Journal of Petrology Advance Access originally published online on September 20, 2006
Journal of Petrology 2006 47(12):2433-2462; doi:10.1093/petrology/egl050
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Coexisting High- and Low-Calcium Melts Identified by Mineral and Melt Inclusion Studies of a Subduction-Influenced Syn-collisional Magma from South Sulawesi, Indonesia
1 MAX PLANCK INSTITUTE FOR CHEMISTRY, GEOCHEMISTRY DIVISION MAINZ, 55020, GERMANY
2 FACULTY OF EARTH AND LIFE SCIENCES, DEPARTMENT OF ISOTOPE GEOCHEMISTRY, FREE UNIVERSITY AMSTERDAM, 1081 HV, THE NETHERLANDS
3 DEPARTMENT OF GEOLOGY AND SOIL SCIENCE, GHENT UNIVERSITY KRIJGSLAAN 281 S8, 9000 GHENT, BELGIUM
4 SCHOOL OF EARTH SCIENCES AND CODES SRC, UNIVERSITY OF TASMANIA HOBART, TAS 7001, AUSTRALIA
5 FACULTY OF GEOSCIENCES, UTRECHT UNIVERSITY UTRECHT, 3508 TA, THE NETHERLANDS
6 DEPARTMENT OF GEOLOGY AND GEOPHYSICS, UNIVERSITY OF ADELAIDE ADELAIDE, SA 5005, AUSTRALIA
RECEIVED NOVEMBER 28, 2005; ACCEPTED AUGUST 22, 2006
 |
ABSTRACT |
|---|
TOP
ABSTRACT
INTRODUCTIONMineral and melt inclusions in olivines from the most Mg-rich magma from the southern West Sulawesi Volcanic Province indicate that two distinct melts contributed to its petrogenesis. The contribution that dominates the whole-rock composition comes from a liquid with high CaO (up to 16 wt %) and low Al2O3 contents (CaO/Al2O3 up to 1), in equilibrium with spinel, olivine (Fo85–91; CaO 0·35–0·5 wt %; NiO 0·2–0·30 wt %) and clinopyroxene. The other component is richer in SiO2 (>50 wt %) and Al2O3 (19–21 wt %), but contains significantly less CaO (<4 wt %); it is in equilibrium with Cr-rich spinel with a low TiO2 content, olivine with low CaO and high NiO content (Fo90–94; CaO 0·05–0·20 wt %; NiO 0·35–0·5 wt %), and orthopyroxene. Both the high- and low-CaO melts are potassium-rich (>3 wt % K2O). The high-CaO melt has a normalized trace element pattern that is typical for subduction-related volcanic rocks, with negative Ta–Nb and Ti anomalies, positive K, Pb and Sr anomalies, and a relatively flat heavy rare earth element (HREE) pattern. The low-CaO melt shows Y and HREE depletion (Gdn/Ybn
41), but its trace element pattern resembles that of the whole-rock and high-CaO melt in other respects, suggesting only small distinctions in source areas between the two components. We propose that the depth of melting and the dominance of H2O- or CO2-bearing fluids were the main controls on generating these contrasting magmas in a syn-collisional environment. The composition of the low-CaO magma does not have any obvious rock equivalent, and it is possible that this type of magma does not easily reach the Earth's surface without the assistance of a water-poor carrier magma. KEY WORDS: melt inclusions; mineral chemistry; olivine; syn-collisional magmatism; ankaramites; low-Ca magma
|
INTRODUCTION |
|---|
Arc magmas are mixtures of silicate liquid and crystals that are often assumed to be phenocrysts that crystallized from the enclosing melt. There is, however, increasing evidence that these crystals belong to more than one population (Tepley et al., 2000
; Turner et al., 2003). They may be phenocrysts in equilibrium with melts that are more or less evolved than, but related to, the host liquid; phenocrysts from melts that are genetically unrelated to the host melt; or they may be xenocrysts from the mantle or crust through which the magma ascended. Even if crystals are indeed phenocrysts of the melt in which they occur, it is questionable whether whole-rock analyses represent liquid compositions, as accumulative processes may also have played a role. Whole-rock analyses cannot, therefore, be relied on to obtain information about the composition of mantle-derived melts without careful study of their ‘crystal cargo’. For arc-related studies this problem is compounded by the fact that most lavas are too evolved to be in equilibrium with a peridotitic mantle assemblage (Davidson, 1996), and must have been affected by processes such as crystal fractionation, magma mixing, crustal contamination and/or degassing.
Early crystallizing minerals such as forsteritic olivine, magnesium-rich clinopyroxene or spinel are more likely to retain information about the primitive magma from which they crystallized (Clynne & Borg, 1997) than whole-rock compositions do. Better still, these mineral phases sometimes contain melt inclusions, which represent small pockets of (primitive) melt, and these can, when erupted and quenched soon enough after entrapment, be used to study these liquids directly (Kamenetsky & Clocchiatti, 1996; Schiano et al., 2000; Sobolev et al., 2000; Frezzotti, 2001; Kent & Elliott, 2002). We combined the study of mineral compositions and melt inclusions in a collision-related high-Mg arc magma from southern Sulawesi (Indonesia). Both the minerals and their melt inclusions preserve evidence of the existence of distinct primitive melts in the system, which mixed to form the magma we find at the Earth's surface. Importantly, and a major focus of this study, is the identification of a melt component for which no equivalent whole-rock compositions have been reported previously. We argue that this might represent small-degree melts that do not reach the surface without the assistance of a ‘carrier magma’.
|
GEOLOGICAL BACKGROUND |
|---|
The general tectonic history of the Western Sulawesi Volcanic Province (Fig. 1), from which the studied sample originates, has been described at length by Elburg & Foden (1999)
and Elburg et al. (2002b, 2003). It belongs to the now extinct part of the Indonesian arc, which was active from the Paleocene to the Miocene. The sample, Tnv96-1, belongs to a 6–9 Ma silica-undersaturated suite of ultrapotassic volcanic rocks with pronounced ‘continental’ isotopic characteristics (87Sr/86Sr = 0·7045–0·7061), indicative of the involvement of subducted continental material in magma genesis. It is a duplicate sample from the same dyke as sample Tnv11 of Elburg & Foden (1999), which cross-cuts slightly deformed lahar-like deposits of unknown age. This dyke looks homogeneous in the field, and petrographic analysis shows that the mineralogy and texture of the two samples is the same. The chemical composition of Tnv96-1 can therefore be assumed to be the same (within error) as that of Tnv11, which is among the three most mafic samples of the suite (20 wt % MgO; Table 1). It can be classified as being both shoshonitic (1·7 wt % K2O at 46·7 wt % SiO2, K2O/Na2O = 1·1) and ankaramitic (phenocrysts of clinopyroxene and olivine only; CaO/Al2O3 >1). Its composition is nepheline-normative (0·6 wt %), with 28 wt % of normative diopside and 36 wt % olivine; normative plagioclase (An73) and orthoclase constitute 22 and 10 wt %, respectively. Its high MgO content, high mg-number [100 x Mg/(Mg + Fetot)] and a mineralogy that is dominated by high mg-number olivine and clinopyroxene crystals (see below) indicate that this sample has undergone little fractionation and is therefore the most likely candidate to give us information about the magma parental to the syn-collisional suite of rocks.
|
|
All samples from this suite show typical subduction-related trace element characteristics, with negative anomalies for Nb, Zr and Ti and positive anomalies for Rb, Sr, Pb and the light rare earth elements (LREE) in normalized trace element diagrams. This potassic magmatism took place shortly after collision of the Sula platform with the Western Sulawesi volcanic arc, which brought west-dipping subduction to a halt (Elburg & Foden, 1999
). The continental crust underlying Western Sulawesi belongs to Sundaland, and was probably cratonized in the Mesozoic. The area from which the sample originates does not show evidence for underthrusting of allochthonous continental fragments, in contrast to Central Sulawesi (Elburg et al., 2003). |
ANALYTICAL TECHNIQUES |
|---|
Major element mineral and glass analyses were performed on a JEOL 8200 electron microprobe at the Max Planck Institute for Chemistry, using natural mineral standards for calibration. Olivine and spinel were analysed with an accelerating voltage of 20 kV and a current of 20 nA. All other phases were analysed with 15 kV and 12 nA. The beam diameter was 2 µm. To obtain paired analyses of olivines and melt inclusions, olivine analyses were performed at distances of
30 µm from the melt inclusion to avoid NiO depletion and CaO enrichment of the olivine by diffusional exchange between the two phases (see discussion on LC olivines below).
Trace elements in the melt inclusions were analysed by laser ablation–inductively coupled plasma–mass spectrometry (LA–ICP–MS), using a Finnigan Element 2 with a Merchantek 213 nm laser, and a Cameca 3f ion microprobe, both at the Max Planck Institute for Chemistry. LA–ICP–MS analyses were performed with He as carrier gas. The instrument was tuned to minimize oxide production, which was always less than 1 rel. % of the signal of its element. Raw Th/U ratios of the NIST 612 glass were 0·91. The laser was operated with a spot size of 60 µm at 6 J/cm2 and resulted in pits with height:width ratios of one. One run consisted of 60 scans of 1·5 s each, of which the first 17 scans were taken with the laser shutter closed, to determine the blank, which was subtracted from the analysis. The analyses were normalized to the Ca content of the inclusion, which had been determined prior to laser ablation by electron microprobe. Melt inclusions with dimensions smaller than 60 µm were also analysed, in which case part of the enclosing olivine was ablated. This is unlikely to have affected either the relative or absolute trace element abundances measured for the melt inclusion significantly, as olivine has extremely low levels of all trace elements analysed, and very minor contents of the normalizing element Ca. Not all scans were used for all inclusions, as the inclusion was sometimes thinner than 60 µm. In that case, only those scans were used where all elements were above the detection limit. NIST SRM612 glass was used for calibration, but was always analysed at the end of the complete set of analyses, to avoid memory effects for elements that are far lower in the glasses analysed than in the calibration standard (e.g. Ta, Nb, Hf). Secondary standards (Kl2-G and StHs6/80-G; Jochum et al., 2000) were used to check the accuracy of the analyses. Ion microprobe measurements of REE were performed following techniques described by Hellebrand et al. (2002). A few inclusions were also analysed for H2O, TiO2, Li and B contents by ion microprobe, following analytical techniques published by Gurenko et al. (2005).
|
MINERAL CHEMISTRY |
|---|
Olivine and clinopyroxene are the most important mineral phases in this sample, and form euhedral crystals. Deformation textures, such as kink bands or undulose extinction, have not been observed in any of the crystals. Chromium-rich spinel occurs as inclusions in olivine and clinopyroxene. Plagioclase, Fe-rich spinel and biotite (rare) occur as microphenocrysts. The groundmass of the sample is largely crystalline, reflecting relatively slow cooling within a dyke. The sample is fairly crystal rich, with large clinopyroxene and olivine crystals constituting 23 and 28 vol. % of the sample, respectively.
