| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Journal of Petrology Volume 42 Number 7 Pages 1249-1278 2001
© Oxford University Press 2001
The Cretaceous Igneous Province of Madagascar: Geochemistry and Petrogenesis of Lavas and Dykes from the Central–Western Sector
1DIPARTIMENTO DI SCIENZE DELLA TERRA, UNIVERSITÀ DI NAPOLI FEDERICO II, VIA MEZZOCANNONE 8, 80134 NAPOLI, ITALY
2SCHOOL OF OCEAN AND EARTH SCIENCE AND TECHNOLOGY, UNIVERSITY OF HAWAII AT MANOA, 1680 EAST–WEST ROAD, HONOLULU, HI 96822, USA
Received February 24, 1999; Revised typescript accepted November 22, 2000
 |
ABSTRACT |
|---|
TOP
ABSTRACT
INTRODUCTIONThe Cretaceous lava sequence and associated mafic dyke swarm in central–western Madagascar (Mailaka and Bemaraha areas) range in composition from picrite basalts to cordierite–orthopyroxene-bearing rhyodacites (MgO from 14 to 0·6 wt %). Petrographic and chemical data indicate the presence of both tholeiitic and transitional magma series, with variable degree of rare earth element enrichment [(La/Nd)n = 1–1·4 for tholeiites vs (La/Nd)n = 0·65–1 for transitional rocks]. Initial (at 88 Ma) 87Sr/86Sr and
Nd range from 0·7044 to 0·7046 and -1·6 to -3·0 in the tholeiitic picrite basalts and basalts, and from 0·7030 to 0·7043 and +7·6 to +3·7 in the transitional picrite basalts and basalts. The rhyodacites have (87Sr/86Sr)88 = 0·7155 and Nd(88) = -10·6. Fractional crystallization of the observed phenocryst phases, starting from the most primitive tholeiitic basalts, combined with moderate amounts of contamination by peraluminous melts derived from partial melting of metapelitic basement rocks, explains the chemical composition of the rhyodacites reasonably well. The different parental magmas of the two series were probably generated by low degrees of partial melting (2·5–5%) of a depleted source (transitional basalts), and higher degrees of partial melting (5–7%) of a source very slightly enriched with a crustally derived component (tholeiitic basalts). Comparison between the samples from the eastern and northern parts of the province indicates that several different parental magmas and mantle sources were involved in the petrogenesis of the Madagascan basalts, and that contributions from mantle chemically equivalent to the modern Marion hotspot were negligible, overall. KEY WORDS: picrite basalts, rhyodacites, fractional crystallization, crustal contamination, mantle sources, Madagascan igneous province
|
INTRODUCTION |
|---|
The extensive Cretaceous volcanic province of Madagascar (Fig. 1a) has been interpreted as a product of the impingement of the Marion plume head upon the base of the northward-moving Madagascan lithosphere (90–125 km thick; Agrawal et al., 1992
; Rakotondraompiana et al., 1999) soon before Madagascar rifted from Greater India (Mahoney et al., 1991; Storey et al., 1995). The rocks of this province are preserved on land as lava flow sequences and intrusive complexes in the Karoo-related epicontinental sedimentary basins of Mahajanga and Morondava (e.g. Besairie & Collignon, 1972), and as lava flows and dyke swarms along the passive margin of the eastern coast (e.g. Brenon & Bussiere, 1959; Nicollet, 1984); the much more voluminous Northern Madagascar Plateau and Conrad Rise constitute probable submarine portions of the province (e.g. Coffin & Eldholm, 1994). Recent age determinations of lavas and dykes in Madagascar (Storetvedt et al., 1992; Storey et al., 1995; Torsvik et al., 1998) range from 92 to 86 Ma, and indicate that the volcanic activity ceased first in the northern part of the island. In a number of places, the dykes intrude, and the lavas lie directly upon, the Precambrian basement.
|
, Precambrian basement; 
, outcrops of felsic rocks; ß, outcrops of basaltic rocks. Samples of the central Mailaka section: M142 (top) to M164 (base). Samples of southern Mailaka (Bemaraha): M121 to M141. Dykes and sills: M101 to M120 (from east to west) and M165 to M173. The basement leucogranite PCM was sampled a few kilometres east of Beravina.Work in other flood basalt provinces has shown that both the mantle and crustal portions of the lithosphere may have a major influence on the composition of erupted magmas. It remains uncertain, however, whether mantle lithosphere acts principally as a contaminant of migrating sublithospheric melts or whether it is a major source of continental flood basalt magmatism (e.g. Ellam & Cox, 1991
; Gallagher & Hawkesworth, 1992; Arndt et al., 1993; Peng et al., 1994; Peate et al., 1999). Similarly, the contributions of hotspot (usually taken to be plume) mantle and non-hotspot asthenosphere to flood basalt magmas appear to vary greatly from province to province (e.g. Kent et al., 1997; Saunders et al., 1997; Sharma, 1997; Peate et al., 1999), such that no universal petrogenetic model has emerged. The geochemical role played by both the Marion hotspot and the continental lithosphere in the Madagascan province remains a major question, in part because detailed geochemical studies are few, important parts of the province remain to be studied, and erosion could have removed much of the volcanic record on land. On the other hand, the presence of rifted volcanic margins on both sides of the island, the thick lithosphere of the central part of the island, and the postulated presence of the Marion hotspot plume head give a very good opportunity to use this igneous province to study the petrogenesis of continental flood basalts. In this paper, we discuss new petrological, major and trace element, and Nd–Sr isotopic data for the previously unstudied central–western region of the Madagascan province, which bear directly on these issues.
|
GEOLOGICAL SETTING: THE CENTRAL–WESTERN REGION |
|---|
The lava sequence of central–western Madagascar was emplaced on the rifted margin of the Morondava basin (Fig. 1a and b), near shore, close to sea level, and above the lower Cretaceous Mailaka Limestone Formation (Besairie & Collignon, 1972
) It is a monotonous-appearing succession of lava flows, many of which have columnar jointing. Less frequent and highly altered pyroclastic deposits (basaltic), red boles, and hydrothermal deposits (agate fields) are intercalated with the lavas. Scattered, volumetrically minor, outcrops of very fresh pitchstones mark the close of magmatic activity in the inland area near Maintirano. The lava succession dips gently eastward, and has a maximum east–west outcrop of 
40 km, with a maximum thickness of 150 m (Besairie & Collignon, 1972). Toward the south, the succession appears to decrease in thickness, and pitchstones are absent. Probable plutonic–volcanic complexes are buried below the sedimentary succession in the seaward areas, as indicated by gravimetric surveys (Besairie & Collignon, 1972). No absolute age determinations of the volcanic rocks of this area are available; however, the biostratigraphic age of this volcanism is Coniacian–Turonian, similar to that of other occurrences throughout the island (see Besairie & Collignon, 1972; Storey et al., 1995).
A dyke swarm, part of the Morondava swarm (Ernst & Buchan, 1997), crops out in the area between the Mailaka limestones and the contact between the Permo-Triassic sedimentary rocks of the Morondava basin and the Precambrian basement (Fig. 1b). The dykes reach 10 m in width, and some are a few kilometres long. They are randomly oriented, and frequently cross-cut each other. Sills of coarse-grained gabbroic rocks intrude the Mailaka limestones. The Precambrian basement of the area comprises a supracrustal sequence of marbles, quartzites and metapelites (Ashwal & Tucker, 1999) and consists of granulite-facies metamorphic rocks, migmatitic–granitoid suites, and metapelites belonging to the kyanite–sillimanite–muscovite–almandine, sillimanite–orthoclase–almandine, and staurolite–almandine zones (Windley et al., 1994; Tucker et al., 1999).
|
GEOCHEMICAL BACKGROUND: MADAGASCAN PROVINCE |
|---|
Representative major and trace element and Sr and Nd isotopic compositions of rocks from previously studied parts of the Madagascan province are given in Table 1. Mahoney et al. (1991)
and Dostal et al. (1992) studied samples from the southern part of the province. Three groups of rocks were identified on a petrological and geochemical basis: (1) tholeiitic basalts (48·8–50 wt % SiO2) in the southeastern coastal outcrops, with low contents of incompatible elements, relatively high initial Nd(88) (+3·3 to +5·8) and low (87Sr/86Sr)88 (0·7032–0·7044); (2) tholeiitic basalts and basaltic andesites (49–52·5 wt % SiO2) in the southwestern outcrops and in the Ejeda–Bekily dyke swarm, which have low Nd(88) (-5·0 to -17·4) and high (87Sr/86Sr)88 (0·7114–0·7213), 207Pb/204Pb and 208Pb/204Pb for their 206Pb/204Pb, and variable TiO2 contents (1·5–3·4 wt %); (3) a dyke swarm of alkali basalts and basanites (44–48 wt % SiO2) with Nd(88) from -2·4 to +4·9, (87Sr/86Sr)88 from 0·7037 to 0·7056, low 206Pb/204Pb (17·4–18·1), and generally low TiO2 (1·0–2·3 wt %) and Nb (10–30 ppm). Those workers identified three source reservoirs in the mantle, which were thought to be responsible for the different isotopic trends observed: (1) a low 206Pb/204Pb reservoir, probably located within the lithospheric mantle; (2) a mid-ocean ridge basalt (MORB)-source mantle reservoir; (3) the Marion hotspot. Mahoney et al. (1991) argued that the lava flows of the southwestern outcrops with high (87Sr/86Sr)88 were the product of extensive crustal contamination.
|
Gioan et al. (1996) provided major and some trace element data for samples from the central and southern part of the Morondava basin (Fig. 1). They identified slightly alkaline rocks (normative nepheline 2–8%) in the central part of the basin, and subalkaline rocks from the southwesternmost outcrops, the latter with compositions similar to those determined for outcrops of the same areas by Dostal et al. (1992).
Storey et al. (1997) focused their study on lava flows and dykes in three transects along the eastern coast (Sambava, Tamatave and Mananjary, Fig. 1). They identified high Fe–Ti (TiO2 > 3 wt %) and Fe–Ti (TiO2 < 2 wt %) tholeiitic basalts. Also identified were high Mg–Ti basalts, which do not seem to be parental magmas of the other rocks. Taken together, the basalts least obviously affected by crustal contamination have a large range in Nd(88) (-5·3 to +8·1), (87Sr/86Sr)88 (0·7031–0·7068) and (206Pb/204Pb)88 (15·82–18·62). To explain most of these lavas, Storey et al. (1997) proposed mixing of partial melts of low-206Pb/204Pb lithospheric mantle with those from MORB-source mantle. Storey et al. (1997) also suggested aggregation of melts formed by variable degrees of partial melting (F) of garnet (F = 4%) and spinel (F = 10%) peridotites to produce the observed compositions of the high Fe–Ti basalts. Relatively few basalts, mainly in the southern part of the eastern coastal belt, were found to have isotopic compositions (in particular, relatively high 206Pb/204Pb) similar to the present-day products of the Marion hotspot. In addition, Storey et al. (1997) identified some relatively high-Si, low-Nb and low-Ti basalts in the Mananjary sector of eastern Madagascar, and argued that these had been affected by crustal contamination. Low-Nb mafic rocks have been described also in the Androy complex of southeastern Madagascar, with more incompatible element-rich basalts resembling some Ejeda–Bekily alkaline dykes (Mahoney et al., in preparation). Such basalt types seem to be completely absent both in the Tamatave and Sambava areas of northeastern Madagascar (Storey et al., 1997).
