Journal of Petrology Advance Access originally published online on March 7, 2006
Journal of Petrology 2006 47(5):1017-1050; doi:10.1093/petrology/egl003
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
The Role of Lithospheric Mantle Heterogeneity in the Generation of Plio-Pleistocene Alkali Basaltic Suites from NW Harrat Ash Shaam (Israel)
1 DEPARTMENT OF GEOGRAPHY, BAR-ILAN UNIVERSITY, RAMAT-GAN 52900, ISRAEL
2 DEPARTMENT OF GEOCHEMISTRY, CENTER OF EARTH SCIENCE, UNIVERSITY OF GÖTTINGEN, 37077 GÖTTINGEN, GERMANY
3 INSTITUTE OF EARTH SCIENCES, THE HEBREW UNIVERSITY OF JERUSALEM, JERUSALEM 91904, ISRAEL
4 INSTITUTE OF MINERALOGY, UNIVERSITY OF HEIDELBERG, D-69120 HEIDELERG, GERMANY
RECEIVED AUGUST 9, 2004; ACCEPTED JANUARY 11, 2006
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONSAMPLING AND ANALYTICAL METHODSPETROGRAPHY AND CHEMISTRY OF...DISCUSSIONCONCLUSIONSREFERENCES |
|---|
Plio-Pleistocene volcanism in the Golan and Galilee (northeastern Israel) shows systematic variability with time and location: alkali basalts were erupted in the south during the Early Pliocene, whereas enriched basanitic lavas erupted in the north during the Late Pliocene (Galilee) and Pleistocene (Golan). The basalts show positive correlations in plots of ratios of highly to moderately incompatible elements versus the concentration of the highly incompatible element (e.g. Nb/Zr vs Nb, La/Sm vs La) and in diagrams of REE/HFSE (rare earth elements/high field strength elements) vs REE concentration (e.g. La/Nb vs La). Some of these correlations are not linear but upward convex. 87Sr/86Sr ratios vary between 0·7031 and 0·7034 and correlate negatively with incompatible element concentrations and positively with Rb/Sr ratios. We interpret these observations as an indication that the main control on magma composition is binary mixing of melts derived from two end-member mantle source components. Based on the high Sr/Ba ratios and negative Rb anomalies in primitive mantle normalized trace element diagrams and the moderate slopes of MREE–HREE (middle REE–heavy REE) in chondrite-normalized diagrams, we suggest that the source for the alkali basaltic end-member was a garnet-bearing amphibole peridotite that had experienced partial dehydration. The very high incompatible element concentrations, low K content, very low Rb contents and steep MREE–HREE patterns in the basanites are attributed to derivation from amphibole- and garnet-bearing pyroxenite veins. It is suggested that the veins were produced via partial melting of amphibole peridotites, followed by complete solidification and dehydration that effectively removed Rb and K. The requirement for the presence of amphibole limits both sources to lithospheric depths. The spatial geochemical variability of the basalts indicates that the lithosphere beneath the region is heterogeneous, composed of vein-rich and vein-poor domains. The relatively uniform 143Nd/144Nd (
Nd = 4·0–5·2) suggests that the two mantle sources were formed by dehydration and partial melting of an originally isotopically uniform reservoir, probably as a result of a Paleozoic thermal event. KEY WORDS: basanites; lithospheric heterogeneity; magma mixing; amphibole peridotite; pyroxenites
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONSAMPLING AND ANALYTICAL METHODSPETROGRAPHY AND CHEMISTRY OF...DISCUSSIONCONCLUSIONSREFERENCES |
|---|
Alkali basalts form a small part of the global basaltic inventory. However, the association with continental rift environments and the rich assemblages of mantle xenoliths they carry make them an important source of information on the composition of the uppermost mantle beneath continental areas and on the mechanisms of basaltic magma generation. A longstanding discussion concerns the question of whether continental alkali basalts are derived from asthenospheric or lithospheric mantle sources or a mixture of both (e.g. Altherr et al., 1990
; Wilson & Downes, 1991; Stein & Hofmann, 1992; Stein et al., 1997; Shaw et al., 2003). Another discussion concerns the mechanism of melting, either a passive, extension-related scenario (e.g. Chapin, 1979; Bohannon et al., 1989; Stein & Hofmann, 1992; Bradshaw et al., 1993; Shaw et al., 2003), or an active scenario, which involves a thermal anomaly, usually a mantle plume (e.g. Gass, 1970; White & McKenzie, 1989; Camp & Roobol, 1992; Wilson & Patterson, 2001). A third possibility, which involves melting of the lithosphere by heat conduction from a sub-lithospheric source, was discussed in the case of small-volume continental flood basalts (Turner et al., 1996; following Gallagher & Hawkesworth, 1992), but is also relevant to the generation of alkali basalts, as discussed below.
A third question, which is closely related to the above two, concerns the involvement of fluids in the generation of alkali basalts. Hydrous fluids are known to be important in the subduction environment, where they control the degree of melting (e.g. Stolper & Newman, 1994). There is growing evidence that hydrous minerals are also a common constituent in the source rocks of continental alkaline lavas (e.g. Francis & Ludden, 1995), but the role of ‘wet melting’ in the petrogenesis of alkali basalts has still not been well established. For instance, the origin of both East African and western Arabian basalts is mainly discussed in terms of melting a dry peridotitic source (e.g. White & McKenzie, 1989; Camp & Roobol, 1992). Stein et al. (1997) were the first to consider the possibility that the presence of hydrous minerals (i.e. amphibole) in the lithospheric mantle, metasomatic remnants of Late Proterozoic subduction, may actually be the main control on partial melting beneath Arabia. Similarly, Späth et al. (2001) considered dehydration of a metasomatized lithosphere as the cause for melting under the Chyulu Hills, southern Kenya.
