Journal of Petrology Advance Access originally published online on April 19, 2008
Journal of Petrology 2008 49(6):1069-1096; doi:10.1093/petrology/egn018
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Geochemistry of the Volcan de l’ Androy Basalt–Rhyolite Complex, Madagascar Cretaceous Igneous Province
1School Of Ocean And Earth Science And Technology, University Of Hawaii, Honolulu, Hi 96822, USA
2Geology Department, University Of Leicester, Leicester Le1 7Rh, Uk
3Department Of Environment, Technology And Social Studies, Roskilde University, Dk-4000, Roskilde, Denmark
4Laboratoire De GéOlogie Appliquée, Université D’ Antananarivo, Antananarivo, Madagascar
RECEIVED JUNE 7, 2007; ACCEPTED MARCH 13, 2008
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONAGE AND FORM OF...SAMPLING AND ANALYTICAL METHODSRESULTSDISCUSSIONCONCLUDING REMARKSSUPPLEMENTARY DATAAPPENDIX: MINERAL-LIQUID...REFERENCES |
|---|
The 4000 km2 Androy massif in southeastern Madagascar is a >2000 m thick sequence of interbedded basalt and rhyolite erupted during a widespread Cretaceous episode of predominantly basaltic volcanism. Two geochemically different groups of basalt, tholeiitic group B1 and mildly alkalic B2, are present, as are two different groups of rhyolite, R1 and R2. Both the basalts and rhyolites appear to have issued from relatively nearby feeders, as compositionally equivalent intrusions are exposed in the vicinity. The R2 rhyolites define a whole-rock Rb–Sr isochron of 84·0 ± 2·4 Ma (2
), the same, within error, as an 40Ar–39Ar sanidine age reported by earlier workers. Plate reconstructions suggest that the area was near the Marion hotspot at this time. Some involvement of hotspot mantle is allowed, but not required, by Nd–Pb–Sr isotope data for the basalts. The two types of basalt may have formed by different amounts of melting of the same mantle source, which remains rather poorly specified, but group B1 was affected much more than B2 by contamination with continental material, probably Archean crust. The R1 rhyolites are petrogenetically related to the B1 basalts, with which they are interbedded. The R2 rhyolites may be derived from melts of frozen high-
Nd B1 basalt coupled with fractionation and assimilation of relatively small amounts of crust. Alternatively, although these rhyolites were erupted significantly later than the B2 basalts, they may have formed through advanced crystal fractionation of B2-type magma and relatively small amounts of crustal assimilation. Separate magmatic plumbing systems appear to have existed more or less contemporaneously in the Androy area. KEY WORDS: large igneous province; rhyolite; mantle plume; basalt; Madagascar
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONAGE AND FORM OF...SAMPLING AND ANALYTICAL METHODSRESULTSDISCUSSIONCONCLUDING REMARKSSUPPLEMENTARY DATAAPPENDIX: MINERAL-LIQUID...REFERENCES |
|---|
Madagascar was the site of extensive basaltic volcanism between about 92 and 84 Ma, shortly before Madagascar and Greater India rifted apart (Storey et al., 1995
; Torsvik et al., 1998). The volcanic episode is postulated to be linked to both rifting and the Marion (also known as Prince Edward) hotspot. Basalt flows along the southeastern coast of Madagascar had a source that was isotopically similar to those of recent volcanoes of the Marion and Crozet (
500 km east of Marion) hotspots, but elsewhere geochemical evidence for a hotspot-type mantle source remains elusive (Mahoney et al., 1991; Storey et al., 1997; Melluso et al., 2001, 2002). The amount of Cretaceous lava removed by erosion is uncertain, but most volcanic sections are less than 200 m thick today. By far the thickest remaining lava pile is in southeastern Madagascar at Volcan de l’Androy (Androy hereafter), where more than 2000 m of flows are present (Battistini, 1959) (Fig. 1a). The projected trends of three major dike swarms converge to the east of the Androy massif (Ernst & Buchan, 1997), and plate reconstructions and radiometric dating suggest that the Marion hotspot was centered near this area toward the end of the magmatic episode (Fig. 1b) (e.g. Storey et al., 1997; Kent et al., 2002).
|
Apart from its great thickness and distinctive location, the Androy massif is unusual in that it is composed of interbedded basalt and abundant phenocryst-poor (generally <10% phenocrysts) rhyolite. Rhyolites are rare in most of the Madagascar province, although some are present along the central part of the rifted eastern margin and in the NW. Previous geochemical work has concentrated on the basalt-dominated flows and dikes along the eastern coast and in the Majunga and Morondava basins of western Madagascar (e.g. Mahoney et al., 1991
; Dostal et al., 1992; Storey et al., 1997; Melluso et al., 1997, 2001, 2002, 2003; but see also Melluso et al., 2005). In this study, we report major element, trace element, and Nd–Sr–Pb isotope data for the Androy massif, and investigate the relationships between the basaltic and silicic rocks.
Silicic rocks are also found in other large igneous provinces, although their abundance varies greatly. Both the Paraná–Etendeka and Karoo–Ferrar provinces are characterized by extensive rhyolite and trachydacite; however, silicic rocks are only minor components of the Deccan and Siberian Traps (e.g. see reviews in Macdougall, 1988). In the Columbia River province, silicic rocks were not erupted during the main Miocene phase of flood basalt formation, but are abundant, and in some intervals predominant, during subsequent development of the Snake River Plain and Yellowstone system (e.g. Doe et al., 1982; Hooper, 1997). Understanding the processes that lead to the formation of bimodal systems in some provinces but not in others is a major goal of studies of large igneous provinces, as is understanding of the interactions between mantle-derived melts and the crust.
|
AGE AND FORM OF THE ANDROY MASSIF |
|---|
|
TOPABSTRACTINTRODUCTIONAGE AND FORM OF...SAMPLING AND ANALYTICAL METHODSRESULTSDISCUSSIONCONCLUDING REMARKSSUPPLEMENTARY DATAAPPENDIX: MINERAL-LIQUID...REFERENCES |
|---|
Cretaceous volcanism began in northern Madagascar and migrated southward over a period of several million years. Storey et al. (1995
) obtained an 40Ar –39Ar age of 84·4 ± 1· 0 (2) Ma for sanidine from a rhyolite (sample AND90-27) near the top of Vohitsimbe, the highest peak of the Androy massif. Plagioclase from a rhyolite dike (AND90-6) at the southeastern edge of the massif yielded an age of 86·3 ± 3·8 Ma, and a dolerite (AND90-30) was dated at 86·3 ± 2·0 Ma. Ages in the 84–86 Ma range also have been reported for basalt from the Ejeda–Bekily dike swarm to the west of Androy and, farther west, for lava flows of the southern Morondava Basin (Storey et al., 1995; Torsvik et al., 1998).
The Androy (‘thorn forest’) massif lies 100–150 km inland from the rifted eastern continental margin of Madagascar and forms a c. 4000 km2 region of elevated topography. The overall structure is that of a large ‘sag’ caldera filled by a saucer-shaped pile of flows resting on Archean basement gneiss of the Androyen System (Battistini, 1959; Storey et al., 1997). The lowermost exposed flows are basalts, termed the Rim Series (Fig. 2), and are exposed best in the NE and NW. At least one rhyolite unit is present in the upper part of the Rim Series. The Rim Series is overlain by the inward-dipping Lower Rhyolites, which in many areas are also the lowest exposed flows in contact with basement. Above the Lower Rhyolites are gently inward-dipping basalt flows, the Main Series, interbedded with and overlain by rhyolites termed the Upper Rhyolites. Both the thickness of rhyolite units and the proportion of rhyolite to basalt increase toward the top of the Main Series. The thickest units are 40–60 m thick, columnar-jointed rhyolites in the topmost portion of the Upper Rhyolites at Vohitsimbe. The Upper Rhyolites, in particular, lack eutaxitic textures (only one Androy silicic sample, AND90-18, shows any eutaxitic texture), suggesting that they may be lava flows rather than pyroclastic deposits, despite their high aspect ratio (generally >50 : 1).
|
The massif is flanked on the west and NE by additional outcrops of rhyolite and by intrusions of microgranite. Dikes of basalt and rhyolite crop out adjacent to the east and SE portions of the massif, and intrude some of the Lower Rhyolites and Rim Series basalt flows. Textbook-type ring faults are not observed, although the microgranites and some of the dikes may represent intrusions that followed such faults. In comparison with the largest present-day basaltic–silicic central volcanic complexes, such as Torfajökull in Iceland (e.g. Gunnarsson et al., 1998
), the Androy complex is several times larger. |
SAMPLING AND ANALYTICAL METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONAGE AND FORM OF...SAMPLING AND ANALYTICAL METHODSRESULTSDISCUSSIONCONCLUDING REMARKSSUPPLEMENTARY DATAAPPENDIX: MINERAL-LIQUID...REFERENCES |
|---|
We collected samples from the massif and nearby dikes and microgranite intrusions (details of locations are given in Electronic Appendix 1, which is available for downloading at http://www.petrology.oxfordjournals.org). Good stratigraphic control is available for the
900 m section of Main Series lavas and Upper Rhyolites at Vohitsimbe, where some 28 flows are exposed. Other flow samples were taken at lower stratigraphic levels, where exposure is more limited and stratigraphic position consequently is known only in a general way.
The methods used for bulk-rock chemical and Nd–Pb–Sr isotopic analysis were those of Storey et al. (1997). Concentrations of major elements and several trace elements were determined for all the samples by X-ray fluorescence spectrometry on agate-ground powders of amygdale- and vein-free chips of rock (Table 1; note that the oxides and totals are reported as volatile-free percentages). Additional trace elements were measured for a subset of powders by instrumental neutron activation analysis (Table 2). Isotope ratios of Nd, Sr, and Pb were measured for a smaller subset chosen to cover the range of major and trace element variation; corresponding parent and daughter element concentrations were determined by isotope dilution (Table 3). As in our past work, these concentrations are used principally for age-correcting isotope ratios; the isotope-dilution data do not necessarily represent bulk-rock values because small (1 mm) hand-picked chips of rock were used for isotope analysis rather than bulk-rock powders (to avoid alteration and phenocrysts). Because the 87Rb/86Sr ratios of most of the rhyolites are very high (as high as 113), the age-corrected (87Sr/86Sr)t values of the rhyolite samples are exceedingly sensitive to the choice of 87Rb decay constant and age [for example, for a 87Rb/86Sr of 113, an age difference of 1 Myr leads to a difference in (87Sr/86Sr)t of 0·0016]. We use Storey et al.'s (1995) sanidine age of 84·4 Ma and, for the decay constant, Kossert's (2003) value of 1· 395 x 10–11 year–1.
