Journal of Petrology | Volume 43 | Number 8 | Pages 1435-1467 | 2002
© Oxford University Press 2002
Petrogenesis of the Back-arc East Scotia Ridge, South Atlantic Ocean
S. FRETZDORFF1,*,
R. A. LIVERMORE2,
C. W. DEVEY3,
P. T. LEAT2 and
P. STOFFERS1
1INSTITUTE OF GEOSCIENCES, UNIVERSITY OF KIEL, OLSHAUSENSTRASSE 40, 24118 KIEL, GERMANY
2BRITISH ANTARCTIC SURVEY, HIGH CROSS, MADINGLEY ROAD, CAMBRIDGE CB3 0ET, UK
3FACHBEREICH 5—GEOWISSENSCHAFTEN, UNIVERSITY OF BREMEN, POSTFACH 330 440, 28334 BREMEN, GERMANY
Received
January 3, 2001;
Revised typescript accepted
January 22, 2002
 |
ABSTRACT
|
|---|
The East Scotia Ridge is an active back-arc spreading centre
located to the west of the South Sandwich island arc in the
South Atlantic Ocean, consisting of nine main segments, E1 (north)
to E9 (south). Major and trace element and Sr–Nd–Pb
isotope compositions are presented, together with water contents,
for lavas sampled along the active ridge axis. Magmatism along
the East Scotia Ridge is chemically heterogeneous, but there
is a common mid-ocean ridge basalt (MORB)-type source component
for all the magmas. An almost unmodified MORB-source mantle
appears to underlie the central part of the back-arc. Subduction
components are found at the northern and southern ends of the
ridge, and there is a marked sediment melt input of up to 2%
in segment E4. Enriched (plume) mantle is present beneath segment
E2 at the northern end of the ridge, suggesting that plume mantle
is flowing westward around the edges of the subducting slab.
The southern part of segment E8 is unique in that its magma
source is similar to sub-arc depleted mantle.
KEY WORDS: geochemistry; petrogenesis; volcanism; back-arc; subduction
|
INTRODUCTION
|
|---|
Fractional melting of mantle peridotite is the primary source
of back-arc magmas and it has been postulated that the major
source component should be similar to that of mid-ocean ridge
basalt (MORB, e.g.
Hawkins, 1976;
Saunders & Tarney, 1979).
However, most back-arc basin basalts have geochemical characteristics
that show that subduction processes are also involved in their
petrogenesis, e.g. basalts erupted in the Mariana Trough and
the Lau Basin (e.g.
Hawkins, 1976;
Hawkins & Melchior, 1985;
Volpe et al., 1987;
Stern et al., 1990;
Gribble et al., 1998;
Turner & Hawkesworth, 1998;
Peate et al., 2001). These characteristics
provide evidence for prior depletion of the mantle source and
enrichment in H
2O and elements such as Ba, Th, U and Pb, thought
to be transported into the mantle wedge from the subducting
slab via sediment melts or aqueous fluids (
Stern et al., 1990;
Saunders et al., 1991;
Woodhead et al., 1993;
Stolper & Newman, 1994;
Pearce et al., 1995;
Gribble et al., 1998). Back-arc
basins can therefore be used to model the transport of slab
components into the mantle wedge over much wider areas in subduction
zone environments than can be achieved by studying volcanic
arcs alone, and can be used to test models of mantle wedge convection.
Furthermore, the transitional character of back-arc basalts
from MORB to arc-like provides critical evidence of the relative
roles of decompression vs volatile fluxed melting of mantle
(
Gribble et al., 1998).
The East Scotia Ridge, an active back-arc spreading centre behind the South Sandwich island arc in the South Atlantic Ocean (Fig. 1), is an interesting place to study subduction-related processes for several reasons: (1) its simple tectonic setting as part of an intra-oceanic arc, without the influence of continental crust; (2) its close proximity to the active island arc, suggesting the possibility of a strong subduction influence on the back-arc mantle source (Livermore et al., 1997); (3) the well-defined chemical signature of the associated island arc (Pearce et al., 1995) and of the sediments subducting at the South Sandwich trench (Barreiro, 1983; Ben Othman et al., 1989; Plank & Langmuir, 1998). Recent campaigns by the British Antarctic Survey (Livermore et al., 1999) and the University of Kiel now provide sampling of the neo-volcanic zone within each segment of the East Scotia Ridge. In this paper, we present major and trace element compositions together with Sr–Nd–Pb isotope data for basalt samples dredged from the ridge axis, to identify petrogenetic processes controlling geochemical variations and to identify the components contributing to the magma source of the East Scotia Ridge lavas.
View larger version (30K):
[in this window]
[in a new window]
|
Fig. 1. (a) Tectonic setting of the East Scotia Ridge within the Scotia Sea area. (b) Sketch map of the South Sandwich subduction zone with the location of the East Scotia Ridge segments (E1–E9) and South Sandwich Islands (after Leat et al., 2000). The arrows mark the spreading direction along the East Scotia Ridge and the relative movement of the subducting plate.
|
|
|
TECTONIC SETTING
|
|---|
The East Scotia Ridge is situated some 250–300 km to the
west of the South Sandwich trench (Fig.
1). It was mapped for
the first time in 1995 using the HAWAII-MR1 swath sonar (
Livermore et al., 1995)
and shown to consist of nine segments, separated
by non-transform offsets. Spreading rates of 60–70 km/Myr
place it in the intermediate range of spreading centres, transitional
between ‘fast’ (axial high) and ‘slow’
(median valley) morphologies. Six segments, E3–E8, are
characterized by faulted median valleys similar to those observed
at the Mid-Atlantic Ridge, whereas segments E2 and E9 both display
axial volcanic ridges (Fig.
2), and are propagating into the
back-arc region (
Livermore et al., 1997;
Bruguier & Livermore, 2001).
Segment E1 extends northward into the South Sandwich
trench in the form of a trough with water depths of up to 5500
m. Side-scan images of this segment show it to be surrounded
by a much more chaotic backscatter pattern compared with the
rest of the East Scotia Ridge, indicating recent establishment
of E1 and/or a more disordered mode of extension (
Livermore et al., 1997).
There is a distinct along-axis trend in axial
depth, from maxima of >4000 m in segments E5 and E6, to minima
of
2600 m in E2 and E9. These characteristics have provided
support for models involving the shallow inflow of Atlantic
mantle around the ends of the subducting slab (
Livermore et al., 1997).
