Mafic granulite xenoliths from the East Indian Shield: Evidence for recycled continental crust in the Archean mantle

Lower-crustal rocks commonly occur in large outcrops (granulite terrains) as well as small fragments (xenoliths) brought to the surface by fast-erupting volcanics. However, granulite xenoliths differ from the exposed granulite terrains in two important aspects. First, many granulite terrains are Archean, whereas very few granulite xenolith localities are situated in the Archean crust. Second, granulite terrains tend to be dominated by evolved compositions, whereas granulite xenoliths are generally dominated by mafic lithologies (Rudnick, 1992).

Turning our attention to the granulite xenoliths reported in the literature, it can be seen that almost all of them are associated with later volcanic eruptions of alkali-basalt and kimberlite (table 7–1 inRudnick, 1992). On the other hand, xenoliths, particularly of granulite-facies assemblages, are rarely associated with plutonic rocks. Recently, cognate mafic granulite xenoliths associated with massif-type charnockite in the Eastern Ghats mobile belt, India, have been described (Bhattacharya et al., 2001; Kar et al., 2003), which, quite significantly, represent one of these rare associations. In these publications, the genetic link between massif-type charnockite and hornblende–mafic granulites, via hornblende dehydration melting and the restitic nature of the hornblende–mafic granulites occurring as cognate xenoliths, were established from mineralogical-geochemical relations (Kar et al., 2003). Alternatively, in view of the density differences, this mafic granulite–charnockite association in the Eastern Ghats belt could be due to mafic enclave patches being caught up within charnockitic plutons. If that was the case, then these enclaves may be considered as xenoliths.

On the basis of geochemical signatures described in the literature (Tomson et al., 2006), charnockite massifs of South India are considered to be deeply eroded arc magma (Santosh et al., 2009). In contrast, the charnockite massifs of the Eastern Ghats have been described as a product of hornblende dehydration melting under granulite-facies conditions (Kar et al., 2003). The tectonic environment of arc collision is the most likely cause of granulite formation in the Eastern Ghats belt, similar to Fiordland, Ironside Mountain (Frost and Frost, 2008). Geochemical signatures, namely, ferroan composition with negative Sr and positive Ti anomalies in the charnockite massif of the Eastern Ghats belt (Kar et al., 2003), are also compatible with such a tectonic environment.

In this communication, we describe the mineralogy, geochemistry, and Sr-Nd isotopic compositions of mafic granulites and pyroxenites occurring as enclaves within massif-type charnockitic rocks of the Eastern Ghats granulite terrain. The geochemical and isotopic signatures in these xenoliths may reveal the nature and mantle source of their protoliths (Condie, 2001), in addition to a possible tectonic environment of mafic magmatism and granulite metamorphism.

The Eastern Ghats granulite belt along the east coast of India, bounded by the Singhbhum craton in the north and Bastar craton in the west, has experienced polyphase deformation and complex, possibly multiple granulite-facies metamorphic events (Bhattacharya et al., 1994; Dasgupta et al., 1994; Sen et al., 1995; Bhattacharya, 1996; Dasgupta and Sengupta, 1998; Bhattacharya and Kar, 2002; Dobmeier and Raith, 2003). Although, generally considered as a Proterozoic terrain, some Archean domains have been recognized, and, more recently, Archean granulite-facies events have also been recorded in this terrain (Ramakrishnan et al., 1998; Rickers et al., 2001; Bhattacharya et al., 2001). The massif-type charnockite and associated mafic granulites form a dominant component in this large granulite terrain; other exposed lithologies include metapelitic granulites, calc-granulites, peraluminous granitoids, and migmatites (Fig. 1). Although contact between charnockite and other important granulite lithologies such as metapelitic granulites and calc-granulites are not exposed, the plutonic nature of the charnockite is evident from expansive bodies of continuously varying compositions, from tonalite through granodiorite to adamellite and granite. Based on Nd model ages, Rickers et al. (2001) described several distinct crustal domains in the Eastern Ghats granulite belt and noted that most of the internal segments are Paleoproterozoic domains (1.8–2.2 Ga), while marginal segments, adjoining the Singhbhum craton in the north and the Bastar craton in the west, are Archean (3.2–3.9 Ga) domains. Although, a Late Mesoproterozoic imprint is overwhelming in the Eastern Ghats belt, 1.6 Ga and 3.0 Ga granulite events have also been suggested (Mezger and Cosca, 1999; Bhattacharya et al., 2001).

