LETTERSPUBLISHED ONLINE: 28 OCTOBER 2012 | DOI: 10.1038/NGEO1614

Compositional evidence for an impact origin of theMoon’s Procellarum basinRyosuke Nakamura1*, Satoru Yamamoto2, Tsuneo Matsunaga2, Yoshiaki Ishihara2,Tomokatsu Morota3, Takahiro Hiroi4, Hiroshi Takeda5, Yoshiko Ogawa6, Yasuhiro Yokota2,Naru Hirata6, Makiko Ohtake7 and Kazuto Saiki8

The asymmetry between the nearside and farside of theMoon is evident in the distribution of mare basalt1, crustalthickness2 and concentrations of radioactive elements3, butits origin remains controversial. According to one attractivescenario, a gigantic impact early in the Moon’s historyproduced the observed dichotomy; the putative 3,000-km-diameter Procellarum basin has been suggested to be a relicof this ancient impact3–5. Low-calcium pyroxene can be formedduring an impact by melting a mixture of crust and mantlematerials6,7 or by excavating differentiated cumulates from thelunar magma ocean8. Therefore, the association of low-calciumpyroxene with a lunar basin could indicate an impact origin.Here we use spectral mapping data from KAGUYA/SELENE(ref. 9) to show that low-calcium pyroxene is concentratedaround two established impact structures, the South Pole–Aitken and Imbrium basins. In addition, we detect a highconcentration of low-calcium pyroxene at Procellarum, whichsupports an impact origin of the ancient basin. We proposethat, in forming the largest known basin on the Moon, theimpact excavated the nearside’s primary feldspathic crust,which derived from the lunar magma ocean. A secondaryfeldspathic crust would have later recrystallized from the sea ofimpactmelt, leading to two distinct sides of theMoon.

The dichotomy of the Earth’s Moon was discovered by thefarside images taken by Luna 3 in 1959. Basaltic mare covers onlya few per cent of the farside, compared with roughly 30% on thenear side1. Subsequent spaceborne observations revealed the front–back asymmetry of the crustal thickness2 and the concentrationof radioactive elements on the nearside3. Several mechanismshave been proposed to produce this dichotomy, such as spatiallyinhomogeneous tidal heating in the Moon-forming stage10 oraccretion of a companion Moon11. Recently, it was found that thenorth–south crustal dichotomy of Mars can be naturally explainedby a giant impact in the ancient age12. Similarly, the Moon’sdichotomy could have resulted from a giant impact. The putativeProcellarum basin, whose diameter is more than 3,000 km withthe centre around (N15, W23), would be the most plausiblecandidate for the ancient impact event3–5. The characteristictopographical impact-basin structures must have been obliteratedby the fluidal nature of the huge impact melt sheet, viscoelastic

1Information Technology Research Institute, National Institute of Advanced Industrial Science and Technology, Umezono 1-1-1, Tsukuba, Ibaraki 305-8568,Japan, 2Center for Environmental Measurement and Analysis, National Institute for Environmental Studies, 16-2 Onogawa, Tsukuba, Ibaraki 305-8506,Japan, 3Graduate School of Environmental Studies, Nagoya University, Furo-cho, Chikusa-ku, Nagoya 464-8601, Japan, 4Department of GeologicalSciences, Brown University, Providence, Rhode Island 02912, USA, 5Department of Earth and Planetary Science, The University of Tokyo, Hongo,Bunkyo-ku, Tokyo 113-0033, Japan, 6The University of Aizu, Ikki-machi, Aizuwakamatsu, Fukushima 965-8580, Japan, 7Institute of Space andAstronautical Science, Japan Aerospace Exploration Agency, 3-1-1 Yoshino-dai, Sagamihara, Kanagawa 229-8515, Japan, 8Department of Earth and SpaceScience, Osaka University, 1-1 Machikaneyama, Toyonaka 560-0043, Osaka, Japan. *e-mail: r.nakamura@aist.go.jp.

