Journal of Petrology Volume 42 Number 2 Pages 407-432 2001
© Oxford University Press 2001
Platinum-Group Elements (PGE) in Basalts of the Seaward-Dipping Reflector Sequence, SE Greenland Coast*
H. PHILIPP,
J.-D. ECKHARDT,
and
H. PUCHELT
INSTITUTE FOR PETROLOGY AND GEOCHEMISTRY, UNIVERSITY OF KARLSRUHE, KAISERSTRAßE 12, D-76128 KARLSRUHE, GERMANY
Received
June 22, 1998;
Revised typescript accepted
May 31, 2000
 |
ABSTRACT
|
|---|
The rift-related, seaward-dipping reflector sequence (SDRS)
SE of Greenland consists of basaltic lavas that exhibit variable
degrees of magmatic differentiation, derived from a heterogeneous
mantle source. Platinum-group elements (PGE) are used to provide
insights into the petrogenetic evolution of the SDRS, and to
characterize the magma sources. Noble metal concentrations correlate
well with indicators for magmatic differentiation (mg-number,
MgO), exhibiting two distinct trends. Concentrations of Ir,
Ru and Rh tend to decrease with progressive differentiation,
indicating compatible behaviour of these elements during fractional
crystallization processes. The variation of Pt and Pd shows
segmented trends. In primitive magmas, Pt and Pd are incompatible
and become enriched in the melt. The primitive magma is S undersaturated,
despite derivation from a depleted mid-ocean ridge basalt source
at a moderate degree of melting, reflecting enhanced S solubility
in the melt caused by high Fe content and elevated temperature.
In the more evolved lavas, Pt and Pd decrease with decreasing
MgO and mg-number. This indicates that S saturation had occurred
with Pt and Pd being incorporated in sulphides, which probably
segregated during ascent. Bulk partition coefficients for the
PGE during partial melting are calculated based on data from
a primitive basaltic unit with MgO
20 wt %, representing a near-primary
magma composition. The determined bulk partition coefficients
for an S-undersaturated melt are about 2 (Ir), 4 (Ru), 1·2
(Rh), 0·5 (Pt) and 0·4 (Pd). This indicates that
Ir, Ru and Rh are compatible during partial melting, whereas
Pt and Pd are incompatible.
KEY WORDS: basalts; Greenland; partial melting; PGE; trace elements
|
INTRODUCTION
|
|---|
Seaward-dipping reflector sequences (SDRSs) are characteristic
features of volcanic rifted margins. The SDRS SE of Greenland
formed during the initial break-up of the North Atlantic in
the late Paleocene–early Eocene (Duncan
et al., 1996
)
and spans the whole period from continental break-up to seafloor
spreading (Brooks & Nielsen, 1982
; Saunders
et al., 1997
).
Large volumes of lava erupted in a short period of time under
subaerial or shallow water conditions along the continental
margins of Norway and Greenland. The high magma production rates,
which led to the formation of magmatic sequences up to 6 km
thick, are inferred to require the presence of a mantle plume
(White & McKenzie, 1989
).
The aims of Ocean Drilling Program (ODP) Legs 152 and 163 were to investigate the origin and nature of the SDRS with regard to the influence of a mantle plume (Larsen et al., 1994a; Duncan et al., 1996). Nine sites were drilled in two transects distal to the ancient plume track at 63°N and 66°N (Fig. 1).
View larger version (73K):
[in this window]
[in a new window]
|
Fig. 1. Map of the North Atlantic, showing the locations of drill sites of ODP Legs 152 and 163 [modified after Larsen et al. (1994a)]. JMFZ, Jan Mayen Fracture Zone; JMR, Jan Mayen Ridge; JMR, Jan Mayen Ridge; A6, A20, A24: selected seafloor spreading magnetic anomalies.
|
|
It has been shown that platinum-group element (PGE) systematics can provide important information about the petrogenesis of basaltic magmas, such as magmatic differentiation, sulphide segregation and mantle source characteristics (e.g. Brügmann et al., 1987; Greenough & Owen, 1992; Barnes & Picard, 1993; Seitz & Keays, 1997; Vogel & Keays, 1997). Many aspects of the geochemistry of PGE remain, however, poorly understood.
In this study, we present PGE data for the entire compositional range of the SDRS, accompanied by major, trace and rare earth element analyses to elucidate the petrogenesis and mantle source characteristics of the magmas. The compositional variation and derivation of the magmas from a heterogeneous mantle source provides an opportunity to describe the PGE systematics of magmatic differentiation and the PGE characteristics of mantle-derived basalts.
|
GEOLOGICAL BACKGROUND AND SAMPLE INFORMATION
|
|---|
Detailed background information on the tectonic and geological
setting of the SE Greenland SDRS appears in the Initial Reports
of ODP Legs 152 and 163 (Larsen
et al., 1994
a; Duncan
et al.,
1996
). In summary, the drilling transects SE of Greenland are
located within an area that was tectonically undisturbed since
the early Proterozoic. The opening of the North Atlantic and
the formation of ocean crust began at
56–53 Ma as indicated
by magnetic anomaly Chron 24r (Vogt & Avery, 1974
; Talwani
& Eldholm, 1977
). The initial break-up of the North Atlantic
was accompanied by the extrusion of huge volumes of lava, resulting
in the formation of the SDRS along the continental margins of
East Greenland and Norway (Fig.
1). Estimates by Larsen &
Jakobsdóttir (1988)
, based on geophysical data, suggest
that the volcanic productivity along the rift zone during break-up
was almost three times greater than that of present-day Iceland.
Based on plate tectonic reconstructions, the Iceland plume was
most probably located beneath central Greenland at the time
of plate separation, although there are some uncertainties about
the exact position of the plume (e.g. Forsyth
et al., 1986
;
White & McKenzie, 1989
; Lawver & Müller, 1994
).
During ODP Legs 152 and 163, a total of nine sites were drilled along two transects at 63°N and 66°N distal to the ancient plume track. A detailed mineralogical and petrological description of the recovered igneous units and their stratigraphic setting has been given by Larsen et al. (1994a), Duncan et al. (1996) and Saunders et al. (1998). In the following discussion we briefly summarize the sites and the petrology of the rocks sampled for this investigation.
