Main

The dichotomy of the Earth’s Moon was discovered by the farside images taken by Luna 3 in 1959. Basaltic mare covers only a few per cent of the farside, compared with roughly 30% on the near side1. Subsequent spaceborne observations revealed the front–back asymmetry of the crustal thickness2 and the concentration of radioactive elements on the nearside3. Several mechanisms have been proposed to produce this dichotomy, such as spatially inhomogeneous tidal heating in the Moon-forming stage10 or accretion of a companion Moon11. Recently, it was found that the north–south crustal dichotomy of Mars can be naturally explained by a giant impact in the ancient age12. Similarly, the Moon’s dichotomy could have resulted from a giant impact. The putative Procellarum basin, whose diameter is more than 3,000 km with the centre around (N15, W23), would be the most plausible candidate for the ancient impact event3,4,5. The characteristic topographical impact-basin structures must have been obliterated by the fluidal nature of the huge impact melt sheet, viscoelastic evolution and thick impact ejecta. On the other hand, we can expect surviving compositional evidence of the mantle exposures and/or large-scale impact melt pool because such an enormous impact must have completely excavated the feldspathic upper crust and produced a tremendous amount of impact melt13. Previous multispectral global surveys14,15 focused on olivine-rich lithologies as possible mantle exposures because equilibrium crystallization and/or density overturn of the lunar magma ocean (LMO) are supposed to produce olivine-rich mantle16. The upper mantle, however, would predominantly consist of low-calcium pyroxene (LCP) if the LMO experienced significant fractional crystallization before the onset of plagioclace floatation8. The differentiation of the huge impact melt provides another chance to generate LCP-rich layers even from olivine-rich mantle6,7,17. Here, therefore, we have targeted the LCP as a diagnostic mafic mineral linked with huge impacts and searched for the spectral signature by using the multi- and hyperspectral data obtained by KAGUYA/SELENE (refs 8, 18).

Figure 1 indicates the global distribution of the mafic exposures dominated by LCP. It should be emphasized that we have selected the spectra whose absorption depth is larger than 10% (see Methods and Supplementary Fig. S2). This selection rule excludes ubiquitous noritic anorthosite19 and heavily space-weathered materials with weaker absorptions. Localized concentrations do occur around the two established largest impact structures, the Imbrium basin and the South Pole–Aitken (SPA) basin. Most of the rest are encircling the putative Procellarum basin. It has been widely supposed that the lunar lower crust has a globally LCP-rich noritic composition19,20,21, but few points are present on the feldspathic highland terrane22. As illustrated in Figs 2 and 3 and Supplementary Fig. S1, most of the LCP-rich materials are exposed on the inner wall or impact ejecta of fresh craters. If we apply the empirical algorithm for these multiband images19, all of the exposures would comprise more than 30% mafic components. Figure 4 compares the spectrum of Apollo sample 14310 (ref. 23) with that of the LCP-rich deposit closest to the sampling point. Also plotted are two representative LCP-rich spectra: Plato M on the northern rim of the Imbrium basin and Antoniadi in the SPA. The rock 14310, collected from the Fra Mauro formation as Imbrium ejecta, is supposed to be noritic impact melt. Their striking spectral similarities suggest that the composition of 14310 (plagioclace = 59%, LCP = 31%) could be analogous to that of the LCP-rich deposits in Fig. 1. Model calculations of radiative transfer also support their compositional similarity (see Supplementary Information).

Figure 1: Location map of the LCP-rich exposures on the Moon.
figure 1

The blue triangles and red crosses denote the sites shown in Figs 2 and 4, respectively, and the rest are shown by yellow diamonds. The olivine-rich exposures18 are represented by green squares for reference. We could confirm the presence of many LCP-rich sites in the SPA basin19,25,30 (Apollo, Bhaba, Finsen, Lyman, Antoniadi, Zeeman and Schrödinger). On the nearside, many LCP-rich sites are located on the northern Imbrium noritic region31, Bullialdus30 and Aratus20, as previously identified by ground-based telescopes and space missions.

Full size image
Figure 2: False-colour composite images of four craters encircling the Procellarum basin.
figure 2

The upper panels show an unnamed crater near Aston (left: W85.6, N33.5) and Epigenes F (right: W8.1, N67.0); the lower panels indicate Heinsius (left: W18.0, S39.7) and Hercules (right: E42.0, N50.3). The image widths are 8 km and reflectance factors are assigned to red (750 nm), green (900 nm) and blue (1,250 nm). The labels show the locations of the Multiband Imager’s spectra shown in Fig. 3.

Full size image
Figure 3: Multiband spectra of the four craters in Fig. 2.
figure 3

Reflectance factor (REFF) measured by the Multiband Imager onboard KAGUYA/SELENE (ref. 28). The pinkish areas (A, E, H1, H2) in Fig. 2 show absorption minima at 900 nm, representing LCP-dominant compositions, whereas the red H3 region shows an absorption minimum at 1,050 nm, indicative of an exposure of olivine.

