Journal of Petrology Advance Access originally published online on November 16, 2006
Journal of Petrology 2007 48(1):185-218; doi:10.1093/petrology/egl061
© The Author 2006. Published by Oxford University Press. All rights reserved. For Permissions, please e-mail: journals.permissions@oxfordjournals.org
Petrogenesis of the Swaziland and Northern Natal Rhyolites of the Lebombo Rifted Volcanic Margin, South East Africa
Jodie A. Miller* and
Chris Harris
Department of Geological Sciences, University of Cape Town, Rondebosch, 7700, South Africa
RECEIVED
JANUARY 30, 2006;
ACCEPTED
SEPTEMBER 20, 2006
 |
ABSTRACT
|
|---|
The Jozini and Mbuluzi rhyolites and Oribi Beds of the southern
Lebombo Monocline, southeastern Africa, have geochemical characteristics
that indicate they were derived by partial melting of a mixture
of high-Ti/Zr and low-Ti/Zr Sabie River Basalt Formation types.
Compositional variations within the different rhyolite types
can largely be explained by subsequent fractional crystallization.
The Sr- and Nd-isotope composition of the rhyolites is unique
amongst Gondwana silicic large igneous provinces, having
Nd values close to Bulk Earth (–0·94 to 0·35)
and low, but more variable, initial
87Sr/
86Sr ratios (0·7034–0·7080).
Quartz phenocryst
18O values indicate that the rhyolite magmas
had
18O values between 5·3 and 6·7
, consistent
with derivation from a basaltic protolith with
18O values between
4·8 and 6·2
. The low-
18O rhyolites (< 6·0
)
come from the same stratigraphic horizon and are overlain and
underlain by rhyolites with more ‘normal’
18O magma
values. These low-
18O rhyolites cannot have been produced by
fractional crystallization or partial melting of mantle-derived
basaltic material. The rhyolites have low water contents, making
it unlikely that the low
18O values are the result of post-emplacement
alteration. Modification of the source by fluid–rock interaction
at elevated temperatures is the most plausible mechanism for
lowering the
18O magma value. It is proposed that the low-
18O
rhyolites were derived by melting of earlier altered rhyolite
in calderas situated to the east, which were not preserved after
Gondwana break-up.
KEY WORDS: rhyolite; Lebombo; stable and radiogenic isotopes; low-18O magmas; partial melting
|
INTRODUCTION
|
|---|
Large Igneous Provinces (LIPs) located near present-day continental
margins provide an important record of crustal and mantle geodynamics
during the pre- to syn-rift stages of the break-up and dispersal
of supercontinents. Most studies of LIPs have tended to focus
on the extensive and generally better preserved mafic sequences
along volcanic rifted margins; however, LIPs can also contain
aereally and volumetrically significant amounts of silicic volcanic
rocks. Understanding the relative timing of the eruptions and
petrogenetic characteristics of the silicic volcanic rocks with
respect to the flood basalts provides important constraints
on the contribution of crustal material to LIP magmatism and
the role of ancient subduction episodes in generating hydrous
lower crustal source materials as well as environmental and
climatic impacts during LIP emplacement (Wignall, 2001
; Bryan
et al., 2002
). Other studies have emphasized the stratigraphic
importance of the silicic eruptive units in constraining the
volcanic stratigraphy over several hundred kilometres within
often monotonous and internally complex flood basalt lava successions,
and for reassembling volcanic provinces now isolated by continental
break-up and seafloor spreading (Milner
et al., 1995
; Bryan
et al., 2002
; Ukstins
et al., 2002
).
The Mesozoic break-up of Gondwana was associated with four LIPs emplaced across southern Africa, Antarctica, and southern South America: the Ferrar (
180 Ma; Encarnacion et al., 1996), the Karoo (183–180 Ma; Allsopp et al., 1984; Duncan et al., 1997; Riley et al., 2004), the Chon Aike (188–178 Ma; V1 of Pankhurst et al., 2000); and the Paraná–Etendeka (135–130 Ma; Hawkesworth et al., 1992) provinces. The Ferrar, Karoo and Paraná–Etendeka provinces are dominated by mafic volcanic rocks with less than 5% by volume silicic material, usually in the form of ignimbrites–rheoignimbrites, rhyolitic lavas, or tuff deposits. In contrast, the Chon Aike province is dominated by silicic volcanic rocks with only rare basalts. Previous geochemical and isotopic studies of the above LIPs have established a reasonably extensive database for the basaltic and rhyolitic components of these provinces. Geochemical studies of the basaltic rocks have been used to identify the potential mantle source(s) involved (e.g. Erlank et al., 1984; Hawkesworth et al., 1984; Peate et al., 1992), whereas the silicic volcanic rocks have been used to evaluate the contribution of crustal material in the petrogenesis of the magmas. Based on these geochemical studies, models for the generation of silicic magmas in LIPs have generally fallen into two groups. The first has invoked either fractional crystallization of basalt accompanied by varying amounts of crustal assimilation (e.g. Garland et al., 1995; Ewart et al., 2004), whereas the second requires partial melting of the crust with superimposed fractional crystallization (e.g. Cleverly et al., 1984; Piccirillo et al., 1988; Pankhurst & Rapela, 1995; Harris et al., 1990; Harris & Milner, 1997).