|
|
|
Mineral compositions were analyzed for a subset of samples chosen using the geochemical data and petrographical observations. Mineral analyses were obtained from polished thin sections with a JEOL JXA-8600S electron microprobe at the Department of Geology, University of Leicester, using an accelerating voltage of 15 kV and a probe current of 30 nA with a beam diameter of 10 µm. Count times of 20 s were used for measurement of characteristic peak intensities, with 10 s at each of the background positions selected on either side of the peak. Quantitative background-corrected results were standardized against a combination of synthetic materials and well-characterized natural minerals and corrected for matrix effects using a ZAF correction procedure. Minimum detection limits, calculated as three standard deviations above background, range from 0·02 wt % for Na2O to 0·05 wt % for FeO and NiO. Olivines from thin sections of AND90-1 and AND90-7 were measured with a probe current of 100 nA and counting times of 80 s on both peak and (total) background for TiO2, Al2O3 and NiO to improve the minimum detection limits for these oxides to 0·017, 0·008 and 0·013 wt %. The complete microprobe dataset is listed in Electronic Appendix 2.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONAGE AND FORM OF...SAMPLING AND ANALYTICAL METHODSRESULTSDISCUSSIONCONCLUDING REMARKSSUPPLEMENTARY DATAAPPENDIX: MINERAL-LIQUID...REFERENCES |
|---|
Petrography and mineralogy
The samples have been subdivided into five types on the basis of whole-rock composition, particularly distinctive trace element characteristics (see below). We distinguish two groups of basalt, B1 and B2, and two groups of rhyolite (sensu lato), R1 and R2. Group R1 is divided further into subgroups R1a and R1b. Group R1a comprises mostly rhyolites, but also includes a small number of trachydacites and microgranites. The compositional differences are reflected in the mineralogy. B1 basalt samples are from flows of the Main Series and Upper Rhyolites and several dikes on the southeastern side of the massif. Group B2 rocks are much less abundant, at least in exposed parts of the massif. As flows, they are confined to the Rim Series, and thus lie stratigraphically below group B1 and the rhyolites. We also sampled a B2 dike on the northeastern side of the massif. Rhyolite group R1 is represented in both the Lower Rhyolites and Upper Rhyolites, and we sampled one R1 flow within a sequence of B2 basalts in the Rim Series. The microgranite intrusions and some of the dikes bordering the massif also belong to R1. In contrast, we found R2 rhyolites only in the Upper Rhyolites, interbedded with R1 rhyolites and B1 basalts. Lacroix (1923
) reported an ankaratrite dike on the west side of Vohitsimbe, but we did not encounter such rocks.
Group B1
Samples belonging to compositional group B1 range from aphyric to moderately porphyritic picrite and basalt, with up to 20% phenocrysts; they also include dolerite. Groundmass textures vary from hypocrystalline to holocrystalline and sub-ophitic; some samples show fluxion (flow) textures. Most of the B1 samples are sparsely porphyritic, with less than 10% phenocrysts. The main phenocryst assemblages are olivine (e.g. samples AND90-1, -7 and -13; these also contain rare pyroxene phenocrysts), plagioclase (AND90-42b), plagioclase + olivine (AND90-8), plagioclase + augite (AND90-51, -56, -57), and olivine + plagioclase + augite (AND90-47A). In many samples, the pyroxene occurs with plagioclase in glomerocrysts. Magnetite or titanomagnetite phenocrysts are present in several samples, and magnesiochromite is embedded in olivine in AND90-7. Alteration in the B1 samples ranges from slight to severe, as reflected crudely by variable whole-rock weight loss on ignition (LOI, –0·45 to 8·75 wt %). The development of calcite in the groundmass is common, and green or brown clay sometimes replaces glassy mesostasis. Some B1 basalts are amygdale-rich, with fillings of calcite, quartz and chlorite.
The main silicate and oxide phases in four B1 samples (AND90-1, -7, -13 and -57) were analyzed by electron microprobe. Unaltered olivine phenocrysts in samples AND90-1 and -7 have high forsterite (Fo) contents. Sample AND90-7 contains the most magnesian olivine, with several crystals exceeding Fo91 [maximum Fo91· 8 (core) ranging to Fo86·2 (rim)]. The CaO content of these phenocrysts is between 0·2 and 0·3 wt %. Nickel oxide mostly ranges from 0·2 to 0·4 wt %, typical of basaltic olivine; however, a few phenocrysts contain more than 0·6 wt % NiO, higher than the most nickel-rich olivine in Hawaiian shield basalts (Sobolev et al., 2005). The crystals are anhedral to subhedral, with textures suggesting fragmentation, but none shows indications of the strain shadows often found in xenocrysts. Small olivine phenocrysts in AND90-13, a ferrobasalt, are more iron-rich (Fo32–34). No olivine was found in sample AND90-57.
Groundmass and microphenocryst plagioclase is in the range of An65–75 in AND90-1 and AND90-7, An51–59 in AND90-13, and An64–70 in AND90-57 (Fig. 3a). Pyroxene compositions vary from Ca46Mg46Fe8 to Ca36Mg36Fe28; that is, from diopside (AND90-1) to augite (AND90-7, -13, and -57; see Fig. 4). Many grains exhibit slight normal crystal zoning. No low-calcium pyroxene was found in the B1 samples.
|
|
Group B2
All the B2 basalts are porphyritic, with up to 25% phenocrysts. The dominant phenocryst assemblages are augite (AND90-79), augite + plagioclase (AND90-75, -76, -78), plagioclase + magnetite (AND90-14), and plagioclase (AND90-73). No olivine was found, reflecting the absence of any near-primitive bulk-rock compositions within this group (see below). Groundmass textures range from hypocrystalline to holocrystalline, sub-ophitic. As with group B1, alteration varies from slight to severe. Green or brown clay sometimes replaces glassy mesostasis. Secondary calcite is scarce. A few samples contain clay- and epidote-filled vesicles (AND90-80, -82) but most samples are amygdale- and vesicle-free. Overall, the B2 samples tend to be somewhat less altered than the B1 samples.
Pyroxene crystals in sample AND90-78 were analyzed by electron microprobe. All are augite, ranging from Ca43Mg47Fe10 to Ca37Mg46Fe17 (Fig. 4). This slight variation reflects core-to-rim zonation. No feldspars of group B2 were analyzed.
Group R1a
Group R1a is composed predominantly of rhyolite but also includes three samples of trachydacite (AND90-18, -22, and -23). The rocks vary from aphyric to moderately phyric, with up to 20% phenocrysts; most samples contain 10% phenocrysts. Plagioclase feldspar is a common phenocryst phase. In some samples, it is accompanied by quartz, ilmenite, magnetite, accessory zircon, and, rarely, sanidine. Pyroxene occurs in several samples. The groundmass, which in some samples shows devitrification textures, comprises mostly feldspar and quartz, ± magnetite. Only one sample (AND90-18) exhibits any eutaxitic texture. Sample AND90-22 is a porphyritic granophyre, with spectacular graphic intergrowths of quartz and alkali feldspar, phenocrysts of quartz and albite, and trace amounts of an altered mafic mineral, possibly originally amphibole. AND90-72 and -81 are samples of two-feldspar (subsolvus) microgranite from the NE quadrant of the Androy massif. They exhibit strong granophyric textures, and contain hedenbergite. Accessory zircon and apatite are also present in these samples. All of the R1a samples are altered; the alteration ranges from minor iron staining and slight alteration of feldspars and mesostasis to complete replacement of feldpars and mesostasis by clay, zeolite, quartz, and calcite. Pyroxene in microgranite AND90-72 is partially replaced by a pale olive-green amphibole, ferroedenite.
The main silicate phases in three samples of rhyolite (AND90-48, -71, -83), a trachydacite (AND90-23), and microgranite (AND90-72) from group R1a were analyzed by electron microprobe. Plagioclase in AND90-23, -72 and -83 is predominantly andesine or oligoclase (Fig. 3b); in addition, alkali feldspar crystals were found in AND90-72 and -23. The feldspar in rhyolite AND90-48 comprises large, euhedral, but slightly rounded phenocrysts of perthitic feldspar. One set of lamellae is pure albite, but we were unable to analyze the other (presumably orthoclase) lamellae. AND90-71, -72 and -83 contain clinopyroxene. The crystals in AND90-71 are 0·1– 0·4 mm across, subhedral, and comprise both augite and pigeonite (Fig. 4). Using the microprobe's electron backscatter facility, it is possible to distinguish the Ca-rich, Fe-poor (dark) and Ca-poor, Fe-rich (light) components. The low-Ca pyroxenes tend to rim the Ca-rich ones. The pyroxene phenocrysts in microgranite AND90-72 are anhedral, dark green hedenbergite, whereas those in rhyolite AND90-83 are partially rounded and resorbed augite, with Fs/En (ferrosilite/enstatite) ratios that are low for rhyolites; they may be xenocrystic. The pyroxene pairs in rhyolite AND90-71 indicate low temperatures of equilibration (very roughly 800–850°C), and the high Ca content of the hedenbergite in microgranite AND90-72 is consistent with the subsolvus textures and equilibration at a temperature of about 600°C (Lindsley & Andersen, 1983).
Group R1b
These high-silica rhyolites were collected from the Vohitsimbe massif. All are porphyritic, with between 20 and 25% phenocrysts comprising alkali feldspar (sanidine or perthite), quartz, and minor magnetite. No plagioclase, pyroxene, or amphibole was seen, although accessory zircon was found in several samples. The groundmass is usually a devitrified assemblage of quartz, feldspar, and magnetite, but in some samples it is cryptocrystalline or vitreous (e.g. AND90-28). AND90-50 is a crystal tuff. Only one R1b sample (AND90-25) exhibits any flow banding and none show eutaxitic textures. Alteration is less than in the R1a rhyolites; feldspars are nearly fresh, and a slight oxidation and development of pale brown clay in the groundmass is the main sign of alteration.
The main silicate phases in four R1b samples (AND90-26, -28, -37 and -45) were analyzed by electron microprobe. Sanidine (An2Ab48Or50) occurs in AND90-26 and -28, whereas AND90-37 and -45 contain partially resorbed and embayed perthite phenocrysts with almost pure end-member Na- and K- feldspar segregations (Fig. 3c).
Group R2
The R2 rhyolites are either aphyric or very sparsely phyric. Where present, phenocrysts are usually subhedral, turbid alkali feldspar with incipient or actual perthite (Fig. 3d), but their abundance is <1% of the total rock. The groundmass comprises microcrystalline to finely crystalline feldspar, quartz, and magnetite. Samples AND90-35, -36, -43, and -53 contain small (< 0·2 mm long) laths and blades of blue–green, pleochroic alkali amphibole, probably arfvedsonite, an amphibole formed under peralkaline conditions. Four samples (AND90-35, -36, -40, -41) contain opaque-rich segregations, and one of these (AND90-40) features a large (3 mm across), skeletal magnetite clot, possibly a pseudomorph after amphibole. Several samples contain trace amounts of minute crystals of zircon.
The groundmass feldspars are pervasively altered, with a dusty, iron-stained appearance, and in some samples the opaque phases (originally magnetite?) are partially replaced by red oxides or oxyhydroxides. Only limited evidence is present for flow alignment of the groundmass phases; most crystal growth appears to have occurred in a stress-free environment. No eutaxitic textures are seen.