Olivine crystals have forsterite (Fo) contents varying between 81 and 94 mol %, although rim compositions can occasionally be as low as 65 mol %. Most compositions, however, lie between 85 and 94 mol % Fo. The olivine analyses scatter widely in diagrams of CaO or NiO vs Fo content. This scatter is interpreted to reflect two main groups, with some compositions intermediate between the end-members (Fig. 2). One group has on average higher Fo (92–94), low CaO (0·05–0·2 wt %) and high NiO contents (0·35–0·5 wt %, one crystal with 0·63 wt %; low-calcium group; LC); the second group has lower average Fo, high CaO (0·35–0·5 wt %) and low NiO contents (<0·3 wt %). The latter group can be further subdivided into a small population of olivines with Fo between 90 and 92 (high-calcium, high-forsterite group; HC-HF) and the majority of crystals with Fo 85–87 (high-calcium, low-forsterite group; HC-LF). The intermediate compositions between the HC and LC groups are denoted as LC-HC olivine crystals; those intermediate between HC-HF and HC-LF are referred to as transitional HC olivine crystals. LC olivine crystals are somewhat less common than HC-LF olivines, and HC-HF olivines make up less than 5% of the total population. LC olivine crystals always show a smooth increase in CaO and decrease in NiO contents towards their rims (Fig. 3); a decrease in Fo content is also observed but is not as pronounced, and resembles the zoning seen in the HC olivines. The latter olivine crystals show only a small increase in CaO and decrease in the (already low) NiO contents towards their rims.
|
|
Spinel occurs as inclusions in all types of olivine phenocrysts, but is least common in the LC olivines. A good correlation exists between the forsterite content of the olivine and the Mg/(Mg + Fe2+) ratio of the enclosed spinel (Fig. 4c). Cr-numbers [100 x Cr/(Cr + Al)] of the spinels in the LC olivine crystals are marginally higher than those in HC-HF olivines (79–82 vs 71–74), whereas those in the HC-LF olivine crystals, which occur most frequently, show a wide variation, with the majority falling between 50 and 60. TiO2 contents of the latter range up to 1·5 wt %, whereas the spinels in the LC and HC-HF olivine crystals reach only 0·8 wt %. Calculated equilibration temperatures, following Ballhaus et al. (1991)
, between olivine and spinel lie around 1000°C for LC and around 1100°C for HC olivines (Table 2). Oxygen fugacities at these temperatures lie 1–3 log units above the fayalite–magnetite–quartz buffer.
|
|
Clinopyroxene phenocrysts show a wide variation in mg-number between 71 and 93. Cr2O3 contents vary between 0 and 1 wt %, showing a positive correlation with mg-number, but a wide range of Cr contents is seen at high mg-number (Fig. 5). Negative correlations exist between mg-number on the one hand and Al2O3 and TiO2 on the other, with the most magnesium-rich clinopyroxenes containing as little as 0·05 wt % TiO2 and 1 wt % Al2O3. Clinopyroxene also occurs as inclusions in HC-LF olivines (Table 3); vice versa, HC-LF olivine inclusions are also found within clinopyroxene crystals. The chemical composition of clinopyroxene crystals enclosed within HC-LF olivine is indistinguishable from that of analysed clinopyroxene phenocrysts, except when the inclusions contain spinel in contact with clinopyroxene; in this case subsolidus equilibration appears to have taken place, depleting the clinopyroxene in Ti and Al and causing an enrichment in Cr. Clinopyroxene inclusions were not seen in HC-HF olivine crystals, but this may be a result of the scarcity of these olivines.
|
|
Orthopyroxene does not occur as a phenocryst phase, but is found as inclusions in LC olivine crystals. The mg-number of the orthopyroxene is similar to that of the surrounding olivine (Table 3). CaO and Al2O3 contents vary up to 1·5 wt % and 2·5 wt %, respectively, and TiO2 contents are no higher than 0·16 wt %. Cr2O3 contents are positively correlated with Al2O3 contents and reach 1%.
Biotite microphenocrysts have a variable composition, with mg-number between 35 and 60, and TiO2 contents between 0·6 and 6·2 wt %. Cl contents are 0·1 wt % or less, and F contents vary between 1 and 1·9 wt %.
Very rare near-spherical Fe–Ni sulphide inclusions were observed in LC olivines.
Plagioclase microphenocrysts have an anorthite content of 48–75 mol %, and the orthoclase component is less than 6 mol %.
Diffusion profile modelling of LC olivine
The HC and LC olivine crystals both show decreasing forsterite contents towards the rims of the crystals, but their profiles for NiO and especially CaO are very different (Fig. 3). The HC olivine crystals have high CaO contents in the core of crystals, and show a very moderate increase in CaO content towards the rim of the crystals, which may be followed by a decrease (Fig. 3a2). The LC olivine crystals have very low CaO contents in the core, which show a pronounced increase towards the rim of the crystals. The length scale over which the increase in CaO takes place in LC olivine is larger than in HC olivine. This increase in CaO is matched by a decrease in NiO in the LC olivine. The contrasting length scales of the change in CaO and NiO content between HC and LC olivine suggest that two different mechanisms may be responsible, i.e. quench crystallization for HC olivine crystals and re-equilibration with a contrasting magma for the LC olivine crystals.
To investigate the cause of the strong decrease in NiO and increase in CaO contents towards the rims of the LC olivine crystals, several microprobe traverses were measured over these crystals. The distribution of CaO and NiO along profiles of the LC olivine crystals strongly resemble diffusion profiles (Fig. 3b2 and b3), and could therefore allow us to calculate the time elapsed since these crystals were incorporated in a melt with contrasting composition. We used the Ni and Ca data provided by Petry et al. (2004) at 1250°C, and the solution to the diffusion equation for a sphere given by Albarède (1995, p. 449):
 |
), as diffusion is approximately six times faster along the c-axis than along the other two crystallographic axes (Petry et al., 2004). In the case where we measured two diffusional profiles orthogonal to each other, and where the shape of the olivine crystal indicated that one must have been subparallel to the c-axis, it is obvious that diffusivity must have been faster along the c-axis. The modelling indicates that the crystals must have resided in the high-Ca magma for 5–10 years. In the modelling for the crystal in Fig. 3, the best result was obtained if the diffusion coefficients for Ni and Ca were 7·5 and 8·3 times faster along the traverse subparallel to the c-axis than orthogonal to this direction; this is a somewhat higher ratio than the six-fold value reported in the literature (Petry et al., 2004). Moreover, the diffusion coefficients for Ni were only twice that of Ca, whereas they are supposed to be close to an order of magnitude different. Although this may reflect the fact that the processes acting on the crystal cannot be approximated by diffusion only, it could also mean that the physical conditions under which diffusion took place were different from those in the experiments from which the diffusion coefficients were determined. Although this modelling will only be an approximation of the processes that affected the LC olivine crystals, it indicates that the LC olivine crystals may have been incorporated within the HC magma, which constitutes the bulk of the sample and with which the LC olivines were clearly not in equilibrium with respect to CaO and NiO content, several years before the magma intruded to form the dyke that was sampled.
|
MELT AND FLUID INCLUSION DESCRIPTION |
|---|
Melt inclusion shape
The three groups of olivine crystals all contain melt inclusions, but those within the HC olivine crystals are significantly larger than those in the cores of the LC olivine crystals, which do not exceed 30 µm diameter. Some LC olivine crystals also contain micro-inclusions of
0·3 µm, consisting of a tiny spinel, a bubble and some glass. When present, these inclusions are so numerous that they give the crystal a cloudy appearance. Only in one olivine (sample 2009) did several inclusions occur in a single plane, resembling a healed crack through the crystal. All other inclusions analysed occurred as discrete bodies, and were therefore interpreted to be primary inclusions, following the criteria set by Roedder (1984). Their shape was generally spherical or ovoid. Inclusions in LC olivines sometimes have a halo of small inclusions around them, and this arrangement was interpreted as indicating decrepitation of the inclusion during decompression.
Fluid inclusions
A selection of 12 fluid inclusions within the three groups of olivine crystals was studied using a heating–freezing stage, to constrain their freezing and homogenization temperatures, as these can give information on the composition of the volatile phase and the pressure at which they have been trapped (Roedder, 1984). The heating–freezing experiments showed that the ice within the bubble thawed at –56·4 ± 0·9°C. This melting temperature is indicative of virtually pure CO2. This result was checked by Raman spectroscopy, which gives information on the molecular bonds present within materials (Burke, 2001). This method yielded peaks of variable intensity at 1288 and 1390 cm–1, which are indicative for CO2. No evidence was found for the presence of an H2O-rich fluid, but one inclusion in an LC olivine that had not been reheated yielded a small peak at 2918 cm–1, indicating the presence of minor CH4.
Homogenization of the volatile phase in the bubbles was hard to observe during heating–freezing experiments, and only one experiment gave an unambiguous result of homogenization into the vapour phase at 27·2°C. This was measured on a fluid inclusion within a rehomogenized melt inclusion within an olivine crystal with characteristics transitional between HC and LC olivine (Fo87, CaO 0·20 wt %). Fluid-only inclusions were measured in two LC olivines which had not been reheated. In these cases, the phase into which homogenization took place was probably vapour (at 10°C) and liquid (at 18°C). Assuming a pure CO2 fluid and using the program LONER18 (Bakker, 2003), these homogenization conditions indicate pressures at an assumed temperature of 1250°C of 98 MPa, 43 MPa and 508 MPa, respectively. This wide range of pressures is probably a reflection of the fact that the olivines did not behave as perfect containers, which is also indicated by the decrepitation marks described above. However, the one result of 5 kbar, although hardly statistically relevant, may suggest that crystallization started at significant depth; this is similar to depths obtained by the same method for the Canary Islands (Nikogosian et al., 2002).
