Journal of Petrology Advance Access originally published online on February 7, 2006
Journal of Petrology 2006 47(5):901-927; doi:10.1093/petrology/egi100
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Mg–Al Sapphirine- and Ca–Al Hibonite-bearing Granulite Xenoliths from the Chyulu Hills Volcanic Field, Kenya
UNIVERSITY OF NEUCHÂTEL, INSTITUTE OF GEOLOGY, RUE EMILE-ARGAND 11, CH-2007 NEUCHÂTEL, SWITZERLAND
RECEIVED APRIL 1, 2004; ACCEPTED DECEMBER 19, 2005
 |
ABSTRACT |
|---|
TOP
ABSTRACT
INTRODUCTIONBasanites of the Chyulu Hills (Kenya Rift) contain mafic Mg–Al and Ca–Al granulite xenoliths. Their protoliths are interpreted as troctolitic cumulates; however, the original mineral assemblages were almost completely transformed by subsolidus reactions. Mg–Al granulites contain the minerals spinel, sapphirine, sillimanite, plagioclase, corundum, clinopyroxene, orthopyroxene and garnet, whereas Ca–Al granulites are characterized by hibonite, spinel, sapphirine, mullite, sillimanite, plagioclase, quartz, clinopyroxene, corundum, and garnet. In the Mg–Al granulites, the first generation of orthopyroxene and some spinel may be of igneous origin. In the Ca–Al granulites, hibonite (and possibly some spinel) are the earliest, possibly igneous, minerals in the crystallization sequence. Most pyroxene, spinel and corundum in Mg–Al and Ca–Al granulites formed by subsolidus reactions. The qualitative P–T path derived from metamorphic reactions corresponds to subsolidus cooling, probably accompanied, or followed by, compression. Final equilibration was achieved at T
600–740°C and P <8 kbar, in the stability field of sillimanite. The early coexistence of corundum and pyroxenes (± spinel), as well as the association of sillimanite and sapphirine with clinopyroxene and the presence of hibonite, makes both types of granulite rare. The Ca–Al hibonite-bearing granulites are unique. Both types enlarge the spectrum of known Ca–Al–Mg-rich granulites worldwide. KEY WORDS: granulite xenoliths; corundum; sapphirine; hibonite; Kenya Rift
|
INTRODUCTION |
|---|
The study of granulites provides important constraints on the nature of high-temperature to ultrahigh-temperature metamorphism in the Earth's crust. On the one hand, the chemical and isotopic compositions of granulites are useful tools for deciphering their origin and geodynamic setting (e.g. Jan & Howie, 1981
; Downes, 1993; Kempton et al., 2001). On the other hand, reaction textures and the P–T paths of granulites serve to unravel metamorphic processes and their causes (e.g. Bohlen, 1987, 1991; Harley, 1989; Kriegsman & Schumacher, 1999; Rickers et al., 2001). The study of granulites from metamorphic terranes often benefits from the incorporation of both approaches (e.g. Möller et al., 1998; Osanai et al., 1998; Bhattacharya & Kar, 2002). Granulites are also a major tool for investigating the composition of the Earth's lower crust. Granulite xenoliths from volcanic rocks are traditionally considered in this context (e.g. Rudnick et al., 1986; Rudnick, 1992; Downes, 1993; Huang et al., 1995; Kempton et al., 2001), with far less attention being paid to their metamorphic reaction paths (Christy, 1989; Dawson et al., 1997). The overwhelming dominance of the geochemical approach to the study of granulitic xenoliths is partly justified, as a given section of the Earth's middle to lower crust, sampled by an ascending magma, may include rocks of different provenance, age and metamorphic evolution, juxtaposed by various tectonic processes (e.g. Downes et al., 1990; Kempton et al., 1990). The studied granulite xenoliths from the Chyulu Hills, East African Rift, can yield information on both issues, the chemical composition of the Earth's lower crust, in this case that beneath the East African Rift, and the nature and conditions of metamorphic processes that affected it. Recent major and trace element studies on minerals and whole-rock samples have shown that the Mg–Al granulites, Ca–Al granulites and garnet–spinel websterite xenoliths of the Chyulu Hills represent a series of gabbroic to troctolitic cumulates that seems to be related to arc magmatism (Ulianov et al., in preparation; see the section on ‘General chemical and petrographic features’). Combined with similarities to certain granulite terranes in Tanzania (see Discussion) this geochemical signature suggests that the xenoliths, and hence the crust and upper mantle beneath the Chyulu Hills, are of Pan-African age.
The igneous minerals in the protoliths of the Mg–Al and Ca–Al granulite xenoliths have, to a large degree, been replaced by subsolidus assemblages that provide important information on the P–T path and on phase relations in mafic Ca–Mg–Al-rich bulk compositions. Meta-igneous granulitic rocks rich in Mg, Al and Ca are rare. They tend to crystallize sapphirine and clinopyroxene, and apparently represent a special paragenetic family of granulites that was recognized by Christy (1989). This includes xenoliths from Stockdale (Meyer & Brookins, 1976), Delegate (Griffin & O'Reilly, 1986) and the Kerguelen Archipelago (Grégoire et al., 1998), and granulites from the metamorphic terranes of Finero (Lensch, 1971; Sills et al., 1983), Roan (Johansson & Möller, 1986), Dunkelsteinerwald (Carswell et al., 1989), the Central Gneiss Belt of the Grenville Province (Grant, 1989), Ullared–Gällared (Möller, 1999) and several other, less well-characterized terranes.
