The Chlorine Isotope Composition of the Moon and Implications for an Anhydrous Mantle

See allHide authors and affiliations

Science  27 Aug 2010:
Vol. 329, Issue 5995, pp. 1050-1053
DOI: 10.1126/science.1192606


Arguably, the most striking geochemical distinction between Earth and the Moon has been the virtual lack of water (hydrogen) in the latter. This conclusion was recently challenged on the basis of geochemical data from lunar materials that suggest that the Moon’s water content might be far higher than previously believed. We measured the chlorine isotope composition of Apollo basalts and glasses and found that the range of isotopic values [from –1 to +24 per mil (‰) versus standard mean ocean chloride] is 25 times the range for Earth. The huge isotopic spread is explained by volatilization of metal halides during basalt eruption—a process that could only occur if the Moon had hydrogen concentrations lower than those of Earth by a factor of ~104 to 105, implying that the lunar interior is essentially anhydrous.

The origin of the Moon is constrained by geophysical models of angular momentum, density and mass, and geochemical arguments based on chemical and isotopic similarities and differences between Earth and the Moon. It is now generally accepted that the Moon was formed from the impact of a Mars-sized body sometime after formation of Earth’s core (1, 2), although a number of observations have not been adequately explained. Geodynamic models indicate that the impactor came from a different region of our solar system and that the bulk of the Moon originated from the impactor rather than Earth (3). This result is at odds with geochemical data, such as the fact that Earth and the Moon have identical oxygen and chromium isotope ratios (4, 5). If the two bodies came from different regions of our solar system, it is expected that they would have different isotope ratios (6). Turbulent mixing between the molten Earth and the Moon immediately after impact has been proposed as a means of homogenization (6), but such intense mixing would also tend to homogenize volatiles between the two bodies. The recognition that the Moon is strongly depleted in volatile elements and is effectively anhydrous was established soon after the return of Apollo samples (7, 8). Several explanations for the absence of hydrogen in the Moon have been suggested: an anhydrous impactor (9), exclusion of water in the Moon because of low water solubility in molten silicate at high temperature and low pressure (10), or Earth and the Moon both having been anhydrous during formation, with volatiles delivered to Earth at a later date (11).

Recently, it was proposed that the water content of the Moon may be far higher than previously recognized. Saal et al. (12) measured water contents of lunar glass beads and concluded that the water content of the Moon may be equivalent to that of typical terrestrial mid-ocean ridge basalts (MORBs). Similar high water contents of lunar basalts have been estimated from H contents of apatite grains (1315). Here, we present Cl isotope data from lunar basalts, volcanic glasses, and apatite grains to assess the water content of the Moon. Chlorine is extremely hydrophilic, tracking closely with water on Earth (16). It is also incompatible in nearly all silicates and extremely volatile. If the Moon were anhydrous, the unique degassing and partition behavior of Cl should lead to Cl isotope fractionations that are distinct from those observed on Earth.

We used gas source mass spectrometry to determine Cl isotope ratios and concentrations of water-soluble leachates and silicate residues from lunar basalts and glasses. A wide range of sample types were chosen to allow us to assess the effects of chemical composition/mineralogy and surface processing. Apatite grains were analyzed in situ with the Cameca 1270 ion microprobe at the University of California, Los Angeles (UCLA). To assess the potential effect of solar wind on the Cl isotope ratio of the lunar regolith and soils, we irradiated a thin film of NaCl with a 100-keV proton beam at the Ion Beam Materials Laboratory, Los Alamos National Laboratory, and then analyzed the irradiated sample for its Cl isotope composition (17).

The Cl isotope compositions of typical terrestrial materials and meteorites are clustered around 0‰ (18) or –1.6‰ (19) with a spread of ±0.5‰ (Fig. 1). Large variations in δ37Cl values are seen only in the specific case where HCl gas is evolved from volcanic fumaroles (16). The enrichment of 37Cl in the vapor is due to the large isotopic fractionation between the solvated Cl ion in an aqueous solution and HCl gas.

Fig. 1

Chlorine isotope composition of selected terrestrial materials and lunar samples (triangles, ion microprobe; diamonds, gas source mass spectrometry). Silicate and leachate samples are shown as solid and open symbols, respectively. Matched leachate-silicate pairs are circled. Terrestrial and meteorite data are from (18, 32, 33), and references therein. Error bars are smaller than the size of symbols. See (18) for meteorite abbreviations.

In contrast to terrestrial basalts, lunar samples show a range of δ37Cl values from –1 to +24‰ (Fig. 1). The large spread is indicative of conditions that do not exist on Earth. We consider three possible explanations for the large spread: (i) initial heterogeneities retained from the Moon-forming event; (ii) isotopic fractionation associated with solar wind and micrometeorite bombardment; and (iii) devolatilization and isotopic fractionation specific to the anhydrous character of lunar basalts.

