Anorthosite-leuconorite suite

Anorthosites typically contain plagioclase (An_55-67),orthopyroxene (X_Mg = 0·55-0·60), subordinate clinopyroxene (X_Mg = 0·68-0·75), biotite, ilmenite and garnet with accessory amounts of apatite, interstitial K-feldspar, quartz and retrograde muscovite, calcite and zoisite. Leuconorites contain higher modal amounts of biotite (up to 10 vol. %), less magnesian pyroxenes (X_MgOpx = 0·37-0·55 and X_MgCpx = 0·60-0·66) and less calcic plagioclase (An₅₀) (Fig. 2). The orthopyroxene megacrysts in anorthosites are commonly mantled by vermicular intergrowths of orthopyroxene and calcic plagioclase (An_70-80). The intergrowths are interpreted to have formed by subsolidus decomposition of aluminous low-Ca pyroxene. Garnet has two modes of occurrence. In anorthosites it occurs as isolated inclusion-free grains (X_Mg = 0·35-0·25) along with plagioclase and exclusive of primary ferromagnesian silicate phases. These garnets typically show coronitic replacement by symplectitic intergrowths of plagioclase (An_77-90) and orthopyroxene (X_Mg = 0·56-0·58). In leuconorites, on the other hand, less magnesian garnet (X_Mg = 0·14) occurs as coronal phase intergrown with quartz that mantles pyroxenes (X_MgCpx = 0·48; X_MgOpx = 0·37) in contact with plagioclase (An₄₇). Similarly, biotite occurring as coarse flakes exclusive of ferromagnesian phases appears to be of igneous origin. However, biotites formed by subsolidus replacement of orthopyroxene, ilmenite and garnet are more common, especially in the leuconorites.

Figure 2. (a) Fe/(Fe + Mg) vs Ca/(Ca + Fe + Mg) diagram illustrating the solid solution ranges of ferromagnesian phases in the high-grade assemblages of ferrodiorites (FD), anorthosites-leuconorites (AN-LN) and ultrabasic rocks (UB) from the Bolangir anorthosite complex. For the sake of clarity, ferropargasitic amphibole has been omitted and the compatibility relations are indicated only by few tie-lines. (b) An-Ab-Or ternary showing the compositions of relict magmatic (plg I) and metamorphic (plg II, Kfsp) feldspars in leuconorites-anorthosites and ferrodiorites.

The extremely rare ultrabasic lenses in the marginal leuconoritic zone of the massif are coarse-grained rocks that exhibit a granoblastic texture. The mineralogy comprises olivine (X_Mg = 0·70-0·62), orthopyroxene (X_Mg = 0·75-0·70), clinopyroxene (X_Mg = 0·85-0·78), clusters of recrystallized magnesio-pargasite (X_Mg = 0·75-0·68) and minor plagioclase (An_55-45). Orthopyroxene is mantled by fine-scale spinel (X_Mg = 0·60-0·54) + plagioclase (An₇₅) ± clinopyroxene intergrowths, which are interpreted to have recrystallized from aluminous low-Ca pyroxene. Matrix plagioclase in contact with pyroxenes shows development of coronitic garnet (X_Mg = 0·48). Dynamic recrystallization caused the plagioclase crystals to coarsen and become clouded with spinel arrayed parallel to their twin lamellae.

Ferrodiorite suite

Ferrodiorites are reported for the first time from anorthosite massifs of the Eastern Ghats Belt. These extremely melanocratic rocks are fine-grained and homogeneous rocks that exhibit a conspicuous porphyritic structure defined by centimetre-sized plagioclase phenocrysts (An_48-40) in a matrix of mafic phases. The porphyritic structure is best preserved in ferrodiorites that occur within the anorthosite and thus largely escaped deformation, whereas ferrodiorites in the strained border zone developed a moderate to strong planar fabric and distinct augen structure.

The ferrodiorites grade in composition and mineralogy from fayalite-bearing, quartz-free and extremely melanocratic varieties to fayalite-absent, less melanocratic plagioclase-rich varieties, some of them quartz bearing. The melanocratic matrix in the fayalite-bearing ferrodiorites is composed of a well-recrystallized granoblastic mosaic of orthopyroxene (X_Mg = 0·28-0·17), olivine (X_Mg = 0·13-0·08), clinopyroxene (X_Mg = 0·37-0·30) and ilmenite, with rare ferropargasite (X_Mg = 0·30-0·20) and annite-rich biotite. Zircon, apatite and metamict thorite are present in more than accessory amounts. They are commonly enclosed in pyroxenes, but never in phenocrystic plagioclase. Large zircon grains are characterized by multiple growth rings with well-defined terminations, overgrown occasionally around subrounded, heavily fractured cores. Pyrrhotite occurs intergrown with chalcopyrite and pentlandite. A conspicuous feature of these rocks is the extensive growth of garnet synchronous with dynamic to static recrystallization of the igneous assemblage. It occurs as coronitic garlands of euhedral crystals formed near the interface of plagioclase phenocrysts with the mafic matrix phases, and also as isolated discrete euhedral crystals in the centres of plagioclase grains. Occasionally, in the cores of weakly strained plagioclase phenocrysts, garnet developed in a skeletal fashion. Coronal garnet commonly is intergrown and separated from the mafic matrix by completely recrystallized, unzoned and strongly antiperthitic plagioclase grains (An_29-33). Phenocrystic plagioclase, however, towards the contacts with garnet, shows distinct compositional zoning with decreasing anorthite content and development of antiperthite exsolutions.