The purpose of this paper is (1) to contribute to the general understanding of mineral assemblages and reactions in Ca–Mg–Al-rich mafic igneous protoliths under granulite-facies conditions and (2) to decipher the reaction paths of the studied granulites and discuss their possible geological significance. For this purpose, mineral assemblages, mineral compositions and reaction sequences are carefully described, P–T conditions are quantified when possible, and a qualitative P–T path is inferred.
|
GEOLOGICAL SETTING |
|---|
The Chyulu Hills volcanic field is situated on the eastern flank of the Kenya rift some 150–200 km east of the rift axis, about 50 km NE of Mt. Kilimanjaro (Fig. 1). It consists of several hundred volcanic cones and lava flows forming a NW–SE-aligned chain
100 km long. For the last decade, the Chyulu Hills have been subject to a number of detailed petrological (focused on xenoliths and alkaline volcanic rocks) and geophysical (gravity and seismological) studies (e.g. Henjes-Kunst & Altherr, 1992; Novak et al., 1997; Ritter & Kaspar, 1997; Späth et al., 2001). The seismic Moho has been detected at a depth of 44 km (the divide is not sharp) with a thickness of 20 km for the lower crust (Ritter & Kaspar, 1997). The lithosphere–asthenosphere boundary lies at 107–120 km depth as suggested by the geobarometric estimates for high-P porphyroclastic garnet lherzolite xenoliths from the Chyulu volcanic rocks (Henjes-Kunst & Altherr, 1992; Altherr et al., in preparation).
|
Volcanic activity within the Chyulu Hills is very young. The compositional spectrum of rocks ranges from foidites through less silica-undersaturated basanites to fractionated alkali basalts (Haug & Strecker, 1995
; Späth et al., 2001). Foidites are the oldest (1·4 Ma), whereas some of the alkali basalts were erupted in historical times (Haug & Strecker, 1995). Despite its young volcanicity, there are no indications for a recent asthenospheric upwelling or upwarping of the Moho beneath the Chyulu Hills volcanic field (Novak et al., 1997; Ritter & Kaspar, 1997). The xenoliths used for this study were sampled from basanites of the Kwa Nthuku volcano in the northwestern part of the Chyulu Hills.
The Chyulu Hills are located within the Pan-African Mozambique mobile belt (Holmes, 1951) extending from the Arabian–Nubian Shield in the north via Ethiopia, Uganda, Kenya and Tanzania to Mozambique, Malawi and Madagascar in the south. The Mozambique belt includes several metamorphic domains with different ages and P–T histories (e.g. Shackleton, 1973; Key et al., 1989; Stern & Dawoud, 1991; Pinna et al., 1993; Appel et al., 1998; Möller et al., 1998, and references therein). It resulted from a complex series of subduction, collision and accretion events between approximately 750 and 615 Ma (e.g. Coolen et al., 1982; Kröner et al., 1987; Stern & Dawoud, 1991; Muhongo & Lenoir, 1994; Möller et al., 1998, 2000; Muhongo et al., 2001; Kröner et al., 2003; Meert, 2003), related to the closure of the Mozambique ocean. The younger events within this time span are commonly termed ‘Pan-African’ (e.g. Möller et al., 2000), whereas the older ones are often related to an ‘East African orogen’ (e.g. Stern, 1994; Meert, 2003). The older history of the Mozambique belt seems to vary with geographical position. Magmatic and metamorphic events have been dated at 800–820 Ma (e.g. Key et al., 1989; Kebede et al., 2001), at approximately 1000–1250 Ma (e.g. Pinna et al., 1993; Kröner et al., 1987; Evans et al., 1999), at around 2000 Ma (e.g. Möller et al., 1995, 1998) and in the Archaean (Bell & Dodson, 1981; Cahen et al., 1984; Muhongo et al., 2001).
Lithologically, the Mozambique belt is dominated by amphibolite-facies and lower grade metamorphic rocks and granitoids, but it also contains numerous granulite complexes. Together with the igneous complexes, the latter were the focus of most of the geochronological studies cited above and the subject of extensive petrological and geochemical investigations (e.g. Maaskant et al., 1980; Sandiford et al., 1987; Gichile, 1992; Pinna et al., 1993; Möller et al., 1995, 2000; Appel et al., 1998). Therefore, the outcropping granulites play a major role in reconstructing Proterozoic metamorphic and geodynamic processes in the Mozambique belt. In contrast, few data are available for granulite xenoliths in volcanic edifices situated within the Mozambique belt of East Africa. They have been described from the Neogene volcanic rocks of the Lashaine volcano in northern Tanzania (Dawson, 1977; Jones et al., 1983). As is the case with the granulite xenoliths of this study, those of Lashaine represent relics of an igneous suite metamorphosed into granulites (Jones et al., 1983), but with a different chemical and mineralogical composition compared with the Chyulu xenoliths.
The local basement of the Chyulu Hills consists of hornblende and hornblende–biotite gneisses, migmatized and granitoid gneisses and subordinate garnet amphibolites and marbles (Saggerson, 1963; Omenge & Okelo, 1992). Granulites are rare. They form thin bands in gneisses, few of which are traceable for great distances (Saggerson, 1963). The granulites are calc-silicate in composition and contain andesine plagioclase, pyroxene (especially clinopyroxene), scapolite, garnet, epidote and in some cases quartz. They are lithologically different from the granulite xenoliths sampled by the young basanitic magmas. Because of the lack of isotopic and age data, both the outcropping basement granulites and the granulite xenoliths of the Chyulu Hills volcanic field cannot be a priori assigned to any of the above-described granulite occurrences in East Africa.
|
ANALYTICAL TECHNIQUES |
|---|
Mineral analyses were made on a CAMECA SX50 microprobe equipped with four wavelength-dispersive spectrometers at the Mineralogisch–Petrographisches Institut, University of Bern, and a CAMECA SX51 microprobe with five wavelength-dispersive spectrometers at the Mineralogisches Institut, University of Heidelberg. No systematic deviations were found in the control analyses performed on both machines. The microprobes were operated at a 15 kV accelerating voltage and a 20 nA beam current. The beam was focused to 1·5–2 µm for mafic phases, and to 5 or 10 µm for plagioclase. PAP correction was applied to the raw data. Natural and synthetic oxide and silicate standards were used for calibration. For spinel, corundum, sillimanite and plagioclase, Fe3+ and Fe2+ were calculated from stoichiometry. For other minerals, all Fe was taken as FeO.
