Journal of Petrology | Pages |
© 1998 Oxford University Press |
The island of Pantelleria (
Fig. 1) is the type locality for pantellerite. This is a peralkaline rhyolite which, at Pantelleria, is extremely enriched in Na, Fe, Cl, and incompatible trace elements. Pre-eruptive H2O contents are moderate to high [2-4 wt %, according to
Kovalenko et al., (1988); 1·4-2·1 wt %, according to
Lowenstern & Mahood, (1991)].
Peralkaline silicic rocks, specifically trachytes and rhyolites with molar (Na2O + K2O)/Al2O3 ratio (agpaitic index, AI) greater than unity, occur as both plutonic (peralkaline granites) and volcanic (peralkaline trachytes, comendites and pantellerites) types. They are found in many different tectonic environments. Typically they occur in non-orogenic continental regions which have been subjected to crustal doming and rifting, e.g. Tibesti, Kenya, Ethiopia, Cameroon and Niger-Nigeria in Africa; East Greenland; British Columbia in North America; China-North Korea in Asia; Pantelleria island in Europe. They are also found commonly in oceanic islands located on actively spreading ridge crests, e.g. Socorro and Easter Islands on the East Pacific Rise; Iceland, the Azores and Ascension on the Mid-Atlantic Ridge; St Paul Island on the Mid-Indian Rise. The Canary Islands, where comendites and pantellerites also occur, constitute an exception in that they are located on a continental slope region. Less commonly, peralkaline silicic rocks occur in island arcs, e.g. Mayor Island in New Zealand, New Guinea and Hokkaido in NE Japan. Still more rarely they occur on continental margins, such as at Nandewar volcano in Australia, the Basin and Range Province in western North America, Volcàn Las Navajas in Mexico, and SW Sardinia in Italy. General studies of peralkaline rocks, with many additional cited references, have been reported by
Bailey et al., (1974),
Sørensen, (1974) and
Fitton & Upton, (1987).
The petrogenesis of peralkaline silicic magmas, and their genetic relationships with the associated mafic magmas, are strongly debated topics in petrology. Hypotheses include: (1) protracted fractional crystallization from alkali basaltic magmas-this hypothesis is supported by the ubiquitous presence of mildly alkali basalts associated with peralkaline rocks
(Ewart et al., 1968;
Barberi
et al., 1975;
Parker, 1983;
Civetta et al., 1984;
Nelson & Hegre, 1990;
Mungall & Martin, 1995); (2) partial melting of alkali-gabbroic cumulates, as proposed for Pantelleria
(Mahood et al., 1990;
Lowenstern & Mahood, 1991); (3) partial melting of different portions of upper lithospheric mantle, lower and upper crust, as proposed for trachytes, pantellerites and comendites, respectively, from Lake Naivasha in Kenya
(Bailey & Macdonald, 1987).
The close association, at Pantelleria, of mildly alkali basalts, peralkaline trachytes and pantellerites provides an important opportunity to address the problem of the petrogenesis of peralkaline silicic magmas. Major and trace element and Sr-Nd-Pb isotope data have been obtained for a variety of mafic and felsic volcanic rocks from Pantelleria, which, when combined with previously published data, allow us to reconstruct the evolution of the magmatic system of the island and to constrain the genetic relationships between the mafic and felsic magmas. Furthermore, these data provide constraints on the characteristics of the mantle source region beneath the Sicily Channel Rift Zone. Basic volcanic rocks from the islands of Pantelleria and Linosa (
Fig. 1) have the least radiogenic Sr isotope compositions of all the Quaternary alkaline volcanic rocks of Italy, and are inferred to represent magmas coming from an `uncontaminated' or minimally contaminated mantle domain, thought to be representative of the mantle of the whole Italian-Tyrrhenian region before overprinting by subduction-related fluid transfer processes
(Hoernle et al., 1995;
D'Antonio et al., 1996). The island of Pantelleria (
Fig. 1) represents the emergent portion of a volcanic edifice that rises ~1000 m above the adjacent sea floor. It is composed dominantly of volcanic rocks which include lavas and pyroclastic deposits, varying in composition from pantellerite, through pantelleritic trachyte and comenditic trachyte, to mildly alkali basalt, in order of decreasing abundance. Exposed mafic volcanic rocks are restricted to the northwestern lobe of the island, although drilling for geothermal research in the northern sector of the island has revealed a basaltic sequence several hundreds of metres thick
(Fulignati et al., 1997). K-Ar determinations on different basaltic units give ages of 118 ± 9, 83 ± 5, and ~29 ka BP
(Civetta et al., 1984). Felsic volcanic rocks range in age from 324 ka BP to 4 ka BP
(Civetta et al., 1984,
, 1988;
Mahood & Hildreth, 1986; this study).
The island is located in the NW-SE trending Sicily Channel Rift Zone (SCRZ)
(Illies, 1981;
Finetti, 1984), which results from transtensional tectonics along the northern margin of the African plate, related to the opening of the Tyrrhenian Sea
(Boccaletti et al., 1987). The SCRZ transects the Pelagian Block, which is a sector of the foreland of the northern part of the African plate preserved after collision with the European plate
(Burollet et al., 1978). The Pelagian Block is composed of continental crust ~40 km thick, which in the SCRZ thins from 25 km at the periphery to 18 km in the axial part
(Colombi et al., 1973;
Cassinis, 1981). Rifting began in the Late Miocene and was intense in the Pliocene to Late Quaternary. It is at present concentrated along the axial ridge area, where the asthenosphere rises up to ~60 km
(Della Vedova et al., 1989). Intense volcanism in the SCRZ has generated two emergent volcanoes which form the islands of Pantelleria and Linosa. Pantelleria is the largest volcano, located on the axial ridge, whereas Linosa occurs at the periphery of the rift zone. Many submerged volcanoes are located either in the rift zone or on the Pelagian Block. Some of these have been active in the last few centuries. Accounts of this very young volcanism have been given by
Imbò, (1965) and
Zarudzki, (1972).
Volcanological and petrological studies on Pantelleria have been carried out since the last century
(Foerstner, 1881;
Washington, 1913-1914;
Zies, 1960,
, 1962,
, 1966;
Carmichael, 1962,
, 1967;
Rittmann, 1967;
Noble & Haffty, 1969;
Romano, 1969;
Villari, 1970,
, 1974;
Korringa & Noble, 1972). More recent studies
(Wright, 1980;
Wolff & Wright, 1981;
Cornette et al., 1982,
, 1983;
Civetta et al., 1984,
, 1988;
Orsi & Sheridan, 1984;
Mahood & Baker, 1986;
Mahood & Hildreth, 1986;
Orsi
et al., 1989,
1991a,
1991b;
Mahood & Stimac, 1990;
Lowenstern & Mahood, 1991;
Lowenstern, 1994) have highlighted the structural and volcanological features of the island, as well as the mineralogical and geochemical characteristics of the volcanic rocks.
The structural setting of the island of Pantelleria is defined by both tectonic and volcano-tectonic lineaments. The tectonic lineaments include faults and fractures related to regional deformation events, which have the same orientation as the rift-bounding faults. NW-SE trending fractures and strike-slip faults are the dominant lineaments, although NE-SW and N-S trending features are common. All these regional features occur outside the area of caldera collapse (see below).
A NE-SW tensile fault system divides the island into two sectors and probably represents a crustal discontinuity along the axial ridge of the rift (
Fig. 1). The northwestern sector includes most of the exposed basaltic rocks, whereas the southeastern sector includes silicic peralkaline rocks. The former has been affected only by NW-SE crustal fractures through which mafic magmas have reached the surface, as testified by the alignment of basaltic vents ranging in age from 80 ka BP to 29 ka BP
(Cornette et al., 1983;
Civetta et al., 1984), and to
The volcano-tectonic features of the island include caldera collapses and resurgence inside the youngest caldera. At least two caldera collapses have affected the island in recent times. The oldest caldera, the La Vecchia caldera, is dated at 114 ka BP
(Mahood & Hildreth, 1986;
Fig. 1). The youngest is related to the eruption of the Green Tuff (50 ka BP;
Orsi & Sheridan, 1984) and has been named the Monastero caldera by
Cornette et al., (1983) and the Cinque Denti caldera by
Mahood & Hildreth, (1983). Inside the Monastero caldera resurgence has taken place with uplifting and tilting of the Montagna Grande block (
Orsi et al., 1991a).
