The Cl isotope composition and halogen contents of Apollo-return samples

Significance

Chlorine isotopes are a sensitive tracer of degassing throughout planetary evolution that provide evidence for the universal depletion of volatiles in the Moon. We show that much of the chlorine in mare basalts is trapped in water-soluble phases from vapor deposition with low isotope values, with the remaining being isotopically heavy from degassing. We also use halogen concentrations and bulk-Cl isotope values to show that most lunar halogen loss and heavy Cl enrichment occurred during the Giant Impact—resulting in a 10× depletion of halogens relative to the Earth. Last, we conclude that lunar apatite has much higher δ³⁷Cl values compared to the bulk rock, likely explained by localized degassing, making their use as direct probes of planetary-scale processes problematic.

Abstract

Lunar mare basalts are depleted in F and Cl by approximately an order of magnitude relative to mid-ocean ridge basalts and contain two Cl-bearing components with elevated isotopic compositions relative to the bulk-Earth value of ∼0‰. The first is a water-soluble chloride constituting 65 ± 10% of total Cl with δ³⁷Cl values averaging 3.0 ± 4.3‰. The second is structurally bound chloride with δ³⁷Cl values averaging 7.3 ± 3.5‰. These high and distinctly different isotopic values are inconsistent with equilibrium fractionation processes and instead suggest early and extensive degassing of an isotopically light vapor. No relationship is observed between F/Cl ratios and δ³⁷Cl values, which suggests that lunar halogen depletion largely resulted from the Moon-forming Giant Impact. The δ³⁷Cl values of apatite are generally higher than the structurally bound Cl, and ubiquitously higher than the calculated bulk δ³⁷Cl values of 4.1 ± 4.0‰. The apatite grains are not representative of the bulk rock, and instead record localized degassing during the final stages of lunar magma ocean (LMO) or later melt crystallization. The large variability in the δ³⁷Cl values of apatite within individual thin sections further supports this conclusion. While urKREEP (primeval KREEP [potassium/rare-earth elements/phosphorus]) has been proposed to be the source of the Moon’s high Cl isotope values, the ferroan anorthosites (FANs) have the highest δ³⁷Cl values and have a positive correlation with Cl content, and yet do not contain apatite, nor evidence of a KREEP component. The high δ³⁷Cl values in this lithology are explained by the incorporation of a >30‰ HCl vapor from a highly evolved LMO.

Over the 50 y since the Apollo missions, chemical and isotopic compositions of lunar materials have revolutionized our collective scientific understanding of planetary materials. A major finding is the recognition that lunar materials are strongly depleted in volatiles relative to Earth and that they exhibit a range of volatile-element stable isotope anomalies (1–16). Studying these isotopic systems allows us to better understand volatile sources and mechanisms by which volatile elements are lost throughout planetary evolution. The chlorine isotope system is of particular interest because the Cl isotope values of the Moon are uniquely high. The high values, and increases in the Cl isotope composition of a planetary body from the nebular baseline (around −5 to −7‰ compared to Earth), are related to the amount of Cl that was lost throughout its accretionary and differentiation history (17, 18).

Lunar materials have extremely elevated δ³⁷Cl values relative to the bulk-Earth value of ∼0‰ (14, 19). Sharp et al. (14) measured the bulk and in situ Cl isotope compositions of a variety of lunar lithologies and observed a range in δ³⁷Cl values from −0.7 to 24.5‰. These authors interpreted these results to suggest that anhydrous degassing of lunar magmas resulted in significant Cl isotope fractionation. Many workers have since focused on in situ measurements of late-crystallizing apatite due to the relative ease of this analysis by ion microprobe methods (i.e., secondary-ion mass spectrometry [SIMS] or NanoSIMS). As a result, bulk-rock Cl isotope measurements are few, with limited comparison to the in situ analyses. The δ³⁷Cl values of lunar apatite have been shown to span a large range from 2.2 up to 81.1‰ (20, 21). Typically, the highest δ³⁷Cl values in lunar apatite are observed within lithologies such as the Mg-suite (22), potassium/rare-earth elements/phosphorus (KREEP)-rich basalts and impact lithologies (10), which have largely been explained in terms of a primeval KREEP (urKREEP) component [residual melt of lunar magma ocean (23)] with a characteristically high δ³⁷Cl value (2, 3). This conclusion was initially supported by trends in bulk-rock rare-earth element (REE) content and average apatite δ³⁷Cl value (2, 3), although this trend is not seen in some lithologies (10, 24).