View larger version (89K):
[in this window]
[in a new window]
|
Fig. 2. HAWAII-MR1 bathymetry of the East Scotia Ridge, showing sample locations occupied during British Antarctic Survey (BAS) cruises JR09 (dredged samples: red triangles), JR12 (wax core samples: red circles; dredged samples: red triangles), JR39b (wax core samples: yellow circles) and the German Polarstern cruise PS47 (dredged samples: red squares).
|
|
The South American plate is subducting beneath the Sandwich plate at a rate of 70–85 km/Myr (Pelayo & Wiens, 1989). Earthquake data indicate that the subducting plate dips at 45–55° to the west, probably steepening slightly in the southern part of the subduction zone (Brett, 1977). The age of the subducting plate is 28–35 Ma below the southern part of the arc, and 50–60 Ma below the northern part of the arc (Sclater et al., 1976; Barker & Lawver, 1988). The sedimentary cover on the subducting plate consists of some 200 m of siliceous sediments in the south, and some 400 m of calcareous and siliceous sediments in the north (Barker, 1995). More than 95% of this sediment input to the trench is subducted, and there is no significant sediment accretion at the trench (Vanneste & Larter, 2002).
The South Sandwich arc consists of 11 main islands forming a distinctly curved island arc, 500 km in length. Most of the islands have abundant evidence of recent volcanic activity, and all are entirely volcanic in origin. The volcanoes belong to the tholeiitic and (relatively rare) calc-alkaline series, and the arc is regarded as a classic example of the primitive stages of island arc development (Baker, 1968, 1990; Pearce et al., 1995). All the arc magmas are depleted in high field strength elements (HFSE) such as Ti, Zr, Hf, Nb and Ta, and heavy rare earth elements (HREE), relative to normal MORB, and are therefore thought to be derived from mantle wedge material that had experienced melt extraction before the magma generation events (Hawkesworth et al., 1977; Pearce et al., 1995). All the magmas are enriched (relative to MORB) in highly incompatible fluid-mobile trace elements, such as Pb, U, Ba and Rb, derived from the subducting plate. Pearce et al. (1995) suggested that variations in incompatible trace element abundances were a result of dynamic melting processes within the sub-arc mantle and that the Nd and Sr isotope covariations of the island arc sample suite indicate the involvement of a subduction component derived from the subducted sediment and altered oceanic crust. We argue below that the sediment-derived component has locally influenced the back-arc magmatism.
Previous geochemical studies
Previous investigations of lavas from the East Scotia Ridge were confined to eight dredge sites (D20–D24, D56, D57, D60) sampled by R.R.S. Shackleton in 1974 and 1981, as well as volcanic rocks from the axis and the lateral flanks of segment E2 and the northern part of E3, sampled by R.R.S. James Clark Ross in 1996 (Fig. 2). Dredge sites were located on the northern tip of segment E3 (D20), east of the southern tip of E2 (D21), west of segment E5 (D22), close to the northern tip of segment E9 (D24), in the centre of segment E9 (D23), and east of the northernmost segment E1 (D56). Apart from the recently sampled segment E2 and the northern part of E3, only the dredge stations D20 and D23 are located in the neo-volcanic axis of the East Scotia Ridge. The petrographic characteristics, major and trace element geochemistry, volatile content, and Sr, Nd, Pb, O and C isotope compositions of lavas dredged at the four locations D20 and D22–24 have been reported in earlier studies (Hawkesworth et al., 1977; Tarney et al., 1977; Saunders & Tarney, 1979; Muenow et al., 1980; Cohen & O’Nions, 1982a; Saunders et al., 1982; Mattey et al., 1984; Newman & Stolper, 1996; Eiler et al., 2000). These studies showed that the lavas have an intermediate geochemical composition between MORB and island-arc tholeiite, leading to the suggestion that fluids and/or sediments derived from the subducting slab influence the mantle source of the East Scotia Ridge basalts (e.g. Saunders & Tarney, 1979; Tarney et al., 1981; Saunders et al., 1982). More recently, Pearce et al. (2001) reanalysed a suite of samples from dredge sites on segments E3 and E9. They showed that the back-arc lavas have Pb and Nd isotope characteristics of South Atlantic, rather than Pacific mantle. This implies that any outflow of Pacific MORB through the Drake Passage as suggested by Alvarez (1982) does not extend as far as the East Scotia Ridge.
Most of the back-arc lavas have lower 206Pb/204Pb ratios than those of the South American–Antarctic Ridge (Pearce et al., 2001). The characteristic geochemical signature of the South American–Antarctic Ridge basalts is interpreted to be the result of a westward asthenospheric flow of enriched mantle from the Bouvet plume along the ridge axis (e.g. Le Roex et al., 1985). Pearce et al. (2001) suggested that this Bouvet plume component is also present in the East Scotia Ridge mantle source as a result of asthenospheric inflow into the back-arc region from the north and south. Leat et al. (2000) carried out a detailed geochemical study of the E2 segment. Their results support the interpretation of Pearce et al (2001) that material from the Bouvet mantle plume is migrating westwards into the back-arc, and showed that lavas from the flanks of segment E2 have a higher plume influence than axial lavas. Furthermore, they demonstrated the contribution of a slab-derived component to most E2 lavas.
Sampling
Sampling stations on the active part of the back-arc spreading ridge were selected on the basis of detailed side-scan sonar and bathymetric data (Livermore et al., 1995; Fig. 2). In 1997–1998, sampling of segments E3–E7 was carried out using a drum-shaped dredge during R.V. Polarstern cruise PS47 (ANT XV/II), with general dredge direction along the spreading axis. Fresh volcanic material, e.g. pillow lavas, sheet flows and lava tubes, most of which preserved fresh glassy margins, were recovered during the 10 dredge deployments during cruise PS47 (Fig. 2). During R.R.S. James Clark Ross cruise JR39b in 1999, fresh volcanic glass fragments were obtained at 28 sites from segments E2 to E9 using a wax corer (Livermore et al., 1999). These fresh volcanic samples, when combined with previous high-density sampling of segment E2 (Leat et al., 2000) during cruise JR12, provide complete coverage of the East Scotia Ridge neo-volcanic zone (Fig. 2, coordinates listed in Table 1).
|
ANALYTICAL METHODS
|
|---|
Macroscopic criteria (vesicularity; size, abundance and variety
of minerals) were used to select different rock types from each
dredge haul, from which glassy pillow rims were separated under
a binocular microscope and washed several times in ultra-pure
water. Glass fragments in wax cups from the wax corer were removed
from the wax with tweezers. The fragments were then placed into
a beaker with water in a microwave oven for 2 min to melt the
wax. The beaker was cooled and the wax removed. The sample was
then washed several times in ultra-pure water.
Analyses of minerals were made at the University of Kiel (Institut für Geowissenschaften) with a Cameca CAMEBAX electron microprobe running at a beam current of 12·5 nA and acceleration voltage of 14–15 kV. Usually, the beam width was about 1–2 µm, and was defocused for plagioclase analyses. Synthetic and natural standards were used to calibrate the microprobe. Repeat measurements show that the relative standard deviations are smaller than ±0·5%, indicating a good reproducibility for all major elements. The major element concentrations of the volcanic glass samples were determined with a Cameca SX50 electron microprobe at GEOMAR (Kiel) (Table 1). A beam current of 10 nA with an accelerating voltage of 15kV was used. The beam was defocused to 5 µm to minimize Na loss during measurements. Glass analyses reported are the averages of more than five individual spot analyses (Table 1) with relative standard deviations smaller than ±0·5%. International glass standards (JDF-D2, CFA47 2) were measured after every 20 analyses to check precision, which is better than ±0·7%.