In the Mahajanga basin (eastern Antanimena plateau to Manasamody, north–central Madagascar; Fig. 1), the rocks range from basalt to basaltic andesite. Melluso et al. (1997) divided them into four groups. Lava flows of eastern Antanimena (SW of Mahajanga; Fig. 1) are characterized by low TiO2 (1·2–1·4 wt %) and Nb contents (5–6 ppm), low light-to-heavy REE (LREE–HREE; lanthanide REE) fractionation [(La/Yb)n = 3·2–3·8, where the subscript n indicates chondrite-normalized], strong negative Nb and Ti anomalies in mantle-normalized incompatible element diagrams and relatively high (87Sr/86Sr)88 (0·707–0·708) (group A). Associated with these are rocks with low Nb (4–8 ppm), low LREE–HREE fractionation [(La/Yb)n = 2·9], and negative Nb anomalies in mantle-normalized element patterns, but with higher TiO2 (>2 wt %) and V (>500 ppm) (group C). These rocks have (87Sr/86Sr)88 (0·7074–0·7075) in the range of group A values. The most abundant group (SE of Mahajanga) is made up of basalts with low ratios of large ion lithophile elements (LILE) to high field strength elements (HFSE) (e.g. K/Nb 150) and (87Sr/86Sr)88 (0·7038) (group B). Other lava flows in the Manasamody area (Fig. 1) show distinct enrichment in incompatible elements [TiO2 > 3 wt %, Nb = 13–25 ppm; Zr/Y > 7; (La/Yb)n = 5·6–7·3], and (87Sr/86Sr)88 from 0·7049 to 0·7053 (group D). These last rocks are likely to be the most incompatible element-enriched tholeiites of the province. The low-Nb basalts (groups A and C) crop out only in the Antanimena plateau, whereas high-Nb and relatively low-87Sr/86Sr basalts (groups B and D) crop out in the Bongolava and Manasamody areas (Fig. 1). Groups B and D are chemically and isotopically (Sr isotopes) similar to some tholeiitic basalts of the eastern coast (Table 1), and, like them, do not show obvious signs of crustal contamination (Melluso et al., 1997).
In summary, there are significant variations in the isotopic and chemical composition of the rocks of previously studied parts of the province. Evidence for hotspot, MORB-source and lithospheric mantle influence has been found for different areas, but a fuller understanding of the relative importance of each source, and the role of lithospheric contamination of magmas (vs lithospheric sources), requires knowledge of the geochemical and petrological characteristics of lavas and dykes in major areas that are still largely unknown in these respects. The Mailaka area of central–western Madagascar, the focus of this paper, is important because it contains one of the largest and best-preserved lava sequences of the province (together with the Androy complex in the SE, and Antanimena) and an associated mafic dyke swarm, which crops out between the lavas and the Precambrian basement (Fig. 1).
|
ANALYTICAL TECHNIQUES |
|---|
A set of 68 samples from stratigraphically controlled transects through the Mailaka lava sequence and from the dykes in the region was analysed in the present study (Fig. 1; Table 2). Most major and trace elements were analysed at Naples on pressed powder pellets with a Philips PW1400 X-ray fluorescence (XRF) spectrometer, following the methods described by Melluso et al. (1997)
. Thirty-five international standards were used for calibration. The average composition of the W2 standard, used as an unknown and monitor of accuracy during the course of the analyses, is shown in Table 2. Precision is estimated to be within 1% for SiO2, TiO2, Al2O3, Fe2O3t and CaO, better than 6% for K2O, ±0·03 wt % for MnO and P2O5, generally better than 5–10% for all trace elements in the observed compositional ranges, and ±1 ppm for Rb and Nb below 10 ppm. Na and Mg were analysed by atomic absorption spectrophotometry at Naples. Typical precision is better than 2% for Mg and better than 6% for Na. LOI (weight loss on ignition) was measured gravimetrically. Major element interlaboratory cross-checking, using XRF analyses on glass discs at SOEST, University of Hawaii, and analyses with inductively coupled plasma-atomic emission spectrometry (ICP-AES) at CRPG, Nancy, was performed (Table 2). REE, U, Th, Pb, Ta and Hf were analysed for a subset of samples by inductively coupled plasma-mass spectrometry (ICP-MS) at CRPG, Nancy. Uncertainties in the measurements are typically within 5–10% (relative) for most elements in the observed compositional ranges (Govindaraju & Mevelle, 1987). Mineral compositions were obtained using a CAMECA SX50 electron microprobe at CNR-CSQEA, Rome. Silicates and oxides were used as standards. The data were reduced utilizing CAMECA’s PAP correction method.
|
Rb, Sr, Sm and Nd concentrations, and Sr and Nd isotope ratios of acid-cleaned and powdered rock chips were analysed by isotope dilution on a VG Sector mass spectrometer at SOEST, University of Hawaii, using techniques described by Mahoney et al. (1991). Typical analytical uncertainties are ±1%, ±0·4%, ±0·2% and ±0·2% for Rb, Sr, Sm and Nd element concentrations, respectively; ±2 x 10-5 for 87Sr/86Sr and ±12 x 10-6 for 143Nd/144Nd. The Sr isotope ratios are reported relative to a value of 87Sr/86Sr = 0·71025 for the NBS 987 standard and Nd isotopes to a 143Nd/144Nd value of 0·511850 for the La Jolla Nd standard. Among the analysed samples, two rocks (M158, a lava flow, and M115, a dyke) have high 87Sr/86Sr (0·7055), coupled with high 143Nd/144Nd (0·5130). They were probably altered by sea water (consistent with the shallow nearshore environment of emplacement) during or after their emplacement, although traces of this contamination are barely seen in thin section. Multistep acid leaching [using the method of Mahoney (1987)] of one of the two samples (M158) reduced the 87Sr/86Sr value to 0·70299. As is well known, the 143Nd/144Nd ratio is not likely to change during moderate rock–sea-water interaction (e.g. McCulloch et al., 1981). Therefore, we preferentially use Nd isotope values in the discussion of the data.
|
CLASSIFICATION AND PETROGRAPHY |
|---|
A notable variety of rock types is present in the Mailaka area. The lava flows and dykes range from basalts to rhyodacites (Table 2; Fig. 2). Basalts and basaltic andesites are predominant. On the basis of their petrographic features, we divide the rocks into two series: a transitional series (TR) ranging from picrite basalt to basalt characterized by the presence of titaniferous salite in the groundmass, and a tholeiitic (THO) series ranging from picrite basalt to rhyodacite characterized by the presence of augite and, in more evolved rocks, also of low-Ca pyroxene. The distinction between the two series is easily seen from their distinct fields in the total alkali–silica diagram (Fig. 2) and also in their CIPW-normative characteristics, the TR series having nepheline or a few percent hypersthene in their norms, thus justifying their transitional chemistry, whereas the THO series are decidedly hypersthene (± olivine, ± quartz)-normative. This distinction corresponds to significant differences in whole-rock elemental trends (see below).
|
The basalts of the TR series are more widespread in the dyke swarm and in the flows of southern Mailaka. In the central Mailaka section, flows of the two series are interlayered, with THO flows dominant.
Central Mailaka section
The rocks of the central Mailaka section range from picrite basalt to rhyodacite. The picrite basalts (MgO > 10 wt %) include both THO and TR types: M162 and M163 (THO) have slightly iddingsitized olivine phenocrysts (with abundant chrome spinel inclusions) set in a fine-grained mesostasis of olivine, plagioclase, augite and iron oxides. M158 (TR) has fresh olivine (± spinel) phenocrysts in a pseudo-ophitic matrix of olivine, plagioclase, Ti-salite and opaque oxides. Alteration effects (not widespread) are seen mostly in the presence of clay minerals, chlorite and calcite in the interstices, and iddingsitization of olivine, often in the mantle of the crystals, which may be covered by fresh rims.
The basalts [M164, M162.1, M157, M150 (THO); M159, M152 (TR)] are mostly aphyric, with rare plagioclase and/or olivine phenocrysts in a matrix of plagioclase, pale green augite and magnetite ± olivine. Small plagioclase laths are sometimes included within larger ophitic augite grains.
The basaltic andesites and andesites [M156, M155, M154, M153, M160, M161, M145 (THO)] are aphyric or sparsely phyric, with rare plagioclase microphenocrysts, and abundant brown glassy matrix with plagioclase, pale green augite, pigeonite and/or orthopyroxene, and opaque oxide microlites. Rare euhedral to subhedral ‘dusty’ plagioclase phenocrysts are often reversely zoned, with a sodic core rich in pyroxene inclusions enclosed in a more calcic, inclusion-free rim (see ‘Mineral chemistry’). The cores are believed to represent xenocrysts. Argillification is sporadically present.
The dacites and rhyodacites (M142, M143, M144, M146, M148, M149) have normative quartz of 20–35% and corundum as high as 3·5%, are vitrophyric and have anhydrous parageneses, with phenocrysts of normally zoned plagioclase, cordierite, quartz, orthopyroxene and ilmenite, and microlites of zircon, set in abundant pale brown glass. Cordierite is euhedral, often with its typical cyclic twinning. Colourless garnet is also present in some samples as very rare corroded cores of aggregates formed by plagioclase, orthopyroxene and cordierite. Fayalite is observed as a very late-crystallized mineral, enclosing plagioclase, or as a euhedral microlite. Clusters of cordierite and poikilitic olivine are also present. No clinopyroxene was found. Minor devitrification of glass and iddingsitization of iron-rich olivine is observed. The composition and relative abundance of the observed phases is distinctly different from those of the other lithotypes (see below). In contrast to the andesites, these rocks are porphyritic. Symplectite textures (vermicular orthopyroxene, plagioclase and cordierite) were observed in a cluster found in one rhyodacite (sample M146). Similar structures were reported by Santosh (1987) in metapelites of southern Kerala (India).
Chemical stratigraphy
The central Mailaka sequence was thoroughly sampled in a road section (samples from M142 to M164 from the top to the base; Fig. 1b, Table 2). The picrite basalt samples are located in the lower part of the sequence, which, however, starts with an olivine-free basalt (M164). Upward in the sequence, more evolved rocks dominate. We infer that this change corresponds to a waning of magmatic activity, and thus to a decrease in mafic magma supplied to shallow reservoirs with time. The abundance of evolved rocks is minor compared with that of the basalts; in central–western Madagascar, evolved rocks make up no more than the upper 20–30 m of the sequence. It is worth noting that andesites (M160, M161) were found in the lower part of the sequence, in flows between the two picrite types, indicating that production of evolved melts was significant early. Three lava flows (M158, picrite basalt; M152 and M159, basalts) of the central Mailaka section belong to the TR series.
South Mailaka (Bemaraha)
Basalts of both series and basaltic andesites of the THO series [M127, M136, M131, M141, M126 (TR basalts); M123, M124, M128, M129, M134, M121, M132, M133, M135, M138 (THO basalts); M137, M139, M140, M122 (THO basaltic andesites)] are interlayered in the scattered outcrops of southernmost Mailaka. They are mostly aphyric or weakly porphyritic lava flows, with dominant plagioclase and olivine phenocrysts, sometimes within ophitic clinopyroxene.
