Journal of Petrology Advance Access originally published online on March 11, 2005
Journal of Petrology 2005 46(7):1367-1391; doi:10.1093/petrology/egi019
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Structure and Dynamics of a Silicic Magmatic System Associated with Caldera-Forming Eruptions at Batur Volcanic Field, Bali, Indonesia
VICTORIAN INSTITUTE OF EARTH AND PLANETARY SCIENCES, SCHOOL OF GEOSCIENCES, MONASH UNIVERSITY, MELBOURNE, VIC. 3800, AUSTRALIA
RECEIVED NOVEMBER 18, 2003; ACCEPTED JANUARY 27, 2005
 |
ABSTRACT |
|---|
TOP
ABSTRACT
INTRODUCTIONThe Batur volcanic field (BVF), in Bali, Indonesia, underwent two successive caldera-forming eruptions that resulted in the deposition of silicic ignimbrites. The magmas erupted during and between these eruptions show a broad range of compositions from low-SiO2 andesite to high-SiO2 dacite. On the basis of their geochemistry and mineralogy these magmas may be assigned to six groups: (1) homogeneous andesites with phenocryst compositions essentially in equilibrium with the whole-rock composition; (2) remobilized crystal-rich low-SiO2 andesites with resorbed phenocrysts in equilibrium with the whole-rock composition; (3) mixed low-SiO2 dacite with a relatively large range of phenocryst compositions, with most phenocrysts slightly too evolved to be in equilibrium with the whole-rock; (4) extensively mixed low-SiO2 dacites with a very large and discontinuous range of phenocryst compositions, with most phenocrysts either more Mg-rich or more evolved than the equilibrium compositions; (5) remobilized crystal-rich low-SiO2 dacites with resorbed and euhedral phenocrysts; (6) homogeneous high-SiO2 dacites lacking evidence for magma mixing and showing narrow ranges of phenocryst compositions in equilibrium with the whole-rock composition. This range of silicic magmas is interpreted to reflect a combination of closed- and open-system fractional crystallization, magma mixing and remobilization of cumulate piles by heating. The variety of magmas erupted simultaneously during the caldera-forming eruptions suggests that the magmatic system consisted of several independent reservoirs of variable composition and degree of crystallization. The magmatic evolution of individual reservoirs varied from closed-system fractional crystallization to fully open-system evolution, thereby resulting in simultaneous production of magmas with contrasted compositions and mineralogy. Extensive emptying of the magmatic system during the caldera-forming eruptions led to successive or simultaneous eruption of several reservoirs.
KEY WORDS: caldera; ignimbrite; magmatic chambers; magma mixing; petrology; Sunda Arc
|
INTRODUCTION |
|---|
Volcanic activity at supra-subduction zone volcanoes is dominantly characterized by eruption of small to moderate volumes (<5 km3) of basaltic to high-silica andesitic magmas. However, eruptions involving large volumes (tens to hundreds of km3) of silicic magmas associated with caldera collapse are not uncommon (e.g. Bacon, 1983
; Aramaki, 1984; Heiken & McCoy, 1984; Druitt et al., 1989; Allen, 2001) and can have dramatic outcomes as illustrated by the 1400 BC Minoan eruption of Santorini (Heiken & McCoy, 1984), the 1815 Tambora (Self et al., 1984) and the 1883 Krakatau eruptions (Self & Rampino, 1981). Understanding the processes that control the genesis and evolution of these large-volume silicic magmatic systems at arc volcanoes is important in terms of hazard assessment.
The large volume of magma erupted during caldera-forming eruptions provides an extensive sample of the subvolcanic magmatic system. In addition, because of the almost instantaneous nature of these eruptions, the magma erupted represents a snapshot of the state of the magmatic system just prior to the eruption. As a result, the deposited ignimbrites are ideal for studying the characteristics of the associated silicic magmatic systems. A large number of studies of this type have been carried out (e.g. Foden, 1986; Bacon & Druitt, 1988; Mandeville et al., 1996), establishing the complexity of these systems and the importance of composition, temperature, crystallinity and magma mixing in the evolution of the magmatic systems. Nevertheless, the origin of large volumes of silicic magmas and the role of fractional crystallization of mantle-derived mafic magmas, crustal anatexis and/or remelting of crustal intrusive bodies in their petrogenesis remains a matter of debate. It is generally assumed that these silicic magmas evolve within a single large magma chamber. However, in the light of recent studies that suggest that the magmatic systems at arc volcanoes are essentially open systems, comprising several interconnected small reservoirs (e.g. Gamble et al., 1999; Dungan et al., 2001; Streck et al., 2002; Reubi & Nicholls, 2005), this may be questionable.
This paper presents a detailed study of the mineralogy and geochemistry of the andesitic to dacitic magmas erupted at Batur volcano, Bali, Indonesia, during two successive catastrophic, caldera-forming, eruptions, and of andesitic to dacitic lava flows that were emplaced within the caldera between the two collapse events. The aim is to establish the characteristics of the magmatic system prior to the two caldera-forming eruptions and to investigate the processes that controlled the genesis and evolution of the voluminous silicic magmatic reservoir.
|
GEOLOGY OF BATUR VOLCANIC FIELD |
|---|
The late Quaternary Batur volcanic field (BVF), situated in northern Bali, Indonesia (Fig. 1) is part of the Sunda arc system, which is associated with northward subduction of the Indo-Australian plate beneath the Eurasian plate (Hamilton, 1979
). The BVF is located about 150 km above the Benioff zone and adjacent to the active Agung volcano and the extinct or dormant Bratan caldera. The crust beneath Bali is about 20 km thick, with an oceanic velocity structure (Curray et al., 1977). Most of the island consists of Tertiary to Quaternary volcanic rocks, with the southern part formed from uplifted coral reefs of Pliocene–Pleistocene age (Purbo-Hadiwidjojo, 1971; Kadar, 1977).
|
The BVF comprises two well-formed nested calderas and an active cone built within the smaller caldera (Fig. 1). Volcanic activity within the BVF may be subdivided into six main periods (Fig. 1) (Reubi & Nicholls, 2004
), as follows.