Melt inclusion rehomogenization
All melt inclusions were recrystallized and therefore had to be rehomogenized before analysis. One set of olivines was doubly polished and reheated under an optical microscope using a Vernadsky-type heating stage (Sobolev et al., 1980) at the Free University of Amsterdam. Full homogenization was not achieved, as a bubble persisted in all inclusions after disappearance of the last daughter mineral, accounting for up to 36 vol. % of the inclusion. Large bubbles were mostly present in both the hypersthene- and nepheline-normative (see below) inclusions within LC olivines, whereas those in HC olivines were less pronounced. No correlation was noted between the composition of the melt inclusions and its bubble size. The presence of high-pressure CO2 in some inclusions (see above) indicates that at least some of the bubble volume can be accounted for by the presence of a volatile phase. Most inclusions also contained a spinel crystal, which made up between 2 and 38 vol. % of the total inclusion. The size of this crystal indicates that it cannot be a daughter crystal, crystallized from the trapped melt. It is more likely that the spinel and the melt were trapped at the same time. It is possible that nucleation of a spinel on a growing olivine crystal may have been the cause of melt entrapment in the first place. Partial rehomogenization temperatures ranged from 1100 to 1250°C, with no clear distinction between the different groups of host olivines. This is the best estimate for the temperature at which the melt fraction became trapped in the olivine host. These temperatures are somewhat higher than given by the olivine–spinel geothermometer (Table 2), indicating that the crystals did not erupt immediately after trapping the melt inclusion, but spent a significant amount of time at lower temperatures. This is also suggested by the modelling of Fe-diffusion profiles in olivine around the inclusions (see below).
Two other sets of melt inclusion-bearing olivines were rehomogenized, by heating in a vertical gas-mixing furnace at 1250°C for 3–10 min [University of Heidelberg, Germany (CO/CO2 = 0·9/0·1) and the University of Tasmania, Australia (Ar atmosphere)]. These inclusions were quenched in water. In some inclusions, homogenization was found to be incomplete, as olivine crystals were present in the inclusion after quenching. After rehomogenization, the olivines were mounted in epoxy and polished to expose the melt inclusions for analysis by electron microprobe, secondary ionization mass spectrometry (SIMS) and LA–ICP–MS.
|
MELT INCLUSION COMPOSITIONS |
|---|
Selected melt inclusion compositions are given in Table 1, and the full dataset is available as an Electronic Appendix in MS Excel format (http://www.petrology.oxfordjournals.org). For ease of discussion, the melt inclusions have been divided into several groups, mainly based upon the forsterite and Ni content of the enclosing olivine (Fig. 6a). Most inclusions analysed were found within HC-LF olivine crystals, as this type of olivine is the most abundant and contains large inclusions. Few came from HC-HF olivine crystals, in keeping with the scarcity of these crystals. The inclusions in LC olivine crystals have been subdivided on the basis of their being hypersthene (LC-Hy) or nepheline-normative (LC-Ne) (see below). Not all olivines fall within the three main groups (LC, HC-HF, HC-LF); some are transitional between HC and LC olivine (indicated as LC-HC inclusions), and others fall between the low- and the high-forsterite groups of HC olivine (indicated as transitional HC inclusions in following diagrams and discussions). Had the Fo and calcium content of the olivine crystals been taken as a basis for subdivision, the groups would have remained largely the same (Fig. 6b). Two glasses adjacent to orthopyroxene inclusions within LC olivines are indicated separately in Fig. 6 (denoted as LC-opx). These melt rims are no more than a few microns thick and their volume is only 10% of that of the associated orthopyroxene (in a two-dimensional view), and it is possible that boundary effects during trapping may have influenced their composition.
|
We will first present the major element composition of the inclusions, uncorrected for any post-entrapment equilibration processes. After this, we will discuss the effects of equilibration with the host olivine.
Major elements
Compositions of melt inclusions are shown in Table 1 and their major element composition has been plotted in Figs 7 and 8. The composition of most melt inclusions is broadly related to that of the enclosing olivine (Fig. 7, Table 1): HC-HF olivine crystals contain melt inclusions with CaO/Al2O3 ratios around one, HC-LF inclusions have CaO/Al2O3 0·5–1 (apart from one inclusion with an unusually high and as yet unexplained CaO/Al2O3 ratio of 1·42), and nearly all hypersthene-normative inclusions in LC and LC-HC olivine crystals fall in the range 0·05–0·4. Ten of the inclusions from the LC-Hy group were analysed in a single crystal, olivine sample 2009, in which the inclusions were located in a plane. These 10 inclusions are very similar, with a standard deviation between 1 and 12 rel. % for all oxides above 0·1%. Considering the correlation between the composition of the melt inclusions and that of the enclosing olivine, it is highly likely that the melt inclusions represent an important component within the magmatic system from which the olivines crystallized, rather than fractions of exotic and ephemeral melt that was trapped accidentally. However, the nepheline-normative inclusions in the LC and transitional olivine crystals do not conform to this general pattern. They generally have low SiO2 contents (<45 wt %) and sometimes have larnite or kalsilite in the norm; they all have higher CaO/Al2O3 ratios (0·6–1) than the hypersthene-normative inclusions, reflecting higher calcium contents.
|
|
For assemblages dominated by clinopyroxene–olivine fractionation, the CaO/Al2O3 ratio can be used as an index of fractionation. As this ratio also distinguishes between LC and HC olivine-hosted inclusions, it has been used as the x-axis in Fig. 8. For HC olivines, K2O and Na2O contents increase with decreasing CaO/Al2O3 ratio, but the hypersthene-normative inclusions in LC olivine crystals do not fall on this trend, as their range in potassium and sodium contents ranges to lower values than expected from the extrapolation of the HC trend. Contents of TiO2 and P2O5 are scattered, with the former oxide reaching lower average levels in the inclusions within LC olivine and the latter higher concentrations. No obvious systematic differences can be seen in S or F concentrations, although both elements are higher in LC-olivine-hosted nepheline-normative (LC-Ne) inclusions. Most of the inclusions in HC olivine do not reach more than 0·32 wt % Cl, but this element can be as high as 0·7 wt % in hy-normative inclusions within LC and LC-HC olivine. The inclusions with high Cl contents have somewhat lower electron microprobe totals, which may indicate that these glasses are rich in H2O. They also have lower K/Cl ratios (<15) than the HC-olivine-hosted inclusions (10–34). These Cl contents are higher than for nearly all other subduction-related glasses, for which Cl contents normally range between 0·1 and 0·5 wt % (Harris & Anderson, 1984
; Sisson & Layne, 1993; Gioncada et al., 1998; Sisson & Bronto, 1998; Luhr, 2001; Kent & Elliott, 2002; Cervantes & Wallace, 2003; Stix et al., 2003; Métrich et al., 2004), apart from some unusually halogen-rich glasses from the Izu volcanic front with up to 0·8 wt % Cl (Straub & Layne, 2003). Fluorine contents are also high compared with the global arc literature (typically <0·1 wt %), although data are scarcer than for Cl. The inclusions from LC olivine 2009 have a lower Cl (0·22 wt %) content than the other hypersthene-normative inclusions, apart from the melts next to the orthopyroxene within LC olivine, which show an even more pronounced depletion in this element (Fig. 8). The glass rims next to orthopyroxene inclusions are also more silicic than the other hypersthene-normative glass inclusions, but their composition is similar in other respects.
Modelling of FeO-diffusion profiles
FeO contents for most melt inclusions were lower than expected for melts in equilibrium with high-forsterite olivine. This probably reflects FeO loss, resulting from re-equilibration with the enclosing olivine (Danyushevsky et al., 2000). This process has major effects on the FeO and MgO concentrations of melt inclusions, and can thereby also change the absolute concentrations of other elements. It is therefore important to try to undo the effect of this re-equilibration process. This can in principle be done by modelling the FeO profiles around the inclusions, following the technique described by Danyushevsky et al. (2002). This is an elaborate process, where the FeO profile in the olivine is modelled as a function of the (known) size of the inclusion, the initial temperature of trapping (approximated by homogenization temperature of the glass), the unknown initial FeO content of the melt inclusion, and the unknown cooling interval and duration of cooling before eruption. The three unknowns in this model are then adjusted until the modelled FeO profile of the olivine and the modelled composition of the melt inclusion match the measured profile and melt inclusion composition.
FeO profiles in olivine were measured around a number of HC- and LC-olivine-hosted melt inclusions. For all HC olivines, FeO contents increased towards inclusions, whereas this type of zoning was less pronounced near inclusions within LC olivine. This zoning is superimposed upon any zoning seen in the olivine crystal as a whole. The modelling focused on the inclusions in the HC olivine crystals, because their initial FeO* concentration was likely to be similar to that of the whole-rock (9 wt %) or other mafic whole-rock samples of the series (9·7–10 wt %), providing a likely starting composition of the melt inclusion. However, we failed to produce a match for both the FeO profiles and the measured melt inclusion composition using 
9 wt % FeO* as the original concentration of the melt inclusions in HC olivines. The profiles could be reproduced only if initial FeO* contents of the melt were taken to be as low as 6–7 wt % (Table 4, Fig. 9). This is lower than expected for a melt in equilibrium with these olivines, and also significantly lower than the whole-rock FeO* content of the host rock, or any other mafic sample from the series. This probably signifies that the inclusions have undergone more than one period of equilibration and Fe loss, and that the profile measured around the inclusion represents only the last equilibration event. The duration of this last equilibration event varied from 4 to 194 days, but no systematic variation of the duration with the forsterite content of the enclosing olivine was observed. The temperature of last equilibration varied between 925 and 1120°C, and showed a negative correlation with the duration of the last equilibration event.
|
|
When the FeO profiles were measured around melt inclusions, it was observed that CaO contents in olivine increased and NiO content decreased towards the melt inclusion (Fig. 10). This was especially noticeable for the inclusions in the LC olivine crystals. This increase in measured CaO content in olivine next to a phase with higher CaO concentrations is sometimes ascribed to secondary fluorescence of the high-Ca phase. However, the anti-correlation of CaO and NiO in the profile, and the fact that Al2O3 concentrations do not show an apparent elevation, argues against secondary fluorescence being the main reason for this CaO increase. We think that post-entrapment equilibration between olivine and the included glass phase also explains the CaO enrichment of the olivine around the melt inclusion, analogous to the observed FeO enrichment.