Whole-rock major element abundances were determined on glass pills [for preparation technique, see Ulianov et al. (2005)] using a CAMECA SX50 microprobe at the Mineralogisch–Petrographisches Institut, University of Bern. The beam was rastered over an area of 15 µm x 20 µm; operating conditions were the same as for minerals. Another set of analyses was obtained by X-ray fluorescence (XRF) on fused discs using a Philips PW2400 instrument at the Institut de Minéralogie et de Pétrographie, University of Fribourg. The results from both techniques appear consistent. In the following, we use the electron microprobe measurements because of their slightly better totals. Cr, Ni, Sr and Ba were also determined by XRF at the University of Fribourg.
|
GENERAL CHEMICAL AND PETROGRAPHIC FEATURES |
|---|
The major element compositions and CIPW norms of the studied granulite xenoliths are presented in Table 1. The rocks are slightly silica-undersaturated, rich in Al and relatively enriched in Ca. The overall high Mg-number and very low abundances of high field strength elements (HFSE) and light rare earth elements (LREE) (Ulianov et al., in preparation) suggest that the rocks represent a sequence of cumulates rather than crystallized melts. The CIPW normative compositions are dominated by plagioclase and olivine, with the normative ‘troctolitic’ component (plagioclase + olivine) ranging from 89·6 to 98·7%. The high Ni and very low Ti abundances in the rocks accord well with a troctolitic mineralogy of the protoliths. The cumulate sequence also includes websteritic lithologies finally equilibrated under mantle pressures and temperatures. They belong to the same major and trace element fractionation trend as the granulites, but are more magnesian and depleted in Si, Al, Ca and alkalis (Ulianov et al., in preparation; see Discussion).
|
The mineral assemblage of the two studied Mg–Al sapphirine-bearing granulites (1904-3 and 1904-7) is spinel, sapphirine, sillimanite, plagioclase, ortho- and clinopyroxene, corundum, and garnet (Fig. 2a and b). The three studied Ca–Al hibonite-bearing granulites (1904-1, 1904-9, 1906-2) always show the assemblage spinel, sillimanite, clinopyroxene, quartz, plagioclase, corundum and garnet (Fig. 2c), but may additionally contain hibonite [Ca(Al,Cr,Ti,Si,Mg,Fe2+)12O19, 1904-1, 1904-9], mullite (1904-9) and sapphirine (1904-1). Both rock types display a wealth of reaction textures.
|
Two out of five granulite xenoliths analysed in this study are banded, composed of layers of more mafic and more felsic granulite. Mafic and felsic layers contain the same minerals but their modes differ. We interpret this banding to be of magmatic origin, as there are no indications of metamorphic differentiation processes or deformation (e.g. foliation) that could, in principle, also produce banding. Fabrics within the layers are isotropic and all of the reaction textures described in the section on textures and mineral compositions have developed in the absence of anisotropic stress.
The studied xenoliths are greatly affected by late-stage alteration. As a rule, garnet is replaced by kelyphite. The kelyphite is always extremely fine-grained. It consists of very Al-rich orthopyroxene, plagioclase and spinel. Clinopyroxene suffered incongruent melting and is often decomposed into a mixture of secondary diopside, plagioclase and quenched melt. Similar phenomena have been described elsewhere; for example, in some lherzolite xenoliths from Hawaii (White, 1966), in garnet lherzolites from South African kimberlites (Carswell, 1975) and in a sapphirine-bearing granulite xenolith from the Delegate pipe in New South Wales, Australia (Griffin & O'Reilly, 1986). These overprints are probably caused by fast heating and decompression in the host basanitic magma. Henceforth, we refer to the primary mineral compositions only (see Tables 2 and 3).
|
|
TEXTURES AND MINERAL COMPOSITIONS |
|---|
Mg–Al sapphirine-bearing granulites
The two studied Mg–Al sapphirine-bearing granulites (specimens 1904-3 and 1904-7) are petrographically similar to each other, although there are some distinctions. Xenolith 1904-3 is rather fine-grained, more mafic and appears mineralogically homogeneous. Xenolith 1904-7 is clearly banded, with more mafic (pyroxene-Rich) and more felsic (plagioclase-Rich) layers. Pyroxenes, plagioclase and garnet dominate, corundum and spinel are essential (5–15 vol. %), sapphirine is relatively rare. The rocks are characterized by various types of reaction domains:
(A) worm-like clusters of clinopyroxene and corundum rimming and often completely replacing spinel (A1, Fig. 3a) and coarse-grained to worm-like aggregates of orthopyroxene, clinopyroxene and spinel often accompanied by corundum (A2, Fig. 3b and c);
|
(B) sapphirine–sillimanite–garnet coronas on spinel (Fig. 3d and e);
(C) garnet developed at the expense of all mafic phases and forming ‘flaky’ garnet–plagioclase and massive garnet–clinopyroxene aggregates after plagioclase (Figs 2a and b and 3a–e);
(D) clusters of Tschermak-rich clinopyroxene intergrown with sillimanite and embedded in plagioclase (Fig. 3f).
Orthopyroxene occurs in two generations. Opx I forms large, often elongated xenomorphic grains (Fig. 2a and b). It is exsolved and contains abundant microlamellae of clinopyroxene visible in high-magnification backscattered electron images only. In specimen 1904-7, Opx I is often deformed. It may show undulose extinction and sometimes initial recrystallization. Analysed with a defocused beam, Opx I shows a pronounced decrease in Ca coupled with a moderate decrease in Al towards the rim (Fig. 4a). The Mg-number [= Mg/(Mg + Fetot)] remains unchanged. Opx II occurs in two subtypes. The first subtype is common in both specimens (Fig. 2a and b). It occurs as rather small, more or less isometric grains with no deformation, recrystallization or exsolution. The second subtype is found in specimen 1904-7 only. It forms aggregates of small strain-free grains apparently produced by recrystallization of Opx I. These aggregates may be rather large and are always corroded by thin rims of clinopyroxene. All transitions in the degree of exsolution from large exsolved grains of Opx I to relatively small exsolution-free grains of Opx II may be found. Opx II in specimen 1904-3 is nearly homogeneous in all elements (Fig. 4b). In specimen 1904-7, the Al content in Opx II either steadily increases from the core rimwards (Fig. 4c) or increases and then again decreases within the rim (Fig. 4e). The Ca content may slightly decrease rimwards (Fig. 4c). The Mg-numbers of orthopyroxene range from 84·0 to 86·0. CaO in Opx I, analysed with a defocused beam, reaches 1·2–1·5 wt %, whereas the CaO content in Opx II is much lower (0·16–0·24 wt %; Table 2).