There are two relevant underlying issues that have not been adequately addressed. The first is that for the limited samples in which both bulk and in situ apatite data are available, the apatite invariably has a higher δ³⁷Cl value. The second is that the range of isotope values in apatite measured within a single thin section can exceed 10‰ (e.g., ref. 2). If the high δ³⁷Cl values of some lunar samples is explained by wide-scale incorporation of a urKREEP component (2, 3), then the variation at the thin-section scale should be minimal. A logical explanation for the range of apatite δ³⁷Cl values at a small scale is that local degassing, at least in part, controls the chlorine isotope composition of apatite. Extensive degassing of halogens is seen on Earth in volcanic systems (25, 26) and is expected in lunar basalts as well. This explanation, however, opens the possibility that the chlorine isotope compositions of late-formed apatite may not be representative of the bulk rock, and instead preserve the composition of the last melt following extensive degassing. To test this hypothesis and to also address the mechanisms of planetary halogen loss, we analyzed a suite of lunar lithologies for their bulk Cl isotope compositions and mare basalts for halogen (F, Cl, Br, I) contents.

Results

Halogen Concentrations.

Cl concentrations were measured on both water-soluble chloride (WSC) and insoluble or structurally bound chloride (SBC) (Table 1). Concentrations of the other halogens (F, Br, I) were measured only on previously leached samples, such that the water-soluble fraction was removed prior to analysis. Previous halogen measurements of lunar samples show that the water-soluble F fraction is minimal (27) and that the Cl/Br ratios are similar between the water-soluble and structural components (28). We therefore assume that our F analyses are a reasonable approximation for the total F in the rock, and that Cl/Br ratios of the insoluble fraction are similar to that of the bulk rock.

Table 1.

Halogen contents and δ³⁷Cl values of lunar materials

Sample	Lithology	Structural Cl		Water-soluble Cl		Structural halogens			Fraction WSC	Bulk δ37Cl
		[Cl], ppm	δ37Cl	[Cl], ppm	δ37Cl	[I], ppb	[Br], ppb	[F], ppm
12002	Low-Ti	2.25	9.40			42.65	138.47	12.81
10017-400	High-Ti-K	4.37	12.53			206.12	55.26	37.63
10017-405	High-Ti-K	3.66	9.23	7.25	12.63	28.94	20.16	38.11	0.66	11.49
10020-255	High-Ti	1.90	8.03	3.34	0.39	16.09	14.67	20.84	0.64	3.16
10044-566	High-Ti	1.63	2.62	3.67	1.60	90.77	56.43	30.78	0.69	1.91
12018-277	Low-Ti	1.93	10.05	3.74	5.03	45.50	29.09	18.80	0.66	6.74
12054-146	Low-Ti	2.05	2.14	9.99	−0.78	31.77	6.45	27.08	0.83	−0.28
12054-150	Low-Ti	5.43	4.29	5.51	0.43	60.74	32.96	31.91	0.50	2.35
12063-343	Low-Ti	1.28	4.19	3.83	0.48	30.27	2.69	21.32	0.75	1.41
14053-305	High-Al	5.80	11.24	5.71	6.11	20.11	14.90	33.49	0.50	8.70
15016-240	Low-Ti	1.12	2.14	1.88	0.84	22.26	12.98	11.12	0.63	1.33
15535-165	Low-Ti	1.52	12.12			27.51	162.36	11.39
15556-258	Low-Ti	1.83	10.57			38.45	153.85	15.37
70215-389	High-Ti	1.49	6.42			15.09	6.53	31.83
70255-56	High-Ti	1.57	7.37			24.13	20.87	24.75
71135-34	High-Ti	3.08	4.01			554.00	34.39	20.02
74275-355	High-Ti	1.55	8.28			20.80	10.16	19.92
76335	Troctolite	1.00	4.07			41.45	9.82	4.66
77215	Cataclastic Norite	2.76	10.26			42.80	15.84	8.77
78235	Shocked Norite	1.40	6.67			267.8	11.43	8.07
60015	Cataclastic	0.52	10.50	1.71	11.37				0.77	11.17
	Anorthosite
60025	FAN	12.10	30.20	3.60	24.45				0.23	28.88
62255	Anorthosite	4.69	25.21	3.44	25.69				0.42	25.41

For mare basalts, WSC contents range from 1.9 to 10.0 ppm while SBC contents are more restricted, ranging from 1.1 to 5.8 ppm (Table 1). For all mare samples, 65 ± 10% of total Cl resides in the water-soluble fraction. If the WSC is derived from vapor exsolution and deposition, then a significant fraction of the degassed component appears to be retained in the lunar samples. This large fraction of WSC is consistent with studies of terrestrial igneous rocks in which up to 90% of Cl is lost by degassing (25, 26, 29). The mare basalts also have more variable Br concentrations relative to Cl (Table 1) and overall have distinctly lower Cl/Br ratios compared to mid-ocean ridge basalt (MORB) (30, 31) (SI Appendix, Fig. S1). The three anorthosite samples average 9 ppm total Cl with variable, but roughly equal proportions in the SBC and WSC.

