Extent of the Hadley-Apennine Igneous Province
The Apennine Bench Formation is extensive, surrounding the Apollo 15 landing site (Hackman 1966; Fig. 2). Photogeological studies indicate that the ABF represents lava plains (e.g., Hawke and Head 1978; Spudis 1978; Blewett and Hawke 2001) and orbital measurements show that it has a composition like that of KREEP basalts (Spudis and Hawke 1986). Considering the abundance of KREEP basalts in 15205, and the inferred stratigraphy of the Apollo 15 region (KREEP basalt beneath the maria; Spudis et al. 1988), it is likely that KREEP basalts of the Apennine Bench Formation underlie the mare basalt lava flows of Palus Putredinis. Ejecta from Autolycus and Aristillus have Th concentrations in the range of KREEP basalts (Blewett and Hawke 2001), suggesting that KREEP basalt underlies those craters. The high Th concentrations could reflect the generally high concentrations of incompatible elements in the Procellarum-KREEP Terrane, but the presence of KREEP basalts in impact melt breccias and soils collected on the Apennine Front and at Station 4 (Fig. 1) suggest that basalts were abundant in the target areas for Autolycus and Aristillus.
Our estimate of the minimum extent of the Apennine Bench Formation and associated KREEP basalts is given by the dashed line in Fig. 2, an area of about 105 km2. The thickness of basalts throughout this large area is uncertain, particularly because there is no photogeological evidence for the source vents or number of separate flows from them. On the basis of geological observations, Spudis et al. (1988) estimated that the Apennine Bench underlying the Apollo 15 mare basalts could be as much as 200 m thick; we take 200 m as an upper limit for the mean thickness. Considering the wide area apparently covered by KREEP basalts, it seems likely that the KREEP lavas are at least 10 m thick and probably thicker. These limits imply a total volume of 1 × 103–2 × 104 km3. For comparison, the well-developed Erastothenian flows in Mare Imbrium have a volume of 4 × 104 km3 (Schaber 1973) and Kilauea volcano has a volume of 2 × 104 km3 (Bargar and Jackson 1973).
KREEP Basalt Bulk Chemical Systematics
To put our new mineral compositional data in context, we summarize the chemical variations among Apollo 15 KREEP basalt samples. Bulk compositions have been studied previously (e.g., Rhodes and Hubbard 1973; Basu and Bower 1976; Ryder 1976, 1987; Irving 1977; Lindstrom et al. 1977; Warren et al. 1983; Ryder et al. 1988; Simon et al. 1988; Ryder and Sherman 1989; Ryder and Martinez 1991). We summarize available data on bulk compositions, including compositions of glass in 15358 and related samples (our data and those from Ryder 1988), fractional crystallization products (quartz monzodiorites, QMD, and related), and experimental data (Hess et al. 1978; Rutherford et al. 1980). A problem with bulk data on KREEP basalts is that the samples are all small; the largest sample, 15386, had a mass of only 7.5 g. Thus, analyses have been performed on small chips, potentially leading to large sampling errors. The QMD samples are even coarser grained than the basalts, so have the largest sampling problems. Furthermore, many analyses of basalt and QMD samples are broad-beam microprobe analyses, which sample only one plane of small fragments, hence are subject to large sampling errors, in addition to uncertainties in the microprobe correction procedures for polymineralic materials. Nevertheless, broad chemical variations are discernible and informative.
KREEP basalt samples show increasing P, K, and Ti with decreasing Mg# (Fig. 7). The TiO2 trend has more scatter, consistent with the range of magma compositions inferred from pyroxene data (Fig. 6C), although it is likely that sampling causes some of the scatter. Expected compositional variations in the three elements as KREEP basalts fractionate are illustrated by a natural experiment: the compositions of yellow glasses in the incompletely crystallized samples in 15358 (Table 5). Pyroxene and plagioclase crystals are zoned in these samples, suggesting that the glass compositions reflect fractional crystallization trends. These trends show uniformly increasing concentrations of P2O5, K2O, and TiO2 with decreasing Mg#, hence with crystallization. Although small amounts of ilmenite have crystallized in these incompletely crystallized basalts, the TiO2 concentrations do not record ilmenite crystallization, even though TiO2 reaches 5 wt% in the residual magma.