Site 988 is located within the northern transect at 66°N (EG 66) close to the ancient plume track (Fig. 1). Here, two igneous units were distinguished in the core recovered from the depth interval 10–32 m below seafloor (mbsf). Unit 1 is a massive and sparsely vesicular plagioclase–pyroxene–olivine-phyric basalt showing moderate alteration. Unit 2 is completely altered and was thus disregarded in our investigations.
Site 989 is located on the southern transect at 63°N (EG 63) closest to the Greenland coast. It was planned to penetrate and sample the oldest parts of the SDRS. However, recent 40Ar/39Ar determinations of the recovered lavas yield unexpected young ages of 57 ± 1·3 Ma (Tegner & Duncan, 1999). Two igneous units were identified: aphyric basalt (Unit 1) and a plagioclase–clinopyroxene–olivine-phyric basalt (Unit 2). Unit 1 is at least 69 m thick and is characterized by constant grain size, vesicularity and mesostasis content. Both units are depleted in incompatible elements (e.g. Zr, Nb, Ti), presumably reflecting derivation of the magmas from a depleted mantle source.
At Site 917, a 778 m thick volcanic sequence with 92 flow units was drilled. Based on trace element and stratigraphic observations, the sequence was subdivided into lower (LS), middle (MS) and upper series (US). The lower series consists of olivine-phyric basalts and two picrite flows (Units 61b and 62). Basalts of the middle series are more evolved and exhibit a significant contamination by continental crust. The base of the middle series is marked by a dacite flow, which was possibly generated by melting or assimilation of pelitic sediments. The middle series is separated from the lavas of the upper series by a sedimentary horizon, which marks a transition in the magma composition. The upper series consists of primitive Mg-rich tholeiites and picrites that formed thin pahoehoe flows during eruption. Oscillation of Ni content implies short-term storage of the magma in small magma chambers. On the basis of 40Ar/39Ar ages, the lavas of the lower and middle series of Site 917 were erupted between 61 and 62 Ma (Sinton & Duncan, 1998), and thus are significantly older than the time of initial seafloor spreading.
Site 990 was positioned seawards of Site 917 to obtain a stratigraphic overlap with the sequence drilled at Site 917. Thirteen igneous units were identified on the basis of phenocryst assemblages or by the presence of vesicular and/or weathered flow tops. The flow units range from moderately evolved aphyric basalts to olivine and plagioclase–olivine–clinopyroxene-phyric basalts. The section is characterized by a systematic decrease of incompatible element contents and increasing Ni and Cr concentrations from top to bottom.
The volcanic succession drilled at Site 918 comprises one sill (Unit 1) and 18 lava flows (Units 2–8). The sill is a plagioclase–pyroxene-phyric basalt and has a trace element composition similar to that of the basalts recovered at Site 988. The other units are defined as aphyric basalts with low trace element abundances. In general, the lavas are moderately evolved and geochemically similar to the basalts at Sites 989 and 990. The restricted chemical composition of the lavas suggests that the magmas were subjected to a uniform degree of differentiation.
|
ANALYTICAL METHODS
|
|---|
A total of 98 whole-rock samples from the SDRS were analysed
for major and trace, rare earth and platinum-group elements.
Major elements were determined by wavelength-dispersive X-ray
fluorescence analysis (Siemens SRS 303 AS) on fused pellets.
Accuracy of the measurement is within 2% relative for all major
elements, based on analyses of the geological reference materials
BE-N and BIR (Govindaraju, 1994
). For drift-correction, BE-N
was run as a reference after every five measurements. Trace
elements (Ni, Zn, Ga, Cr, Rb, Sr, Y, Zr, Nb, V) were analysed
by wavelength-dispersive X-ray fluorescence analysis (Siemens
SRS 303 AS) on pressed pellets. Analyses are accurate to 5%
relative, except for Nb, which is <10%. Analytical accuracy
and precision were determined in the same way as for major elements.
The elements Cu and Ba were determined by energy-dispersive
XRF (Tracor Spectrace 5000) on the same pressed pellets used
for wavelength-dispersive XRF. Detection limits are 4·5
ppm (Cu) and 2·9 ppm (Ba) at measuring times of 1000
s/spectrum (Kramar, 1997
). Accuracy was checked by repeated
analysis of the international standard BE-N and is within 8%
for Cu and 2% for Ba. The trace elements Co, Sc, Hf, Ta and
Th were determined by instrumental neutron activation analysis
(INAA) following the method of Kramar & Puchelt (1982)
.
Sample and standard capsules were irradiated simultaneously
in the TRIGA reactor at Heidelberg (Germany) at a thermal neutron
flux of 8
x 10
12 n/cm
2 per s for 5 h. Two detector systems were
used for measurement: a coaxial Ge(Li)-detector for the high-energy
range (150–2000 keV) and a planar high-purity Ge-detector
for the energy range of 30–210 keV. Rare earth elements
(REE) were analysed by ICP-MS (Fisons PQ2) after HClO
4–HF
digestion. Precision of all ICP-MS analyses is within 2–6%
based on replicate measurements of working standards and an
international reference material (BR). Platinum-group elements
were determined by ICP-MS (Fisons PQ2) after NiS fire assay
preconcentration of the noble metals from 25–35 g of dry
powder. Average method detection limits (3
above background),
calculated for 30 g of rock powder are 0·08 ppb Ir, 0·09
ppb Ru, 0·04 ppb Rh, 0·40 ppb Pt and 0·43
ppb Pd. Replicate analyses of the international reference material
WPR-1 (Govindaraju, 1994
), in-house standard ‘147-898D-10W’
(Puchelt
et al., 1996
) and selected samples are reported in
Table
1.
View this table:
[in this window]
[in a new window]
|
Table 1: Results of replicate analyses of selected samples from Site 990, international standard WPR-1 (Govindaraju, 1994) and in-house standard (Puchelt et al., 1996)
|
|
|
RESULTS
|
|---|
Major elements
Whole-rock chemical compositions, recalculated on volatile-free
basis, are listed in the Appendix (Table
A1). The reported values
represent the average composition for each of the investigated
igneous units. The complete dataset can be accessed as an Excel
spreadsheet on the
Journal of Petrology Web site at
http://www.petrology.oupjournals.org.