- Building of a basaltic to dacitic stratovolcano (Penulisan volcano) and a small parasitic cone on the SE flank. This activity started at least 510 kyr BP (Wheller & Varne, 1986).
- Collapse of the first caldera (CI), associated with the eruption of dacitic ignimbrite (Ubud Ignimbrite) dated by 14C at 29 300 years BP (Sutawidjaja, 1990).
- Formation of an andesitic to dacitic lava-dome complex (Bunbulan lava-dome complex) and a small dacitic tuff cone (Payang tuff cone) within CI.
- Collapse of the second caldera (CII), again accompanied by eruption of andesitic to dacitic ignimbrite (Gunungkawi Ignimbrite), dated by 14C at 20 150 years BP (Sutawidjaja, 1990).
- Andesitic to dacitic explosive activity within CII, producing pyroclastic fall deposits [Peneloken and Penulisan fall deposits of Sutawidjaja (1990)].
- Building of the historically active, 1700 m high, basaltic andesite Batur stratovolcano within CII.
Detailed descriptions of the two ignimbrites have been given by Reubi & Nicholls (2004); only the key characteristic are summarized below.
The Ubud Ignimbrite consists dominantly of pyroclastic flow deposits with minor pumice fall deposits. The intra-caldera succession comprises up to 16 non-welded to densely welded pyroclastic flow units. The outflow succession covers most of southern Bali (Fig. 1) and comprises up to five flow units. The deposits are typically ash-rich, lithic clast-poor and contain 2–15% of pumices. Welding grades from incipient in distal sections to partially welded in more proximal settings. The pumices show a range of textures from highly vesicular (up to 75 vol. %) to black glassy fiamme but have a consistent mineralogy. There is no stratigraphic correlation between the intracaldera and the outflow successions, suggesting that the latter record an earlier phase of eruption, the products of which are buried beneath the observed succession within the caldera. The total volume of observed deposits is 18 km3 [13 km3 Dense Rock Equivalent (DRE)]. However, the volume of the associated caldera suggests that up to 60 km3 could have been erupted (see Reubi & Nicholls, 2004).
The Bunbulan lava-dome complex occurs in the NE sector of CI and forms the NE wall of CII (Fig. 1). The succession consists of six superimposed lava flows with minor intercalated pyroclastic flow and fall deposits. The dacitic lava flows at the base of the succession are thick (up to 30 m), massive and non-vesicular whereas the higher andesitic flows are thin (5–6 m), non- to moderately vesicular and have brecciated bases.
The Gunungkawi Ignimbrite intra-caldera succession consists of interbedded accretionary lapilli-bearing ash surge, ash fall, pumice lapilli fall and thin pyroclastic flow deposits, overlain by a thick and massive pyroclastic flow deposit comprising 20–30% of pumices. The outflow succession occurs in central, southern, and northern Bali and comprises a single flow unit, which is underlain by intercalated pumice-rich and ash-rich pyroclastic flow deposits in northern Bali. The deposits are non-welded. Two distinct types of pumices are observed. The first type is grey, moderately to highly vesicular (50–85 vol. %), crystal-poor (2–5%) and occasionally banded, with alternating black and grey bands. The second type is black, moderately vesicular (50–65 vol. %) and moderately crystal-rich (20–25%). The total volume of the observed deposits is 7 km3 (4 km3 DRE) but a volume of up to 9 km3 DRE may be expected from the size of the associated caldera.
The Peneloken and Penulisan fall deposits cover the rims of the CI and CII calderas and the SW flank of BVF (Fig. 1) (Sutawidjaja, 1990). The Peneloken fall deposits consist of interbedded ash-rich, accretionary lapilli-bearing fall deposits and pumice lapilli to bomb fall deposits. The Penulisan fall deposits consist of interbedded, normally graded pumice lapilli to coarse ash fall deposits. Pumices are highly vesicular (70–85 vol. %), crystal poor (<5%), and white to pink in the Peneloken fall deposits and moderately vesicular (50–60 vol. %), crystal poor (<5%) and dark grey to yellow in the Penulisan fall deposits.
|
ANALYTICAL METHODS |
|---|
Whole-rock major element contents were determined by X-ray fluorescence (XRF) spectrometry on fused glass beads at the University of Melbourne, using a Siemens SRS3000 instrument. Trace element contents were determined by inductively coupled plasma-mass spectrometry (ICP-MS) at Monash University on a Finnigan-MAT ELEMENT high resolution instrument. Samples were dissolved by HF–HNO3, HNO3 and HCl acid digestion. Precision for trace elements is typically better than ±2·5%. Mineral analyses were carried out on a CAMECA SX-50 electron microprobe at the University of Melbourne, using ZAF on-line data reduction and matrix correction procedures. The accelerating voltage used was 15 kV and the beam current 25 nA for minerals and 10 nA for the groundmass glass, with 10 s counting time. Analytical uncertainty is typically <1% for major elements.
