Remote detection of_magmatic_water_in_bullialdus_crater_on_the_moon

LETTERSPUBLISHED ONLINE: 25 AUGUST 2013 | DOI: 10.1038/NGEO1909

Remote detection of magmatic water in BullialdusCrater on the MoonR. Klima1*, J. Cahill1, J. Hagerty2 and D. Lawrence1

Once considered dry compared with Earth, laboratory analysesof igneous components of lunar samples have suggested thatthe Moon’s interior is not entirely anhydrous1,2. Water andhydroxyl have also been detected from orbit on the lunarsurface, but these have been attributed to nonindigenoussources3–5, such as interactions with the solar wind. Magmaticlunar volatiles—evidence for water indigenous to the lunarinterior—have not previously been detected remotely. Herewe analyse spectroscopic data from the Moon MineralogyMapper (M3) and report that the central peak of BullialdusCrater is significantly enhanced in hydroxyl relative to itssurroundings. We suggest that the strong and localizedhydroxyl absorption features are inconsistent with a surficialorigin. Instead, they are consistent with hydroxyl bound tomagmatic minerals that were excavated from depth by theimpact that formed Bullialdus Crater. Furthermore, estimatesof thorium concentration in the central peak using data fromthe Lunar Prospector orbiter indicate an enhancement inincompatible elements, in contrast to the compositions ofwater-bearing lunar samples2. We suggest that the hydroxyl-bearing material was excavated from a magmatic source that isdistinct from that of samples analysed thus far.

Hydrogen, hydroxyl (OH−) and water have been detected on thelunar surface from orbit using a range of different instruments3–7.Hypotheses for globally distributed surficial H2O/OH− moleculesor larger deposits of water ice in the permanently shadowed regionsof the Moon include that OH− and H2O molecules are producedin situ through the interaction of solar-wind-derived protons withlunar soils3,5,8 and/or that they are delivered by comets and otherimpactors9. Broadly distributed molecular OH− or H2O formedby solar-wind interaction with surface silicates is expected to beloosely bound to the lunar surface8 andmay dissociate from surfacesoils during solar heating, migrating along ballistic trajectoriesuntil ultimately becoming cold-trapped in permanently shadowedregions and/or buried10,11. New research suggests that hydroxylformed by solar-wind bombardment may also become embeddedin agglutinates during micrometeorite impacts as part of the spaceweathering process12. Until now, bound, magmatic lunar volatileshave not been detected remotely anywhere on theMoon.

The 61-km-diameter Bullialdus Crater, centred at 20.7◦ S,337.8◦ E in Mare Nubium (Fig. 1), lies along the southern edgeof the Procellarum KREEP (potassium, rare earth elements andphosphorus) Terrane (PKT), a region on the nearside of theMoon that is highly enriched in incompatible elements13. LunarProspector measurements suggest that Bullialdus Crater coincideswith a localized concentration of thorium14. Bullialdus Crateris mineralogically distinct and its central peak has long beenrecognized as exhibiting strong spectral signatures typical of norite,

1Johns Hopkins University Applied Physics Laboratory, Laurel, Maryland 20723, USA, 2US Geological Survey, Astrogeology Science Center, Flagstaff,Arizona 86001, USA. *e-mail: [email protected]

an intrusively formed igneous rock dominated by orthopyroxene(Ca-poor pyroxene) and anorthite (Ca-rich plagioclase feldspar)15.Stratigraphic interpretations of materials exposed in BullialdusCrater walls, coupled with the presence of exhumed noriticmaterials in its central peak, led previous investigators to suggestthat the Bullialdus impact excavated a layered mafic intrusion15,16.Data fromM3 provide an opportunity to examine Bullialdus Craterboth at high spatial (∼140m per pixel) and spectral (20–40 nmsampling) resolution. They also provide, for the first time, theopportunity to characterize visible to near-infrared wavelengthreflectance spectra (that is, 0.6–3 µm). The 3 µmregion in particularis critical for near-infrared volatile assessment of the Moon or anyother airless body because both water and hydroxyl are stronglyabsorbing in this portion of the electromagnetic spectrum.