The volcanic history of the island is characterized by large explosive eruptions, some of which produced caldera collapses, alternating with periods dominated by less energetic eruptions. The history before 50 ka BP cannot be reconstructed in detail because only remnants of the erupted products are exposed. This is due either to repeated collapse of the central part of the island and erosion along the coastal cliffs, or to blanketing of the whole island by the Green Tuff erupted at ~50 ka BP
(Cornette et al., 1983). The history since this last large eruption has been subdivided by
Civetta et al., (1984,
, 1988) into six silicic cycles sometimes intercalated with basaltic eruptions. The Green Tuff is considered representative of the first silicic cycle. It is the product of a complex eruption including ignimbrites, fall and surge horizons
(Orsi & Sheridan, 1984). The chemical composition of the Green Tuff varies from the base upwards from pantellerite to comenditic trachyte
(Civetta et al., 1984). All the other silicic cycles, dated at around 35-29, 22, 20-15, 14-12 and 10-4 ka BP, are characterized by eruptive products ranging in composition from pantellerite to pantelleritic trachyte, or to comenditic trachyte. Even though it is not always possible to arrange all the analysed samples in stratigraphic succession, for many cycles it has been demonstrated that the most differentiated magmas were erupted early in the cycle. This has been interpreted as the consequence of eruptions tapping a zoned magma chamber at progressively deeper levels during each eruptive cycle
(Civetta et al., 1988). Lava flows, pumice-fall deposits and ignimbrites from Pantelleria have been sampled for geochemical and isotopic analysis. In addition, some trachytic enclaves, thought to represent fragments of a crystal-rich, lower portion of a stratified magma chamber
(Mahood & Hildreth, 1986), have also been collected from lavas and pyroclastic units.
Mineral phases were analysed by combined WDS-EDS techniques using a CAMECA SX50 electron microprobe at the Centro di Studi per il Quaternario e l'Evoluzione Ambientale-CNR (Rome). Data reduction was made using the ZAF4/FLS software by Link Analytical. Representative analyses are reported in
Tables 1 and
2. Table 1. Representative analyses of olivine, clinopyroxene and amphibole in mafic and felsic volcanic rocks from Pantelleria.
Table 2. Representative microprobe analyses of feldspars, opaques, aenigmatite and apatite in mafic and felsic volcanic rocks from Pantelleria.
Both whole-rock lavas and pumice fragments, and glass separated from pumice fragments were ground in an agate mill, after careful washing in distilled water to remove any seawater-derived salt deposits. Major element compositions and Sc abundances were determined by inductively coupled plasma-atomic emission spectrometry (ICP-AES), and the remainder of the trace elements by inductively coupled plasma-mass spectrometry (ICP-MS) at the Centre de Recherches Pétrographiques et Géochimiques of Nancy, Cedex (France). Precision is 0·5% for major element oxides and variable in the range 2-5% for trace element of 50-150 ppm, 2-10% for trace element contents of 10-50 ppm, and 5-25% for trace element contents of 0-5 ppm
(Govindaraju & Mevelle, 1987; J. Morel, personal communication, 1997). Four samples were analysed for Rb and Sr by isotope dilution techniques at the Istituto di Geocronologia e Geochimica Isotopica of Pisa (Italy). Data are listed in
Tables 3 and
4. Table 3. Major (wt %) and trace element (ppm) compositions of mafic volcanic rocks from Pantelleria.
Both whole rocks (basalts and comenditic trachytes) and separated feldspars (comenditic trachytes and pan- tellerites) were analysed for Sr isotopes, whereas only whole rocks were analysed for Nd and Pb isotopes (
Table 5). Feldspar separates were leached in hot distilled 6·15 M HCl for 10 min and repeatedly washed with double distilled H2O. This procedure was adopted as the investigated felsic volcanic rocks have low Sr contents, and thus are strongly susceptible to post-depositional contamination by groundwater-transported Sr, which, at Pantelleria, contains appreciable amounts of Sr of seawater origin. Strontium and neodymium were extracted by conventional cation exchange chromatographic techniques, after dissolution with Merck Suprapur HF-HNO3-HCl mixtures. Sr and Nd isotope ratios were measured in dynamic mode as metal species on a VG 354 double-collector, thermal ionization mass spectrometer at the University `Federico II' of Naples. Three duplicate samples were also analysed for Nd at the Department of Geological Sciences, University of California, Santa Barbara, using a Finnigan MAT 261 multiple collector, thermal ionization mass spectrometer running in dynamic mode, giving results similar within analytical uncertainty (
Table 5). Replicate analyses of the NBS-987 and the La Jolla international reference standards gave average values of 0·71027 ± 0·00001 (2[sgr]m, standard deviation of the mean, n = 12; Naples), 0·511864 ± 0·000006 (2[sgr]m, n = 10; Naples) and 0·511854 ± 0·000020 (2[sgr]m, n = 7; Santa Barbara), respectively. Blanks were on the order of 3 ng for Sr and 0·4 ng for Nd. Table 5. Sr, Nd and Pb isotope compositions; Rb, Sr, Sm, Nd, U, Th and Pb contents; and some trace element ratios of selected mafic and felsic volcanic rocks from Pantelleria and Linosa.
Pb isotope ratios, and Pb, U and Th concentrations were determined at the Department of Geological Sciences, University of California, Santa Barbara. About 70 mg of powder was spiked with 205Pb-235U-230Th and then dissolved in distilled HF-HNO3-HCl mixtures, and finally redissolved in 1 M HBr. Three samples were prepared as duplicates by leaching powders with cold 2 M HNO3. These gave significantly different values for Pb isotope ratios, especially with respect to the 206Pb/204Pb ratio (
Table 5), thus suggesting that some of the samples may have been affected by foreign lead contamination to some extent (see discussion below). Lead was extracted by chromatographic exchange in Dowex 1 anion resin, using standard HBr and HCl elution procedures. All work was done in laminar flow hoods, using sub-boiling distilled acids and water for reagents. Blanks were of the order of 0·3 ng. Pb was measured using the silica gel-phosphoric acid method, on Re filaments. Isotopic analyses were made on a Finnigan MAT 261 multiple collector, thermal ionization mass spectrometer operating in a static mode, in which all five Pb isotopes were collected simultaneously. Replicate analyses of Pb standard NBS 981 indicated that Pb ratios are accurate to within ~±0·02% (2[sgr]) per mass unit after applying mass discrimination corrections of 0·13 ± 0·01% per mass unit. U and Th were extracted by chromatographic exchange in Dowex 1 anion resin, using standard HNO3 and HF-HCl elution procedures, loaded on Re filaments with colloidal graphite, and measured as metal species.
14C dating on selected palaeosol and charcoal samples younger than 10 ka BP has been carried out by Teledyne Brown Engineering Environmental Services, New Jersey, USA. The Libby half-life of 5568 years has been used to calculate ages. Data are reported in
Table 7 (below). The mafic volcanic rocks are classified mostly as basalt, according to the total alkali-silica diagram (TAS,
Le Bas et al., 1986;
Fig. 2a). Only one sample, Opl 1013, falls in the field for hawaiite. In the literature, these rocks have been described as either transitional
(Villari, 1974;
Civetta et al., 1984) or alkali basalts
(Mahood & Baker, 1986). Projection of their compositions on an R
1-R2 diagram
(De La Roche et al., 1980,
Fig. 3) clearly shows that the alkalinity of these rocks is highly variable, and that the samples range from transitional basalt, through alkali basalt, to basanite. This variability is also matched by a corresponding change in their CIPW normative compositions, with hypersthene in the transitional basalts, and up to 8·6% nepheline in the most alkaline mafic volcanic rock (sample Sic 98b in
Civetta et al., 1984; see also
Table 3). Interestingly, sample Opl 1013, classified as hawaiite according to the TAS grid, falls in the transitional basalt field, near the boundary with the trachybasalt field.
The felsic volcanic rocks, most of which are peralkaline (AI higher than unity), are classified as comenditic trachyte, pantelleritic trachyte and pantellerite, according to the Al2O3 vs FeOt grid
(Macdonald, 1974;
Fig. 2b). Only two samples, Sic 6 and Sic 50, fall within the field for comendite, although very close to the crossing of the discriminant boundaries. In the R1-R2 diagram (
Fig. 3) they fall in the fields for trachyte, quartz trachyte and alkaline rhyolite. Interestingly, not all the pantellerites fall in the alkaline rhyolite field; half lie in the quartz trachyte field, and two in the trachyte field. The basalts and the hawaiite are weakly to highly porphyritic (5-20% by volume phenocrysts), with plagioclase > olivine > clinopyroxene ± Ti-magnetite, set in a microcrystalline groundmass consisting of plagioclase, olivine, clinopyroxene, ilmenite, Ti-magnetite and apatite; analcite occurs in rocks with high normative ne contents (e.g. sample Opl 120b). In some cases, two types of groundmass can be recognized under the microscope. One type appears dark, having abundant small Fe-Ti oxide grains. The other type has less abundant, larger Fe-Ti oxide grains, thus appearing lighter than the former. Sometimes, both dark and light groundmasses are interfingered in the same rock (e.g. samples Sic 17a, a basalt, and Opl 1013, the hawaiite).
The comenditic trachytes are usually highly porphyritic (20-40% by volume phenocrysts), with abundant alkali-feldspar, clinopyroxene, olivine, Ti-magnetite and ilmenite, occurring both in aggregates and as single crystals. The groundmass is almost exclusively vesicular glass, with sparse microlites of alkali-feldspar, clinopyroxene, alkali-amphibole and Fe-Ti oxides.
The comendites are porphyritic, with predominant alkali-feldspar and subordinate clinopyroxene, rare olivine and aenigmatite, set in a vesicular glassy groundmass with rare microlites.
