystalline feldspathic impact melt with rounded clasts of magnesian troctolitic anorthosites, anorthositic troctolites, and rare anorthosite, many containing spinel (Figs. 1 and 2; see Takeda et al. 2007). The analyzed section contains three large clasts of anorthositic troctolite with granulitic textures (i.e., are composed of equant mineral grains of similar sizes; Fig. 1). The mineral grains are chemically homogeneous. However, the clasts also show some relict igneous textures, like subhedral relics of euhedral plagioclase laths with interstitial sub-ophitic pyroxenes and olivine. Some olivine grains are partially surrounded by orthopyroxene (Fig. 2b), suggesting the crystallization reaction olivine + melt→orthopyroxene (e.g., Morse 1980). Also, spinel grains are commonly enclosed in plagioclase (Fig. 2c), suggesting the crystallization reaction spinel + melt→anorthite in compositions near the anorthite-olivine join (e.g., Morse 1980). Thus, it seems likely that these clasts were originally impact melts, which crystallized and were metamorphosed so as to chemically homogenize their mineral and partially erase their igneous textures.

Dhofar 309—Clast D1

Clast D1 contains 64% plagioclase (An96.8), 25% olivine (Fo85.7), 10% orthopyroxene, and 0.5% of both augite and Mg-Al spinel, all of constant composition (Table 2, Fig. 3). We obtained REE abundances for plagioclase, orthopyroxene, augite, and olivine. The calculated REE pattern of clast D1 is essentially flat at ∼2 × CI, with a positive Eu anomaly at about ∼10 × CI (Fig. 4). This pattern is essentially identical to that of the bulk meteorite (Fig. 5; i.e., Dhofar 489 which is paired with Dhofar 309; Korotev et al. 2006).

Dhofar 309—Clast D2

Clast D2 contains 62% plagioclase (An96.3), 25% olivine (Fo83.2), 12% orthopyroxene, 0.3% augite, and 0.7% spinels. Mineral compositions do not vary (Table 2, Fig. 3). D2 contains two distinct spinels, a chromite and an Mg-Al spinel, which fall close to the MgAl2O4–FeCr2O4 join. The spinels are not chemically zoned, and there is no sign that one crystallized before the other; their compositions may represent chemical equilibrium. The low-Cr spinel in D2 has approximately twice as much Cr as the spinel in D1, suggesting that the presence of two spinels in D2 (and the absence of two spinels in D1) represents a real difference in the clasts’ Cr contents.

Secondary ion mass spectrometry analyses were obtained on all mineral phases except spinels (Table 1). The reconstructed bulk composition of clast D2 has nearly a flat REE pattern at ∼2 × CI, with a slight depletion in light REEs, and with Eu at ∼10 × CI (Fig. 4). This pattern is very similar to those of clast D1 and of the bulk meteorite (Fig. 5).

Dhofar 309–Clast D3

Clast D3 is a troctolite, containing no augite and only small proportions of orthopyroxene and spinel (Table 1). It contains 72% plagioclase (An97.2), 28% olivine (Fo87.9), one grain of orthopyroxene, and four minuscule grains of chromite. Mineral compositions do not vary (Fig. 4). The Mg′ of its olivine and An content of plagioclase place it in the field of “magnesian-suite” rocks (Fig. 3). We did not obtain REE data for this clast.

Discussion, Part I: Geochemistry of Magnesian Anorthositic Granulites

Chemically, the clasts of magnesian anorthositic granulite in ALHA81005 and Dhofar 309 are distinct from other lunar materials, including magnesian anorthositic granulites collected during the Apollo missions (Lindstrom and Lindstrom 1986). Here, we show that the meteoritic magnesian anorthositic granulite are (1) little contaminated by meteoritic components, and therefore represent true lunar compositions; (2) represent a significantly restricted range of compositions; and (3) are chemically distinct from other lunar lithologies and compositional components.

Meteoritic Contamination

Lunar granulites, including those described here, are regarded as impactites, and could be significantly contaminated by impactor material, most likely chondritic in composition. Siderophile elements (e.g., Ni, Co, Ir, etc.) are the most sensitive tracers of chondritic meteorite contributions in lunar rocks; pristine lunar rocks contain extremely low abundances of siderophiles (Warren 1993).

