Journal of Petrology Advance Access originally published online on April 17, 2007
Journal of Petrology 2007 48(6):1155-1183; doi:10.1093/petrology/egm013
© The Author 2007. Published by Oxford University Press. All rights reserved. For Permissions, please e-mail: journals.permissions@oxfordjournals.org
Constraining Fluid and Sediment Contributions to Subduction-Related Magmatism in Indonesia: Ijen Volcanic Complex
H. K. Handley1,*,
C. G. Macpherson1,
J. P. Davidson1,
K. Berlo2 and
D. Lowry3
1Department of Earth Sciences, Durham University, Durham DH1 3LE, UK
2Department of Earth Sciences, Bristol University, Bristol BS8 1RJ, UK
3Department Of Geology, Royal Holloway University Of London, Egham Tw20 0ex, UK
RECEIVED
JULY 10, 2006;
ACCEPTED
MARCH 7, 2007
 |
ABSTRACT
|
|---|
Ijen Volcanic Complex (IVC) in East Java, Indonesia is situated
on thickened oceanic crust within the Quaternary volcanic front
of the Sunda arc. The 20 km wide calderas complex contains around
22 post-caldera eruptive centres, positioned either around the
caldera-rim (CR) or along a roughly NE–SW lineament inside
the caldera (IC). The CR and IC lavas exhibit separate differentiation
histories. Major element and trace element modelling shows that
fractionation of olivine, clinopyroxene, Fe–Ti oxide ±
plagioclase is important in the CR group, whereas plagioclase
is the dominant fractionating mineral in the same assemblage
for the IC group. Spatial controls on magmatic differentiation
highlight important structural controls on magma storage and
ascent at IVC. Mantle-like
18O values, restricted ranges in
Sr, Nd and Hf isotope ratios, and a lack of correlation between
isotope ratios and indices of differentiation in IVC lavas are
consistent with magmatic evolution through fractional crystallization.
Furthermore, the small ranges in isotopic ratios throughout
the complex indicate little heterogeneity in the mantle. IVC
lavas possess higher
176Hf/
177Hf and
143Nd/
144Nd isotope ratios
than other volcanoes of Java, representing the least contaminated
source so far analysed and, therefore, give the best estimate
yet of the pre-subduction mantle wedge isotopic composition
beneath Java. Trace element and radiogenic isotope data are
consistent with a two-stage, three-component petrogenetic model
for IVC, whereby an Indian-type mid-ocean ridge basalt (I-MORB)-like
fertile mantle wedge is first infiltrated by a small percentage
of fluid, sourced from the altered oceanic crust, prior to addition
of <1% Indian Ocean sediment dominated by pelagic material.
KEY WORDS: differentiation; geochemistry; source components; Sr, Nd, Hf and O isotopes; Sunda arc
|
INTRODUCTION AND SETTING
|
|---|
Understanding magma genesis and evolution in subduction zone
environments is crucial to understanding the formation of the
continents and crustal recycling in the mantle. However, determining
original source compositions of volcanic arc rocks can be challenging,
as lavas rarely reach the Earth's surface without experiencing
some processes that modify their composition; for example, crystal
fractionation, magma mixing or crustal contamination (Davidson,
1996
; Thirlwall
et al., 1996
; Mandeville
et al., 1996
; Reubi
& Nicholls, 2004
; Davidson
et al., 2005
). In most petrogenetic
models, island arc magmas originate from a mantle wedge that
is modified by slab-derived components (Hawkesworth
et al.,
1991
; McCulloch & Gamble, 1991
; Pearce & Peate, 1995
).
The magnitude and nature of the contribution from the subducting
slab is debated, although there is a consensus that a slab-derived
fluid is involved, consisting of fluids from either the altered
oceanic crust (Hawkesworth
et al., 1997
; Turner
et al., 1997
;
Turner & Foden, 2001
) or overlying sediments (Class
et al.,
2000
), and other components, possibly partial melts, derived
from subducted sediment (Edwards
et al., 1993
; Elliott
et al.,
1997
; Turner & Foden, 2001
; Vroon
et al., 2001
).
Ijen volcanic complex (IVC) is located on the eastern edge of Java within the Quaternary volcanic front of the Sunda arc (Fig. 1). The arc forms the western part of the Indonesian subduction zone system, which extends over 3000 km from the Andaman Islands north of Sumatra to Flores in the Banda Sea. It formed as a result of the northward subduction of the Indo-Australian Plate beneath the Eurasian Plate at a rate of around 6–7 cm/year (DeMets et al., 1990; Tregoning et al., 1994). Despite this apparently simple tectonic setting, understanding magma genesis and evolution at the Sunda arc is complicated by the variable nature of the arc crust, the changing age of the subducting oceanic crust, which increases eastwards from 
80 to 130 Ma (Plank & Langmuir, 1998), and variation in the type and amount of subducted sediment, which ranges from bimodal siliceous ooze and clays with minor turbidites beneath the western part of the arc to strongly clay-dominated material with secondary carbonate and minor siliceous ooze in the east (Plank & Langmuir, 1998). Along the length of the arc the composition and thickness of the overriding Eurasian plate is also thought to change, from continental in the west to oceanic in the east (Hamilton, 1979). Beneath Java, the crust is 20 km thick and has a seismic velocity structure intermediate between continental and oceanic (Whitford, 1975). The Java crust is said to consist of ophiolite slivers, mélange and older volcanic rocks (Hamilton, 1979). The Sunda Land (SE Asian continental part of the Sunda block–Eurasian plate with pre-Tertiary basement) boundary shown in Fig. 1 suggests that in western Java the island arc is built on continental material. In eastern Java beneath the IVC (outside the hypothesized Sunda Land boundary) the overriding plate comprises thickened oceanic crust. However, recent research on inherited zircons (Smyth, 2005) provides evidence for old continental basement beneath the Southern Mountains Arc (south of the active volcanic axis) in East Java.