|
WHOLE-ROCK COMPOSITIONS AND TEMPORAL EVOLUTION |
|---|
A total of 34 samples of the pumices from the two ignimbrites, the Peneloken and Penulisan fall deposits, and the Bunbulan lavas were analysed for their whole-rock compositions. Representative whole-rock major and trace element analyses are listed in Table 1. The complete dataset can be downloaded from the Journal of Petrology website at http://www.petrology.oupjournals.org/. Selected Harker diagrams are plotted in Fig. 2, and Th, Sr and Ba contents, and La/Yb and Zr/Nb ratios are plotted against SiO2 in Fig. 3. The analysed samples have tholeiitic major element compositions and trace element compositions typical of subduction-related magma. They range from medium-K andesite to medium- to high-K dacite [classification of Peccerillo & Taylor (1976)
] (Fig. 2a). The Ubud Ignimbrite pumices range in composition from low-SiO2 to high-SiO2 (65–68 wt % SiO2) dacite (on a volatile-free basis). The Gunungkawi Ignimbrite grey pumices are high-SiO2 andesite to high-SiO2 dacite (60–67 wt % SiO2), whereas the black crystal-rich pumices are low-SiO2 andesite to low-SiO2 dacite (56–63 wt % SiO2). The Bunbulan lavas show a wide range of composition, from andesite to rhyolite (58–71 wt % SiO2) (in the following description and discussion, the only rhyolitic lava, B12, is considered to be part of the high-SiO2 dacite group). The Peneloken and Penulisan fall deposit pumices are high-SiO2 andesite to high-SiO2 (61–68 wt % SiO2) dacite in composition.
|
|
|
Concentrations of compatible elements (e.g. MgO, CaO, P2O5, Sr) decrease rapidly with increasing SiO2 abundance whereas concentrations of incompatible elements (e.g. K2O, Zr, Ba, Th) increase. Al2O3 abundance decreases only slightly with increasing SiO2 abundance. All samples define linear arrays, except two Bunbulan lavas (B52, B12) that have lower K2O and higher MgO and Al2O3 at similar SiO2 contents. Two pumices from the Ubud Ignimbrite densely welded intra-caldera facies show abnormally high contents of K2O that are not correlated with higher contents of incompatible trace elements, except Rb in one of these pumices. These high contents are believed to result from post-depositional vapour-phase alteration.
All samples show similar primitive mantle normalized trace element patterns typical of subduction-related magmas (Fig. 4; Table 1). The magnitudes of the negative Ti anomalies increase with increasing SiO2. Eu and Sr show weak positive anomalies in the low-SiO2 andesites (Eu/Eu* = 1·1–1·0) that progressively become negative anomalies with increasing SiO2 content (Eu/Eu* = 0·9–0·7 in the high-SiO2 dacites). The dacites have incompatible trace element ratios similar to those of the andesites (Fig. 3), but have slightly lower middle to heavy rare earth element (MREE/HREE) ratios (e.g. Dy/Yb ranges from 2·1 in the andesite to 1·5 in the dacite).
|
On a plot of SiO2 content versus the relative stratigraphic positions of the samples within and between the different units, no clear overall trend toward more or less silicic compositions with time is observed (Fig. 5). Within each unit, the most silicic magmas tend to be erupted during the early phases of activity and generally become progressively more mafic with time. Significant and rapid variations are observed within the two ignimbrites and within the Peneloken fall deposits.
|
|
GLASS COMPOSITIONS |
|---|
Matrix glass compositions from the Ubud and Gunungkawi Ignimbrite pumices are presented in Table 2 and the variations in MgO and Al2O3 contents relative to SiO2 are shown in Fig. 6. The glasses show much the same compositional range as the whole-rock compositions. Glasses from the low-SiO2 andesite sample #2.31 and the high-SiO2 dacite sample #2.35 are systematically more SiO2-rich than the whole-rock compositions. The low-SiO2 dacitic pumices (samples #2.18, B16, #2.30 and B59) have glass compositions that range from slightly lower SiO2 to higher SiO2 contents than their respective whole-rock compositions. The two colour-banded, low-SiO2, dacitic pumices (samples #2.30 and B59) show the broadest range of glass compositions. However, no systematic variations in glass composition were observed between the black and grey zones.
|
|
PETROGRAPHY |
|---|
Ubud Ignimbrite
Pumices in the Ubud Ignimbrite are petrographically homogeneous and contain 8–11% phenocrysts (vesicle-free basis) with plagioclase the dominant phase, followed by olivine, clinopyroxene, Ti-magnetite, orthopyroxene and ilmenite (Table 3). Apatite is a common accessory phase present as inclusions within the phenocrysts. Plagioclase typically displays oscillatory zoning (Fig. 7a). Occasional plagioclases with sieve-textured cores containing abundant large glass inclusions are observed within the pumices from the intracaldera succession (low-SiO2 dacite). Crystal clots comprising plagioclase, clinopyroxene, Ti-magnetite, ± orthopyroxene or plagioclase, olivine, Ti-magnetite, ± clinopyroxene are frequent.
|
|
|
Bunbulan lava-dome complex
The Bunbulan lavas are porphyritic with plagioclase, olivine, clinopyroxene, Ti-magnetite and ± ilmenite (4–14% phenocrysts) and have microlitic textures. Plagioclase is the dominant phase in all the lavas and generally displays oscillatory zoning (Fig. 7b). Plagioclases with sieve-structured cores containing abundant glass inclusions are rare in the high-SiO2 dacitic lavas, but are more common in the low-SiO2 dacites and andesites. Clinopyroxene is more abundant than olivine except in the high-SiO2 dacites (Table 3). Clinopyroxene phenocrysts in the andesite B7 are commonly mantled by pigeonite. Ilmenite is present only in the high-SiO2 dacites. Apatite is a common accessory phase. Crystal clots comprising plagioclase, clinopyroxene, Ti-magnetite, ± olivine are common in these lavas.