H2O analyses were obtained by Fourier transform infrared (FTIR) spectroscopy at the University of Kiel using the methods and calibration described by Stolper (1982). Double-polished glass thin sections, 100–250 µm thick, were analysed with a Bruker IFS 66v/S FTIR spectrometer equipped with a Bruker A 590 Infrared Microscope with a mercury–cadmium–telluride detector. Optically clear areas of known thickness (±5 µm) were measured with aperture size of 120 µm diameter with 256 scans per spot. Concentrations were calculated using the Beer–Lambert law (e.g. Stolper, 1982) with assumed densities of 2·7 g/cm3. Replicate analyses of different glass fragments from the same specimen were typically reproducible to ±10%. Comparison of the H2O concentrations of two basalt glasses (EN112 4D-10; P1505-1) analysed by FTIR spectroscopy at the University of Tulsa (USA) (Michael, 1995) with the FTIR analyses at University of Kiel show reproducibility better than ±5%.
Trace element concentrations were determined by inductively coupled plasma mass spectrometry (ICP-MS) with a VG PlasmaQuad PQ1 system at the University of Kiel, using preparation and measurement methods described by Garbe-Schönberg (1993). Analytical accuracy and reproducibility are estimated from measurements of international rock standards BHVO-1 and BIR, and duplicate analyses of samples (Table 1). The accuracy of the standards is within ±10% of the suggested working values (Jenner et al., 1990; Govindaraju, 1994), but generally better than ±5% for rare earth elements (REE). Duplicate analyses show reproducibility to be better than 3%, except for the transition metals (e.g. Cr, Ni, Cu, Zn, Ga) and elements with very low concentrations (e.g. Cs), which show deviations of up to 20% (Table 1).
Sr, Nd and Pb isotope ratios of volcanic glasses were measured on a MAT262-RPQ2+ thermal ionization mass spectrometer at GEOMAR in static mode. Glass chips were leached for 1 h in 6N HCl before sample dissolution, and analytical procedures followed those described by Hoernle & Tilton (1991). 87Sr/86Sr ratios were normalized to 86Sr/88Sr = 0·1194 within-run and 143Nd/144Nd ratios to 146Nd/144Nd = 0·7219. The long-term reproducibility of NBS 987 in this laboratory is 87Sr/86Sr = 0·710254 ± 17 (n = 83) and of La Jolla 143Nd/144Nd = 0·511845 ± 11 (n = 123). The in-house SPEX monitor, calibrated against La Jolla with 143Nd/144Nd = 0·511706 ± 12 (n = 9), yielded 143Nd/144Nd = 0·511705 ± 11 (n = 11) over the course of this study. The long-term reproducibility of NBS 981 (n = 63) in this laboratory is 206Pb/204Pb = 16·896 ± 5, 207Pb/204Pb = 15·437 ± 7 and 208Pb/204Pb = 36·524 ± 21. During the course of this study, NBS 981 analyses (n = 8) gave 206Pb/204Pb = 16·893 ± 10, 207Pb/204Pb = 15·433 ± 13 and 208Pb/204Pb = 36·514 ± 41, and were corrected to the values of Todt et al. (1996). Total chemistry Pb blanks were <0·3 ng and thus negligible. In Table 2 we show the isotope analyses of selected East Scotia Ridge samples and the 2
standard deviations.
|
RESULTS
|
|---|
Petrography and mineral chemistry
We present a general summary of the major petrographic features
and mineralogical characteristics of the dredged lavas from
segments E3–E7 in Table
3. Individual mineral analyses
are not presented here, but can be downloaded from the
Journal of Petrology Web site at
http://www.petrology.oupjournals.org.
The phenocrysts are dominated by plagioclase and olivine phenocrysts
with subordinate clinopyroxene, chrome spinel and magnetite.
The most visible differences between the samples are in both
their phenocryst contents and their vesicularity. Samples from
segment E3 (96DS, 97DS, 98DS) contain some plagioclase and olivine
phenocrysts (5%) and few vesicles, whereas the lavas from segment
E4 (99DS, 100DS) are highly vesicular, but have an almost aphyric
groundmass and sparse olivine and plagioclase phenocrysts (<3%).
Most of the samples from dredge locations in segments E5 (104DS),
E6 (106DS, 107DS) and E7 (109DS, 110DS) are slightly vesicular
and exhibit a far higher plagioclase phenocryst content (30%)
than the other segments.
Olivine phenocrysts are mostly rounded or highly skeletal crystals. Subordinate olivine is present in the matrix and glass rims. Generally, individual olivine phenocrysts show a comparable range of Fo89 (cores) to Fo83 (rims). There are no distinct chemical differences between phenocrysts and matrix minerals. Calculations show that most olivines are in equilibrium with the matrix glass at atmospheric pressure for a Fe/Mg KD of 0·3 (Roeder & Emslie, 1970; Ulmer, 1989).
Plagioclase phenocrysts are up to 5 mm long in volcanic rocks from segments E5–E7 and with smaller sizes of up to 2 mm in the lavas of segments E3 and E4. Some phenocrysts show zonation from calcic cores to more sodic rims. Core compositions fall in the range An92–85 and rims, as well as unzoned phenocrysts, have An contents of 78–87. Matrix plagioclase crystals are the least calcic with the range An70–82.
Generally, opaque minerals are scarce in East Scotia Ridge lavas. Titanomagnetite, magnetite and chrome spinels mostly form inclusions in olivine phenocrysts in the lavas of segments E3 and E4. The cr-number in the spinels ranges from 62 to 66.
Magmatic geochemistry
Major element analyses of the dredged and wax core samples are presented in Table 1. Figure 3 shows all the East Scotia Ridge lavas as well as the chemical trends of the subduction-related magmas from the South Sandwich Islands in the K2O–SiO2 classification diagram of Peccerillo & Taylor (1976). The majority of the back-arc samples range from basalt to basaltic andesite and are primarily low-K tholeiites, although some show medium-K compositions. For a given SiO2 abundance, samples from segments E2 and E4 are significantly enriched in K2O relative to the other segments. MgO contents for the majority of segments lie in the restricted range of 6–9 wt % and have mg-number between 60 and 70. However, segments E2 and E8 have yielded samples with MgO contents as low as 3 wt % and with mg-number as low as 40. These two segments also show systematic variations of CaO with MgO (Fig. 4a). In the plot of Al2O3 against MgO, each segment forms a distinct trend, parallel to those from the other segments and to the Mid-Atlantic Ridge (Fig. 4b).
