Journal of Petrology Advance Access originally published online on August 27, 2007
Journal of Petrology 2007 48(10):1999-2031; doi:10.1093/petrology/egm048
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Wet and Dry Basalt Magma Evolution at Torishima Volcano, Izu–Bonin Arc, Japan: the Possible Role of Phengite in the Downgoing Slab
1Institute for Research on Earth Evolution (IFREE), Japan Agency for Marine–Earth Science and Technology (JAMSTEC), Yokosuka 237-0061, Japan
2Institute of Geoscience, Geological Survey of Japan/Aist, Tsukuba 305-8567, Japan
3Smithsonian Institution, Washington, DC 20560, USA
RECEIVED OCTOBER 18, 2006; ACCEPTED JULY 27, 2007
 |
ABSTRACT |
|---|
TOP
ABSTRACT
INTRODUCTIONThe arc-front volcanoes of Sumisu (31·5°N, 140°E) and Torishima (30·5°N, 140·3°E) in the central Izu–Bonin arc are similar in size and rise as relatively isolated edifices from the seafloor. Together they provide valuable along-arc information about magma generation processes. The volcanoes have erupted low-K basalts originating from both wet and dry parental basaltic magmas (low-Zr basalts and high-Zr basalts, respectively). Based on models involving fluid-immobile incompatible element ratios (La/Sm), the parental basalts appear to result from different degrees of partial melting of the same source mantle (
20% and 10% for wet and dry basalt magmas, respectively). Assuming that the wet basalts contain greater abundances of slab-derived components than their dry counterparts, geochemical comparison of these two basalt types permits the identification of the specific elements involved in fluid transport from the subducting slab. Using an extensive set of new geochemical data from Torishima, where the top of the downgoing slab is about 100 km deep, we find that Cs, Pb, and Sr are variably enriched in the low-Zr basalts, which cannot be accounted for by fractional crystallization or by differences in the degree of mantle melting. These elements are interpreted to be selectively concentrated in slab-derived metasomatic fluids. Variations in K, high field strength element and rare earth element concentrations are readily explained by variations in the degree of melting between the low- and high-Zr basalts; these elements are not contained in the slab-derived fluids. Rb and Ba exhibit variable behaviour in the low-Zr basalts, ranging from immobile, similar to K, to mildly enriched in some low-Zr basalts. We suggest that the K-rich mica, phengite, plays an important role in determining the composition of fluids released from the downgoing slab. In arc-front settings, where slab depth is 
100 km, phengite is stable, and the fluids released from the slab contain little K. In back-arc settings, however, where the slab is at 100–140 km depth, phengite is unstable, and K-rich fluids are released. We conclude that cross-arc variations in the K content of arc basalts are probably related to differing compositions of released fluids or melts rather than the widely held view that such variations are controlled by the degree of partial melting. KEY WORDS: arc volcano; degrees of melting; mantle wedge; water; wet and dry basalts
|
INTRODUCTION |
|---|
The Izu–Bonin volcanic arc is an excellent example of an intra-oceanic convergent margin system (Fig. 1). The geochemistry of volcanic front magmatism at Sumisu volcano is well known from studies of Sumisu island, as well as dredging and ROV or submersible dives in the adjacent Sumisu submarine caldera (Tamura et al., 2005
; Shukuno et al., 2006; Tani et al., 2007). Evidence for involvement of both dry and wet basalts as parental liquids in the genesis of Sumisu volcano was presented by Tamura et al. (2005). The chief aims of this study are: (1) to interpret new data from the nearby arc-front volcano Torishima; (2) to consider the petrogenesis of the two types of basalt to gain an insight into the nature of the slab-derived component; (3) to evaluate the role of phengite in the subducting slab. Torishima lies 110 km SSE of Sumisu, and the two volcanoes rise as relatively isolated edifices of similar size from the arc-front ridge (Fig. 1). A combined study of these two volcanic systems can provide a relatively simple 2-D along-arc perspective of magma genesis in an environment little contaminated by complex pre-volcanic histories.
|
Elliott et al. (1997
) documented island-to-island geochemical variations among the Mariana arc volcanoes to the south. Their two-component model suggests that the main differences in the composition of the lavas erupted from different islands of the Marianas are due to the addition of variable amounts of melted sediment to their sources. Signatures of aqueous fluid addition are inversely correlated with indices of sediment melt contribution, which is likely only if the aqueous fluid flux across the entire arc is fairly constant. Our study of Torishima and that of Sumisu by Tamura et al. (2005) focuses on variations within and between these two volcanoes, based on samples obtained from both the subaerial and submarine parts.
From this study, we conclude that the arc-front volcano Torishima contains basalts originating from both wet and dry parental basalt magmas, a situation similar to that found at Sumisu. The extensive dataset for Torishima makes it possible to compare these two types of basalt in detail. Based on the assumption that wet basalts contain more slab-derived fluids than dry basalts, it is shown that Cs, Pb and Sr are selectively concentrated in the slab-derived fluids below this frontal volcano, whereas potassium, high field strength elements (HFSE) and rare earth elements (REE) are not. The behaviour of Rb and Ba is intermediate between that of Th and K and Cs and Sr, suggesting that they are mildly immobile in the aqueous fluid. We suggest that phengite (a potassium-rich mica) plays a role in temporarily immobilizing K, partitioning the large ion lithophile elements (LILE) into fluids and causing fractionation of LILE (Cs/Rb and Cs/Ba) below the volcanic front (Schmidt, 1996; Domanik & Holloway, 1996; Melzer & Wunder, 2000; Green & Adam, 2003).