|
Clinopyroxene forms two (1904-3) or three (1904-7) generations. The first generation is in equilibrium with Opx II and will therefore be further referred to as Cpx II (Fig. 2a and b). Cpx II grains occur as part of reaction domains A (Fig. 3a–c). They do not show any exsolution effects, with the larger grains showing chemically homogeneous cores. Towards the rim, an increase in Al, Na and Mg-number and a decrease in Ca are observed (Fig. 4e and f). The zonation is systematic and regular in specimen 1904-3, where all grains of Cpx II display very similar profiles (Fig. 4e). In specimen 1904-7, however, the zonation is irregular (Fig. 4f). The rim contents of Al vary over a wide range from profile to profile. Cpx III occurs in reaction domains C (Fig. 2a) or as mantles on Opx II, and is compositionally similar to the rims or intermediate zones of Cpx II grains (Table 2). Cpx IV is unusual. Found in specimen 1904-7 only, it forms part of reaction domains D (Fig. 3f). It is extremely rich in Tschermak components and has 18–23 wt % Al2O3 (Table 2). The Mg-numbers of clinopyroxene range from 85·9 to 92·2. CaO ranges from 17·0 to 21·6 wt % (Table 2).
Spinel forms two textural generations. Spl I is common and occurs as large, more or less isometric grains in reaction domains A and B (Fig. 3a–c). Two subtypes of Spl I can be distinguished. Spl Ia is present in specimen 1904-7 only, where it forms individual grains enclosed in plagioclase and partially or completely replaced by clinopyroxene and corundum (reaction domain A1, Fig. 3a). It does not show any evidence for earlier coexistence with ortho- and clinopyroxene. Spl Ib is ubiquitous and occurs as part of reaction domains A2 (Spl + Opx + Cpx ± Crn, Fig. 3b and c) and B (Spl + Spr + Sil + Grt, Fig. 3d and e) in both specimens. Chemically, the two subtypes of spinel are not well distinguishable. Cr tends to concentrate in Spl Ia (0·30–0·77 wt % Cr2O3 at 0·90–1·30 wt % NiO), whereas Ni preferably concentrates in Spl Ib (up to 1·74 wt % NiO, but at 0·10–0·35 wt % Cr2O3). The Mg-number ranges from 71·8 to 76·6 (Table 1). Spl I is not zoned but some grains may show a decrease in the Mg-number restricted to the outermost 10–20 µm. Spl II forms very thin rims on sillimanite in some of reaction domains B (Fig. 3d and e). It cannot be analysed, because of its small size.
Corundum forms worm-like intergrowths with clinopyroxene and platy aggregates with spinel in reaction domains A1 (Fig. 3a), as well as large individual grains in reaction domains A2 (Figs 2a and 3c). Corundum contains up to 0·50 wt % Fe2O3 and 0·16 wt % Cr2O3 (Table 2). Some grains are slightly zoned whereby Cr and Fe increase and Al decreases rimwards.
Plagioclase forms large grains and granular aggregates with frequent sillimanite inclusions. In the rim parts of such aggregates, it is often intimately intergrown with garnet forming ‘flaky’ Grt–Pl aggregates. Plagioclase also occurs as thin envelopes mimicking the borders of some Cpx II grains. Plagioclase is relatively Na-rich (An21–16 for 1904-3; An40–30 for 1904-7; Table 2). It may show some enrichment in Na in the envelopes around clinopyroxene II (Table 2).
Garnet replaces almost all other minerals, forming a network-like structure around pyroxene, spinel and corundum grains, ‘flaky’ aggregates with plagioclase and massive aggregates with Cpx III (Figs 2a and b and 3a–e). As garnet itself is normally replaced by kelyphite, only rare relicts are available for analysis. The Mg-number ranges from 62·6 to 70·1 and CaO from 4·2 to 6·0 wt % (Table 2).
Sapphirine forms small platy inclusions in the outermost zones of spinel I grains or rims spinel I in reaction domains B (Fig. 3d and e). In the latter case, sapphirine apparently resorbs spinel and, in turn, is being replaced by sillimanite or directly by garnet. Sapphirine compositions are close to the ideal composition 7(Fe,Mg)O x 9(Fe,Al,Cr)2O3 x 3SiO2 (Table 2). NiO reaches 0·50 wt % and is the only minor component.
Sillimanite rims sapphirine and spinel in reaction domains B (Fig. 3d and e) and forms frequent needle-like inclusions in plagioclase (Figs 2b and 3f). If associated with spinel and sapphirine, sillimanite tends to replace spinel, always replaces sapphirine and, in turn, is resorbed by garnet. Sillimanite hosted in plagioclase appears to be in textural equilibrium with the latter. Fe2O3 contents are <0·65 wt % (Table 2).
Ca–Al hibonite-bearing granulites
The three studied samples (1904-1, 1904-9, 1906-2) are very similar to each other. Xenolith 1904-1 is relatively coarse-grained (plagioclase grains reach 3–4 mm, mafic minerals are smaller in size) and banded. Xenoliths 1904-9 and 1906-2 are finer grained and homogeneous. Specimen 1906-2 is more mafic compared with the others. Major minerals in all samples are clinopyroxene and plagioclase, while corundum and garnet are essential (10–15 vol. % each) and quartz, spinel and sillimanite subordinate (5–7 vol. %). In addition, sapphirine in sample 1904-1 and hibonite in samples 1904-1 and 1904-9 occur as rare accessories. Mullite is very rare. It is encountered in one specimen (1904-9), and only in three grains.