View larger version (22K):
[in this window]
[in a new window]
|
Fig. 4. Variation of CaO and Al2O3 (wt %) vs wt % MgO for samples from the East Scotia Ridge. In (a), the continuous and dashed lines represent calculated liquid compositions from crystal fractionation models determined with the GPP program of Geist et al. (1985). In (b), the continuous lines with arrows show the crystal fractionation trends of the E2 (Leat et al., 2000), E3, E7 and E8 segments, and Mid-Atlantic Ridge lavas (MAR: Schilling et al., 1983). Each segment forms a distinct trend, parallel to those from the other segments and to the Mid-Atlantic Ridge. The dashed line with an arrow is perpendicular to the fractionation trends and may reflect either the effect of varying amounts of water on plagioclase crystallization or different Al2O3 contents in primary magmas of each segment (see text for details).
|
|
Water contents of selected East Scotia Ridge magmas cover the range of 0·2–2·3 wt % (Table 1). Water contents of MORB vary between 0·05 and 1·3 wt % (Michael, 1995), although most N-MORB have contents between 0·1 and 0·4 wt % (Michael & Chase, 1987; Dixon et al., 1988; J. E. Dixon, personal communication, 2001). All samples from East Scotia Ridge segments E6 and E7 show H2O contents within this N-MORB range. Some samples from segments E2, E3, E4 and E9, however, contain up to 2·2 wt % H2O (Table 1). It was not possible to measure the H2O content of all the volcanic glasses along the back-arc, because of the presence of vesicles and microcrystals. We measured the H2O content of only two E2 samples from the centre of the segment (WX42, WX43; Table 1).
Selected volcanic glass samples were analysed for trace elements and compositions are listed in Table 1. Comparison of the relative abundances of highly and moderately incompatible trace elements along the East Scotia Ridge is facilitated by using N-MORB-normalized (Hofmann, 1988) abundance profiles (Fig. 5). Relative to N-MORB, the East Scotia Ridge volcanic glasses have higher abundances of large ion lithophile elements (LILE) such as K, Pb, Rb and Ba, but are less enriched than bulk South Atlantic sediment (from Plank & Langmuir, 1998). In contrast, the bulk sediment pattern has lower HREE abundances than the back-arc magmas. Trace element patterns of the South Sandwich Islands samples are sub-parallel to those of the back-arc glasses, although the island arc magmas tend to be more depleted in Nb, Ta, Zr and Hf.
Isotopic data for selected East Scotia Ridge samples are given in Table 2. The back-arc magmas are less radiogenic in terms of Sr and Pb than those from the South Sandwich Islands (for discussion, see section on components contributing to the back-arc source and Figs 15 and 16) and in the case of segments E6 and E7 show isotopic ratios as low as an average MORB (from Cohen & O’Nions, 1982b). The Nd isotopic ratios of the back-arc magmas are comparable with those from the island arc. Both island arc and back-arc are much less radiogenic in Sr and more radiogenic in Nd compared with bulk South Atlantic sediment. The Pb isotopes in the sediments are similar to those in the island-arc magmas and lie at the radiogenic extreme of the back-arc magma fields.
View larger version (33K):
[in this window]
[in a new window]
|
Fig. 15. 143Nd/144Nd (a) and 87Sr/86Sr (b) vs 206Pb/204Pb for the East Scotia Ridge lavas (E2: Leat et al., 2000), South Sandwich island arc (Pearce et al., 1995), South American–Antarctic Ridge (SAAR; Dickey et al., 1977; Le Roex et al., 1985; Ito et al., 1987; Kurz et al., 1998), southern Pacific MORB (Castillo et al., 1998), bulk South Atlantic sediment (Ben Othman et al., 1989; Plank & Langmuir, 1998), Bouvet Island (Sun, 1980; Weaver et al., 1987), and Discovery [average composition from Douglass et al. (1999)]. For samples DR.157 and DR.158 dredged on the flanks of the E2 segment, data are from D. Harrison (unpublished data, 2000; location sites shown in Fig. 2). Data from additional E3 (D20) and E9 (D23) samples, indicated with a ‘P’, are from Pearce et al. (2001). Three end-members are suggested to be involved in the petrogenesis of the East Scotia Ridge lavas and calculated mixing trends are indicated. Literature data: ‘MORB source’ from McKenzie & O’Nions (1995) and Cohen & O’Nions (1982b); ‘plume mantle’ from Sun & McDonough (1989), McDonough & Sun (1995) and McKenzie & O’Nions (1995); ‘sediment melt’ from Ben Othman et al. (1989), Plank & Langmuir (1998) and Class et al. (2000). Data used for mixing calculations in (a) MORB source–bulk sediment: 206Pb/204Pb: 17·8 and 18·6; 0·05, 10 and 3 ppm; 143Nd/144Nd: 0·5132 and 0·5125; 1·2, 34 and 90 ppm; MORB source–plume mantle: 206Pb/204Pb: 17·8 and 19·5; 0·05 and 0·15 ppm; 143Nd/144Nd: 0·5132 and 0·5129; 1·2 and 1·25 ppm; MORB source + 0·4% bulk sediment–plume mantle: 206Pb/204Pb: 17·94 and 19·5; 0·06 and 0·15 ppm; 143Nd/144Nd: 0·5130 and 0·5129; 1·23 and 1·25 ppm; in (b) MORB source–bulk sediment: 206Pb/204Pb: 17·8 and 18·6; 0·05 and 10·3 ppm; 87Sr/86Sr: 0·7025 and 0·7080; 14·7, 170 and 440 ppm; MORB source + 0·4% bulk sediment–plume mantle: 206Pb/204Pb: 17·94 and 19·5; 0·05 and 0·15 ppm; 87Sr/86Sr: 0·7030 and 0·7035; 14·7 and 19·9 ppm.
|
|
View larger version (34K):
[in this window]
[in a new window]
|
Fig. 16. 208Pb/204Pb (a) and 207Pb/204Pb (b) vs 206Pb/204Pb for the East Scotia Ridge lavas (E2; Leat et al., 2000), South Sandwich island arc (Pearce et al., 1995), South American–Antarctic Ridge (Dickey et al., 1977; Le Roex et al., 1985; Ito et al., 1987; Kurz et al., 1998), Pacific MORB (Castillo et al., 1998), bulk South Atlantic sediment (Barreiro, 1983; Ben Othman et al., 1989; Plank & Langmuir, 1998), and Bouvet Island (Sun, 1980; Kurz et al., 1998), and Discovery (average composition from Douglass et al., 1999). For samples DR.157 and DR.158, dredged on the flanks of the E2 segment, data are from D. Harrison (unpublished data, 2000; location sites shown in Fig. 2). Data from additional E3 (D20) and E9 (D23) samples, indicated with a ‘P’, are from Pearce et al. (2001). Data used for mixing calculations in (a) MORB source–bulk sediment: 206Pb/204Pb: 17·8 and 18·6; 0·05 and 10 ppm; 208Pb/204Pb: 37·4 and 38·6 ppm; in (b) 207Pb/204Pb: 15·45 and 15·62 ppm.
|
|
|
MODELLING FRACTIONAL CRYSTALLIZATION
|
|---|
The large range of major element compositions covered by the
samples from segments E2 and E8 (see, for example, Figs
3 and
4) can be modelled at least partially as the product of simple
crystal fractionation. Unfortunately, as these segments were
sampled only with a wax corer, it is impossible to check our
results against the petrography of the lavas. We have performed
crystal fractionation calculations using the least-squares GPP
program of Geist
et al. (1989) and mineral compositions measured
on phenocrysts from other segments. The calculated fractionation
trends are shown in the CaO vs MgO diagram (Fig.