Transverse geochemical variations across the Izu–Bonin arc and other volcanic arcs have been discussed extensively in the literature (Kuno, 1960, 1966; for summary, see Gill, 1981, p. 209; Tatsumi & Eggins, 1995; Hochstaedter et al., 2000, 2001). Based on data compiled from the Izu–Bonin arc volcanoes, we suggest here that such cross-arc variations need not be the result of greater degrees of partial melting beneath frontal regions (e.g. Sakuyama & Nesbitt, 1986; Kushiro, 1994) but may instead be related to the transient stability of phengite in the downgoing slab.
|
ANALYTICAL METHODS |
|---|
After initial splitting and jaw crushing, all samples were pulverized in an agate ball mill. Major and trace elements were determined by X-ray fluorescence (XRF) at IFREE, JAMSTEC. Trace elements were analyzed on pressed powder discs, and major elements were determined on fused glass discs. A mixture of
0·4 g of each powdered sample and 4 g of anhydrous lithium tetraborate (Li2B4O7) was used; no matrix correction was applied because of the high dilution involved. The REE and some trace elements were determined by inductively coupled plasma mass spectrometry (ICP-MS) using a VG Elemental® PQ3 instrument enhanced with a chicane lens system, following the procedures described by Chang et al. (2003). Accurately weighed 100 mg aliquots of powdered samples were digested in screw-cap PFA vials with mixed acids (0·65 ml HClO4 and 2 ml HF) on a hot plate at 130–140°C for 3 days. After drying, 1 ml HClO4 was added, the vials were heated at 160°C overnight, and they were then opened to evaporate the solutions at a gradually increased temperature of up to 190°C to drive off excess HF and convert the fluoride into chlorides. The residues were taken into 2 ml of 6 mol/l HNO3, moderately heated and then evaporated at 120°C to incipient dryness. The final dissolution was performed with 10 ml of 2% HNO3. Prior to analysis, In and Bi were added to aliquots of sample solutions as internal standards to correct signal drifts during measurements. Typical dilution factors from this preparation procedure were about 7500 for the samples analyzed. The ICP-MS system was optimized to a sensitivity of c. 120 000 c.p.s./ppb and 100 000 c.p.s./ppb for 115In and 209Bi, respectively, and Ce oxide formation of less than 0·5%. Measured Eu and Gd concentrations were corrected for oxide and hydroxide interferences. The analytical results of repeated analyses of well-established reference standards (JB-2 and BHVO-1) agreed very well with recommended values, proving that the REE and trace element determinations were performed at high accuracy and reproducibility. Representative major and trace element data for basaltic rocks from Torishima volcano are reported in Table 1. The entire XRF dataset and modal mineralogy of volcanic rocks from Torishima volcano, ranging from basalt to rhyolite, are available as an Electronic Appendix, which may be downloaded from http://www.petrology.oxfordjournals.org/. All major element data have been normalized to 100% on a volatile-free basis, with total iron calculated as FeO.
|
Electron microprobe analyses were carried out using the JAMSTEC JEOL JXA-8900 Superprobe equipped with five wavelength-dispersive spectrometers (WDS). Olivine analyses were made with a counting time of 100 s, an accelerating voltage of 20 kV, a beam current of 25 nA and a probe diameter of 5 µm to ensure reliable Ni values. Pyroxene and plagioclase analyses were made with a counting time of 20 s, an accelerating voltage of 15 kV and a beam current of 15 nA. Representative mineral compositions in basalts and basaltic andesites from Torishima volcano are given in Table 2.
|
|
TORISHIMA VOLCANO |
|---|
The bathymetric map shows that the main edifice of Torishima volcano rises from an ocean depth of
1000 m (Fig. 2). The volcano is about 20 km in diameter, and Torishima island lies along the south rim of a submarine caldera. The surrounding sea floor was dredged during the December 2002 cruise of the R.V. Kairei (KR02-16) (Fig. 2); Torishima island was studied and sampled during a 2 week field expedition in 2003. Fifty-six dredge samples and 141 samples from the island were analyzed by XRF. Torishima seamount (D14 in Fig. 2), an isolated knoll 20 km west of Torishima island, consists of andesite and dacite scoriae (60·7–66·6 wt % SiO2) characterized by systematically higher TiO2 and K2O contents than Torishima volcano at a given SiO2 content. Thus, the Torishima knoll is not considered to be part of the Torishima magmatic system and is not considered further. Most of Torishima island consists of older prehistoric lavas and dykes (<50 wt % SiO2), which are overlain by younger prehistoric lavas (51–53 wt % SiO2) and then by basaltic andesite lava flows of the 1939 eruption (54·5 wt % SiO2) (Table 1, Fig. 3).
|
|
Collectively, more than 70% of the samples from Torishima volcano are basalts and basaltic andesite lava flows and dykes (<55% SiO2). Samples containing >55 wt % SiO2 consist mostly of scoria and pumice and were collected both on land and in dredge hauls from the nearby sea floor. Figure 4 shows the variation of MgO vs SiO2 for the analysed samples from Sumisu and Torishima volcanoes. Both volcanoes display compositions ranging from basalt to rhyolite. At Sumisu, a
6 wt % SiO2 gap separates andesites from dacites [see Shukuno et al. (2006) for a possible explanation of this gap]; at Torishima, pumices from the both the sea floor and the island bridge this gap (Fig. 4).
|
Basalts from Torishima volcano
Torishima contains two types of basalt, comparison between which provides important insights into the role of slab-derived components in arc-front basalt genesis.
