Magnesium stable isotopes support the lunar magma ocean cumulate remelting model for mare basalts

Significance

Soon after the return of the first lunar samples by the Apollo missions, it became clear that the Moon is a highly differentiated object, with a plagioclase-rich crust formed by mineral flotation in a very early magma ocean. The younger lunar mare basalts were interpreted to result from remelting of these lunar magma ocean (LMO) cumulate layers. Here we report significant Mg isotope variations produced by the Moon’s early magmatic differentiation. These results support models in which lunar basalts are formed by partial melting of distinct cumulate sources produced during the LMO crystallization and imply that the bulk lunar Mg isotope composition is similar to that of the inner solar system.

Abstract

We report high-precision Mg isotopic analyses of different types of lunar samples including two pristine Mg-suite rocks (72415 and 76535), basalts, anorthosites, breccias, mineral separates, and lunar meteorites. The Mg isotopic composition of the dunite 72415 (δ²⁵Mg = −0.140 ± 0.010‰, δ²⁶Mg = −0.291 ± 0.018‰), the most Mg-rich and possibly the oldest lunar sample, may provide the best estimate of the Mg isotopic composition of the bulk silicate Moon (BSM). This δ²⁶Mg value of the Moon is similar to those of the Earth and chondrites and reflects both the relative homogeneity of Mg isotopes in the solar system and the lack of Mg isotope fractionation by the Moon-forming giant impact. In contrast to the behavior of Mg isotopes in terrestrial basalts and mantle rocks, Mg isotopic data on lunar samples show isotopic variations among the basalts and pristine anorthositic rocks reflecting isotopic fractionation during the early lunar magma ocean (LMO) differentiation. Calculated evolutions of δ²⁶Mg values during the LMO differentiation are consistent with the observed δ²⁶Mg variations in lunar samples, implying that Mg isotope variations in lunar basalts are consistent with their origin by remelting of distinct LMO cumulates.

The early work on lunar anorthosites separated from the Apollo 11 lunar regolith led to the development of the lunar magma ocean (LMO) hypothesis in the early 1970s (1, 2). This hypothesis has later been elaborated by lunar meteorites and remote sensing (3), chronology of pristine lunar rocks (4, 5), and complementary Eu anomalies in anorthosites and mare basalts (6). Fractional crystallization of the LMO resulted in the formation of a mafic mantle and a feldspathic crust, with late-stage ilmenite-rich cumulates and the materials enriched in potassium (K), rare earth elements (REE), and phosphorus (P) (KREEP) crystallizing beneath the lunar anorthositic crust (7, 8).

Many experimental and theoretical models along with elemental and isotopic data of lunar samples, including both stable and radiogenic isotopes, were used to determine the Moon’s origin and magmatic evolution as well as the origin of lunar basalts, which are presumably partial melts of the cumulates produced during the LMO crystallization (9–19). Over the last 2 decades, the nontraditional stable isotopes have also provided new insights into the accretion and magmatic evolution of planetary bodies including the Moon (20–35). Most of the data (except for volatile elements such as K and Zn) show that among bodies in the solar system the bulk silicate Earth (BSE) and bulk silicate Moon (BSM) are uniquely similar, despite some variations among the lunar rocks. Magnesium, a major element with three stable isotopes, is potentially an important tool to study the Moon’s early magmatic differentiation because its isotopic fractionation is only influenced by mineral crystallization and is not affected by core formation processes (23, 31, 35–39). Most studies indicate that Mg isotopic compositions in the inner solar system are homogeneous and vary perhaps as a result of igneous differentiation processes (22, 23, 36, 38–42). This homogeneity of Mg isotope compositions among planetary bodies has been questioned by a recent high-precision Mg isotope study (29) based on the idea that some variations should be expected due to vapor loss from growing planetesimals (29) or sorting and physical separation of CAIs and chondrules in the protoplanetary disk (37, 43). However, the extent and mechanisms of Mg isotopic fractionation during the magmatic evolution of planetary bodies remain unexplored. More recent studies of Mg isotopes in the Moon, asteroid Vesta, ureilites, and Martian meteorites report significant Mg isotope variations perhaps related to their igneous differentiation history (22, 23, 31, 35, 44). Most Mg isotopic studies of lunar samples show a significant dichotomy between low- and high-Ti basalts (23, 37, 39). The isotope dichotomy in lunar basalts is also seen for other elements and is suggested to be the result of heterogeneities produced by the lunar magmatic differentiation (18, 20, 27, 45); however, this assumption for Mg has not yet been well studied. Moreover, the isotopic composition of the lunar basalts is commonly used to estimate the isotopic composition of the BSM, but it remains unclear whether the basalts are the most representative of the BSM (27, 32). To evaluate such possibilities and constrain the behavior of Mg isotopes during the lunar magmatic differentiation, we analyzed a suite of representative samples including pristine anorthositic and Mg-suite rocks, lunar basalts, and mineral separates (Materials and Methods).