4a), together
with the mineral contents of the cumulate. The trends can be
modelled as the result of either one- (E2) or two-stage (E8)
crystal fractionation. Olivine, plagioclase, and subordinate
chrome spinel can produce the trend between 9 and 7 wt % MgO
in the E8 magmas, whereas below 7 wt % MgO the decrease in CaO
in both E2 and E8 reflects the appearance of clinopyroxene in
the fractionating assemblage.
One of the most differentiated glass samples of segment E2 (WX33, Leat et al., 2000) could be produced from the assumed parental basalt (WX21, Leat et al., 2000) by a total of 50% fractional crystallization: 27·3% clinopyroxene, 20·8% plagioclase, 2·4% olivine, 0·5% chrome spinel (R2 = 0·24). The first step fractionation model for samples from segment E8 yielded 10% crystal fractionation from the assumed parental (WX58) to daughter magma composition (WX62): 5% plagioclase, 3·5% olivine, 0·8% chrome spinel (R2 = 0·33). During the second step fractionation, the most differentiated glass sample of segment E8 (WX60) could be produced by 40% fractional crystallization: 19·5% plagioclase, 17·5% clinopyroxene, 4·4% olivine, 0·1% chrome spinel (R2 = 0·11).
It is important to note that elevated water contents increase the mineral–liquid Kd of plagioclase and suppress plagioclase crystallization (e.g. Eggler & Burnham, 1973; Michael & Chase, 1987; Sisson & Grove, 1993; Hergt & Farley, 1994). At present, these effects cannot be quantified in a rigorous way and so the calculations discussed above should only be taken as indicative of the fractional crystallization processes that may have affected the lavas.
The East Scotia Ridge lavas have high Al2O3 for a given MgO content compared with the Mid-Atlantic Ridge (MAR) glasses (Fig. 4b). High Al2O3 concentrations compared with MORB have also been observed in other back-arc regions (e.g. Mariana Trough, Sumisu–Torishima back-arc rifts) and have been attributed to the greater H2O content in back-arc magmas during crystallization (Sinton & Fryer, 1987; Fryer et al., 1990). The differences in Al2O3 contents between segments shown in Fig. 4b may reflect either the effect of varying amounts of water on plagioclase crystallization or differences in Al2O3 contents of magmas in different segments. Experimental work (Gaetani & Grove, 1998) has shown that varying amounts of water in the source at the time of melting will not affect the Al content of the primary magma and so an investigation of the relationship between Al and H2O in the magmas should be able to distinguish between these two possibilities. Moreover, the measured H2O content of the glasses should reflect unperturbed magmatic values, as degassing during ascent and eruption has little effect on dissolved H2O contents unless eruption depths are less than a few hundred metres (e.g. Stolper & Newman, 1994). This is confirmed for the East Scotia Ridge by our observations that there is no correlation between eruption depth and measured water contents in the glasses.
First, however, we must remove the effects of low-pressure crystal fractionation on Al2O3 and H2O in the erupted magmas. For Al2O3, this involves calculating Al8 (Al2O3 content of the magma recalculated to 8 wt % MgO along the fractionation vector for each segment) in a manner similar to the Fe8 or Na8 values of Klein & Langmuir (1987). We chose samples for H2O determinations that are relatively primitive (>6% MgO), and so do not expect the effects of crystal fractionation on water contents to be large. We used the quantified fractionation assemblages calculated above and the Rayleigh distillation equation (see, e.g. Cox et al., 1979) to work out the effects of fractionation on magmatic water contents, assuming complete incompatibility of water in the cumulate assemblage. Figure 6 shows the relationship between (H2O)8 and Al8. We see a positive correlation between these two parameters which, in view of the possibilities outlined above, we interpret as reflecting the effect of magmatic water content on plagioclase solubility in the melt. Michael & Chase (1987) have shown that the MORB crystallization sequence is olivine 
olivine + plagioclase olivine + plagioclase + clinopyroxene. Our interpretation of the correlation between (H2O)8 and Al8 is that magmas with higher water contents progress further along the olivine-only fractionation path (shown in Fig. 4b) and so reach higher Al2O3 contents before plagioclase starts to precipitate.
View larger version (11K):
[in this window]
[in a new window]
|
Fig. 6. Fractionation corrected Al8 (wt %) concentration vs (H2O)8 (wt %) content for the East Scotia Ridge glasses. The Al2O3 content of the magma has been recalculated to 8 wt % MgO along the fractionation vector for each segment in a manner similar to Fe8 or Na8 of Klein & Langmuir (1987). Quantified fractionation assemblages (see fractional crystallization calculation in text and in Fig. 4a) and the Rayleigh distillation equation (see, e.g. Cox et al., 1979) have been used to work out the effects of fractionation on magmatic water contents assuming complete incompatibility of water in the cumulate assemblage. Data for dredged samples from segments E3 (D20) and E9 (D23) (see Fig. 2) are from Muenow et al. (1980) and are indicated with a ‘D’.
|
|
In general, segments E2 and E8 are most strongly affected by fractional crystallization processes, with differentiated samples reaching MgO contents as low as 3 wt %. Interestingly, the samples with lowest MgO concentrations erupt near the topographic highs at the centres of these two segments. Surprisingly, lavas of segment E9, which shows a very pronounced topographic high, are not strongly fractionated. Geophysical investigations of segment E2 (Livermore et al., 1997) showed that there is a melt lens situated directly beneath the topographic high at the segment centre, 3 km beneath the sea floor. Leat et al. (2000) suggested that the highly fractionated magmas observed on the high probably formed within the seismically imaged melt lens. If this is the case, then we might expect further melt lenses to exist beneath the topographic highs on segments E8 and E9. The absence of highly fractionated magmas on E9, despite the existence of a pronounced axial high (Bruguier & Livermore, 2001), suggests that sea-floor topography is not always a reliable indicator of the presence of a persistent magma chamber.
|
CHEMICAL VARIATIONS AND CHARACTERIZATION OF COMPONENTS ALONG THE RIDGE
|
|---|
Chemical variations along the ridge
There are systematic variations in composition of lavas along
the East Scotia Ridge. Na
8 values, which indicate degree of
mantle partial melting (
Klein & Langmuir, 1987), are roughly
positively correlated with axial depth along the ridge, in that
the highest Na
8 samples are generally from deeper parts of the
ridge (Fig.