The volcano is made up of three basalt groups of differing ages, which are separated by two small, but well-defined, gaps in SiO2 (50–51 wt % and 53–54·5 wt %) (Fig. 5a). Silica content increases from the older prehistoric units (<50 wt % SiO2) through the younger prehistoric units (51–53 wt % SiO2) to the products of the 1939 eruption (54·5 wt % SiO2) (Figs 3 and 5a). Our field study of Torishima island suggests that the main body of the island, and possibly the entire Torishima volcano, consists of older prehistoric lavas and dykes (Fig. 3). Thus, basalts containing <50 wt % SiO2 are the most voluminous. Using FeO*/MgO as a proxy for differentiation, however, the basalts form two trends, which are distinguished by their differing Zr contents at the same FeO*/MgO (Fig. 5b). Basalts with SiO2 < 50 wt % have lower Zr at the same FeO*/MgO than basalts and basaltic andesites with >51 wt % SiO2, which are comparable respectively to the low-Zr and high-Zr basalts of Sumisu (Tamura et al., 2005). Most basaltic volcanism in Torishima volcano, therefore, consists of low-Zr basalts, which change to high-Zr basalts in the later stages of activity without apparent mixing (Fig. 5b).
|
Our data for the Torishima island basalts are comparable with those of Taylor & Nesbitt (1998
) and Gill et al. (1994). The data of Taylor & Nesbitt (1998) encompass the range of our low-Zr basalts to high-Zr basalts whereas Gill et al. (1994) reported data for the high-Zr basalt that erupted in 1939. |
PETROGRAPHY AND OLIVINE CHEMISTRY OF LOW-ZR AND HIGH-ZR BASALTS FROM TORISHIMA VOLCANO |
|---|
Figure 6 shows mineral assemblages and modes (vol.%) of 37 high-Zr basalts and 93 low-Zr basalts, in order of increasing FeO*/MgO. The FeO*/MgO values of the low-Zr basalts and high-Zr basalts from Torishima have similar ranges from 1·31 to 2·78 and from 1·13 to 2·68, respectively. Most low-Zr basalts, however, have FeO*/MgO <2, whereas many of the high-Zr basalts are more differentiated and have FeO*/MgO >2.
|
With respect to mafic mineral phenocrysts, 90 of 93 low-Zr basalts contain both olivine and augite. Two samples, with 1· 8–1· 9 FeO*/MgO, contain only olivine phenocrysts; the most differentiated sample (FeO*/MgO = 2·78) contains orthopyroxene and augite phenocrysts without olivine. In contrast, the high-Zr basalts show a wider variation in mafic phenocryst assemblages. Fifteen of 37 high-Zr basalts contain olivine + augite, nine contain olivine + augite + orthopyroxene, nine contain augite + orthopyroxene, two contain only augite, and one sample each contains only olivine or orthopyroxene. Four samples, with FeO*/MgO <2·0, contain orthopyroxene phenocrysts, suggesting earlier crystallization of orthopyroxene in high-Zr basalts. No pigeonite has been observed as a phenocryst in either the low-Zr basalts or high-Zr basalts.
Figure 7 shows olivine, plagioclase, orthopyroxene, clinopyroxene, and ‘total phenocrysts’ contents (vol.%) vs FeO*/MgO in the whole-rocks. The phenocryst content of the low-Zr basalts ranges from 27 to 51%, averaging 40%. The phenocryst content of high-Zr basalts averages 30 vol.%, about 10% lower than for the low-Zr basalts. Moreover, many high-Zr rocks contain <20 vol.% of phenocrysts, and one sample (FeO*/MgO = 2·0) is aphyric (1% phenocrysts). Generally, the low-Zr basalts contain more abundant phenocrysts than the high-Zr basalts, but orthopyroxene appears only in the high-Zr basalts at FeO*/MgO <2. Plagioclase abundances are similar in the low-Zr and high-Zr basalts, except in some phenocryst-poor rocks and in the 1939 lava flows (Fig. 7).
|
At Sumisu, many high-Zr basalts contain only olivine + plagioclase, and augite appears only in the more evolved lavas of the high-Zr basalts. Most Torishima high-Zr basalts, however, contain augite phenocrysts. Figure 8 shows photomicrographs of Torishima basalts and basaltic andesites. Some plagioclases in low-Zr basalts contain inclusions of augite (Fig. 8a), whereas, in contrast, some augites in the high-Zr basalts contain inclusions of plagioclase (Fig. 8b and c). Augite phenocrysts containing plagioclase inclusions have not been observed in the low-Zr basalts. The presence or absence of plagioclase inclusions in augite or vice versa may not necessarily provide strong evidence of the order of crystallization, and the inclusions could be xenocrystal. We consider it likely that augite crystallized after plagioclase in the Torishima high-Zr basalts. It is possible that the crystallization sequences of the high-Zr basalts in both Sumisu and Torishima are the same, but they differ from those of the low-Zr basalts in these two areas.