The F contents of lunar materials range from 4.7 to 38.1 ppm, with highlands lithologies exhibiting a marked depletion relative to mare basalts (Fig. 1). F contents appear to be in-part related to Ti contents (SI Appendix, Fig. S2), with high-Ti basalts being the most F-rich from 20.0 to 38.1 ppm. A best fit to the mare samples is a Cl/F = 0.32 (R² = 0.62). The bulk Cl/F ratios are higher than the average Cl/F ratios of lunar apatite, consistent with the result that the majority of Cl resides in the water-soluble fraction (Table 1). The Cl/F ratios of the insoluble fraction (rinsed samples) overlap the average Cl/F ratio of lunar apatite, further supporting this conclusion. Last, the F and Cl concentration of lunar samples are depleted by an order of magnitude relative to Earth.

Fig. 1.

F vs. Cl contents (ppm) of terrestrial mid-ocean ridge basalts (MORB) (whole rock, black pentagons; glasses, gray pentagons) and lunar materials. The blue diamonds are the total bulk Cl contents of mare basalts (WSC + SBC). The red diamonds are the structurally bound Cl contents. Highlands F and Cl abundances are from the SBC fraction only. Data of MORB collected from the PETDB database. Apatite F and Cl data gathered from refs. 2, 32, and 33 and are seen in the gold line. Blue line represents the best fit for mare basalts. The faded gray horizontal bar is the accepted range of Cl abundances in MORB (30). Higher Cl values are likely contamination. F and Cl errors are smaller than symbol size.

Cl Isotope Values.

Chlorine isotope values are reported in per mil notation (‰) relative to standard mean ocean chloride (SMOC) (34). The average δ³⁷Cl value for total Cl in mare basalts (sum of WSC + SBC) is 4.1 ± 4.0‰ (n = 9). The structural component averages 7.3 ± 3.5‰ (n = 17), while the water-soluble fraction averages 1.8 ± 2.5‰ (n = 8) (excluding an anomalous WSC value at 12.6‰, 10017-405) (Fig. 2). There is no statistically significant difference between the δ³⁷Cl values of the structurally bound component for high-Ti basalts (7.3 ± 3.1‰, n = 8) and low-Ti basalts (6.9 ± 4.1‰, n = 8) [two tailed t test: t₍₁₄₎ = 0.249, P value = 0.81]. Fig. 2 clearly indicates that the water-soluble fraction for most of the samples is significantly lower than that of the structurally bound Cl. Excluding WSC in sample 10017-405, the isotopic compositions of the WSC and SBC are distinctly different in mare basalts [two-tailed t test: t₍₂₄₎ = 4.007, P value = 0.0006]. The three ferroan anorthosite (FAN) samples have much higher δ³⁷Cl values than the mare basalts of 10.5‰, 30.2‰, and 25.2‰ in the SBC, and 11.4‰, 24.5‰, and 25.7‰ in the WSC in samples 60015, 60025, and 62255, respectively. The three non-FAN highlands samples (two norites and a troctolite: 76335, 78235, and 77215) have SBC δ³⁷Cl values of 4.07‰, 6.67‰, and 10.3‰, respectively, similar to the mare basalts.

Fig. 2.

Chlorine isotope values of planetary and terrestrial materials. MORB data from ref. 19. Bulk lunar data from refs. 14, 15. SIMS data: low-Ti and high-Ti basalts from refs. 3 and 32, high-Al from refs. 2 and 10, highlands from refs. 2 and 35, and breccias from ref. 36. Chondrite data from ref. 19. Martian data from refs. 18 and 37. Iron meteorite data from ref. 17. In-situ lunar apatite δ³⁷Cl values are red crosses. Bulk SBC and WSC δ³⁷Cl values are plotted as black and blue pentagons, respectively. Faded purple bar represents the estimated δ³⁷Cl value of the nebula (17). Faded green bar represents the estimated δ³⁷Cl value of the bulk Earth (19).

The δ³⁷Cl values of published SIMS analyses of apatite grains from high-Ti and low-Ti mare basalts averages 11.56 ± 4.0‰ (2, 3, 32), a value that is higher and significantly statistically different from the bulk data of 3.2 ± 3.0‰ excluding the WSC of 10017-405 [two-tailed t test: t₍₇₈₎ = 5.491, P value = 0.0001], and the structurally bound values of 7.3 ± 3.5‰ [two-tailed t test: t₍₈₇₎ = 4.075, P value = 0.0001]. The discrepancy between bulk and in situ δ³⁷Cl values is even more striking for the high-Al and three highlands samples where apatite has δ³⁷Cl values clustering from 20 to 30‰ compared to the SBC average of 8.1‰. Clearly, the δ³⁷Cl values of apatite in these samples are not representative of the bulk rock. Last, no trend is observed between Cl concentration and δ³⁷Cl values [except for FANs (see Fig. 4)], whereas a poor correlation is observed between F contents and bulk total δ³⁷Cl values (SI Appendix, Fig. S3).

Fig. 4.

Plot of 1/[Cl, Zn] (ppm) vs. δ³⁷Cl and δ⁶⁶Zn value for FANs. The y intercept gives the δ³⁷Cl value of the infiltrating gas at 31‰. An initial concentration of 1.5 ppm (before infiltration of Cl vapor) gives a δ³⁷Cl value of pristine FAN of 4‰. Zn data from ref. 7. The red line is the best for the Cl data with an R² = 0.9982. The blue line is the best fit for the Zn data with an R² = 0.8847.