7a), and shallow segments E2 and E8 have low Na
8 values. Most samples from the central segments E4–E7 have
relatively high and roughly constant Na
8. Segment E9 also has
high Na
8, although it is as shallow as segment E2. The relationships
between segments E2–E8 indicate that axial morphology
may have been grossly controlled by degree of partial melting.
Nb/Yb, a ratio unaffected by additions from the subducting slab,
but sensitive to the degree of partial melting, is approximately
positively correlated with axial depth (Fig.
7b). Lavas with
generally low, N-MORB-like, Nb/Yb ratios dominate in rift-like
segments E5–E7 near the centre of the ridge, whereas high
Nb/Yb lavas occur in the shallow, inflated segments E2 and E9.
However, high Nb/Yb samples have also erupted in segment E4,
which is rift-like and deep, but not in segment E8, which is
shallow. The high Nb/Yb ratios in segment E2 are not the result
of low degrees of partial melting of MORB-source mantle, because
low degrees of partial melting would be characterized by both
high Nb/Yb and high Na
8. This is not observed (Fig.
7). Because
Nb is not likely to have been introduced from the subducting
slab (
Pearce & Peate, 1995), segments E2, E9 and part of
E4 seem to have been supplied by a more enriched, higher Nb/Yb
mantle source than MORB. The southern part of segment E8 contains
low Na
8 lavas that also have Nb/Yb ratios significantly below
the N-MORB ratio. These characteristics are consistent with
relatively high degrees of melting of a MORB or sub-arc depleted
mantle source.
Variations in REE along the ridge are shown in Fig. 8. Light REE (LREE) abundances can be increased by input from the subducting slab as well as affected by mantle composition and degree of partial melting (Pearce & Peate, 1995). HREE abundances are not affected by additions from the subducting slab. There are significant variations in REE along the East Scotia Ridge. All samples from segments E2 and E9 as well as samples from the overlapping region of segments E3 and E4 (hereafter E3–E4 overlapper) are enriched in LREE. Segments in the central part of the back-arc have low (La/Sm)N ratios comparable with N-MORB. The along-ridge variations in the (La/Sm)N ratio closely follow those in Nb/Yb (Figs 7 and 8), consistent with different mantle sources underlying different parts of the ridge. However, variable inputs from the subducting slab may also be present, especially in segments at the ends of the ridge. The (Dy/Yb)N ratios of samples from the ridge closely follow variations in Na8, and are lowest in segment E8, where they are similar to N-MORB (Figs 7 and 8). (Dy/Yb)N ratios are high in segment E9, moderately high in the central segments E5–E7 and relatively low in segment E2. The low (Dy/Yb)N ratios in segments E2 and E8 indicate that these segments had higher degrees of partial melting than other segments of the ridge, consistent with the low Na8 values of segments E2 and E8 (Figs 7 and 8).
View larger version (21K):
[in this window]
[in a new window]
|
Fig. 8. (a), (b) Chondrite-normalized (Sun & McDonough, 1989) REE ratios and water depth as a function of latitude along the East Scotia Ridge. The vertical lines mark the boundaries of the main ridge segments. In (a) and (b), dashed lines show the N-MORB average of Hofmann (1988). (For E2 segment data, see Fig. 7.)
|
|
The positive correlation of (La/Sm)N with 87Sr/86Sr (Fig. 9) reinforces our conclusion that the trace element variations cannot be the result of partial melting alone. The increase in (La/Sm)N with increasing 87Sr/86Sr could be a result either of addition of radiogenic Sr with LREE via subducted sediments, or the presence of an enriched (OIB) mantle component, such as Bouvet mantle plume, under segments E2, E3, E4 and E9. In the following sections, we distinguish the separate mantle and slab-derived components in the back-arc source.
View larger version (31K):
[in this window]
[in a new window]
|
Fig. 9. Variation of (La/Sm)N vs 87Sr/86Sr for lavas from the East Scotia Ridge (E2: Leat et al., 2000), South American–Antarctic Ridge (Dickey et al., 1977; Le Roex et al., 1985), South Sandwich island arc (Pearce et al., 1995), Bouvet Island (Le Roex & Erlank, 1982; Kurz et al., 1998), and the (La/Sm)N (Hofmann, 1988) and 87Sr/86Sr (Ito et al., 1987) composition of N-MORB.
|
|
Characterization of components contributing to the back-arc magma source
It has been argued that the characteristic enrichments in volcanic arc rocks of most highly incompatible trace elements (relative to MORB) result from the contribution of at least two components: sediment or sediment melt and aqueous fluid (Ellam & Hawkesworth, 1988; Ryan et al., 1995; Turner et al., 1996; Elliot et al., 1997; Turner & Hawkesworth, 1997; Class et al., 2000). The sediment component is characterized by high La/Yb, La/Sm and Th/Nb, and is variously regarded as addition of bulk sediment or partial melt of subducted sedimentary rocks. The slab-derived fluid has high B, B/Be, U/Th and Ba/Th, and may consist of variable proportions derived from dehydration of basaltic crust and from dewatering of sediments (Ishikawa & Tera, 1999; Class et al., 2000). Cross-arc geochemical traverses indicate that the fluid component dominates at the volcanic front, where hydration of the mantle causes the greatest amount of melting of the mantle wedge (Ryan et al., 1995). This suggests that subduction components in back-arc spreading centres may be poor in the fluid component.
The addition of subduction components to a mantle source can be assessed by comparing abundances of trace elements in lavas to a reference composition. Pearce & Peate (1995) compared arc lavas with a global MORB array and estimated the percentage contribution from subducted material for various elements. They found that LILE such as Ba, Rb, K, Pb, Th, U and Sr are dominated by subducted material, having a subduction zone contribution of >80%. Class et al. (2000) normalized to a local composition to isolate particular components. In Fig. 10, we have compared lavas from segments E2–E9 with sample WX48, which has N-MORB-like values of (La/Sm)N, Nb/Yb, H2O and 87Sr/86Sr (0·58, 0·75, 0·18 wt % and 0·702511, respectively) (Tables 1 and 2). Enrichment factors were calculated by dividing measured element/Yb ratios by the element/Yb ratio of sample WX48, normalized to the Nb/Yb ratio of the sample. This method effectively removes the effects of fractional crystallization and different degrees of partial melting for elements of similar incompatibility to Nb. It also removes the effects of different mantle compositions such as plume vs MORB-source mantles. The LILE are ordered in Fig. 10 according to relative importance of the subduction component to their abundances in arcs from Ba (highest) to Sr (lowest) as determined by Pearce & Peate (1995).