Dykes and sills
The dykes are coarse grained to very fine grained, and range in composition from Mg-rich basalt to basaltic andesite. Both THO and TR types are present, but TR types are dominant (Table 2). Textures are very similar to those observed in the lava sequence of Mailaka, except for the range in grain sizes. Basalt dykes rich in olivine and groundmass titaniferous pyroxene [M165, M115 (TR)] have textures and chemical compositions identical to those observed in the M158 flow. This feature may indicate that at least one of these dykes was a feeder for the M158 lava flow. The most differentiated basaltic andesite (M108, THO) is a fine-grained sample with microphenocrysts of plagioclase and intergranular pyroxene in a glassy mesostasis containing skeletal Fe–Ti oxides. The dyke M169 has very large plagioclase phenocrysts and is evidently cumulitic.
Coarse-grained gabbro sills (M120, M119) intruded the Mailaka limestones. Sample M120 is ophitic with large plagioclase grains enclosed in pale, purple clinopyroxene. Olivine in this sample is Fe rich and poikilitic. It crystallized, together with oxides, after plagioclase. M119 is also ophitic but lacks late-crystallized olivine.
|
MINERAL COMPOSITIONS |
|---|
Pyroxene
As noted above, the two types of calcic pyroxene provide a first-order distinction between basalts belonging to the two magma series.
The augite–Fe-augite trend of the THO series shows a decrease in pyroxene Ca content with increasing differentiation from basalts, through basaltic andesites, to andesites (Fig. 3). The picrite basalt M162 has the most Mg-rich augite compositions (Ca43Mg46Fe11 to Ca37Mg40Fe23; it should be recalled that in this sample clinopyroxene is present only as groundmass microlites). The scatter of the data in Fig. 3 is likely to be the result of fast cooling. Limited Ti enrichment is observed in the augite trend (as high as 2 wt % TiO2 in augite of basaltic andesite M122). Pigeonite is a rare phase in the basaltic andesites, whereas moderately Mg-rich orthopyroxene is present in the andesites (Fig. 3a, Table 3). The rhyodacites have phenocrysts of unzoned, Fe-rich orthopyroxene {mg-number = [atomic Mg/(Mg + Fe + Mn)] = 0·26–0·30}, with very low wollastonite content (<2 mol %). Orthopyroxene has a compositional gap from andesites to rhyodacites (Fig. 3).
|
|
Early- to late-crystallized pyroxene compositions of the TR series define a trend from diopside to hedenbergite (Ca45Mg44Fe11 to Ca47Mg6Fe47; Fig. 3a). This trend displays significant Ti enrichment in intermediate compositions (up to 4·73 wt % TiO2 in the basalt M104), followed by decreasing Ti toward the most Fe-rich compositions, and smooth Na enrichment (Na2O = 2·7 wt % in the groundmass of M168; Table 3), typical of weakly alkaline series evolving to residual peralkaline melts (e.g. Stephenson & Upton, 1982; Kerr, 1998). Ca-poor pyroxene is absent in the TR rocks.
All the most Mg-rich clinopyroxenes of both series crystallized at very low pressure (<0·6 kbar), according to the barometer calibrated on Ca- and Mg-rich clinopyroxene structure (Nimis, 1999).
Oxides
Chromium-rich spinels are found as inclusions in olivine of the most Mg-rich basalts. They have a significant range in cr-number [molar Cr/(Cr + Al) = 0·68–0·23; Cr2O3 up to 31·4 wt %; Table 3] and increase in TiO2 and total iron (FeOt) with decreasing Mg/(Mg + Fe), similar to the spinels in the picrite basalts of the Gujarat area of the Deccan Traps (Melluso et al., 1995; Fig. 3b). Ilmenite is a rare accessory phase. Groundmass titanomagnetite is low in Al2O3 (<3 wt %) and high in ulvöspinel (68–91 mol %) (Table 3). When found together with Ti-magnetite (M120 sill) the calculated equilibration temperatures and oxygen fugacities are slightly below the quartz–fayalite–magnetite (QFM) buffer, and range from 771 to 750°C and from -16·6 to -16·7 log units, respectively, reflecting subsolidus re-equilibration.
Olivine
Olivine shows a generally continuous range from Fo83 to Fo6 (Table 4; Fig. 3a). The picrite basalts M162 (THO), M158 and M115 (TR) have the most Fo-rich phenocryst core compositions (Fo83 and Fo81, respectively). In both picrite basalt types, olivine also is present in the groundmass (Fo73–41). The largest and most continuous variation (Fo70–6) is found in basalt M105 (TR). The sample of a gabbroic sill (M120) has late Fe-rich olivine (Fo37–30). Olivine is absent in the basaltic andesites and andesites, but reappears in the rhyodacites. These have olivine with uniformly Fe-rich composition (Fo15–12), crystallized after plagioclase, cordierite and orthopyroxene (Table 4).
|
Feldspar
Plagioclase has a significant compositional range in the THO (An68–40) and TR (An72–9) basalts, and in the basaltic andesites (An64–24), andesites (An67–33) and rhyodacites (An67–40). Despite their differentiated composition, the rhyodacites have Ca-rich plagioclase, and this mineral is their only calcium-bearing phenocryst. Some feldspars within the andesites and basaltic andesites are cloudy, and show reverse zoning (e.g. An34, cores, to An51, rims, in M139; Table 4). Plagioclase compositions of both the THO and TR series show a slight increase of FeOt from the most calcic plagioclase to a composition of about An45, beyond which FeOt decreases. On the other hand, the iron content of plagioclase in the rhyodacites is distinctly lower at a given anorthite content (<0·4 wt %; Fig. 3c). This difference clearly indicates that the calcic plagioclase of the rhyodacites is not a xenocryst phase inherited from less evolved rocks.
Cordierite, garnet and glass
Cordierite is present in the rhyodacites as an equilibrium phenocryst phase. The atomic Mg/(Mg + Fe) ratio of this mineral has a narrow range (from 0·47 to 0·41; Table 4) and is distinctly higher than in the coexisting orthopyroxene (mg-number = 0·30–0·25) and olivine (Fo15–12; see above); this probably indicates early cordierite crystallization, and confirms the crystallization sequence. No chemical analyses of the very rare colourless garnet are available. Glass is abundant in the evolved rocks (andesites to rhyodacites), and is almost invariably rhyolitic (Table 4). The Al2O3 contents of this glass are lower than the bulk-rock Al2O3 values observed in the most felsic rhyodacites (12–13 vs 14 wt %).
Summary of the petrographic characteristics
The strongly differing compositions of the pyroxenes of the THO and TR rocks indicate independent crystallization paths from different parental magmas. The presence in the rhyodacites of calcic plagioclase, iron-rich orthopyroxene and cordierite phenocrysts [the latter two phases having higher Mg/(Mg + Fe) than olivines and pyroxenes found in some basalts and andesites], and rare garnet, indicate that these rocks cannot be related to the least evolved samples by closed-system fractionation.
|
MAJOR AND TRACE ELEMENT CHEMISTRY |
|---|
The most primitive basalts have high Mgv [atomic Mg/(Mg + Fe2+) with Fe2O3/FeO = 0·15], 0·64–0·69, not too different from the values of magmas in equilibrium with mantle peridotite (Mgv ( 0·68–0·76; Frey et al., 1978
). However, the most Mg-rich olivine found in sample M162 has a forsterite content of Fo83, which is too low to be in equilibrium with the host rock Mgv of 0·69, assuming an Fe–Mg distribution coefficient between olivine and liquid of 0·33 (Roeder & Emslie, 1970). This fact indicates that the most Mg-rich samples have excess olivine phenocrysts. Subtraction of 11·7% of Fo83 olivine from sample M162 yields a bulk-rock MgO content of 10 wt % and an Mgv of 0·62, which would be in equilibrium with Fo83 olivine. The still high MgO of the recalculated composition of M162 of 10 wt % may be close to a magmatic value, as suggested by the highly Mg-rich nature of the groundmass clinopyroxene and the presence of olivine in the groundmass.
Major and trace element variations, using Zr as a differentiation index, are shown in Fig. 4. Data for the two rock series plot in distinct fields in most major and trace element variation diagrams, particularly in the diagrams with SiO2, TiO2, K2O, Na2O and Sc. On the other hand, there is broad overlap in the diagrams for Y vs Zr and Nb vs Zr. CaO behaves similarly to MgO, except for some scatter in the most Zr-poor samples. The increase of Al2O3 and decrease of Sc with increasing Zr in the basalts of the TR series, when compared with the decrease of Al2O3 from the most Zr-poor samples, and the scattered behaviour of Sc of the THO series, could indicate lower plagioclase/pyroxene ratios in the mineral assemblages removed from magmas of the TR series relative to those of the THO series. This interpretation is supported by the low Cr (10 ppm) and high Sr (550 ppm) contents observed in the most differentiated TR basalt (M168), relative to values in rocks at the same MgO level of the THO series (see Table 2). In the THO series, TiO2 content never exceeds 2·5 wt %, and is slightly lower than in the TR rocks at similar Zr; the most pronounced differences are seen in the most Mg-rich basalts (1·6 wt % in the TR picrite basalts vs 0·8 wt % in the THO picrite basalts; Table 2). Na2O contents increase toward mildly differentiated rocks, and then tend to be roughly constant towards the rhyodacites. Rb, Y and Nb increase from basalts to rhyodacites to values as high as 180, 60 and 25 ppm, respectively. The marked increase in the Rb/Zr ratio from basalts (0·05–0·07) to rhyodacites (0·43–0·46) is noted, and is clear evidence of selective enrichment of Rb during differentiation, both elements being strongly incompatible with the observed phenocryst phases.
|
The lavas and dykes of the TR series have convex-upward chondrite-normalized REE patterns [(La/Nd)n = 0·65–1], and LREE contents generally lower than 20 times the chondrite average (Fig. 5; Table 5). The most differentiated TR basalt, M168, has a pattern subparallel to that of the more Mg-rich basalts, and with higher REE contents (Fig. 5).
|
|
The THO picrite basalt M162 and basalts M129, M102 and M164 have slightly LREE-enriched chondrite-normalized patterns [(La/Nd)n = 1–1·4; (La)n = 24]. The LREE–HREE fractionation, the HREE contents and the negative Eu anomaly increase toward the rhyodacites [(La)n 200; (La/Yb)n = 7·6; Eu/Eu* (normalized Eu/interpolated Eu between normalized Sm and Gd) = 0·46 in M146 and M148]. The andesite M160 has a significant positive Eu anomaly (Eu/Eu* = 1·15), relatively high Sr, Ba and Na2O contents (440 ppm, 1270 ppm and 4·8 wt %, respectively), and very low Cr, Cu and Ni (all <10 ppm; Table 2 and 5). Given the aphyric nature of this sample, these features could indicate contamination with Na-, Sr-, Eu- and Ba-rich melts or rocks and/or derivation from different parental magmas.