Figure 7. Bulk rock chemical compositions of KREEP basalt and quartz monzodiorite (QMD) clasts (sources below), yellow glass in quenched rocks from 15358 (Ryder 1988; and our new data), and experimental data that approximate equilibrium liquid lines of descent for 15382 (Hess et al. 1978) and 15386 (Rutherford et al. 1980). KREEP basalt and QMD data from: Rhodes and Hubbard (1973), Basu and Bower (1976), Irving (1977), Lindstrom et al. (1977), Warren et al. (1983), Simon et al. (1988), Ryder et al. (1988), Papike et al. (1998), Ryder and Sherman (1989).
Download figure to PowerPoint
The liquid line of descent of a magma with the composition of 15386 (Rutherford et al. 1980) peaks at a TiO2 concentration of 4.9 wt% (at 1086 °C), then declines (Fig. 7C), clearly indicating ilmenite crystallization. Magma with the composition of KREEP basalt 15382 reaches a peak in TiO2 at 1080 °C (Hess et al. 1978). Although the experimental data on the liquid line of descent were collected at conditions closer to equilibrium than was the case for the yellow glass mesostases in the quenched basalts, both show generally the same trends of increasing concentrations of incompatible elements with crystallization. The experimental melts reach silicate liquid immiscibility at 1014 °C (15386) to 1035 °C (15382), shown in Fig. 7 by the dashed lines joining two points. One melt is enriched in SiO2 and K2O, the other in P2O5 and TiO2 (and in FeO, although the Fe/Mg ratio is not strongly affected by the immiscibility). These data have implications for the genesis of KREEP differentiates, as discussed below.
The QMD samples, generally considered to be formed by fractional crystallization of KREEP basalt magmas (e.g., Ryder and Martinez 1991), deviate from the general compositional trends (Fig. 7). A portion of the variation may be caused by nonrepresentative sampling, but probably not all. For example, the data show a general increase in K2O and P2O5 with decreasing Mg#, consistent with fractional crystallization. On the other hand, P2O5 is low in some samples and TiO2 is systematically low in the samples. These trends hint at fractionation of phosphate minerals and ilmenite. We discuss the possible causes of these compositional variations below.
Mineral Compositions: Partial Melting and Fractional Crystallization
The cores of orthopyroxene and plagioclase crystals in KREEP basalts record the compositions of the magmas in which they crystallized. The rapid cooling of lava flows prevents significant elemental diffusion. Minor and trace element concentrations (such as Cr, Al, and Ti in pyroxene and Fe in plagioclase) might be affected by lava cooling rates and crystal growth rates, but Watson (1996) argues that this is not likely to happen in naturally occurring basalts. However, Grove and Bence (1977) found correlations between Cr, Ti, and Al concentrations in low-Ca clinopyroxenes and cooling rates in an experimental study of low-Ti quartz-normative mare basalts. These minor elements are subject to mutual substitutions during crystallization, thus complicating assessing the compositions of basalt parent magmas. In general, the substitutions increase with increasing cooling rates, as shown by largest concentrations of all three elements being found in the first pyroxenes to crystallize in the most rapidly cooled experimental samples. Our data (Fig. 6A–C) correspond to the first pyroxene to crystallize in each KREEP basalt clast studied, but we find no correlation between concentrations of Cr, Ti, and Al with plagioclase grain size, a rough monitor of cooling rate (Table 4). Thus, we conclude that our measurements (Fig. 6) of minor elements in pyroxene and plagioclase provide useful information about the parent magmas of each basalt sample.
