Analysis of lunar samples: Implications for planet formation and evolution

See allHide authors and affiliations

Science  19 Jul 2019:
Vol. 365, Issue 6450, pp. 240-243
DOI: 10.1126/science.aaw7580


The analysis of lunar samples returned to Earth by the Apollo and Luna missions changed our view of the processes involved in planet formation. The data obtained on lunar samples brought to light the importance during planet growth of highly energetic collisions that lead to global-scale melting. This violent birth determines the initial structure and long-term evolution of planets. Once past its formative era, the lunar surface has served as a recorder of more than 4 billion years of interaction with the space environment. The chronologic record of lunar cratering determined from the returned samples underpins age estimates for planetary surfaces throughout the inner Solar System and provides evidence of the dynamic nature of the Solar System during the planet-forming era.

Planetary scientists seek to understand how planets formed and evolved to their present states. Because Earth is geologically active, much of the terrestrial record from the time of planet formation has been overwritten so many times that it is now hard to separate the role of ancient events from more recent ones. In contrast, many features of the Moon have been preserved from this formative era. Some can be studied remotely, but others were revealed only after the lunar samples returned during the Apollo and Luna missions were analyzed in terrestrial laboratories.

Magma oceanography

Prior to the Apollo landings, the following statement represented the prevailing view of terrestrial planet formation: “It seems possible and indeed probable that the earth could and did accumulate below the melting point of silicates throughout its entire growth from a small size to its present one” (1). This reflected the opinion that planets grew by the gentle accumulation of asteroid-sized planetesimals, so that high temperatures only occurred locally in some of the larger impacts. The samples collected during the Apollo 11 mission in 1969 quickly disproved the idea that planetary bodies, even ones the size of our Moon, started out cold. Bulk samples from the Apollo 11 landing site contained a minor component (4%) of anorthosite, an igneous rock type that is known, but uncommon, on Earth. Anorthosite consists predominantly of a single mineral, plagioclase feldspar. A handful of small mineral grains led to the suggestion that the entire lunar highlands were dominated by anorthosite (2, 3). Based on the elevation of the highlands relative to the flat plains of basaltic lavas that infilled the major impact basins on the Moon’s Earth-facing side (originally misidentified by early astronomers as seas, leading them to be called mare basins), the assumption of an anorthositic composition for the highlands implied a highlands crust some 25 km thick (2). The current best estimate for lunar nearside crustal thickness is ~30 km based on lunar gravity data measured more than 40 years later (4).

How could essentially all of the lunar highlands crust be composed of anorthosite? In molten rock (magma) of composition similar to that of the bulk Moon, anorthite plagioclase, a calcium-aluminum silicate, has a lower density than the magma, and so would float. An obvious way to form an anorthositic crust is to assume that a large portion of the Moon was at one time molten, with the anorthositic crust forming by flotation above the crystallizing magma in much the same way that an iceberg floats on the ocean. This idea led to what became known as the “magma ocean” model for early lunar differentiation (Fig. 1).

Fig. 1 Diagram of the lunar magma ocean model.

Convection (arrows) within the magma ocean circulates the magma. Crystallization of dense olivine and pyroxene (green) and buoyant plagioclase (gray) separates the magnesium-rich minerals of the interior from the anorthositic crust. The thickness of the lunar crust is exaggerated by about a factor of 10 for display.

From a magma with the composition expected for the bulk Moon, the dense magnesium silicates olivine and pyroxene would have been the first minerals to crystallize (59). Settling of these crystals to the bottom of the magma ocean left the remaining magma richer in calcium and aluminum, leading to crystallization of plagioclase. Olivine and pyroxene crystallization also drove the remaining magma to higher concentrations of iron, which served to further increase its density. By the time the magma became saturated in plagioclase, as it crystallized, its buoyancy with respect to the magma caused the plagioclase to float to the surface of the magma ocean to form the highlands crust. As the crust formed, it trapped beneath it the remaining magma that was enriched in all elements that were not incorporated into the minerals that had crystallized by that time. These elements include potassium (K), the rare earth elements (REE), and phosphorus (P), leading to the acronym KREEP for the rocks formed from this residual magma (10). Because KREEP is rich in radioactive elements such as uranium, thorium, and potassium, whose concentrations can be measured from orbit because they emit gamma rays (11), we now have maps of the lunar surface that show that KREEP is concentrated dominantly on the Earth-facing side of the Moon, in the region of the major mare basalt-filled impact basins (1113) (Fig. 2).