The pantelleritic trachytes are porphyritic with alkali-feldspar, clinopyroxene, aenigmatite, olivine, ilmenite and magnetite, set in a vesicular glassy groundmass with very few microlites.
The pantellerites are characterized by alkali-feldspar and clinopyroxene often in aggregates, aenigmatite commonly as single crystals, sometimes embayed quartz, as single crystals or in monomineralic aggregates, or even intergrown with anorthoclase. Typical accessory phases are pyrrhotite, substituting for Ti-magnetite and ilmenite in rocks with SiO2 > 67 wt %, and very scarce apatite. The groundmass consists of vesicular, light-coloured glass, with sparse microlites of alkali-feldspar, quartz, clinopyroxene, aenigmatite and alkali-amphibole. Electron microprobe compositions of representative mineral phases in the mafic and felsic volcanic rocks are reported in
Tables 1 and
2. Additional data are available from the authors on request. The compositional variations of olivines, pyroxenes and feldspars are illustrated in
Figs 4 and
5.
Clinopyroxene in the mafic rocks is highly variable in terms of its Ca, Ti and Al content (
Table 1 and
Fig. 5a). For example, the Ti/Al ratio varies widely even in clinopyroxenes from the same rock, from ~0·20 to ~0·46 (
Fig. 4). According to the classification grid recommended by
Morimoto, (1988;
Fig. 5a) the crystals can be classified as diopside and augite. These two compositions are simultaneously present in the same rock, both as distinct and zoned crystals, and both as early and late crystallization phases. In particular, in the most alkali-rich basalt (Opl 120b), there is a group of Al, Ti-poor diopsides showing a limited range of Fe* (= Fe2+ + Fe3+ + Mn) at constant Ca, and a group of Al, Ti-rich augites, exhibiting a trend of increasing Fe* with decreasing Ca. Augite is the commonest clinopyroxene in the less alkaline basalt (Sic 17a) and in the hawaiite (Opl 1013), although some phenocrysts straddle the boundary between diopside and augite. In the felsic volcanic rocks the pyroxene appears to continue the compositional range of that of the mafic volcanic rocks. It is an augite richer in Fe*, with variable Ca, Ti and Al contents. Such compositional variations in the clinopyroxenes in the volcanic rocks of Pantelleria have not been reported previously in the literature. The most extreme compositions of clinopyroxene found in the felsic volcanic rocks correspond to those previously reported in the literature as sodic ferrohedenbergite
(Carmichael, 1962;
Sutherland, 1974).
Olivine in the mafic volcanic rocks varies regularly from chrysolite, through hyalosiderite, to hortonolite (
Fig. 5b, and
Table 1). The total range is Fo85·6-46·5, extending that reported by
Mahood & Baker, (1986) in rocks of similar compositions (Fo82-70). The most Mg-rich compositions occur in the most primitive basalt (Opl 120b), and the most Fe-rich compositions occur in the groundmass microlites in the hawaiite (Opl 1013). No strong zoning has been detected in the phenocrysts. In the felsic volcanic rocks the olivine ranges from ferrohortonolite up to almost pure fayalite; the latter is found in pantellerites (total range is Fo22·5-5·9).
Feldspar (
Table 2) in the mafic volcanic rocks is plagioclase which varies from phenocrysts of bytownite, through phenocrysts and microphenocrysts of labradorite, to microlites of andesine, the total range being from An81·1 to An40·8 (
Fig. 5c). Rare partially resorbed, sieve-textured crystals of bytownite compositions are present. Normal zoning is typical.
Mahood & Baker, (1986) reported a range from An76 to An55 in rocks of similar composition. In the felsic volcanic rocks feldspar becomes richer in the orthoclase component and poorer in both the anorthite and albite components; and the K/Na ratio increases with progressive differentiation and peralkalinity of the magma. The composition ranges within the field for anorthoclase from Or16 to Or37, approaching the limiting composition for feldspar in pantelleritic magmas (Or38) experimentally determined by
Carmichael & MacKenzie, (1963). No sanidine was observed, at variance with the data of previous workers
(Sutherland, 1974;
Mahood & Stimac, 1990); those compositions were most probably the result of analytical bias.
Fe-Ti oxides in the mafic volcanic rocks include Ti-magnetite along with ilmenite in those with transitional affinity (
Table 2). Where available, coexisting pairs of Ti-magnetite and ilmenite in both the dark and light groundmasses have been used to determine oxygen fugacity [f(O2)] and temperature (T), according to the solution model developed by
Andersen & Lindsley, (1988). The results do not show any significant difference in the f(O2)-T of the two types of groundmass, all samples plotting between the fayalite-magnetite quartz (FMQ) and magnetite-wüstite (MW) synthetic oxygen buffers ([Delta]FMQ = -0·47 to -0·85), similar to many continental basalts in intra-plate regimes (e.g. Paranà,
Bellieni et al., 1988; Antarctica,
Brotzu et al., 1988; Deccan,
Sethna & Sethna, 1988). The calculated temperatures are in the range 1079-940°C. In the felsic volcanic rocks, Ti-magnetite is much more abundant than ilmenite, and no pairs of the two oxides have been found suitable for the f(O2)-T calculations.
Carmichael, (1967) reported a temperature of 1025°C and an oxygen fugacity of 10-11·2 for a pantellerite, in the range observed here for mafic volcanic rocks.
Aenigmatite (
Table 2) shows only a small range in composition, similar to that in volcanics from many different localities. In the more evolved and peralkaline rocks aenigmatite is richer in Fe and Na, and poorer in Mg, Ca and Al. It replaces Fe-rich olivine and Ti-magnetite in the more strongly peralkaline rocks.
Apatite in the felsic volcanic rocks is fluorapatite, with up to 5·7% of fluorine (
Table 2).
Amphibole has been found sporadically as microlites in a few felsic volcanic rocks. According to the currently used nomenclature
(Leake, 1978) it is an alkali-amphibole classifiable as calcian ferro-eckermannite (
Table 1). The major and trace element compositions of selected mafic and felsic volcanic rocks from Pantelleria are listed in
Tables 3 and
4, respectively. Analyses are arranged in order of decreasing mg-number [= 100Mg/(Mg + Fe);
Wilkinson, 1982] and increasing AI, respectively. Most of the analysed samples belong to depositional units younger than 50 ka BP. Some are older than 50 ka BP, according to either radiometric age determinations
(Civetta et al., 1984,
, 1988;
Orsi et al., 1991b) or their stratigraphic position.
Selected major and trace elements are plotted against weight percent SiO2, used as a differentiation index, in
Figs 6 and
7, respectively, and against mg-number in
Fig. 8 for mafic volcanic rocks alone, to highlight the possible differences between them. A compositional gap between mafic and felsic volcanic rocks occurs. A few rocks with SiO2 contents ranging between 51 and 62 wt % have been found, mainly as xenoliths in felsic deposits
(Villari, 1974;
Civetta et al., 1984). They probably are the result of mixing between basic and felsic magmas
(Mahood & Baker, 1986). Lack of compositions in the range 51-62 wt % SiO2 is indicative of a real silica gap and not due to sampling bias. Therefore, the igneous rocks of Pantelleria are a bimodal suite of mafic (basalt and hawaiite) and felsic volcanic rocks (comenditic trachyte, pantelleritic trachyte, comendite and pantellerite).
Mafic volcanic rocks
The basalts and hawaiite of Pantelleria are all somewhat differentiated. No sample has >6·9% MgO, and Ni and Cr contents are always <100 and <190 ppm, respectively (
Table 3). Variation diagrams (
Fig. 8) show a general decrease of MgO, increase of Fe2O3tot and TiO2, and scattered behaviour of SiO2, Al2O3 and P2O5, with decreasing mg-number. Furthermore, basalts older than 50 ka BP (the age of emplacement of the Green Tuff) have distinctly lower contents of SiO2 and Al2O3, and higher contents of Fe2O3tot, MgO, alkalis (these are not shown) and, in particular, TiO2 and P2O5, relative to basalts younger than 50 ka BP. The two groups of basalts can be distinguished on the basis of their age of eruption and TiO2 and P2O5 contents: a high-TiO2 (>3 wt %) and -P2O5 ( \>= 0·9 wt %) group, erupted before 50 ka BP, and a low-TiO2 (<3 wt %) and -P2O5 ( \<= 0·75 wt %) group, erupted after 50 ka BP. The hawaiite plots either alone or along with the low-TiO2-P2O5 group. Trace element contents also display differences between the two groups, exhibiting noticeable variations. The high-TiO2-P2O5 basalts (older than 50 ka BP, hereafter referred to as high-Ti-P) have twice the trace element contents, including Sr, Y, Zr, Nb, Ba and light rare earth elements (LREE), relative to the low-TiO2-P2O5 basalts (younger than 50 ka BP, hereafter referred to as low-Ti-P), particularly at lower mg-numbers (
Fig. 8 and
Table 3). However, even though the decrease of Sr and Ba reflects the dominant fractionation of plagioclase within the two mafic groups, the decrease of Zr, Nb and La shown by the high-Ti-P basalts cannot be accounted for by fractionation of plagioclase and mafic phases. REE patterns (
Fig. 9) for representative basalts show moderate fractionation with enrichment of LREE relative to middle REE (MREE), and heavy REE (HREE) (LaN/LuN = 9·3/19·3; TbN/YbN = 1·82/2·63); the high-Ti-P basalts generally have the highest REE abundances and ratios. Almost all basalts display a slightly positive Eu anomaly (Eu/Eu* = 1·08/1·34;
Table 5). This feature has been recognized in mildly alkali basalts associated with pantellerites from other localities around the world, e.g. at the Boina Centre in Afar Rift, Ethiopia
(Barberi et al., 1975) and at Mayor Island, New Zealand
(Ewart et al., 1968). In the basalts from Pantelleria, the observed partially resorbed, sieve-textured plagioclase crystals may account for the positive Eu anomalies. Interestingly, most of the REE patterns cross each other, ruling out derivation from a single parent melt. Shallow-level fractional crystallization, under relatively dry conditions, of plagioclase, augite, olivine, and Fe-Ti oxides can explain most of the observed variations, although the differences in P2O5, TiO2, and incompatible trace element contents call for at least two distinct evolutionary paths. Moreover, the REE patterns crossing each other cannot be explained by a fractional crystallization process, but may reflect variable degree of partial melting.