Discussion

Previous Mg isotopic studies have found no measurable Mg isotope fractionations in terrestrial whole rocks formed by partial melting and magmatic differentiation (41, 42, 46). However, significant intermineral Mg isotope fractionations have been found among terrestrial minerals due to both kinetic and equilibrium effects (43, 47–51). In contrast to terrestrial rocks, the lunar samples studied here (Fig. 1 and SI Appendix, Tables S1 and S2) show Mg isotope variations similar to what has been reported for other differentiated bodies (22, 23, 31, 35). This Mg isotopic variation could be a tracer of planetary differentiation. In the standard LMO model, fractional crystallization resulted in flotation of low-density minerals like feldspar forming the anorthositic crust, whereas sinking of the denser ferromagnesian minerals produced olivine and pyroxene layers (7, 8, 52). Here we first discuss the Mg isotope compositions of the Mg-suite (the most Mg-rich samples) and FAN samples that are thought to be the direct products of LMO crystallization (53).

Fig. 1.

The δ²⁶Mg values of terrestrial and lunar samples analyzed in this study (SI Appendix, Tables S1 and S2). The solid and dotted lines are Mg isotopic composition of the BSM and two SEs, respectively (δ²⁶Mg = −0.291 ± 0.018‰). The literature data are from refs. 23 (Moon), 35 (Mars), and 41 (Earth and chondrites). The shaded blue and yellow boxes are Mg isotopic compositions measured for low-Ti and high-Ti basalts, respectively (this study).