Previous interpretations of this area suggest that the Bullialdusimpactor penetrated the basalt-flooded Nubium Basin, excavatinga range of intrusive crustal rocks. The central peaks of Bullialdusare mineralogically diverse, with the westernmost peak exhibitinga more clinopyroxene-rich spectral signature (Fig. 1e, cyan) thanthe northern peaks. On the basis of radiative transfer models ofClementine reflectance data, this western portion of the peak hasbeen classified as anorthositic gabbronorite, whereas the northernpeaks have been classified as anorthositic norite or norite17. Norite(Fig. 1e, yellow) dominates the bulk of the central peak, thoughthere is a region of anorthositicmaterial (Fig. 1e, dark blue) exposedtowards the centre of the peak. Anorthositicmaterial is also exposedin the crater rim and proximal ejecta (Fig. 1d, black). A map of the2.8 µm band depth (Fig. 1f,g) reveals that hydroxyl is detected inonly the central peak and is explicitly found in association with thenoritic and anorthositic peak material.

In laboratory measurements, bound OH− can be distinguishedfrom adsorbed OH− and H2Omolecules by examining the spectralshape. Unfortunately, the specific spectral shape of the hydroxylabsorption band cannot be precisely characterized by M3 becauseof the 40 nm spectral sampling in the 3 µm region. The surficialOH− and H2O molecules detected previously by M3 typicallyexhibit a broad absorption beyond 2.8 µm (ref. 8). In contrast,the hydroxyl absorption observed in the central peak of BullialdusCrater is significantly stronger and sharper than is observed atsimilar latitudes in other nearby terrains (Fig. 2 and SupplementaryFig. S1). The absorption observed in both the anorthositic andnoritic material exhibits a clear band minimum at 2.8 µm. Theband shape is significantly sharper than laboratory measurementsof agglutinates12, but is consistent in both energy and bandshape with OH− measured in transmission spectra of internallybound hydroxyl in nominally anhydrous terrestrial minerals18 suchas anorthite and orthopyroxene, and minor hydrated terrestrialminerals such as apatite19 (Supplementary Fig. S2).

NATURE GEOSCIENCE | VOL 6 | SEPTEMBER 2013 | www.nature.com/naturegeoscience 737

LETTERS NATURE GEOSCIENCE DOI: 10.1038/NGEO1909

23° W 22° W 21° W

76

5

1 2

43

b

22° S

d

f

c

e

g

23° W 22° W 21° W

23° W 22° W 21° W

Bullialdus

Mare Nubium

a

8

20° S

21° S

22° S

20° S

21° S

20° S

21° S

22° S

Figure 1 | Context and spectral parameter maps of Bullialdus Crater. a, LROC wide-angle camera (WAC) view of Bullialdus Crater. Scale bar, 100 km.Image credit: NASA/GSFC/ASU. b,c, M3 750 nm albedo map of Bullialdus Crater (80 km image width; b) and central peak (scale bar, 5 km; c). d,e, M3

false colour composite (Supplementary Table S1) highlighting mineralogy of the crater (d) and central peak (e). f,g, 2.8 µm OH− absorption strength inBullialdus Crater (f) and central peak (g). Absorption strength grades black (0%) to white (4%). Numbers on a–c indicate locations from which Fig. 2spectra were extracted.

If the hydroxyl observed at Bullialdus Crater is solar-windimplanted, it could either be locked stably within agglutinates12or it could be unstable, adhered loosely to mineral grains butcontinuously forming in the solar-wind flux10. The former wouldimply that the central peak of Bullialdus is more enriched inagglutinates (or older) than any other material within at least 10◦latitude in any direction, which is inconsistent with surfacematuritymaps of the region20. The strongest water/hydroxyl absorptionbands are found along and just below the peak ridge on bothsides and coincide with bands of large (tens to ∼150-m-diameter)boulders and potentially bedrock (Fig. 3). The coarse texture andhigh albedo of the OH−-rich regions of the central peak suggestthat it has not experienced unusually high degrees of agglutinationrelative to its surroundings.