Summary
The two groups of basalt cannot categorically be delineated on the basis of their mineralogy, but olivine was not found in group B2, consistent with the more evolved character of this group (see below). The high-Fo olivine in AND90-1 and -7 is consistent with the relatively high MgO content of these B1 picrites (see below), suggesting that they are not very fractionated. The high NiO content of some of the olivine could suggest derivation from a pyroxene-rich mantle source (see Sobolev et al., 2005).
The R1 rhyolites (and microgranites) are mostly porphyritic, with the dominant phenocrysts being feldspars and quartz. In contrast, the R2 rhyolites are mostly aphyric, and the only phenocryst is alkali feldspar, typically perthitic. In the R1a rhyolites, plagioclase is dominant over alkali feldspar, whereas plagioclase is not found in the R1b rhyolites, consistent with their more evolved, high-silica compositions (see below). Perthitic unmixing is common in feldspars of groups R1b and R2. The majority of the phenocrysts displaying such unmixing are invariably large (> 1 mm) and equant, and often show resorption. We infer that they are probably xenocrystic. The R1a rhyolites host a range of pyroxenes, including augite, hedenbergite and pigeonite. Neither R1b nor R2 rhyolites contain pyroxene, but several R2 rhyolites contain small crystals of arfvedsonite. Many rhyolite samples contain accessory zircon, and apatite was identified in the R1a rhyolites.
Chemical effects of alteration
Some care must be taken in interpreting the chemical data, for alteration has variably but visibly affected all the samples, as noted above. Alteration is reflected in a very general way in LOI values, which range widely from –0·45 to 8·75 wt % for the basaltic samples and from 0·18 to 3·31 wt % for the silicic ones.
As is common in subaerial weathering environments (e.g. Beane et al., 1986; Price et al., 1991), alteration appears to have variably affected K, Na, Rb, and Ba contents in the basaltic samples. Correlations between Ca and more alteration-resistant elements such as Ti or Fe are poor (Fig. 5c), suggesting that Ca contents have been affected significantly in some of the basalts. The scatter is diminished if samples with LOI >3 wt % are excluded (Fig. 5d); correlations involving P also improve. An extreme example of the effects of alteration-related element redistribution, coupled with outcrop-scale variability in crystal content, is provided by samples AND90-42a and -42b (Fig. 6). These samples were taken from a weathered outcrop of the same basalt flow about 15 m apart. Both are visibly very altered and have high LOI values (5·88 and 3·13 wt %, respectively). They have similar Th, Nb, Ta, Zr, Hf, Ti, Ni, and Co contents, but differ greatly in Rb, Ba, K, Sr, Y, Na, and Cr, and to a lesser extent in several other trace elements, including some of the rare earth elements (REE).
|
|
Some of the variability in Ca, K, and Na contents in the rhyolites (Figs 5c, d and 7e) can be ascribed to post-eruptive mobility, and mobility of SiO2 is indicated by SiO2 values greater than 78 wt % in several quartz-poor rhyolites (78 wt % is the maximum observed for unaltered quartz-free rhyolite; e.g. Barker, 1981
). In addition, the REE patterns of AND90-32 and -61 display sizeable negative Ce anomalies (Fig. 8h). These anomalies may reflect alteration in which Ce was mobilized in its +4 oxidation state, although the alteration in these samples appears superficially rather similar to that in some samples without Ce anomalies; alternatively, the anomalies may result from removal of trace amounts of unusual REE-rich accessory minerals (which, however, we did not see in the thin sections) during late-stage magmatic evolution. Miller & Harris (2007) also noted Ce anomalies in several of their samples from the Lebombo area of the Karoo province.
|
|
Geochemistry of the basaltic rocks
Group B1
This group includes both basalt and basaltic andesite. Most of the B1 rocks have low MgO (1· 83–5·59 wt %), Mg-number (23–53, = [100 x molar Mg/(Mg + Fe2+)], assuming Fe2O3 is 13% of total iron oxide), and Ni (6–101 ppm). However, dike sample AND90-1 and flow sample AND90-7 have much higher values (12·14 and 13·19 wt % MgO, Mg-numbers of 70 and 71, and 305 and 382 ppm Ni, respectively), consistent with their high-Fo olivine phenocrysts. Silica content ranges from 48·5 to 58·0 wt %, but three of the four samples with SiO2 > 54·6 wt % have very high LOI (4·13–7·50 wt %) and we suspect that alteration may have modified their SiO2 values. Data for all but four B1 samples fall below the line dividing tholeiitic and alkalic Hawaiian lavas in Fig. 5a; the high total alkali contents in the other four samples are likely to reflect local alteration-related enrichment (e.g. AND90-42b; Fig. 6).
All of the B1 rocks are moderately enriched in the light REE (LREE) relative to the middle (MREE) and heavy REE (HREE) (Fig. 8f). For example, chondrite-normalized (La/Sm)ch = 1·7–3·2 and (La/Yb)ch = 3·5–8·6; the lowest values (and flattest REE patterns) are exhibited by the two high-MgO samples. Europium anomalies range from absent to slightly negative. Primitive-mantle-normalized incompatible element patterns exhibit a Nb–Ta trough, with normalized (Nb/La)pm between 0·40 and 0·70 (Fig. 8a). Other notable features are sizeable peaks at Pb and troughs at P. With rare exceptions (e.g. high-MgO sample AND90-1), Sr is variably depleted relative to Nd. Broadly similar patterns are seen elsewhere in southern Madagascar in lava flows in the Morondava Basin and tholeiitic dikes of the Ejeda–Bekily swarm (Mahoney et al., 1991; Dostal et al., 1992).
Age-corrected isotope ratios define an array that partly overlaps, but is distinct from, that of the tholeiitic lavas of southwestern Madagascar. The B1 samples cover a wide range of Nd(t) and (87Sr/86Sr)t, from –2·5 to –13·2 and 0·70915 to 0·71500, respectively (Fig. 9a), and the values correlate roughly with degree of LREE enrichment [e.g. increasing (87Sr/86Sr)t and decreasing Nd(t) are accompanied by decreasing 147Sm/144Nd and increasing (La/Sm)ch]. Both the highest and lowest Nd(t) values are from dikes, whereas four of the five B1 lavas analyzed from the Vohitsimbe section display only a narrow range of Nd(t), from –6·5 to –7· 0. Relative to ocean ridges and islands, all of the B1 rocks have high (207Pb/204Pb)t (15·82–15·96) and (208Pb/204Pb)t (39·39–40·94) for their (206Pb/204Pb)t values (18·09–18·92). Their Pb isotope ratios do not correlate well with each other (Fig. 9d–f) or with either Nd (Fig. 9b) or Sr isotope ratios. This cannot be ascribed simply to over- or under-correction in the age adjustment for Pb as a result of alteration-related modification of U/Pb and/or Th/Pb. For example, values of 207Pb/204Pb have changed little since 84 Ma from radiogenic ingrowth (235U, the parent isotope of 207Pb, is only 0·7 atom% of 238U) and are thus insensitive to post-eruptive alteration of the U/Pb ratio; yet the full range of (207Pb/204Pb)t from 15·82–15·96 is observed among samples with closely similar Nd(t) values, between –6·5 and –7·2 (Fig. 9c).
|
Group B2
Like most of the B1 samples, the B2 rocks are evolved, with MgO between 3·55 and 6·60 wt %, Mg-number between 36 and 54, and 27–95 ppm Ni. However, they are lower in SiO2 (46·92–50·65 wt %) and mildly alkalic (Fig. 5a). They also are more LREE-enriched (Fig. 8g), with (La/Sm)ch = 3·0–3·7 and (La/Yb)ch = 11–14. Europium anomalies are absent. Values of (Nb/La)pm are higher (0·85–0·92) than in group B1, and the incompatible-element patterns lack Nb–Ta troughs (Fig. 8b). Strontium troughs are also lacking, and peaks at Pb vary from absent to modest; however, as with the B1 rocks, a sizeable trough at P is consistently present. Broadly similar patterns are displayed by some alkalic basalt dikes in the Ejeda–Bekily swarm. However, the dikes tend to have large peaks at Ba (Dostal et al., 1992
) that are not present in the B2 patterns.
The B2 basalts are also isotopically distinct from the Ejeda–Bekily dikes (Fig. 9). They differ from the B1 samples in having higher Nd(t) (+0·6 to –2·4) and lower (87Sr/86Sr)t (0·70579–0·70734). Ratios of (206Pb/204Pb)t (18·20–18·63) are within the B1 range but (207Pb/204Pb)t (15·66–15·70) is much lower; (208Pb/204Pb)t also extends to lower values (to 39·29).
Geochemistry of the silicic rocks
Most of the silicic samples are subaluminous; a few are mildly peralkaline. Like the basalts, they define two compositionally distinct groups (see Storey et al., 1997). Group R1 has much lower Th, Nb, Ta, Zr, Hf, Y, and HREE concentrations than R2 (e.g. Nb = 27–105 ppm in R1 vs 182–215 ppm in R2). The R1 samples also tend to have lower Na2O and higher TiO2, P2O5, and CaO for a given SiO2 content than the R2 samples (Fig. 7).
Group R1
Two compositional sub-types are present within R1: R1a and R1b. A diagnostic feature of R1a is a trough at Nb and Ta in Fig. 8c, with (Nb/La)pm = 0·16–0·68. In this important respect, the R1a incompatible-element patterns are similar to those of the B1 basalts. Large troughs at Ti, P, and Sr are present, as are a variably negative Eu anomaly and a prominent peak at Pb. Several R1a samples also have pronounced troughs at Ba. In addition to rhyolite with SiO2 > 70 wt %, R1a includes three samples of trachydacite with 65·00–66·29 wt % SiO2 (Fig. 5a). Trachydacite AND90-23 has an incompatible-element pattern closely resembling those of several of the rhyolites. The pattern of AND90-18 is similar, but exhibits the largest Nb–Ta trough of group R1a; this sample also has the highest (La/Yb)ch ratio and highest Zr, Hf, and MREE and LREE concentrations of the group.
The R1b samples are all rhyolites, and were found only in the Vohitsimbe section, interbedded with R1a and B1 units. They reach higher SiO2 and lower TiO2 and total iron than group R1a (Figs 5 and 7), are characterized by distinctly higher (Nb/La)pm (1· 3–4·2; Fig. 8d), and their incompatible element patterns display variable-size peaks at Zr and Hf relative to Nd and Sm, which are lacking in R1a patterns. Troughs at Ba are generally larger than seen among the R1a samples. So too, in many cases, are those at Ti, P, Sr, and Eu (all relative to Zr–Hf). Also, the R1b rhyolites are much less enriched in the LREE relative to the HREE [(La/Yb)ch = 0·97–3·4 for R1b vs 6·0–19 for R1a; see Fig. 8h and i], whereas concentrations of Yb, Lu, and Y tend to be somewhat higher than in most R1a samples at similar SiO2 (e.g. Fig. 7h).