The clasts of magnesian anorthositic granulite must contain only limited proportions of chondritic material. We could analyze their olivine and pyroxenes for the siderophile elements Ni and Co; the clasts contain only minuscule proportions of metal and sulfide (Table 2). By modal recombination, the clasts contain little Ni or Co (20–55 ppm Ni and 9–27 ppm Co; Table 3) and have non-chondritic Ni/Co of ∼0.04 to 0.4 × CI (Table 3). These abundances and ratios suggest relatively little chondritic component, such that indigenous Ni and Co are dominant. Nickel and Co are surely present in Fe-sulfide and metal, but their proportions are so low that they can affect bulk Ni and Co abundances minimally. For further explanation see the Appendix.

Bulk Composition

In this study, granulitic clasts were classified as magnesian and anorthositic based on two criteria: Mg′ > 70, and abundant plagioclase (>50% volume). Nevertheless it turned out that most of the analyzed granulitic clasts were magnesian anorthositic with very limited ranges of Mg′ (80–86; Table 3) and plagioclase abundance (62–72%; Table 2). REE abundances in these clasts are essentially identical, with HREE near 2 × CI and Eu near 10 × CI (Figs. 4 and 5); LREE are slightly and variably depleted compared to HREE (Figs. 4 and 5), but these variations may be within analytical uncertainties.

Although the clasts of magnesian anorthositic granulite are similar, they are not identical in mineralogy or in bulk chemistry. Mineralogically, the clasts from Dhofar 039 contain less augite than those from ALHA81005 (Table 2). Clast D3 is distinct from the others in that it contains very little pyroxene of either species (Table 2), and D2 is distinct by containing both Cr-rich and Cr-poor (Al-rich) spinels (Table 1).

With respect to bulk chemistry, the magnesian anorthositic granulite clasts show some significant variations in element abundances (Table 3). Titanium is approximately twice as abundant in the clasts in ALHA 81005 than it is in those in Dhofar 309. Chromium also varies significantly, from 0.06 to 0.41 wt% Cr2O3 (Table 3). This variation in Cr content is not solely from uncertainty in abundance of spinel (major host of Cr). The Cr-poor spinel in Clast D2 is richer in Cr than the Cr-poor spinels in the other clasts, confirming that the differences in Cr abundance are not merely results of sampling uncertainty of uncommon spinel grains. This range of compositions means that the clasts magnesian anorthositic granulites do not represent a simple, or single lithology or cogenetic sequence; instead they likely represent a range of related lithologies or chemical varieties.

Chemical Distinctness

As seen above, the compositions of magnesian anorthositic granulite clasts are distinct from other lunar compositions, and therefore must represent a different petrogenesis. In this section, we concur with and re-enforce Korotev’s inference in 1983 that “. . . the 81005 granulitic clasts cannot be explained as a mixture of known pristine rock types and . . . may represent an endogenous igneous rock type dissimilar to any currently recognized in the Apollo collection.” In the next section we will place the composition of these magnesian anorthositic granulites in the context of global lunar geochemistry.

Mg′ Versus An

The principal classification scheme of lunar non-mare rocks is to use the Mg′ (molar Mg/(Mg + Fe) ratio) of the mafic minerals and the An number (molar Ca/(Na + Ca) ratio) of the plagioclase (e.g., Raedeke and McCallum 1980). This presentation clearly distinguishes rocks of the ferroan anorthosite suite from those of the magnesian-suite plutonic rocks (e.g., gabbros and troctolites). Magnesian anorthositic granulites from both the Apollo collection and the lunar meteorites are clearly distinct from ferroan anorthosites, but plot in or near the area of magnesian-suite rocks (e.g., Goodrich et al. 1984; Lindstrom and Lindstrom 1986).

Mg′ Versus Al2O3

A related presentation of bulk chemistry is Mg′ versus Al2O3 (Fig. 6, after Warren 2005), which ignores the An content of plagioclase but includes its abundance by the proxy of Al abundance. Figure 6 clearly shows that the magnesian anorthositic granulites (Apollo and meteoritic) are distinct from the ferroan anorthosite suite in having lower Al2O3 and higher Mg′ than the former (Table 3; Lindstrom and Lindstrom 1986).

Figure 6.