evolution and thick impact ejecta. On the other hand, we canexpect surviving compositional evidence of the mantle exposuresand/or large-scale impact melt pool because such an enormousimpact must have completely excavated the feldspathic upper crustand produced a tremendous amount of impact melt13. Previousmultispectral global surveys14,15 focused on olivine-rich lithologiesas possible mantle exposures because equilibrium crystallizationand/or density overturn of the lunar magma ocean (LMO) aresupposed to produce olivine-rich mantle16. The upper mantle,however, would predominantly consist of low-calcium pyroxene(LCP) if the LMO experienced significant fractional crystallizationbefore the onset of plagioclace floatation8. The differentiation ofthe huge impact melt provides another chance to generate LCP-richlayers even from olivine-rich mantle6,7,17. Here, therefore, we havetargeted the LCP as a diagnostic mafic mineral linked with hugeimpacts and searched for the spectral signature by using the multi-and hyperspectral data obtained byKAGUYA/SELENE (refs 8,18).

Figure 1 indicates the global distribution of the mafic exposuresdominated by LCP. It should be emphasized that we have selectedthe spectra whose absorption depth is larger than 10% (seeMethodsand Supplementary Fig. S2). This selection rule excludes ubiquitousnoritic anorthosite19 and heavily space-weathered materials withweaker absorptions. Localized concentrations do occur around thetwo established largest impact structures, the Imbrium basin andthe South Pole–Aitken (SPA) basin. Most of the rest are encirclingthe putative Procellarum basin. It has been widely supposed that thelunar lower crust has a globally LCP-rich noritic composition19–21,but few points are present on the feldspathic highland terrane22.As illustrated in Figs 2 and 3 and Supplementary Fig. S1, most ofthe LCP-rich materials are exposed on the inner wall or impactejecta of fresh craters. If we apply the empirical algorithm for thesemultiband images19, all of the exposures would comprise morethan 30% mafic components. Figure 4 compares the spectrum ofApollo sample 14310 (ref. 23) with that of the LCP-rich depositclosest to the sampling point. Also plotted are two representativeLCP-rich spectra: Plato M on the northern rim of the Imbriumbasin and Antoniadi in the SPA. The rock 14310, collected fromthe Fra Mauro formation as Imbrium ejecta, is supposed to benoritic impact melt. Their striking spectral similarities suggest thatthe composition of 14310 (plagioclace= 59%, LCP= 31%) could

NATURE GEOSCIENCE | VOL 5 | NOVEMBER 2012 | www.nature.com/naturegeoscience 775

© 2012 Macmillan Publishers Limited. All rights reserved.

LETTERS NATURE GEOSCIENCE DOI: 10.1038/NGEO1614

Figure 1 | Location map of the LCP-rich exposures on the Moon. 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 ofmany LCP-rich sites in the SPA basin19,25,30 (Apollo, Bhaba, Finsen, Lyman, Antoniadi, Zeeman and Schrödinger). On the nearside, many LCP-rich sites arelocated on the northern Imbrium noritic region31, Bullialdus30 and Aratus20, as previously identified by ground-based telescopes and space missions.

A

H1

H2

H3

E

Figure 2 | False-colour composite images of four craters encircling the Procellarum basin. 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 imagewidths 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 MultibandImager’s spectra shown in Fig. 3.

776 NATURE GEOSCIENCE | VOL 5 | NOVEMBER 2012 | www.nature.com/naturegeoscience

© 2012 Macmillan Publishers Limited. All rights reserved.

NATURE GEOSCIENCE DOI: 10.1038/NGEO1614 LETTERS

0.25

0.20

0.15REFF

0.10

0.05500 1,000

Wavelength (nm)

A (Aston)E (Epigenes F)H1 (Heinsius)H2 (Hercules)H3 (Hercules)

1,500

Figure 3 | Multiband spectra of the four craters in Fig. 2. Reflectancefactor (REFF) measured by the Multiband Imager onboardKAGUYA/SELENE (ref. 28). The pinkish areas (A, E, H1, H2) in Fig. 2 showabsorption 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.

be analogous to that of the LCP-rich deposits in Fig. 1. Modelcalculations of radiative transfer also support their compositionalsimilarity (see Supplementary Information).