In this study we report the temporal and spatial variations in bulk-rock chemical and Nd–Sr isotopic composition of Plio-Pleistocene alkali basalts from the northwestern part of Harrat ash Shaam, one of the largest volcanic provinces in the Arabian plate (Fig. 1). We show that most of the variability in the trace element composition of these basalts cannot be accommodated by simple models of variable degree melting of a homogeneous mantle source. Instead, we attribute this variability to mixing of melts derived from partial melting of pyroxenite veins and their hydrated host peridotites. We also suggest that the veins were produced during earlier magmatic processes within the lithosphere. Finally, we discuss the implication of the model for the melting regime beneath northern Arabia.
|
Geological background
This study focuses on basalts that erupted during the Plio-Pleistocene in the Golan and the Galilee areas of northern Israel, which are located at the northwestern part of Harrat ash Shaam, the largest of about 20 Cenozoic volcanic fields in the western Arabian Peninsula (Fig. 1). The volcanism in Arabia is Oligocene to Holocene in age (31 Ma to recent; Menzies et al., 1997
; although ages as old as Middle Eocene have also been reported, e.g. Pallister, 1987), and is associated with rifting along the Red Sea. Volcanism in the Harrat ash Shaam commenced at about 25 Ma (Barberi et al., 1980; Ilani et al., 2001) and continued to recent times (Dubertret, 1955). At its northwestern edge, the Ash Shaam volcanic field is bisected by the Dead Sea Transform fault (Fig. 1), which separates the volcanism of the Galilee to the west from that of the Golan to the east. Volcanism in this area occurred in two main periods: Middle Miocene (17–9 Ma) and Plio-Pleistocene (5·5 Ma to recent, Mor, 1993; Shaliv, 1991). The center of the Middle Miocene volcanism is located in the Yizre'el Valley and southeastern Galilee (Fig. 1, Shaliv, 1991; Weinstein, 2000). The Plio-Pleistocene volcanism includes basaltic flows and cinder cones in the Golan and the eastern Galilee as well as basalts along the Dead Sea Transform. The latter are mostly known from subsurface drilling and are intercalated with lacustrine sediments (Mittlefehldt & Slager, 1986). This study focuses on the Plio-Pleistocene basalts from the Golan and Galilee.
Overall in the study area, there is a trend of decreasing age from south to north (e.g. Mor, 1993; Heimann et al., 1996; Fig. 1). West of the Dead Sea Transform, the focus of the Middle Miocene volcanism was in the Yizre'el Valley, locally reaching a thickness of more than 800 m. Early Pliocene activity was concentrated in eastern Lower Galilee (west of the Sea of Galilee), and Late Pliocene lavas erupted further north, in the Upper Galilee (Fig. 1). The thickness of the Galilee Pliocene volcanic sections usually does not exceed 100 m. East of the Dead Sea Transform, the southern Golan plateau is covered by a 20–200 m section of Early Pliocene basalts (Fig. 1). Further north, in the Yahudiya plateau of the central western Golan, there is an 
150 m thick Late Pliocene basaltic section. This section is probably underlain in the subsurface by Early Pliocene basalts, as indicated by a 3·3 Ma basaltic flow in the bottom of one of the local canyons (sample Wy-292, Table 1) and from seismic reflection profiles (Dafny et al., 2003). Pleistocene basalts are found only in the eastern and northern Golan (Fig. 1). In most of the eastern and northern Golan, the volcanic section is relatively thick (up to 700 m in the Bashanit area; Dafny et al., 2003; Fig. 1), but the age of the subsurface volcanic rocks is not known (there is one small exposure of Early Pliocene basalts in the northernmost part of the Golan; Heimann et al., 1996). A similar decreasing age pattern is observed along the Dead Sea Transform. Basalts beneath and around the Sea of Galilee are of Early Pliocene age (Heimann, 1990; Heimann et al., 1996; Hurwitz et al., 2002), whereas Late Pliocene and Early Pleistocene basalts are found only further north, in northern Korazim and in the subsurface and around the Hula Valley (Fig. 1, Heimann, 1990; Mor, 1993).
|
The main active tectonic features in the region are the Red Sea rift and the Dead Sea Transform fault (inset to Fig. 1). The Red Sea trends N25W. Rifting started at 26 Ma, accompanied (or pre-dated) by uplift and volcanism on the Arabian side (e.g. Omar & Steckler, 1995
; Menzies et al., 1997). Seafloor spreading commenced at about 5 Ma along the southern part of the axial trough. The Dead Sea Transform fault separates the Arabian plate, which includes the Golan, from the Sinai sub-plate including the Yizre'el Valley and the Galilee. The fault trends north–south and the total amount of left-lateral slip along it is about 105 km (Quennell, 1958; Freund et al., 1970; Freund, 1978; Garfunkel et al., 1981). It is commonly accepted that this displacement has accumulated during the past 20 Myr (Bartov et al., 1980) and that an offset of 60 km had already occurred by the end of the Miocene (5 Ma), before the Early Pliocene volcanism started (Freund, 1978; Garfunkel et al., 1981). A series of pull-apart basins (e.g. the Dead Sea, the Sea of Galilee and the Hula Valley, Fig. 1) was developed along the Dead Sea Transform as a result of sideways stepping of the major fault planes (see Garfunkel, 1981; Garfunkel & Ben Avraham, 2001). Pliocene and Plio-Pleistocene basalts are buried in the Sea of Galilee and the Hula Valley basins, respectively, under hundreds to thousands of meters of fluvio-lacustrine sediments. |
SAMPLING AND ANALYTICAL METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONSAMPLING AND ANALYTICAL METHODSPETROGRAPHY AND CHEMISTRY OF...DISCUSSIONCONCLUSIONSREFERENCES |
|---|
Ninety-seven samples were collected from volcanic outcrops in the Golan and the Galilee, representing the entire Plio-Pleistocene sequence on both sides of the Dead Sea Transform. About 100 g of fresh rock pieces were crushed and ground to fine powder (<200 mesh) for chemical and isotopic analyses, and thin sections were prepared for petrographic studies.
Concentrations of major elements were determined by X-ray fluorescence (XRF) using a wavelength Philips PW-1404 system at Guttenberg University, Mainz, using glass discs. Calibration curves were established using international standards [recommended values from Govindaraju (1989) and Jochum et al. (1990)]. Fe was determined as Fe2O3(total); Fe2+ was assumed to be 90% of the total iron. Accuracy was within 1–2% for most elements (accuracies of Na and Mn are 4% and 5%, respectively). Reproducibility was measured by running duplicates within run and in separate runs and was found to be better than 2% for most elements and 4% for Na. In 12 samples, concentrations of major elements were determined by inductively coupled plasma atomic emission spectrometry (ICP-AES; Weinstein et al., 1994), and three of these samples were remeasured by XRF. The two methods agree within a few per cent for most of the elements, except for P concentrations, which were consistently higher (by 6–9%) in the ICP-AES measurements. For these three samples, we use the XRF data. Loss on ignition (LOI) was determined by heating 1 g of a sample to about 1100°C, and was corrected for oxidation of ferrous iron during heating. Calculated LOI values are considered as maximum values, as many basalts are relatively oxidized.
