Lunar Fe and Ti Abundances: Comparison of Lunar Prospector and Clementine Data

See allHide authors and affiliations

Science  04 Sep 1998:
Vol. 281, Issue 5382, pp. 1493-1496
DOI: 10.1126/science.281.5382.1493


The Lunar Prospector neutron spectrometer data correlate well with iron and titanium abundances obtained through analysis of Clementine spectral reflectance data. With the iron and titanium dependence removed, the neutron spectrometer data also reveal regions with enhanced amounts of gadolinium and samarium, incompatible rare earth elements that are enriched in the final phases of magma crystallization. These regions are found mainly around the ramparts of the Imbrium impact basin but not around the other basins, including the much larger and deeper South Pole–Aitken basin. This result confirms the compositional uniqueness of the surface and interior of the Imbrium region.

The surface of the moon provides a record of the early evolution of the Earth-moon system through the period of heavy bombardment ∼4 billion years ago. In contrast, most terrestrial rocks are much younger. Throughout the moon's history large impacts have excavated material from the lower crust (and possibly the mafic upper mantle) and deposited it on the surface. Basaltic volcanism during and after the heavy bombardment epoch flooded some impact basins with material derived from partial melting of the upper mantle and lower crust. Study of these materials thus provides a window into the moon's interior. Whereas most of our understanding of lunar composition is derived primarily from returned Apollo and Luna samples, the question remains to what extent these samples are representative of the whole moon.

A major step toward a global assessment of lunar surface chemistry was provided by analysis of spectral reflectance data returned by the Clementine mission. Comparison of the spectral reflectance properties and chemical compositions of lunar soils returned by the Apollo and Luna missions, and remote measurement of the spectral properties of the Apollo and Luna sample collection sites by Clementine, led to the development of algorithms that derive the abundance of FeO and TiO2 from spectral properties of lunar soils and surface units with 1 to 2 weight % accuracy. By applying these algorithms to Clementine global imaging, it has been possible to infer the quantitative abundance of FeO and TiO2 within ±80° latitude (1–4) at resolutions approaching 100 m. The approach was calibrated with lunar samples returned from a relatively small area of the lunar nearside, and it is possible that areas distant from the landing sites, such as the farside, might have different mineralogies. Thus, the inferred FeO and TiO2 values might be spurious there (5). Here, we test the validity of the Clementine spectral reflectance (CSR) method with independent analyses of Fe and Ti, using data from the Lunar Prospector (LP) neutron spectrometer (6, 7).

The neutron spectrometer measures neutrons at thermal (0.001 to 0.3 eV), epithermal (0.3 eV to 500 keV), and fast (500 keV to 8 MeV) energies. Lunar neutrons are created by the interaction of galactic cosmic rays with the nuclei in the lunar regolith; they are produced in the fast regime at high energies as a direct result of spallation. These neutrons inelastically scatter off of other nuclei in the soil, losing energy as they pass through the epithermal regime. When their energies approach that corresponding to the temperature of the ambient regolith (thermal regime), the neutrons are captured by nuclei that have large cross sections for thermal neutron absorption (8). Iron and titanium are the most abundant elements with large absorption cross sections; consequently, they have considerable influence on thermal neutron fluxes. In addition, Fe and Ti evidently produce more fast neutrons than elements with lower atomic numbers (9). Consequently, when LP is above the Fe- and Ti-rich maria, the neutron spectrometer detects a higher fast neutron flux and a lower thermal neutron flux (6, 7). We report on neutron data acquired during the first 6 months of the LP mapping mission, beginning 16 January 1998.

The neutron spectrometer measures the net production of fast neutrons from the elements in the regolith below LP and the net flux of thermal neutrons resulting from production, moderation, and absorption due to all species. Consequently, it is not possible to directly infer the abundances of Fe and Ti separately with neutron spectrometer data alone. Instead, we use the most recently generated CSR Fe and Ti maps from (4) to predict how much neutron production and absorption should be seen and compare that prediction to what was actually observed. Any differences must be due to either a surplus or a lack of neutron absorbers, that is, Fe or Ti in unseen mineral phases, or other possible neutron absorbers.