As the depths of excavation and melting are larger than themaximum lunar crustal thickness, a tremendous amount of impactmelt would be formed by assimilating the feldspathic lunar crustand mafic-rich upper mantle13. The differentiation of this meltwould have produced the LCP-dominant lithology6,7. Alternatively,LCP-rich exposures could be composed of magnesian-suiteplutons24 or ultramafic LMO cumulates8,25 excavated by hugeimpacts. The magnesian-suite parent magma may have selectivelyintruded into the rim of Imbrium and SPA where the crust wasthinned by large impacts. The origin of the LCP-rich rocks, in everycase, remains to be related to huge impacts as long as no othertectonic mechanisms produce the spatial pattern seen in Fig. 1.Therefore, LCP-rich sites surrounding the Procellarum basin couldbe linked with the biggest impact on theMoon.

The observed coexistence of LCP and olivine (Fig. 1 and Her-cules crater in Fig. 2) would provide crucial clues to better under-stand their formation process (see Supplementary Information).It should be noted that LCP exposures are not found aroundsmaller olivine-bearing basins, such as Crisium andMoscovience18.One simple interpretation is that those basins could be largeenough to excavate olivine-rich mantle, but too small to generateLCP through significant differentiation of the huge impact meltsheets6,25. Another possibility is that LCP is effectively formed in twosuccessive impacts. In fact, Fig. 1 shows a significant concentrationof LCP-rich points on the multi-ring basins Apollo, Antoniadi andSchrödinger in the SPA. Imbrium could be another example ofthe second impact following the preceding larger impact, that is,the Procellarum basin.

The coincidence between the global thorium concentrationand the circular outline of a putative Procellarum basin suggeststhat a single giant impact could be responsible for producing theProcellarum basin3. An old and gigantic impact on the Procellarumregion accounts for various aspects of the lunar dichotomy in aconsistent way. First, it can produce the observed offset between thecentre-of-mass and centre-of-figure by stripping off the feldspathicupper crust on the impact side5. As the minimum moment ofinertia would be directed along the Earth–Moon axis by thespin–orbital evolution, the excavated hemisphere or the antipodalwould eventually face the Earth26. If the impact occurred before

Refle

ctan

ce a

fter

con

tinuu

m r

emov

al 1.0

0.9

0.8

0.7

0.6600 800 1,000 1,200

Wavelength (nm)

1,400

Apollo 14310Fra MauroPlato MAntoniadi

Figure 4 | Continuum-removed spectra of the Apollo sample 14310(ref. 23) and that of the nearest LCP-rich point in Fra Mauro. Also plottedare spectra for Plato M and Antoniadi25. Their footprint locations aredenoted by red crosses in Fig. 1. All of the spectra show short centralwavelengths (<925 nm) for the 1-µm absorption bands. The 1.3 µmabsorptions are very faint, whereas the 2-µm absorption bands are evidentwithin the limited wavelength coverage. All of these are characteristicfeatures of LCP (ref. 25). The flat bottom of the 1 mm absorption band for14310 possibly indicates poor data quality of the laboratory measurements.

complete crystallization of the LMO, the resulting impact meltmust have incorporated urKREEP, the late-stage liquid fromLMO crystallization6,7. In contrast, a later impact on the presentSPA region would show a lower KREEP concentration, providedurKREEP had finished lateral migration to the Procellarum and/orvertical settling with mantle overturn16. The Procellarum impactitself might have triggered the lateral migration. The front–backasymmetry of the mare distribution could have resulted fromthe different crustal thickness2 and/or KREEP concentration inProcellarum KREEP terrane27 (PKT).

According to the present standard impact theories, the Pro-cellarum and SPA impact should have completely expelled thefeldspathic crust derived from the LMO (see Supplementary Infor-mation), whereas gravity data indicate a thin low-density crust inthe PKT and SPA (ref. 2). In addition, a previous multiband surveyfound crater central peaks and inner walls consisting of anorthositicmaterials within SPA and PKT (ref. 28). Why does this feldspathiccrust exist if ancient gigantic impacts removed the primary crustcrystalized from the LMO? The answer would lie in the vast volumeof melt sheet produced by the impacts. The thick melt sheets, ex-tending hundreds of kilometres laterally and a few tens of kilometresin depth, would have differentiated to generate secondary crustpredominantly consisting of anorthosite6,29 (see SupplementaryInformation). It is likely that the crust observed on PKT and SPAat present is not a remnant of the primordial crust solidified fromthe LMO, but a secondary product from the impactmelt.