View this table:
[in this window]
[in a new window]
|
Table A1: Average major element oxides content (wt %) of investigated units, recalculated on a volatile-free basis
|
|
The samples are classified as tholeiitic basalts on the basis
of their SiO
2–(Na
2O + K
2O) variations. The lavas exhibit
different degrees of differentiation as illustrated by the
mg-number
[defined as
mg-number = 100MgO/(MgO + FeO); molecular oxide
proportions and Fe
2O
3/FeO = 0·15]. Samples from Site
917 show a wide range in
mg-number, from 34 in the most evolved
lava to
80 in some picritic units. In contrast, samples from
Sites 988, 989, 918 and 990 span a narrow range in
mg-number
(46–65).
Figure 2 illustrates the variation of selected major elements with MgO; SiO2, TiO2, Na2O and K2O concentrations increase with decreasing MgO. Samples from Sites 988, 989, 918 and 990 exhibit a narrow range of compositions and always show lower concentrations than samples from Site 917 at corresponding MgO values. Fe2O3(total) contents (not shown) range between 10 and 15 wt % and display no systematic change with MgO. However, in samples from Site 917, Fe2O3(total) contents decrease from primitive to evolved lavas. Al2O3 contents are 15 wt %, decreasing towards higher MgO concentrations in samples from the upper series of Site 917. The content of CaO gradually increases with decreasing MgO, but decreases below 7 wt % MgO. Below this MgO content, CaO concentrations decrease with decreasing MgO. Samples from Sites 988, 989, 918 and 990 lie on the maximum of the MgO–CaO trend, defined by Site 917 samples.
Trace elements
Trace element concentrations are listed in the Appendix (Table A2). Compatible element concentrations (Ni, Cr and Co) show a strong correlation with MgO (Fig. 3a). Concentrations range from 1030 ppm (Ni) and 1440 ppm (Cr) in a picritic unit of the upper series of Site 917, to 33 ppm for Ni and 23 ppm for Cr in the most evolved samples. The Site 917 basalts feature highly variable compatible trace element abundances, whereas basalts from Sites 918, 988, 989 and 990 are characterized by a much narrower range in concentrations. Incompatible element concentrations show a systematic increase with decreasing MgO. Different trends for incompatible elements are apparent for Site 917 basalts and the basalts from Sites 918, 989 and 990, as illustrated in Fig. 3b. For example, Zr shows a much stronger enrichment in Site 917 samples compared with the other sites. In contrast, the Nb data display an enrichment trend that is defined by samples from Sites 918, 989 and 990. Only few samples from the lower series of Site 917, Site 918 (Unit 1) and Site 988 show a stronger enrichment of Nb. The scatter of Zr and Nb concentrations at MgO contents between 12 and 15 wt % probably reflects either variable degrees of partial melting of a single source or derivation from a heterogeneous mantle source. The variation of Sc with MgO displays enrichment of Sc with decreasing MgO content, whereas below 7–9 wt % MgO, Sc contents decrease with decreasing MgO. Lavas from Sites 989 and 918 and some lava flows from Site 990 exhibit particularly high Sc concentrations (45–50 ppm) similar to some basalts from the Vøring Plateau (Viereck et al., 1989) and the Rockall Plateau (Joron et al., 1984).
View larger version (22K):
[in this window]
[in a new window]
|
Fig. 3. Variations of (a) Ni and Cr and (b) Zr, Nb, Rb and Sc with MgO. Data are given in the Appendix (Table A2).
|
|
Multi-element diagrams, normalized to normal mid-ocean ridge
basalt (N-MORB) (Sun & McDonough, 1989
) emphasize the distinct
chemical compositions of the basalts from the various sites.
Samples from Sites 988 and 918 (Unit 1) show a clear enrichment
in incompatible elements with high Nb contents (17·2–29·0
ppm) and low Zr/Nb ratios (
6–8), similar to ocean island
basalts (OIB) with average Zr/Nb
5·8 (Sun & McDonough,
1989
; Fig. 4a). In contrast, samples of Sites 918 (Units 8b,
11b, 12b, 14), 989 and 990 have a signature similar to N-MORB
with Zr/Nb
12–22. However, the samples from Sites 989
and 990 show a remarkable enrichment of Rb, Ba and K (Fig. 4a).
Multi-element diagrams characterize and summarize the average
incompatible element distribution for the three igneous series
of Site 917 (Fig. 4b and c). Samples from the upper series of
Site 917 display element abundances close to N-MORB, except
for Rb, Ba and K, which show enrichment up to
10
x N-MORB (Fig.
4b). In contrast, all lavas from the middle and lower series
of Site 917 are characterized by a strong enrichment in the
large ion lithophile elements (LILE) and light rare earth elements
(LREE). Additionally, samples from Site 917 (middle series)
are characterized by a pronounced negative Nb–Ta anomaly.
Within the lower series of Site 917, five units (Units 60, 61b,
62, 68, 70) are distinct with regard to their Zr/Nb ratio, which
is <15, although unit 70 (Zr/Nb
15) seems to be similar to
the high Zr/Nb subgroup, i.e. showing high La and Ce abundances
and an Nb–Ta depletion.
View larger version (25K):
[in this window]
[in a new window]
|
Fig. 4. N-MORB normalized trace element patterns for samples of (a) Sites 918, 988, 989 and 990, (b) Site 917 (Middle and Upper Series) and (c) Site 917 (Lower Series), with normalization values after Sun & McDonough (1989).
|
|
Rare earth elements
The REE abundances in the investigated igneous units are listed
in the Appendix (Table
A3). Chondrite-normalized (Boynton, 1984
)
REE patterns also emphasize the distinct geochemical composition
of basalts from Sites 918, 989 and 990 and those from Sites
988 and 918 (Unit 1; Fig. 5a). Samples from Sites 988 and 918
(Unit 1) are characterized by high LREE contents (La 52
x C1
and 68
x C1), LREE enrichment ([La/Yb]
C1 3·2) and high
heavy rare earth element (HREE) abundances (Yb
17
x C1). REE
patterns of samples from Sites 918, 989 and 990 are similar
to those of N-MORB with low LREE abundances and relatively unfractionated
patterns between middle REE (MREE) and HREE ([Sm/Yb]
C1 0·8
± 0·12). Figure 5b shows the REE patterns of Site
917 samples, grouped into middle and upper series. The primitive
lavas of the upper series of Site 917 cover a wide range of
chondrite-normalized REE contents (e.g. La 5·7–25·7;
Eu 8·9–31·2; Yb 5·0–18·1)
but are characterized by similar La/Yb ratios ([La/Yb]
C1 1·1
± 0·2). In contrast, lava flows from the middle
series from Site 917 exhibit much higher chondrite-normalized
LREE abundances (La
42–110) and a strong LREE enrichment
([La/Yb]
C1 3·5–14·2).