|
Correction for FeO loss
As the original FeO content of the melt inclusions could not be retrieved by modelling the diffusion profile around the inclusions, the true initial FeO content of the melt could not be determined accurately. The inclusions were therefore recalculated using different initial FeO contents, varying between 9·5 and 6·5 wt % for the inclusions within HC olivine (i.e. varying between values for whole-rocks, and those obtained from the modelling), and between 4 and 7 wt % for inclusions in LC olivines (as Fe values were on average lower, and Fe loss appeared to be less pronounced in these inclusions). The maximum discrepancy observed between the measured value and a corrected value was 8 rel. % for SiO2, and up to 23 rel. % for TiO2, Al2O3, CaO, Na2O, K2O and P2O5, when a melt inclusions with a measured FeO content of 2·5 wt % was corrected to 7 wt %. Most corrections were, however, significantly smaller. To give an impression of the effect of the corrections, Fig. 11 shows selected Harker variation diagrams for the uncorrected data, the data corrected to a ‘best guess’ FeO content, and the data corrected to 7 wt % FeO. The best guess amounted to 9·5 wt % FeO for the inclusions in HC-LF and transitional HC olivines; 9 wt % FeO for the inclusions in the HC-HF olivines; 4·0 wt % FeO for the Hy-normative inclusions in LC olivine; 4·8 wt % FeO* for the Ne-normative LC-olivine hosted inclusions; and variable FeO contents between 4·4 and 7 wt % for the inclusions in transitional HC-LC olivines.
|
An important result of the systematic assessment of the FeO loss correction was that this correction did not affect the silica saturation of the inclusions: inclusions within HC olivines remained silica-undersaturated irrespective of assumed FeO* content (olivine, nepheline with or without leucite in the norm); hypersthene-normative inclusions within LC olivines remained hy-normative, and ne-normative inclusions within LC olivines remained ne-normative.
The FeO correction does, of course, change the FeO and MgO content of the inclusions significantly. As the values for these oxides are a direct reflection of the correction, they can no longer give us any relevant geological information, and have therefore not been used. As we cannot be sure that our assumed FeO contents are correct, we prefer to use the uncorrected data for further discussion. The CaO/Al2O3 ratio was used as a fractionation index in Fig. 8; the variation of this ratio with the recalculated SiO2 content is given in Fig. 11.
Trace elements and water
Trace element data, normalized to normal mid-ocean ridge basalt (N-MORB; Sun & McDonough, 1989) are shown in Fig. 12. Melt inclusions from LC and HC olivines have similar patterns for all elements except for the heavy REE (HREE) and Y (Figs 12–14). All melt inclusions show the hallmarks of subduction-related magmas, with strong enrichments in the large ion lithophile elements (LILE), such as Rb, Ba and Sr, and relative depletions for the high field strength elements (HFSE), such as Nb, Ta and Ti. The inclusions from LC and LC-HC olivine crystals generally have higher contents of LILE and LREE than the HC-olivine-hosted inclusions (Fig. 13). The former show much lower contents of Y and the HREE (Table 1), leading to Gd/Yb and Sr/Y ratios well in excess of those for the inclusions in HC olivine or for MORB. There is a good correlation between Zr, Th, the LILE and HREE (Fig. 13). No correlation is seen between trace element concentrations and the forsterite content of the olivine in which these inclusions occur.
|
|
|
The rare earth element pattern of an orthopyroxene (opx b) included in an LC olivine shows a somewhat HREE-depleted pattern, at very low absolute levels of REE (Fig. 14). When orthopyroxene–melt distribution coefficients from McKenzie & O'Nions (1991)
are used to calculate the REE pattern of the melt in equilibrium with this orthopyroxene, the result shows a clear resemblance to the melt inclusions in LC olivine crystals, with LREE enrichment and HREE depletion. The pattern shows some spikes and troughs, which are interpreted to reflect inaccuracy in the measurements of these elements in the orthopyroxene because of very low concentrations (0·3–0·02 times chondrite).
Ion microprobe measurements for H2O, Li and B were made late in the study, when few inclusions of sufficient size and quality were left for analysis. Therefore, only one Hy-normative inclusion in LC olivine was analysed (from olivine 2009), which did not have high Cl contents. Water contents in the 10 analysed inclusions varied between 0·56 and 0·94 wt %, and show a correlation with the K2O content of the glass (Table 5), but without any distinctions between the groups (as far as can be seen in this limited dataset). We are not completely convinced that these values represent primary water concentration of the magmas. The decrepitation marks around some inclusions indicate that volatiles may have been lost, and recent experiments have shown that water loss by proton diffusion through olivine enclosed in magmas that have been affected by degassing can happen in a matter of hours (Hauri, 2002).
|
Lithium contents in the inclusions in which H2O was measured varied between 7 and 20 ppm, with one outlier in an HC-LF-olivine-hosted inclusion at 33 ppm. Apart for the analysed LC-olivine-hosted inclusion, a correlation seems to exist between Na2O and Li contents of the inclusions (Table 5). Boron contents vary between 17 and 52 ppm, and a correlation exists between B and the Fo content of the enclosing olivine, with the LC-olivine-hosted inclusion having the highest B content. B and Li contents were also measured in the host olivine of one of the HC-HF group and the LC olivine, yielding olivine–liquid distributions coefficients for Li of 0·4 and 0·27, and for B of 0·009 and 0·018. These values are slightly higher than but within the same order of magnitude as the values reported by Zanetti et al. (2004)
for a hydrous basanite melt (0·22 and 0·0058 for Li and B, respectively). |
DISCUSSION |
|---|
Identification of contrasting magmas
Inclusion relationships have shown that the following phases occur together, and are therefore likely to be genetically related:
- HC-LF olivine; high-Ti spinel; clinopyroxene
- HC-HF olivine; medium-Ti spinel;
- LC olivine; low-Ti spinel; orthopyroxene.
The measured CaO and NiO contents of both the LC and HC olivine crystals are unusual when compared with olivine crystals from ‘normal’ arc volcanic rocks, which generally have 0·15–0·28 wt % CaO and 0·15–0·40 wt % NiO at Fo86–88 [Elburg, unpublished data for North Sulawesi and Alor (Indonesia), Ambryn and Yasur (Vanuatu), Santorini (Greece); Nakamura, 1995; Clynne & Borg, 1997; Tamura et al., 2000; Reubi et al., 2002; Smith & Leeman, 2005]. The HC olivine crystals have high CaO and low NiO contents compared with normal arc volcanic rocks, and, conversely, LC olivine has lower CaO and higher NiO contents. The smooth variations of CaO, NiO and forsterite contents within the HC group as a whole appear to suggest that they belong to a single fractionation series, and we propose that they co-crystallized with clinopyroxene and high-Ti spinel. This apparent fractionation series (see discussion below) accounts for the majority of phenocrysts within the sample. The melt inclusions in the HC olivine crystals are also closer to the whole-rock composition than those in LC olivine, and this suggests that the whole-rock composition (Tnv11, Table 1) is dominated by the contribution from HC melt and crystals.
If the HC-HF and HC-LF olivine crystals are members of a single fractionation series, then LC olivine cannot belong to this fractionation series, since LC olivine has far higher NiO and lower CaO contents at similar forsterite contents than HC olivine. Moreover, pyroxene enclosed within LC olivine is always orthopyroxene rather than clinopyroxene. It is therefore unlikely that LC olivine crystals were in equilibrium with the same magmas from which the HC phenocrysts formed.
Ca and Ni in olivine and melt inclusions: origin of Ne-normative LC inclusions
With respect to their low CaO and high NiO and MgO contents, LC olivine crystals resemble olivine from mantle xenoliths and peridotites (Dick, 1989; Pearce et al., 2000; Griselin, 2001; McInnes et al., 2001), even though the latter seldom exceeds 93 mol % Fo and 0·4 wt % NiO. Forsterite-rich low-Ca olivine crystals within volcanic rocks are often interpreted as mantle xenocrysts (Ramsay et al., 1984; Cameron, 1985; Boudier, 1991; Rohrbach et al., 2005); however, we suggest that the LC olivine crystallized from a melt and is not a mantle-derived xenocryst (Kamenetsky et al., 2006).
A strong argument for the magmatic interpretation is the coherence between the CaO content of the melt inclusions and their enclosing olivines. Using the olivine–melt distribution coefficients from Beattie et al. (1991), we can calculate the CaO content of the olivine with which the melt inclusions would have been in equilibrium (Fig. 15a). The distribution of calcium between olivine and melt is dependent on the MgO distribution, which is influenced by the FeO* recalculation procedure. The calculation has been performed with the ‘best guess’ FeO* recalculation, but the trends are similar if the raw data are used. The calculated olivine CaO contents correspond reasonably well to the measured CaO contents of the enclosing olivine. An exception is formed by the LC-olivine-hosted nepheline-normative melts, which give olivine CaO contents that are significantly higher than the measured concentrations. The generally good match between the calculated and actual CaO contents cannot be a result of later equilibration between olivine and melt, as CaO enrichment of the olivine close (<15 µm) to the inclusion (Fig. 10) demonstrates that diffusional equilibration for CaO has been limited, in keeping with the slow diffusion for this element in olivine (Petry et al., 2004). Although CaO concentrations in the olivine around the inclusions are low compared with those in the inclusions (0·4 wt % vs 4 wt %), the small size of the hy-normative inclusions in LC-olivine results in the CaO loss being non-negligible, and it could be as high as 15 rel. % for a typical inclusion with a diameter of 25 µm and a measured CaO content of 4 wt % (see the Appendix). This could explain why the calculated CaO contents of the enclosing olivine are generally slightly low for the hypersthene-normative inclusions in LC olivine, whereas no systematic offset is seen for the inclusions in the HC olivines. As the exact amount of CaO loss depends on the initial CaO content of the melt inclusion, its size and the diffusion profile around the inclusion, we did not apply a correction for this effect, as diffusion profiles were not measured around all inclusions.
|
The CaO contents of the melt inclusions within the olivines support the idea that the HC and Hy-normative LC inclusions could represent the melt from which the olivine crystallized. The NiO contents of the olivine can be used in a manner similar to the CaO content to calculate the Ni concentration of the parental melt. The LC olivine crystals contain significantly more NiO than HC olivine crystals, which could be interpreted as crystallization from a more NiO-rich melt. However, if we assume that the parental melt had a bulk composition similar to that of the melt inclusions (again corrected for FeO* loss), then the different melt structure between the inclusions in the two major groups of olivines would cause a significant difference in the Ni olivine–melt distribution coefficient. This compositional effect is far more pronounced for Ni than it is for Ca. Therefore, calculated Ni contents in the melts parental to LC and HC-HF olivine would not differ much, ranging between 200 and 300 ppm (Fig. 15b). If uncorrected melt inclusion compositions are used, the calculated Ni values for both groups lie between 100 and 200 ppm. Three of the Ne-normative glasses in LC olivine have higher calculated Ni contents around 350 ppm. Considering the fact that the Ne-normative inclusions did not appear to be in equilibrium with their host olivine for Ca, these calculated Ni contents are unlikely to be correct. The whole-rock contains significantly more Ni (664 ppm; Elburg & Foden, 1999
) than the calculated Ni contents of the melt inclusions. This is likely to reflect olivine accumulation, indicating that this whole-rock compositions cannot be equated with a melt composition.