Discussion

Halogen Degassing.

The Cl/F ratios of mare basalts are similar to those of the least altered MORB basalts, but their concentrations are an order of magnitude lower (Fig. 1). As F is thought to be less volatile than Cl, this result suggests that ∼90% of halogens were lost during an extremely energetic process, such as the Giant Impact, assuming that the lunar protolith was similar to Earth. Although we anticipated that Cl/Br and Cl/F ratios would correlate with δ³⁷Cl values given the fact that Br is significantly heavier than Cl, and F is less volatile, no trends are observed (SI Appendix, Figs. S4 and S5). There is also no apparent relationship in the Cl abundance and δ³⁷Cl values, and only a weak relationship between F abundance and δ³⁷Cl value (SI Appendix, Fig. S3), which we interpret as a manifestation of the high δ³⁷Cl values from F-rich apatite grains. Last, Br contents are more variable than Cl, which may be related to differing degrees of retention as the nonsoluble “structurally bound” component. Unfortunately, we were unable to measure water-soluble bromide in any sample as they had already been processed prior to this work, leaving our Br/Cl relationship ambiguous at present.

Chlorine Isotope Composition of Mare Basalts.

The majority of Cl in mare basalts is hosted in the water-soluble fraction (Table 1). The water-soluble and structurally bound fractions average 5.0 and 2.5 ppm Cl, respectively. The δ³⁷Cl values of the WSC average 1.8 ± 2.5‰ (excluding 10017-405) and are far lower than those of the SBC (7.3 ± 3.5‰) (Fig. 3). Similar differences have been seen in δ⁶⁶Zn values of water-soluble and structural Zn in Rusty Rock (66095), although the effect is far less pronounced (5), and importantly, the δ⁶⁶Zn values are anomalously low compared to Earth, while the δ³⁷Cl values are high. If the WSC and SBC were in isotopic equilibrium, we would expect their δ³⁷Cl values to be almost identical, as the fractionation between all chloride-bearing phases is very small at high temperatures (38). Instead, the relatively low isotopically light WSC component is most readily explained by preferential local degassing of ³⁵Cl and subsequent deposition of the light vapor as a water-soluble salt during the waning stages of crystallization. The degassing may occur along fractures or into fluid inclusions and/or vesicles (39). Documentation of visible water-soluble Cl-bearing phases, however, is scarce due to their deliquescent nature and resulting poor preservation. Cl-bearing phases such as akaganéite, lawrencite, and Cl-bearing sulfates have been described in Rusty Rock 66095 (15, 40), and over 60% of Apollo 16 rocks contain a chlorine-bearing “rust” such as akagenéite, which is thought to the be hydrated by-product of FeCl₂ (41). Pyroclastic glasses have also been observed to be coated in NaCl (42). In general, the trace amounts of Cl-bearing water-soluble phases have simply not been described in lunar materials due to either a lack of attention or trace deliquescent occurrences being erased by sample preparation/storage procedures.

Fig. 3.

Rayleigh distillation models of chlorine degassing into a vacuum assuming α values following Graham’s Law and Eq. 1. The red lines show the isotopic composition of the residual fraction (i.e., the degassing reservoir). The blue lines show the instantaneous isotopic composition of vapor derived from the residual melt. The black lines show the isotopic composition of the cumulative vapor component. The isotopic compositions of lunar materials are readily achieved by the degassing of FeCl₂ or HCl.

Arguably the most striking observation in this work is the fact that in no samples measured to date do the bulk δ³⁷Cl values match those of apatite. While the higher Cl/F ratio of the bulk rock compared to lunar apatite supports the observation that significant Cl resides outside of the apatite in the mare basalts (Fig. 1), the poor relationship with δ³⁷Cl values and F contents suggests that other F-bearing phases are also present. As K-rich mesostasis glass is commonly cited in the literature to occur alongside apatite (43), we measured the Cl and F contents of trapped immiscible melts in sample 10050 (low-K, high-Ti basalt) by electron microprobe. Both the K-rich and Fe-rich melts contained significant amounts of F and Cl (1,480, 160, and 5,450, 230 ppm F and Cl, respectively) (SI Appendix, Fig. S6). Although no Cl isotope analyses of these Cl-rich glasses have been made, these phases clearly represent another source of Cl and F in lunar materials and partly explain the poor relationship with F contents and δ³⁷Cl values.

In terms of the individual components, lunar apatite has the highest δ³⁷Cl values in mare basalts, averaging 12.6 ± 4.8‰ (high-Ti) and 9.9 ± 4.1‰ (low-Ti). Apatite crystallizes after most of the Cl has been lost to the vapor phase or sequestered in other minerals or glass. The δ³⁷Cl values of the apatite are therefore not representative of the bulk-rock or lunar lithologies in general, but rather record the δ³⁷Cl values of the last stages of crystallization. The loss of a light Cl-bearing vapor raises the δ³⁷Cl value of the residual Cl, which is ultimately incorporated in late-forming apatite and glass. The large spread in δ³⁷Cl values and Cl/F in apatite measured within individual thin sections or even single grains is consistent with apatite forming during the waning stages of crystallization and the final phases of local vaporization.