Felsic volcanic rocks
The felsic volcanic rocks of Pantelleria cover the compositional spectrum from comenditic trachyte (SiO2 ~63 wt %), through pantelleric trachyte (SiO2 ~66 wt %), to pantellerite [SiO2 between 67 and 72 wt %;
Table 4 and
Civetta et al., (1984)].
The comenditic trachyte is separated from the mafic volcanic rocks by a silica gap of 11 wt %. Conversely, the compositional trend from comenditic trachyte to pantellerite is continuous and characterized by an increase of SiO2 from 62 to 72 wt %, accompanied by a decrease of Al2O3, P2O5, CaO, MgO, TiO2 and Fe2O3tot, a slight decrease of Na2O, and an increase of K2O followed by a decrease (
Fig. 6 and
Table 4). Overall, the felsic volcanic rocks are characterized by a regular and strong enrichment in LREE, Rb, Zr, Nb, Y, and Th, whereas Ba, Sr, Co, Ni, Sc decrease from comenditic trachyte to pantellerite (
Fig. 7 and
Table 4). These variations are well correlated with the peralkalinity of the rocks, as the agpaitic index increases with increasing incompatible trace element contents.
The main features of the chondrite-normalized REE patterns (
Fig. 9) as differentiation increases from comenditic trachyte to pantellerite are: (1) increase of total concentration of REE, (2) slight increase of LREE relative to HREE enrichment (LaN/LuN ranges from 8·07 to 10·29), (3) development of a progressively negative Eu anomaly owing to the dominant feldspar fractionation, with Eu/Eu* ranging from 1·04-0·76 in comenditic trachytes to 0·48-0·38 in pantellerites (
Table 5). The overall chemical variations observed in the whole basalt-comenditic trachyte-pantellerite sequence at Pantelleria can be attributable at a first sight to fractionation process(es) occurring in a closed-system crustal reservoir(s). The available chemical data can be used to test hypotheses of derivation of: (1) the felsic magmas from the mafic magmas, and (2) the more differentiated pantelleritic magmas from the less differentiated comenditic trachytic magmas.
Mass balance calculations, based on both the major and trace element compositions of whole-rock samples, and mineral chemical data, were performed to test quantitatively the hypothesized petrogenetic processes. Major element compositions were modelled using the method of
Stormer & Nicholls, (1978), which consists of subtracting suitable minerals from a rock representing a parental magma, to create a calculated daughter magma. The composition of this calculated daughter magma is then compared with the composition of a rock chosen as representative of a possible natural daughter magma. The match between calculated and observed major element compositions of the daughter magma is considered to be acceptable when the sum of the squares of residuals ([Sigma]r2) is less than unity. Trace element compositions of daughter magmas were calculated starting from the results of the major element modelling, using the Rayleigh fractionation equation, and partition coefficients taken from the literature on similar rocks
(Lemarchand et al., 1987;
Mahood & Stimac, 1990;
Caroff et al., 1993). The match between calculated and observed trace element compositions of the daughter magma is considered to be acceptable when the differences are within the analytical uncertainty. The results are given in
Table 6. Table 6. Results of mass balance calculations on volcanic rocks of Pantelleria.
Two different petrogenetic processes have been modelled. The first one involves derivation of the more differentiated magmas, comenditic trachytes and pantellerites, from a basaltic parental magma, using the most evolved basalts as starting compositions (i.e. samples Sic 17a, low-Ti-P, and Opl 1001, high-Ti-P). The results of the calculation demonstrate that it is possible to derive a comenditic trachytic magma from both a low-Ti-P and a high-Ti-P basaltic magma by subtracting about 83 and 77%, respectively, of a mineral assemblage involving dominant plagioclase and clinopyroxene, and subordinate Ti-magnetite, olivine, apatite and ilmenite. Moreover, a pantelleritic magma can also be obtained from a basaltic parent magma by subtracting ~95% of a similar mineral assemblage, with addition of alkali-feldspar and aenigmatite, but without ilmenite. This process has not been modelled using trace element compositions, as partition coefficients are likely to be very variable. The modelling based on major element compositions gives good results, with [Sigma]r2 <0·19 for the basalt to comenditic trachyte transitions, and <0·03 for the basalt to pantellerite transitions. The type and relative percentages of mineral phases involved are in substantially good agreement with the petrographic evidence. The trace element composition of the calculated daughter magmas is also in good agreement with the observed data. This result is of some importance in that it sustains the possibility of a genetic linkage between mafic and felsic volcanics at Pantelleria.
The second attempt at modelling involved deriving a pantelleritic magma from a comenditic trachytic parent magma by subtracting a mineral assemblage with highly dominant anorthoclase and subordinate aenigmatite, olivine, clinopyroxene or amphibole, and apatite. The results require high amounts of crystallization, namely ~77% when clinopyroxene is involved ([Sigma]r2 <0·22), and ~82% when amphibole is involved ([Sigma]r2 <0·12). Both are in agreement with the relatively high amount of anorthoclase-dominated mineral assemblages found in comenditic trachytic samples. Good matches have also been obtained based on trace element compositions. Similar results were obtained by
Civetta et al., (1984) using mineral compositions taken from the literature. It is noteworthy that fractionation models involving amphibole give a better match between calculated and observed daughter magma compositions, suggesting a possible fundamental role for that mineral phase. This has been shown to be true in mafic to felsic peralkaline rocks suites from Terceira island, Azores
(Mungall & Martin, 1995).
The two modelled processes, namely (1) derivation of comenditic trachytic and pantelleritic magmas by fractionation of a basaltic parent magma and (2) derivation of pantelleritic magmas from a comenditic trachytic parent magma, are both mathematically possible. Obviously, these results do not prove that basaltic magmas at Pantelleria have generated comenditic trachytic and in turn pantelleritic magmas by protracted processes of fractional crystallization, but simply confirm that this possibility is not ruled out mathematically. Geological, geophysical, geochemical and petrographic evidence support this possibility, as it will be clear in the following discussion. The isotopic compositions of strontium, neodymium and lead, and Rb, Sr, Sm, Nd, U, Th and Pb concentrations in representative mafic and felsic volcanic rocks are reported in
Table 5. The 87Sr/86Sr ratios of the felsic volcanic rocks were age-corrected using the ages reported in
Table 4. The correction is necessary given the high Rb/Sr ratios of these very young samples, and is in the range 0·00001-0·00005. All the 87Sr/86Sr ratios were normalized to the accepted value of 0·710235 for the NBS-987 international reference standard. Age-corrected Nd and Pb isotope ratios do not vary by more than 1 part in 108 and 104, respectively, and therefore are not reported.
Previous Sr isotope studies of basalts, comenditic trachytes and pantellerites from Pantelleria were reported by
Barberi et al., (1969),
Korringa & Noble, (1972) and
Civetta et al., (1984). The published Sr isotope ratios for the basalts
(Civetta et al., 1984) are lower than those reported in this study by ~0·00015 because the old isotope data were normalized to a different reference standard (Eimer & Amend, 87Sr/86Sr = 0·70800), whereas the new analyses are normalized to NBS-987 = 0·710235.
The 87Sr/86Sr ratios of the mafic volcanic rocks (basalts and the hawaiite) (
Table 5) range between 0·70299 and 0·70320. The Sr isotope composition is positively correlated with both compatible (e.g. Sr) and incompatible (e.g. La and Nb) trace element contents (
Fig. 10).
