Over the years since the Apollo era, the current lunar interior structure has been investigated using seismic, gravity, and magnetic field data. The Apollo seismic network recorded about 1,800 meteoroid impacts, 28 energetic shallow moonquakes (with body wave magnitudes up to five and hypocenters about 100 km below the surface), and about 7,000 extremely weak deep moonquakes that were located about halfway to the center of the Moon (e.g., Wieczorek 2009). The deep moonquakes are very enigmatic in that their occurrences are correlated with the tides raised by the Earth, they involve very low stress drops (less than 1 bar), they appear to originate from about 300 “nests” that are repeatedly activated, and almost all of these nests are located on the Moon’s nearside hemisphere (Nakamura 2003; Bulow et al. 2007; Qin et al. 2012). The nearside distribution of the deep moonquakes could perhaps indicate that the farside hemisphere is seismically inactive. Such a hypothesis is possible, especially if one considers that almost all of the Moon’s mare basalts erupted on its nearside hemisphere. Alternatively, it is possible that the deepest interior of the Moon attenuates seismic signals originating from farside events. In support of this hypothesis, seismic arrivals along ray paths passing through regions deeper than 700 km radius are not detected, suggesting mantle in this region hot enough to be seismically attenuating and possibly partially molten (Nakamura 2005). One scenario that could be consistent with the above observations is that numerous magma-filled fractures currently located in the deep mantle are capable of relieving the small stresses induced by Earth-raised tides.

The Lunar Laser Ranging mission (Williams et al. 2001) and recent Gravity Recovery and Interior Laboratory (GRAIL) mission (Zuber et al. 2012) provide us the gravity measurements. Especially the Love numbers from the GRAIL mission are unprecedentedly accurate. These Love numbers have been used to invert the bulk solid-body “dissipation factor” Q (Williams et al. 2001) as well as the tidal tomography (Zhong et al. 2012). The Q number is around ~30, which is compatible to that of the Earth and Mars. This Q number infers a core–mantle boundary (CMB) temperature ~1,400 °C (Nimmo et al. 2012). This temperature is below the harzburgite mantle solidus while higher than ilmenite-mixed harzburgite solidus (Zhang et al. 2013a, see Fig. 1). Joint inversion of seismic and gravitational data (Khan et al. 2006) suggests a CMB temperature range from 1,170 °C to 1,500 °C, consistent with that inferred from the Q facter. To gain a partial melt, the lunar deep mantle has to be chemically heterogeneous (either ilmenite rich or water rich (Evans et al. 2014)).

Fig. 1
figure 1

The present-day mantle temperature profiles from various studies (Modified from Zhang et al. 2013a)

Full size image

The surface and orbit magnetic records have also been used to explore the present-day mantle temperature (Sonett 1982). The electrical conductivity profile of the Moon has been estimated by using electromagnetic-sounding techniques, whereby the relationship between simultaneous time variations in the ambient magnetic field as measured from orbit and the resulting field measured on the surface was analyzed (Wieczorek 2009; Hood et al. 1982). This electrical conductivity profile suggests the lunar mantle temperature from 1,100 °C to 1,250 °C (see Fig. 1).

The lunar mantle is energetically interacting with its core. A number of seismic and gravity studies identify the lunar core with a radius of about 250–430 km (Wieczorek 2009; Weber et al. 2011; Garcia et al. 2011). The core is believed to have liquid outer and solid inner regions based on seismic reflections at an inner–outer core boundary. Weber et al. (2011) estimated a 240 km radius solidified inner core within a core of 340 km radius. The existence of a liquid outer core has also been suggested to explain the lunar rotational dissipation inferred from the Lunar Laser Ranging data (Williams et al. 2001). Seismic travel time inversion (Lognonne and Johnson 2007) suggested a density for the lunar core consistent with the presence of light elements which would lower the melting temperature allowing for a fluid outer core. This light element is most likely sulfur (Righter and Drake 1996; Weber et al. 2011; Stevenson and Yoder 1981).

Besides the above observed present-day structures, several paleo-structures related to the lunar evolution history have also been observed. The Moon is believed to pass through the giant impact (Canup 2012), magma ocean (Warren 1985), late heavy bombardment (Hartmann 2003), and the potassium, rare earth element, and phosphorus (KREEP) migration (Warren 1988) during its early evolution. The magma ocean solidification determines the layering of mantle mineralogy, while the KREEP migration process distributes the heat-producing elements and largely influences the lunar mantle evolution to the present day. The asymmetric distribution of mare basalts at the Procellarum KREEP Terrane is an indicator of the mantle evolution. The age of basalt samples from the Mare region spans from 3.9 to 3.2 Ga (Nyquist and Shih 1992). Mare basalts were erupted primarily on the current lunar nearside. These age and surface distribution of the mare magmatism provide the major constraint on the evolution of the Moon.

The presence of a paleomagnetic field also provides a physical constraint on core and mantle evolution. Intrinsic remnant magnetism of the lunar crust has been long recognized from the Apollo samples. Recent paleointensity measurements on selected lunar samples using recently developed methodologies (e.g., Lawrence et al. 2008) have indicated that a core dynamo may have existed on the Moon from at least 4.2 to 3.6 Ga (Garrick-Bethell et al. 2009; Shea et al. 2012; Tikoo et al. 2012). The existence of a convectively driven core dynamo requires a minimum heat flux from the core (e.g., Stevenson et al. 1983; Stegman et al. 2003) or an alternative energy source for the dynamo (Dwyer et al. 2011; Le Bars et al. 2011).