The calculations suggest that the nepheline-normative melt inclusions in LC olivine are not in equilibrium with their host. The two main options for the formation of these inclusions are that they represent aliquots of HC melt that has been accidentally trapped in LC olivine; or that they do not represent liquid compositions. The first option seems unlikely, as the ne-normative LC inclusions have different Sr/Y and Gd/Yb ratios from true HC melt inclusions (Fig. 13a). We therefore think that the ne-normative melt inclusions in LC olivine represent inclusions from which some melt has escaped after partial crystallization. This would enrich the inclusion in olivine and clinopyroxene components (Portnyagin et al., 2005), whereas silicic liquid would escape. Because clinopyroxene has a higher liquid–solid distribution coefficient for the HREE, this process can also explain the reduced Sr/Y and Gd/Yb ratios compared with the hy-normative inclusions. Phlogopite might also have been a residual phase, causing an enrichment in K2O and F, whereas the higher sulphur contents could be explained by the formation of iron sulphides.
Distinction between melt inclusions in HC-HF and HC-LF olivine
The smooth variation of CaO and NiO with forsterite contents between HC-HF and HC-LF olivine is suggestive of a single fractionation series, but the very minor depletion in NiO between the two subgroups is surprising, since both clinopyroxene and olivine have high mineral/liquid distribution coefficients for this element [around two for clinopyroxene and 8–18 for olivine, using the formulation from Beattie et al. (1991)]. However, olivine–liquid distribution coefficients distribution increase strongly with fractionation, so NiO contents in olivine will decrease far more slowly during fractionation than the Ni content of the liquid with which it is in equilibrium.
More doubt is cast upon the fractionation relationship between the melts in HC-HF and HC-LF olivines by the variation of P2O5 vs CaO/Al2O3 (as a potential fractionation index) (Fig. 8), where P2O5 contents within HC-HF-olivine-hosted inclusions vary between 0·3 and 1·05 wt % at CaO/Al2O3 = 1, and within HC-LF-olivine-hosted inclusions between 0·44 and 1·03 wt % at CaO/Al2O3 = 0·65 (with 1 S.D. on P2O5 analyses 0·02 wt % at the 0·2 wt % level). Absolute levels of P2O5 (or any other oxide) are influenced by the FeO correction, but this affects concentrations by less than 25% relative, instead of the 300% observed here. Oxide ratios are not influenced by the FeO recalculation and can therefore be used more reliably to assess fractionation behaviour. A diagram of K2O/Al2O3 vs CaO/Al2O3 (Fig. 16) shows that K2O/Al2O3 ratios increase when CaO/Al2O3 ratios decrease between HC-HF- and HC-LF-hosted melt inclusions. This general trend would be expected for a fractionating assemblage of olivine (K2O/Al2O3 = 0, CaO/Al2O3 = 8–22) and clinopyroxene (K2O/Al2O3 = 0–0·02, CaO/Al2O3 = 5–25 at mg-number >84), but the K2O/Al2O3 ratio increases too fast compared with the decrease in CaO/Al2O3 for a fractionating olivine–clinopyroxene assemblage. The fractionating assemblage should have a CaO/Al2O3 of <2·5 at K2O/Al2O3 = 0, whereas all measured clinopyroxene crystals have CaO/Al2O3 >3. Although one could decrease the CaO/Al2O3 ratio of the fractionating assemblage by adding spinel, the Al2O3 content of spinel in equilibrium with forsterite-rich olivine is only 11–13 wt %, and one would need an unrealistically high percentage of spinel (20–30 wt % of the fractionating assemblage) to fit the K2O/Al2O3 to the CaO/Al2O3 ratios for fractionating from HC-HF to intermediate HC compositions.
|
2·5 at K2O/Al2O3 = 0. Measured CaO/Al2O3 ratios for clinopyroxene, which should be the main control on CaO/Al2O3 variations, vary between 25 (at mg-number 92) and 5 (at mg-number 84) at K2O/Al2O3 <0·01. Simple crystal fractionation cannot, therefore, explain the range of HC melts trapped in olivine.Simple mixing calculations also show that it is not possible to simultaneously match the (FeO*-corrected) CaO, Al2O3, K2O and SiO2 concentrations of melt inclusions in HC-HF olivines by adding any combination of analysed clinopyroxene, olivine and spinel to HC-LF olivine-hosted melt inclusions. With these calculations, Al2O3 contents are too low and SiO2 contents too high when CaO and K2O contents match measured values.
In short, there is too much scatter in oxides such as P2O5 and TiO2 at a given index of fractionation for the melts in the HC olivines to belong to a single simple fractionation series. Moreover, the trends in elemental ratios cannot be explained by fractionation of an assemblage of olivine, clinopyroxene ± Cr-spinel. It is likely that fractionation effects are superimposed upon variations in the mineralogy of the source of the melts.
LC melts
Although high CaO/Al2O3 melts have been found in many tectonic environments (Schiano et al., 2000; Kogiso & Hirschmann, 2001; Portnyagin et al., 2005), high-silica, low CaO/Al2O3 primitive melts are less common. They are either low-Ca boninites, found in hot subduction environments (Crawford et al., 1989), lamproites (Foley et al., 1987) or adakites (Defant & Drummond, 1990). For comparison, Figs 8 and 11 indicate the compositions of representative examples of these magmas. The melt inclusions found in LC olivine crystals cannot be low-Ca boninites, as the latter are characterized by lower Al2O3, Na2O, K2O and TiO2 contents (Kamenetsky et al., 2002) than the LC melts. Moreover, HREE depletion is not a characteristic of boninitic melts. Adakitic melts are well known for their HREE depletion, and correspondingly high Sr/Y and Gd/Yb ratios (Martin et al., 2005). However, these melts display lower K2O, higher Na2O contents and higher CaO/Al2O3 ratios than the LC melt inclusions described here. The closest match to the LC melts in terms of lamproitic rocks is given by orendites from central Italy (Peccerillo et al., 1988), although these contain higher K2O and less Na2O than the melt inclusions analysed. Orendites also display HREE depletion, although not as extreme as some of our LC melt inclusions.
The most likely cause of the observed HREE depletion in melt inclusions and orthopyroxenes within LC olivines is equilibration with garnet. Silica-rich melts with low Ca/Al ratios can be formed by crystallization of clinopyroxene, orthopyroxene and garnet from more primitive subduction-related magmas at high pressure and pH2O (Müntener et al., 2001). However, this process does not agree with the melt inclusions occurring in olivines with extremely high forsterite contents (Fo94), as fractional crystallization would rapidly lower the mg-number of the melt. Also, the high forsterite content of enclosing olivine does not favour generation of the liquids by direct melting of the subducted oceanic crust, or by remelting of lower crustal cumulates, unless the magmas or their sources were extremely oxidized and most Fe would occur as Fe3+. Although clinopyroxene can also retain HREE, the middle REE (MREE)/HREE ratios in the LC melt inclusions are too high to be explained by retention in clinopyroxene only (Blundy et al., 1998). The most plausible way to generate the LC-olivine-hosted melt inclusions is by partial melting of a mantle source with residual garnet, or equilibration between melt and a garnet-bearing mantle lithology. If the mantle source started out with HREE contents similar to those of a MORB source, simple batch melt modelling indicates that this source must have contained 20–25 wt % garnet in its residue during partial melting [using the garnet–melt Kd values of McKenzie & O'Nions (1991)]. The presence of garnet in the residue implies either that partial melting took place in a normal peridotitic mantle at depths greater than 90 km, or that the source had a composition that stabilized garnet relative to spinel at lower pressures. A high alumina content, as seen in pyroxenites, is one way of stabilizing garnet (Hirschmann & Stolper, 1996). This agrees with the high Al2O3 content of the melt inclusions.
We do not know of any whole-rock analysis that is a perfect match for the LC melt inclusions, but it is possible that the existence of this type of melts is more common than hitherto recognized, as only detailed mineralogical and melt inclusion studies allow us to recognize their existence. Low-Ca olivine is increasingly being reported from subduction settings (Ramsay et al., 1984; Rohrbach et al., 2005), but is commonly interpreted as a mantle xenocryst, and not studied to any great extent. We suggest that some characteristics of these low Ca/Al melts stop them from ever reaching the surface. Although the one measured H2O content was not very high, we failed to analyse inclusions with high Cl contents. It is possible that these have higher water contents, and that the melts reached water saturation before reaching the surface, resulting in degassing and freezing of the melts below the surface, similar to the model proposed by Tamura & Tasumi (2002) for the Izu–Bonin arc.
Similarities and contrasts between LC and HC melts and their sources
The melt inclusions within the LC and HC olivine crystals display very similar trace element patterns, apart from the HREE and Y. They show similar relative depletions in the HFSE and enrichments in the LILE, and in the relative levels of Cs–Rb–Ba enrichment, which are the elements that tend to be variable in arc volcanic rocks, possibly reflecting selective retention in mica or amphibole (Elburg et al., 2002a). Also, the good correlation for the dataset as a whole between elements that are thought to be mobile in the fluid phase (such as Rb), and are consequently enriched in arc volcanic rocks relative to MORB, and a fluid-immobile element such as Zr (Fig. 13) suggests that the mantle sources of the LC and HC melts may have had a similar trace element contents.
