Geochimica el Cosmt~'himica Atla Vol. 53.pp. 529-541 0016-7037/89/$3.00 + .00 Copyright © 1989 Pergamon Press plc. Printed in U.S.A. Metasomatic products of the lunar magma ocean: The role of KREEP dissemination CLIVE R. NEAL and LAWRENCEA. TAYLOR Department of Geological Sciences, University of Tennessee, Knoxville, TN 37996, U.S.A. (Received July 6, 1988; accepted in revised form December 6, 1988) Abstract--The orion of the incompatible element-rich lunar component, KREEP, is in the Lunar Magma Ocean (LMO). The fractionated residual melt after crystallization of the LMO represents "urKREEP" (after WARRENand WAS,SON, 1979). The percentage of this residual melt is low enough to be within the realm of silicate liquid immiscibility (SLI). This process has the ability to split the KREEP signature into K- and REEP-Fractions, which are manifest as lunar granite (K) and phosphate phases present in highland litholngies or as quartz ferrotroctolite in lunar soils (REEP). We envisage this as a localized, but significant process since only a small portion of urKREEP undergoes SLI. Norms of experimental and Apollo 15 basaltic immiscible glasses suggest that the REEP-Fraction found in the lunar highlands has undergone post-SLl fractionation of at least fayalite. This significantly reduces the density of the REEP-Fraction and coupled with its low viscosity (10-15 poise), it can percolate upward, metasomatizing the lunar crust. The higher viscosity of the granitic melt (~30000 poise) inhibits its mobility, and it forms "pods" in the lower crust (as required for VHK basalt petrogenesis). With the identification of KREEPy components, the composition of urKREEP can be calculated. Using experimental evidence, the KREEP components may be recombined to give the pre-SLI composition, or liquid-liquid Kd's can be used to calculate a pre-SLl composition from lunar granite. Both calculated urKREEP compositions are lower in MgO and A1203 and higher in FeO and P2Os than reported low- and high-K KREEP compositions. The calculated REE abundances are higher and the REE profiles are slightly more LREE-enriched than the previously reported KREEP compositions. However, the REE profile calculated using liquid-liquid Kd's is concave-upwards, compared to the LREE-enriched profile produced by recombining the KREEP components. Pre-SLI whitlockite/apatite fractionation occurred prior to immiscibility in order to generate the concave-upwards profile. The LREE-enriched profile represents the pre-SL! magma prior to phosphate fractionation. The presence of"superKREEPy" rock types can be accounted for by K- and REEP-Fraction assimilation, and the compositions of olivine vitrophyres can now be modeled using analyzed lunar rock types without the inference of a mythical "high-Mg" component. Other, more widespread KREEPy rocks (e.g., Apollo 14 breccias, Apollo 15 KREEP basalts, LKFM basalts) are produced by incorporation of the more widespread pre-SLI urKREEP component. The splitting of KREEP into identifiablelithologlcalcomponents allows the petrogenesisand role of KREEP in lunar evolution to be better understood. INTRODUCTION THE PETROGENESIS OF a widespread lunar component which is enriched in incompatible elements has been the subject of lively debate since the early 1970's. This component was first identified by MEYER and HUBBARD (1970) in Apollo 12 soils. Subsequently, it was given the acronym "KREEP" (Hun- BARD et al., 1971; MEYER et al., 1971) to account for its elevated K, rare-earth element (REE), and P contents. Oc- currences of KREEP have been documented from every landing site (e.g., APOLLOSOlE SURVEY, 1971). Furthermore, a KREEPy component has been found in basalts (e.g., IRV- ING, 1977; RYDER, 1987; SALPAS eta[., 1988; NEAL et al., 1988a) and breccias (e.g., McKAY and WEILL, 1977; SALPAS et al., 1987), as well as soils (e.g., MORRIS et al., 1977; KO- ROTEV, 1981). It is interesting to note that a terrestrial ana- logue to KREEP may be found in kimberlites, which are considered to be the products of mantle melting and meta- somatism (e.g., WYLLIE, 1980). There are two basic models proposed for the generation of KREEP, based upon the Lunar Magma Ocean hypothesis (LMO): 1) partial melting of an "ANT-suite" (Anorthosite, Norite, Troctolite) cumulate (e.g., WALKER el a[., 1972; PRINZ et al., 1973; NAVA and PHILPOTTS, 1973); or 2) as a residual magma after extreme fractional crystallization (e. g., DOWTY et aL, 1976; SHIH, 1977; WARREN and WASSON, 1979; WARREN, 1985). AS an offshoot of model 2), CRAW- FORD and HOLL[STER (1977) suggested that the residual magma underwent liquid immiscibility, forming granitic and 529 high-Fe, KREEPy melts. WARREN and WASSON (1979) demonstrated that KREEP-normalized trace element patterns for a variety of"KREEPy" materials from different localities were essentially flat (i.e., constant trace element ratios). They argued that this indicated a uniform, almost Moon-wide KREEP reservoir and concluded that ICREEP could not be produced by partial melting. Based on this evidence, WARREN and WASSON (1979) argued for KREEP to be the residual from the LMO. They used the German prefix "ur" (meaning primeval) to describe KREEP produced in this way. However, while we accept the essential features of WARREN and WASSON'S(1979) model for urKREEP petrogenesis, we do not believe that it goes far enough in explaining many KREEP-related phenomena. For example, the geochemical modelling developed to account for the observed composi- tions of soils, breccias, and olivine vitrophyres (impact melts) requires the mixing of three major components by meteorite impact: anorthosite, KREEP, and a high-Mg rock-type (e.g., WXNKE et al., 1976; WASSONet al., 1977; ALLEN et al., 1979; KOROTEV et al., 1980; SHERVAIS et al., 1988). It is well es- tablished that the lunar highlands are dominated by anor- thosite, troctolite, and norite, but it appears that these fith- ologies cannot account for the high MG# of soils, breccias, and especially olivine vitrophyres (e.g., WASSON el all., 1977; SHERVAISel al., 1988). This is because 50 to 90% ofa KREEP component is required by such models in order to generate the observed incompatible element abundances. However, such large quantities of KREEP serve to greatly reduce the Mg content of the mixture. Therefore, in order to generate