The massif-type charnockite has the common imprint of a streaky gneissic foliation, designated S₁, which is axial planar to some rootless folds represented by hornblende–mafic granulite (Fig. 2A). This indicates that protoliths of hornblende–mafic granulite were present before development of S₁ and hence could be considered as inclusions in the host charnockite. Also, there are some discordant blocks within the charnockitic gneiss (Fig. 2B).

The most commonly observed enclaves have the assemblage: hornblende-clinopyroxene-orthopyroxene-plagioclase–Fe-Ti oxides ± quartz ± garnet. In these hornblende–mafic granulites, many hornblende grains have embayed grain boundaries with pyroxene and plagioclase in the embayed portions, indicating growth of pyroxene and plagioclase at the expense of hornblende (Fig. 3A). Additionally, quartzofeldspathic films at hornblende margins are sometimes observed (Fig. 3B). These petrographic features indicate prograde heating in these enclaves, which could be related to crystallization of charnockitic magma. Additionally, some pyroxenite xenoliths are found in the marginal segments, around Jenapore in the north and Jaypore in the west. These have the assemblage: clinopyroxene-orthopyroxene–Fe-Ti oxides ± plagioclase ± quartz.

Mineral composition data were obtained by electron microprobe analysis (EMPA) at the Institute Instrumentation Centre, Indian Institute of Technology (IIC, IIT), Roorkee, and Central Petrological Laboratory (CPL), Geological Survey of India, Kolkata, using Jeol Jxa 8600 and CAMECA Sx 100 machines, respectively. Operating conditions used at IIC, IIT, Roorkee, were 15 kV accelerating voltage, 20 nA sample current, and 3 μm beam diameter; while conditions were 15 kV accelerating voltage, 12 nA sample current, and 1 μm beam diameter at CPL, Kolkata.

Bulk chemical analysis was carried out by X-ray fluorescence (XRF) spectrometry at the National Geophysical Research Institute, Hyderabad. Operating condition for XRF spectrometer was 20/40 kV for major oxides and 50/60 kV for trace elements. Nominal analysis time was 300 s for all major oxides and 100 s for each trace element. The overall error in accuracy (% relative standard deviation) for major and minor oxides was less than 5% and that for trace elements was less than 12%. The average precision was better than 1.5%.

Trace elements including rare earth elements (REEs) and isotopic analyses were carried out at the Institute Instrumentation Centre, IIT, Roorkee, by means of an inductively coupled plasma–mass spectrometer (ICP-MS), and a thermal ionization mass spectrometer, respectively. For ICP-MS analysis, the average precision was ±4.1% RSD.

For isotope dilution and isotopic composition, Rb, Sr, and REE fractions were separated from the rock solutions using Bio-Rad AG50 X 8 ion exchange resin in silica glass columns. Sm and Nd were separated from REE fractions using prepacked LN Spec Resins (bought from Eichrom Technology INC, Dorien, Illinois, USA). The total procedure blank in the laboratory was less than 8 ng of Sr and ∼1 ng of Nd during the period of analysis. Rb, Sr, Sm, and Nd abundances were determined by isotope dilution method. All isotopic measurements were done on a Thermo Electron TRITON T1 fully automatic variable multicollector mass spectrometer at the Institute Instrumentation Centre, IIT, Roorkee, based on 2σ (standard error) statistics. Measured ratios for isotopic composition were normalized to ⁸⁶Sr/⁸⁸Sr = 0.1194 for Sr and ¹⁴⁶Nd/¹⁴⁴Nd = 0.7219 for Nd. The measured ratio of ⁸⁷Sr/⁸⁶Sr for SRM-987 Sr standard was 0.710248 ± 20 (2σ) (quoted value 0.710245) and that for ¹⁴³Nd/¹⁴⁴Nd for Ames Nd standard was 0.512138 ± 8 (2σ) (quoted value 0.512138).