View larger version (24K):
[in this window]
[in a new window]
|
Fig. 5. Chondrite-normalized REE patterns for (a) Sites 918, 988, 989 and 990, (b) Site 917 (Middle and Upper Series) and (c) Site 917 (Lower Series) with normalizing values after Boynton (1984).
|
|
Figure 5c shows the chondrite-normalized REE patterns for samples
from the lower series from Site 917, distinguished by their
Zr/Nb ratio. The REE abundances for lava flows with Zr/Nb >
22 cover a wide range, e.g. (18–253)
x C1 for La, accompanied
by strong LREE enrichment ([La/Yb]
C1 3·2 ± 1·1).
By contrast, lava flows (units 60, 61b, 62 and 68) with Zr/Nb
10·6–13·5 are characterized by lower LREE
contents and slight LREE enrichment ([La/Yb]
C1 2·7 ±
0·8). The lava sample from unit 70 (Zr/Nb
15) exhibits
more similarity to the high Zr/Nb sub-group with La
184
x C1
and [La/Yb]
C1 9·7. All samples from Site 917 are characterized
by a negative slope between the MREE and the HREE ([Sm/Yb]
C1 1·81 ± 0·44).
Platinum-group elements
The average PGE concentrations and Pd/Ir, Pt/Ir and Pt/Pd ratios of the igneous units are listed in Table 2. In comparison with previously reported values of MORB (Crocket & Teruta, 1977; Hamlyn et al., 1985; Plessen & Erzinger, 1997) and OIB (Fryer & Greenough, 1992) from other locations, the basalts of this SDRS have higher noble metal abundances (e.g. Ir 
2·73 ppb; Pt 14·3 ppb; Pd 16·5 ppb). These contents are, however, similar to PGE abundances reported for tholeiites and picrites from the Skaergaard intrusion (Nielson & Brooks, 1995). The picritic units of the upper series of Site 917 have the highest Ir, Ru and Rh concentrations, whereas the middle and lower series of Site 917 display significantly lower abundances. The elements Ir, Ru and Rh show a systematic decrease in concentrations with progressive magmatic differentiation (Ir 2·73–0·10 ppb; Ru 1·73–0·10 ppb; Rh 0·89–0·05 ppb) indicating compatible behaviour of these elements during magma evolution (Fig. 6).
Bivariate plots of MgO vs Pt and Pd (Fig. 7) illustrate the geochemical behaviour of these elements. In the primitive picritic and olivine-phyric basalts from the upper series of Site 917, Pt and Pd behave incompatibly as the magma evolved. Pt and Pd contents increase to 15 ppb in the lavas from the upper series of Site 917, whereas in lower and middle series samples from Site 917, Pt and Pd concentrations generally do not exceed 5 ppb. In samples from Sites 918, 988, 989 and 990, Pt and Pd contents exhibit a strong decline within a narrow range of MgO contents. Pd and Pt concentrations decrease from 10–15 ppb at 9 wt % MgO to 5 ppb at 6 wt % MgO. Most samples are characterized by a uniform Pt/Pd ratio of 0·89. It should be noted that most samples from Site 917 (middle series) show significantly lower Pt/Pd values, whereas samples from Sites 988 and 918 (Unit 1), and some samples from the lower series of Site 917 display higher values (Fig. 8).
The samples are characterized by distinct primitive mantle-normalized siderophile element patterns (Barnes et al., 1988). Samples from Site 988 and Unit 1 from Site 918 have parallel patterns with nearly unfractionated Ir/Ru (0·9) and Pt/Pd (0·8) ratios and a strong enrichment in copper (Fig. 9a). Samples from Sites 918 (Units 8b, 11b, 12b, 14) and 989 show a remarkable similarity and plot within the field defined by samples from Site 990 (Fig. 9b). Samples from these three Sites (918, 989, 990) are characterized by a slight Ru anomaly and a marked fractionation between Ir, Ru and Rh on the one hand and Pt and Pd on the other. Primitive mantle normalized patterns for samples from Site 917 are illustrated in Fig. 9c. Site 917 (upper series) samples display flat patterns with a slight fractionation between Ir, Ru, Rh and Pt, Pd. Additionally, the lavas from the upper series of Site 917 feature only minor fractionation between Ni and Ir ([Ni/Ir]PM 0·9), Ni and Ru ([Ni/Ru]PM 1·1) and between Cu and Pd ([Cu/Pd]PM 1·2). The patterns for the upper series of Site 917 are similar to patterns for komatiites from Gorgona Island, Colombia (Brügmann et al., 1987) and Kambalda, West Australia (Keays, 1982). In contrast, samples from the middle series from Site 917 are characterized by lower PGE and Ni abundances, except for Cu, which is close to primitive mantle values. Figure 9d illustrates the primitive mantle-normalized siderophile element patterns for Site 917 (lower series) samples, distinguished by their Zr/Nb ratios. Platinum-group elements, Ni and Cu show nearly the same abundances as in samples from the middle and upper series of Site 917, whereas samples with low Zr/Nb ratios are characterized by a negative Ru anomaly. These samples show a wide range in Rh and Pt abundances, whereas Pd abundances are rather similar in these samples.
View larger version (38K):
[in this window]
[in a new window]
|
Fig. 9. Mantle-normalized (PM) pattern for PGE, Ni, and Cu. (a) Sites 988 and 918 (Unit 1); (b) Sites 989, 990 and 918 (Units 8b, 11b, 12b, 14); (c) Site 917 (Middle and Upper Series); (d) Site 917 (Lower Series). Normalization values after Barnes et al. (1988) are 2000 ppm Ni, 4·4 ppb Ir, 5·6 ppb Ru, 1·6 ppb Rh, 8·3 ppb Pt, 4·4 ppb Pd and 28 ppm Cu.
|
|
|
DISCUSSION
|
|---|
Major and trace elements
The volcanic succession drilled at the southern transect EG
63 represents the transition from continental flood basalt volcanism
to oceanic volcanism (Fitton
et al., 1998
b; Fram
et al.,1998
).