The melanocratic matrix of the fayalite-absent ferrodiorites consists of orthopyroxene (X_Mg = 0·42-0·25), subordinate clinopyroxene (X_Mg = 0·53-0·35) formed through granular and lamellar exsolution from orthopyroxene, garnet (X_Mg = 0·17-0·09) and ilmenite, with rare ferropargasite and annitic biotite. Commonly the rocks are strongly deformed and exhibit blastomylonitic fabrics, with shape-preferred orientation of matrix orthopyroxene laths, well-recrystallized flattened aggregates of strongly antiperthitic plagioclase (An_58-50) developed from former plagioclase phenocrysts, and a few large ribbons of quartz. Yet, some less deformed specimens still preserve a distinct porphyritic texture. Typically these aggregates contain interstitial vermicules of quartz adjacent to K-feldspar granules. Garnet occurs exclusively as a coronitic phase that girdles the plagioclase aggregates and, in some cases, clusters to dense aggregates. Ubiquitous accessory phases are zircon, apatite, metamict thorite and the sulphides pyrrhotite, chalcopyrite and pentlandite.

Near the contacts with the garnetiferous granite gneisses, silicic ferrodiorites grade into coarse-grained ferromonzodioritic rocks. The contact relations with the gneisses, however, are not well exposed. Ferromonzodiorites are not found exclusive of ferrodiorites. These mesocratic rocks show a streaky to laminated gneissic foliation and distinct mylonitic microstructures, with extreme stretching of the former coarse quartz grains and aggregates of mafic minerals. They are distinguished from the silicic ferrodiorites by their uniform, more felsic composition and the presence of perthitic K-feldspar in the assemblage with orthopyroxene (X_Mg = 0·44-0·16), subordinate clinopyroxene (X_Mg = 0·58-0·26), garnet (X_Mg = 0·20-0·06), plagioclase (An_37-30), quartz and ilmenite. The former large grains of plagioclase and K-feldspar have been completely recrystallized to polygonal mosaics, whereas quartz forms large xenoblastic to platy grains with subgrain structure and undulose extinction. As in the ferrodiorites, garnet occurs exclusively as coronal phase girdling feldspar aggregates where in contact with the pyroxenes. Accessory phases are zircon, apatite, metamict thorite, pyrrhotite, chalcopyrite and pentlandite.

The low-variance assemblages of the ferrodiorite suite as well as the extremely ferrous compositions of the phases have allowed us to constrain with reliance the P-T-X_fluid conditions of the coronitic reaction stage of high-grade metamorphism. The results of the thermodynamic evaluation obtained for 14 fayalite-bearingferrodiorite samples using a multi-reaction approach (Berman, 1988, , 1991) indicate remarkably consistent conditions of chemical equilibration, i.e. 750-800°C, 7-8 kbar, a(H₂O) of 0·15-0·25 and logf(O₂) of -18 bar. The P-T estimates together with the reaction textures suggest a mid-crustal emplacement of the ferrodioritic melts (and the anorthosite pluton) followed by near-isobaric cooling from the igneous (~1100°C) to the regional thermal regime. A detailed presentation of the thermobarometric evaluation will be given in a separate paper.

Border zone granites

The felsic rocks bordering the massif constitute a homogeneous suite of coarse-grained garnetiferous gneisses, broadly granitic in composition. Minerals present in varying proportion are megacrystic K-feldspar, quartz, plagioclase, porphyroblastic garnet, orthopyroxene, clinopyroxene, biotite, ilmenite, apatite and zircon. The streaky gneissose fabric of the rocks, defined by drawn-out aggregates of ferromagnesian silicates, quartz lenticles and K-feldspar augen, is overprinted, but not obliterated by a granoblastic annealing texture. Significant in this context is the nature of occurrence of garnet, which forms pre- to syn-kinematic porphyroblasts that are occasionally flattened close to the massif. The garnets commonly enclose subrounded grains of quartz, K-feldspar, biotite, plagioclase, ilmenite, zircon and apatite. The poikiloblastic nature of the garnets and their inclusion paragenesis suggest that they have formed through incongruent biotite melting reactions in metagraywacke protoliths, where they became entrained in the segregating felsic melts (e.g. Vielzeuf & Montel, 1994; Patiño-Douce & Beard, 1995, , 1996).