Concentrations of trace elements (V, Cr, Co, Ni, Cu, Zn, Ga, Rb, Sr, Y, Zr, Nb, Ba) were also determined by XRF using powder pellets. Accuracy is usually better than 10%relative. Ba concentrations measured on one of the standards (BHVO-1) were very different from the recommended values (86 and 139 ppm, respectively), whereas six other standards yielded accuracies better than 10%. This suggests that Ba data should be treated cautiously. Precision of trace elements was also better than 10% relative, except for Ba, for which precision was 16%.
Concentrations of Sc, Cs, Hf, Ta, Pb, Th, U and 14 rare earth elements (REE) in 40 samples were determined by inductively coupled plasma mass spectrometry (ICP-MS) using a VG Plasma Quad II+ system (University of Göttingen) with a peak jump mode [following the procedure described by Gao et al. (1999)]. Blanks are negligible for most of the elements, except for Gd and Dy (in some of the runs), which are, therefore, not used below for quantitative modeling. Accuracies are better than 10% for most REE and U (for Tb and Tm accuracy is 20%), whereas those for Ta and Hf are 15% (low stability of HFSE in the solution) and those for Pb and Th are 25% and 30%, respectively (a strong memory effect for Th). Therefore, both the Pb and Th data should be interpreted cautiously. Sc was consistently higher (by about 20%) than the recommended values for international standards as a result of interference with HCO2. Cs was lower by 20%. Therefore, both Sc and Cs measured by ICP-MS are not reported in this study.
The concentrations of Sc, Cs, Hf, Ta, Th, U and 11 REE in 15 samples were determined by instrumental neutron activation analysis (INAA) in the University of Karlsruhe, following the procedure described by Kramar & Puchelt (1982). Accuracies are between 0 and 15%. Comparative analyses of six samples (three of them from the current study) show that there is a good agreement between INAA and ICP-MS measurements for La, Ce, Nd, Sm, Eu, Gd, Yb, Lu, Th and U (<20% difference), whereas Tb, Ho, Ta and Hf are in significant disagreement [for the last two this is probably because of the problem of high field strength element (HFSE) stabilization in solutions during ICP-MS analyses]. For the three comparative samples, the reported data (Table 1) are those from the INAA (except for the Pb and Pr data, which are from the ICP-MS measurements).
Rock powder (100 mg) was leached for about 40 min in warm 0·5N HCl to remove secondary Sr. Leaching experiments on some of the samples showed that the main difference in the Sr isotope ratios of the treated basalts is between no leaching and 0·5N HCl, whereas the effect of changing to stronger acid (e.g. 6N HCl, Table 1) was within the analytical error of the laboratory (an average difference of 2·3 x 10–5). After washing and drying, the samples were digested by 3 ml of 24N HF + 0·5 ml of 14N HNO3 for at least 8 h at 120°C, and evaporated at 140°C (some of the samples were heated for an additional 2 h at 180°C to break the fluorides). The evaporates were then dissolved in diluted HNO3 (120°C), checked for complete dissolution, evaporated and redissolved by 6N HCl (70–80°C). After evaporation to dryness, the residue was taken up by 3·2 ml of 2·6N HCl and centrifuged in Teflon tubes. Solutions (2 ml) were loaded on columns consisting of 15 ml cation exchange (preconditioned by 20 ml of 2·6N HCl). The columns were then washed by additional 42 ml of 2·6N HCl (eluting most of the elements except for Sr and the REE), and then most of the Sr was eluted from the column by another 6 ml of 2·6N HCl. After 10 ml wash (14 ml in the last few samples) by 6N HCl, the REE were eluted by an additional 6 ml of 6N HCl. After drying, both Sr and REE aliquots were treated with HNO3 to destroy residual resin and dried again. REE aliquots were taken by 0·5 ml of 0·18N HCl and loaded on columns consisting of 0·5 g orthophosphoric acid (1·5N) mixed with 5 g Teflon beads (after conditioning the columns with 0·18N HCl). The columns were washed with an additional 7·5 ml of 0·18N HCl, and the Nd was eluted with an additional 6 ml and dried. All chemical procedures were carried out in a clean environment, and all the acids used were double distilled. Blanks were negligible. Isotopic analyses were carried out on a Finnigan MAT 262 mass spectrometer at the Geochemistry Institute, University of Göttingen, equipped with nine Faraday cups and using a multicollector static mode for all elements. For Sr isotope analysis, about 4 ng were loaded with phosphoric acid onto Ta single filaments (except for the standard NBS-987, which was loaded onto Re filaments). Sr analyses were corrected for mass fractionation by normalizing to 88Sr/86Sr = 8·375209. The NBS-987 standard was routinely measured, giving a mean 87Sr/86Sr value of 0·710229 and a reproducibility (2
) of 18 x 10–6 (n = 13). The mean of all NBS-987 measured in the laboratory during the period September 1995 to July 1996 was 0·710244 (2 = 32 x 10–6). For Nd isotope analysis, about 1 ng was loaded onto Re double filaments. Isotope ratios were normalized to 146Nd/144Nd = 0·7219. The measured mean 143Nd/144Nd for the La Jolla Nd standard was 0·511848 with a reproducibility of 16 x 10–6 (2, n = 11), whereas the mean for the period September 1995 to July 1996 was 0·511851 (2 = 18 x 10–6). All isotopic ratios measured (Sr, Nd) are considered as initial ratios because of the relatively young ages (0–5 Ma) of the basalts studied and their relatively low Rb/Sr ratios.