The rhyolites of the Lebombo Monocline form part of the Karoo volcanic province and show two unusual characteristics compared with compositionally similar rocks from other LIPs: first, their initial Sr- and especially Nd-isotope ratios are not elevated in comparison with the associated basalts, which indicates little or no input of old continental crust (Hawkesworth et al., 1984; Harris & Erlank, 1992); second, some of the Lebombo rhyolites appear to have crystallized from low-18O magmas (Harris & Erlank, 1992). Since the Harris & Erlank (1992) study was undertaken, identifying the low-18O character of the rhyolites, there has been considerable progress in quantifying oxygen isotope fractionation between phenocryst phases and melts of different compositions (e.g. Stolper & Epstein, 1991; Matthews et al., 1994; Palin et al., 1996; Zhao & Zheng, 2003; Bindeman et al., 2004) and in our understanding of the generation of low-18O primary rhyolite magmas (e.g. Balsley & Gregory, 1998; Bindeman & Valley, 2001, 2003; Boroughs et al., 2005). Whereas low-18O rhyolite magmas are relatively common in some caldera systems, they are still a relatively rare phenomenon world-wide (Balsley & Gregory, 1998). In this sense, the Lebombo rhyolites are significant because, although not all of them have low 18O values, and the 18O values are not as low as in some other examples (e.g. Yellowstone Plateau volcanic field, Wyoming: Hildreth et al., 1984), they are the only low-18O magmas associated with a large igneous province.
Harris & Erlank (1992) analysed a comparatively small number of samples from a wide geographical area along the Lebombo Monocline and hence had little or no information on the stratigraphic variation in both the radiogenic and stable isotope data. Those workers suggested that the low 18O values of the Lebombo rhyolites resulted from high-temperature interaction between circulating meteoric water and the source prior to partial melting. In this study, a much more expanded geochemical and isotope dataset has been compiled, along three stratigraphic sections through the rhyolites of Swaziland and northern Natal. These sections represent the thickest sequences of rhyolites in the Lebombo Monocline and can be used to assess stratigraphic and spatial variations in their major- and trace-element and isotope geochemistry. In addition, the original dataset of Harris & Erlank (1992) has been re-evaluated in light of the more recent information on oxygen isotope fractionation in rhyolite magmas and within the context of the new data obtained in this study. This new expanded dataset is then used to assess the petrogenetic processes that produced the rhyolite magmas and to reach some conclusions about the physical environment in which they formed. In addition, we have attempted to constrain further the composition of the underplated material to determine if it is represented by any of the exposed basalt lavas.
|
RHYOLITES OF THE LEBOMBO MONOCLINE
|
|---|
The Karoo continental flood basalt province of southern Africa
represents the earliest phase of igneous magmatic activity associated
with initial continental thinning and subsequent break-up of
Gondwana during the Mesozoic (Eales
et al., 1984
). The Lebombo
Monocline, which represents the rifted volcanic margin of SE
Africa, is a 600 km long, north–south-trending, linear
feature parallel to the eastern margin of the Archaean Kaapvaal
Craton (Cleverly
et al., 1984
; Watkeys, 2002
), containing the
most extensive rhyolite sequences associated with the Karoo
volcanic province (
Fig. 1a). Together with the Save–Mwenezi
rift arm located in southern Zimbabwe, the total preserved volume
of rhyolite is thought to be
35 000 km
3 (Cleverly
et al., 1984
).
The Lebombo Monocline probably marks the transition from normal
thickness continental crust in the west to thinned continental
crust in the east, associated with rifting of Antarctica from
Africa during Gondwana break-up. The rhyolites are thought to
have been produced by decompression partial melting during the
initiation of rifting (Cleverly
et al., 1984
). The voluminous
Karoo eruptions consist predominantly of mafic rocks (tholeiitic
basalts), with the main phase of volcanism dated by the Ar–Ar
method at
182 Ma (Duncan
et al., 1997
). Lesser amounts of silicic
rocks were erupted slightly later at 178–180 Ma (Duncan
et al., 1997
; Marsh
et al., 1997
). These ages were confirmed
by Riley
et al. (2004
), who showed that U–Pb sensitive
high-resolution ion microprobe (SHRIMP) ages for rhyolites interbedded
within the Sabie River Basalt Formation (SRBF) and the overlying
Jozini rhyolites were the same within error, with the latter
being 182·1 ± 2·9 Ma.
Stratigraphic relationships
The stratigraphy of the Karoo sequence in the Lebombo can be
summarized as a thick sequence of basalts of the Sabie River
and Letaba formations that overlie sedimentary rocks of the
Karoo Supergroup (Cox, 1972
; Cox & Bristow, 1984
; Eales
et al., 1984
). These basalts are overlain by rhyolites of the
Jozini and Mbuluzi formations (hereafter termed Jozini and Mbuluzi
rhyolites), which are in turn overlain by the Movene Basalt
Formation, which is in turn capped by Cretaceous sedimentary
rocks (
Fig. 1b); Eales
et al., 1984
). Both the Jozini and Mbuluzi
rhyolites occur as sheets representing individual cooling units,
which have features characteristic of both lavas and ignimbrites,
and are considered to have formed as anhydrous, high-temperature
(1000–1100°C) ash-flow tuffs (Betton, 1978
; Cleverly,
1979
). Textures typical of ignimbrites are preserved at the
top and base of each unit (Bristow, 1976
; Cleverly, 1977
) and
the lava-like features are considered to have formed in response
to post-eruptive welding and remobilization of the degassed
ignimbrite deposits (Cleverly, 1979
). The plagioclase-phyric
Jozini rhyolites, which are volumetrically the more significant
unit, directly overlie the Sabie River basalts along most of
the length of the Lebombo. Together with the sparsely quartz-phyric
Mbuluzi rhyolites (and associated subsidiary quartz-phyric Oribi
Beds), which are laterally discontinuous and occur principally
in Swaziland, the two formations comprise a maximum thickness
of around 5000 m of continuous rhyolite with no interbedded
basalts. Cleverly (1977
), in his original mapping in NE Swaziland,
distinguished basic, intermediate and acid rhyolite types (on
the basis of wt % SiO
2 and phenocryst populations) and mapped
more than 20 distinct flows up to 200 m thick, which pinch and
swell and extend laterally in excess of 60 km (
Fig. 1c). Flow
I, which is a relatively continuous flow in the region of this
study and directly overlies the Sabie River Basalt Formation,
was mapped as a basic rhyolite and probably equates to the Jozini
rhyolites. Flows II and upwards (
Fig. 1c) were mapped as both
basic and intermediate rhyolites but probably largely represent
Mbuluzi rhyolites. The interbedded and lenticular acid rhyolites
(Flows VI and IX) are known as the Oribi Beds and are quartz-phyric.