Figure
10 shows the variation of MgO content with stratigraphic
position. The earliest lavas (Site 917, Lower and Middle Series)
comprise a pre-break-up continental succession (Fitton
et al.,
1998
b) ranging from high-Mg basalts (MgO
15 wt %) to evolved
basalts with MgO
5–6 wt %. Fitton
et al. (1998
b)
have
attributed this feature to magma evolution in crustal magma
chambers with decreasing supply of primitive melts, attended
by an increasing influence of crustal contamination. By contrast,
the lava units from the upper series of Site 917 are thought
to represent break-up related magmatism (Larsen
et al., 1994
b)
and are characterized by high-Mg basalts and the absence of
evolved basalts. The rapid oscillation in Ni content with stratigraphic
height in the upper series implies short-term storage in smaller
reservoirs accompanied by crystallization and accumulation of
olivine phenocrysts (Larsen
et al., 1994
a; Fitton
et al., 1998
b).
Lavas cored at Site 990 are geochemically similar to the igneous
units 8b, 11b, 12b and 14 recovered at Site 918, which are interpreted
as an oceanic succession (Larsen
et al., 1994
b; Fitton
et al.,
1998
b). These lavas are of restricted geochemical composition
(MgO
7–9 wt %), suggesting that the magmas were derived
from re-established permanent magma reservoirs associated with
a spreading axis, which prevented the eruption of primary magmas
(Larsen
et al., 1994
b). The inferred stratigraphic position
of the uppermost units recovered at Site 917 and the depth of
penetration at Site 990 (342 mbsf) define the transition from
continental to oceanic succession to less than
100 m thickness
(Duncan
et al., 1996
). According to seismic data (Duncan
et al., 1996
), the lavas drilled at Site 989 should be located
stratigraphically below the lower series from Site 917 and should
have been erupted before the 61–60 Ma old lower and middle
series lavas from Site 917. However, the geochemical composition
of the samples from Site 989 compares more closely with the
oceanic volcanism, represented by the lavas from Sites 918 and
990. This is in agreement with the
40Ar/
39Ar age of 57·1
± 1·3 Ma reported by Tegner & Duncan (1999)
for lavas from Site 989.
View larger version (14K):
[in this window]
[in a new window]
|
Fig. 10. Stratigraphic variation of MgO in the basaltic sequences of the SDRS SE of Greenland. (Note that the vertical axis reflects relative stratigraphic position and is not a continuous depth profile.)
|
|
The variation of CaO/Al2O3 ratio with MgO for the lavas from the SDRS (Fig. 11) is particularly useful in revealing the fractionation assemblage because of the sensitivity of CaO/Al2O3 to the proportions of olivine, clinopyroxene and plagioclase (Fram & Lesher, 1997) and suggests that many compositional variations within the lavas from the SDRS are due to crystal fractionation processes. Fractionation of olivine decreases MgO at constant CaO/Al2O3 ratio, whereas plagioclase fractionation increases both MgO and CaO/Al2O3. By contrast, removal of clinopyroxene from the melt results in decreasing MgO and CaO/Al2O3 ratio. Samples from the upper series from Site 917 with MgO > 10 wt % display uniform CaO/Al2O3 (0·65 ± 0·06) implying that fractionation–accumulation of olivine has occurred. Liquid lines of descent (LLD), defined by the results of low-pressure melting experiments on two samples from the lower series (86R7) and upper series (11R4) from Site 917 show two different trends (Thy et al., 1998). Plagioclase crystallization is documented by the increase of the CaO/Al2O3 ratio at MgO 9 wt %. When augite appears in the fractionation assemblage at 7 wt % MgO, the slopes change. The LLD for sample 87R7 shows a strong decline, reflecting crystallization of augite with high CaO/Al2O3 ratio (7·7–9·6; Thy et al., 1998), whereas the trend for sample 11R4 evolves towards moderate increase of CaO/Al2O3 and implies crystallization of augite with significant lower CaO/Al2O3 ratio (5·1–5·8; Thy et al., 1998). The crystallization sequence of olivine, plagioclase and augite for a primitive aphyric lava (9·1 wt % MgO) from the lower series of Site 917 (87R7) was defined by the low-pressure melting experiments of Thy et al. (1998). For sample 11R4 with 10·7 wt % MgO they obtained a crystallization order with olivine and plagioclase appearing together at 1228 ± 5°C, followed by augite at 1182 ± 5°C (Thy et al., 1998). The same conclusion on the fractionation assemblage can be drawn by the variation of Sc with MgO (Fig. 3b). The concentration of Sc during magma differentiation is sensitive to the fractionation assemblage, because Sc is moderately incompatible in olivine (D = 0·1–0·3; Beattie et al., 1991), highly incompatible in plagioclase (D = 0·02; Sun & Nesbitt, 1979) and compatible in clinopyroxene (D = 2–5; Gallahan & Nielsen, 1992). The gradual increase of Sc with decreasing MgO between 25 and 7 wt % can be ascribed to the fractionation–accumulation of olivine ( ± plagioclase), whereas the decline at MgO 7 wt % is attributed to the fractionation of clinopyroxene together with olivine ( ± plagioclase). Unusually high Sc concentrations in the samples from Sites 918, 989 and 990 cannot be due to only the fractionation of olivine and plagioclase. Larsen et al. (1998) have proposed a two-stage mantle melting process to explain the petrogenesis of the magma. First, a high-pressure melt fraction has to be removed, with Sc remaining in garnet (D = 2·27; Ulmer, 1989) in the residue. Second, this residue, enriched in Sc, melts at lower pressure in the spinel stability field, followed by fractionation of 20–40% of olivine + plagioclase (Fitton et al., 1998b).