Analytical methods

Major and trace element concentrations of 93 rock samples comprising 39 anorthosites-leuconorites, 2ultramafic xenoliths, 3 high-Ti ferrodioritic rocks, 24ferrodiorites, 7 ferromonzodiorites, 11 garnetiferous granitoid gneisses bordering the massif, and 7 migmatitic country gneisses were determined by X-ray fluorescence (XRF) spectrometry on fused glass discs using a Philips-PW 1480 model. Peak overlap and matrix effects were corrected using the software package OXIQUANT (Vogel & Kuipers, 1987). Accuracy strongly depends on the concentration; for major elements the relative accuracy is generally better than 2% and for the analysed trace elements <3%, and much less for trace elements in the concentration range <10 ppm. Selected samples were analysed for rare earth elements (REE) by instrumental neutron activation analysis (INAA) (Bonn) and inductively coupled plasma atomic emission spectrometry (ICP-AES) (Marburg). REE patterns were normalized to chondrite values after Taylor & McLennan, (1985). The data set is supplemented by whole-rock oxygen isotope analyses of several samples from each rock group. O-isotope compositions were analysed on 8-10 mg aliquots of whole-rock powders, using a modified Clayton & Mayeda, (1963) type extraction line. Purified F₂ gas (Asprey, 1976) was used to liberate the oxygen, which was then converted to CO₂ and measured on a SIRA-9 mass spectrometer (VG Instruments). All analyses were replicated at least two times; the reported [delta]¹⁸O values represent mean values with 1[sgr] deviations of ±0·1%°. Tables 1 and 2 provide whole-rock analytical data of representative samples from each suite.

Table 1. Major and trace element chemistry and oxygen isotope data of representative samples from the Bolangir anorthosite complex and associated rock suites

Table 2. Rare earth element data for representative samples from the Bolangir anorthosite complex and associated rock suites

Anorthosite-leuconorite suite

The anorthosite-leuconorite suite is characterized by intermediate mg-numbers (30-65) and low K₂O (0·60-0·90 wt %). Within the suite TiO₂, FeO, MnO and MgO increase from anorthosite to leuconorite, whereas Al₂O₃, CaO and Na₂O decrease (Fig. 3). CaO/(CaO + Na₂O + K₂O) ratios fall in a narrow range (0·71-0·76) (Fig. 4a), consistent with the uniform plagioclase composition throughout the massif. The rocks are depleted in all plagioclase-incompatible elements, whereas the abundances of Sr (430-580 ppm) and Ba (190-310 ppm) are high (Fig. 5). The REE patterns are typical for massif-type anorthosites [ Ashwal, (1993) and references therein], i.e. REE abundances 0·5-20 times the chondritic values, with (La/Lu)_N in the range of 35-5 and marked positive Eu anomalies (Eu* = 14·2-5·0) (Fig. 6). Abundances of the REE increase from anorthosite to leuconorite, whereas Eu* and (La/Lu)_N decrease with decreasing Al₂O₃. [delta]¹⁸O values cluster between 6·7 and 7·5%° (Fig. 7).

Figure 3. Major element vs Al₂O₃ diagrams for the Bolangir anorthosite complex and the associated rock suites. Symbols: open squares, anorthosites; open squares with cross, leuconorites; filled squares with white cross, low-Zr ferrodiorites; filled diamonds, fayalite-bearing ferrodiorites; filled diamonds with white cross, ferrodiorites with entrained anorthositic plagioclase xenocrysts; filled circles, silicic ferrodiorites; centred circles, ferromonzodiorites; stars, ultrabasic xenoliths; open crosses, border zone granites; crosses, country gneisses.

Figure 4. Bivariate diagrams of CaO/(CaO + Na₂O + K₂O) and mg-number vs FeO* + MnO + MgO + TiO₂ illustrating the compositional trends and relationships of leuconorites-anorthosites, ferrodiorites, ferromonzodiorites and border zone granites at Bolangir. The compositional variation of the ferrodiorite suite in terms of the femic components is caused by progressive accumulation of plagioclase towards the silicic members (a), and therefore keeps the mg-number unaffected (b). The ferromonzodiorites evolved from the plagioclase-laden ferrodioritic melt through hybridization with a felsic melt component (the border zone granites) (a). (For further discussion see text.) Symbols as in Fig. 3; Plag I, composition of plagioclase phenocrysts in the ferrodiorite suite.

Figure 5. Trace element vs Al₂O₃ diagrams for the Bolangir anorthosite complex and the associated rock suites. Symbols as in Fig. 3.

Figure 6. Chondrite-normalized REE patterns for representative rock samples of the Bolangir anorthosite complex and the associated rock suites. The REE are normalized to the average values of C1 chondrites suggested by Taylor & McLennan, (1985). The analytical data are given in Table 2. Also shown are bivariate plots of Ce_N, Yb_N and Eu_N vs the sum of femic components (symbols as in Fig. 3).