Despite a number of petrogenetic studies of Javan volcanoes
(Edwards
et al., 1993
; Turner & Foden, 2001
; Gertisser &
Keller, 2003
; and references therein), uncertainty still prevails
over the nature of the subduction component and the mechanisms
by which it is added to the mantle wedge. For example, Edwards
et al. (1993
) proposed a homogeneous slab contribution along
the Sunda arc, whereas Turner & Foden (2001
) identified
along-arc heterogeneity in this component. The IVC is located
on the eastern edge of Java within the Quaternary volcanic front
of the Sunda arc (
Fig. 1) and is therefore in an ideal location
to investigate the mantle source composition beneath Java; the
relatively thin crust on which the volcano sits minimizes the
likelihood of high-level continental crust contamination (compared
with studies in West Java) and potentially allows us to constrain
better crustal contributions in the mantle source. Furthermore,
apparent spatial controls on some major and trace element variations
at IVC mean that the volcanic complex provides a useful opportunity
to investigate the relationship between volcanic structure and
shallow-level processes influencing the composition of arc lavas.
The aims of this study are, therefore, two-fold: (1) to identify
the impact, if any, of volcanic and subjacent structure upon
magmatic differentiation; (2) to characterize the IVC magma
source and place constraints on the nature of the slab component(s)
and the transfer mechanism(s) to the mantle wedge.
|
OVERVIEW OF THE IJEN VOLCANO COMPLEX
|
|---|
The Ijen volcanic complex consists of several stratovolcanoes
and cinder cones constructed within and around a 20 km wide
caldera (
Fig. 2). The complex takes its name from Kawah Ijen
volcano, the only volcano of the complex that is currently active.
Based on an episode of caldera collapse, the stratigraphic succession
has been divided into pre-caldera, caldera and post-caldera
groups (van Bemmelen, 1949
). Little is known about the age,
structure or volcanic history of the pre-caldera cone (Old Ijen);
it is thought to be either early (van Bemmelen, 1949
) or late
(Sitorus, 1990
; K–Ar dating) Pleistocene. Sitorus (1990
)
considered the pre-caldera volcanic structure to be a large,
single, composite volcano whereas Berlo (2001
), suggested that
the pre-caldera collapse volcano may have had multiple vents.
Kemmerling (1921
) has suggested the possibility that twin volcanoes
made up Old Ijen, because of the oval shape of the subsequent
caldera. The deposits of Old Ijen disconformably overlie Miocene
limestone (van Bemmelen, 1949
) and consist of pyroclastic flow
deposits, a series of pumice airfall deposits and several lava
flows. The caldera wall forms a prominent arcuate ridge in the
north, but elsewhere in the complex the caldera rim is buried
by the post-caldera volcanoes (
Fig. 2). The collapse of the
pre-caldera cone(s) is thought to have occurred in stages more
than 50 000 years ago (Sitorus, 1990
). The caldera group consists
of two main units, comprising large volumes of caldera-forming
ignimbrite and minor layers of pumice airfall and lahar deposits.
Most of the eruptive products were deposited to the north (van
Bemmelen, 1949
). Shortly after caldera collapse a resurgent
dome (Blau) formed inside the caldera. The post-caldera group
is characterized by many types of eruption (phreatomagmatic,
phreatic, strombolian and plinian) and comprises 58 lithological
volcanic units, erupted from 22 separate vents (Sitorus, 1990
).
The post-caldera eruptive centres can be further divided on
the basis of their geographical position within the complex:
those located on the caldera rim (CR), dominantly composite
cones, or intra-caldera (IC) centres situated inside the southern
part of the caldera along a ENE–WSW lineation (
Fig. 2).
The latter are predominantly cinder cones with minor composite
cones and a lava dome. The IC volcanoes are generally younger
(with the exception of Blau) than the CR volcanoes (Sitorus,
1990
; Berlo, 2001
). Faulting is evident within the caldera complex
in the north (
Fig. 2) and large faults have affected the caldera-rim
volcanoes of Merapi, Ringgih and Djampit. Present volcanic activity
is limited to Ijen crater (Kawah Ijen), an acidic crater lake
with a maximum diameter of 1·2 km and a depth of around
200–300 m. The last major eruption created phreatomagmatic
air fall and pyroclastic flow deposits
2590 years ago. Historical
eruptions, which were mainly phreatic, occurred in 1796, 1817,
1917, 1936, 1952, 1993, 1994, 1999 and 2000 (Volcanic Survey
of Indonesia:
http://www.vsi.esdm.go.id/volcanoes/ijen/history.html).
|
SAMPLE GROUPINGS
|
|---|
This study builds on the established stratigraphy and preliminary
geochemical work of Sitorus (1990
). Few samples were available
from the pre-caldera and caldera groups, therefore emphasis
is on the geochemistry of the post-caldera rocks. Based on temporal
variations in geochemistry, particularly in SiO
2, FeO, CaO,
Ni and Sr contents, Sitorus (1990
) divided the post-caldera
samples into six groups, but admitted that the variations are
not entirely consistent through time, either within or between
groups. Therefore, in this study the majority of the samples
have been simply divided into CR and IC groups, based on their
spatial distribution rather than age (
Fig. 2). The centres of
Kawah Ijen and Blau are considered separately from the two main
post-caldera groups. Kawah Ijen is located on the western flank
of Merapi (a CR volcano) in line with the other intra-caldera
centres and appears to have geochemical affinities to both groups.