 Bulk molar Mg′ versus Al2O3 content for meteoritic magnesian anorthositic granulites (MAG) and other lunar rocks and meteorites (after Warren 2005). Meteoritic magnesian anorthositic granulites analyzed here are represented by filled circles (Table 3); small filled circles are magnesian anorthositic granulite clasts in Dhofar 309 pairs from Takeda et al. (2006, 2008). Feldspathic lunar meteorites as filled triangles, mixed composition lunar meteorites (with significant proportions of mare basalts and highland material: Korotev et al. 2009) as open triangles; lunar mare meteorites as down-pointing solid triangles, data from compilations of R. Korotev. Bulk ALHA81005 and Dhofar 309 (actually, the paired meteorite Dhofar 489) denoted. Apollo ferroan anorthosites shown as open crosses (Warren 1993). Lines show calculations of mixtures among end-members: ferroan anorthosite (Dhofar 081 bulk as the example), average meteorite magnesian anorthositic granulite from this work, a typical low-Ti mare basalt (from lunar meteorites), and a hypothetical magnesian peridotite (Warren 2005). Solid line is mixing between ferroan anorthosite and magnesian anorthositic granulite or peridotite; the two mixing lines coincide on this projection. Long-dash line is mixing between ferroan anorthosite and low-Ti mare basalt; short-dash line between meteoritic magnesian anorthositic granulite and low-Ti mare basalt.

REE Patterns

The rare earth element patterns of lunar rocks give crucial clues to their origins, and militate against simple relationships between these magnesian anorthositic granulites and other lunar rock types (Fig. 5). The magnesian anorthositic granulites all have positive Eu anomalies (e.g., Eu/Sm ratios) comparable to those of ferroan anorthosite suite rocks, but the Apollo magnesian anorthositic granulites have greater REE abundances overall (Fig. 5, Lindstrom and Lindstrom 1986; see Fig. 8.8 of Heiken et al. 1991). Also, the Apollo magnesian anorthositic granulites are more strongly enriched in light REE (e.g., La/Sm ratio) than most ferroan anorthosite suite rocks or even the REE-rich KREEP component (Lindstrom and Lindstrom 1986).

Figure 8.

 Ti-Sc-Sm geochemistry of meteoritic magnesian anorthositic granulite clasts and Apollo lunar materials. Magnesian anorthositic granulite (MAG) clasts are analyzed here as filled circles; magnesian anorthositic granulite clasts in ALHA81005 analyzed by Goodrich et al. (1984) as open circles; Mg-gabbronorites as filled squares (James and Flohr 1983); Apollo troctolites (T), Apollo magnesian anorthositic granulite as crosses (Lindstrom and Lindstrom 1986); and dunite 74215 (D), and bulk ALHA81005 and Dhofar 309 as triangles. a) Ti/Sm mass ratio versus Mg′ (figure after Norman and Ryder 1980; Cohen et al. 2004). b) Sc/Sm mass ratio versus Mg′ (figure after Norman and Ryder 1980; see Warren and Wasson 1980).

Compared to Apollo magnesian anorthositic granulites, the REE in the clasts analyzed here and by Goodrich et al. (1984) are relatively unfractionated—most REE at 1–2 × CI, and Eu near 10 × CI (Fig. 5). This REE pattern is similar to those of many ferroan anorthosites, and helps distinguish the magnesian anorthositic granulites from other magnesian rock types, like Apollo magnesian anorthositic granulites and the magnesian-suite plutonics.


The element abundance ratio Th/Sm is a good discriminant among pristine lunar rock types (Fig. 7a; after Korotev et al. 2003), combining a measure of incompatible element abundances (Sm) with a measure of enrichment of highly incompatible elements (Th/Sm). KREEP has the highest Sm abundance and Th/Sm ratio (∼2 × CI) of known pristine lunar materials (Fig. 7a), which are thought to have arisen via extreme mineral/melt fractionations (Korotev et al. 2003). Apollo magnesian-suite plutonic rocks have a similar Th/Sm, which suggests that their source magmas were originally KREEP-rich or assimilated KREEP after their formation (Snyder et al. 1995; Shearer et al. 2006). The bulk compositions of lunar feldspathic meteorites have a value of Th/Sm that is near that of KREEP (Fig. 7a; Korotev et al. 2003), which could imply that they contain some admixed KREEP component. Mare basalts generally have moderate-to-high Sm abundances and sub-chondritic Th/Sm, representing relatively low degree partial melting of LREE-depleted mantle sources, such as magma-ocean cumulates (Shearer et al. 2006). Ferroan anorthosites have low Sm abundances, and variably low Th/Sm, representing an absence of KREEP, and low but variable proportions of mafic minerals. The Apollo magnesian anorthositic granulite are distinct from these pristine lithologies, in having: (1) REE abundances too low for a close relationship with Apollo magnesian-suite rocks, (2) Mg′ too high for a close relationship with ferroan anorthosites (Lindstrom and Lindstrom 1986; Korotev and Jolliff 2001), and (3) Th/Sm ratios at, and above, that of KREEP (Fig. 7a; Korotev et al. 2003).

Figure 7.