K–Ar ages were determined by A. Heimann (Geological Survey of Israel). Details of the analytical procedures have been given by Heimann et al. (1996).
|
PETROGRAPHY AND CHEMISTRY OF THE GOLAN–GALILEE BASALTS |
|---|
|
TOPABSTRACTINTRODUCTIONSAMPLING AND ANALYTICAL METHODSPETROGRAPHY AND CHEMISTRY OF...DISCUSSIONCONCLUSIONSREFERENCES |
|---|
General description and petrography of the Golan–Galilee basalts
The volcanic rocks of the Galilee and the Golan are mainly composed of basaltic flows. Scoriaceous rocks are mostly restricted to the young cinder cones of the northern Golan (Pleistocene) and of the Upper Galilee (Late Pliocene), although a few scoria exposures are also found amongst the Early Pliocene rocks of the southern Golan. Tuff deposits are very rare, and are not included in this study.
We subdivide the Plio-Pleistocene Golan–Galilee volcanism into six units according to age and location (Fig. 1). These are, in order: Early Pliocene basalts from the southeastern (Lower) Galilee, Late Pliocene basalts from the northeastern (Upper) Galilee, Early Pliocene basalts from the southern Golan, Late Pliocene basalts from the western Golan, Early Pleistocene basalts from the eastern and northern Golan, and Late Pleistocene basalts and scoria also from the eastern and northern Golan. In most figures, we refer to both Golan Pleistocene units as one group.
Most Golan–Galilee samples typically show porphyritic to glomeroporphyritic textures, with the phenocrysts usually dominated by olivine (Fo78 to Fo85, 5–20% of the rock; Table 1), which is often partly or completely replaced by iddingsite. Clinopyroxene phenocrysts occur in smaller proportions (<5%), and they are rare in the Early and Late Pliocene basalts. Early Pleistocene basalts from the Golan are the only group that shows similar proportions of olivine and clinopyroxene phenocrysts (2–3%). Plagioclase phenocrysts (labradorite, usually <1%) are observed in about 20% of the samples, mainly those of Early Pliocene age. The groundmass of all Early Pliocene and of Golan Late Pliocene basalts is mostly medium grained, whereas that of Galilee Late Pliocene and of Golan Pleistocene basalts is usually fine grained, with a pilotaxitic texture dominated by plagioclase laths and prismatic crystals of clinopyroxene. Intergranular nepheline is also common. The groundmass of the scoriaceous rocks usually shows a hyalopilitic texture, although the glass is usually replaced. Calcite amygdules are common in the Pliocene basalts.
Mantle xenoliths are common in the Golan and Galilee basalts, though usually restricted to the basanitic units. Spinel peridotites are found both in the Golan Late Pleistocene and in the Upper Galilee Late Pliocene basanites. Pyroxenites are found just in one Pleistocene maar from the northern Golan (Birket Ram, Fig. 1). The xenolith assemblage of this maar is especially large and variable, containing peridotites, dry and hydrous pyroxenites, mafic granulites and megacrysts of kaersutite and microcline (Mittlefehldt, 1984; Stein, 1987). Mafic granulites (as well as anorthoclase megacrysts) were also found in an Early Pliocene Lower Galilee site (Gazit, 2005).
Major and trace element geochemistry of the basalts
The bulk-rock major and trace element compositions of representative Golan–Galilee basalts are given in Table 1. The complete dataset is available for downloading from the Journal of Petrology website at http://www.petrology.oupjournals.org. The samples mainly plot in the basanite, hawaiite and alkali basalt fields of the total alkalis vs silica (TAS) diagram (Fig. 2), and show a trend of increasing total alkali concentration (Na2O + K2O) with decreasing silica. SiO2 concentrations range between 43% and 51% and total alkalis between 3·5% and 7·5%. The spectrum of compositions (alkali basalt–hawaiite–basanite) is similar to that of the Neogene–Quaternary basalts from northeastern Jordan (another part of the Harrat ash Shaam volcanic field; Shaw et al., 2003), although the Golan–Galilee basalts are slightly higher in SiO2 (Fig. 2). Miocene basalts from the Yizre'el Valley and the southeastern Galilee (Fig. 1) are somewhat different. These extend to lower SiO2 concentrations than the Plio-Pleistocene basalts (including nephelinites with 38–40% SiO2, not shown in Fig. 2; Weinstein, 2000) but show no systematic increase of alkalis with decreasing SiO2 (Fig. 2).
|
Most Golan–Galilee samples have MgO >7% (up to 11·6%) and Mg numbers from 0·55 to 0·66 (Fig. 3a). The hawaiites typically have lower MgO concentrations, implying a relatively higher degree of differentiation. Alumina (12–15% Al2O3), and to a lesser extent Na2O (2·6–5·1%), correlate negatively with MgO (Fig. 3b and c). All other major elements (e.g. CaO, Fig. 3d) do not show any correlation with MgO (either general or within-suite). Nickel (3–326 ppm), Cr (5–423 ppm) and, to a lesser extent, Co (22–68 ppm) correlate positively with MgO concentration (Fig. 3g and h), reflecting olivine and clinopyroxene fractionation. On the other hand, incompatible elements, as well as TiO2 and P2O5, do not correlate with MgO (Figs 3e, f and 4a–d), implying that the variability in their concentrations is not controlled by fractional crystallization. However, they correlate positively with each other and negatively with SiO2 (Fig. 4e–h).