Blau, although easily identifiable as a volcanic centre inside
the caldera, is located north of the linear trend of the other
intra-caldera volcanoes and geochemical data show scatter between
the two groups. Stratigraphic relationships, geomorphology (deep
incised valleys) and K–Ar dating suggest that Blau is
significantly older than the rest of the intra-caldera centres
(Sitorus, 1990
).
A total of 43 samples from the IVC were obtained from the collection of Sitorus (1990) as rock powders, with the exception of five samples, for which hand specimens were also obtained. Major element data, mineralogical data and petrographic descriptions of the IVC rocks are taken from Sitorus (1990) and presented in Electronic Appendices 1–3, respectively (which may be downloaded from the Journal of Petrology website at http://petrology.oxfordjournals.org).
|
ANALYTICAL TECHNIQUES
|
|---|
Five bulk-rock and 43 powdered samples were available for trace
element and isotopic study. The generally low range of loss
on ignition values (0·39–1·25 wt %) for
the post-caldera rocks with the exception of three samples (2·24,
2·26 and 4·39 wt % LOI) indicates that most of
the samples are fresh. Four of the five caldera samples with
reported major element data (Electronic Appendix 1) have high
LOI values and were not considered suitable for trace element
and isotopic analysis.
Trace element concentrations in 43 IVC rocks were determined using a PerkinElmer ELAN 6000 quadrupole inductively coupled plasma mass spectrometry (ICP-MS) system at Durham University. Full details of the analytical procedure and instrument operating conditions have been given by Ottley et al. (2003) and Handley (2006). Calibration of the ELAN was achieved via the use of in-house standards and international reference materials [W2, BHVO-1, AGV1, BE-N and BIR1, using the accepted standard values given by Potts et al. (1992)], together with procedural blanks (three per batch). Internal and external reproducibility of standard values are generally less than 3% relative standard deviation (Ottley et al., 2003). Comparison of Zr and Ti elemental concentrations in IVC rocks obtained by X-ray fluorescence (Sitorus, 1990) and ICP-MS (this study) show excellent agreement, with mean relative differences of 6% and 7%, respectively.
Sr, Nd and Hf isotope analysis was undertaken at the Arthur Holmes Isotope Geology Laboratory (AHIGL), Durham University. The sample dissolution procedure and separation of Hf and Nd from rock samples is based on that presented by Dowall et al. (2003). The chemical separation procedure for Sr in this study follows that outlined by Charlier et al. (2006). Sr, Nd and Hf fractions were measured for isotope ratios using the AHIGL ThermoElectron Neptune multi-collector ICP-MS system. Detailed instrument operating conditions (including cup configurations and interference corrections) have been given by Dowall et al. (2003) and Nowell et al. (2003). Samples were analysed on several separate occasions. Data quality was monitored over the separate sessions by frequent analysis of standard reference materials throughout each run. Measured values for the NBS 987, J&M and JMC 475 standards ± 2SD error over the period of study were 0·710262 ± 11 (n = 53), 0·511106 ± 10 (n = 74) and 0·282146 ± 4 (n = 51), respectively. Data are plotted relative to NBS 987, J&M and JMC 475 standard values of 0·71024 (Thirlwall, 1991), 0·511110 (Royse et al., 1998) and 0·282160 (Nowell et al., 1998), respectively. Total procedural blanks for Sr, Nd and Hf were determined by ICP-MS on the PerkinElmer ELAN 6000 quadrupole ICP-MS system at Durham University and were below 1·2 ng for Sr, 219 pg for Nd and 73 pg for Hf. These values are considered insignificant in relation to the quantity of Sr, Nd and Hf typically processed from IVC rocks (12 µg, 3 µg and 500 ng, respectively).
Oxygen isotope analyses of mineral separates were performed by laser-fluorination at Royal Holloway, University of London using the procedures outlined by Macpherson et al. (2000). Oxygen results are reported as per mil deviations relative to the standard mean ocean water (SMOW) standard. In-house standard values of SC olivine 2 and GMG II over the period of study were +5·24
± 0·04 (1
, n = 6) and +5·69 ± 0·08 (1, n = 15); within 0·01 of accepted values. Precision and accuracy within runs were similar to the overall standard results. Oxygen yields were greater than 98% for olivine, 93% for plagioclase and between 92 and 94% for clinopyroxene. Although the yields for plagioclase and clinopyroxene are slightly low, they were consistent over the three sessions. Replicate analyses of olivine, plagioclase and clinopyroxene were within 0·06, 0·09 and 0·04, respectively.
|
GEOCHEMISTRY
|
|---|
Major element data
Major element data are reported in Electronic Appendix 1 and
illustrated in
Fig. 3. Although SiO
2 is commonly used as the
index of differentiation for typically evolved arc lavas (Davidson
et al., 2005
), MgO was chosen here as (1) IVC includes a large
number of relatively less evolved (SiO
2 < 55 wt %) rocks,
and (2) there appears to be some division in SiO
2 contents between
the CR and IC post-caldera groups at lower MgO content (
Fig. 3a).