If OH− in Bullialdus Crater is loosely adhered to the surface,the band depths of the OH− absorption might weaken later in thelunar day, as the surface becomes warmer, or when the Moon iswithin the Earth’s magnetotail and shielded from the solar wind.The central peak of Bullialdus was imaged by M3 at different timesof the lunar day. Hydroxyl absorptions are deepest near the apexesof the peaks and weaken with increasing distance downslope. Theyoccur symmetrically about the central peak ridges, but becauseof the illumination geometry in each of the optical periods, wecompare a slightly weaker absorption on a flatter slope that isnot in shadow during any of the three optical periods. After

photometric correction, there is little to no change in absorptionband depth observed to suggest hydroxyl mobility (Fig. 2 andSupplementary Fig. S3).

A previous study10 showed that the likelihood of OH−adsorption may depend on lithology. The central peak of AristillusCrater (33.9◦N, 1.2◦ E) has a similar lithology to Bullialdus andis also the focus of a thorium anomaly. Given that Aristillus ismore than 10◦ farther from the equator, it should be cooler andmore hospitable to OH− adsorption than Bullialdus, resulting ina central peak that is coated with OH−. However, it does notexhibit a prominent regional 3 µm anomaly (Fig. 2). The geologiccontext coupled with the stability of the 2.8 µm absorption bandat Bullialdus Crater leads us to conclude that the water/hydroxyldetected in the central peak is most likely bound within primarymagmatic materials rather than implanted by the solar wind orwithin glassy agglutinates.

We can place limits on the amount of water/hydroxyl presentby making simplifying assumptions about the nature of the surfaceand/or mineralogy. On the basis of radiative transfer modelling,another study21 suggested that Bullialdus Crater central peak as awhole contains about 57% mafic minerals and 43% plagioclase.Using this bulk mineralogy, coupled with an orthopyroxenecomposition of Mg75 (ref. 22), we can derive the single-scatteringalbedo of the surface, from which the absorbance can becalculated23–25 (Supplementary Fig. S4). If we assume that the local

738 NATURE GEOSCIENCE | VOL 6 | SEPTEMBER 2013 | www.nature.com/naturegeoscience

NATURE GEOSCIENCE DOI: 10.1038/NGEO1909 LETTERS1 Central peak2 Central peak

5 Crater floor6 Crater wall

7 Mare Nubium soil6 Highland soil south

3 Central peak4 Central peak

0.001,000 1,500 2,000

Wavelength (nm)

2,500 3,000 1,000 1,500 2,000

Wavelength (nm)

2,500 3,000

1,000 1,500 2,000

Wavelength (nm)

2,500 3,000 1,000 1,500 2,000

Wavelength (nm)

2,500 3,000

0.05

0.10

0.15

0.20

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.2

0.4

0.6

0.8

1.0

1.2

1.4

Scal

ed r

efle

ctan

ce

Refle

ctan

ceSc

aled

ref

lect

ance

Scal

ed r

efle

ctan

ce

0.00

0.05

0.25

0.15

0.20

0.10

Optical period 1bOptical period 2aOptical period 2c1

Bullialdus central peak, 20.8° S, 22.3° WAristillus central peak, 33.9° N, 1.2° EHighlands, 59.0° S, 43.5° W

a b

c d

Figure 2 | Spectra of Bullialdus Crater and the surrounding region. a, Strong absorption bands at 1,000 and 2,000 nm are indicative of pyroxene, whereasflatter, brighter spectra are probably dominated by anorthosite. b, Bullialdus spectra scaled to unity at 2.736 µm. Only the spectra extracted from thecentral peak exhibit a hydroxyl absorption. c, Spectra from Bullialdus Crater central peak compared with the higher latitude, noritic central peak of AristillusCrater and a higher-latitude region of highlands exhibiting a typical surficial water signature. d, Repeat spectral imaging during three optical periods. c,d arescaled at 2.736 µm.

regolith exhibits a grain size between 15 and 150 µm,we estimate themaximumabundance in the central peak to be 80±40 ppmaswater.

