There are three main approaches which have been used to estimate the chemical composition of the upper mantle: (1) using analysis of primitive peridotites; (2) using mantle melt-residue relations; (3) using cosmochemical constraints.
The first approach makes use of petrological and geochemical analyses of tectonically exposed mantle rocks such as massif peridotites and ophiolite bodies, abyssal peridotites and xenoliths in kimberlites and alkali basalts. For example, Jagoutz et al. (1979) used the average composition of six seemingly fertile spinel lherzolite xenoliths (Fig. 1.4) which were believed to have experienced only very small degrees of partial melting. High pressure and high temperature experiments (Fujii and Scarfe, 1985; Hirose, 1997; Hirose and Kawamoto, 1995; Hirose and Kushiro, 1993) have shown that 10-30% partial melting of lherzolite is able to produce ordinary basaltic melts; lherzolites, therefore, represent a fertile bulk composition. By contrast, harzburgites or dunites represent the most melt-depleted refractory mantle bulk compositions (Fig. 1.4). Care must be taken to choose such primitive samples. On the one hand, most natural peridotites are depleted in incompatible elements to different extents, i.e. they have lower contents of CaO, Al2O3, Na2O, etc. than the fertile mantle is expected to have (Palme and O'Neill, 2014). On the other hand, some peridotites with the highest CaO and Al2O3 may also not represent the pristine peridotite because they show evidence of metasomatism which may replenish incompatible elements after melt extraction (Palme and O'Neill, 2014). However, aside from a few inclusions in diamonds that may have a deeper origin, most mantle rock samples come from depths less than ~ 200 km, therefore studies on natural samples can constrain only the composition of the upper mantle. Isotope and trace element heterogeneities found in ocean island basalts (OIB), which are considered to derive from plumes rising from the Earth’s lower mantle, have led geochemists to argue for an undepleted and undegassed reservoir as a result of limited
16 mass exchange between the lower and the upper mantle (Albarède and van der Hilst, 2002;
Arevalo and McDonough, 2010; Jochum et al., 1983).
Figure 1.4 Mineralogical classification ternary diagram for peridotites and pyroxenites. Peridotite have > 40% olivine. The shaded field represents the range of values for most upper mantle peridotite samples. The arrows indicate the melting trend from lherzolite (L) to harzburgite (H) to dunite (D) (Modified from McDonough and Rudnick, 1998).
Rather than directly using the chemical composition of natural samples, the second approach is based on melt-residue relations. Pyrolite, a theoretical model mantle composition conceived by Ringwood as the source rock for mid-oceanic ridge basalts (MORB), was constructed by mixing a mantle-derived magma (basaltic or komatiitic) with a refractory residue (harzburgite or dunite) in proportions so that the resultant model mantle would contain 3-4 wt.% CaO and Al2O3 and olivine of approximately Fo89 composition (Ringwood, 1975; Sun, 1982). The term ‘Pyrolite’ refers, thus, to a model-dependent composition instead of a rock type and consists of a mineralogy dominated by olivine >
pyroxene and capable of yielding basaltic magmas during partial melting. Whether this mineralogical model is also capable of describing the composition of the lower mantle is still a matter of debate. However, if the mantle convects and mixes as a single unit then the asthenospheric mantle that melts beneath ridges should have the same composition as the lower mantle.
17 The third approach consists of constructing compositional models based on chondritic meteorite compositions. Chondrite classification is based on bulk chemistry, oxygen isotopic composition, mineralogy, petrology and proportions of various chondritic components (Krot et al., 2014). Fourteen groups of chondrites have been recognized and thirteen of them comprise three major classes: carbonaceous (C), ordinary (O), and enstatite (E), each of which contains different groups. If we assume these chondrites to be the possible building blocks of the Earth, the mass ratio between the Earth’s core and mantle as well as their chemical compositions could be calculated based on the bulk composition of these meteorites. CI carbonaceous chondrites may be the most suitable for this purpose as they are the most primitive chondritic meteorites having a composition which closely matches that of the solar photosphere (Allègre et al., 1995; Li and Fei, 2014; Lodders, 2003). Most meteorite-based Earth models assume that the refractory lithophile elements have chondritic ratios but models then differ in the way they consider major element abundances.
Some models assume that the Earth has a bulk major element composition equal to that of CI carbonaceous chondrites or enstatite chondrites (Allègre et al., 1995; Javoy, 1995), whereas other models assume that the Earth is depleted in major elements (50%
condensation temperature TC=1355-1250 K) relative to the refractory lithophile elements (TC=1850-1355 K) (McDonough and Sun, 1995; Palme and O'Neill, 2014). As Mg and Si are depleted in the upper mantle relative to refractory lithophile elements, when compared to all chondritic meteorites, the first class of models appeal either to a superchondritic Mg/Si ratio or to an additional reservoir of Si either in the core or lower mantle. Such models are, therefore, often cited as evidence that the lower mantle is chemically different from that of the upper mantle. The question of whether the lower mantle is isochemical with the upper mantle is therefore a complex issue which is still controversial. Some believe peridotitic or pyrolitic materials are dominant in the whole mantle (e.g. Ringwood, 1962), while others claim a more Fe and Si-rich lower mantle (e.g. Anderson, 1989; Hart and Zindler, 1986; Liu, 1982). Although it is difficult to unambiguously resolve such a controversy based on the current seismological observations and mineral -physics data (e.g. Bina, 2003; Mattern et al., 2005), most geophysical observations such as seismic velocities and electrical conductivity
18 measurements are in reasonable agreement with a pyrolitic whole mantle composition to a first approximation, although the uncertainties are currently very large. Indeed, seismic tomography supports ‘whole-mantle’ circulation with oceanic lithosphere subducted into the lower mantle and a return flow of upwelling plumes into the upper mantle (Bercovici and Karato, 2003). This evidence, however, does not exclude the possibility that the mass exchange between the upper and lower mantle has been only partial during the history of the Earth and that very deep regions of the lower mantle may have indeed a different composition (Ballmer et al., 2017; Kellogg et al., 1999).
Estimated bulk silicate Earth/upper mantle compositions from different studies using various methods are compared in Table 1.1, which show remarkable similarity to one another in major element concentrations. As mentioned above, models assuming similar bulk Earth major element concentrations with chondritic meteorites require a lower Fe and higher Si content for the lower mantle (Liu, 1982).
Table 1.1 Major element composition of pyrolite calculated by different studies.
1 2 3 4 5 6 7 8 9 10 11a 11b Allègre et al. (1995); 10 Palme and O'Neill (2014); 11 Liu (1982). Mg#, molar Mg/(Mg+Fe). aupper mantle; bmore silica-rich lower mantle.
19