Fig. 2 Map of the distribution of thorium (Th) on the lunar surface.

Data from the Lunar Prospector and Clementine missions (82) allow division of the lunar crust into three compositionally distinct terranes, outlined here in white. On the near side is the Th-rich Procellarum-KREEP Terrane (PKT); the far side is dominated by the Th-poor Feldspathic Highlands Terrane (top) penetrated by the impact that formed the South Pole–Aitken basin (bottom) that exposed slightly more Th-rich material, but nowhere near as Th-rich as in the PKT. [Figure provided by Bradley Jolliff updated and modified from the figure shown in (12)]

The restricted areal distribution of KREEP could reflect the presence of a KREEP layer beneath the whole crust that was only excavated to the surface by impacts large enough to penetrate through the crust. However, the largest impact basin on the Moon, the South Pole–Aitken basin, shows only slightly elevated abundances of thorium (Fig. 2). Alternatively, both the apparently asymmetric distribution of KREEP and the thicker lunar crust on the farside than on the nearside (4) have led to proposals that the asymmetry originated in the magma ocean era by either gravitational or thermal interaction with Earth (14, 15). The concentration of radioactive KREEP into a smaller portion of the Moon also could have provided a heat source to explain both mare basalt volcanism that continued for more than a billion years after the Moon solidified (16) and the higher number of large impact basins facing Earth than on the lunar farside (17).

The simple magma ocean model described above almost certainly underestimates the complexity of the process. Several investigations have explored more complicated evolution scenarios, including the consequences of density-driven overturn of different layers in the lunar interior after magma ocean crystallization (1820). Nonetheless, the basic magma ocean model explains many characteristics of the lunar samples. Later Apollo missions that landed on, or close to, the highlands returned substantial amounts of anorthosite. The global importance of anorthosite has been confirmed through compositional analysis of the lunar surface from orbit (13), as well as by meteorites from the Moon that provide samples of the lunar crust likely far removed from the Apollo landing sites (21). Some mare basalts have compositions similar to terrestrial basalts, but others have distinctively high contents of titanium (5, 22). These unusual lunar lava compositions most likely reflect a wide range in the composition of the rocks in the lunar interior that melted to create these lava flows (5, 7, 9). This type of compositional diversity in the lunar interior is predicted by the magma ocean model, as it fits the sequence of dense mineral accumulation that would occur as the magma ocean cooled and crystallized (6).

Further evidence in support of a lunar magma ocean comes from measurements of the isotopic composition of elements that receive the decay products of naturally occurring radioactive isotopes. Unless a magma ocean has a thick insulating crust, or atmosphere, above it, heat loss to space causes cooling and crystallization of the magma in a few million years or less (23, 24)—a time that is short relative to the chronological resolution of many radioactive clocks used to determine the age of rocks. If they formed by direct crystallization of the lunar magma ocean, the rocks of the highlands crust should show little variation in age, corresponding to the date of magma ocean crystallization. The data, however, show a dispersion in ages for different highland rocks (25). Whether this reflects the range in actual crystallization ages, or resetting of the radiometric clocks in the samples by slow cooling deep in the crust or by impact-related metamorphism, is unclear. Modern applications of radiometric dating techniques are narrowing the age range of the crustal rocks to 4.36 to 4.40 billion years (25), 150 to 200 million years after the start of Solar System formation. Resolving the age range of the lunar crust is made difficult by the small number of returned crustal samples appropriate for dating and the fact that the oldest crust on the Moon has been repeatedly pummeled by impacting objects. Obtaining a clear answer to the age of the magma ocean, and the lunar crust in general, likely will require a broader collection of crustal rocks to be returned from the Moon.