|
Olivines
Olivine phenocrysts occur in both the low- and high-Zr basalts from Torishima (Figs 6 and 7). Figure 9 shows the variation of NiO vs Fo [100Mg/(Mg + Fe)]. The low-Zr basalts are richer in olivine phenocrysts than the high-Zr basalts (Fig. 7), but both low- and high-Zr basalts have similar olivine compositions in terms of Fo and NiO contents (Fig. 9a). Low-Zr Torishima basalts contain as much as 7 vol.% olivine (Fig. 7), but the NiO contents of these olivines are low (Fig. 9a), suggesting a high degree of olivine fractionation before eruption. At Sumisu, olivines in the low-Zr basalts have lower NiO contents and/or have higher Fo values than those in the high-Zr basalts at given Fo and NiO contents, respectively. Based on this olivine chemistry, Tamura et al. (2005
) suggested that low-Zr basalts are the products of higher degrees of partial melting than the high-Zr basalts. Unfortunately, all olivines from Torishima are fairly evolved, and most basalts contain augite phenocrysts. Thus the differences in olivine chemistry between the primary low-Zr and high-Zr basalts have been obscured by fractionation of both olivine and augite.
|
We selected the most magnesian low-Zr basalts (31, 33, 35.5, 103 and 117) and high-Zr basalts (11, D15-R7, D19-R10, D19-R12 and D19-R13) to minimize the effect of augite fractionation. Figure 9b shows the composition of olivine phenocrysts in the low-Zr basalts (red field) and high-Zr basalts (blue field) compared with calculated equilibrium olivines (squares) and olivine fractionation trends (lines) determined for magnesian low-Zr basalts (black squares; 31, 33, 35.5, 103 and 117) and magnesian high-Zr basalts (white squares; 11, D15-R7, D19-R10, D19-R12 and D19-R13). The numbers on the lines indicate the percentage of added equilibrium olivine. Calculated olivine compositions in the low-Zr basalts are more magnesian than those in high-Zr basalts at a given NiO content. Low-Zr basalts contain olivine phenocrysts whose compositions overlap those of the calculated equilibrium olivines, and continuously extend towards more iron-rich compositions with decreasing NiO content and also extend to more iron-rich compositions at the same NiO content. The latter could be the effect of the augite fractionation. On the other hand, although the high-Zr basalts contain several olivine phenocrysts that are more magnesian than the calculated equilibrium olivines, most olivine compositions from the high-Zr basalts overlap those of the calculated equilibrium olivines and continuously extend toward more iron-rich compositions. It is possible that several magnesian olivines, which are more magnesian than the equilibrium high-Zr olivine trends, could be xenocrysts from older low-Zr basalt magmas because the majority of actual high-Zr basalt olivines plot on the extension of the calculated high-Zr basalt olivine trends. The primary olivine compositions at 0·4 wt % NiO calculated for the low-Zr basalts (31, 33, 35.5, 103 and 117) range from Fo92·8 to Fo93·3, with an average of Fo93. In contrast, those calculated for the high-Zr basalts (11, D15-R7, D19-R10, D19-R12 and D19-R13) are Fo91 on average and range from Fo90·3 to Fo91· 4. Thus, we can conclude from Fig. 9 and the primary olivine compositions calculated from the whole-rock compositions [see Tamura et al. (2005
) for details], that both low-Zr and high-Zr basalts were derived from depleted, and thus magnesiun-rich, mantle sources and that the source mantle of the low-Zr basalts is more magnesian, and thus more depleted, than that of the high-Zr basalts. |
GEOCHEMISTRY OF LOW-Zr VS HIGH-Zr BASALTS FROM TORISHIMA VOLCANO |
|---|
Figure 10 shows the variation of selected major elements (SiO2, Al2O3, MgO, TiO2, FeO*, CaO, Na2O and K2O; Fig. 10a) and trace elements (Ni, Sr, Y, Rb, Cu and Zn; Fig. 10b) vs FeO*/MgO in the high-Zr and low-Zr Torishima basalts. The K2O contents of the high-Zr basalts are almost twice those of the low-Zr basalts at the same FeO*/MgO. SiO2, Na2O, Cu and Y also show systematic differences between these two basalt types, but on smaller scales. They do not display systematic differences in Al2O3, MgO, TiO2, Ni, Sr, and Zn at a given FeO*/MgO.