Textures are characterized by felsic and mafic domains. The felsic domains consist of plagioclase and the mafic domains are dominated by clinopyroxene, corundum, garnet, spinel, sillimanite and quartz (Fig. 2c). Both domains are normally separated by a fine-grained corona that consists of garnet and recrystallized plagioclase, grading locally, as in the Mg–Al granulites, into a massive zone of intimately intergrown micrograins of garnet and clinopyroxene (Figs 2c and 5a and b).
Clinopyroxene can be divided into three generations. Cpx I occurs in close association with quartz. Together, they form granular cores of large complex clinopyroxene grains (Fig. 5a and b). Typically, such cores contain either one to several large (0·2–0·3 mm) or 10–20 small (<0·1 mm) drop-like inclusions of quartz with different optical orientations. Worm-like intergrowths of quartz and Cpx I may occur as well. Apart from this texture, the rocks are devoid of quartz. The core is always mantled by an outer zone of clinopyroxene that lacks quartz. Normally, the whole grain is embedded in plagioclase. Cpx II forms smaller grains that do not have quartz. Cpx III forms very small (<10 µm) grains intergrown and texturally equilibrated with garnet in the massive fine grained Grt–Cpx zones developed after plagioclase (Fig. 2c).
|
In general, clinopyroxene is characterized by moderate to high Mg-number (86·3–91·2 in Cpx I, 83·6–90·7 in Cpx II and III). CaO ranges from 18·05 to 21·68 wt % (Table 3). Chemically, Cpx I is easily distinguishable from Cpx II and III. It is more magnesian than Cpx II, more Ca-rich, less aluminous and much less tschermakitic than Cpx II and III (Table 3). Cpx II is zoned (Fig. 6). If no inclusions are present, the core parts of the largest grains display plateaux in Al (Fig. 6a and b). Closer to the rims, the Al content increases and then always slightly decreases again within the outermost 10 µm. The Na zonation mimics that of Al but is less extreme; Ca shows the inverse pattern to Na (Fig. 6). The Mg-number increases rimwards, sometimes with a slight tendency to decrease again within the outermost rim (Fig. 6). In smaller grains, Cpx II shows the same style of zonation, but with no plateau (Fig. 6c–e). Cpx III has a composition intermediate between the core and the rim zones of Cpx II (Table 3).
|
|
Plagioclase forms large isometric or elongated grains with abundant needle-like inclusions of sillimanite and late inclusion-free envelope-like zones mimicking the borders of clinopyroxene II in contact with the ‘flaky’ Grt–Pl or massive Grt–Cpx aggregates (Figs 2c and 5a). As an exception, plagioclase occurs as amoeboidal inclusions in clinopyroxene. Plagioclase is Na-rich (An39–34 for 1904-1; An35–32 for 1904-9; An31–26 for 1906-2). It is slightly inhomogeneous and shows a tendency towards higher Ab in the envelopes (Table 3).
Corundum forms grains of variable shape. It is always closely associated with clinopyroxene and often occurs as inclusions in the latter (Fig. 2c). Isolated grains of corundum are Cr-poor and almost unzoned. Rare corundum grains intergrown with Spl II demonstrate strong chemical zonation towards Spl II: the content of Cr2O3 reaches 7·0 wt %, whereas the Al content decreases (Table 3).
Garnet replaces almost all other minerals and does not form independent grains. Most garnet is developed at the expense of corundum and plagioclase. As a rule, garnet is altered into kelyphite, only rare relics are preserved. The Mg-number ranges from 58·6 to 65·5, and CaO content varies from 7·90 to 11·45 wt % (Table 3).
Spinel can be subdivided into three generations. Spl I is common. It forms relatively large grains always associated with sillimanite, sometimes also with sapphirine, hibonite or mullite. The textural relationships between Spl I, Sil and other minerals are not easily discernible. Almost all grains of Spl I are completely mantled by platy crystals of sillimanite (Fig. 5c and d). Besides that, sillimanite tends to separate spinel into individual blocks, and all transitional forms from single grains of Spl I with a Sil mantle to well-developed mosaic intergrowths of Spl I and Sil are observed (e.g. Fig. 5c). Some of these aggregates show clear textural indications of sillimanite replacing spinel. In others, these minerals seem to be in textural equilibrium. Sapphirine, mullite and/or hibonite, if present, occur in the core areas of the Spl–Sil aggregates (Fig. 5d and e). Of these minerals, only mullite apparently resorbs spinel (Fig. 5e). Hibonite shows linear contacts with spinel with no clear indications of resorption, although spinel often rims hibonite and not vice versa. The aggregates are always corroded by garnet (Fig. 5c–e). Except for one case of direct contact with spinel, garnet resorbs only the outer sillimanite mantle, but not the inner mineral assemblage (Fig. 5c–e). In thin section, the described aggregates are always closely associated with Cpx I and II (Figs 2c and Fig. 5d and e).
Spinel II is rare. It forms small grains localized directly at the rims of the larger corundum grains, or penetrating corundum from the rim inwards (Fig. 5f). Late spinel III occurs as very thin mantles picking out borders of corundum and sillimanite in the various aggregates (e.g. Fig. 5d).