View larger version (31K):
[in this window]
[in a new window]
|
Fig. 10. Estimation of the addition of subduction-related components (Ba, Rb, K, Pb, Th, U, Sr), termed ‘enrichment factors’, to the East Scotia Ridge magma source relative to a MORB-like ‘baseline’ sample (WX48) from segment E5. Enrichment factors were calculated by dividing measured element/Yb ratios by the element/Yb ratio of sample WX48 normalized to the Nb/Yb ratio of the sample. This method removes the effects of fractional crystallization, different degrees of partial melting for elements of similar incompatibility to Nb and the effects of different mantle compositions. It should be noted that the scale of the plot showing segment E4 extends to higher values than the other diagrams. (See text for details.)
|
|
Enrichment factors are 
1 in nearly all cases. The enrichment factors of <1 for Sr in some samples probably reflect removal of Sr in plagioclase during fractional crystallization. On the basis of these calculations (Fig. 10), the central part of the East Scotia Ridge (segments E5–E7) shows only a minor influence of a subduction component. Samples from segment E7 consistently have the lowest enrichment factors, below two in most cases. Samples from segments E2–E5, E8 and E9 show highly variable subduction enrichment factors varying from one to 10 and even up to 15 in segment E4, and tend to be highest in the segments near the northern and southern ends of the ridge. Samples with high enrichment factors are interpreted to have very large contributions of LILE from subducted material. There are differences in the relative enrichment of LILE between samples within segments and from segment to segment. In general, Ba, Rb and Th are more enriched than K, U, Pb and Sr (Fig. 10). Th is relatively more enriched in the northern segments E2, E3 and E4 than the southern segments E8 and E9, suggesting regional variations in the importance of different processes or slab-derived components.
The relationships of LILE in the back-arc are shown in Fig. 11. Samples from the South Sandwich Islands form a negative correlation between Ba/Th and Th/Nb. This was interpreted by Leat et al. (2000) as a result of variations in the relative amounts of two different slab-derived components in the arc. The high Ba/Th component was interpreted to be aqueous fluid, and the high Th/Nb component was suggested to be sediment. Samples from the East Scotia Ridge are scattered between MORB (and plume mantle) and the arc, and trend away from MORB toward both arc components. An implication of Fig. 11 is that the two components present in the arc also influence compositions in the back-arc. Some samples from E2, E4 and E8 plot toward the high Th/Nb components, whereas most of the rest of the samples trend toward the high Ba/Th component. In this plot, it appears that the fluid component is dominant in most back-arc samples. If the identification of the origin of the components is correct, the few high Th/Nb samples contain a significant input from sediment or sediment melt, whereas the majority of samples are dominated by aqueous fluid from the slab.
View larger version (42K):
[in this window]
[in a new window]
|
Fig. 11. Variation of Ba/Th vs Th/Nb ratios for lavas from the East Scotia Ridge, South Sandwich island arc (Pearce et al., 1995), Bouvet Island (Le Roex & Erlank, 1982; Weaver et al., 1987), and the composition of N-MORB (Hofmann, 1988). Ba/Th and Th/Nb ratios for samples DR.157 and DR.158 are from D. Harrison (unpublished data, 2000; location sites shown in Fig. 2). The continuous lines with arrows are putative enrichments interpreted to have been caused by aqueous fluid (increasing Ba/Th for constant low Th/Nb ratios) and sediment addition (increasing Th/Nb for constant low Ba/Th ratios) from the subducting slab (see text for details).
|
|
The behaviour of these two possible components is further assessed in Fig. 12. Both Ba/Th and Th/Nb correlate positively with H2O abundances. In the case of Ba/Th, this is consistent with transport of Ba in an aqueous fluid. The positive correlation of Th/Nb with H2O suggests that Th is transported into the mantle source either in an aqueous fluid, or in a hydrous sediment melt. Several workers have argued that Th is transported from the slab in silicate melts (e.g. Elliott et al., 1997; Turner & Hawkesworth, 1997; Class et al., 2000). This is consistent with experimental data indicating that Th is preferentially partitioned into silicate melts but not into aqueous fluids (Brenan et al., 1995; Johnson & Plank, 1999), suggesting that Th requires sediment melting to be efficiently transferred to the arc. However, the partition behaviour of Th in aqueous fluids remains uncertain (Pearce & Peate, 1995) and some workers have suggested that, at high temperature, Th could be transported in an aqueous fluid (Stolper & Newman, 1994).
View larger version (13K):
[in this window]
[in a new window]
|
Fig. 12. Variation of (a) Ba/Th and (b) Th/Nb against the (H2O)8 (wt %) content (see Fig. 6 for calculation procedure) for the East Scotia Ridge glasses.
|
|
We have already shown, using Nb/Yb ratios, that there are variations
in the composition of the back-arc mantle source independent
of subduction components (Fig.
7). In Fig.
13 we investigate
the compositional effects of the slab-derived components and
plume vs MORB-source mantles on the back-arc lavas. Trends A
and B in Fig.
13 diverge strongly from N-MORB. Trend A is characterized
by positive correlation of Nb/Yb with Th/Yb, increasing Ta/Nd
at constant Th/Nb, and slightly increasing Th/Nd with decreasing
143Nd/
144Nd. This trend could be the result of MORB–plume
source mixing and/or variable degrees of mantle partial melting.
Trend B displays increases in Th/Yb and Th/Nb with moderate
increases in Nb/Yb and Ta/Nd, and strong negative correlation
between Th/Nd and
143Nd/
144Nd. We interpret this trend to represent
MORB–subduction component mixing. Trend A is defined mostly
by samples from segments E2 and E9 and therefore occurs symmetrically
at the northern and southern ends of the spreading centre. This
supports proposals that plume-influenced mantle is migrating
into the back-arc around the lateral edges of the subducting
plate (
Livermore et al., 1997;
Leat et al., 2000).