All Torishima basalts are strongly depleted in light REE (LREE), compared with middle REE (MREE) and heavy REE (HREE) (Fig. 11). Importantly, the low-Zr basalts at Torishima are more LREE depleted than the high-Zr basalts, a feature also observed at Sumisu volcano (Tamura et al., 2005). Figure 11 shows that the REE patterns of the low-Zr and high-Zr basalts from Torishima are comparable with those at Sumisu.
|
|
Figure 12 shows the variation of La/Sm vs Gd/Yb and La/Sm vs Zr/Y for Torishima (this study) and Sumisu (Tamura et al., 2005
). La/Sm is positively correlated with Gd/Yb and Zr/Y at both Torishima and Sumisu, and the ranges and values are similar. La/Sm values at Sumisu range from 0·7 to 1· 05, and the low-Zr basalts have lower ratios, ranging from 0·7 to 0·9. At Torishima, La/Sm values range from 0·7 to 1· 0, and, again, the low-Zr basalts in Torishima have lower ratios, ranging from 0·7 to 0·8. It is important to note, however, that although the shapes of the REE patterns are similar (Fig. 11), the absolute REE concentrations are lower in the Torishima low-Zr suite than in the Sumisu low-Zr suite. The low-Zr suite of basalts are amongst the most LREE-depleted basalts found anywhere on Earth to date, despite having experienced differentiation that has given rise to whole-rock MgO contents ranging from 7·6 to 5·0 wt % and olivines with Fo85 to Fo65. The extreme depletion is indicated by the rocks having less than five times the mantle concentrations of HREE and HFSE and sub-chondritic Nb. Furthermore, their incompatible element concentrations are about half of the concentrations in the most depleted mid-ocean ridge basalt (MORB) known, despite being more differentiated. Such extremely depleted rocks have been found in only one other area, the northernmost islands of Tonga, especially on Tafahi (J. B. Gill, personal communication).
|
|
DEGREES OF MELTING OF THE SOURCE OF THE TORISHIMA BASALTS |
|---|
The degree of partial melting of the source of the primary magmas can be estimated using the ratios of trace element concentrations in lavas with different compositions, as has been described by Maaløe (1994
) and Maaløe & Pedersen (2003). To accomplish this, those workers assumed (1) the likely modal mineralogy of the source mantle, and (2) the relative proportions of the mineral constituents that enter the melt. Based on these assumptions, the estimated degree of melting depends on the trace element ratios in two selected lava compositions and not on their absolute values. Tamura et al. (2005) constrained the degree of melting of the source mantle beneath Sumisu using three ratios (La/Sm, Sm/Yb, and Zr/Y) in lavas of differing composition without making the two assumptions listed above. Although they were unable to obtain unique solutions, some consistent constraints emerged.
Here, we better constrain the degree of partial melting of the source of the primary magmas, using La/Sm in lavas of differing composition and making the two assumptions listed above. This approach, assuming that the effects of slab-derived components on LREE abundances are minimal, is tenable at Sumisu and Torishima because (1) the lavas of both volcanoes are heavily depleted in LREE compared with MREE and HREE (Fig. 11), and (2) La/Sm is positively correlated with Gd/Yb and Zr/Y in both the Torishima and Sumisu basalts (Fig. 12). First, we assume that the relative proportions of the mineral constituents enter the melt according to the reaction of Gaetani & Grove (1998), for non-modal batch melting of hydrous lherzolite at 1· 2 GPa: 0·62 cpx + 0·51 opx + 0·12 sp = 0·25 ol + 1· 0 melt. Second, although the Izu–Bonin arc basalts are among the most depleted on Earth, we assume that residual clinopyroxene plays a major role in fractionating La from Sm and Zr from Y (Tamura et al., 2005). Thus the melting residues of the Izu–Bonin basalts could range from cpx-free harzburgite to lherzolite containing small amounts of clinopyroxene. Hirschmann et al. (1999) showed that when cpx is exhausted from the residue during isobaric batch melting of spinel peridotite at 1 GPa, the isobaric productivity calculated by MELTS decreases by more than a factor of four, and therefore extensive partial melting of harzburgitic residues is inhibited. Thus, we assume that all clinopyroxene could be exhausted to leave a harzburgitic residue in the source of the low-Zr primary basalts. This residue is just at the point of the cpx-out reaction without further melting of the harzburgitic residue, when the melt fraction is 0·2.
Gaetani et al. (2003) and McDade et al. (2003a) showed that under both hydrous and anhydrous conditions, there are differences in the partitioning behaviour of trace elements during mantle melting. For hydrous harzburgite, which is assumed to be the residue of the primary low-Zr basalts, hydrous melting partition coefficients for orthopyroxene are taken from McDade et al. (2003a); these are similar to the anhydrous ones of McDade et al. (2003b). Those for olivine are taken from McDade et al. (2003b), assuming that H2O has a negligible effect on olivine–melt partition coefficients. Residual clinopyroxene plays a major role in fractionating La from Sm, and thus the residue of the primary high-Zr basalts should contain some clinopyroxene. Although the high-Zr basalts are not thought to be dry, they should contain less water than the low-Zr basalts. As a result, for the source mantle of the high-Zr basalts, anhydrous melting partition coefficients, taken from McDade et al. (2003b), are used for olivine, clinopyroxene and orthopyroxene. When the olivine content of the residue of the high-Zr basalts is assumed to be more than 70 wt %, the likely modal mineralogy of the original mantle can be determined. Spinel is not included in the calculations because spinel–melt partition coefficients are assumed to be negligible for La and Sm.
Interactive calculations were performed to minimize the difference between the original mantle source mineralogy (olivine, orthopyroxene and clinopyroxene) of the low-Zr basalts and that of the high-Zr basalts.