Another prediction of the magma ocean model is that the anorthositic crust, KREEP, and the accumulated crystals in the lunar interior that would later be melted to produce mare basalt lavas all formed over a relatively short time interval. Radioactive dating of mare and KREEP basalts indicates eruption ages from 4.25 to 2.9 billion years ago (8). At the time of their eruption, however, the basalts were characterized by a wide range in the relative abundances of the isotopes that are produced by the radioactive decay of naturally occurring radioactive elements (26). This is true also for one radiometric system, the 103 million–year half-life decay of 146Sm to 142Nd. In this system, the parent isotope, 146Sm, effectively became extinct about 4 billion years ago. Nonetheless, lunar lavas younger than 4 billion years show a range in 142Nd/144Nd (2729), where 144Nd is a stable isotope of neodymium (Fig. 3). The inherited variability in the isotopic composition of these elements at the time of mare basalt eruption shows that the compositional variability of the basalts can be traced back to the characteristics of the rocks in the lunar interior that were melted to generate these lavas. Combining the radiogenic isotope data for all analyzed lunar igneous rocks (e.g., Fig. 3) suggests that their source regions formed in an initial event that occurred between 4.37 and 4.40 billion years ago (2731), similar to the ages of crustal anorthosites. Such a result is consistent with the idea that the main compositional structure of the lunar crust and mantle was determined by an event of duration less than a few millions of years, as is predicted by the magma ocean model.

Fig. 3 Illustration of a common source history for lunar igneous rocks.

The neutron fluence–corrected 142Nd/144Nd (μ142Nd is the deviation in the sample’s 142Nd/144Nd relative to the laboratory standard in parts per million) is plotted against the 147Sm/144Nd ratio of the sample (highlands rocks) or calculated for the source of the basalts according to their initial 143Nd/144Nd ratio. Both Nd isotope ratios are modified by the radioactive decay of 146Sm (half-life 103 million years) to 142Nd and that of 147Sm (half-life 106 billion years) to 143Nd. The lines on the diagram show the slopes expected for different ages (Byr, billion years) for the process that created the range in Sm/Nd ratios that resulted in the initial isotopic range in these rocks. The alignment of the data for mare basalts, KREEP-rich rocks, and various highland samples along a single line of slope corresponding to an age of 4.38 billion years indicates that the source materials of these lavas formed during a single short-duration event. [Data from (27, 29, 83, 84)]

An energetic birth

The magma ocean model implies that some energetic process involving high temperatures was involved in Moon formation. The nature of that process became clearer from comparison of the composition of lunar and terrestrial rocks. Early analyses of mare basalts were used to argue that the composition of the bulk Moon was similar to that of Earth’s mantle (32). Earth, Mars, and their presumed meteoritic building blocks have subtle stable isotopic differences that reflect imperfect mixing of the contributions of the many stellar nucleosynthetic events in the galaxy that created the elements in our Solar System before the planets began to assemble (3335). Earth and the Moon, however, appear to share identical isotopic composition (36, 37). This holds even for tungsten (38), which tracks the timing of core formation due to the radioactive decay of 182Hf to 182W (39). The timing of core formation is unlikely to be the same on two bodies of such different size as Earth and the Moon. The Moon thus appears to share a very close genetic relationship with Earth.

The idea that the Moon formed from Earth dates back to at least 1879 (40, 41) with the suggestion that the Moon was flung out by centrifugal force from a fast-spinning Earth. Planetary-scale fission of this nature likely requires much more energy than can be provided by spin alone. The compositional similarities of lunar and terrestrial rocks, however, revived the discussion of whether and how the Moon could have been derived from Earth. The focus turned rapidly to the question of whether a large impact into Earth could put enough material into stable Earth orbit to form the Moon (42). The answer appears to be “yes” (4244), but exactly how is still being debated. Early models of the Moon-forming impact involved a glancing collision of the proto-Earth with an object roughly the size of Mars (45). Such an impact could place enough material into Earth orbit to form the Moon. The rapid reassembly of this material into a single body would result in a largely molten Moon, consistent with the evidence for a lunar magma ocean. Another attractive feature of the impact model is that it provides an explanation for the low bulk density of the Moon that is related to its small core, and hence its low iron/silicon ratio. Impact simulations show that the dense iron metal component is preferentially captured by Earth as the iron cores of both impactor and target merge (46, 47). In a glancing-blow impact, however, the major fraction of the Moon is made up of material from the impactor’s mantle, not Earth, thus requiring that the impactor was compositionally similar to Earth. This is not impossible, but neither is it likely given the randomness of planetary accretion and the timing of core formation on different planetary bodies. To address this issue, more recent giant-impact models explore still more energetic collisions (46, 47). The most energetic of these (48) suggests that the outer layers of Earth expanded as a silicate vapor cloud of sufficient diameter to allow Moon formation from the condensed vapor as it cooled. These more energetic impacts mix Earth’s mantle and the impactor to the point where any isotopic distinctions between target and impactor before the impact would be no longer resolvable.