Gunungkawi Ignimbrite
The grey, occasionally colour-banded, moderately to highly vesicular, high-SiO2 andesitic to dacitic pumices that dominate in the Gunungkawi Ignimbrite contain 6–12% phenocrysts (vesicle-free basis) with plagioclase the dominant phase, followed by olivine, clinopyroxene, Ti-magnetite and ilmenite (Table 3). The majority of plagioclases are optically unzoned (Fig. 7c) or slightly oscillatory zoned. Plagioclases with sieve-textured cores containing large glass inclusions are occasionally observed. Plagioclase commonly occurs in crystal clots with olivine, Ti-magnetite, ± clinopyroxene. Apatite is present as inclusions within the phenocrysts. Angular lithic fragments up to 5 mm in size are common within these pumices and comprise coarse-grained granodiorites, microdiorites, intersertal textured andesites and porphyritic andesites.
The black, moderately vesicular, crystal-rich, low-SiO2 andesitic to low-SiO2 dacitic pumices contain 30–35% phenocrysts of, in order of abundance, plagioclase, olivine, clinopyroxene and Ti-magnetite (Table 3). In the low-SiO2 andesites, the phenocrysts have globular morphologies indicative of extensive resorption. Plagioclases show particularly pronounced resorption textures and have sieve-structured cores full of very large glass inclusions (Fig. 7d). Euhedral phenocrysts are rare. In the low-SiO2 dacite, resorbed phenocrysts are also observed but euhedral crystals are dominant.
Peneloken and Penulisan fall deposits
The high-SiO2 dacitic and high-SiO2 andesitic pumices in the Peneloken fall unit contain <10% phenocrysts of plagioclase, olivine, clinopyroxene, Ti-magnetite, ± ilmenite. Plagioclase typically displays oscillatory zoning. Occasional plagioclases with sieve-structured cores are observed within the high-SiO2 andesitic pumices.
|
MINERALOGY AND MINERAL CHEMISTRY |
|---|
Two samples of pumice from the Ubud Ignimbrite (samples #2.18 and #2.35), four from the Gunungkawi Ignimbrite (samples #2.31, B16, #2.30 and B59), and four samples of lavas from the Bunbulan lava-dome complex (samples B7, B52, B11 and B12) were selected for detailed study of the compositional variation within the phenocryst assemblage. These samples were chosen to typify the ranges of composition and petrography observed within the various units and provide insights into temporal and mineralogical variation within the Batur silicic magmatic system.
Plagioclase
Representative analyses of plagioclase crystals are presented in Table 4. The complete dataset can be downloaded from the Journal of Petrology website. Histograms of plagioclase phenocryst core compositions are shown in Fig. 8. Also illustrated in Fig. 8 are ranges of plagioclase compositions in equilibrium with the whole-rock compositions calculated using 
values in the range 2–5·5, based on the experimental studies of Baker & Eggler (1987), Sisson & Grove (1993) and Panjasawatwong et al. (1995).
|
|
Four distinct composition distribution patterns are observed. (1) The high-SiO2 dacites (#2.35, B11 and B12) show essentially single populations centred on Ca-number [100 Ca/(Ca + Na)] 33–36, overlapping the range of equilibrium compositions. Plagioclase phenocrysts are largely normally zoned with superimposed oscillatory zoning. (2) The low-SiO2 dacites (#2.18 and B16) display a broad main population around Ca-number 40–56, within or slightly more Na-rich than the range of equilibrium compositions, and a few Ca-rich (Ca-number 61–92) outliers. Crystals from the main populations are normally or slightly reversely zoned, with superimposed oscillatory zoning. The Ca-rich outliers are normally or occasionally reversely zoned and have Ca-rich rims (Ca-number up to 89). (3) The low-SiO2 dacites (B59, #2.30 and B52) show large peaks around Ca-number 31–36, which could clearly not be in equilibrium with melts similar to the whole-rock compositions, a series of small peaks between Ca-number 47 and 61 within the range of equilibrium compositions, and occasional Ca-rich (Ca-number 61–92) outliers. Plagioclase phenocrysts show normal or slight reverse zoning with superimposed oscillatory zoning. Crystals that form the small peaks between Ca-number 47 and 61 have rims that are distinctively more Ca-rich than the crystals that form the large peaks. (4) The low-SiO2 andesite #2.31 and the andesite B7 show large and continuous ranges of composition with major peaks at around Ca-number 71 and 51, respectively. The majority of crystals in these samples could be in equilibrium with melts similar to the whole-rock compositions. Plagioclase phenocrysts are generally normally zoned with superimposed oscillatory zoning. Occasional slightly reversely zoned crystals showing an increase of up to 6 Ca-number from core to rim are also observed. In the low-SiO2 andesite #2.31, the rare euhedral plagioclases show a similar range of composition to the crystals with globular morphologies.