To compare the CSR maps [which have a surface resolution of 0.25° (7.5 km)] to the neutron observations from LP, we convolved the maps with the effective surface response functions of the neutron detector. For thermal neutrons, the relevant surface area has an effective footprint diameter of about 700 km for a 100-km orbit. The fast neutron footprint is smaller, 350 km in diameter.

The fast neutron flux should correlate with the Fe and Ti content in the regolith. The maps shown in (7) are in qualitative agreement with the published CSR Fe and Ti abundances (4). Overall, the correlation coefficient between the fast neutron count rate and the CSR Fe+Ti weight % values is 0.811. For a more restricted latitude range (where possible problems associated with lighting in the CSR data are reduced), ±60°, the correlation improves to 0.887. For a region including the eastern nearside maria and some farside highlands (40° to 180°E longitude, ±60° latitude), the correlation coefficient is 0.930, whereas for the nearside maria region alone (90°W to 90°E longitude, −30° to 60° latitude), it is 0.941.

LP detected a low flux of thermal neutrons over the maria and South Pole–Aitken basin (7), indicative of the presence of high concentrations of Fe and Ti. The net thermal neutron absorption in a given volume of regolith is related to the macroscopic absorption cross sectionEmbedded Image(1)wherefi,σa i , andAi are the weight fraction, thermal neutron absorption cross section, and atomic mass of elementi, respectively, and N A is Avogadro's number. Table 1 shows the contribution of the various elements to the total macroscopic absorption cross section for three bulk regolith compositions: an Apollo 11 high-Ti mare soil, an Apollo 16 low-Fe highlands soil, and an Apollo 14 KREEP basalt (KREEP is a component of some lunar rocks with a composition enriched in potassium, K, rare earth elements, REE, and phosphorus, P). Fe and Ti are the biggest contributors to absorption in the mare soil; Fe is also the chief absorber (despite its low abundance) in the highlands soil, but Ca is almost as important. Silicon, Al, and the other major elements play a lesser, but not negligible, role in thermal neutron absorption.

Table 1

Contributions to Σeff, thermal neutron absorption. Ap, Apollo mission.

View this table:

Because Ca abundance is so variable on the moon, we must include its effects in our calculations. We make use of an observed inverse correlation between CaO and FeO in lunar samples to do this (10). In samples with very low FeO abundances, CaO reaches about 20% by weight; as FeO approaches about 12%, CaO abundance drops linearly to about 10% and stays at about that level in materials with higher FeO abundances. We assume that the net absorption due to Si, Al, and other major elements is approximately constant at 18 × 10–4 cm2/g.

Certain REEs have anomalously high cross sections for absorption of thermal neutrons because of nuclear resonances (11). Consequently, the effects of these rare earth elements can be disproportionate to their low chemical abundances. Gadolinium and Sm, in particular, can significantly affect Σeff in regions where they are abundant (>5 μg/g; see Table 1). Gadolinium and Sm are most abundant in KREEP, along with other incompatible elements such as Th, K, and U. These elements are enriched in the final phases of magma crystallization, and for this reason KREEP is thought to be derived from the lower crust. In the Apollo 14 KREEP basalts, the Gd and Sm contribution to neutron absorption is comparable with the total from all other elements and increases Σeff from 50.8 × 10−4 to 93.2 ×10−4 cm2/g (12). The contributions of Sm and Gd to the Apollo 11 mare and Apollo 16 highlands Σeff values are about 10 and 13%, respectively (Table 1). We have not included the effects of Gd and Sm in our estimates of Σeff based on CSR-derived values of Fe and Ti abundances.

Simulations of neutron transport in lunar and martian regoliths have shown that the ratio of the fluxes of fast to thermal neutronsj f /j th is linearly related to Σeff (12). This is what would be expected: the more neutron absorbers in the regolith, the fewer the number of thermal neutrons observed per fast neutron created. Thus, j f /j th is the neutron measurement proxy that directly relates to the macroscopic absorption cross section Σeff.

The macroscopic absorption cross section Σeff inferred from the CSR Fe and Ti results (Fig. 1) is high over western Oceanus Procellarum (50°W 15°N), Mare Imbrium (25°W 35°N), and Mare Tranquillitatis (25°E 10°N). The value of Σeff is also high in the South Pole–Aitken basin mainly because of the basin's high Fe content. The ratio of the fast to thermal neutron fluxes,j f /j th, as observed by LP, shows similar patterns but also some differences, particularly around the periphery of Mare Imbrium (Fig. 2). To compare these two data sets, we plot the neutron flux ratioj f/j th versus the inferred CSR macroscopic absorption cross section Σeff(Fig. 3). The observed absorption of thermal neutrons is often greater than would be expected on the basis of CSR Fe, Ti, and Ca.