The highland Mg-suite samples are plutonic rocks with distinctive characteristics of a high Mg # (60–95, where Mg # = molar [MgO/MgO + FeO] × 100); an enrichment in KREEP material; and a depletion in Cr, Ni, and Co (5, 14). They are likely formed by partial melting of a hybrid parent magma produced in the early lunar magmatic evolution (5, 14). The model that fits best with the petrology, geochemistry, and chronology of Mg-suite rocks (14) predicts that their parental magmas were formed from less dense early cumulates at high temperature (∼1,400–1,800 °C). Then, these rising hot and low-density magmas were mixed with KREEP and plagioclase at the base of the crust. We studied two Mg-suite rocks, 72415 and 76535. The lunar dunite 72415, one of the oldest lunar samples of the Mg-suite, contains chromite symplectites indicative of crystallization at 40–50 km rather than at a shallow depth of ∼1 km (5, 54, 55). The deep cumulate origin of this sample has been questioned by Ryder (55) based on its slightly zoned olivines (Fo # 86–89) with relatively high CaO (∼0.1%) compared with plutonic olivines. However, based on the existence of slightly zoned olivines (Fo # 84–88) in pallasites (56) and high CaO content of San Carlos olivine, Wang et al. (27) argued that none of these observations provides strong evidence against deep origin of the dunite 72415. The Fe isotopic composition of this dunite is distinctly lighter (∼0.35–0.45‰ in δ⁵⁶Fe) than that of the BSE and BSM (27, 32). It is suggested that the Fe isotope fractionation during the early crystallization produced isotopically light olivines of the deeper mantle, which balance the heavy Fe isotope composition of lunar basalts (27). This, in turn, leads to similar Fe isotopic compositions of the BSE and BSM. Also, Sossi and Moynier (32) analyzed more Mg-suite rocks and found that the compositions of the BSE and BSM are broadly indistinguishable, but the Fe isotopic composition of the dunite 72415 was still lighter (∼0.40‰ in δ⁵⁶Fe) than that of the BSM. They interpreted the Fe isotopic composition of this sample as an anomaly resulting from Fe–Mg diffusion in olivine rather than equilibrium fractionation during olivine crystallization. The lack of Fe isotope fractionation during olivine crystallization from a basaltic melt experiment (57) may also rule out the equilibrium Fe isotope fractionation via olivine crystallization in the early LMO. However, the δ²⁶Mg value of the dunite 72415 measured in this study (−0.291 ± 0.018‰) agrees well with the previous estimate of Mg isotopic composition of the Moon (−0.26 ± 0.16‰) (23). Therefore, the suggested diffusive fractionation origin for the light Fe isotopic composition of the dunite 72415, which is not recorded in the Mg isotopic composition of this rock, still remains enigmatic. The δ²⁶Mg values of this dunite and the pristine troctolite 76535 (−0.336 ± 0.031‰) overlap with the estimated values of the Moon (−0.26 ± 0.16‰) (23), the Earth (−0.25 ± 0.07‰), chondrites (−0.28 ± 0.06‰) (41), and Mars (−0.271 ± 0.040‰) (35). These two Mg-suite samples of deep origin (72415 and 76535) have high modal abundance of olivine; hence, the similarity between their Mg isotopic compositions and that of the BSM (23) hints at no Mg isotopic fractionation during crystallization of the most magnesian olivine from the LMO (Fig. 1). The lack of Mg isotope fractionation during olivine crystallization is also confirmed by the δ²⁶Mg value of the olivine separate (SI Appendix, Table S2) and will be evaluated in our isotopic model in the next section. These results agree with the Fe isotopic study of Mg-suite samples (32) suggesting that the Mg-suite rocks may be the most representative of the BSM for both Mg and Fe isotopic compositions. Our results contradict a recent study of Mg isotope compositions of planetary bodies (29) suggesting that all differentiated bodies are isotopically heavier (∼0.02 ± 0.010‰, 2σ_m) than chondrites. The latter was explained by equilibrium isotope fractionation between silicate liquid and vapor lost during the Moon’s accretion (29). Nevertheless, significant Mg isotope variations among the rocks from different planetary bodies such as Mars, the asteroid Vesta, and the Moon do exist implying that igneous differentiation fractionates stable isotopes. Therefore, a model that can link measured isotopic compositions of individual rocks to the bulk compositions of their source planets/planetesimals is a prerequisite for evaluating potential differences in stable Mg isotope compositions among planetary bodies in the solar system.

Ferroan-anorthosites (FAN) are assumed to be flotation cumulates of a global LMO. The three FAN samples studied here are pristine highland rocks with very low siderophile and incompatible element abundances (58). Two anorthosites, 60015 and 60025, are enriched in heavy Mg isotopes (Δ²⁶Mg_sample-BSM ∼ 0.3‰). However, anorthosite 62236 has a δ²⁶Mg value of −0.249‰ similar to that of the BSM (Fig. 1). Considering the fractionation factor for plagioclase-melt (Δ²⁶Mg_{plagioclase-melt} = 0.869‰) (SI Appendix, Table S7), the δ²⁶Mg values of liquids in equilibrium with samples 60015 and 60025 could be −0.849 and −0.787‰, respectively. These values can reflect a late stage of LMO evolution as is shown in Isotopic Fractionation Model. However, the LMO origin of younger FAN samples is more controversial (59–61). The third FAN sample 62236 is a noritic anorthosite with 83% plagioclase (Pl), 7% orthopyroxene (Opx), 5% clinopyroxene (Cpx), and 5% olivine (Ol) (58). The MgO contents (wt %) of the pyroxenes and plagioclase of this sample (Opx, 22.1; Cpx, 14–21; and Pl, 0.00; ref. 62) indicate that Opx, Cpx, and Ol are the main hosts of Mg, resulting in a Mg isotopic composition similar to that of the BSM. The young age of this sample (63) also implies its formation by processes other than a simple flotation from the LMO. The lack of significant correlation between δ²⁶Mg and MgO in the low-Ti basalts, dunite, and troctolite implies no major Mg isotope fractionation by crystallization of olivine cumulates, which account for most of the Mg in the BSM. However, the significant variation in δ²⁶Mg of anorthosites and high-Ti lunar basalts with lower MgO content suggests possible Mg isotope fractionation during crystallization of clinopyroxene and plagioclase from the LMO (SI Appendix, Fig. S5).