Civetta et al., (1984) found that the 87Sr/86Sr ratio of the low-Sr pantellerites was in the range 0·70335-0·70852. This wide range in Sr isotope composition was interpreted to be the result of a combined effect of fractional crystallization and contamination with upper-crustal igneous rocks and/or marine sediments. Given the very low Sr content of the pantellerites, even small amounts of contamination by a radiogenic crustal component would have resulted in an increase in 87Sr/86Sr ratio. The leaching procedure, used in the present study before the determination of Sr isotope composition in the Sr-poor felsic volcanic rocks and their feldspar separates, efficiently removed most of the foreign Sr, thus resulting in a much lower range of 87Sr/86Sr ratios in these rocks than previously reported (
Table 5). The 87Sr/86Sr ratio of the feldspar separated from the comenditic trachytes and pantellerites ranges from 0·70308 to 0·70389, with the majority of the samples with 87Sr/86Sr <0·70324. This range partially overlaps that measured in the mafic volcanic rocks (0·70299-0·70320).
In plots of 87Sr/86Sr vs TiO2, Nb, La and Sr (
Fig. 10) two groups of samples are evident: one group, comprising all the mafic volcanic rocks, the comenditic trachytes and some pantellerites, is less enriched in radiogenic Sr, with 87Sr/86Sr ratios between 0·7030 and 0·7033. The other group, comprising only some pantellerites, is more enriched in radiogenic Sr. This group shows 87Sr/86Sr ratios always higher than 0·7034, and negative correlations between 87Sr/86Sr ratio and La and Nb contents. These higher Sr isotopic ratios could be due to the effect of foreign strontium on the Sr-poor volcanic rocks, incompletely removed by the leaching procedure adopted, as shown by the relationship Sr-87Sr/86Sr (
Fig. 10). Processes of crustal contamination during storage in magma chamber(s) are thought to be ruled out by the negative correlations observed between Sr isotope ratio and La and Nb contents, which are the reverse of what would be commonly expected. Thus, no significant differences in terms of Sr isotope composition appear between the mafic and felsic volcanic rocks. On the other hand, a difference in 87Sr/86Sr ratio is evident for the two groups of basalts, the high-Ti-P and the low-Ti-P: 87Sr/86Sr ratio is generally higher in the high-Ti-P basalts, and correlates positively with TiO2 and La, suggesting source heterogeneity.
As for Sr isotopes, no significant difference in Nd isotope composition has been found between the mafic and felsic volcanic rocks, although the highest 143Nd/144Nd ratios have been detected in the latter. The 143Nd/144Nd ratio of basalts and hawaiite ranges from 0·51292 to 0·51299; that of the comenditic trachytes from 0·51299 to 0·51302, and that of the pantellerites from 0·51297 to 0·51301 (
Table 5).
Pb isotope compositions vary considerably within the mafic volcanic rocks, despite the minor ranges detected in Sr and Nd isotopes. 206Pb/204Pb ratio ranges from 19·13 to 19·94; 207Pb/204Pb ratio from 15·60 to 15·67; 208Pb/204Pb ratio from 39·08 to 39·62. Comenditic trachytes and pantellerites display smaller variations in 206Pb/204Pb ratio, ranging from 19·48 to 19·69; 207Pb/204Pb ratio, ranging from 15·59 to 15·65; and 208Pb/204Pb, ranging from 39·02 to 39·24 (
Table 5). Similar ranges for Pb isotopes were reported by
Mahood et al., (1990; e.g. 206Pb/204Pb = 19·05-19·98), whereas
Esperança & Crisci, (1995) reported values for 206Pb/204Pb ratio as low as 18·48. Our study has demonstrated that similar or even lower values for Pb isotope ratios can been obtained on some samples when analysed without acid leaching (particularly samples Sic 5 and Sic 17a,
Table 5). Thus, we suggest that some basaltic units on Pantelleria might have been affected by some kind of contamination by foreign lead, possibly supplied by the active geothermal system of the island
(Fulignati et al., 1997).
It is interesting to note that a correlation exists between the age of eruption, geochemical characteristics, and the Sr, Nd and Pb isotope ratios of the basaltic rocks. In fact, basalts erupted before 50 ka BP (i.e. Opl 1001, Opl 120b, Sic 5 and Sic 3,
Tables 3 and
5), all having high TiO2 and P2O5 contents, show 87Sr/86Sr ratio in the range 0·70310-0·70320, 143Nd/144Nd ratio in the range 0·51294-0·51298, and
206Pb/204Pb ratio in the range 19·59-19·94. Conversely, basalts erupted after 50 ka BP (i.e. Sic 16, Sic 17a Sic 48 and Opl 104b) have low TiO2 and P2O5 contents, and 87Sr/86Sr ratio in the range 0·70299-0·70311, 143Nd/144Nd ratio in the range 0·51292-0·51299, and
206Pb/204Pb ratio in the range 19·13-19·72. Furthermore, high-Ti-P basalts have LaN/LuN and TbN/YbN ratios generally higher than low-Ti-P basalts (
Table 5 and
Fig. 9). Thus, on average the high-Ti-P group includes basalts with higher 87Sr/86Sr, 206Pb/204Pb, LaN/LuN and TbN/YbN, and less variable 143Nd/144Nd ratios, relative to the low-Ti-P group of basalts.
Such relationships between geochemical and Sr-Nd-Pb isotopic characteristics, and age of eruption in the volcanic rocks of Pantelleria have not been reported by previous work and will be discussed in the following section, with reference to their implications for the source region characterization. The petrogenesis of the peralkaline felsic magmas at Pantelleria, and their genetic relationship with the associated mildly alkaline mafic magmas, has been explained in the literature essentially by two different mechanisms: (1) a protracted fractional crystallization process starting from a parental alkali basaltic magma, which gives rise to trachytic and then pantelleritic magmas (e.g.
Civetta et al., 1984); (2) partial melting of an alkali gabbroic cumulate, giving rise to a trachytic magma which in turn, by a subsequent fractional crystallization process, gives rise to all the felsic peralkaline magmas, up to the pantellerites, as suggested by
Mahood et al., (1990) and
Lowenstern & Mahood, (1991).
Any genetic hypothesis that can be envisaged to link the felsic to the mafic magmas has to take into account the following geological, petrographical and geochemical evidence:
(1) The volume of peralkaline volcanics exposed on the island is much greater than that of the mafic volcanic rocks. On the other hand, a 1·2 km deep drill hole in the northern sector of the island
(Fulignati et al., 1997) has shown that the eruption of huge volumes of basaltic magma pre-dates the peralkaline magmatism. Furthermore, mafic volcanism is widespread all along the rift of the Sicily Channel, and occurred in the neighbourhood of the island even in recent times
(Beccaluva et al., 1982;
Calanchi et al., 1989). Linosa, the other volcanic island in the Sicily Channel (
Fig. 1), located on a flank of the rift, is composed mainly of mafic alkaline volcanic rocks. In a study of the alkali basalt-pantellerite sequence erupted at Las Navajas volcano in Mexico,
Nelson & Hegre, (1990) noticed that the volume of mafic alkaline volcanics is much higher than in other areas of peralkaline volcanism in Mexico, where only mildly peralkaline comendites occur, whereas pantellerites are absent. Those researchers suggested that high abundance of mafic alkaline magma could be the fundamental factor in providing the initial volume of magma necessary to drive the differentiation process to the extreme peralkaline compositions.
(2) Mafic volcanic rocks occur only outside the caldera on the island of Pantelleria. This distribution provides good evidence for the existence of a shallow reservoir in which mafic magmas might have stagnated and fractionated to produce felsic magmas
(Civetta et al., 1984,
1988;
Mahood & Hildreth, 1986;
Orsi et al., 1991b).
(3) Glomeroporphyritic clots of feldspar, augite, Fe-Ti oxide ± olivine occur in the least evolved comenditic trachytes. This can be taken as evidence for fractional crystallization processes from more mafic parental magmas, as suggested by
Parker, (1983) for the Paisano volcano in Texas.
(4) The trends described by major and trace elements when plotted against differentiation indices are continuous, at least within the felsic volcanic rocks (
Figs 6 and
7). The felsic volcanic rocks fall along the thermal valley between the feldspar minimum (Or30) on the Ab-Or join and the minimum on the Qz-feldspar cotectic boundary at P(H2O) = 1000 kg/cm2 when projected into the NaAlSi3O8-KAlSi3O8-SiO2-H2O-Na2SiO3- NaFeSi2O6 system
(Carmichael & MacKenzie, 1963;
Fig. 11). It is clear that all the felsic volcanic rocks constitute a single progressive trend, starting from comenditic trachytes, through pantelleritic trachytes, up to pantellerites. It is noteworthy that the higher the silica content (and AI) of the rock, the closer is the rock to the thermal minimum; moreover, the majority of the samples showing quartz phenocrysts plot within the 720°C isotherm, i.e. closer to the thermal minimum. In particular, quartz appears as a phenocryst, sometimes intergrown with anorthoclase, suggesting a cotectic relation, in samples with normative Qz >27% and SiO2 content >70·4 (on an anydrous basis). The trend of whole-rock and glass samples is mirrored by that of the anorthoclase phenocrysts analysed from these rocks, whose compositions are projected on the Ab-Or join in
Fig. 11, becoming progressively richer in the orthoclase component, approaching the limiting composition of Or38
(Carmichael & MacKenzie, 1963), as already discussed in the mineral chemistry section. This strongly suggests that anorthoclase fractionation was important in the petrogenesis of all the felsic volcanic rocks at Pantelleria. This is also in agreement with the strong depletion of Ba and Sr displayed by the felsic volcanic rocks, subtracted from the liquid by the crystallizing feldspar (
Table 4 and
Fig. 7). On the other hand, all the incompatible trace elements are more and more strongly enriched as fractionation proceeds, thus indicating that the fractionating phases (anorthoclase and minor clinopyroxene and/or amphibole) do not allow significant incorporation of such elements in their lattices. Incompatible trace element pairs display linear covariations passing through zero throughout the whole basalt-comenditic trachyte-pantellerite sequence, with very good correlation coefficients (R = 0·97-0·99), suggesting a closed-system differentiation process (
Fig. 12). Very striking is the behaviour of REE, which show substantially similar patterns from basalts through comenditic trachytes and pantelleritic trachytes to pantellerites, with a negative Eu anomaly in the felsic volcanic rocks, deepening regularly from comenditic trachytes to pantellerites, also suggesting alkali-feldspar as the dominant fractionating phase (
Fig. 9). The observed slight change in LaN/Lu
N ratio from mafic (9·28-19·31;
Table 5) to felsic volcanic rocks (8·07-10·29), can be accounted for by the involvement of apatite and clinopyroxene during the fractional crystallization process, which have different distribution coefficients for LREE and MREE.