View larger version (18K):
[in this window]
[in a new window]
|
Fig. 11. Plot of MgO vs CaO/Al2O3 ratio. (Note that samples with LOI > 6 wt % have been omitted.) Liquid lines of descent (LLD) are after Thy et al. (1998)
|
|
On the basis of the Zr/Y vs Nb/Y variation diagram (Fig. 12) introduced by Fitton et al. (1997), different magma sources can be distinguished, which have been involved in the formation of the SDRS: an N-MORB source and the heterogeneous Iceland plume, comprising both a depleted and an enriched component (Hemond et al., 1993; Thirlwall et al., 1994). Fitton et al. (1997) have shown that samples from the Iceland neovolcanic zone plot within the field defined by
and
whereas N-MORB data (Hofmann, 1988
; Sun & McDonough,
1989
) plot below the lower boundary. The earliest lavas of the
SDRS (Site 917 lower and middle series) also plot below the
lower boundary and are therefore assumed be derived from an
N-MORB source (Fitton
et al., 1997
, 1998
b). Samples from the
middle series and some samples from the lower series from Site
917 plot on a trend towards the composition of Archaean gneiss
(Blichert-Toft
et al., 1995
; Fitton
et al., 1998
b) implying
assimilation of continental crust. Fitton
et al. (1998
a)
have
recognized two distinct crustal contaminants: (1) lower-crustal
basic granulite affected the earliest lavas, represented by
the lower series from Site 917; (2) upper-crustal amphibolite-facies
gneiss contaminated the middle series lavas from Site 917. They
attributed this feature to storage of magmas at progressively
shallower levels in the crust associated with progressive lithospheric
extension. It is remarkable that some samples from the lower
series of Site 917 (Units 60, 61b, 62 and 68; Zr/Nb = 10·6–13·5)
plot within the enriched array of the Iceland neovolcanic field.
This implies that they were derived either from an enriched
plume source, or from a depleted plume source by lower degree
of partial melting as suggested by Larsen & Saunders (1998)
.
In Fig.
12 samples from the break-up related upper series of
Site 917 cluster around the N-MORB field and exhibit REE patterns
typical for N-MORB. Based on unusually high Zr/Nb ratios (
24–60)
in the lavas from Site 917 (upper series), Larsen
et al. (1998)
concluded that these melts must have been produced from a mantle
that was more depleted than normal MORB mantle. Samples from
the post-break-up lavas from Sites 918 and 990 and samples from
Site 989 plot in the depleted part of the Iceland array and
are thus inferred to be derived from the depleted Iceland plume
source, despite being LREE depleted and otherwise similar to
N-MORB. Samples from Sites 988 and the sill from Site 918 (Unit
1) plot in the enriched array of the field defined by the Iceland
neovolcanic zone and are characterized by LREE enrichment.
View larger version (27K):
[in this window]
[in a new window]
|
Fig. 12. Nb/Y vs Zr/Y discrimination diagram highlights the different magma sources of basalts from the SDRS SE of Greenland. References for N-MORB, Archaean gneiss and Iceland neovolcanic zone are given in the text.
|
|
Platinum-group elements
Element mobility during alteration and hydrothermal processes can cause problems for petrogenetic studies, especially when trace element ratios are used to determine fractionation trends and source characteristics. Crocket & Teruta (1977), Barnes et al. (1985) and Crocket (1990) have shown that PGE can be mobilized under metasomatic and hydrothermal conditions. The PGE concentrations of the SDRS samples, when plotted against indicators of alteration (e.g. LOI, H2O+), however, show no changes in PGE concentrations as a result of alteration processes.
Crystal fractionation
The compatible behaviour of the elements Ir and Ru during crystal fractionation is well documented (e.g. Barnes et al., 1985; Brügmann et al., 1987; Barnes & Picard, 1993; Zhou, 1994; Wyman et al., 1995), whereas investigations describing the compatible behaviour of Rh are extremely rare in the literature (Wyman et al., 1995; Momme et al., 1999). It remains a subject of debate which phase is responsible for concentrating Ir, Ru and Rh. Barnes et al. (1985) rejected substitution of Ir and Ru in olivine and chromite during crystal fractionation as a mechanism for fractionating the PGE. On the other hand, on the basis of their investigations of komatiite suites from Alexo (Ontario) and Gorgona Island (Colombia), Brügmann et al. (1987) suggested that olivine dominantly incorporates Ir and Ru with a partition coefficient (olivine/liquid) of 1·8 ± 0·6 for Ir and 1·6 ± 0·6 for Ru. Alternatively, Keays (1982) suggested that Ir is precipitated in an early stage of magma evolution as an Ir–Os alloy, possibly acting as a nucleus for olivine crystallization. This is in accordance with the experimental results of Amossé & Alibert (1993) and Amossé et al. (1997) on the solubility of Ir in silicate melts. In contrast, O’Neill et al. (1995) have suggested that Ir–Os alloys cannot precipitate from silicate melts, as a result of the high solubility of Ir at oxygen fugacities (fO2) appropriate for terrestrial basalts. Additionally, they concluded that the presence of Ir as Ir2+ in silicate melts at geologically relevant temperatures and oxygen fugacity raises the possibility for substitution of Ir for Mg in ferromagnesian silicates. In their work on the partitioning of Ru, Rh and Pd between spinel and silicate melt, Capobianco & Drake (1990) and Capobianco et al. (1994) have shown that Ru and Rh are highly compatible in spinel, with spinel–melt partition coefficients of 20–4000 (Ru) and 90–370 (Rh). In contrast, Pd appears to be incompatible in spinel, with Dspinel/liquid = 0·02–0·7. This observation contrasts sharply with the general classification given by Barnes et al. (1985), who assigned Rh to the Pd subgroup (PPGE, consisting of Rh, Pd and Pt) of the PGE. Although the experiments were carried out at high oxygen fugacity, Capobianco & Drake (1990) and Capobianco et al. (1994) concluded that spinel has the ability to fractionate Ru and Rh from Pd. This was confirmed by Zhou et al. (1998), who have shown that high-Cr and high-Al chromitites from orogenic peridotites are enriched in Ir, Ru and Rh relative to Pt and Pd. Amossé et al. (1997) suggested the influence of oxygen fugacity on the geochemical behaviour of Rh. They have shown that at fO2 < 10-7 bar, Rh is stable as Rh2+ and behaves like Pd, whereas at fO2 > 10-7 bar Rh becomes trivalent and thus can be incorporated in the spinel lattice.