Three low-Ti lunar basalts have average δ²⁶Mg of −0.285 ± 0.109‰ similar to those of the BSM (this study and ref. 23), Mars (35), Earth, and chondrites (41). In contrast, high-Ti basalts have substantially lower δ²⁶Mg values (−0.694 to −0.312‰) with the average δ²⁶Mg of −0.462 ± 0.084‰ (Fig. 1 and SI Appendix, Table S1), which is similar to earlier reports (23, 37, 39). Stable isotope studies of other elements such as O, Ti, Li, and Fe have also shown a dichotomy between low- and high-Ti basalts with low-Ti basalts being similar to the BSE and high-Ti basalts departing toward heavier/lighter isotope compositions (17, 18, 20, 34, 45, 64, 65). Sedaghatpour et al. (23) predicted the isotopically light ilmenite produced at the late stage of LMO solidification to be the main source of the isotopically light Mg observed in high-Ti basalts. However, our analyses of lunar ilmenites show insignificant Mg isotope fractionation by ilmenite crystallization at the late stage of LMO solidification (SI Appendix).

The δ²⁶Mg values of two splits of sample 15555, −0.723 ± 0.037‰ (15555, 19) and −0.778 ± 0.017‰ (15555, 999), are significantly lighter than other low-Ti basalts. However, another split of this sample has an unusually heavy δ²⁶Mg value of −0.02 ± 0.03‰ (23), which was explained by the heterogeneous mineral distribution in this sample. Oxygen isotopic and chemical variations are also observed in different chips of this sample, which consists of olivine, pyroxene, and plagioclase (17, 66). Because chemical variations among Apollo 15 olivine-normative mare basalts are mostly controlled by olivine, it was suggested that representative samples of Apollo 15 olivine-normative rocks should be >1 g and ideally >5 g (66). Moreover, a recent Mg isotopic study of a zoned olivine grain from sample 15555 has shown a significant effect of the Mg–Fe intermineral diffusion (67). A similar kinetic Mg–Fe isotope variation has also been observed in olivine megacrysts from a Martian meteorite (68). During the Mg–Fe diffusion, the mineral becomes isotopically lighter than coexisting mineral or melt if Mg diffuses into it and becomes isotopically heavy if Mg diffuses out, due to the faster diffusion of light isotopes. Therefore, different mineral abundances in small subsamples could affect the isotopic compositions of different splits of the basalt 15555. We have dissolved ∼0.053 g of 15555, 19 and ∼1.004 g of 15555, 999, much larger amounts than the ∼0.010 g subsample dissolved in ref. 23. The relatively consistent δ²⁶Mg values in our two larger subsamples suggest that the δ²⁶Mg value of −0.778‰ measured in the largest subsample may be representative of the bulk 15555. However, this isotopic composition is still anomalous among the low-Ti basalts.

Among the breccias analyzed in this study, only sample 14321, 1803 has a significantly lighter Mg isotopic composition than the estimated BSM (Δ²⁶Mg_14321-BSM = −0.869‰) (Fig. 1). Chemical and isotopic compositions of this sample indicate that it is derived from KREEP and high-Al basalt-rich impactites formed in pre-Imbrian craters (69, 70). The light Mg isotopic composition of the sample 14321 is not likely a result of evaporation/deposition during the impact events, because Mg is a moderately refractory element (71), and its isotope composition is unlikely to be fractionated by the impact events. In addition, none of the other breccias and impact melts (this study and ref. 23) shows a significant Mg isotopic fractionation relative to the BSM. Based on the isotope fractionation model in the following section, KREEP components are likely to have very low δ²⁶Mg. Thus, the KREEP-rich nature of this sample may explain its extreme Mg isotope composition. However, analysis of the KREEP basalt 15386 yielded δ²⁶Mg of −0.349 ± 0.038‰, which is not as low as could be expected if it was only related to KREEP components.

Three of the lunar meteorites analyzed here have δ²⁶Mg values that are ∼0.10‰ different from that of the BSM (δ²⁶Mg_NWA7007 = −0.193 ± 0.016‰, δ²⁶Mg_Dhofar1625 = −0.382 ± 0.032‰, and δ²⁶Mg_NWA6570 = −0.369 ± 0.027‰) (Fig. 1). Because our meteorite samples were small (2–8 mg) and probably not representative of the bulk compositions, it is likely that measured isotopic variations are due to different proportions of major minerals within these samples.