 Th-La-Sm geochemistry of meteoritic magnesian anorthositic granulite (MAG) clasts, compared to other lunar materials. a) Chondrite-normalized Th/Sm (and La/Sm) versus Sm; figure after Korotev et al. (2003). Magnesian anorthositic granulite clasts are analyzed here as large filled circles (La/Sm, see Fig. 7b), magnesian anorthositic granulite clasts in ALHA 81005 analyzed by Goodrich et al. (1984) as small filled circles. For comparison, Apollo magnesian anorthositic granulites are shown as open circles (Lindstrom and Lindstrom 1986), and as small gray filled circles (Hudgins et al. 2008). b) Abundances of Th and La in magnesian anorthositic granulite clasts in ALHA 81005 (Goodrich et al. 1984) and selected lunar meteorites (Laul et al. 1983; Korotev et al. 2006; pers. comm.); graph after Korotev and Jolliff (2001) and Hudgins et al. (2008). Th/La values in the magnesian anorthositic granulite clasts (filled circles, see inset of area near origin) and meteorites with the lowest Th and La abundances (Northwest Africa [NWA] 482 and Dhofar 489 = Dhofar 309) are approximately at the CI value. Samples richer in La and Th (open circles) approach the KREEP value of Th/La of ∼1.3 × CI.

Several clasts of magnesian anorthositic granulite have been analyzed for Th and Sm (Goodrich et al. 1984), and those clasts fall in the field of ferroan anorthosites on Fig. 7a—low Sm contents and near-chondritic ratios of Th/Sm. Obviously, they are not ferroan anorthosites (because of their Mg′ and lower plagioclase abundance), and are quite distinct from the Apollo magnesian granulites, which have super-chondritic Th/Sm.

For the magnesian anorthositic granulite clasts studied here, Th was below detection limits and so the Th/Sm ratio could not be calculated. However, we can use La/Sm as a proxy for Th/Sm, because La and Th are both highly incompatible, and track each other closely in most igneous fractionations (i.e., Th/La should be approximately 1.0 × CI). This value is distinct from that of KREEP at Th/La ∼ 1.3–1.4 × CI, which is inferred to represent extreme silicate-melt fractionations (e.g., Heiken et al. 1991; Gnos et al. 2004).

The magnesian anorthositic granulites that have been analyzed for both Th and La have La/Th near the CI value (Figs. 7a and 7b; Goodrich et al. 1984). Thus, for the clasts analyzed here, we estimate the CI-normalized abundance of Th, Th*, as being identical to that La (Fig. 7a).

Thus, the criteria of Mg′ and Th/Sm versus Sm show that meteoritic magnesian anorthositic granulite has a distinct geochemistry, unlike those of: KREEP, the ferroan anorthosite suite, the magnesian-suite plutonics, the mare basalts, the bulk compositions of feldspathic lunar meteorites, and the Apollo magnesian anorthositic granulites (Fig. 7a).


Abundances of Ti, Sc, and Sm are also useful as discriminators among lunar rock types (Norman and Ryder 1980; Warren and Wasson 1980; James and Flohr 1983). Samarium acts here as a proxy for incompatible elements in general. Abundances of Ti are strongly affected by fractionation of ilmenite, both accumulation into a source region and crystallization from magma. In the absence of ilmenite, Ti is moderately incompatible like Sm; thus, fractionation of olivine, pyroxene, and plagioclase would have little effect on a magma’s Ti/Sm ratio. Scandium abundances are affected primarily by pyroxene fractionation; Sc is compatible in pyroxenes but incompatible in other basalt minerals.

The Ti/Sm ratio divides pristine highlands rocks into two groups (Fig. 8a; Norman and Ryder 1980; James and Flohr 1983). One group includes nearly all ferroan anorthosite suite, Mg-gabbronorites (61224,11; 67667; 67915), the dunite 72415, and several spinel-troctolites (67435, 73235); these rocks have Ti/Sm near that of CI chondrites, implying that their parent magmas and mantle sources were not affected by ilmenite fractionation. The second group of rocks includes KREEP-rich lithologies and the Mg-norites are characterized by low Ti/Sm ratios, which suggests that their sources were affected by fractionation and removal of ilmenite. It is possible that the Mg-norite plutonic rocks formed from primitive basaltic magmas that assimilated some KREEP, inheriting their high Mg′ from the basalt, and high REE abundances and low Ti abundances from the KREEP (Fig. 8a; James and Flohr 1983).