Of the other minor subsidiary rhyolite units, the most distinctive
are the Mkutshane Beds, which occur towards the bottom of the
Sabie River Basalt Formation. They are geochemically distinct
from either the Jozini or the Mbuluzi rhyolites and in particular
have elevated radiogenic isotope signatures indicative of crustal
input (Cleverly
et al., 1984
).
Petrography
Previous studies have revealed that, with the exception of the phenocryst populations, there are few petrographic differences between the different rhyolite units occurring in the Lebombo (Cleverly, 1977; Cleverly et al., 1984). In general, all the rhyolites are porphyritic and contain phenocrysts of one or more of plagioclase, clinopyroxene, magnetite, quartz and sanidine, as well as apatite, zircon, olivine and ilmenite (Cleverly et al., 1984). Phenocrysts are present in varying amounts but rarely exceed 30% by volume and more commonly make up 15–20% by volume of the rhyolites (Cleverly et al., 1984). The phenocrysts are usually set in a fine-grained groundmass of devitrified glass and, in the case of plagioclase and clinopyroxene, show complex zoning patterns and compositional differences specific to the individual rhyolite sample, confirming their origin as phenocrysts and not xenocrysts (Cleverly, 1977, 1979; Betton, 1978; Cleverly et al., 1984). Minor secondary hydrous phases are present as alteration products of both phenocrysts and groundmass (Harris & Erlank, 1992).
Plagioclase is the most common as well as dominant phenocryst, occurring as subhedral crystals 1–5 mm in size, with cores of oligoclase–andesine strongly zoned to more sodic rim compositions. Clinopyroxene phenocrysts occur in all but the most silica-rich rhyolites and exhibit complex zoning patterns (Betton, 1979; Cleverly et al., 1984). They are much smaller than the plagioclase phenocrysts, being only 0·1–1 mm in length, and make up a much smaller proportion of the phenocryst population. Quartz phenocrysts are usually subhedral but clearly show the outline of pyramid faces. In some rocks, grains are partially resorbed and small magmatic inclusions are visible, clearly indicating that the quartz is igneous in origin. Quartz phenocrysts are not uncommon in the Mbuluzi rhyolites, ranging from trace amounts to just over 5% in sample SR24 (Table 1). Quartz phenocrysts are more abundant in the Oribi Beds, ranging from 4·1 to 8·3 vol.% but are almost completely absent from the Jozini rhyolites. The presence or abundance of quartz phenocrysts is one of the main tools used to distinguish between the rhyolite units and is particularly important in identifying the units from within the Mbuluzi rhyolites that can be classified as the Oribi Beds. The Oribi beds (samples MG16-20 and SR35) have a distinctive texture, with large phenocrysts of quartz (up to 3 mm diameter), plagioclase and alkali feldspar (up to 4 mm diameter) set in a fine-grained groundmass.
View this table:
[in this window]
[in a new window]
|
Table 1: Whole-rock major and trace element geochemistry for Mbuluzi, Jozini and Oribi Bed rhyolites from the Lebombo Monocline
|
|
|
SAMPLING TRAVERSES
|
|---|
A total of 46 samples were collected for geochemical analysis
from three sample traverses made approximately perpendicular
to the north–south strike of the rhyolites (
Fig. 1c).
These samples are made up of 27 samples from the Mbuluzi rhyolites,
six from the Oribi Beds and 12 from the Jozini rhyolites. One
sample of the Mkutshane Beds was also collected. The most northerly
traverse was made along the Mbuluzi River between Mlawula Station
in Swaziland and the Mozambique border (
Fig. 1c), and is referred
to as the Mbuluzi section (all samples prefixed MB). To the
south of this, a second traverse was made to the north and west
of the town of Siteki (
Fig. 1c), and is referred to as the Siteki
section (all samples prefixed SR). Samples from the two Swaziland
traverses have been placed into their correct stratigraphic
order using Global Positioning System (GPS) coordinates and
the maps of Cleverly (1977
). In this region between the Mbuluzi
River and Siteki, the outcrop distribution of the individual
rhyolite flows appears to be controlled by the pre-existing
topography, and in some cases successive flows truncate previous
flows, causing lateral discontinuities in some of the flows
(
Fig. 1c). As a result, all of the samples from the Mbuluzi
section occur in flows that are stratigraphically below the
flows from which the samples of the Siteki section were collected,
with the exception of samples SR28 and SR40 (
Fig. 1c). However,
the actual stratigraphic thickness of rhyolite in the two areas
is more or less equal (
Fig. 1c). To the south, a third traverse
was made through the Jozini rhyolites and the samples were collected
along the road through Jozini (samples prefixed J) towards the
Mozambique border, south of Swaziland (
Fig. 1b).
|
ANALYTICAL TECHNIQUES
|
|---|
Quartz phenocryst volume percent was determined by point counting
an average of
3000 points per thin section. Whole-rock major-
and some trace-elements (Nb, Zr, Y, Rb, and Sr) were determined
by X-ray fluorescence (XRF) at the University of Cape Town (UCT)
and the University of Stellenbosch using an X’unique spectrometer;
errors and detection limits are equal to or lower than those
of le Roex
et al. (1981
). All other trace element and rare earth
element (REE) concentrations were determined by inductively
coupled plasma mass spectrometry (ICP-MS), using a three-step
HF–HNO
3 acid digestion procedure in sealed and heated
Savilex vessels. Dried samples were taken up in a 5% HNO
3 solution
containing 10 ppb Re, Rh, In and Bi as internal standards and
analysed on a Perkin Elmer/Sciex Elan 6000 ICP-MS system in
the Department of Geological Sciences at UCT. Further details
of the analytical techniques have been given by le Roex
et al.