Lavas from IVC display a broad range in SiO
2 content from 48
to 63 wt % and relatively low MgO contents (< 5·8
wt %) typical of island arc volcanoes, which suggests that no
true, primary basalts were erupted within the complex. MgO correlates
positively with Fe
2O
3, TiO
2 and CaO, whereas Na
2O, SiO
2 and
K
2O generally increase with decreasing MgO. Al
2O
3 and P
2O
5 data
are considerably more scattered, especially at lower MgO contents
(
Fig. 3).
The IC samples form a coherent, linear array in the majority
of variation diagrams, with a narrower range of SiO
2 and MgO
than for the CR volcanic rocks. IC and CR rocks generally appear
to follow different trends on plots of SiO
2, Al
2O
3, K
2O and
CaO against MgO (
Fig. 3a–d), with the exception of four
CR samples (KI 202, KI 136, KI 164 and KI 108) that follow the
IC trend. The IC group has higher SiO
2 and K
2O contents than
the CR group at similar MgO and displays strong negative correlations,
whereas within the CR group there is a weaker and less obviously
negative correlation between SiO
2 and K
2O and MgO (
Fig. 3a and
3c). The more differentiated CR rocks (below
4 wt % MgO) have
elevated CaO and Al
2O
3 contents when compared with the IC samples
at the same degree of differentiation (
Fig. 3b and
3d). Samples
from Kawah Ijen and Blau appear to straddle the two trends in
the majority of the variation diagrams (e.g.
Fig. 3b–d).
Trace element geochemistry
Ni (<40 ppm) and Cr (<60 ppm) concentrations are low in IVC rocks (Table 1) and Ni is positively correlated with MgO (e.g. Fig. 4). There is a large decrease in Ni (and Cr) abundance from the most mafic rocks to those with intermediate MgO contents, whereas the more evolved lavas display more constant concentrations. Large ion lithophile element (LILE; Rb, Cs, Ba and including U and Th) concentrations (e.g. Fig. 4b and 4c) generally increase with decreasing MgO. Variation of Rb with MgO mirrors that of K2O (Fig. 3c), where IC group contents are higher than the relatively constant concentrations in the CR group. Sr concentrations, on the other hand, are more elevated in the CR group and generally increase with decreasing MgO, whereas Sr concentrations in the IC rocks are in the main lower at similar MgO (Fig. 4d); Al2O3 and CaO exhibit similar behaviour. The IC samples have higher abundances of high field strength elements (HFSE), such as Zr, Hf, Nb and Ta, than the CR rocks at a given MgO (Fig. 4e and 4f). The variation of these elements also mirrors that of K2O. Rare earth element (REE) variation with MgO is illustrated in Fig. 4g and 4h. Light REE (LREE, e.g. La) concentrations in the IVC rocks generally increase with decreasing MgO (Fig. 4g); however, heavy REE (HREE) variation trends are more difficult to discern; there is a wide variation in Yb at low MgO content (Fig. 4h).
Mid-ocean ridge basalt (MORB)-normalized trace element abundance
patterns (
Fig. 5) are typical of island arc volcanoes, with
enrichment of the more mobile LILE and LREE relative to HFSE
and HREE. The highly to moderately incompatible elements are
all enriched relative to MORB, similar to those from other Sunda
arc volcanoes (Turner & Foden, 2001
) and local sediments
(Vroon
et al., 1995
). The IC group are generally more enriched
than the CR group. This enrichment is emphasized in the inset
diagram of
Fig. 5, where the IC are normalized to the CR group.
There is a general decrease in the degree of enrichment of the
IC as element incompatibility decreases, with the exception
of Ba and Sr, which are considerably less enriched relative
to their neighbouring elements, and in the case of Sr, which
is actually depleted relative to the CR group. There is greater
enrichment of Zr and Hf in the IC group compared with the more
compatible elements, whereas the Ti concentrations of the IC
and CR rocks are identical.
Radiogenic isotopes
The Sr–Nd–Hf isotope compositions of the IVC samples
are listed in
Table 1 and plotted in
Fig. 6. In contrast to
their major and trace element compositions, the intra-caldera
and caldera-rim groups cannot be distinguished from one another
on the basis of radiogenic isotopes. The relatively large dataset
displays a restricted range in
87Sr/
86Sr (0·704169–0·704483),
143Nd/
144Nd (0·512814–0·512895) and
176Hf/
177Hf
(0·283078–0·283133), and plots within the
range of other Indonesian island arc volcanoes. The Sr and Nd
isotope ratios of lavas from the complex are among the least
radiogenic Sr and most radiogenic Nd isotopic values reported
for Javanese volcanoes (
Fig. 6a); only samples from Guntur (Edwards,
1990
) and Galunggung (Gerbe
et al., 1992
) have higher
143Nd/
144Nd
and lower
87Sr/
86Sr ratios. The Sr isotope ratios of the IVC
lavas are consistent with the previously recognized eastward
decrease in
87Sr/
86Sr from West Java to Bali (Whitford, 1975
).
In Nd–Hf isotope space the IVC lavas lie on the edge of
the Java field close to the Indian Ocean MORB field (
Fig. 6b).