The mafic rocks have low Zr/Y (2·5–5; except for two samples with Zr/Y = 6) and Ti/V (21–39), and high Zr/Nb (13–42) and Hf/Ta (7–16); the highest values of the last two ratios are found in the THO samples and are distinctly higher than the values observed in basalts along the eastern coast (Zr/Nb = 11–24 and Hf/Ta = 4·5–8·1; Storey et al., 1997). The THO mafic rocks are also characterized by the lowest Ti/Zr ratios (60–80), compared with the values of their TR counterparts (>100) (see also Fig. 4). The low contents of Nb, Zr (2·3–6·4 and <100 ppm, respectively) and REE (see above) are all typical of magmas derived from incompatible element-depleted mantle (see le Roex, 1987; le Roex et al., 1989; Sun & McDonough, 1989).
Primitive-mantle-normalized patterns of the mafic rocks (Fig. 6) have marked peaks in Ba and Sr, and normalized abundances for the other elements generally lower than 10 times primitive mantle. No negative Ti or Nb anomalies are present in the TR rocks, which have a smooth pattern from Nd to Yb, whereas the THO basalts M162, M129 and M102 have negative Nb and small negative Ti anomalies. Neither the TR nor THO patterns have analogues in the basalts of the eastern coast or northwestern Madagascar (Fig. 6). In some TR samples (e.g. M111), the abundances of many elements are almost identical to those of average N-MORB (Sun & McDonough, 1989), excepting the higher abundances of Rb, Ba and Sr, and the steeper trend in the least incompatible elements.
|
|
Sr–Nd ISOTOPIC COMPOSITIONS |
|---|
As with the major and trace element variations, a wide overall range is observed in the Sr and Nd isotopic compositions (Table 5). The basalts of the TR series have the highest
Nd(88) (as high as +7·9, M105, basalt) and the lowest (87Sr/86Sr)88 (0·70298, leached M158, picrite basalt; Sr(88) = -20·08). The isotopic range of this series partially overlaps with that of rocks from the eastern coast of Madagascar (Nd(88) = -5·3 to +8·1; Mahoney et al., 1991; Storey et al., 1997; Fig. 7). The isotopic composition of the TR rocks is similar to present-day values observed in some recent Marion hotspot lavas (with maximum Nd = +7·4 and minimum 87Sr/86Sr = 0·70295, Sr(88) = -20·6; Mahoney et al., 1992; Fig. 7) and is within the range for Indian MORB (Nd(88) = -3·9 to +11·3; Mahoney et al., 1992).
|
The THO picrite basalt M162 and the basalt M129 have distinctly more radiogenic Sr and less radiogenic Nd [(87Sr/86Sr)88 = 0·70442 and 0·70459; Sr(88) = +0·26 and +2·73; Nd(88) = -1·2 and -3·0, respectively], compared with their TR counterparts at the same MgO level. The THO mafic rocks plot close to the Sr–Nd isotopic range of basalts recovered in the 39°-41°E sector of the Southwest Indian Ridge (Mahoney et al., 1992).
The picrite basalts and basalts of the TR series show a range in (87Sr/86Sr)88 from 0·70298 to 0·70546 but, with the exception of M168, only a very narrow range in Nd(88), from +7·3 to +7·9 (Table 5; Fig. 7). The variations in (87Sr/86Sr)88 of the mafic rocks appear largely to be related to interaction with sea water (see the ‘Analytical techniques’ section) and, therefore, only the (87Sr/86Sr)88 value of the acid-leached split of M158 is considered to be magmatic. The most differentiated TR basalt, M168, has lower Nd(88) (+4·0) than the more Mg-rich basalts, at similar (87Sr/86Sr)88 (0·70431; Sr(88) = -1·29), and could be affected by minor crustal contamination.
The THO trend from basaltic andesite to rhyodacite shows increasingly radiogenic Sr and unradiogenic Nd with increasing magmatic evolution [(87Sr/86Sr)88 = 0·7067 in basaltic andesites to 0·7155 in rhyodacites; Sr(88) = +32·7 to 157·6, and Nd(88) = -3·4 to -10·6]. The increase in (87Sr/86Sr)88 and decrease in Nd(88) from basalts to rhyodacites point to an open magmatic system, indicating significant interaction of magmas with low-Nd, high-87Sr/86Sr crustal materials. The Sr–Nd isotopic composition of the rhyodacites overlaps with the values of the basalts of the southwestern part of the plateau (Mahoney et al., 1991) (Fig. 7a).
|
MAGMATIC EVOLUTION |
|---|
Evidence of different parental magmas, fractionation and open-system processes
No simple crystal fractionation schemes (involving olivine and Cr-spinel) can explain the compositional difference between the THO picrite basalts (e.g. M162 and M163) and the TR picrite basalts (e.g. M158, M115 and M165), consistent with their markedly different contents of elements incompatible with olivine, and their
Nd values. Some of the differences in major element contents (e.g. SiO2, K2O, Al2O3) could be broadly compatible with addition of crustal material to a nepheline-normative basalt to produce a tholeiitic basalt. However, addition of crustal material to a nepheline-normative basalt to obtain M162 and M163 does not explain the much lower TiO2 contents of these two samples, even when olivine accumulation is taken into account.
The rocks of the THO series follow a smooth trend in the pseudoternary diagram diopside–olivine–SiO2 (Grove et al., 1982; Fig. 8). This trend indicates early olivine (± plagioclase) fractionation (or enrichment) and, after a kink, plagioclase + Ca-rich clinopyroxene ± olivine ± low-Ca pyroxene fractionation, towards rhyodacite compositions. Data for the rhyodacites plot on the quartz–orthopyroxene cotectic, outside the olivine–quartz divide. This evolutionary path is typical of a low-pressure liquid line of descent, as data for the samples plot very close to or along the 1 atm anhydrous cotectics, as do data for the tholeiitic basalts of the eastern part of the province (Fig. 8). Data for the TR series plot, with much scatter, close to the olivine–clinopyroxene join or to nepheline-normative liquid lines of descent (i.e. close to, or left of, the olivine–Ca-clinopyroxene join), thus confirming the different magmatic evolution of these rocks, as seen also in the contrasting pyroxene compositions.
|
Modelling of open-system processes: THO series
We have attempted to model the transition from the mafic to felsic rocks in the THO series with the MELTS program of Ghiorso & Sack (1995), and have compared the results with those of major element mass-balance calculations using the observed phenocryst compositions (Stormer & Nicholls, 1978). Results of both calculations are summarized in Table 6. The MELTS fractionation trends of various compositions at low pressures and oxygen fugacities, when plotted on an A/CNK vs SiO2 plot (Fig. 9), do not fit the data of the THO series. MELTS calculations starting from more differentiated compositions (other basaltic andesites and andesites; Fig. 9) also fail to produce peraluminous compositions similar to those of the rhyodacites. Moreover, in some runs the mineral assemblages required by MELTS have much lower augite/Ca-poor pyroxene ratios than those obtained by major element mass-balance calculations (and actually observed in thin section; Table 6).
|
|
The paragenesis and mineral composition of the rhyodacites do not conform to the trends observed for the basalt–andesite series, particularly the anomalously high anorthite content observed in the plagioclase (Fig. 3c) and the presence of the Al-rich mineral cordierite. These latter features strongly suggest that the rhyodacites are the product of interaction of magmas with metapelitic basement.
The Mailaka rhyodacites could potentially represent anatectic melts. However, this possibility poses immediate problems, such as the fact that the rhyodacites have Zr and Y contents much higher than those normally observed in anatectic, S-type leucogranites or shales worldwide (e.g. Whalen et al., 1987; Condie, 1993; Bea et al., 1994; Schnetger, 1994). Morever, the high Y and HREE contents and the low LREE/HREE ratios of the rhyodacites (see Fig. 5) should indicate that crustal melting to form the rhyodacites did not occur in the presence of residual garnet (e.g. Bea et al., 1994; Schnetger, 1994), yet garnet is an important phase in the metamorphic rocks of the nearest exposed crystalline basement (see Windley et al., 1994). Nor does it appear likely that the rhyodacites are derived from total or near-total melting of felsic crust, as the exposed basement of southeastern Somalia and Tanzania (the nearest pre-Gondwana break-up analogues; Lenoir et al., 1994; Appel, 1996; Möller et al., 1998) lacks rocks with such high Y contents. A comparison of the rhyodacites and equivalent peraluminous granites of the Madagascan basement (Tables 2 and 5, and see below) shows that the granites have much lower Y, Zr and Nb contents (4·8, 109 and 3·1 ppm, respectively), lower HREE (e.g. Lu = 0·07 ppm) and higher (La/Yb)n ratios (58), constituting evidence of their formation in the presence of residual garnet (± rutile, ± zircon). The smooth, increasing contents (starting from basalts) of Y, Zr, Nb and HREE with progressive degree of evolution (Figs 4 and 5) are pertinent to this issue. Assuming Nb is the most incompatible element of this group (a reasonable assumption for rocks with olivine, pyroxene, feldspar and minor oxide phenocrysts), and that Nb, an HFSE, is much less likely to increase significantly in concentration as a result of crustal contamination or alteration than, say, the LILE, we calculate that 80% of fractionation is needed to go from basalt M162 to rhyodacites [using the formula Nb0/Nbl f (where Nb0 is Nb in basalt, Nbl is Nb in rhyodacite and f is residual liquid fraction)]. With this amount of fractionation, the HREE and Y have roughly constant calculated D (D is the bulk solid/liquid partition coefficient at f = 0·2) values (0·15–0·3), Eu has calculated D = 0·66, whereas Rb, Th, U, Pb, La, Ce, Pr and Nd have calculated D < 0 (Table 7). The elements with calculated D < 0 (that is, more enriched than a perfectly incompatible element, an impossible situation for Rayleigh fractionation) are thought to be enriched by external contributions. The relatively high calculated DEu is due mostly to the effect of plagioclase fractionation, and the low, and relatively similar, calculated D of the HREE and Y (0·14–0·24) is thought to be simply related to their degree of incompatibility in the observed phenocryst phases [see the review by Green (1994)]. We deduce that the contamination process had a negligible influence on the HREE contents. This is possible if mantle-derived magmas mixed with felsic anatectic melts generated in the presence of residual garnet, which strongly partitioned HREE into the solid residue in the crust. Apparently, crystal fractionation of the observed phenocrysts from basaltic parental magmas was the dominant process leading to the production of differentiated melts, and the concentrations of incompatible elements such as the HFSE and the HREE were modified little (or not at all) during contamination.
|
The small volume of felsic rocks in the upper part of the Mailaka sequence, the increase of some incompatible trace elements well beyond levels expected from Rayleigh fractionation, and, finally, the increasingly radiogenic Sr and unradiogenic Nd isotopic ratios accompanying the transition from the most primitive to the most evolved rocks, indicate that concomitant assimilation and fractional crystallization models (AFC; e.g. DePaolo, 1981) may well explain the observed geochemical variations.