To constrain whether the measured Mg isotope compositions were formed during the LMO crystallization, we have analyzed several lunar mineral separates (SI Appendix, Table S2) and have modeled the Mg isotope fractionation during the LMO crystallization using the estimated mineral-melt fractionation factors based on these lunar mineral measurements and terrestrial mineral data in the literature.

Isotope Fractionation Model

We model Mg isotope fractionation during the LMO differentiation to test if the isotopic dichotomy seen in low- and high-Ti basalts is consistent with the cumulate remelting model (9, 12, 18). In this study, we used the Snyder et al. (11) model for the LMO differentiation. The combination of both equilibrium and fractional crystallization (suggested in ref. 11) was not used in the Fe isotope model of ref. 27; hence, to be consistent with this model and evaluate our model with different major elements, we present results for both calculated Mg and Fe isotope ratios.

In ref. 11, Mg and Fe concentration evolutions are given as a function of percent solidification (PCS) defined by PCS = 100(1 − F_m), where F_m is the mass fraction of residual melt. This model has six stages with the first two being equilibrium crystallization of Ol (up to PCS = 40) followed by Opx until PCS = 76. To use their result (figure 3 in ref. 11), each stage is approximated with a constant bulk solid–melt partition coefficient (Dⁱ_s/m value) for element i. The concentration evolution of i in the magma (Cⁱ_m) for equilibrium crystallization stages is given by

$$\frac{C_m^i}{C_T^i}=\frac{1}{F_m+D_{s/m}^i\left(1-F_m\right)},$$ [1]

where Cⁱ_T is the concentration of i in the total (T) or bulk system. Then, the rest of the concentration evolution is controlled by fractional crystallization forming Opx (PCS = 76–78%), Ol + Pl + pigeonite (Pig) (PCS = 78–86%), Cpx + Pl + Pig (PCS = 86–95%), and Cpx + Pl + Pig + ilmenite (Ilm) (PCS = 95–99.5%). We use constant D values for each stage in the Rayleigh fractional crystallization law:

$$\frac{C_m^i}{C_T^i}=F_m^{\left(D_{s/m}^i-1\right)}.$$ [2]

This allows us to reproduce the Snyder et al. (11) model to calculate the Mg and Fe concentration of the melt at different stages of LMO crystallization. The parameters used for this calculation are given in SI Appendix, Tables S5 and S6. The results in Fig. 2A show Mg and Fe concentration evolutions of the LMO that closely reproduce the curves in ref. 11. For an isotope ratio (i/j), where i and j are different isotopes of an element E, we need to define the mass fraction of the reference isotope j in the melt:

$$f_m^j=\frac{C_m^j{F}_m}{C_T^j}=\frac{C_m^j\left[1-\left(PCS/100\right)\right]}{C_T^j}.$$ [3]

The following equations are used to calculate the isotopic compositions of the melt and cumulate as a function of PCS at each stage that is approximated with a constant fractionation factor $\left({\alpha }_{s/m}^{i/j}\right)$ for different isotope ratios between solid (s) and melt (m). The evolution of the isotope ratios in the melt for equilibrium crystallization in δ notation is

$${\delta }^{i/j}{E}_m=\frac{\left[\mathrm{1,000}+{\delta }_T^{i/j}\right]}{\left[f_m^j+\left(1-f_m^j\right){\alpha }_{s/m}^{i/j}\right]}-\mathrm{1,000.}$$ [4]

The evolution of the isotope ratios in the melt for the stages with Rayleigh fractional crystallization is

$${\delta }^{i/j}{E}_m=\left[\mathrm{1,000}+{\delta }_T^{i/j}\right]\left[{\left(f_m^j\right)}^{\left({\alpha }_{s/m}^{i/j}-1\right)}-\mathrm{1,000}\right].$$ [5]

In both equilibrium and fractional crystallization, the cumulate isotope ratio evolutions are given by

$${\delta }^{i/j}{E}_{cum}-{\delta }^{i/j}{E}_m=\mathrm{1,000}\ln \left({\alpha }_{s/m}^{i/j}\right).$$ [6]