The pre-caldera group has relatively low
87Sr/
86Sr ratios (0·704169–0·704238)
and are indistinguishable from one another in
143Nd/
144Nd (0·512892–0·512895).
However, their range in
176Hf/
177Hf (0·283081–0·283118)
is comparable with the range of Hf isotope ratios seen in the
volcanic complex as a whole (
Fig. 6b inset). IC
143Nd/
144Nd
values are similar to those of the CR group with no obvious
division between the two groups. The IC lavas display a more
limited range in
176Hf/
177Hf ratios, but plot in the centre
of the larger spread of the CR group (
Fig. 6b inset).
View larger version (44K):
[in this window]
[in a new window]
[Download PowerPoint slide]
|
Fig. 6. (a) Variation of 143Nd/144Nd vs 87Sr/86Sr for the IVC. Inset: Nd and Sr isotope diagram for the IVC groups. The lack of any distinction in isotope ratio between the IC and CR groups should be noted. Data sources: I-MORB: Price et al. (1986), Ito et al. (1987), Rehkämper & Hofmann (1997) and Chauvel & Blichert-Toft (2001); N-MORB: Ito et al. (1987) and Chauvel & Blichert-Toft (2001); altered oceanic crust (AOC): Staudigel et al. (1995); other Java volcanic rocks: Whitford et al. (1981), White & Patchett (1984), Edwards (1990, including Guntur tholeiites), Gerbe et al. (1992, Galunggung) and Gertisser & Keller (2003); Indian ocean sediments: Ben Othman et al. (1989) and Gasparon & Varne (1998). (b) Variation of 176Hf/177Hf vs143Nd/144Nd for the IVC. Inset: Hf–Nd isotope diagram of IVC volcanic rocks separated by eruptive group. Data sources: I-MORB: Salters (1996), Nowell et al. (1998) and Chauvel & Blichert-Toft (2001); N-MORB: I-MORB references listed for (a) plus Salters & Hart (1991); OIB: Patchett & Tatsumoto (1980), Patchett (1983), Stille et al. (1986), Salters & Hart (1991), Nowell et al. (1998) and Salters & White (1998); Java: White & Patchett (1984) and Woodhead et al. (2001); Banda: White & Patchett (1984); Indian ocean sediments: White et al. (1986), Ben Othman et al. (1989) and Vervoort et al. (1999). Dividing line for Indian and Pacific MORB provenance from Pearce et al. (1999). Maximum 2 external errors for each geographical group are shown in the inset diagrams.
|
|
Oxygen isotope data
The
18O values of clinopyroxene, olivine and plagioclase are
+5·38 to +5·58
(
n = 5), +5·02
(
n = 1) and
+6·08
(
n = 1), respectively (
Table 2). The range of clinopyroxene
18O values is narrow and comparable with the average mantle
18O value of +5·57 ± 0·32
(Ionov
et al.,
1994
; Mattey
et al., 1994
). IVC
18O clinopyroxene values are
lower than those of plagioclase and higher than those of olivine;
with coexisting mineral pair
cpx–ol and
plag–cpx values of 0·36
and 0·5
, respectively, suggesting
isotopic equilibrium at typical magmatic temperatures for andesite
liquids (Macpherson & Mattey, 1998
; Macpherson
et al., 1998
).
The limited dataset displays no noticeable difference in the
cpx
18O values of the different geographical groups.
The oxygen isotope ratios of clinopyroxene phenocrysts from
IVC rocks are within the range reported for clinopyroxene from
the primitive Galunggung lavas from West Java (+5·3 to
+5·6
, Harmon & Gerbe, 1992
), and clinopyroxene and
olivine
18O values lie at the lower end of the Banda arc
18O
range (+5·18 to +7·04
and +4·92 to +5·59
,
respectively; Vroon
et al., 2001
). The
18O value for the only
plagioclase sample analysed from the IVC is within error of
the
18O range of Galunggung lavas (+5·6 to +6·0
;
Harmon & Gerbe, 1992
) and is lower than
18O values recorded
in plagioclase phenocrysts from Merapi volcanic rocks (+6·5
to +7·00
, Gertisser & Keller, 2003
). The
18O values
of the IVC rocks also are close to the values postulated for
the upper mantle in other subduction zones (Smith
et al., 1996
;
Thirlwall
et al., 1996
; Macpherson & Mattey, 1998
; Macpherson
et al., 1998
, 2000
; Eiler
et al., 2000
).
|
DISCUSSION
|
|---|
Magmatic differentiation processes
Differentiation processes can significantly change the composition
of magmas as they rise through the lithosphere towards the surface.
It is important to identify the impact of these processes on
the composition of the magma to remove uncertainty in establishing
the composition and magnitude of slab-derived contributions
to the mantle source in subduction zones. Low MgO, Ni and Cr
abundances in IVC volcanic rocks indicate that they are not
primary mantle melts, and that magma compositions were, therefore,
modified
en route to the Earth's surface.