An AFC model taking into account the Nd isotopic and the Nb and Nd elemental variations from basalts to rhyodacites is presented in Fig. 10a and b. The composition of the contaminant was assumed to be that of a peraluminous, two-mica leucogranite, sampled in the basement close to the study area, and having characteristics typical of pelite-derived anatectic granites (e.g. 143Nd/144Nd = 0·51120 ± 1, 87Sr/86Sr = 0·72127 ± 1, Nd = 16·4 ppm, Sm/Nd = 0·13; Tables 2 and 5). Indeed, averages of Archaean shales have closely comparable Nd concentrations and isotopic compositions (e.g. 143Nd/144Nd = 0·51113; Nd = 22·1 ppm; Sm/Nd = 0·18; Vervoort et al., 1999). The choice of this anatectic melt for the contaminant is also constrained by the similar isotope values of other metamorphic rocks of the Madagascan basement (see Paquette et al., 1993; Tucker et al., 1999).
|
As is well known, the fractionation–assimilation trends are a function of the ratio between the mass assimilated and the mass cumulated from the magma [and of the composition of the contaminant as well; see DePaolo (1981)
]. This ratio is expressed as r (assimilation rate/accumulation rate, equal to mass assimilated/mass cumulated if contamination and assimilation occur simultaneously). The THO series data are clustered in a range of values of r between 0·13 and 0·35, with most data plotting close to or along the model curve at r = 0·2 (Fig. 10a and b). The rhyodacites thus could be the product of 80% fractionation coupled with 16% assimilation of a crustal melt, starting from a basaltic parental magma. It is worth noting that the sample with the lowest 143Nd/144Nd is the andesite M160, which has some peculiar stratigraphic and chemical characteristics (its position in the lower part of the Mailaka sequence, far from the rhyodacites, and its high Ba, Eu, Na and Sr contents; see above) that could reflect a larger contribution from (different?) anatectic melts.
Major element mass-balance calculations were made using the basalt M162 and the rhyodacite M148, the composition obtained by Vielzeuf & Holloway (1988) at 800°C during partial melting of pelitic compositions at 8 kbar as the major element composition of the crustal melt, and the composition of observed phenocryst phases. The calculations indicate that the major element composition of M148 could be obtained by the addition of 8–11% of the peraluminous crustal melt, with 84% fractionation of olivine (22·1%), plagioclase (45·7%), augite (25·7%) and oxides (6·5%) (Table 6).
Estimates of liquidus temperatures based on major element compositions were made according to calibrations useful for anhydrous tholeiitic melts [olivine–liquid equilibrium (Roeder & Emslie, 1970) and MgO–temperature relationships (Eggins, 1993) for Mg-rich compositions; equilibrium between olivine, plagioclase, augite and low-Ca pyroxene (Grove & Juster, 1989), and MgO–temperature relationships (Toplis & Carroll, 1995) for Mg-poor basalts and basaltic andesites to felsic rocks (see these references for the pertinent algorithms and uncertainties in the calibrations)]. The values range from 1230–1200°C for the most Mg-rich basalts, not modified by olivine enrichment, to 1100–1150°C for basaltic andesites (which are close to the reaction point olivine + liquid = pigeonite), and to 1040–1060°C for the rhyodacites (with Fe-orthopyroxene, quartz and Fe-olivine). Similar temperatures (1000–1100°C) have been obtained with two-pyroxene thermometry for the felsic rocks of the Paranà–Etendeka province (see Bellieni et al., 1986; Garland et al., 1995; Harris & Milner, 1996; Ewart et al., 1998). We also obtained similar temperature estimates with the MELTS program, except for the most silica-rich compositions (<900°C). The temperatures obtained for the Madagascan rocks can be used to place bounds on the maximum amount of crust that could be added to a crystallizing mantle-derived melt. A maximum ratio between the mass of assimilated crust and the mass of magma of 0·28 is indicated using the model developed by Grunder (1995) and Tegner et al. (1999). This value is higher than that obtained with the AFC model, and from major element mass-balance calculations (<0·2). This result suggests that cooling of 250–200°C during crystallization of mafic magma could provide sufficient energy to obtain the amount of crustal melts required in the chemical models summarized above.
Modelling the evolution of the TR series
The limited Nd isotopic variation in the TR rocks suggests that closed-system models of magmatic evolution may be appropriate for this series, although modelling may be complicated by the effects of variable interaction with sea water on some major and trace elements (e.g. Rb, Ba, Pb, Sr and K) and on Sr isotopes. Starting from the sample M111, 50% fractionation of olivine (22·8%), clinopyroxene (45·9%) and plagioclase (31·3%) is needed to reach the chemical composition of M168 (
R2 = 0·26; Table 6). The data are broadly in agreement with the observed trace element variations. Assuming an oxidized environment, the MELTS crystallization path produced too high an Fe–Ti enrichment before oxide saturation in the model starting from M158 (up to 21 wt % Fe2O3t and 2·8 wt % TiO2 at fO2 set at QFM + 1 and 0·5 kbar, with corresponding SiO2 = 45·1 wt % and MgO = 5·3 wt %). Toplis & Carroll (1996) also noted anomalous Fe–Ti enrichment with MELTS modelling in their experimental study on tholeiitic compositions. The low-temperature compositions calculated by MELTS from M158 are analogues of peralkaline rhyolites, following subtraction of olivine (6·7%), plagioclase (37·4%), clinopyroxene (26·2%), orthopyroxene (2·6%), pigeonite (2·7%), oxides (12·3%) and apatite (0·33%). These residual compositions have an Agpaitic Index [molar (Na + K)/Al] = 2·1 at SiO2 = 70 wt %. This result is in agreement with the Na-enrichment trend of the late-crystallized pyroxenes of the TR series, but it is noted that subcalcic pyroxenes are completely absent from the paragenesis found in thin section.
Summing up: (1) the rocks of the THO and TR series follow distinct low-pressure fractionation trends towards different residual magmas, dominated by fractionation of gabbroic assemblages, but characterized by different plagioclase/pyroxene ratios; (2) the evolved tholeiitic rocks have been moderately contaminated by melts derived from the metapelitic basement; (3) the chemical composition of the most felsic rocks represents, to a large extent, extensive fractionation of mafic parents (compare contents and variation of Zr, Y, HREE and Nb); (4) contamination of mantle-derived basalts by bulk metapelitic crust is not sufficient to explain the chemical variations observed; contamination processes involving (peraluminous) partial melts are more likely.
|
MANTLE SOURCE CHARACTERISTICS |
|---|
The mafic rocks of the TR series are characterized by low Zr and Nb contents, concave-downward chondrite-normalized REE patterns, low La/Nb (0·8–1·3), Ti/V and (87Sr/86Sr)88, and high Zr/Nb, Ti/Zr, Lu/Hf and
Nd(88). As noted above, the chemical and isotopic characteristics of the TR rocks largely match those of average N-MORB magmas (see le Roex, 1987; Sun & McDonough, 1989; Mahoney et al., 1992). Overall, the chemical and isotopic characteristics of the TR rocks, with the exception of some elements that could have been modified by alteration, indicate that they experienced little crustal contamination. The slight alkaline affinity of these samples, in contrast to the typical tholeiitic affinity of MORB, may be due to deeper final equilibration of the TR primary magmas with the surrounding peridotite, and/or could indicate that the TR basalts are the product of slightly lower melt fractions than typical MORB (see below).
The presence of somewhat MORB-like rocks is important, because it may indicate that an asthenospheric mantle source was important in central–western Madagascar. Pronounced geochemical similarities are observed between the TR rocks and some broadly MORB-like suites found in other continental flood basalt provinces. In particular, concave-downward REE patterns [e.g. (La/Nd)n = 0·83] (unlike the eastern Madagascan tholeiites) and somewhat ‘depleted’ Sr–Nd isotopic compositions (Nd(i) = +1 to +4; Sr(i) = -4 to -9) are observed in dykes of the Rooi Rand swarm of the Karoo province and in some dykes of the Falkland Islands (Duncan et al., 1990; Mitchell et al., 1999). Very similar chemical and isotopic characteristics are also seen in the Horingbaai swarm of the southern Etendeka (Namibia), particularly the Sr–Nd isotopic range (Nd(i) = +6 to +8; Sr(i) = -15 to -22; Duncan et al., 1990), and the nepheline-normative nature. These swarms are interpreted to have been largely derived from asthenospheric mantle beneath thinned lithosphere, slightly after the main phase of activity in the Karoo and Paraná–Etendeka provinces (Duncan et al., 1990).
The mafic rocks of the THO series, despite being more Si rich than their TR counterparts at the same MgO content, have slightly LREE-enriched patterns, high La/Nb (1·5–2·6) and small negative Nb and Ti anomalies in their mantle-normalized diagrams (Fig. 6), all features that are typical of low-Ti continental flood basalts and subduction-related mafic rocks. Because these features are observed in the most mafic rocks, they are likely to be primary features inherited from the source regions. Enrichment occurring during a mantle-derived tholeiitic magma’s ascent into the crust cannot be excluded a priori, but requires the determination of the chemical characteristics of supposed pristine parental magmas, which in this case are not observed. At present, we have no reason to suppose that the most mafic THO basalts were contaminated more during the passage through the crust than their TR counterparts, which are shown to be LREE depleted even at low MgO levels (e.g. M168; Fig. 5). The chemical and isotopic differences between the TR and the THO basalts are much larger that those observed between the THO basalts and the THO basaltic andesites, which did undergo extensive coupled fractionation and contamination (see above). Mixing (without fractionation) between a TR basalt and crust to obtain THO basalts involves 10–20% added crust (Fig. 10c). This is taken to be unreliable, given the Mg-rich nature of the THO basalts and of their observed phases.
Pressures of equilibration with peridotitic mantle sources; and partial melting models
By adding olivine until reaching an MgO content of 14 wt %, the most mafic rocks of central western Madagascar (THO and TR) could be in equilibrium with mantle olivine of Fo88–89 composition. According to the empirical relationship of Albarède (1992) based on the SiO2 and MgO contents of experimentally equilibrated melts, and to the pressure–composition relationship obtained by Hirose & Kushiro (1993) in their experimentally obtained melts from peridotite sources, the recalculated compositions have a significant range of equilibration pressures. The TR samples M115, M165 and M158 could have equilibrated with peridotitic mantle at 27 kbar (85 km depth), not too far from the garnet–spinel transition, whereas the THO basalts M162 and M163 could have equilibrated at 18 kbar (60 km depth). The samples M111, M167 and M172 (TR) have a lower pressure of equilibration (21–22 kbar; 70 km depth). All the recalculated analyses have relatively high Fe2O3t contents (12·7–15·2 wt %, vs 12·7–15·4 wt % for the unnormalized analyses; see Table 2), which are within the range of relatively Fe-rich mafic magmas (e.g. Scarrow & Cox, 1995; Turner & Hawkesworth, 1995) and close to or within the range of compositions obtained from melting of the peridotite HK-66 (Fe2O3t = 9–14 wt %; Hirose & Kushiro, 1993). The mafic rocks of both series have CaO/TiO2 (5–11), Al2O3/TiO2 (8–16) and CaO/Al2O3 (0·6–0·8) matching values found in the low-Ti picrite basalts of the northwestern Deccan Traps (9·6 ± 2·3, 13·5 ± 3·3, and 0·72 ± 0·05, respectively), which are inferred to be derived from basaltic-major-element- and trace-element-depleted mantle sources (Melluso et al., 1995).