In principle, one might expect that the bulk hydrogen signaturecould be detected using neutron spectroscopy in a mannersimilar to how enhanced hydrogen signatures were detected at thelunar poles6. However, the small spatial scale of the Bullialduscentral peak (∼5 km width) and the low hydrogen concentrations([H] = (2/18)(80 ± 40) = 9 ± 4 ppmH) preclude the detectionof a hydrogen signature from either the Lunar Prospector orLunar Exploration Neutron Detector data because the hydrogensensitivity of these data sets is tens of ppm H and >100 ppmH, respectively26,27.

A general Th enhancement (5–6 ppm) focused on BullialdusCrater suggests that the excavated rocks also contain increasedamounts of KREEP. The broad spatial resolution (>30 km) of theTh data14 precludes a unique quantification of the Th content ofthe central peak. However, using forward modelling techniquesthat have been validated28 elsewhere on the Moon, the Thconcentrations of the central peak region can be non-uniquelyestimated. Given the distinct mineralogical character of the centralpeak, we have carried out forwardmodelling of the entire Bullialdusregion to infer the Th concentration of the central peak region(Supplementary Information). Our modelling (Fig. 4) indicates

that the central peak could have a Th concentration as high as16.5±2.5 ppm. Based on a Clementine-derived FeO estimate, thecentral peak has FeO abundances ranging from 8 to 10wt%. Whencompared with values in the lunar sample suite these Th valuesare consistent with alkali-suite lithologies such as alkali norites29,though the composition of the pyroxenes22 is more magnesian thantypical for alkali-suite samples.

The association of OH− with KREEP is enigmatic. Though ithas been suggested that any native lunar OH−, such as KREEP,would be concentrated in late-stagemagma ocean fluids8, studies ofhydroxyl in lunar apatites with varying KREEP abundance reveal ananticorrelation between KREEP and OH−, suggesting complexityand heterogeneity in the volatile distribution in the lunar interior2.Bullialdusmay have exposed further source-region heterogeneity byexcavatingmaterial that is enriched in both volatiles andKREEP.

The Bullialdus-forming impact is likely to have excavatedabout 6–9 km deep into the lunar crust21. However, the impactoccurred near the edge of Mare Nubium, near the inner ringof the basin. This fortuitous placement of Bullialdus Crater mayexplain its unusual composition. One possible scenario is thatmaterial from a pluton, originally residing within the lower crust,was uplifted to 6–9 km beneath the surface during the Nubiumbasin-forming impact and subsequently excavated by the Bullialdus

NATURE GEOSCIENCE | VOL 6 | SEPTEMBER 2013 | www.nature.com/naturegeoscience 739

LETTERS NATURE GEOSCIENCE DOI: 10.1038/NGEO1909

Noritic Anorthositic

a b c

Figure 3 | Geology of Bullialdus Crater central peak. a, Perspective view of central peak, looking towards the south. LROC narrow-angle camera (NAC)images (Supplementary Fig. S5) overlain by 2.8 µm OH− band depth map from Fig. 1 and draped over LROC WAC 100 m digital elevation model30

(vertically exaggerated×6). The strongest absorptions occur where large boulders are observed on either side of the ridges. Arrows indicate where theimages in b,c are located. b,c, Subset from LROC NAC image showing the surface geology in noritic (b) and anorthositic (c) spectral regions. Scale bars,300 m.

6.24

6.09

5.95

5.80

5.65

5.51

5.36

5.21

5.07

4.92

4.77

Th abundance (ppm

)

16.66

15.00

13.33

11.67

10.00

8.33

6.67

5.00

3.33

1.67

0.00

Th abundance (ppm

)

Th abundance (ppm)

18.18

17.13

16.09

15.04

14.00

12.95

11.91

10.86

9.81

8.77

7.72

FeO abundance (w

t%)

¬20

¬20

¬20

¬20 ¬20

¬20

0.412 14 16 18 20

0.6

0.8

1.0

1.2

1.4

1.6

χ2

a

c

b

d

Figure 4 | Thorium content of Bullialdus Crater central peak. a, LP-GRS Th map (0.5◦×0.5◦) of Bullialdus Crater. Scale bar, 50 km. b, Clementine FeOmap of the crater. c, Forward modelling results for the region. Central peak is outlined in black in a–c. d, χ2 error analysis of the central peak, indicating thatthe optimum Th value for the peak is 16.5±2.5 ppm Th.