The overall calculated bulk δ³⁷Cl value of the mare basalts averages 4.1‰, higher than the Earth value, but far lower than the apatite values (Fig. 2). This higher δ³⁷Cl value of the bulk Moon relative to Earth, alongside the poor relationship with F/Cl and δ³⁷Cl values despite the fact the F is less volatile, suggests that lunar halogen depletion and ³⁷Cl enrichment likely resulted from a planet-wide event, presumably the Giant Impact.

Chlorine Isotope Composition of the Highlands Lithologies.

The SBC δ³⁷Cl values from troctolite 76335 (4.1‰) and two norites 77215 and 78235 (10.3‰ and 6.7‰) measured are similar to the mare basalts. In contrast, the δ³⁷Cl values of apatite grains from the highlands lithologies (30.6 ± 2.3‰) and the Al-rich basalts (24.9 ± 6.0‰) are far higher than those of the mare basalts (Fig. 2). The low bulk SBC δ³⁷Cl values of highlands troctolites and norites compared to their apatite requires, from mass balance constraints, that the majority of the SBC has a low δ³⁷Cl value and, like the mare basalts, mainly resides somewhere besides the apatite. This conclusion is supported by the fact that the highlands lithologies have far less F, but the same amount of Cl, relative to basalts (SI Appendix, Fig. S7). Assuming that the majority of F resides in apatite, then the amount of apatite in these rocks is low compared to the mare basalts. Only the FANs have anomalously high bulk chlorine isotope values, similar to those found in highlands apatite. The δ³⁷Cl values of the SBC and WSC components from the three FANs are similar to each other and exceptionally high (10.5 to 30.2‰) compared to all other lithologies.

It has been suggested that the high δ³⁷Cl values of the highland rocks apatite is due to the incorporation of a high δ³⁷Cl-bearing urKREEP component (2, 3). urKREEP is the hypothetical reservoir representing the final residue of LMO crystallization (23) and is therefore exceptionally enriched in incompatible trace elements. It has been proposed that urKREEP magma(s) underwent significant degassing and preferentially lost ³⁵Cl to space (3). These high ³⁷Cl/³⁵Cl melts were then subsequently incorporated into the highlands lithologies raising their δ³⁷Cl values. The correlation between the REE abundances and the δ³⁷Cl values of apatite have been used in support of this argument (2, 3). The causal relationship between urKREEP incorporation and high δ³⁷Cl values, however, has not been established. One would expect that if the addition of the isotopically heavy and Cl-rich urKREEP occurred, then the Cl content of the rocks would also increase. In other words, in order to raise the δ³⁷Cl values of the urKREEP-contaminated lithologies, it is necessary to introduce additional heavy Cl. However, the Cl contents of the highlands lithologies are lower, not higher than the mare basalts. Additionally, the FANs are the only lithology to record the high bulk δ³⁷Cl values characteristic of urKREEP, yet they do not contain apatite, nor do they have clear evidence of a urKREEP component.

An alternative explanation for the high δ³⁷Cl values of the highland lithologies is their markedly different cooling history compared to mare basalts. The norites, troctolites and anorthosites all cooled very slowly as part of a planet-wide cooling event on the order of 10 °C/Ma (44). In contrast, the mare basalts cooled on the order of ∼1 to 20 °C/h (45). The slowly cooled highlands lithologies can therefore be expected to have lost far more Cl during degassing, compared to rapidly cooled mare basalts. Conversely, the high δ³⁷Cl values of high-Al basalt apatite may also be in-part explained by vapor-phase metasomatism (10). As described below, the effect on the Cl isotope composition of the different rock types can be evaluated using the concept of Rayleigh distillation.

Partial Cl-Retention and Vapor Isotopic Compositions.

The isotopic composition of a degassing magma, and the derived vapor can be modeled by use of the Rayleigh distillation equation given by the following:

$${\delta }_{\mathrm{Final}}={\delta }_{\mathrm{Initial}}+\left(\mathrm{1,000}+{\delta }_{\mathrm{Initial}}\right)\times \left(F^{\left(\alpha -1\right)}-1\right),$$ [1]

where δ_Initial and δ_Final are the initial and instantaneous isotopic composition of the melt, F is the fraction of the element remaining and α is the fractionation factor between vapor and melt, given by the following:

$${\alpha }_{\mathrm{Vapor-melt}}=\frac{\mathrm{1,000}+{\delta }_{\mathrm{Vapor}}}{\mathrm{1,000}+{\delta }_{\mathrm{Melt}}}.$$ [2]