The results of non-modal fractional melting models for the mafic magmas of both suites are shown in Fig. 11. We used only those elements resistant to low-temperature alteration. The TR rocks have LREE-depleted patterns, and therefore an LREE-depleted mantle source [the ‘depleted earth’ of McKenzie & O’Nions (1991, 1995)] was used as the starting composition. The results indicate that the mafic TR basalts could be produced by 2·5–5% melting of a residual spinel-bearing facies, and that melts from garnet-bearing sources could be present only in very minor amounts (20% of the total melt fraction; Fig. 11). The low degrees of partial melting obtained for the TR basalts are in good agreement with their slightly alkaline nature and with the calculated equilibration pressure obtained from major element modelling. We also note that present-day Marion hotspot lavas could be formed from the same source only at very low degrees of melting (<1%) with a variable contribution of melts from both spinel- and garnet-bearing sources (Fig. 11). This suggests that the present-day Marion hotspot magmas (mostly alkali basalts and basanitoids; Mahoney et al., 1992) were generated from a source more incompatible element enriched than that of the TR rocks, although sharing very similar Sr–Nd isotopic ratios.
|
The THO mafic rocks plot in the very low-degree melt region of model melts from depleted mantle in Fig. 11, and, therefore, a different source is more likely. Their LREE enrichment relative to the TR basalts, their subalkaline affinity, higher La/Nb, and low Ti/Zr ratios are all indications that the THO basalts ultimately are derived from higher degrees of partial melting (likely at shallower depths) of a compositionally different source. The slightly lower Nb/Yb ratios of the THO mafic rocks (2·7–3·3) than the TR rocks (3·3–4·6), at roughly the same Yb, could also indicate slightly higher degrees of melting for the THO group. We suppose that the slight LREE enrichment over HFSE occurred in the mantle, and therefore have modelled the source of the THO rocks by adding a very small degree of crust-derived component (1·1%) to the depleted source of the TR rocks (Fig. 11). This increases the La/Nb of the source to a value similar to that of the mafic THO samples (1·4–1·5). The model indicates that the THO rocks could be 5–7% melts of such a source with spinel in the residue.
Figure 11 shows that the basalts of eastern Madagascar cannot be formed from the same depleted mantle source composition as the TR mafic rocks, nor from that of the THO basalts. The low (Tb/Yb)n ratio of most Madagascan mafic rocks is evidence that spinel-bearing sources gave the most important contribution to the melts. Only a few basalts (some in the Mananjary and Tamatave sector and group D of the Mahajanga basin; Melluso et al., 1997) could have been produced by mixing of melts from both garnet- and spinel-bearing sources (see also Storey et al., 1997).
Major element mass-balance calculations starting from a spinel peridotite [bulk-rock chemistry and mineral compositions of McKenzie & O’Nions (1991)] indicate 7% partial melting to produce M158 (recalculated at 14 wt % MgO, see above; R2 = 0·04). The melting fraction reduces to 5% if M158 is produced by a garnet-bearing source (R2 = 0·02). On the other hand, 8% partial melting is required to produce the recalculated composition of M162 (R2 = 0·02) from a spinel lherzolite. A cpx-poor lherzolite is always the residue.
Therefore, taking these results and also the silica-saturation characteristics into account, it is likely that the THO primary rocks represent slightly higher-degree melts equilibrated with mantle at shallower depths than the TR analogues. Modelling indicates that both TR and THO rocks are derived from generally small extents of melting of incompatible element-depleted sources. The TR rocks could be formed by low degrees of melting (2–5%) of a source similar in many respects to that of N-MORB, whereas the THO rocks could be the product of slightly higher degrees of melting (5–7%) of the same source, to which a very small amount (1·1%) of an LREE-enriched component was added.
Controls on the degree of lithospheric extension and on melt production
The Karoo-related Morondava and Mahajanga sedimentary basins of western and northern Madagascar could have played a major role in determining the characteristics of melts produced in the Madagascan province. These basins are regions of rifted, thin lithosphere (formed in the early Mesozoic, long before the 88 Ma volcanism) beneath which melting could occur and through which melts might ascend comparatively easily. The rifted eastern coast of Madagascar is characterized by a narrow continental shelf and the absence of a well-developed sedimentary basin (e.g. Besairie & Collignon, 1972; Piqué, 1999); during Cretaceous rifting it could have provided an even easier route for migration of sublithospheric melts. As is well known, the thickness of the lithosphere and the temperature of the mantle can strongly influence the extent of melting beneath continental regions. We assume that the flows and dykes of northwestern and central–western Madagascar represent magmas that did not flow laterally for great distances on the surface or in the lithosphere, respectively. The thermal structure and the melt production rate of the mantle suggested by White & McKenzie (1995) could be pertinent to this respect. The rocks of central–western Madagascar are >1000 km from the presumed 88 Ma position of the Marion hotspot (Storey et al., 1995), and the mantle beneath the central–western part of the province would be only slightly hotter (by 
100°C) than normal asthenosphere (with a potential temperature of 1300°C), and much cooler than any plume head. A relatively small amount of lithospheric stretching [with extension factor ß (the ratio of initial to extended lithospheric thickness) 2] with this mantle potential temperature (i.e. 1400°C) would be sufficient to produce the observed thickness of the lava pile (<500 m) and would account for the generally low degrees of partial melting estimated for the mafic rocks of central–western Madagascar [see fig. 5 of White & McKenzie (1995)].
|
REGIONAL CHEMICAL CORRELATIONS |
|---|
The presence of the weakly alkaline basaltic dykes and lavas of the TR series in central–western Madagascar is an important finding because only tholeiitic rocks have been found in the eastern coastal outcrops (Dostal et al., 1992
; Storey et al., 1997), and from northwestern Madagascar (Melluso et al., 1997). Somewhat similar transitional alkaline compositions may be present in lavas of the central Morondava basin, to the south of our study area (Gioan et al., 1996), whereas major chemical and isotopic differences exist between the TR basalts and the Ejeda–Bekily alkali basalts and basanites (Mahoney et al., 1991; Dostal et al., 1992), precluding a common source (compare Figs 6, 7 and 11). The variation in Zr and Nb contents at different differentiation levels (as indicated by the MgO content) of the data available for the Madagascan province (Fig. 12) is particularly interesting. We note the following features: (1) the very large range in the trace element content at moderate MgO; (2) the group D tholeiitic basalts of the Mahajanga basin have almost the same Zr and Nb contents as the strongly evolved Mailaka rhyodacites; (3) very few basalts of the eastern coast, if any, match the composition of the TR and THO basalts. Therefore, chemically different parental magmas must have existed (Fig. 12; Table 1).
|
Melluso et al. (1997) showed that there is a marked compositional change in the chemistry of the basalts just south of Mahajanga, with low-Nb, low-Zr basalts (group A and C basalts) present only to the SW (see also Fig. 1). The THO rocks of this study and the groups A and C basalts exhibit some differences in major element composition at the same SiO2 and MgO levels, although they share similar compositional features, such as the relatively low Zr and Nb and the low LREE enrichment. Most low-Nb, low-Zr basalts crop out in the western part of the province and in some of the southwestern areas, whereas none are found in the Sambava and Tamatave regions. Even though the recent volcanoes of the Marion hotspot include lavas with isotopic values essentially identical to those of the TR samples [assuming the (87Sr/86Sr)88 of the acid-leached lava is representative of the series’ unaltered magmatic composition; Fig. 7], the Marion hotspot lavas are unlikely to be products of the same incompatible element-depleted sources as the TR mafic lavas, as discussed in the section on melting modelling. The chemical characteristics of the TR rocks imply that their mantle source has no trace of incompatible element-enriched plume-type or lithospheric material. On the other hand, the composition of the mafic THO rocks has been modelled with the input of very small amounts of a crust-derived melt component to a depleted mantle source. This could be related to contamination during magma ascent through the lithosphere, or to direct melting of portions of lithosphere modified by earlier addition of crustal materials. In this respect, we note that some of the southwestern basalts with very high 87Sr/86Sr and low 143Nd/144Nd (Mahoney et al., 1991) have proved to be very difficult to model by AFC, utilizing the known Madagascan crustal rocks as contaminants (Fig. 10a and b), and require exceedingly high amounts of mixing with crustal materials (>50%; Fig. 10c), with the conservative hypothesis of no concomitant fractionation (as these rocks are still chemically basalts). Therefore, their crust-like isotopic compositions could be considered also as partially inherited from the source regions. Also, the relatively high contents of some alteration-mobile incompatible elements (Ba, Rb, K, Pb, Sr) in both the mafic THO and TR rocks seem unlikely to be totally a result of secondary effects, and could partially be a primary feature.
The Nd model ages of the mafic THO samples (assuming evolution from a depleted, N-MORB-type, mantle reservoir) are 1 Ga (Table 8) and could indicate the timing of enrichment events in the lithospheric mantle or be the result of mixing between old, subduction-related, crustal material and depleted mantle, as has been hypothesized for some Indian Ocean basalts close to Madagascar (Rehkämper & Hofmann, 1997). In contrast, the Mesozoic Nd model ages of the mafic TR rocks (200 Ma; Table 5) again suggest a negligible role of old ‘enriched’ components, in the source of these magmas.
Storey et al. (1997) argued that isotopic components thought to be lithospheric appear to have been important along the eastern coast and in the south. They also noted that rare melts with Nd–Pb–Sr isotopic compositions resembling those of the present-day Marion hotspot (Fig. 7) could have reached the surface in a few areas in southeastern Madagascar, whereas, farther north (Tamatave and Sambava sectors), melts coming mostly from the lithospheric mantle appear to have been much more prevalent than those from the plume or the MORB-type asthenosphere presumably surrounding it. The geochemical characteristics of the group B and D basalts of the northwestern Mahajanga basin could be genetically related to those of the rocks of the eastern part of the province, thus reinforcing the hypothesis of their derivation from a slightly enriched lithospheric source (Melluso et al., 1997).