The representative mineral compositions are presented in Table 1.

Clinopyroxenes are Mg-rich augite, but variable X_Mg values between 0.61 and 0.75 in the different suites are observed. However, in all the suites, clinopyroxenes are the most magnesian, compared to the coexisting hornblende and orthopyroxene.

Orthopyroxenes are enstatite, but of variable X_Mg in the different suites, e.g., between 0.47 and 0.63. They are the least magnesian.

Hornblendes are Ti-rich pargasites, but they have variable composition, with X_Mg lying between 0.58 and 0.70 in different suites.

Plagioclase feldspars are mostly andesine, but they have variable X_Ca values in different suites: at Jenapore between 0.28 and 0.34; at Jaypore between 0.35 and 0.36; at Naraseraopet between 0.48 and 0.50; at Paderu between 0.56 and 0.57; and at Sunki between 0.89 and 0.91. This suite is highly calcic.

In the Eastern Ghats belt, most of the mafic granulites described in the literature are two-pyroxene granulites, although secondary hornblende has also been reported in some cases (Dasgupta et al., 1991, 1993). From the South Indian granulite terrain, Prakash (1999) has described some prograde hornblende with corroded margins. Similarly, from the Eastern Ghats belt, Kar et al. (2003) have reported prograde hornblende in mafic granulites within massif-type charnockite. The X_Mg variation between coexisting Clinopyroxene > hornblende > orthopyroxene in the mafic granulite xenoliths (Table 1) from the Eastern Ghats granulite belt is also consistent with the experimental results of hornblende dehydration melting (Patino Douce and Beard, 1995). At 7 kbar and 900 °C, X_Mg in the coexisting phases has been found to be: clinopyroxene 0.67; hornblende 0.62; and orthopyroxene 0.56. Along with the textural relation (Fig. 3A), the X_Mg variation between coexisting phases confirms the prograde nature of the hornblende in the mafic granulite xenoliths. Also, subtle Mg enrichment at hornblende grain margin is consistent with prograde heating (Rapp and Watson, 1995; Springer and Seck, 1997).

These mineralogical data together with the field relations and petrographic features indicate the xenolithic nature of the hornblende mafic granulites.

The bulk compositional data of the xenoliths are presented in Table 2. The hornblende–mafic granulites are characterized by variable normative An: at Jenapore between 13 and 23; at Jaypore between 13 and 22; at Naraseraopet between 14 and 24; at Paderu between 13 and 32; and at Sunki between 24 and 47. This bulk compositional variation corresponds to plagioclase composition in different suites (Table 1). Most of the samples are olivine normative (5–24); only a few have small normative quartz (<5). Also, normative hypersthene (0–44) and diopside (9–45) are highly variable. Bulk composition is similar to low-K tholeiite, although there is a notable compositional variation, particularly in Mg#, lying between 40.12 and 73.71. The little variation in CN/CNK molecular (CaO + Na₂O)/(CaO + Na₂O + K₂O) values between 0.91 and 0.99 coupled with wide variation in Mg# would indicate some inhomogeneity in source compositions. However, these chemical variations cannot be assigned to a previously differentiated suite, because no co-variation between Niggli alk with either SiO₂ or Mg# has been observed. A Mg# versus SiO₂/Al₂O₃ diagram suggests that in spite of the compositional variations, the hornblende–mafic granulite xenoliths, though somewhat evolved, are of basaltic melt compositions (Fig. 4). It is important to note that mafic granulite xenoliths from western Hungary have been argued, on similar ground, as being solidified melts rather than mafic cumulates (Kempton and Harnon, 1992). On the other hand, pyroxenites with very little quartzofeldspathic material are clearly the result of pyroxene accumulation.