(2001
). Unless stated otherwise, whole-rock data plotted on
the diagrams have been recalculated to anhydrous totals before
plotting. Sample SR35, which has an unusual REE pattern, was
run in duplicate, with no significant difference between the
analyses.
Mineral separates (quartz, feldspar, and magnetite) for O-isotope analysis were hand-picked from sieved, jaw-crush material. Quartz phenocrysts were easily recognized as they are relatively large and equidimensional, and were cleaned in warm dilute HF to remove any adhering material. In one sample (MG17) it was possible to distinguish phenocryst quartz from quartz of secondary origin and both were analysed. The hand specimen contains narrow veins and small cavities lined with quartz crystals. The secondary quartz was distinct, as crystal fragments showed prism faces and lacked melt inclusions, which were typically visible in the phenocrysts. In all other samples where quartz was separated it was clearly of phenocryst origin. Although pyroxene phenocrysts are present in some of the samples, it was impossible to separate sufficient material for analysis.
Oxygen isotope ratios were determined at UCT following the method of Clayton & Mayeda (1963) using ClF3 as the oxidizing reagent (Borthwick & Harmon, 1982). Samples were degassed under vacuum at 200°C for 2 h prior to addition of ClF3. The samples were then reacted for at least 4 h at 550°C and the liberated O2 was converted to CO2 using a hot platinized carbon rod. Isotope ratios were measured on a Finnigan MAT 252 mass spectrometer. Data are reported in the familiar notation where 18O is (Rsample/Rstandard – 1) x 103 and R is the ratio of 18O/16O. At UCT, a quartz standard (MQ) calibrated against NBS-28 was analysed in duplicate with each run of eight samples and used to convert the raw data to the SMOW scale using the 18O value of 9·64% for NBS-28 recommended by Coplen et al. (1983). The average difference between duplicates of MQ analysed during the course of this study is 0·15.
Hydrogen isotopes were determined at UCT using the method of Vennemann & O’Neil (1993). Whole-rock samples were degassed on the vacuum line at 200°C prior to pyrolysis. An internal water standard (CTMP, D = –9) was used to calibrate the data to the SMOW scale and a second water standard (DML, D –300) was used to correct for scale compression (e.g. Coplen, 1993). Typical reproducibility of internal biotite standards during the period of analysis was ± 2 (1
). Water contents were determined either from the voltage measured on the mass 2 collector or (in the case of large samples) from the pressure measured during sample inlet using identical inlet volume to standards of known number of micromoles. Repeated measurements of water standards of known mass suggest that the typical relative error for the water content is 3% at the 2 mg level.
The Sr- and Nd-isotope data were obtained using a VG Sector mass spectrometer in the Department of Geological Sciences at UCT, following standard chemical separation procedures described by le Roex & Lanyon (1998). The standard SRM987 gave 0·71024 ± 1 and 0·71025 ± 9 during the Sr-isotope runs and the La Jolla Nd-isotope standard gave values of 0·51178 ± 1, 0·51178 ± 1, 0·51177 ± 1 during the Nd-isotope runs. The radiogenic isotope data were normalized to values of 0·71026 (SRM987) for Sr isotopes and 0·51184 (La Jolla) for Nd isotopes.
|
RESULTS
|
|---|
Major, trace and rare earth element data for the Mbuluzi rhyolites
and Oribi Beds as well as one sample of Jozini rhyolite from
Swaziland (Mbuluzi and Siteki Traverse) and the Jozini rhyolites
in northern Natal (Jozini traverse) are presented in
Table 1.
Major element geochemistry
The samples analysed in this study have SiO2 contents (wt %) that vary between 60·14% and 76·25%, and almost all the samples are classified as rhyolites on a total alkalis–silica (TAS) diagram (Fig. 2a). The Oribi Beds have a significantly higher average wt % SiO2 (74·24%) than the Mbuluzi and Jozini rhyolites (69·95% and 67·70%, respectively). As the loss on ignition (LOI) for all of the samples is relatively low, recalculation to anhydrous totals increases these averages only marginally (an average of 1· 4 wt %). Six samples should be classified as high-K dacites rather than rhyolites (Fig. 2b), and one sample (MG22) of the Jozini rhyolite falls within the high-K andesite field.
Major element oxides vs SiO
2 are presented in
Fig. 3 and show
similar trends to those described and discussed by Cleverly
et al. (1984
). The new data illustrate several important features;
in particular the good negative correlations between TiO
2, Fe
2O
3*,
P
2O
5 and SiO
2 and less strong negative correlations between
CaO and MgO and SiO
2. The Jozini rhyolites generally plot at
the low-SiO
2, high-Fe
2O
3 end of the data array, whereas the
Oribi Beds plot at the high-SiO
2, low-Fe
2O
3 end of the data
array. There is a general decrease in Na
2O with increasing SiO
2,
whereas K
2O shows a positive correlation with SiO
2. The weak
negative correlation between Al
2O
3 and SiO
2 is probably due
to the constant sum effect, with the Oribi Beds having lower
Al
2O
3 than the other samples. One sample, MG22, has very high
Fe
2O
3, CaO, MgO and P
2O
5 (along with low Zr and Nb—see
below) and plots in the high-K andesite field, possibly indicating
less differentiated composition.