The compositional differences throughout the province could be linked to the north–south-trending pattern of some of the lineaments in the Precambrian basement, and to the presence of major shear zones running NW–SE and north–south (e.g. the Ranotsara shear zone; Fig. 1). Therefore, a regional pattern to the distribution of basalt types in the different sectors of the Madagascan province can be inferred with the new dataset added in this study. The combined data now available clearly indicate that magmas with the same chemical composition did not feed lavas and dykes from both sides of the Madagascan province. The concentration of the major present-day outcrops around the edges of the island (most of which, in the western half of the island, are within the coastal basins) offers no evidence that a formerly thick, now eroded, volcanic sequence lay above the thicker lithosphere of the interior (Fig. 1). We suspect that lavas (that originated from the western, southern, and eastern dyke swarms) may not have covered the interior to any great extent, but were confined mainly to the outer regions of thin lithosphere, rifted during late Cretaceous time or earlier, in late Palaeozoic–Mesozoic time.
|
CONCLUSIONS |
|---|
We identified two magmatic series in the central–western sector of the Madagascan province. Neither series has an analogue in the rocks of the eastern coast, nor in the outcrops of northwestern Madagascar. The transitional alkaline basalts strongly resemble MORB, both on a chemical and Sr–Nd isotopic basis, and seem to have been generated by low degrees of partial melting of a depleted mantle source. The tholeiitic basalts cannot be generated by reasonable degrees of melting of the same source, and need a small contribution of LREE-enriched, probably crust-derived, melt. Some mafic rocks of the transitional–alkaline series fractionated to evolved basaltic compositions. Crystal fractionation and crustal assimilation of the mafic tholeiitic magmas led ultimately to peraluminous rhyodacites. This work provides the first compelling evidence of extensive, low-pressure, differentiation processes in this province. The chemical and isotopic compositions of the mafic magmas in different areas of the province indicate that the magma supply systems of the Madagascan province were geochemically distinct, indicating independent parental magmas and heterogeneity in the source regions. This could also cast doubt on whether the Madagascan province is to be considered as a single flood basalt province. Further work on the other volcanic rocks of the province will better define the relative contributions of lithospheric vs asthenospheric vs plume sources, but, currently, the data indicate at most a minor contribution of mantle chemically equivalent to that feeding modern lavas of the hotspots linked to flood basalt provinces, as in some other Gondwana flood basalt sequences (e.g. Paranà, Marques et al., 1999
; Peate et al., 1999).
|
ACKNOWLEDGEMENTS |
|---|
We thank Renato Alaimo for his valuable logistical support in Madagascar; Antonio Canzanella, Vincenzo Monetti, Marcello Serracino and Khalil Spencer for help with the analytical work; Roger Rambeloson, Voahangy Ratrimo and Dieudonné Razafimahatratra for help with the fieldwork; and Alessio Langella for making the field trip more enjoyable. Peter Appel (Kiel University) is also thanked for providing his unpublished analyses of Tanzanian metapelites. Massimo D’Antonio very kindly performed the Sr–Nd isotopic determinations on the Precambrian leucogranite PCM. ‘Ciro’ Ricci and ‘PJ’ Cappelletti patiently drafted some figures. Early versions of this work benefited from critical reading of, and interesting discussions with, Peter Hooper and Enzo Piccirillo, and constructive reviews of Ray Kent, Alan Saunders, Marjorie Wilson and an anonymous reviewer. The research was funded by the Italian agencies MURST and CNR.
|
FOOTNOTES |
|---|
*Corresponding author. E-mail: melluso{at}cds.unina.it.
|
REFERENCES |
|---|
Agrawal, P. K., Pandey, O. P. & Negi, J. G. (1992). Madagascar: a continental fragment of the paleo-super Dharwar craton of India. Geology 20, 543–546.
Albarède, F. (1992). How deep do common basaltic magmas form and differentiate? Journal of Geophysical Research 97, 10997–11009.[Web of Science]
Appel, P. (1996). Hochdruckgranulite und Eklogite im Mozambique Belt von Tansania: eine geochemische und petrologische Studie. Doctorate dissertation, University of Kiel.
Arndt, N. T., Czamanske, G. K., Wooden, J. L. & Fedorenko, V. (1993). Mantle and crustal contributions to continental flood volcanism. Tectonophysics 223, 39–52.
Ashwal, L. D. & Tucker, R. D. (1999). Geology of Madagascar: a brief outline. Gondwana Research 2, 335–339.
Bea, F., Pereira, M. D. & Stroh, A. (1994). Mineral/leucosome trace element partitioning in a peraluminous migmatite (a laser ablation–ICP-MS study). Chemical Geology 117, 291–312.[Web of Science]
Bellieni, G., Comin Chiaramonti, P., Marques, L. S., Melfi, A. J., Nardy, A. J. R., Papatrechas, C., Piccirillo, E. M., Roisenberg, A. & Stolfa, D. (1986). Petrogenetic aspects of acid and basaltic lavas from the Paranà plateau (Brazil): geological, mineralogical and petrochemical relationships. Journal of Petrology 27, 915–944.
Besairie, H. (1964). Geological map of Madagascar. Tananarive: Service Géologique de Madagascar.
Besairie, H. & Collignon, M. (1972). Géologie de Madagascar—les terraines sédimentaires. Annales Géologiques de Madagascar 35, 553 pp.
Bowen, N. L. (1928). The Evolution of the Igneous Rocks. Princeton, NJ: Princeton University Press.
Boynton, W. V. (1984). Cosmochemistry of the rare earth elements: meteorite studies. In: Henderson, P. (ed.) Rare Earth Element Geochemistry. Amsterdam: Elsevier, pp. 63–114.
Brenon, P. & Bussiere, P. (1959). Le volcanisme à Madagascar. Bulletin Volcanologique 21, 77–93.
Coffin, M. F. & Eldholm, O. (1994). Large igneous provinces: crustal structure, dimensions, and external consequences. Reviews of Geophysics 32, 1–36.[Web of Science]
Condie, K. C. (1993). Chemical composition and evolution of the upper continental crust: contrasting results from surface samples and shales. Chemical Geology 104, 1–37.[Web of Science]
DePaolo, D. J. (1981). Trace element and isotopic effects of coupled fractional crystallization and crustal contamination processes. Earth and Planetary Science Letters 53, 189–202.[Web of Science]
Dostal, J. Dupuy, C., Nicollet, C. & Cantagrel, J. M. (1992). Geochemistry and petrogenesis of upper Cretaceous basaltic rocks from southern Madagascar. Chemical Geology 97, 199–218.
Duncan, A. R., Armstrong, R. A., Erlank, A. J., Marsh, J. S. & Watkins, R. T. (1990). MORB-related dolerites associated with the final phases of Karoo flood basalt volcanism in southern Africa. In: Parker, A. J., Rickwood, P. C. & Tucker, D. H. (eds) Mafic Dykes and Emplacement Mechanisms. Rotterdam: Balkema, pp. 119–129.
Eggins, S. M. (1993). Origin and differentiation of picritic arc magmas, Ambae (Aoba), Vanuatu. Contributions to Mineralogy and Petrology 114, 79–100.
Ellam, R. & Cox, K. G. (1991). An interpretation of Karoo picrite magmas in terms of interaction between asthenospheric magmas and the mantle lithosphere. Earth and Planetary Science Letters 105, 330–342.[Web of Science]
Ernst, R. E. & Buchan, K. L. (1997). Giant radiating dyke swarms: their use in identifying pre-Mesozoic Large Igneous Provinces and mantle plumes. In: Mahoney, J. J. & Coffin, M. F. (eds) Large Igneous Provinces: Continental, Oceanic and Planetary Flood Volcanism. Geophysical Monograph, American Geophysical Union 100, 297–333.
Ewart, A., Milner, S. C., Armstrong, R. A. & Duncan, A. R. (1998). Etendeka volcanism of the Goboboseb mountains and Messum igneous complex, Namibia. Part II: voluminous quartz latite volcanism of the Awahab magma system. Journal of Petrology 39, 227–253.
Frey, F. A., Green, D. H. & Roy, S. D. (1978). Integrated models of basalt petrogenesis: a study of quartz-tholeiites to olivine melilitites from south eastern Australia utilizing geochemical and experimental petrological data. Journal of Petrology 19, 463–513.
Gallagher, K. & Hawkesworth, C. J. (1992). Dehydration melting and the generation of continental flood basalts. Nature 358, 37–43.
Garland, F., Hawkesworth, C. J. & Mantovani, M. S. M. (1995). Description and petrogenesis of the Paranà rhyolites, southern Brazil. Journal of Petrology 36, 1193–1227.
Ghiorso, M. S. & Sack, R. O. (1995). Chemical mass transfer in magmatic processes. IV. A revised and internally consistent thermodynamic model for the interpolation and extrapolation of liquid–solid equilibria in magmatic systems at elevated temperatures and pressures. Contributions to Mineralogy and Petrology 119, 197–212.[Web of Science]
Gioan, P., Rasendrasoa, J., Bardintzeff, J.-M. & Rasamimanana, G. (1996). Nouvelles données petrographiques et structurales sur le magmatism du Sud du bassin de Morondava (Sud-Ouest de Madagascar). Journal of African Earth Sciences 22, 597–608.[Web of Science]
Govindaraju, K. & Mevelle, G. (1987). Fully automated dissolution and separation methods for inductively coupled plasma atomic emission spectrometry rock analysis. Application to the determination of the rare earth elements. Journal of Analytical Atomic Spectroscopy 2, 615–621.
Green, T. H. (1994). Experimental studies of trace-element partitioning applicable to igneous petrogenesis—Sedona 16 years later. Chemical Geology 117, 1–36.[Web of Science]
Grove, T. L. & Juster, T. C. (1989). Experimental investigations of low-Ca pyroxene stability and olivine–pyroxene–liquid equilibria at 1-atm in natural basaltic and andesitic liquids. Contributions to Mineralogy and Petrology 103, 287–305.
Grove, T. L., Gerlach, D. C. & Sando, T. W. (1982). Origin of calcalkaline series lavas at Medicine Lake Volcano by fractionation, assimilation and mixing. Contributions to Mineralogy and Petrology 80, 160–182; 114, 422–424.[Web of Science]
Grunder, A. (1995). Material and thermal roles of basalt in crustal magmatism: case study from eastern Nevada. Geology 23, 952–956.
Harris, C. & Milner, S. M. (1996). Crustal origin for the Paraná rhyolites: discussion of ‘Description and petrogenesis of the Paraná rhyolites, southern Brazil’ by Garland et al., 1995. Journal of Petrology 37, 299–302.
Hirose, K. & Kushiro, I. (1993). Partial melting of dry peridotites at high pressures: determination of compositions of melts segregated from peridotite using aggregates of diamonds. Earth and Planetary Science Letters 114, 477–489.[Web of Science]
Kent, R. W., Saunders, A. D., Kempton, P. D. & Ghose, N. C. (1997). Rajmahal Traps, eastern India: mantle sources and melt distribution at a volcanic rifted margin. In: Mahoney, J. J. & Coffin, M. F. (eds) Large Igneous Provinces: Continental, Oceanic and Planetary Flood Volcanism. Geophysical Monograph, American Geophysical Union 100, 145–182.
Kerr, A. C. (1998). Mineral chemistry of the Mull–Morvern Tertiary lava succession, western Scotland. Mineralogical Magazine 62, 295–312.[Abstract]
Kostopoulos, D. K. & James, S. D. (1992). Parameterization of the melting regime of the shallow upper mantle and the effects of variable lithospheric stretching on the mantle modal stratification and trace element concentrations in magmas. Journal of Petrology 33, 665–691.
Le Bas, M. J., Le Maitre, R. W., Streckeisen, A. & Zanettin, P. (1986). A chemical classification of volcanic rocks based on the total alkali–silica diagram. Journal of Petrology 27, 745–750.
Lenoir, J.-L., Küster, D., Liegeois, J.-P., Utke, A., Haider, A. & Matheis, G. (1994). Origin and regional significance of the late Precambrian and early Palaeozoic granitoids in the Pan-African belt of Somalia. Geologische Rundschau 83, 624–641.[Web of Science]
le Roex, A. P. (1987). Source regions of mid-ocean ridge basalts: evidence for enrichment processes. In: Menzies, M. A. & Hawkesworth, C. J. (eds) Mantle Metasomatism. New York: Academic Press, pp. 389–422.
le Roex, A. P., Dick, H. J. B. & Fisher, R. L. (1989). Petrology and geochemistry of MORB from 25°E to 46°E along the Southwest Indian Ridge: evidence for contrasting styles of mantle enrichment. Journal of Petrology 30, 947–986.