Our mineral data and published bulk compositions of KREEP basalts suggest a relatively small range in Mg#, a point made previously by Ryder (1988). The Mg# of the most magnesian pyroxene ranges from 85.5 to 77.1 (Fig. 6). We can estimate the corresponding Mg# in the magma by using the orthopyroxene-melt partition coefficient. Using Bédard’s (2007) parameterization of a large database of experimental data, we use a value of 0.28, corresponding to a melt MgO concentration of 8 wt% (the partition coefficient varies with MgO content of the melt). This partition coefficient and the observed range in mineral composition translate to a range in Mg# in the magmas of 62–49. Bulk compositions of KREEP basalt fragments suggest a somewhat greater range in magma Mg# (Fig. 7), from 73.3 to 34.6. However, all but two of the 43 samples plotted are in the range from 73.3 to 50.2, suggesting that the most evolved KREEP basalt fragments might be related to the more fractionated quartz monzodiorites (see next section). Dropping the two samples with the lowest bulk Mg#, we calculate that the most magnesian pyroxenes in these fragments would have Mg# in the range 90–79. The most magnesian pyroxene Mg# is higher than the most magnesian orthopyroxene core we measured (85.5, Fig. 6), but in reasonable agreement considering that the bulk compositional data were not obtained on the same set of KREEP basalt fragments as were our mineral data. The important point is that KREEP basalts have a limited range in Mg#, indicating small amounts of fractional crystallization or small differences in partial melting of mantle source regions. If the variation was caused by fractional crystallization, it almost certainly took place in subsurface magma chambers because plagioclase and pyroxene do not readily separate inside small basaltic magma bodies such as lava flows (Mangan and Marsh 1992). Note that KREEP basalts have similar major element compositions to terrestrial basalts, hence similar initial viscosities, so the analogy to terrestrial basalts is valid. This is completely different from lava flows with a long interval (10–100 °C) of crystallization of pyroxene or pyroxene and olivine, such as mare basalts, most Martian meteorite, and terrestrial highly mafic flows.
Minor elements in plagioclase show clear and logical trends with increasing Ab content (Fig. 6D and 6E). The strong correlation with K2O most likely reflects crystal-chemical control, but it is consistent with fractional crystallization. FeO also increases with Ab content, thus recording the increase in Fe and Na with crystallization. There is a weak correlation between Ab in plagioclase and Mg# in the most magnesian pyroxene (Fig. 6F), consistent with fractional crystallization, but the significant variation in Ab at a given Mg# indicates that the magmas must vary in their initial Mg/Fe, Na/Ca, and the extent to which plagioclase and pyroxene co-crystallized or one preceded the other.
The main characteristic of our mineral data set is the large ranges in minor element concentration for a given Mg# in pyroxene or Ab content of plagioclase (for FeO), well outside the range expected for analytical uncertainties (±0.05 wt%, 2-sigma). Concentrations of these elements are affected by both partial melting and fractional crystallization, but fractionation of a single magma ought to have produced less scatter, as shown by the data for clast #18 in 15205 (Fig. 6A–C). Thus, we suggest that the lack of strong correlation with the fractionation parameters Mg# and albite content indicates differences in source region compositions and percentage of partial melting. Such differences are not necessarily large. All sources might have consisted of orthopyroxene and plagioclase with or without olivine, and variable amounts of ilmenite, phosphates, and other minerals. Small amounts of ilmenite in the source might have caused differences in TiO2; exhaustion of ilmenite during partial melting would then dilute an initially higher TiO2 content to a lower one. A similar argument can be made for Cr2O3, with its variability caused by exhaustion of olivine in the source region as the amount of partial melting increased (Cr partitions strongly into olivine compared to silicate melt).