Following Tamura et al. (2005),
|
|
|
|
|
|
Figure 13 shows the degree of melting for the high-Zr basalt (FH) vs La/Sm in the low-Zr basalt assuming FL = 0·2. FH is determined using La/Sm values for pairs of high-Zr basalts and low-Zr basalts. The La/Sm ratios of the high-Zr basalts and their sample numbers are shown in the diagram. The La/Sm values of the low-Zr basalts range from 0·71 to 0·83. This variation could be caused by different degrees of partial melting within the low-Zr mantle source. However, we assumed that 20% partial melting produced each of the low-Zr basalt primary magmas. Thus, the degree of melting of the source of the high-Zr basalt increases with La/Sm in the paired low-Zr basalt. Table 3 shows the average from calculations made using 30 paired low-Zr and high-Zr basalts. The average FH is 0·11 (Table 3). When 20% partial melting is assumed to produce the primary magmas of the low-Zr basalt, only 11% melting is required to yield the high-Zr basalts (Table 3).
|
|
|
FLUID-DERIVED COMPONENTS IN THE SOURCE OF THE TORISHIMA BASALTS |
|---|
The positive correlations between La/Sm vs Gd/Yb and La/Sm vs Zr/Y (Fig. 12) at Sumisu and Torishima volcanoes and the calculations using La/Sm values (Fig. 13) suggest that the low-Zr basalts result from higher degrees of partial melting (
20%) than the high-Zr basalts (11%). We therefore use Zr/Y as a proxy of the degree of melting of the mantle source. When we construct a diagram using incompatible element (A) on a Zr/Y vs A/Y plot, there should be a positive correlation if element A is more incompatible than Y in terms of partial melting of the source mantle. On the other hand, low-Zr basalts may have higher water contents than the high-Zr basalts under mantle conditions (Tamura et al., 2005), where water is derived from the subducting lithosphere beneath the frontal volcanoes. Thus, any deviation from the partial melting trend will be the result of water, which can contain several specific elements, being added to the mantle source (e.g. Elliott, 2003). Figure 14 shows the variation of Zr/Y vs A/Y (A = Rb, Ba, K2O, Pb, and Sr) for the low-Zr and high-Zr basalts from Torishima volcano. K2O/Y shows a positive correlation with Zr/Y, and the low-Zr basalts have lower K2O/Y values than the high-Zr basalts. This is consistent with variations in the degree of melting in the source mantle. On the other hand, Pb/Y and Sr/Y, and to a lesser degree Ba/Y and Rb/Y, deviate from this melting trend, showing relative enrichment of the low-Zr basalts in Pb and Sr compared with Y. Figure 15, presenting incompatible element data for representative samples of the low-Zr and high-Zr basalts, shows trends similar to those in Fig. 14. Elements assumed to be more incompatible than Sm, during partial melting of the source mantle and fractional crystallization of the parental basalt magmas, might be expected to form positive trends on these diagrams. Th and K form such trends, but Cs and Sr clearly do not. The lack of correlation for Cs and Sr could be explained by the addition of an aqueous fluid derived from the slab to the source mantle. The behaviour of Rb and Ba is intermediate between that of Th and K and Cs and Sr, suggesting that they are mildly immobile in the slab-derived aqueous fluids. Thus, we conclude that the REE, HFSE (Zr and Th), and K, and to a lesser degree Rb and Ba, follow degrees-of-melting expectations between high- and low-Zr basalts, but as shown in Figs 14 and 15, Cs, Pb and Sr do not. The latter elements are contained in arc-front slab-derived fluids. Rb and Ba exhibit variable behaviour in the low-Zr basalts, ranging from immobile, similar to K, to mildly enriched in some low-Zr basalts.
|
|
|
DISCUSSION |
|---|
Effect of H2O on phase stability of low- and high-Zr Torishima basalts
Basalts from Izu-Oshima, near the northern end of the Izu–Bonin arc, have compositions similar to the high-Zr basalts from Torishima, and also contain low-Ca pyroxene phenocrysts; the water content of these basalts was estimated to be <0·7% (Fujii et al., 1988
). In the Lesser Antilles volcanic arc, Macdonald et al. (2000) suggested that the relatively early appearance of orthopyroxene in the low-K suite of the Soufriere of St. Vincent, compared with St. Kitts and Montserrat, might indicate that the magmas were less hydrous, whereas its non-appearance in the C-series of Grenada is consistent with more hydrous crystallization. The appearance of orthopyroxene, however, may depend at least as much on the activity of silica as water.
We have no direct evidence for the H2O content of the Torishima basalts, but phenocryst populations and experiments by other workers provide important information. There is apparent incompatibility of Al in both high- and low-Zr suites (Fig. 10a). Arguably, this could mean that the water content is high enough to suppress plagioclase crystallization in both. Conversely, it could mean that plagioclase has accumulated in both, especially the high-Zr basalts. All low-Zr basalts in Sumisu and in Torishima contain olivine + clinopyroxene + plagioclase phenocrysts (except for the single differentiated sample shown in Fig. 6). In contrast, many Sumisu high-Zr basalts contain only olivine + plagioclase, and in Torishima, clinopyroxene appears as a phenocryst later than plagioclase in the crystallization sequence of the high-Zr basalts (Fig. 8). Nineteen of 37 high-Zr Torishima basalts contain orthopyroxene phenocrysts. Moreover, four Torishima high-Zr samples, having FeO*/MgO <2·0, contain orthopyroxene phenocrysts, suggesting that orthopyroxene crystallizes earlier in the high-Zr basalts (Figs 6 and 7). Similar relationships have been discussed in part by Tamura et al. (2005).