Figure 7. Whole-rock oxygen isotope data for representative samples of the Bolangir anorthosite complex and associated rock suites. Symbols as in Fig. 3.

Melanocratic vein-like domains within the massif are low-Zr ferrodiorites. Although indistinguishable in the field from the border zone high-Zr ferrodiorites, low-Zr ferrodiorites differ significantly in their higher contents of SiO₂, Al₂O₃, TiO₂, MgO, CaO and P₂O₅ and lower abundances of FeO, MnO and high field strength elements (HFSE) (Table 1; sample 2111B). The chemical features closely resemble members of the Labrieville and St Urban oxide-apatite-rich gabbronorites (OAGNs) (Owens & Dymek, 1992), Fe-Ti-P-rich (FTP) rocks from the Adirondack Mountains (McLelland et al., 1994) and the oxide-rich ferrodiorites from the Laramie Anorthosite Complex (Kolker et al., 1990; Mitchell et al., 1996). Whereas Owens & Dymek, (1992) interpreted OAGNs to be derived through fractionation from a mangeritic magma, both McLelland et al., (1994) and Mitchell et al., (1996) considered the rocks to be residual melts of anorthosite crystallization. The mode of occurrence of low-HFSE ferrodiorites at Bolangir would be in line with the latter interpretation.

The extremely rare ultrabasic lenses within the leuconoritic marginal zone of the massif have SiO₂ contents varying from 44 to 46 wt % and mg-numbers between 65 and 75. Samples with low SiO₂ are olivine bearing. Cr₂O₃ and Ni contents are high (1500-1800 ppm and 490-700 ppm, respectively), whereas K₂O, Ba, Rb and Sr contents are low (Table 1, Figs 3 and 4). The rocks are characterized by flat REE patterns showing weakly positive to weakly negative Eu anomalies (Fig. 5). [delta]¹⁸O values are low (5·5%°).

Ferrodiorite suite

The ferrodiorites constitute a suite of extremely iron-enriched fayalite-bearing to more silicic fayalite-free rock types within the SiO₂ range from 37 to 58 wt % and a complementary FeO* range from 39 to 10 wt %. In comparison with world-wide occurrences of chemically similar rocks associated with massif-type anorthosites (Ashwal & Seifert, 1980; Goldberg, 1984; Hill, 1988; Duchesne et al., 1989; Kolker et al., 1990; Owens & Dymek, 1992; McLelland et al., 1994; Mitchell et al., 1996), the Bolangir ferrodiorites have exceptionally high abundances of HFSE (Zr 5500-1400 ppm; Y 240-80 ppm; Sc 80-20 ppm; Nb 280-70 ppm; Th 190-65 ppm; Y 240-75 ppm) and the trivalent REE (La 480-220 ppm; Ce 1100-550 ppm; Yb 22-10 ppm) (Tables 1 and 2; Fig. 7a). To this extent the rocks have no known terrestrial (or lunar) equivalents.

Variation diagrams for the ferrodiorite suite show strongly correlated linear trends for major and trace elements (Figs 3 and 5). Elements in the fine-grained melanocratic matrix (the former melt phase) such as FeO* (39-10 wt %), MnO (0·6-0·15 wt %), Zn (500-140 ppm), TiO₂ (4·6-1·0 wt %), HFSE and REE decrease with increasing alumina content, whereas elements hosted in the plagioclase phenocrysts such as Al₂O₃ (7·9-18·0 wt %), Na₂O (1·0-3·5 wt %), CaO (4·7-7·0 wt %) and Sr (110-360 ppm) increase. The linear trends on the variation diagrams, in line with the porphyritic structure of the rocks, may therefore be attributed to the observed progressive accumulation of plagioclase phenocrysts towards the more silicic members of the suite. Further support for such an interpretation is provided by a bivariate plot of (FeO* + MnO + MgO + TiO₂) vs CaO/(CaO + Na₂O + K₂O), in which the two variables represent the melanocratic melt component and the Ca component of feldspar, respectively (Fig. 4a). The majority of ferrodiorite samples define a linear array that points to the average composition of phenocrystic plagioclase. Two ferrodiorite samples (K110 and B684) that contain anorthositic plagioclase xenocrysts, however, fall on a mixing-line which connects the compositional fields of fayalite-bearing ferrodiorites and anorthosites-leuconorites. The progressive, presumably gravitational accumulation of plagioclase towards the silicic members of the ferrodiorite suite is also reflected by the REE data: whereas Eu that resides in the plagioclase phenocrysts increases, the abundances of the trivalent REE decrease with decreasing melanocratic fraction (Fig. 6). Further, the almost constant mg-number (12-16), throughout the ferrodiorite suite suggests that fractionation of femic phases was insignificant (Fig. 4b). Yet, as a few samples with higher mg-number (20-26) contain recrystallized orthopyroxene aggregates replacing large phenocrysts, fractionation of this femic phase cannot be ruled out completely.