Despite these similarities, there are significant distinctions between the melt inclusions in LC and HC olivines, with inclusions from LC olivine being calcium poor, silica saturated, chlorine rich and HREE depleted, and those in HC olivine calcium rich and silica undersaturated. HC-olivine-hosted melt inclusions only show a very slight depletion in HREE compared with a MORB source, indicating that garnet is a minor residual phase, if at all. The high CaO contents indicate that the source may have contained significant amounts of clinopyroxene, which entered the melt. Even though the residual mineral assemblages for the LC and HC sources are obviously very different, the similarity in trace element patterns of the melts points towards a close relationship between the sources.
Anomalous source composition
We undertook some exploratory runs with the program MELTS (for pressures <10 kbar) and pMELTS (pressures 10–20 kbar) (Ghiorso et al., 2002) to investigate the possibility of deriving high CaO/Al2O3 and low CaO/Al2O3 melts from similar sources. Several other workers have attempted to model high CaO/Al2O3 melts with the MELTS program (Schiano et al., 2000; Kogiso & Hirschmann, 2001), and found, not unexpectedly, that clinopyroxenite sources give the best results. We obtained the best result with a clinopyroxenite composition that is somewhat more aluminous than the compositions used by other workers (Schiano et al., 2000; Kogiso & Hirschmann, 2001) to model the Ne-normative high CaO/Al2O3 melts of the HC olivines (Table 6). If we keep most elements the same, but lower the CaO and increase the Al2O3 content of this source, its bulk composition changes from being nepheline-normative to being hypersthene-normative. This then has the desired effect of giving melt compositions that are hypersthene-normative, and leaving a garnet-bearing residue at 20 kbar. This is the smallest change in the source composition that would give the desired effect of obtaining liquids that are very different in terms of major elements. As subduction-zone fluids are thought to be rich in components derived from the breakdown of aluminosilicates (Manning, 2004), it is feasible that these changes to the mantle are brought about by interaction with this type of fluid.
|
It is also possible that the distinctions in the LC and HC source's residual mineral assemblage reflect changes in pressure instead of chemical composition. However, an important effect of a pressure increase is to generate melts that are more silica undersaturated (O'Hara, 1968
), whereas the LC melts are actually more silica oversaturated than the HC melts. A greater depth of melt generation is therefore unlikely to be the only reason for the presence of garnet in the source. Because partial melting has the opposite effect on SiO2 for Hy- and Ne-normative source compositions, the melts of the HC source become lower in silica with decreasing degree of partial melting. To obtain melt compositions with a SiO2 content around 46–47 wt %, as in the melt inclusions in HC-HF olivines, the (Ne-normative) source needs to be more Si rich than the (Hy-normative) source of the LC melt inclusions (with >50 wt % SiO2). The degree of melting also needs to be larger than for the LC inclusions, to obtain the desired high Ca/Al ratios. To avoid having to postulate extremely high temperatures, it is most likely that this took place at shallower depth than for the LC inclusions. A similar melt to the one produced at 1450°C and 20 kbar can be produced at 5 kbar at only 1275°C. The larger degree of partial melting leads to rather low K2O contents in the melt compared with observed values (1·4 vs >3·5 wt %) if the K2O content of the source is not increased.
As it remains problematic to find a perfect match for this type of liquid by single-stage melting models, it is possible that some of the liquids' geochemical characteristics were obtained during later equilibration with garnet- or clinopyroxene host-rocks. As the equilibrated liquids were still capable of crystallizing Fo90 olivine, the equilibration process cannot have involved significant crystallization and must therefore have happened at high temperatures.
Influence of volatile phases
The presence and composition of a volatile phase during partial melting may also have played a role in generating the distinctions between HC and LC liquids. Unfortunately, the pMELTS software can incorporate only H2O in partial melting models, not CO2, so we cannot assess this option by computer modelling. It is well known that the presence of water expands the stability field of olivine, and causes the eutectic composition in a water-bearing mantle assemblage to have a relatively high SiO2 content (Nicholls, 1974), whereas CO2 has the opposite effect (Mysen & Boettcher, 1975). We notice that LC melt inclusions have higher Cl contents and lower electron microprobe totals than our HC inclusions, and this could point towards a higher dissolved water content in the HC melt component. Alternatively, the high H2O and SiO2 contents could both be the result of low degrees of partial melting, rather than the high water content being the cause of the high SiO2 content. The importance of CO2 in the magmatic system is evident from our fluid inclusion studies. CO2-rich fluids have also been proposed as an explanation for high-calcium subduction-related magmatism (Green et al., 2004), although the melts generated in that study were generally hypersthene- instead of nepheline-normative, like our HC melts. The melts produced by Mysen & Boettcher (1975) from a lherzolitic source rock under high-CO2 conditions were, however, nepheline-normative.
The association of high-Ca melts and water-rich low-Ca melts could also have a bearing on clinopyroxenite–dunite massifs hosting Alaskan- (or Ural)-type Pt deposits, commonly associated with chromitites (Malitch & Thalhammer, 2002). The clinopyroxenite–dunite association indicates a possible genetic relationship to the high-Ca melts associated with the HC olivines. Chromitite segregation, however, appears to be favoured by the presence of a water-rich fluid (Matveev & Ballhaus, 2002), which may point towards a role for a water-rich melt similar to the one associated with the LC olivines. Interestingly, recent work on silicate inclusions within Pt–Fe alloys (Johan, 2006) indicates the presence of low-calcium, high-aluminium (CaO/Al2O3 = 0·28), silicic, water-rich HP melts associated with Uralian–Alaskan-type Pt mineralization.
Preferred model
Whether the coexistence of primitive high-Ca and low-Ca melts is typical for post-subduction magmatism, of which the studied sample is an example, needs further investigation. We envisage a model in which the subducting slab stalls as a result of continental collision, causing an increase in slab temperature with time, as also proposed by Elburg & Foden (1999) for South Sulawesi on the basis of isotope data. This temperature increase could cause the breakdown of hydrous minerals and carbonates that would otherwise have remained stable in the slab at this depth. Water-bearing fluids (or hydrous silicic melts) from the slab would metasomatize the mantle and cause melting. Considering the fact that the presence of water significantly reduces the mantle solidus, and the fact that the mantle near the slab would be warmer than during a normal non-collisional subduction setting, melting took place at relatively high pressures, and in equilibrium with garnet. This would be the source for LC melts. As water would react with the surrounding mantle and become incorporated in melt or water-bearing mineral phases, the remaining fluids would become enriched in CO2 (Schuiling, 2004) and continue upwards through the mantle wedge. Melting may then take place at shallower depths in the presence of a CO2-rich fluid, and generate HC melts. These can continue their upward journey to the surface, but the hydrous LC melts become water saturated at shallower depths and freeze. Therefore, they never erupt, but can only reach the surface if they are carried by their companion HC magma.
|
CONCLUSIONS |
|---|
Sample Tnv96-1, on which this mineral and melt inclusion study was performed, is one of several mafic ankaramitic samples within a suite of alkali-rich undersaturated magmas that erupted between 6 and 9 Ma in South Sulawesi. Although their extremely mafic whole-rock compositions reflect cumulate processes, the Sr, Nd and Pb isotopic composition of these samples falls within the range of this suite of samples from South Sulawesi (Elburg & Foden, 1999
). The results from the present study can therefore be interpreted to give information on the magma parental to this syn- to post-collisional suite of magmas. The parental magmas are clearly represented by the silica-undersaturated high-Ca suite of melt inclusions, rather than the silica-saturated low-Ca suite. However, the close resemblance in LILE, LREE and HFSE trace element signature between the two types of melt inclusions suggests a genetic relationship between the two melt types, rather than accidental admixture of xenolithic olivine into a high-Ca magma. In light of the good correlation between the composition of the melt inclusion and enclosing olivine, it is unlikely that local melting processes in a cumulate-bearing magma chamber can be blamed for the unusual melt compositions, as suggested by Danyushevsky et al. (2004). The depletion in Y and HREE seen in the high-silica, low-Ca melt inclusions argues for the presence of garnet in the source. The high silica content of these inclusions suggests high water pressure during melt generation. However, the exsolved volatile phase in these inclusions, like that in the high-Ca inclusions, is virtually pure CO2. This may reflect the lower solubility of CO2, and thereby enhanced exsolution at lower pressures, in silicate melts, but also loss of H2O by decrepitation and proton diffusion. It is unavoidable that the latter process has been operational, considering the high diffusivity of hydrogen in olivine and the long storage of these olivines at magmatic temperatures, as shown by the diffusional profiles around the melt inclusions. The high-Ca inclusions have lower silica contents, and are likely to have formed by melting from an H2O-poor, CO2-rich source. The fact that many of the most mafic magmas in arc environments show ankaramitic characteristics may reflect the more limited role of degassing-induced fractional crystallization in these H2O-poor magmas. In contrast, the absence of the high-silica magmas, similar to the low-Ca melt inclusion, as a whole-rock composition within the arc environment may reflect the unavoidable freezing of this magma type because of degassing during ascent and depressurization.