Although the role of giant impacts in more general aspects of planet formation remains unclear, the evidence for a lunar magma ocean and the Earth-Moon compositional similarity—both derived primarily from analysis of returned lunar samples—displaced the idea of cold, gentle planet assembly with one dominated by highly energetic impacts. As a result, the role of giant impacts is now an intrinsic part of models for planet formation (49), as are magma oceans as the initial stage of planet differentiation (23).

Keeping hydrated

Another early indication that the Moon formed hot came from the depletion of moderately volatile elements—such as the alkali metals, halogens, and lead—in lunar samples relative to Earth (32), which is itself depleted in these elements relative to the Sun and primitive meteorites (50). Given the very high temperatures (several thousand kelvin) expected in a giant impact, more highly volatile compounds such as water were presumed to have been completely lost from the Moon (51). Hydrous minerals in lunar basalts are extremely rare, but whether this resulted from completely dry parental magmas or the loss of water from the magma during its eruption into the vacuum of the lunar surface was not clear. Deposits of volatile-driven pyroclastic volcanism—for example, the orange soils noticed by Apollo 17 astronaut Harrison Schmitt—were found at most Apollo sites and now have been mapped more widely from orbit (52). The identity of the gas fueling their explosive eruption is not clear. The water and CO2 contents of samples of the pyroclastic glasses were too low to be quantified by the analytical tools available during the Apollo era. Improvements in the detection limits for hydrogen provided by secondary ion mass spectrometry allowed hydrogen to be detected in these pyroclastic materials (53). The moderate water contents so calculated were later supported by direct measurements on inclusions of glass formed when small amounts of melt were trapped inside minerals that had crystallized in the magma chamber prior to eruption (54). The melt inclusions were kept from outgassing by the crystal that encapsulates them. Related work on the mineral apatite, a calcium phosphate that can incorporate water into its crystal structure, also has provided evidence for moderate water contents in lunar magmas (55). The presence of water in lunar magmas has prompted reexamination of the fate of volatile elements during the giant impact (56). More generally, results from lunar sample analysis have driven a reinvestigation of the mechanism by which the terrestrial planets obtained water and carbon, with the implication that these volatile components may have been present when the terrestrial planets formed, rather than having been added later by accretion of icy bodies such as comets or icy asteroids (57).

Sampling the space environment

The Moon-forming giant impact was not the last collision to be experienced by either Earth or the Moon. Only a few parts per million of the surface area of Earth dates to before 3.8 billion years ago, but most of the lunar surface is that old or older. The Moon’s surface thus provides the record of bombardment from the era when impacts were a common force influencing planetary surfaces.

At the smallest scale, the Moon’s surface is continually bombarded by energetic atoms from cosmic rays and the solar wind. The surface traps some of these atoms and thus serves as a long-term record of the composition of the solar wind (51, 5860). Examination of lunar samples showed the presence of surface coatings of iron and titanium, caused by irradiation with solar wind protons, which, coupled with the small grain sizes of the lunar regolith, contributed to the Moon’s low surface reflectance (albedo) (61). This “space weathering” occurs on the surface of any Solar System body that is not protected by either an atmosphere or a magnetic field, with consequences for the interpretation of spectroscopic data for other airless bodies in the Solar System, particularly asteroids (62). Lunar “swirls,” patterned color differences on the lunar surface that reflect focused particle irradiation directed by lunar magnetic fields, are another expression of this phenomenon (51, 63). Recent measurements of the remanent magnetism in lunar samples suggest that the Moon had a global magnetic field of a few tens of microteslas (Earth’s magnetic field ranges from about 25 to 65 microteslas at the surface) from at least 4.25 to 2.5 billion years ago (64).

At much larger scales, remote observation of the Moon shows the lunar highlands to be more heavily cratered than the mare basins. This provides a relative chronology for the lunar surface, showing the mare to be younger than the highlands. Converting this relative crater chronology into an absolute chronology for the lunar surface became possible through analysis of the samples returned by Apollo. Age determinations for rocks from the lunar surface (65) allowed calibration of the lunar impact flux through time (Fig. 4). This flux estimate can be extrapolated throughout the inner Solar System, allowing the cratering record for other planets and moons to be turned from relative to absolute chronologies for their surfaces (66). Such data provide basic information on the rate of planetary resurfacing either destructively by erosion, or constructively by volcanism, with the latter providing key information on the dynamics of planetary interiors.