|
|
Primitive mantle normalized incompatible trace element patterns (Fig. 5a–c) are similar to those of ocean island basalts (OIB, Fig. 5c), with strong enrichment in most highly incompatible elements. This enrichment is stronger in the basanites than in the alkali basalts. The REE show a continuous decrease of normalized concentrations from La to Yb (Fig. 6; except for a slight positive Eu anomaly in some of the samples), with (La/Yb)CN ratios of 5–20 in the alkali basalts and 18–40 in the basanites (CN indicates chondrite normalized). Concentrations of heavy REE (HREE) in all Golan–Galilee basalts are lower than in typical mid-ocean ridge basalt (MORB) and OIB. Slopes of middle REE (MREE) to HREE are similar to those of OIB in the alkali basalts (Tb/Yb(CN) ratios of 1·6–3·4), but higher in the basanites (2·8–4·2). All Golan–Galilee basalts show prominent anomalies in primitive mantle normalized trace element diagrams (Fig. 5). These include positive anomalies in Ba, Nb, Sr, P and Zr, and negative anomalies in Rb. The positive P anomalies and negative Rb anomalies are more prominent in the basanites than in the alkali basalts (Fig. 5a and b), and the Ba and Nb anomalies do not always exist. In addition, the basanites show negative anomalies of K and Pb (Fig. 5a and b). In the alkali basalts and in few basanites, there are also negative Th and U anomalies. Basalts from northeastern Jordan show similar Nb, Sr, P, Rb and K anomalies, but they do not show positive Ba and P anomalies (Shaw et al., 2003
). In general, the ratios of the REE to their neighboring HFSE in Fig. 5 (e.g. La/Nb, Sm/Zr and Eu/Ti) increase from the alkali basalts to the basanites (Fig. 7a and b). Similarly, P/Zr ratios also increase from alkali basalts to the basanites (Fig. 7c). Basalts from northeastern Jordan show similar trends of increasing La/Nb from alkali basalts to basanites (Fig. 7a; Shaw et al., 2003), although their overall La (and Nb) concentrations are lower than in the Golan–Galilee. On the other hand, in volcanic rocks from the Dead Sea area (Shaw, 2003) La/Nb ratios decrease from alkali basalts to basanites (Fig. 7a).
|
) and Late (
). Dashed lines are for alkali basalts and continuous lines for basanites. Primitive Mantle values are from McDonough & Sun [1995
|
|
A central observation in this study is the patterns defined by the Golan–Galilee basalts in plots of some incompatible element ratios vs concentration (e.g. La/Nb vs La, Nb/Zr vs Nb and La/Sm vs La, Fig. 7a–e). In these, the Pliocene alkali basalts define steeper trends than the Pleistocene Golan basanites. The general trends defined by the Golan samples in these diagrams are convex-upward, which we interpret below as mixing curves.
Spatial and temporal variability
In both the Golan and Galilee, basalt chemistry appears to change with both time and geographic location, from alkali basalts in the south to mainly basanites in the north. In the Golan, Early Pliocene lavas from the south are mainly alkali basalts, with a few hawaiites (Fig. 2). In the western Golan, the Late Pliocene basalts are also alkali basalts, but with lower silica contents (almost all samples contain normative nepheline compared with only 60% of the Early Pliocene samples). Pleistocene lavas from the northern and eastern Golan are mainly basanites and hawaiites (Fig. 2). In the Galilee, Early Pliocene lavas from the south (Lower Galilee) are mainly alkali basalts and hawaiites, whereas Late Pliocene lavas from the north (Upper Galilee) are all basanitic (Fig. 2).
The spatial–temporal geochemical variability is also reflected in the concentrations of incompatible elements and in H/I ratios (where H indicates a hygromagmatophile, highly incompatible element, and I indicates a moderate incompatible element, e.g. Nb/Zr and La/Sm). Basanites (Golan Pleistocene and Galilee Late Pliocene) have higher incompatible element concentrations and ratios than the alkali basalts (Early Pliocene and most of the Golan Late Pliocene samples; see Figs 5a–c and 7d, e). Also, most of the basanites contain >2·5% TiO2 and >0·9% P2O5 (averages 2·8% and 1·2%, respectively) whereas the alkali basalts have lower concentrations (averages of 2·3% and 0·5%, respectively; Table 1).
Ratios of light REE (LREE) and MREE to HFSE of similar compatibility (e.g. La/Nb, Sm/Zr and Eu/Ti) also increase in the Golan, from the southern Early Pliocene alkali basalts to the northern and eastern Pleistocene basanites (Table 2, Fig. 7a and b). The behavior of P is similar to that of the REE; P/Zr increases from 7–15 in the southern Golan to 15–30 in the northern Golan (Fig. 7c). There is no systematic variability (between suites or within an individual suite) in MgO and compatible element concentrations, except for the Pleistocene basanites from the northern Golan, for which Weinstein et al. (1994) observed that Late Pleistocene basalts have higher MgO and Ni concentrations than Early Pleistocene basalts.
|
An interesting spatial–temporal pattern is shown in the trace element and P concentrations of basalts from the central Golan (Fig. 8). More than half of the samples were taken from flows along a 150 m high cliff section in the Zavitan river in the Yahudiya plateau (western Golan, Fig. 1), spanning a period from the Early to Late Pliocene (3·3–1·8 Ma, Table 1 and Fig. 8). The rest of the samples were collected along a line drawn from the Zavitan river to the Bashanit area in the eastern Golan (A–A' in Fig. 1). The samples along this line are all Pleistocene in age (1·3–0·4 Ma, Table 1 and Fig. 8), but there is no clear trend of decreasing age toward the NE (Fig. 8), even though the easternmost sample is also the youngest. In general, the concentration of P and incompatible elements in the Pleistocene samples is higher than in the Pliocene ones. However, whereas during the Late Pliocene there is a clear temporal trend of increasing concentrations with time, in the Pleistocene samples the main change is with location (towards the NE) and not with time (e.g. concentrations of all three elements increase from the 0·7 Ma sample to the 1·3 Ma one), even though the youngest sample also has the highest concentrations. We note that the location of the vents that fed the Late Pliocene basalts of the Yahudiya plateau is unknown. They could either be further to the east (now covered by Pleistocene basalts, Fig. 1) or unexposed local fissures in the western Golan. Thus, the control on the variation in basalt composition in the central Golan is both temporal and spatial.
|
Unlike the Golan–Galilee, Neogene–Quaternary basalts from northeastern Jordan do not show any temporal or spatial patterns (Shaw et al., 2003
).