The factor-of-two variation in Al2O3 in the most magnesian pyroxenes (Fig. 6B) is surprising in light of bulk rock Al2O3 ranging from 15 to 20 wt%, with most values in the range from 15 to 18 wt% (Fig. 7). Orthopyroxene Al2O3 concentrations might reflect crystal growth factors, although Al2O3 does not correlate with grain size. Higher Al2O3 in pyroxene might instead indicate pyroxene crystallization before plagioclase, when Al2O3 in the lava was highest. In turn, this may reflect different proportions of plagioclase and orthopyroxene in the source region. Different amounts of partial melting might have led to total melting of plagioclase in the source, and subsequent decrease in Al2O3 in the magma as melting continued. Whatever the details, the most straightforward explanation for minor element variations in pyroxene and plagioclase is that the source region varied in composition, although not enough to lead to drastically different magma compositions: they are all still KREEP basalts.
Differing source compositions are consistent with the weak correlation between albite in the most calcic plagioclase and the Mg# of the most magnesian pyroxene (Fig. 6F). For a given Mg# in pyroxene, co-existing plagioclase ranges in Ab, typically 4 mol%, but up to 6 mol%. Compositions of a suite of basalts formed by partial melting of a common source region ought to show decreasing Ab with increasing Mg#, as the data suggest, but the scatter also indicates variation in the source composition.
Formation of the KREEP Basalt Differentiates
Ryder (1976), Irving (1977), and Ryder and Martinez (1991) drew attention to extensively fractionated samples from Apollo 15. As discussed in the previous section, the relative compositional variation among KREEP basalts might be caused by differences in the amount of partial melting. On the other hand, the samples loosely classified as quartz monzodiorites (QMD) clearly show that extensive fractional crystallization must have operated: they have much lower Mg# than do the basalts, and in general, elevated incompatible element concentrations (Fig. 7). Here, we use published experimental data (Hess et al. 1978; Rutherford et al. 1980) and yellow glasses in 15358 (Ryder 1988; and our data) to track likely compositional trends for fractional crystallization (Fig. 7).
The yellow residual glasses in KREEP basalt clasts in 15358 represent a closer approach to fractional crystallization than do the close-to-equilibrium experiments reported by Hess et al. (1978) and Rutherford et al. (1980). The yellow glasses show continuous increase in the incompatible elements K and P, not surprising for a crystallizing system. Titanium (TiO2) also increases with crystallization (decreasing Mg#), indicating that no significant ilmenite crystallized. In fact, small amounts of ilmenite are present in the basalt clasts in 15358, indicating ilmenite saturation, but not enough crystallization to lead to a measurable decrease in Ti in the residual melt (hence in the preserved glass). In the experiments conducted with a starting composition similar to 15386, ilmenite crystallizes when the TiO2 concentration in the melt reaches about 5 wt%, it eventually reaches a point of silicate liquid immiscibility when the Mg# is less than about 10. The 15382 composition probably also saturates in ilmenite at roughly 5 wt% TiO2, but reaches immiscibility earlier, when the Mg# is about 25. None of the yellow glasses contain blebs of high-Fe and high-Si immiscible melts as present in the Hess et al. (1978) and Rutherford et al. (1980) experiments.
The quartz monzodiorite does not lie on the trends indicated by the yellow glasses or the experiments (Fig. 7). Ignoring the sampling problems in analyzing small, relatively coarse-grained rock fragments, the QMD compositions are most consistent with fractionation of ilmenite and phosphate minerals, as shown by their low P2O5 contents compared with the yellow glass and experimental trends (Fig. 7). Potassium (K2O) shows both enriched and depleted concentrations, suggesting silicate liquid immiscibility. Lunar samples record immiscibility in action in an Apollo 14 QMD rock fragment (Jolliff et al. 1999) and in an Apollo 12 felsite (Warren et al. 1987) in which phosphate crystallization preceded immiscibility. However, P2O5 and TiO2 ought to be enriched in the Fe-rich melt, not depleted. The depth at which fractionation took place may play a role in the path taken by KREEP basalt magmas: Rutherford et al. (1996) showed that silicate liquid immiscibility is suppressed at 3 kbar (corresponding to a depth of about 60 km). Thus, if the magma bodies that produced the QMDs were deep, liquid immiscibility may not have occurred. Alternatively, the differences in element concentrations could be caused by fractional crystallization, but of different starting magmas, one low in K2O and another higher, hence leading to a range in K2O concentrations of the QMD.