The R1a isotope data form a trend that, with decreasing Nd(t), departs from the B1 array toward higher (87Sr/86Sr)t and reaches the lowest Nd(t) and by far the highest (87Sr/86Sr)t values of any of the Androy rocks (Fig. 9a). The most extreme values are exhibited by a dike on the southeastern side of the massif (AND90-6) and a flow within the upper part of the Rim Series on the northeastern side (AND90-83); this flow is stratigraphically the lowest of the rhyolite flows we sampled. These two rhyolites have closely similar Nd(t) = –17· 0 and –17·3 and (87Sr/86Sr)t = 0·73602 and 0·73622. They also have very similar age-corrected Pb isotope ratios, virtually identical incompatible-element patterns, and very similar major element compositions. The AND90-6 dike thus may be a feeder of the AND90-83 flow. Their (206Pb/204Pb)t and (207Pb/204Pb)t values are within the B1 basalt range, but their (208Pb/204Pb)t values (41·19, 41·39) are higher (Fig. 9b–e). The other three R1a samples we analyzed isotopically, including trachydacites AND90-18 and -23, have lower (208Pb/204Pb)t (40·18–40·58) and (87Sr/86Sr)t (0·7121–0·7194), and less negative Nd(t) (–7·5 to –11· 6). In this respect, they are more similar to the B1 basalts.
We measured isotope ratios for two R1b rhyolites (AND90-27, AND90-29), both from the upper levels at Vohitsimbe (Fig. 10). Values for both samples are very similar to those for the highest-Nd(t) R1a rhyolite (AND90-48, also from Vohitsimbe) in all panels of Fig. 9. Moreover, values for these three rhyolites are similar to those of four of the five Vohitsimbe B1 basalts we analyzed isotopically [e.g. Nd(t) = –7·5 to –7· 8 for the rhyolites and –6·5 to –7· 0 for the basalts]. The one B1 lava from a stratigraphic level below the Vohitsimbe section that we analyzed isotopically also has similar Nd(t) (–7·2) (Fig. 10).
|
Group R2
Although the R2 rhyolites have much higher concentrations of HREE, Th, Nb, Ta, Zr, Hf, and Y than group R1, their incompatible element patterns are similar in overall shape to those of the R1b samples, with peaks at Nb–Ta and Zr–Hf, and large troughs at Ba, Sr, P, Eu, and Ti (Fig. 8e). As with group R1b, the relative enrichment in the LREE (e.g. (La/Yb)ch = 2· 4–4·4) is significantly less than for the R1a samples (Fig. 8j).
The four R2 flows we analyzed isotopically are all from the Vohitsimbe section, and have closely similar age-corrected isotope ratios. These samples define a Rb–Sr isochron with initial (87Sr/86Sr)t = 0·71348 ± 0·00054 and an age of 84·0 ± 2·4 Ma (2); the mean square of weighted deviates (MSWD) is 2·3 (Fig. 11). No correlation of age-corrected (87Sr/86Sr)t with 1/Sr or 87Rb/86Sr is present, suggesting that the isochron is not a mixing line. Indeed, the Rb–Sr age is identical, within error, to Storey et al.'s (1995) 40Ar–39Ar age of 84·4 ± 1· 0 Ma (2) for sanidine from Vohitsimbe R1b rhyolite AND90-27. (Interestingly, a Rb–Sr age of 84·6 ± 1· 5 Ma is obtained from R1b samples AND90-27 and -29, R1a sample AND90-48, and B1 samples AND90-23 and -49B, all of which are from the Vohitsimbe section and have, in addition to Sr, similar age-corrected Nd and Pb isotope ratios; however, the MSWD is large, at 11· 5.)
|
Age-corrected
Nd(t) values of three of the R2 samples, AND90-34, -43, and -53, are identical within error (–2·4 to –2·6) and that for AND90-41 (–2·0) is only slightly higher. However, unlike their measured 87Sr/86Sr and 87Rb/86Sr values, variation in measured 143Nd/144Nd and 147Sm/144Nd is small, so that no Sm–Nd isochron is developed. In Fig. 9a, the age-corrected Nd–Sr isotope data for the R2 rhyolites cluster to the right of the array formed by the BI and B2 basalts. Age-corrected Pb isotope ratios cover a slightly larger total range than the present-day values (e.g. 0·10 vs 0·05 for 208Pb/204Pb), suggesting some relatively recent disturbance of U/Pb and Th/Pb ratios in some of the R2 samples, probably as a result of weathering-related alteration (e.g. Rosholt et al., 1971; Doe et al., 1982). Nevertheless, the total variation in age-corrected (206Pb/204Pb)t, (207Pb/204Pb)t, and (208Pb/204Pb)t is small: 18·61–18·71, 15·80–15·82, and 39·78–40·00, respectively (Fig. 9b–e). |
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONAGE AND FORM OF...SAMPLING AND ANALYTICAL METHODSRESULTSDISCUSSIONCONCLUDING REMARKSSUPPLEMENTARY DATAAPPENDIX: MINERAL-LIQUID...REFERENCES |
|---|
Origin of the basaltic rocks
A substantial role for continental components in the genesis of group B1 is indicated by the combination of variable but low
Nd(t), high (87Sr/86Sr)t, (207Pb/204Pb)t, and (208Pb/204Pb)t, and incompatible element patterns with Nb–Ta troughs and Pb peaks. The same is true for tholeiitic basalts in southwestern Madagascar (Mahoney et al., 1991; Dostal et al., 1992). Assuming the B1 magmas originated from a relatively homogeneous mantle source with higher Nd than measured for any of our samples, the Pb isotope data imply considerable heterogeneity in 206Pb/204Pb in the low-Nd continental material that affected the magmas (Fig. 9b). In contrast, the coherent array in Fig. 9a indicates relative Nd and Sr isotope homogeneity for this material. The isotopic and chemical characteristics of basement crust and lithospheric mantle in this region of Madagascar are essentially unknown. However, a rough positive correlation is present between (87Sr/86Sr)t and SiO2, and a negative one between Nd(t) and SiO2 (Fig. 12a), suggesting that the low-Nd material was crust rather than lithospheric mantle. Some indirect support that it was not mantle peridotite may be provided by the R1 rhyolites, which are interbedded with the B1 basalts, appear to be related petrogenetically to them (see below), and reach even lower Nd(t); rhyolites do not form by melting of peridotite (e.g. Yoder, 1973). Because B1 (and B2) dikes are present in and around the Androy massif, the basalt flows were probably erupted from relatively nearby feeders. Given that basement exposed in the vicinity is largely Archean, the continental material affecting the B1 basalts is also likely to have been Archean.
|
Isotope ratios among the B1 basalts do not correlate with Mg-number (Fig. 12b), implying that no particular impediment existed to post-contamination differentiation of B1 magmas and/or that contamination could occur at different stages of differentiation (for example, contamination might occur at a deeper crustal level, followed by magma movement to a shallower level where further fractional crystallization could occur). The apparent abundance of B1 basalts with
Nd(t) near –7 in the thick Vohitsimbe section (Fig. 10) may reflect an open magmatic system in which recharge, fractionation, and contamination reached a near-steady state (e.g. O'hara & Mathews, 1981; Wooden et al., 1993; Peng et al., 1994) during the later stages of Androy volcanism. That the lowest-Nd Androy rocks are rhyolites and that rhyolites underlie and are interbedded with the B1 basalts even raises the question of whether contamination of B1 magmas could have occurred through admixture with low-Nd R1-type silicic liquids rather than by bulk or selective assimilation of crustal wall rock. Indeed, the B1 and R1a data define a negative correlation of Nd(t) with ratios such as (Nb/La)pm and (Nb/Pb)pm (Fig. 12c and d), qualitatively consistent with such mixing. However, leaving aside questions of the physical feasibility of such a process on the large scale appropriate for Androy (e.g. Snyder, 2000), we can reject mixing between B1 magmas and the lowest-Nd R1-type liquids as the main contamination mechanism. For example, although the R1 rhyolites reach higher (208Pb/204Pb)t and (87Sr/86Sr)t than the basalts, the basalts attain higher (207Pb/204Pb)t than any of the rhyolites (Fig. 9).
The characteristics of group B2—higher Nd(t) and lower (87Sr/86Sr)t, (207Pb/204Pb)t, (208Pb/204Pb)t, and SiO2 than B1, incompatible element patterns that lack Nb–Ta troughs and lack significant Pb peaks—suggest that the B2 basalts are less contaminated than the B1 rocks. The stratigraphic location of the B2 flows below the B1 flows suggests the possibility that B2 magma could have been the high-Nd(t) mixing end-member for group B1. Such a relationship is permitted by our isotopic results but not supported by the chemical data. For example, although the B2 and B1 data lie along a single trend in Fig. 12c, they diverge at a high angle in Fig. 12d. Likewise, B1 sample AND90-1 and B2 sample AND90-78 have similar Nd isotope ratios (within 0·9 Nd units of each other), but the B1 rock has much lower concentrations of incompatible elements (Table 2) and an incompatible-element pattern with a flatter overall slope from Lu to Rb. With the exception of Nb and Ta, contamination by typical Archean mafic to felsic crust (e.g. Rudnick & Fountain, 1995) would increase, not decrease, the magmatic concentrations of the highly incompatible elements and would not flatten the slope of the pattern from Lu to Rb. However, it remains possible that the B2 magmas were generated by smaller amounts of partial melting in the same mantle source as the B1 basalts. The overall slopes (although not all the details) of the incompatible element patterns of the B2 basalts and least contaminated, high-Nd(t) B1 basalts are matched reasonably well by model patterns that assume a single source and different amounts of partial melting in the garnet and spinel facies (e.g. Fig. 13).
|
Were any of the Androy basalts derived from Marion hotspot mantle? As noted above, plate reconstructions suggest that the hotspot was near this part of Madagascar at 84 Ma (Fig. 1b). The two picritic high-MgO B1 samples (AND90-1 and -7) with forsteritic olivine (up to Fo91· 8) are consistent with a (thermal) plume source, and the high NiO content of their olivine phenocrysts could suggest a source relatively rich in pyroxenite (see Sobolev et al., 2005
). None of the Androy samples have isotopic signatures resembling those of recent products of the Marion hotspot; nor are they similar to those of the Crozet hotspot, 500 km east of Marion. However, the B2 basalts possess more ocean-island-like isotopic ratios and incompatible element patterns than those of group B1, and in all panels of Fig. 9 the B2 data points, particularly those of the two highest-Nd(t) samples, lie broadly between the B1 (and rhyolite) data and the fields for Marion and Crozet. The same is true in Fig. 12c, which shows Nd(t) vs (Nb/La)pm; the Nb/La ratio is relatively insensitive to variation in the amount of mantle melting, as Nb and La have very similar bulk distribution coefficients in most mantle rock types. Thus, it is possible that Androy magmas contained a hotspot-derived component, although the isotopic data certainly do not require this.