Individual Spl I grains are nearly homogeneous but they show considerable inter-grain variations in Cr2O3 contents and Mg-number, depending on their textural position (Table 3). Cr2O3 typically ranges from 0·3 to 6·0 wt %, but is lower (0·3–2·7 wt %) in Spl I associated with hibonite and much higher (up to 22 wt %) in some of the Spl I–Sil aggregates. Mg-number ranges from 66·5 to 70·9 (Table 3). NiO and ZnO are generally high (1·1–2·3 and 1·0–2·6 wt %, respectively). Spinel II is also nearly homogeneous in individual grains. It is extremely rich in Cr (33–40 wt % Cr2O3) and has lower NiO contents and Mg-number than Spl I (Fig. 7; Table 3). The composition of spinel III is not easily measurable because of small grain size.
|
Sillimanite forms platy aggregates with spinel I (Fig. 5c), outer zones on spinel I and its aggregates with hibonite, mullite and sapphirine (Fig. 5d and e), as well as frequent needle-like inclusions in plagioclase (e.g. Fig. 5a). In the aggregates, sillimanite apparently resorbs hibonite, may resorb spinel I, sapphirine and mullite (Fig. 5d and e) and, in turn, is resorbed by garnet. Sillimanite shows a spectrum of compositions from nearly ideal in most cases to Cr-rich (up to
2·0 wt % Cr2O3) in contact with Cr-rich Spl I. Fe2O3 may amount to 0·65 wt % (Table 3). Sapphirine occurs as a rare accessory phase and is identified with confidence in specimen 1904-1 only. In specimen 1906-2, it may be present as micro-relicts enclosed in sillimanite in the outer parts of Spl–Sil aggregates, but the grains are too small to be analysed by electron microprobe. Sapphirine in specimen 1904-1 is closely associated with spinel, hibonite and sillimanite (e.g. Fig. 5d). It may show straight borders with Spl I with no indications of resorption, but more often it resorbs Spl I. In turn, it is always resorbed by sillimanite, and in some cases directly by garnet. The analysed sapphirine is close to the ideal composition 7(Fe,Mg)O x 9(Fe,Al, Cr)2O3 x 3SiO2. Cr2O3 may reach 1·12 wt %, and NiO 0·47 wt % (Table 3).
Mullite occurs in specimen 1904-9 only. It is found in the cores of three spinel–hibonite–sillimanite aggregates, one of which is shown in Fig. 5e. Mullite apparently resorbs hibonite and spinel I and, in turn, is replaced by sillimanite. Chemically, it belongs to the 2:1 type (2Al2O3:1SiO2). In the mullite solid solution series Al4+2xSi2–2xO10–x (Cameron, 1977), it is characterized by x 0·40 (Table 3). The mullite is nearly homogeneous. It may contain limited amounts of Cr2O3 and FeOt (<0·50 wt % each; see Table 3).
Hibonite occurs as accessory phase in xenoliths 1904-1 and 1904-9. Its absence in xenolith 1906-2 may be due to the relative depletion of its bulk chemical composition in Al and Ca compared with xenoliths 1904-1 and 1904-9 (see Table 1). Hibonite forms small isometric or elongated, often complexly indented grains intergrown with spinel I in the cores of complex corona textures made of hibonite, spinel, mullite, sapphirine and sillimanite (Fig. 5d and e). Neither hibonite nor spinel and sapphirine replaces each other. At the same time, the petrographic relationships of hibonite with spinel and sapphirine point to early crystallization of hibonite, as both spinel and sapphirine tend to localize on the border of hibonite grains, while hibonite never rims either spinel or sapphirine (Fig. 5d and e). Mullite, when present, apparently resorbs hibonite, just as sillimanite does (Fig. 5e).
Chemically, the studied hibonite is close to the ideal formula Ca(Al,Ti,Si,Mg,Fe2+)12O19. Ca, Al, Si, Mg and Fe contents vary in a narrow range. SiO2 ranges from 2·32 to 3·10 wt %. Some TiO2 is always present (0·05–0·74 wt % in specimen 1904-1, 0·26–2·29 wt % in specimen 1904-9). Significant intergrain variations and a systematic rimwards increase in Ti are also common. Cr2O3 ranges from 0·03 to 0·52 wt %, and ZnO does not exceed 0·25 wt % (Table 3). Ca contents vary from 1·016 to 1·048 c.p.f.u. and are somewhat in excess compared with the ideal stoichiometry of hibonite. An excess of Ca has been found in both terrestrial (e.g. Maaskant et al., 1980) and meteoritic (e.g. Keil & Fuchs, 1971; MacDougall, 1979) hibonite and is common for this mineral. This indicates that an exchange of (Mg,Fe2+) and Ca in site A of the hibonite structure is unlikely. At the same time, the classical substitution (Ti,Si) + (Mg,Fe2+) 
2(Al,Cr) in site B (e.g. Burns & Burns, 1984) and the common exchange Ti 2(Mg,Fe2+) were operative (Ulianov et al., 2005). The studied hibonite is very poor in REE (
REE from 1·5 to 30 µg/g) with positive Eu anomalies and variable enrichment in LREE and very low abundances (<1–2 µg/g) of Nb, Ta, Zr, Hf and Th (Ulianov et al., 2005).
|
PRE-METAMORPHIC VS METAMORPHIC MINERAL ASSEMBLAGES |
|---|
The studied xenoliths are considered to have undergone nearly complete subsolidus recrystallization. Although their CIPW normative compositions are troctolitic, modal olivine is absent. It seems to have completely reacted out during subsolidus cooling, as discussed below. The plagioclase grains actually present show indications of metamorphic re-equilibration and thus do not represent igneous relicts, texturally or compositionally.
In the Mg–Al sapphirine-bearing granulites, the only potential igneous relicts are the largest grains of orthopyroxene (Opx I), of which the reintegrated composition is very rich in calcium and aluminum (Table 2), and possibly a minor part of spinel texturally independent of pyroxenes (Spl Ia). Spinel Ia compositions, however, re-equilibrated under subsolidus conditions and are almost indistinguishable from those of metamorphic spinel Ib (see the section on ‘Textures and mineral compositions’). Sapphirine never forms individual grains and is found only as a constituent of the complex corona textures developed around the grains of Spl I, most of which are metamorphic in origin (see the section on ‘Metamorphic reactions’). Therefore, we do not interpret sapphirine as magmatic, although it may be a liquidus phase in basic Al-rich bulk compositions (Liu & Presnall, 1990, 2000). Corundum appears texturally equilibrated with metamorphic spinel and pyroxenes (Fig. 3c; see the section on ‘Metamorphic reactions’) and is hence not considered to be igneous even though it may also be a liquidus phase in basic Al-rich bulk compositions (Liu & Presnall, 1990, 2000). Sillimanite and garnet form only reaction rims and are clearly of metamorphic origin.