Olivine
Ranges of olivine phenocryst core and rim compositions are shown in Fig. 9a, and representative analyses are included in Table 5, together with the calculated ranges of olivines in equilibrium with the respective whole-rock compositions. The complete dataset can be downloaded from the Journal of Petrology website. The distribution coefficient Fe/MgKD min/liq ranges from 0·29 to 0·34 for basaltic to basaltic andesite compositions (Roeder & Emslie, 1970; Ulmer, 1989; Sisson & Grove, 1993), but increases to 0·68 for silica-rich (rhyolite) compositions (Kilinc & Gerke, 2003). The calculations were performed using Fe/MgKD min/liq 0·29–0·68 and assuming Fe2+ = 0·9
Fe in the melt.
|
|
As for plagioclase, the studied samples can be divided into four groups. (1) The high-SiO2 dacites (#2.35, B11 and B12) show small ranges of olivine composition from Fo42 to Fo30, within the range of equilibrium compositions. (2) The low-SiO2 dacites (#2.18, B52 and B16) show olivine composition from Fo56 to Fo35 overlapping the range of equilibrium compositions. B16 also contains more Mg-rich outliers to Fo58. (3) The low-SiO2 dacites B59 and #2.30 contain olivines clearly too Fe-rich(Fo34–30) to have been in equilibrium with melts similar to their host rocks and they also contain Mg-rich olivines (Fo76–62). (4) The low-SiO2 andesite #2.31 shows a continuous range (Fo67–59) similar to the expected equilibrium compositions. The rare olivines in the andesite B7 are Fo55 in composition. In all samples, olivine phenocrysts are normally zoned.
Pyroxene
Representative pyroxene phenocryst analyses are presented in Table 6. The complete dataset can be downloaded from the Journal of Petrology website. The mg-numbers [mg-number = Mg/(Mg + Fe2+) x 100] for pyroxenes are shown in Fig. 9b, along with mg-number ranges for pyroxenes in equilibrium with whole-rock compositions, calculated using Fe2+/MgKDmin/whole-rock = 0·23–0·3 (Sisson & Grove, 1993) and assuming Fe2+ = 0·9Fe in the magma. These values were obtained from experiments on basaltic to basaltic andesite melt compositions producing highly aluminous clinopyroxenes and may not be appropriate for dacitic melts and clinopyroxenes with low Al2O3 contents. Clinopyroxenes are augites with 45–35% Wo and 43–29% En. The orthopyroxene crystals found in Ubud pumices are hypersthenes with 4–3% Wo and 61–45% En. The low-Ca clinopyroxene rims in the Bunbulan andesite B7 are pigeonite with 11–9% Wo and 54–52% En.
|
The high-SiO2 dacites #2.35, B11 and B12 show clinopyroxene compositions with mg-number 67–53, too Mg-rich to have been in equilibrium with melts similar to their host rocks. The orthopyroxene in sample #2.35 has mg-number 52–50, within the range of equilibrium compositions. The low-SiO2 dacites #2.18 and B16 show broad clinopyroxene (and orthopyroxene in #2.18) composition ranges from mg-number 83 to 51. Most crystals are more Mg-rich than the calculated equilibrium compositions. The low-SiO2 dacites B59, #2.30 and B52 contain two distinct groups; the principal one consists of Fe-rich clinopyroxenes (mg-number 61–54), the second one comprises occasional Mg-rich clinopyroxenes (mg-number 83–76). The low-SiO2 andesite #2.31 and the andesite B7 show large ranges in clinopyroxene composition with mg-number 86–68. Most crystals are more Mg-rich than the equilibrium compositions. The pigeonite rims in B7 have lower mg-number, 63–60. Pyroene phenocrysts are mostly normally zoned in all samples except sample #2.31. However, occasional slightly reversely zoned phenocrysts showing an increase of up to three in mg-number from core to rim are also observed. Assuming the Fe3+ calculated by stoichiometry, the high mg-number of the clinopyroxene rims in sample #2.31 reflects increase in Fe3+ contents from the core to the rim. High mg-numbers of clinopyroxenes reflect, in part, their high Fe3+ contents (assuming the Fe3+ calculated by stoichiometry). Nevertheless, even if total FeO contents are considered in the calculation of mg-numbers, clinopyroxenes remain more Mg-rich than the calculated compositions in equilibrium with the host rocks compositions.
Fe–Ti oxides
Representative compositions of Ti-magnetite and ilmenite are presented in Table 7. The complete dataset can be downloaded from the Journal of Petrology website. Ti-magnetite in the high-SiO2 dacites #2.35, B11 and B12 ranges in composition from 69 to 55 mol % ulvöspinel, whereas the ilmenite content in the rhombohedral phase ranges from 96 to 93 mol %. In the low-SiO2 dacites #2.18, #2.30, B59, B52 and B16, the composition of Ti-magnetite varies from 66 to 27 mol % ulvöspinel and the composition of the rhombohedral phase varies from 98 to 93 mol % ilmenite. The low-SiO2 andesite #2.31 and the andesite B7 show ranges of titanomagnetite composition of 52–43 and 69–43 mol % ulvöspinel, respectively.
|
|
ESTIMATION OF INTENSIVE PARAMETERS |
|---|
Temperature and fO2
Temperatures were calculated for coexisting orthopyroxene and clinopyroxene cores in the Ubud Ignimbrite pumices (Table 6). Temperatures calculated are between 1010 and 940°C in the high-SiO2 dacite #2.35 and between 980 and 880°C for the low-SiO2 dacite #2.18. Compositions of pigeonite rims and adjacent clinopyroxene outer zones in the andesite B7 suggest temperatures between 1080 and 1100°C.