Significantly, the B2 incompatible-element patterns all exhibit a prominent trough at P (Fig. 8b). Melting in the garnet stability field may produce a P trough (Haggerty et al., 1994); however, modern Marion hotspot lavas, which appear to have formed in the presence of residual garnet (Storey et al., 1997; Janney et al., 2005), lack P troughs. Moderate to large amounts of magmatic contamination by many types of continental crust can produce a trough at P, but also commonly generate a large peak at Pb, a Nb–Ta trough, and sometimes a trough at Sr or Ti (e.g. Cox & Hawkesworth, 1984; Wooden et al., 1993; Peng et al., 1994), none of which are present in the B2 patterns. Weathering-related alteration can redistribute P in basalts (e.g. Fodor et al., 1992) and, as noted above, the basalts with the highest LOI values exhibit greater variability in P content. However, despite the variable alteration in the B2 samples (e.g. LOI ranges from 0·24 to 2·81 wt % among the B2 samples analyzed for trace elements), the size of the P trough is rather constant. Phosphorus troughs can also develop by removal of apatite during differentiation, but P increases with decreasing MgO in both groups B2 and B1 (Fig. 5e), implying little removal of apatite from the basaltic magmas; none was seen in thin section. We conclude that the P trough is probably largely a mantle source feature. If so, the B2 basalts were derived from a source at least partly distinct from that feeding recent Marion hotspot magmas. The difference could indicate a component in the hotspot source during the middle Cretaceous that is no longer present. Alternatively, as proposed for the alkalic dikes in the Ejeda–Bekily swarm (Mahoney et al., 1991; Dostal et al., 1992), the B2 basalts may represent partial melts of lithospheric mantle, or hotspot-derived magmas that interacted significantly with such mantle. Interestingly, most of the alkalic Ejeda–Bekily dikes have a P trough, although their isotopic differences from the B2 basalts (Fig. 9) indicate a separate petrogenesis. Phosphorus troughs are also a feature of all the B1 patterns (e.g. Fig. 8a), supporting the possibility that groups B1 and B2 are related by different amounts of partial melting of a common source.
Origin of the rhyolites
Rhyolites in predominantly basaltic continental volcanic provinces are attributed to some combination of extreme crystal fractionation (with or without assimilation of wall rock) and crustal melting (with or without mixing of melts with more mafic mantle-derived magmas). In the category of crustal melting, both melting of old crustal rock and of solidified basalt intruded within or underplated to the crust during the same volcanic episode can occur (e.g. Sethna & Battiwala, 1974; Doe et al., 1982; Cleverly et al., 1984; Bellieni et al., 1986; Lightfoot et al., 1987; Harris et al., 1990; Leeman et al., 1992; Garland et al., 1995; Cameron et al., 1996; Harris & Milner, 1997; Kirstein et al., 2000; Riley et al., 2001; Miller & Harris, 2007). Modeling indicates that large amounts of rhyolite can be generated when basalt repeatedly invades a region of crust over a period of 105–106 years, and suggests that partial melting of basalt and/or older crustal wall rock will typically occur together with the simultaneous generation of highly fractionated liquids by crystallization of basaltic magma (Annen & Sparks, 2002; Annen et al., 2006). Evidence from O isotopes, U-series isotopes, and combined major and trace element compositions in Iceland, where old continental crust is not a factor, indicates that either fractionation or melting can be the dominant process, depending on the volcanic system (e.g. Sigmarsson et al., 1991; Furman et al., 1992; Jónasson, 1994; Selbekk & Trønnes, 2007).
Group R1
As with the basalts, we infer from the presence of R1 dikes that the R1 flows were erupted from relatively nearby vents. The overlap of the Vohitsimbe R1a and R1b isotope ratios with B1 values, the Nb–Ta troughs in the R1a and B1 incompatible-element patterns, and the interbedding of B1 and R1 flows strongly suggest that group R1 is related petrogenetically to group B1. A petrogenetic association is also suggested by element ratios such as Nb/Pb (e.g. Fig. 12d), which should not be affected much by fractionation of most of the minerals observed in these rhyolites (i.e. plagioclase, Fe–Ti oxides, clinopyroxene, apatite, and quartz). Within the Vohitsimbe section, four of the five B1 samples and all three R1 samples that we analyzed isotopically possess similar age-corrected Nd, Sr, and Pb isotope ratios; for example, the full range of Nd(t) measured is only 1· 3 units (from –6·5 to –7· 8) (Fig. 10). These R1 rhyolites are therefore likely to have been derived by advanced crystallization of B1 magma or remelting of earlier-intruded B1 basalt followed by fractionational crystallization, with relatively little additional assimilation of old crustal wall rock.
Neither process is constrained very strongly by our data. No isotopically equivalent rocks were found among the few samples of intermediate composition, and major element modeling, in particular, is hampered by the alteration effects discussed above. However, simple major-element mass balance following Stormer & Nicholls (1978) suggests that 75–95% crystal fractionation of an assemblage consisting of plagioclase, clinopyroxene, magnetite and/or ilmenite, alkali feldspar, and apatite would be needed to produce the Vohitsimbe R1 rhyolites from the B1 basalts (the lowest residuals are achieved by assuming an intermediate stage corresponding to one of the rare trachydacites, but it should be remembered that the Nd(t) values of the trachydacites (–9·4, –11· 6) indicate a significantly greater crustal component than in the Vohitsimbe R1 rhyolites). The scarcity of intermediate compositions (Fig. 5a), termed the ‘silica gap’, is a feature common to many basalt–rhyolite associations, and often taken as evidence for rhyolite formation by melting of basaltic precursors. However, it also may simply reflect the onset of significant amounts of magnetite fractionation, which can cause a rapid increase in SiO2 content (Garland et al., 1995). Indeed, many of the rhyolites contain magnetite phenocrysts, and diagrams of Nb vs Zr and Y, elements that are incompatible in magnetite, show a nearly continuous trend for the B1 and R1a samples (Fig. 14).
|
Simple trace-element models provide some useful insights. We were unable to produce model incompatible-element patterns resembling those of the R1 rhyolites by assuming partial melting of solidified B1 basalt with no subsequent fractionation. However, the key features of the Vohitsimbe R1a patterns can be produced at least semi-quantitatively by fractional crystallization of isotopically similar B1 magmas (e.g. Fig. 15a). Partial melting of B1 basalt followed by fractional crystallization of an assemblage of minerals present in these rocks can also yield model patterns with the main features of the Vohitsimbe R1a patterns, but only if the basalt is less evolved than our isotopically appropriate samples (e.g. Fig. 15b). Fits of similar quality are obtained for a range of postulated mineral proportions in the source. No hydrous phases were used in the modeling, as primary hydrous minerals are lacking in both the R1 rhyolites and basalts; nor were any exotic accessory phases included. The total percentages of fractional crystallization and partial melting are in the ranges estimated in other studies (e.g. Cleverly et al., 1984
; Furman et al., 1992; Jónasson et al., 1990; Jónasson, 1994; Cameron et al., 1996; Melluso et al., 2005).
|
We stress that the examples shown in Fig. 15 should be viewed as illustrative only; they are by no means unique, and such modeling clearly is far too simple to be strictly correct. It assumes a single bulk fractionating assemblage and a single set of constant mineral–melt distribution coefficients, whereas both the fractionating minerals and distribution coefficients change during evolution from basaltic to rhyolitic compositions. Published distribution coefficients for a given rock type vary considerably for some elements (see the Appendix for the values used). Also, even ignoring post-eruptive changes caused by alteration, the basalt samples used in constructing the model patterns are highly unlikely to represent the specific starting compositions from which the actual rhyolites were derived. Errors in any of the above are propagated, and in some cases magnified, through the calculations. Thus, it is difficult to determine whether an origin by fractional crystallization or by partial melting of B1 basalt followed by fractional crystallization is more likely for the Vohitsimbe R1a rhyolites. We note, however, that the amounts of partial melting required in the models are all rather high (25% in the example illustrated in Fig. 15b). The need for a suitable heat source may pose a difficulty in this case, given the >100 km distance of the Androy area from the zone of major Cretaceous rifting. Indeed, although the analogy with Iceland is imperfect, partial melting of basalt is more important in the formation of Icelandic rhyolites erupted near rift zones, whereas fractional crystallization tends to dominate for rhyolites erupted far from the rift zones (e.g. Selbekk & Trønnes, 2007
).
What about the R1b rhyolites? Their incompatible-element patterns are distinctly unlike those of isotopically similar R1a or B1 samples, most notably at Nb–Ta and Zr–Hf, and in the lesser enrichment of the LREE and MREE relative to the HREE (Fig. 8). Their higher average SiO2 and the lack of plagioclase phenocrysts in the samples we analyzed by electron microprobe indicate that R1b tends to be more evolved than R1a. Samples AND90-31 and -32 provide an important clue. They were taken from different parts of the same flow, yet AND90-31 has R1b characteristics and AND90-32 is an R1a rhyolite. The differences in the REE suggest control by a REE-rich mineral. Zircon and apatite are two of the more common such phases, and were observed in some of the Androy samples. The large troughs at P in the R1a and R1b (and R2) patterns in Fig. 8 provide strong evidence of apatite control. Similarly, the nearly constant Zr with increasing SiO2 content among the combined R1a and R1b data at SiO2 > 70 wt % in Fig. 7g is consistent with at least a minor role for zircon removal in the rhyolite stage of magma genesis; so, too, is the position of several R1b data points to the left of and above the B1–R1a array in Fig. 14b. However, zircon fractionation cannot produce the distinctive humps at Zr–Hf (Fig. 8d) or, because of zircon's greater affinity for HREE vs LREE, the relatively flat slopes of the R1b REE patterns (Fig. 8i). Apatite can produce features qualitatively similar to those observed, but unrealistically large amounts of apatite removal are required; much greater, for example, than permitted by the P2O5 contents of the Androy samples. However, as illustrated in Fig. 15c, removal from AND90-32 of a tiny amount (0·13%) of allanite, a REE-rich accessory phase common in silicic igneous rocks, produces a pattern very similar to that of AND90-31 (except for Ce, which, as noted above, may have been lowered by alteration in AND90-32). Likewise, Fig. 15d shows that removal of an even smaller amount (0·04%) of allanite from R1a sample AND90-48 produces a pattern closely resembling that of isotopically similar R1b sample AND90-29. REE concentrations in the rhyolites are in the range estimated for saturation of allanite (Miller & Mittlefehldt, 1983). We did not observe allanite (or monazite or other, more exotic accessory minerals) in thin section, but detection of such phases is difficult because they are never present in more than trace amounts, typically as tiny crystals lining or included in the major phenocrysts (e.g. Miller & Mittlefehldt, 1983; Wolff & Storey, 1984; Michael, 1988). As such, trace minerals such as allanite may largely have been retained along with most of the phenocrysts in the rhyolite magma source.