The pyroxenites (Table 2) have ultramafic bulk composition, with high normative Ol, between 13.78 and 25.55; high normative Di, between 13.36 and 61.15; and low normative Or, between 0.12 and 3.41. However, wide variation in Mg#, between 34.4 and 72.21, and only a little variation in CN/CNK values, between 0.96 and 1.0, indicate some heterogeneity in source rock composition.

The hornblende–mafic granulites are relatively rich but highly variable in transitional trace elements: Cr: 63–976 ppm, except sample JN1/2 with 1834 ppm; Ni: 20–71, except a few samples with high-Ni contents between 180 and 1094 ppm. Base metals are: Cu 6–254 ppm, average 50 ppm; Zn 57–169 ppm, average 97 ppm; and Ti 2278–12,108 ppm, average 7818 ppm (Table 3). The Cr and Ni contents of most samples are typical of primitive to slightly fractionated mantle-derived magmas. The mafic granulite xenoliths described here are distinctive in their mineralogy, compared to those described from different parts of the world, in that abundant prograde hornblende is present and garnet is typically absent. Other examples are some of the mafic granulite xenoliths described from the Deccan Trap, India, and from Fidra, Scotland, which are actually two-pyroxene granulites without garnet (Desai et al., 2004; Downes et al., 2001). Mafic granulite xenoliths from the northern Baltic Shield on the other hand are typically garnet granulites (Kempton et al., 2001). The mafic granulite xenoliths described here are relatively poor in incompatible elements: Rb 1–15 ppm, average 6.8 ppm, except JN2/1 and Paderu suite samples, which have an average of 61 ppm. The high Rb values in the Paderu samples relate to additional prograde biotite in them. Ba content varies from 62 to 625 ppm, with an average of 241 ppm, except two samples with high-Ba contents of 1996 and 2056 ppm. The higher contents of Rb and Ba in some samples could relate to more modal feldspars in them. However, no significant depletion in Sr at 73–654 ppm, average 170 ppm, is observed. Variations of trace-elements Sr, Ce, Ni, and Zr are not related to variation in Mg#, and this further attests to the fact that the chemical variability in these mafic granulite xenoliths is not the result of a previous magmatic differentiation and/or crystallization of minerals. The low-Ni content is comparable to that of average arc basalt (McCulloch and Gamble, 1991). In order to characterize the tectonic setting of the basalts, they have been classified into three groups, which are considered separately—one group consists of the hornblende granulites of the Archean crustal domains of Jenapore and Jaypore (Rickers et al., 2001), the second group consists of samples of the Paleoproterozoic crustal domains of Sunki and Naraseraopet, and the third group consists of samples of Paderu, which have additional prograde biotite in them.

For the Archean domain, the trace-element distribution relative to primitive mantle (Fig. 5A) shows a conspicuous peak in Ba, negative Th and K anomalies, and a strong negative Nb anomaly, all characteristic of arc basalts (McCulloch, 1993; Condie, 2001). A significant negative Zr anomaly and positive Eu anomaly are also similar to those observed in arc-derived basalts. Also high La/Nb ratios, mostly >1, are compatible with arc basalts (Rudnick, 1995; Condie, 1999). However, unlike arc basalts, a positive Ta anomaly could have been the effect of later granulite-facies metamorphism. The other significant contrast with arc basalts is in the negative Sr anomaly, and this could be related to the subsequent granulite-forming reactions, when plagioclase in these xenoliths did not equilibrate (Johannes, 1983; Patino Douce and Beard, 1995). Low values of Nb/U (average 2.5) further indicate crustal contamination.