Two-oxide thermometry calculations for the dacitic pumices and lavas are presented in Table 7. Ilmenite is present only in small quantities in the investigated samples and all analysed crystals were groundmass crystals. Only a few contiguous pairs were found. Most temperatures were calculated using non-contiguous groundmass ilmenites and magnetites that are in equilibrium according to the criterion established by Bacon & Hirschmann (1988). The ranges of temperature obtained always bracketed the temperatures obtained from contiguous pairs. Temperatures calculated for coexisting ilmenite and magnetite are substantially lower (up to 150°C) than estimates derived from two-pyroxene thermometry. The lower temperatures suggested by these methods are likely to reflect the xenocrystic origin of the pyroxenes (see below).
For the Ubud Ignimbrite high-SiO2 dacite #2.35, the estimated temperature range is 846–834°C. Temperatures between 780 and 736°C were obtained for the Bunbulan high-SiO2 dacite lava B11. Fe–Ti oxides in the Gunungkawi low-SiO2 dacites B59 and #2.30 yielded temperatures within the ranges 835–822 and 830–793°C, respectively. Temperatures for the low-SiO2 dacite B16 range from 867 to 853°C. The pumice or fiamme from the Ubud Ignimbrite densely welded intracaldera facies (sample #2.18) and two of the Bunbulan lavas (B12 and B52) gave temperatures <650°C, indicative of post-emplacement low-temperature re-equilibration of the Fe–Ti oxides.
Estimated oxygen fugacities are between –0·3 and –0·4 log units below QFM (quartz–fayalite–magnetite) for the high-SiO2 dacite #2.35, between –0·4 and –0·6 for the low-SiO2 dacite B16 and between –0·7 and –1·1 for the low-SiO2 dacites B59 and #2.30.
Volatile contents
The H2O contents of the magmas that erupted to produce the two ignimbrites were estimated by the plagioclase–melt hygrometer of Housh & Luhr (1991) for plagioclase rims and groundmass glass. Calculations were carried out using the ranges of temperature estimated from Fe–Ti thermometry that are considered to be representative of the conditions that occurred late in the crystallization history (i.e. crystallization of the phenocryst rims) and pressures of 1·5 kbar (see below). Only plagioclase rims and adjacent groundmass glass compositions in equilibrium according to their Ca-numbers (for Ca/NaKD min/liq values between 2 and 5·5) were used in the calculations. Nevertheless significant differences (up to 1 wt % H2O) were observed between the average melt H2O contents calculated from the independent anorthite and albite exchange reactions. The Ubud Ignimbrite high-SiO2 dacite #2.35 yielded average values of 4·0 and 5·0 wt % for the Ab and An solution models, respectively. The Gunungkawi Ignimbrite low-SiO2 dacites B59 and #2.30 gave average values of 5·0 and 5·9, whereas average values of 4·8 and 5·5 wt % were obtained for the low-SiO2 dacites B16. Despite the large uncertainties, these results suggest that the Ubud Ignimbrite magmas had lower H2O content (
4–5 wt %) than the Gunungkawi Ignimbrite magmas (5–6 wt %).
Pressure
No suitable mineral barometer assemblage is present in the Batur silicic rocks. The only constraint comes from the absence of amphibole, indicating that these magmas achieved water saturation at pressures below the lower stability limit of amphibole under water-saturated conditions in dacitic melts (1·4 kbar at 850°C; Gardner et al., 1995).
|
DISCUSSION |
|---|
Magmatic processes
Magma mixing
The mineralogical and mineral chemical data presented above suggest that a large part of the silicic magmas at Batur volcano were the products of magma mixing or mingling. The low-SiO2 dacites B59 and #2.30 show the most compelling evidence for an origin by magma mixing, i.e. very large and discontinuous ranges of phenocryst compositions with most phenocrysts either too evolved or too mafic to be in equilibrium with melts similar to the whole-rock compositions, and large ranges of groundmass glass compositions extending from more mafic to more silicic than the whole-rock compositions. The low-SiO2 dacite B52 also has a large and discontinuous range of plagioclase compositions suggesting an origin by magma mixing. The low-SiO2 dacites #2.18 and B16 show large ranges of phenocryst and groundmass glass compositions, suggesting that they are the products of magma mixing. Nevertheless, a significant part of phenocrysts in these samples are in or near equilibrium with melts similar to the whole-rock compositions, suggesting either that the differences in composition of the two magmas that mixed were small or that only small volumes of more mafic magma were involved.