(5) Further evidence for genetic links between the felsic volcanic rocks is provided by the observation that many explosive eruptions at Pantelleria (e.g. the Green Tuff eruption) gave rise to products that are compositionally zoned from first erupted pantellerite to last erupted pantelleritic trachyte, or comenditic trachyte
(Civetta et al., 1984,
, 1988).
(6) Rocks trending to or of intermediate composition (i.e. mugearites, benmoreites) are absent at Pantelleria. Bimodality seems to be a common feature of peralkaline magmatism, as it is shared by other localities around the world, for instance the Paisano volcano (Trans-Pecos Texas;
Parker, 1983) and the Kenya Rift Valley
(Macdonald et al., 1994). However, examples of complete basalt-pantellerite sequences are not uncommon. At Las Navajas volcano, located in an area at present undergoing extension in Mexico, a complete hawaiite-mugearite-benmoreite-trachyte-peralkaline rhyolite association occurs and has been attributed to the low-pressure fractional crystallization of a single alkali basaltic parent magma
(Nelson & Hegre, 1990). At the Boina volcanic centre (Afar Rift, Ethiopia), a basalt-ferrobasalt-dark trachyte-trachyte-comendite- pantellerite sequence has been recognized and attributed to a single low-pressure fractional crystallization process by
Barberi et al., (1975). These workers have shown that a critical change in P(O2), involving a sudden drop followed by a rapid increase, occurred during the crystallization of Fe-Ti oxides near the transition to peralkalinity. They claimed that this crucial stage would control marked chemical variations (Eu content, Fe2+/Fe3+ ratio) in a restricted crystallization interval, responsible for the scarcity of rocks of intermediate composition commonly found at this stage, the so-called Daly gap. The less evolved trachytes occurring at Pantelleria are peralkaline (AI >1,
Table 4), however, and thus the critical change in P(O
2) must have occurred already, and therefore this hypothesis cannot be verified. Other workers have attributed the virtual absence of intermediate volcanic rocks to eruption dynamics
(Mungall & Martin, 1995). Magmas of intermediate composition are likely to contain such a huge number of crystals that their high viscosity prevents them from rising to the surface
(Marsh, 1981). In fact, the very few rocks of benmoreitic composition found at Pantelleria are strongly porphyritic xenoliths
(Civetta et al., 1984), which may represent the products of solidification of magmas that were not able to reach the surface because of their high crystal content. The critical change in P(O2), the complex eruptive dynamics, and the distinct rheologic properties of the intermediate magmas may all contribute to the presence of a Daly gap in the eruptive products of Pantelleria.
The systematic study of the temporal and stratigraphical relationships of the mafic and felsic volcanic rocks of Pantelleria and their Sr, Nd and Pb isotope compositions has shown that the majority of the <50 ka BP comenditic trachytes and pantellerites have Sr, Nd and Pb isotope ratios in the same range as those of the basalts of similar age (
Table 5). This supports least-squares modelling results, based on major and trace element data (
Table 6), showing that it is possible to derive a comenditic trachyte magma from a basaltic parent magma, and a pantelleritic magma from a comenditic trachyte magma by simple fractional crystallization processes. Modelling the derivation of felsic peralkaline magmas by partial melting of mafic igneous rocks would require a much better knowledge of the partitioning of elements between crystals, liquid and vapour phases during partial melting and subsequent ascent than is currently available, as discussed by
Macdonald, (1987). Such modelling has been demonstrated to be unsuccessful in some cases (e.g. at Terceira island, Azores;
Mungall & Martin, 1995), particularly in that the final felsic magmas would be far too enriched in Sr and REE. Consequently, it has not been considered worth while in the present study.
Lowenstern & Mahood, (1991) suggested a model whereby partial melting of alkali gabbros at Pantelleria generated pantelleritic trachytic magmas, which in turn gave rise to more differentiated pantelleritic magmas by low-pressure fractional crystallization, and to complementary cumulates erupting as crystal-rich trachytes. A cumulative origin for comenditic trachytes is ruled out by REE evidence (Eu/Eu* = 1·04 to -0·76,
Table 5). Lowenstern & Mahood's model is based on measurements of the magmatic H2O content of pantellerites, and on the assumption that H2O behaves as a strongly incompatible element, as high-field strength elements do; those workers claimed that a fractional crystallization process from alkali basaltic parent magma would result in too high a water content in the final pantelleritic magmas, ~8 wt %, against observed contents ranging between 1·4 and 2·1 wt %. Our modelling based on major and trace element compositions demonstrates that amphibole can play an important role during the fractional crystallization from comenditic trachyte to pantellerite (14-18%,
Table 6), thus reconciling the observed low H2O content of pantellerites.
Mungall & Martin, (1995) pointed out the importance of amphibole fractionation in driving the residual magmas toward peralkalinity.
Thus, by combining the geological, petrographical and geochemical evidence, the results of least-squares calculations and the isotopic data, we favour the hypothesis of derivation of the felsic magmas by process(es) of fractional crystallization of the observed mineral phases from mafic magmas similar to those erupted in the northwestern corner of Pantelleria, thus corroborating previous suggestions based on geological and geochemical grounds
(Civetta et al., 1984,
, 1988). However, mineral chemical data, particularly those for clinopyroxenes, suggest a more complex evolutionary history for the basaltic magmas. In fact, a pressure range of 0-4 kbar has been calculated for crystallization of clinopyroxene in the mafic volcanic rocks of Pantelleria using the clinopyroxene geobarometer recently proposed by
Nimis, (1995). In particular, Al, Ti-poor diopsides appear to have crystallized at higher pressure than Al, Ti-rich augites. Thus, the simultaneous presence of diopside and augite in the basalts of Pantelleria suggests either (1) polybaric fractionation of homogeneous basaltic magmas carrying clinopyroxenes having crystallized in the pressure range 0-4 kbar, or (2) mixing between basaltic magmas carrying clinopyroxene phenocrysts showing different evolutionary trends, i.e. an Fe-enrichment trend at high, constant Ca contents and an Fe-enrichment trend at decreasing Ca content, the former typical of moderately alkali basalts and the latter typical of transitional, less alkali basalts (e.g. Dal
Negro et al., 1982). Whatever the origin of the petrographic evidence for mineralogic disequilibrium may be, it must have occurred while the mafic magmas were rising through the lithosphere and fractionating. Geochemical and Sr-Pb isotope data suggest that the basalts of Pantelleria are not similar to one another, but have been derived from distinct parental magmas coming from a heterogeneous mantle source, as will be discussed subsequently. During the last 50 ka of volcanic history six eruptive cycles have occurred on Pantelleria, recognized on the basis of field relationships, the occurrence of palaeosols and geochronological data
(Civetta et al., 1984,
, 1988;
Orsi et al., 1991b). During each cycle, progressively less differentiated magmas were erupted with time, suggesting that the eruptions were fed by a reservoir filled by magma geochemically less differentiated downwards. The longer the repose time after each eruptive cycle, the more differentiated was the magma erupted when activity resumed during the following cycle.
The six cycles were distinguished by studying the silicic products (mostly comenditic trachytes and pantellerites) emplaced during the last 50 ka. It has been shown also that the geochemical features of the basaltic volcanic rocks erupted on the island of Pantelleria before and after 50 ka BP are different in terms of TiO2, P2O5 and incompatible trace element contents, as well as in some isotopic characteristics. This geochemical difference can be related to the important volcano-tectonic event that characterized the volcanic history of Pantelleria at 50 ka BP; namely, the formation of the Monastero caldera, which followed the eruption of the Green Tuff
(Cornette et al., 1983). This caldera formation must have been triggered by a regional tectonic event also responsible for the magmatic system being isolated from the feeding zone that furnished the high-Ti-P basaltic melts. Afterwards, the magmatic system was fed only by low-Ti-P basaltic melts, perhaps coming from shallower depths. A similar situation has been noted from the island of Stromboli, where the chemical composition of the erupted magma changed after a major caldera collapse (
Francalanci et al., 1993a).