The fractionation α is determined for loss from a melt into a vacuum, governed by Grahams law such that

$${\alpha }_{\mathrm{Vapor-melt}}=\sqrt{\frac{M_2}{M_1}},$$ [3]

where M₂ is the molar mass of the volatilizing light isotopologue with ³⁵Cl and M₁ is the molar mass of the heavy isotopologue with ³⁷Cl. For FeCl₂, M₂ = 55.8+35.0+35.0 and M₁ = 55.8+35.0+37.0, such that α= 0.992. This is the maximum fractionation for an ideal case where the devolatilizing species passes into a vacuum. The actual fractionation factor will be smaller (α closer to 1). These equations govern how the isotopic composition of instantaneous vapor, the integrated vapor and the residual melt change as degassing proceeds (Fig. 3). Last, the isotopic composition of the cumulative vapor is given by Eq. 4:

$${\delta }_{\mathrm{Cumulative}\hspace{0.17em}\mathrm{Vapor}}=\frac{{\delta }_{\mathrm{Initial}}-F\left({\delta }_{\mathrm{Final}}\right)}{1-F}.$$ [4]

When magmatic degassing occurs, the initial vapor has a δ³⁷Cl value that is 1000lnα lower than the magma (−27.8‰ for HCl and −8.0‰ for FeCl₂ using α values following Grahams law). The instantaneous δ³⁷Cl values of the vapor are shown by the blue curves in Fig. 3. The change in the isotopic composition of the residual melt are shown by the red curves. The black curves represent the cumulative δ³⁷Cl values of the vapor phase as it accumulates, i.e., a case where all vapor is retained in the system. Loss of 50% of the FeCl₂ to the vapor phase will raise the residual Cl by ∼5.6‰ (point A in Fig. 3). The instantaneous vapor coming off the melt at this point will have a δ³⁷Cl value ∼8.0‰ lighter than the residual (point B). If all Cl vapor that was evolved during degassing was retained in the system, then the total vapor component (equal to the water-soluble chloride) would have a δ³⁷Cl value of −2.4‰ (point C). In this example, during the earliest stages of degassing, the cumulative vapor component represents a negligible total fraction of Cl. As degassing proceeds, however, the cumulative vapor incorporates an increasing fraction of the total Cl and ultimately approaches the isotopic composition of the initial undegassed melt.

If, instead, the system is open to vapor loss during the early, high-temperature phase of cooling, then different isotope values will be achieved compared to the closed-system case above. This is what we expect for the slowly cooled highlands rocks. During the initial stages of degassing, temperatures would be too high for the vapor phase to condense (5), and the light vapor component would not be retained. Given the extremely slow cooling rates, the vapor would migrate out of the system and be lost from the system with the δ³⁷Cl value of the rock progressively increasing. Loss of a significant fraction of the light vapor, as expected in the slowly cooled highlands rocks, results in high ³⁷Cl/³⁵Cl in the residue, and thus explains the elevated bulk values and the very high δ³⁷Cl values of their late-formed apatite grains.

Anomalous Cl and Zn Isotope Compositions of FANs.

The FANs are thought to represent the primitive lunar crust that formed by the accumulation of buoyant plagioclase over the lunar magma ocean (LMO) (46, 47). Standard LMO crystallization models state that the primitive lunar crust would have formed a thermally insulating “lid” on the crystallizing magma ocean, potentially limiting volatile loss to space (2, 4, 48). However, volatiles would have continually percolated through the crust postcrystallization (41), possibly during a global mode of volcanic “heat-pipes” transporting melt to the surface (49), or be lost by crust-breaching impacts (2). The bulk δ³⁷Cl values of FANs are the highest measured in planetary materials to date and positively correlate with Cl contents, opposite from what would be expected from a simple open system degassing trend (Fig. 4). Instead, the data suggest incorporation and deposition of a heavy Cl-bearing vapor into the FANs. This is consistent with Cl being incorporated in the FANs only after most had passed through the lithology as a gas phase. Zn isotope compositions on the other hand, have a negative correlation with concentration (7) suggesting that the FANs incorporated an early degassed light Zn-bearing phase.

Exotic sources of Cl and Zn to the FANs such as chondrites or regolith can be discounted on the basis that both Cl and Zn isotope values, as well as elemental concentrations, are a poor match. A heavy urKREEP source could also be invoked to explain the FANs high δ³⁷Cl values. Some KREEPy lithologies (i.e., Mg-suite) are thought to represent plutonic emplacements into the lunar crust and therefore the proximity of the FANs and urKREEP is certainly feasible. McCubbin et al. (2015) estimates KREEP melts likely contained 1.4 wt% Cl when apatite began to crystallize which record the characteristically high Cl isotope compositions. Mixing with 0.1% KREEP with a δ³⁷Cl value of +30‰ could explain the highest [Cl] and δ³⁷Cl reported here, however, this mixing component is not seen in elevated REE patterns or any other geochemical system to our knowledge. Additionally, the FANs were likely emplaced prior to urKREEP formation (50), and their Zn isotope compositions are also distinct from KREEPy samples, with comparable [Zn] but with significantly lower δ⁶⁶Zn values (7). In total, assimilation of urKREEP is inconsistent with both Cl and Zn isotope data (51). We instead propose that the Cl and Zn isotope compositions resulted from vapor deposition from the degassing of underlying magma(s).