The graph of Ti/Sm versus Mg′ (Fig. 8a), shows that the analyzed clasts of magnesian anorthositic granulite are distinct from known pristine rocks. These meteoritic clasts have Ti/Sm like the ferroan anorthosite rocks and Mg-gabbronorites, but are more magnesian than either (Fig. 8b). The Mg-gabbronorites and the granulite clasts have similar Sc/Sm (Norman and Ryder 1980; Warren and Wasson 1980; James and Flohr 1983), and the former have far more augite than the latter. In Ti/Sm and Mg′, the only Apollo pristine rocks that are similar to the clasts of magnesian anorthositic granulite are several spinel troctolites (67435, 73235) and the dunite 72415. The clasts of magnesian anorthositic granulite are distinct, however, in having much higher Sc/Sm than the troctolites (Fig. 8b). In fact, dunite 72415 is the only pristine Apollo sample with Sc/Sm, Ti/Sm, and Mg′ near those of the magnesian anorthositic granulites.

Discussion, Part II: Global Distribution and Significance

As shown above, magnesian anorthositic granulite clasts from lunar meteorites represent lithologies or chemical compositions that are distinct from known pristine lithologies, and are not simple mixtures of known pristine lithologies or compositions. In this section, we show that magnesian anorthositic granulites are found in many meteorites, and thus are from varied locations across the lunar surface. The next paragraphs will demonstrate that (1) ALHA81005 and Dhofar 309 represent different sites on the lunar surface; (2) similar granulites are present in other lunar meteorites, and so must be present across much of the surface; (3) the compositions of the lunar highlands and mixed-composition lunar meteorites are reasonably represented as mixtures of mare basalts, ferroan anorthosite suite rocks, magnesian anorthositic granulite, and basalt (Fig. 6).

Pairing and Source-Pairing

The clasts of magnesian anorthositic granulite in ALHA81005 and Dhofar 309 are so similar in composition that one might ask if their host meteorites came from the same site on the Moon—i.e., ejected in the same impact event. The answer is almost certainly “no,” because the two meteorites were ejected from the Moon at different times, ALHA81005 at ∼60,000 yr ago (Eugster 2003), and Dhofar 309 at ∼300,000 yr ago (Nishiizumi and Caffee 2006).

Beyond this simple negation, bulk chemistry and clast proportions suggest that Dhofar 309 and ALHA 81005 are from different sites. Dhofar 309 is significantly richer in Al and Mg than ALHA81005 (Fig. 6), and has lower abundances of incompatible elements (Korotev 2006; Korotev et al. 2006). Dhofar 309 contains clasts only of ferroan anorthosites and magnesian anorthositic rocks (Takeda et al. 2006, 2007, 2008), while ALHA81005 contains these and many others (Treiman and Drake 1983; Goodrich et al. 1984).

Magnesian Anorthositic Granulite in Other Lunar Meteorites

Feldspathic lunar meteorites probably come from more than 20 sites on the Moon (Korotev 2005; Nishiizumi and Caffee 2006), and there is evidence that magnesian anorthositic granulite is present at most of them. Feldspathic lunar meteorites with reported clasts of magnesian granulites Mg′ > 75, include MacAlpine Hills (MAC) 81004/5 (Koeberl et al. 1991; Neal et al. 1991; Taylor 1991; Warren and Kallemeyn 1991), Queen Alexandra Range (QUE) 93069 (Koeberl et al. 1996); Pecora Escarpment (PCA) 02007 (Day et al. 2006; Korotev et al. 2006), Dhofar 025 (Cahill et al. 2004), Dhofar 081 (Cahill et al. 2004), Dar al Gani 400 (Semenova et al. 2000), Yamato-791197 (Lindstrom et al. 1986), Y86032 (Koeberl et al. 1990; Nyquist et al. 2006; Takeda et al. 2007), Y-983885 (Arai et al. 2005), and Sayh al Uhaymir (SaU) 300 (Hsu et al. 2008).

For some of these meteorites, available data suggest that their magnesian anorthositic granulite clasts are similar to those described here. Y-86032 has bulk REE abundances at ∼2–4 × CI (Eu at 10 × CI; Koeberl et al. 1990), and clasts of magnesian feldspar-rich material (Koeberl et al. 1990; Nyquist et al. 2006); if these clasts were REE-rich (like the Apollo magnesian granulites), it seems likely that the bulk REE abundances would not be so low. Similarly, bulk Y-791197 and bulk MAC 88104/5 are only slightly richer in REE than ALHA 81005 (Lindstrom et al. 1986; Koeberl et al. 1991), which suggests again that their magnesian anorthositic granulite clasts could not be highly enriched in the REE.