Mahoney, J. J. (1987). An isotopic survey of Pacific oceanic plateaus: implications for their nature and origin. In: Keating, B. H., Fryer, P., Batiza, R. & Boehlert, G. W. (eds) Seamounts, Islands and Atolls. Geophysical Monograph, American Geophysical Union 43, 207–220.
Mahoney, J. J., le Roex, A. P., Peng, Z. X., Fisher, R. L. & Natland, J. H. (1992). Southwestern limits of Indian Ocean Ridge mantle and the origin of low 206Pb/204Pb mid-ocean ridge basalt: isotope systematics of the central southwest Indian Ridge (17°–50°E). Journal of Geophysical Research 97, 19771–19790.
Mahoney, J. J., Nicollet, C. & Dupuy, C. (1991). Madagascar basalts: tracking oceanic and continental sources. Earth and Planetary Science Letters 104, 350–363.
Marques, L. S., Dupré, B. & Piccirillo, E. M. (1999). Mantle source compositions of the Paraná Magmatic province (southern Brazil): evidence from trace element and Sr–Nd–Pb isotope geochemistry. Journal of Geodynamics 28, 439–458.
McCulloch, M. T., Gregory, R. T., Wasserburg, G. J. & Taylor, H. P., Jr (1981). Sm–Nd, Rb–Sr and 16O/18O isotope systematics in an oceanic crustal section: evidence from the Samail ophiolite. Journal of Geophysical Research 86, 2721–2735.[Web of Science]
McKenzie, D. P. & O’Nions, R. K. (1991). Partial melt distributions from inversion of rare earth element concentrations. Journal of Petrology 32, 1021–1092.
McKenzie, D. P. & O’Nions, R. K. (1995). The source regions of ocean island basalts. Journal of Petrology 36, 133–159.
Melluso, L., Beccaluva, L., Brotzu, P., Gregnanin, A., Gupta, A. K., Morbidelli, L. & Traversa, G. (1995). Constraints on the nature of the mantle sources of the Deccan Traps from the petrology and geochemistry of the basalts of Gujarat state (Western India). Journal of Petrology 36, 1393–1432.
Melluso, L., Morra, V., Brotzu, P., Razafiniparany, A., Ratrimo, V. & Razafimahatratra, D. (1997). Geochemistry and petrogenesis of the Cretaceous flood basalt sequence of northern Madagascar. Journal of African Earth Sciences 24, 371–390.
Mitchell, C., Ellam, R. M. & Cox, K. G. (1999). Mesozoic dolerite dykes of the Falkland islands: petrology, petrogenesis and implications for geochemical proviciality in Gondwanaland low-Ti basaltic rocks. Journal of the Geological Society, London 156, 901–916.
Möller, A., Mezger, K. & Schenk, V. (1998). Crustal age domains and the evolution of the continental crust in the Mozambique belt of Tanzania: combined Sm–Nd, Rb–Sr and Pb–Pb isotopic evidence. Journal of Petrology 39, 749–783.
Nicollet, C. (1984). Le volcanisme dans le Sud-Ouest de Madagascar. Journal of African Earth Sciences 2, 383–388.[Web of Science]
Nimis, P. (1999). Clinopyroxene geobarometry of magmatic rocks. Part 2. Structural geobarometers for basic to acid, tholeiitic and mildly alkaline magmatic systems. Contributions to Mineralogy and Petrology 135, 62–74.[Web of Science]
Paquette, J. L., Nédélec, A., Moine, B. & Rakotondrazafy, M. (1993). U–Pb, single zircon evaporation and Sm–Nd isotopic study of a granulite domain in SE Madagascar. Journal of Geology 102, 523–538.
Peate, D. W., Hawkesworth, C. J., Mantovani, M. S. M., Rogers, N. W. & Turner, S. P. (1999). Petrogenesis and stratigraphy of the high-Ti Urubici magma type in the Paraná flood basalt province and implications for the nature of ‘Dupal’-type mantle in the South Atlantic region. Journal of Petrology 40, 451–473.
Peng, Z. X., Mahoney, J. J., Hooper, P. R., Harris, C. & Beane, J. E. (1994). 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. Geochimica et Cosmochimica Acta 58, 267–288.[Web of Science]
Piqué, A. (1999). L’évolution géologique de Madagascar et la dislocation du Gondwana: une introduction. Journal of African Earth Sciences 28, 919–930.[Web of Science]
Rakotondraompiana, S. L., Albuy, Y. & Piqué, A. (1999). Modèle de lithosphere pour l’île de Madagascar (océan Indien occidental): nouvelle interprétation des données gravimétriques. Journal of African Earth Sciences 28, 961–973.
Rehkämper, M. & Hofmann, A. W. (1997). Recycled ocean crust sediment in Indian Ocean MORB. Earth and Planetary Science Letters 147, 93–106.
Roeder, P. L. & Emslie, R. F. (1970). Olivine–liquid equilibrium. Contributions to Mineralogy and Petrology 29, 275–289.[Web of Science]
Santosh, M. (1987). Cordierite gneisses of southern Kerala, India: petrology, fluid inclusions and implications for crustal uplift history. Contributions to Mineralogy and Petrology 96, 343–356.
Saunders, A. D., Fitton, J. G., Kerr, A. C., Norry, M. J. & Kent, R. W. (1997). The North Atlantic igneous province. In: Mahoney, J. J. & Coffin, M. F. (eds) Large Igneous Provinces: Continental, Oceanic and Planetary Flood Volcanism. Geophysical Monograph, American Geophysical Union 100, 45–93.
Scarrow, J. H. & Cox, K. G. (1995). Basalts generated by decompressive adiabatic melting: a case study from Skye, NW Scotland. Journal of Petrology 36, 1–26.
Schnetger, B. (1994). Partial melting during the evolution of the amphibolite- to granulite-facies gneisses of the Ivrea Zone, northern Italy. Chemical Geology 113, 71–101.
Sharma, M. (1997). Siberian Traps. In: Mahoney, J. J. & Coffin, M. F. (eds) Large Igneous Provinces: Continental, Oceanic and Planetary Flood Volcanism. Geophysical Monograph, American Geophysical Union 100, 273–295.
Stephenson, D. & Upton, B. G. J. (1992). Ferromagnesian silicates in a differentiated alkaline complex: Kûngnât Fjeld, South Greenland. Mineralogical Magazine 46, 283–300.
Storetvedt, M., Mitchell, J. G., Abranches, M. C., Maaloe, S. & Robin, G. (1992). The coast-parallel dykes of east Madagascar: age of intrusion, remagnetization and tectonic aspects. Journal of African Earth Sciences 15, 237–249.
Storey, M., Mahoney, J. J., Saunders, A. D., Duncan, R. A., Kelley, S. P. & Coffin, M. F. (1995). Timing of hot spot-related vulcanism and the breakup of Madagascar and India. Science 267, 852–855.
Storey, M., Mahoney, J. J. & Saunders, A. D. (1997). 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. Geophysical Monograph, American Geophysical Union 100, 95–122.
Stormer, J. C., Jr & Nicholls, J. (1978). XLFrac: a program for interactive testing of magmatic differentiation models. Computers and Geosciences 4, 143–159.
Sun, S.-s. & McDonough, W. F. (1989). Chemical and isotopic systematics of oceanic basalts: implications for mantle composition and processes. In: Saunders, A. D. & Norry, M. J. (eds) Magmatism in the Ocean Basins. Geological Society, London, Special Publications 42, 313–345.
Tegner, C., Robins, B., Reginiussen, H. & Grundvig, S. (1999). Assimilation of crustal xenoliths in a basaltic magma chamber: Sr and Nd isotopic constraints from the Hasvik Layered Intrusion, Norway. Journal of Petrology 40, 363–380.
Toplis, M. J. & Carroll, M. R. (1995). An experimental study of the influence of oxygen fugacity on Fe–Ti oxide stability, phase relations, and mineral–melt equilibria in ferro-basaltic systems. Journal of Petrology 36, 1137–1170.
Toplis, M. J. & Carroll, M. R. (1996). Differentiation of ferro-basaltic magmas under conditions open and closed to oxygen: implications for the Skaergaard intrusion and other natural systems. Journal of Petrology 37, 837–858.
Torsvik, T. H., Tucker, R. D., Ashwal, L. D., Eide, E. A., Rakotosolofo, N. A. & de Wit, M. J. (1998). Late Cretaceous magmatism of Madagascar: paleomagnetic evidence for a stationary hotspot. Earth and Planetary Science Letters 164, 221–232.
Tucker, R. D., Ashwal, L. D., Handke, M. J., Hamilton, M. A., Le Grange, M. & Rambeloson, R. A. (1999). U–Pb geochronology and isotope geochemistry of the Archean and Proterozoic rocks of north–central Madagascar. Journal of Geology 107, 135–153.
Turner, S. F. & Hawkesworth, C. J. (1995). The nature of the sub-continental mantle: constraints from the major-element composition of continental flood basalts. Chemical Geology 120, 295–314.
Vervoort, J. D., Patchett, P. J., Blichert-Toft, J. & Albarède, F. (1999). Relationships between Lu–Hf and Sm–Nd isotopic systems in the global sedimentary system. Earth and Planetary Science Letters 168, 79–99.
Vielzeuf, D. & Holloway, J. R. (1988). Experimental determination of the fluid-absent melting relations in the pelitic system. Contributions to Mineralogy and Petrology 98, 257–276.[Web of Science]
Whalen, J. B., Currie, K. L. & Chappell, B. W. (1987). A-type granites: geochemical characteristics, discrimination and petrogenesis. Contributions to Mineralogy and Petrology 95, 407–419.[Web of Science]
White, R. S. & McKenzie, D. P. (1995). Mantle plumes and flood basalts. Journal of Geophysical Research 100, 17543–17585.
Windley, B. F., Razafiniparany, A., Razakamanana, T. & Ackermand, D. (1994). Tectonic framework of the Precambrian of Madagascar and its Gondwana connections: a review and reappraisal. Geologische Rundschau 83, 642–659.[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]
|
|
|
|
 |
|
 J. J. Mahoney, A. D. Saunders, M. Storey, and A. Randriamanantenasoa Geochemistry of the Volcan de l' Androy Basalt-Rhyolite Complex, Madagascar Cretaceous Igneous Province J. Petrology, June 1, 2008; 49(6): 1069 - 1096. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
L. MELLUSO, V. MORRA, P. BROTZU, S. TOMMASINI, M. R. RENNA, R. A. DUNCAN, L. FRANCIOSI, and F. D'AMELIO Geochronology and Petrogenesis of the Cretaceous Antampombato-Ambatovy Complex and Associated Dyke Swarm, Madagascar J. Petrology, October 1, 2005; 46(10): 1963 - 1996. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
L. Melluso, L. Melluso, V. Morra, P. Brotzu, L. Franciosi, A.M.P. Lieberknecht, and L. Bennio Geochemical provinciality in the Cretaceous basaltic magmatism of Northern Madagascar: mantle source implications Journal of the Geological Society, May 1, 2003; 160(3): 477 - 488. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||