In the Paleoproterozoic domains, trace-element distributions relative to primitive mantle (Fig. 5B) show a conspicuous peak in Ba, negative Th and K anomalies, and a strong negative Nb anomaly, characteristic of arc basalts (McCulloch, 1993; Condie, 2001). Significant negative Zr anomalies are similar to those observed in arc-derived basalts. Also, high La/Nb ratios, mostly >1, are compatible with arc basalts (Rudnick, 1995; Condie, 1999). However, positive Ta anomalies and the lack of significant positive Sr anomalies could have resulted from granulite-facies metamorphism. Particularly, the lack of significant positive Sr anomalies could be related to subsequent granulite-forming reactions, when plagioclase in these xenoliths did not equilibrate (Johannes, 1983). In this connection, it is important to note that plagioclase compositions in hornblende-mafic granulite xenoliths in this domain differ significantly from sample to sample (X_ca = 0.50–0.89). Very low values of Nb/U (average 0.92) are indicative of crustal contamination. In the Paderu suite xenoliths, with additional prograde biotite, trace-element distributions relative to primitive mantle (Fig. 5C) show Rb enrichment, no Ba spike or negative Th anomalies, but negative Sr anomalies. Although negative Nb anomalies and high La/Nb ratios are similar to those observed in arc basalts, most of the incompatible element contents are high, similar to those of the ocean-island basalts (OIBs). The tectonic setting of the basalts (protoliths of mafic xenoliths) of this suite cannot be conclusively defined on the basis of the trace-element pattern. Crustal contamination in the mantle source region might have caused such aberrations.

Pyroxenites are rich in transitional trace elements: Cr between 158 and 1574 ppm, average 827 ppm; Ni between 39 and 694 ppm, average 231 ppm; base metals: Cu between 11 and 118 ppm, average 39 ppm; Zn between 58 and 318 ppm, average 175 ppm (Table 3). On the other hand, they are poor in incompatible elements: Rb between 2 and 12 ppm, average 7 ppm; Ba between 4 and 505 ppm, average 226 ppm; and Sr between 9 and 165 ppm, average 93 ppm. These trace-element signatures are compatible with their mineralogy, namely, two-pyroxene assemblage with very little feldspar in them.

In the Archean domain, the hornblende–mafic granulite xenoliths are light (L) REE enriched (La/Yb)_N = 1.8–15.3; while LREE fractionation, (La/Sm)_N = 1.6–3.8, is more than heavy (H) REE fractionation, (Gd/Lu)_N = 0.93–3.5. The chondrite-normalized plot shows significant negative Eu anomalies, with Eu/Eu* between 0.18 and 0.35 (Fig. 6A). Negative Eu anomalies and LREE enrichment are also characteristics of the plagioclase-rich garnet granulite xenoliths from the northern Baltic Shield (Kempton et al., 2001). However, Baltic Shield xenoliths with garnet are more enriched in LREE, (La/Yb)_N = 10.5–23.8, while the present suite xenoliths are garnet-free, and instead have abundant prograde hornblende. In the Paleoproterozoic domain, the mafic xenoliths are LREE enriched, (La/Yb)_N = 1.5– 10.3; while LREE fractionation, (La/Sm)_N = 1.02–3.7, is more than HREE fractionation, (Gd/Lu)_N = 0.76–2.1. Many of the samples show small negative Eu anomalies between 0.4 and 0.9 (Fig. 6B). Four samples with positive Eu anomalies could reflect effects of granulite-facies metamorphism, namely, lack of plagioclase equilibration; incidentally, variable and more calcic plagioclase feldspars are recorded from these suites (Table 1). The Paderu mafic xenoliths, with additional biotite, are also LREE enriched, (La/Yb)_N = 2.7–5.6; while LREE fractionation, (La/Sm)_N = 1.9–2.9, is more than HREE fractionation, (Gd/Lu)_N = 1.1–1.7. The chondrite-normalized plot shows small negative Eu anomalies, with Eu/Eu* between 0.33 and 0.95 (Fig. 6C).

Sr-Nd isotopic composition of the mafic xenoliths is presented in Table 4.