Magnesian Anorthositic Granulite as a Chemical Component

Magnesian anorthositic granulite lithologies may be significant components in mixed highlands-mare lunar meteorites. The bulk compositions and petrographies of these “mixed” meteorites imply that they are impact-generated mixtures of mare and highland components (Korotev et al. 2009). As shown in Fig. 6, the compositions of mixed lunar meteorites fall among the mixing lines between ferroan anorthosite, magnesian anorthositic granulite, and meteoritic lunar mare basalts (blue, black, and green lines of Fig. 6). In this model, a hypothetical magnesian peridotite of the lunar mantle (Warren 2005) could serve as well as magnesian anorthositic granulite (Fig. 6). However, no lunar meteorite composition requires admixture of peridotite rather than of magnesian anorthositic, because no meteorite compositions fall to the left of the mixing line between magnesian anorthositic granulite and mare basalts (Fig. 6).

In fact, several mixed-composition lunar materials can be represented as binary mixtures of magnesian anorthositic granulite and mare basalt (Fig. 6), including gabbronorites from ALHA81005 (Goodrich et al. 1984; Maloy et al. 2004; Treiman et al. 2005), and the meteorite Dhofar 961 (Korotev et al. 2007). Dhofar 961 is noteworthy because its source may be the South Pole/Aitken basin (Jolliff et al. 2007, 2008; Korotev et al. 2007). The compositions of Dhofar 961 and its mafic clasts fall essentially on the mixing line between magnesian anorthositic granulite and mare basalt (Fig. 6). If Dhofar 961 is from the South Pole/Aitken basin, then the mafic material in the basin is likely to be basaltic and not peridotitic (Lucey 2004). The lack of a significant ferroan anorthosite component in the Dhofar 961 composition (Fig. 6) suggests that the South Pole/Aitken area includes little ferroan anorthosite, as is the case for the Imbrium and Serenitatis basins (Spudis et al. 1991; Ryder et al. 1997).

Remote Sensing Evidence for Magnesian Anorthositic Granulite

The feldspathic lunar meteorites come from random sites in the lunar highlands (Warren 2005), most of which is in Feldspathic Highlands Terrane (FHT). Recently acquired remote sensing data from the Lunar Prospector spacecraft are consistent with a significant proportion of magnesian anorthositic granulite in the FHT.

The Lunar Prospector spacecraft acquired gamma-ray spectra from the lunar surface, and yielded partial chemical analyses at a coarse spatial scale (Lawrence et al. 1998). These data imply that the average composition of the FHT surface is ∼4.5 wt% FeO, ∼28 wt% Al2O3, and Th <1 ppm (Jolliff et al. 2000; Korotev et al. 2003; Gillis et al. 2004). This composition is consistent with a mixture of magnesian anorthositic granulite and ferroan anorthosite suite rocks, as are most of the feldspathic lunar meteorites (discussion above; Fig. 6; Takeda et al. 2006). Lunar Prospector did not determine abundances of Mg, so the Mg′ of the FHT is unconstrained.

The Clementine spacecraft acquired optical reflectance spectra of the lunar farside highlands in a several bands in the UV-NIR (McEwen and Robinson 1997); these data are also consistent with (but not definitive of) abundant magnesian anorthositic granulite in the lunar highlands. Tompkins and Pieters (1999) analyzed these reflectance data for central peaks of impact craters and basins; the rationale being that central peak rocks are samples of the lunar subsurface, now exposed at the surface. They found that much of the highlands crust is anorthositic (>90% plagioclase), that a large portion of it is anorthositic but with more mafic minerals (80–90% plagioclase), and that a smaller but still significant portion contains even more mafic minerals (60–80% plagioclase; anorthositic norites, gabbronorites, and gabbros). Tompkins and Pieters (1999) suggest that these latter rock types may reflect either a decrease in crustal plagioclase content with depth, or the presence of plagioclase-rich plutons analogous to those of the Apollo magnesian-suite. The reflectance data are not sensitive enough to Mg′ to distinguish between these possible origins. However, these anorthositic lithologies are broadly consistent with those of the magnesian anorthositic granulite clasts (Table 2), and are again consistent with magnesian anorthositic granulite being widespread in the highlands crust.

Discussion, Part III: Origin of Magnesian Anorthositic Granulite

The origins of magnesian anorthositic granulites and their magnesian anorthositic precursors are not clear. Although these materials are widespread across the lunar highlands, and abundant in some areas (like the Dhofar 309 source; Fig. 6), they are not directly addressed in current hypotheses of lunar evolution.