Correlations of various major and trace elements with indices of differentiation (i.e. MgO) in IVC lavas (Figs 3 and 4) suggest that the concentrations of some elements are controlled by shallow-level differentiation processes such as fractional crystallization, magma mixing or contamination. Major and trace element variations for most of the post-caldera samples can be described as one of two trends, which are exemplified in the CaO vs MgO diagram (Fig. 3d), where a low-Ca group, dominantly consisting of the IC samples, displays a positive correlation, and a high-Ca group, composed of most CR samples along with some Kawah Ijen and Blau samples, shows little change in CaO with decreasing MgO. Similar trends are apparent in plots of Al2O3 and Sr against MgO (Figs 3b and 4d). Higher modal plagioclase abundances (10%, Table 3), and a less pronounced negative Eu anomaly in high-Ca lavas (Fig. 7) suggest that the different geochemical trends exhibited by the low-Ca and high-Ca groups may be due to more extensive plagioclase fractionation in the former, or plagioclase accumulation in the latter. Possible evidence for open-system processes in the petroenesis of the IVC lavas includes: coexistence of phenocrysts (plagioclase and clinopyroxene) that display normal, oscillatory and reverse zoning (Sitorus, 1990); clinopyroxene overgrowths on orthopyroxene phenocrysts; and bimodal distributions in plagioclase core compositions (Sitorus, 1990, see Electronic Appendix 2; Berlo 2001). One group of plagioclase crystal cores cluster below An60, and the other generally above An80. However, bimodal plagioclase compositions are found within volcanic rocks from both the high- and low-Ca groups, indicating that although mixing may occur, it is not the dominant control on chemical variation, otherwise we would only expect to find bimodal plagioclase populations in high-Ca rocks. Also, higher modal plagioclase abundance (Table 3; Electronic Appendix 3) does not necessarily indicate accumulation; it may just be a result of a greater degree of crystallization or more plagioclase in the modal equilibrium assemblage. Variations in the fractionating mineral assemblage can produce contrasting differentiation trends on variation diagrams (e.g. Davidson, 1996). Crystallization vectors drawn in Fig. 3b and 3d illustrate the magmatic evolution predicted for primitive IVC magma during fractionation of different mineral assemblages. The different trends observed in the high- and low-Ca groups can result from differences in the phases that make up the fractionating assemblage and their proportions. Assemblages consisting of clinopyroxene and olivine ± plagioclase can replicate the high-Ca array (Fig. 3b and 3d), whereas the same assemblage but with more plagioclase fractionation relative to clinopyroxene and olivine is suggested for the low-Ca trend (see labelled arrows, Fig. 3b and 3d). Less plagioclase fractionation in the evolution of the high-Ca lavas, relative to the low-Ca lavas, is also consistent with the variations of Al2O3, Sr and Ba with respect to MgO, along with a smaller negative Eu anomaly.
Fractional crystallization
The XLFRAC, least-squares major element modelling technique
(see Stormer & Nicholls, 1978
) was used to ascertain whether
the high-Ca and low-Ca differentiation trends can be explained
by differences in the amount of plagioclase fractionation, and
also to determine whether compositional variation within and
between geographical groups (e.g. IC, CR, Kawah Ijen) can be
explained by fractional crystallization. Models 1 and 2 in
Table 3 suggest that the most evolved post-caldera rocks of the high-Ca
and low-Ca groups can be produced by fractionation of plagioclase,
clinopyroxene, olivine and Fe–Ti oxide from the least
evolved samples of their respective groups (
r2 = 0·2
and 0·19, respectively). Models of fractionation to other
high-Ca and low-Ca group daughter compositions (models 3–6
and 7–9, respectively) yield excellent
r2 values (<0·16).
All of these models also suggest that significantly less plagioclase
fractionation is required from the parent magma to reach high-Ca
daughter compositions (10–12%) compared with those with
low-Ca (
20–37%). Therefore, we suggest that the different
(high- and low-Ca) trends (e.g.
Fig. 3b and
3d) are due to a
greater amount of plagioclase fractionation in the IC rocks
than in the CR rocks.
Least-squares analysis demonstrates that intra-group CR fractionation models yield low r2 values (models 2–9, Table 3). A good solution (r2 = 0·02) is also obtained for a model using a less evolved IC sample as parent to a more evolved IC rock (model 10). Low r2 values are obtained in model 11, between the two Kawah Ijen samples, which straddle the high-Ca and low-Ca trends on several major element diagrams. Some inter-group models can also generate very acceptable results (models 13–15), suggesting that members of the different geographical groups can be related to each other through fractional crystallization. However, least-squares modelling cannot generate low r2 in models where IC rocks represent the parent magma composition and CR is the daughter, unless crystals are accumulated rather than removed (models 16 and 17).
Utilizing the phase proportions and degree of crystallization predicted from the major element modelling (Table 3) it is possible to test the conclusions of least-squares analysis by modelling trace element concentrations using the Rayleigh fractionation equation Cl = CoF(D–1), where Cl and Co represent the concentration of an element in the daughter and parental liquids, respectively, F is the fraction of liquid remaining and D is the bulk distribution coefficient. The distribution coefficients used in modelling are given in Table 4. The results of selected trace element models are given in Table 5 and show excellent agreement between calculated and observed daughter concentrations. The strongest agreement (model 10) is between the intra-caldera samples KI 92 and KI 34 in the low-Ca group, where calculated values lie within 10% of the measured concentrations. In summary, trace element modelling validates models of fractional crystallization developed from major element data; this suggests that fractional crystallization is the main control on melt evolution at IVC. It also confirms that the separate trends of the low-Ca and high-Ca groups are likely to be due to variable influence of plagioclase in the fractionating mineral assemblage and are not a result of plagioclase accumulation in the high-Ca group.