We prefer a model in which the LC and HC melts are formed by melting of normal (non-pyroxenitic) mantle, which was modified by fluids or melts from the stalled subducted slab. These fluids were dominated by H2O for the LC melts, and by CO2 for HC melts, which also formed at shallower depths than the LC melts. Whether the coexistence of these two magma types is a local peculiarity or is more common in (post-collisional) subduction environments, and a potential link with Alaskan-type Pt deposits, needs further investigation.
|
SUPPLEMENTARY DATA |
|---|
Supplementary data for this paper are available at Journal of Petrology online.
|
APPENDIX: MODELLING OF CAO LOSS TO OLIVINE |
|---|
The distance–CaO concentration gradient in the olivine around the inclusion was described by the function
, which is the solution to diffusion in a semi-infinite medium with constant surface concentration (Albarède, 1995). C0 was taken to be the CaO content at an ‘infinite’ distance from the melt inclusion, and Cint was taken as the CaO content closest to the inclusion. The value for Dt was adjusted to match the calculated profile to the observed one. The actual values for D and t are of no consequence for this model. The amount of CaO lost to the olivine can be calculated from a hypothetical ‘boundary layer’, which is a layer with a thickness 
with a homogeneous concentration Cint. This is the analytical equivalent of calculating the surface under the diffusional profile from x = 0 and x = 
, and Cint to C0. Therefore, the amount of CaO lost to the olivine is CaOlost = (4/3)
(
– r)3(Cint – C0), with r being the radius of the melt inclusion. The original CaO concentration of the melt inclusion is then [(4/3)(r)3CaOmelt + (3·25/2·77) CaOlost]/(4/3)(r)3. The factor (3·25/2·77) accounts for the assumed density difference between the melt inclusion and the olivine. |
ACKNOWLEDGEMENTS |
|---|
This research was funded by a European Union Marie Curie Fellowship to the first author. Al Hofmann is thanked for providing the opportunity to pursue this research at the Max Planck Institute for Chemistry in Mainz. The electron microprobe facilities at the Max Planck Institute were financed by a Wolfgang Paul award to A. Sobolev. Stephan Klemme helped with the homogenization experiments in Heidelberg; Ronald Bakker provided access to and help with the Raman spectrometer in Leoben. Andrey Gurenko gave advice on preparation of the melt inclusions, and Nora Groschkopf assisted with electron microprobe analyses. Eric Hellebrand volunteered his help for the ion microprobe analyses, and Brigitte Stoll and Kirsten Herweg helped with the LA–ICP–MS analyses. Igor Nikogosian was supported by ISES (Netherlands Research Centre for Integrated Solid Earth Science). Leonid Danyushevsky is acknowledged for lively discussions and providing the FeO-recalculation program. This paper benefited from thorough reviews by M. Portnyagin and Y. Tamura, and meticulous editorial handling by John Gamble.
*Corresponding author. Present address: Department of Geology and Soil Science, Ghent University, Krijgslaan 281 S8, 9000 Ghent, Belgium. Telephone: +32-9-2644566. Fax: +32-9-2644984. E-mail: Marlina.Elburg{at}Ugent.be
|
REFERENCES |
|---|
Albarède F. (1995) Introduction to Geochemical Modelling(Cambridge University Press, Cambridge) pp. 543.
Bakker R. J. (2003) Package FLUIDS1. Computer programs for analysis of fluid inclusion data and for modelling bulk fluid properties. Chemical Geology 194:3–23.[CrossRef][Web of Science]
Ballhaus C., Berry R. F., Green D. H. (1991) High pressure experimental calibration of the olivine–orthopyroxene–spinel oxygen geobarometer: implications for the oxidation state of the upper mantle. Contributions to Mineralogy and Petrology 107:27–40.[CrossRef][Web of Science]
Beattie P., Ford C., Russel D. (1991) Partition coefficients for olivine–melt and orthopyroxene–melt systems. Contributions to Mineralogy and Petrology 109:212–224.[CrossRef][Web of Science]
Blundy J. D., Robinson J. A. C., Wood B. J. (1998) Heavy REE are compatible in clinopyroxene on the spinel lherzolite solidus. Earth and Planetary Science Letters 160:493–504.[CrossRef][Web of Science]
Boudier F. (1991) Olivine xenocrysts in picritic magmas: an experimental and microstructural study. Contributions to Mineralogy and Petrology 109:114–123.[CrossRef][Web of Science]
Burke E. A. J. (2001) Raman microspectrometry of fluid inclusions. Lithos 55:139–158.[CrossRef][Web of Science]
Cameron W. E. (1985) Petrology and origin of primitive lavas from the Troodos ophiolite, Cyprus. Contributions to Mineralogy and Petrology 89:239–255.[CrossRef][Web of Science]
Cervantes P. and Wallace P. J. (2003) Role of H2O in subduction-zone magmatism: new insights from melt inclusions in high-Mg basalts from central Mexico. Geology 31:235–238.
Clynne M. A. and Borg L. A. (1997) Olivine and chromian spinel in primitive calc-alkaline and tholeiitic lavas from the southernmost Cascade Range, California: a reflection of relative fertility of the source. Canadian Mineralogist 35:453–472.[Web of Science]
Costa-Rodriguez F. and Chakraborty S. (2004) Decadal time gaps between mafic intrusion and silicic eruption obtained from chemical zoning patterns in olivine. Earth and Planetary Science Letters 227:517–530.[CrossRef][Web of Science]
Crawford A. J., Falloon T. J., Green D. H. (1989) Classification, petrogenesis and tectonic setting of boninites. Boninites and Related Rocks (Unwin HymanIn Crawford A. J. (Ed.). , London)2–49.
Danyushevsky L. V., Della-Pasqua F. N., Sokolov S. (2000) Re-equilibration of melt inclusions trapped by magnesian olivine phenocrysts from subduction-related magmas: petrological implications. Contributions to Mineralogy and Petrology 138:68–83.[CrossRef][Web of Science]
Danyushevsky L. V., Sokolov S., Falloon T. J. (2002) Melt inclusions in olivine phenocrysts: using diffusive re-equilibration to determine the cooling history of a crystal, with implications for the origin of olivine-phyric volcanic rocks. Journal of Petrology 43:1651–1671.
Danyushevsky L. V., Leslie R. A. J., Crawford A. J., Durance P. (2004) Melt inclusions in primitive olivine phenocrysts: the role of localized reaction processes in the origin of anomalous compositions. Journal of Petrology 45:2531–2553.
Davidson J. P. (1996) Deciphering mantle and crustal signatures in subduction zone magmatism. In Bebout G. E., Scholl D. W., Kirby S. H., Platt J. P. (Eds.). Subduction: Top to Bottom. Geophysical Monograph, American Geophysical Union 96: pp. 251–262.
Defant M. J. and Drummond M. S. (1990) Derivation of some modern arc magmas by melting of young subducted lithosphere. Nature 347:662–665.[CrossRef]
Dick H. J. B. (1989) Abyssal peridotites, very slow spreading ridges and ocean ridge interaction. In Saunders A. D. and Norry M. J. (Eds.). Magmatism in the Ocean Basins. Geological Society, London, Special Publications 42: pp. 71–105.
Elburg M. A. and Foden J. (1999) Geochemical response to varying tectonic settings: an example from southern Sulawesi (Indonesia). Geochimica et Cosmochimica Acta 63:1155–1172.[CrossRef][Web of Science]
Elburg M. A., van Bergen M. J., Hoogewerff J., Foden J., Vroon P. Z., Zulkarnain I., Nasution A. (2002a) Geochemical trends across an arc–continent collision zone: magma sources and slab–wedge transfer processes below the Pantar Strait volcanoes (Indonesia). Geochimica et Cosmochimica Acta 66:2771–2789.[CrossRef][Web of Science]
Elburg M. A., van Leeuwen T., Foden J., Muhardjo J. (2002b) Origin of geochemical variability by arc–continent collision in the Biru area, southern Sulawesi (Indonesia). Journal of Petrology 43:1–24.
Elburg M. A., van Leeuwen T., Foden J., Muhardjo J. (2003) Spatial and temporal isotopic domains of contrasting igneous suites in Western and Northern Sulawesi, Indonesia. Chemical Geology 199:243–276.[Web of Science]
Evensen N. M., Hamilton P. J., O'Nions R. K. (1978) Rare-earth abundances in chondritic meteorites. Geochimica et Cosmochimica Acta 42:1199–1212.[CrossRef][Web of Science]
Foley S. F., Venturelli G., Green D. H., Toscani L. (1987) The ultrapotassic rocks: characteristics, classification, and constraints for petrogenetic models. Earth-Science Reviews 24:81–134.
Frezzotti M.-L. (2001) Silicate–melt inclusions in magmatic rocks: applications to petrology. Lithos 55:273–299.[CrossRef][Web of Science]
Ghiorso M. S., Hirschmann M. M., Reiners P. W., Kress V. C. I. (2002) The pMELTS: a revision of MELTS for improved calculation of phase relations and major element partitioning related to partial melting of the mantle to 3 GPa. Geochemistry, Geophysics, Geosystems 3: 10.1029/2001GC000217.
Gioncada A., Clocchiatti R., Sbrana A., Bottazzi P., Massare D., Ottolini L. (1998) A study of melt inclusions at Vulcano (Aeolian Islands, Italy): insights on the primitive magmas and on the volcanic feeding system. Bulletin of Volcanology 60:286–306.[CrossRef][Web of Science]
Green D. H., Schmidt M. W., Hibberson W. O. (2004) Island-arc ankaramites: primitive melts from fluxed refractory lherzolitic mantle. Journal of Petrology 45:391–403.
Griselin M. (2001) Geochemical and isotopic study of the Xigaze and Luobusa ophiolite massifs. Ph.D. thesis,Free University Amsterdam216.
Gurenko A. A., Belousov A. B., Trumbull R. B., Sobolev A. V. (2005) Explosive basaltic volcanism of the Chikurachki Volcano (Kurile Islands, Russia): petrology, geochemistry and volatile loading revealed from inclusions in minerals. Journal of Volcanology and Geothermal Research 147:203–232.[CrossRef]
Harris D. M. and Anderson A. T. J. (1984) Volatiles H2O, CO2, and Cl in subduction related basalt. Contributions to Mineralogy and Petrology 87:120–128.[CrossRef][Web of Science]
Hauri E. H. (2002) SIMS analysis of volatiles in volcanic glasses: 2. Isotopes and abundances in Hawaiian melt inclusions. Chemical Geology 183:115–141.[CrossRef][Web of Science]
Hellebrand E., Snow J. E., Hoppe P., Hofmann A. W. (2002) Garnet-field melting and late-stage refertilization in ‘residual’ abyssal peridotites from the Central Indian Ridge. Journal of Petrology 43:2305–2338.
Hirschmann M. M. and Stolper E. M. (1996) A possible role for garnet pyroxenite in the origin of the ‘garnet signature’ in MORB. Contributions to Mineralogy and Petrology 124:185–208.[CrossRef][Web of Science]
Jochum K. P., Dingwell D. B., Rocholl A., et al. (2000) The preparation and preliminary characterisation of eight geological MPI-DING reference glasses for in-situ microanalysis. Geostandards Newsletter 24:87–133.[CrossRef][Web of Science]
Johan Z. (2006) Platinum-group minerals from placers related to the Nizhni Tagil (Middle Urals, Russia) Uralian–Alaskan-type ultramafic complex: ore-mineralogy and study of silicate inclusions in (Pt, Fe) alloys. Mineralogy and Petrology 87:1–30.