Magnesian Anorthositic Granulite Compared to Other Highlands Lithologies

From our chemical data, the meteoritic magnesian anorthositic granulites do not appear closely related to pristine lunar lithologies: ferroan anorthosite suite rocks, magnesian-suite plutonics (Mg-norite and Mg-gabbronorite), troctolites, dunites, nor mare basalts. They are most similar to some plutonic rocks of the Apollo magnesian suite (spinel troctolites and Mg-gabbronorites), but are geochemically distinct.

The clasts of magnesian anorthositic granulites are similar to rocks of the ferroan anorthosite suite in containing abundant calcic plagioclase, and having key element abundance ratios near chondritic (Th/Sm, Ti/Sm and Sc/Sm; Figs. 3, 4, 6, 7, and 8). However, magnesian anorthositic granulite cannot form in the same manner as ferroan anorthosites within magma ocean models of the lunar petrogenesis. Ferroan anorthosites are inferred to be remnants of plagioclase-rich floatation cumulates, “rockbergs,” that accreted at the surface of a highly differentiated, Fe-rich magma ocean (e.g., Warren 1985; Shearer and Papike 1999). Probably, the precursors for magnesian anorthositic granulites could not form in this process, because a suitably magnesian parental magma would not have been dense enough to float its plagioclase.

The other major non-mare rock types are the magnesian-suite plutonics: Mg-norites, Mg-gabbronorites, and the troctolite/dunite group (James and Flohr 1983; Hess 1994; see Figs. 3 and 6), which are inferred to represent melts from magma ocean cumulates (e.g., Warren 1985; Shearer and Papike 1999). Magnesian anorthositic granulites are distinct from these lithologies. The Mg-norites have higher REE abundances, are LREE-enriched, and have lower Ti/Sm and Sc/Sm (Figs. 5, 7, and 8). The Mg-gabbronorites contain less olivine, more high-Ca pyroxene, and are generally less magnesian (Table 2, Fig. 7; James and Flohr 1983). Magnesian anorthositic granulites are similar to the Apollo troctolites in most respects (Mg′, mineral proportions, Th/Sm, and Ti/Sm; Table 2, Figs. 7 and 8). However, the clasts of magnesian anorthositic granulites have Sc/Sm much higher than do the Apollo troctolites (Fig. 8b), reflecting the latters’ kinship to KREEP. The only lithology with Mg′ and trace element ratios comparable to those of the meteoritic magnesian anorthositic granulites is the unique, and petrologically unrelated, dunite 72415.

Magnesian Anorthositic Granulites Derived from Post-LMO Plutons

If the precursors of magnesian anorthositic granulites could not form directly from a magma ocean (like ferroan anorthosites), they can perhaps be ascribed to post-lunar magma ocean (LMO) plutons that lacked a significant KREEP component (Figs. 6, 7, and 8). It has not been clear if KREEP is an essential ingredient in formation of post-LMO plutons (with its K, U, and Th providing the heat for remelting LMO cumulates) or whether KREEP is a passive “passenger,” assimilated and transported by magmas that were originally KREEP-free (e.g., Shearer and Papike 1999; Korotev and Jolliff 2001; Korotev et al. 2003). If the clasts of magnesian anorthositic granulites represent relics of post-LMO plutons (granulated, partially melted, and/or metamorphosed), then it seems reasonable that KREEP was not an essential factor in their development. In this scenario, post-LMO plutons may or may not contain KREEP, depending on whether or not they passed through the northern nearside of the Moon, where KREEP appears to be concentrated. To test this hypothesis, one would want to examine true plutonic rocks from KREEP-free regions of the lunar crust.


The clasts of magnesian anorthositic granulite from lunar meteorites are distinct from those of known lunar pristine materials, and cannot be represented as mixtures of known pristine materials. Thus, magnesian anorthositic granulites must represent or include a chemical component that is not represented by known pristine samples (Korotev et al. 1983a, 1983b; Korotev and Jolliff 2001).

In major element composition mineralogy, the meteoritic magnesian anorthositic granulites are similar to the Apollo magnesian anorthositic granulites and to magnesian-suite plutonic troctolites. These granulite clasts are clearly distinct from anorthosites, Mg-norites, Mg-gabbronorites, and dunites. In trace element (REE) composition, the meteoritic magnesian anorthositic granulites are most similar to ferroan anorthosites. They are clearly distinct from: Apollo magnesian granulites in having lower REE abundance and no LREE enrichment; Mg-norites in Th/Sm, Ti/Sm, and Sc/Sm; and from spinel troctolites in Sc/Sm.