In view of the previously defined crustal domains (Rickers et al., 2001; Dobmeier and Raith, 2003), these are grouped into two isotopically distinct groups, namely, Paleoproterozoic and Archean, respectively. Archean domains (north and west margin of Eastern Ghats belt) also contain some pyroxenite xenoliths.

Whole-rock Rb-Sr and Sm-Nd isotopic data could not be used for precise age determination of the granulite-facies event. In fact, whole-rock isotopic systems are useful for dating cogenetic magmatic suites, while dating of metamorphism commonly uses minerals, like garnet (Sm-Nd), zircon, and monazite (U-Pb). In this study, our main interest is to find the age of the protoliths, or the age of mafic magmatism, which can best be estimated by Nd model ages. The Nd model age or depleted mantle age (T_DM) was calculated with reference to CHUR (chondritic uniform reservoir) values of ¹⁴³Nd/¹⁴⁴Nd as 0.513151 and ¹⁴⁷Sm/¹⁴⁴Nd as 0.222 and using a decay constant λ of 6.54 ×10^{−12 yr−1} (DePaolo, 1988). However, some samples, marked with an asterisk in Table 4, give aberrant results, indicating mineralogical reconstitution during granulite-facies metamorphism. For these samples, a two-stage model age was calculated using Milisenda et al.’s formulation (Milisenda et al., 1994). For the northern and central Eastern Ghats belt, 1.0 Ga as the dominant granulite-forming event is considered, and hence Nd model dates for the samples with aberrant results were recalculated for this 1.0 Ga granulite event. The Naraseraopet suite belongs to the southern Eastern Ghats belt, for which a 1.6 Ga granulite-forming event has been reported (Mezger and Cosca, 1999).

For the Paleoproterozoic domains, T_DM ranges between 2.01 and 2.8 Ga, with an average of 2.5 ± 0.03 Ga. In general, T_DM represents the time of separation of the protolith of a rock from the depleted mantle source. On this basis, the average crustal residence time of all the rocks in the Paleoproterozoic domain can be taken to be close to 2.5 Ga. It is noted that modified calculation, using Milisenda's formula for some samples, actually removed the aberrations, and the results became consistent with the rest of the data set.

For the Archean domains, T_DM ranges between 3.0 and 3.7 Ga, with an average age of 3.3 ± 0.01 Ga, while for the pyroxenites, this range is between 3.2 and 4.1 Ga, with an average of 3.9 Ga. The average crustal residence ages for the two can thus be taken to be 3.3 and 3.9 Ga, respectively.

Mafic granulites associated with massif-type charnockites have so far been described as basic members of intrusive/magmatic charnockites (Subba Rao and Divakara Rao, 1988; Sheraton et al., 1996; Young et al., 1997). In contrast, the mafic granulites described here have been interpreted as xenoliths within the host massif-type charnockites on the basis of the field relations, and petrographic and mineralogical evidences.

Magnesian composition (Mg# between 40 and 74) with positive Sr and negative Nb and Ti anomalies is compatible with arc magmatism in subduction zone (Tomson et al., 2006; Santosh et al., 2009). The primitive mantle–normalized Nb/U ratio could be an important tracer of mantle source composition of oceanic basalts; the mean value for 166 mid-ocean-ridge basalts is 47 ± 11, and the mean value for 500 “non–enriched mantle–type” OIBs is 52 ± 15 (Hofmann, 2004). In contrast, a mean value of 8 in continental crust was reported by Rudnick and Fountain (1995). The low values in the hornblende–mafic granulite xenoliths (average 1.55) in this study clearly point to recycled continental crust in the mantle source of the xenoliths.

Foundering of lower continental crust into underlying convecting mantle has been proposed to explain the evolved composition of Earth's continental crust (Kay and Kay, 1991; Rudnick, 1995), and rare direct evidence of this process has been provided from North China craton (Gao et al., 2004).