View larger version (27K):
[in this window]
[in a new window]
[Download PowerPoint slide]
|
Fig. 3. Major element variation diagrams: (a) Al2O3, (b) TiO2, (c) Fe2O3*, (d) MgO, (e) CaO, (f) Na2O, (g) K2O and (h) P2O5 vs SiO2 (wt%) for the Lebombo Rhyolites. Also shown for comparison are the average values for the Mbuluzi and Jozini Formation rhyolites and the Oribi Beds from Cleverly et al. (1984). Fe2O3* is all Fe reported as Fe2O3.
|
|
Trace element geochemistry
Comparison of trace element contents with wt % SiO
2 indicates
that correlations within each of the three rhyolite groups are
rather poor, especially for the Jozini rhyolites. In general,
SiO
2 correlates negatively with Sr and Zr and positively with
Nb and Rb. However, there is no clear distinction between the
three rhyolite groups except in the case of Zr, which acts as
an effective discriminator (
Fig. 4a). The Jozini rhyolites have
consistently higher Zr contents (1100–1220 ppm) than the
Mbuluzi rhyolites (800–1150 ppm), with the Oribi Beds
having the lowest Zr concentrations (450–700 ppm). One
sample of the Jozini rhyolite with a lower SiO
2 content also
has a lower Zr content (923 ppm). Overall, there is a strong
negative correlation between Rb and Sr (
Fig. 4c,
Table 1). Although
there is considerable overlap in the ranges of the three rhyolite
types, in general the Jozini rhyolites have the lowest Rb/Sr
ratios (average
0·7) whereas the Oribi Beds have the
highest Rb/Sr ratios (average
1·7), indicating that Rb/Sr
generally increases with increasing silica content. Correlations
within the three rhyolite types between Zr and Y (
Fig. 4b) or
Nb (
Fig. 4d) are poor and serve mainly to highlight the usefulness
of Zr in distinguishing between the rhyolites. Although the
Oribi Beds have relatively low and constant Zr/Y of around four,
the Jozini rhyolites have a wide range of Zr/Y because of the
relatively constant Zr content combined with variable Y contents,
with a similar although less pronounced pattern also being observed
in the Mbuluzi rhyolites (
Fig. 4c).
Figure 4 indicates that
the Jozini rhyolites have a relatively constant Zr/Nb of slightly
less than 15, whereas the Oribi Beds and the Mbuluzi rhyolites
show a fair degree of scatter in Zr/Nb, but have average Zr/Nb
of 6·9 and 12·3, respectively.
Chondrite-normalized REE patterns (
Fig. 5) for the three rhyolite
types are approximately parallel, moderately light REE (LREE)-enriched
with distinct negative Eu anomalies (average Eu/Eu* = 0·64).
The magnitude of the Eu anomaly decreases with increasing silica
content (
Fig. 6) for the rhyolites as a whole, but the correlation
within the Jozini and Mbuluzi rhyolites is poor. One Oribi Bed
rhyolite (SR35) has an extremely unusual REE pattern with strongly
depleted middle REE (MREE), a strong negative Eu anomaly and
a slightly positive Ce anomaly (
Fig. 5d). A repeat analysis
of this sample gave the same concentrations and it is, therefore,
concluded that REE pattern is not an artefact of incomplete
dissolution of the sample and is possibly related to secondary
alteration processes. The Jozini rhyolites have slightly lower
REE contents than the Mbuluzi rhyolites. The Oribi Beds samples
(excluding sample SR35) have similar REE patterns.
Figure 5 highlights the similarity in the REE patterns for the three
rhyolite groups. A minority of samples show negative Ce anomalies,
with four samples having Ce/Ce* <0·8, whereas one
sample had a pronounced positive Ce anomaly of 1· 49.
View larger version (25K):
[in this window]
[in a new window]
[Download PowerPoint slide]
|
Fig. 5. Chondrite-normalized REE diagrams for rhyolites of (a) the Mbuluzi Formation, (b) the Oribi Beds and (c) the Jozini Formation. Samples of the Mbuluzi Formation from both the Mbuluzi and Siteki traverses are shown on the same diagram (a) and exhibit no differences in REE patterns. (d) REE patterns for the three rock groups superimposed, illustrating the similarity of REE patterns. The unusual REE pattern for one sample of the Oribi beds is SR35. Chondrite normalization factors taken from Sun & McDonough (1989).
|
|
Oxygen and hydrogen isotopes
Oxygen and hydrogen isotope data are presented in
Table 2. The
18Owr values show a considerably larger range (4·2–8·5
)
than the
18O quartz values (5·9–7·3
) or
the
18O plagioclase values, except for one sample (J60) with
a very high
18O feldspar value of 13·0
. Unfortunately,
most of the rhyolites did not contain more than one separable
phenocryst phase. In only three samples was it possible to determine
the
18O values of both quartz and plagioclase and the erratic
values of
quartz–feldspar (1· 9, –0·3
and –1·1
) indicate that equilibrium was not maintained,
as a result of differential resetting of the feldspar. For sample
MG17 both phenocryst quartz and vein quartz were analysed. The
former had a
18O value of 6·7
whereas the latter had
a
18O value of 8·7
. The whole-rock
D values range from
–67 to –111
, with an average of –96 ±
12
(1
). The water content (wt % H
2O
+) determined during the
H-isotope determination ranges from 0·3 to 1·2
wt %, with an average of 0·62 ± 0·27 (1
).
The stable isotope data show several important features.
18Owr
values plotted against wt % SiO
2 (
Fig. 7a) indicates that the
Jozini rhyolites have higher
18Owr values [average
18O = 9·9
± 0·6 (1
)
] than the Mbuluzi rhyolites or the Oribi
Bed rhyolites [combined average
18O = 6·4 ± 1·
9 (1
)
] at a given SiO
2 content. No correlation exists between
18Owr and
18O quartz values (
Fig. 7b), with the bulk of the
variation occurring in whole-rock rather than quartz values.
18Owr also shows no correlation with either wt % H
2O
+ (
Fig. 7c)
or whole-rock
D (
Fig. 7d).
Figure 7 indicates that rocks with
<0·4 wt % H
2O
+ have consistently low
D values, with
values between –90 and –116
, whereas samples with
>0·4 wt % H
2O
+ have a much larger range of
D values.