Sr and Nd isotope geochemistry
The Sr–Nd isotopic compositions of the Golan–Galilee alkali basalts and basanites are given in Table 1 and plotted in Fig. 9a and b. Nd isotope ratios are remarkably uniform in the range Nd = 4·0–5·2, whereas 87Sr/86Sr varies between 0·7031 and 0·7034, a composition close to that of PREMA (PREvalent MAntle, Wörner et al., 1986). A similar range of Sr–Nd isotope compositions has been reported for Miocene basalts from the Yizre'el Valley, southeastern Galilee and the southern Golan (Nd 4·1–5·4 and 87Sr/86Sr 0·7031–0·7034, Weinstein, 2000; Stein & Hofmann, 1992). The isotopic compositions of the Golan–Galilee basalts fall within the field of basalts from Saudi Arabia (Altherr et al., 1990, Fig. 9a), although the latter show larger ranges in both 87Sr/86Sr and Nd (0·7028–0·7035 and 3·6–7·6, respectively), extending towards the Red Sea MORB field (e.g. Volker et al., 1993). Basalts from northeastern Jordan (Shaw et al., 2003) show a significantly wider range of Sr isotope compositions (0·70305–0·70415), whereas their Nd isotope ratios (Nd = 4·1–6·4) are relatively similar to those of the Golan–Galilee basalts. Interestingly, Sr and Nd isotopes are strongly negatively correlated in the Jordanian basalts (Shaw et al., 2003), probably as a result of contamination with an upper crust component. This trend is not observed in the Golan–Galilee and is just slightly shown by the Saudi Arabian basalts (Altherr et al., 1990).
|
, Yizre'el Valley and southeastern Galilee; •, one southern Golan sample) and Mezozoic samples (The Sr isotope ratios of the Golan–Galilee basalts are negatively correlated with incompatible element concentrations (e.g. Nb and Sr, Fig. 10a and b) and positively correlated with Rb/Sr ratios (Fig. 10c) and 1/Sr (Fig. 10d).
|
, Y. Weinstein, unpublished data, 1998). Also shown in (c) are the compositions of calculated APR and PYV sources (
; Rb/Sr ratios are from Tables 4 and 5, and the 87Sr/86Sr values are those of Wy-337 and Wy-52, respectively; see text). |
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONSAMPLING AND ANALYTICAL METHODSPETROGRAPHY AND CHEMISTRY OF...DISCUSSIONCONCLUSIONSREFERENCES |
|---|
In the following discussion we focus on the nature and the causes of the compositional variability observed in the Golan–Galilee basalts. Stein & Goldstein (1996)
and Stein et al. (1997) have proposed that all Phanerozoic basalts from Arabia have a lithospheric rather than an asthenospheric mantle source. Following this approach, we show below that the compositional variability of the Golan–Galilee basalts reflects the heterogeneous nature of the underlying lithospheric mantle.
Alteration and crustal contamination processes
Alteration probably did not have a significant effect on the chemistry of the Golan–Galilee basalts. Almost all the samples are fresh, showing no petrographic evidence for alteration except for the presence of iddingsite that partially or completely replaces olivine. However, the scatter shown by the Galilee basanites in plots of Rb (and to a lesser extent K) versus SiO2 (Fig. 4h) or versus other incompatible elements (not shown) implies that this group was subjected to some secondary process that affected the concentration of the more mobile elements. As the Golan basalts show good correlations in the above plots (Fig. 4h) and do not show any other signs of alteration, we chose to focus on this group in the petrogenetic discussion below.
The degree of crustal contamination of the Golan–Galilee basalts, as well as of other Phanerozoic basalts from Israel (Stein & Hofmann, 1992), is probably fairly small on the basis of their Sr and Nd isotope compositions (Fig. 9b). Upper crustal contamination should result in a significantly larger range of variability and higher Sr isotope ratios. A contribution from the lower crust cannot be absolutely ruled out, but should probably result in some variability in Nd isotope composition (Fig. 9b), unlike the uniform values observed in the Golan–Galilee basalts (Fig. 8a). Similarly, Altherr et al. (1990) showed that most Saudi Arabian basalts carry near pristine mantle isotopic signatures and ruled out crustal contamination. We note, however, that some Arabian basalts, e.g. from Yemen and especially those from nearby northeastern Jordan, do show a crustal Nd–Sr isotopic signature (Baker et al., 1997; Shaw et al., 2003; Fig. 9a).
Differentiation of the primary magmas
The Golan–Galilee basalts are clearly not primary in composition but are, nevertheless, relatively primitive. Their Mg numbers are between 0·66 and 0·37 (Table 1), MgO contents are usually <12 wt %, and Ni and Cr are <350 ppm and <500 ppm, respectively. This indicates that all the basalts have experienced a certain degree of crystal fractionation. However, most of the incompatible elements show no correlation with MgO concentration (e.g. Fig. 4a–d). This indicates that the wide concentration range of incompatible elements (e.g. a factor of five in Nb and nine in La concentrations) in the Golan–Galilee basalts was not generated by differentiation processes. Rather, the variability in incompatible element concentrations could have originated from mantle source heterogeneity or from variable degree partial melting. The negative correlation of incompatible elements (e.g. Nb, La, Ba, Rb) with SiO2 (Fig. 4e–h) also supports this conclusion.
Because all of the Golan–Galilee magmas underwent differentiation, their original (primary) incompatible element composition must have been modified to some extent. Based on the good correlation of Ni and MgO (Fig. 3g), we assume that most primitive magmas followed more or less similar differentiation paths. Therefore, to better discuss melting regimes and mantle source compositions, we have corrected the Golan–Galilee basalt compositions to a reference value of 10% MgO by adding various minerals to the bulk-rock composition, following the approach of Klein & Langmuir (1987). This correction procedure should restore the magma composition close enough to that of the primary magmas (e.g. Takahashi & Kushiro, 1983; Falloon & Green, 1988). About 10% of the samples have MgO >10% (Table 1, Fig. 3) and required no correction.
The dominance of olivine and clinopyroxene in the phenocryst population and the good correlation of Ni, Cr and Mg numbers with MgO concentrations (Fig. 3a, g and h) suggest that these minerals were the main fractionating phases. Plagioclase fractionation may be ruled out based on the negative correlation between Al2O3 and MgO (Fig. 3b), the high Sr concentrations, the rarity of Eu negative anomalies (Fig. 6) and the minor presence of plagioclase phenocrysts in the basalts. Amphibole is also ruled out because of the absence of amphibole phenocrysts. As indicated by the fractionation simulations shown in Fig. 3a, b, d and e, clinopyroxene cannot be the main fractionating phase. On the other hand, olivine alone cannot account for the trend observed in the CaO vs MgO diagram (Fig. 3d) and probably also not for that in the Al2O3 vs MgO diagram (Fig. 3b, too steep). Altogether, the data trends are best explained by a fractionating mineral assemblage with an olivine:clinopyroxene ratio of about 60:40, and we use these proportions to correct to 10 wt % MgO. In a few samples, where large deviations from the 60:40 curve were noticed, the proportions were changed accordingly. As the distribution coefficients of both clinopyroxene and olivine are <1 for all incompatible elements, the effect of the correction procedure on the incompatible elements is mainly one of dilution, by up to 17–18% for the highly incompatible elements (e.g. La) and less for others.