View larger version (21K):
[in this window]
[in a new window]
|
Fig. 13. Variation of Th/Yb vs Nb/Yb (a), Ta/Nd vs Th/Nb (b) and 143Nd/144Nd vs Th/Nd (c) for the East Scotia Ridge lavas (E2: Leat et al., 2000), the South Sandwich island arc (Pearce et al., 1995), Bouvet Island (Sun, 1980; Weaver et al., 1987) and bulk South Atlantic sediment (Plank & Langmuir, 1998). For samples DR.157 and DR.158, dredged on the flanks of the E2 segment, data are from D. Harrison (unpublished data, 2000; location sites shown in Fig. 2). Calculated mixing trends (grey lines) between a MORB-like end-member and enriched source components (Bouvet Island: Trend A; bulk South Atlantic sediment: Trend B) are shown. It should be noted that, in (a) and (b), samples from the southern part of segment E8 tend towards the data field of the South Sandwich island arc, which is suggested to be generated from sub-arc mantle modified by subduction-related components from the downgoing slab.
|
|
The subduction contribution (Trend B) is seen in the high Th/Yb and Th/Nb of samples from segments E2, E3, E4 and E8. Lavas from the southern part of E8 samples are distinctive in that they have low (Dy/Yb)N ratios (Fig. 8) and low Nb/Yb and Ta/Nd ratios (Fig. 13). Their mantle source is therefore interpreted to have experienced previous partial melt extraction, and hence depletion in incompatible elements. This is the only place in the spreading centre where previously depleted mantle appears to have been a magma source. Such depleted mantle is also inferred to have formed the source of all the South Sandwich arc lavas (Pearce et al., 1995). Samples from segments E2, E3 and E4 are enriched in the sediment component, trending to high Th/Nb and Th/Nd ratios, at Nb/Yb and Ta/Nd ratios that are higher than MORB. We interpret these compositions to have been formed by addition of the sediment component to an N-MORB-like source mantle. Some samples of segment E2, especially the low-Na8 group, as well as several lavas dredged on the flanks of the axial high, plot between the two mixing lines (Fig. 13).
The two trends are also clearly seen in plots of U8 and Nb8 vs (H2O)8 (Fig. 14). Trend A shows strong increase in U8 and Nb8 with moderate increase in (H2O)8. This trend is defined by most samples from segment E9 as well as the South American–Antarctic Ridge (note: we have no water contents for the most plume-influenced E2 samples, and so they do not appear in Figs 12 and 14). The South American–Antarctic Ridge is not influenced by the South Sandwich subduction system, but it has been suggested that the ambient MORB-source mantle beneath the ridge has been modified by mantle migrating westward from the Bouvet mantle plume (Le Roex et al., 1985; Kurz et al., 1998). Trend B in Fig. 14 shows constant, MORB-like Nb8 and increasing U8 with increasing (H2O)8 content, implying that magmas of the segments concerned (E2–E5, E9) were produced from sources that experienced addition of water in a subduction component.
View larger version (21K):
[in this window]
[in a new window]
|
Fig. 14. Variation of U8 (a) and Nb8 (b) vs (H2O)8 (wt %) content (see Fig. 6 for calculation procedure) for the East Scotia Ridge (ESR) volcanic glasses compared with (a) Mariana Trough lavas (Volpe et al., 1987; Stolper & Newman, 1994) and (b) South American–Antarctic Ridge lavas (Le Roex et al., 1985; Michael, 1995). It should be noted that two clear trends can be distinguished in both diagrams. Trend A is observed in segment E9 and South American–Antarctic Ridge lavas showing an increase of U8 and Nb8 at nearly constant water contents, whereas Trend B (samples from segments E2–E5) shows in (a) a positive correlation of H2O with U8 and in (b) constant Nb8 with increasing water content. Samples having MORB-like geochemical character are indicated. Water contents for dredged samples from segment E3 (D20; indicated with a ‘D’) are from Muenow et al. (1980); U ppm from Cohen & O’Nions (1982a) and Nb ppm from Saunders & Tarney (1979) and are recalculated to 8 wt % MgO (see Fig. 6 for calculation procedure).
|
|
To place quantitative constraints on the contribution of the various source components to the different ridge segments, we examined the Sr–Nd–Pb isotope systematics. Figures 15 and 16 show our preferred mixing models between the three end-members involved in the genesis of the East Scotia Ridge lavas: (1) N-MORB-source mantle forming the ambient asthenosphere (Cohen & O’Nions, 1982b); (2) OIB-source mantles (McDonough & Sun, 1995) similar to the source of Bouvet Island magmas; (3) sediment components based on South Atlantic bulk sediment (Barreiro, 1983; Ben Othman et al., 1989; Plank & Langmuir, 1998) using two different estimates of the trace element abundances of sediment melt (Class et al., 2000). We also show the composition of the Discovery mantle plume, which is situated at about 44°S, close to the Mid-Atlantic Ridge (Douglass et al., 1999), and which is a similar distance from the East Scotia Ridge to the Bouvet mantle plume.
In Figs 15 and 16, the East Scotia Ridge samples plot close to the MORB end-member and between the MORB end-member and the sediment and mantle plume compositions. This indicates that the MORB component is the most widespread magma source for the East Scotia Ridge, with additional contributions from variable amounts of the sediment and mantle plume components, consistent with interpretations of trace element compositions (Figs 11, 13 and 14). The East Scotia Ridge samples have similar 143Nd/144Nd to, but lower 87Sr/86Sr than, the South Sandwich island arc. Apart from four samples from the flanks of segment E2, all the East Scotia Ridge samples have lower 206Pb/204Pb than the South Sandwich arc. Mixing curves show that the Sr, Nd and Pb isotopic compositions of all the East Scotia Ridge samples (apart from the high 206Pb/204Pb samples from segment E2) can be modelled by addition of up to 2% sediment melt to MORB-source mantle (Figs 15 and 16). Most samples from the central segments E3–E7 can be modelled by addition of <0·4% sediment melt to MORB-source mantle. A group of samples from the flanks of segment E2 shows a clear involvement of a component having 206Pb/204Pb ratios of 19·3, believed to have migrated from the Bouvet mantle plume (Leat et al., 2000; Pearce et al., 2001). However, it should be noted that it is difficult to distinguish between mixing of sediment and mixing of plume mantle components with a MORB-source mantle using Pb isotope systematics (Fig. 16).
|
DISTRIBUTION OF MANTLE COMPONENTS
|
|---|
The East Scotia Ridge is tectonically simple, and there are
some features of the distribution of magma types within it that
appear to be systematic (Fig.
17). The central segments of the
ridge are sourced from MORB-source mantle that was influenced
only to a minor degree by components from the slab. Departures
from MORB-like compositions are strongest close to both northern
and southern ends of the ridge. The range in water contents
of the magmas, correlating with the range in trace element abundances
from MORB-like to arc-like, suggests that decompression melting
dominates melt production in the central part of the ridge,
and that volatile-fluxed melting is more important at the ends
of the ridge.
View larger version (26K):
[in this window]
[in a new window]
|
Fig. 17. Sketch of the South Sandwich subdution zone showing the distribution of components contributing to the source of individual back-arc segments. (See text for details.)
|
|
Segment E2
This segment was discussed by Leat et al. (2000), who suggested that the segment tapped both N-MORB and plume mantle sources, and that both aqueous fluid and subducted sediment components had contributed to the range of compositions. The mantle plume
TOP
ABSTRACT
INTRODUCTION