Although there is a lack of correlation generally between
18Owr
values and wt % SiO
2, the Oribi beds show a fairly good positive
correlation (
Fig. 7a). This cannot be related to the
18O value
of the magma because the correlation between
18O values for
whole-rocks and quartz is, if anything, negative (
Fig. 7b).
Positive correlations between
18Owr values and wt % SiO
2 as
a result of low-temperature alteration are to be expected (Harris,
1989
) in glassy rocks because glass is susceptible to low-temperature
exchange, and the amount of glass increases with increasing
silica content. This would also explain the weak positive correlation
between whole-rock
D values and SiO
2 (
Fig. 7f) in that hydration
water might be expected to have less negative
D values.
Strontium and neodymium isotopes
Strontium and neodymium isotope data for 15 of the rhyolites are reported in Table 3 and shown graphically in Fig. 8. Samples were selected on the basis of the presence of quartz phenocrysts and include four samples of the Oribi Beds, 10 samples of the Mbuluzi rhyolites and one sample from the Mkutshane Beds. No samples of the Jozini rhyolites were analysed as these were analysed previously by Harris & Erlank (1992). An important feature of the data from this study is that there is very little variation in initial 143Nd/144Nd ratios (0·51236–0·51243), with Nd values close to Bulk Earth (average Nd = –0·03). Initial 87Sr/86Sr values are much more varied, ranging from 0·7035 to 0·7080. With two outlying samples omitted, the Rb–Sr data form an isochron that yields an age comparable with those found in previous studies, indicating that the Rb–Sr system has been relatively undisturbed since the time of eruption. The four Oribi Beds samples show substantially less variation in initial 87Sr/86Sr ratios (0·7039–0·7052) than the Mbuluzi rhyolites (0·7035–0·7080). The single sample of the Mkutshane Beds has highly radiogenic initial 87Sr/86Sr and 143Nd/144Nd isotope ratios (0·7147 and 0·51159, respectively) and is clearly crustal in origin, as shown by Betton (1979) on the basis of Pb and Sr isotope data. There is no correlation between Sm/Nd and measured 143Nd/144Nd, indicating that there is no isochron relationship for this system. The inset diagram in Fig. 8 has an expanded Nd scale and shows that there is no correlation between Nd and initial Sr-isotope ratio. The initial 87Sr/86Sr and Nd composition of a variety of basaltic and felsic magma types from southern African and adjacent Gondwana volcanic provinces is also shown in Fig. 8 for comparison. The compositional variation of the Lebombo rhyolites is unlike that of any of the other magma types and this feature will be discussed in more detail below.
|
DISCUSSION
|
|---|
Stratigraphic variations
Lateral variations in the thickness and continuity of the individual
rhyolite flows (
Fig. 1c) make the assessment of systematic variations
in geochemistry with stratigraphic height problematic.
Figure 9 shows data from the combined Mbuluzi and Siteki traverses for
various geochemical parameters as a function of flow number.
On first pass there seems little evidence for systematic changes
in the chemical composition of the various rhyolite flows with
stratigraphic height. The samples of Oribi rhyolite from Flows
VI and IX indicate differences that would be consistent with
an increasing degree of fractional crystallization between the
flows, in particular higher SiO
2 contents (
Fig. 9a) coupled
with lower Fe
2O
3 contents (
Fig. 9b), although it should be noted
that this is based on the one sample of the Oribi beds from
Flow IX. The Rb/Sr ratios show a small variation in that the
middle flows have slightly elevated ratios compared with the
flows at the base and top of the sequence examined (
Fig. 9c).
The Zr/Nb ratios are constant and illustrate again the different
Zr/Nb ranges for the Oribi, Mbuluzi and Jozini rhyolites (
Fig. 9d).
In contrast, the variation in stable isotope data with respect
to stratigraphic height clearly indicates lower
18O quartz values
associated with Flows IX, X and a
2 (
Fig. 9e). Although Flows
IX and X do not appear to be connected to Flow a
2, all three
flows occur at the same stratigraphic position, towards the
base of the Mbuluzi Formation (
Fig. 1c).
18O quartz values in
these flows vary between 5·9 and 6·3
. In the case
of Flows X and IX, the flows immediately above and below have
18O quartz values of 7·3
, whereas for Flow a
2, the surrounding
flows have
18O quartz values around 7·1
. Moreover, the
18O quartz values appear to record a bimodal
18O distribution,
as the higher
18O quartz values all fall between 6·9
and 7·3
, although two samples have
18O quartz values
of 6·7
. In comparison, the
18Owr values (
Fig. 9f) show
both high and low values that seem to oscillate with stratigraphic
height (see
Table 1 and
Fig. 1c). This is particularly clear
when looking at the Mbuluzi section, where from east to west
the
18Owr values start at 6·7
, increase to 8·4
,
decrease to 3·9
, increase again to 6·1
, before
decreasing again to 4·3
and ending at 7·3
. This
last sample is of the Jozini rhyolite, which, in general, has
higher
18Owr values than the Mbuluzi rhyolite and Oribi Beds.
This pattern of oscillating
18Owr values in part explains the
range in
18Owr values for some individual flows, and is best
illustrated in Flow VII, which incorporates the topmost section
of the Mbuluzi traverse. The difference in the pattern of
18Owr
and
18O quartz values indicates that whereas the
18O quartz
values reflect original magmatic values, the
18Owr values reflect
hydrothermal alteration processes whose timing with respect
to rhyolite eruption must be determined. The significance of
these features is discussed in more detail below. The rhyolite
Sr values are higher in the Siteki traverse than in the Mbuluzi
traverse (
Fig. 9f). However, this pattern is obscured by the
within-flow variation, which is often considerable, and the
small number of data points. Likewise, the samples from the
Jozini traverse do not show any significant stratigraphic variation,
although there are some indications that samples from near the
base of individual flows (J51, J57 and J59) have lower than
normal SiO
2 contents.