Globally distributed lobate scarps (a type of tectonic landforms) are observed (Binder 1982) outside the mare-filled basins, by the Lunar Reconnaissance Orbiter Camera (http://lunar.gsfc.nasa.gov/mission.html). The ages of these lobate scarps are from 3.5 Ga to the present day, identified with associated crater ages as well as the cross-cutting relationships (Watters et al. 2010). Most of these deformations are associated with the lunar global contraction through the cooling of the Moon. The absence of identifiable thrust faults limits the total net contraction that has occurred (Watters et al. 2010), indicating a decrease in radius less than 1 km after 3.8 Ga (MacDonald 1960; Solomon 1977; Watters et al. 2010), while, gravity anomalies derived from the GRAIL data suggest the presence of early volume expansion of the Moon at ~3.8 Ga (Andrews-Hanna, et al. 2013).

To understand the present-day structures and reconcile with lunar evolutions inferred from the paleo-records, four different classes of models have been proposed. Wasson and Warren (1980) first suggested that the crystallization of the magma ocean was asymmetric. A locally thicker crust on the farside would give rise to a thinner underlying magma ocean, and thus concentrate KREEP-rich materials on the nearside, where the crust is thinnest. This model requires an initially thicker crust on the farside, possibly caused by a giant impact on the nearside (Neumann et al. 1996; Byrne 2007), or a global convection pattern within the magma ocean that could have transported the crust preferentially to the farside (Loper and Werner 2002).

Second, impacts are often invoked to explain KREEP localization. Ghods and Arkani-Hamed (2007) showed that impacts would generate subsurface thermal anomalies that might redistribute KREEP-rich materials located beneath the crust. Depending on their size, the impacts would either mix this layer into the underlying mantle or simply concentrate KREEP at the edge of the basin, thus enhancing volcanic activity there. Another view involving impact is from Jutzi and Asphaug (2011), who claimed that a low-velocity impact from a 1,200 km diameter companion on our moon might have displaced the global KREEP-rich layer to the nearside.

Third, Zhang et al. (2013a) assumed that a global layer of dense, late-stage ilmenite cumulates quickly sank to the core–mantle boundary, carrying along with it a large fraction of the KREEP layer. This layer involves to a large extent (degree-1) upwelling caused by thermal expansion due to the radioactive heating, explaining both the present-day distribution of heat sources and the timing of volcanism. Their model is also applied to explain the lunar core dynamo history and the surface contraction history (Zhang et al. 2013b). A more recent study by Qin et al. (2012) showed that the correlation between deep moonquakes and mare basalt could be the consequence of this ilmenite-cumulate layer, if it was enriched in water. A related model, proposed by Parmentier et al. (2002), showed that the downwelling of a mixed ilmenite-cumulate layer itself could follow a degree-1 pattern and concentrate KREEP beneath the PKT (though, see also Elkins-Tanton et al. (2002) for comments about that model).

Finally, the influence of the PKT on lunar evolution has been investigated by Wieczorek and Phillips (2000) and Hess and Parmentier (2001) before and by Laneuville et al. (2013) recently. Wieczorek and Phillips (2000) developed an axially symmetric 3-D conduction model which showed that partial melting of the underlying mantle is an inevitable outcome of a thick KREEP layer on the nearside hemisphere and that volcanism should span most of the lunar history. This result was confirmed by Hess and Parmentier (2001) in a 1-D thermal conduction study but they also noted that the wide, partially molten region caused by that layer could form an impenetrable barrier to the eruption of mare basalts. They concluded that the hypothesis of a thickened KREEP layer below the PKT imposed strong constraints on the concentration of heat sources in the PKT and crustal thickness to remain consistent with both geological and petrological observations. Recently, Grimm (2013) reanalyzed the results of Wieczorek and Phillips (2000), showed that such models also predict large gravity or topography anomalies and electrical conductivity signatures that may be inconsistent with observations, and suggested that these studies used too simplified conduction models. Adding the convection to the new model, Laneuville et al. (2013) investigated the influences of the PKT with the lunar mantle convection. They find that in addition to localizing most of the melt production on the nearside, such an enriched concentration of heat sources in the PKT crust has an influence down to the core–mantle boundary and leaves a present-day temperature anomaly within the nearside mantle.

Two of above four views well reconcile the present-day structures with the inferred lunar evolution. They are of Laneuville et al. (2013) and Zhang et al. (2013a). The difference between them is the origin of the mare basalts: shallow versus deep (Shearer et al. 2006). Laneuville et al. (2013) keep the KREEP material located in a 10 km thick spherical cap with an angular radius of 40° at the base of the crust which heats due to radioactive decay, which, in turn, causes the deeper mantle convection and melts the underlying mantle material. This model explains some of the mare basalts as well as the connection between the mare basalts and Procellarum KREEP Terrane. Zhang et al. (2013a) assume that some of the IC sinks into the lunar mantle and is subsequently brought up to a depth of several hundred kilometers by fluid dynamic processes, leading to decompression melting (Hess and Parmentier 1995; Zhong et al. 2000). The mixture of ilmenite-cumulate and olivine orthopyroxene, formed during the sinking of ilmenite-cumulate in an olivine and pyroxene mantle, would settle deep in the mantle because it was more chemically dense (Hess and Parmentier 1995; Elkins-Tanton et al. 2002; Zhong et al. 2000).

While the details of these contrasting models can be improved, it is necessary to distinguish between them in order to understand the thermal history and current structure of the Moon. These different models also make specific testable predictions, including significantly reduced deep moonquake activities on the farside, relative to the nearside, and distinct seismic structure at large depths on the nearside, all of which can be tested with future deployment of seismometers, particularly on the farside.