Crater-forming impact. Alternatively, deep fracturing during andfollowing the Nubium-Basin impact may have facilitated intrusionof hydroxyl-bearing magma into the upper (∼10-km-deep) crust.Our derived water/hydroxyl abundance is within the range ofOH− concentrations for nominally anhydrous crustal pyroxenes onEarth18. High concentrations (>1,000 ppm weight) are strongly afunction of orthopyroxene aluminium content and typically format pressures greater than 1GPa. A pressure of 1GPa would implythat any OH−-bearing pyroxenes crystallized at a depth of nearto 200 km, well below the excavation depth of Bullialdus or eventhe Nubium impact. A shallower origin is possible if the OH− ishosted within a minor hydrous mineral, as fluid inclusions, or ina quenched silicate melt. If the minor phase makes up 2% of the

bulk rock, it would require ∼0.2–0.6 wt% OH− to account forthe observed signature. This concentration is within the range ofOH− concentrationsmeasured in apatites frommare basalt samples(some >1.5 wt% OH−).

MethodsAll M3 and Lunar Reconnaissance Orbiter camera (LROC) data used in this projectare available in the NASA planetary data system at http://pds.nasa.gov. BullialdusCrater was fully imaged by M3 during three optical periods: 1b, which wasilluminated from the east (average phase angle 60◦) and has a spatial resolution of140m per pixel; 2a, which was illuminated from the west (average phase angle 68◦)and has a spatial resolution of 140m per pixel; and 2c1, which was illuminated fromthe west (average phase angle 40◦). Aristillus Crater was imaged in optical periods 1band 2c1. Data were mosaicked for each of the optical periods and then parameters

740 NATURE GEOSCIENCE | VOL 6 | SEPTEMBER 2013 | www.nature.com/naturegeoscience

NATURE GEOSCIENCE DOI: 10.1038/NGEO1909 LETTERSwere selected to isolate and characterize key absorption features for more easyvisualization of spectral variability in its spatial context (Supplementary Table S1).

Estimation of hydroxyl abundance in the central peak materials requiresconversion of spectra into an absorption coefficient, knowledge of the molarabsorption coefficient for the local lithology and an estimate of the path lengththrough the minerals. These calculations further require knowledge of surfacetemperature, particle size and other scattering properties. We can place limits onthe amount of hydroxyl present by making several simplifying assumptions aboutthe nature of the surface and mineralogy. One advantage to using an absorptioncoefficient as derived from the single-scattering albedo of a particulate surface is thatwe are not subject to orientation effects, because the observed signal is inherently anintegrated absorption average combining all crystallographic directions.

Level 2c1 data were specifically selected for modelling hydroxyl abundancebecause they were obtained at the lowest phase angle, which minimizes shadowingand multiple scattering effects4 on the steep slopes of the central peak. We convertthe spectra to single-scattering albedo using themethods of ref. 23 and then solve forthe absorption coefficient using a particle size of 45 µm. To derive single-scatteringalbedo, we calculate the albedo factor23 using the incidence and emission anglesand the reflectance values at each wavelength. Albedo factor is then convertedto single-scattering albedo23. If we approximate the mean path ray length (inmicrometres) through a single particle as 45 µm, we can approximate absorbance23.This requires the use of a refractive index (n), which we calculate24 using pyroxenewith an Mg# of 0.75 as an estimate for the composition of the central peak ofBullialdus22. Single-scattering albedo and calculated absorbance spectra for thecentral peak spectra are shown in Supplementary Fig. S3.