For the Paleoproterozoic domains, the average crustal residence time is 2.5 b.y., and this may be interpreted as the age of mafic magmatism in this domain. The initial ⁸⁷Sr/⁸⁶Sr ratios for all the Paleoproterozoic samples, calculated at 2.5 Ga, range between 0.70647 and 0.72343; while their ε_Nd values calculated at 2.5 Ga range between −6.09 and +2.55. The large variation in the initial ⁸⁷Sr/⁸⁶Sr ratios, in particular, is caused in part by the decay of ⁸⁷Rb, which might have occurred during later high-temperature metamorphism, which is commonly recorded in the Eastern Ghats belt (Mukhopadhyay and Basak, 2009, and references therein). The high ⁸⁷Sr/⁸⁶Sr values and negative or low ε_Nd values indicate that they preserved time-integrated LREE enrichment, which is also evident in Figure 6B. In the Sr-Nd correlation diagram (Fig. 7A), most of the samples plot in quadrant IV, indicating sources that were enriched in Rb and depleted in Sm. However, some samples plot in quadrant I, which represents magma sources with coupled enrichment of both Sm and Rb. Since such coupled enrichments are contrary to the geochemical properties of these elements, the elevated ⁸⁷Sr/⁸⁶Sr ratios presumably reflect Sr derived from continental crust. It is noteworthy that basaltic rocks rarely plot in quadrant I, one exceptional case being that of the Lesser Antilles (Faure, 1986). Our preferred interpretation for such a pattern in the Sr-Nd correlation diagram is contamination of the mantle source with recycled continental crust, which is also evident in the geochemical signatures discussed in the previous section. Most importantly, at 2.5 Ga, high and variable ⁸⁷Sr/⁸⁶Sr ratios and negative or slightly positive ε_Nd values of some samples indicate an enriched mantle source. However, strongly positive ε_Nd values in the majority of samples indicate significant juvenile crustal addition around 2.5 Ga. Incidentally, this is identical with the 2.5 Ga globally prominent crust-forming event (Condie et al., 2009).

For Archean domains, the average crustal residence time is 3.3 b.y., and this may be interpreted as the age of mafic magmatism in the marginal segments (Singhbhum and Bastar craton margin) of the Eastern Ghats belt. The initial ⁸⁷Sr/⁸⁶Sr ratios calculated for all the Archean samples at 3.3 Ga range between 0.70103 and 0.71974. The ε_Nd values calculated at 3.3 Ga range between −5.95 and +3.67. At 3.3 Ga, most of the samples have negative ε_Nd values, and this, together with high radiogenic Sr, is indicative of recycled continental crust in the mantle source region for these samples as well.

It is important to note that mafic magmatism of corresponding ages is reported from the adjacent cratons. In the Singhbhum craton, mafic magmatism spread from ca. 3.3 Ga (oldest enclaves of orthoamphibolites) to ca. 0.1 Ga (newer dolerite dike swarms), as summarized by Bose (2009). And Neoarchean (ca. 2.6 Ga) mafic magmas have recently been reported from the Bastar craton (Srivastava, 2006).

In the Sr-Nd correlation diagram (Fig. 7B), most of the samples plot in quadrant IV, except one, which plots in quadrant I. The enriched Rb and depleted Sm values for the samples in quadrant IV could be interpreted as the characteristics of the source, which remained isolated for a sufficient interval of time to acquire the distinctive isotopic compositions of Sr and Nd. However, in view of the geochemical signatures, particularly low Nb/U ratios, our preferred interpretation is contamination of the mantle source by recycled continental crust. However, newly formed crust at 3.3 Ga (positive ε_Nd values) is also evident in this shield region.

For the pyroxenite xenoliths, the initial ⁸⁷Sr/⁸⁶Sr ratios and ε_Nd values calculated at 3.9 Ga range between 0.70227 and 0.741 and between −7.3 and + 6.2, respectively. The pyroxenite xenoliths could be cumulates, and yet their isotopic signatures are in no way distinguishable from those of the hornblende–mafic granulite xenoliths. Their enriched compositions at 3.9 Ga in the marginal segment also indicate recycled continental crust in the mantle source region.