In contrast, the high-SiO2 dacites (#2.35, B11 and B12) are characterized by narrower ranges in plagioclase and olivine compositions that are in equilibrium with the host lava compositions, and groundmass glasses that are more evolved than the whole-rock compositions or, in the case of the B11 and B12 lavas, groundmass plagioclases that have similar to lower Ca-number than the phenocrysts. These features suggest that the high-SiO2 dacites represent homogeneous magmas. However, most pyroxenes in these samples are clearly more mafic than the olivines and the compositions in equilibrium with their host rocks (Fig. 9). Although the apparent lack of equilibrium with the host-rock compositions may result from calculations using distribution coefficients inappropriate for dacitic melts, the large offset in mg-numbers between the olivines and pyroxenes suggests that they did not crystallize in the same melts, implying a xenocrystic origin for the pyroxenes. Considering the absence of evidence for magma mixing, it is likely that pyroxenes represent xenocrysts carried by the dacitic melts, probably restite crystals inherited from an early phase of differentiation. The low-SiO2 andesite #2.31 and the andesite B7 also have groundmass glasses or groundmass plagioclases more evolved than the whole-rock or plagioclase phenocryst compositions, respectively, and show phenocrysts dominantly in equilibrium, suggesting that the low-SiO2 andesites and andesites also represent essentially homogeneous magmas. However, the relatively large phenocryst composition ranges and phenocryst resorption textures observed in the Gunungkawi Ignimbrite, low-SiO2, crystal-rich andesite #2.31 suggest that the melt did not evolve along a ‘liquid line of descent’.
Causes of phenocryst resorption in the Gunungkawi Ignimbrite crystal-rich pumices
Phenocryst resorption textures are ubiquitous in the Gunungkawi Ignimbrite black crystal-rich pumices (samples #2.31 and #2.9). Plagioclase shows the most extensive resorption features, but olivine and clinopyroxene also commonly show rounded shapes, suggesting that sub-liquidus crystal compositions were taken out of their domains of stability late in the evolution of these magmas. Resorbed plagioclases did not crystallize a rim following the resorption event, indicating that it happened just prior to the eruption. This may reflect several mechanisms: (1) mixing with a more mafic magma; (2) heating; (3) addition of volatiles to the melt; (4) decompression. As discussed above, magma mixing may be ruled out, in this case, on the basis of the evolved composition of the groundmass glass and by the absence of compositional contrast between resorbed and euhedral phenocrysts. Decompression affects the whole volume of magma and is, therefore, likely to result in resorption of all the phenocrysts present, which is not consistent with the presence of euhedral crystals in these samples. Couch et al. (2001) have shown that heating of a partially crystallized magma body, for example by intrusions of hot mafic magma in the vicinity, can result in the development of a thermal gradient and convection. Convection occurs through generation of localized plumes, preventing complete homogenization and resulting in juxtaposition of crystals with varying thermal histories. This ‘self-mixing’ process could explain the occurrence of both resorbed and euhedral phenocrysts as well as the large ranges of phenocryst compositions observed in these samples. Increase in melt volatile content may occur if mafic magma accumulates at the base of the magma body and cools. Volatiles exsolved during crystallization of the mafic magma are then transferred to the resident melt. However, this process is normally accompanied by heating and will, therefore, occur only in conjunction with the ‘self-mixing’ process, enhancing remelting.
In summary, the resorption textures observed in the Gunungkawi crystal-rich pumices are thought to result from heating and self-mixing in a crystal-rich magma body. A similar process is believed to have strongly influenced the evolution of the historical basaltic andesite magmas at Batur volcano (Reubi, 2004; Reubi & Nicholls, in preparation) and has also been proposed as the origin of several crystal-rich magmas erupted at arc volcanoes (e.g. Matthews et al., 1999; Murphy et al., 2000; Couch et al., 2001).
Fractional crystallization
On the basis of the evidence presented in the previous sections, we conclude that magma mixing was involved in the generation of all magmas except the andesite and high-SiO2 dacite magmas, which are considered likely to be products of simple fractional crystallization. It should be noted that the high-SiO2 dacites contain unequilibrated pyroxenes interpreted as xenocrysts and are therefore not strict products of fractional crystallization. However, as a result of the very low modal proportion of pyroxene (Table 3), whole-rock compositions may be considered as representing homogeneous magmas. Consequently, only these two compositions have been used in our modelling. Previous studies of the overall magmatic evolution of Batur volcanic field have shown that the silicic magmas belong to the same suite as the basaltic andesite magmas erupted during the last 150 years (Wheller & Varne, 1986; Reubi & Nicholls, 2005). Oxygen, strontium and thorium isotope ratios of Batur dacites and historical basaltic andesites are very similar (Whitford, 1975; Wheller & Varne, 1986; Rubin et al., 1989), indicating melting of a similar mantle source and closed-system evolution. Fractional crystallization modelling indicates that the andesite (B7) can be related to the historical basaltic andesite by 33% fractionation of plagioclase, clinopyroxene, olivine and Ti-magnetite (Reubi & Nicholls, 2005). However, the large 226Ra excesses measured in the historical basaltic andesites (Rubin et al., 1989) indicate that these magmas were produced less than 8000 years ago, which implies that they cannot be relics of the magmatic system that produced the silicic magmas. Therefore, the historical basaltic andesites represent only model compositions for basaltic andesite magmas present within the magmatic system during the period of major silicic activity.