The baseline-removed hydroxyl absorption intensity of CP1 was fit by aGaussian curve and then the centre frequency (in wavenumbers) of the OH−stretching absorption was used to calculate the integrated molar absorptioncoefficient (εi; ref. 25). As this calibration for εi uses units of water per mol ratherthan H or OH−, we used the formulation of the Beer–Lambert law shown inequation (1) to determine the concentration of water. The area of the Gaussian wasused as the integrated absorption intensity (Ai), the path length of 45 µm was usedfor thickness (t ) and density of 3.0 g cm−3 was used for the bulk rock density (ρ).This results in a concentration (c) of 80 ppm for the chosen particle size of 45 µm.

c =Ai

t×p×εi×1.8 (1)

To ensure that this estimate was not highly dependent on the composition ofthe mineral hosting the hydroxyl, we calculated the absorption coefficient using therefractive index of anorthite and quartz as well. In those cases, the baseline-removedabsorption coefficients were indistinguishable from the orthopyroxene and resultedin the same concentration at a fixed grain size. As the scattering properties of therocks are controlled by the dominant mineral phase(s), we did not carry out ananalysis using apatite. For a given concentration of water, the strength of an OH−absorption in apatite is comparable to the OH− strength in a nominally anhydrousmineral18,19. As any minor phase would comprise only a few per cent of the totalrock, the amount of hydroxyl in that phase would be significantly higher (that is, weare calculating the abundance in the bulk path length, not in the host phase per se).

In the laboratory, water content calculations based on infrared spectra aresubject to typically 10–30% relative uncertainty depending on the strength of theabsorption band and the approximation of the molar absorption coefficient12,25.As the calculation of single-scattering albedo is heavily dependent on grain size,additional uncertainty is introduced. If the average grain size of the surface is closerto 15 µm, the water abundance would be approximately 20% higher, whereas if itis closer to 150 µm the water content would be approximately 20% lower, resultingin a range of∼45–125 ppm.

Received 19 April 2013; accepted 5 July 2013; published online25 August 2013

References1. Saal, A. E. et al. Volatile content of lunar volcanic glasses and the presence of

water in the Moon’s interior. Nature 454, 192–195 (2008).2. McCubbin, F. M. et al. Fluorine and chlorine abundances in lunar apatite:

Implications for heterogeneous distributions of magmatic volatiles in the lunarinterior. Geochim. Cosmochim. Acta 75, 5073–5093 (2011).

3. Pieters, C. M. et al. Character and spatial distribution of OH/H2O onthe surface of the Moon seen by M3 on Chandrayaan-1. Science 326,568–572 (2009).

4. Clark, R. N. Detection of adsorbed water and hydroxyl on the Moon. Science326, 562–564 (2009).

5. Sunshine, J. M. et al. Temporal and spatial variability of lunar hydration asobserved by the Deep Impact spacecraft. Science 326, 565–568 (2009).

6. Feldman,W. et al. Fluxes of fast and epithermal neutrons from lunar prospector:Evidence for water ice at the lunar poles. Science 281, 1496–1500 (1998).

7. Colaprete, A. et al. Detection of water in the LCROSS ejecta plume. Science330, 463–468 (2010).

8. McCord, T. B. et al. Sources and physical processes responsible for theOH/H2O in the lunar soil as revealed by the Moon Mineralogy mapper (M3).J. Geophys. Res. 116, E00G05 (2011).

9. Arnold, J. R. Ice in the lunar polar regions. J. Geophys. Res. 84,5659–5668 (1979).

10. Hibbitts, C. C. et al. Thermal stability of water and hydroxyl on the surface ofthe Moon from temperature-programmed desorption measurements of lunaranalog materials. Icarus 213, 64–72 (2011).

11. Crider, D. H. & Vondrak, R. R. The solar wind as a possible source of lunarpolar hydrogen deposits. J. Geophys. Res. 105, 26773–26782 (2000).

12. Liu, Y. et al. Direct measurement of hydroxyl in the lunar regolith and theorigin of lunar surface water. Nature Geosci. 5, 779–782 (2012).

13. Jolliff, B. L., Gillis, J. J., Haskin, L. A., Korotev, R. L. & Wieczorek, M. A.Major lunar crustal terranes: Surface expressions and crust–mantle origins.J. Geophys. Res. 105, 4197–4216 (2000).

