50th Lunar and Planetary Science Conference 2019 (LPI Contrib. No. 2132) 2851.pdf

THE LUNAR MAGMA OCEAN IS DEAD, LONG LIVE THE LUNAR MAGMA OCEAN! S. M. Elardo. Department of Geological Sciences, University of Florida, Gainesville, FL 32611, USA. [email protected]. Invited Abstract – Special Session: 50 Years of Lunar Science: The Apollo Legacy

Introduction: For the past 50 years, the lunar July 1969. In the intervening ~50 years since its incep- magma ocean (LMO) model has been the paradigm un- tion, the LMO model has continuously evolved as lunar der which all lunar data regarding the igneous evolution science has progressed. The picture of lunar differentia- of the Moon, derived from samples or otherwise, have tion that has emerged from decades of samples studies, been interpreted. It has also served as the foundation for remote observations, and geophysical modeling is one understanding planetary differentiation solar system- of enormous complexity. The simple LMO layer-cake wide. Whereas plate tectonics and erosion have erased models of sequential crystallization and crust formation or severely metamorphosed the oldest crust on Earth, are “dead” and have given way to more multifaceted and and mantle convection has mixed away much, but not realistic models describing complex dynamic processes all of the original heterogeneity in Earth’s mantle de- in the early, cooling Moon. However, the basic tenants rived from differentiation, the Moon preserves many of of the LMO, wide-spread and deep melting, crust for- the direct products of the LMO. This is known, of mation through flotation, and a chemically heterogene- course, because of the return of ~842 pounds of lunar ous mantle, live on, having survived decades of scru- samples to Earth by the six successful Apollo landings. tiny. The details of basic LMO crystallization models The Apollo samples revealed a geologic environment are thoroughly detailed elsewhere [e.g., 3]; therefore, in that is far more limited in terms of both lithologic and this brief review, I will discuss a few areas of LMO- mineralogical diversity compared to Earth. However, related research that have been the focus of recent scru- the antiquity of the samples combined with certain tell- tiny and progress, and/or are areas that will require sig- tale geochemical and isotopic features, reveal that the nificant attention in the future, all having been driven or Moon’s anorthositic crust is the complementary reser- heavily informed by studies of the Apollo sample suite. voir to much of, if not all of the Moon’s mantle, and Timing and Duration of the LMO: Questions have indicate that most of crust was extracted from the mantle persisted regarding the timing of the onset of the LMO early and quickly in a widespread melting event that (i.e., the formation of the Moon) and the duration of would come to be known as the lunar magma ocean. LMO crystallization for decades based primarily on am- biguity in the interpretation of isochron and model ages derived from analyses of Apollo samples [4]. Recent progress in determining the ages of primary crustal rocks, the earliest crust-building plutonic rocks, and model ages for mantle reservoirs have almost ubiqui- tously produced ages within error of 4.38 – 4.34 Ga [e.g., 4 - 9], suggesting a younger Moon/LMO than pre- viously thought. Although a few zircon analyses and zir- con Lu-Hf model ages have been interpreted as indicat- ing an older age for lunar formation/the LMO [e.g., 10 - 12] and highlight the ambiguity inherent in interpreta- tion, most lines of evidence point toward a young Moon and a rapidly crystallizing LMO when taken at face value. Although models have been presented to recon- cile petrologic constraints with isotopic constraints [e.g., 13], the similarity between ages for both mantle source regions and crustal rocks remains potentially Figure 1: Early schematic model of the LMO from [1], pre- problematic, especially in light of essentially concord- sented roughly 6 months after the return of Apollo 11. ant ages for zircons of the Jack Hills, Australia [14], which contain trace element and O isotopic evidence for The original LMO models (e.g., Fig. 1) were devel- relatively cool, wet, compositionally evolved magma- oped remarkably quickly, with the first two papers [1, tism at Earth’s surface at ~4.37 Ga. 2] detailing early forms of the model being presented at Mechanisms and Timing of Volatile Loss: Imme- the Apollo 11 Lunar Science Conference in January diately after the return of the Apollo samples it became 1970, a mere 6 months after the return of Apollo 11 in abundantly clear that the Moon is depleted in geochem- ically volatile elements relative to Earth and some other 50th Lunar and Planetary Science Conference 2019 (LPI Contrib. No. 2132) 2851.pdf