The first of these possibilities is unlikely. Volatilization and fractionation could certainly occur during accretion of the Moon with preferential loss of the light isotope during volatile escape to space. However, maintaining such large isotopic variations in a system that was certainly molten and went through a magma ocean stage (20) would not be expected and has not been observed for other volatile elements, such as K (21).

The second scenario is similarly unlikely. Because micrometeorite bombardment and solar wind sputtering can preferentially remove the light isotopes of volatile elements, we chose samples that have experienced both minimum and maximum degrees of surface processing (Table 1) to test for a correlation between the extent of surface exposure (soil maturity) and δ37Cl values. Lunar soil sample 64501,232, with a high maturity index and the highest δ34S value (+11.7‰) of any of our samples, has an intermediate δ37Cl value, similar to that of immature soil sample 61220,39 (22), arguing against preferential 35Cl loss due to formation and removal of light HCl (HCl gas) to space. In addition, the δ37Cl value of the surface volatile chlorides are always lower than the rock-bound Cl in the same sample, in some cases as much as 10‰. If the surface chlorides interacted with solar wind, they should be enriched in 37Cl, not depleted. Apatite grains in KREEP (potassium–rare earth element–phosphorus) basalt sample 72275,491 were almost certainly shielded from solar wind, and yet this sample has the highest δ37Cl value of all measured samples. As a final test, a thin film of NaCl was found to be isotopically unfractionated after bombarded by a 5 × 1017 ions/cm2 proton source (Table 1). Preferential formation of H35Cl during proton bombardment should raise the δ37Cl values of the residual salt, a trend that is not seen here.

Table 1

Cl contents and δ37Cl values for lunar samples and irradiated salt experiment. Initial Cl contents were calculated assuming Rayleigh distillation, an α value of 0.99, and an initial δ37Cl value of 0‰. SMOC, standard mean ocean chloride. All apatite grains were analyzed by secondary ion mass spectrometry.

View this table:

The third possibility, that devolatilization of anhydrous lunar basalts caused the range of δ37Cl values is supported by several trends. First, the lowest measured δ37Cl values of all samples overlap the range for both terrestrial and meteorite samples. This is consistent with the idea of a well-mixed Cl isotope ratio for the inner solar system. All higher δ37Cl values, as discussed below, are explained by fractionation during degassing. Second, the δ37Cl values of the leached material, representing vapor deposition of the volatile chloride species, are always lower than the coexisting silicate. These data strongly support the idea that the light isotope of Cl is lost during devolatilization.

In terrestrial magmas and lavas, there is always an excess of H relative to Cl, and the dominant volatile chloride species is HCl gas. The fractionation between HCl gas and a basalt melt must be close to 0‰, because the δ37Cl values of terrestrial basalts have very little variation, despite varying Cl contents. Two factors control Cl isotope fractionation during volatilization on Earth: (i) the preferential loss of 35Cl to the vapor phase (due to its higher translational velocities and vapor pressure relative to 37Cl), and (ii) the preferential incorporation of 37Cl into HCl gas (due to the covalent character of the bond) (16, 23). The two factors tend to cancel one another, resulting in near-zero fractionation between melt and vapor. In a water-free melt, the volatile chloride species are metal chlorides (24, 25). Under such conditions, the bonding of Cl in the vapor and melt phase is similar, so that the preferential kinetic loss of the light isotope is not cancelled out by a different bonding energies of the melt and vapor phase. Vapor loss should increase the δ37Cl value of the residual magma, which explains the extraordinary isotopic range of lunar samples. NaCl, ZnCl2, FeCl2, and other metal chlorides have been observed as surface coatings on lunar volcanic materials (26, 27), consistent with deposition of metal chloride vapor phases. Thermodynamic calculations also support metal halide vapor as the dominant Cl species when the Cl/H ratio is substantially greater than 1 (8).