SI Appendix, Tables S7–S9 list the isotope fractionation factors between minerals and melt used in our calculation, the Mg and Fe concentrations for crystallizing minerals, and the resulting bulk isotope fractionation factors, respectively. The calculated curves for the evolution of ²⁶Mg/²⁴Mg and ⁵⁶Fe/⁵⁴Fe of the LMO are shown in δ values in Fig. 2B. The solid and dashed curves represent melt and cumulate compositions, respectively, that evolve during the LMO solidification. Lunar basalts have been produced by partial melting of the LMO cumulates (9, 12, 18, 72). Hence, melting of LMO cumulates following the dashed curves in Fig. 2B tends to produce melts that are similar to the LMO melt shown by solid curves. This is because the isotope fractionation during melting to form lunar basalts involves the same minerals (the source cumulate minerals), but the process is a melt-mineral fractionation instead of a mineral-melt fractionation. Our calculations show that primarily crystallization of Opx and Cpx with slightly heavier Mg isotopic compositions compared with that of the BSM can produce a light residual melt up to −1.716‰ by the end of LMO crystallization (PCS = 99.5%) (Fig. 2B). The blue and yellow rectangles on the Mg isotope data in Fig. 2B show the measured ranges of δ²⁶Mg values of low- and high-Ti lunar basalts, respectively. The measured values of these basalts overlap the Mg isotope evolution curves of the melt and cumulates in our model, which shows that the large isotopic variation of lunar basalts can be the result of the magmatic processes producing these basalts. This implies that a simple average of the isotopic compositions of the basalts is not likely to be representative of the BSM. A similar argument can be made for any other differentiated body, which is ignored by Hin et al. (29). The isotopic compositions of the lunar basalts are, however, consistent with the cumulate remelting model. In particular, the positions of low- and high-Ti basalt fields in Fig. 2B are consistent with the type of source mineral assemblages inferred from the Lu–Hf and Sm–Nd isotope systematics of these basalts (12, 15) as well as some other stable isotopic studies of these samples (17, 18, 20, 27, 45). Based on these studies (12, 15), low-Ti basalts can be produced from an assemblage of olivine and orthopyroxene with trace amount of clinopyroxene that crystallized early in the history of the LMO, whereas high-Ti basalts can be produced from a variety of late ilmenite-bearing mineral assemblages. The calculated enrichment of the heavy Fe isotopes at late stages in the LMO is also shown in Fig. 2B. This enrichment is caused primarily by crystallization of olivine and pyroxenes that produced the cumulate sources of low-Ti basalts. Crystallization of ilmenite with high Fe content at the end of LMO solidification produced the cumulate sources of high-Ti basalts. The blue and yellow rectangles on the Fe isotope data in Fig. 2B show the measured ranges of δ⁵⁶Fe values for low-Ti and high-Ti lunar basalts, respectively, which also overlap with the Fe isotope fractionation curves for the LMO crystallization. This model shows that the crystallization of pyroxenes significantly affects both Mg and Fe isotopic compositions of the source of low-Ti basalts. On the other hand, clinopyroxene crystallization could be the most important factor for explaining the light Mg isotopic composition of high-Ti basalts, whereas ilmenite crystallization has a larger effect on the Fe isotopic composition of these basalts. Lunar basalts show an Fe isotopic composition that is generally heavier than that of the BSE, with the Fe isotopic composition of low-Ti basalts being closer to the BSE and lighter (∼0.1‰ in δ⁵⁶Fe) than high-Ti basalts (27, 32, 45). It has been suggested that the heavy Fe isotope composition of the Moon estimated based on the basalts is due to either evaporation of Fe by the giant impact or isotopic fractionation during the lunar magmatic differentiation (45, 64, 73–75). The isotopic fractionation during the lunar magmatic differentiation is favored in our model because the more recent Fe isotope data of Mg-suite rocks as better analogs of the BSM suggest similar Fe isotopic compositions for the BSE and BSM (27, 32). Magnesium loss during the Moon formation has also been suggested as a cause of small variation seen between the BSM and chondrites (29). However, our study demonstrates that accurate accounting for the Mg and Fe isotopic fractionation during the LMO differentiation is needed before any inferences about evaporative loss of these elements from the Moon can be made.