planetary bodies. The Moon-forming giant impact has anorthositic terranes on the farside formed earlier in the long thought to be the event that depleted lunar materi- LMO crystallization sequence than complementary ter- als in volatile elements, but a combination of new anal- ranes on the nearside, supporting asymmetric crystalli- yses of Apollo samples, experiments, and geochemical zation models like that of Loper and Werner [21]. modeling have cast doubt on that idea, or at the least Asymmetric crustal growth models also imply that re- have challenged its completeness as an explanation. Re- sidual liquids in the LMO should be concentrated on the cent high-resolution measurements of Zn, K, and S iso- nearside, opposite to the thickening crust on the farside. topic compositions of lunar samples [15-17] have re- The concentration of KREEP-rich materials on the near- vealed differences between Earth and the Moon that side, which were sampled repeatedly by Apollo, sup- may be indicative of volatile loss in the aftermath of the ports such models. giant impact. However, other workers have argued that The traditional post-Apollo view of cumulate over- degassing of volatile and moderately volatile elements turn post-LMO has also been challenged by recent stud- from the LMO can partially or fully explain the isotopic ies. The geodynamic study of Boukaré et al. [24] sug- compositions of Zn [18] and Cl [19, 20]. The Cl isotopic gested that density-driven overturn of the LMO cumu- compositions of mare basalts correlate with abundances late pile may have begun before the end of magma ocean of incompatible trace elements, with KREEP-rich sam- crystallization. They argued that cooling of the LMO ples having the most fractionated Cl, which supports may have been sufficiently fast before the formation of models of Cl loss late in the LMO crystallization se- a coherent plagioclase-rich crust to trap large quantities quence instead of during lunar formation and accretion. of residual liquid in the LMO cumulate pile. If true, the Gaining a better understanding of the timing of volatile viscosity of the cumulate pile may have been low loss during lunar evolution will undoubtedly lead to bet- enough to trigger overturn of the mantle before the end ter models of volatile depletion in the terrestrial planets of the LMO, which would have important consequences and targeted analyses of lunar sample are key. for the geochemistry of the mantle and residual LMO. Crystallization Dynamics and the Development After a coherent crust forms, crystallization slows, al- of Hemispheric Asymmetries: The Moon has a num- lowing more time for trapped liquid to be efficiently re- ber of physical and geochemical asymmetries that may moved from the cumulate pile, which is consistent with be linked to complexities in the differentiation process. geochemical constraints from Apollo samples [25], and Most notable are hemispheric asymmetries in crustal thus increasing viscosity. thickness and near-surface distribution of the KREEP References: [1] Wood et al. (1970) Apollo 11 Lunar reservoir. In simple LMO models, crustal thickness is Sci. Conf., Vol 1, 965-988 [2] Smith et al. (1970) Apollo not predicted to vary greatly over the lunar surface 11 Lunar Sci. Conf., Vol 1, 897-925.. [3] Shearer et al. based on the assumption of a homogeneous, cooling (2006) RiMG, Vol. 60, Ch. 4. [4] Borg et al. (2014) LMO. Preferential crust-building on the farside would MaPS, 45, 619-647. [5] Carlson et al. (2014) Phil. require a physical forcing in order to delay plagioclase Trans. Royal Soc. A., 372. [6] Gaffney and Borg (2014) crystallization on the nearside or to force plagioclase ac- GCA, 140, 227-240. [7] McLeod et al. (2016) GCA, 187, cumulation on the farside. Loper and Werner [21] pro- 350-374. [8] Connolly and Bizzaro (2016) EPSL, 452, posed that during LMO crystallization, a pattern of tilted 36-43. [9] Snape et al. (2016) EPSL, 451, 149-158. [10] convection arose due to small temperature contrasts be- Nemchin et al. (2009) Nat. Geosci., 2, 133-136. [11] tween the near and farsides. Such a convective pattern Grange et al. (2011) GCA, 75, 2213-2232. [12] Barboni would force plagioclase accumulation to occur on the et al. (2017) Sci. Adv., 3(1), e1602365. [13] Elkins-Tan- slightly cooler farside, thus preferentially thickening the ton et al. (2011) EPSL, 304, 326-336. [14] Valley et al. crust there relative to the nearside. (2014) Nat. Geosci., 7, 219-223. [15] Paniello et al. Compositional evidence supports models of asym- (2012) Nature, 490, 376-380. [16] Wang and Jacobsen metric crustal growth. In the Apollo sample collection, (2016) Nature, 538, 487-490. [17] Wing and Farquhar ferroan anorthosites (FANs) have a mean Mg# of ~60 (2015) GCA, 170, 266-280. [18] Dhaliwal et al. (2018) and span a range in compositions from ~40 – 70 [22]. In Icarus, 300, 249-260. [19] Boyce et al. (2015) Sci. Adv., simple LMO models, the average composition of FANs 1, e1500380. [20] Barnes et al. (2016) EPSL, 447, 84- should be relatively consistent over the lunar surface. 94. [21] Loper and Werner (2002) JGR-Planets, 107, However, using gamma ray spectrometer data from the E6, 5046. [22] Warren (1993) Am. Min., 78, 360-376. Kaguya Spectral Profiler, Ohtake et al. [23] found that [23] Ohtake et al. (2012) Nat. Geosci., 5, 384-388. [24] Mg# range was shifted to higher values in the farside Boukaré et al. (2018) EPSL, 491, 216-225. [25] highlands relative to the nearside highlands terranes, McCubbin et al. (2015) Am. Min., 100, 1668-1707. with the peak in the farside dataset at an Mg# of ~65 vs. Acknowledgements: S. M. E. acknowledges NASA ~55 on the nearside. This Mg# distribution implies that Solar System Workings grant NNX16AQ17G.