MethodsThroughout this paper, we focus on the wavelengths between 510 and 1,600 nmwhere we can get good spectra from both the Spectral Profiler18 and the MultibandImager28 onboard KAGUYA/SELENE. From the complete Spectral Profiler data setincluding about 69 million points, we have selected the spectra with the followingfive conditions: the absolute radiance at 512 nm is larger than 23Wmm−1 m−2 sr−1;the difference in the continuum-removed reflectance Rc is smaller than 0.005between 971 and 980 nm (the continuum is calculated by an automatic algorithmas a tangent line to the target spectrum and continuum removal denotes thedivision of the original spectrum by the continuum line); both the minimumand 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 rejectthe low-quality data. The rest are the criteria to pick up clear indications of LCPwhose central wavelength of the 1 and 2mm absorptions are located at shorter

NATURE GEOSCIENCE | VOL 5 | NOVEMBER 2012 | www.nature.com/naturegeoscience 777

© 2012 Macmillan Publishers Limited. All rights reserved.

LETTERS NATURE GEOSCIENCE DOI: 10.1038/NGEO1614

wavelengths compared with LCP (ref. 25). This screening procedure results in528 data points in Fig. 1. The Spectral Profiler covered more than 20% of thelunar surface even at the equator18 during the total 1.5-year mission period.Subsequently, we have conducted an areal survey around the Spectral Profilerdetection points by using Multiband Imager multispectral images with the samealgorithm for Clementine UVVIS (ref. 31).

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

Received 15 May 2012; accepted 21 September 2012;published online 28 October 2012

References1. Head, J. W. III & Wilson, L. Lunar mare volcanism: Stratigraphy, eruption,

conditions, and the evolution of secondary crusts. Geochim. Cosmochim. Acta56, 2155–2174 (1992).

2. Ishihara, Y. et al. Crustal thickness of the Moon: Implications for farside basinstructures. Geophys. Res. Lett. 36, L19202 (2009).

3. Feldman, W. C. et al. Global distribution of lunar composition: New resultsfrom Lunar Prospector. J. Geophys. Res. 107, 5016 (2002).

4. Whitaker, E. A. The lunar Procellarum basin, in multi-ring basins. LPSC12,Part A (1981).

5. Byrne, C. J. A large basin on the near side of the Moon. Earth Moon Planets101, 153–158 (2007).

6. Hess, P. C. Petrogenesis of lunar troctolites. J. Geophys. Res. 99,19083–19093 (1994).

7. Warren, P. H., Claeys, P. & Cedillo-Pardo, E. in The Cretaceous-Tertiary Eventand Other Catastrophes in Earth History. Boulder, Colorado (eds Ryder, G.,Fastovsky, D. & Gartner, S.) (Geological Society of America, Special Paper 307,1996).

8. Elkins-Tanton, L. T., Burgess, S. & Yin, Q-Z. The lunar magma ocean:Reconciling the solidification process with lunar petrology and geochronology.Earth Planet. Sci. Lett. 304, 326–336 (2011).

9. Matsunaga, T. et al. Discoveries on the lithology of lunar crater central peaksby SELENE Spectral Profiler. Geophys. Res. Lett. 35, L23201 (2008).

10. Garrick-Bethell, I., Nimmo, F. & Wieczorek, M. A. Structure and formation ofthe lunar farside highlands. Science 330, 949–951 (2010).

11. Jutzi, M. & Asphaug, E. Forming the lunar farside highlands by accretion of acompanion moon. Nature 476, 69–72 (2011).

12. Nimmo, F. et al. Implications of an impact origin for the martian hemisphericdichotomy. Nature 453, 1220–1223 (2008).

13. Cintala, M. J. & Grieve, R. A. F. Scaling impact-melt and crater dimensions:Implications for the lunar cratering record. Meteor. Planet. Sci. 33,889–912 (1998).

14. Lucey, P. G., Taylor, G. J., Hawke, B. R. & Spudis, P. D. FeO and TiO2

concentrations in the South Pole-Aitken basin—implications for mantlecomposition and basin formation. J. Geophys. Res. 103, 3701–3708 (1998).

15. Pieters, C. M. et al. Rock types of South Pole-Aitken basin and extent of basalticvolcanism. J. Geophys. Res. 106, 28001–28022 (2000).