Green (1982) showed experimentally that, under dry conditions in the 0–5 kbar pressure range, plagioclase is the major liquidus, or near-liquidus, phase in tholeiitic basalts and is subsequently joined by olivine and later by pyroxene. When 5% H2O is added to a tholeiite, however, the appearance of plagioclase is markedly depressed, and pyroxene together with olivine become important near-liquidus phases from 2 to 14 kbar (Green, 1982). Sisson & Grove (1993) demonstrated that at 2 kbar, with the addition of H2O, the assemblage plagioclase + olivine + high-Ca pyroxene appears at high melt fractions, apparently close to the liquidus of high-alumina basalts. On the other hand, at low to moderate pressures, anhydrous high-alumina basalts crystallize plagioclase or olivine and plagioclase over extended temperature intervals before being joined by high-Ca pyroxene (Grove et al., 1982; Baker & Eggler, 1983; Bartels et al., 1991). Cotectic growth of augite is a likely consequence of an elevated magmatic water content that promotes expansion of the clinopyroxene phase volume. Kushiro (1969) showed that H2O destabilizes low-Ca pyroxene in the system forsterite–diopside–silica. Sisson & Grove (1993) established that orthopyroxene crystallization from arc magmas is inhibited by high water contents; low-Ca pyroxene crystallizes in high-alumina basalts in 1 atm anhydrous experiments (Grove et al., 1982) but does not in 2 kbar H2O-saturated experiments (Sisson & Grove, 1993). These experimental results suggest that the phenocryst assemblage olivine + plagioclase crystallizes under dry conditions and thus the appearance of high-Ca pyroxene, together with olivine and plagioclase, requires a few per cent of water. Moreover, the early appearance of low-Ca pyroxene might be taken as an indication that the magmas were less hydrous.
Collectively, these findings qualitatively suggest that low-Zr basalt magmas containing phenocrystic olivine + clinopyroxene were hydrous, but that cpx-free or opx-bearing high-Zr magmas were dry.
Water and degrees of melting
Elliott et al. (1997) assumed that there is a fairly constant aqueous fluid flux along the entire Mariana arc, and they did not consider the degrees of melting of mantle sources. For example, in plots of fluid-immobile incompatible elements, the Mariana volcanic rocks display arrays that they inferred to reflect variable mixing of a sediment component with a depleted mantle wedge. Part of the variation, however, may have resulted from the effect of differing degrees of melting of a similar mantle wedge. Straub et al. (2004) separated low-K from very-low-K suites in tephra from ODP site 782A located 120 km east of the Izu–Bonin Quaternary volcanic front. They concluded that the differences reflect the mass fraction of K-bearing fluid added to the wedge, based on the assumption that both suites resulted from an uniform (20%) degree of partial melting (batch melting) of the mantle source. In the South Sandwich Island arc, Pearce et al. (1995) concluded that the arc-melting event requires an average of 20% melting and suggested that some of the back-arc lavas in this area could have been derived by a smaller percentage of melting of a less depleted source. In the Aleutian arc, based on modal batch-melting models of lherzolite, Jicha et al. (2004) indicated that the compositions of the Roundhead and Shishaldin lavas require a 1·5–2·0% partial melt of a slightly modified MORB-source mantle, whereas Seguam lavas require a larger (1–5%) fluid addition to the mantle wedge and 22% partial melting of the fluid-enriched source. Luhr (1992) studied two contrasting Mexican volcanoes (Colima and Ceboruco) and concluded that the mantle wedge beneath Colima is more strongly affected by fluids rising from the slab, and this effect leads to higher percentages of partial melting (by a factor of 2·5) as compared with the Ceboruco source. A similar correlation between magmatic water content and degrees of partial melting was deduced by Stolper & Newman (1994) for basalts from the Mariana Trough, Kelley et al. (2006) for submarine basalts at several back-arc basins, and by Eiler et al. (2000), based chiefly on results from the Vanuatu arc. Tamura et al. (2005, and this study) have used data from Sumisu and Torishima to suggest that, within a single volcanic complex, the water content in the mantle source region can be heterogeneous, and this results in different degrees of partial melting of the source. These papers deal with basalts, but Kushiro (1990), Baker et al. (1994) and Tamura (1994) suggested that a heterogeneous distribution of H2O in the mantle results in the production of a spectrum of mantle melts ranging from wet (calc-alkaline) to dry (tholeiitic).
In summary, from both regional and local points of view, a positive correlation might exist between the proportion of water in the source mantle and its degree of melting; the maximum degree of melting of the arc source mantle is about 20%.