Structural controls on magma ascent and storage
The location of the CR volcanoes around the caldera rim suggests
that ring fractures may facilitate the movement of magma below
these volcanoes. This has been proposed for other post-caldera
vents located along radial fractures within volcanic complexes,
such as Roccamonfina volcano in Italy (Giannetti, 2001
) and
Chichontepec volcanic centre in El Salvador (Rotolo & Castorina,
1998
). The linear orientation (roughly NE–SW, Berlo, 2001
)
of the intra-caldera volcanoes also suggests that magma below
the central part of the caldera is utilizing lines of weakness
and further highlights the structural control on the location
of post-caldera volcanoes at IVC.
The different fractionation trends exhibited by the low-Ca (dominantly the intra-caldera rocks) and high-Ca (dominantly the caldera-rim rocks) suggest that spatial variations in chemistry within the volcanic complex might be linked to sub-volcanic structure. The contrast in geochemistry is proposed to result from differences in the amount of plagioclase fractionation from the respective magmas. Experimental evidence (Grove et al., 2003) suggests that plagioclase crystallization is suppressed by high water contents and high pressure in basaltic andesite magmas. Therefore, the more extensive plagioclase fractionation inferred for the IC group magmas could result from either shallower level (i.e. lower pressure) storage or lower water contents (or both) in the IC magmas relative to the CR magmas. The volatile content of magmas is also controlled by pressure; thus, the suppression of plagioclase crystallization in the CR group could be a result of magmatic differentiation at deeper levels in the crust below the caldera-rim volcanoes (Fig. 8). Magmas stored at greater depth, with higher volatile contents, might be expected to erupt more explosively than those stored at shallower levels, from which eruptions might be more passive. The presence of large stratovolcanoes on the caldera-rim and dominantly small cinder cones inside the caldera at IVC is consistent with the type of eruption, compatible with this interpretation. Therefore, at IVC it is proposed that volcanic structure exerts some control on the depths at which the rising magma can pond. Kawah Ijen erupts lavas belonging to both the low-Ca and high-Ca trends. Kawah Ijen lies at the intersection between the caldera rim and the northeastern edge of the intra-caldera lineation (Fig. 2) and could, therefore, tap both shallow and deeply ponded magmas (Fig. 8). There is no temporal division in the production of high-Ca and low-Ca lava types at Kawah Ijen, and no evidence for physical mingling or mixing (Berlo, 2001), implying separate magmatic pathways for magmas erupted from different depths (Fig. 8).
View larger version (22K):
[in this window]
[in a new window]
[Download PowerPoint slide]
|
Fig. 8. Simplified diagram of magma petrogenesis at IVC, focusing on crustal level differentiation. Cross-section shown is oriented roughly parallel to the lineation of the intra-caldera volcanoes. The diagram shows the envisaged deeper level storage of magma below the CR volcanoes, and shallower level storage plus more extensive plagioclase fractionation below the IC volcanoes. Kawah Ijen is thought to tap both the shallow and deep reservoirs and transport the magma through unconnected pathways. It should be noted that Fe–Ti oxide is also fractionating in both the shallow and deep reservoirs.
|
|
Role of crustal contamination
There is considerable evidence from other island arcs (Thirlwall
& Graham, 1984
; Davidson
et al. 1987
; Ellam & Harmon,
1990
; Davidson, 1996
; Smith
et al., 1996
; Thirlwall
et al.,
1996
; Macpherson
et al., 1998
) for the contamination of primary
magmas by the arc crust. Contamination is also believed to be
an important process, responsible for modifying isotope ratios,
in the western Sunda arc (Gasparon
et al., 1994
; Gasparon &
Varne, 1998
) and at Sangeang Api volcano in the East Sunda arc
(Turner
et al., 2003
). Therefore, prior to discussing magma
source compositions it is important to assess the role of crustal
contamination at IVC. The absence of correlations between IVC
Sr, Nd and Hf isotope ratios and indices of differentiation
(e.g. MgO,
Fig. 9) and the restricted ranges of Sr, Nd and Hf
isotope ratios in IVC samples (
Table 1) are consistent with
a negligible input of isotopically distinct crust during differentiation.
Clinopyroxene
18O values in the IVC lavas are homogeneous (+5·38
to +5·58
,
Table 2) and low, lying within the range of
mantle
18O values (+5·57 ± 0·32
) reported
by Mattey
et al. (1994
) and Ionov
et al. (1994
). Therefore,
it is unlikely that the lavas have been contaminated by upper
crustal materials, which typically possess high
18O values,
such as those reported by Gertisser & Keller (2003
) for
the local upper crust (calcareous sediments) in Central Java
(+20·5
and +18·9
). Furthermore, only igneous cumulate
xenoliths have been found within the Ijen complex lavas; continental-type
crustal xenoliths have not been detected (Sitorus, 1990
; Berlo,
2001
). Monomineralic plagioclase xenoliths are present in lavas
erupted from Anyar (IC) and Blau, and cumulate xenoliths containing
plagioclase, olivine, clinopyroxene and Ti-magnetite have been
found in Rante (CR) and Kawah Ijen lavas. A cumulate containing
olivine, ortho- and clinopyroxene was also found in lava from
Kukusan (Berlo, 2001
).
