The role of ilmenite in the source region for mare basalts: Evidence from niobium, zirconium, and cerium in picritic glasses

Abstract

To investigate models for the generation of lunar high Ti-basalts, we have analyzed lunar picritic glasses for Zr, Nb, Ce, and Ti at high precision using ion microprobe techniques. The picritic magmas represented by these glasses have experienced minor crystallization, which has allowed us to partially eliminate the effects of post-melting processes commonly experienced by crystalline high-Ti mare basalts. The Nb Zr for these glasses ranges from .05-.11. The high-Ti glasses generally tend to have higher values of Nb Zr (.072-.109) than the very low-Ti glasses (.048-.085). The crystalline mare basalts tend to have slightly higher Nb Zr than glasses with similar Ti from the same site. For example, the Apollo 17 (A17) high-Ti basalts have Nb Zr of approximately .09. whereas, the A17 high-Ti glasses have Nb Zr of .07. KREEP has Nb Zr of approximately .06. Thus Zr is fractionated from Nb to different degrees in the various picritic magmas. The concentrations of Zr, Nb, and Ce increase from the very low-Ti glasses to the high-Ti glasses, and along the trajectory the Nb Ce and Zr Ce increase. Nb Ce (.25-1.6) and Zr Ce (4–15) for the picritic glasses overlap with KREEP ( Nb Ce = .36 and Zr Ce =5 ). Zr Ti and Nb Ti show a wide range of variation in these glasses. Both Zr Ti and Nb Ti in the glasses range from approximately .0014 to slightly less than .0003. The Zr Ti and Nb Ti for these glasses overlap with that of the crystalline mare basalts. Generally, with increasing Ti, Nb, and Zr, Zr Ti and Nb Ti decrease. The exceptions to this are the Apollo 14 (A14) glasses that exhibit an increase in Nb Ti and Zr Ti . Based on these data for the picritic glasses and experimentally determined partition coefficients for Nb, Zr, and Ce, the mantle sources for these picritic magmas are slightly to moderately fractionated from C1 chondrite and previous estimates of the bulk silicate Moon. Our best fit model for our data and this observation is that both the very low-Ti and high-Ti picritic magmas were derived through small to moderate degrees of nonmodal melting of lunar mantle sources consisting of a mixture of late-stage LMO cumulates (derived after >95% crystallization of the LMO) and early to intermediate LMO cumulates (derived prior to 80% crystallization of the LMO). The early LMO cumulates had Nb Zr , Zr Ce , and Nb Ce ratios near C1 chondrite, whereas these ratios were fractionated in the late-stage LMO cumulates. This hybridization of mantle sources occurred during large scale overturning of the LMO cumulate pile. The source for the low-Ti picritic magmas had very minor amounts of ilmenite, whereas the source for the high-Ti picritic magmas probably contained less than 6% ilmenite. For all the picritic magmas, ilmenite was exhausted from the residua during melting. Models suggesting that the high-Ti magmas are derived through the assimilation of an ilmenite-bearing cumulate layer or preferential assimilation of ilmenite by low-Ti primary magmas are not consistent with the Nb, Zr, Ce, and Ti data magmas (Hubbard and Minear, 1975; Wagner and Grove, 1993, 1995). In particular, the preferential assimilation by very low-Ti picritic magmas of ilmenite with expected Nb Ce (20,000–22,000) and Nb Zr (55) signatures would displace the resulting high-Ti magma too far from our observed data. Large scale overturning of the LMO cumulate pile also accounts for the trace element signatures found in the A14 picritic glasses. The evolved signature found in these primitive very low-Ti picritic glasses is most likely a product of KREEP incorporation into the LMO cumulate source rather than either contamination by evolved ilmenite-bearing cumulates or incorporation of higher proportions of locally derived intercumulus melt.