The 60:40 ol:cpx ratio assumed in the fractionating assemblage is very different from the usually observed strong dominance of olivine in the phenocryst population of basalts (Table 1). The phenocryst assemblage probably reflects a later, shallower, crystallization stage in which olivine is expected to dominate (e.g. Takahashi & Kushiro, 1983; Sack et al., 1987; Weinstein et al., 1994).
Melting of a heterogeneous lithospheric mantle
On the basis of the above discussion, we consider that the compositional spectrum observed in the Golan–Galilee basalts, especially that of incompatible elements, was not produced by crystal fractionation processes, but reflects mantle source heterogeneity and/or partial melting processes.
Considering the young age of the Golan–Galilee basalts and their low Rb/Sr ratios, the variability of Sr isotope ratios (Fig. 9a) precludes the derivation of the parent magmas from a homogeneous mantle source. Furthermore, the negative correlation of Sr isotope ratios with incompatible element concentrations (e.g. Nb and Sr, Fig. 10a and b) and the positive correlation with 87Rb/86Sr and with 1/Sr (Fig. 10c and d) suggest that the spectrum of melt compositions was produced by binary mixing. Mixing is further supported by the upward-convex patterns in some of the H/I vs H diagrams (e.g. Fig. 11a–d) that differ from the linear trends expected in a case of variable degrees of melting (see Hofmann & Feigenson, 1983). The linear trends observed in diagrams of MREE/HREE vs LREE/HREE (e.g. Tb/Yb vs La/Yb, Fig. 11f) also differ from those expected in variable degree of partial melting.
|
The incompatible element characteristics of the Golan–Galilee basalts (Figs 7 and 11a–e) could alternatively be considered as a combination of two linear arrays, each of which originates from variable degree melting of a distinct mantle source. However, in this scenario we should expect each group to cluster around the 87Sr/86Sr ratio of its mantle source. The continuous variation of Sr isotope ratios and their correlation with incompatible element concentrations (Fig. 10a and b) suggests that this is not the case. In addition, it is hard to fit the trace element data with a partial melting model involving two distinct mineralogical assemblages. Source composition and mineralogy that fit a given set of element ratios and concentrations cannot produce the trends of other sets of elements. For example, in the diagram of Nb/Zr vs Nb (Fig. 11a), the relatively moderate trend defined by the more enriched samples (mainly the basanites) is similar (parallel) to the slope of a spinel peridotite melting curve (curve 1 in Fig. 11), whereas the steep slope of the less enriched samples (<40 ppm Nb) is similar to that of a pyroxenite curve (curve 5). On the other hand, in the diagram of La/Nb vs La (Fig. 11b), the trend of the enriched samples is similar to that of an amphibole lherzolite (curve 3) or of a phlogopite peridotite melting curve (curve 4), whereas the steeper slope of the alkali basalts could not be produced by any source mineralogy. The opposite is true for La/Sm vs La (Fig. 11c), where the relatively flat slope defined by the enriched samples does not fit any of the melting curves. Moreover, the positive correlation shown by both basanites and alkali basalts in Sr/Nb vs Sr (Fig. 11e) does not match any of the melting simulations. Therefore, we consider that the incompatible element characteristics of the Golan–Galilee basalts are mainly controlled by binary mixing of partial melts from discrete mantle sources, especially in the case of the Golan basalts. Mixing of discrete mantle melts, as opposed to mixing of mantle components (e.g. variable degree of source enrichment), is our preferred model because we anticipate that in the latter the correlation of 87Sr/86Sr with incompatible element concentrations (Fig. 10a and b) would be erased during partial melting processes. Nevertheless, the rarity of reversely zoned phenocrysts in the Golan and Galilee basalts (Weinstein et al., 1994
) implies that mixing occurred early in the ‘life’ of the magma, close to the source area, before the phenocrysts started crystallizing. Mixing as the main control on basalt chemistry was also identified by Shaw et al. (2003) in the neighboring northeastern Jordanian basalts. We therefore conclude that the compositions of the Golan–Galilee basalts reflect mixing of melts derived from two distinct mantle sources. One end-member is the main source for the alkali basalts, and is characterized by relatively low concentrations of incompatible elements, low H/I ratios (e.g. Nb/Zr) and low REE/HFSE ratios (e.g. La/Nb). The second end-member is enriched in incompatible elements and in the above ratios, and is the main source for the basanites. The compositional range observed in the basanites (e.g. Fig. 11a–c and f) suggests that the enriched source is heterogeneous; namely, it includes a spectrum of enriched compositions.
It is interesting to note that alkali basalts and basanites from the Dead Sea area of Jordan (Shaw, 2003) show a very clear upward-convex trend on a diagram of Nb/Zr vs Nb (Fig. 7d). Miocene basalts from the Yizre'el Valley and the southeastern Lower Galilee (Weinstein, 2000) also show an upward-convex pattern on the same diagram (Fig. 7d). On the other hand, basalts from northeastern Jordan, which were also interpreted in terms of binary mixing by Shaw et al. (2003), do show a positive correlation on this diagram but not an upward-convex pattern (Fig. 7d). In all three datasets, an upward-convex pattern is not identified on the La/Nb vs La diagram (Fig. 7a), probably because of the relatively small range of La concentrations in these datasets. However, whereas the Israeli Miocene and the northeastern Jordanian basalts do show a positive correlation (similar to the Golan–Galilee basalts), the Dead Sea samples show a clear negative correlation of La/Nb ratios with La concentrations (Fig. 7a). If the Nb/Zr vs Nb trends (Fig. 7d) are taken as evidence for binary mixing in petrogenesis of the Dead Sea basalts, then the enriched end-member in this case must have had a lower La/Nb ratio than the alkali basaltic end-member.