The fundamental point is that the QMD is clearly more evolved than are the KREEP basalts, indicating fractional crystallization. The presence of fine exsolution lamellae in the pyroxenes in all QMD fragments is indicative of formation in a small magma body (Ryder and Martinez 1991). The cooling rate of 0.01 °C/yr determined for Apollo 14 sample 14161,7373 (Jolliff et al. 1999) indicates crystallization in a sill at least 1 km thick. If applicable to Apollo 15 QMD samples, the magmatic system must include km-sized sill-like magma bodies. All QMD pieces have been found in the samples collected on the Apennine Front (impact melt breccia 15405 and soil 15434), indicating that the region excavated by Aristillus and/or Autolycus was the probable site of both KREEP basalt volcanism and subsurface magma bodies.
A Long-Lived Igneous Province or Two Episodes of Magmatism?
Only three Apollo 15 KREEP basalt samples have been dated, 15382, 15386, and 15434,73 (a fragment from coarse-fines soil 15434,4). Rb-Sr ages, updated using a 87Rb decay constant of 1.402 Ga−1 (Begemann et al. 2001), of the three samples are 3.87 ± 0.05 Ga (15382; Papanastassiou and Wasserburg 1976), 3.88 ± 0.05 Ga (15434,73; Nyquist et al. 1975) to 3.91 ± 0.04 Ga (15386; Nyquist et al. 1975). Ryder (1994) suggests that all three fragments date a single event (eruption of basaltic lava), which he estimated as occurring at 3.89 ± 0.02 Ga. This is almost certainly the age of the Apennine Bench Formation, which clearly postdates formation of the Imbrium Basin (Spudis and Hawke 1986; cf. Deutsch and Stöffler 1987). At the other end of the age spectrum, the QMD from 15405 has a U-Pb age of 4.30 Ga (Meyer et al. 1996). Because the QMD formed by fractionation of a KREEP basalt magma, at least some KREEP volcanism must substantially pre-date formation of the Imbrium Basin. The old age for the QMD shows that the epoch of KREEP basalt volcanism was either long or pulsed with one event at 4.3 Ga and another at 3.9 Ga. The old QMD age raises the question of the original pre-Imbrium location of the small magma body (or bodies) in which quartz monzodiorites formed.
Role of the Imbrium and Other Large Impacts
As suggested by Spudis (1978) and evaluated in detail by Ryder (1994), formation of the KREEP basalts that compose the Apennine Bench Formation may have been triggered by the formation of the Imbrium Basin. Ryder (1994) notes that the Apollo 15 KREEP basalts are 3.89 ± 0.02 Ga old (age corrected for a different 87Rb decay constant, as noted above) and that the age of the Imbrium Basin is indistinguishable from that age, although the basalts clearly formed after the basin was created. Ryder (1994) argued that large amounts of partially melted rock existed in the lower crust or upper mantle, kept above the solidus by the high concentrations of the heat-producing elements K, Th, and U present in KREEP. The Imbrium impact thus could have released magma already existing, but too dense or too viscous to migrate through the low-density crust. The presence of melt at depth in the Procellarum-KREEP terrane is supported by the thermal modeling of Wieczorek and Phillips (2000).