Fig. 4 Impact rate on the lunar crust through time.

Calibration points are based on crater densities from different age terranes sampled at the Apollo (designated by A) and Luna 16 landing sites. The gray peak at 3.9 billion years represents the hypothesized spike in impacts associated with the late heavy bombardment. [Figure adapted from (85)]

Individual lunar craters have had their ages determined via geochronologic studies of lunar breccias and particularly impact melts. Unexpectedly, ages near 3.9 billion years occurred repeatedly (67). The same age was also seen in a number of lunar highlands rocks (68). This result was used to suggest that several of the large lunar basins on the Moon’s nearside formed at roughly the same time because of a large increase in the influx rate of bombarding planetesimals. The idea of a “terminal lunar cataclysm” or “late heavy bombardment,” as this event has been called (69), led to a wide variety of hypothesized causes and consequences including orbital migration of the gas giant planets (70), an explanation for the limited amount of preserved terrestrial crust older than 3.9 billion years (71), and the delayed start to life on Earth (72, 73).

Recent analyses of lunar and meteoritic samples have questioned whether the late heavy bombardment is required by the lunar impact data (74). The debate is informed by two approaches. One is the measurement in lunar impact rocks of the abundance of a group of elements, called the highly siderophile elements (siderophile indicates a solubility preference for iron metal rather than silicate), which are present at 10−6 g/g or higher levels in meteorites, but at 10−9 g/g or lower abundances in planetary mantles that have lost most of these elements to the core. The goal of this approach is to use the abundance patterns of these elements in lunar rocks to discriminate the types of impactors involved and thus distinguish one impact from another (7578). An impact the size of a major basin-forming event (e.g., the Imbrium basin) will scatter ejecta over a good portion of the lunar surface, so associating any particular surface-collected rock formed by impact with the crater that ejected it is not straightforward. The other approach examines the probability that the record of older impacts is overprinted by younger impacts until the rate of impact falls below some critical level (69), in which case the “late heavy bombardment” may simply reflect the end of a rapidly declining rate of impact (Fig. 4). Humans conducted field work on the Moon for less than 12 days, covering a very limited portion of the Earth-facing side of the Moon. Refining the chronologic sequence of large impact events on the Moon requires more detailed field studies conducted in future visits that cover a larger portion of the lunar surface.

Lunar samples never stop giving

Many of the discoveries discussed above were made shortly after the lunar samples became available for analysis. Others, however, were delayed until analytical techniques, and the knowledge base needed to ask new questions, advanced. Examples of the long-term yield from lunar samples are the determination of their water content, continuing refinements in the ages of lunar samples and particularly the evidence for an early lunar differentiation event, and measurements that show a long-lived lunar magnetic field. Some outstanding problems likely will require new sample collection efforts—for example, determining the true age range of the lunar crust and hence the age of the Moon, and resolving whether the late heavy bombardment reflected a restricted interval of enhanced impacts or the tail of a rapidly declining impact flux. With the still improving ability to make measurements on smaller samples, another opportunity lies in exploring the archived quantity of lunar samples. Interior images of breccias using x-ray tomography (79) can search for rock clasts within the breccias that might be of use for additional age determinations, particularly of the ancient lunar crust. Determining whether the data obtained from Apollo and Luna samples can be extrapolated to the whole Moon, or whether they constitute a biased record heavily influenced by the nearside basin-forming events, can only be addressed by sample returns from new locations well removed from the Apollo and Luna sites. The recent Chang’e 4 landing in the South Pole–Aitken basin on the far side of the Moon has returned data suggesting that this impact penetrated into the lunar mantle (80). Much can be learned by remote analysis, and indeed the many lunar orbital missions conducted since Apollo inform the choices for where to sample next, but many of the results discussed above required the most advanced analysis possible in terrestrial laboratories, and hence the return of lunar samples to Earth (81).

References and Notes

Acknowledgments: I thank two anonymous reviewers for suggestions that substantially improved the paper. Funding: Writing of the paper was supported by the Carnegie Institution for Science. Competing interests: None.
View Abstract

Stay Connected to Science

Navigate This Article