Aqueous fluids derived from the subducting slab beneath volcanic front volcanoes
Volcanic front magmatism along subduction zones is believed to be strongly influenced by a ubiquitous ‘slab-fluid’ component (e.g. Pearce, 1982; Hawkesworth et al., 1993; Elliott, 2003). On the other hand, nearly H2O-free magmas have been described along the Cascade arc and Izu–Bonin–Mariana arc (e.g. Nakano & Yamamoto, 1991; Sisson & Bronto, 1998; Kohut et al., 2006). Tamura et al. (2005), furthermore, showed that both dry and wet primary basalts exist in the Sumisu magmatic system. The same two types of basalt are distributed widely on Torishima island. We can therefore compare wet and dry basalts from a single volcano, making it possible to identify the specific elements involved in fluid transport from the subducting slab. Figures 14–15
show that at Torishima Cs, Pb, and Sr were carried by these fluids, but K, Th and REE were not. Rb and Ba are mildly immobile. Interestingly, results from Torishima are consistent with the widely held views on the role of slab-fluid components in arc volcanoes (Elliott, 2003), based upon a comparison of basalt compositions in various arc settings with those erupted along mid-ocean ridges. Moreover, trace-element variations within the low-Zr basalts can be produced by the addition of fluid-derived components, so we can, therefore, estimate the proportion of a specific element derived from the fluids. For example, the Sr/Y of the low-Zr basalts ranges from 14 to 19 (Fig. 14) at similar Zr/Y values. If this variation is produced by the addition of Sr to the source mantle by fluids, 60–70% of the Sr in the low-Zr magmas could have been derived from the subducted slab (Fig. 14).
Based on isotopic and trace element data from the northern Izu–Bonin arc, Ishizuka et al. (2003) suggested the following: (1) about 35% of the Sr is derived from fluids from altered oceanic crust (AOC) at the volcanic front; (2) contributions from the slab are more significant for Pb, and the volcanic front requires 70–80% of the Pb from the slab-derived component; (3) Nd is almost exclusively mantle-derived in the volcanic front lavas, suggesting that the contribution of Nd from the slab-derived fluid is negligible. As shown in Figs 14 and 15, the wide and continuous range of variation in these fluid-mobile elements in the low-Zr basalts (at similar Zr/Y and La/Sm values) may reflect a range of fluid volumes added to the mantle wedge. Moreover, the positive trends displayed by the high-Zr basalts (Figs 14 and 15) are not always independent of the fluid additions, because relatively dry high-Zr magmas could contain some slab-derived fluids. Comparison of the results of Ishizuka et al. (2003) and this study may, therefore, not be appropriate, but at least the high mobility of Sr and Pb and the immobility of REE are mutually consistent.
Immobile potassium, LILE fractionation and the possible role of phengite
One new finding of this study is that K appears to be immobile in the slab below Torishima. Experimental studies on the stability of phengitic muscovite indicate that K-mica in the slab persists to depths much greater than the zone of melt generation beneath arcs (Domanik & Holloway, 1996; Schmidt, 1996). Schmidt (1996) and Schmidt & Poli (1998) showed that dehydration at low pressures is almost entirely associated with potassium-free phases, whereas at pressures above 30 kbar (100 km depth) the role of phengite becomes increasingly important. Johnson & Plank (1999) performed dehydration and melting experiments on subducted sediments. They concluded that, with increasing temperature, Rb and K become less compatible in the subducted sediment, and the greatest change in their D values coincides with mica breakdown between 800 and 900°C at 2–3 GPa. This is consistent with our findings from the arc-front Torishima volcano, which show that wet (low-Zr) basalts do not contain excess K2O compared with dry (high-Zr) basalts. Our study suggests the compatible behaviour of K in the downgoing slab beneath Torishima and Sumisu as compared with other incompatible elements (Figs 14 and 15). We suggest that this is the result of phengite stability at temperatures <800°C in the subducting slab below Torishima.
Experimentally determined partition coefficients for LILE between fluids and phengite at high pressures (2–4 GPa) and low temperatures (600–700°C) show that Cs strongly fractionates into the coexisting fluids (Melzer & Wunder, 2000; Green & Adam, 2003). Thus Rb/Cs and Ba/Cs ratios may provide discriminants for the micas involved (T. H. Green, personal communication, 2006). Figure 16 shows the variation of Ba/Cs vs Rb/Cs for Torishima basalts. The high-Zr basalts exhibit a positive correlation across a wider range of values than the low-Zr basalts, which plot in the lower portion of the high-Zr basalt field. The low-Zr basalts are enriched in the fluid-mobile element Cs (Fig. 15), and, moreover, they show LILE fractionation, resulting in lower Ba/Cs and Rb/Cs in the low-Zr basalts (Fig. 16). LILE variation and fractionation could have resulted from the addition of fluids derived from the subducting slab. The source mantle may have been variably enriched in Cs, reflected by the large variation in the high-Zr basalts. The narrow range shown by the low-Zr basalts may have arisen when the fluid-derived Cs dominated. It is also possible that some Ba and Rb may have also been added to the source mantle. Fluid addition to the mantle source will decrease Ba/Cs and Rb/Cs, because Cs does not partition into phengite and strongly partitions into the fluid, compared with Rb and Ba. We therefore suggest that H2O-rich fluids originating from the subducting Pacific plate below the volcanic fronts of the Izu–Bonin arc may contain very little potassium. Instead, these fluids carry a large amount of Cs and, to a lesser degree, Rb and Ba, and cause fractionation of Ba/Cs and Rb/Cs, because K-bearing phengite is stable in the downgoing slab beneath the arc front. In contrast, potassium will be increasingly concentrated