In comparison with the Bolangir anorthosites, the abundances of all elements except the plagioclase-compatible elements such as CaO, Al₂O₃, Na₂O, K₂O, Sr and Ba are markedly higher in the ferrodiorites (Fig. 8b). Significantly, the chondrite-normalized REE spectra of the ferrodiorites exhibit pronounced negative Eu anomalies that complement the positive Eu anomalies in the REE spectra of anorthosites and leuconorites (Fig. 6). It is obvious that the processes that gave rise to the ferrodiorite melts must have involved extensive plagioclase fractionation. The fact that ferrodiorites of this kind always occur in intimate spatial association with massif-type anorthosites strongly argues for a close genetic relationship, i.e. a derivation of the ferrodiorite melts by continuous anorthosite crystallization. Further, mineralogical features such as the sharp increase of Fe/(Fe + Mg) of ferromagnesian phases and the marked decrease of An content in plagioclase from anorthosites-leuconorites to ferrodiorites (Fig. 2b) would be compatible with a derivation of the ferrodiorite melts by continuous anorthosite crystallization.

Figure 8. Spider diagrams illustrating (a) the compositional ranges of Bolangir high-Zr ferrodiorites and anorthosite-associated ferrodiorites from world-wide occurrences and (b) the complementary chemistry of ferrodiorites and anorthosites-leuconorites at Bolangir.

The enrichment of the high-Zr ferrodiorites in ¹⁸O by ~1%° relative to the anorthosites-leuconorites (Fig. 7), however, is incompatible with plagioclase fractionation alone, as the isotopic fractionation between basic melt and plagioclase is slightly negative at T < 1400°C (Kyser et al., 1982; Kyser, 1986) and thus would drive the isotopic composition of the residual ferrodioritic melt to lower [delta]¹⁸O values. If the observed positive shift resulted from fractional crystallization processes in the ascending anorthosite diapir, the bulk isotopic fractionation factor must have been positive, implying co-crystallization of phases with positive melt-crystal fractionation factors (pyroxenes and especially ilmenite). The major contribution thus would come from the precipitation of ilmenite, a phase which resembles magnetite in the fractionation behaviour. The largest isotopic fractionation in basaltic magmas occurs between plagioclase and magnetite, displaying [delta]¹⁸O values of 2-2·5%° (Hoernes & Friedrichsen, 1977; Taylor & Sheppard, 1986). The intimate association of massive Fe-Ti oxide bodies with ferrodiorites in many massif-type anorthosite complexes gives evidence for such a fractionation process [ Ashwal, (1993) and references therein]. In this context it is important to note that `vein-like' bodies of low-Zr ferrodiorites within the anorthosite massif exhibit similarly low or even lower [delta]¹⁸O values compared with their anorthosite host. This suggests that these rocks represent ilmenite- and apatite-rich accumulations, rather than residual liquids and thus could account for the observed enrichment of ¹⁸O in the high-Zr ferrodiorites. In addition, the high-Zr ferrodiorites, following their segregation and extraction from the ascending anorthosite diapir, might have experienced isotopic exchange with the adjacent felsic crustal melts. Considering the large surface relative to the small volume of the ferrodioritic sheets, such a process might be especially effective.

Ferromonzodiorites

The transitional nature of the ferromonzodiorites observed in the field is also reflected in the elemental variation diagrams (Figs 3 and 5). Except for significantly higher contents of K₂O, Rb and Ba, the ferromonzodiorites share the major and trace element characteristics of the most silicic ferrodiorites. The REE spectra, on the other hand, resemble those of the border zone granites (Fig. 6). Compared with the ferrodiorites, the REE spectra are less fractionated and lack a Eu anomaly. In the CaO/(CaO + Na₂O + K₂O) vs (FeO* + MnO + MgO + TiO₂) diagram (Fig. 4a) the ferromonzodiorite samples plot along a linear array that links the silicic ferrodiorites with the spatially associated granites. We interpret the ferromonzodiorites to have formed by local hybridization of the ferrodioritic melts with the felsic magmas bordering the massif. The large spread in [delta]¹⁸O values from 7·5 to 9%° (Fig. 7) is compatible with this interpretation.

Border zone granites

The garnetiferous granite gneisses are peraluminous [Al₂O₃/(CaO + Na₂O + K₂O) = 1-1·05] and characterized by high contents of K₂O (4·4-7·5 wt %) and Ba (1500-2900 ppm) as well as high K₂O/Na₂O ratios (2·0-3·5) (Table 1; Figs 3 and 5). Rare earth elements show LREE-enriched, but weakly fractionated (La_N/Yb_N = 20-25) patterns with small positive to no Eu anomalies (Fig. 6). The [delta]¹⁸O values (7·5-9·2%°) are typical for crustally derived granitic melts (Fig. 7).