Our data show that Ir, Ru and Rh display compatible behaviour during magma evolution, exhibiting good correlations with MgO, Ni and Cr (Fig. 13). The conclusion that these three elements are compatible in olivine, however, is inconsistent with a number of petrological and geochemical studies showing that olivine did not contain significant amounts of Ir, Ru and Rh and is therefore not responsible for the fractionation of these three elements (e.g. Mitchell & Keays, 1981; Gueddari et al., 1996; Lorand et al., 1999). Therefore it is suggested that olivine fractionation is not responsible for the compatible behaviour of Ir, Ru and Rh in the lavas from the SDRS. The good correlation of Ir, Ru and Rh with Cr further indicates that spinel could also be responsible for the compatibility of these three elements. However, Demant (1998) has shown that the chemical composition of the spinels from Site 917 is in the range of MORB, crystallizing at temperature of 1200°C and an oxygen fugacity below the nickel–nickel oxide oxygen buffer. This indicates that the oxygen fugacity was too low to form Rh3+, such that the compatible behaviour of Rh according to Amossé et al. (1997) cannot be attributed to the fractionation of spinel.
A number of investigations have shown that Ir, Ru and Rh are not incorporated in the lattice of spinel, but are concentrated in trace inclusions such as sulphides or alloys (e.g. Stockmann & Hlava, 1984; Walker et al., 1996), implying that spinel itself has no bearing on the fractionation of these three elements. For example, Barnes et al. (1985) and Peck & Keays (1990a) suggested that Ir and Ru were partitioned into early formed alloys during magma fractionation (evolution). The positive correlation of Ir with Ru and Rh over the entire compositional range indicates that all three elements are probably hosted by the same phase. On the basis of these results, it seems probable that Ir, Ru and Rh in the SDRS lavas were fractionated through crystallization of Rh-bearing Ir–Ru alloys.
The geochemical behaviour of Pd and Pt during magma differentiation is controlled by their abilities to dissolve in silicate, oxide, metallic and sulphide phases. Keays (1995) pointed out the importance of the sulphur (S) content of a melt for the partitioning behaviour of Pt and Pd during magma differentiation. In S-undersaturated melts Pd and Pt behave as incompatible elements, becoming enriched during magma evolution. In S-saturated melts, Pd and Pt are preferentially partitioned into sulphides as a result of their high partition coefficients (D[Pt,Pd]sulphide-silicate 104; Campbell et al., 1983; Peach et al., 1990, 1994). Gravity settling of the sulphides thus results in a depletion of Pt and Pd in the residual silicate magma. Figure 7 demonstrates that Pd and Pt behave incompatibly in samples from the upper series of Site 917, indicating that this magma was S undersaturated. In Site 918, 988, 989 and 990 samples, Pt and Pd correlate positively with MgO and exhibit considerably high Pt and Pd contents, whereas samples from the lower and middle series of Site 917 display no systematic MgO–Pd and MgO–Pt variation, accompanied by lower Pt and Pd abundances. Barnes & Picard (1988) and Vogel & Keays (1997) have shown that the Cu/Pd ratio is an useful tool to test the S saturation of a magma. In S-undersaturated melts, the Cu/Pd ratio remains relatively constant during crystal fractionation, because Cu and Pd both behave as incompatible elements in S-undersaturated systems (Barnes & Picard, 1993). In S-saturated melts, the Cu/Pd ratio tends to greater values because of the significantly higher sulphide–silicate partition coefficient for Pd (D 104; Campbell et al., 1983; Peach et al., 1990, 1994) relative to Cu (D 245–1383; Rajamani & Naldrett, 1978; Peach et al., 1990). Samples from the upper series of Site 917 are characterized by uniform Cu/Pd ratios (8·2 x 103) showing minor variation of Cu/Pd over the entire range of MgO. Only the most evolved basalt sample (unit 33; MgO 8·3 wt %) shows a higher Cu/Pd ratio (13 x 103), implying that S saturation has occurred. Samples from Sites 918, 988, 989 and 990 exhibit a systematic variation of Cu/Pd ratio with decreasing MgO content (Fig. 14). At an MgO content of 8 wt %, Cu/Pd ratios increase, indicating that in these lavas S saturation had occurred, leading to sulphide liquid unmixing and thus a decrease in the Pd and Pt concentrations. In contrast, samples from the lower and middle series of Site 917 are, in general, characterized by significantly higher Cu/Pd ratios, implying that in these lavas S saturation and thus segregation of sulphides had occurred.
View larger version (12K):
[in this window]
[in a new window]
|
Fig. 14. Plot of Cu/Pd vs MgO, showing the variation of Cu and Pd as a result of sulphide precipitation.
|
|
Our data show that Pt is less enriched than Pd in samples from the upper series of Site 917, indicating that Pt is less incompatible during magma evolution. This suggests that Pt is slightly compatible in silicate or oxide phases during differentiation of S-undersaturated magmas. Another aspect might be the fractionation of Fe–Pt alloys, coprecipitating with Ir–Ru–Os alloys, as shown by experiments of Fleet et al. (1991) and Borisov & Palme (1997). These observations agree with data for chromites separated from cumulates from the Heazlewood River ultramafic complex, Tasmania (Australia; Peck & Keays, 1990b), in which Pt abundances were found to be enhanced relative to Pd within the chromite-bearing cumulates.
Partial melting and magma source characteristics
The incompatible behaviour of Pt and Pd indicates that the magmas from Site 917 (upper series) saturated at the time of separation from their source. Two mechanisms are proposed by which S-undersaturated melts can be produced. Hamlyn et al. (1985) and Hamlyn & Keays (1986) proposed that magmas generated by low to moderate degrees of partial melting of undepleted mantle are S saturated, leaving PGE-rich sulphides behind in the residue. Further melting of this depleted source leads to dissolution of PGE-rich sulphide components and produces melts that are enriched in Pt and Pd, but impoverished in S, certain incompatible lithophile elements (e.g. Ti, HREE) and moderately compatible chalcophile elements (e.g. Cu). Alternatively, S-undersaturated melts can also be produced by high degrees (>25%) of partial melting of peridotite as shown for komatiitic and picritic magmas (Keays, 1995). In contrast, for the postulated degree of partial melting for the generation of primary MORB magmas (8–20%; Klein & Langmuir, 1987), the S content in the upper mantle (200 ppm; Morgan, 1986) and the S solubility in basaltic melts (800 ppm at 9 wt % FeO; Haughton et al., 1974), it is probable that primary MORB melts are S saturated (Peach et al., 1990).