Full size image
Figure 4: Continuum-removed spectra of the Apollo sample 14310 (ref. 23) and that of the nearest LCP-rich point in Fra Mauro.
figure 4

Also plotted are spectra for Plato M and Antoniadi25. Their footprint locations are denoted by red crosses in Fig. 1. All of the spectra show short central wavelengths (<925 nm) for the 1-μm absorption bands. The 1.3 μm absorptions are very faint, whereas the 2-μm absorption bands are evident within the limited wavelength coverage. All of these are characteristic features of LCP (ref. 25). The flat bottom of the 1 mm absorption band for 14310 possibly indicates poor data quality of the laboratory measurements.

Full size image

As the depths of excavation and melting are larger than the maximum lunar crustal thickness, a tremendous amount of impact melt would be formed by assimilating the feldspathic lunar crust and mafic-rich upper mantle13. The differentiation of this melt would have produced the LCP-dominant lithology6,7. Alternatively, LCP-rich exposures could be composed of magnesian-suite plutons24 or ultramafic LMO cumulates8,25 excavated by huge impacts. The magnesian-suite parent magma may have selectively intruded into the rim of Imbrium and SPA where the crust was thinned by large impacts. The origin of the LCP-rich rocks, in every case, remains to be related to huge impacts as long as no other tectonic mechanisms produce the spatial pattern seen in Fig. 1. Therefore, LCP-rich sites surrounding the Procellarum basin could be linked with the biggest impact on the Moon.

The observed coexistence of LCP and olivine (Fig. 1 and Hercules crater in Fig. 2) would provide crucial clues to better understand their formation process (see Supplementary Information). It should be noted that LCP exposures are not found around smaller olivine-bearing basins, such as Crisium and Moscovience18. One simple interpretation is that those basins could be large enough to excavate olivine-rich mantle, but too small to generate LCP through significant differentiation of the huge impact melt sheets6,25. Another possibility is that LCP is effectively formed in two successive impacts. In fact, Fig. 1 shows a significant concentration of LCP-rich points on the multi-ring basins Apollo, Antoniadi and Schrödinger in the SPA. Imbrium could be another example of the second impact following the preceding larger impact, that is, the Procellarum basin.

The coincidence between the global thorium concentration and the circular outline of a putative Procellarum basin suggests that a single giant impact could be responsible for producing the Procellarum basin3. An old and gigantic impact on the Procellarum region accounts for various aspects of the lunar dichotomy in a consistent way. First, it can produce the observed offset between the centre-of-mass and centre-of-figure by stripping off the feldspathic upper crust on the impact side5. As the minimum moment of inertia would be directed along the Earth–Moon axis by the spin–orbital evolution, the excavated hemisphere or the antipodal would eventually face the Earth26. If the impact occurred before complete crystallization of the LMO, the resulting impact melt must have incorporated urKREEP, the late-stage liquid from LMO crystallization6,7. In contrast, a later impact on the present SPA region would show a lower KREEP concentration, provided urKREEP had finished lateral migration to the Procellarum and/or vertical settling with mantle overturn16. The Procellarum impact itself might have triggered the lateral migration. The front–back asymmetry of the mare distribution could have resulted from the different crustal thickness2 and/or KREEP concentration in Procellarum KREEP terrane27 (PKT).

According to the present standard impact theories, the Procellarum and SPA impact should have completely expelled the feldspathic crust derived from the LMO (see Supplementary Information), whereas gravity data indicate a thin low-density crust in the PKT and SPA (ref. 2). In addition, a previous multiband survey found crater central peaks and inner walls consisting of anorthositic materials within SPA and PKT (ref. 28). Why does this feldspathic crust exist if ancient gigantic impacts removed the primary crust crystalized from the LMO? The answer would lie in the vast volume of melt sheet produced by the impacts. The thick melt sheets, extending hundreds of kilometres laterally and a few tens of kilometres in depth, would have differentiated to generate secondary crust predominantly consisting of anorthosite6,29 (see Supplementary Information). It is likely that the crust observed on PKT and SPA at present is not a remnant of the primordial crust solidified from the LMO, but a secondary product from the impact melt.

Methods

Throughout this paper, we focus on the wavelengths between 510 and 1,600 nm where we can get good spectra from both the Spectral Profiler18 and the Multiband Imager28 onboard KAGUYA/SELENE. From the complete Spectral Profiler data set including about 69 million points, we have selected the spectra with the following five conditions: the absolute radiance at 512 nm is larger than 23 W mm−1 m−2 sr−1; the difference in the continuum-removed reflectance Rc is smaller than 0.005 between 971 and 980 nm (the continuum is calculated by an automatic algorithm as a tangent line to the target spectrum and continuum removal denotes the division of the original spectrum by the continuum line); both the minimum and next-lowest Rc occur below 925 nm; the minimum Rc is smaller than 0.9; Rc at 1,403 nm is larger than that at 1,508 nm. The first two conditions reject the low-quality data. The rest are the criteria to pick up clear indications of LCP whose central wavelength of the 1 and 2 mm absorptions are located at shorter wavelengths compared with LCP (ref. 25). This screening procedure results in 528 data points in Fig. 1. The Spectral Profiler covered more than 20% of the lunar surface even at the equator18 during the total 1.5-year mission period. Subsequently, we have conducted an areal survey around the Spectral Profiler detection points by using Multiband Imager multispectral images with the same algorithm for Clementine UVVIS (ref. 31).

The original data can be found online at the SELENE data archive (http://l2db.selene.darts.isas.jaxa.jp).