Least-squares mixing models were formulated to assess the possibility that the homogeneous magmas are related by fractional crystallization. Calculations were carried out using the average composition of the different phenocryst phases present in the parent magmas, except for apatite, for which the composition from Parat et al. (2002) has been used. The models were tested by comparing the apparent bulk partition coefficients (D) calculated using the method of Allègre & Minster (1978) with possible bulk partition coefficients for the mineral assemblage obtained from the least-squares models using ranges of published mineral partition coefficients for andesitic and dacitic magmas (GERM compilation, available at http://earthref.org/GERM/, and references therein). Additionally, residual melt fractions (F) were calculated using the Rayleigh fractionation equation and the calculated apparent D values for comparison with F values obtain from least-squares models. Results are listed inTable 8.
|
The fractionation models indicate that the compositions of the high-SiO2 dacites can be produced by 41–43% fractional crystallization of the andesite B7, except for the sample B12. The high-SiO2 dacite B12 belongs to a distinct magmatic series, likely to be related to pre-caldera amphibole-bearing andesites (Reubi & Nicholls, 2005
). The low-SiO2 andesite #2.31 cannot produce the high-SiO2 dacites. Compared with the trends produced by the other samples, #2.31 is enriched in elements compatible in plagioclase (e.g. Al2O3, Sr) and depleted in elements incompatible in plagioclase (e.g. MgO, FeO, V) (Figs 2 and 3). This could indicate either that #2.31 was produced by fractionation of an assemblage dominated by clinopyroxene and olivine, suggesting fractionation at higher pressure and/or lower H2O content compared with the other andesites or that it accumulated plagioclase. Considering the high proportion of plagioclase phenocrysts in this sample (Table 3) and their broad range of composition (Fig. 8), the second hypothesis is more likely.
Models for magmatic differentiation
Rayleigh fractionation models for the homogeneous magmas are presented in Fig. 10, along with possible mixing trends. As suggested by their mineralogy, two distinct types of low-SiO2 dacite, represented by different mixing trends, are observed (Fig. 10a). The first type (e.g. B59) could have resulted from mixing between the andesite and high-SiO2 dacite magmas. The second type of low-SiO2 dacite (e.g. sample #2.18) lay closer to the fractionation trend. In regard to the continuous ranges of phenocryst compositions observed in the second type (Figs 8 and 9), we propose that they represent the products of crystallization within compositionally zoned magma storage reservoirs that were partially homogenized prior to, or during, eruption. The low-SiO2 dacite B52 resulted from mixing between a high-SiO2 dacite similar to B12 and a mafic component compositionally distinct from the magmas erupted during and between the caldera forming eruptions, probably an amphibole-bearing andesite (Reubi & Nicholls, 2005).
|
The existence of only two homogeneous compositions related by fractional crystallization among the silicic magmas at Batur (i.e. andesite and high-SiO2 dacite) suggests that the fractionation mechanism produced composition gaps (
10 wt % SiO2). Additionally, the distinct phenocryst populations in the two homogeneous compositions (Fig. 8) signify very efficient segregation of the residual melt from the phenocrysts. Several fractionation mechanisms may have produced the compositional gaps: critical crystallinity of a propagating solidification front (Marsh, 1981, 1996), convective crystal fractionation along magma reservoir walls (McBirney et al., 1985), crystallization along a liquidus with shallow slope (Grove & Donnelly-Nolan, 1986), and gas-driven filter pressing (Sisson & Bacon, 1999). The percentage of crystallization calculated to produce the high-SiO2 dacites magmas from the andesites at Batur (40%) is similar to the expected critical crystallinity of these magmas[35–45 vol. % (Marsh, 1981; Brophy, 1991)], suggesting that extraction of residual magmas from a rigid crystal network (crystal-mush), as in the solidification front or gas-driven filter pressing models, produced the composition gaps in this case. Subsequent mixing of the homogeneous andesite and high-SiO2 dacite magmas produced the continuous range of silicic compositions observed at Batur.
An alternative model to fractional crystallization capable of explaining compositional gaps may involve remelting of plutonic bodies formed by previous batches of magma of the same origin (e.g. Marsh, 1996). However this mechanism produces crystal-rich magma containing abundant resorbed restite crystals (Matthews et al., 1999; Murphy et al., 2000). Consequently, remelting is unlikely to have produced the homogeneous silicic magmas at Batur that are crystal-poor and show very low proportions of resorbed crystals. Nevertheless, as discussed above, the crystal-rich Gunungkawi Ignimbrite black pumices (samples #2.31 and #2.9) show textural evidence suggestive of heating indicating that remelting may have occurred and produced some of the silicic magmas at Batur.
In summary, we propose the following steps for the genesis of Batur silicic magmas (Fig. 11): (1) closed-system crystal fractionation (?35%) of basaltic to basaltic andesite melts under variable pressure produced the range of andesite magmas; (2) closed-system fractional crystallization (35–45%) of the andesite magmas produced the high-SiO2 dacite magmas; (3) mixing between these two groups of magmas produced the mixed low-SiO2 dacite magmas; (4) mixing between high-SiO2 andesite and low-SiO2 dacite melts, possibly during overturn of a compositionally zoned reservoir resulting from open-system fractional crystallization, yielded the slightly mixed low-SiO2 dacite magmas; (5) heating and self-mixing of low-SiO2 andesite cumulate piles produced the crystal-rich low-SiO2 andesite magmas; (6) heating and self-mixing of low-SiO2 dacite cumulate piles produced the crystal-rich low-SiO2 dacite magmas.
|
Temporal variations in magma composition and structure of the magmatic system
The relative temporal relationships between the different types of silicic magmas inferred from the stratigraphic positions of the samples provide useful information regarding the structure of the magmatic system. It should be noted that the limited number of samples studied in detail for each deposit implies that the magmatic system may not be fully characterized by our data. However, it constrains the minimum number of distinct magmas present within the magmatic system before each major eruption and provides an insight into the complexity of this system. The deposits from the early phase of the first caldera-forming eruption (Ubud Ignimbrite) comprise high-SiO2 dacite pu