In the Ca–Al hibonite-bearing granulites, hibonite and some spinel are the only phases that could be igneous relicts; hibonite in terms of composition and textures, and spinel only texturally. The metamorphic transformation of the initial troctolitic assemblage olivine + plagioclase on cooling must have formed pyroxenes and spinel, as in the Mg–Al granulites (see the section on ‘Metamorphic reactions’). Spinel grains intergrown with or rimmed by sillimanite are common in the studied specimens, although some of them may texturally represent relicts of the igneous stage. When hibonite is present together with spinel in such aggregates, spinel grains tend to localize on the border of hibonite grains and sometimes rim hibonite, whereas hibonite never rims spinel (see Fig. 5d and e, and the section on ‘Textures and mineral compositions’). Therefore, hibonite can be considered as the earliest mineral of the reaction sequence most likely to represent an igneous relict. The latter hypothesis is confirmed by experiments that show hibonite to be a liquidus phase in experimental systems relevant to Ca,Al-rich inclusions in chondrite meteorites (Drake & Boynton, 1988; Beckett & Stolper, 1994; Kennedy et al., 1994). Some hibonites from the liquidus experiments are markedly enriched in silica (up to 1·7 wt %) despite the strong silica undersaturation of the coexisting melt (Beckett & Stolper, 1994). This suggests that high silica contents (as in the studied hibonites; see Table 2) may be characteristic of igneous hibonite.
Hibonite in the studied xenoliths is characterized by significant chemical variations, particularly in Ti and REE. Hibonites with very low Ti contents may retain their pristine (igneous) chemical composition, whereas high-Ti hibonites were apparently re-equilibrated during metamorphism (Ulianov et al., 2005).
|
THERMOBAROMETRIC CONSTRAINTS |
|---|
In the Mg–Al sapphirine-bearing granulites, quartz is completely absent and thus no geobarometers based on Px–Pl–Grt–Qtz assemblages can be used. In the Ca–Al hibonite-bearing granulites, quartz is not equilibrated with garnet and plagioclase, thus excluding the application of Grt–An–Cpx–Qtz or Grt–An–Sil–Qtz geobarometers. Barometry based on the reaction Spl + Qtz = Grt + Sil (Bohlen et al., 1986
; Nichols et al., 1992) is hampered by the lack of data on the equilibrium composition of garnet in the course of early metamorphism, to which this reaction is relevant. Given the presence of sillimanite and the absence of kyanite or textures indicating growth of sillimanite at the expense of kyanite in the studied xenoliths, the P–T evolution of the granulites seems limited to the sillimanite stability field. We thus fixed a maximum pressure of 8 kbar for all temperature calculations.
For the Mg–Al granulite xenoliths, we calculated temperatures using the cores of Opx I with the exsolved lamellae of Cpx. We obtained estimates of 1090 and 1135°C with the Ca-in-Opx thermometer of Brey & Köhler (1990) (see Table 4). These temperatures probably correspond to an early stage of subsolidus cooling rather than to the magmatic crystallization of the protolith because the original composition of Opx I was modified by diffusion owing to mineral re-equilibration on cooling. For the latest stage of equilibration recorded in the pyroxene rims and in the aggregates of Cpx III and Grt, the temperature estimates range widely from 514 to 784°C, depending on the geothermometer used (Table 4). Discrepancies between the estimates based on the two-pyroxene solvus and on the Fe–Mg exchange between Grt and Cpx may indicate either unreliability of solvus geothermometry in this range of temperatures, or differences in the diffusivities of Ca, Fe and Mg. In specimen 1904-7, where phases are poorly equilibrated in general, the discrepancies are especially pronounced.
|
For the Ca–Al granulite xenoliths, temperatures may be evaluated with confidence only for the latest stage of equilibration recorded in the pyroxene rims and in the aggregates of Cpx III and Grt. The estimates from two conventional Fe–Mg Cpx–Grt geothermometers (Ai, 1994
; Krogh Ravna, 2000) are shown in Table 4. Almost no discrepancies between these two geothermometers are observed. Specimens 1904-9 and 1906-2 are ‘hotter’ (725–741°C) than specimen 1904-1 (635–639°C). |
METAMORPHIC REACTIONS IN THE Mg–Al SAPPHIRINE-BEARING GRANULITES |
|---|
The major metamorphic features of these rocks that have to be explained are reaction domains A and B as well as the ubiquitous formation of garnet (see the section on ‘Textures and mineral compositions’).
Reaction domain A1: formation of worm-like clusters of Cpx + Crn rimming and often completely replacing Spl (Fig. 3a). Spinel Ia in the cores of the clusters is early and has reacted with enclosing plagioclase. Obviously, the sapphirine- and sillimanite-forming reactions of domain B (described below) could not initiate, because of the absence of adjacent pyroxenes. Instead, the reaction Spl + Pl = Cpx + Crn took place (see Fig. 7). The lack of pyroxenes around Spl Ia grains may also suggest that these grains were not produced by the reaction Ol + Pl = Opx + Cpx + Spl but that they represent earlier, perhaps igneous relicts (see the section on ‘Pre-metamorphic vs metamorphic mineral assemblages’). Metamorphic Cpx–Crn aggregates formed by the reaction Spl + Pl = Cpx + Crn have, to our knowledge, not yet been described from localities elsewhere. Assemblages of Spl and Pl formed at the expense of Cpx and Crn are known from mafic rocks within the Horoman Peridotite complex, as described by Morishita & Arai (2001). The reaction Spl + Pl = Cpx + Crn in the CMAS system has been calculated by those workers at 800°C at the pressures relevant here (8 kbar). In the Mg–Al granulites, this reaction was crossed towards the low-temperature side (Cpx + Crn) upon cooling.