Kamenetsky V. and Clocchiatti R. (1996) Primitive magmatism of Mt Etna: insights from mineralogy and melt inclusions. Earth and Planetary Science Letters 142:553–572.[CrossRef][Web of Science]
Kamenetsky V. S., Sobolev A. V., Eggins S. M., Crawford A. J., Arculus R. J. (2002) Olivine-enriched melt inclusions in chromites from low-Ca boninites, Cape Vogel, Papua New Guinea: evidence for ultramafic magma, refractory mantle source and enriched components. Chemical Geology 183:287–303.[CrossRef][Web of Science]
Kamenetsky V. S., Elburg M. A., Arculus R. J., Thomas R. (2006) Magmatic origin of low-Ca olivine in subduction-related magmas: co-existence of contrasting magmas. Chemical Geology 233:346–357.[CrossRef][Web of Science]
Kent A. J. R. and Elliott T. R. (2002) Melt inclusions from Marianas arc lavas: implications for the composition and formation of island arc magmas. Chemical Geology 183:263–286.[CrossRef][Web of Science]
Kogiso T. and Hirschmann M. M. (2001) Experimental study of clinopyroxenite partial melting and the origin of ultra-calcic melt inclusions. Contributions to Mineralogy and Petrology 142:347–360.[Web of Science]
Luhr J. F. (2001) Glass inclusions and melt volcatile contents at Paricutin Volcano, Mexico. Contributions to Mineralogy and Petrology 142:261–283.[Web of Science]
Malitch K. N. and Thalhammer O. A. R. (2002) Pt–Fe nuggets derived from clinopyroxenite–dunite massifs, Russia: a structural, compositional and osmium-isotope study. Canadian Mineralogist 40:395–418.
Manning C. E. (2004) The chemistry of subduction-zone fluids. Earth and Planetary Science Letters 23:1–16.
Martin H., Smithies R. H., Rapp R., Moyen J.-F., Champoin D. (2005) An overview of adakite, tonalite–trondhjemite–granodiorite (TTG) and sanukitoid: relationships and some implications for crustal evolution. Lithos 79:1–24.[CrossRef][Web of Science]
Matveev S. and Ballhaus C. (2002) Role of water in the origin of podiform chromitite deposits. Earth and Planetary Science Letters 203:235–243.[CrossRef][Web of Science]
McInnes B. I. A., Gregoire M., Binns R., Herzig P., Hannington M. D. (2001) Hydrous metasomatism of oceanic sub-arc mantle, Lihir, Papua New Guinea: petrology and geochemistry of fluid-metasomatised mantle wedge xenoliths. Earth and Planetary Science Letters 188:169–183.[CrossRef][Web of Science]
McKenzie D. and O'Nions R. K. (1991) Partial melt distributions from inversion of rare earth element concentrations. Journal of Petrology 32:1021–1091.
Métrich N., Allard P., Spilliaert N., Andronico D., Burton M. (2004) 2001 flank eruption of the alkali- and volatile-rich primitive basalt responsible for Mount Etna's evolution in the last three decades. Earth and Planetary Science Letters 228:1–17.[CrossRef][Web of Science]
Müntener O., Kelemen P. B., Grove T. L. (2001) The role of H2O during crystallization of primitive arc magmas under uppermost mantle conditions and genesis of igneous pyroxenites. Contributions to Mineralogy and Petrology 141:643–658.[Web of Science]
Mysen B. O. and Boettcher A. L. (1975) Melting of a hydrous mantle: II. Geochemistry of crystals and liquids formed by anatexis of mantle peridotite at high pressures and high temperatures as a function of controlled activities of water, hydrogen, and carbon dioxide. Journal of Petrology 16:549–593.
Nakamura M. (1995) Residence time and crystallization history of nickeliferous olivine phenocrysts from the northern Yatsugatake volcanoes, central Japan; applcation of growth and diffusion model in the system Mg–Fe–Ni. Journal of Volcanology and Geothermal Research 66:81–100.[CrossRef][Web of Science]
Nicholls I. A. (1974) Liquids in equilibrium with peridotitic mineral assemblages at high water pressures. Contributions to Mineralogy and Petrology 45:289–316.[CrossRef][Web of Science]
Nikogosian I. K., Elliott T., Touret J. L. R. (2002) Melt evolution beneath thick lithosphere: a magmatic inclusion study of La Palma, Canary Islands. Chemical Geology 183:169–193.[CrossRef][Web of Science]
O'Hara M. J. (1968) The bearing of phase equilibria studies in synthetic and natural systems on the origin and evolution of basic and ultrabasic rocks. Earth-Science Reviews 4:69–133.
Pearce J. A., Barker P. F., Edwards S. J., Parkinson I. J., Leat P.T. (2000) Geochemistry and tectonic significance of peridotites from the South Sandwich arc-basin system, South Alantic. Contributions to Mineralogy and Petrology 139:36–53.[CrossRef][Web of Science]
Peccerillo A., Poli G., Serri G. (1988) Petrogenesis of orenditic and kamafugitic rocks from central Italy. Canadian Mineralogist 26:45–65.[Web of Science]
Petry C., Chakraborty S., Palme H. (2004) Experimental determination of Ni diffusion coefficients in olivine and their dependence on temperature, composition, oxygen fugacity, and crystallographic orientation. Geochimica et Cosmochimica Acta 68:4179–4188.[CrossRef][Web of Science]
Portnyagin M. V., Mironov N. L., Matveev S. V., Plechov P. Y. (2005) Petrology of avachites, high-magnesian basalts of Avachinsky Volcano, Kamchatka: II. Melt inclusions in olivine. Petrology 13:358–388.
Ramsay W. R. H., Crawford A. J., Foden J. D. (1984) Field setting, mineralogy, chemistry, and genesis of arc picrites, New Georgia, Solomon Islands. Contributions to Mineralogy and Petrology 88:386–402.[CrossRef][Web of Science]
Reubi O., Nicholls I. A., Kamenetsky V. S. (2002) Early mixing and mingling in the evolution of basaltic magmas: evidence from phenocryst assemblages, Slamet Volcano, Java, Indonesia. Journal of Volcanology and Geothermal Research 119:255–274.[CrossRef]
Roedder E. (1984) Fluid Inclusions. Mineralogical Society of America, Reviews in Mineralogy 12: pp. 646.
Rohrbach A., Schuth S., Ballhaus C., Münker C., Matveev S., Qopoto C. (2005) Petrological constraints on the origin of arc picrites, New Georgia Group, Solomon Islands. Contributions to Mineralogy and Petrology 149:685–698.[CrossRef][Web of Science]
Schiano P., Eiler J.M., Hutcheon I. D., Stolper E. M. (2000) Primitive CaO-rich, silica-undersaturated melts in island arcs: evidence for the involvement of clinopyroxene-rich lithologies in the petrogenesis of arc magmas. Geochemistry, Geophysics, Geosystems 1: 1999GC000032.
Schuiling R. D. (2004) Thermal effects of massive CO2 emissions associated with subduction volcanism. Comptes Rendus, Geoscience 336:1053–1059.[CrossRef]
Sisson T. W. and Bronto S. (1998) Evidence for pressure-release melting beneath magmatic arcs from basalt at Galunggung, Indonesia. Nature 391:883–886.[CrossRef]
Sisson T. W. and Layne G. D. (1993) H2O in basalt and basaltic andesite glass inclusions from four subduction-related volcanics. Earth and Planetary Science Letters 117:619–635.[CrossRef][Web of Science]
Smith D. R. and Leeman W. P. (2005) Chromian spinel–olivine phase chemistry and the origin of primitive basalts of the southern Washington Cascades. Journal of Volcanology and Geothermal Research 140:49–66.[CrossRef][Web of Science]
Sobolev A. V., Dmitriev L. V., Barsukov V. L., Nevsorov V. N., Slutsky A. B. (1980) The formation conditions of high magnesium olivines from the monomineralic fraction of Luna-24 regolith. Geochimica et Cosmochimica Acta pp. 105–116.
Sobolev A. V., Hofmann A. W., Nikogosian I. K. (2000) Recycled oceanic crust observed in ‘ghost plagioclase’ within the source of Mauna Loa lavas. Nature 404:986–990.[CrossRef][Medline]
Stix J., Layne G. D., Williams S. N. (2003) Mechanisms of degassing at Nevado del Ruiz volcano, Colombia. Journal of the Geological Society, London 160:507–521.
Straub S. M. and Layne G. D. (2003) Decoupling of fluids and fluid-mobile elements during shallow subduction: evidence from halogen-rich andesite melt inclusions from the Izu volcanic arc front. Geochemistry, Geophysics, Geosystems 4: 2002GC000349.
Sun S.-s. and McDonough W. F. (1989) Chemical and isotopic systematics of oceanic basalts: implications for mantle composition and processes. In Saunders A. D. and Norry M. J. (Eds.). Magmatism in the Ocean Basins. Geological Society, London, Special Publications 42: pp. 313–345.
Tamura Y. and Tatsumi Y. (2002) Remelting of an andesitic crust as possible origin for rhyolitic magma in oceanic arcs: an example from the Izu–Bonin arc. Journal of Petrology 43:1029–1047.
Tamura Y., Yuhara M., Ishii T. (2000) Primary arc basalts from Daisen Volcano, Japan: equilibrium crystal fractionation versus disequilibrium fractionation during supercooling. Journal of Petrology 41:431–448.
Tepley F. J., Davidson J. P., Tilling R. I., Arth J. G. (2000) Magma mixing, recharge and eruption histories recorded in plagioclase phenocrysts from El Chichón Volcano, Mexico. Journal of Petrology 41:1397–1411.
Turner S., George R., Jerram D. A., Carpenter N., Hawkesworth C. J. (2003) Case studies of plagioclase growth and residence times in island arc lavas and a model to reconcile discordant age information. Earth and Planetary Science Letters 214:279–294.[CrossRef][Web of Science]
Zanetti A., Tiepolo M., Oberti R., Vannucci R. (2004) Trace-element partitioning in olivine: modelling of a complete data set from a synthetic hydrous basanite melt. Lithos 75:39–54.[CrossRef][Web of Science]
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  G. Moore Interpreting H2O and CO2 Contents in Melt Inclusions: Constraints from Solubility Experiments and Modeling Reviews in Mineralogy and Geochemistry, January 1, 2008; 69(1): 333 - 362. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||