In
Fig. 13a and b the Sr isotope composition and the Zr content of the products erupted during the last 50 ka are plotted against the time of eruption. It is evident that the comenditic trachytic magmas (Zr content <1000 ppm) are erupted at 50 (Green Tuff), 35 (Montagna Grande) and at ~6 ka BP. This implies that only during these cycles was the magma chamber tapped at the deepest levels. Interestingly, inspection of the plot of
Fig. 13b shows that the degree of evolution of the erupted magmas, represented by the Zr content of the most differentiated rock of each cycle, increases from 50 to 4 ka BP. As already noticed, the 87Sr/86Sr ratio is higher in the most differentiated pantellerites, as the Sr content in these rocks is very low, and surficial contamination has been more efficient.
The youngest eruptive cycle occurred at 10-4 ka BP, and has been investigated in more detail through field work, 14C dating (
Table 7) and geochemical analysis of volcanic rocks collected at different stratigraphic height for each unit.
Figure 13c shows the Zr content of the products of this cycle against the age of eruption. The plot shows that the degree of evolution of the magma erupted from 10 to 8 ka BP increased with time, then it decreased from 8 to 6 ka BP, and then it increased again from 6 to 4 ka BP. The chemistry-time relationships highlight that the magmatic system has been tapped at progressively deeper levels in the time span 8-6 ka BP, during which it has not been able to give rise to a highly differentiated cap, probably because the eruption rate has been higher than the differentiation rate. Conversely, between 6 and 4 ka BP the differentiation rate was dominant with respect to the eruption rate, and more strongly differentiated magmas were erupted again. Table 7. 14C dating results on selected samples younger than 10 ka.
Notwithstanding their relatively evolved character, the isotopic and incompatible trace element characteristics of the mafic volcanic rocks allow us to infer the main characteristics of the source region of the parent magmas feeding the Pantelleria system. According to REE data (
Table 5;
Fig. 9), this source region should be located in the spinel-garnet lherzolite transition zone (see, e.g.
McKenzie & O'Nions, 1991), as low-Ti-P and high-Ti-P basalts show a moderate MREE to HREE fractionation, with TbN/YbN in the range 1·82-2·18, and 2·21-2·63, respectively. These ranges are comparable with those of alkali basalts from Hawaii (TbN/YbN = 1·89-2·45;
Frey et al., 1991), which are commonly considered to have been generated in a garnet-bearing lherzolitic mantle
(Frey et al., 1991;
McKenzie & O'Nions, 1991).
On the basis of major and trace element geochemistry, it has been pointed out that the basalts of Pantelleria have a range of Ti, P and other incompatible trace element abundances and ratios at a given MgO content (
Table 3 and
Fig. 8). Furthermore, they have different REE concentrations with highly variable LaN/LuN (9·28-19·31), so that the REE patterns of distinct basalt samples cross each other (
Fig. 9), and fractionated TbN/YbN ratios (1·82-2·63). Lastly, they display a wide range of Pb and, to a minor extent, Sr and Nd isotope compositions (
Table 5). All these geochemical and isotopic features are correlated with the age of emplacement. Although different LaN/LuN and TbN/YbN ratios could in part be due to the effect of variable degree of partial melting of a garnet-bearing lherzolitic mantle source, this cannot account for the isotopic variability of the most primitive magmas. Thus, it is most likely that the analysed samples represent distinct parental basaltic magmas, derived from a rather heterogeneous mantle source, with different sectors activated at distinct times as testified by differences in the geochemical and isotopic characteristics of the erupted basalts. A heterogeneous mantle source has also been proposed by
Mahood & Baker, (1986), based on major and trace element data for the Pantelleria basalts, although no chemistry-time relationships have been recognized by those workers. Alternatively, the range in isotope ratios has been attributed to shallow-level processes (e.g. crustal contamination) rather than mantle source heterogeneity by
Esperança & Crisci, (1993,
, 1995).
The combination of lower Sr isotope ratios, and higher Nd isotope, Rb/Sr and LREE/HREE ratios (
Table 5) than those for Bulk Earth suggests that the basaltic magmas originated from a mantle source characterized by a time-integrated depletion in incompatible elements, which has been recently enriched in some incompatible elements, such as Rb and LREE (
Fig. 14). In terms of Sr and Nd isotope ratios, this source appears fairly homogeneous, and not very different from that of enriched mid-ocean ridge basalts (E-MORB). In fact, all samples from Pantelleria lie in the upper left, depleted quadrant in the 87Sr/86Sr vs 143Nd/144Nd ratios diagram, extending from the lower end of the data field for Pacific and Atlantic MORB, and overlapping that for tholeiitic volcanic rocks of Mt Etna (eastern Sicily) and alkali syenitic intrusives of Pietre Nere (a locality on the Adriatic side of Italian peninsula,
Fig. 1). Four representative Na-alkali basaltic samples from Linosa, located in the Sicily Channel southwest of Pantelleria (
Fig. 1), plot close to the Pantelleria field and suggest a common source for the basaltic magmas of the whole Sicily Channel Rift Zone.
It is interesting to note that the Pb isotope ratios of some of the mafic and felsic volcanics of Pantelleria are strongly radiogenic, similar to values for ocean island basalts (OIB) characterized by a component with high U/Pb ratio (HIMU) in their source region, such as at St Helena and Tubuaii Island
(Zindler & Hart, 1986). This is clearly evident in diagrams of
207Pb/204Pb and 208Pb/204Pb vs 206Pb/204Pb (
Fig. 15): data points for Pantelleria and Linosa define an extended trend crossing the Northern Hemisphere Reference Line (NHRL,
Hart, 1984), only partially overlapping the data field for Atlantic and Pacific MORB, with most of the samples strongly displaced towards the inferred position of HIMU. Interestingly, similarly radiogenic values for Pb isotopes in southern Italy have been recorded at Mt Etna volcano and Mts Iblei in eastern Sicily
(Carter & Civetta, 1977;
Esperança et al., 1994;
Hoernle et al., 1996), as well as at the island of Ustica (L. Civetta et al., unpublished data, 1998), located north of Sicily (
Fig. 1). Moreover, basalts from all these areas display Sr and Nd isotope ratios in the same range as basalts from Pantelleria (
Fig. 14)
(Armienti et al., 1989;
D'Orazio, 1994;
Marty et al., 1994). All these basaltic rocks converge towards, or are part of, the Low Velocity Component (LVC), represented as a box in
Figs 14-
16. The LVC, recently recognized by
Hoernle et al., (1995), is a common asthenospheric component found throughout the eastern Atlantic, western Mediterranean, and western and central Europe. The LVC is similar to HIMU, although differing from it because of less radiogenic 206Pb/204Pb ratios.
Further evidence for the involvement of distinct components, both depleted and enriched, in the source region of SCRZ mafic magmas comes from the 87Sr/86Sr and 143Nd/144Nd vs 206Pb/204Pb diagrams (
Fig. 16a and b). Mafic volcanic rocks from Pantelleria and Linosa lie between the inferred positions of the Depleted MORB-source Mantle (DMM) and HIMU components of
Zindler & Hart, (1986), or alternatively, close to the LVC of
Hoernle et al., (1995). However, the trend does not point to the position of the DMM component in
Fig. 16, suggesting the possible minor involvement of a third, enriched component. This could be either (1) an EM 1-like component (Enriched Mantle 1 component;
Zindler & Hart, 1986), i.e. ancient pelagic sedimentary material recycled in the deep mantle, or (2) an EM 2-like component, i.e. ancient terrigenous sedimentary material recycled in the deep mantle, or relatively young sedimentary material recently subducted. In fact, several trace element ratios
(Weaver, 1991;
Chauvel et al., 1992) give further evidence for the involvement of both HIMU and EM 1 components in the source region of the Pantelleria basalts. Values of Ce/Pb, Zr/Nb, K/Nb, Ba/Th and Ba/La ratios listed in
Table 5 suggest strongly the involvement of HIMU in the source region of high-Ti-P basalts of Pantelleria. Furthermore, Ba/La and Ba/Nb ratios favour the involvement of EM 1 and/or EM 2 components, although the low Th/La ratios exclude EM 2. The trace element ratios of the low-Ti-P basalts also require both HIMU and EM 1, although the latter is more supported than the former, in agreement with the lower 206Pb/204Pb ratios. In detail, the Ce/Pb ratio has been shown to be particularly indicative of a HIMU affinity in oceanic basalts
(Chauvel et al., 1992): it ranges from 27 to 55 at Tubuaii Island and from 26 to 56 at Pantelleria (
Table 5). The overall distribution of incompatible elements in the basalts of Pantelleria (
Fig. 17) is surprisingly similar to that of basalts from oceanic islands with typical HIMU isotopic signatures, such as Tubuaii (Pacific Ocean). On the other hand, the Pantelleria basalts also show enrichment in K, Rb and Ba relative to that of HIMU OIB, a feature typical of basalts with an EM 1 signature.