The S-undersaturated lavas from the upper series of Site 917 are inferred to be derived from a MORB mantle at degrees of partial melting between 12 and 15% (Fitton et al., 1998b; Fram et al., 1998). Haughton et al. (1974) have shown that the S solubility of a silicate melt is strongly dependent on its FeO content. The S capacity of a silicate melt increases from 800 ppm at 9 wt % FeO to 1100–1200 ppm at FeO 11–12 wt % at 1200°C. Further, there is a strong increase in the S solubility with rising temperature of the melt (Wendlandt, 1982). Thy et al. (1998) have calculated primary melt compositions for upper series lavas from Site 917 with FeOtotal 11 wt % and estimated 1-atm liquidus temperatures of 1380°C. Although these temperatures are not direct estimates of mantle temperature, they suggest potential mantle temperatures 100–150°C above those of the MORB source (Fram et al., 1998). Therefore we suggest that the S solubility in the magmas from Site 917 (upper series) was significantly enhanced to produce S-undersaturated magmas even at moderate degrees of partial melting.
Demant (1998) reported olivine compositions of Fo90–91 in lavas from units 14 and 16 (Site 917, upper series), suggesting that these lavas represent unfractionated mantle melts. These samples, therefore, allow us to investigate the geochemical behaviour of PGE during partial melting and the production of S-undersaturated magmas. Because partition coefficients of PGE for individual minerals are still missing, we calculate the bulk distribution coefficients using the measured or most probable values for the parameters as discussed in the following paragraph. In consequence, our calculated D values are correct only under these conditions and neglect partition of PGE into different minerals. Bulk distribution coefficients for the PGE are calculated using the modal batch melting equation
where
Di is the bulk partition coefficient of element
i,
Cs,i is the
concentration of element
i in the source,
Cl,i is the concentration
of element
i in the melt and
F is the degree of partial melting.
The concentrations of the PGE in the melt were calculated from
our data, assuming that unit 16 from Site 917 represents a primary
magma composition. In the absence of any direct data for PGE
in the MORB source, we used in a first approach an initial source
concentration for PGE based on the least depleted Pyrenean orogenic
spinel lherzolites (TUR 7, DES 7; Pattou
et al., 1996
) and slightly
depleted spinel lherzolites from Kilbourne Hole, New Mexico
(Morgan, 1986
). The degree of partial melting is assumed to
be
15% (Fitton
et al., 1998
b).
The calculated partition coefficients (Table 3) are in accordance with the results of Barnes & Picard (1993) in that they emphasize the distinct geochemical character of Ir and Ru on the one hand and Pt and Pd on the other. Ir and Ru show bulk partition coefficients of 1·7–2·3 and 3·8–4·3, respectively. Furthermore, Rh shows compatible behaviour with a calculated partition coefficient of D 1·2. The partition coefficient calculated for Pt (0·47) is slightly higher than the D value for Pd (0·38). This implies that Pt is more compatible than Pd during partial melting processes and must therefore, in the absence of sulphides, be slightly incorporated in silicate and/or oxide phases.
View this table:
[in this window]
[in a new window]
|
Table 3: Calculated bulk partition coefficients for different PGE concentrations from different spinel lherzolites at 15% partial melting
|
|
In contrast, samples from Sites 989, 990 and 918 are thought to be derived from a depleted plume source, despite being LREE depleted and otherwise resembling typical N-MORB. The depletion of trace elements is explained by the segregation of small amounts of melt, enriched in incompatible elements and LREE (Elliott et al., 1991). Thus, partial melting of the remaining residue results in the generation of an incompatible element depleted magma. This is supported by the lower TiO2 concentrations (0·98 wt %) and low abundances of incompatible trace elements (Zr 61 ppm) relative to average OIB (Sun & McDonough, 1989). This implies that the residue became successively enriched in Pd and Pt if the degree of partial melting was insufficient to dissolve all sulphides and completely deplete the residue in sulphur (Hamlyn et al., 1985; Keays, 1995). This gives rise to partial melts that are S undersaturated at even lower degrees of partial melting. The variation of Pt and Pd with MgO implies that the magma was S saturated at the time of crystallization. However, we suggest that the parental magma that formed the lavas at Sites 918, 989 and 990 was initially S undersaturated because of the moderate to high Pt and Pd contents, accompanied by low Cu/Pd ratios [(8·4–10·4) x 103] in the most primitive samples. Thus, the most primitive lavas in this oceanic succession represent the onset of S saturation.
|
CONCLUSIONS
|
|---|
Basalts from different units of the SDRS are assumed to be derived
from heterogeneous mantle sources, as illustrated by distinct
incompatible trace element ratios (e.g. Zr/Nb, Zr/Y, Nb/Y) and
REE patterns and the absence of a parallel differentiation trend.
Compositional variations within individual cogenetic basaltic
suites are best accounted for by crystal fractionation.
The elements Ir, Ru and Rh display overall compatible behaviour during crystal fractionation, with the highest concentrations in the primitive basalts. The geochemical behaviour of Pt and Pd is bimodal. In the lavas from Site 917 (upper series) they behave as incompatible elements during crystal fractionation, suggesting that the magma was S undersaturated. The basalts from Sites 918, 988, 989 and 990 are characterized by low Cu/Pd in the most primitive lavas, moderate to high Pt and Pd concentrations and compatible behaviour of Pt and Pd during magma differentiation. This suggests that the primary magmas were S undersaturated and reached S saturation during ascent, resulting in sulphide segregation.
The lava sample from unit 16 (Site 917 upper series) is thought to represent a near-primary magma composition and was used to determine the behaviour of the PGE during partial melting. The incompatible behaviour of Pt and Pd in this basalt implies that the primary melt was S undersaturated. Therefore, we calculate bulk partition coefficients for the PGE for the system ‘silicate solid–silicate melt’. For Ir, Ru and Rh, we calculate bulk partition coefficients of 1·2–4·3, whereas our data suggest D values <1 for Pd and Pt. It is shown that Pt is less incompatible than Pd during partial melting, and is probably incorporated in silicate and/or oxide phases.
TOP
ABSTRACT
INTRODUCTION