Fig. 2.

Calculated evolution of δ²⁶Mg and δ⁵⁶Fe values vs. PCS during the LMO differentiation. The model is based on the magma ocean crystallization model of ref. 11 with equilibrium crystallization up to 76% PCS of the LMO followed by fractional crystallization (Cpx, clinopyroxene; Ilm, ilmenite; Ol, olivine; Pig, pigeonite; Pl, plagioclase). (A) The Fe and Mg concentration evolutions during LMO crystallization from ref. 11 are reproduced to have the Mg and Fe contents of the melt and cumulate at different stages to calculate the Mg and Fe isotopic evolutions. (B) The solid and dashed lines show evolution of Fe and Mg isotope compositions of residual melts and instantaneous cumulates, respectively. The shaded blue and yellow boxes are Mg and Fe isotopic compositions measured for low-Ti and high-Ti basalts, respectively (this study and refs. 35 and 39). For low-Ti basalts, δ²⁶Mg is in the range of −0.394 to −0.220‰, and δ⁵⁶Fe is in the range of 0.038–0.110‰. For high-Ti basalts, δ²⁶Mg is in the range of −0.312 to −0.694‰, and δ⁵⁶Fe is in the range of 0.130–0.212‰. The results show that Fe and Mg isotope variations in lunar basalts are consistent with the LMO cumulate remelting model for their origin (9, 10, 12).

Although a new model by Lin et al. (13) makes improvements upon (11), our results are not particularly sensitive to the differences between these two models as can be seen from Eq. 3. Because the changes in fractionating mineral assemblages are sufficiently similar, the differences between the two models are insignificant. We also tested our model considering only fractional crystallization through the LMO crystallization suggested recently (19), which gives similar results for low- and high-Ti basalts except for the less depleted residual melt (−0.76‰) by the end of LMO crystallization (PCS = 99.5%).

Conclusions

Our model shows that both Mg and Fe isotope variations in lunar basalts can be explained by the LMO cumulate remelting model. The Mg isotope measurements of lunar samples combined with this model support the notion that (i) the BSM and BSE have similar Mg isotopic compositions, (ii) Mg isotopes preserve the signature of lunar magmatic differentiation, and (iii) the isotopic composition of the lunar basalts may not be the best representative of the BSM. Magnesium isotopic compositions of the low-Ti basalt 15555 and the KREEP basalt 15386 do not fit our calculated model. Furthermore, although Mg isotopic composition of lunar breccia 14321 fits the model, its chemical composition with high Mg content does not match the late cumulate of LMO crystallization with low Mg content. These anomalies might be related to different and more complex origins for these samples.

Materials and Methods

Samples.

Different types of lunar samples including pristine anorthositic and Mg-suite rocks, lunar basalts, few mineral separates, and several Mg standards are studied here. Petrology, mineralogy, and chemical compositions of these samples are available in the lunar sample compendium (76) and the R. Korotev list of lunar meteorites (meteorites.wustl.edu/lunar). A brief description of these samples is given in SI Appendix.

Analytical Methods.

The sample dissolution process was done in a mixture of HF-HCl-HNO₃ using a CEM MARS 6 microwave digestion system through a three-step procedure. Ion-exchange chromatography procedure established in our group (38) was used for Mg purification. Magnesium isotope ratios were measured with a Nu Plasma II MC-ICPMS in low-resolution mode and wet plasma analysis using the sample-standard bracketing method. The results are reported in δ notation relative to the DSM3 standard in per mil (‰) (SI Appendix, Tables S1 and S2). The uncertainties of the measurements are reported as two SEMs (2σ_m). The long-term reproducibility, which is the 2SD of multiple isotopic measurements of the standards over 18 mo, is better than ±0.08‰ (SI Appendix, Fig. S3 and Table S4). The Mg isotopic compositions of our in-house pure standards (38), Cambridge1 (77), and the US Geological Survey standards (78) are within error of the recommended values (SI Appendix, Table S4). The Mg isotopic compositions of all standards and samples measured here lie on a single mass-dependent fractionation curve following the exponential law (SI Appendix, Fig. S4), which is in agreement with the previous studies (47). More details of the analytical method are given in SI Appendix.