Unlike Sr isotopes, Nd isotope compositions are uniform in all the Golan–Galilee basalts (Nd = 4·6 ± 0·6), suggesting a genetic relationship between the two mantle sources. Based on this observation, we suggest that the two mantle sources may have been derived from a common reservoir, which separated into the above two sources. The following discussion focuses on the characterization of the two sources and on their evolution.
Source of the alkali basalts
The melt derived from the alkali basaltic source can be closely characterized by sample Wy-337 from the southern Golan, as it has the lowest concentrations of highly incompatible elements. On a primitive mantle normalized trace element diagram (Fig. 5c), Wy-337 displays prominent peaks at Ba, Nb, Sr, Ti and Zr, and negative anomalies at Rb and U. Because there is no major mantle mineral that can significantly fractionate Rb from Ba, we attribute the low Rb/Ba (Fig. 5a, Table 2) to preferential removal of Rb by hydrous fluids during the petrogenesis of the basalts or the evolution of their mantle sources. Equilibration of the source with fluids may also cause the negative U anomaly (high D(fluid–rock), Keppler, 1996) and the positive Nb anomaly, which as a HFSE has a tendency to be strongly retained in the rock (Keppler, 1996; see Table 3) compared with other elements (including the REE).
|
The relatively high Sr/Ba and the higher than PM Nb/Ba of Wy-337 (0·71 and 1·10, respectively, PM-normalized ratios) suggest that Ba may have been buffered by a residual hydrous phase. This phase was probably an amphibole rather than a phlogopite, as buffering by phlogopite would cause extremely low Ba concentrations in the partial melts (DBa(phl-melt) >1; Adam et al., 1993
; LaTourrette et al., 1995), and the Nb/Ba ratio would, therefore, be much higher than that of the primitive mantle. The presence of phlogopite is also not favored by the relatively low K/Nb ratio (<1, PM ratios) in the alkali basalts (Fig. 5a–c).
It is, therefore, suggested that the source for the alkali basaltic end-member melt was an amphibole-bearing peridotite (designated as APR) that underwent partial dehydration, which caused the preferred removal of Rb and U. The presence of amphibole limits the depth of the source to <90 km, as 30 kbar is the maximum pressure for amphibole stability (Green, 1973; Brey et al., 1983; Wallace & Green, 1988). The low Sm/Nd (0·73 PM, Table 2) and the relatively high Tb/Yb (1·7 PM, Table 2 and Fig. 5c) of Wy-337 suggest buffering by garnet, which implies that the APR source was located deeper than 60 km (20 kbar, Mysen & Boettcher, 1975). Combining the two constraints, the APR source appears to have been located at 60–90 km depth.
A source depth shallower than 90 km implies a lithospheric rather than a deeper mantle origin, as the low heat flow beneath Israel (40 mW/m2, Eckstein, 1979) predicts a lithosphere thickness in excess of 100 km. Unlike McGuire (1988), Nasir (1992) and Shaw et al. (2003), we consider that the elevated temperatures calculated for peridotitic xenoliths (e.g. 1015°C at 15 kbar in western Arabia, McGuire, 1988; McGuire & Bohannon, 1989; Nasir, 1992; Stein et al., 1993) hosted in Arabian Cenozoic basalts represent local heating by magmas (although not the host magmas) and not regional elevated geotherms (see Stein et al., 1993).
A lithospheric mantle source for the Israeli and Arabian basalts was previously suggested by Stein & Goldstein (1996), who showed that all juvenile Late Proterozoic magmas from the Arabian–Nubian Shield (ANS) and Phanerozoic alkali basalts from the Arabian plate lie on the same trends in Nd, Sr and Pb isotope evolution diagrams. These trends differ significantly from the growth curves of asthenospheric MORB-source mantle. Moreover, isotope ratios of Israeli Mezozoic basalts (Triassic to Cretaceous) are very similar to those of the Plio-Pleistocene Golan–Galilee ones (corrected initial 87Sr/86Sr ratios of 0·7029–0·7031 and Nd of 4·5–5·4, Fig. 9a), probably implying a closer genetic relationship between these two groups than between the coeval Golan–Galilee and Arabian basalts. Stein & Goldstein (1996) concluded that the sources of these magmas have remained associated with the ANS since its formation during the Late Proterozoic Pan-African orogeny and that these sources are, therefore, located in the lithospheric mantle.
Following Stein et al. (1997), it is suggested that the incompatible element characteristics of the APR source were acquired during Late Proterozoic Pan-African events as a consequence of fluxing of the newly formed mantle wedge by slab-derived fluids. In the slab, the fluids equilibrated with eclogite and peridotite. In the wedge, a hydration front propagated upwards, leaving behind amphibole peridotite. Trace elements carried by the fluids partitioned between the fluid and the hydrated peridotite. This chromatographic effect (Navon & Stolper, 1987) led to the stripping of Nb and other trace elements with high D(amphibole peridotite–fluid) from the migrating fluid and their preferential retention at deeper wedge levels. Lanthanum and other trace elements with lower D(peridotite–fluid) remained in the fluid until shallower wedge levels. As a result, Nb-enriched zones were produced in deeper parts of the wedge, whereas Nb-depleted zones were generated at shallower levels or away from the fluid migration channels. The low La/Nb and Th/Nb ratios of the alkali basaltic melts imply derivation from a Nb-enriched zone.
In Table 4 and Fig. 12 we present a possible model for the petrogenesis of the alkali basaltic Golan end-member (Wy-337). It includes: (1) equilibration of hydrous fluids with a subducting slab (an arbitrary eclogite/peridotite ratio of 1:1 and a fluid/rock ratio of 0·005); (2) infiltration of these fluids and hydration of the mantle wedge, producing amphibole peridotites with Nb-enriched and Nb-depleted zones; (3) partial dehydration of the amphibole peridotite (fluid/rock ratio of 0·005), which significantly reduced, although it did not eliminate, the content of amphibole in the source rock and which caused the large decrease in the Rb content of the source rock; finally, (4) partial melting (13%), which produced the alkali basaltic end-member melt. Further details of the model can be found in the notes to Table 4.
|
). Also shown is the composition of the amphibole peridotite source after partial dehydration (APR). (See Table 4 and text for details.)
|