The R1 samples from stratigraphic levels below the Vohitsimbe section have more negative Nd(t) (–9·4 to –17·3) than measured in the Vohitsimbe R1 rhyolites. Although the actual end-member compositions are unknown, extrapolation of the B2–B1–R1a array in Fig. 12c and the B1–R1 array in Fig. 12d to global average (Nb/La)pm and (Nb/Pb)pm values of Archean crust suggests a very rough lower limit on Nd of about –25 for the crustal end-member. However, the offset of the lowest-Nd(t) R1 data to much greater (87Sr/86Sr)t than the lowest-Nd(t) B1 data (Fig. 9a) is accompanied by significantly higher (208Pb/204Pb)t, yet lower (207Pb/204Pb)t, relative to (206Pb/204Pb)t (Fig. 9d and e). These Pb isotope differences indicate that the offset cannot be caused only by differences in coupled fractionation and assimilation during rhyolite vs basalt generation. Rather, these R1 rocks must contain a component of old crust isotopically somewhat different from the low-Nd material affecting the B1 basalts.
Without calling upon any input at all from a basaltic precursor, model fractionated melts of global average Archean intermediate crust yield incompatible-element patterns with characteristics broadly resembling those of the two lowest-Nd(t) rhyolites; that is, those with Nd(t) of –17· 0 and –17·3 (e.g. Fig. 16). The assumptions used in constructing the model pattern in Fig. 16 (40% melting followed by 40% fractional crystallization of plagioclase, clinopyroxene, magnetite, and a small amount of apatite, using the same mineral distribution coefficients as for Fig. 15) are again clearly oversimplifications; also, the specific crustal material involved clearly was not global average Archean crust. Nevertheless, the result suggests that Archean crust may be the overwhelmingly dominant component in the two lowest-Nd rhyolites. Although both orthogneiss and paragneiss are present in the regional basement (Dissanayake & Chandrajith, 1999; Kröner et al., 1999; de Wit et al., 2001), the low Al2O3 content (11 wt %) of these rhyolites favors orthogneiss over paragneiss or other aluminous rock of sedimentary origin for this component (e.g. Thy et al., 1990; Beard et al., 1994).
|
The trachydacites contain less of this crustal component, as indicated by their
Nd(t) values (–9·4 and –11· 6), but more of it than the Vohitsimbe R1 rhyolites. The exact petrogenetic relationship of the trachydacites to the other R1 rocks is unclear, in part because there are so few trachydacites. Data for sample AND90-23 generally lie along or near the ends of trends described by the R1a rhyolites in chemical variation diagrams, broadly consistent with this rock representing a less advanced degree of fractional crystallization in the R1a series. The same appears to be true for the more limited data we have for AND90-22. In contrast, as noted above, AND90-18 is extreme in that it has a deeper Nb–Ta trough, greater (La/Yb)ch, and greater concentrations of Zr, Hf, and the MREE and LREE than any of the R1a rhyolites (e.g. Fig. 7g). The high Zr and Hf could indicate a small amount of ‘excess’ zircon in the split of this sample that we used for chemical analysis, but this cannot explain the other features of this unusual sample.
Group R2
Data for the R2 rhyolites are also offset to higher (87Sr/86Sr)t at similar Nd(t) relative to the isotopic array defined by the B2 and B1 basalts (Fig. 9a). Unlike the low-Nd R1 samples, however, the R2 data lie consistently between values for the B2 and B1 basalts in Pb–Pb and Nd–Pb isotope space. This feature suggests some sort of petrogenetic relationship to the basalts, and that low-Nd material different from that affecting the basalts is not necessarily required for the R2 rhyolites. That is, their high relative (87Sr/86Sr)t may be a result of coupled assimilation and fractional crystallization in which Sr behaved as a more compatible element than in the basaltic magmas and/or in which the ratio of crystallization rate to assimilation rate was higher. The poorly constrained variables noted in the preceding paragraphs hinder meaningful modeling, but the general effect would be to produce higher (87Sr/86Sr)t for a given Nd(t) and Pb isotope ratio than in the basalts (James, 1981; DePaolo, 1981). The reason is that Sr has a significantly larger bulk distribution coefficient in silicic systems than in basaltic ones, because of the increased importance of feldspar in the fractionating mineral assemblage and because Sr becomes increasingly compatible in feldspar (and several other minerals) as magmatic silica content increases (e.g. Irving, 1978). The deep troughs at Sr in the R2 (and R1) incompatible-element patterns testify that Sr indeed behaved as a rather strongly compatible element during formation of the rhyolites.
The interbedding of R2 and B1 flows at Vohitsimbe (e.g. Fig. 10) suggests the possibility that the R2 rhyolites could be derived from B1 basalt. However, Nd(t) in the R2 samples is as high as or higher than the highest B1 values; thus, the R2 rhyolites cannot have originated by any combination of assimilation of typical (i.e. low-Nd) Archean crust and either fractionation or melting of B1 basalt, at least not B1 basalt like any of our samples. If B1-type basalt with higher Nd(t) was available, could it have been parental to the R2 rhyolites? We were unable to generate R2-like incompatible-element patterns by modeling crystal fractionation of B1 compositions, with or without an added Archean crustal component, principally because concentrations of highly incompatible elements in group B1 are too low relative to those of the less incompatible elements. In contrast, model patterns exhibiting the main features of the R2 patterns can be obtained from fractionated melts of B1 basalt with similar Nd(t) (–3·3 in AND90-1) to the R2 values (–2·0 to –2·6; Fig. 17a). With the distribution coefficients used (Appendix), the chief misfit is at Sr. As with the R1b rhyolites, the high relative Nb–Ta and Zr–Hf of group R2 can be accounted for by removal of a very small amount of allanite. On this basis, it is possible that the R2 rhyolites were derived from partial melts of high-Nd B1 basalt, coupled with rather extensive fractionation and assimilation of relatively small amounts of ancient crust.
|
Group B2's higher
Nd(t) values and incompatible-element concentrations, together with its mildly alkalic nature, might seem to make it a better candidate than group B1 for generating the R2 rhyolites. However, combined melting and fractionation models involving <30% partial melting of B2 basalt and assimilation of 0–30% of average Archean mafic, intermediate, or felsic crust did not provide good matches for the R2 incompatible-element patterns. In particular, highly incompatible elements tended to be over-enriched whereas HREE contents were consistently too low. Better fits were achieved by increasing the amount of partial melting to >50%, but melt percentages greater than 30% are thought to cause disintegration of the basalt protolith, yielding a mafic mush rather than silicic liquid (e.g. Jónasson, 1994). In contrast, advanced crystal fractionation of B2 magma, with or without a small amount of Archean crust, produces a much better fit to the R2 patterns, although HREE concentrations again tend to be rather low (e.g. Fig. 17b). In short, the trace element evidence appears approximately equally permissive of either a B1-melting (followed by crystal fractionation) origin or a B2-fractionation origin for group R2. A difficulty with deriving the R2 rhyolites via fractionation of B2 basalt, however, is that the B2 flows are confined to the lowermost portion of the volcanic stratigraphy, whereas the R2 rhyolites are in the upper part, within a thick sequence of B1 and R1 flows. Thus, B2-type magma would have had to be present, but unerupted, throughout the Androy episode or else reappear toward the end of volcanism, perhaps in response to decreasing amounts of partial melting in the mantle source. If so, such magmas must not have interacted much with B1 magmas, which were present during eruption of the R2 rhyolites (e.g. Fig. 10).
|
CONCLUDING REMARKS |
|---|
|
TOPABSTRACTINTRODUCTIONAGE AND FORM OF...SAMPLING AND ANALYTICAL METHODSRESULTSDISCUSSIONCONCLUDING REMARKSSUPPLEMENTARY DATAAPPENDIX: MINERAL-LIQUID...REFERENCES |
|---|
The fact that B1 basalts, R1 rhyolites, and R2 rhyolites are interbedded appears to require the existence of separate, simultaneously active magmatic plumbing systems, at least at the depths where rhyolite generation was occurring. Likewise, at least some interbedding exists between the B2 and R1 flows. We did not find any R2 dikes, but the presence of R1, B1, and B2 dikes indicates that the basalt and R1 rhyolite flows are probably locally derived, which in turn suggests that the different feeder systems were located fairly close to one another. Unfortunately, the mineral assemblages present in the Androy rocks are not very useful for estimating depths of magma equilibration. The normative compositions of the rhyolites are variable, partly because of alteration, but in the rhyolite tetrahedron (Tuttle & Bowen, 1958
) the majority of values for both groups R1 and R2 plot between the 0 and 0·5 GPa cotectics, qualitatively consistent with equilibration at upper-crustal depths. The saucer shape of the Androy structure is consistent with at least one sizeable magma reservoir in the upper crust, whereas the lack of classic caldera-collapse features appears to make a large reservoir at very shallow depths unlikely.
Androy represents the greatest concentration of rhyolite in the Cretaceous system of Madagascar, but far greater volumes of rhyolite were erupted in the Karoo (>35 000 km3; Cleverly et al., 1984) and Paraná (>16 000 km3; Bellieni et al., 1986; Garland et al., 1995) flood basalt provinces. In both provinces, the rhyolites were formed near the end of volcanism and appear to have erupted largely from zones where crustal extension was greatest. Most of the rhyolite production is attributed to thinning-related decompression and melting of solidified basalt that was underplated earlier to the lower crust. A similar origin may apply for the rhyolites on the central eastern coast of Madagascar, where dikes indicate pre-breakup crustal extension of as much as 20% (M. Storey et al., unpublished data). Like the Karoo and Paraná rhyolites, the Androy rhyolites were formed near the end of the volcanic episode (Storey et al., 1995). However, the Androy massif lies some 100–150 km inland from the rifted margin in a region where extension was limited. Indeed, no major dike swarm transects this area, although the trends of three major swarms converge to the east (Ernst & Buchan, 1997). We infer that extension of basement crust in the Androy area was sufficient to permit repeated intrusion of basaltic magma but limited enough to promote development of relatively stable crustal magma networks and reservoirs where formation of rhyolite could occur through a combination of melting of basalt and magmatic differentiation.
Why some flood basalt provinces produce voluminous rhyolite whereas others (e.g. the Siberian and Deccan Traps) produce only minor rhyolite may ultimately be related to the supply rate of basaltic magma. In many provinces, large-volume flow units of differentiated basalt, commonly in excess of 1000 km3 (and perhaps as large as 10 000 km3; Self et al., 2006), testify to very large, basalt-dominated magma reservoirs. In such systems, formation of rhyolite, by assimilation or fusion of crust or by crystal fractionation, is likely to be swamped by the basaltic component. Only when the input of basaltic magma slows or temporarily stops may more silicic melts begin to dominate a system. However, such melts are likely to be incorporated in the next large batches of mantle-derived melts. Thus, normally we would expect to see eruption of crustally contaminated basalts, as in the Deccan and Siberian Traps. Until magmatism as a whole begins to die down (for example, when a region moves away from a hotspot), rhyolitic magmas are unlikely to be produced or, if they are, will tend be incorporated in the more dominant basalt component. The amount of mantle melt that is generated will be a function of lithospheric extension, as well as source temperature and source composition. In southeastern Madagascar, relatively small amounts of extension were occurring near a probable plume, similar to the situation in the rhyolite-rich Snake River Plain and Yellowstone areas of North America. Fewer rhyolites are found along the rifted eastern margin of Madagascar, where extension was far greater but basalt is abundant. Much of this basalt shows evidence of significant crustal contamination (Storey et al., 1997; J. Mahoney, unpublished data).