14. Lawrence, D. J. et al. Small-area thorium features on the lunar surface.J. Geophys. Res. 108, JE002050 (2003).

15. Pieters, C. M. Bullialdus—strengthening the case for lunar plutons. Geophys.Res. Lett. 18, 2129–2131 (1991).

16. Tompkins, S., Pieters, C. M., Mustard, J. F., Pinet, P. & Chevrel, S. D.Distribution of materials excavated by the lunar crater Bullialdus andimplications for the geologic history of the Nubium region. Icarus 110,261–274 (1994).

17. Cahill, J. T. & Lucey, P. G. Radiative transfer modeling of lunar highlandsspectral classes and relationship to lunar samples. J. Geophys. Res. 112,10007 (2007).

18. Johnson, E. A. inWater in Nominally Anhydrous Minerals Vol. 62 (eds Keppler,H. & Smyth, J. R.) 117–154 (Mineralogical Society of America, 2006).

19. Wang, K. L., Zhang, Y. X. & Naab, F. U. Calibration for IR measurements ofOH in apatite. Am. Mineral. 96, 1392–1397 (2011).

20. Lucey, P. G., Blewett, D. T., Taylor, G. J. & Hawke, B. R. Imaging of lunarsurface maturity. J. Geophys. Res. 105, 20377–20386 (2000).

21. Cahill, J. T. S., Lucey, P. G. &Wieczorek, M. A. Compositional variations of thelunar crust: Results from radiative transfer modeling of central peak spectra.J. Geophys. Res. 114, 09001 (2009).

22. Klima, R. L. et al. New insights into lunar petrology: Distribution andcomposition of prominent low-Ca pyroxene exposures as observed by theMoon Mineralogy Mapper (M3). J. Geophys. Res. 116, E00G06 (2011).

23. Hapke, B.Theory of Reflectance and Emittance Spectroscopy 2nd edn (CambridgeUniv. Press, 2012).

24. Lucey, P. G. Model near-infrared optical constants of olivine and pyroxene as afunction of iron content. J. Geophys. Res. 103, 1703–1713 (1998).

25. Libowitzky, E. & Rossman, G. An IR absorption calibration for water inminerals. Am. Mineral. 82, 1111–1115 (1997).

26. Lawrence, D. J. et al. Improved modeling of Lunar Prospector neutronspectrometer data: Implications for hydrogen deposits at the lunar poles.J. Geophys. Res. 111, E08001 (2006).

27. Mitrofanov, I. G. et al. Lunar exploration neutron detector for the NASA lunarreconnaissance orbiter. Space Sci. Rev. 150, 183–207 (2010).

28. Hagerty, J. J. et al. Refined thorium abundances for lunar red spots:Implications for evolved, nonmare volcanism on the Moon. J. Geophys. Res.111, E06002 (2006).

29. Papike, J. J., Ryder, G. & Shearer, C. K. in Planetary Materials Vol. 36 (ed.Papike, J. J) 1–189 (Mineralogical Society of America, 1998).

30. Scholten, F. et al. GLD100: The near-global lunar 100m raster DTM fromLROCWAC stereo image data. J. Geophys. Res. 117, E00H17 (2012).

AcknowledgementsWe thank the NASA Lunar Advanced Science and Engineering Program (NNX10AH62Gto RK/JHUAPL), the NASA National Lunar Science Institute Polar Exploration Node(NNA09DB31A to JHUAPL) and the NASA Planetary Mission Data Analysis Program(NNH09AL42I to JH/USGS) for supporting this research. We are also grateful to theNASADiscovery Program, Indian Space ResearchOrganization andM3 team.

Author contributionsAll authors contributed extensively to this work. R.K. wrote the main manuscript withcomments and feedback from the whole team. R.K. led analysis and modelling of M3

data, J.C. led processing and analysis of LROC data and J.H. and D.L. contributed thethorium analyses and forward modelling.

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.K.

Competing financial interestsThe authors declare no competing financial interests.

NATURE GEOSCIENCE | VOL 6 | SEPTEMBER 2013 | www.nature.com/naturegeoscience 741