Petrographically and chemically, the granite gneisses resemble megacrystic K-feldspar granites that occur independent of anorthosite massifs as numerous plutonic bodies in the central and eastern domains of the Eastern Ghats Belt (Mukhopadhyay & Bhattacharya, 1997). Mukhopadhyay, (1995) interpreted one such granitoid body at Salur to have formed by incomplete separation of restite phases (garnet, plagioclase, orthopyroxene and quartz) in melts derived by high degrees of incongruent melting of biotite-rich metagreywacke protoliths. A crustal origin of the melts through extensive anatexis of psammitic to pelitic metasediments is clearly indicated by the Nd-Sr isotope systematics of several porphyritic granitoid complexes (Krause et al., 1996; O. Krause, personal communication, 1997). Furthermore, the distinctive chemistry of the granites bordering the massif and the quartzofeldspathic gneisses farther away (Figs 3 and 5), the sharp physical boundary between the two suites, and the absence of metasedimentary gneiss and mafic granulite enclaves (similar to those in the quartzofeldspathic gneisses) in the granitoids strongly suggest that the granitic magmas were not in situ melts, but brought up from deeper crustal zones.

Ferrodiorites: immiscible melts or residual melts?

Liquid immiscibility between Fe-rich basic melts and K-rich silicic melts has been shown to occur in synthetic systems (Roedder, 1956, , 1979; McBirney, 1975; Watson, 1976; Philpotts, 1981) as well as in natural systems (Roedder & Weiblen, 1970; Philpotts, 1982). Freestone, (1978) demonstrated that the addition of TiO₂ (3 wt %) and P₂O₅ (1 wt %) in the silica-leucite-fayalite pseudoternary system causes the two-liquid field to expand towards the leucite apex (Fig. 9). The compositions of the Bolangir ferrodiorites and border zone granites plot in two distinct fields that lie on the two-liquid solvus of Freestone, (1978). By implication, the ferrodiorites and granites may be regarded as melts related through unmixing. However, the following arguments contradict such an interpretation. First, Watson, (1976) and Ryerson & Hess, (1978) demonstrated that Ca, Ba and Sr among other elements are partitioned into the Fe-rich melt. By contrast, the abundances of these elements are lower in the ferrodiorites (Table 1). Second, experimental work of Ellison & Hess, (1991) showed that REE are preferentially partitioned into the Fe-rich liquid, and two-liquid partition coefficients for the REE are similar at a given temperature. Although the abundances of trivalent REE are higher in the ferrodiorites than in the granites in accordance with the experimental results, Eu abundances are higher in the granites (Fig. 6). Third, the garnet in the ferrodiorites is exclusively coronitic and therefore of metamorphic origin, whereas porphyroblastic garnet in the granites is a product of the incongruent melting process that generated the granitic melt. The absence of porphyroblastic garnet in ferrodiorites is difficult to explain, if the two melts shared a common parentage before liquid unmixing and the ferrodiorites formed by segregation of the denser Fe-rich melt. In view of the above arguments, the close match of the compositional fields for ferrodiorites and border zone granites with the two-liquid solvus appears to be fortuitous. Instead, it is reasonable to assume that the ferrodiorites are melts residual to extensive plagioclase fractionation in basic magmas. Such a process should cause the residual melts to become enriched in all plagioclase incompatible elements. This is evidenced by the complementary compositional nature of the anorthosite-leuconorite cumulates and the ferrodiorite melts (Fig. 8b).

Figure 9. (FeO* + MgO + TiO₂ + P₂O₅) - (Al₂O₃ + Na₂O + K₂O) - SiO₂ ternary diagram showing the compositions of anorthosites-leuconorites, ferrodiorites, ferromonzodiorites and border zone granites of the Bolangir anorthosite complex. Also shown is the two-liquid field from the system fayalite-leucite-silica (Roedder, 1951) and its expansion towards the leucite apex resulting from the addition of 3 wt % TiO₂ and 1 wt % P₂O₅ (Freestone, 1978).

The problem of high concentrations of HFSE and REE in the ferrodiorites

The geochemical attributes of the anorthosite complex (anorthosite-leuconorite-ferrodiorite suite) at Bolangir, such as the low overall abundances of large ion lithophile elements (LILE), HFSE and transition elements, the intermediate ranges of mg-number and SiO₂, and the moderately fractionated REE patterns, suggest that the magma parental to the suite was a fractionated basic melt. Following the experimental work of Fram & Longhi, (1992), the presence of plagioclase of intermediate composition and the possible existence of aluminousorthopyroxene in the anorthosite indicate liquidus temperatures in excess of 1200°C at pressures of 10-12 kbar. The pressure range is consistent with the estimates obtained from the Mg-olivine + quartz Mg-orthopyroxene equilibrium for the most magnesian sample (ultrabasic xenolith B 502B) using the thermodynamic data base of Berman, (1988, , 1991). The estimated pressure, 8·7 kbar at 1000°C, is a minimum considering quartz is absent in the assemblage. These P-T values rule out a crustal origin for the anorthosite parental magma. The occurrence of ultrabasic xenoliths ([delta]¹⁸O = 5·4%°), although very rare, within the leuconoritic outer part of the anorthosite massif would also point to a mantle origin. However, the average [delta]¹⁸O value of 7·1%° for the anorthosites would argue for some crustal contamination of the parental magma or an ¹⁸O enrichment in the subcontinental mantle (Harmon & Hoefs, 1995; Hoefs, 1997) where the parental magma was formed. In conclusion, we consider that the parental magma to the anorthosite suite resembled high-alumina basalt in composition (see Fram & Longhi, 1992; Ashwal, 1993; Mitchell et al., 1995).