Isotopic compositions of mafic granulite xenoliths in the present study provide clear evidence of recycled continental crust in the mantle source region. For the marginal segments, this continental recycling must have occurred prior to 3.3 Ga, and this implies growth of the continental crust in the early Archean in these segments. From the age distribution in the continental crust, Condie (1998) proposed that continents grew episodically with major periods of growth at 2.7, 1.9, and 1.2 Ga. These are also major orogenic periods worldwide. From the Eastern Ghats granulite belt, Rickers et al. (2001) considered two high-grade events at 1.6 and 1.0 Ga, and none of these orogenic periods was associated with new crustal growth, as indicated by negative ε_Nd values in that report. However, both positive and negative ε_Nd values at 2.5 and 3.3 Ga, recorded from the mafic xenoliths in the present study, indicate both preexisting (recycled?) and juvenile continental crustal growth in these periods. Most importantly, 2.5 Ga is close to the major orogenic period worldwide (Condie et al., 2009).

With trace-element and isotopic evidence of recycled continental crust in the mantle source region of the basaltic melts (protolith of the mafic xenoliths), it is imperative to consider partial melting models. Because REEs are mostly immobile during melting, we consider only REE composition of the mafic xenoliths. The assumed source mineralogy is typical of garnet peridotite xenoliths: olivine = 60; orthopyroxene = 15; clinopyroxene = 20; garnet = 5. Although higher proportions of garnet are also reported (Xu Yigang, 2000), the modal proportions assumed here give the closest match for the mafic xenoliths. REE composition of the mantle source is estimated as: 70% of garnet peridotite plus 30% of global average subducted sediments (data sources: McDonough and Sun, 1995; Plank and Langmuir, 1998, respectively) and 15% melting give the best match. Batch melting calculations were performed using NEWPET computer program, and selected results are given in Table 5. In the chondrite-normalized plot, although significant spread in most of the REEs is observed in the mafic xenoliths, 15% melting of a source composed of 70% garnet peridotite and 30% of GLOSS (Global subducting sediments) results in a melt composition that passes through the xenoliths ensemble (Fig. 8).

Two granulite-forming events at ca. 3.0 and 1.6 Ga are commonly ascribed to collision tectonics (Santosh et al., 2009). These were preceded by subduction and associated slab melting at ca. 3.3 and 2.5 Ga, respectively, as indicated by the arc magmatic signatures in the mafic granulite xenoliths.

Prograde heating in the hornblende-bearing mafic granulite xenoliths is characteristically associated with the massif-type charnockites of the Eastern Ghats granulite belt, India.

Protoliths of the hornblende-bearing xenoliths are interpreted as arc-derived basalts, based on trace-element characteristics. Certain trace-element ratios also point to recycled continental crust in the mantle source of the xenoliths.

Sr-Nd isotopic data from the xenolith suites indicate two periods of mafic magmatism, based on average crustal residence ages in the Eastern Ghats belt: ca. 3.3 and 2.5, respectively. Further, all the xenolith suites have evolved isotopic compositions, which, along with negative ε_Nd values in some samples, indicate a recycled continental crust in their mantle source. The time of 2.5 Ga is particularly associated with significant new crustal growth, indicated by positive ε_Nd values of the majority of the samples.

Since collision tectonics are commonly considered as a possible mechanism for granulite metamorphism, a tectonic scenario for the southeastern Indian Peninsula, namely, two cycles of subduction-collision, may be envisaged.

The Indian Statistical Institute, Calcutta, provided the infrastructural facilities. Department of Science and Technology, government of India, provided financial support for the analytical work, in the form of a research project. We gratefully acknowledge the analytical facilities provided by the National Geophysical Research Institute, Hyderabad, Indian Institute of Technology, Roorkee, and Geological Survey of India.