It is noteworthy that most trace element ratios in the basalts of Pantelleria are far lower than values typical of continental crust, thus ruling out any appreciable involvement of continental crustal material in the pe- trogenesis of the magmas, as well as any surface contamination by it, supporting combined Pb-Sr-Nd isotope data and field evidence for lack of sedimentary xenoliths in any outcropping volcanic units. In particular, Ce/Pb ratios (26-56,
Table 5) are far higher than the average ratios for both typical continental crust (Ce/Pb = 4;
Hofmann, 1988) and Calabrian crust (Ce/Pb = 9·5;
Caggianelli et al., 1991;
Rottura et al., 1991;
Francalanci et al., 1993b), the latter thought to be a possible contaminant in the area. The only exception is Ba/Th ratios (70-150,
Table 5), higher than values typical of HIMU, although in the range of EM 1.
Inspection of
Figs 14 and
15 suggests that the source region of the Pantelleria magmas is not exclusive to the magmas feeding its system, nor to those feeding the volcanism of the whole Sicily Channel Rift Zone, but rather it is widespread in the whole southern sector of Italy, given a similar range in values for Sr, Nd and Pb isotope ratios of the igneous rocks of eastern Sicily and Ustica Island, and even Pietre Nere. Different hypotheses can be envisaged to explain the distinctive isotopic features of this mantle source. It should have retained high U/Pb and Th/Pb ratios for a time long enough to develop high 206Pb/204Pb and 208Pb/204Pb ratios. A similar situation could occur either in subducted material brought deep into the mantle and successively recycled in the melting region, as commonly suggested for the enriched components found in the source of OIB, including the component HIMU (e.g.
Zindler & Hart, 1986;
Chauvel
et al., 1992), or in the subcontinental lithospheric portion of the upper mantle, which is suspected to participate in the melting process in continental environments and thought to be capable of developing strong geochemical and isotopic heterogeneities [e.g.
Menzies & Hawkesworth, (1987) and quoted references]. On the other hand, the lower 87Sr/86Sr, and higher 143Nd/
144Nd values than Bulk Earth indicate that the source of the magmatism of the Sicily Channel, and probably also that of eastern Sicily, has been also depleted in incompatible elements for a very long time. This, along with the strong homogeneity of Sr and Nd isotope ratios, casts some doubt on the lithospheric mantle as the source region, as long-term depletion in incompatible elements would have produced Nd isotopic characteristics more radiogenic than those observed.
Melting of the subcontinental lithospheric mantle as a process responsible for the generation of the parental magmas has been suggested for the Mts Iblei
(Esperança et al., 1994;
Beccaluva et al., 1998) and the Pantelleria
(Esperança & Crisci, 1993,
, 1995) basalts. Esperança & Crisci attributed the isotopic heterogeneity observed in the mafic volcanic rocks of Pantelleria to shallow rather than deep processes, claiming that the source region of the parental magmas is isotopically similar to the source region of the Tyrrhenian seafloor magmas, and that much of the Pb in the mafic magmas of Pantelleria could have been contributed by lithospheric mantle and Calabrian-type crust. The results of the present study have shown a completely different scenario, on the basis of geochemical and isotopic evidence. REE data (TbN/YbN = 1·82-2·63) suggest that magma generation occurred in the spinel-garnet transition zone, namely at a depth of 70-80 km at least
(McKenzie & O'Nions, 1991). Thus magma generation should have occurred well within the asthenosphere, given the reduced thickness of the lithosphere beneath the island because of stretching (60 km;
Della Vedova et al., 1989). Furthermore, recent studies based on geochemical and thermodynamic constraints
(Arndt & Christensen, 1992;
Anderson, 1994) have shown the unlikelihood of the lithosphere participating extensively in the melting process in continental areas. Thus, we favour a model whereby the source region of the magmatism of Pantelleria is the asthenosphere. Combined Sr-Nd-Pb isotope data (
Figs 14-
16) rule out any similarity between Tyrrhenian seafloor and Pantelleria source regions, as well as any appreciable involvement of Calabrian crust. Pb isotope data are striking in this respect, as contamination, or contribution to the source, by Calabrian crust would have modified easily the Pb isotope compositions of the resulting `contaminated' magmas, given the peculiar Pb isotopic characteristics and the high Pb contents of the Calabrian crust (1·5-34 ppm, average 16 ppm;
Caggianelli et al., 1991;
Rottura et al., 1991;
Francalanci et al., 1993b).
At least three distinct components are involved in the source region of the mafic magmas of Pantelleria: one is a MORB-like component, and can be thought as the uppermost portion of the asthenosphere. The second important component is thought to be responsible for the characteristic HIMU signature of some of the basalts, as suggested by the Pb isotope data and many trace element ratios. Furthermore, the presence of a third enriched component, EM 1-like, as suggested by the combined Sr-Pb-Nd relationships (
Fig. 15) and many trace element ratios (
Table 5 and
Fig. 17), is also compatible with a deep asthenospheric mantle origin
(Weaver, 1991;
Chauvel et al., 1992;
Hart et al., 1992).
A mantle plume with HIMU-EM 1 characteristics, interacting with the shallower DMM asthenospheric mantle, seems to be a suitable model to explain the origin of the SCRZ magmatism and much of its isotopic and geochemical characteristics. Preliminary 3He/4He data on Pantelleria (L. Civetta et al., unpublished data, 1998) give values similar to those available for Mt Etna basalts (~6·6 RA, where RA = the atmospheric value of 3He/4He,
Graham et al., 1992
a;
Marty et al., 1994), which resemble those of Pantelleria in terms of Pb-Nd-Sr isotope compositions. These are within the range of 3He/4He values for HIMU islands (5-7 RA;
Graham et al., 1992b;
Hart et al., 1992), and do not exclude a HIMU plume origin for the Mt Etna and SCRZ magmas as well. Furthermore, low 3He/4He values exclude any significant involvement in the petrogenesis of the Pantelleria mafic magmas of the mantle component recently proposed by
Hart et al., (1992), the Focus Zone (FOZO), which is supposed to carry a 3He-enriched signature (3He/4He = 10-35 RA). The mantle plume hypothesized as the main source of the parent magmas of Pantelleria might be either a steady-state plume, similar to those suggested to be located under newly forming continental rifts
(White & McKenzie, 1989), or linked to the large-scale mantle upwelling present beneath the eastern Atlantic, western Mediterranean, and western and central Europe
(Hoernle et al., 1995,
, 1996). Petrographic features, major and trace element, and Sr-Nd-Pb isotope data for mafic and felsic volcanic rocks from the island of Pantelleria, modelled using major and trace element data, allow us to establish that: (1) the mafic volcanic rocks are derived from a number of basaltic parental magmas with different degree of alkalinity, i.e. from hy- to ne-normative; (2) the felsic volcanic rocks are the products of prolonged closed-system fractional crystallization of parental mildly alkali basalts; (3) the source region for the mafic parental magmas was heterogeneous, probably because of the involvement of at least two distinct geochemical components: a MORB-type, relatively depleted component, and a HIMU-type enriched component, both located in the asthenospheric mantle; (4) a further enriched component, possibly EM 1-type, also of deep asthenospheric origin, could have been involved in the petrogenesis of mafic parental magmas, as suggested by combined Sr-Nd-Pb isotopic and incompatible trace element data; (5) the petrological data, supported by the geophysical evidence, allow us to propose a model for derivation of the mafic parental magmas of Pantelleria from an asthenospheric mantle plume, bringing the HIMU-EM 1 signature, rising through, and interacting with, the shallower depleted asthenospheric mantle bringing the MORB signature.
The authors wish to thank Marcello Serracino for his kind assistance with microprobe analysis; Riccardo Petrini for making some isotope dilution analyses of Rb and Sr; Mariagiovanna Capone, Sandro de Vita and Antonio Carandente for their help in collecting and preparing some samples; Federico Cella, Sandro Conticelli and Aldo Cundari for stimulating discussions and useful suggestions. Mariagiovanna Capone is thanked also for drawing
Fig. 1. Pb isotope analyses at UCSB by M.D. were supported by a grant from the Italian National Research Council (CNR), which is gratefully acknowledged. The research was carried out with the financial support of MURST (40%) and CNR to L.C. and G.O. Criticisms and suggestions by E. M. Piccirillo, K. Hoernle, S. Esperança and the Editor Marjorie Wilson greatly improved the manuscript and are appreciated.
INTRODUCTION
GEOLOGICAL BACKGROUND
SAMPLE SELECTION AND ANALYTICAL TECHNIQUES
Table 4. Major (wt %) and trace element (ppm) compositions of felsic volcanic rocks from Pantelleria.
CLASSIFICATION
PETROGRAPHY
MINERAL CHEMISTRY DATA
GEOCHEMISTRY
Major and trace element chemistry
Low-pressure evolution
Sr, Nd and Pb isotope compositions
DISCUSSION
Genetic relationships between mafic and felsic volcanic rocks
Chemistry vs time relationships
Characterization of the magma source region
CONCLUSIONS
ACKNOWLEDGEMENTS
REFERENCES