Among these choices, meteoritic magnesian anorthositic granulites are most similar to the lunar troctolites and the Mg-gabbronorites, both of which are inferred to come from basaltic plutons emplaced after solidification of the lunar magma ocean (e.g., Shearer and Papike 1999; Shearer et al. 2006). Thus, it seems reasonable to infer that magnesian anorthositic granulites represent rock derived from similar intrusions, although not of the same chemistry as the plutons sampled in the Apollo and Luna collections.

Magnesian anorthositic granulites are found in several distinct lunar meteorites, and so may represent a widespread lithology on the moon. The meteorites studied here, ALHA81005 and Dhofar 309, were ejected from the Moon in different impact events and thus sampled different places on the moon. Magnesian anorthositic granulites are present in other lunar meteorites, and bulk compositions of highlands lunar and mixed-composition lunar meteorites (Korotev et al. 2009) can be represented reasonably as mixtures of ferroan anorthosite suite rocks, magnesian anorthositic granulite, and mare basalt. In addition, remote sensing evidence suggests that magnesian anorthositic granulites are present across much of the lunar highlands.

Thus, inferences re-enforce and expand on those of earlier works: these granulite clasts represent a distinct group of lunar materials, and their compositions are not merely mixtures of known lunar pristine materials (e.g., Korotev et al. 1983a, 1983b; Lindstrom and Lindstrom 1986; Korotev and Jolliff 2001; Korotev et al. 2003). Quoting Korotev et al. (1983a), “. . . the early lunar crust contained . . . a significant proportion of both ferroan and magnesian anorthositic norites. Such a conclusion is at variance with models that treat materials of anorthositic norite composition as mixtures of anorthosite plus norite, troctolite, and dunite. Once again, sampling of a new . . . lunar site has brought evidence for new rock types, emphasizing both the variety of compositions of endogenous igneous rocks and the incomplete nature of the sampling of Moon’s surface.” Our work and that of others (e.g., Takeda et al. 2006, 2007, 2008) shows that magnesian anorthositic compositions are common across the lunar highlands and must be considered in interpreting remote sensing data. The origins of magnesian anorthositic granulites, magnesian anorthositic granulites, and their precursor lithology (or lithologies) remain uncertain, pending acquisition of new meteorite samples, new analyses of those and samples in hand, new remote sensing data from the Moon, and (most definitively) returned lunar samples.


This research was inspired by R. Korotev (Washington University), and encouraged by D. Lindstrom and M. Lindstrom. Dr. M. Nazarov of the Vernadsky Institute of Geochemistry and Analytical Chemistry, Moscow graciously loaned the thin section of Dhofar 309. The Meteorite Curatorial Facility of Johnson Space Center, Houston loaned the ALHA81005 section (with thanks to the Meteorite Working Group and K. Righter, curator). R. Korotev also provided us his unpublished compilation of lunar meteorite analyses. Assistance with EMP analyses and mapping came from C. Schwandt (Jacobs Sverdrup), and assistance with SIMS analyses came from Paul Burger (University of New Mexico). We are grateful for careful reviews by T. Arai, J. Hudgins, and R. Korotev. Maloy was supported by a graduate fellowship from the Lunar and Planetary Institute. Treiman’s and Gross’s participations were supported through the NASA CAN with the Lunar and Planetary Institute, and the NASA Cosmochemistry Program (grant NNX08AH78G). Shearer’s participation and use of the UNM SIMS facility were supported by grants from the NASA Cosmochemistry Program. Lunar and Planetary Institute Contribution #1514.

Editorial Handling

Dr. Randy Korotev


Siderophiles in Metal and Sulfides

In the clasts of magnesian granulite studied here, metal and sulfide have such low abundances that they must contribute insignificantly to the bulk Ni and Co abundances of the clasts. For instance, consider clast D2, which contains 0.001% metal and 0.003% troilite (FeS) by volume. If that metal were pure Ni, it would contribute ∼25 ppm Ni to the bulk clast’s abundance. However, the metal is mostly Fe (we lack quantitative analyses). If it is like metal in ordinary chondrites, 6–7% Ni (Brearley and Jones 1998), it would contribute only ∼2 ppm Ni to the bulk. Similarly, if the sulfide in D2 were pure NiS (which it is not), it would contribute only ∼40 ppm Ni to the bulk; if the sulfide is like that in ordinary chondrites, with Fe/Ni >100 (Brearley and Jones 1998), then it would contribute negligible Ni to the bulk. Cobalt is distinctly more lithophile than Ni, so that even more of it will be in the silicates than the metal or sulfide.