While it has generally been assumed that Zn and Cl both degas as chlorides (and ZnS), the isotopic compositions of the FANs require that these two elements are decoupled (see also ref. 5). Assuming that Zn and Cl had similar volatilities, they would vaporize and condense over roughly the same temperature interval. If this were the case, low δ⁶⁶Zn values would occur alongside low δ³⁷Cl values, which is clearly not seen in the data for the FANs or other samples such as Rusty Rock. Furthermore, it is difficult to explain the very low δ⁶⁶Zn values of −13.7‰ found in Rusty Rock, or −11.4‰ in 65315 using a simple Rayleigh fractionation model if ZnCl₂ is assumed to be the degassing species and the initial δ⁶⁶Zn value of the source was close to 0‰. This is because the maximum 1000lnα value for ZnCl₂ is only −7.4‰, far less than the ∼13–14‰ fractionation for Rusty Rock (5). The trends seen in Fig. 4 make coupled degassing equally unlikely for the FANs.

We instead propose that the low δ⁶⁶Zn values are better explained by the degassing of Zn^o from the underlying magma and subsequent deposition in the FANs at relatively high temperatures, as suggested by recent thermodynamic modeling (52). Importantly, the 1000lnα value of Zn^o degassing is −15.4‰, compared to only −7.4‰ for ZnCl₂ (assuming Grahams law applies). In contrast to Zn^o gas, which condenses at around 700 °C, metal chloride species are volatile above 500 °C (52, 53). In the most extreme case where Cl is transported as HCl, Cl would never condense. It would only become incorporated in a rock when it reacts to form a solid chloride phase (e.g., FeCl₂). Zn^o gas derived from the underlying magma ocean would degas through the FANs where the high condensation temperatures of ZnS or Zn^o would lead to only a small amount of Zn migrating to the higher and colder levels of the FANs. As the FANs cool, the migration distance of the Zn decreases from the underlying magma, such that only the early, light Zn would be found in the upper level of the FANs, which is presumably what is sampled in the Apollo collection.

Cl differs from Zn because it has a lower condensation temperature. As long as the anorthositic lid remained hot, Cl would not condense, but rather pass through the FANs. Retention of this vapor percolating through the FANs would only have occurred when temperatures dropped sufficiently for Cl-bearing vapors to condense (e.g., ref. 5). If most of the Cl had already been lost from the underlying magma before it could condense in the FANs, then the δ³⁷Cl value of the vapor that was ultimately incorporated would be isotopically heavy (Fig. 5). In the case of HCl degassing, loss of 90% of the total Cl would raise the vapor value to over 30‰. For FeCl₂, the amount lost would be closer to 99%. Addition of the heavy vapor would have increased both the Cl concentration and the δ³⁷Cl value, and indeed, a negative linear mixing trend of 1/[Cl] vs. δ³⁷Cl is seen in the limited data for the FANs, supporting this hypothesis (Fig. 4). In total, the partial Cl-retention model in the FANs suggests they were only partially effective in limiting volatile-loss to the surface and future modeling of LMO volatile-loss should account for the continued loss of volatiles after FAN crystallization (54).

Fig. 5.

Modified Rayleigh distillation model showing partial chlorine retention. Line legend is the same as in Fig. 3. In this model, we choose to use HCl as an isotopologue capable of reaching the largest range of δ³⁷Cl values. Three different arbitrary points are chosen for schematic purposes (A, B, and C) in which to vapor derived from an underlying melt is retained. If vapor begins to be retained at point A and is completely incorporated, it will have a δ³⁷Cl value of +6‰ with 80% of initial [Cl]. Likewise, for B with a δ³⁷Cl of +20‰ and 50% of initial [Cl], and last C with a δ³⁷Cl of +34‰ and 30% initial [Cl]. The partial retention of chlorine following this model will show increasing δ³⁷Cl with [Cl].

Conclusion

There are several important findings in this work. First, the bulk Cl isotope compositions of lunar mare basalts is 3–4‰ higher than the bulk Earth and the halogen concentration of lunar lithologies is an order of magnitude lower than MORB. We propose that these differences are the result of an early planetary-wide halogen depletion and isotope fractionation associated with the Giant Impact, similar to what is seen for other volatile elements (e.g., refs. 9, 55).