Reaction domain A2: formation of pyroxenes and spinel (± corundum). The coarse-grained, worm-like or symplectitic aggregates of Opx II, Cpx II and Spl Ib (Figs 2a and 3b) are interpreted to have formed at the expense of the primary igneous assemblage olivine + plagioclase. Such Opx–Cpx–Spl textures are well documented from several localities (e.g. Tenthorey et al., 1996; Müntener et al., 2000; Montanini & Tribuzio, 2001). The reaction Ol + Pl = Opx + Cpx + Spl in the CMAS system is univariant and has been determined experimentally by many workers (Kushiro & Yoder, 1966; Herzberg, 1978; Gasparik, 1984a). The high-temperature assemblage Ol + Pl is stable at 1250–1300°C at relatively low pressures (8 kbar; Kushiro & Yoder, 1966). In the Mg–Al granulites, this assemblage was replaced by Opx + Cpx + Spl, probably during cooling after igneous crystallization. In Fig. 7, this reaction is shown schematically, because iron in natural systems, preferably accommodated by spinel, displaces the reaction to lower pressures, whereas sodium in plagioclase causes the opposite effect, with the additional chemical components making the reaction at least trivariant.
The Opx–Cpx–Spl aggregates may contain corundum forming individual grains (Figs 2a and 3c) and, less frequently, intergrowths with Spl I. Because no indications for the growth of corundum at the expense of pyroxenes or spinel (and vice versa) in these domains are observed, corundum seems to have formed in equilibrium with pyroxenes and spinel. This points to the reaction Ol + Pl = Cpx + Opx + Spl + Crn that replaces Ol + Pl = Cpx + Opx + Spl upon cooling, provided the pressure is relatively low. To our knowledge, this reaction has not been described from experiments or from nature, nor has it been calculated in any grid. It must, however, take place within the stability field of Opx + Crn, which is limited to temperatures below 850°C at the pressures of interest (8 kbar) in MAS grids (Gasparik, 2003; Ouzegane et al., 2003; Kelly & Harley, 2004). The feasibility of this reaction for natural systems depends on the bulk chemical composition (e.g. Na and Fe contents) and its variations in local domains of the studied rocks, which affects the stability of the assemblage Cpx + Opx + Spl + Crn in P–T space.
Reaction domain B: formation of sapphirine and sillimanite. The sapphirine–sillimanite coronas on spinel I (Fig. 3d and e) occur only in cases where the initial spinel is situated near both plagioclase and pyroxene. The key to understanding this texture is given by Fig. 3d and e. Although spinel I appears strongly resorbed by garnet, both sapphirine and sillimanite are present even on the smallest relicts of Spl I (Fig. 3d). This texture points to the co-crystallization of the assemblages Grt + Spr and Grt + Sil after spinel I, plagioclase and pyroxene. It argues against the formation of only sapphirine and sillimanite after spinel, with subsequent resorption of the resulting Spl–Spr–Sil coronas by garnet. A possible sequence of reactions during cooling, leading to the Spr–Sil–Grt corona textures after spinel, can be inferred from a Schreinemakers analysis in the CMAS system (Christy 1989; Fig. 7) as follows:
 | (1) |
 | (2) |
Reactions (1), nearly degenerate, probably acted simultaneously and are responsible for sapphirine and garnet, whereas reaction (2) produced sillimanite and garnet once spinel was mantled by sapphirine and garnet. Reaction coronas similar to those described here are known in their fully developed forms only in granulite xenoliths from the Stockdale kimberlite, Kansas (Meyer & Brookins, 1976).
Reaction domain C: formation of garnet. The occurrence of garnet in the corona textures with spinel, sillimanite and sapphirine (Fig. 3d and e) can be satisfactorily explained by the Schreinemakers diagram (Fig. 7), as discussed above. Garnet developed at the expense of corundum (Fig. 2a), can be attributed to the reaction Crn + Opx ± Cpx = Grt. The resorption of pyroxene grains located close to corundum by garnet can also be explained by this reaction. Garnet resorbing pyroxenes and spinel in reaction domains A (Fig. 3b) cannot have formed by the reaction Opx + Cpx + Spl = Grt, as the latter requires additional silica in order to consume spinel. We speculate, therefore, that plagioclase participated in the reaction Opx + Cpx + Spl + Pl = Grt (Gasparik, 1984a). The ‘flaky’ and massive zones of garnet or garnet and clinopyroxene propagating into and resorbing plagioclase (Fig. 2a) may also be explained by this reaction or by Opx + Spl + Pl = Grt + Cpx (Green & Ringwood, 1967) taking place simultaneously to the development of garnet in reaction domains A and B. Similar reaction zones of garnet after spinel, pyroxenes and plagioclase in metamorphosed metabasites have been described from many occurrences of mafic granulites throughout the world (e.g. Griffin, 1971; Irving, 1974; Indares & Rivers, 1995; Attoh, 1998; Indares, 2003), including meta-anorthosites and related rocks of eastern and northeastern Tanzania (Appel, 1996; Appel et al., 1998). Most of the above reactions can easily take place during cooling, either isobaric or concomitant to/followed by compression, although only the garnet-forming reaction Crn + Opx ± Cpx = Grt clearly necessitates compression (Fig. 7).
Although garnet seems generally stable and no textural evidence for its breakdown has been found, in specimen 1904-7 it may have started to decompose to the very end of the reaction path. A possible reaction (Pl + Grt = Cpx + Sil) involves also plagioclase and accounts for the formation of Cpx IV with its exceptional concentrations of Al (reaction domain D). Similar mineral relations are observed in sapphirine-bearing granulites from Stockdale (Meyer & Brookins, 1976).
It is worth stressing the importance of local chemical heterogeneities and additional chemical components in the studied rocks, which control the P–T dependence of the described mineral assemblages. For example, the formation of spinel–sillimanite–sapphirine–garnet corona textur