Elkins-Tanton et al. (2004) modeled post-impact melting processes and aside from the production of impact melt, identified two main sources of melting. One form is pressure-release melting beneath the basin, which effectively moves the solidus deeper into the Moon, inducing a zone of melting. This effect is more efficient if the material is already near or above the solidus, as Ryder (1994) argued. Pressure-release melting is prompt and indistinguishable in time from formation of the basin. The second melting process results from convection induced by the formation of the basin; this causes pressure-release melting of deep mantle rock as it rises adiabatically. This phase could have lasted a few hundred million years and may have led to eruption of mare basalt magmas.
The KREEP basalts that are possibly related to the formation of the Imbrium Basin are only those present within the Apennine Bench Formation. The age of the QMD in impact melt rock 15405 implies a pre-Imbrium age of 4.3 Ga for that episode of KREEP magmatism. Thus, although not yet dated directly, the KREEP basalt clasts within 15405 may be significantly older than the 3.89 Ga age of the KREEP basalts of the Apennine Bench Formation. Besides showing that KREEP volcanism began early in lunar history, this inferred older episode provides constraints on the formation of the Mg- and alkali-suites of nonmare rocks, assuming that they are related to KREEP magmatism. Mg-suite and alkali-suite rocks, and zircon analyses from breccias, show a range in age from about 4.49 Ga to 3.9 Ga (Nemchin et al. 2008, 2009; Edmunson et al. 2009; Taylor et al. 2009; samples are mostly from the Apollo 14 and 17 sites). The zircon data set shows peaks in the distribution, suggesting specific epochs of increased rates of magmatism. Given the possible relation between emplacement of KREEP basalts of the Apennine Bench Formation and the Imbrium Basin-forming impact, these peaks in magmatic activity similarly might be caused by large impacts, as argued by Nemchin et al. (2008), although episodic volcanism caused by internal processes, such as heating in the mantle and mantle dynamics, cannot be ruled out.
Genesis of the Basalts Composing the Apennine Bench Formation
Both mineral and bulk compositions indicate that the KREEP basalt source region was highly magnesian, with Mg# in the range 80–90. This relation reflects a long-standing paradox of KREEP basalt—magnesian magma rich in incompatible trace elements. The KREEP basalt source regions were probably dominated by orthopyroxene and plagioclase (e.g., Ryder 1987). The high Mg# suggests that the source region for Apollo 15 KREEP basalts consisted of magnesian, low-density mafic minerals formed relatively early during crystallization of the lunar magma ocean, but rose to the upper mantle (possibly the lower crust) by overturn of the mantle as denser minerals richer in Fe sank. Such a magnesian source would produce magnesian magmas poor in trace elements unless it was contaminated by late-stage fractionates rich in incompatible elements (e.g., urKREEP; Warren 1986, 1988). Such contamination, however, could have occurred by sinking of urKREEP and dense, mafic silicates and ilmenite, followed by partial melting (e.g., Shearer et al. 1991). Straightforward sinking of solid masses of rock is unlikely (Elkins-Tanton et al. 2002), so perhaps impacts could enhance the process by initiating overturn or even driving shallow materials deeper into the Moon. On the other hand, a problem with such hybrid sources is that a range of partial melting, as seems to be called for by variations in the compositions of KREEP basalt magmas, would result in fractionation among the REE and of Rb/Sr (e.g., Warren and Wasson 1979; Ryder 1994), which is not observed among the KREEP basalts. An alternative to hybrid mantle sources is the interesting idea (Elardo et al. 2011) that rising diapirs of magnesian early magma ocean cumulates mixed in the solid state with urKREEP and plagioclase-rich crust, then eventually partially melt to form the parent magmas of Mg-suite rocks and KREEP basalts. In the diapir-crust mixing model, large impacts might have aided the mixing of materials to produce KREEP basalt source regions. In both assimilation and diapir-crust mixing, KREEP basalt and Mg-suite magmas form close to the crust-mantle boundary from sources that were dominated by olivine, orthopyroxene, and plagioclase, but probably varied in abundances of minor minerals. The extent to which large impacts played a role in mixing materials to make source regions or in causing pressure-release melting is unknown.