To explain the unusual chemistry of the ferrodiorites, i.e. their high abundances of HFSE and REE, it is necessary to constrain the possible process of magmatic differentiation. Three mechanisms may be envisaged: (1) closed-system fractionation of high-alumina basalt; (2) closed-system fractionation of high-alumina basalt contaminated by crustal sources either by assimilation or magma mixing before anorthosite crystallization; (3) closed-system fractionation of high-alumina basalt followed by crustal contamination of the anorthosite residual melts.

To test which of the mechanisms could produceresidual liquids that chemically resemble the Bolangirferrodiorites, we have performed a series of fractional crystallization calculations using the element pairs Sr (compatible) vs Zr (incompatible), La vs Eu (both incompatible) and Ce vs (Ce/Yb), and mineral-melt partition coefficients of these elements for the relevant mineral phases in basic melts (Simmons & Hanson, 1978; Henderson, 1986). The initial element concentrations in the parental magma were approximated to the average element abundances of high-alumina gabbros reported from the Laramie Complex [ Mitchell et al., (1996); Sr 463 ± 35 ppm, Zr 37 ± 22 ppm, La_CN 18·8 ± 10, Ce_CN 15·6 ± 9, Eu_CN 13·9 ± 4, Yb_CN 3·10 ± 1·8]. The approximation was necessary in the absence of chemically similar rocks from the anorthosite massifs in the Eastern Ghats Belt.

Here we refer only to the results of one of the schemes that leads to maximum enrichment of HFSE and REE in the computed residual melts. The scheme is as follows. As a first step, the element abundances in the basic melt in equilibrium with the assumed solid assemblage A (Pl 93%, Opx 3%, Cpx 1%, Bt 1%, Grt 0·5%, Ilm 1%, Apt 0·5%) are computed for equilibrium crystallization in the range from 5 to 98%. In the next step, the melt composition at 70% crystallization is taken as the starting melt composition, and concentrations of the relevant elements in the melt (up to 98% crystallization) in equilibrium with a solid assemblage modified to more basic (leuconorite) composition (assemblage B: Pl 87%, Opx 6%, Cpx 2%, Bt 2%, Grt 0·5%, Ilm 2%, Apt 0·5%) are computed. This exercise is repeated by taking the melts at 45% and 35% crystallization as starting compositions for the next set of calculations and by further modifying the solid assemblage to progressively more basic composition (assemblage C: Pl 74%, Opx 13%, Cpx 5%, Bt 3%, Grt 0·8%, Ilm 3·4%, Apt 0·8%; assemblage D: Pl 66%, Opx 18%, Cpx 6%, Bt 4%, Grt 1%, Ilm 4%, Apt 1%). The modal mineralogies of assumed solids closely resemble the mineral assemblages in anorthosites and leuconorites in the massif. Discrete garnet grains that occur exclusive of ferromagnesian phases are tentatively taken to be part of the igneous assemblage. Neglecting the small modal amount of garnet in the computations causes the coexisting melts to become marginally enriched in Zr and Ce/Yb, especially at high degree of fractionation, but the deviations are unlikely to change the inferences made below.

The computed fractionation trends conform with the observed elemental variations in the anorthosite-leuconorite suite and the ferrodiorites. The absolute element abundances of Zr and REE in the ferrodiorites, however, are approached only at a very high degree of crystallization (>98%) of the parental magma. But at such high degree of fractionation, the computed Zr and REE abundances in the anorthosite-leuconorite are significantly different from the measured abundances of these elements. It thus appears that crustal sources are likely to have contaminated either the parental magma before or during anorthosite formation, or the residual Fe-rich melts subsequent to their extraction, or both. Strong evidence for crustal contamination is the exceptionally high Th/U ratios (up to 200) of the ferrodiorites. Such ratios would point to assimilation of granulite-grade supracrustal rocks that are typically depleted in U, but carry monazite as an important accessory phase. In the Laramie anorthosite complex, Mitchell et al., (1996) have demonstrated from isotope systematics that the late stage Fe-enriched residual melts extracted from the anorthosite have variably been contaminated via crustal contamination. However, those workers did not preclude the possibility of crustal contamination of the parental magma.