Geochemical modelling
Origin of the rhyolites by partial melting
Cleverly et al. (1984) pointed out that the Lebombo rhyolites form a coherent group of rocks, a fact borne out by the new extended dataset, and considered that the close association with the underlying basalts indicates a clear petrogenetic link. Derivation by partial melting of pre-Mesozoic crust was rejected on the basis of radiogenic isotope data, and Cleverly et al. (1984) considered the possibility that the rhyolite magmas formed either by partial melting of a basaltic source or by prolonged fractional crystallization of a basaltic magma. They further considered the possibility that the range in chemical composition within the rhyolites might be due to varying degrees of partial melting of a basaltic parent or by fractional crystallization of the rhyolite magma once formed. Cleverly et al. (1984) concluded that the rhyolite magmas were produced by decompression partial melting of a basaltic sill complex that was coeval with the basalt magmatism. During crustal thinning prior to continental separation at 155 Ma (Watkeys, 2002), the base of the crust would have been uplifted by as much as 20 km. Provided the sill complex had not cooled significantly below its solidus, the drop in pressure would have been sufficient to initiate partial melting. Since 1984, the geochronological database on the Lebombo volcanic rocks has been greatly expanded. The most recent age determinations (Riley et al., 2004) favour the eruption of the entire 12 km sequence of basalts and rhyolites within 1–2 Myr at about 180–182 Ma. Although this would require rapid crustal thinning, it would also mean that the basaltic sill complex remained close to its solidus, rendering it susceptible to decompression partial melting.
It is not the purpose of this study to repeat the partial melting and fractional crystallization modelling of Cleverly et al. (1984) and Betton (1979). Nevertheless, various questions remain unanswered. First, Cleverly et al. (1984) did not discuss in any detail the composition of the source material and its relationship with the exposed basalt types and intrusions. Second, the petrogenetic relationships between the Jozini and Mbuluzi Formation rhyolites and the Oribi beds require further consideration, in the light of the greatly expanded dataset from this study, in particular the addition of detailed REE data.
Comparison of basalt and rhyolite magma types
The SRBF forms the major thickness of basalts that underlie the rhyolites along the entire length of the Lebombo (Fig. 1b). If the rhyolites were derived by partial melting of underplated basaltic material, the SRBF is the most likely surface representative of this basaltic source. Based on major and trace element geochemistry, previous studies (e.g. Duncan et al., 1984; Sweeney & Watkeys, 1990; Sweeney et al., 1994) have shown that there are various basaltic magma types in the Lebombo, the most significant division being between the high- and low-Ti/Zr (HTZ and LTZ) types. Sweeney et al. (1994) further subdivided the HTZ basalts into a high- and low-Fe type (HTZ-HF and HTZ-LF). These basalt types are geographically distinct in that the HTZ types are found in the northern Lebombo whereas the LTZ type is confined to the southern Lebombo. The transition from LTZ to HTZ basalts is apparently gradational and occurs north of Swaziland along the Sabie River (Sweeney et al., 1994). In that region, the general sequence of eruption is first HTZ-LF, followed by LTZ, followed by HTZ-HF, although in some areas the presence of interbedded basalt types indicates that simultaneous eruption occurred at times (Sweeney et al., 1994).
The geochemical differences between the basaltic magma types are clearly illustrated using a combination of trace element and major element data in Fig. 10. The HTZ and LTZ basalt types have distinct TiO2 and K2O contents (Fig. 10a and c), whereas the distinction between HTZ-HF and HTZ-LF as well as that with the LTZ type are shown more clearly by comparing Zr/Y ratios (Fig. 10d). In contrast to the basalt magma types, the rhyolite magma types show no clear difference in TiO2 or K2O (Fig. 10a and c). However, the distinct differences between the Jozini and Mbuluzi rhyolites, in terms of Zr and Zr/Y (Fig. 10b and d), mirror the differences seen in the different basalt types. The relationship between Zr and Y (Fig. 10d) is particularly interesting in that the Jozini rhyolites have Zr/Y 10, as is the case for the low-Fe variant of the HTZ basalts (Sweeney et al., 1994). On the other hand, the Oribi beds have similar Zr/Y ratios to the LTZ basalts, whereas the Mbuluzi rhyolites have fairly consistent Zr/Y ratios that are similar to those of the HTZ-HF type. The Mwenezi rhyolites have lower Zr and Y contents, but similar Zr/Y to the Jozini rhyolites (Fig. 10d). If the rhyolites were produced by partial melting of underplated basalt without subsequent modification of their trace element compositions by fractional crystallization or crustal contamination, an obvious first-order conclusion is that the high-Zr/Y Jozini and Nuanetsi rhyolites were produced by partial melting of the HTZ-LF basalts whereas the Mbuluzi rhyolites were produced by partial melting of the HTZ-HF basalts, and only the Oribi beds were derived by partial melting of an LTZ basalt source. However, given that the concentration of Zr has probably been affected by fractional crystallization of zircon, this hypothesis should be treated with caution and will be discussed further below.
View larger version (34K):
[in this window]
[in a new window]
[Download PowerPoint slide]
|
Fig. 10. Comparison of basalt and rhyolite compositions from Karoo database (courtesy of A. R. Duncan). The new data presented in this paper are shown as the grey field. (a) TiO2 vs wt% SiO2; (b) Zr vs wt% SiO2; (c) K2O vs wt% SiO2; (d) Y vs Zr (ppm).
|
|
Variation in rhyolite composition
Cleverly
et al. (1984
) successfully modelled the variation within
the Mbuluzi rhyolites from 70 to 72 wt % SiO
2 by about 9% crystallization
of an assemblage consisting of 69% plagioclase (An
34), 11% pyroxene,
and 20% magnetite. The proportions of plagioclase to pyroxene
match the typical phenocryst assemblages in the rhyolites, but
magnetite is much less abundant in the rhyolites themselves.