Second, we find a large isotope fractionation between the water-soluble and structurally bound chloride within individual samples. Approximately 65% of Cl in the mare basalts is hosted in a water-soluble with an average δ³⁷Cl value of 1.8 ± 2.5‰ (excluding 10017–405). The remaining insoluble or structurally bound chloride averages δ³⁷Cl values of 7.3 ± 3.5‰.The large difference in the isotopic composition of these two coexisting Cl reservoirs cannot be explained in terms of equilibrium, high-temperature isotope fractionation (38). Instead, the low δ³⁷Cl values of the vapor component are explained by kinetic isotope fractionation during local degassing and subsequent deposition. The δ³⁷Cl values of apatite grains are significantly higher than the bulk δ³⁷Cl values in all lithologies. Rather than being representative of the bulk rock, apatite δ³⁷Cl values record information about the final stages of crystallization. Local degassing during late-stage crystallization explains the wide range of δ³⁷Cl values seen in apatite grains within individual thin sections.

Third, the highlands rocks have extremely elevated δ³⁷Cl values in apatite, but only moderately elevated bulk δ³⁷Cl values (Fig. 2). We explain the high δ³⁷Cl values of the late-formed apatite as being a result of the slow cooling rate for these plutonic lithologies resulting in extensive Cl degassing prior to apatite formation. Compared to the mare basalts, cooling of the magma ocean and overlying solid lithologies was many orders of magnitude slower. The very slow cooling would result in a larger fraction of Cl to be lost to the vapor phase, thereby causing the residual melt to have increasingly high δ³⁷Cl values. This effect is especially notable for the late-formed apatite which records the highest δ³⁷Cl values (Fig. 4).

Finally, the FANs uniquely have extremely high δ³⁷Cl values for both the WSC and SBC (Fig. 2), although they have no urKREEP component, nor apatite grains. This result is at odds with the hypothesis that the high δ³⁷Cl values of highlands rocks is related to incorporation of urKREEP. Instead, we suggest that their extremely high δ³⁷Cl values are explained as the product of condensation of a heavy Cl-bearing vapor. The anorthosite layer above the crystallizing magma ocean would have been extremely hot post crystallization. The Cl-bearing vapor derived from the underlying magma would therefore pass through the anorthositic lid without condensing. Only after extensive degassing would the anorthosite layer cool sufficiently for the Cl-bearing vapor to condense. By the time this occurred, the vapor would have evolved to a very high δ³⁷Cl value (>30‰) which is supported in the positive correlation between Cl content and δ³⁷Cl values. In contrast, Zn has a much higher condensation temperature (over 700 °C for ZnS). Only the earliest, and thereby the isotopically lightest Zn^o vapor would penetrate to the upper layers of the FANs resulting in the incorporation of Zn with exceptionally low δ⁶⁶Zn values. Once the FANs cooled below 700 °C, Zn migration distances would be severely curtailed. The very low δ⁶⁶Zn values of FANs (and Rusty Rock) are thereby best explained by degassing of Zn^o where the 1000lnα value is significantly higher than for higher mass Zn-bearing compounds.

Materials and Methods

Cl isotope compositions of a suite of lunar materials were measured following the methodology of Sharp et al. (14) at the University of New Mexico. Water-soluble chloride (WSC) was collected by leaching powders with de-ionized water overnight. Structurally bound chloride (SBC) was extracted by pyrohydrolysis on the residual powders following WSC leaching. Leached powders were rinsed an additional 3 times before drying to remove any residual WSC component, dried and then loaded into quartz tubes. Samples were then subsequently pyrolyzed by melting with an oxy-propane torch in a stream of water vapor within a distillation apparatus (2). An aliquot of parent solutions were measured at National Aeronautics and Space Administration (NASA) Johnson Space Center for Cl, Br and I contents by ICPMS following the methodology of ref. 56. Average 1σ deviation on Cl, Br and I contents were 2.2, 11.4, and 1.14%, respectively. Additional information regarding halogen measurements can be found in the SI Appendix. Fluorine contents were measured at UNM using Ion Chromatography.

A split of the sample solutions were reacted with 5 mL of 50% HNO₃ overnight in order to remove sulfur, followed by an addition of 1 mL of 0.4 M AgNO₃ overnight in a light-free environment in order to precipitate AgCl. AgCl was then filtered, dried, and loaded into 8 mm Pyrex tubes. Sample tubes were then evacuated, 10 μL of CH₃I was added and subsequently flame sealed. Sealed tubes were then reacted at 80 °C for 48 h to produce CH₃Cl as an analyte. Introduction of CH₃Cl into the mass spectrometer took place by cracking sealed sample tubes in a stream of He and collecting the CH₃Cl in a liquid N₂ trap. This trap was subsequently warmed, and the sample gas was passed through a GC column held at 80 °C to separate excess CH₃I from the CH₃Cl analyte. A small “sniffer” capillary was used to detect when CH₃Cl had completely left the column and was trapped in a second liquid N₂ trap. Flow direction was then reversed prior to the incorporation of CH₃I, which resulted in a pure CH₃Cl sample. Trapped CH₃Cl was then released by warming the LN₂ trap and introduced via an open split into the mass spectrometer. Chlorine isotope compositions were measured in continuous flow on a Delta^PLUSXL. Measurements were standardized relative to SMOC with a long-term reproducibility of this method being less than ±0.25‰ in our laboratory.

Data Availability.

All study data are included in the article and SI Appendix.