Finally, the reason a hotspot-type isotopic signature like that seen in many of the southeastern coastal basalts is lacking at Androy—even though plate reconstructions indicate that the area was near the Marion hotspot at 84 Ma—may also be related to the location of the massif in an area of limited extension. Extension-related thinning and removal of lithospheric mantle was probably less than it was closer to the rifted eastern margin of Madagascar (e.g. see Storey et al., 1997, Fig. 13), leaving more potentially fusible material in the lithospheric mantle beneath Androy and/or permitting greater interaction between lithospheric mantle and magmas ascending from the convecting mantle.
|
SUPPLEMENTARY DATA |
|---|
|
TOPABSTRACTINTRODUCTIONAGE AND FORM OF...SAMPLING AND ANALYTICAL METHODSRESULTSDISCUSSIONCONCLUDING REMARKSSUPPLEMENTARY DATAAPPENDIX: MINERAL-LIQUID...REFERENCES |
|---|
Supplementary data for this paper are available at Journal of Petrology online.
|
APPENDIX: MINERAL–LIQUID DISTRIBUTION COEFFICIENTS USED FOR RHYOLITE MODELING |
|---|
|
TOPABSTRACTINTRODUCTIONAGE AND FORM OF...SAMPLING AND ANALYTICAL METHODSRESULTSDISCUSSIONCONCLUDING REMARKSSUPPLEMENTARY DATAAPPENDIX: MINERAL-LIQUID...REFERENCES |
|---|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
For plagioclase, the coefficients for Sr and Eu shown in parentheses were used for modeling fractional crystallization to rhyolite compositions; coefficients not in parentheses were used for modeling partial melting of mafic or intermediate crust. SC, stoichiometric constituent. Distribution coefficients are from the GERM website (GERM, 2006; http://www.earthref.org).
|
ACKNOWLEDGEMENTS |
|---|
We are grateful to A. Razafiniparany for invaluable assistance in arranging permits for the field program, to R. N. Wilson for assistance with the microprobe analysis, and to D. VonderHaar and N. Hulbirt for help with other aspects of the work. We thank W. Le Masurier for useful comments, and J. Geldmacher, C. Harris, and L. Melluso for detailed, thoughtful reviews. Funding was provided by the Natural Environment Research Council, UK (GR3/7484) and US National Science Foundation (EAR-9104897).
*Corresponding author. Telephone: 1-808-956-8705. E-mail: jmahoney{at}hawaii.edu
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONAGE AND FORM OF...SAMPLING AND ANALYTICAL METHODSRESULTSDISCUSSIONCONCLUDING REMARKSSUPPLEMENTARY DATAAPPENDIX: MINERAL-LIQUID...REFERENCES |
|---|
Annen C, Sparks RSJ. Effects of repetitive emplacement of basaltic intrusions on thermal evolution and melt generation in the crust. Earth and Planetary Science Letters (2002) 203:937–955.[CrossRef][Web of Science]
Annen C, Blundy, Sparks RSJ. The genesis of intermediate and silicic magmas in deep crustal hot zones. Journal of Petrology (2006) 47:505–539.
Barker F. Inroduction to special issue on granites and rhyolites: a commentary for the nonspecialist. Journal of Geophysical Research (1981) 86:10131–10135.
Battistini R. La structure du massif volcanique de l’Androy (Madagascar). Bulletin de la Société Géologique de France (1959) 7:187–191.
Beane JE, Turner CA, Hooper PR, Subbarao KV, Walsh JN. Stratigraphy, composition and form of the Deccan basalts, Western Ghats, India. Bulletin of Volcanology (1986) 48:61–83.[CrossRef]
Beard JS, Lofgren GE, Sinha AK, Tolio RP. Partial melting of apatite-bearing charnockite, granulite, and diorite: melt compositions, restite mineralogy, and petrologic implications. Journal of Geophysical Research (1994) 99:21591–21603.[CrossRef]
Bellieni G, Comin-Chiaramonti P, Marques LS, Melfi AJ, Nardy AJR, Papatrechas C, Piccirillo EM, Roisenberg A, Stolfa D. Petrogenetic aspects of acid and basaltic lavas from the Paraná plateau (Brazil): mineralogical and petrochemical aspects. Journal of Petrology (1986) 27:915–944.
Besairie H. Madagascar, Carte Géologique. (1964) Tananarive: Service Géologique de Madagascar.
Cameron KL, Parker DF, Sampson DE. Testing crustal melting models for the origin of flood rhyolites: a Nd–Pb–Sr isotopic study of the Tertiary Davis Mountains volcanic field, west Texas. Journal of Geophysical Research (1996) 101:20407–20422.[CrossRef]
Cleverly RW, Betton PJ, & Bristow J. Geochemistry and petrogenesis of the Lebombo rhyolites. Petrogenesis of the Volcanic Rocks of the Karoo Province—Erlank AJ, ed. (1984) 13. Geological Society of South Africa Special Publication. 171–194.
Cox KG, Hawkesworth CJ. Relative contributions of crust and mantle to flood basalt magmatism, Mahabaleshwar area, Deccan Traps. Philosophical Transactions of the Royal Society of London, Series A (1984) 310:627–641.[CrossRef]
DePaolo DJ. Trace element and isotopic effects of combined wallrock assimilation and fractional crystallization. Earth and Planetary Science Letters (1981) 53:189–202.[CrossRef][Web of Science]
de Wit MJ, Bowring SA, Ashwal LD, Randrianasolo LG, Morel VPI, Rambeloson RA. Age and tectonic evolution of Neoproterozoic ductile shear zones in southwestern Madagascar, with implications for Gondwana studies. Tectonics (2001) 20:1–45.[CrossRef][Web of Science]
Dissanayake CB, Chandrajith R. Sri Lanka–Madagascar Gondwana linkage: evidence for a Pan-African mineral belt. Journal of Geology (1999) 107:223–235.[CrossRef][Web of Science]
Doe BR, Leeman WP, Christiansen RL, Hedge CE. Lead and strontium isotopes and related trace elements as genetic tracers in the upper Cenozoic rhyolite–basalt association of the Yellowstone Plateau volcanic field. Journal of Geophysical Research (1982) 87:4785–4806.
Dostal J, Dupuy C, Nicollet C, Cantagrel JM. Geochemistry and petrogenesis of Upper Cretaceous basaltic rocks from southern Malagasy. Chemical Geology (1992) 97:199–218.[CrossRef][Web of Science]
Ernst RE, Buchan KL. Giant radiating dyke swarms: their use in identifying pre-Mesozoic large igneous provinces and mantle plumes. Large Igneous Provinces: Continental, Oceanic, and Planetary Flood Volcanism—Mahoney JJ, Coffin MF, eds. (1997) 100. American Geophysical Union Monograph. 297–333.
Fodor RV, Frey FA, Bauer GR, Clague DA. Ages, rare-earth element enrichment, and petrogenesis of tholeiitic and alkalic basalts from Kahoolawe Island, Hawaii. Contributions to Mineralogy and Petrology (1992) 110:442–462.[CrossRef][Web of Science]
Furman T, Frey FA, Meyer PS. Petrogenesis of evolved basalts and rhyolites at Austurhorn, southeastern Iceland: the role of fractional crystallization. Journal of Petrology (1992) 33:1405–1445.
Garland FE, Hawkesworth CJ, Mantovani MSM. Description and petrogenesis of Paraná rhyolites. Journal of Petrology (1995) 36:1193–1227.
GeoReM. Geological and Environmental Reference Materials Data Base. (2006) http://georem.mpch-mainz.gwdg.de/start.asp.
GERM. Geochemical Earth Reference Model. Partition coefficient (Kd) database. (2006) http://www.earthref.org.
Gunnarsson B, Marsh BD, Taylor HP. Generation of Icelandic rhyolites: silicic lavas from the Torfajökull central volcano. Journal of Volcanology and Geothermal Research (1998) 83:1–45.[CrossRef][Web of Science]
Haggerty SE, Fung AT, Burt DM. Apatite, phosphorus and titanium in eclogitic garnet from the upper mantle. Geophysical Research Letters (1994) 21:1699–1702.[CrossRef][Web of Science]
Harris C, Milner S. Crustal origin for the Paraná rhyolites: discussion of ‘Description and petrogenesis of the Paraná rhyolites, southern Brazil’ by Garland et al. Journal of Petrology (1997) 38:299–302.[CrossRef][Web of Science]
Harris C, Whittingham AM, Milner SC, Armstrong RA. Oxygen isotope geochemistry of the silicic volcanic rocks of the Etendeka–Paraná province: source constraints. Geology (1990) 18:1119–1121.
Hart SR. A large-scale isotope anomaly in the Southern Hemisphere mantle. Nature (1984) 309:753–757.[CrossRef]
Hooper PR. The Columbia River flood basalt province: current status. Large Igneous Provinces: Continental, Oceanic, and Planetary Flood Volcanism.—Mahoney JJ, Coffin MF, eds. (1997) 100. American Geophysical Union Monograph. 1–28.
Irving AJ. A review of experimental studies on crystal–liquid trace element partitioning. Geochimica et Cosmochimica Acta (1978) 42:743–770.[CrossRef][Web of Science]
James D. The use of combined oxygen and radiogenic isotopes as indicators of crustal contamination. Annual Review of Earth and Planetary Sciences (1981) 9:311–344.[CrossRef][Web of Science]
Janney PE, le Roex AP, Carlson RW. Hafnium isotope and trace element constraints on the nature of mantle heterogeneity beneath the central Southwest Indian Ridge (13°E to 47°E). Journal of Petrology (2005) 46:2427–2464.
Jónasson K. Rhyolite volcanism in the Krafla central volcano, north-east Iceland. Bulletin of Volcanology (1994) 56:516–628.[CrossRef][Web of Science]
Jónasson K, Holm PM, Pedersen AK. Petrogenesis of silicic rocks from the Króksfjördur central volcano, NW Iceland. Journal of Petrology (1990) 33:1345–1369.
Kent RW, Pringle MS, Müller RD, Saunders AD, Ghose NC. 40Ar/39Ar geochronology of the Rajmahal basalts, India, and their relationship to the Kerguelen Plateau. Journal of Petrology (2002) 43:1141–1153.
Kirstein LA, Peate DW, Hawkesworth CJ, Turner SP, Harris C, Mantovani MSM. Early Cretaceous basaltic and rhyolitic magmatism in southern Uruguay associated with the opening of the South Atlantic. Journal of Petrology (2000) 41:1413–1438.
