Interaction between mafic and felsic melts are reported in all studied plutons in the Borborema Province (Mariano & Sial, 1990; Neves & Vauchez, 1995; Neves & Mariano, 1997; Campos et al., 2002). Absence of cumulates, relatively constant K2O with increasing SiO2 (Figure 3a), linear trends in major- (e.g., Figure 3c and d) and trace-elements (e.g. Figure 4c and d) against MgO diagrams, and lack of negative Eu anomalies in REE diagrams (Figure 5a), indicate that samples with higher SiO2 reflect mixing with granitic melts rather than fractional crystallization. The relatively low δO18 whole-rock values (< 8.2 %), although higher than typical mantle values of 5-6 %, indicate that assimilation of high δO18 crustal rocks was minor if at all.
Given the relatively high MgO and low SiO2 in most dioritic samples, mafic or ultramafic sources are required. Since geochemical and isotopic characteristics are not compatible with an origin by partial melting of normal mantle peridotitic sources, other possibilities include mafic lower continental crust, oceanic crust, mantle wedge above subducting oceanic plate, and subcontinental lithospheric mantle.
Partial melting of mafic protoliths appears unlikely because most dehydration-melting experiments of metabasalts have produced tonalitic to granitic melts rather than dioritic ones (e.g. Rushmer, 1991; Patiño-Douce & Beard, 1995). Wolf & Wyllie (1994) and Rapp & Watson (1995) report experimental melts produced at temperatures above 1000°C with SiO2 contents similar to those of the studied diorites. However, their TiO2 and MgO contents are lower and Na2O/K2O ratios much higher (>> 2-3) than in the Borborema Province diorites (Table 1). Also, the high temperatures required to produce the low- SiO2 experimental melts are hard to attain at crustal pressures, even by repeated injections of basaltic magmas at the Moho, the most efficient mechanism to heat up the lower crust (Petford & Gallagher, 2001; Annen & Sparks, 2002). Mafic components within the subcontinental mantle, rather than in the lower continental crust, could be considered as an alternative. However, partial melting of eclogites, the stable mafic rock type at upper mantle conditions, produces melts strongly depleted in heavy REEs and Y due to partitioning of these elements into garnet (Green, 1994; Rapp & Watson, 1995; Rapp et al., 2003). This is inconsistent with the observed REE patterns and concentration of HREE (ca. 5-10 times chondritic values) (Figure 5). Similar reasoning can be applied against partial melts from subducted oceanic crust, from which they can also be distinguished by negative εNd(t) values.
The major-element chemistry of our most primitive dioritic samples, except for their high K2O content, are similar to: (a) mid-ocean ridge basalt (MORB) from a segment of the southwestern Indian ridge characterized by high Na2O (3.3-4.0 wt%) and Al2O3 (16.1-18.3 wt%), low TiO2 (1.2-1.5 wt%), and low CaO/Al2O3 (0.56-0.66) (Meyzen et al., 2003); (b) experimental melts produced by partial melting of lherzolite under water-undersaturated conditions (0.5 and 0.9 wt% H2O added) at 1100°C and 1 GPa (Hirose and Kawamoto, 1995). The unusual MORB compositions have been attributed by Meyzen et al. (2003) to partial melting of a mantle depleted in clinopyroxene, due to earlier melting events. This explanation can also account for the higher CaO content of the experimental melts as compared with the Borborema diorites (Table 1). Thus, a refractory source appears to be involved in the genesis of the diorites. This source could only be subcontinental lithospheric mantle, depleted due to early melt extraction of continental crust. Because the high contents of incompatible elements cannot be explained by interaction with granitic magmas (e.g., Ba is higher in the diorites than in the coexisting granites; Figure 4b; Neves et al., 2000), metasomatic enrichment of the source following melt depletion is required. Calculated compositions produced by dehydration melting experiments of phlogopite-bearing peridotite at 1075°C and 1GPa under water-undersaturated condition (Conceição & Green, 2004) are similar to the Borborema diorites, except for lower FeO and higher MgO and CaO (Table 1). This represents a strong argument favorable to the genesis of the diorites by small degree partial melting (5-8 %; Conceição & Green, 2004) of a metasomatized mantle source.
Nd model ages of diorites suggest that the metasomatic event occurred in the Paleoproterozoic (Table 1), during which much of continental crust in Borborema Province was produced (Van Schmus et al., 1995; Brito Neves et al., 2000, Neves, 2003 and references therein). The possibility that the old TDM ages reflect crustal contamination through assimilation of supracrustal rocks can be ruled out because these rocks have lower TDM ages than the diorites (typically 1.0-1.3Ga; Van Schmus et al., 1995; 2003)
The source proposed for Tibetan shoshonites is also lithospheric mantle metasomatized well before (0.9-1.3 Ga; Turner et al., 1996; Chung et al., 1998) the partial melting event. Higher alumina contents in the Borborema diorites (Figure 3b) might reflect higher degrees of partial melting from a source with similar composition and/or an alumina-poorer source for the shoshonites. The first possibility is consistent with higher LILE contents of the shoshonites.