Figure 1

(top). Macroscopic absorption cross section Σeff as determined from CSR Fe and Ti abundances. The original 0.25° by 0.25° map was smoothed with the thermal neutron response function but is binned in 2° by 2° (∼60 km by 60 km) pixels. Local highs in Σeff can be seen over the nearside maria and over the South Pole–Aitken basin, where Fe and Ti concentrations are higher than the surrounding highlands.

Figure 2

(bottom). Ratio of the LP fast-to-thermal neutron count rates,j f /j th. This quantity should be directly related to the abundance of thermal neutron absorbers in the regolith. Highs similar to Fig. 1 are found over the maria and the South Pole–Aitken basin. Data from the first 6 months of LP mapping are shown.

Figure 3

Fast-to-thermal neutron count rate ratioj f/j th plotted versus CSR-derived Σeff for data from ±60° latitudes (black points). Red points represent data from a more restricted region (see text). The solid line is an ideal linear relation that would be expected if Fe, Ti, and Ca are the most important absorbing species. Flux ratio values to the left of this line indicate that the calculated values of Σeff based on CSR data are too low, and other neutron absorbers are present.

The overall global correlation between Σeff calculated with the CSR Fe and Ti values and the proxy for macroscopic absorption coefficient,j f/j th, is 0.849. By restricting the latitude range to ±60° (black points in Fig. 3), the correlation improves to 0.903. A still more limited region covering the eastern maria and some farside highlands, 20°E to 180°E longitude and ±30° latitude (red points in Fig. 3), yields a correlation coefficient of 0.978. Thus, our calculated values of Σeff based on CSR Fe and Ti abundances correlate well with j f/j th in some regions but not as well in others.

A map of the difference ΔΣeff between this linear relation and the calculated Σeff points (Fig. 3) reflects the implied abundance of neutron absorbers other than Fe and Ti (Fig. 4). The highest values are found over the rim of the Imbrium basin, including the Apennine mountains in the east up through the Alps, across to the Jura mountains in the west, down through the Aristarchus plateau and through the Fra Mauro formation to the south. LP gamma-ray spectrometer maps show that these regions also have high Th and K concentrations (13). This similarity implies that the absorbing species are affiliated with the incompatible elements found in KREEP; the correlation coefficient between the gamma-ray spectrometer Th data (13) and ΔΣeff is 0.93. The values of ΔΣeff are consistent with the range (10 × 10−4 to 42 × 10−4 cm2/g) of Gd and Sm contributions listed in Table 1. The K in KREEP plays only a very minor role in Σeff (Table 1). Therefore, we conclude that ΔΣeff reflects primarily the concentration of Gd and Sm and, thus, is a tracer for KREEP.

Figure 4

Map of ΔΣeff, a measure of thermal neutron absorbers in addition to Fe, Ti, and other major elements implied by CSR-derived abundances. The greatest values of ΔΣeff are found surrounding the Imbrium basin, approximately colocated with regions rich in KREEP.

The estimated thermal neutron macroscopic absorption coefficient that would be expected on the basis of concentrations of Fe and Ti derived from CSR data, coupled with an estimate of Ca concentrations, is in reasonable agreement with the LP neutron spectrometer results. Discrepancies arise in regions that have significant levels of KREEP, where Gd and Sm are major thermal neutron absorbers. Thus, the CSR method appears to be a reliable technique for obtaining FeO and TiO2 abundances moon-wide. The inferred KREEP-rich regions form a ring around the Imbrium impact site and are directly related to either excavation of this lower crustal chemistry or to volcanism that extruded KREEP-rich lava on the surface. On the other hand, the much larger, deeper South Pole–Aitken impact basin shows little KREEP enhancement (14). This result appears to confirm the uniqueness of the Imbrium lower crustal chemistry and suggests that the moon may have considerable regional compositional heterogeneity at depth.

  • * To whom correspondence should be addressed. E-mail: relphic{at}


Stay Connected to Science

Navigate This Article