Kossert K. Half-life measurements of 87Rb by liquid scintillation counting. Applied Radiation and Isotopes (2003) 59:377–382.[CrossRef][Web of Science][Medline]
Kröner A, Windley BF, Jaeckelt P, Brewer TS, Razakamanana T. New zircon ages and regional significance for the evolution of the Pan-African orogen in Madagascar. Journal of the Geological Society, London (1999) 156:1125–1135.
Lacroix A. Mineralogie du Madagascar, Volume 3 (1923) Paris: August Challamel.
Le Bas MJ, Le Maitre RW, Streckeisen A, Zanettin B. A chemical classification of volcanic rocks based on the total alkali–silica diagram. Journal of Petrology (1986) 27:745–750.
Leeman WP, Oldow JS, Hart WK. Lithosphere-scale thrusting in the western U.S. Cordillera as constrained by Sr and Nd isotopic transition in Neogene volcanic rocks. Geology (1992) 20:63–66.
Lightfoot PC, Hawkesworth CJ, Sethna SF. Petrogenesis of rhyolites and trachytes from the Deccan Traps: Sr, Nd and Pb isotope and trace element evidence. Contributions to Mineralogy and Petrology (1987) 95:44–54.[CrossRef][Web of Science]
Lindsley DH, Andersen DJ. A two-pyroxene thermometer. Journal of Geophysical Research (1983) 88:887–906.[CrossRef]
Macdonald GA, Katsura T. Chemical composition of Hawaiian lavas. Journal of Petrology (1964) 5:2–133.
Macdougall JD, ed. Continental Flood Basalts. (1988) Dordrecht: Kluwer Academic.
Mahoney JJ, Nicollet C, Dupuy C. Madagascar basalts: tracking oceanic and continental sources. Earth and Planetary Science Letters (1991) 104:350–363.[CrossRef][Web of Science]
Mahoney JJ, le Roex AP, Peng ZX, Fisher RL, Natland JH. Southwestern limits of Indian Ocean ridge mantle and the origin of low-206Pb/204Pb MORB: isotope systematics of the central Southwest Indian Ridge (17°–50°E). Journal of Geophysical Research (1992) 97:19771–19790.[CrossRef]
Mahoney JJ, White WM, Upton BGJ, Neal CR, Scrutton RA. Beyond EM-1: lavas from Afanasy–Nikitin Rise and the Crozet Archipelago, Indian Ocean. Geology (1996) 24:615–618.
Melluso L, Morra V, Brotzu P, Razafiniparany A, Ratrimo V, Razafimahatratra D. Geochemistry and Sr-isotopic composition of the Cretaceous flood basalt sequence of northern Madagascar: petrogenetic and geodynamic implications. Journal of African Earth Sciences (1997) 24:371–390.[CrossRef]
Melluso L, Morra V, Brotzu P, Mahoney JJ. The Cretaceous igneous province of central–western Madagascar: evidence for heterogeneous mantle sources, crystal fractionation and crustal contamination. Journal of Petrology (2001) 42:1249–1278.
Melluso L, Morra V, Brotzu P, D’Antonio M. Petrogenesis of the Late Cretaceous tholeiitic magmatism in the passive margins of northeastern Madagascar. Volcanic Rifted Margins—Menzies MA, Klemperer SL, Ebinger CJ, Baker J, eds. (2002) 362. Geolocial Society of America, Special Papers. 83–98.
Melluso L, Morra V, Brotzu P, Franciosi L, Petteruti, Lieberknecht AM, Bennio L. Geochemical provinciality in the Cretaceous magmatism of northern Madagascar, and mantle source implications. Journal of the Geological Society, London (2003) 160:477–488.
Melluso L, Morra V, Brotzu P, Tommasini S, Renna MR, Duncan RA, Franciosi L, D’Amelio F. Geochronology and petrogenesis of the Cretaceous Antampombato–Ambatovy complex and associated dyke swarm. Journal of Petrology (2005) 46:1963–1996.
Michael PJ. Partition coefficients for rare earth elements in mafic minerals of high silica rhyolites: the importance of accessory mineral inclusions. Geochimica et Cosmochimica Acta (1988) 52:275–282.[CrossRef][Web of Science]
Miller CF, Mittlefehldt DW. Depletion of light rare-earth elements in felsic magmas. Geology (1983) 10:129–133.[CrossRef][Web of Science]
Miller JA, Harris C. Petrogenesis of the Swaziland and northern Natal rhyolites of the Lebombo rifted volcanic margin, south east Africa. Journal of Petrology (2007) 48:185–218.
Niu Y-L, Waggoner DG, Sinton JM, Mahoney JJ. Mantle source heterogeneity and melting processes beneath seafloor spreading centers: East Pacific Rise 18°–19°S. Journal of Geophysical Research (1996) 101:27711–27733.[CrossRef]
O'hara MJ, Mathews RE. Geochemical evolution in an advancing, periodically replenished, periodically tapped, continuously fractionated magma chamber. Journal of the Geological Society, London (1981) 138:237–278.
Peng ZX, Mahoney JJ, Hooper PR, Harris C, Beane JE. A role for lower continental crust in flood basalt genesis? Isotopic and incompatible element study of the lower six formations of the western Deccan Traps, India. Geochimica et Cosmochimica Acta (1994) 58:267–288.[CrossRef][Web of Science]
Price RC, Gray CM, Wilson RE, Frey FA, Taylor SR. The effects of weathering on rare earth elements, Y, and Ba in Tertiary basalts from southeastern Australia. Chemical Geology (1991) 93:245–265.[CrossRef][Web of Science]
Riley TR, Leat PT, Pankhurst RJ, Harris C. Origins of large volume rhyolitic volcanism in the Antarctic Peninsula and Patagonia by crustal melting. Journal of Petrology (2001) 42:1043–1065.
Rosholt J, Prijana N, Noble DC. Mobility of uranium and thorium in glassy and crystallized silicic volcanic rocks. Economic Geology (1971) 66:1061–1069.
Rudnick R, Fountain DM. Nature and composition of the continental crust: a lower crustal perspective. Reviews of Geophysics (1995) 33:267–309.[CrossRef][Web of Science]
Salters VJM, Stracke A. Composition of the depleted mantle. Geochemistry, Geophysics, Geosystems (2004) 5. Q05004, doi:05010.01029/02003GC000597.
Selbekk RS, Trønnes RG. The 1362 A.D. Öræfajökull eruption, Iceland: petrology and geochemistry of large-volume homogeneous rhyolite. Journal of Volcanology and Geothermal Research (2007) 160:42–58.[CrossRef][Web of Science]
Self S, Widdowson M, Thordarson T, Jay AE. Volatile fluxes during flood basalt eruptions and potential effects on the global environment: a Deccan perspective. Earth and Planetary Science Letters (2006) 248:517–531.[CrossRef]
Sethna SF, Battiwala HK. Metasomatized basalt xenoliths in the trachyte of Manori–Gorai, Bombay, and their significance. Geological, Mining and Metallurgical Society of India, Golden Jubilee Volume (1974) 337–346.
Shaw DM. Trace element fractionation during anatexis. Geochimica et Cosmochimica Acta (1970) 34:237–243.[CrossRef][Web of Science]
Sigmarsson O, Hémond C, Condomines M, Fourcade S, Oskarsson N. Origin of silicic magma in Iceland revealed by Th isotopes. Geology (1991) 19:621–624.
Snyder D. Thermal effects of the intrusion of basaltic magma into a more silicic magma chamber and implications for eruption triggering. Earth and Planetary Science Letters (2000) 175:257–273.[CrossRef][Web of Science]
Sobolev AV, Hofmann AW, Sobolev SV, Nikogosian IK. An olivine-free mantle source of Hawaiian shield basalts. Nature (2005) 434:590–597.[CrossRef][Medline]
Storey M, Mahoney JJ, Saunders AD, Duncan RA, Kelley SP, Coffin MF. Timing of hotspot-related volcanism and the breakup of Madagascar and India. Science (1995) 267:852–855.
Storey M, Mahoney JJ, Saunders AD. Cretaceous basalts in Madagascar and the transition between plume and continental lithosphere mantle sources. In: Mahoney, J. J. & Coffin, M. F. (eds). Large Igneous Provinces: Continental, Oceanic, and Planetary Flood Volcanism. American Geophysical Union Monograph (1997) 100:95–122.
Stormer JC, Nicholls J. XLFRAC: a program for the interactive testing of magmatic differentiation models. Computers and Geosciences (1978) 4:143–159.[CrossRef]
Sun S-S, McDonough WF. Chemical and isotopic systematics of oceanic basalts: implications for mantle composition and processes. Magmatism in the Ocean Basins—Saunders AD, Norry MJ, eds. (1989) 42. Geological Society, London, Special Publications. 313–345.
Thy P, Beard JS, Lofgren GE. Experimental constraints on the origin of Icelandic rhyolites. Journal of Geology (1990) 98:417–421.[Web of Science]
Todt W, Cliff RA, Hanser A, & Hofmann A. Evaluation of a 202Pb–205Pb double spike for high-precision lead isotopic analyses. In: Earth Processes: Reading the Isotopic Code—Basu A, Hart S, eds. (1996) 95. American Geophysical Union Monograph. 429–437.
Torsvik TH, Tucker RD, Ashwal LD, Eide EA, Rakotosolofo NA, de Wit MJ. Late Cretaceous magmatism in Madagascar: palaeomagnetic evidence for a stationary Marion hotspot. Earth and Planetary Science Letters (1998) 164:221–232.[CrossRef][Web of Science]
Tuttle OF, Bowen NL. Origin of granite in the light of experimental studies in the system NaAlSi3O8–KAlSi3O8–SiO2–H2O. Geological Society of America, Memoirs (1958) 74.
Wolff JA, Storey M. Zoning in highly alkaline magma bodies. Geological Magazine (1984) 21:563–575.
Wooden JL, Czamanske GK, Fedorenko VA, Arndt NT, Chauvel C, Bouse RM, King BSW, Knight RJ, Siems DF. Isotopic and trace-element constraints on mantle and crustal contributions to Siberian continental flood basalts. Noril'sk area, Siberia. Geochimica et Cosmochimica Acta (1993) 57:3766–3704.
Yoder HS Jr. Contemporaneous basaltic and rhyolitic magmas. American Mineralogist (1973) 58:153–171.[Web of Science]
York D. Least-squares fitting of a straight line with correlated errors. Earth and Planetary Science Letters (1969) 5:320–324.[CrossRef][Web of Science]
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  L. Melluso, H. C. Sheth, J. J. Mahoney, V. Morra, C. M. Petrone, and M. Storey Correlations between silicic volcanic rocks of the St Mary's Islands (southwestern India) and eastern Madagascar: implications for Late Cretaceous India-Madagascar reconstructions Journal of the Geological Society, March 1, 2009; 166(2): 283 - 294. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||