Similar to conclusions reached in studies of other Sunda arc
volcanoes (Gerbe
et al., 1992
; Elburg
et al., 2002
; Gertisser
& Keller, 2003
), the evidence above suggests that any interaction
of IVC magmas with the arc crust during differentiation has
had a negligible impact on their geochemistry. Fractional crystallization,
therefore, appears to be the dominant differentiation process
controlling geochemical variations in IVC lavas. Higher Sr and
lower Nd and Hf isotope ratios in IVC rocks relative to MORB
point towards contamination of the IVC source by an isotopically
distinct crustal component.
Magma source components
Most models of magma petrogenesis at island arcs involve three main source components: (1) the mantle wedge; (2) the subducting slab (oceanic crust and associated sediments); (3) the arc lithosphere. The majority of island arc magmas are thought to originate in the mantle wedge (Ringwood, 1974; Ellam & Hawkesworth, 1988; McCulloch & Gamble, 1991), which has been inferred by several workers to be similar to the source of MORB (Davidson, 1987; Woodhead et al., 1993; Gamble et al., 1996; Turner et al., 2003). It has been proposed that there might be contribution from an enriched mantle component [ocean island basalt (OIB) source] in some Sunda arc lavas (Wheller et al., 1987; Edwards et al., 1991, 1993; van Bergen et al., 1992), although a MORB-source-like mantle source has been advocated for the Sunda arc by others (e.g. White & Patchett, 1984; Turner & Foden, 2001; Elburg et al., 2002). Helium isotope values of olivine crystals in mantle xenoliths and island arc volcanic rocks also implicate a MORB-source-like mantle source in the western Sunda arc (Hilton & Craig, 1989; Gasparon et al., 1994).
Mantle source characteristics
The HREE and HFSE are thought to remain relatively immobile compared with other elements, such as LILE, during slab dehydration (Tatsumi et al., 1986; Kessel et al., 2005) and abundances are generally too low in oceanic and continental crust (Taylor & McLennan, 1985) to significantly alter ratios of these elements in magmas derived from the mantle. We can use these trace elements, therefore, to help ascertain the pre-subduction composition of the mantle wedge. HREE concentrations are 10–15 times chondrite values in IVC lavas and display relatively flat profiles (Fig. 10). This suggests that garnet is not an important residual mineral in the source region and that the IVC magmas are derived largely from a shallow mantle source, above the garnet–spinel transition for wet peridotite. The La/Lu ratios of CR (41) and IC (44) lavas are very similar, which is consistent with a similar source mineralogy and degree of partial melting for these two post-caldera groups. The large difference in La/Lu ratios within the three pre-caldera group samples (34–80), when compared with the fairly constant ratio of the post-caldera volcanics (42) might be due to variable degrees of partial melting and/or different source mineralogies for the pre-caldera rocks. If post-caldera rocks from the IC and CR volcanoes, Kawah Ijen and Blau share the same source then they should possess similar ratios of immobile trace elements generally assumed to be unmodified by subduction processes (Table 6). Zr/Nb [17–23 (except one CR sample at 14)] and Ta/Nb (0·07–0·13) ratios do not change significantly with differentiation (Fig. 11a and 11b) and are relatively homogeneous in all IVC eruptive rocks. Zr/Nb ratios are similar to those of MORB [normal and Indian type (N-MORB and I-MORB)] whereas Ta/Nb ratios are comparable with both MORB and OIB (Table 6; e.g. Fig. 11). The similarity between HFSE and HREE concentrations in primitive IVC basalts and I-MORB, rather than N-MORB, is also apparent in Fig. 12. Therefore, we suggest that the mantle wedge beneath the IVC is similar to the source of I-MORB. Several studies utilizing Pb isotopic and trace element data have reached a similar conclusion for magmatism in SE Asia throughout the Cenozoic (Taylor et al., 1994; Hickey-Vargas, 1998; Macpherson & Hall, 2001, 2002; Elburg et al., 2002; Macpherson et al., 2003). The HFSE and HREE concentrations and ratios of IVC lavas also indicate that the mantle wedge is not significantly depleted beneath East Java, in contrast to others arcs such as Izu–Bonin and Mariana (Fig. 12), where mantle sources are thought to have experienced melt extraction prior to their involvement in arc petrogenesis (Woodhead et al., 1993; Elliott et al., 1997; Taylor & Nesbitt, 1998).
Sr is fluid-mobile during slab dehydration, therefore, Sr isotope
data are not well suited for identifying the isotopic composition
of the precursor mantle wedge. Experimental work (Tatsumi
et al., 1986
; Brenan
et al., 1995
; You
et al., 1996
) and studies
of arc lavas (McCulloch & Gamble, 1991
; Pearce & Peate,
1995
; Münker
et al., 2004
) indicate that Hf and Nd are
relatively immobile during slab dehydration, particularly with
respect to the formation of aqueous fluids (Kessel
et al., 2005
).
However, Woodhead
et al. (2001
) have cast doubt upon the status
of Hf as a truly fluid-immobile element; they see evidence in
the New Britain subduction system for the transport of Hf in
aqueous fluids when Nd is immobile. Their observations are based
on the contrast in Hf isotope ratios of the mantle wedge beneath
New Britain and the subducting crust of the Woodlark Basin.
As we advocate an I-MORB composition for both the mantle wedge
and the down-going subducted crust, any Hf mobility should have
negligible i
TOP
ABSTRACT
INTRODUCTION AND SETTING