The effects of crustal contamination of parental magma on ferrodiorite chemistry can be estimated by comparing the element abundances in the assumed high-alumina gabbro with lower-crustal rocks. Taking tonalite gneisses as lower-crustal analogues (Wedepohl et al., 1991; La_CN 60, Ce_CN 45, Eu_CN 11, Yb_CN 4·3; Sr 375 ppm, Zr 137 ppm) it is observed that except for Eu and HREE, which are comparable, LREE and Zr abundances are lower in high-alumina gabbros, whereas Sr is more enriched. Clearly, bulk crustal contamination of the anorthosite parental magma will cause the residual melts to become more enriched in REE (except Eu) and Zr with respect to the computed abundances using uncontaminated parental magma. But at high degree of fractionation necessary to attain the measured abundances in the ferrodiorites, the computed REE and Zr contents in the coexisting anorthosite-leuconorite solids will be drawn well outside the compositional fields limited by measured abundances. The effect will be more pronounced if the magma fractionates by a Rayleigh process. Clearly, fractionation of high-alumina basalts crustally contaminated at the source or fractionation with continuous crustal contamination cannot explain the Zr and REE chemistry of anorthosite-leuconorite solids as well as ferrodiorite melts. Alternatively, the paradox can be eliminated by decoupling the residual melts from the solids in equilibrium at an advanced stage of fractionation, i.e. till the computed concentrations of elements in the anorthosite-leuconorite solids are matching the measured concentrations. Following their extraction, the residual melts may then evolve by crustal contamination.

Considering a late-stage contamination of the ferrodioritic melts, the low concentrations of SiO₂, K₂O, Rb and Ba and the unusually high abundances of HFSE and REE cannot be explained by bulk crustal assimilation. For the same reasons, magma mixing with the border zone granitic melts can be ruled out. On the other hand, the two melts in physical contact, initially in chemical disequilibrium, could undergo diffusion-controlled element exchange until the melt compositions approach those of immiscible melts (Freestone, 1978; Roedder, 1979; Philpotts, 1982) (Fig. 9). In fact, the element ratios of the two melts are comparable with experimental two-liquid partitioning coefficients (Watson, 1976). Diffusional exchange could also account for the observed enrichment of ¹⁸O in the extracted ferrodioritic melts relative to the parent anorthosite.

As an alternative explanation, selective assimilation of accessory phases from crustal sources that preferentially host HFSE and REE may have caused the unusually high abundances of these elements in the ferrodiorites. The most likely accessory phases are zircon, monazite and apatite. These phases, being heavier than the felsic melts, could have been incorporated into the Fe-enriched residual melts through gravitational settling from the overlying granitic magma. Because of the extreme REE and HFSE abundances in the accessory phases, only very small amounts of assimilation would cause large spikes of these elements in the ferrodioritic melts. The effectiveness of such a process clearly depends on the viscosity of the granitic melt. The settling velocities of accessory phases (zircon, monazite) in highly viscous granitic melts could be assessed by using Stoke's law. Assuming a melt viscosity [eta] of 4 log units and a density of the melt of 2·4 g/cm³ (Johannes & Holtz, 1996), model calculations yield settling velocities for zircon (r = 250 µm; [rho] = 4·7 g/cm³) and monazite (r = 500 µm; [rho] = 5·1 g/cm³) of 0·98 m/yr and 4·39 m/yr, respectively. Although these settling velocities are significantly lower than those reported for silicate and oxide phases in basaltic systems (Wager & Brown, 1967), it still appears possible that monazite, and to a lesser extent, zircon grains could have sunk from the overlying viscous and less dense felsic melt into the less viscous and dense ferrodioritic melt. In as much as Fe-rich residual melts are typically strongly undersaturated with respect to Zr, P and REE (Dickinson & Hess, 1982, , 1985; Watson & Harrison, 1983, , 1984), the entrained accessory phases are likely to dissolve, causing the melts to become spiked in Zr, P and zircon, monazite and apatite hosted trace elements such as REE, Y, Th and U. With regard to Zr, for instance, all ferrodiorite samples show Zr concentrations that are close to or below the Zr saturation limits, as experimentally determined by Watson & Harrison, (1983, , 1984) in melts of comparable major element chemistry at a temperature of 1000°C; that is, presumably below the liquidus temperature of these rocks. In accordance with this, zircon grains in the ferrodiorites typically show magmatic morphology and internal growth zonation, and only very rarely contain inherited cores.

Ferromonzodiorites: products of magma mixing?

Field relations and petrographic features suggest that the ferromonzodiorites developed from the plagioclase-enriched members of the ferrodiorite suite through hybridization with a felsic counterpart. Magma mixing with the overlying granite melt should cause the hybrid melts to fall on a mixing line joining the compositional fields of the Fe-rich and the felsic magmas. The variation diagrams, however, show that such relations are distinct only for K₂O, Rb and Ba (Figs 3 and 4), which argues for a dominance of the alkali feldspar component in this hybridization process. Radiogenic isotope data are needed to further clarify the nature of the ferromonzodiorites.