Melt extraction compositional trends

While all spinel-lherzolite facies suites show remarkably similar compositional trends as a function of depletion, some garnet peridotite xenoliths in kimberlites and lamproites from ancient cratonic lithospheric keels show signih-cantly different trends (e.g., see Boyd, 1989 Chapters 2.05 and 2.08). Most of these xenoliths are extremely depleted extrapolation of the trends back to the PM MgO of 36.7% gives similar concentrations of Si02, EeO AI2O3, and CaO to the spinel Iherzolites (O Neill and Palme, 1998) the difference in their chemistry is due to a different style of melt extraction, and not a difference in original mantle composition. 

Direct evidence for the compositional effects of partial melt extraction is preserved in samples of upper-mantle lithosphere with a range of ages, including Archean cratonic mantle, Proterozoic subcontinental mantle, and modern oceanic mantle. Samples of upper mantle are collected as xenoliths, peridotites dredged from oceanic fracture zones, and slices of upper mantle tectonically exposed at the surface, and extensive samples exist from both oceanic and continental settings (see Chapters 2.04 and 2.05). Here, data sets are assembled for oceanic and subcontinental mantle lithosphere, and compositional trends are compared to those predicted for partial melt extraction from fertile peridotite in order to deduce the role that melt extraction has played in producing compositional variability in upper-mantle lithosphere, and to place constraints on the thermal evolution of the mantle. 

Figure 6 Major-element oxides (wt.%) versus FeO as a function of pressure (GPa) and degree of batch melt extraction (sources the 1 GPa and 2 GPa trends are based on the Kinzler and Grove (1992a, 1993) model for melting of primitive mantle of McDonough and Sun (1995) (composition 1, Table 1), and the trends at higher pressures are based on the data of Walter (1998) for melting of fertile peridotite KR4003).

Figure 10 shows major-element oxides versus Mg for off-craton and oceanic mantle, as well as some estimated compositions for primitive mantle (Table 1). As expected from the normative plots, the two sets of mantle compositions have distinct trends for all oxides. Previous models for primitive upper mantle have a range in Mg from 89 to 90, and Figures 9 and 10 show that the oceanic and off-craton trends also converge within this range. Assuming that the off-craton and abyssal mantle trends are due primarily to melt extraction from a common protolith, then the intersection of the trends should provide a good estimate for the composition of fertile upper mantle for major elements. 

Figure 13 Normative spinel Iherzolite mineral abundances (wt.%) versus rock Mg for oceanic mantle (as in Figure 7) relative to trends for 0-25% batch melt extraction at 0.5-2 GPa. The starting composition is fertile upper mantle as determined in this study (Table 1, 8), and residues are calculated using the melting model of Kinzler and Grove (1992a, 1993).

Eased on two main lines of evidence, Niu et al. (1997) concluded that abyssal peridotites are the end products of melt extraction followed by variable amounts of olivine crystallization. First, in their set of reconstructed compositions they found that model fractional and batch melt extraction trends could not reproduce major and minor element variations in their data set. Most importantly, they found that melt extraction models failed to account for the strong positive correlation between FeO and MgO, as well as incompatible minor-element concentrations. Specifically, at a given Na20 or Ti02 content, abyssal peridotites are enriched in MgO relative to model melt extraction residues. Niu et al. (1997) showed that these compositional anomalies can be reconciled by a model of melt extraction followed by olivine crystallization, with more MgO-enriched samples having more accumulated olivine. If correct, this model has important implications for understanding melt extraction at oceanic ridges, and it has recently been the focus of re-evaluation. 

Figure 18 Normative olivine and opx (garnet Uierzo-Ute norm) versus rock Mg showing low-T cratonic mantle relative to batch melt extraction trends from 3 GPa to 7 GPa (see Figure 6). Open circles are compositions from the Kaapvaal, Siberian, and Slave cratons, and the filled circles are compositions from the Tanzanian, Canadian, and Greenland cratons.

---