16. Hess, P. C. & Parmentier, E. M. A model for the thermal and chemicalevolution of the Moon’s interior: Implications for the onset of mare volcanism.Earth Planet. Sci. Lett. 134, 501–514 (1995).

17. Spray, J. G., Thompson, L. M., Biren, M. B. & O’Connell-Cooper, C. TheManicouagan impact structure as a terrestrial analogue site for lunar andmartian planetary science. Planet. Space Sci. 58, 538–551 (2010).

18. Yamamoto, S. et al. Possible mantle origin of olivine around lunar impactbasins detected by SELENE. Nature Geosci. 3, 533–536 (2010).

19. Tompkins, S. & Pieters, C. M. Mineralogy of the lunar crust: Results fromClementine.Meteor. Planet. Sci. 34, 25–41 (1999).

20. Spudis, P. D., Hawke, B. R. & Lucey, P. G. Proc. of 18th Lunar and PlanetaryScience Conference 155–168 (Cambridge Univ. Press, 1988).

21. Pieters, C. Noritic anorthosite is the most common highland rock type seen onthe surface of the lunar nearside. Rev. Geophys. 24, 57–588 (1986).

22. Jolliff, B. L. et al. Major lunar crustal terranes: Surface expressions andcrust-mantle origins. J. Geophys. Res. 105, 4197–4216 (2000).

23. Gancarz, A. J., Albee, A. L. & Chodos, A. A. Comparative petrology of Apollo16 sample 68415 and Apollo 14 samples 14276 and 14310. Earth Planet. Sci. Lett.16, 307–330 (1972).

24. Shearer, C. K. & Papike, J. J. Early crustal building processes on the moon:Models for the petrogenesis of the magnesian suite. Geochim. Cosmochim. Acta69, 3445–3461 (2005).

25. Nakamura, R. et al. Ultramafic impactmelt sheet beneath the South Pole-Aitkenbasin on the Moon. Geophys. Res. Lett. 362, L22202 (2009).

26. Wieczorek, M. A. & Le Feuvre, M. Did a large impact reorient the Moon? Icarus200, 358–366 (2009).

27. Wieczorek, M. A. & Phillips, R. J. The Procellarum KREEP terrane:Implications for mare volcanism and lunar evolution. J. Geophys. Res. 105,20417–20430 (2000).

28. Ohtake, M. et al. The global distribution of pure anorthosite on the Moon.Nature 461, 236–240 (2009).

29. Mohit, P. S. & Phillips, R. J. Viscoelastic evolution of lunar multiring basins.J. Geophys. Res. 111, E12001 (2006).

30. Klima, R. L. et al. New insights into lunar petrology: Distribution andcomposition of prominent low-Ca pyroxene exposures as observed by theMoon Mineralogy Mapper. J Geophys. Res. 116, E00G06 (2011).

31. Isaacson, P. J. & Pieters, C. M. Northern imbrium noritic anomaly. J Geophys.Res. 114, E09007 (2009).

AcknowledgementsThe images and spectra used here were acquired by the JAXA lunar orbiterKAGUYA/SELENE. We thank Fujitsu and the JASCO Corporation for their dedicatedefforts in developing the Spectral Profiler.

Author contributionsR.N. and S.Y. performed the spectral analysis and model calculations. S.Y., T.M., Y.Y.and Y.O. carried out the data reduction and instrument calibration. S.Y. and N.H.contributed to the qualitative estimate of the impact melt production. T.H. contributed tothe comparison of the spaceborne lunar spectra and laboratory spectra of returned Apollosamples. Y.I., T.M., H.T. and K.S. solidified the results of this paper from geophysical andmineralogical points of view. T.M. and M.O. served as principal investigators to acquirethe images and spectra from the Spectral Profiler and the Multiband Imager onboardKAGUYA. R.N., S.Y. and T.M. worked jointly to write the paper. All authors discussedthe interpretation of the results and commented on themanuscript.

Additional informationSupplementary information is available in the online version of the paper. Reprints andpermissions information is available online at www.nature.com/reprints. Correspondenceand requests for materials should be addressed to R.N.

Competing financial interestsThe authors declare no competing financial interests.

778 NATURE GEOSCIENCE | VOL 5 | NOVEMBER 2012 | www.nature.com/naturegeoscience

© 2012 Macmillan Publishers Limited. All rights reserved.