The change in the δ37Cl value of the melt during vapor loss of a metal chloride can be modeled assuming Rayleigh distillation, where δfinal = δinitial + [(1000 + δinitial)(F(α–1) – 1)], F is the fraction of Cl remaining in the silicate melt, and α is the fractionation between vapor and melt given by α = [(1000 + δvapor)/(1000 + δsilicate melt)]. For volatilization of a liquid into a vacuum, α=m1/m2, where m1 and m2 are the masses of the light and heavy isotopolog of the vaporizing species (28). For the common chloride species NaCl, FeCl2, and ZnCl2, α is 0.983, 0.992, and 0.993, respectively. These fractionations are similar to the measured Δ37Cl (silicate−vapor) value from samples 64501,232 (α = 0.990) and 61220,39 (α = 0.992). Actual fractionations less than those computed from relative mass differences have been observed experimentally (29) and are expected when diffusion rates are insufficient to maintain isotopic homogeneity in the melt. Nonetheless, the lighter isotope diffuses faster through the melt, so that fractionation during volatilization will still occur. Volatilization can only increase the δ37Cl value of the residual Cl in the melt. The lowest value of all samples is –0.7‰, measured on leachates of a fire-fountain glass (sample 74002,36). This value is similar to bulk Earth and meteorite samples (18) and may be slightly lighter than bulk Moon because it is a vapor condensate sample and is likely the product of some fractionation. This sample is also anomalously volatile rich and is the only sample that has indigenous hydrogen (12). As a result, the volatile Cl species could have been HCl, and little fractionation during degassing would be expected. Mare basalt sample 12040,214 has both δ37Cl and δ34S values of 0‰, suggesting minor fractionation during degassing. We conclude that the undifferentiated Moon appears to have a δ37Cl value close to 0‰, similar to Earth.

Initial Cl contents were calculated assuming Rayleigh distillation and an initial δ37Cl value of 0‰ (17). The amount of Cl vapor loss necessary to reach a 37Cl enrichment equivalent to a δ37Cl value of +20‰ ranges from 63 to 98% for a value of α between 0.983 and 0.992, respectively (Fig. 2). These are minimum degrees of degassing because diffusion-limited exchange will cause a reduction in the fractionation between melt and vapor. The calculated Cl contents of the undegassed basalts (and apatite grains for the ion microprobe analyses) have a positive correlation with the δ37Cl values (Table 1). It is reasonable that samples with high initial Cl contents would lose proportionally more Cl than Cl-poor ones, and therefore evolve to higher δ37Cl values.

Fig. 2

Calculated trajectories of δ37Cl values of basalt during loss of metal chloride volatiles. A δ37Cl value of 20‰ is obtained at >70% (NaCl) to >90% (FeCl2 or ZnCl2) volatilization (F < 0.3) assuming an initial δ37Cl value of 0‰. The α values of 0.983 and 0.992 are obtained from the relationshipEmbedded Image for NaCl and FeCl2 or ZnCl2, respectively.

The H contents of lunar magmas must have been substantially lower than their Cl contents, or else HCl gas rather than metal chlorides would have been the volatilizing species, and no isotopic fractionation would have occurred. Excepting one pigeonite basalt sample, the calculated initial Cl concentrations of three undifferentiated lunar basalts are all close to 40 parts per million (ppm) (Table 1). Cl volatilization is expected to have occurred only after 90 to 95% crystallization has taken place (14), concentrating Cl in the remaining liquid. Primitive melts therefore had Cl concentrations that were lower by at least a factor of 10, or ~4 ppm. Finally, primitive melts generated in the mantle should be concentrated in incompatible elements relative to bulk lunar mantle by a factor of ~10 (30), so that the lunar mantle should have a Cl concentration of ~400 parts per billion (ppb), with H contents around 10 ppb (given that the Cl/H atomic mass unit ratio is 35.45/1.01), versus the bulk Earth value of ~260 ppm (17, 31).

Isotopic anomalies in lunar samples have been seen in other elemental systems, but never with the extreme variations observed for the Cl isotopes. What makes Cl unique is its hydrophilic affinity and high volatility relative to all other isotopic systems studied in lunar samples. It is therefore an extremely sensitive indicator of H contents. If lunar basalts had initial H contents anywhere close to terrestrial magmas (e.g., MORB), then the extreme fractionations observed in the lunar materials would not have occurred. We suggest that the Moon was anhydrous as initially proposed (7) and that the high H contents calculated for some lunar samples are atypical. High H contents are found only in anomalous samples with extreme volatile contents and are likely the product of igneous processes that resulted in extreme volatile enrichment. In contrast to other samples, they have δ37Cl values close to 0‰ and are inconsistent with the high and variable δ37Cl values reported for the majority of lunar samples.

Supporting Online Material

Materials and Methods

Table S1


References and Notes

  1. See supporting material on Science Online.
  2. Supported by NASA Origins grant 07-SSO07-0077 (Z.D.S.), NASA LASER grant NNX08AY786 (C.K.S.), and a NASA Cosmochemistry grant (K.D.M.). The UCLA ion microprobe facility is partially supported by a grant from the NSF Instrumentation and Facilities Program. We thank lunar curator G. Lofgren and staff of the Johnson Space Center, Houston. P. Burger provided imaging and quantitative elemental analyses of apatite. The Ion Beam Materials Laboratory at Los Alamos is partially supported by DOE Office of Basic Energy Sciences programs.
View Abstract

Stay Connected to Science

Navigate This Article