On the basis of their modelling, Cleverly
et al. (1984
) suggested
that either substantial thicknesses of magnetite cumulates must
exist beneath the Lebombo, or that fractional crystallization
is not a viable mechanism for generating the variation in chemical
composition. In focusing on the Mbuluzi rhyolites, Cleverly
et al. (1984
) did not model the compositional variation in the
various rhyolite types as products of different amounts of fractional
crystallization of a single parental felsic magma.
Quantitative modelling of major elements as part of this work concentrated on producing a silica-rich Oribi bed from a typical Mbuluzi Formation rhyolite. Sample SR32 was chosen as a typical Mbuluzi rhyolite as it is petrographically fresh and has relatively low wt % SiO2 (70·28% normalized to 100% volatile-free with total Fe as FeO) compared with the Mbuluzi rhyolites as a whole. Compositionally, it is very similar to a number of the Jozini rhyolites and therefore no attempt was made to relate the Jozini and Mbuluzi rhyolites by fractional crystallization. Sample MG16 (SiO2 75·68%), which was chosen to represent the Oribi beds, contains 6 vol.% quartz phenocrysts. This sample has an LOI of 0·66 wt %, suggesting that alteration had little or no effect on the major element composition. Consideration of the relationship between bulk-rock and phenocryst compositions (Fig. 11) indicates that the strong negative correlation between Fe2O3 and SiO2 (Fig. 11a) is due to fractionation of clinopyroxene ± magnetite. The strong positive correlation between TiO2 and Fe2O3 (Fig. 11b) also requires some fractionation of Ti-magnetite, although the proportions of pyroxene and magnetite depend strongly on the exact composition of the pyroxene chosen. A combination of pyroxene and plagioclase fractionation accounts for the weak negative correlation between Na2O and SiO2 (Fig. 11c), whereas the positive correlation between K2O and SiO2 (Fig. 11d) indicates that some fractionation of K-rich alkali feldspar was also involved. The results of one model are shown in Table 4 and indicate that Oribi rhyolite could be generated from Mbuluzi rhyolite by 44·6% fractional crystallization of an assemblage consisting of 19% Fe-rich clinopyroxene, 59% plagioclase (An20), 14% sanidine (Or93), 7% Ti-magnetite, and 1% apatite. This is compatible with the phenocryst assemblage observed in the rhyolites, and in particular the magnetite content is more appropriate than the original model of Cleverly et al. (1984). The addition of quartz as a fractionating phase did not result in more tightly constrained models, and it would therefore appear that quartz fractionation was not important.
View this table:
[in this window]
[in a new window]
|
Table 4: Major element least-squares fractional crystallization model relating Mbuluzi rhyolite (SR32) to Oribi bed rhyolite (MG16) based on major element mixing model defined by Cleverly et al. (1984)
|
|
The trace element variations are more difficult to explain in
terms of fractional crystallization. Qualitatively, the major
element model described above is consistent with the negative
correlation between Sr concentrations and SiO
2 contents and
the positive correlation between both Rb and Nb concentrations
and wt % SiO
2. On the other hand, Zr concentrations show a generally
negative correlation with wt % SiO
2 (
Fig. 4a), as does Zr/Nb.
This suggests that Nb is behaving as an incompatible element
whereas Zr is not, presumably because of fractionation of zircon.
Numerous small crystals of zircon are visible in many of the
rhyolites. The Zr saturation concentration for rhyolites depends
on chemical composition and temperature (Watson & Harrison,
1983
). Application of the equations of Watson & Harrison
(1983
) shows that the M-value [mole fraction Na + K + 2Ca/(Al
x Si)] of the three rhyolite types is between 1·0 and
1·7 (mean 1·44). None of the rhyolites would have
been zircon saturated at 1020°C, which is close to the proposed
eruption temperature of the rhyolite magmas (Betton, 1978
).
Nearly all of the rhyolites would have been saturated in zircon
at 930°C, for which the Zr concentration is about 600 ppm
for an M-value of 1·44. It therefore seems feasible that
zircon fractionation occurred during fractional crystallization
of the rhyolite magmas. The Zr/Y ratios show a scattered relationship
with SiO
2 but also suggest that Y is behaving more incompatibly
than Zr. The positive correlation between Rb/Sr ratio and SiO
2 (
Fig. 4b) and the positive correlation between the size of the
negative Eu-anomaly and SiO
2 (
Fig. 6) is consistent with the
importance of plagioclase fractionation throughout the evolution
of the rhyolites.
In the light of the above discussion, it is clear that the high Zr/Nb ratios (15) and Zr/Y ratios (10) that characterize the Jozini rhyolites and the low Zr/Nb (7) and Zr/Y (4) ratios that characterize the Oribi Beds are probably the result of fractional crystallization of zircon and do not reflect derivation from source materials of different compositions. The calculated trace element concentrations of MG16 based on the major element fractional crystallization model are shown in Table 4. Good matches were obtained for Sr and Ba, but the match for Rb was poorer, which might be due to the susceptibility of this element to alteration. The concentrations of incompatible elements such as La and Y (but not Ce) and Nb are overestimated by the model and Zr is underestimated by a factor of more than two. Fractionation of zircon and other accessory minerals could potentially explain the lack of fit of the trace element model. Zircon is present in most of the rocks and, as reported by Riley et al. (2004), a very small proportion of the zircons are inherited. However, there is no correlation between La/Yb and SiO2, indicating that zircon fractionation had a minimal effect on trace element abundances, other than Zr. The Kd values for the heavy REE (HREE) in zircon are much larger than those for the LREE (e.g. Mahood & Hildreth, 1983). The rhyo
TOP
ABSTRACT
INTRODUCTION
[SmN x GdN]. Chondrite normalization factors are from Sun & McDonough (1989
